Puetz - Relativistic Geodesic

Fundamental Theories of Physics 196
Dirk Puetzfeld
Claus Lämmerzahl Editors
Relativistic
Geodesy
Foundations and Applications
Fundamental Theories of Physics
Volume 196
Series editors
Henk van Beijeren, Utrecht, The Netherlands
Philippe Blanchard, Bielefeld, Germany
Bob Coecke, Oxford, United Kingdom
Dennis Dieks, Utrecht, The Netherlands
Bianca Dittrich, Waterloo, Canada
Detlef Dürr, Munich, Germany
Ruth Durrer, Geneva, Switzerland
Roman Frigg, London, United Kingdom
Christopher Fuchs, Boston, USA
Domenico J. W. Giulini, Bremen, Germany
Gregg Jaeger, Boston, USA
Claus Kiefer, Cologne, Germany
Nicolaas P. Landsman, Nijmegen, The Netherlands
Christian Maes, Leuven, Belgium
Mio Murao, Bunkyo-ku, Tokyo, Japan
Hermann Nicolai, Potsdam, Germany
Vesselin Petkov, Montreal, Canada
Laura Ruetsche, Ann Arbor, USA
Mairi Sakellariadou, London, UK
Alwyn van der Merwe, Denver, USA
Rainer Verch, Leipzig, Germany
Reinhard F. Werner, Hannover, Germany
Christian Wüthrich, Geneva, Switzerland
Lai-Sang Young, New York City, USA
The international monograph series “Fundamental Theories of Physics” aims to
stretch the boundaries of mainstream physics by clarifying and developing the
theoretical and conceptual framework of physics and by applying it to a wide range
of interdisciplinary scientific fields. Original contributions in well-established fields
such as Quantum Physics, Relativity Theory, Cosmology, Quantum Field Theory,
Statistical Mechanics and Nonlinear Dynamics are welcome. The series also
provides a forum for non-conventional approaches to these fields. Publications
should present new and promising ideas, with prospects for their further
development, and carefully show how they connect to conventional views of the
topic. Although the aim of this series is to go beyond established mainstream
physics, a high profile and open-minded Editorial Board will evaluate all
contributions carefully to ensure a high scientific standard.
More information about this series at http://www.springer.com/series/6001

Dirk Puetzfeld Claus Lämmerzahl
•
Editors
Relativistic Geodesy
Foundations and Applications
123
Editors
Dirk Puetzfeld Claus Lämmerzahl
ZARM ZARM
University of Bremen University of Bremen
Bremen, Germany Bremen, Germany
ISSN 0168-1222 ISSN 2365-6425 (electronic)

Fundamental Theories of Physics
ISBN 978-3-030-11499-2 ISBN 978-3-030-11500-5 (eBook)
https://doi.org/10.1007/978-3-030-11500-5
Library of Congress Control Number: 2018967737
© Springer Nature Switzerland AG 2019

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Recent years have seen the advent of high precision measuring methods, in par-
ticular, modern clocks reached an unprecedented level of accuracy and stability.
This was accompanied by important developments in the fields of atom and laser
interferometry. Laser interferometers improved by several orders of magnitude and
interferometry in space nowadays is a mature technology ready for practical
applications. All this is of direct importance for many fields of physics, and con-
sequently for geodesy.
The high precision of these new experimental capabilities made clear that
geodesy can no longer rely solely on Newtonian concepts, which are still used
within the field. Geodetical models and the interpretation of data within these
models therefore inevitably require concepts which go beyond the Newtonian
picture of space and time. The theoretical underpinning of geodesy should therefore
be based on the special and the general theory of relativity, the latter still represents
the most successful gravity theory to the present date. This new “relativistic geo-
desy” is the topic of the present volume.
In 2016, we organized1 an international conference in Bad Honnef (Germany) on
the Relativistic Geodesy: Foundations and Applications. The conference brought
together the leading experts in their respective fields and was very well received by
the speakers as well as by the audience. We would like to thank the WE-Heraeus
Foundation for the generous support of this conference. Our thanks also go to the
Physikzentrum Bad Honnef where the conference took place.
The positive reception and the feedback after the conference made clear that
there is a strong demand for an up-to-date volume, covering the methods employed
in current research in the context of the relativistic geodesy. This book intends to
give such a status report. It hopefully is of value for the experts working in this field
and may also serve as a guideline for students. At the same time, we should warn
potential readers that it is not intended to serve as a replacement for a textbook on
either of the subjects of gravitational physics or geodesy. But we hope that it
bridges some of the gaps between the relativity and the geodesy communities, in
1
http://puetzfeld.org/relgeo2016.html.
v
vi Preface
particular, when it comes to implementation and application of relativistic concepts

and methods in the field of geodesy.
The present volume is based on the lectures given at the conference and gives an
overview over the following topics:
• Time and frequency metrology
• Chronometric geodesy
• (Clock) gradiometry
• Satellite experiments
• Navigation systems
• Tests of gravity by means of geodetic measurements
In covering these topics, definitions and methods from relativistic gravity are
introduced. Emphasis is put on the coverage of the geodetically relevant concepts in
the context of Einstein’s theory (e.g., role of observers, use of clocks, and definition
of reference systems). Furthermore, fundamental questions in the context of the
measuring process, as well as approximation methods which make certain calcu-
lations feasible, are discussed in detail.
We as editors are deeply indebted to the contributors to this volume, who made
great efforts to present their respective areas of research in an accessible way to a
broader audience. We hope that the material presented in here will prove to be
useful as a reference for experienced researchers, as well as serve as an inspiration
for younger researchers who want to enter the exciting emerging field of relativistic
geodesy.
Bremen, Germany Dirk Puetzfeld2

Claus Lämmerzahl
2
D. P. acknowledges the support by the Deutsche Forschungsgemeinschaft (DFG) through the
grant PU 461/1-1.
Contents
Time and Frequency Metrology in the Context

of Relativistic Geodesy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Andreas Bauch
Chronometric Geodesy: Methods and Applications . . . . . . . . . . . . . . . . 25
Pacome Delva, Heiner Denker and Guillaume Lion
Measuring the Gravitational Field in General Relativity:
From Deviation Equations and the Gravitational Compass
to Relativistic Clock Gradiometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Yuri N. Obukhov and Dirk Puetzfeld
A Snapshot of J. L. Synge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Peter A. Hogan
General Relativistic Gravity Gradiometry . . . . . . . . . . . . . . . . . . . . . . . 143
Bahram Mashhoon
Reference-Ellipsoid and Normal Gravity Field
in Post-Newtonian Geodesy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Sergei Kopeikin
Anholonomity in Pre-and Relativistic Geodesy . . . . . . . . . . . . . . . . . . . . 229
Erik W. Grafarend
Epistemic Relativity: An Experimental Approach to Physics . . . . . . . . . 291
Bartolomé Coll
Use of Geodesy and Geophysics Measurements to Probe
the Gravitational Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
Aurélien Hees, Adrien Bourgoin, Pacome Delva, Christophe Le
Poncin-Lafitte and Peter Wolf
Operationalization of Basic Relativistic Measurements . . . . . . . . . . . . . . 359
Bruno Hartmann
vii
viii Contents
Can Spacetime Curvature be Used in Future Navigation Systems? . . . . 379

Hernando Quevedo
World-Line Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
Jan-Willem van Holten
On the Applicability of the Geodesic Deviation Equation
in General Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
Dennis Philipp, Dirk Puetzfeld and Claus Lämmerzahl
Measurement of Frame Dragging with Geodetic Satellites Based
on Gravity Field Models from CHAMP, GRACE and Beyond . . . . . . . 453
Rolf König and Ignazio Ciufolini
Tests of General Relativity with the LARES Satellites . . . . . . . . . . . . . . 467
Ignazio Ciufolini, Antonio Paolozzi, Erricos C. Pavlis, Richard Matzner,
Rolf König, John Ries, Giampiero Sindoni, Claudio Paris
and Vahe Gurzadyan
Contributions
• A. Bauch Time and frequency metrology in the context of relativistic geodesy

• P. Delva, H. Denker, G. Lion Chronometric geodesy: methods and applications
• Y. N. Obukhov, D. Puetzfeld Measuring the gravitational field in General
Relativity: From deviation equations and the gravitational compass to rela-
tivistic clock gradiometry
• P. A. Hogan A Snapshot of J. L. Synge
• B. Mashhoon General Relativistic Gravity Gradiometry
• S. Kopeikin Reference-ellipsoid and normal gravity field in post-Newtonian
geodesy
• E. W. Grafarend Anholonomity in Pre and Relativistic Geodesy
• B. Coll Epistemic relativity: An experimental approach to physics
• A. Hees, A. Bourgoin, P. Delva, C. Le Poncin-Lafitte, P. Wolf Use of geodesy
and geophysics measurements to probe the gravitational interaction
• B. Hartmann Operationalization of basic relativistic measurements
• H. Quevedo Can spacetime curvature be used in future navigation systems?
• J.-W. van Holten World-line perturbation theory
• D. Philipp, D. Puetzfeld, C. Lämmerzahl On the applicability of the geodesic
deviation equation in General Relativity
• R. König, I. Ciufolini Measurement of frame dragging with geodetic satellites
based on gravity field models from CHAMP, GRACE and beyond
• I. Ciufolini Tests of General Relativity with the LARES satellites
ix
Contributors
Andreas Bauch Physikalisch-Technische Bundesanstalt, Braunschweig, Germany

Adrien Bourgoin Dipartimento di Ingegneria Industriale, University of Bologna,
Bologna, Italy
Ignazio Ciufolini Dip. Ingegneria dell’Innovazione, Università del Salento, Lecce,
Italy;
Centro Fermi, Roma, Italy;
Centro Fermi - Museo Storico della Fisica e Centro Studi e Ricerche “Enrico
Fermi”, Rome, Italy
Bartolomé Coll Relativistic Positioning Systems, Department of Astronomy &
Astrophysics, University of Valencia, Burjassot, Valencia, Spain
Pacome Delva SYRTE Observatoire de Paris, Université PSL, CNRS, Sorbonne
Université, LNE, Paris, France
Heiner Denker Institut für Erdmessung, Leibniz Universität Hannover (LUH),
Hannover, Germany
Erik W. Grafarend Department of Geodesy and Geoinformatics, Faculty of
Aerospace Engineering and Geodesy, Faculty of Mathematics and Physics,
Stuttgart, Germany
Vahe Gurzadyan Center for Cosmology and Astrophysics, Alikhanian National
Laboratory, Yerevan, Armenia
Bruno Hartmann Humboldt University, Berlin, Germany
Aurélien Hees SYRTE, Observatoire de Paris, Université PSL, CNRS, Sorbonne
Peter A. Hogan School of Physics, University College Dublin, Belfield, Dublin 4,
Ireland
xi
xii Contributors
Rolf König Helmholtz-Zentrum Potsdam Deutsches GeoForschungsZentrum

GFZ, Wessling, Germany
Sergei Kopeikin Department of Physics and Astronomy, University of Missouri,
Columbia, MO, USA
Claus Lämmerzahl Center of Applied Space Technology and Microgravity
(ZARM), University of Bremen, Bremen, Germany
Christophe Le Poncin-Lafitte SYRTE, Observatoire de Paris, Université PSL,
CNRS, Sorbonne Université, LNE, Paris, France
Guillaume Lion LASTIG LAREG IGN, ENSG, Univ Paris Diderot, Sorbonne
Paris Cité, Paris, France
Bahram Mashhoon Department of Physics and Astronomy, University of
Missouri, Columbia, MO, USA;
School of Astronomy, Institute for Research in Fundamental Sciences (IPM),
Tehran, Iran
Richard Matzner Theory Group, University of Texas at Austin, Austin, TX, USA
Yuri N. Obukhov Theoretical Physics Laboratory, Nuclear Safety Institute,
Russian Academy of Sciences, Moscow, Russia
Antonio Paolozzi Scuola di Ingegneria Aerospaziale, Sapienza Università di
Roma, Rome, Italy
Claudio Paris Centro Fermi - Museo Storico della Fisica e Centro Studi e
Ricerche “Enrico Fermi”, Rome, Italy
Erricos C. Pavlis Joint Center for Earth Systems Technology (JCET), University
of Maryland, Baltimore County, MD, USA
Dennis Philipp Center of Applied Space Technology and Microgravity (ZARM),
University of Bremen, Bremen, Germany
Dirk Puetzfeld Center of Applied Space Technology and Microgravity (ZARM),
University of Bremen, Bremen, Germany
Hernando Quevedo Instituto de Ciencias Nucleares, Universidad Nacional
Autónoma de México, Mexico, DF, Mexico;
Dipartimento di Fisica and ICRANet, Università di Roma “La Sapienza”, Rome,
Italy;
Department of Theoretical and Nuclear Physics, Kazakh National University,
Almaty, Kazakhstan
John Ries Center for Space Research, University of Texas at Austin, Austin, TX,
USA
Contributors xiii
Giampiero Sindoni Scuola di Ingegneria Aerospaziale, Sapienza Università di

Roma, Rome, Italy
Jan-Willem van Holten Nikhef, Amsterdam, The Netherlands
Peter Wolf SYRTE, Observatoire de Paris, Université PSL, CNRS, Sorbonne
Time and Frequency Metrology
in the Context of Relativistic Geodesy
Andreas Bauch
Abstract A status report is given on current practice and trends in time and frequency
metrology. Emphasis is laid on such fields of activity that are of interest in the context
of relativistic geodesy. In consequence, several topics of a priori general relevance
will not be dealt with. Clocks and the means of comparing their reading are equally
important in practically all applications and thus dealt with in this contribution. The
performance of commercial atomic clocks did not change significantly during the
last 20 years. Progress is noted in the direction of miniaturization, leading to the
wide-spread use of chip-scale atomic clocks. On the other hand, research institutes
invested considerably into the perfection of their instrumentation. Cold-atom caesium
fountain clocks realize the SI-second with a relative uncertainty of close to 1 × 10−16 ,
and with a relative frequency instability of the same magnitude after averaging over
a few days only. Optical frequency standards are getting closer to being useful in
practice: outstanding accuracy combined with improved technological readiness can
be noted. So one necessary ingredient for relativistic geodesy has become available.
Satellite-based time and frequency comparison is here still somewhat behind: Time
transfer with ns-accuracy and frequency transfer with 1 × 10−15 per day relative
instability have become routine. Better performance requires new signal structures
and processing schemes, some appear on the horizon.
1 Introduction
From the author’s perspective, relativistic geodesy requires the following actions:
Two “super-clocks” have to be operated simultaneously. Their frequency accuracy
and stability, and the frequency difference or frequency ratio prevailing when both
are operated side-by-side need to be determined at the outset. Then one of them is
kept at a known “height”, the other one at an unknown “height”. Now they need to be
A. Bauch (B)
Physikalisch-Technische Bundesanstalt, Bundesallee100,
38116 Braunschweig, Germany
e-mail: andreas.bauch@ptb.de
URL: http://www.ptb.de/time
© Springer Nature Switzerland AG 2019 1

D. Puetzfeld and C. Lämmerzahl (eds.), Relativistic Geodesy,
Fundamental Theories of Physics 196, https://doi.org/10.1007/978-3-030-11500-5_1
2 A. Bauch
compared with sufficient accuracy in order to determine the height-difference to the

desired accuracy, based on the frequency difference or ratio observed while they are
operated at their new locations. Validity of General Relativity is taken for granted.
For an exercise of this kind to be meaningful, the comparisons at the two sites should
be made with an uncertainty of 10−17 or below, which indeed requires “super” clocks
and comparison means and remains a real challenge as we will see. In the literature
one finds a few early citations of “relativistic geodesy” or “chronometric leveling”,
one referred to quite often is [1]. At the time of writing a very detailed elaboration
on the matter is underway, based on campaigns running during 2011–2015 [2].
Coordinated and disciplined operation of atomic clocks and of time transfer sys-
tems is on the other hand integral part of time and frequency metrology, comprising
in general:
(a) Development of quartz oscillators and atomic clocks and their operation;
(b) Characterization of the properties of oscillators and clocks;
(c) Realization of time scales, e.g. national legal time;
(d) Time and frequency comparison of clocks and time scales, locally and remotely;
(e) Dissemination of time-of-day, time interval, and standard frequency to the public.
In the context of relativistic geodesy, and for economy of writing, it is admissible

to skip some subjects. But it is essential to understand the status and the progress to
be expected in several aspects of (a), (c) and (d). The subjects will be laid down in
four sections before some conclusive statements are given. From the onset it should
be clear that a certain simplification is unavoidable. Optical frequency standards and
fiber-based frequency transfer for the comparison of optical clocks will be dealt with
in separate articles in this volume and covered in a few words only in this one.
2 Characterization of Time and Frequency Signals
This brief section is intended to introduce the vocabulary used further on and to
introduce the two quantities used in the characterization of frequency standards and
clocks. The devices will then be described subsequently. Let us start with statistical
signal properties. The frequency of oscillators and clocks is subject to systematic
and random variations with respect to their intended nominal output value. Many
measures for quantitative characterization are extensively covered in the literature
[3, 4],1 but all are based on the following formal description of the observed signal.
1 Note:Although this is a review article, my intention was not to provide an exhaustive list of
references, but rather to limit myself to text books and previous review articles of other authors with
few exceptions.
Time and Frequency Metrology … 3
A frequency standard outputs a (nearly) sinusoidal signal voltage described by
V (t) = [V0 + e(t)] × sin{2π ν0 t + φ(t)}, (1)
where ν0 , φ(t), V0 , and e(t) are the nominal frequency, the instantaneous phase fluctu-
ations, the nominal signal amplitude, and its temporal variations, respectively. Further
practical quantities are the instantaneous phase-time variations, x(t) = φ(t)/(2π ν0 ),
and the instantaneous normalized frequency departure y(t) = (dφ/dt)/(2π ν0 ). Both
can be analyzed in the time domain and in the frequency domain. For the remainder
of this article the restriction to time-domain quantities is justified, which are based on
mean frequency values ȳ(τ ) measured during an averaging time τ . The most popular
measure is the Allan variance

σ y2 (τ ) = ( ȳk+1 (τ ) − ȳk (τ ))2 /2. (2)
Here the ȳk -values are understood as a contiguous (no dead time) series of data, and
the brackets signify an infinite time average, including normalization. In practice,
a finite sum of terms is only available. Ideally the number of samples at the longest
averaging time τ should be ten or larger. A double-logarithmic plot of σ y (τ ) versus
τ helps to discriminate among some causes of instability in the clock signal because
they lead to a different slope of the plot. If shot noise of the detected atoms is
the dominating noise source, the frequency noise is white and σ y (τ ) decreases like
τ −1/2 . In this case, σ y (τ ) agrees with the classical standard deviation of the sample.
Long-term deviations from this τ −1/2 -behaviour are quite common and indicate that
parameters defining ν0 are not stable. In such a case the classical standard deviation
would diverge with increasing τ and increasing observation time. In Fig. 1 I show
schematic examples of the frequency instability expected or observed for a variety of
atomic frequency standards. More detailed plots of that kind are shown subsequently
as Figs. 2 and 3.
With the exemption of the active hydrogen maser, the following expression relates
the observed σ y (τ ) to operational parameters of a frequency standard typically for a
wide range of averaging times τ :
η 1
σ y (τ ) = ×√ . (3)
Q × (S/N ) τ/s
Here η is a numerical factor of the order of unity, depending on the shape of the
resonance line and of the method of frequency modulation to determine the line
center. Q is the line quality factor (transition frequency / line width of the observed
transition), and S/N is the signal-to-noise ratio for a 1 Hz detection bandwidth.
The frequency standards discussed subsequently differ in the combination of the
quantities Q, S and N . To understand the leap from microwave to optical frequency
standards seen in Fig. 1, a look at Q is helpful.2 In a caesium fountain clock the
2 In
order to understand the word “leap”: Only very few fountain clocks achieve an instability as
shown, see discussion in Sect. 3.2.
4 A. Bauch
10 GHz transition is detected with a 1 Hz line width, whereas, to give orders of

magnitude, a 500 THz optical transition is detected with a line width of 5 Hz, so the
Q-values differ by 104 .
Now we turn to the characterization of systematic effects. The term accuracy is
often used in a broad sense to express the agreement between the clock’s average
output frequency and its nominal value conforming to the SI second definition. But
according to the rules of metrology [5], accuracy should not be combined with a
quantitative statement. Nevertheless, product manufacturers often state the accuracy
of their devices as a range of output frequencies to be expected, usually without giv-
ing details about the causes of potential frequency deviations. A detailed uncertainty
estimate, on the other hand, is provided for so-called primary clocks and optical fre-
quency standards. It reflects the quantitative knowledge of all (known) effects which
may cause the output frequency to deviate from the transition frequency of unper-
turbed atoms (or ions) at rest. The components of the uncertainty due to individual
effects and finally the combined uncertainty are stated. The combined uncertainty
reported for a primary clock expresses the potential deviation of its second-ticks from
the SI-second. For all other devices it reflects the state of knowledge of systematic
effects, whereas the absolute value of the transition frequency involved can be given
only with the (sometimes larger) uncertainty of the primary clock that served as
reference for its measurement.
Fig. 1 Relative frequency instability σ y (τ ) of different atomic frequency standards, from top
to bottom: (typical) rubidium standard (grey); commercial caesium standard type 5071A, high-
performance option (long dash); PTB primary clock CS2 (solid); passive hydrogen masers (short
dash); active hydrogen maser (dash-dot); PTB CSF2, state 2016 (dash-dot-dot); single ytterbium
ion optical frequency standard (bold dots); strontium optical lattice clock (dots)
3 Atomic Clocks
Atomic properties such as energy differences between atomic eigenstates and thus
atomic transition frequencies are believed to be natural constants and thus not to
depend on space and time (apart from relativistic effects). They are governed by
fundamental constants which describe the interactions among particles and fields.
This basic principle governs all kinds of atomic clocks. The most detailed treatment
of the underlying physics is given in the books of Vanier and Audoin [6] and Vanier
and Tomescu [7], in less depth in Audoin and Guinot [8] and in [9], where the reader
can also find detailed explanations of their function.
3.1 Commercial Clocks
3.1.1 Rubidium Gas Cell and Miniaturized Frequency Standards
For completeness they have to be mentioned here, as they are produced in large quanti-
ties and are indispensible in the fields of telecommunication, power grid management,
navigation, just wherever the performance of a quartz oscillator is insufficient. The
atomic reference transition is the 6.84 GHz hyperfine splitting frequency in 87 Rb.
Several manufacturers share the large market. The devices differ in performance,
size, power consumption, and we give in the following some performance figures as
a guideline: As we will see, they have little importance in the context of relativistic
geodesy, maybe just in the background to keep the infrastructure functioning.
Rubidium gas cell frequency standards come in packages between half a liter
and less than 100 cm3 and have a power consumption between 5 and 20 Watt. The
relative deviation of the output frequency (typically 10 MHz) from its nominal value
is of the order 10−9 and difficult to predict. During a month the value may change
by 1 to 30×10−11 due to aging. The relative frequency instability is of order 10−11
at τ = 1 s and white noise characteristics prevail up to 1000 s or even 10000 s of
averaging, depending very much on the stability of the environment. Unless very
special care in the packaging is taken, the devices are sensitive to external magnetic
fields and temperature changes. A very important application is their use as so-
called GNSS-disciplined oscillators: The offset as well as the long-term aging and
sensitivity to external perturbations is suppressed by steering the output frequency
to a reference signal received from a GNSS, today still most common from the US
Global Positioning System GPS (see Sect. 5.1). Because of their low weight and
power consumption rubidium clocks appeared particularly suited for use on board of
satellites. Space qualified versions are today operated in the navigation satellites of
all global navigation satellite systems (GNSS), serving as the source for the synthesis
of the GNSS signals, i.e. carrier frequency and modulation.
So-called chip-scale atomic clocks are on the market since a number of years
and represent an attractive alternative to the rubidium clocks. The atomic reference
6 A. Bauch
transition is the 9.19 GHz hyperfine splitting frequency in 133 Cs that has tradition-
ally been used in the caesium beam clocks discussed in the following section. In
a very compact package and at a power consumption of order 100 mW they out-
perform quartz oscillators and almost reach the rubidium performance [10]. Their
main application is in battery-powered and hand-held devices. Their development
was sponsored by the US military for future use in hand-held GPS receivers. But in
fact, they are now deployed in thousands as part of undersea reflection seismology
sensor installations deployed by oil exploration companies.
3.1.2 Commercial Caesium Clocks
Caesium atomic beam clocks have been produced commercially since the late 1950s,
starting with the so-called Atomichron of the National Company [11]. When design-
ing commercial clocks, a compromise between weight, volume, power consumption,
and performance and cost is unavoidable. Several manufacturers have participated
in this business over the years [12], but today essentially all production of instru-
ments for civil use is in hands of Microsemi (www.microsemi.com). 25 years since
its first appearance on the market, the model 5071A, initially developed by Hewlett-
Packard, then branded as Agilent, later produced by Symmetricom, a firm taken
over by Microsemi recently, is the work horse in the timing community. Standard
and high-performance versions of this clock are on the market. Part of the improved
specifications of the latter versus the former are due to a larger atomic flux employed
which entails a larger S/N ratio. The price to be paid (literally) is a faster depletion
of the caesium reservoir, thus a reduced period of warranty. Recently Oscilloquartz
(ADVA Optical Networking, www.oscilloquartz.com) announced the forthcoming
release of a commercial beam clock, using the technique of optical pumping for state
selection and detection. An instability lower by a factor of three than for conven-
tional commercial caesium clocks was reported at conferences, but no experience on
accuracy and long-term performance has been published yet.
I give examples of the observed performance of clocks of type 5071A operated
at PTB in laboratory environment during 2015. In Fig. 2 (left) records of the clock
rate with reference to UTC(PTB) are shown. The clocks designated C1, C8, and
C9 are high-performance versions, C6 is a standard performance version. In this
context, UTC(PTB) can be regarded as an ideal reference, its scale-unit being very
close to the SI-second (see Sect. 4). The maximum rate we note is that of the clock
C9 of about 4400 ns/360 days, corresponding to a relative frequency difference of
142 × 10−15 . This is a typical value for this type of clock, for which the manufacturer
specifies the magnitude of the offset from the nominal frequency (accuracy) as below
500 × 10−15 . We note in case of clock C8 that its rate (slope of the plot) changed
during the year. The relative frequency instability values of the four clocks are shown
in the right part of Fig. 2. The clocks’ frequency instability is governed by white
frequency noise for averaging times up to a few days of averaging. The so-called
flicker floor is substantial for the device C8 and also noticeable for others. The specs
shown are from a current sales brochure whereas the clocks C8 and C9 are more than
Fig. 2 Left: rate of four 5071 commercial caesium atomic clocks with reference to UTC(PTB)
during 2015 (Modified Julian Day number MJD 57384 corresponds to 2015-12-28): C1 (solid
line), C8 (dotted line), C9 (dashed line), C6 (dash-dot-dot line). Right: relative frequency instability
of the clocks derived from the data shown left, specifications from the 2015 brochure of Microsemi
for the standard performance clock (open square) and the high-performance clock (open circle)
20 years old and each already needed beam tube replacement twice. So the slight
violation of the current specs is not surprising. The standard performance clock
C6 is substantially more stable than the current specs predict. In summary one can
say that the performance of these devices is remarkable and very useful in general
time-keeping activities, but nevertheless their instability is prohibitive to use them in
serious quantitative tests of relativity and also in the context of relativistic geodesy.
PTB continues to operate its legacy CS1 and CS2 caesium atomic beam clocks
as the last ones world-wide of a previously larger ensemble of that kind. They were
developed with the intention to surpass the limitations of commercial clocks and are
each unique specimen. Their uncertainty for the realization of the SI second has been
well developed and published. It amounts to 8 × 10−15 for CS1 and 12 × 10−15 for
CS2 [13]. Their relative frequency instability is not so different from those of the
commercial devices at short averaging times, but in the long term no flicker level
above (1-2)×10−15 can be noted. CS1 and CS2 constitute a back-up reference for
the realization of PTB’s time scale UTC(PTB), see Sect. 4.
3.1.3 Hydrogen Masers
The ground state hyperfine splitting of the hydrogen atom corresponds to a transition
line at a frequency of 1.4 GHz. Research into the use of this atomic transition in a
frequency standards started at Harvard University in the 1950s. In the active maser, as
it is called, stimulated emission inside a high-Q cavity which encloses the hydrogen
atoms kept in a storage bulb is used to detect the atomic transition [14]. In the passive
maser the transition is probed by injecting radiation into the cavity and observing the
effect on the atoms. Limited by the difficulty to control a variety of perturbing effects,
the maser output frequency reflects the unperturbed hyperfine splitting frequency of
hydrogen atoms only with an uncertainty of order 10−11 . But, as already shown
8 A. Bauch
Fig. 3 Left: Frequency steering applied to the PTB active hydrogen maser used for generation
of UTC(PTB) during one year, ending at MJD 57709 (2016-11-17); right: Relative frequency
instability of the data from the second half of the period shown in the plot left; original data (solid
symbols), frequency drift removed (open symbols)
in Fig. 1, both types of standards share a remarkably low frequency instability at

averaging times up to one day, the lowest for any commercially available frequency
standard at present.
Based on a long tradition, masers are today produced in Russia and in the US. A
Swiss firm combines Russian physics packages with Swiss electronics. Small scale
production of masers is reported from China and Japan, but the products are not
used outside the respective country. Only one Russian manufacturer currently offers
a passive maser commercially. However a space qualified variant serves as local fre-
quency source in satellites of the European satellite navigation system Galileo. In the
future also satellites of the Russian counterpart GLONASS shall be equipped with
passive masers. Traditionally, active masers have served as frequency references in
very-long baseline interferometry observatories since their existence. Nowadays sev-
eral National Metrology Institutes (NMI) realize their reference time scales based on
an active hydrogen maser, steered in frequency with respect to a superior “primary”
reference. The same strategy is followed for the physical realization of the system
time of Galileo [15]. These applications are not hampered by the frequency drift typ-
ically associated with masers. Without precautions, the drift caused by the aging of
the mechanical cavity structure typically is of order 10−15 /day. Cavity auto-tuning
reduces the frequency drift to the order 10−16 /day which is then caused by other
effects [14]. In anticipation of the next paragraph and Sect. 4, Fig. 3 illustrates the
performance of the active maser that has been used as physical source for the gener-
ation of PTB’s reference time scale UTC(PTB) during one year until mid November
2016. The UTC(PTB) steering is derived from daily comparisons with one or both
caesium fountain clocks of PTB. The regression line represents the linear part of
the maser frequency drift of 1.03 × 10−16 /day. During the second half of the period
shown, the steering was based on the average of the two fountains, with particularly
higher data availability of fountain CSF2, which clearly improved the day-to-day
stability. For this period the combined frequency instability is shown in the right
plot, based on the original data and with the linear drift removed, respectively.
Active masers constitute also an important infrastructure in laboratories operating
optical frequency standards. Direct comparisons of remote masers, however, cannot
answer questions in the context of relativistic geodesy because of the lack of accuracy.
3.2 Cold-Atom Fountain Clocks
Laser cooling to μK temperature is the key to the success of the fountain concept [16].
In a fountain the laser cooled cloud of atoms is launched upwards with a velocity vs
and the microwave excitation is performed during the ballistic flight, as illustrated in
Fig. 4. The atoms come to rest under the action of gravity at a height of H = vs2 /(2g).
With a height of the fountain setup of about 1 m and vs = 4.4 m/s the total time of
flight, back to the starting point, is about 0.9 s. On their way the atoms interact twice
with the field sustained in the microwave cavity, on their way up and then on their
way down, separated in time by the so-called interaction time. This is typically 0.5 s,
leading to a width of the observed resonance of 1 Hz. During clock operation, the
transition probability is determined changing the probing frequency f p from cycle
to cycle alternately on either side of the central resonance feature. The difference
of successive measurements is numerically integrated and represents the difference
Fig. 4 Operation of a fountain frequency standard, illustrated in a time sequence from left to right.
Arrows represent laser beams (white if they are blocked); a Loading of a cloud of cold atoms; b
Launch of the cloud by de-tuning of the frequency of the vertical lasers; c Cloud expansion during
the ballistic flight; d Second passage of the atoms through the microwave cavity and probing of the
state population by laser irradiation and fluorescence detection
10 A. Bauch
between the frequency source that drives the synthesis electronics producing the
probing signal and the observed caesium transition frequency.
The relative difference between the observed and the unperturbed transition fre-
quency due to several systematic effects amounts to about 10−13 only, much less than
the 2 × 10−10 in beam clocks, and evaluation of several caesium fountains proved that
they realize the SI second with an uncertainty in the low 10−16 range. See [17] for an
overview on fountains and [18] for a detailed uncertainty evaluation for PTB’s second
fountain clock CSF2. In details, its uncertainty depends on the operational conditions
that slightly change from period to period. During September 2016, CSF1 and CSF2
were operated with a stated uncertainty of 3.5 × 10−16 and 2 × 10−16 , respectively.
During the last 24 months, including October 2016, data from 10 caesium fountain
frequency standards were published in the context of collaboration with the Bureau
International des Poids et Mesures (BIPM), see next section, with stated uncertainties
ranging between 0.17 × 10−15 and 2 × 10−15 . Data are shown in Fig. 5 in the next
section.
The frequency instability of caesium fountain clocks depends largely on the oper-
ational parameters, mostly on the atom number in the cloud, but also on the source
of microwave radiation that irradiates the atoms. A few fountains are operated inten-
tionally with very low atom numbers, thereby minimizing the frequency shift due to
cold-atom collisions [7, 17]. On the other hand, low frequency instability is desir-
able, in particular when the frequency of reference transitions of optical clocks shall
be measured in SI Hz as realized with the fountains. Here PTB has pioneered the
routine use of an optically stabilized microwave oscillator [19] instead of a quartz-
oscillator based microwave synthesis. The short term stability of the microwave
signal is provided by a 1.5 μm cavity-stabilized fiber laser via a commercial femto-
second frequency comb. In the long-term, the microwave oscillator involved is locked
to the hydrogen maser to enable fountain frequency measurements with respect to
the maser (see Fig. 3). In this setup the instability contribution of the microwave
oscillator via the so-called Dick-effect [7, 17] becomes negligible and the overall
instability is mostly limited by the number of detected atoms. The respective curve
(CSF2 2016) in Fig. 1 represents this situation.
Fountains operated in several institutes have been compared among each other
over years, and the comparison uncertainty is the combined statistical and system-
atic uncertainty of the standards and the comparison techniques. In the most recent
long term study [20] FO-2 of SYRTE/Observatoire de Paris (OP) was compared
with NIST-F1 of the National Institute of Standards and Technology (NIST), USA,
between 2006 and 2012, just to give one example. The difference FO2 - NIST-F1 was
determined as −0.12 × 10−15 , based on 24 scattered data sets over periods where
both fountains were operated simultaneously or quasi-simultaneously. The uncer-
tainty that accompanies this value is not trivial to determine as the properties of the
fountains and of the time transfer changed over the years. From [20], Table 1, I esti-
mate 0.51 × 10−15 for the systematic part and 0.22 × 10−15 for the statistical part,
where I assume the statistical contribution to decrease with the square-root of the
number of comparisons. Each fountain frequency was corrected for the gravitational
red shift, so as if both were operated at zero height. The corrections amounted to
6.54 × 10−15 for FO-2 and 179.95 × 10−15 for NIST-F1 (Boulder, Colorado, height
some 1600 m above sea level). So the comparisons represent an - even quite crude -
type of relativistic geodesy experiment, although it was never named like that. Deter-
mination of the red-shift correction at the Boulder site is quite challenging and was
described in [21].
In fact, the French FO-2 is a double-fountain in which also rubidium atoms (87 Rb)
are launched and their ground-state hyperfine transition at 6.8 GHz is observed. The
use of Rb atoms instead of Cs is motivated by the fact that at a given number density
of atoms in the cloud, the frequency shift due to cold atom collisions is considerably
smaller. This would allow a better frequency stability, and this was the impetus for
the United States Naval Observatory (USNO) to build a group of Rb-fountains that
has shown indeed excellent performance [22].
3.3 Optical Frequency Standards
Frequency standards in the infra-red and in the visible range of the electromagnetic
spectrum have been developed and used since decades. The most prominent use
is as wavelength standards in practical length metrology and for the realization of
the meter. Since the new definition of the meter became effective in 1983, this SI
unit should be realized according to a mise-en-pratique. One method is the use of
radiations whose wavelength in vacuum or whose frequency is stated in a list that
is periodically updated and that can be retrieved from the BIPM web-site and found
in the Appendix 2 of the BIPMs SI-brochure [23]. Detailed account of frequency
standards used for this purpose can, e.g., be found in [24] .
It has been predicted for quite some time that the performance - in terms of accu-
racy and frequency instability - of a laser as a frequency standard, when stabilized
to a suitable and narrow optical transition between a metastable state and the ground
state, might surpass that of a frequency standard in the radio-frequency region. Rea-
sons for that are at hand and were mentioned before (Sect. 1). The increase in Q-factor
leads to a much reduced frequency instability at short averaging times which paves
the way to explore systematically frequency shifting effects in a conveniently short
measurement time. In addition, the magnitude of systematic energy level shifts is
of the same order as in microwave atomic frequency standards, so that the relative
uncertainty is dramatically reduced.
Three ingredients for an optical frequency standard to become feasible have been
perfected during the last years: the required stable interrogation oscillator (clock
laser), the optical frequency comb for counting the cycles of optical frequencies
(glorified with the Nobel prize given to T. Hänsch in 2005 [25]), and confinement
of laser-cooled atomic species to a range whose dimensions are smaller than the
wavelength in radio-frequency or laser traps. The basic technology was described
in the textbooks [7, 24], and a detailed survey on optical frequency standards with
more than 200 references included was recently provided by Ludlow et al. [26].
12 A. Bauch
Two distinct kinds of optical frequency standards have been developed, differing in
the kind of trap used. Charged atoms (ions) can easily be trapped with electric fields
without significantly disturbing their atomic energy levels and thus the resonance
(clock) frequency. A “single ion at rest in free space” [27] represents the ultimate
isolation of a spectroscopic object and thus the smallest systematic frequency shifts.
But the frequency control of the clock laser is inevitably based on the (weak) signal
obtained from one ion only. Neutral atoms need to be trapped using forces exerted by
attacking the charge distribution inside the atom, which inevitably has an influence
on the electronic structure. Seminal work of H. Katori showed how to set up a trap
made from standing laser fields which shifts the two energy levels defining the clock
transition equally [28]. In such an optical lattice ensembles of several thousands of
atoms can be cooled and stored.
The storage and laser cooling methods are applicable to a great variety of atoms
and ions in different charge states. When selecting an atom for an optical clock, the
properties of the reference transition therefore play an important role as is described
in detail in [26]. The most significant results from trapped ion frequency standards
up to now have involved the ions 27 Al + , 40 Ca + , 88 Sr + , 115 I n + , 171 Y b+ (Q), 171 Y b+
(O) (see explanations below), and 199 H g + . Neutral-atom based frequency standards
employed 87 Sr , 88 Sr , 171 Y b, and 199 H g. Measurements of the optical transition fre-
quencies of these species with respect to caesium fountain clocks in the relative
uncertainty range of 1016 have been reported (except for 115 I n). In PTB, two dif-
ferent experiments (with the ions Y b+ and with Sr atoms) succeeded in reaching
this accuracy range. The 171 Y b+ possesses two suitable transitions, the 2 S1/2 - 2 D3/2
quadrupole transition at 436 nm (Q) and the 2 S1/2 - 2 F7/2 octupole transition (O) at
467 nm. At the time of writing, the octupole transition frequency can be realized
with an uncertainty of 3 × 10−18 [29], and the ratio between the two transition fre-
quencies is known with a relative uncertainty of about 10−16 . The uncertainty for the
realization of the clock transition frequency in 87 Sr was estimated as 2 × 10−17 , and
measurements of the frequency ratio 171 Y b+ (O)/87 Sr have been performed several
times during the last years. This kind of ratio measurements are instrumental in fun-
damental physics studies, as discussed at the end. A transportable variant of the Sr
optical frequency standard has been developed at PTB [30] and was recently used in
a kind of demonstration exercise for chronometric leveling. Publication of results is
pending.
This subsection was written as the last one of the manuscript, but nevertheless it
will inevitably fail to be up-to-date at the time of its publication. The rate of progress
is very high and the number of groups actively involved is quite large. Whether
1 × 10−18 relative uncertainty will have been reached already? I dare not make a
prediction.
4 International Time Scales and Their Local Realizations
International Atomic Time (Temps Atomique International, TAI) and Coordinated

Universal Time (UTC) are maintained and disseminated by the BIPM Time Depart-
ment. About 75 NMIs and astronomical and scientific laboratories that operate
atomic clocks of different kind and time transfer equipment are the players in the
activity. Each of them, designated “k”, realizes an approximation to UTC, denoted
UTC(k), which is used as the reference for local clock comparisons. Data of the
kind [U T C(k) − clocki (k)] are provided to the BIPM where i designates an indi-
vidual clock (total number involved in 2016 about 400). Time transfer using cal-
ibrated equipment provides the differences between the UTC(k) time scales with
ns-uncertainty (see next section), and respective data are provided to the BIPM as
well.
Almost all atomic clocks involved are commercial caesium clocks and hydrogen
masers, as described above. In addition, about 12 laboratories develop and maintain
nowadays caesium fountain clocks. By combining the clock and time transfer data
using the algorithm ALGOS, an averaged time scale, called free atomic time scale,
is calculated. The algorithm was designed to provide a reliable scale with optimized
frequency instability for one month averaging time [31]. Individual clocks contribute
with statistical weights that are based on their performance during the last 12 months.
In a second step, the relative departure of the scale of the free atomic time scale
from the SI second is determined from data of the primary frequency standards.
The departure is ideally brought to zero by a very gentle frequency adjustment, and
the resulting scale is TAI. In Fig. 5 the comparison of the 10 fountain clocks for
which data were made public during the last two years, including October 2016,
with TAI is shown. Ideally, the data points would scatter around zero. The dispersion
of the points mostly reflects the instability introduced by the comparison between the
respective fountain, its local intermediate reference (hydrogen maser, UTC(k)), and
TAI. The individual uncertainty contributions are reported in the monthly Circular T
and explained in an explanatory supplement for which a link is provided on the web
[32].
Some atomic transitions were recognized as so-called secondary representations
of the second by a Working Group of the Consultative Committee (CC) for Length
(CCL) and the CC for Time and Frequency [33]. The idea behind is to draw upon
stable and accurate frequency standards for the monitoring of TAI and at the same time
to prepare a solid data base for a decision about a future re-definition of the second.
The 6.8 GHz ground-state hyperfine transition frequency in 87 Rb is one of them. Data
from the French FO-2 have been reported in Circular T since 2012. In 2016 the first
data from an optical frequency standard having the transition 5s 2 1 S0 − 5s5 p 3 P0 in
87
Sr at 429 THz as reference were submitted by SYRTE / OP to the BIPM. It can be
expected that other research teams will follow this example during the coming years.
UTC, the final product, has the same scale unit as TAI, but differs from TAI
by an integer number of seconds, introduced as “leap seconds” on request from
the International Earth Rotation and References Systems Service (IERS) [8, 31].
14 A. Bauch
Fig. 5 Comparison of primary frequency standards with TAI during two years, ending MJD 57689
(2016-10-28), data taken from Circular T [32]. The standards are operated at Istituto Nazionale di
Ricerca Metrologica, Torino, Italy (IT), LNE-SYRTE / Paris Observatory, NIST, National Institute
of Metrology (NIM), Beijing, PR China, National Physical Laboratory (NPL), Teddington, UK,
PTB, and VNIIFTRI, Mendeleevo, Russia (SU)
Dissemination of UTC by the BIPM happens in the form of time series [U T C −

U T C(k)] for selected dates in the past month being published in the Circular T
[32]. The Circular T provides the traceability to international standards (time unit,
frequency, and time scale) for each participating institute. The NMIs in turn are
responsible for the dissemination of their UTC(k) in their respective countries and
thus provide traceability to calibration services, academic institutes, and to the com-
mon public.
For some applications the publication of Circular T only once per month with
the last reported value from typically more than ten days in the past was felt as
an inconvenience. The BIPM Time Department implemented a rapid realization of
UTC, called UTCr, which has been published every week since July 2013 [34].
UTCr gives daily values of [U T Cr − U T C(k)] for a subset of laboratories which
committed to submit data daily (and thus fully automatic) to the BIPM. The difference
[U T C − U T Cr ] is at the few-nanoseconds level, but inevitably not zero: the clock
ensemble is different (smaller) and also the time link data used are not the same
as for UTC generation. So typically once per month a small time step aligns UTCr
with UTC, which is considered as a nuisance for other applications for which a
monotonous [U T Cr − U T C(k)] would be desirable.
There is no strict rule or recommendation governing the offset between UTC and
UTC(k). For many years an offset of 100 ns was felt appropriate, more recently
several institutes strive to stay within 10 ns. End of October 2016 the (absolute)
difference [U T C − U T C(k)] was below 20 ns for 39 institutes. To stay within 10 ns
is definitely facilitated if the time scale is built from the average of a large ensemble
of atomic clocks - as it used to be the situation at USNO for many years and continues
to be today - or from a hydrogen maser steered towards a long-term stable, accurate
Fig. 6 Comparison of UTC

with UTC(PTB) (black) and
UTC(OP) (grey) during eight
years
reference, such as a fountain clock. PTB was the first institute to go this way [35],
starting in 2010. Figure 6 illustrates the success of this strategy that was adopted two
years later also by the French laboratory LNE-SYRTE / OP.
5 Satellite-Based Time and Frequency Transfer
The comparison of distant clocks has always been an important part of time metrol-
ogy. A comparison on a local and regional scale can be achieved with electrical
signals transported in cables. Unsurpassed accuracy could be demonstrated during
recent years by using optical fibers to transport either stabilized laser radiation or
modulated laser signals [36, 37] even over 1000 km distances (see the contribution
by G. Grosche in these proceedings.). On a global scale, however, the use of radio
signals from or via satellites remains the first choice [38, 39]. Subsequently, two
satellite-based methods are presented, the reception of signals of Global Naviga-
tion Satellite Systems (GNSS), and Two-Way Satellite Time and Frequency Transfer
(TWSTFT).
5.1 GNSS-Based Time and Frequency Transfer
The primary purpose of all Global Navigation Satellite Systems (GNSS) is to serve as
a positioning and navigation system. But each system relies on accurate timing, more
precise, the satellite ranges used to calculate position are derived from propagation
time measurements of the signals transmitted from each satellite in view. The sig-
nals broadcast by GNSS satellites are derived from onboard atomic clocks (caesium
beam, rubidium gas cell frequency standards, passive hydrogen masers). Details of
16 A. Bauch
signal properties and on-board configuration of the existing GNSS are inter alia well
explained in [39] and in text books on GNSS, e.g. [40]. A brief description of the
usage of the signals follows, adapted to GPS and the European counterpart Galileo,
and noting that the Russian system GLONASS differs in many details [39]. The
GNSS carriers are phase modulated with pseudo-random noise codes (PRN-codes).
These are binary codes with different chip codes, unique for each satellite. All satel-
lites transmit their signals on the same frequencies. A receiver generates a local copy
of the PRN-code derived from its internal oscillator. This local copy is electronically
shifted in time by δt and correlated with the incoming antenna signal. If the received
satellite PRN-coded signal and the shifted replica match, the receivers tracking loops
can lock to the satellite signal. When this has happened data, usually called the navi-
gation message, can be read by the receiver, reporting the almanac, orbit parameters
and parameters that refer the individual satellite clock to the underlying GNSS time.
In this contribution we neglect the type of receiver that has de facto the widest
general use, but that is inappropriate for accurate clock comparisons. It combines
the recorded value δt with the navigation message to discipline the frequency of its
inbuilt quartz oscillator to GNSS time and delivers standard frequency output and
a one pulse per second electrical signal (1 PPS) representing GNSS time. Called
GNSS disciplined oscillator (GNSSDO)3 it is the common instrumentation in cali-
bration laboratories, industry, wherever such signals are needed. Another variety of
such instruments outputs the time-of-day information, converted from the navigation
message, either in a clock display, in standard electrical time codes like IRIG, or for
distribution in the Internet or in local area networks using the Network Time Protocol
(NTP).
In what might be called “scientific” timing receivers, the measured time offset δt
for each satellite in view with respect to the local reference signal connected to the
receiver is stored. The information contained in the navigation message is used to
provide output data in the form of local reference (local time scale) minus GNSS
time. Modern receivers are capable to measure also the phase difference between the
received carrier signal and the local reference once code lock had been established.
Such GNSS carrier phase measurements are two orders of magnitude more precise
than the code data. Code and carrier phase measurement results are usually output
in the so-called receiver independent exchange format (RINEX) [41]. The current
version that is adapted to the multitude of GNSS and transmitted signals is 3.03, but
older versions (GPS + GLONASS only) are still common.
Precise point positioning (PPP) is a code and carrier phase-based analysis tech-
nique that has become very popular and most often combines GPS observations at
the two transmit frequencies f 1 = 1575.42 MHz (L1) and f 2 = 1227.60 MHz (L2).
PPP builds on the precise satellite orbits and clock products generated by the Inter-
national GNSS Service IGS (see www.igs.org). The position of the antenna of an
isolated GPS receiver is provided by PPP with high accuracy on a global scale. At
the same time, the difference between local reference clock and IGS time, a time
3 For improved hold-over capability, some models include a Rb frequency standard (see Sect. 3.1)
that is steered with a few hours time constant.
scale generated by IGS, is calculated and reported in the output data. A software
package in frequent use has been developed by National Resources Canada and the
software has been generously made available to several timing laboratories for local
installations, and an online service is also available [42].
Code-based time transfer in the popular common-view (CV) method has been used
already for decades and has still its merits. It is built upon simultaneous reception of
the transmitted signal from the same satellite by two receivers on Earth. Thereby the
impact of common errors in the GPS signals caused by errors in the satellite position,
instabilities of the satellite clocks, and the effects of the intentional degradation
(known as “selective availability”) that had been applied to the GPS signal until May
2000 are strongly reduced. Receivers of the first generation used for time comparison
were single-channel, single-frequency (L1) receivers.4 The propagation of GNSS
signals are affected by atmospheric effects. The ionosphere provokes delays that
can be modeled on a global scale only to a limited extent. Substantial errors occur,
particularly during periods of high solar activity and when the receiver is at low
latitude. As the ionosphere shows dispersion, group velocity and phase velocity are
affected with opposite sign and depend on the carrier frequency. This property is
used in advanced receivers that receive and process signals on both frequencies f 1
and f 2 to determine the ionospheric delay in situ. Data generated in this way are
labeled as L3P-data.
With increasing availability and accuracy of IGS products, the common-view
method has almost been replaced by GPS all-in-view (AV), which is in practice
simpler to implement. After exchange of the (standardized [43]) data files among the
laboratories, the individual observation data are corrected for the above mentioned
effects based on IGS products before averages over convenient intervals are formed.
Subtraction of corresponding data allows the comparison of the local time scales or
frequency standards. Comparisons within Europe practically give the same results in
CV and AV, even without the use of external products. AV is, however, particularly
useful in intercontinental comparisons and thus widely used today by BIPM in its
undertaking to realize TAI. Directives on a common format and standard formulae
and parameters for code-based data evaluation were developed jointly by the BIPM
and the CCTF [43].
Figure 7 illustrates the advantage of dual-frequency reception in a comparison
between PTB Braunschweig and IMBH Sarajevo, Bosnia-Herzegovina. Data collec-
tion happened during 2014 in support of the operations establishment and receiver
calibration of the time laboratory at the Bosnian NMI. L3P data are more noisy on
short averaging times, but are free from daily variations seen in the single frequency
data. Daily patterns can be explained with insufficient modeling of the ionosphere.
4 As an aside, I remember well that my start as PhD student at PTB in 1983 coincided with the
installation of the first receiver of this kind in the PTB time-unit laboratory, a single-channel GPS
receiver for time transfer provided by the then National Bureau of Standards (now NIST). This
receiver is now at display in the Deutsche Uhrenmuseum (German clock museum), Furtwangen,
Black Forest.
18 A. Bauch
Fig. 7 GPS CV time comparisons between the NMI of Bosnia-Herzegovina, IMBH, Sarajevo, and
PTB; open symbols: L1C single frequency data, full symbols: dual frequency L3P data obtained
from the same receivers. The offset is due to incomplete delay calibration of the IMBH receiver at
the time of data taking in 2014
Fig. 8 GPS comparison T = U T C(O P) − U T C(P T B), evaluated by BIPM, GPS P3 (black)
and PPP (grey) data (left), frequency comparison instability (right)
The second example,5 shown as Fig. 8 contrasts L3P (code-based) and PPP eval-
uation of data collected with receivers at Observatoire de Paris and PTB during
October 2016. The kink in the PPP data around the middle of the period points to an
interruption of continuous recording of observations at one (or both) receivers. The
PPP solution then continues with a new estimate of the integer phase offset which is
affected by the more noisy code-based measurements. The instability (kink removed)
achieved with both methods converges at averaging times exceeding one day which
reflects the instability of the two time scales that are compared (see also Fig. 6).
The so-called Modified Allan variance (modσ y ) is a variant of the Allan variance
5 Source of data: BIPM ftp server at ftp://ftp2.bipm.org/pub/tai/timelinks/lkc/, data in monthly files,
Read Me file provided.

explained in Sect. 2 [4]. In general, it has been found that frequency comparisons
between distant clocks using PPP links show an instability of about 1 × 10−15 at
1 day averaging time and a few 10−16 at 5–10 days averaging.
Joint work of BIPM and the French Centre National d’Etudes Spatiales (CNES)
has led to improved capabilities of PPP for longer averaging times than one day. Their
approach is called Integer PPP (IPP) and avoids the treatment of a priori unknown
number of carrier cycles between the space craft antenna and the receiver on Earth
(the “phase ambiguities”) as floating point numbers as it is done in the “standard”
PPP [44]. They demonstrated comparisons with an instability of 2 × 10−15 at 5 h
averaging and 10−16 at 4 days. The data analysis is currently still quite laborious,
but work is ongoing to make IPPP a useful tool, potentially also in the context of
relativistic geodesy. As a demonstration of IPPP performance, a comparison between
two fountains during 60 days could reveal whether the 10−17 region could be reached.
But this has not happened yet.
5.2 Two-Way Satellite Time and Frequency Transfer
Two-way satellite time and frequency transfer (TWSTFT) is based on the exchange of
signals between two active terminals A and B. Signals propagate between A and B in
both directions, simultaneously, and at each terminal the time-of-arrival of the signal
from the other side is recorded. All propagation delays cancel to first order when the
two measurement results are combined to provide the time difference between the
clocks connected to the two terminals. To connect terminals on Earth, TWSTFT is
made using fixed satellite services in the Ku-band and the X-band, and geo-stationary
telecommunication satellites serve as relay [38, 39, 45]. There are still several effects
causing non-reciprocities which are discussed in detail in the literature [46]. Most
of the propagation related effects can, however, account for small non-reciprocities
only, typically not exceeding 0.1 ns, depending on the geometry of the locations of
the stations and the satellite, and the transmission frequencies.
A brief description of the established services follows. Pseudorandom noise (PRN)
binary phase-shift keying (BPSK) modulated carriers are transmitted. The phase
modulation is synchronized with the local clock’s 1 PPS output. Each station uses a
dedicated PRN code for its BPSK sequence in the transmitted signal. The receiving
equipment is capable to generate the BPSK sequence of each remote station and to
reconstitute a 1 PPS tick from the received signal. This is measured by a time-interval
counter (TIC) with respect to the local clock. Following a pre-arranged schedule, both
stations of a pair lock on the code of the corresponding remote station for a specified
period, measure the signal’s time of arrival, and store the results. After exchanging
the data records the difference between the two clocks is computed. Within Europe,
both stations are within the same antenna footprint of the satellite, and signals are
routed through the same transponder electronics. In this favourable case, the link
delays can be calibrated and time transfer with uncertainty of 1 ns or even slightly
below is feasible. Satellites rarely have an antenna footprint that is wide enough
20 A. Bauch
Fig. 9 TWSTFT
comparisons
UTC(PTB)-UTC(k) during
September 2016, looking at
the plots from top
downwards: “k” = INRIM
(Torino, Italy), NIST,
SYRTE/OP and USNO
to cover both Europa and US or Europe and Asia, respectively, so that the above
condition is not fulfilled. Here another method has to be used, and for many links
between the Europe, the US and Asia GPS data have been used for delay calibration.
TWSTFT has as well proven to provide a relative uncertainty for frequency transfer
of about 1 × 10−15 at averaging times of one day [47]. TWSTFT is therefore used in
the international network of time keeping institutions supporting the realization of
TAI [32].
On the other hand, admittedly, TWSTFT as used today has some weakness, as
can be identified by looking at some time transfer results shown in Fig. 9. Each of
the data points represents the result of a two-minute data collection of time scale
comparison between PTB and a remote institute, two in Europe and two in the US.
Nominally there are 12 such measurements per day per link. In the plots we note
different levels of noise, and apparently systematic variations with daily period, the
strength of which is not constant with time. A lot of studies went into the cause of such
“diurnals”, and the above mentioned non-reciprocities were suspected as causes. But
recently evidence was found that they are likely caused by the receiving electronics
which cannot always cope with the changing receive frequency. It is modulated in
a daily rhythm due to the classical Doppler effect proportional to the (small, order
m/s) line-of-sight velocity of the geostationary satellite with respect to the receiving
antenna. This effect and a good part of measurement noise could be suppressed or
reduced if a larger phase-shift keying rate would be used. This would spread the
signal power in a wider band, but would require a larger portion of the transponder
bandwidth reserved - and paid for - for the application. But the cost is currently
prohibitive to establish a continuous all-year service of such kind.
In addition to the routine comparisons with institutes in Europe and the USA, PTB
supported recently two experiments aiming at improvements of the performance of
TWSTFT links. One experiment involved European institutes in a collaboration
funded by the European Commission, and consisted of the transmission/reception
of signals with a 20-fold wider spectral distribution in the Ku-band region for a
few weeks, thereby circumventing some limitations mentioned before. The results
showed a good part of the expected reduction in short-term measurement noise, but no
significant improvement in the long term. The other one was managed and evaluated
by National Institute of Communications Technology (NICT), Japan. Here it could
be demonstrated that transmission of signals in a quite small spectral band, but using
the carrier phase as the measurement quantity provides frequency comparison with
less than 10−15 instability when averaging longer than 20 000 s at quite favorable
operational cost, which is considered as important as the performance itself. Research
on that subject is ongoing.
6 Look Ahead and Conclusions
The “Atomic Clock Ensemble in Space” (ACES) unfortunately remains “on the
horizon” only. The launch of the scientific mission that relies on the availability
of the International Space Station (ISS) experienced once more delay into 2018.
The ACES project will involve a space segment and a complex ground segment
[48, 49]. The space segment will comprise PHARAO (Projet d’Horloge Atomique
par Refroidissement d’Atomes en Orbite), a primary frequency standard based on
laser-cooled caesium atoms, and an active hydrogen maser. An on-board time scale
will be generated that should reflect the short-term instability of the maser and the
long-term characteristics and accuracy of PHARAO. Connection to the ground is
going to be provided by the so-called Microwave Link (MWL) and the European
Laser Timing (ELT) space terminals. All that will be installed on the Columbus
External Payload Facility. The MWL follows the principle of TWSTFT, now between
space and ground, but the transmitted signals are going to occupy a hundred times
wider bandwidth than those used routinely in these days. The MWL shall be used
to compare the space clocks with high-performance ground clocks at seven sites
worldwide where ground terminals are going to be installed. The ACES campaign
of 18 months duration shall be used for a couple of fundamental physics studies and
international clock comparisons. The projected frequency transfer capabilities are
competitive with the uncertainty of a few optical frequency standards and could be
used to verify the relativistic red shift of the frequency standards at the sites at the
10−17 level.
In general, satellite-based time and frequency transfer serves plenty of applica-
tions, and in particular with the advent of new freely available signals and newer
modulation schemes on GNSS some improvement can be expected. But it seems
unlikely that the gap in performance (stability) between optical frequency standards
and the GNSS comparison techniques can be significantly reduced. The GNSS all-
in-view technique is unique: it allows comparisons among laboratories wherever
they are located on Earth, and installation of receiver and antenna is quite simple.
TWSTFT can bridge approximately 10 000 km because both sites must simulta-
neously be in the field of view of the same satellite. The synchronization of the
ground stations of the deep space tracking networks maintained by NASA and ESA
22 A. Bauch
has practically to rely on GNSS comparisons as they are separated by about 120
degree in longitude on the globe. For sure, TWSTFT offers high potentials in terms
of achievable measurement noise and accuracy, but it remains open to see who is
going to pay the bill.
“More accurate clocks - What are they needed for?” has been partially answered
in [50], and many publications demonstrate the interdependence of metrology and
fundamental science. The research into atomic frequency standards and time and
frequency dissemination can help improving our understanding of the laws of physics
in general. Without any pretension of completeness let me first mention pulsar timing
as a fascinating branch of radio-astronomy [51]. Although I doubt that pulsar time
scales [52] are going to have properties adequate for replacing atomic time scales, we
remember that pulsar timing was the basis for the indirect proof of the existence of
gravitational radiation emitted by the binary system of two neutron stars [53, 54]. The
emission of gravitational waves was predicted and the accompanying orbital energy
loss became observable as an orbital period change. A very active field of study in
these days is search for variations of fundamental constants. Several parameters that
are usually designated as “constants”, such as charge and mass of the electron as well
as the fine structure constant α are predicted to vary on cosmological time scales,
and laboratory searches of the - if at all - tiny temporal variations today involve
atomic frequency standards [55]. As explained in [55], atomic transition frequencies
depend in a different way on these constants. Measurement of the ratio of atomic
transition frequencies of different atomic species (or of transition frequencies in the
unit Hertz as realized with caesium fountain clocks) repeatedly over time allows
under certain circumstances to determine limits on the temporal variations in our
days. The uncertainty of such ratio measurements got lower and lower with time
during recent years, mostly due to the dramatic improvement in the performance of
optical frequency standards, as laid out in Sect. 3.3. Just to give one example, at the
time of writing (April 2017), the tightest limit of variation of α is in the low 10−18
per year, and not statistically significant [56]. Complemented by searches involving
astrophysical data, such laboratory searches may in the future point to new physics
beyond the standard model.
Disclaimer
The mentioning of individual products and their manufacturers is not to be understood
as endorsement by PTB. Data obtained at PTB reflect the properties of the selected
equipment and its installation conditions and may deviate from observations made
at other sites.
Acknowledgements This review paper reports mostly on achievements of colleagues from all over
the world. The fruitful collaboration belonged to the pleasures of the author’s business life. Special
thanks go to Ekkehard Peik, Dirk Piester and Stefan Weyers of PTB for critical reading of the
manuscript.
References
1. A. Bjerhammar, On relativistic geodesy. Bull. Geod. 59, 207 (1985)

2. H. Denker et al., Geodetic methods to determine the relativistic redshift at the level of 10−18 in
the context of international timescales - a review and practical results. J. Geod. 92, 487 (2018)
to be published
3. Standard definitions of Physical Quantities for Fundamental Frequency and Time Metrology,
IEEE-Std 1189–1988 (IEEE, 1988)
4. W.J. Riley, Handbook of Frequency Stability Analysis, NIST Special Publication, vol. 1065
(NIST, 2008)
5. Joint Committee for Guides in Metrology, Evaluation of Measurement Data - Guide to the
Expression of Uncertainty in Measurement. JCGM 100:2008 (2008)
6. J. Vanier, C. Audoin, The Quantum Physics of Atomic Frequency Standards (Adam Hilger,
Bristol, 1989)
7. J. Vanier, C. Tomescu, The Quantum Physics of Atomic Frequency Standards - Recent Devel-
opments (CRC Press, Taylor & Francis Group, Boca Raton, 2016)
8. C. Audoin, B. Guinot, The Measurement of Time (Cambridge University Press, Cambridge,
2001)
9. E.F. Arias, A. Bauch, Metrology of time and frequency, in Handbook of Metrology ed. by M.
Gläser, M. Kochsiek (WILEY-VCH Verlag, Weinheim, 2010), p. 315
10. R. Lutwak et al., The MAC - a miniature atomic clock. Proceedings 2005 International Fre-
quency Control Symposium and Exhibition (2005), p. 752
11. P. Forman, Atomichron®: the atomic clock from concept to commercial product. Proc. IEEE
73, 1181 (1985)
12. L.S. Cutler, Fifty years of commercial caesium clocks. Metrologia 42, S90 (2005)
13. A. Bauch, The PTB primary clocks CS1 and CS2. Metrologia 42, S43 (2005)
14. N.F. Ramsey, Experiments with separated oscillatory fields and hydrogen masers (Nobel Lec-
ture). Rev. Mod. Phys. 62, 541 (1990)
15. R. Piriz et al., The time validation facility (TVF): an all-new key element of the Galileo
operational phase. Proceedings of the Joint IEEE International Frequency Control Symposium
and the European Frequency and Time Forum, Denver (2015), p. 320
16. H.J. Metcalf, P. van der Straten, Laser-Cooling and Trapping (Springer, New York, 1999)
17. R. Wynands, S. Weyers, Atomic fountain clocks. Metrologia 42, S64 (2005)
18. V. Gerginov et al., Uncertainty evaluation of the caesium fountain clock PTB-CSF2. Metrologia
47, 65 (2010)
19. B. Lipphard, V. Gerginov, S. Weyers, Optical stabilization of a microwave oscillator for fountain
clock interrogation. IEEE Trans. Ultrason. Ferroelectr. Freq. Control. 99, 99 (2016)
20. G. Petit, G. Panfilo, Comparison of frequency standards used for TAI. IEEE Trans. Intrumen-
tation Meas. 62, 1550 (2013)
21. N.K. Pavlis, M.A. Weiss, The relativistic redshift with 3 × 10−17 uncertainty at NIST, Boulder,
Colorado, USA. Metrologia 40, 66 (2003)
22. S. Peil et al., Evaluation of long term performance of continuously running atomic fountains.
Metrologia 51, 263 (2014)
23. BIPM, Le Système international d’unités (The International System of Units), 8th edition.
BIPM, Pavillon de Breteuil, F-92312 Sèvres Cedex, France, 2006, 2014 update)
24. F. Riehle, Frequency Standards: Basics and Applications (Wiley VCH, Weinheim, 2006)
25. T.W. Hänsch, Passion for precision (Nobel lecture). Rev. Mod. Phys. 78, 1297 (2006)
26. A.D. Ludlow et al., Optical atomic clocks. Rev. Mod. Phys. 87, 637 (2015)
27. H. Dehmelt, Coherent spectroscopy on single atomic system at rest in free space. J. Phys.
(Paris) 42, C8–299 (1981)
28. H. Katori, Spectroscopy of Strontium atoms in the Lamb-Dicke confinement, in Proceedings of
the 6th Symposium on Frequency Standards and Metrology (2001). ((St. Andrews, Scotland),
the proceedings were published by World Scientific, Singapore, in 2002, p. 323)
24 A. Bauch
29. N. Huntemann et al., Single-ion atomic clock with 3 × 10−18 systematic uncertainty. Phys.
Rev. Lett. 116, 063001 (2016)
30. S.B. Koller et al., Transportable optical lattice clock with 7 × 10−17 uncertainty. Phys. Rev.
Lett. 118, 073601 (2017)
31. B. Guinot, E.F. Arias, Atomic time-keeping from 1955 to the present. Metrologia 42, S20
(2005)
32. BIPM Time Department. Circular T. BIPM, http://www.bipm.org/en/bipm-services/
timescales/time-ftp/Circular-T.html
33. P. Gill, F. Riehle, On secondary representations of the second, in Proceeding of the 20th
European Frequency and Time Forum (2006), p. 282
34. G. Petit et al., UTCr: a rapid realization of UTC. Metrologia 51, 33 (2014)
35. A. Bauch et al., Generation of UTC(PTB) as a fountain-clock based time scale. Metrologia 49,
180 (2012)
36. C. Lisdat et al., A clock network for geodesy and fundamental science. Nat. Commun. 7, 12443
(2016)
37. Ł. Śliwczyński et al., Fiber-optic time transfer for UTC-traceable synchronization for telecom
networks. IEEE Commun. Stand. Mag. 1, 66 (2017)
38. J. Levine, A review of time and frequency transfer methods. Metrologia 45, S162 (2008)
39. ITU Study Group 7, ITU Handbook: Satellite Time and Frequency Transfer and Dissemination
(International Telcommunication Union, Geneva, 2010)
40. E.D. Kaplan, C.J. Hegarty (eds.), Understanding GPS, Principles and Applications, 2nd edn.
(Artech, Boston, London, 2006)
41. RINEX Working Group and RTCM-SC104, RINEX The Receiver Independent Exchange For-
mat Version 3.02. International GNSS Service (2013), https://igscb.jpl.nasa.gov/igscb/data/
format/rinex302.pdf
42. J. Kouba, P. Heroux, Precise point positioning using IGS orbit and clock products. GPS Solut.
5(2), 12 (2002)
43. P. Defraigne, G. Petit, CGGTTS-Version 2E: an extended standard for GNSS time transfer.
Metrologia 52, G1 (2015)
44. G. Petit et al., 10−16 frequency transfer by GPS PPP with integer ambiguity resolution. Metrolo-
gia 52, 301 (2015)
45. D. Kirchner, Two-way satellite time transfer. Review of Radio Science 1996–1999 (1999), p. 27
46. A. Bauch, D. Piester, M. Fujieda, W. Lewandowski, Directive for operational use and data
handling in two-way satellite time and frequency transfer (TWSTFT). BIPM, Rapport 2011/01,
2011 (2011)
47. A. Bauch et al., Comparison between frequency standards in Europe and the USA at the 10−15
uncertainty level. Metrologia 43, 109 (2006)
48. L. Cacciapuoti, Chr. Salomon, Space clocks and fundamental tests: the ACES experiment. Eur.
Phys. J. Spec. Top. 172, 57 (2009)
49. L. Cacciapuoti, Chr. Salomon, Atomic clock ensemble in space. J. Phys. Conf. Ser. 327, 012049
(2011)
50. E. Peik, A. Bauch, More accurate clocks - What are they needed for? PTB-Mitteilungen 119(2),
16 (2009)
51. I.H. Stairs, Testing general relativity with pulsar timing. Living Rev. Relativ. (2003), http://
relativity.livingreviews.org/open?pubNo=lrr-2003-5&=node5.html
52. G. Hobbs, Development of a pulsar-based time-scale. Mon. Not. R. Astron. Soc. 427, 2780
(2012)
53. R.A. Hulse, The discovery of the binary pulsar (Nobel Lecture). Rev. Mod. Phys. 66, 699
(1994)
54. J.H. Taylor, Binary pulsars and relativistic gravity (Nobel Lecture). Rev. Mod. Phys. 66, 711
(1994)
55. N. Huntemann et al., Improved limit on a temporal variation of m p /m e from comparisons of
Y b+ and Cs atomic clocks. Phys. Rev. Lett. 113, 210802 (2014)
56. C. Lisdat, Private communication (2017)
Chronometric Geodesy: Methods
and Applications
Pacome Delva, Heiner Denker and Guillaume Lion
1 Introduction
The theory of general relativity (GR) was born more than one hundred years ago,
and since the beginning has striking prediction success. Einstein proposed three
effects for its experimental verification, all verified shortly after their prediction:
the perihelion precession of Mercury’s orbit, the deflection of light by the Sun,
and the gravitational redshift of spectral lines of stars. Other predictions from GR
had to wait decades before being confirmed experimentally. It is only in 1959 that
the gravitational redshift is confirmed in a laboratory experiment by Pound, Rebka
and Snider [1–4]. Two gamma-ray emitting iron nuclei at different heights were
compared, verifying GR prediction with a relative accuracy of 10% (and later <1%).
In parallel, the era of atomic time began in 1955 with the caesium frequency standard
built by Essen and Parry at the National Physical Laboratory (NPL) [5, 6]. Since then,
the accuracy and stability of atomic clocks were constantly ameliorated, with around
one order of magnitude gained every ten years (see Fig. 1).
In this context, the unit of time of the International System of Units (SI), the
second, was officially defined with respect to a specified hyperfine transition of the
caesium atom in the year 1967–1968.1 Moreover, as local atomic timescales were
1 Resolution 1 of the 13th General Conference on Weights and Measures (CGPM) [7].
P. Delva (B)
SYRTE Observatoire de Paris, Université PSL, CNRS, Sorbonne Université,
LNE 61 avenue de l’Observatoire, 75014 Paris, France
e-mail: pacome.delva@obspm.fr
H. Denker
Institut für Erdmessung, Leibniz Universität Hannover (LUH),
Schneiderberg 50, 30167 Hannover, Germany
G. Lion
LASTIG LAREG IGN, ENSG, Univ Paris Diderot, Sorbonne Paris Cité,
35 rue Hélène Brion, 75013 Paris, France

26 P. Delva et al.
Fig. 1 Accuracy records for microwave and optical clocks. From the first Cs clock by Essen
and Parry in the 1950’s, an order of magnitude was gained every ten years. The advent of optical
frequency combs boosted the performances of optical clocks, and they recently surpassed microwave
clocks
developed thanks to commercial caesium clocks [8] as well as laboratory caesium

standards, there was a need to compare these different timescales in order to build
a mean international atomic time, which could be adopted by everyone. This was
done at the beginning by the United States Naval Observatory (USNO) and the
Bureau International de l’Heure (BIH) with radio time signals, which allowed time
comparisons with uncertainties of the order of 1 ms [9]. A jump in accuracy occured
with the first demonstration from the Hewlett-Packard company of the possibility
of using commercial flights to transport its caesium clocks, allowing time transfer
with an uncertainty of around 1 µs [9]. This eventually led to the famous test of
Hafele and Keating [10, 11], who flew four caesium beam clocks around the world
on commercial jet flights during October 1971. They predicted and measured the
desynchronization of the proper times of these commercial atomic clocks with respect
to the USNO atomic scale, and thus verified the gravitational redshift effect with an
accuracy of around 12%.2
2 Thisis based on the numbers given in table 1 of [10] and table √ 1 of [11]: the relative accuracy of
the gravitational part of the relativistic shift effect is taken as ( 182 + 102 + 72 ns)/(179 ns).
Chronometric Geodesy: Methods and Applications 27
Following the Hafele and Keating experiment, Briatore and Leschiutta did the first
experimental measurements of the gravitational redshift with a direct comparison
of ground caesium beam atomic frequency standards [12]. The two clocks were
separated in heights by h = 3250 m, predicting a desynchronization of t/t ≈
gh/c2 ≈ 30.6 ns d−1 , where g and c are the local gravity and the velocity of light
in vacuum. The measurement gave (36.5 ± 5.8) ns d−1 , giving a relative accuracy of
around 20%. Now, we can say that this experiment is amongst the first demonstrations
of chronometric levelling (see Sect. 2.4): the clock comparison measured a difference
of altitude between the two clocks of (3880 ± 620) m, to be compared with the
otherwise measured value of 3250 m.
Now that the atomic clock accuracy reaches the low 10−18 in fractional frequency
(see Fig. 1), and can be compared to this level over continental distances with optical
fibres (see Sect. 3.3), the accuracy of chronometric levelling reaches the cm level
and begins to be competitive with classical geodetic techniques such as geometric
levelling and GNSS/geoid levelling. Moreover, the building of global timescales
requires now to take into account these effects to the best possible accuracy. It is
the topic of this chapter to explain how atomic clock comparisons and the building
of timescales can benefit from the latest developments in physical geodesy for the
modelization and realization of the geoid, as well as how classical geodesy could
benefit from this new type of observable, which are clock comparisons that are
directly linked to gravity potential differences.
In Sect. 2 we introduce fundamental concepts of GR concerning the measurement
of time, relativistic reference systems and we review the recent literature of chrono-
metric geodesy. In Sect. 3 we introduce the theory of frequency standard comparisons,
beginning with the Einstein equivalence principle, followed by the description of the
frequency techniques, and finally, we describe clock syntonization and the realization
of timescales. Section 4 describes the geodetic methods for determining the gravity
potential, namely the geometric levelling approach and the GNSS/geoid approach, as
well as considerations about the uncertainties of these methods. In Sect. 5 we describe
the European project International Timescales with Optical Clocks (ITOC) where
unified relativistic redshift corrections were determined for several atomic clocks
in European national metrology institutes. Finally, in Sect. 6 we present numerical
simulations exploring what could be the contribution of clock comparisons for the
determination of the geoid.
2 The Relativistic Framework
2.1 Observers and the Spacetime Manifold
The theoretical background of chronometric geodesy is general relativity (GR). In

GR theory, space and time are bound together in a continuous entity named space-
time. Spacetime geometry specifies how matter and energy behave, while matter
28 P. Delva et al.
and energy distribution tells how spacetime geometry is curved. This is a non-linear
process and the link between geometry on one side, and matter/energy on the other
side is given by the Einstein equations. Gravitation is no longer a force as in New-
tonian theory, but the manifestation of the variation of the background geometry.
Variations of spacetime can be induced by a choice of coordinates, causing inertial
effects, which act in a way similar but not equivalent to gravitation. The presence
of energy/matter gives rise to curvature of the background geometry, and therefore
gravity. However, gravitational effects can never be completely disentangled from
inertial effects.
A spacetime is formally described by a four-dimensional differentiable manifold
M endowed with a pseudo-Riemannian metric g. A point of the manifold is called
an event. Let us define an open subset U ∈ M and an event P in this open subset. A
chart or coordinate system {x α }α=0...3 can be defined in U; it maps point P ∈ U to a
point {x αP } ∈ R4 . The four real numbers x αP are the coordinates of event P. Generally
a coordinate system cannot be defined on the whole manifold. An atlas is a collection
of charts with some properties, which cover the whole manifold.
As the manifold is smooth, the difference vector dx between two infinitesimally
close events may be defined. Then each event P is associated to a vector space
T P (M), called the tangent space, which contains the set of all possible four-vectors
dx . A basis {eα } of the tangent space is usually called a frame. The introduced
coordinates induce a coordinate basis eα = {∂ α } P ≡ {∂/∂ x α } P . However, a frame
does not need to be associated with any coordinate system.
The metric tensor g is a symmetric bilinear scalar function of two vectors. Given
two vectors v and w, the metric tensor returns a scalar called the dot product:
= v · w
g(v , w) =w · v = g(w,
v ). The metric can be characterized by its action
on a basis of the tangent space. For example, gαβ ≡ ∂ α · ∂ β are the components of
the metric tensor in the natural frame associated with the coordinate system {x α }.
The infinitesimal interval
ds 2 ≡ dx · dx = gαβ dx α dx β (1)
between two neighboring events is invariant under coordinate transformation. Trajec-

tories of observers are defined by worldlines C, which are parameterized 1-D curves
in the manifold M.
One of the most striking consequences of GR is the fact that coordinates do
not have a direct physical interpretation as in the Newtonian theory. Indeed, one
has to distinguish proper quantities from coordinate quantities. A proper quantity
is the result of a physical measurement in a real or Gedanken experiment. It is
mathematically described by a scalar, a quantity which is invariant under general
coordinate transformations. However, a scalar quantity is not necessarily a physical
measurement. The latter needs to be defined with respect to a reference frame adapted
to the observer. The proper time τ of a clock is implicitly defined with the relation
ds 2 = −c2 dτ 2 , (2)
where c is a constant velocity characterizing the spacetime, which can be identified

with the velocity of light in vacuum [13].
Proper time is invariant under general coordinate transformations. The proper
times of two different clocks can be compared by means of a time transfer technique,
while their proper frequencies can be compared with frequency transfer techniques
(see Sect. 3). Time and frequency comparisons are linked, but both approaches lead
to different formalisms and experimental techniques. Usually time transfer is more
challenging, as it necessitates the knowledge of many instrumental delays with accu-
racy, while they can often be neglected in the frequency transfer.
When the spatial separation of two clocks is much smaller than the typical length of
the background curvature of spacetime, then curvature effects can safely be neglected.
This is a consequence of the Einstein equivalence principle (see e.g. [14]). This type
of measurement will be termed local comparison of clocks. On the contrary, if the
distance between both clocks is of the order or bigger than the typical length of the
background curvature, the result of the comparison will have curvature perturbations,
which depend on both the locations of the clocks and the particular time or frequency
transfer technique. This type of measurement will be termed non-local or distant
comparison of clocks.
The relation between proper time and coordinate time can be deduced from
Eqs. (1)–(2):
1
dτ = −dx · dx . (3)
c
Let us integrate this relation along the worldline C : x α = f α (λ) parameterized by

λ, between two events A and B belonging to C (see Fig. 2). The associated tangent
˙ ≡ d/dλ, and dx = v dλ, such that:
vector is v = f˙α ∂ α , where ()
B λB
1
τ (A, B) ≡ dτ = dλ −gαβ f˙α f˙β . (4)
A c λA
The parameter λ is usually chosen as the coordinate x 0 = ct in the context of rel-

ativistic time and frequency transfer. It is clear from this formula that the proper
time elapsed between two events A and B depends on the worldline C, i.e. on the
trajectory of the observer between these two events. One consequence of GR is that
the parameter λ cannot be adapted such that both proper and coordinate time be equal
everywhere. This is possible only in special relativity where there is no curvature in
a well chosen reference system.
2.2 Simultaneity and Synchronization
Let us define an observer O with trajectory C parameterized by its proper time τ ,

and an event M which does not belong to C (see Fig. 2). How can we define an event
on C which is simultaneous with event M. This is possible in the Newtonian space
30 P. Delva et al.
Fig. 2 Illustration of the

Einstein synchronization
convention
which is Euclidean, and therefore time is absolute, i.e. independent from the observer.
However, in GR proper time is defined only along the worldline of the observer and
is not a global property of spacetime. Only the light cone is a fundamental element
of a given spacetime. The light cone is the collection of vectors v ∈ T P (M) which
satisfies v · v = 0. It is independent of the observer and divides the tangent space
in three parts, past and future containing time vectors which satisfy v · v < 0, and a
third part containing space vectors which satisfy v · v > 0. It is supposed here that the
metric tensor has a signature (−, +, +, +), i.e. at least one basis of the tangent space
exists for which v · w = −v 0 w 0 + v 1 w 1 + v 2 w 2 + v 3 w 3 , where v , w
∈ T P (M).
With the notion of the light cone, spacetime can be time-oriented, but it does not
say which set of events can be considered simultaneous. Indeed in GR simultaneity
can only be conventional and not an intrinsic property of spacetime. Einstein has
suggested an operational definition of simultaneity. Suppose that an observer O is
equipped with a clock and a system to send and receive electromagnetic signals. A
signal is sent at event A ∈ C, received and reflected with no delay at event M and
finally received at event B ∈ C (see Fig. 2). The proper times along C corresponding to
events A and B, respectively τ A and τ B , are measured with the clock. By convention,
the event M which is simultaneous to M along the observer trajectory corresponds
to the proper time:
1 1
τ= (τ A + τ B ) = τ A + (τ B − τ A ) . (5)
2 2
Fig. 3 Einstein
synchronization convention
is not transitive: events B
and C are defined as
simultaneous with A thanks
to the convention, while D is
defined as simultaneous to
C. However, events B and D
do not coincide
Fig. 4 Slow clock transport

This convention is usually called Einstein synchronization. It is a geometrical con-

vention based on the concept of light cones, and an operational convention based on
the exchange of electromagnetic signals. However this convention is not transitive.
This is illustrated in Fig. 3. Events B and C are simultaneous to A (with Einstein
synchronization); D is simultaneous to C, but D and B events do not generally coin-
cide. Therefore this convention is not practical to define global timescales such as
the TAI (Temps Atomique International). This problem was discussed in [15–17] in
the context of satellite clock synchronization. However, in these articles the problem
was thought as “synchronization errors”, but it was in fact well understood in the
context of GR, as noted in [18–20].
Another convention is the slow clock transport synchronization. Let us define
three clocks A, B and M with corresponding worldlines C A , C B and C M (see Fig. 4).
Clocks A and M are compared locally at event A0 such that clock M proper time is
set to τ M = τ0M = τ0A at this event. Then clock M goes toward clock B and crosses
worldline C B at event B1 , where τ M = τ1M and τ B = τ1B . The event B0 on C B is
defined simultaneous to event A0 on C A by the slow clock transport synchronization
convention with:
τ0B = lim [τ1B − (τ1M − τ0M )] , (6)
v→0
32 P. Delva et al.
Fig. 5 Coordinate
where v is the coordinate velocity of clock M. The limit condition of null veloc-
ity is not feasible in a real experiment. Therefore, operationally, this convention
depends on the particular trajectory of the mobile clock M, and reaches a differ-
ent synchronization than the Einstein synchronization convention, as shown in [16].
However, in special relativity, i.e. with a null background curvature, it can be shown
that both synchronization conventions are equivalent. If the spacetime geometry and
the mobile clock trajectory are sufficiently known, then in the weak-field and low
velocity approximation it is possible to use the coordinate time of clock M at events
A0 and B1 instead of its proper time, so that the convention will not depend on the
particular trajectory of the mobile clock. However it will depend on the relativistic
coordinate system chosen to calculate the coordinate time. The inaccuracy of the
time transfer operated with this convention can be assessed with the closing relation:
τ0A = limv→0 [τ1A − (τ1M − τ0M )].
Finally, we define the coordinate synchronization convention: two events P1 and
P2 of coordinates {x1α } and {x2α } are considered to be simultaneous if the values
of their time coordinates are equal: x10 = x20 (see Fig. 5). This definition follows
the definition of simultaneity adopted in special relativity in [21, 22]. It is conve-
nient to introduce three-dimensional hypersurfaces with constant time coordinate t:
t ≡ {P ∈ M, x 0P = ct}. By choosing a particular relativistic reference system we
introduce a conventional foliation of spacetime, giving the hypersurfaces of simul-
taneity. The synchronization of clocks with this convention obviously depends on the
chosen reference system. It is the most commonly used convention for the building of
timescales such as TAI and Global Navigation Satellite System (GNSS) timescales.
Indeed, this convention is very similar to what is well known in Newtonian physics
where the foliation of spacetime is absolute. For this convention to become opera-
tional it is necessary to define conventional relativistic reference systems. In special
relativity, for clocks which are at rest with respect to an inertial reference system, the
Einstein synchronization is a convenient procedure to achieve coordinate synchro-
nization of clocks.
It was proposed in [23] to build a global “coordinate time grid” on and around
the Earth, therefore realizing the idea of coordinate synchronization convention for
clocks, without any problem of transitivity. The authors proposed to take as a refer-
ence a clock on the geoid,3 i.e. to choose a conventional reference system R such
that the proper time of a clock at rest on the geoid coincides with its coordinate time
in R. We will see later that this choice is convenient because it implies a simple
link between the relativistic correction of a clock, in order to realize the coordinate
time synchronization, and its altitude. The authors in [23] detailed several opera-
tional methods of time transfer using the coordinate synchronization convention:
portable clocks, one-way and two-way synchronization with electromagnetic sig-
nals. The same authors in [24] estimated the main limitation on the determination of
coordinate time: the knowledge of the geoid.
It is interesting to note that the question of synchronization of clocks in non-
inertial reference systems raised a controversy in the 80’s, driven by the development
of Navstar Global Positioning System (GPS) and the need for a global timescale
on Earth. This is reviewed in [25], where the author concludes: “In principle, the
curved Schwarzschild space cannot be imbedded in a four-dimensional flat space
without the addition of more dimensions. Thus the theoretical basis for the GPS
navigational scheme would appear to be flawed, and a new algorithm would have to
be constructed”. Indeed, coordinate time synchronization can only be theoretically
realized in approximation schemes, e.g. post-Newtonian approximation as reviewed
in [26] for GPS. A different relativistic approach to this problem has been initiated
in [27], where the idea is to give to a constellation of satellites the possibility to
constitute by itself a primary and autonomous positioning system, without any need
for synchronization of the clocks. Such a relativistic positioning system is defined
with the introduction of emission coordinates, which have been re-introduced by
several authors in the context of navigation systems [28–36].
A resolution concerning the global “coordinate time grid” was proposed by
N. Ashby at the International Astronomical Union (IAU) Symposium No. 114,
reported in [37]:
1. To adopt the coordinate time system (as approved by the Consultative Committee
for the Definition of the Second (CCDS) and the International Radio Consultative
Committee (CCIR)) as a global time scale for the Earth;
2. To continue further investigations for the determination and adjustment of the
International Atomic Time (TAI) and the Terrestrial Dynamic Time (TDT).
3 In the Newtonian sense, the geoid is the equipotential of the Earth’s gravity (Newtonian) potential,
which best coincides with the (mean) surface of the oceans.

34 P. Delva et al.
The resolution was not adopted, but the chairman of the Scientific Organizing Com-
mittee, J. Kovalesky, considered that specialists in Celestial Mechanics and Astrom-
etry needed more time to study the problem in competent commissions of IAU.
Following this Symposium, several authors have contributed to the definition of
global coordinate times [38–42].
As the definition of coordinate timescales necessitates the definition of a four
dimensional relativistic reference system, the IAU working group had several com-
plementary tasks in hand (Resolution C2 of the IAU General Assembly in 19854 ):
1. the definition of the Conventional Terrestrial and Conventional Celestial Refer-
ence Systems,
2. ways of specifying practical realizations of these systems,
3. methods of determining the relationships between these realizations, and
4. a revision of the definitions of dynamical and atomic time to ensure their consis-
tency with appropriate relativistic theories
Moreover, the President of the International Association of Geodesy (IAG) was
invited to “appoint a representative to the working group for appropriate coordina-
tion on matters relevant to Geodesy”. This work eventually led to the set of IAU
Resolutions in 1991 and 2000 that define the present reference systems.
2.3 Relativistic Reference Systems
Several approaches have been considered for the definition of relativistic reference
systems. Generalised Fermi coordinates were considered in [43–45]. However, the
use of Fermi coordinates is not adapted to self-gravitating bodies for which mass-
energy contributes to the determination of the initial metric g when solving the
Einstein equations. For this reason, harmonic coordinates are preferred and rec-
ommended for the definition and realization of relativistic celestial reference sys-
tems [46–50], where the frame origin can be centered at the center-of-mass of a
massive body. One drawback of harmonic frames is that the harmonic gauge condi-
tion does not admit rigidly rotating frames [51, chapter 8]. Other recent approaches
are based on a perturbed Schwarzschild metric [52], or on the Kerr metric [53] in the
different context of a slowly rotating astronomical object.
Following the pioneering works, a set of Resolutions was adopted at the IAU
General Assembly in Manchester in the year 2000 [54]:
• B1.3: definition of the Barycentric and Geocentric Celestial Reference Systems
(BCRS and GCRS)
• B1.4: form of the Earth post-Newtonian potential expansion
4 All IAU Resolutions can be found at http://www.iau.org/administration/resolutions/

general_assemblies/.
• B1.5: time transformations and realization of coordinate times in the Solar System
(uncertainty < 5 × 10−18 in rate and 0.2 ps in phase amplitude for locations farther
than a few Solar radii from the Sun)
• B1.9: definition of Terrestrial Time (TT)
We summarize here very briefly these resolutions. A relativistic reference system
is implicitly defined by giving the components of the metric tensor in this reference
system, in addition to a conventional spatial origin and orientation for the spatial
part of the frame, and a conventional time origin for the time coordinate (the time
orientation is trivial). The metric tensor is a solution of the Einstein equations in
the low velocity and weak gravitational field approximation, for an ensemble of N
bodies.
The Solar System Barycentric Celestial Reference System (BCRS), recommended
by the IAU Resolutions, can be used to model light propagation from distant celestial
objects and the motion of bodies within the Solar System. It is defined with:
2w(x ) 2w(x )2
g00 = −1 + c2
− c4
, (7)
g0i = − c43 wi (x ) , (8)

2w(x )
gi j = δi j 1 + c2
, (9)
where x ≡ {ct, x i }, with i = 1 . . . 3, w and wi are scalar and vector potentials. Its
origin is at the barycenter of the Solar System masses, while the orientation of the spa-
tial axes is fixed up to a constant time-independent rotation matrix about the origin
(a natural choice is the International Celestial Reference System (ICRS) orienta-
tion which is fixed with respect to distant quasars). The coordinate time t is called
Barycentric Coordinate Time (TCB). The origin of TCB is defined with respect to
TAI: its value on 1977 January 1, 00:00:00 TAI (JD = 2,443,144.5 TAI) must be 1977
January 1, 00:00:32.184.
The unit of measurement of TCB should be chosen so that it is consistent with the
SI second. An interesting discussion about timescale units can be found in [55]. As
coordinate times such as TCB are not proper times, they cannot be directly measured
by clocks. They are calculated using the corresponding metric components, e.g.
Eqs. (7)–(9) for TCB, in combination with Eq. (4), which has to be inverted. Indeed,
the basic observables to build timescales are the readings of proper times on clocks,
which are local experiments. If the clocks are realizing the SI second, then the
timescales calculated from these measurements are also in SI units, and the unit of
such time coordinate could be named “SI-induced second”.
The second relativistic reference system, recommended by the IAU Resolutions,
is the Geocentric Celestial Reference System (GCRS). It can be used to model phe-
nomenon in the vicinity of the Earth, such as its gravity field, artificial satellites
orbiting the Earth or Earth rotation. It is defined with:
36 P. Delva et al.

2V ( X) 2
2V ( X)
G 00 = −1 + c2
− c4
, (10)
i
G 0i = − c43 V ( X) , (11)

G i j = δi j 1 + 2Vc(2X) , (12)
≡ {cT, X i }, and V and V i are scalar and vector potentials. Note that we
where ( X)
use notation V instead of usual notation W because W is commonly used in geodesy
for the gravity potential. The frame origin is at the centre of mass of the Earth, and
the orientation of the spatial axes is fixed with respect to the spatial part of the BCRS.
The coordinate time T is called Geocentric Coordinate Time (TCG). It has the same
origin and unit as TCB.
TCG is the proper time of a clock at infinity, and is not convenient because
its rate differs from the one of clocks on the ground. Therefore IAU Resolu-
tions introduced Terrestrial Time (TT), which differs from TCG by a constant rate
L G = 6.969290134 × 10−10 :
d(TT)
= 1 − LG. (13)
d(TCG)
The origins of TT and TCG are defined so that they coincide with TCB in origin:
TT (resp. TCG) = TAI + 32.184 s on 1977 January 1 st, 0 h TAI. TT is a theoretical
timescale and can have different realizations, e.g. TT(BIPM), or TT(TAI) = TAI +
32.184 s. (see e.g. [56]).
2.4 Chronometric Geodesy
Chronometric geodesy is the use of clocks to determine the spacetime metric. Indeed,
the gravitational redshift effect discovered by Einstein must be taken into account
when comparing the frequencies of distant clocks. Instead of using our knowledge
of the Earth’s gravitational field to predict frequency shifts between distant clocks,
one can revert the problem and ask if the measurement of frequency shifts between
distant clocks can improve our knowledge of the gravitational field. To do simple
orders of magnitude estimates it is good to have in mind some correspondences:
ν
1m↔ ∼ 10−16 ↔ W ∼ 10 m2 s−2 , (14)
ν
where 1 m is the height difference between two clocks, ν is the frequency difference
in a frequency transfer between the same two clocks, and W is the gravity potential
difference (see Sect. 4.1) between the locations of these clocks.
From this correspondence, we can already recognize two direct applications of
clocks in geodesy: if we are capable of comparing clocks to 10−16 accuracy, we can
determine height differences between clocks with one meter accuracy (levelling), or
determine geopotential differences with 10 m2 s−2 accuracy.
To the knowledge of the authors, the latter technique was first mentioned in the
geodetic literature by Bjerhammar [57] within a short section on a “new physi-
cal geodesy”. Vermeer [58] introduced the term “chronometric levelling”, while
Bjerhammar [59] discussed the clock-based levelling approach under the title “rel-
ativistic geodesy”, and also included a definition of a relativistic geoid. The term
“chronometric” seems well suited for qualifying the method of using clocks to deter-
mine directly gravity potential differences, as “chronometry” is the science of the
measurement of time. However the term “levelling” seems to be too restrictive with
respect to all the applications one could think of using the results of clock compar-
isons. Therefore we will use the term “chronometric geodesy” to name the scientific
discipline that deals with the measurement and representation of the Earth, includ-
ing its gravity field, with the help of atomic clocks. It is sometimes also named
“clock-based geodesy”, or “relativistic geodesy”. However this last designation is
improper as relativistic geodesy aims at describing all possible techniques (including
e.g. gravimetry, gradiometry, VLBI, Earth rotation, …) in a relativistic framework.
The natural arena of chronometric geodesy is the four-dimensional spacetime. At the
lowest order, there is proportionality between relative frequency shift measurements
– corrected from the first order Doppler effect – and (Newtonian) gravity potential dif-
ferences. To calculate this relation one does not need the theory of general relativity,
but only to postulate Local Position Invariance. Therefore, if the measurement accu-
racy does not reach the magnitude of the higher order terms, it is perfectly possible
to use clock comparison measurements – corrected for the first order Doppler effect
– as a direct measurement of (differences of) the gravity potential that is considered
in classical geodesy. Comparisons between two clocks on the ground generally use
a third reference clock in space, or an optical fibre on the ground (see Sect. 3.3).
In his article, Martin Vermeer explores the “possibilities for technical realisation
of a system for measuring potential differences over intercontinental distances” using
clock comparisons [58]. The two main ingredients are, of course, accurate clocks
and a mean to compare them. He considers hydrogen maser clocks. For the links he
considers a 2-way satellite link over a geostationary satellite, or GPS receivers in
interferometric mode. He has also considered a way to compare proper frequencies
of the different hydrogen maser clocks. Today this can be overcome by comparing
primary frequency standards (PFS, see Sect. 3.2), which have a well defined proper
frequency based on the transition of Caesium 133 used for the definition of the
second. Secondary frequency standards (SFS), i.e. standards based on a transition
other than the defining one, may also be used if the uncertainty in systematic effects
has been fully evaluated, and the frequency measured against PFS.
With the advent of optical clocks, it often happens that the evaluation of system-
atics can be done more accurately than for PFS. This was one of the purposes of
the European project5 of “International timescales with optical clocks” [60], where
optical clocks based on different atoms are compared to each other locally, and to
5 projects.npl.co.uk/itoc.
38 P. Delva et al.
PFS. Within this project, a proof-of-principle experiment of chronometric geodesy

was done by comparing two optical clocks separated by a height difference of around
1000 m, using an optical fibre link [61].
Few authors have seriously considered chronometric geodesy in the past. Follow-
ing Vermeer’s idea, the possibility of using GPS observations to solve the problem
of determining geoid heights has been explored in [62]. The authors considered two
techniques based on frequency comparisons and direct clock readings. However,
they leave aside the practical feasibility of such techniques. The value and future
applicability of chronometric geodesy has been discussed in [63], including direct
geoid mapping on continents and joint gravity-geopotential surveying to invert for
subsurface density anomalies. They find that a geoid perturbation caused by a 1.5 km
radius sphere with 20 percent density anomaly buried at 2 km depth in the Earth’s
crust is already detectable by atomic clocks with an achievable accuracy of 10−18 .
The potentiality of the new generation of atomic clocks has been shown in [64],
based on optical transitions, to measure heights with a resolution of around 30 cm.
The possibility of determining the geopotential at high spatial resolution thanks to
chronometric geodesy is thoroughly explored and evaluated in [65]. This work will
be detailed in Sect. 6.
2.5 The Chronometric Geoid
Arne Bjerhammar in 1985 gives a precise definition of the “relativistic geoid”

[59, 66]:
The relativistic geoid is the surface where precise clocks run with the same speed and the
surface is nearest to mean sea level.
This is an operational definition, which has been translated in the context of post-
Newtonian theory [47, 67]. In these two articles a different operational definition
of the relativistic geoid has been introduced based on gravimetric measurements: a
surface orthogonal everywhere to the direction of the plumb-line and closest to mean
sea level. The authors call the two surfaces obtained with clocks and gravimetric
measurements the “u-geoid” and the “a-geoid”, respectively. They have shown that
these two surfaces coincide in the case of a stationary metric. In order to distin-
guish the operational definition of the geoid from its theoretical description, it is
less ambiguous to give a name based on the particular technique to measure it. The
term “relativistic geoid” is too vague as Soffel et al. [67] have defined two different
ones. The names chosen by Soffel et al. are not particularly explicit, so instead of
“u-geoid” and “a-geoid” one can call them “chronometric geoid” and “gravimetric
geoid” respectively. There can be no confusion with the geoid derived from satellite
measurements, as this is a quasi-geoid that does not coincide with the geoid on the
continents [68]. Other considerations on the chronometric geoid can be found in [51,
69, 70].
Fig. 6 A photon of frequency ν A is emitted at point A toward point B, where the measured frequency
is ν B . a) A and B are two points at rest in an accelerated frame, with acceleration a in the same
direction as the emitted photon. b) A and B are at rest in a non accelerated (locally inertial) frame
in presence of a gravitational field such that g = −
a
We notice that the problem of defining a reference isochronometric surface is

closely related to the problem of realizing Terrestrial Time (TT). This is developed
in more details in Sect. 3.5.
Recently, extensive work has been done aiming at the development of an exact
relativistic theory for the Earth’s geoid undulation [71], as well as developing a theory
for the reference level surface in the context of post-newtonian gravity [72, 73]. This
goes beyond the problem of the realization of a reference isochronometric surface
and tackles the tough work of extending all concepts of classical physical geodesy
(see e.g. [68]) in the framework of general relativity.
3 Comparisons of Frequency Standards
3.1 The Einstein Equivalence Principle
Let’s consider a photon emitted at a point A in an accelerated reference system toward

a point B, which lies in the direction of the acceleration (see Fig. 6). We assume that
both points are separated by a distance h 0 , as measured in the accelerated frame.
The photon time of flight is δt = h 0 /c, and the frame velocity during this time
increases by δv = aδt = ah 0 /c, where a is the magnitude of the frame acceleration
a . The frequency at point B (reception) is then shifted because of the Doppler effect,
compared to the frequency at point A (emission), by an amount:
νB δv ah 0
=1− =1− 2 . (15)
νA c c
40 P. Delva et al.
Now, the Einstein Equivalence Principle (EEP) postulates that a gravitational

field g is locally equivalent to an acceleration field a = −
g . We deduce that in a
non-accelerated (locally inertial) frame in presence of a gravitational field g:
νB gh 0
=1− 2 , (16)
νA c
where g = | g |, ν A is the photon frequency at emission (strong gravitational poten-

tial) and ν B is the photon frequency at reception (weak gravitational potential). As
ν B < ν A , it is usual to say that the frequency at the point of reception is “red-shifted”.
One can consider it in terms of conservation of energy. Intuitively, the photon that
goes from A to B has to “work” to be able to escape the gravitational field, then it
looses energy and its frequency decreases by virtue of E = hν, with h the Planck
constant.
If two ideal clocks are placed in A and B and the clock at A (strong gravitational
potential) is used to generate the signal ν A , then the signal received at B (weak
gravitational potential) has a lower frequency than a signal locally generated by the
clock at B.
3.2 Relativistic Frequency Transfer
Let’s consider two atomic frequency standards (AFS) A and B which deliver the
proper frequencies f A and f B . These two frequencies can be different if the two
AFS are based on different atom transitions. Following the Bureau International des
Poids et Mesures (BIPM) we name primary frequency standards (PFS) the AFS
based on the atom of Caesium 133, more commonly named Caesium Fountains.
The best PFS have a very low relative accuracy in the range 10−15 – 10−16 (see
e.g. [74]). Then, we name secondary frequency standards (SFS) the AFS which are
based on a different atom than the Caesium atoms. The Consultative Committee for
Length (CCL)-Consultative Committee for Time and Frequency (CCTF) frequency
standards working group is in charge of producing and maintaining a single list of rec-
ommended values of standard frequencies for the practical realization of secondary
representations of the second.6 SFS can have a relative accuracy down to the range
10−17 – 10−18 [75–77]. See also [78, 79], where a method is presented for analysing
over-determined sets of clock frequency comparison data involving standards based
on a number of different reference transitions.
The goal of a frequency comparison between two AFS A and B is to determine the
ratio of their frequencies f A / f B . The most used technique for frequency comparison
nowadays is the transmission of an electromagnetic signal between A and B, reaching
the following formula:
fA f A ν A νB
= , (17)
fB ν A νB f B
6 See http://www.bipm.org/en/publications/mises-en-pratique/standard-frequencies.html.
where ν A is the proper frequency of the photon at the time of emission t A , and ν B
is the proper frequency of the same photon at the time of reception t B . The ratio
ν A / f A is known or measured, ν B / f B is measured, while ν A /ν B has to be modelled
and calculated.
Let S(x α ) be the phase of the electromagnetic signal emitted by clock A. It can be
shown that light rays are contained in hypersurfaces of constant phase. The frequency
measured by A/B is:
1 dS
ν A/B = , (18)
2π dτ A/B
where τ A/B is the proper time along the worldline of clock A/B. We introduce the
A/B
wave vector kα = (∂α S) A/B to obtain:
1 A/B α
ν A/B = k u A/B , (19)
2π α
where u αA/B = dx αA/B /dτ is the four-velocity of clock A/B. Finally, we obtain a
fundamental relation for the frequency transfer:
νA k Auα
= αB αA . (20)
νB kα u B
This formula does not depend on a particular theory, and thus can be used to perform
tests of general relativity. It is needed in the context of chronometric geodesy in order
to calculate the gravity potential difference between two clocks for which the ratio
f A / f B is well known.
Introducing v i = dx i /dt and k̂i = ki /k0 , it is usually written as:
k̂iA v iA
νA u0 k A 1 +
= 0A 0B c
. (21)
νB u B k0 1 + k̂iB v iB
c
From Eq. (18) we deduce that:

−1
νA dτ B dτ dτ dt B
= = . (22)
νB dτ A dt A dt B dt A
The derivative (dt B /dt A ) is affected by processes in the frequency transfer itself
and depends on the particular technique used for the frequency comparison. It is
considered in more details in Sect. 3.3.
The derivatives (dτ/dt) in (22) do not depend on the frequency transfer technique
but just on the state (velocity and location) of the emitting and receiving AFS. In
Sects. 4 and 5 we focus on the best practical determination of these terms. Indeed,
calculation of these terms is limited in accuracy by our knowledge of the Earth’s
42 P. Delva et al.
gravitational field. We note that quantity (ν A /ν B ) in (22) is a scalar, therefore invariant

under general coordinate transformation (see Sect. 2.1). However the splitting of
this quantity as written on the right-hand side of (22) is not invariant and depends
on the particular relativistic reference system used for the splitting. The choice of a
particular relativistic reference system gives a conventional meaning to simultaneity:
two events are simultaneous if they have the same time coordinate t (see Sect. 2.2)
for free and guided propagation.
For applications on the Earth, such as (ground) clock syntonization (Sect. 3.4)
and the realization of a worldwide coordinate time (Sect. 3.5), a natural choice of
a relativistic reference system is the spatial part of the geocentric celestial refer-
ence system (GCRS) together with the terrestrial time (TT) as a coordinate time
(see Sect. 2.3). Following [80, 81], the coordinate to proper time transformation can
be written down to a relative accuracy of 10−18 as:
dτ 1
= 1 + L G − 2 W static + W temp . (23)

dT T c
where L G is a constant defined in Sect. 2.3, and W = V + Z is the gravity (gravita-
tional plus centrifugal) potential, commonly used in geodesy (see e.g. [68, 82]). The
gravity potential is split into a static part W static and a part varying with time W temp .
Neglected terms in Eq. (23) are terms in c−4 or smaller as well as one term of order
c−2 resulting from the coupling of higher order multipole moments of the Earth to the
external tidal gravitational field. All the neglected terms in the transformation (23)
amount in the vicinity of the Earth to a few parts in 10−19 or less [80].
The static part of the gravity potential W static can be derived from geometric
levelling or GNSS positions combined with a gravimetric geoid model. For instance,
the best unified evaluation of the static gravity potential for several AFS in Europe
was one of the main purposes of the ITOC project (see Sect. 5). The time varying
part W temp is dominated by solid Earth and ocean tides and further discussed in Sect.
3.6.
3.3 Frequency Transfer Techniques
We discuss in this section the foundations of two frequency transfer techniques

widely used, based on the propagation of an electromagnetic signal either in free
space or in an optical fiber. Free and guided propagation lead to different theoretical
modelling approaches of the frequency transfer. We limit the presented results to
one-way transfer and give appropriate references for two-way transfer techniques.
Free space time and frequency transfer can be realized using radiofrequency sig-
nals (of order 1–10 GHz) with well established techniques [83], and in the opti-
cal domain with lasers [84]. GNSS (Global Navigation Satellite Systems) [85–89]
and TWSTFT (Two-Way Satellite Time and Frequency Transfer) [83, 90–93] have
been widely used for years to perform clock comparisons and establish international
timescales such as TAI [56]. The ACES MWL (MicroWave Link) [94] is being
developed in the frame of the ACES (Atomic Clocks Ensemble in Space) experi-
ment [95, 96]. New techniques using two-way laser links have been developed and
operated, such as T2L2 (Time Transfer by Laser Light) [97–101], and others are in
development, such as ELT [102, 103], which is part of the ACES experiment.
Existing free space frequency transfer techniques are in the range 10−15 –10−16
for the fractional frequency accuracy and stability, with the goal of being in the 10−17
range for the ACES experiment. However, they are not sufficient for the comparisons
of optical clocks, which have fractional frequency accuracy and stability in the 10−17 –
10−19 range [76, 77, 104–107]. Therefore, phase-coherent optical links have been
developped using principally an optical fibre as a medium for the propagation [74,
75, 108, 109], attaining spectacular stability and accuracy in the range 10−19 and
below. However, phase coherent free space optical links are also being developed
[110–113]. It is not clear yet if these techniques will be able to be as good as optical
fibre techniques, mainly because of the effect of atmospheric turbulence [114–116].
3.3.1 Free Space Propagation Comparisons
In the case of propagation in free space, if we suppose that the spacetime is stationary,
i.e. ∂0 gαβ = 0, then it can be shown that k0 is constant along the light ray, meaning
that k0A = k0B . Then, from Eqs. (21) and (22) we deduce that
k̂iA v iA
dt B 1+
= c
. (24)
dt A k̂iB v iB
1+ c
The quantity dt B /dt A in Eq. (22) can be computed with several methods. Two
different approaches are presented in some detail in Appendix A of [117]: a direct
integration of the null geodesic equations, and a simpler way, which is the differ-
entiation of the time transfer function. This second method is quite powerful: a
general method has been developed to calculate the time transfer function as a Post-
Minkowskian (PM) series up to any order in G, the gravitational constant [118, 119].
See for example [120] for the calculation of the one-way frequency shift up to the 2-
PM approximation. This method does not require the integration of the null geodesic
equations. The frequency shift is expressed as integral of functions defined from the
metric and its derivatives and performed along a Minkowskian straight line.
Let A be the emitting station, with GCRS position x A (t), and B the receiving
station, with position xB (t). We use t = TCG and hence the calculated coordinate
time intervals are in TCG. The corresponding time intervals in TT are obtained by
multiplying with (1 − L G ). We denote by t A the coordinate time at the instant of
emission of an electromagnetic signal, and by t B the coordinate time at the instant of
reception. We define r A = | x A (t A )|, r B = |
x B (t B )| and R AB = |
x B (t B ) − x A (t A )|, as
well as the coordinate velocities v A = d x A /dt (t A ) and vB = d xB /dt (t B ). Then the
frequency ratio can be expressed as [117]:
44 P. Delva et al.

v 2B
νA 1− 1
c2 2
+ U E (
xB) q
A
= , (25)
νB 1− 1 v 2A
+ U E (
x A) q B
c2 2
where U E is the Newtonian potential of the Earth, and, if the desired accuracy is
greater than 5 × 10−17 ,
N AB ·
vA v A +(r A +r B ) N AB ·
4G M E R AB N A · vA
qA = 1 − c
− c3 (r A +r B )2 −R2AB
, (26)
N AB ·
vB v B −(r A +r B ) N AB ·
4G M E R AB N A · vB
qB = 1 − c
− c3 (r A +r B )2 −R2AB
, (27)
where N AB = ( x B (t B ) − x A (t A ))/R AB , G is the gravitational constant, and M E is

the mass of the Earth.
Note that formulas (26) and (27) have been obtained by assuming that the field of
the Earth is spherically symmetric. If an accuracy lower than 5 × 10−17 is required,
it is necessary to take into account the J2 terms in the Newtonian potential.
The terms of order c−1 correspond to the relative Doppler effect between the
clocks. Terms of order c−2 in Eq. (25) are the classic second-order Doppler effect
and gravitational redshift.7 Terms of order c−3 amount to less than 3.6 × 10−14 for
a satellite in Low-Earth Orbit and 2.2 × 10−15 for the ground. Terms of order c−4 ,
omitted in Eq. (25), can reach a few parts in 10−19 in the vicinity of the Earth [80].
3.3.2 Fibre Propagation Comparisons
If the signal propagates in an optical fibre, the term (dt B /dt A ) has been calculated
up to order c−3 in [122] for one-way and two-way time and frequency transfers. The
result for one-way frequency transfer is:

dt B 1 L
∂n ∂T 1 L
∂ v · sl
=1+ + nα dl + 2 dl , (28)
dt A c 0 ∂t ∂t c 0 ∂t
where L is the total rest length of the fibre at time of emission t A , n is the effective
refractive index of the fibre, α is the linear thermal expansion coefficient of the fibre,
T is the temperature of the fibre as a function of time and location, and v and sl are
the velocity and tangent vector fields of the fibre, respectively.
Up to the second order it does not depend on the gravitational field, as for the free
propagation in vacuum. The first order term is due to the variation of the fibre length
(e.g. due to thermal expansion) and of its refractive index. For a 1000 km fibre with
refractive index n = 1.5 this term is equal to 2 × 10−13 , but this term cancels in a
two-way frequency transfer. The second order term is the derivative of the Sagnac
7 One can notice that the separation between a gravitational red-shift and a Doppler effect is specific
to the chosen coordinate system. One can read the book by Synge [121] for a different interpretation
in terms of relative velocity and Doppler effect only.
effect, which is of order 10−19 or less for a 1000 km fibre. The sign of this term
depends on the direction of propagation of the signal in the fibre, such that it adds
up when doing two-way frequency transfer. Finally the neglected third order term is
of the order of 10−22 for a 1000 km fibre.
3.4 Clock Syntonization
Clock syntonization necessitates to calculate the derivatives (dτ/dt) from (22).

Using (23) and neglecting all terms smaller than 10−18 we deduce:
−1
dτ dτ WA − WB
=1+ , (29)
dT T A dT T B c2
where W = W static + W temp . Therefore syntonization necessitates the knowledge of

the difference of the gravity potential between locations A and B. Two widely used
geodesy techniques can be used to determine the static part of this difference: geo-
metrical levelling and GNSS positions combined with a gravimetric geoid model.
Geometrical levelling has the advantage to be very accurate on short distances (typ-
ically 0.2–1.0 mm for a 1 km double run levelling) and should be preferred when
comparing clocks within the same institute (local comparison). However, geomet-
rical levelling accumulates errors with increasing distance (up to several dm over
1000 km distance) and hence the GNSS/geoid approach should be preferred for
comparisons between different institutes.
Direct geometrical levelling between two points A and B does not necessitate
a point of reference and leads to a high accuracy. However, when determining the
height of the clocks with respect to the national height system, the reference point of
the zero altitude can be very far away from the clocks and therefore the link to the zero
altitude may lead to a bias in the determination of the height of the clock. Moreover,
the reference point of the zero altitude can be different from one country to the other,
because it can be based on different realizations of mean sea level. This leads to
the problem of unifying national height systems [123]. The GNSS/geoid approach
allows the derivation of the height system bias term for a particular country. It is
therefore possible to correct for the bias in the geometrical levelling technique for
international clock comparisons. However long distance errors cannot be avoided in
geometrical levelling for distant comparisons, for which the GNSS/geoid approach
is more adapted.
The GNSS/geoid method is based on the assumption that the gravitational poten-
tial is regular (zero) at infinity. This has the advantage that when using one gravimetric
model, the zero origin of the gravitational potential is coherent between all locations
covered by the model. High quality regional models exist for Europe and a new one
was developed during the course of the ITOC project (EGG2015, see Sect. 5.3).
Indeed, this technique is highly dependent on the quality and coverage of the ground
46 P. Delva et al.
gravimetric observables, and particular care should be taken in the use of the gravime-
try dataset. This method allows the derivation of absolute potential values with about
2–3 cm accuracy in terms of heights (best case scenario, i.e., accurate GNSS posi-
tions, sufficient terrestrial data around sites of interest, and state-of-the-art global
satellite geopotential utilized). Detailed considerations about the uncertainties of the
two approaches, geometric levelling and GNSS/geoid, can be found in [81].
3.5 The Realization of Terrestrial Time (TT)
The realization of TT necessitates the knowledge of the absolute gravity potential.

The TAI is the most commonly used realization of TT [56, 124]. First, comparisons
of about 400 atomic clocks around the world in around 70 laboratories are used
to calculate the free atomic scale (EAL), a fly-wheel timescale. In a second step,
around 15 AFS are used to steer the unit of EAL such that its scale corresponds to the
definition of the second. Direct comparisons between AFS are not necessary in this
process. Instead, each laboratory compares its AFS to a (master) clock participating
in EAL.
In 1980, the definition of TAI was given by the Consultative Committee for the
Definition of the Second as:
TAI is a coordinate time scale defined in a geocentric reference frame with the SI second as
realized on the rotating geoid as the scale unit.
This reference to the geoid was very ambiguous. Indeed, the value of the gravity
potential on the geoid, Wgeoid , depends on the global ocean level which changes with
time.8 In addition, there are several methods to realize the geoid as “closest to the
mean sea level” so that there is yet no adopted standard to define a reference geoid
and Wgeoid value (see e.g. a discussion in [125]). Several authors have considered the
time variation of Wgeoid , see e.g. [126, 127], but there is some uncertainty in what is
accounted for in such a linear model. A recent estimate by Dayoub et al. over 1993–
2009 gives dWgeoid /dt = −2.7 × 10−2 m2 s−2 yr−1 , mostly driven by the sea level
change of +2.9 mm/yr. However, the rate of change of the global ocean level could
vary during the next decades, and predictions are highly model dependent [128].
Nevertheless, to state an order of magnitude, considering a systematic variation in
the sea level of order 2 mm/yr, different definitions of a reference surface for the
gravity potential could yield differences in the redshift correction of the AFS of
order 2 × 10−18 in a decade, which is of the same order than the best current SFS
accuracies [76, 77].
However, this ambiguity disappeared with the new definition of TT adopted with
IAU resolution B1.9 (2000) [54] (see Sect. 2.3). If TAI is a realization of TT then
one has to apply a relative frequency correction, or redshift correction, to the AFS
8 Here we use notation Wgeoid instead of the commonly used W0 , in order to emphasize that there
is no generally accepted conventional and unified value of the geoid gravity potential value.
frequency such that

dτ
=1, (30)
dT T
This equation is exact in GR. Given a model of the spacetime metric and of the AFS
worldlines, it implicitly defines an isochronometric hypersurface, i.e. an hypersurface
where all clocks run at the same rate as TT. This hypersurface can be foliated using
TT coordinate time as a collection of 3D hypersurfaces T T with constant TT (see
Sect. 2.1). Following Eq. (23), the total correction (to be added) to the AFS relative
frequency in order to realize TT is:
W0(IAU) − W
=− , (31)
c2
where W0(IAU) = c2 L G = 62636856.000519 m2 s−2 , and W is the gravity potential at

the clock location for the considered epoch. We have seen that it is usual in geodesy
to separate the problem of modelization of the gravity field in a static part and a part
varying with time (see Eq. (23) and Sect. 4). This splitting is conventional and it
should be done with care as several conventions exist (see Sect. 5.2). Then the AFS
correction can be split in a static part and a part varying with time:
W0(IAU) − W static W temp

= static + temp = − 2
+ . (32)
c c2
Keeping only the static part of the gravity field, the problem becomes stationary
and the isochronometric hypersurface T T is uniquely determined for clocks fixed on
the Earth’s surface. In the weak gravitational field and low velocity limit, it coincides
at the lowest order with the Newtonian equipotential of the gravity field with exact
value W0(IAU) = c2 L G . Higher order relativistic corrections (terms of order c−4 in )
are of the order 2 × 10−19 or below in the AFS relative frequency [70].
We emphasize that the realization of TT does not necessitate any longer the real-
ization of a geoid. The reference equipotential is just an equipotential with a well
defined value, W0(IAU) , which is constant in time and exactly known: its value is a
convention. However if this reference equipotential is defined theoretically with no
ambiguity, it needs to be realized in the same way as the geoid, leading to inaccu-
racies in its realization, mainly due to the imperfect knowledge of the Earth’s mass
distribution. In this context, it is interesting to note that clocks in orbit around the
Earth are less sensitive to the Earth’s gravitational field, and thus to errors in its
modelization.
As an illustration, let’s take a clock in a satellite following an approximately
circular orbit of radius a around the Earth. Approximating the Earth gravitational
potential along the satellite trajectory with V = G M/a, where √ G M is the Earth
gravitational parameter, then the velocity of the clock is v = G M/a. In order for
the clock to realize TT, one needs V + v 2 /2 ≈ c2 L G , i.e. a ≈ 9543 km, and a good
knowledge of the trajectory of the satellite. It is shown in [80] that at this altitude the
48 P. Delva et al.
effect of solid Earth tides, ocean tide, polar motion, and changes in the atmospheric
pressure are below 10−18 in fractional frequency. Moreover, tidal effects can also be
calculated with uncertainties below 10−18 in fractional frequency.
The definition of the scalar potential in the context of relativistic reference frames,
from which the redshift correction formula (31) is deduced, is coherent in the New-
tonian limit with the assumption done in classical geodesy that the Newtonian poten-
tial is regular at infinity. Therefore the GNSS/geoid method is very well adapted to
the determination of the redshift corrections in the context of relativistic reference
frames. As discussed, when using national height systems one has to calculate cor-
rections such that the assumption of regularity is fulfilled over the area covered by
the clock comparisons. This will be illustrated in detail in Sect. 5.
Finally, according to Eqs. (29) and (31), syntonizing two AFS necessitates to
determine the relative gravity potential between the locations of both clocks, while
the realization of TT necessitates the determination of the absolute gravity potential
at the location of the contributing AFS. If the redshift correction (31) is known for
two clocks, it is easy to obtain Eq. (29) in order to syntonize them. Therefore, both
the problem of syntonization and the realization of TT can be tackled by determining
the absolute gravity potential at the locations of the contributing AFS.
3.6 Temporal Variations of the Gravity Field
For the temporal variations of the gravity field W temp , one can refer to [129], where
all corrections bigger than 10−18 in relative frequency are modelled and evaluated.
The dominant effect is the gravity potential variation induced by solid Earth tides,
which can be (at most) 10−16 for clock syntonization on international scales, and
10−17 for the realization of TT. The second major contributor is the induced signal
of ocean tides. However, both solid Earth and ocean tide signals can be modelled
down to an accuracy of a few parts in 1019 .
Several other time-variable effects can affect the clock comparisons at the 10−18
level, such as solid Earth pole tides, non-tidal mass redistributions in the atmosphere,
the oceans and the continental water storage, as well as secular signals due to sea
level changes and glacial isostatic adjustment. Non-tidal mass redistribution effects
on the gravity field are strongly dependent on location and/or weather conditions. As
clock comparisons now approach the 10−18 stability, it will be necessary to develop
guidelines in order to include these effects for the syntonization of clocks and their
contribution to the realization of TT. Recent analysis of optical clock comparisons
have included temporal variations [130, 131].
4 Geodetic Methods for Determining the Gravity Potential
This section describes geodetic methods for determining the gravity potential, needed
for the computation of the relativistic redshift corrections for optical clock observa-
tions. The focus is on the determination of the static (spatially variable) part of the
potential field, while temporal variations in the station coordinates and the poten-
tial quantities (with the largest components resulting from solid Earth and ocean
tide effects, see [129]) are assumed to be taken into account through appropriate
reductions or by using sufficiently long averaging times. This is common geodetic
practice and leads to a quasi-static state (e.g. by referring all quantities to a given
epoch), such that the Earth can be considered as a rigid and non-deformable body,
uniformly rotating about a body-fixed axis. Hence, all gravity field quantities includ-
ing the level surfaces are considered in the following as static quantities, which do
not change in time. On this basis, the static and temporal components of the gravity
potential can be added to obtain the actual potential value at time t, as needed, e.g.,
for the evaluation of clock comparison experiments.
In this context, a note on the handling of the permanent (time-independent) parts
of the tidal corrections is appropriate; for details, see, e.g., [132, 133], or [134]. The
International Association of Geodesy (IAG) has recommended that the so-called
“zero-tide system” should be used (resolutions no. 9 and 16 from the year 1983;
cf. [135]), where the direct (permanent) tide effects are removed, but the indirect
deformation effects associated with the permanent tidal deformation are retained.
Unfortunately, geodesy and other disciplines do not strictly follow the IAG resolu-
tions for the handling of the permanent tidal effects, and therefore, depending on
the application, appropriate corrections may be necessary to refer all quantities to a
common tidal system (see below and the aforementioned references).
In the following, some fundamentals of physical geodesy are given, and then
two geodetic methods are described for determining the gravity potential, con-
sidering both the geometric levelling approach and the GNSS/geoid approach
(GNSS – Global Navigation Satellite Systems), together with corresponding uncer-
tainty considerations.
4.1 Fundamentals of Physical Geodesy
Classical physical geodesy is largely based on the Newtonian theory with Newton’s
law of gravitation, giving the gravitational force between two point masses, to which
a gravitational acceleration (also termed gravitation) can be ascribed by setting the
mass at the attracted point P to unity. Then, by the law of superposition, the grav-
itational acceleration of an extended body like the Earth can be computed as the
vector sum of the accelerations generated by the individual point masses (or mass
elements), yielding
50 P. Delva et al.

r − r
b = b (r) = −G dm , dm = ρdv , ρ = ρ(r ) , (33)
Earth r − r 3
where r and r are the position vectors of the attracted point P and the source point
Q, respectively, dm is the differential mass element, ρ is the volume density, dv is
the volume element, and G is the gravitational constant. The SI unit of acceleration
is ms−2 , but the non-SI unit Gal is still frequently used in geodesy and geophysics
(1 Gal = 0.01 m s−2 , 1 mGal = 1 ×10−5 m s−2 ). While an artificial satellite is only
affected by gravitation, a body rotating with the Earth also experiences a centrifugal
force and a corresponding centrifugal acceleration z, which is directed outwards and
perpendicular to the rotation axis:
z = z( p) = ω2 p . (34)
In the above equation, ω is the angular velocity, and p is the distance vector from
the rotation axis. Finally, the gravity acceleration (or gravity) vector g is the resultant
of the gravitation b and the centrifugal acceleration z:
g = b+ z . (35)
As the gravitational and centrifugal acceleration vectors b and z both form con-
servative vector fields or potential fields, these can be represented as the gradient of
corresponding potential functions by
g = ∇W = b + z = ∇VE + ∇ Z E = ∇(VE + Z E ) , (36)
where W is the gravity potential, consisting of the gravitational potential VE and the
centrifugal potential Z E . Based on Eqs. (33)–(36), the gravity potential W can be
expressed as

ρdv ω2 2
W = W (r) = VE + Z E = G + p , (37)
Earth l 2
where l and p are the lengths of the vectors r − r and p, respectively. All potentials
are defined with a positive sign, which is common geodetic practice. The gravita-
tional potential VE is assumed to be regular (i.e. zero) at infinity and has the important
property that it fulfills the Laplace equation outside the masses; hence it can be rep-
resented by harmonic functions in free space, with the spherical harmonic expansion
playing a very important role. Further details on potential theory and properties of
the potential functions can be found, e.g., in [134, 136, 137].
The determination of the gravity potential W as a function of position is one of the
primary goals of physical geodesy; if W (r) were known, then all other parameters
of interest could be derived from it, including the gravity vector g according to
Eq. (36) as well as the form of the equipotential surfaces (by solving the equation
W (r) = const.). Furthermore, the gravity potential is also the ideal quantity for
describing the direction of water flow, i.e. water flows from points with lower gravity
potential to points with higher values. However, although the above equation is
fundamental in geodesy, it cannot be used directly to compute the gravity potential
W due to insufficient knowledge about the density structure of the entire Earth; this
is evident from the fact that densities are at best known with two to three significant
digits, while geodesy generally strives for a relative uncertainty of at least 10−9 for all
relevant quantities (including the potential W ). Therefore, the determination of the
exterior potential field must be solved indirectly based on measurements performed
at or above the Earth’s surface, which leads to the area of geodetic boundary value
problems (GBVPs; see below).
The gravity potential is closely related to the question of heights as well as level
or equipotential surfaces and the geoid, where the geoid is classically defined as
a selected level surface with constant gravity potential W0 , conceptually chosen to
approximate (in a mathematical sense) the mean ocean surface or mean sea level
(MSL). However, MSL does not coincide with a level surface due to the forcing
of the oceans by winds, atmospheric pressure, and buoyancy in combination with
gravity and the Earth’s rotation. The deviation of MSL from a best fitting equipotential
surface (geoid) is denoted as the (mean) dynamic ocean topography (DOT); it reaches
maximum values of about ±2 m and is of vital importance for oceanographers for
deriving ocean circulation models [138].
On the other hand, a substantially different approach was chosen by the IAG
during its General Assembly in Prague, 2015, within “IAG Resolution (No. 1) for
the definition and realization of an International Height Reference System (IHRS)”
[139], where a numerical value W0(IHRS) = 62, 636, 853.4 m2 s−2 (based on obser-
vations and data related to the mean tidal system) is defined for the realization of
the IHRS vertical reference level surface, with a corresponding note, stating that
W0(IHRS) is related to “the equipotential surface that coincides (in the least-squares
sense) with the worldwide mean ocean surface, the most accepted definition of the
geoid” [140]. Although the classical geodetic geoid definition and the IAG 2015
resolution both refer to the worldwide mean ocean surface, so far no adopted stan-
dards exist for the definition of MSL, the handling of time-dependent terms (e.g.,
due to global sea level rise), and the derivation of W0 , where the latter value can be
determined in principle from satellite altimetry and a global geopotential model (see
[141, 142]). Furthermore, the IHRS value for the reference potential is inconsistent
with the corresponding value W0(IAU) used for the definition of TT (see Sect. 3.5);
Petit et al. [124] denote these two definitions as “classical geoid” and “chronometric
geoid”, respectively.
In this context, it is somewhat unfortunate that the same notation (W0 ) is used to
represent different estimates for a quantity that is connected with the (time-variable)
mean ocean surface, but this issue can be resolved only through future international
cooperation, even though it seems unlikely that the different communities are willing
to change their definitions. In the meantime, this problem has to be handled by a
simple constant shift transformation between the different level surfaces, associated
with a thorough documentation of the procedures and conventions involved. It is
52 P. Delva et al.
clear that the definition of the zero level surface (W0 issue) is largely a matter of
convention, where a good option is probably to select a conventional value of W0
(referring to a certain epoch) with a corresponding zero level surface, and to describe
then the potential of the time-variable mean ocean surface for any given point in time
as the deviation from this reference value.
4.2 The Geometric Levelling Approach
The classical and most direct way to obtain gravity potential differences is based on
geometric levelling and gravity observations, denoted here as the geometric levelling
approach. Based on Eq. (36), the gravity potential differential can be expressed as
∂W ∂W ∂W
dW = dx + dy + dz = ∇W · ds = g · ds = −g dn , (38)
∂x ∂y ∂z
where ds is the vectorial line element, g is the magnitude of the gravity vector, and
dn is the distance along the outer normal of the level surface (zenith or vertical),
which by integration leads to the geopotential number C in the form
P P
C (i) = W0(i) − W P = − dW = g dn , (39)
P0(i) P0(i)
where P is a point at the Earth’s surface, (i) refers to the selected zero level or
height reference surface (height datum) with the gravity potential W0(i) , and P0(i) is
an arbitrary point on that level surface. Thus, in addition to the raw levelling results
(dn), gravity observations (g) are needed along the path between P0(i) and P, for
details, see, e.g., [137]. The geopotential number C is defined such that it is positive
for points P above the zero level surface, similar to heights. It should be noted
that the integral in Eq. (39) and hence C is path independent, as the gravity field is
conservative. Furthermore, the geopotential numbers can be directly linked to the
redshift correction according to Eq. (32) if one takes W0(IAU) as zero reference zero
level reference potential.
However, regarding height networks, the zero level surface and the corresponding
potential is typically selected in an implicit way by connecting the levelling to a
fundamental national tide gauge, but the exact numerical value of the reference
potential is usually unknown. As mean sea level deviates from a level surface within
the Earth’s gravity field due to the dynamic ocean topography (see Fig. 7), this leads
to inconsistencies of more than 0.5 m between different national height systems
across Europe, the extreme being Belgium, which differs by more than 2 m from
all other European countries due to the selection of low tide water as the reference
(instead of mean sea level).
Geometric levelling (also called spirit levelling) itself is a quasi-differential tech-
nique, which provides height differences δn (backsight minus foresight reading) with
Fig. 7 Illustration of several quantities involved in gravity field modelling
respect to a local horizontal line of sight. The uncertainty of geometric levelling is

rather low over shorter distances, where it can reach the sub-millimetre level, but
it is susceptible to systematic errors up to the decimetre level over 1000 km dis-
tance (see Sect. 4.4). In addition, the non-parallelism of the level surfaces cannot be
neglected over larger distances, as it results in a path dependence of the raw levelling
results ( dn = 0), but this problem can be overcome by using potential differences,
which are path independent because the gravity field is conservative ( dW = 0).
For this reason, geopotential numbers are almost exclusively used as the foundation
for national and continental height reference systems (vertical datum) worldwide,
but one can also work with heights and corresponding gravity corrections to the raw
levelling results (cf. [137]).
Although the geopotential numbers are ideal quantities for describing the direction
of water flow, they have the unit m2 s−2 and are thus somewhat inconvenient in
disciplines such as civil engineering. A conversion to metric heights is therefore
desirable, which can be achieved by dividing the C values by an appropriate gravity
value. Widely used are the orthometric heights (e.g. in the USA, Canada, Austria, and
Switzerland) and normal heights (e.g. in Germany, France and many other European
countries). Heights also play an important role in gravity field modelling due to the
strong height dependence of various gravity field quantities.
The orthometric height H is defined as the distance between the surface point
P and the zero level surface (geoid), measured along the curved plumb line, which
54 P. Delva et al.
explains the common understanding of this term as “height above sea level” [137].
All relevant height and gravity field related quantities, are illustrated in Fig. 7. The
orthometric height can be derived from Eq. (39) by integrating along the plumb line,
giving
H (i)
C (i) 1
H (i) = , ḡ = (i) g dH , (40)
ḡ H 0
where ḡ is the mean gravity along the plumb line (inside the Earth). As ḡ cannot be
observed directly, hypotheses about the interior gravity field are necessary, which is
one of the main drawbacks of the orthometric heights. Therefore, in order to avoid
hypotheses about the Earth’s interior gravity field, the normal heights H N were
introduced by Molodensky (e.g. [143]) in the form
H N (i)
C (i) 1
H N (i) = , γ̄ = N (i) γ dHN , (41)
γ̄ H 0
where γ̄ is a mean normal gravity value along the normal plumb line (within the
normal gravity field, associated with the level ellipsoid), and γ is the normal gravity
acceleration along this line. Consequently, the normal height H N is measured along
the slightly curved normal plumb line [137]. This definition avoids hypotheses about
the Earth’s interior gravity field, which is the main reason for adopting it in many
countries. Indeed, the value γ̄ can be calculated analytically, as the normal gravity
potential of the level ellipsoid U is known analytically (see next section), but γ̄ is
slightly depending on the chosen reference ellipsoid. However, the normal height
does not have a simple physical interpretation, in contrast to the orthometric height
(“height above sea level”). Nevertheless, the normal height can be interpreted as the
height above the quasigeoid, which is not a level surface and also has no physical
interpretation (see [137]).
While the orthometric and normal heights are related to the Earth’s gravity field
(so-called physical heights), the ellipsoidal heights h, as derived from GNSS obser-
vations, are purely geometric quantities, describing the distance (along the ellipsoid
normal) of a point P from a conventional reference ellipsoid. As the geoid and quasi-
geoid serve as the zero height reference surface (vertical datum) for the orthometric
and normal heights, respectively, the following relation holds
h = H (i) + N (i) = H N (i) + ζ (i) , (42)
where N (i) is the geoid undulation, and ζ (i) is the quasigeoid height or height
anomaly; for further details on the geoid and quasigeoid (height anomalies) see,
e.g., [137]. Equation (42) neglects the fact that strictly the relevant quantities are
measured along slightly different lines in space, but the maximum effect is only at
the sub-millimetre level (for further details cf. [134]).
Lastly, the geometric levelling approach gives only gravity potential differences,
but the associated constant zero potential W0(i) can be determined by at least one
(better several) GNSS and levelling points in combination with the (gravimetrically
derived) disturbing potential, as described in the next section. Rearranging the above
equations gives the desired gravity potential values in the form
W P = W0(i) − C (i) = W0(i) − ḡ H (i) = W0(i) − γ̄ H N (i) , (43)
and hence the geopotential numbers and the heights H (i) and H N (i) are fully
equivalent.
4.3 The GNSS/Geoid Approach
The gravity potential W cannot be derived directly from Eq. (37) due to insufficient
knowledge about the density structure of the entire Earth, and therefore it must
be determined indirectly based on measurements performed at or above the Earth’s
surface, which leads to the area of geodetic boundary value problems. In this context,
gravity measurements form one of the most important data sets. However, since
gravity (represented as g = |g| = length of the gravity vector g) and other relevant
observations depend in general in a nonlinear way on the potential W , the observation
equations must be linearized. This is done by introducing an a priori known reference
potential and corresponding reference positions. Regarding the reference potential,
traditionally the normal gravity field related to the level ellipsoid is employed, where
the ellipsoid surface is a level surface of its own gravity field. The level ellipsoid
is chosen as a conventional system, because it is easy to compute (from just four
fundamental parameters; e.g. two geometrical parameters for the ellipsoid plus the
total mass M and the angular velocity ω), useful for other disciplines, and also utilized
for describing station positions (e.g. in connection with GNSS or the International
Terrestrial Reference Frame – ITRF). However, today spherical harmonic expansions
based on satellite data could also be employed (cf. [134]).
The linearization process leads to the disturbing (or anomalous) potential T
defined as
TP = W P − U P , (44)
where U is the normal gravity potential associated with the level ellipsoid. Accord-
ingly, the gravity vector and other gravity field observables are approximated by
corresponding reference quantities based on the level ellipsoid, leading to gravity
anomalies g, height anomalies ζ , geoid undulations N , etc. The main advantage
of the linearization process is that the residual quantities (with respect to the known
ellipsoidal reference field) are in general four to five orders of magnitude smaller
than the original ones, and in addition they are less position dependent.
Hence, the disturbing potential T takes over the role of W as the new fundamental
target quantity, to which all other gravity field quantities of interest are related.
Accordingly, the gravity anomaly is given by
56 P. Delva et al.
∂T 1 ∂γ ∂γ (i)
g P = g P − γ Q = − + T− W0 − U 0 , (45)
∂h γ ∂h ∂h
where g P is the gravity acceleration at the observation point P (at the Earth’s surface
or above), γ Q is the normal gravity acceleration at a known linearization point Q
(telluroid, Q is located on the same ellipsoidal normal as P at a distance H N above
the ellipsoid, or equivalently U Q = W P ; for further details, see [134]), the partial
derivatives are with respect to the ellipsoidal height h, and δW0(i) = W0(i) − U0 is
the potential difference between the zero level height reference surface (W0(i) ) and
the normal gravity potential U0 at the surface of the level ellipsoid. Equation (45)
is also denoted as the fundamental equation of physical geodesy; it represents a
boundary condition that has to be fulfilled by solutions of the Laplace equation for
the disturbing potential T , sought within the framework of GBVPs. Moreover, the
subscripts P and Q are dropped on the right side of Eq. (45), noting that it must be
evaluated at the known telluroid point (boundary surface).
In a similar way, Bruns’s formula gives the height anomaly or quasigeoid height
as a function of T in the form
T W (i) − U0 T δW0(i)
ζ (i) = h − H N (i) = − 0 = − = ζ + ζ0(i) , (46)
γ γ γ γ
implying that ζ (i) and ζ are associated with the corresponding zero level surfaces
W = W0(i) and W = U0 , respectively. The δW0(i) term is also denoted as height system
bias and is frequently omitted in the literature, implicitly assuming that W0(i) equals
U0 . However, when aiming at a consistent derivation of absolute potential values,
the δW0(i) term has to be taken into consideration.
Hence, all linearized gravity field observables are linked to the disturbing poten-
tial T , which has the important property of being harmonic outside the Earth’s surface
and regular (zero) at infinity. Consequently, solutions of T are developed in the frame-
work of potential theory and GBVPs, i.e. solutions of the Laplace equation are sought
that fulfil certain boundary conditions. Now, the first option to compute T is based
on the well-known spherical harmonic expansion, using coefficients derived from
satellite data alone or in combination with terrestrial data (e.g., EGM2008; EGM –
Earth Gravitational Model [144]), yielding
n max
a n+1
n
T (θ, λ, r ) = T nm Y nm (θ, λ) (47)
n=0
r m=−n
with

cos mλ for m ≥ 0
Y nm (θ, λ) = P n|m| (cos θ ) , (48)
sin |m|λ for m < 0

G M C nm for m ≥ 0
T nm = , (49)
a S nm for m < 0
where (θ, λ, r ) are spherical coordinates, n, m are integers denoting the degree and
order, G M is the geocentric gravitational constant (gravitational constant G times
the mass of the Earth M), a is in the first instance an arbitrary constant, but is
typically set equal to the semimajor axis of a reference ellipsoid, P nm (cos θ ) are the
fully normalized associated Legendre functions of the first kind, and C nm , S nm
are the (fully normalized) spherical harmonic coefficients (also denoted as Stokes’s
constants), representing the difference in the gravitational potential between the real
Earth and the level ellipsoid.
Regarding the uncertainty of a gravity field quantity computed from a global
spherical harmonic model up to some fixed degree n max , the coefficient uncertainties
lead to the so-called commission error based on the law of error propagation, and the
omitted coefficients above degree n max , which are not available in the model, lead to
the corresponding omission error. With dedicated satellite gravity field missions such
as GRACE (Gravity Recovery and Climate Experiment) and GOCE (Gravity field and
steady-state Ocean Circulation Explorer), the long wavelength geoid and quasigeoid
can today be determined with low uncertainty, e.g., about 1 mm at 200 km resolution
(n = 95) and 1 cm at 150 km resolution (n = 135) from GRACE (e.g. [145]), and
1.5 cm at about 110 km resolution (n = 185) from the GOCE mission (e.g. [146,
147]). However the corresponding omission error at these wavelengths is still quite
significant with values at the level of several decimetres, e.g., 0.94 m for n = 90,
0.42 m for n = 200, and 0.23 m for n = 360. For the ultra-high degree geopotential
model EGM2008 [144], which combines satellite and terrestrial data and is complete
up to degree and order 2159, the omission error is 0.023 m, while the commission
error is about 5 to 20 cm, depending on the region and the corresponding data quality.
The above uncertainty estimates are based on the published potential coefficient
standard deviations as well as a statistical model for the estimation of corresponding
omission errors, but do not include the uncertainty contribution of G M (zero degree
term in Eq. (47)); hence, the latter term, contributing about 3 mm in terms of the
height anomaly (corresponds to about 0.5 ppb; see [148, 149]), has to be added in
quadrature to the figures given above. Further details on the uncertainty estimates
can be found in [134].
Based on these considerations it is clear that satellite measurements alone will
never be able to supply the complete geopotential field with sufficient accuracy,
which is due to the signal attenuation with height and the required satellite altitudes
of a few 100 km. Only a combination of the highly accurate and homogeneous (long
wavelength) satellite gravity fields with high-resolution terrestrial data (mainly grav-
ity and topography data with a resolution down to 1–2 km and below) can cope with
this task. In this respect, the satellite and terrestrial data complement each other in an
ideal way, with the satellite data accurately providing the long wavelength field struc-
tures, while the terrestrial data sets, which have potential weaknesses in large-scale
accuracy and coverage, mainly contribute the short wavelength features. However,
in the future, also height anomalies derived from common GNSS and clock points
may contribute to regional gravity field modelling (see Sect. 6).
Consequently, regional solutions for the disturbing potential and other gravity field
parameters have to be developed, which typically have a higher resolution (down to
58 P. Delva et al.
1–2 km) than global spherical harmonic models. Based on the developments of
Molodensky (e.g., [143]), the disturbing potential T can be derived from a series
of surface integrals, involving gravity anomalies and heights over the entire Earth’s
surface, which in the first instance can be symbolically written as
T = M(g) , (50)
where M is the Molodensky operator and g are the gravity anomalies over the
entire Earth’s surface.
Further details on regional gravity field modelling are given in [81, 134], including
the solution of Molodensky’s problem, the remove-compute-restore (RCR) proce-
dure, the spectral combination approach, data requirements, and uncertainty esti-
mates for the disturbing potential and quasigeoid heights. These investigations show
that quasigeoid heights can be obtained today with an estimated uncertainty of 1.9 cm,
where the major contributions come from the spectral band below spherical harmonic
degree 360. Furthermore, this uncertainty estimate represents an optimistic scenario
and is only valid for the case that a state-of-the-art global satellite model (e.g. a 5th
generation GOCE model [147]) is employed and sufficient high-resolution and high-
quality terrestrial gravity and terrain data sets (especially gravity measurements with
a spacing of a few kilometers and an uncertainty lower than 1 mGal) are available
around the point of interest (e.g. within a distance of 50–100 km), see also [150,
151]. Fortunately, such a data situation exists for most of the metrology institutes
with optical clock laboratories – at least in Europe. Furthermore, the perspective
exists to improve the uncertainty of the calculated quasigeoid heights [81].
Now, once the disturbing potential values T are computed, either from a global
geopotential model by Eq. (47), or from a regional solution by Eq. (50) based on
Molodensky’s theory, the gravity potential W , needed for the relativistic redshift
corrections, can be computed most straightforwardly from Eq. (44) as
W P = U P + TP , (51)
where the basic requirement is that the position of the given point P in space must
be known accurately (e.g. from GNSS observations), as the normal potential U is
strongly height-dependent, while T is only weakly height dependent with a maximum
vertical gradient of a few parts in 10−3 m2 s−2 per metre. The above equation also
makes clear that the predicted potential values W P are in the end independent of
the choice of W0 and U0 used for the linearization. Furthermore, by combining
equation (51) with (46), and representing U as a function of U0 and the ellipsoidal
height h, the following alternative expressions for W (at point P) can be derived as

W P = U0 − γ̄ (h − ζ ) = U0 − γ̄ h − ζ (i) + δW0(i) , (52)
which demonstrates that ellipsoidal heights (e.g. from GNSS) and the results from
gravity field modelling in the form of the quasigeoid heights (height anomalies) ζ or
the disturbing potential T are required, whereby a similar equation can be derived
for the geoid undulations N . Consequently, the above approach (Eqs. (51) and (52))
is denoted here somewhat loosely as the GNSS/geoid approach, which is also known
in the literature as the GNSS/GBVP approach (the geodetic boundary value problem
is the basis for computing the disturbing potential T ; see, e.g., [152, 153]).
The GNSS/geoid approach depends strongly on precise gravity field modelling
(disturbing potential T , metric height anomalies ζ or geoid undulations N ) and pre-
cise GNSS positions (ellipsoidal heights h) for the points of interest, with the advan-
tage that it delivers the absolute gravity potential W , which is not directly observable
and is therefore always based on the assumption that the gravitational potential is
regular (zero) at infinity (see above). In addition, the GNSS/geoid approach allows
the derivation of the height system bias term δW0(i) based on Eq. (46) together with at
least one (better several) common GNSS and levelling stations in combination with
the gravimetrically determined disturbing potential T .
4.4 Uncertainty Considerations
The following uncertainty considerations are based on heights, but corresponding

potential values can easily be obtained by multiplying the meter values with an
average gravity value (e.g. 9.81 ms−2 or roughly 10 ms−2 ). Regarding the geometric
levelling and the GNSS/geoid approach, the most direct and accurate way to derive
potential differences over short distances is the geometric levelling technique, as
standard deviations of 0.2–1.0 mm can be attained for a 1 km double-run levelling
with appropriate technical equipment [137]. However, the uncertainty of geometric
levelling depends on many factors, with some of the levelling errors behaving in
a random manner and propagating with the square root of the number of individ-
ual set-ups or the distance, respectively, while other errors of systematic type may
propagate with distance in a less favourable way. Consequently, it is important to
keep in mind that geometric levelling is a differential technique and hence may be
susceptible to systematic errors; examples include the differences between the sec-
ond and third geodetic levelling in Great Britain (about 0.2 m in the north–south
direction over about 1000 km distance [154]), corresponding differences between an
old and new levelling in France (about 0.25 m from the Mediterranean Sea to the
North Sea, also mainly in north–south direction, distance about 900 km [155]), and
inconsistencies of more than ±1 m across Canada and the USA (differences between
different levellings and with respect to an accurate geoid [156–158]). In addition, a
further complication with geometric levelling in different countries is that the results
are usually based on different tide gauges with offsets between the corresponding
zero level surfaces, which, for example, reach more than 0.5 m across Europe. Fur-
thermore, in some countries the levelling observations are about 100 years old and
thus may not represent the actual situation due to possibly occurring recent vertical
crustal movements.
With respect to the GNSS/geoid approach, the uncertainty of the GNSS posi-
tions is today more or less independent of the interstation distance. For instance, the
60 P. Delva et al.
station coordinates provided by the International GNSS Service (IGS) or the IERS
(e.g. ITRF2008) reach vertical accuracies of about 5–10 mm (cf. [159–161]). The
uncertainty of the quasigeoid heights (height anomalies) is discussed in the previous
subsection, showing that a standard deviation of 1.9 cm is possible in a best-case
scenario. Moreover, the values are nearly uncorrelated over longer distances, with a
correlation of less than 10% beyond a distance of about 180 km [81]. Aiming at the
determination of the absolute gravity potential W according to Eq. (51) or (52), which
is the main advantage of the GNSS/geoid technique over the geometric levelling
approach, both the uncertainties of GNSS and the quasigeoid have to be considered.
Assuming a standard deviation of 1.9 cm for the quasigeoid heights and 1 cm for the
GNSS ellipsoidal heights without correlations between both quantities, a standard
deviation of 2.2 cm is finally obtained (in terms of heights) for the absolute potential
values based on the GNSS/geoid approach. Thus, for contributions of optical clocks
to international timescales, which require the absolute potential W P relative to a con-
ventional zero potential W0 (see Sect. 3.5), the relativistic redshift correction can be
computed with an uncertainty of about 2×10−18 . This is the case more or less every-
where in the world where high-resolution regional gravity field models have been
developed on the basis of a state-of-the-art global satellite model in combination with
sufficient terrestrial gravity field data. On the other hand, for potential differences
over larger distances of a few 100 km (i.e. typical distances between different metrol-
ogy institutes), the statistical correlations of the quasigeoid values virtually vanish,
which then leads√ to a standard deviation for the potential difference of 3.2 cm in terms
of height, i.e. 2 times the figure given above for the absolute potential (according
to the law of error propagation), which again has to be considered as a best-case
scenario. This would also hold for intercontinental connections between metrology
institutes, provided again that sufficient regional high-resolution terrestrial data exist
around these places. Furthermore, in view of future refined satellite and terrestrial
data, the perspective exists to improve the uncertainty of the relativistic redshift cor-
rections from the level of a few parts in 1018 to one part in 1018 or below. According
to this, over long distances across national borders, the GNSS/geoid approach should
be a better approach than geometric levelling (see also [81]).
5 Relativistic Redshift Corrections for the Realization of

TT from Geodetic Methods
An atomic clock, in order to contribute to the realization of Terrestrial Time (TT),

needs to be corrected for the relativistic redshift (see Sect. 3.5, Eq. (31)). In this
section we present some results from the ITOC (International Timescales with Optical
Clocks) project, in particular those linked to the determination of unified relativistic
redshift corrections for several European metrology institutes.
5.1 The ITOC Project
The ITOC project ([60]; see also http://projects.npl.co.uk/itoc/) was a 3 years (2013–
2016) EURAMET joint research project funded by the European Community’s
Seventh Framework Programme, ERA-NET Plus. This project was done in the
context of a future optical redefinition of the SI second (see e.g. [162–165]). An
extensive programme of comparisons between high accuracy European optical
atomic clocks has been performed, verifying the estimated uncertainty budgets of
the optical clocks. Relativistic effects influencing clock comparisons have been eval-
uated at an improved level of accuracy, and the potential benefits that optical clocks
could bring to the field of geodesy have been demonstrated.
Several optical frequency ratio measurements as well as independent absolute
frequency measurements of optical lattice clocks have been made locally at the fol-
lowing NMIs (National Metrology Institutes): INRIM (Istituto Nazionale di Ricerca
Metrologica, Torino, Italy), LNE-SYRTE (Laboratoire national de métrologie et
d’essais – Système de Références Temps-Espace, Paris, France), NPL (National
Physical Laboratory, Teddington, UK), and PTB (Physikalisch-Technische Bundes-
anstalt, Braunschweig, Germany), all of whom operate one or more than one type
of optical clock, as well as Caesium primary frequency standards (see e.g. [78, 106,
107, 166]). Distant comparisons have also been performed between the same labo-
ratories with a broadband version of two-way satellite time and frequency transfer
(TWSTFT).
A proof-of-principle experiment has been realized to show that the relativistic
redshift of optical clocks can be exploited to measure gravity potential differences
over medium–long baselines. A transportable 87 Sr optical lattice clock has been
developed at PTB [105]. It has been transported to the Laboratoire Souterrain de
Modane (LSM) in the Fréjus road tunnel through the Alps between France and
Italy. There it was compared, using a transportable frequency comb from NPL, to
the caesium fountain primary frequency standard at INRIM, via a coherent fibre link
and a second optical frequency comb operated by INRIM. A physical model has been
formulated to describe the relativistic effects relevant to time and frequency transfer
over optical fibre links, and has been used to evaluate the relativistic corrections
for the fibre links now in place between NPL, LNE-SYRTE and PTB, as well as to
provide guidelines on the importance of exact fibre routing for time and frequency
transfer via optical fibre links (see [122] and Sect. 3.3.2).
Within the ITOC project, the gravity potential has been determined by IfE/LUH
(Institut für Erdmessung, Leibniz Universität Hannover) with significantly improved
accuracy at the sites participating in optical clock comparisons within the project
(INRIM, LNE-SYRTE, LSM, NPL and PTB). Levelling measurements and grav-
ity surveys have been performed at INRIM, LSM, OBSPARIS, NPL and PTB, the
latter including at least one absolute gravity observation at each site. These mea-
surements have been integrated into the existing European gravity database and used
for the computation of a new version of the European Gravimetric (Quasi) Geoid,
EGG2015 (see [167] and Sect. 5.3). Time-variable gravity potential signals induced
62 P. Delva et al.
by tides and non-tidal mass redistributions have also been calculated for the optical
clock comparison sites [129]. Finally, the potential contributions of combined GNSS
and optical clock measurements for determining the gravity potential at high spatial
resolution have been studied theoretically, which will be presented in Sect. 6.
5.2 The GNSS and Levelling Campaigns
Within the ITOC project, GNSS and levelling observations were performed at the
NMIs INRIM, LNE-SYRTE, NPL, and PTB, as well as the collaborator LSM (not
an NMI) to calculate the relativistic redshift corrections. First of all, some general
recommendations were developed for carrying out the measurements to ensure accu-
racies in the millimetre range for the levelling results and better than one centimetre
for the GNSS (ellipsoidal) heights (see [81]). In general, it is recommended to install
fixed markers in all local laboratories close to the clock tables to allow an easy height
transfer to the clocks (e.g. with a simple spirit level used for building construction),
and to connect these markers by geometric levelling with millimetre uncertainty to
the existing national levelling networks and at least two (better several) GNSS sta-
tions. This is to support local clock comparisons at the highest level, and to apply the
GNSS/geoid approach to obtain also the absolute potential values for remote clock
comparisons and contributions to international timescales, while at the same time
improving the redundancy and allowing a mutual control of GNSS, levelling, and
(quasi)geoid data.
Fig. 8 Map showing the

locations of the INRIM,
LNE-SYRTE, NPL, PTB,
and LSM sites
The actual levelling and GNSS measurements were mainly taken by local survey-
ors on behalf of the respective NMIs, and the NMIs provided all results to Leibniz
Universität Hannover (LUH; Institut für Erdmessung) for further processing and
homogenisation. The locations of the above mentioned ITOC clock sites are shown
in Fig. 8.
The coordinates of all GNSS stations were referred to the ITRF2008 at its asso-
ciated standard reference epoch 2005.0, with three-dimensional Cartesian and ellip-
soidal coordinates being available. The geometric levelling results were based in the
first instance on the corresponding national vertical reference networks, which are
the following:
• DHHN92 is the official German height reference system; it is based on the Ams-
terdam tide gauge and consists of normal heights.
• NGF-IGN69 is the official French height reference system; it is based on the
Marseille tide gauge and also consists of normal heights. In addition, selected
levelling lines were re-observed since 2000, which lead to the so-called (NGF)–
NIREF network, differing from the old network mainly by a south-to-north trend
of 31.0 mm per degree latitude [155].
• ODN (Ordnance Datum Newlyn, established by Ordnance Survey) is the height
reference system for mainland Great Britain; it is based on the Newlyn tide gauge
and consists of orthometric heights.
• IGM is the Italian height reference system (established by Istituto Geografico
Militare); it is based on the Genova tide gauge and consists of orthometric heights.
The different zero level surfaces of the above national height systems (datum) were
taken into account by transforming all national heights into the unified European
Vertical Reference System (EVRS) using its latest realization EVRF2007 (European
Vertical Reference Frame 2007). The EVRF2007 is based on a common adjustment of
all available European levelling networks in terms of geopotential numbers, which are
finally transformed into normal heights. The measurements within the UELN (United
European Levelling Network) originate from very different epochs, but reductions
for vertical crustal movements were only applied for the (still ongoing) post-glacial
isostatic adjustment (GIA) in northern Europe; for further details on EVRF2007,
see [168]. However, as GIA hardly affects the aforementioned clock sites, while
other sources of vertical crustal movements are not known, the EVRF2007 heights
are considered as stable in time in the following.
For the conversion of the national heights into the vertical reference frame
EVRF2007, nearby common points with heights in both systems were utilized;
this information was kindly provided by Bundesamt für Kartographie und Geodäsie
(BKG) in Germany (M. Sacher, personal communication, 9 October 2015). If such
information is not available, the CRS-EU webpage (Coordinate Reference Systems
in Europe; http://www.crs-geo.eu, also operated by BKG) can be used, which gives,
besides a description of all the national and international coordinate and height ref-
erence systems for the participating European countries, up to three transformation
parameters (height bias and two tilt parameters) for the transformation of the national
heights into EVRF2007 and a statement on the quality of this transformation.
64 P. Delva et al.
The EVRF2007 heights were computed as
H N (EVRF2007) = H (national) + H (national) , (53)
where H (national) is a constant shift for each NMI site. The following offsets
H (national) were employed:
• PTB: +0.006 m (DHHN92),

• LNE-SYRTE: −0.479 m (NGF-IGN69),
• NPL: −0.010 m (ODN, based on the official EVRF2007 results),
−0.144 m (ODN, own estimate, see below)
• INRIM: −0.307 m (IGM),
• LSM: −0.307 m (IGM).
The accuracy of the above transformation depends on the accuracy of the input heights
as well as the number of identical points, giving RMS residuals of the transformation
between 2 mm (Germany) and 35 mm (Italy). A further note is necessary for the
computation of the NPL offset. The offset of −0.010 m is based on the official
EVRF2007 heights, which rely on hydrodynamic levelling (see [169]), but do not
include the 1994 channel tunnel levelling. Therefore, a first attempt was made to
consider the new channel tunnel levelling as well as the new levelling measurements
in France (NIREF, see above); this was done by starting with an offset of −0.479 m
for LNE-SYRTE, plus a correction for the NGF-IGN69 tilt between LNE-SYRTE
and Coquelle (channel tunnel entrance in France) of −0.065 m (south-north slope
= −31.0 mm per degree latitude, latitude difference = 3.105◦ [155]), plus an offset
of +0.400 m for the difference between ODN and NGF-IGN69 from the channel
tunnel levelling [170], resulting in an offset of −0.144 m for NPL. In Sect. 5.4 it is
shown that the new offset leads to a better agreement between the geometric levelling
and the GNSS/geoid approach.
Further details on the local levelling results and corresponding GNSS observations
at some of the aforementioned clock sites can be found in [81]. In general, the
uncertainty of the local levellings is at the few millimeters level, and the uncertainty
of the GNSS ellipsoidal heights is estimated to be better than 10 mm. Moreover,
care has to be taken in the handling of the permanent parts of the tidal corrections
(for details, see, e.g., [132–134]. The IAG has recommended to use the so-called
“zero tide system” (resolutions no. 9 and 16 from the year 1983; cf.[135]), which is
implemented in the European height reference frame EVRF2007 and the European
gravity field modelling performed at LUH (e.g. EGG2015, see below). On the other
hand, most GNSS coordinates (including the ITRF and IGS results) refer to the “non-
tidal (or tide-free) system”. Hence, for consistency with the IAG recommendations
and the other quantities involved (EVRF2007 heights, quasigeoid), the ellipsoidal
heights from GNSS were converted from the non-tidal to the zero-tide system based
on the following formula from [133] with
h zt = h nt + 60.34 − 179.01 sin2 φ − 1.82 sin4 φ [mm] , (54)

where φ is the ellipsoidal latitude, and h nt and h zt are the non-tidal and zero-tide
ellipsoidal heights, respectively. Hence, the zero-tide heights over Europe are about
3–5 cm smaller than the corresponding non-tidal heights.
5.3 The European Gravimetric Quasigeoid Model EGG2015
The latest European gravimetric quasigeoid model EGG2015 [167] was employed
to determine absolute potential values based on Eqs. (51) and (52), as needed for
the derivation of the relativistic redshift corrections in the context of international
timescales. The major differences between EGG2015 and the previous EGG2008
model [134] are the inclusion of additional gravity measurements carried out recently
around the aforementioned ITOC clock sites [60] and the use of a newer geopoten-
tial model based on the GOCE satellite mission instead of EGM2008. The new
gravity measurements around the clock sites were carried out by LUH, taking at
least one absolute gravity observation (with the LUH FG5X-220 instrument) plus
additional relative gravity observations (relative to the established absolute points)
around all ITOC sites. The total number of new gravity points is 36 for INRIM,
100 for LNE-SYRTE, 123 for LSM, 66 for NPL, and 46 for PTB, where most of the
measurements were taken around LSM due to the high mountains and corresponding
strong gravity field variations. Overall, the purpose of the new gravity measurements
was threefold, namely to perform spot checks of the largely historic gravity data
base (consistency check), to add new observations in areas void of gravity data so
far (coverage improvement), and to serve for future geodynamic and meteorologi-
cal purposes (infrastructure improvement), with the ultimate goal of improving the
reliability and accuracy of the computed quasigeoid model.
EGG2015 was computed from surface gravity data in combination with topo-
graphic information and the geopotential model GOCO05S [146] based on the RCR
technique. The estimated uncertainty (standard deviation) of the absolute quasigeoid
values is 1.9 cm; further details including correlation information can be found in [81,
134].
5.4 Gravity Potential Determination
First, a consistency check between the GNSS and levelling heights at each clock site
was performed by evaluating the differences between the GNSS ellipsoidal heights
and the normal heights from levelling, computed as ζGNSS = h zt − H N (EVRF2007) ,
also denoted as GNSS/levelling quasigeoid heights (h zt is referring to ITRF2008,
epoch 2005.0, zero-tide system; H N (EVRF2007) is based on EVRF2007; see Sect. 5.2).
As the distances between the GNSS stations at each NMI site are typically only a
few 100 m, the quasigeoid at each site can be approximated in the first instance by a
horizontal plane, but a more general and better way (especially for larger interstation
66 P. Delva et al.
distances) is the comparison with a high-resolution gravimetric quasigeoid model,

such as EGG2015. After computing the differences (ζGNSS − ζEGG2015 ), the main
quantities of interest are the residuals about the mean difference, giving RMS values
of 11 mm (max. 16 mm) for INRIM, 5 mm (max. 5 mm) for LNE-SYRTE, 6 mm
(max. 8 mm) for NPL, and 4 mm (max. 6 mm) for PTB, while for LSM only a control
through two RTK (Real Time Kinematics - a differential GNSS technique) positions
exists, giving a RMS difference of 17 mm (max. 29 mm); this proves an excellent
consistency of the GNSS and levelling results at all clock sites. Although initial
results were worse for the PTB and LNE-SYRTE sites, the problem was traced to
an incorrect identification of the corresponding antenna reference points (ARPs); at
the PTB site, an error of 16 mm was found for station PTBB, and at LNE-SYRTE,
there was a difference of 29 mm between the ARP and the levelling benchmark
and an additional error in the ARP height of 8 mm at station OPMT. It should be
noted that, due to the high consistency of the GNSS and levelling data at all sites,
even quite small problems in the ARP heights (below 1 cm) could be detected and
corrected after on-site inspections and additional verification measurements. This
also strongly supports the recommendation to have sufficient redundancy in the
GNSS and levelling stations.
Now, in order to apply the GNSS/geoid approach according to Eqs. (51) and (52),
ellipsoidal heights are required for all stations of interest. However, initially GNSS
coordinates are only available for a few selected points at each NMI site, while
for most of the other laboratory points near the clocks, only levelled heights exist.
Therefore, based on Eq. (46), a quantity δζ is defined as

δζ = h − H N (i) − ζ (i) , (55)
which should be zero in theory, but is not in practice due to the uncertainties in
the quantities involved (GNSS, levelling, quasigeoid). However, if a high-resolution
quasigeoid model is employed (such as EGG2015), the term δζ should be small and
represent only long-wavelength features, mainly due to systematic levelling errors
over large distances as well as long-wavelength quasigeoid errors. In this case, an
average (constant) value δζ (based on the common GNSS and levelling benchmarks)
can be used at each NMI site to convert all levelled heights into ellipsoidal heights
by using
h (adj) = H N (i) + ζ (i) + δζ = H N (i) + ζ + ζ0(i) + δζ , (56)
which is based on Eq. (42). This has the advantage that locally (at each NMI)
the consistency is kept between the levelling results, on the one hand, and the
GNSS/quasigeoid results on the other hand. Consequently, the final potential differ-
ences between stations at each NMI are identical for the GNSS/geoid and geometric
levelling approach, which is reasonable, as locally the uncertainty of levelling is
usually lower than that of the GNSS/quasigeoid results.
Based on the ellipsoidal heights (according to Eq. (56)) and the EVRF2007 normal
heights (based on Eq. (53)), the gravity potential values can finally be derived for
all relevant stations, using both the geometric levelling approach (Eq. (43)) and
the GNSS/geoid approach (Eq. (51) or (52)). The results from both approaches are
provided in the form of geopotential numbers according to Eq. (39) with
C (IAU) = W0(IAU) − W P (57)
where the conventional value W0(IAU) = c2 L G ≈ 62, 636, 856.00 m2 s2 is used, fol-
lowing the IERS2010 conventions and the IAU resolutions for the definition of
TT (see Sect. 3.5). The geopotential numbers C are more convenient than the absolute
potential values W P due to their smaller numerical values and direct usability for the
derivation of the (static) relativistic redshift corrections according to Eq. (32). The
geopotential numbers C derived from Eq. (57) are typically given in the geopoten-
tial unit (gpu; 1 gpu = 10m2 s−2 ), resulting in numerical values of C that are about
2% smaller than the numerical height values. Regarding the geometric levelling
approach, the value W0(EVRF2007) = 62, 636, 857.86 m2 s−2 based on the European
EUVN_DA GNSS/levelling data set from [171] is utilized in Eq. (43), giving C (lev) .
For the GNSS/geoid approach according to Eq. (51) or (52), the disturbing potential
T or the corresponding height anomaly values ζ are taken from the EGG2015 model,
and the normal potential U0 = 62, 636, 860.850 m2 s−2 , associated with the surface
of the underlying GRS80 (Geodetic Reference System 1980; see [172]) level ellip-
soid, is used, resulting in C (GNSS/geoid) . Furthermore, the mean normal gravity values
γ̄ are also based on the GRS80 level ellipsoid; for further details, see [81]. Taking
all this into account, leads to the following discrepancies between the geopotential
numbers from the GNSS/geoid and the geometric levelling approach, defined in the
sense C = C (GNSS/geoid) − C (lev) :
• PTB: −0.017 gpu,

• LNE-SYRTE: −0.109 gpu,
• NPL: −0.275 gpu (with ODN offset based on official EVRF2007 results),
−0.144 gpu (with ODN offset based on own estimate, see above),
• INRIM: +0.019 gpu,
• LSM: −0.087 gpu.
The above results show first of all that the two approaches differ at the few decimetre
level over Europe, that the consideration of the new French and channel tunnel lev-
elling leads to a better agreement, and that the implementation of the national height
system offsets was done correctly, recalling, e.g., that the difference between the
French and German zero level surfaces is about half a metre. However, as the above
differences C are directly depending on the chosen reference potential W0(EVRF2007)
for EVRF2007, potential differences between two stations and the corresponding dis-
crepancies between the GNSS/geoid and the geometric levelling approach are dis-
cussed as well in the following. Regarding potential differences, the discrepancies
between both approaches amount to −0.106 gpu for the connection INRIM/LSM,
−0.036 gpu for INRIM/PTB, −0.092 gpu for PTB/LNE-SYRTE, −0.166 gpu for
LNE-SYRTE/NPL (−0.035 gpu based on own ODN offset, see above), as well
68 P. Delva et al.
as −0.258 gpu for PTB/NPL (−0.127 gpu based on own ODN offset, see above),
respectively.
Regarding the significance of the aforementioned discrepancies in the potential
differences between both geodetic approaches (levelling, GNSS/geoid), these have
to be discussed in relation to the corresponding uncertainties of levelling, GNSS, and
the quasigeoid model. Denker et al. [81] discuss the uncertainties (standard devia-
tion) from single line levelling connections and the EVRF2007 network adjustment,
indicating a factor 2.5 improvement due to the network adjustment. The EVRF2007
network adjustment gives a standard deviation of about 20 mm for the height con-
nection PTB/LNE-SYRTE, while the corresponding standard deviations for the con-
nections PTB/NPL and LNE-SYRTE/NPL are both about 80 mm (M. Sacher, BKG,
Leipzig, Germany, personal communication, 10 May 2017), the latter being domi-
nated by the uncertainty of the hydrodynamic levelling across the English Channel.
However, these internal uncertainty estimates from the network adjustment do not
consider any systematic levelling error contributions. On the other hand, the GNSS
ellipsoidal heights have uncertainties below 10 mm, the uncertainty of EGG2015 has
been discussed above, yielding a standard deviation of 19 mm for the absolute values
and about 27 mm for corresponding differences over longer distances, and therefore
some of the larger discrepancies between the two geodetic approaches (levelling
versus GNSS/geoid) have to be considered as statistically significant. Hence, as sys-
tematic errors in levelling at the decimetre level exist over larger distances in the
order of 1000km (e.g. in France, UK, and North America; see above), it is hypoth-
esized that the largest uncertainty contribution to the discrepancies between both
geodetic approaches comes from geometric levelling (see also [81]). Consequently,
geometric levelling is recommended mainly for shorter distances of up to several ten
kilometres, where it can give millimetre uncertainties, while over long distances, the
GNSS/geoid approach should be a better approach than geometric levelling, and it
can also deliver absolute potential values needed for contributions to international
timescales.
5.5 Unified Relativistic Redshift Corrections
The results from the gravity potential determination from both the geometric lev-
elling and the GNSS/geoid approach are given in Table 1 for the two ITOC sites
PTB and LNE-SYRTE as typical examples; further results for the other ITOC sites
are foreseen for a separate publication, and corresponding results for further sites in
Germany are documented in [81]. Based on the discussion in the preceding section
as well as Sect. 4.4, Table 1 gives the relativistic redshift corrections only for the
GNSS/geoid approach, which can be considered as the recommended values. The red-
shift corrections are based on the conventional value W0(IAU) , following the IERS2010
conventions and the IAU resolutions for the definition of TT (see Sect. 3.5), using
equation (32). The uncertainty of the given relativistic redshift corrections based
on the GNSS/geoid approach amounts to about 2 × 10−18 (see above). All opera-
tions from the measurements to the final values of the unified relativistic redshift
corrections are summarized in the flowchart given in Fig. 9.
Finally, as the results from the geometric levelling approach and the GNSS/geoid
approach are presently inconsistent at the decimetre level across Europe, the more or
less direct observation of gravity potential differences through optical clock compar-
isons (with targeted fractional accuracies of 10−18 , corresponding to 1cm in height) is
eagerly awaited as a means for resolving the existing discrepancies between different
geodetic techniques and remedying the geodetic height determination problem over
large distances. A first attempt in this direction was the comparison of two strontium
optical clocks between PTB and LNE-SYRTE via a fibre link, showing an uncer-
tainty and agreement with the geodetic results of about 5 × 10−17 [75]. This was
mainly limited by the uncertainty and instability of the participating clocks, which
is likely to improve in the near future.
Furthermore, for clocks with performance at the 10−17 level and below, time-
variable effects in the gravity potential, especially solid Earth and ocean tides, have
to be considered and can also serve as a method of evaluating the performance of the
optical clocks (i.e. a detectability test). Recent analysis of optical clock comparisons
already included temporal variations [130, 131]. Then, after further improvements
in the optical clock performance, conclusive geodetic results can be anticipated in
the future, and clock networks may also contribute to the establishment of the Inter-
national Height Reference System (IHRS).
6 Contribution of Chronometric Geodesy to the

Determination of the Geoid
In geodesy, geoid determination is understood as the determination of the shape

and size of the geoid with respect to a well-defined coordinate reference system,
which usually means the determination of the height of the geoid (geoid height)
above a given reference ellipsoid. The problem is solved within the framework of
potential theory and GBVPs, where the task is to find a harmonic function (i.e. the
disturbing potential T ) everywhere outside the Earth’s masses (possibly after mass
displacements and reductions), which fulfills certain boundary conditions. In princi-
ple, all measurements that can be mathematically linked to the disturbing potential
T (e.g. gravity anomalies, vertical deflections, gradiometer observations, and point-
wise disturbing potential values itself), can contribute to the solution, but in practice
gravity measurements usually play the main role in combination with topographic
and global satellite gravity information (also denoted as the gravimetric method, see
above). A very flexible approach, with the capability to combine all the aforemen-
tioned (inhomogeneous) measurements of different kinds and the option to predict
(output) heterogeneous quantities related to T , is the least-squares collocation (LSC)
method [173].
70
Table 1 Ellipsoidal coordinates (latitude, longitude, height; φ, λ, h (ad j.) ) referring to ITRF2008 reference frame (epoch 2005.0; GRS80 ellipsoid; zero-
tide system), normal heights H N (EVRF2007) based on EVRF2007, geopotential numbers based on the geometric levelling (C (lev) ) and GNSS/geoid approach
(IAU)
(C (GNSS/geoid) ) relative to the IAU2000 conventional reference potential W0 and differences C thereof, as well as the relativistic redshift correction based
on the GNSS/geoid approach
Station φ λ h (ad j.) H N (i) C (lev) C (GNSS/geoid)C Redshift
[◦ ] [’] [”] [◦ ] [’] [”] [m] [m] [10 m2 s−2 ] [10 m2 s−2 ] [10 m2 s−2 ] [10−16 ]
PTB, Braunschweig, Germany
PTBB 52 17 46.28177 10 27 35.08676 130.201 87.442 85.617 85.600 −0.017 −95.243
LB03 52 17 49.94834 10 27 37.63590 143.514 100.758 98.684 98.667 −0.017 −109.782
AF02 52 17 30.90851 10 27 28.21874 123.716 80.945 79.242 79.225 −0.017 −88.150
MB02 52 17 47.22270 10 27 50.49262 144.932 102.173 100.072 100.055 −0.017 −111.326
KB01 52 17 45.2 10 27 33.1 119.627 76.867 75.241 75.224 −0.017 −83.698
KB02 52 17 46.3 10 27 35.1 119.708 76.949 75.321 75.304 −0.017 −83.787
LNE-SYRTE, Paris, France
100 48 50 7.99682 2 20 8.38896 105.652 61.394 60.039 59.930 −0.109 −76.845
A 48 50 10.90277 2 20 10.55555 130.964 86.706 84.868 84.759 −0.109 −76.851
OPMT 48 50 9.31198 2 20 5.77891 122.546 78.288 76.611 76.502 −0.109 −81.712
P. Delva et al.
Fig. 9 Flowchart: from measurements to the determination of a unified relativistic redshift clock
correction at European scale. It is possible to extend this chart to the worldwide scale wherever a
high quality gravimetric model of the geoid exists
Regarding the use of clocks for gravity field modelling and geoid determination,
this always implies that also precise positions of the clock points with respect to
a well-defined reference system are required. This concerns mainly the ellipsoidal
height, which should be available with the same (or lower) uncertainty than the clock-
based physical heights or potential values, such that gravity field related quantities
N = h − H or ζ = h − HN (cf. Eq. (42)) can be obtained, establishing a direct link
to the disturbing potential T (e.g. through Eq. (46)); this is exactly the same situa-
tion as a combination of GNSS and geometric levelling (so-called GNSS/levelling),
as employed since many years (e.g. [174, 175]). Consequently, always clock plus
72 P. Delva et al.
GNSS measurements are required for gravity field modelling and geoid determina-
tion. Furthermore, in view of further improved clocks at (or below) the 10−18 level, it
should be noted that an ellipsoidal height uncertainty of 5–10 mm is about the limit
of what is achievable with GNSS today, requiring static and sufficiently long obser-
vation sessions and an appropriate post-processing. Clock measurements alone are
directly equivalent to the results from geometric levelling and gravity measurements
and hence can be considered as a height (but not a geoid) determination technique; if
clocks can be compared with a (space) reference clock with known potential value,
then this could help to realize the geoid, i.e. to find its position with respect to a given
measurement point on the Earth’s surface, but this still does not mean that one would
know the coordinates of the corresponding geoid point (i.e. its ellipsoidal height or
geoid height).
Distant clock comparisons and GNSS measurements provide a new kind of geode-
tic observable, which is complementary to the classical geodetic measurements (ter-
restrial and satellite gravity field observations). We have seen in Sect. 4.3 that satellite
and terrestrial data (mainly gravity and topography) complement each other, with
the satellite data providing the long wavelength field structures, while the terrestrial
data contributes to the short wavelength features. Indeed, terrestrial data (gravity and
topography) is most sensitive to small-scale spatial variations of the gravity potential.
For this reason, insufficiently dense terrestrial data can lead to significant errors in
the determination of the geoid.
By nature, potential data are smoother and more sensitive to mass sources at large
scales than gravity data. They can complement the information given by the gravity
data in the same way as the satellite data does, but on smaller scales. Therefore they
could provide the medium wavelength field structure, in between the spectral infor-
mation of classical terrestrial data and satellite data. They could reduce the error in
the determination of the geoid where gravity data are too sparse to reconstruct the
medium wavelengths field structures. Indeed, gravity data are sometimes sparsely
distributed: the plains are generally densely surveyed, while the mountainous regions
are poorly covered because some areas are mostly inaccessible by conventional grav-
ity surveys. Clock and GNSS data nearby these inaccessible areas could reduce the
error in the determination of the geoid.
To illustrate the potential benefits of clocks and GNSS in geodesy, the determi-
nation of the geopotential at high spatial resolution, about 10 km, was investigated
in [65]. The tested region is the Massif Central in France. It is interesting because it is
characterized by smooth, moderate altitude mountains and volcanic plateaus, leading
to variations of the gravitational field over a range of spatial scales. In such type of
region, the scarcity of gravity data is an important limitation in deriving accurate
high resolution geopotential models.
6.1 Methodology
The simulations are based on synthetic data (gravity and potential/clock data) and
consist in comparing the quality of the geopotential reconstruction solutions from
the gravity data, with or without taking into account clock data. In the following,
“clock data” is considered as disturbing potential T values derived from clock and
GNSS measurements as outlined above (on the basis of Eq. (46)). The synthetic
gravity and potential data are sampled by using a state of the art geopotential model
[176, EIGEN-6C4] up to degree and order 2000 (i.e. 10 km resolution), and some
spatial distribution of points. The solutions are estimated thanks to an inversion
method, requiring a covariance model to interpolate the data, and they are compared
to a reference model. In more details, the numerical process is presented below and
sketched up in Fig. 10:
1. Step 1: Generation of the reference model of the disturbing potential T with pro-
gram GEOPOT [177], which allows to compute the gravity field related quantities
at given locations by using mainly a geopotential model. The long wavelengths
of the gravity field covered by the satellites and longer than the extent of the local
area are removed, providing centered or close to zero data for the determination
of a local covariance function. The terrain effects are removed with the help of
the topographic potential model dV_ELL_RET2012 [178];
2. Step 2: Generation of the synthetic data δg and T from a realistic spatial distri-
bution. A white noise is then added to δg and T , with a standard deviation of
0.1 m2 s−2 (i.e. 1 cm on the geoid) for clocks and 1 mGal for gravimetric mea-
surements;
3. Step 3: Estimation of the disturbing potential T from the synthetic data δg only
and then in combination with the synthetic data T on the 10-km grid using the
Least-Squares Collocation (LSC) method. In this step, a logarithmic 3D covari-
ance function is employed [179]. This model has the advantage to provide the
auto-covariances (ACF) and cross-covariances (CCF) of the potential T and its
derivatives in closed-form expressions. Parameters of this model are adjusted to
the empirical ACF of δg with the program GPFIT [180]. Note that low frequen-
cies are included in this covariance function, which were not removed as done in
step 1.
4. Final step : Evaluation of the potential recovery quality for selected data situations
by comparing the statistics of the residuals δ between the estimated values T and
the reference model T .
Let us underline that in this work, we use synthetic potential data while a network
of clocks would give access to potential differences between the clocks. We indeed
assume that the clock-based potential differences have been connected to one or a
few reference points, without introducing additional biases larger than the assumed
clock uncertainties. In order to have more realistic simulations, we should add
the noise due to uncertainty of the geometric coordinates of the clock, especially
the vertical component. This is a work in progress. However, if this error is below
74 P. Delva et al.
Fig. 10 Scheme of the numerical approach used to evaluate the contribution of atomic clocks
the accuracy of the clock, i.e. 0.1 m2 s−2 (1 cm on the geoid), it will not change the
main conclusions of this work.
6.2 Data Set
The locations of the gravimetric data are chosen to reproduce a realistic distribution of
measurements. Their spatial distribution can be obtained from the BGI (International
Gravimetric Bureau) database, then under-sampled by using a data reduction process,
as plotted in blue in Fig. 11. For this test case, the clock measurements (red markers)
are put only where existing land gravity data are located and in areas where the
gravity data coverage is poor. Moreover, in order to avoid clock points to be too
close to each other, a minimal distance is defined between them.
6.3 Contribution of Clocks
In Fig. 12, it is shown that adding the clock-based potential values to the
existing gravimetric data set can notably improve the reconstruction of the poten-
tial T . In Fig. 12a, the 4374 gravimetric data are used as input and the dis-
turbing potential is estimated with a bias μT ≈ 0.041 m2 s−2 (4.1 mm) and a
rms σT ≈ 0.25 m2 s−2 (2.5 cm). By combining the gravimetric measurements and
the 33 potential measurements, see Fig. 12a, the bias is improved by one order of
magnitude (μT ≈ −0.002 m2 s−2 or −0.2 mm) and the standard deviation by a fac-
tor 3 (σT ≈ 0.07 m2 s−2 or 7 mm). From the comparison of Fig. 12a, b it is clear that
the pure gravimetric solution exhibits a significant trend, which may be related to
Fig. 11 Spatial distribution of 4374 gravity data and 33 clock data used in the synthetic tests
the data collection area and covariance function used, while the additional potential
data effectively remove this trend.
Another important conclusion stemming from our simulations is that for solving
the problem of gravity field recovery, it is not required to have a dense clock network.
As shown in [65], only a very few percent of clock measurements compared to the
number of needed gravity data is sufficient. A more detailed study discussing the role
of different parameters, such as the noise level in the data, effects of the resolution
of gravity measurements and modeling errors can be found in [65].
As a result of this work, ways to optimize clock location points have begun in
order to answer to a practical question: where to put the geopotential measurements
to minimize the residuals and improve further the determination of the gravity field?
76 P. Delva et al.
(a) Without clock data.
(b) With clock data.
Fig. 12 Accuracy of the disturbing potential T reconstruction on a regular 10-km grid in Massif
Central, obtained by comparing the reference model and the reconstructed one. In figure (a), the
estimation is realized from the 4374 gravimetric data δg only, and in figure (b) by adding 33 potential
data T to the gravity data. To avoid edge effects in the estimated potential recovery, a grid edge
cutoff of 30 km has been removed in the solutions. Figures published in [65].
This is important when the gravimetric measurements can be tarnished by correlated

errors. For this, we have implemented the optimization of a spatial distribution of
clocks completing a pre-existing gravimetric network, by using the genetic algorithm
-MOEA (Multi Objective Evolutionary Algorithm, see [181]).
7 Conclusions
We presented in this chapter what is chronometric geodesy, introducing notions and

methods, both theoretical and experimental. The interest in this rather new topic
is raised by the tremendous ameliorations of atomic clocks in the last decade; it
is at the crossroads of general relativity and physical geodesy. On the one hand,
physical geodesy is essential in order to model the relativistic redshift in distant

clock comparisons, as well as to establish global timescales such as the TAI. On the
other hand, when the limitations of physical geodesy are reached in terms of method
inaccuracies, then the clock comparison observables, which give directly gravity
potential differences, could bring something new for physical geodesy.
The question whether these ideas will emerge one day as operational methods
depends to a large part on technological challenges. The development of sufficiently
accurate and transportable optical clocks is not a barrier, and several projects go in
this direction [105, 182, 183]. The frequency transfer method is more challenging,
especially on global scales (see Sect. 3.3). Optical fibre transfers fully meet the
expectations of current and future optical clocks, but are limited to continental scales
and are available only along predefined paths. Phase coherent free space optical links
are being developed, but are currently limited by the effect of atmospheric turbulence.
This method would be more adapted to global scales, especially if we think about
some islands in the middle of the ocean, which are unlikely to be linked with an
optical fibre.
Finally, we have to speak about the stability and integration of time of optical
clocks. Some recent techniques such as three-dimensional optical lattice clocks [104]
allow to greatly improve the integration times necessary to attain some specified
accuracy for the clock. This would permit, in a distant comparison, to obtain the
variations of the gravity potential with a good time resolution, and could lead to new
ideas for the study of geophysical phenomenon.
Acknowledgements The authors would like to thank Jérôme Lodewyck (SYRTE/Paris Observa-
tory) for providing Fig. 1, and Martina Sacher (Bundesamt für Kartographie und Geodäsie, BKG,
Leipzig, Germany) for providing information on the EVRF2007 heights and uncertainties, the
associated height transformations, and a new UELN adjustment in progress.
This research was supported by the European Metrology Research Programme (EMRP) within
the Joint Research Project “International Timescales with Optical Clocks” (SIB55 ITOC), as well
as the Deutsche Forschungsgemeinschaft (DFG) within the Collaborative Research Centre 1128
“Relativistic Geodesy and Gravimetry with Quantum Sensors (geo-Q)”, project C04. The EMRP
is jointly funded by the EMRP participating countries within EURAMET and the European Union.
We gratefully acknowledge financial support from Labex FIRST-TF and ERC AdOC (Grant No.
617553).
References
1. R.V. Pound, G.A. Rebka, Resonant absorption of the 14.4-kev γ ray from 0.10-μsec fe57 .
Phys. Rev. Lett. 3(12), 554–556 (1959)
2. R.V. Pound, G.A. Rebka, Gravitational red-shift in nuclear resonance. Phys. Rev. Lett. 3(9),
439–441 (1959)
3. R.V. Pound, G.A. Rebka, Apparent weight of photons. Phys. Rev. Lett. 4(7), 337–341 (1960)
4. R.V. Pound, J.L. Snider, Effect of gravity on gamma radiation. Phys. Rev. 140(3B), B788–
B803 (1965)
5. Norman F. Ramsey, History of early atomic clocks. Metrologia 42(3), S1 (2005)
6. Sigfrido Leschiutta, The definition of the ‘atomic’ second. Metrologia 42(3), S10 (2005)
78 P. Delva et al.
7. J. Terrien, News from the international bureau of weights and measures. Metrologia 4(1), 41
(1968)
8. Leonard S. Cutler, Fifty years of commercial caesium clocks. Metrologia 42(3), S90 (2005)
9. B. Guinot, E.F. Arias, Atomic time-keeping from 1955 to the present. Metrologia 42(3), S20
(2005)
10. J.C. Hafele, R.E. Keating, Around-the-world atomic clocks: predicted relativistic time gains.
Science 177(4044), 166–168 (1972)
11. J.C. Hafele, R.E. Keating, Around-the-world atomic clocks: observed relativistic time gains.
Science 177, 168–170 (1972)
12. L. Briatore, S. Leschiutta, Evidence for the Earth gravitational shift by direct atomic-time-
scale comparison. Nuovo Cim. B 37(2), 219–231 (1977)
13. Jean-Marc Lévy-Leblond, One more derivation of the Lorentz transformation. Am. J. Phys.
44(3), 271–277 (1976)
14. C.W. Misner, K.S. Thorne, J.A. Wheeler, Gravitation (W. H. Freeman, San Francisco, 1973)
15. Jeffrey M. Cohen, Harry E. Moses, New test of the synchronization procedure in noninertial
systems. Phys. Rev. Lett. 39(26), 1641–1643 (1977)
16. J.M. Cohen, H.E. Moses, A. Rosenblum, Clock-transport synchronization in noninertial
frames and gravitational fields. Phys. Rev. Lett. 51(17), 1501–1502 (1983)
17. J.M. Cohen, H.E. Moses, A. Rosenblum, Electromagnetic synchronisation of clocks with
finite separation in a rotating system. Class. Quantum Grav. 1(6), L57 (1984)
18. M.F. Podlaha, Note on the Cohen, Moses and Rosenblum letter about the slow-clock transport
synchronization in noninertial reference systems. Lett. Nuovo Cim. 40(7), 223–224 (1984)
19. N. Ashby, D.W. Allan, Coordinate time on and near the Earth. Phys. Rev. Lett. 53(19), 1858–
1858 (1984)
20. N. Ashby, D.W. Allan, Coordinate time on and near the Earth (erratum). Phys. Rev. Lett.
54(3), 254–254 (1985)
21. M. Born, Einstein’s Theory of Relativity (Dover Publications, New York, 1962)
22. C. Møller, Theory of Relativity, 2nd edn. (Oxford University Press, Oxford, 1976)
23. N. Ashby, D.W. Allan, Practical implications of relativity for a global coordinate time scale.
Radio Sci. 14(4), 649–669 (1979)
24. D.W. Allan, N. Ashby, Coordinate Time in the Vicinity of the Earth, vol. 114 (1986), pp.
299–312
25. A.J. Skalafuris, Current theoretical attempts toward synchronization of a global satellite net-
work. Radio Sci. 20(6), 1529–1536 (1985)
26. N. Ashby, Relativity in the global positioning system. Living Rev. Relativ. 6, 1 (2003)
27. B. Coll, A Principal Positioning System for the Earth, vol. 14 (2003), pp. 34–38,
arXiv:gr-qc/0306043
28. Carlo Rovelli, GPS observables in general relativity. Phys. Rev. D 65(4), 044017 (2002)
29. Marc Lachièze-Rey, The covariance of GPS coordinates and frames. Class. Quantum Grav.
23(10), 3531 (2006)
30. B. Coll, J.A. Morales, Symmetric frames on Lorentzian spaces. J. Math. Phys. 32(9),
2450–2455 (1991)
31. M. Blagojević, J. Garecki, F.W. Hehl, Y.N. Obukhov, Real null coframes in general relativity
and GPS type coordinates. Phys. Rev. D 65(4), 044018 (2002)
32. P. Delva, U. Kostić, A. Čadež, Numerical modeling of a Global Navigation Satellite System
in a general relativistic framework. Adv. Space Res. 47(2), 370–379 (2011)
33. D. Bini, A. Geralico, M.L. Ruggiero, A. Tartaglia, Emission versus Fermi coordinates: appli-
cations to relativistic positioning systems. Class. Quantum Grav. 25(20), 205011 (2008)
34. A. Tartaglia, Emission coordinates for the navigation in space. Acta Astronaut. 67(5), 539–545
(2010)
35. N. Puchades, D. Sáez, Relativistic positioning: four-dimensional numerical approach in
Minkowski space-time. Astrophys. Space Sci. 341(2), 631–643 (2012)
36. D. Bunandar, S.A. Caveny, R.A. Matzner, Measuring emission coordinates in a pulsar-based
relativistic positioning system. Phys. Rev. D 84(10), 104005 (2011)
37. V.A. Brumberg, General discussion, in Relativity in celestial mechanics and astrometry, pro-
ceedings of the IAU symposium No.114, ed. by J. Kovalesky and V.A. Brumberg (D. Reidel
publishing company, 1986)
38. B. Guinot, Is the International Atomic Time TAI a coordinate time or a proper time? Celest.
Mech. 38, 155–161 (1986)
39. B. Guinot, P.K. Seidelmann, Time scales - their history, definition and interpretation. Astron.
Astrophys. 194, 304–308 (1988)
40. T.-Y. Huang, B.-X. Xu, J. Zhu, H. Zhang, The concepts of International Atomic Time (TAI)
and Terrestrial Dynamic Time (TDT). Astron. Astrophys. 220, 329–334 (1989)
41. V.A. Brumberg, S.M. Kopeikin, Relativistic time scales in the solar system. Celest. Mech.
Dyn. Astron. 48, 23–44 (1990)
42. S.A. Klioner, The problem of clock synchronization: a relativistic approach. Celest. Mech.
Dyn. Astron. 53(1), 81–109 (1992)
43. N. Ashby, B. Bertotti, Relativistic perturbations of an earth satellite. Phys. Rev. Lett. 52(7),
485–488 (1984)
44. T. Fukushima, The Fermi coordinate system in the post-Newtonian framework. Celest. Mech.
44, 61–75 (1988)
45. N. Ashby, B. Bertotti, Relativistic effects in local inertial frames. Phys. Rev. D 34(8), 2246–
2259 (1986)
46. S.M. Kopejkin, Celestial coordinate reference systems in curved space-time. Celest. Mech.
44, 87–115 (1988)
47. M.H. Soffel, Relativity in Astrometry, Celestial Mechanics, and Geodesy (Springer, Berlin,
1989)
48. V.A. Brumberg, S.M. Kopejkin, Relativistic theory of celestial reference frames, in Reference
Frames, Astrophysics and Space Science Library, vol. 154, ed. by J. Kovalevsky, I.I. Mueller,
B. Kolaczek (Springer, Netherlands, 1989), pp. 115–141
49. V.A. Brumberg, S.M. Kopejkin, Relativistic reference systems and motion of test bodies in
the vicinity of the earth. Nuovo Cim. 103(1), 63–98 (1989)
50. T. Damour, M. Soffel, C. Xu, General-relativistic celestial mechanics. I. Method and definition
of reference systems. Phys. Rev. D 43(10), 3273–3307 (1991)
51. S.M. Kopeikin, M. Efroimsky, G. Kaplan, Relativistic Celestial Mechanics of the Solar System
(Wiley, New York, 2011)
52. U. Kostić, M. Horvat, A. Gomboc, Relativistic positioning system in perturbed spacetime.
Class. Quantum Grav. 32(21), 215004 (2015)
53. D. Bini, B. Mashhoon, Relativistic gravity gradiometry. Phys. Rev. D 94(12), 124009 (2016)
54. M. Soffel, S.A. Klioner, G. Petit, P. Wolf, S.M. Kopeikin, P. Bretagnon, V.A. Brumberg,
N. Capitaine, T. Damour, T. Fukushima, B. Guinot, T.-Y. Huang, L. Lindegren, C. Ma, K.
Nordtvedt, J.C. Ries, P.K. Seidelmann, D. Vokrouhlický, C.M. Will, C. Xu, The IAU 2000
resolutions for astrometry, celestial mechanics, and metrology in the relativistic framework:
explanatory supplement. Astron. J. 126(6), 2687 (2003)
55. S.A. Klioner, N. Capitaine, W.M. Folkner, B. Guinot, T.-Y. Huang, S.M. Kopeikin, E.V. Pit-
jeva, P.K. Seidelmann, M.H. Soffel, Units of relativistic time scales and associated quantities,
in Relativity in Fundamental Astronomy: Dynamics, Reference Frames, and Data Analysis,
Proceedings of the International Astronomical Union, vol. 5 (2009), pp. 79–84
56. E.F. Arias, G. Panfilo, G. Petit, Timescales at the BIPM. Metrologia 48(4), S145 (2011)
57. A. Bjerhammar, Discrete approaches to the solution of the boundary value problem in physical
geodesy. Bolletino di geodesia e scienze affini 2, 185–241 (1975)
58. M. Vermeer, Chronometric levelling, Technical report, Finnish Geodetic Institute, Helsinki
(1983)
59. A. Bjerhammar, On a relativistic geodesy. Bull. Geodesique 59(3), 207–220 (1985)
60. H.S. Margolis, R.M. Godun, P. Gill, L.A.M. Johnson, S.L. Shemar, P.B. Whibberley,
D. Calonico, F. Levi, L. Lorini, M. Pizzocaro, P. Delva, S. Bize, J. Achkar, H. Denker, L. Tim-
men, C. Voigt, S. Falke, D. Piester, C. Lisdat, U. Sterr, S. Vogt, S. Weyers, J. Gersl, T. Lindvall,
M. Merimaa, International timescales with optical clocks (ITOC), in European Frequency and
80 P. Delva et al.
Time Forum International Frequency Control Symposium (EFTF/IFC), 2013 Joint (2013), pp.
908–911
61. J. Grotti, S. Koller, S. Vogt, S. Häfner, U. Sterr, C. Lisdat, H. Denker, C. Voigt, L. Timmen,
A. Rolland, F.N. Baynes, H.S. Margolis, M. Zampaolo, P. Thoumany, M. Pizzocaro, B. Rauf,
F. Bregolin, A. Tampellini, P. Barbieri, M. Zucco, G.A. Costanzo, C. Clivati, F. Levi, D.
Calonico, Geodesy and metrology with a transportable optical. Nat. Phys. 14(5), 437 (2018)
62. V.A. Brumberg, E. Groten, On determination of heights by using terrestrial clocks and GPS
signals. J. Geod. 76(1), 49–54 (2002)
63. R. Bondarescu, M. Bondarescu, G. Hetényi, L. Boschi, P. Jetzer, J. Balakrishna, Geophysical
applicability of atomic clocks: direct continental geoid mapping. Geophys. J. Int. 191(1),
78–82 (2012)
64. C.W. Chou, D.B. Hume, T. Rosenband, D.J. Wineland, Optical clocks and relativity. Science
329(5999), 1630–1633 (2010)
65. G. Lion, I. Panet, P. Wolf, C. Guerlin, S. Bize, P. Delva, Determination of a high spatial
resolution geopotential model using atomic clock comparisons. J. Geod. (2017), pp. 1–15
66. A. Bjerhammar, Relativistic geodesy. Technical Report NON118 NGS36, NOAA Technical
Report (1986)
67. M. Soffel, H. Herold, H. Ruder, M. Schneider, Relativistic theory of gravimetric measurements
and definition of the geoid. Manuscr. Geod. 13, 143–146 (1988)
68. B. Hofmann-Wellenhof, H. Moritz, Physical Geodesy (Springer Science & Business Media,
2006)
69. S.M. Kopejkin, Relativistic manifestations of gravitational fields in gravimetry and geodesy.
Manuscr. Geod. 16, 301–312 (1991)
70. J. Müller, M. Soffel, S.A. Klioner, Geodesy and relativity. J. Geod. 82(3), 133–145 (2007)
71. S.M. Kopeikin, E.M. Mazurova, A.P. Karpik, Towards an exact relativistic theory of Earth’s
geoid undulation. Phys. Lett. A 379(26–27), 1555–1562 (2015)
72. S.M. Kopeikin, W. Han, E. Mazurova, Post-Newtonian reference ellipsoid for relativistic
geodesy. Phys. Rev. D 93(4), 044069 (2016)
73. S.M. Kopeikin, Reference ellipsoid and geoid in chronometric geodesy. Front. Astron. Space
Sci. 3 (2016)
74. J. Guéna, S. Weyers, M. Abgrall, C. Grebing, V. Gerginov, P. Rosenbusch, S. Bize, B. Lip-
phardt, H. Denker, N. Quintin, S.M.F. Raupach, D. Nicolodi, F. Stefani, N. Chiodo, S. Koke,
A. Kuhl, F. Wiotte, F. Meynadier, E. Camisard, C. Chardonnet, Y. Le Coq, M. Lours, G.
Santarelli, A. Amy-Klein, R. Le Targat, O. Lopez, P.E. Pottie, G. Grosche, First international
comparison of fountain primary frequency standards via a long distance optical fiber link.
Metrologia 54(3), 348 (2017)
75. C. Lisdat, G. Grosche, N. Quintin, C. Shi, S.M.F. Raupach, C. Grebing, D. Nicolodi, F. Stefani,
A. Al-Masoudi, S. Dörscher, S. Häfner, J.-L. Robyr, N. Chiodo, S. Bilicki, E. Bookjans, A.
Koczwara, S. Koke, A. Kuhl, F. Wiotte, F. Meynadier, E. Camisard, M. Abgrall, M. Lours, T.
Legero, H. Schnatz, U. Sterr, H. Denker, C. Chardonnet, Y. Le Coq, G. Santarelli, A. Amy-
Klein, R. Le Targat, J. Lodewyck, O. Lopez, P.-E. Pottie, A clock network for geodesy and
fundamental science. Nat. Commun. 7, 12443 (2016)
76. M. Schioppo, R.C. Brown, W.F. McGrew, N. Hinkley, R.J. Fasano, K. Beloy, T.H. Yoon,
G. Milani, D. Nicolodi, J.A. Sherman, N.B. Phillips, C.W. Oates, A.D. Ludlow, Ultrastable
optical clock with two cold-atom ensembles. Nat. Photonics 11(1), 48–52 (2017)
77. N. Huntemann, C. Sanner, B. Lipphardt, Chr. Tamm, E. Peik, Single-ion atomic clock with
3 × 10−18 systematic uncertainty. Phys. Rev. Lett. 116(6), 063001 (2016)
78. H.S. Margolis, P. Gill, Least-squares analysis of clock frequency comparison data to deduce
optimized frequency and frequency ratio values. Metrologia 52(5), 628 (2015)
79. H.S. Margolis, P. Gill, Determination of optimized frequency and frequency ratio values from
over-determined sets of clock comparison data. J. Phys.: Conf. Ser. 723(1), 012060 (2016)
80. P. Wolf, G. Petit, Relativistic theory for clock syntonization and the realization of geocentric
coordinate times. Astron. Astrophys. 304, 653 (1995)
81. H. Denker, L. Timmen, C. Voigt, S. Weyers, E. Peik, H.S. Margolis, P. Delva, P. Wolf, G. Petit,
Geodetic methods to determine the relativistic redshift at the level of 10−18 in the context of
international timescales – a review and practical results. J. Geod. 92(5), 487–516 (2018)
82. W. Torge, Geodesy, 2nd edn. (Berlin; New York, W. de Gruyter, 1991)
83. A. Bauch, Time and frequency comparisons using radiofrequency signals from satellites.
Comptes Rendus Phys. 16(5), 471–479 (2015)
84. E. Samain, Clock comparison based on laser ranging technologies. Int. J. Mod. Phys. D 24(08),
1530021 (2015)
85. G. Petit, A. Kanj, S. Loyer, J. Delporte, F. Mercier, F. Perosanz, 1 × 10−16 frequency transfer
by GPS PPP with integer ambiguity resolutio. Metrologia 52(2), 301 (2015)
86. S. Droste, C. Grebing, J. Leute, S.M.F. Raupach, A. Matveev, T.W. Hänsch, A. Bauch, R.
Holzwarth, G. Grosche, Characterization of a 450 km baseline GPS carrier-phase link using
an optical fiber link. New J. Phys. 17(8), 083044 (2015)
87. J. Leute, N. Huntemann, B. Lipphardt, C. Tamm, P.B.R. Nisbet-Jones, S.A. King, R.M. Godun,
J.M. Jones, H.S. Margolis, P.B. Whibberley, A. Wallin, M. Merimaa, P. Gill, E. Peik, Frequency
comparison of 171 Yb+ ion optical clocks at ptb and npl via GPS PPP. IEEE Trans. Ultrason.
Ferroelectr. Freq. Control 63(7), 981–985 (2016)
88. P. Dubé, J.E. Bernard, M. Gertsvolf, Absolute frequency measurement of the 88 Sr + clock
transition using a GPS link to the SI second. Metrologia 54(3), 290 (2017)
89. C.F.A. Baynham, R.M. Godun, J.M. Jones, S.A. King, P.B.R. Nisbet-Jones, F. Baynes, A.
Rolland, P.E.G. Baird, K. Bongs, P. Gill, H.S. Margolis, Absolute frequency measurement of
the optical clock transition in with an uncertainty of using a frequency link to international
atomic time. J. Mod. Opt. 65(2), 221–227 (2018)
90. D. Kirchner, Two-way time transfer via communication satellites. Proc. IEEE 79(7), 983–990
(1991)
91. D. Piester, A. Bauch, L. Breakiron, D. Matsakis, B. Blanzano, O. Koudelka, Time transfer
with nanosecond accuracy for the realization of International atomic time. Metrologia 45(2),
185 (2008)
92. M. Fujieda, T. Gotoh, J. Amagai, Advanced two-way satellite frequency transfer by carrier-
phase and carrier-frequency measurements. J. Phys.: Conf. Ser. 723(1), 012036 (2016)
93. H. Hachisu, M. Fujieda, S. Nagano, T. Gotoh, A. Nogami, T. Ido, St Falke, N. Huntemann,
C. Grebing, B. Lipphardt, Ch. Lisdat, D. Piester, Direct comparison of optical lattice clocks
with an intercontinental baseline of 9000 km. Opt. Lett. 39(14), 4072–4075 (2014)
94. F. Meynadier, P. Delva, C. le Poncin-Lafitte, C. Guerlin, P. Wolf, Atomic clock ensemble in
space (ACES) data analysis. Class. Quantum Grav. 35(3), 035018 (2018)
95. L. Cacciapuoti, Ch. Salomon, Space clocks and fundamental tests: The ACES experiment.
Eur. Phys. J. Spec. Top. 172(1), 57–68 (2009)
96. Ph. Laurent, D. Massonnet, L. Cacciapuoti, C. Salomon, The ACES/PHARAO space mission.
Comptes rendus de l’Académie des sciences. Physique 16(5), 540–552 (2015)
97. P. Exertier, E. Samain, C. Courde, M. Aimar, J.M. Torre, G.D. Rovera, M. Abgrall, P. Uhrich,
R. Sherwood, G. Herold, U. Schreiber, P. Guillemot, Sub-ns time transfer consistency: a direct
comparison between GPS CV and T2L2. Metrologia 53(6), 1395 (2016)
98. G.D. Rovera, M. Abgrall, C. Courde, P. Exertier, P. Fridelance, Ph. Guillemot, M. Laas-Bourez,
N. Martin, E. Samain, R. Sherwood, J.-M. Torre, P. Uhrich, A direct comparison between two
independently calibrated time transfer techniques: T2L2 and GPS Common-Views. J. Phys.:
Conf. Ser. 723(1), 012037 (2016)
99. E. Samain, P. Vrancken, P. Guillemot, P. Fridelance, P. Exertier, Time transfer by laser link
(T2L2): characterization and calibration of the flight instrument. Metrologia 51(5), 503 (2014)
100. E. Samain, P. Exertier, C. Courde, P. Fridelance, P. Guillemot, M. Laas-Bourez, J.-M. Torre,
Time transfer by laser link: a complete analysis of the uncertainty budget. Metrologia 52(2),
423 (2015)
101. P. Exertier, E. Samain, N. Martin, C. Courde, M. Laas-Bourez, C. Foussard, Ph. Guillemot,
Time transfer by laser link: data analysis and validation to the ps level. Adv. Space Res. 54(11),
2371–2385 (2014)
82 P. Delva et al.
102. J. Kodet, M. Vacek, P. Fort, I. Prochazka, J. Blazej, Photon Counting Receiver for the Laser
Time Transfer, Optical Design, and Construction, vol. 8072 (2011), pp. 80720A (International
Society for Optics and Photonics, 2011)
103. I. Prochazka, J. Kodet, J. Blazej, Note: space qualified photon counting detector for laser time
transfer with picosecond precision and stability. Rev. Sci. Instrum. 87(5), 056102 (2016)
104. S.L. Campbell, R.B. Hutson, G.E. Marti, A. Goban, N. Darkwah Oppong, R.L. McNally,
L. Sonderhouse, J.M. Robinson, W. Zhang, B.J. Bloom, J. Ye, A fermi-degenerate three-
dimensional optical lattice clock. Science 358(6359), 90–94 (2017)
105. S.B. Koller, J. Grotti, St. Vogt, A. Al-Masoudi, S. Dörscher, S. Häfner, U. Sterr, Ch. Lisdat,
Transportable optical lattice clock with 7 × 10−17 uncertainty. Phys. Rev. Lett. 118(7), 073601
(2017)
106. R. Tyumenev, M. Favier, S. Bilicki, E. Bookjans, R. Le Targat, J. Lodewyck, D. Nicolodi,
Y. Le Coq, M. Abgrall, J. Guéna, L. De Sarlo, S. Bize, Comparing a mercury optical lattice
clock with microwave and optical frequency standards. New J. Phys. 18(11), 113002 (2016)
107. J. Lodewyck, S. Bilicki, E. Bookjans, J.-L. Robyr, C. Shi, G. Vallet, R. Le Targat, D. Nicolodi,
Y. Le Coq, J. Guéna, M. Abgrall, P. Rosenbusch, S. Bize, Optical to microwave clock frequency
ratios with a nearly continuous strontium optical lattice clock. Metrologia 53(4), 1123 (2016)
108. K. Predehl, G. Grosche, S.M.F. Raupach, S. Droste, O. Terra, J. Alnis, Th Legero, T.W.
Hänsch, Th Udem, R. Holzwarth, H. Schnatz, A 920-kilometer optical fiber link for frequency
metrology at the 19th decimal place. Science 336(6080), 441–444 (2012)
109. O. Lopez, A. Haboucha, B. Chanteau, C. Chardonnet, A. Amy-Klein, G. Santarelli, Ultra-
stable long distance optical frequency distribution using the Internet fiber network. Opt.
Express 20(21), 23518–23526 (2012)
110. N. Chiodo, K. Djerroud, O. Acef, A. Clairon, P. Wolf, Lasers for coherent optical satellite
links with large dynamics. Appl. Opt. 52(30), 7342–7351 (2013)
111. K. Djerroud, O. Acef, A. Clairon, P. Lemonde, C.N. Man, E. Samain, P. Wolf, Coherent optical
link through the turbulent atmosphere. Opt. Lett. 35(9), 1479–1481 (2010)
112. F.R. Giorgetta, W.C. Swann, L.C. Sinclair, E. Baumann, I. Coddington, N.R. Newbury, Optical
two-way time and frequency transfer over free space. Nat. Photonics 7(6), 434–438 (2013)
113. J.-D. Deschênes, L.C. Sinclair, F.R. Giorgetta, W.C. Swann, E. Baumann, H. Bergeron,
M. Cermak, I. Coddington, N.R. Newbury, Synchronization of distant optical clocks at the
femtosecond level. Phys. Rev. X 6(2), 021016 (2016)
114. L.C. Sinclair, F.R. Giorgetta, W.C. Swann, E. Baumann, I. Coddington, N.R. Newbury, Optical
phase noise from atmospheric fluctuations and its impact on optical time-frequency transfer.
Phys. Rev. A 89(2), 023805 (2014)
115. L.C. Sinclair, W.C. Swann, H. Bergeron, E. Baumann, M. Cermak, I. Coddington, J.-D.
Deschênes, F.R. Giorgetta, J.C. Juarez, I. Khader, K.G. Petrillo, K.T. Souza, M.L. Dennis,
N.R. Newbury, Synchronization of clocks through 12 km of strongly turbulent air over a city.
Appl. Phys. Lett. 109(15), 151104 (2016)
116. C. Robert, J.-M. Conan, P. Wolf, Impact of turbulence on high-precision ground-satellite
frequency transfer with two-way coherent optical links. Phys. Rev. A 93(3), 033860 (2016)
117. L. Blanchet, C. Salomon, P. Teyssandier, P. Wolf, Relativistic theory for time and frequency
transfer to order. Astron. Astrophys. 370(1), 10 (2001)
118. C. Le Poncin-Lafitte, B. Linet, P. Teyssandier, World function and time transfer: general
post-Minkowskian expansions. Class. Quantum Grav. 21(18), 4463 (2004)
119. P. Teyssandier, C. Le Poncin-Lafitte, General post-Minkowskian expansion of time transfer
functions. Class. Quantum Grav. 25(14), 145020 (2008)
120. A. Hees, S. Bertone, C. Le Poncin-Lafitte, Frequency shift up to the 2-PM approximation,
in SF2A-2012: Proceedings of the annual meeting of the French Society of Astronomy and
Astrophysics (2012), pp. 145–148, arXiv:1210.2577
121. J.L. Synge, Relativity: The General Theory, 1st edn. (North-Holland Publishing Company,
Amsterdam, 1960)
122. J. Gers̆l, P. Delva, P. Wolf, Relativistic corrections for time and frequency transfer in optical
fibres. Metrologia 52(4), 552 (2015)
123. A. Rülke, G. Liebsch, M. Sacher, U. Schäfer, U. Schirmer, J. Ihde, Unification of European

height system realizations. J. Geod. Sci. 2(4), 343–354 (2013)
124. G. Petit, P. Wolf, P. Delva, Atomic time, clocks, and clock comparisons in relativistic space-
time: a review, in Frontiers in Relativistic Celestial Mechanics - Volume 2: Applications and
Experiments, ed. by S.M. Kopeikin. De Gruyter Studies in Mathematical Physics (De Gruyter,
2014), pp. 249–279
125. L. Sánchez, Towards a vertical datum standardisation under the umbrella of Global Geodetic
Observing System. J. Geod. Sci. 2(4), 325–342 (2012)
126. M. Burs̆a, S. Kenyon, J. Kouba, Z. S̆íma, V. Vatrt, V. Vtek, M. Vojtís̆ková, The geopotential
value W0 for specifying the relativistic atomic time scale and a global vertical reference
system. J. Geod. 81(2), 103–110 (2006)
127. N. Dayoub, S.J. Edwards, P. Moore, The Gauss-Listing geopotential value W0 and its rate
from altimetric mean sea level and GRACE. J. Geod. 86(9), 681–694 (2012)
128. S. Jevrejeva, J.C. Moore, A. Grinsted, Sea level projections to AD2500 with a new generation
of climate change scenarios. Glob. Planet. Chang. 80–81, 14–20 (2012)
129. C. Voigt, H. Denker, L. Timmen, Time-variable gravity potential components for optical clock
comparisons and the definition of international time scales. Metrologia 53(6), 1365 (2016)
130. T. Takano, M. Takamoto, I. Ushijima, N. Ohmae, T. Akatsuka, A. Yamaguchi, Y. Kuroishi,
H. Munekane, B. Miyahara, H. Katori, Real-time geopotentiometry with synchronously linked
optical lattice clocks. Nat. Photonics 10(10), 662–666 (2016)
131. P. Delva, J. Lodewyck, S. Bilicki, E. Bookjans, G. Vallet, R. Le Targat, P.-E. Pottie, C. Guerlin,
F. Meynadier, C. Le Poncin-Lafitte, others, Test of special relativity using a fiber network of
optical clocks. Phys. Rev. Lett. 118(22), 221102 (2017)
132. J. Mäkinen, J. Ihde, The permanent tide in height systems, in Observing our Changing Earth,
International Association of Geodesy Symposia (Springer, Berlin, 2009), pp. 81–87
133. J. Ihde, J. Mäkinen, M. Sacher, Conventions for the definition and realization of a European
Vertical Reference System (EVRS) – EVRS Conventions 2007. EVRS Conventions V5.1,
Bundesamt für Kartographie and Geodäsie, Finnish Geodetic Institute, 2008-12-17. Technical
report, 2008
134. H. Denker, Regional gravity field modeling: theory and practical results, in Sciences of
Geodesy, vol. II, ed. by G. Xu (Springer, Berlin, 2013), pp. 185–291
135. Bull. Geodesique 58(3), 309–323 (1984)
136. W.A. Heiskanen, H. Moritz, Physical Geodesy (W.H. Freeman and Company, San Francisco,
London, 1967)
137. W. Torge, J. Müller, Geodesy, 4th edn. (De Gruyter, Berlin, Boston, 2012)
138. F. Condi, C. Wunsch, Gravity field variability, the geoid, and ocean dynamics, in V Hotine-
Marussi Symposium on Mathematical Geodesy, International Association of Geodesy Sym-
posia, vol. 127 (Springer, Berlin, 2004), pp. 285–292
139. H. Drewes, F. Kuglitsch, J. Adám, S. Rózsa, The Geodesist’s handbook 2016. J. Geod. 90(10),
907–1205 (2016)
140. J. Ihde, R. Barzaghi, U. Marti, L. Sànchez, M. Sideris, H. Drewes, C. Foerste, T. Gruber,
G. Liebsch, G. Pail, Report of the ad-hoc group on an international height reference system
(IHRS), in Reports 2011-2015, Number 39 in IAG Travaux (2015), pp. 549–557
141. M. Burša, J. Kouba, M. Kumar, A. Müller, K. Raděj, S.A. True, V. Vatrt, M. Vojtíšková,
Geoidal geopotential and world height system. Stud. Geophys. Geod. 43(4), 327–337 (1999)
142. L. Sánchez, R. C̆underlk, N. Dayoub, K. Mikula, Z. Minarechová, Z. S̆íma, V. Vatrt,
M. Vojtís̆ková, A conventional value for the geoid reference potential W0 . J. Geod. 90(9),
815–835 (2016)
143. M.S. Molodenskii, V.F. Eremeev, M.I. Yurkina, Methods for Study of the External Gravita-
tional Field and Figure of the Earth (Israel Program for Scientific Translations, Jerusalem,
1962)
144. N.K. Pavlis, S.A. Holmes, S.C. Kenyon, J.K. Factor, The development and evaluation of the
Earth Gravitational Model 2008 (EGM2008). J. Geophys. Res. 117(B4), B04406 (2012)
84 P. Delva et al.
145. T. Mayer-Gürr, N. Zehentner, B. Klinger, A. Kvas, ITSG-Grace2014: a new GRACE gravity

field release computed in Graz, 2014. GRACE Science Team Meeting (GSTM), Potsdam, 29
Sept.-01 Oct. 2014
146. T. Mayer-Gürr, G. Team, The combined satellite gravity field model GOCO05s, 2015. EGU
General Assembly 2015, Vienna (2015)
147. J.M. Brockmann, N. Zehentner, E. Höck, R. Pail, I. Loth, T. Mayer-Gürr, W.-D. Schuh,
EGM_TIM_RL05: an independent geoid with centimeter accuracy purely based on the GOCE
mission. Geophys. Res. Lett. 41(22), 8089–8099 (2014)
148. D.E. Smith, R. Kolenkiewicz, P.J. Dunn, M.H. Torrence, Earth scale below a part per billion
from Satellite Laser Ranging, in Geodesy Beyond 2000, International Association of Geodesy
Symposia, vol. 121 (Springer, Berlin, 2000), pp. 3–12
149. J.C. Ries, The scale of the terrestrial reference frame from VLBI and SLR, 2014. IERS Unified
Analysis Workshop, Pasadena, CA, 27–28 June 2014
150. R. Forsberg, Modelling the fine-structure of the geoid: methods, data requirements and some
results. Surv. Geophys. 14(4–5), 403–418 (1993)
151. C. Jekeli, O. Error, Data requirements, and the fractal dimension of the geoid, in VII Hotine-
Marussi Symposium on Mathematical Geodesy, International Association of Geodesy Sym-
posia, vol. 137 (Springer, Berlin, 2012), pp. 181–187
152. R. Rummel, P. Teunissen, Height datum definition, height datum connection and the role of
the geodetic boundary value problem. Bull. Geodesique 62(4), 477–498 (1988)
153. B. Heck, R. Rummel, Strategies for solving the vertical datum problem using terrestrial and
satellite geodetic data, in Sea Surface Topography and the Geoid, International Association
of Geodesy Symposia, vol. 104 (Springer, New York, 1990), pp. 116–128
154. J. Kelsey, D.A. Gray, Geodetic aspects concerning possible subsidence in southeastern Eng-
land. Phil. Trans. R. Soc. Lond. A 272(1221), 141–149 (1972)
155. P. Rebischung, H. Duquenne, F. Duquenne, The new French zero-order levelling network –
first global results and possible consequences for UELN, 2008, EUREF Symposium, Brussels,
June 18-21, 2008 (2008)
156. M. Véronneau, R. Duvai, J. Huang, A gravimetric geoid model as a vertical datum in Canada.
Geomatica 60, 165–172 (2006)
157. D. Smith, M. Véronneau, D.R. Roman, J. Huang, Y. Wang, M. Sideris, Towards the unification
of the vertical datums over the North American continent, 2010. IAG Comm. 1 Symposium
2010, Reference Frames for Applications in Geosciences (REFAG2010), Marne-La-Vallée,
France, 4–8 Oct. 2010 (2010)
158. D.A. Smith, S.A. Holmes, X. Li, S. Guillaume, Y.M. Wang, B. Bürki, D.R. Roman, T.M.
Damiani, Confirming regional 1 cm differential geoid accuracy from airborne gravimetry: the
Geoid slope validation survey of 2011. J. Geod. 87(10–12), 885–907 (2013)
159. Z. Altamimi, X. Collilieux, L. Métivier, ITRF2008: an improved solution of the international
terrestrial reference frame. J. Geod. 85(8), 457–473 (2011)
160. Z. Altamimi, P. Rebischung, L. Métivier, X. Collilieux, ITRF2014: a new release of the
International Terrestrial Reference Frame modeling nonlinear station motions. J. Geophys.
Res. Solid Earth 121, 6109–6131 (2016)
161. M. Seitz, D. Angermann, H. Drewes, Accuracy assessment of the ITRS 2008 realization
of DGFI: DTRF2008, in Reference Frames for Applications in Geosciences, International
Association of Geodesy Symposia, vol. 138 (Springer, Berlin, 2013), pp. 87–93
162. H. Margolis, Timekeepers of the future (2014), https://www.nature.com/articles/nphys2834
163. F. Riehle, Towards a redefinition of the second based on optical atomic clocks. Comptes
Rendus Phys. 16(5), 506–515 (2015)
164. P. Gill, Is the time right for a redefinition of the second by optical atomic clocks? J. Phys.:
Conf. Ser. 723(1), 012053 (2016)
165. F. Riehle, Optical clock networks. Nat. Photonics 11(1), 25–31 (2017)
166. S. Falke, N. Lemke, C. Grebing, B. Lipphardt, S. Weyers, G. Vladislav, N. Huntemann, C.
Hagemann, A. Al-Masoudi, S. Häfner, S. Vogt, S. Uwe, C. Lisdat, A strontium lattice clock
with 3 × 10−17 inaccuracy and its frequency. New J. Phys. 16(7), 073023 (2014)
167. H. Denker, A new European Gravimetric (Quasi)Geoid EGG2015, 2015. XXVI General
Assembly of the International Union of Geodesy and Geophysics (IUGG), Earth and Envi-
ronmental Sciences for Future Generations, Prague, Czech Republic, 22 June–02 July 2015
(Poster)
168. M. Sacher, J. Ihde, G. Liebsch, J. Mäkinen, EVRF2007 as realization of the European vertical
reference system, 2008, EUREF Symposium, Brussels, Belgium, 18–21 June 2008 (2008)
169. D.E. Cartwright, J. Crease, A comparison of the geodetic reference levels of England and
France by means of the sea surface. Proc. R. Soc. Lond. A 273(1355), 558–580 (1963)
170. M. Greaves, R. Hipkin, C. Calvert, C. Fane, P. Rebischung, F. Duquenne, A. Harmel,
A. Coulomb, H. Duquenne, Connection of British and French levelling networks – Applica-
tions to UELN, 2007, EUREF Symposium, London, June 6–9, 2007 (2007)
171. A. Kenyeres, M. Sacher, J. Ihde, H. Denker, U. Marti, EUVN densification action – final
report. Technical report (2010), https://evrs.bkg.bund.de/SharedDocs/Downloads/EVRS/
EN/Publications/EUVN-DA_FinalReport.pdf?__blob=publicationFile&v=1
172. H. Moritz, Geodetic reference system 1980. J. Geod. 74(1), 128–133 (2000)
173. H. Moritz, Advanced Physical Geodesy (Wichmann, Karlsruhe, 1980)
174. H. Denker, Hochauflösende regionale Schwerefeldbestimmung mit gravimetrischen und
topographischen Daten. PhD thesis, Wiss. Arb. d. Fachr. Verm.wesen d. Univ. Hannover,
Nr. 156, Hannover, 1988
175. H. Denker, Evaluation and improvement of the EGG97 quasigeoid model for Europe by GPS
and leveling data, in Second Continental Workshop on the Geoid in Europe, Proceed., Report
of Finnish Geodetic Institute, Masala, vol. 98:4, ed. by M. Vermeer, J. Ádám (1998), pp.
53–61
176. C. Förste, S.L. Bruinsma, O. Abrikosov, J.-M. Lemoine, J.C. Marty, F. Flechtner, G. Balmino,
F. Barthelmes, R. Biancale, EIGEN-6C4 The latest combined global gravity field model
including GOCE data up to degree and order 2190 of GFZ Potsdam and GRGS Toulouse
(2014)
177. D.A. Smith, There is no such thing as “The” EGM96 geoid: subtle points on the use of a global
geopotential model. Technical report, IGeS Bulletin No. 8, International Geoid Service, Milan,
Italy, 1998
178. S.J. Claessens, C. Hirt, Ellipsoidal topographic potential: New solutions for spectral forward
gravity modeling of topography with respect to a reference ellipsoid. J. Geophys. Res. Solid
Earth 118(11), 5991–6002 (2013)
179. R. Forsberg, A new covariance model for inertial gravimetry and gradiometry. J. Geophys.
Res. 92(B2), 1305–1310 (1987)
180. R. Forsberg. An overview manual for the GRAVSOFT – Geodetic Gravity Field Modelling
Programs. Technical report, DNSC – Danios National Space Center, 2003
181. D. Coulot, P. Rebischung, A. Pollet, L. Grondin, G. Collot, Global optimization of GNSS
station reference networks. GPS Solut. 19(4), 569–577 (2015)
182. J. Cao, P. Zhang, J. Shang, K. Cui, J. Yuan, S. Chao, S. Wang, H. Shu, X. Huang, A compact,
transportable single-ion optical clock with 7.8 × 1017 systematic uncertainty. Appl. Phys. B
123(4), 112 (2017)
183. M. Yasuda, T. Tanabe, T. Kobayashi, D. Akamatsu, T. Sato, A. Hatakeyama, Laser-controlled
cold ytterbium atom source for transportable optical clocks. J. Phys. Soc. Jpn. 86(12), 125001
(2017)
Measuring the Gravitational Field
in General Relativity: From Deviation
Equations and the Gravitational
Compass to Relativistic Clock
Gradiometry
Yuri N. Obukhov and Dirk Puetzfeld
Abstract How does one measure the gravitational field? We give explicit answers
to this fundamental question and show how all components of the curvature tensor,
which represents the gravitational field in Einstein’s theory of General Relativity,
can be obtained by means of two different methods. The first method relies on the
measuring the accelerations of a suitably prepared set of test bodies relative to the
observer. The second method utilizes a set of suitably prepared clocks. The methods
discussed here form the basis of relativistic (clock) gradiometry and are of direct
operational relevance for applications in geodesy.
1 Introduction
The measurement of the gravitational field lies at the heart of gravitational physics
and geodesy. Here we provide the relativistic foundation and present two methods
for the operational determination of the gravitational field.
In Einstein’s theory of gravitation, i.e. General Relativity (GR), the gravitational
field manifests itself in the form of the Riemannian curvature tensor Rabc d [1, 2].
This 4th-rank tensor can be defined as a measure of the noncommutativity of the
parallel transport process of the underlying spacetime manifold M. In terms of the
covariant derivative ∇a , and for a mixed tensor T c d , it is introduced via
(∇a ∇b − ∇b ∇a ) T c d = Rabe c T e d − Rabd e T c e . (1)
Y. N. Obukhov
Theoretical Physics Laboratory, Nuclear Safety Institute, Russian Academy of Sciences,
B. Tulskaya 52, 115191 Moscow, Russia
e-mail: obukhov@ibrae.ac.ru
D. Puetzfeld (B)
Center of Applied Space Technology and Microgravity (ZARM), University of Bremen,
28359 Bremen, Germany
e-mail: dirk.puetzfeld@zarm.uni-bremen.de
URL: http://puetzfeld.org

88 Y. N. Obukhov and D. Puetzfeld
Note that our curvature conventions differ from those in [2, 3], see also Tables 1 and 2
in Appendix for a directory of symbols used throughout the article. General Relativity
is formulated on a four-dimensional (pseudo) Riemannian spacetime with the metric
gab of the signature (+1, −1, −1, −1) which is compatible with the connection in the
sense of ∇c gab = 0. Therefore the curvature tensor in Einstein’s theory has twenty
(20) independent components for the most general field configurations produced by
nontrivial matter sources, whereas in vacuum the number of independent components
reduces to ten (10). As compared to Newton’s theory, the gravitational field thus has
more degrees of freedom in the relativistic framework.
The smooth tensor field gab (x c ) introduces the metricity relations on the spacetime
manifold M: an interval (“distance”) between any two close points x ∈ M and x +
d x ∈ M is defined by
ds 2 = gab d x a d x b . (2)
The metric and connection (g, ∇) underlie the formalism of Synge’s world function
[2] which plays a crucial role in the methods of measurement of the gravitational
field in GR and in its natural extensions.
A central question in General Relativity, and consequently in relativistic geodesy,
is how these components of the gravitational field can be determined in an operational
way.
1.1 Method 1: Measuring the Gravitational Field by Means

of Test Bodies
Method 1 utilizes a suitably prepared set of test bodies in order to determine all
components of the curvature of spacetime and thereby the gravitational field. This
method relies on the measurement of the acceleration between the test-bodies and
the observer. Historically, Felix Pirani [1] was the first to point out that one could
determine the full Riemann tensor with the help of a (sufficiently large) number of
test bodies in the vicinity of observer’s world line. Pirani’s suggestion to measure
the curvature was based on the equation which describes the dynamics of a vector
connecting two adjacent geodesics in spacetime. In the literature this equation is
known as a Jacobi equation, or a geodesic deviation equation; its early derivations
in a Riemannian context can be found in [4–6].
A modern derivation and extension of the deviation equation, based on [7], is
presented in the next section. In particular, it is explicitly shown, how a suitably
prepared set of test bodies can be used to determine all components of the curvature
of spacetime (and thereby to measure the gravitational field) with the help of an
exact solution for the components of the Riemann tensor in terms of the mutual
accelerations between the constituents of a cloud of test bodies and the observer.
This can be viewed as an explicit realization of Szekeres’ “gravitational compass”
[8], or Synge’s “curvature detector” [2]. In geodetic terms, such a solution represents
a realization of a relativistic gradiometer or tensor gradiometer, which has a direct
Measuring the Gravitational Field in General Relativity … 89
Fig. 1 Method 1: Sketch of

the operational procedure to
measure the curvature of
spacetime. An observer
moving along a world line Y
monitors the accelerations
(m,n) A to a set of suitably
a
prepared test bodies (hollow
circles). The number of test
bodies required for the
determination of all
curvature components
depends on the type of the
underlying spacetime
operational relevance and forms the basis of relativistic gradiometry. The operational
procedure, see Fig. 1, is to monitor the accelerations of a set of test bodies w.r.t. to
an observer moving along a reference world line Y . A mechanical analogue would
be to measure the forces between the test bodies and the reference body via a spring
connecting them.
Method 1 relies on the standard geodesic deviation equation. A modern covariant
derivation of this equation, as well as its generalization to higher orders will be
provided to make the presentation self-contained. Furthermore, we provide an explicit
exact solution for the curvature components in terms of the mutual accelerations
between the constituents of a cloud of test bodies and the observer. Our presentation
is mainly based on [7].
1.2 Method 2: Measuring the Gravitational Field by Means

of Clocks
Method 2 utilizes a suitably prepared set of clocks to determine all components of

the gravitational field in General Relativity, see Fig. 2. In contrast to the gravitational
compass, the method relies on the frequency comparison between the clocks from
the ensemble and the one carried by the observer. We call such an experimental
setup a clock compass, in analogy to the usual gravitational compass, or in geodetic
language a clock gradiometer.
Fig. 2 Method 2: Sketch of

the operational procedure to
measure the curvature of
spacetime. An observer with
a clock moving along a
world line Y compares his
clock readings C to a set of
suitably prepared clocks in
the vicinity of Y . The
number of clocks required
for the determination of all
curvature components
depends on the type of the
underlying spacetime
We base our review on [9] and pay particular attention to the construction of the
underlying reference frame. As in the case of the gravitational compass, our results
are of direct operational relevance for the setup of networks of clocks, for example
in the context of relativistic geodesy.
2 Theoretical Foundations
In this section we present the theoretical foundations for method 1 and method 2.
For method 1 we start by comparing two general curves in an arbitrary spacetime
manifold and work out an equation for the generalized deviation vector between
those two curves in Sect. 2.1.
For method 2 we first show in Sect. 2.2 how the metric along an arbitrary world
line can be expressed in terms of geometrical and kinematical parameters. This result
is then used in Sect. 2.3 to derive the frequency ratio of two clocks moving on two
general curves, again within an arbitrary spacetime manifold.
2.1 Method 1: Comparison of Two General Curves
Consider two curves Y (t) and X (t˜) with general parameters t and t˜, i.e. are not
necessarily the proper time on the given curves. Now we connect two points x ∈ X
and y ∈ Y on the two curves by the geodesic joining the two points (we assume that
this geodesic is unique).
For the geodesic connecting the two general curves Y (t) and X (t˜) we have the
world function introduced as an integral
y 2

σ (x, y) := dτ (3)
2
x
the geodesic curve γ connecting the spacetime points x and y. Here dτ =

over
gab u a u b dλ is the differential of the proper time along the geodesic which is defined
as the curve γ = {x a (λ)}, such that the tangent vector u a = d x a /dλ is parallely trans-
ported Du a /dλ = 0, and = ±1 for timelike/spacelike curves. Along with the world
function σ (x, y), another important bitensor is the parallel propagator g y x (x, y) that
allows for the parallel transportation of objects along the unique geodesic that links
the points x and y. For example, given a vector V x at x, the corresponding vec-
tor at y is obtained by means of the parallel transport along the geodesic curve
as V y = g y x (x, y)V x . For more details see, e.g., [2, 10] or Sect. 5 in [3]. A com-
pact summary of useful formulas in the context of the bitensor formalism can also be
found in the appendices A and B of [11]. Note that we will use the condensed notation
when the spacetime point to which an index of a bitensor belongs can be directly read
from an index itself. Indices attached to the world-function always denote covariant
derivatives, at the given point, i.e. σ y := ∇ y σ , hence we do not make explicit use of
the semicolon in case of the world-function.
Conceptually, the closest object to the connecting vector between the two points
is the covariant derivative of the world function: σ y . Note though that σ y is just
tangent at that point (its length being the the geodesic length between y and x), only
in flat spacetime it coincides with the connecting vector. Keeping in mind such an
interpretation, let us now work out a propagation equation for this “generalized”
connecting vector along the reference curve, cf. Fig. 3. Following our conventions
the reference curve will be Y (t) and we define the generalized connecting vector to
be:
η y := −σ y . (4)
Taking its covariant total derivative, we have
D y1 D
η = − σ y1 Y (t), X (t˜)
dt dt
∂Y y2 ∂ X x2 d t˜
= −σ y1 y2 − σ y1 x2
∂t ∂ t˜ dt
d t˜
= −σ y1 y2 u y2 − σ y1 x2 ũ x2 , (5)
dt
Fig. 3 Sketch of the two

arbitrarily parametrized
world lines Y (t) and X (t˜),
and the geodesic connecting
two points on these world
line. The (generalized)
deviation vector along the
reference world line Y is
denoted by η y
where in the last line we defined the velocities along the two curves Y and X . As
usual, σ y x1 ...y2 ... := ∇x1 . . . ∇ y2 . . . (σ y ) denote the higher order covariant derivatives
of the world function. We continue by taking the second derivative of (5), which
yields
D 2 y1 d t˜
2
η = −σ y1 y2 y3 u y2 u y3 − 2σ y1 y2 x3 u y2 ũ x3
dt dt
2
d t˜
−σ y1 y2 a y2 − σ y1 x2 x3 ũ x2 ũ x3
dt
2
d t˜ d 2 t˜
−σ y1 x2 ã x2 − σ y1 x2 ũ x2 2 , (6)
dt dt
here we introduced the accelerations a y := Du y /dt, and ã x := D ũ x /d t˜. Equation

(6) is already the generalized deviation equation, but the goal is to have all the
quantities therein defined along the reference wordline Y .
We now derive some auxiliary formulas, by introducing the inverse of the second
derivative of the world function via the following equations:
−1y −1x
σ 1
x σ y2
x
= δ y1 y2 , σ 1
y σ x2
y
= δ x1 x2 . (7)
−1
Multiplication of (5) by σ x3 y1 then yields
d t˜ −1 −1 Dσ y1
ũ x3 = − σ x3 y1 σ y1 y2 u y2 + σ x3 y1
dt dt
y1
x3 Dσ
= K y2 u − H y1
x3 y2
. (8)
dt
In the last line we defined two auxiliary quantities K x y and H x y – the notation
follows the terminology of Dixon. Equation (8) allows us to formally express the the
velocity along the curve X in terms of the quantities which are defined at Y and then
“propagated” by K x y and H x y . Using (8) in (6) we arrive at:
D 2 y1
η = −σ y1 y2 y3 u y2 u y3 − σ y1 y2 a y2
dt 2
Dσ y4
−2σ y1 y2 x3 u y2 K x3 y4 u y4 − H x3 y4
dt
y4
Dσ
−σ y1 x2 x3 K x2 y4 u y4 − H x2 y4
dt
y5
Dσ
× K x3 y5 u y5 − H x3 y5
dt

D Dσ y3
−σ y1 x2 K x2 y3 u y3 − H x2 y3 . (9)
dt dt
We may derive an alternative version of this equation – by using (8) multiplied by

dt/d t˜ – which yields
dt Dσ y1 dt
ũ x3 = K x3 y2 u y2 − H x3 y1 , (10)
d t˜ dt d t˜
and inserted into (6):
2
D 2 y1 d t˜
η = −σ y1 y2 y3 u y2 u y3 − σ y1 y2 a y2 − σ y1 x2 ã x2 (11)
dt 2 dt

Dσ y4
−2σ y2 x3 u
y1 y2
K y4 u − H y4
x3 y4 x3
dt
y4
Dσ
−σ y1 x2 x3 K x2 y4 u y4 − H x2 y4
dt
y5
Dσ
× K x3 y5 u y5 − H x3 y5
dt
2˜ y3
y1 dt d t x2 Dσ
−σ x2 K y3 u − H y3
x2 y3
. (12)
d t˜ dt 2 dt
Note that we may determine the factor d t˜/dt by requiring that the velocity along the
curve X is normalized, i.e. ũ x ũ x = 1, in which case (8) yields
d t˜ Dσ y2
= ũ x1 K x1 y2 u y2 − ũ x1 H x1 y2 . (13)
dt dt
Expansion of quantities on the reference world line The generalized (exact) devi-
ation Eqs. (9) and (12) contain quantities which are not defined along the reference
curve, in particular the covariant derivatives of the world function. We now make use
of the covariant expansions of these quantities, which read (for details, see [12]):
∞
1 y0
σ y0 x1 = g y x1 − δ y0 y + α y y2...yk+1 σ y2 · · · σ yk+1 , (14)
k=2
k!
∞
1 y0
σ y0 y1 = δ y0 y1 − β y1 y2 ...yk+1 σ y2 · · ·σ yk+1 , (15)
k=2
k!

1 y0
g y0 x1 ;x2 = g y x1 g y x2 R y y y3 σ y3
2
∞
1 y0
+ γ y y y3 ...yk+2 σ y3 · · ·σ yk+2 , (16)
k=2
k!

1 y0
g y0
x1 ;y2 = g y x1 R y y2 y3 σ y3
2
∞
1 y0
+ γ y y2 y3 ...yk+2 σ · · ·σ

y3 yk+2
. (17)
k=2
k!
The coefficients α, β, γ in these expansions are polynomials constructed from the

Riemann curvature tensor and its covariant derivatives. The first coefficients read (as
one can also check using computer algebra [13]):
1
α y0 y1 y2 y3 = − R y0 (y2 y3 )y1 , (18)
3
2 y0
β y1 y2 y3 = R (y2 y3 )y1 ,
y0
(19)
3
1
α y1 y2 y3 y4 = ∇(y2 R y0 y3 y4 )y1 ,
y0
(20)
2
1
β y1 y2 y3 y4 = − ∇(y2 R y0 y3 y4 )y1 ,
y0
(21)
2
7 3
α y0 y1 y2 y3 y4 y5 = − R y0 (y2 y3 |y | R y y4 y5 )y1 − ∇(y5 ∇ y4 R y0 y2 y3 )y1 , (22)
15 5
8 2
β y0 y1 y2 y3 y4 y5 = R y0 (y2 y3 |y | R y y4 y5 )y1 + ∇(y5 ∇ y4 R y0 y2 y3 )y1 , (23)
15 5
1
γ y0 y1 y2 y3 y4 = ∇(y3 R y0 |y1 |y4 )y2 , (24)
3
1 1
γ y1 y2 y3 y4 y5 = R y0 y1 y (y3 R y y4 y5 )y2 + ∇(y5 ∇ y4 R y0 |y1 y2 |y3 ) .
y0
(25)
4 4
These results allow us to derive the third derivatives of the world function appearing
in (9) and (12), i.e. we have up to the second order in the deviation vector:

2 y0 1 1
σ y0
y1 y2 = − R (y2 y3 )y1 σ − y3
∇ y R y0 (y3 y4 )y1
3 2 2 2

1
− ∇ y3 R (y2 y4 )y1 σ y3 σ y4
y0
3
1
− λ y0 y1 y2 y3 y4 y5 σ y3 σ y4 σ y5 + O(σ 4 ), (26)
6
2 y0 1
σ y0 y1 x2 = g y x2 R (y y3 )y1 σ y3 − ∇(y R y0 y3 y4 )y1 σ y3 σ y4
3 4

1 y0
+ μ y1 y y3 y4 y5 σ σ σ y3 y4 y5
+ O(σ 4 ), (27)
6

y y 1 y0 1 y0
σ y0 x1 x2 = −g x1 g x2 R y y y3 − R (y y3 )y σ y3

2 3

1 1
+ ∇(y R |y |y4 )y + ∇(y R y3 y4 )y σ y3 σ y4
y0

y0

6 3 4

1 y0
+ ν y y y3 y4 y5 σ σ σ

y3 y4 y5
+ O(σ 4 ). (28)
6
Here we introduced a compact notation for the combinations of the second covariant
derivatives of the curvature and the quadratic polynomial of the curvature tensor (in
symbolic form, “∇∇ R + R · R”):

λ y0 y1 y2 y3 y4 y5 = β y0 y1 y3 y4 y5 ;y2 + β y0 y1 y2 y3 y4 y5 − 3β y0 y1 y (y3 β y |y2 |y4 y5 ) , (29)
y
μ y0
y1 y2 y3 y4 y5 =β y0
y1 y2 y3 y4 y5 − 3β y0
y1 y (y3 α |y2 |y4 y5 ) , (30)
y
ν y0
y1 y2 y3 y4 y5 =γ y1 y2 y3 y4 y5 + α y1 y2 y3 y4 y5 − 3α
y0 y0 y0
y1 y (y3 α |y2 |y4 y5 )
1 y
− R y1 y2 (y3 α y0 |y |y4 y5 ) . (31)
4
Substituting the coefficients of the expansions (14)–(16) we obtain the explicit (com-
plicated) expressions which we do not display here.
For the symmetrized versions of (26) and (28) we obtain
1 y0
σ y0 (y1 y2 ) = R (y y )y σ y3
3 1 2 3
1 1
− ∇(y1 R y0 |y3 y4 |y2 ) + ∇ y3 R y0 (y1 y2 )y4 σ y3 σ y4
4 3
1 y0
− λ (y1 y2 )y3 y4 y5 σ y3 σ y4 σ y5 + O(σ 4 ), (32)
6
y 2
σ y0 (x1 x2 ) = g (x1 g y x2 ) − R y0 (y y )y3 σ y3
3

1 1
+ ∇ y3 R y0 (y y )y4 − ∇(y R y0 |y3 y4 |y ) σ y3 σ y4
4 3

1
− ν y0 (y y )y3 y4 y5 σ y3 σ y4 σ y5 + O(σ 4 ), (33)
6
−1
Furthermore we need the expansions of K x y and σ x y = − H x y :

−1x 1 y
σ 1
y2 = −g x1
y δy y2 −
R (y3 y4 )y2 σ y3 σ y4
6

1 y
+ ∇(y3 R y4 y5 )y2 σ σ σ
y3 y4 y5
+ O(σ 4 ), (34)
12

1
K x1 y2 = g y δ y y2 − R y (y3 y4 )y2 σ y3 σ y4
x1

2

1 y
+ ∇(y3 R y4 y5 )y2 σ σ σ
y3 y4 y5
+ O(σ 4 ). (35)
6
From this one can derive the recurring term in (12) up to the needed order, i.e.

Dσ y2 Dσ y
K x1 y2 u y2 − H x1 y2 = g x1 y u y −
dt dt
y2
1 1 Dσ
− R y (y3 y4 )y2 σ y3 σ y4 u y2 −
2 3 dt

1 y
+ ∇(y3 R y4 y5 )y2 u σ σ σ
y2 y3 y4 y5
+ O(σ 4 ). (36)
6
With these expansions at hand we are finally able to develop the deviation Eq. (12)
up to the third order.
Denote ã y1 = g y1 x2 ã x2 in accordance with the definition of the parallel propagator,
and introduce
φ y1 y2 y3 y4 y5 y6 = λ y1 y2 y3 y4 y5 y6 − 2μ y1 y2 y3 y4 y5 y6 + ν y1 y2 y3 y4 y5 y6 . (37)
The deviation equation up to the third order reads

2
D 2 y1 d t˜ dt d 2 t˜ y1 Dη y1 dt d 2 t˜
η = ã y1 − a y1 + u +
dt 2 dt d t˜ dt 2 dt d t˜ dt 2

Dη y2
−η y4 R y1 y2 y3 y4 u y2 u y3 + 2u y3
dt

1 1
+η η u u
y4 y5 y2 y3
∇ y R y4 y5 y3 − ∇ y4 R y2 y3 y5
y1 y1
2 2 3
2
2˜
1 y1 1 y2 d t˜ y2 dt d t
+ R y4 y5 y2 a + ã y2
−u
3 2 dt d t˜ dt 2

1
− η y4 η y5 η y6 φ y1 y2 y3 y4 y5 y6 u y2 u y3
6
2
2˜
1 d t˜ y2 dt d t
− ∇(y4 R y5 y6 )y2 a + ã
y1 y2 y2
−u
2 dt d t˜ dt 2

1 Dη y y2 y3
− uy η η − ∇(y R y1 y2 y3 )y1 + ∇ y2 R y1 (y y )y3
2 dt

1 2 Dη y2 Dη y3 y4 y1
− ∇(y R y1 |y2 y3 |y ) − η R y2 y3 y4 + O(σ 4 ). (38)
3 3 dt dt
We would like to stress that the generalized deviation Eq. (38) is completely general.
In particular, it allows for a comparison of two general, i.e. not necessarily geodetic,
world lines in spacetime. Various special cases of (38) qualitatively reproduce all the
previous results in the literature, see in particular [14–20].
2.2 Method 2: Reference Frame (Inertial and Gravitational

Effects)
The above discussion of the deviation equation made clear that a suitable choice
of coordinates is crucial for the successful determination of the gravitational field.
In particular, the operational realization of the coordinates is of importance when it
comes to actual measurements.
From an experimentalists perspective so-called (generalized) Fermi coordinates
appear to be realizable operationally. There have been several suggestions for such
coordinates in the literature in different contexts [2, 21–46]. In the following we
are going to derive the line element in the vicinity of a world line, representing an
observer in an arbitrary state of motion, in generalized Fermi coordinates.
Fermi normal coordinates Following [24] we start by taking successive derivatives
of the usual geodesic equation. This generates a set of equations of the form (for
n ≥ 2)
dn xa d x b1 d x bn
n
= −b1 ...bn a ··· , (39)
ds ds ds
where the objects with n ≥ 3 lower indices are defined by the recurrent relation
b1 ...bn a := ∂(b1 b2 ...bn ) a − (n − 1) c(b1 ...bn−2 a bn−1 bn ) c (40)

from the components of the linear connection bc a . A solution x a = x a (s) of the
geodesic equation may then be expressed as a series

d x a

a s 2 d 2 x a s 3 d 3 x a
x = x 0+s
a
+ + + ···
ds 0 2 ds 2 0 6 ds 3 0
s2 0 a b c s3 0 a b c d
= q a + sv a − bc v v − bcd v v v − · · · , (41)
2 6
a 0
where in the last line we used q a := x a |0 , v a := ddsx 0 , and ... a := ... a |0 for con-
stant quantities at the point around which the series development is performed.
Now let us setup coordinates centered on the reference curve Y to describe an
adjacent point X . For this we consider a unique geodesic connecting Y and X . We
define our coordinates in the vicinity of a point on Y (s), with proper time s, by using
a tetrad λb (α) which is Fermi transported along Y , i.e.
X 0 = s, X α = τ ξ b λb (α) . (42)
Here α = 1, . . . , 3, and τ is the proper time along the (spacelike) geodesic connecting
Y (s) and X . The ξ b are constants, and it is important to notice that the tetrads are
functions of the proper time s along the reference curve Y , but independent of τ . See
Fig. 4 for further explanations. By means of this linear ansatz (42) for the coordinates
in the vicinity of Y , we obtain for the derivatives w.r.t. τ along the connecting geodesic
(n ≥ 1):
dn X 0 d Xα d n+1 X α
= 0, = ξ b λb (α) , = 0. (43)
dτ n dτ dτ n+1
In other words, in the chosen coordinates (42), along the geodesic connecting Y and
X , one obtains for the derivatives (n ≥ 2)
d X b1 d X bn
b1 ...bn a ··· = 0. (44)
dτ dτ
This immediately yields
β1 ...βn a = 0, (45)
along the connecting curve, in the region covered by the linear coordinates as defined
above.
The Fermi normal coordinate system cannot cover the whole spacetime manifold.
By construction, it is a good way to describe the physical phenomena in a small region
around the world line of an observer. The smallness of the corresponding domain
depends on the motion of the latter, in particular, on the magnitudes of acceleration
|a| and angular velocity |ω| of the observer which set the two characteristic lengths:
Fig. 4 Construction of the coordinate system around the reference curve Y . Coordinates of a point
X in the vicinity of Y (s) – s representing the proper time along Y – are constructed by means
of a tetrad λb (α) . Here τ is the proper time along the (spacelike) geodesic connecting Y and X .
By choosing a linear ansatz for the coordinates the derivatives of the connection vanish along the
geodesic connecting Y and X
tr = c2 /|a| and rot = c/|ω|. The Fermi coordinate system X α provides a good
description for the region |X |/ 1. For example, this condition is with a high accu-
racy valid in terrestrial laboratories since tr = c2 /|g⊕ | ≈ 1016 m (one light year), and
rot = c/|⊕ | ≈ 4 × 1012 m (27 astronomical units). Note, however, that for a par-
ticle accelerated in a storage ring ≈ 10−6 m. Furthermore, the region of validity of
the Fermi coordinate system is restricted by the strength of the gravitational field in
the region close to the reference curve, grav = min{|Rabcd |−1/2 , |Rabcd |/|Rabcd,e |},
so that the curvature should have not yet caused geodesics to cross. We always assume
that there is a unique geodesic connecting Y and X .
Explicit form of the connection At the lowest order, in flat spacetime, the connection
of a noninertial system that is accelerating with a α and rotating with angular velocity
ωα at the origin of the coordinate system is
00 0 = αβ c = 0, 00 α = a α , 0α 0 = aα , 0β α = −εα βγ ωγ . (46)

Hereafter εαβγ is the 3-dimensional totally antisymmetric Levi-Civita symbol, and

the Euclidean 3-dimensional metric δαβ is used to raise and lower the spatial (Greek)
indices, in particular aα = δαβ a β and εα βγ = δ αδ εδβγ . For the time derivatives we
have
∂0 00 0 = ∂0 αβ c = 0, ∂0 00 α = ∂0 a α =: bα ,

∂0 0α 0 = bα , ∂0 0β α = −εα βγ ∂0 ωγ =: −εα βγ ηγ . (47)
From the definition of the curvature we can express the next order of derivatives of
the connection in terms of the curvature:
∂α 00 0 = bα − aβ εβ αγ ωγ ,
∂α 00 β = − R0α0 β − εβ αγ ηγ + aα a β − δαβ ωγ ωγ + ωα ωβ ,
∂α 0β 0 = − R0αβ 0 − aα aβ ,
∂α 0β γ = − R0αβ γ + εγ αδ ωδ aβ . (48)
Using (45), we derive the spatial derivatives
2
∂α βγ d = Rα(βγ ) d , (49)
3
see also the general solution given in the Appendix B of [7].
Explicit form of the metric In order to determine, in the vicinity of the reference
curve Y , the form of the metric at the point X in coordinates y a centered on Y , we
start again with an expansion of the metric around the reference curve
1
gab | X = gab |Y + gab,c Y y c + gab,cd Y y c y d + · · · . (50)
2
Of course in normal coordinates we have gab |Y = ηab , whereas the derivatives of
the metric have to be calculated, and the result actually depends on which type of
coordinates we want to use. The derivatives of the metric may be expressed just by
successive differentiation of the metricity condition ∇c gab = 0:
gab,c = 2 gd(a b)c d ,

gab,cd = 2 ∂d ge(a b)c e + ∂d c(a e gb)e ,
..
. (51)
In other words, we can iteratively determine the metric by plugging in the explicit
form of the connection and its derivatives from above.
In combination with (46) one finds:
g00,0 = g0α,0 = gαβ,0 = gαβ,γ = 0,

g00,α = 2aα , g0α,β = εαβγ ωγ . (52)
For the second order derivatives of metric we obtain, again using (51) in combi-
nation with (47), (48), and (52):
g00,00 = g0α,00 = gαβ,00 = gαβ,γ 0 = 0, g00,α0 = 2bα ,

g0α,β0 = −εγ βδ ηδ gαγ = εαβγ ηγ ,
g00,αβ = − 2R0βα 0 + 2aα aβ − 2δαβ ωγ ωγ + 2ωα ωβ ,
4 2
g0α,βγ = − Rα(βγ ) 0 , gαβ,γ δ = Rγ (αβ)δ . (53)
3 3
Note that R0βα 0 + Rα0β 0 + Rβα0 0 ≡ 0, in view of the Ricci identity. Since Rβα0 0 =
0, we thus find R0βα 0 = R0(βα) 0 .
As a result, we derive the line element in the Fermi coordinates (up to the second
order):

ds 2 X (y 0 , y α ) = (dy 0 )2 1 + 2aα y α + 2bα y α y 0

+(aα aβ − δαβ ωγ ωγ + ωα ωβ − R0αβ0 )y α y β
2
+2dy 0 dy α εαβγ ωγ y β + εαβγ ηγ y β y 0 − Rαβγ 0 y β y γ

3
α β 1 γ δ
−dy dy δαβ − Rγ αβδ y y + O(3). (54)
3
It is worthwhile to notice that we can recast this result into

ds 2 X (y 0 , y α ) = (dy 0 )2 1 + 2a α y α + (a α a β − δαβ ωγ ωγ + ωα ωβ
2
−R0αβ0 )y α y β + 2dy 0 dy α εαβγ ωγ y β − Rαβγ 0 y β y γ

3
α β 1 γ δ
−dy dy δαβ − Rγ αβδ y y + O(3), (55)
3
by introducing a α = aα + y 0 ∂0 aα = aα + y 0 bα and ωα = ωα + y 0 ∂0 ωα = ωα + y 0
ηα which represent the power expansion of the time dependent acceleration and
angular velocity.
2.3 Method 2: Apparent Behavior of Clocks
The results from the last section may now be used to describe the behavior of clocks
in the vicinity of the reference world line, around which the coordinates were con-
structed.
There is one interesting peculiarity about writing the metric like in (54), i.e. one
obtains clock effects which depend on the acceleration of the clock (just integrate
along a curve in those coordinates and the terms with a and ω will of course contribute
to the proper time along the curve). This behavior of clocks is of course due to the
choice of the noninertial observer, and they are only present along curves which
do not coincide with the observers world line. Recall that, by construction, one has
Minkowski’s metric along the world line of the observer, which is also the center
of the coordinate system in which (54) is written – all inertial effects vanish at the
origin of the coordinate system.
Flat case We start with the flat spacetime and switch to a quantity which is directly
measurable, i.e. the proper time quotient of two clocks located at Y and X . It is
worthwhile to note that for a flat spacetime, Ri jk l = 0, the interval (54) reduces to
the Hehl-Ni [47] line element of a noninertial (rotating and accelerating) system:

ds 2 X (y 0 , y α ) = (1 + a α y α )2 (dy 0 )2 − δαβ (dy α + εα μν ωμ y ν dy 0 )
×(dy β + εβ ρσ ωρ y σ dy 0 ) + O(3), (56)
From (54) we derive

2 2
ds| X dy 0
= 1 − δαβ v α v β + 2aα y α + 2bα y α y 0
ds|Y ds|Y

+y α y β aα aβ − δαβ ωγ ωγ + ωα ωβ

+2v α εαβγ y β ωγ + y 0 y β ηγ + O(3) (57)
1
= 1+ α β
2aα y α + 2bα y α y 0
1 − δαβ v v

+y α y β aα aβ − δαβ ωγ ωγ + ωα ωβ

+2v α εαβγ y β ωγ + y 0 y β ηγ + O(3). (58)
Here we introduced the velocity v α = dy α /dy 0 . Defining
V α := v α + εα βγ ωβ y γ , (59)
we can rewrite the above relation more elegantly as

2 2
ds| X dy 0
= (1 + a α y α )2 − δαβ V α V β + O(3). (60)
ds|Y ds|Y
Equation (58) is reminiscent of the situation which we encountered in case of the

gravitational compass, i.e. we may look at this measurable quantity depending on
how we prepare the
2
ds| X
C y α , y 0 , v α , a α , ωα , bα , ηα := . (61)
ds|Y
Curved case Now let us investigate the curved spacetime, after all we are interested
in measuring the gravitational field by means of clock comparison. The frequency
ratio becomes:
2
ds| X 1
= 1+ 2aα y α + 2bα y α y 0
ds|Y 1 − δαβ v α v β

+y α y β aα aβ − R0αβ0 − δαβ ωγ ωγ + ωα ωβ
4
+2v α εαβγ y β ωγ + y 0 y β ηγ − v α y β y γ Rαβγ 0
3

1
+ v α v β y γ y δ Rγ αβδ + O(3). (62)
3
Analogously to the flat case in (61), we introduce a shortcut for the measurable fre-
quency
ratio in a curved background, denoting its dependence on different quantities
as C y α , y 0 , v α , a α , ωα , bα , ηα , Rαβγ δ .
Note that in the flat, as well as in the curved case, the frequency ratio becomes
independent of bα and ηα on the three-dimensional slice with fixed y 0 (since we can
alwayschoose our coordinate time parameter y 0 = 0), i.e. we have C (y α , v α , a α , ωα )
and C y α , v α , a α , ωα , Rαβγ δ respectively.
3 Operational Determination of the Gravitational Field
3.1 Method 1: Relativistic Gradiometry/Gravitational

Compass
The determination of the curvature of spacetime in the context of deviation equations

has been discussed in [2, 8, 48]. In particular, Szekeres coined in [8] the notion of a
“gravitational compass.” From now on we will adopt this notion for a set of suitably
prepared test bodies which allow for the measurement of the curvature and, thereby,
the gravitational field.
The operational procedure is to monitor the accelerations of a set of test bodies
w.r.t. to an observer moving on the reference world line Y . A mechanical analogue
would be to measure the forces between the test bodies and the reference body via a
spring connecting them.
Rewriting the deviation equation We now describe the configurations of test bodies
which allow for a complete determination of all curvature components in a Rieman-
nian background spacetime. For concreteness, our analysis will be based on the stan-
dard geodesic deviation equation, as well as one of its generalizations. Our starting
point is the standard geodesic deviation equation, i.e.
D2 a
η = R a bcd u b ηc u d . (63)
ds 2
Since we want to express the curvature in terms of measured quantities, i.e. the
velocities and the accelerations, we rewrite this equation in terms of the standard
(non-covariant) derivative w.r.t. the proper time.
In order to simplify the resulting equation we employ normal coordinates, i.e. we
have on the world line of the reference test body
2
ab c |Y = 0, ∂a bc d |Y = Ra(bc) d . (64)
3
In terms of the standard total derivative w.r.t. to the proper time s, the deviation
Eq. (63) takes the form:
d2 a |Y 2 a
η = R bcd u b ηc u d . (65)
ds 2 3
However, what actually seems to be measured by a compass at the reference point Y
is the lower components of the relative acceleration. For the lower index position, in
terms of the ordinary derivative in normal coordinates, the deviation Eq. (63) takes
the form
d2 |Y 4
ηa = Rabcd u b ηc u d . (66)
ds 2 3
Explicit compass setup Let us consider a general 6-point compass. In addition to
the reference test body on the world line we will use the following geometrical setup
of the 5 remaining test bodies:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 0 0
⎜ 1 ⎟ ⎜ ⎟ ⎜ ⎟
η =⎜
(1) a ⎟ , (2) ηa = ⎜ ⎟ , (3) ηa = ⎜ 0 ⎟ ,
0
⎝0⎠ ⎝1⎠ ⎝0⎠
0 0 1
⎛ ⎞ ⎛ ⎞
0 0
⎜ 1 ⎟ ⎜ ⎟
η =⎜
(4) a ⎟ , (5) ηa = ⎜ 0 ⎟ . (67)
⎝1⎠ ⎝1⎠
0 1
In addition to the positions of the compass constituents, we have to make a choice for
the velocity of the reference test body/observer. In the following we will use (m) dif-
ferent compasses, each of these compasses will have a different velocity (associated)
with the reference test body. In other words, we consider (m) different compasses
or reference test bodies, all of which are located at the world line reference point Y
(at the same time), and all these (m) observers measure the relative accelerations to
all five test bodies placed at the positions given in (67). The left-hand sides of (66)
are the measured accelerations and in the following we refer to them by (m,n) Aa .
Furthermore, we also introduced the compass index (m) u a for the velocities. In other
words, for (m) compasses and (n) bodies in one compass, we have the following set
of equations:
(m,n) |Y 4
Aa = Rabcd (m) u b (n) ηc (m) u d . (68)
3
What remains to be chosen, apart from the (n = 1 . . . 5) positions of bodies in one
compass, is the number (m) and the actual directions in which each compass/observer
shall move. Of course in the end we want to minimize both numbers, i.e. (m) and
(n), which are needed to determine all curvature components.
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
c10 c20 c30
⎜ 0 ⎟ (2) a ⎜ c21 ⎟ (3) a ⎜ 0 ⎟
u =⎜
(1) a ⎟ ⎜ ⎟ ⎜ ⎟
⎝ 0 ⎠ , u = ⎝ 0 ⎠ , u = ⎝ c32 ⎠ ,
0 0 0
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
c40 c50 c60
⎜ 0 ⎟ (5) a ⎜ c51 ⎟ (6) a ⎜ 0 ⎟
u =⎜
(4) a ⎟ ⎜ ⎟ ⎜ ⎟
⎝ 0 ⎠ , u = ⎝ c52 ⎠ , u = ⎝ c62 ⎠ . (69)
c43 0 c63
The c(m)a here are just constants, chosen appropriately to ensure the normalization
of the 4-velocity of each compass.
In summary, we are going to consider (m) = 1 . . . 6 compasses, each of them with
6-points, where the five reference points are always the (n) = 1 . . . 5 from (67).
Explicit curvature components The 20 independent components of the curvature
tensor can be explicitly determined in terms of the accelerations (m,n) Aa and veloc-
ities (m) u a by making use of the deviation Eq. (68) with the help of the compass
configuration given in (67) and (69). The result reads as follows:
3 (1,1) −2
01 : R1010 = A1 c10 , (70)
4
3 −2
02 : R2010 = (1,1) A2 c10 , (71)
4
3 −2
03 : R3010 = (1,1) A3 c10 , (72)
4
3 (1,2) −2
04 : R2020 = A2 c10 , (73)
4
3 (1,2) −2
05 : R3020 = A3 c10 , (74)
4
3 (1,3) −2
06 : R3030 = A3 c10 , (75)
4
3 (2,1) −1 −1 −1
07 : R2110 = A2 c21 c20 − R2010 c21 c20 , (76)
4
3 (2,1) −1 −1 −1
08 : R3110 = A3 c21 c20 − R3010 c21 c20 , (77)
4
3 (3,1) −2 −1
09 : R0212 = A0 c32 + R2010 c32 c30 , (78)
4
3 (2,2) −2 2 −2 −1
10 : R1212 = A2 c21 − R2020 c20 c21 − 2R0212 c21 c20 , (79)
4
3 (3,2) −1 −1 −1
11 : R3220 = A3 c32 c30 − R3020 c32 c30 , (80)
4
3 (4,1) −2 −1
12 : R0313 = A0 c43 + R3010 c43 c40 , (81)
4
3 (2,3) −2 2 −2 −1
13 : R1313 = A3 c21 − R3030 c20 c21 − 2R0313 c21 c20 , (82)
4
3 (4,2) −2 −1
14 : R0323 = A0 c43 + R3020 c43 c40 , (83)
4
3 (4,2) −2 −2 2 −1
15 : R2323 = A2 c43 − R2020 c43 c40 + 2R3220 c43 c40 , (84)
4
3 (5,3) −1 −1 1 −1 −1 2
16 : R3132 = A3 c52 c51 − R3030 c52 c51 c50
8 2
−1 −1 1 −1
−R0313 c52 c50 − R0323 c51 c50 − R1313 c52 c51
2
1 −1
− R2323 c52 c51 , (85)
2
3 −1 −1 1 −1 −1 2
17 : R1213 = (6,1) A1 c63 c62 − R1010 c63 c62 c60
8 2
−1 −1 1 −1
+R2110 c63 c60 + R3110 c62 c60 − R1212 c63 c62
2
1 −1
− R1313 c63 c62 , (86)
2
There are still 3 components of the curvature tensor missing. To determine them,
we notice that the following relation between the remaining equations is at our
disposal:
3 (2,2) −1 −1 −1 −1
R0312 − R0231 = A3 c20 c21 − R3020 c20 c21 − R3121 c21 c20 , (87)
4
3 −1 −1 −1 −1
R0231 − R0123 = (4,1) A2 c40 c43 − R2010 c40 c43 − R2313 c43 c40 . (88)
4
Subtracting (87) from (88) and using the Ricci identity we find:
1 (4,1) −1 −1 1 −1 −1
18 : R0231 = A2 c40 c43 − (2,2) A3 c20 c21
4 4
1 −1 −1
+ R3020 c20 c21 + R3121 c21 c20
3
−1 −1
−R2010 c40 c43 − R2313 c43 c40 , (89)
1 −1 −1 1 −1 −1
19 : R0312 = (4,1) A2 c40 c42 + (2,2) A3 c20 c21
4 2
1 −1 −1
− 2R3020 c20 c21 + 2R3121 c21 c20
3
−1 −1
+R2010 c40 c43 + R2313 c43 c40 . (90)
Finally, by reinsertion of (87) in one of the remaining compass equations, one obtains:
3 (4,1) −1 −1 −1 3 −1 −1
20 : R3212 = A3 c20 c21 c50 c52 − (5,2) A3 c51 c52
4 4
−1 −1 −1
+R3121 c52 c51 − c50 c21 c20 + R3220 c50 c51
−1
−1 −1

+R3020 c50 c52 c50 c51 − c20 c21 . (91)
By examination of the components given in (70)–(91), we conclude that for a full

determination of the curvature one needs 13 test bodies, see Fig. 5 for a sketch of the
solution.
Vacuum spacetime In vacuum the number of independent components of the cur-
vature is reduced to the 10 components of the Weyl tensor Cabcd . Replacing Rabcd
in the compass solution (70)–(91), and taking into account the symmetries of Weyl
we may use a reduced compass setup to completely determine the gravitational field,
i.e.
3 (1,1) −2
01 : C1010 = A1 c10 , (92)
4
3 (1,1) −2
02 : C2010 = A2 c10 , (93)
4
3 (1,1) −2
03 : C3010 = A3 c10 , (94)
4
3 (1,2) −2
04 : C2020 = A2 c10 , (95)
4
3 (1,2) −2
05 : C3020 = A3 c10 , (96)
4
Fig. 5 Symbolical sketch of

the explicit compass solution
in (70)–(91). In total 13
suitably prepared test bodies
(hollow circles) are needed
to determine all curvature
components. The reference
body is denoted by the black
circle. Note that with the
standard deviation equation
all (1...6) u a , but only (1...3) ηa
are needed in the solution
3 (2,1) −1 −1 −1
06 : C2110 = A2 c21 c20 − C2010 c21 c20 , (97)
4
3 −1 −1 −1
07 : C3110 = (2,1) A3 c21 c20 − C3010 c21 c20 , (98)
4
3 −2 −1
08 : C0212 = (3,1) A0 c32 + C2010 c32 c30 , (99)
4
1 (4,1) −1 −1 1 −1 −1
09 : C0231 = A2 c40 c43 − (2,2) A3 c20 c21
4 4
1 −1 −1

+ C3020 c20 c21 + c21 c20
3
1 −1 −1

− C2010 c40 c43 + c43 c40 , (100)
3
1 −1 −1 1 −1 −1
10 : C0312 = (4,1) A2 c40 c42 + (2,2) A3 c20 c21
4 2
2 −1 −1

− C3020 c20 c21 + c21 c20
3
1 −1 −1

+ C2010 c40 c43 + c43 c40 . (101)
3
All the other components of the Weyl tensor are obtained from the above by making
use of the double-self-duality property Cabcd = − 41 abe f cdgh C e f gh , where abcd is
the totally antisymmetric Levi-Civita tensor with 0123 = 1, and the Ricci identity.
See Fig. 6 for a sketch of the solution.
Fig. 6 Symbolical sketch of

the explicit compass solution
in (92)–(101) for the vacuum
case. In total 6 suitably
prepared test bodies (hollow
circles) are needed to
determine all components of
the Weyl tensor. The
reference body is denoted by
the black circle. Note that in
vacuum, with the standard
deviation equation, all
(1...4) u a , but only (1...2) ηa are
needed in the solution
3.2 Method 2: Relativistic Clock Gradiometry/Gravitational

Clock Compass
Now we turn to the determination of the curvature in a general spacetime by means

of clocks. We consider the non-vacuum case first, when one needs to measure 20
independent components of the Riemann curvature tensor Rabc d .
Again we start by rearranging the system (61):

(n) α (n) β 4 1
y y −R0αβ0 − Rγ αβ0 (m) v γ + Rαγ δβ (m) v γ (m) v δ
3 3
= B((n) y α , (m) v α , ( p) a α , (q) ωα ), (102)
where

B(y α , v α , a α , ωα ) := (1 − v 2 ) (C − 1) − 2aα y α − y α y β aα aβ

−δαβ ωγ ω + ωα ωβ − 2v α εαβγ y β ωγ .
γ
(103)
Analogously to our analysis of the gravitational compass [7], we may now consider
different setups of clocks to measure as many curvature components as possible. The
system in (102) yields (please note that only the position and the velocity indices are
indicated):
01 : R1010 = (1,1) B, (104)

3 −1 −1
02 : R2110 = c22 c42 (c22 − c42 )−1 (1,1) Bc22
2
− (1,1) Bc42
2
4

(1,2) (1,4)
+ Bc42 −
2
Bc22 ,
2
(105)

−1 −1
03 : R1212 = −3c22 c42 (c22 − c42 )−1 (1,1)
Bc22 − (1,1) Bc42

(1,2) (1,4)
+ Bc42 − Bc22 , (106)

3 −1 −1
04 : R3110 = c33 c63 (c33 − c63 )−1 (1,1) Bc33
2
− (1,1) Bc63
2
4

(1,3) (1,6)
+ Bc63 −
2
Bc33 ,
2
(107)

−1 −1
05 : R1313 = −3c33 c63 (c33 − c63 )−1 (1,1)
Bc33 − (1,1) Bc63

+(1,3) Bc63 − (1,6) Bc33 , (108)

3 −1 −1 (1,5) 4
06 : R1213 = c52 c53 − B + R1010 −
R2110 c52
2 3

4 1 1
− R3110 c53 − R1212 c52
2
− R1313 c53
2
, (109)
3 3 3
07 : R2020 = (2,2) B, (110)

3 −1 (2,1) 1
08 : R0212 = c11 B − R2020 + R1212 c11 2
, (111)
4 3

3 −1 −1 −1 (2,2)
09 : R3220 = c33 c53 (c33 − c53 ) 2
Bc33 − (2,2) Bc53
2
4

+(2,3) Bc53
2
− (2,5) Bc33
2
, (112)

−1 −1 −1 (2,2)
10 : R2323 = −3c33 c53 (c33 − c53 ) Bc33 − (2,2) Bc53

+(2,3) Bc53 − (5,2) Bc33 , (113)

3 −1 −1 (2,6) 4 4
11 : R3212 = c61 c63 − B + R2020 + R0212 c61 − R3220 c63
2 3 3

1 1
− R1212 c61
2
− R2323 c63
2
, (114)
3 3
12 : R3030 = (3,3) B, (115)

3 −1 (3,1) 1
13 : R0313 = c11 B − R3030 + R1313 c112
, (116)
4 3

3 −1 (3,2) 1
14 : R0323 = c22 B − R3030 + R2323 c22 ,
2
(117)
4 3

3 −1 −1 4 4
15 : R3132 = c41 c42 − (3,4) B + R3030 + R0313 c41 + R0323 c42
2 3 3

1 1
− R1313 c412
− R2323 c42
2
, (118)
3 3

1 (4,1) 4
16 : R2010 = B − R1010 − R2020 − R0212 c11
2 3

4 1
− R2110 c11 + R1212 c112
, (119)
3 3

1 (5,2) 4
17 : R3020 = B − R2020 − R3030 − R0323 c22
2 3

4 1
− R3220 c22 + R2323 c222
, (120)
3 3

1 (6,1) 4
18 : R3010 = B − R1010 − R3030 − R0313 c11
2 3

4 1
− R3110 c11 + R1313 c112
. (121)
3 3
Introducing abbreviations

3 −1 (4,3)
K 1 := c33 − B + R1010 + 2R2010 + R2020
4

4 1
− (R3110 + R3220 )c33 − (R1313 + 2R3132 + R2323 )c33 ,
2
(122)
3 3

3 −1
K 2 := c11 − (5,1) B + R2020 + 2R3020 + R3030
4

4 1
+ (R0212 + R0313 )c11 − (R1212 + 2R1213 + R1313 )c11 ,
2
(123)
3 3

3 −1
K 3 := c22 − (6,2) B + R1010 + 2R3010 + R3030
4

4 1
− (R2110 + R0323 )c22 − (R1212 + 2R3212 + R2323 )c22 ,
2
(124)
3 3
we find the remaining three curvature components
1
19 : R1023 = (K 3 − K 1 ) , (125)
3
1
20 : R2013 = (K 2 − K 1 ) , (126)
3
1
21 : R3021 = (K 3 − K 2 ) . (127)
3
See Fig. 7 for a symbolical sketch of the solution. The B’s in these equations can be
explicitly resolved in terms of the C’s
(1,1)
(1,1)
B = 1 − c11
2
C −1 , (128)
(1,2)
(1,2)
B = 1 − c22
2
C −1 , (129)
(1,3)
(1,3)
B = 1 − c33
2
C −1 , (130)
(1,4)
(1,4)
B = 1 − c41
2
− c42
2
C −1 , (131)
(1,5)
(1,5)
B = 1 − c52
2
− c53
2
C −1 , (132)
(1,6)
(1,6)
B = 1 − c61
2
− c63
2
C −1 , (133)
(2,1)
(2,1)
B = 1 − c11
2
C −1 , (134)
(2,2)
(2,2)
B = 1 − c22
2
C −1 , (135)
(2,3)
(2,3)
B = 1 − c33
2
C −1 , (136)
(2,5)
(2,5)
B = 1 − c52
2
− c53
2
C −1 , (137)
(2,6)
(2,6)
B = 1 − c61
2
− c63
2
C −1 , (138)
(3,1)
(3,1)
B = 1 − c11
2
C −1 , (139)
(3,2)
(3,2)
B = 1 − c22
2
C −1 , (140)
(3,3)
(3,3)
B = 1 − c33
2
C −1 , (141)
(3,4)
(3,4)
B = 1 − c41
2
− c42
2
C −1 , (142)
(4,1)
(4,1)
B = 1 − c11
2
C −1 , (143)
(4,3)
(4,1)
B = 1 − c33
2
C −1 , (144)
(5,1)
(5,1)
B = 1 − c11
2
C −1 , (145)
(5,2)
(5,2)
B = 1 − c22
2
C −1 , (146)
(6,1)
(6,1)
B = 1 − c11
2
C −1 , (147)
(6,2)
(6,2)
B = 1 − c22
2
C −1 . (148)
Vacuum spacetime In vacuum the number of independent components of the curva-

ture is reduced to the 10 components of the Weyl tensor Cabcd . Replacing Rabcd in the
compass solution (104)–(121), and taking into account the symmetries of the Weyl
Fig. 7 Symbolical sketch of the explicit solution for the curvature (104)–(127). In total 21 suitably
prepared clocks (hollow circles) are needed to determine all curvature components. The observer
is denoted by the black circle. Note that all (1...6) v a , but only (1...3) y a are needed in the solution
tensor, we may use a reduced clock setup to completely determine the gravitational
field. All other components may be obtained from the double self-duality property
Cabcd = − 41 εabe f εcdgh C e f gh .
01 : C2323 = −(1,1) B, (149)

3 −1 −1 −1 (1,1)
02 : C0323 = c22 c42 (c22 − c42 ) 2
Bc22 − (1,1) Bc42
2
4

+(1,2) Bc42
2
− (1,4) Bc22
2
, (150)

−1 −1
03 : C3030 = 3c22 c42 (c22 − c42 )−1 (1,1)
Bc22 − (1,1) Bc42

(1,2) (1,4)
+ Bc42 − Bc22 , (151)
(2,2)
04 : C2020 = B, (152)

3 −1 (1,3) 1
05 : C3220 = c33 B + C2323 − C2020 c33 ,
2
(153)
4 3

3 −1 (2,1) 1
06 : C0313 = − c11 B − C2020 − C3030 c11 ,
2
(154)
4 3

3 −1 −1 (1,5) 4
07 : C3020 = − c52 c53 B + C2323 + C0323 c52
2 3

4 1 1
− C3220 c53 − C3030 c52 − C2020 c53 ,
2 2
(155)
3 3 3

3 −1 −1 (2,6) 4 4
08 : C3212 = − c61 c63 B − C2020 + C0313 c61 + C3220 c63
2 3 3

1 1
− C3030 c61
2
+ C2323 c63
2
, (156)
3 3

3 −1 −1 (3,4) 4 4
09 : C3132 = − c41 c42 B − C3030 − C0313 c41 − C0323 c42
2 3 3

1 1
− C2020 c41
2
+ C2323 c42
2
. (157)
3 3
With the abbreviations

3 −1
K 1 := c33 − (4,3) B − C2323 + 2C3132 + C2020
4

1
+ (C2020 − 2C3132 − C2323 )c33 ,
2
(158)
3

3 −1
K 2 := c11 − (5,1) B + C2020 + 2C3020 + C3030
4

1
+ (C3030 − 2C3020 + C2020 )c11
2
, (159)
3

3 −1 (6,2)
K 3 := − c22 B + C2323 − 2C3212 − C3030
4

1
− (C3030 − 2C3212 − C2323 )c22
2
, (160)
3
the remaining three curvature components read
1
10 : C1023 = (K 3 − K 1 ) , (161)
3
1
11 : C2013 = (K 2 − K 1 ) , (162)
3
1
12 : C3021 = (K 3 − K 2 ) . (163)
3
A symbolical sketch of the solution is given in Fig. 8.
Fig. 8 Symbolical sketch of the explicit vacuum solution for the curvature (149)–(163). In total 11
suitably prepared clocks (hollow circles) are needed to determine all curvature components. The
observer is denoted by the black circle. Note that all (1...6) v a , but only (1...3) y a are needed in the
solution
4 Summary
4.1 Method 1: Summary
In the framework of Synge’s world function approach, we have derived a generalized

covariant deviation Eq. (12) which is valid for arbitrary world lines and in general
background spacetimes. Making use of systematic expansions of the exact devia-
tion equation up to the third order in the world function, we obtain the final result
(38) which can be viewed as a generalization of the well-known geodesic deviation
equation. Furthermore, our results encompass several suggestions for a generalized
deviation equation from the literature as special cases, and therefore may serve a
unified framework for further studies.
In Sect. 3.1 we have shown how deviation equations can be used to determine the
curvature of spacetime. For this we extended the notion of a gravitational compass
[8] and worked out compass setups for general as well as for vacuum spacetimes.
One setup is based on the standard geodesic deviation equation, and another is based
on the next order generalization given in which goes beyond the linearized case.
For both cases we provided the explicit compass solution which allows for a full
determination of the curvature.
In contrast to the general considerations in [2, 8] we give an explicit exact solution

for the compass setup. With the standard deviation equation, as well as with the
generalized deviation equation, we need at least 13 test bodies to determine all
curvature components in a general spacetime. For the standard deviation we therefore
obtain the same number of bodies as in [48], however it is worthwhile to note that
no explicit solution was given in [48] for a non-vacuum spacetime. In the case of
a generalized deviation equation our findings are at odds with the results in [48].
However, this discrepancy in the generalized case comes as no surprise since the
generalized equation used in [48] – which was previously derived in [49] – differs
from our equation. In vacuum spacetimes, we have explicitly shown that the number
of required test bodies is reduced to 6, for the standard deviation equation, and to 5,
for the generalized deviation equation.
Furthermore, it is interesting to note that in the case of the standard deviation
equation, the opinion of the authors [2, 8] differs when it comes to the number
of required test bodies. This seems to be related to the counting scheme and the
interpretation of the notion of a compass. Since no explicit compass solutions were
given in [2, 8], one cannot make a comparison to our results. In the case of [48],
we were not able to verify that the given solution does fulfill the compass equations
derived in that work. However, the agreement on the number of required bodies in
combination with the standard deviation is reassuring.
Our results are of direct operational relevance and form the basis for many experi-
ments. Important applications range from the description of gravitational wave detec-
tors to the study of satellite configurations for gravitational field mapping in rela-
tivistic geodesy.
4.2 Method 2: Summary
Section 3.2 describes an experimental setup which we call a clock compass, in anal-
ogy to the usual gravitational compass [7, 8]. We have shown that a suitably prepared
set of clocks can be used to determine all components of the gravitational field, i.e.
the curvature, in General Relativity, as well as to describe the state of motion of a
noninertial observer.
Working out explicit clock compass setups in different situations, we have demon-
strated that in general 6 clocks are needed to determine the linear acceleration as
well as the rotational velocity, while 4 clocks will suffice in case of the velocity.
Furthermore, we prove that one needs 21 and 11 clocks, respectively, to determine
all curvature components in a general curved spacetime and in vacuum. In view
of possible future experimental realizations it is interesting to note that restrictions
regarding the choice of clock velocities in a setup lead to restrictions regarding the
number of determinable curvature components.
Our results are of direct operational relevance for the setup of networks of clocks,
especially in the context of relativistic geodesy. In geodetic terms, the given clock
configurations may be thought of as a clock gradiometers. Taking into account the
steadily increasing accuracy of clocks [50], these results should be combined with
those from a gradiometric context, for example in the form of a hybrid gravitational
compass – which combines acceleration as well as clock measurements in one setup.
Another possible application is the detection of gravitational waves by means of
clock as well as standard interferometric techniques. An interesting question is a
possible reduction of the number of measurements by a combination of different
techniques.
5 Outlook: Operational Determination of the Gravitational

Field in Theories Beyond GR
In the previous sections, we have shown how the deviation equation as well as an
ensemble of clocks can be used to measure the gravitational field in GR. However, the
results were limited to theories in a Riemannian background. While such theories are
justified in many physical situations, several modern gravitational theories [51–53]
reach significantly beyond the Riemannian geometrical framework. In particular it is
already well-known [12, 54, 55], that in the description of test bodies with intrinsic
degrees of freedom – like spin – there is a natural coupling to the post-Riemannian
features of spacetime. Therefore, in view of possible tests of gravitational theories
by means of structured test bodies, a further extension of the deviation equation to
post-Riemannian geometries is needed.
In the following we present a generalized deviation equation in a Riemann–Cartan
background, allowing for spacetimes endowed with torsion, the presentation is based
on [56]. This equation describes the dynamics of the connecting vector which links
events on two general (adjacent) world lines. Our results are valid for any theory in
a Riemann–Cartan background, in particular they apply to Einstein–Cartan theory
[57] as well as to Poincaré gauge theory [58, 59]. Interestingly, Synge was apparently
the first who derived the deviation equation for the Riemann–Cartan geometry [60].
5.1 World Function and Deviation Equation
Let us briefly recapitulate the relevant steps which lead to the generalized deviation
equation: We want to compare two general curves Y (t) and X (t˜) in an arbitrary
spacetime manifold. Here t and t˜ are general parameters, i.e. not necessarily the
proper time on the given curves. In contrast to the Riemannian case, see Sect. 2.1,
we now connect two points x ∈ X and y ∈ Y on the two curves by the autoparallel
joining the two points (we assume that this autoparallel is unique). An autoparallel is
a curve along which the velocity vector is transported parallel to itself with respect to
the connection on the spacetime manifold. In a Riemannian space autoparallel curves
coincide with geodesic lines. Along the autoparallel we have the world function σ ,
Fig. 9 Sketch of the two

arbitrarily parametrized
world lines Y (t) and X (t˜),
and the (dashed) autoparallel
connecting two points on
these world line. The
generalized deviation vector
along the reference world
line Y is denoted by η y
and conceptually the closest object to the connecting vector between the two points
is the covariant derivative of the world function, denoted at the point y by σ y , cf.
Fig. 9.
At this point, the generality of our derivation of the deviation equation from
Sect. 2.1 pays of, i.e. the exact deviation equation given in Eq. 12 can be directly used
in the case in which the connecting curve is an autoparallel therefore, to recapitulate,
we have:
2
D 2 y1 d t˜
2
η = −σ y1
y y
2 3
u y2 y3
u − σ y1
y2
a y2
− σ y1
x 2
ã x2
dt dt
y4
Dσ
−2σ y1 y2 x3 u y2 K x3 y4 u y4 − H x3 y4
dt
y4
Dσ
−σ y1 x2 x3 K x2 y4 u y4 − H x2 y4
dt
y5
Dσ
× K x3 y5 u y5 − H x3 y5
dt
2˜ y3
dt d t x2 Dσ
−σ y1 x2 K x2
y3 u y3
− H y3 . (164)
d t˜ dt 2 dt
Again the factor d t˜/dt by requiring that the velocity along the curve X is normalized,
i.e. ũ x ũ x = 1, in which case we have
d t˜ Dσ y2
= ũ x1 K x1 y2 u y2 − ũ x1 H x1 y2 . (165)
dt dt
Equation (164) is the exact generalized deviation equation, it is completely general

and can be viewed as the extension of the standard geodesic deviation (Jacobi)
equation to any order.
5.2 World Function in Riemann–Cartan Spacetime
In order to arrive at an expanded approximate version of the deviation equation,

we need to work out the properties of a world function based on autoparallels in a
Riemann–Cartan background. In contrast to a Riemannian spacetime a Riemann–
Cartan spacetime is endowed with an asymmetric connection ab c , and there will
be differences when it comes to the basic properties of a world function σ based
on autoparallels. We base our presentation on [56], other relevant references which
contain some results in a Riemann–Cartan context are [61–68].
For a world function σ based on autoparallels, we have the following basic rela-
tions in the case of spacetimes with asymmetric connections:
σ x σx = σ y σ y = 2σ, (166)
σ x2 σx2 x1 = σ x1 , (167)
σx1 x2 − σx2 x1 = Tx1 x2 x3 ∂x3 σ. (168)
Note in particular the change in (168) due to the presence of the spacetime torsion
Tx1 x2 x3 , which leads to σx1 x2 = σx2 x1 , in contrast to the symmetric Riemannian case,
s
in which σ x1 x2 = σ x2 x1 holds.1
In many calculations the limiting behavior of a bitensor B... (x, y) as x approaches
the references point y is required. This so-called coincidence limit of a bitensor
B... (x, y) is a tensor
[B... ] = lim B... (x, y), (169)

x→y
at y and will be denoted by square brackets. In particular, for a bitensor B with

arbitrary indices at different points (here just denoted by dots), we have the rule [2]

[B... ];y = B...;y + B...;x . (170)
1 We use “s” to indicate relations which only hold for symmetric connections and denote Riemannian
objects by the overbar.

We collect the following useful identities for the world function σ :
[σ ] = [σx ] = [σ y ] = 0, (171)
[σx1 x2 ] = [σ y1 y2 ] = g y1 y2 , (172)
[σx1 y2 ] = [σ y1 x2 ] = −g y1 y2 , (173)
[σx3 x1 x2 ] + [σx2 x1 x3 ] = 0. (174)
Note that up to the second covariant derivative the coincidence limits of the world
function match those in spacetimes with symmetric connections. However, at the
next (third) order the presence of the torsion leads to
1
[σx1 x2 x3 ] = Ty1 y3 y2 + Ty2 y3 y1 + Ty1 y2 y3 = K y2 y1 y3 , (175)
2
where in the last line we made use of the contortion2 K ab c = ab c − ab c . With the
help of (170) we obtain for the other combinations with three indices:
[σ y1 x2 x3 ] = −[σ y1 y2 x3 ] = [σ y1 y2 y3 ] = K y2 y1 y3 . (176)
The non-vanishing of these limits leads to added complexity in subsequent calcula-

tions compared to the Riemannian case.
At the fourth order we have
K y1 y y2 K y3 yy4 + K y1 y y3 K y2 yy4 + K y1 y y4 K y2 yy4

+[σx4 x1 x2 x3 ] + [σx3 x1 x2 x4 ] + [σx2 x1 x3 x4 ] = 0, (177)
and in particular
1 1
[σx1 x2 x3 x4 ] = ∇ y K y3 y2 y4 + K y4 y2 y3 + ∇ y3 3K y2 y1 y4 − K y1 y2 y4
3 1 3
1
+ ∇ y4 3K y2 y1 y3 − K y1 y2 y3 + π y1 y2 y3 y4 , (178)
3
1 1
[σx1 x2 x3 y4 ] = − ∇ y1 K y3 y2 y4 + K y4 y2 y3 − ∇ y3 3K y2 y1 y4 − K y1 y2 y4
3 3
1
+ ∇ y4 K y1 y2 y3 − π y1 y2 y3 y4 , (179)
3
1 1
[σx1 x2 y3 y4 ] = ∇ y1 K y4 y2 y3 + K y3 y2 y4 − ∇ y4 K y1 y2 y3
3 3
1
− ∇ y3 K y1 y2 y4 + π y1 y2 y4 y3 , (180)
3
2 The contortion K y2 y1 y3 should not be confused with the Jacobi propagator K x y .

1 1 1
[σx1 y2 y3 y4 ] = − ∇ y1 K y3 y4 y2 + K y2 y4 y3 + ∇ y3 K y1 y4 y2 + ∇ y2 K y1 y4 y3
3 3 3
+∇ y4 K y3 y1 y2 − π y1 y4 y3 y2 , (181)
1 1 1
[σ y1 y2 y3 y4 ] = ∇ y4 −2K y2 y3 y1 + K y1 y3 y2 − ∇ y2 K y4 y3 y1 − ∇ y1 K y4 y3 y2
3 3 3
−∇ y3 K y2 y4 y1 + π y4 y3 y2 y1 , (182)
1
π y1 y2 y3 y4 := K y1 y2 y K y3 y4 y + K y4 y3 y − K y1 y3 y K y4 y2 y + K yy2 y4
3
−K y1 y4 y K y3 y2 y + K yy2 y3 − 3K y2 y1 y K y3 y4 y + K y3 y1 y K yy2 y4

+K y4 y1 y K yy2 y3 + R y1 y3 y2 y4 + R y1 y4 y2 y3 . (183)
Again, we note the added complexity compared to the Riemannian case, in which
s
we have [σx1 x2 x3 x4 ] = 13 R y2 y4 y1 y3 + R y3 y2 y1 y4 at the fourth order. In particular, we
observe the occurrence of derivatives of the contortion in (178)–(182).
Finally, let us collect the basic properties of the so-called parallel propagator
g y x := e(a) ex(a) , defined in terms of a parallelly propagated tetrad e(a) , which in turn
y y
allows for the transport of objects, i.e. V y = g y x V x , V y1 y2 = g y1 x1 g y2 x2 V x1 x2 , etc.,

along an autoparallel:
g y1 x g x y2 = δ y1 y2 , g x1 y g y x2 = δ x1 x2 , (184)
σ x ∇x g x1 y1 = σ y ∇ y g x1 y1 = 0,
σ x ∇x g y1 x1 = σ y ∇ y g y1 x1 = 0, (185)
σx = −g y x σ y , σ y = −g x y σx . (186)
Note in particular the coincidence limits of its derivatives

g x0 y1 = δ y0 y1 , (187)
x0
g x0 y1 ;x2 = g y1 ;y2 = 0, (188)
x0
g y1 ;x2 x3 = − g x0 y1 ;x2 y3 = g x0 y1 ;x2 x3
1
= − g x0 y1 ;y2 y3 = R y0 y1 y2 y3 . (189)
2
In the next section we will derive an expanded approximate version of the devi-
ation equation. For this we first work out the expanded version of quantities around
the reference world line Y . In particular, we make use of the covariant expansion
technique [2, 3] on the basis of the autoparallel world function.
Expanded Riemann–Cartan deviation equation For a general bitensor B... with a

given index structure, we have the following general expansion, up to the third order
(in powers of σ y ):
B y1 ...yn = A y1 ...yn + A y1 ...yn+1 σ yn+1

1
+ A y1 ...yn+1 yn+2 σ yn+1 σ yn+2 + O σ 3 , (190)
2
A y1 ...yn := B y1 ...yn , (191)

A y1 ...yn+1 := B y1 ...yn ;yn+1 − A y1 ...yn ;yn+1 , (192)

A y1 ...yn+2 := B y1 ...yn ;yn+1 yn+2 − A y1 ...yn y0 σ y0 yn+1 yn+2
−A y1 ...yn ;yn+1 yn+2 − 2 A y1 ...yn (yn+1 ;yn+2 ) . (193)
With the help of (190) we are able to iteratively expand any bitensor to any order,
provided the coincidence limits entering the expansion coefficients can be calcu-
lated. The expansion for bitensors with mixed index structure can be obtained from
transporting the indices in (190) by means of the parallel propagator.
In order to develop an approximate form of the generalized deviation Eq. (164)
up to the second order, we need the following expansions of the derivatives of the
world function:

σ y1 y2 = g y1 y2 + K y2 y1 y3 σ y3 + O σ 2 , (194)

σ y1 x2 = −g y1 x2 + gx2 y K y3 yy1 σ y3 + O σ 2 , (195)

1
σ y1 y2 y3 = K y2 y1 y3 + ∇ y4 K y2 y3 y1 + K y1 y3 y2
3
− ∇ y2 K y4 y3 y1 − ∇ y1 K y4 y3 y2 − 3∇ y3 K y2 y4 y1

+ 3π y4 y3 y2 y1 σ y4 + O σ 2 , (196)

1
σ y1 y2 x3 = gx3 y3 K y2 y3 y1 − ∇ y3 K y2 y4 y1 + K y1 y4 y2
3
− ∇ y2 K y3 y4 y1 − ∇ y1 K y3 y4 y2

+ 3π y3 y4 y2 y1 σ y4 + O σ 2 , (197)

σ y1 x2 x3 = gx2 y2 gx3 y3 K y3 y1 y2

1
+ ∇ y2 K y4 y3 y1 + K y1 y3 y4
3

− ∇ y4 K y2 y3 y1 + 3K y3 y1 y2

− ∇ y1 K y2 y3 y4 + 3π y2 y3 y4 y1 σ y4 + O σ 2 . (198)
The Jacobi propagators are approximated as follows

H x1 y2 = g x1 y2 + K y3 y2 x1 σ y3 + O σ 2 , (199)

K x1 y2 = g x1 y2 + K y2 x1 y3 + K y3 y2 x1 σ y3 + O σ 2 , (200)
which in turn allows for an expansion of the recurring term entering (164):
Dσ y2 Dσ y

K x1 y2 u y2 − H x1 y2 = g x1 y u y − + Ty2 y3 y u y2 σ y3 +O σ 2 . (201)
dt dt
Synchronous parametrization Before writing down the expanded version of the
generalized deviation equation, we will simplify the latter by choosing a proper
parametrization of the neighboring curves. The factors with the derivatives of the
parameters t and t˜ appear in (164) due to the non-synchronous parametrization of the
two curves. It is possible to make things simpler by introducing the synchronization
of parametrization. Namely, we start by rewriting the velocity as
dY y d t˜ dY y
uy = = . (202)
dt dt d t˜
That is, we now parametrize the position on the first curve by the same variable t˜
that is used on the second curve. Accordingly, we denote
dY y
uy = . (203)
d t˜
By differentiation, we then derive
2
d 2 t˜ y d t˜
ay = u + ay, (204)
dt 2 dt
where
D y D2Y y
ay = u = . (205)
d t˜ d t˜2
Analogously, we derive for the derivative of the deviation vector
2
D2η y d 2 t˜ Dη y d t˜ D2η y
= + . (206)
dt 2 dt 2 d t˜ dt d t˜2
Now everything is synchronous in the sense that both curves are parametrized by t˜.
As a result, the exact deviation Eq. (164) is recast into a simpler form
D 2 y1
η = −σ y1 y2 a y2 − σ y1 x2 ã x2 − σ y1 y2 y3 u y2 u y3
d t˜2

Dσ y4
−2σ y1 y2 x3 u y2 K x3 y4 u y4 − H x3 y4
d t˜
y4
Dσ
−σ y1 x2 x3 K x2 y4 u y4 − H x2 y4
d t˜
y5
Dσ
× K x3 y5 u y5 − H x3 y5 . (207)
d t˜
Explicit expansion of the deviation equation Substituting the expansions (194)–

(201) into (207), we obtain the final result
D 2 y1 y3
y1 y2 Dη
η = ã y1
− a y1
+ T y y u − y1 y2 y3 y4 u y2 u y3
d t˜2
2 3
d t˜
2
+ K y2 y4 a − K y4 y2 ã η + O σ ,
y1 y2 y1 y2 y4
(208)

where we introduced the abbreviation

y1 y2 y3 y4 := 2π y3 y4 y2 y1 − π y4 y3 y2 y1 − π y2 y3 y4 y1 + Ty y2 y1 Ty4 y3 y
−2∇ y2 K (y1 y3 )y4 + ∇ y1 K y2 y3 y4 − ∇ y4 K y2 y3 y1 . (209)
It should be understood that the last expression is contracted with u y2 u y3 and hence

the symmetrization is naturally imposed on the indices (y2 y3 ).
Equation (208) allows for the comparison of two general world lines in Riemann–
Cartan spacetime, which are not necessarily geodetic or autoparallel. It therefore
represents the generalization of the deviation equation derived in Eq. (38).
Riemannian case A great simplification is achieved in a Riemannian background,
when
y1 y2 y3 y4 = 2π y3 y4 y2 y1 − π y4 y3 y2 y1 − π y2 y3 y4 y1
= R y1 y3 y2 y4 , (210)
and (208) is reduced to
D 2 y1 s y1
η = ã y1 − a y1 − R y2 y3 y4 u
y2
u y3 η y4 + O σ 2 . (211)
d t˜2
Along geodesic curves, this equation is further reduced to the well-known geodesic
deviation (Jacobi) equation.
Choice of coordinates In order to utilize the deviation equation for measurements or
in a gravitational compass setup [2, 7, 8, 48], the occurring covariant total derivatives
need to be rewritten and an appropriate coordinate choice needs to be made. The left-
hand side of the deviation equation takes the form:
D 2 ηa ◦◦ ◦
2
= u̇ b ∇b ηa + η a −2u b ba d ηd −u b u c cb d ∂d ηa
dt
−u b u c ηe ∂c ba e − cb d da e − ca d bd e . (212)
◦
Here we used η a := dηa /dt for the standard total derivative.
Observe that the first term on the right-hand side vanishes in the case of autopar-
allel curves (u̇ a := Du a /dt = 0). Also note the symmetrization of the connection
imposed by the velocities in some terms.
Rewriting the connection in terms of the contortion and switching to normal coor-
dinates [23, 24, 69–73] along the world line, which we assume to be an autoparallel,
yields
D 2 ηa |Y ◦◦ ◦
2
= η a +2u b K ba d ηd +u b u c K cb d ∂d ηa
dt
2
+u u ηe ∂c K ba − R c(ba) + K cb K da − K ca K bd .
b c e e d e d e
(213)
3
Note the appearance of a term containing the partial (not ordinary total) derivative
of the deviation vector, in contrast to the Riemannian case.
The first term in the second line may be rewritten as an ordinary total derivative, i.e.
◦
u b u c ηe ∂c K ba e = u b ηe K bae , but this is still inconvenient when recalling the compass
equation, which will contain terms with covariant derivatives of the contortion.
5.3 Operational Interpretation
Some thoughts about the operational interpretation of the coordinate choice are in
order. In particular, it should be stressed that we did not specify any physical theory in
which the deviation Eq. (207) should be applied. Or, stated the other way round, the
derived deviation equation is of completely geometrical nature, i.e. it describes the
change of the deviation vector between points on two general curves in Riemann–
Cartan spacetime. From the mathematical perspective, the choice of coordinates
should be solely guided by the simplicity of the resulting equation. In this sense, our
previous choice of normal coordinates appears to be appropriate. But what about the
physical interpretation, or better, the operational realization of such coordinates?
Let us recall the coordinate choice in General Relativity in a Riemannian back-
ground. In this case normal coordinates also have a clear operational meaning, which
is related to the motion of structureless test bodies in General Relativity. As is well
known, such test bodies move along the geodesic equation. In other words, we could
– at least in principle – identify a normal coordinate system by the local observation
of test bodies. If other external forces are absent, normal coordinates will locally
– where “locally” refers to the observers laboratory on the reference world line –
lead to straight line motion of test bodies. In this sense, there is a clear operational
procedure for the realization of normal coordinates.
Here we are in a more general situation, since we have not yet specified which
gravitational theory we are considering in the geometrical Riemann–Cartan back-
ground. The physical choice of a gravity theory will be crucial for the operational
realization of the coordinates. Recall the form of the equations of motion for a very
large class [12, 55] of gravitational theories, which also allow for additional internal
degrees of freedom, in particular for spin. In this case the equations of motion are no
longer given by the geodesic equation or, as it is sometimes erroneously postulated
in the literature, by the autoparallel equation. In such theories, test bodies exhibit
an additional spin-curvature coupling, which leads to non-geodesic motion, even
locally.
In the context of gravitational theories beyond GR, one should therefore be aware
of the fact, that for the experimental realization of the normal coordinates, one now
has to make sure to use the correct equation of motion and, consequently, the correct
type of test body. Taking the example of a theory with spin-curvature coupling, like
Einstein–Cartan theory, this would eventually lead to the usage of test bodies with
vanishing spin – since those still move on standard geodesics, and therefore lead to
an identical procedure as in the general relativistic case, i.e. one adopts coordinates
in which the motion of those test bodies becomes rectilinear.
5.4 Summary
This concludes or outlook and the generalization of the deviation equation to a

Riemann–Cartan geometry. The generalization should serve as a foundation for the
test of gravitational theories which make use of post-Riemannian geometrical struc-
tures. As we have discussed in detail, the operational usability of the Riemann–Cartan
deviation equation differs from the one in a general relativistic context, which was
also noticed quite early in [51]. In contrast to the Riemannian case, an algebraic real-
ization of a gravitational compass [2, 7, 8, 48] on the basis of the deviation equation
is out of the question due to the appearance of derivatives of the torsion even at the
lowest orders. It remains to be shown which additional concepts and assumptions
are needed in order to fully realize a gravitational compass in a Riemann–Cartan
background.
Acknowledgements This work was supported by the Deutsche Forschungsgemeinschaft (DFG)

through the grant PU 461/1-1 (D.P.). The work of Y.N.O. was partially supported by PIER (“Part-
nership for Innovation, Education and Research” between DESY and Universität Hamburg) and by
the Russian Foundation for Basic Research (Grant No. 16-02-00844-A).
Appendix
A Directory of Symbols
Table 1 Directory of symbols

Symbol Explanation
Geometrical quantities
gab Metric
√
−g Determinant of the metric
δba Kronecker symbol
εabcd , εαβγ (4D, 3D) Levi-Civita symbol
s, τ Proper time
x a , ya Coordinates
λb (α) (Fermi propagated) tetrad
Y (s), X (τ ) (Reference) world line
ξa Constants in spatial Fermi coordinates
ab c , ab c (Levi-Civita) connection
Tab c , K ab c Torsion, contortion
∗ c Derivative of connection (normal coordinates)
ab...
Rabc d , Cabc d Riemann, Weyl curvature
σ World function
ηy Deviation vector
g y0 x0 Parallel propagator
K x y, H x y Jacobi propagators
Misc
ua , ab 4-velocity, 4-acceleration
v α , ωα , V α (Linear, rotational, combined) 3-velocity
α
b ,η α Derivative of (linear, rotational) acceleration
Operators
∂i , “,” Partial derivative
∇i , “;” Covariant derivative
D
ds = “˙” Total covariant derivative
d
=“ ◦” Total derivative
ds
“[. . . ]” Coincidence limit
“” Riemannian object
Table 2 Directory of symbols (continued)

Symbol Explanation
Auxiliary quantities (Method 1)
(m,n) A , (m,n, p) A Accelerations of compass constituents
a a
α y0 y1 ...yn , β y0 y1 ...yn , γ y0 y1 ...yn Expansion coefficients
c(m)a , d(m)a Constants
φ y1 y2 ... , λ y1 y2 ... , μ y1 y2 ... , Abbreviations
i1 ... , i1 ...
Auxiliary quantities (Method 2)
C Frequency ratio
A, B, K 1,2,3 Abbreviations
Auxiliary quantities (Outlook)
A y1 ...yn Expansion coefficient
π y1 y2 y3 y4 Abbreviation
References
1. F.A.E. Pirani, On the physical significance of the Riemann tensor. Acta Phys. Pol. 15, 389
(1956)
2. J.L. Synge, Relativity: The General Theory (North-Holland, Amsterdam, 1960)
3. E. Poisson, A. Pound, I. Vega, The motion of point particles in curved spacetime. Living Rev.
Relativ. 14(1), 7 (2011)
4. T. Levi-Civita, Sur l’écart géodésique. Math. Ann. 97, 291 (1927)
5. J.L. Synge, The first and second variations of the length integral in Riemannian space. Proc.
Lond. Math. Soc. 25, 247 (1926)
6. J.L. Synge, On the geometry of dynamics. Phil. Trans. R. Soc. Lond. A 226, 31 (1927)
7. D. Puetzfeld, Y.N. Obukhov, Generalized deviation equation and determination of the curvature
in general relativity. Phys. Rev. D 93, 044073 (2016)
8. P. Szekeres, The gravitational compass. J. Math. Phys. 6, 1387 (1965)
9. D. Puetzfeld, Y.N. Obukhov, C. Lämmerzahl, Gravitational clock compass in general relativity.
Phys. Rev. D 98, 024032 (2018)
10. B.S. DeWitt, R.W. Brehme, Radiation damping in a gravitational field. Ann. Phys. (N.Y.) 9,
220 (1960)
11. D. Puetzfeld, Y.N. Obukhov, Covariant equations of motion for test bodies in gravitational
theories with general nonminimal coupling. Phys. Rev. D 87, 044045 (2013)
12. D. Puetzfeld, Y.N. Obukhov, Equations of motion in metric-affine gravity: a covariant unified
framework. Phys. Rev. D 90, 084034 (2014)
13. A.C. Ottewill, B. Wardell, Transport equation approach to calculations of Hadamard Green
functions and non-coincident DeWitt coefficients. Phys. Rev. D 84, 104039 (2011)
14. D.E. Hodgkinson, A modified theory of geodesic deviation. Gen. Relativ. Gravit. 3, 351 (1972)
15. S.L. Bażański, Kinematics of relative motion of test particles in general relativity. Ann. H.
Poin. A 27, 115 (1977)
16. A.N. Aleksandrov, K.A. Piragas, Geodesic structure: I. Relative dynamics of geodesics. Theor.
Math. Phys. 38, 48 (1978)
17. B. Schutz, On generalized equations of geodesic deviation, in Galaxies, Axisymmetric Systems,
and Relativity, vol 17, ed. by M.A.H. MacCallum (Cambridge University Press, Cambridge,
1985), p. 237
18. C. Chicone, B. Mashhoon, The generalized Jacobi equation. Class. Quantum Gravity 19, 4231
(2002)
19. T. Mullari, R. Tammelo, On the relativistic tidal effects in the second approximation. Class.
Quantum Gravity 23, 4047 (2006)
20. J. Vines, Geodesic deviation at higher orders via covariant bitensors. Gen. Relativ. Gravit. 47,
59 (2015)
21. E. Fermi, Sopra i fenomeni che avvengono in vicinanza di una linea oraria. Atti. Accad. Naz.
Lincei Cl, Sci. Fis. Mat. Nat. Rend. 31, 21, 51, 101 (1922)
22. E. Fermi, Collected Papers, vol 1, ed. by E. Amaldi, E. Persico, F. Rasetti, E. Segrè (University
of Chicago Press, Chicago, 1962)
23. O. Veblen, Normal coordinates for the geometry of paths. Proc. Natl. Acad. Sci. (USA) 8, 192
(1922)
24. O. Veblen, T.Y. Thomas, The geometry of paths. Trans. Am. Math. Soc. 25, 551 (1923)
25. J.L. Synge, A characteristic function in Riemannian space and its application to the solution
of geodesic triangles. Proc. Lond. Math. Soc. 32, 241 (1931)
26. A.G. Walker, Relative coordinates. Proc. R. Soc. Edinb. 52, 345 (1932)
27. F.K. Manasse, C.W. Misner, Fermi normal coordinates and some basic concepts in differential
geometry. J. Math. Phys. 4, 735 (1963)
28. C.W. Misner, K.S. Thorne, J.A. Wheeler, Gravitation (Freeman, San Francisco, 1973)
29. W.-T. Ni, On the proper reference frame and local coordinates of an accelerated observer in
special relativity. Chin. J. Phys. 15, 51 (1977)
30. B. Mashhoon, Tidal radiation. Astrophys. J. 216, 591 (1977)
31. W.-T. Ni, M. Zimmermann, Inertial and gravitational effects in the proper reference frame of
an accelerated, rotating observer. Phys. Rev. D 17, 1473 (1978)
32. W.-Q. Li, W.-T. Ni, On an accelerated observer with rotating tetrad in special relativity. Chin.
J. Phys. 16, 214 (1978)
33. W.-T. Ni, Geodesic triangles and expansion of the metrics in normal coordinates. Chin. J. Phys.
16, 223 (1978)
34. W.-Q. Li, W.-T. Ni, Coupled inertial and gravitational effects in the proper reference frame of
an accelerated, rotating observer. J. Math. Phys. 20, 1473 (1979)
35. W.-Q. Li, W.-T. Ni, Expansions of the affinity, metric and geodesic equations in Fermi normal
coordinates about a geodesic. J. Math. Phys. 20, 1925 (1979)
36. N. Ashby, B. Bertotti, Relativistic effects in local inertial frames. Phys. Rev. D 34, 2246 (1986)
37. A.M. Eisele, On the behaviour of an accelerated clock. Helv. Phys. Acta 60, 1024 (1987)
38. T. Fukushima, The Fermi coordinate system in the post-Newtonian framework. Celest. Mech.
44, 1024 (1988)
39. O. Semerák, Stationary frames in the Kerr field. Gen. Relativ. Gravit. 25, 1041 (1993)
40. K.-P. Marzlin, Fermi coordinates for weak gravitational fields. Phys. Rev. D 50, 888 (1994)
41. D. Bini, A. Geralico, R.T. Jantzen, Kerr metric, static observers and Fermi coordinates. J. Math.
Phys. 22, 4729 (2005)
42. C. Chicone, B. Mashhoon, Explicit Fermi coordinates and tidal dynamics in de Sitter and Gödel
spacetime. Phys. Rev. D 74, 064019 (2006)
43. D. Klein, P. Collas, General transformation formulas for Fermi-Walker coordinates. Class.
Quantum Gravity 25, 145019 (2008)
44. D. Klein, P. Collas, Exact Fermi coordinates for a class of space-times. J. Math. Phys. 51,
022501 (2010)
45. P. Delva, M.-C. Angonin, Extended Fermi coordinates. Gen. Relativ. Gravit. 44, 1 (2012)
46. S.G. Turyshev, O.L. Minazzoli, V.T. Toth, Accelerating relativistic reference frames in
Minkowski space-time. J. Math. Phys. 53, 032501 (2012)
47. F.W. Hehl, W.-T. Ni, Inertial effects of a Dirac particle. Phys. Rev. D 42, 2045 (1990)
48. I. Ciufolini, M. Demianski, How to measure the curvature of space-time. Phys. Rev. D 34, 1018
(1986)
49. I. Ciufolini, Generalized geodesic deviation equation. Phys. Rev. D 34, 1014 (1986)
50. R. Rodrigo, V. Dehant, L. Gurvits, M. Kramer, R. Park, P. Wolf, J. Zarnecki (eds.), High
Performance Clocks with Special Emphasis on Geodesy and Geophysics and Applications to
Other Bodies of the Solar System, vol. 63, Space Sciences Series of ISSI (Springer, Netherlands,
2018)
51. F.W. Hehl, P. von der Heyde, G.D. Kerlick, J.M. Nester, General relativity with spin and torsion:
foundations and prospects. Rev. Mod. Phys. 48, 393 (1976)
52. M. Blagojević, F.W. Hehl, Gauge Theories of Gravitation: A Reader with Commentaries (Impe-
rial College Press, London, 2013)
53. V.N. Ponomarev, A.O. Barvinsky, Y.N. Obukhov, Gauge Approach and Quantization Methods
in Gravity Theory (Nauka, Moscow, 2017)
54. F.W. Hehl, Y.N. Obukhov, D. Puetzfeld, On Poincaré gauge theory of gravity, its equations of
motion, and gravity probe B. Phys. Lett. A 377, 1775 (2013)
55. Y.N. Obukhov, D. Puetzfeld, Multipolar test body equations of motion in generalized gravity
theories, in Equations of Motion in Relativistic Gravity, vol. 179, Fundamental Theories of
Physics, ed. by D. Puetzfeld, et al. (Springer, Cham, 2015), p. 67
56. D. Puetzfeld, Y.N. Obukhov, Deviation equation in Riemann-Cartan spacetime. Phys. Rev. D
97, 104069 (2018)
57. A. Trautman, Einstein-Cartan theory, in Encyclopedia of Mathematical Physics, vol. 2, ed. by
J.-P. Francoise, G.L. Naber, S.T. Tsou (Elsevier, Oxford, 2006), p. 189
58. Y.N. Obukhov, Poincaré gauge gravity: selected topics. Int. J. Geom. Methods Mod. Phys. 03,
95 (2006)
59. Y.N. Obukhov, Poincaré gauge gravity: an overview. Int. J. Geom. Methods Mod. Phys. 15,
Supp. 1 (2018) 1840005
60. J.L. Synge, Geodesics in non-holonomic geometry. Math. Ann. 99, 738 (1928)
61. W.H. Goldthorpe, Spectral geometry and S O(4) gravity in a Riemann-Cartan spacetime. Nucl.
Phys. B 170, 307 (1980)
62. H.T. Nieh, M.L. Yan, Quantized Dirac field in curved Riemann-Cartan background: I. Sym-
metry properties, Green’s function. Ann. Phys. (N.Y.) 138, 237 (1982)
63. N.H. Barth, Heat kernel expansion coefficient: I. An extension. J. Phys. A Math. Gen. 20, 857
(1987)
64. S. Yajima, Evaluation of the heat kernel in Riemann-Cartan space. Class. Quantum Gravity 13,
2423 (1996)
65. S.S. Manoff, Auto-parallel equation as Euler-Lagrange’s equation in spaces with affine con-
nections and metrics. Gen. Relativ. Gravit. 32, 1559 (2000)
66. S.S. Manoff, Deviation equations of Synge and Schild over spaces with affine connections and
metrics. Int. J. Mod. Phys. A 16, 1109 (2001)
67. B.Z. Iliev. Deviation equations in spaces with a transport along paths. JINR Commun. E2-94-40,
Dubna, 1994 (2003)
68. R.J. van den Hoogen, Towards a covariant smoothing procedure for gravitational theories. J.
Math. Phys. 58, 122501 (2017)
69. T.Y. Thomas, The Differential Invariants of Generalized Spaces (Cambridge University Press,
Cambridge, 1934)
70. J.A. Schouten, Ricci-Calculus. An Introduction to Tensor Analysis and its Geometric Applica-
tions, 2nd edn. (Springer, Berlin, 1954)
71. I.G. Avramidi, A covariant technique for the calculation of the one-loop effective action. Nucl.
Phys. B 355, 712 (1991)
72. I.G. Avramidi, Covariant methods for the calculation of the effective action in quantum field
theory and investigation of higher-derivative quantum gravity. Ph.D. thesis, Moscow State
University (1986), English version arXiv:hep-th/9510140
73. A.Z. Petrov, Einstein Spaces (Pergamon, Oxford, 1969)
A Snapshot of J. L. Synge
Peter A. Hogan
Abstract A brief description is given of the life and influence on relativity theory
of Professor J. L. Synge accompanied by some technical examples to illustrate his
style of work.
1 Introduction
When I was a postdoctoral fellow working with Professor Synge in the School of
Theoretical Physics of the Dublin Institute for Advanced Studies he was fifty–one
years older than me and he remained research active for another twenty years. John
Lighton Synge FRS was born in Dublin on 23rd. March, 1897 and died in Dublin
on 30th. March, 1995. As well as his emphasis on, and mastery of, the geometry
of space–time he had a unique delivery, both verbal and written, which I will try to
convey in the course of this short article. But first the basic facts of his academic
life are as follows: He was educated in St. Andrew’s College, Dublin and entered
Trinity College, University of Dublin in 1915. He graduated B.A. (1919), M.A.
(1922) and Sc.D. (1926). He was Assistant Professor of Mathematics in the Uni-
versity of Toronto (1920–1925), subsequently returning to Trinity College Dublin
as Professor of Natural Philosophy (1925–1930) and then left for the University of
Toronto again to take up the position of Professor of Applied Mathematics (1930–
1943). From there he went to Ohio State University as chairman of the Mathematics
Department (1943–1946) followed by Head, Mathematics Department at Carnegie
Institute of Technology, Pittsburgh (1946–1948) before returning to Dublin to estab-
lish his school of relativity in the Dublin Institute for Advanced Studies. He officially
retired when he was seventy–five years old.
Synge was prolific, publishing 250 papers and 11 books. In 1986 he wrote, but did
not publish, some informal autobiographical notes [1], which he described as being
for his family and descendants and to aid obituary writers, and which are deposited in
P. A. Hogan (B)
School of Physics, University College Dublin, Belfield, Dublin 4, Ireland
e-mail: peter.hogan@ucd.ie

132 P. A. Hogan
the library of the School of Theoretical Physics of the Dublin Institute for Advanced
Studies.
Early in his career he published his first important paper: “On the Geometry of
Dynamics”, Phil. Trans. Roy. Soc. A 226 (1926), 31–106. Of this work he said [1];
“I sent a copy to T. Levi-Cività, and in return he sent me a copy of a paper by him,
just appearing. Our papers had in common the equation of geodesic deviation, now
familiar to relativists, but he had done it using an indefinite line element, appropriate
to relativity, whereas my line element was positive definite.”
2 A Scheme of Approximation
Synge placed great emphasis on working things out for oneself, writing that [2] “the
lust for calculation must be tempered by periods of inaction, in which the mechanism
is completely unscrewed and then put together again. It is the decarbonisation of the
mind.” As an illustration of this activity I give a weak field approximation scheme
published in 1970 by Synge [3] which has the advantage that it can be described
without reference to an example. This is a topic which, by 1970, had become a stan-
dard entry in textbooks on general relativity and one might be forgiven for thinking
that by then the last word had been said on it.
We first need some basic objects and notation. In Minkowskian space-time Synge
liked to use imaginary time (which some people find maddening!) and to write the
position 4–vector in rectangular Cartesians and time as
√
xa = (x, y, z, it) with a = 1, 2, 3, 4 and i = −1,
with the index in the covariant or lower position. The Minkowskian metric tensor in
these coordinates has components δab (the Kronecker delta). If the metric tensor of
a space–time has components of the form
gab = δab + γab ,
then he defined the “truncated Einstein tensor” Ĝ ab via
G ab = L ab + Ĝ ab ,
where G ab is the Einstein tensor calculated with the metric gab and
1 1
L ab = (γab,cc + γcc,ab − γac,cb − γbc,ca ) − δab (γcc,dd − γcd,cd ) .
2 2
The energy–momentum–stress tensor of matter giving rise to a gravitational field has
components T ab . With these preliminaries Synge’s strategy is as follows:
A Snapshot of J. L. Synge 133
(1) Given T ab , generate a sequence of metrics gab = δab + γab (M = 0, 1, 2, . . . , N );

M M
(2) Approximations are introduced by expressing the components T ab in terms of a
small parameter;
(3) Integrability conditions, equivalent to the equations of motion, are imposed to
terminate the sequence at a term which satisfies Einstein’s field equations with a
predicted order of approximation in terms of the small parameter.
The sequence is constructed as follows: with
∗ 1
γab = γab − δab γcc (M = 0, 1, 2, . . . N ) ,
M M 2 M
and
ab
H ab = T ab + (8 π)−1 Ĝ (γ ) (M = 0, 1, 2, . . . N ) ,
M M M
define the sequence {γab } by

M
∗
γab = 0 and γab = 16 π K rabs H r s (M = 1, 2, 3, . . . N ) .
0 M M−1
Here K rabs is an operator defined by
K rabs = −δar δbs J + J (δar Dbs + δbs Dar − δab Dr s )J ,
with Da = ∂/∂xa , Dab = ∂ 2 /∂xa ∂xb . The operator J is the inverse d’Alembertian:

1 x , t − |
f ( x − x |)
J f (
x , t) = − d3 x .
4π x − x |
|
Synge proved that the integrals involved in the implementation of the operator K
converge if the physical system is stationary (T ab ,4 = 0) for some period in the past.
He called this property J –convergence.
Approximations are introduced as follows: all {γab } defined above satisfy the coor-
M
dinate conditions
∗
γab,b = 0 (M = 0, 1, 2, . . . N ) .
M
Introduce approximations by assuming T ab = O(k) for some dimensionless param-

eter k then
γab − γab = O(k M ) (M = 1, 2, 3, . . . , N ) ,

M M−1
134 P. A. Hogan
from the definition of γab , and

ab ab
Ĝ − Ĝ = O(k M+1 ) ,
M M−1
from the quadratic nature of Ĝ ab . To obtain a solution of Einstein’s field equations in

the Nth. approximation, terminate the sequence {γab } at the N th. term by imposing
M
the Integrability Conditions/Equations of Motion in the N th. approximation:
ab
H ab
,b ≡ T ab ,b + (8 π)−1 Ĝ ,b =0.
N −1 N −1
Now
∗
γab = −16 π J H ab
,
N N −1
and
G ab + 8 π T ab = O(k N +1 ) ,
N
showing that Einstein’s field equations are approximately satisfied in this sense. This
scheme was subsequently utilised for the study of equations of motion in general
relativity [4–7].
In an amusing spin-off Synge [8] constructed the following divergence–free
pseudo–tensor: first write the vanishing covariant divergence of the energy–
momentum–stress tensor in the equivalent forms
T ab |b = 0 ⇔ T ab ,b + K a = 0 .
Here K a = cba
T cb + cb
b
T ac is not a tensor (so the position of the index a is not
significant; bc are the components of the Riemannian connection calculated with
a
the metric tensor gab ). Then define the pseudo–vector
Q a = J K a ⇒ Q a = K a ,
(with J the operator introduced above and the Minkowskian d’Alembertian oper-
ator) and define the pseudo–tensor
ϕab = Q a,b + Q b,a − δab Q c,c .
It thus follows that
ϕab,b = Q a + Q b,ab − Q c,ca = Q a = K a = −T ab ,b .

Hence
τ ab = T ab + ϕab = τ ba ,
is a pseudo–tensor with vanishing divergence (τ ab ,b = 0). However Synge offered,

in his characteristic style, these words of warning: “I refrain from attaching the words
momentum and energy to this pseudo–tensor or to integrals formed from it, because
I believe that we are barking up the wrong tree if we attach such important phys-
ical terms to mathematical constructs which lack the essential invariance property
fundamental in general relativity.”
3 Lorentz Transformations
Synge gave a succinct description of his early education when he wrote [1]: “Although
there are great gaps in my scientific equipment - like Hadamard, I could never get
my teeth into group theory - I think I have ranged more widely than most. I might
easily have stuck to classical subjects in which I was well trained as an undergraduate
(dynamics, hydrodynamics, elasticity), but I wanted to take part in the new subjects,
and in due course I mastered relativity but not quantum theory.” True to this back-
ground, when considering Lorentz transformations, Synge thought of the analogy
with “the kinematics of a rigid body with a fixed point” (in [9]) and thus the construc-
tion of a general rotation in three dimensional Euclidean space in terms of the Euler
angles. For Lorentz transformations the analogy requires six transformations of an
orthonormal tetrad to another orthonormal tetrad, involving three pseudo angles (the
arguments of hyperbolic functions) and three Euclidean angles. While this perspec-
tive is interesting the resulting formalism is not well suited to discussing the detailed
effect of Lorentz transformations on the null cone. In the second edition of his text on
special relativity Synge thanked I. Robinson and A. Taub “for pointing out an error
in Chap. IV of this book as first published (1955): singular Lorentz transformations
were overlooked.”
Taub was using spinors but Robinson had encountered the singular case in a
novel way [10–12]: Robinson was interested in the Schwarzschild solution in the
limit m → +∞. Starting with the Eddington–Finkelstein form

(d x 2 + dy 2 ) 2m
ds 2 = −r 2 2 + 2 du dr + 1 − du 2
1 + 4 (x + y )
1 2 2 r
and, using a clever coordinate transformation, Robinson wrote this in the form

r2 2
ds 2 = − (dξ 2
+ dη 2
) + 2 du dr + λ 2
− du 2 , λ = m −1/3
cosh2 λξ r
136 P. A. Hogan
Taking the limit λ → 0 (⇔ m → +∞) this becomes
2
ds 2 = −r 2 (dξ 2 + dη 2 ) + 2 du dr − du 2
r
This is another (different from Schwarzschild) Robinson–Trautman [13] type D vac-
uum space–time. The metric tensor has one term singular at r = 0. This line element
can be written in the form
ds 2 = −T 4/3 (d X 2 + dY 2 ) − T −2/3 d Z 2 + dT 2
which is a Kasner [14] solution of Einstein’s vacuum field equations. If we remove

the term singular at r = 0 above we have a line element
ds 2 = −r 2 (dξ 2 + dη 2 ) + 2 du dr
This is flat space–time and r = 0 is a null geodesic. Hence
ξ →ξ+a , η →η+b, u →u , r →r
where a, b are real constants, constitutes a Lorentz transformation leaving only the
null direction r = 0 invariant. This is a singular Lorentz transformation (or null rota-
tion) and the example moreover shows that such transformations exist and constitute
a two–parameter Abelian subgroup of the Lorentz group.
4 Synge on an Observation of E. T. Whittaker
I mentioned at the outset that Professor Synge remained research active well into
old age. To demonstrate this I want to give an example of some work carried out
when he was eighty–eight years old. For several years, starting in the early 1980s,
he and I found it convenient to correspond via letter. This allowed easy exchange
of the results of calculations before the age of email. He typed his letters, including
equations, on an ancient machine which he had used for years. The example I want
to give involves an observation due to E. T. Whittaker and to do it justice I must first
give a fairly extensive introduction.
Whittaker (in [15, 16]) was concerned with the Liénard–Wiechert electromagnetic
field of a moving charge e so we will need some notation which we can briefly
summarise as follows:
(1) Line element: ds 2 = ηi j d X i d X j = −d X 2 − dY 2 − d Z 2 + dT 2 .
(2) World line of charge: X i = wi (u) ; v i (u) = dwi /du with v i vi = +1 (⇒
v i = 4–velocity, u = arc length or proper time); a i = dv i /du = 4–acceleration ⇒
a i vi = 0)
(3) Retarded distance of X i from X i = wi (u):
r = ηi j (X i − wi (u))v j ≥ 0 ; ηi j (X i − wi (u))(X j − w j (u)) = 0 .
Let X i − wi (u) = r k i then k i ki = 0 and k i vi = +1. Parametrise the direction of

k i by x, y such that

1 1
k =
i
P0−1 −x, −y, −1 + (x 2 + y 2 ), 1 + (x 2 + y 2 ) ,
4 4
and then the normalisation k i vi = +1 implies

1 2 1 2
P0 = x v (u) + y v (u) + 1 − (x + y ) v (u) + 1 + (x + y ) v 4 (u) .
1 2 2 3 2
4 4
Whittaker observed that the Liénard–Wiechert 4–potential
e vi
Ai = ⇒ Ai ,i = 0 = Ai ,
r
could be written, modulo a gauge transformation, in the form
e vi
Ai = = K i j F, j + ∗ K i j G , j ,
r
where K i j = −K ji is a constant real bivector, with ∗ K i j = 21 i jkl K kl its dual, and

F, G are real–valued functions each satisfying the Minkowskian wave equation
F = 0 and G = 0 .
To establish this in coordinates x, y, r, u we need

∂ P2 ∂ki ∂ ∂ki ∂ ∂ ∂ ∂
=− 0 + + vi + ki − (1 − r ai k i ) ,
∂X i r ∂x ∂x ∂y ∂y ∂r ∂u ∂r
and

P2 ∂2 ∂2 ∂2 2 ∂ ∂2 2 ∂
= − 20 + − (1 − 2 ai k r )
i
+ +2 + .
r ∂x 2 ∂ y2 ∂r 2 r ∂r ∂u∂r r ∂u
In coordinates x, y, r, u Whittaker’s two wave functions are [17]

e y
F = − log(x 2 + y 2 ) and G = −e tan−1 ,
2 x
(two harmonic functions) and thus
138 P. A. Hogan

∂F e P02 ∂ki ∂ki
= x +y ,
∂ Xi r (x 2 + y 2 ) ∂x ∂y
and

∂G e P02 ∂ki ∂ki
= x −y .
∂ Xi r (x 2 + y 2 ) ∂y ∂x
Define
K i j = δ3i δ4 − δ4i δ3 and L i j = δ1i δ2 − δ2i δ1 = ∗ K i j ,

j j j j
then
e vi
Ai = K i j F, j + ∗ K i j G , j = + η i j , j ,
r
with

= e log{r P0−1 x 2 + y 2 } .
Whittaker pointed out that this decomposition is analogous to the splitting of a plane
light wave into two plane polarised components. A notable fact is that almost every
vacuum Maxwell field can be resolved into two parts in this way. The presentation
of Whittaker’s observation in coordinates x, y, r, u facilitates the derivation of the
explicit decomposition (see [17]) for the Goldberg–Kerr electromagnetic field [18].
The second, and final, part of the introduction, to enable us to appreciate Synge’s
contribution, involves a simple proof of this decomposition of a vacuum Maxwell
field in general.
We are working in Minkowskian space–time and we shall write the line element
as given above in rectangular Cartesian coordinates and time X i = (X, Y, Z , T ) with
i = 1, 2, 3, 4. In addition we shall make use of the following basis vector fields:
∂ ∂ ∂ ∂ ∂ ∂
ki = + , li =− + ,
∂X i ∂Z ∂T ∂X i ∂Z ∂T
∂ ∂ ∂ ∂ ∂ ∂
mi = +i , m̄ i = −i .
∂X i ∂X ∂Y ∂X i ∂X ∂Y
All scalar products (with respect to the Minkowskian metric) of the pairs of these
vectors vanish except k i li = +2 and m i m̄ i = −2. In what follows a complex self–
dual bivector satisfies: Ai j = −A ji and ∗ Ai j = i Ai j and a complex anti–self–dual
bivector satisfies: Bi j = −B ji and ∗ Bi j = −i Bi j , with the star denoting the Hodge
dual. A basis of complex anti–self–dual bivectors is given by
m i j = m i k j − m j ki , n i j = m̄ i l j − m̄ j li ,
and
li j = m i m̄ j − m̄ i m j + li k j − l j ki .
Let Fi j = −F ji be a candidate for a real Maxwell bivector. Since Fi j + i ∗ Fi j is an

anti–self–dual complex bivector it can be expanded on the basis above as
Fi j + i ∗ Fi j = φ0 n i j + φ1 li j + φ2 m i j ,
where φ0 , φ1 , φ2 are complex–valued functions of X i . Maxwell’s Equations
(F i j + i ∗ F i j ), j = 0 ,
imply integrability conditions for the existence of a complex–valued function Q(X i )

such that:
(a) Q is a wave function: Q = 0 ⇔ m̄ i m j Q ,i j = k i l j Q ,i j ;
(b) φ0 = 41 k i m j Q ,i j , φ1 = 14 k i l j Q ,i j , φ2 = − 41 l i m̄ j Q ,i j .
Let l¯i j denote the complex conjugate of li j , then l¯i j is self–dual. Define
1 1
Wi j = l¯i p Q , pj − l¯j p Q , pi = −W ji .
4 4
Since Q is a wave function it follows that Wi j is anti–self–dual. Expressing Wi j on
the anti–self–dual bivector basis, and using (b) above, results in
Wi j = Fi j + i ∗ Fi j .
Hence with 41 l¯i j = K i j − i ∗ K i j and Q = U + i V , we can write
Fi j = Ai, j − A j,i with Ai = K i j U, j + ∗ K i j V, j .
Thus in general an analytic solution of Maxwell’s vacuum field equations on

Minkowskian space–time can be constructed from a pair of real wave functions
U, V and a constant real bivector K i j = −K ji . The classic paper on this type of
result for zero rest mass, spin s fields is that of Penrose [19] (see also Stewart [20]).
When I wrote out this proof (incorporated into [21]) and sent it to Synge his
response was characteristic. He worked it all out for himself and sent me the following
proof in December, 1985:
Synge’s proof begins with
Lemma: With X i = (X, Y, Z , T ), ηi j = diag(−1, −1, −1, +1), Fi j = −F ji a
Maxwell field so that F i j , j = 0 ; F j,k + Fki, j + F jk,i = 0 then
Fi j = 0 at T = 0 ⇒ Fi j = 0 for all T .
“You cannot make energy out of nothing” (Synge).

140 P. A. Hogan
Corollary: If Fi j and Hi j are Maxwell fields then
Fi j = Hi j at T = 0 ⇒ Fi j = Hi j for all T .
With these preliminaries Synge stated the following:

Theorem: Given a Maxwell field Fi j and
Hi j = K i l U,l j + ∗ K i l V,l j − K j l U,li − ∗ K j l V,li ,
with K i j = −K ji = constants and U, V wave functions, then Hi j is a Maxwell field

and there exists K i j , U, V such that
Hi j = Fi j at T = 0 .
Comment: Clearly Hi j is a solution of Maxwell’s equations. The choice of K i j , U, V

is not unique. The theorem demands only their existence.
Proof Choose K i j = δ3i δ4 − δ3i δ4 then ∗ K i j = δ1i δ2 − δ2i δ1 and writing out Hi j =
j j j j
Fi j at T = 0 we find the following pairs of equations for the Cauchy data U, V, U,4 ,
V,4 for the wave functions at T = 0: (all equations evaluated at T = 0)
(A): (U,4 ),1 = F13 + V,23 and (U,4 ),2 = F23 − V,13 ;
(B): (V,4 ),1 = F24 − U,23 and (V,4 ),2 = −F14 + U,13 ;
(C): U,11 + U,22 = −F34 and V,11 + V,22 = −F12 .
If the equations (A) are consistent and if the equations (B) are consistent then (A),
(B) and (C) can in principle be solved for the Cauchy data. The consistency follows
from the assumption that Fi j is a Maxwell field since then (A) implies that
(U,4 ),12 − (U,4 ),21 = F13,2 − F23,1 + V,232 + V,131 = F13,2 + F32,1 + F21,3 = 0 ,
and (B) implies that
(V,4 ),12 − (V,4 ),21 = F24,2 + F14,1 − U,232 − U,131 = F24,2 + F14,1 + F34,3 = 0 ,
and the theorem is established.
5 Epilogue
When visitors came to the Center for Relativity in the University of Texas at Austin,
Alfred Schild, the founder of the Center and one of Synge’s former collaborators
[22] would enthusiastically point out to them that this was where Roy Kerr found
his solution. This raises the question: what were the stand–out works produced in
Professor Synge’s school of relativity in Dublin? I discussed this with George Ellis
Fig. 1 J. L. Synge
1897–1995
some time ago and we concluded that Felix Pirani’s study of the physical signifi-
cance of the Riemann tensor [23] and Werner Israel’s proof of the uniqueness of the
static black hole (uncharged [24] and charged [25]) are arguably the most profound
products of Synge’s school.
When Synge turned ninety years of age a small conference was organised in his
honour. His status within Ireland was reflected in the report in a national newspaper
which stated: “President Hillery [Head of State] attended a special event in the Dublin
Institute for Advanced Studies yesterday to wish a happy 90th. birthday to Professor
Emeritus J. L. Synge, Ireland’s most distinguished mathematician of the present
century. Although he has been retired for fifteen years, the professor, a nephew of
the playwright J. M. Synge, published three papers last year and has two more at
present in the course of publication” [Irish Times, 23rd. March, 1987].
My photograph of Professor Synge (Fig. 1) was taken in July, 1987 in my back
garden. Also present were two of Synge’s former students, Dermott Mc Crea (see [4,
5, 7] for example) and Stephen O’Brien (of the O’Brien–Synge junction conditions
[26]) together with Bill Bonnor who was visiting from the University of London.
References
1. J.L. Synge, Autobiography. Dublin Institute for Advanced Studies, unpublished (1986)
2. J.L. Synge, Relativity: The General Theory (North-Holland Publishing Company, Amsterdam,
1966)
3. J.L. Synge, Proc. R. Ir. Acad. A59, 11 (1970)
4. P.A. Hogan, J.D. McCrea, GRG J. 5, 77 (1974)
5. J.D. McCrea, G.M. O’Brien, GRG J. 9, 1101 (1977)
6. G.M. O’Brien, GRG J. 10, 129 (1979)
7. J.D. McCrea, GRG J. 13, 397 (1981)
142 P. A. Hogan
8. J.L. Synge, Nature 215, 102 (1967)

9. J.L. Synge, Relativity: The Special Theory (North-Holland Publishing Company, Amsterdam,
1965)
10. Private communication to the author from I. Robinson
11. W. Rindler, A. Trautman, Gravitation and Geometry, Bibliopolis, Naples (1987), p. 13
12. C. Barrabès, P.A. Hogan, Advanced General Relativity: Gravity Waves, Spinning Particles, and
Black Holes (Oxford University Press, Oxford, 2013)
13. I. Robinson, A. Trautman, Proc. R. Soc. A265, 463 (1962)
14. E. Kasner, Trans. Am. Math. Soc. 27, 155 (1925)
15. E.T. Whittaker, Proc. Lond. Math. Soc. 1, 367 (1903)
16. E.T. Whittaker, A History of the Theories of Aether and Electricity (Nelson, London, 1958), p.
410
17. G.F.R. Ellis, P.A. Hogan, Ann. Phys. (N.Y.) 210, 178 (1991)
18. J.N. Goldberg, R.P. Kerr, J. Math. Phys. 5, 172 (1964)
19. R. Penrose, Proc. R. Soc. A284, 159 (1965)
20. J.M. Stewart, Proc. R. Soc. A367, 527 (1979)
21. P.A. Hogan, J. Math. Phys. 28, 2087 (1987)
22. J.L. Synge, A. Schild, Tensor Calculus (University of Toronto Press, Toronto, 1949)
23. F.A.E. Pirani, Acta Phys. Pol. 15, 389 (1956)
24. W. Israel, Phys. Rev. 164, 1776 (1967)
25. W. Israel, Commun. Math. Phys. 8, 245 (1968)
26. J.L. Synge, S. O’Brien. Commun. Dubl. Inst. Adv. Stud. A9 (1952)
General Relativistic Gravity
Gradiometry
Bahram Mashhoon
Abstract Gravity gradiometry within the framework of the general theory of rela-
tivity involves the measurement of the elements of the relativistic tidal matrix, which
is theoretically obtained via the projection of the spacetime curvature tensor upon
the nonrotating orthonormal tetrad frame of a geodesic observer. The behavior of
the measured components of the curvature tensor under Lorentz boosts is briefly
described in connection with the existence of certain special tidal directions. Rel-
ativistic gravity gradiometry in the exterior gravitational field of a rotating mass is
discussed and a gravitomagnetic beat effect along an inclined spherical geodesic
orbit is elucidated.
1 Newtonian Gravity Gradiometry
Consider a distribution of matter of density ρ(t, x) and the corresponding Newtonian

gravitational potential (t, x) in an inertial frame of reference. In a source-free
region of space, we imagine two nearby test masses m a and m b that fall freely in
the potential along trajectories xa (t) and xb (t), respectively. Choosing one of
these as the reference trajectory, we are interested in the relative motion of these test
particles. With xb (t) as the fiducial path, let us define ξ(t) = xa (t) − xb (t). Newton’s
second law of motion implies that the instantaneous deviation vector ξ(t) between
the neighboring paths satisfies the tidal equation
d 2 ξi
+ κi j ξ j + O(|ξ|2 ) = 0 , (1)
dt 2
B. Mashhoon (B)
Department of Physics and Astronomy, University of Missouri,
Columbia, MO 65211, USA
e-mail: MashhoonB@missouri.edu
B. Mashhoon
School of Astronomy, Institute for Research in Fundamental Sciences (IPM),
P. O. Box 19395-5531, Tehran, Iran

144 B. Mashhoon
where −κi j ξ j is the first-order tidal acceleration and
∂2
κi j (t, x) = (2)
∂x i ∂x j
is the symmetric tidal matrix evaluated along the reference trajectory. In the Newto-
nian theory of gravitation, gravity gradiometry involves the measurement of κi j (t, x),
which is the gradient of the acceleration of gravity and can be determined, in princi-
ple, by means of Eq. (1).
The tidal matrix in Eq. (2) is independent of the test masses m a and m b as a
consequence of the principle of equivalence. The principle of equivalence of gravi-
tational and inertial masses ensures the universality of the gravitational interaction.
The modern history of the science of gravity gradiometry can be traced back to the
pioneering efforts of L. Eötvös, who employed a torsion-balance method to test the
principle of equivalence (1889–1922).
In Eq. (2), Poisson’s equation for , ∇ 2 = 4πG ρ, reduces to Laplace’s equa-
tion, ∇ 2 = 0, in the source-free region under consideration. In this case, ∇ 2 κi j = 0
and hence each element of the Newtonian tidal matrix is a harmonic function. More-
over, tidal matrix (2) is traceless; therefore, the shape of a tidally deformed test body
would generally tend to either a cigar-like or a pancake-like configuration when tides
are dominant, since the symmetric and traceless tidal matrix can in general have
either two positive and one negative or one positive and two negative eigenvalues,
respectively.
In recent years, gravity gradiometers of high sensitivity have been developed;
indeed, the Paik gravity gradiometer employs superconducting quantum interfer-
ence devices [1–3]. Furthermore, gravity gradients can now be measured via atom
interferometry as well [4, 5]. Gravity gradiometry has many important practical
applications. The magnitude of a gravity gradient is usually expressed in units of
Eötvös, 1 E = 10−9 s−2 .
To extend the treatment of gravity gradiometry to the relativistic domain, it is
necessary to introduce the quasi-inertial Fermi normal coordinate system that can
provide a physically meaningful interpretation of the measurement of relative motion
within the framework of general relativity (GR). In GR, masses m a and m b follow
geodesics and a hypothetical observer comoving with the fiducial test mass m b would
set up in the neighborhood of the reference trajectory a laboratory where the motion
of m a could be monitored. Such a quasi-inertial frame is represented by the Fermi
normal coordinate system [6–8]. In our treatment of Fermi coordinates in the next
section, we employ an extended framework [9–12], since in practice nongravitational
accelerations and rotations may be present.
General Relativistic Gravity Gradiometry 145
2 Fermi Coordinates
To develop the relativistic analogs of Eqs. (1) and (2), we consider a congruence
of future-directed timelike paths representing the world lines of test masses in a
gravitational field. Next, we choose a reference path in the congruence and estab-
lish a local quasi-inertial Fermi system of geodesic coordinates in its neighbor-
hood. This is necessary in order to provide a physically meaningful interpretation
of the measurement of relative motion from the standpoint of the observer comov-
ing with the reference test mass along the fiducial world line x̄ μ (τ ). The observer
has proper time τ and carries an orthonormal tetrad frame λμ α̂ (τ ) along x̄ μ ; that
is, gμν λμ α̂ λν β̂ = ηα̂β̂ , where gμν is the spacetime metric and ηα̂β̂ is the Minkowski
metric given by diag(−1, 1, 1, 1) in our convention. Here, λμ 0̂ (τ ) = d x̄ μ /dτ is the
observer’s temporal axis and its local frame is carried along its path according to
Dλμ α̂
= φα̂ β̂ λμ β̂ , (3)
dτ
where φα̂β̂ is the observer’s antisymmetric acceleration tensor. Greek indices run
from 0 to 3, while Latin indices run from 1 to 3. The signature of the spacetime
metric is +2 and units are chosen such that c = G = 1, unless specified otherwise.
In close analogy with the electromagnetic field tensor, we can decompose the
acceleration tensor into its “electric” and “magnetic” components, namely, φα̂β̂ →
(−A, ), where A(τ ) is a spacetime scalar that represents the translational accel-
eration of the fiducial observer and (τ ) is a spacetime scalar that represents its
rotational acceleration. More precisely, the reference observer in general follows an
accelerated world line with
d 2 x̄ μ ν
μ d x̄ d x̄
σ
+ νσ = Aμ , (4)
dτ 2 dτ dτ
where
Aμ = Aî λμ î (5)
and is the angular velocity of the rotation of the observer’s spatial frame with
respect to a locally nonrotating (i.e. Fermi–Walker transported) frame.
At each event x̄ μ (τ ) along the reference world line, we imagine all spacelike
geodesic curves that start out from this event and are normal to the reference world
line. These generate a local hypersurface. Let x μ be an event on this hypersurface
sufficiently close to the reference world line such that there is a unique spacelike
geodesic of proper length σ that connects x̄ μ (τ ) to x μ . We define ξ μ to be a unit
spacelike vector that is tangent to the unique spacelike geodesic at x̄ μ (τ ), so that
ξμ (τ ) λμ 0̂ (τ ) = 0. Then, to event x μ one assigns Fermi coordinates X μ̂ , where
X 0̂ := τ , X î := σ ξ μ (τ ) λμ î (τ ) . (6)
146 B. Mashhoon
The reference observer has Fermi coordinates X μ̂ = (τ , 0, 0, 0) and is thus perma-

nently fixed at the spatial origin of the Fermi coordinate system. Henceforth, we find
it convenient to express Fermi coordinates as X μ̂ = (T, X), where |X| = σ. When
σ = 0, X î /σ = ξ μ (τ ) λμ î (τ ), for i = 1, 2, 3, are the corresponding direction cosines
at proper time τ along x̄ μ .
The Fermi coordinate system is admissible in a cylindrical domain along x̄ μ of
radius |X| ∼ R, where R is a certain minimal radius of curvature of spacetime along
the reference world line.
The spacetime metric in Fermi coordinates is given by
ds 2 = gμ̂ν̂ (T, X) d X μ̂ d X ν̂ , (7)
where
g0̂0̂ = −P 2 + Q 2 − R0̂î 0̂ ĵ X î X ĵ + O(|X|3 ) , (8)

2
g0̂î = Q î − R X ĵ X k̂ + O(|X|3 ) , (9)
3 0̂ ĵ î k̂
1 ˆ
gî ĵ = δî ĵ − Rî k̂ ĵ lˆ X k̂ X l + O(|X|3 ) . (10)
3
Here, P and Q,
P := 1 + A(T ) · X , Q := (T ) × X , (11)
are related to the local translational and rotational accelerations of the reference
observer, respectively, and
Rα̂β̂ γ̂ δ̂ (T ) := Rμνρσ λμ α̂ λν β̂ λρ γ̂ λσ δ̂ (12)
is the projection of the Riemann curvature tensor along x̄ μ on the tetrad frame of the
reference observer.
Fermi coordinates are invariantly defined and can have advantages over other
physically motivated coordinate systems such as radar coordinates [13]; therefore,
they have been applied in many different contexts. For instance, Fermi coordinates
have been employed to elucidate dynamics of astrophysical jets [14–18].
3 Relativistic Gravity Gradiometry
In Einstein’s GR, gravity gradiometry involves the measurement of the gravitational

field, which is represented by the Riemannian curvature of spacetime. When an
observer measures a gravitational field, the curvature tensor must be projected onto
the tetrad frame of the observer.
It is now straightforward to express the equation of motion of any other test mass
in the Fermi coordinate system and study the motion of the test mass relative to the
fiducial test mass that follows world line x̄ μ . This general framework is necessary
in practice, since the motion of the reference test mass may involve translational
and rotational accelerations of nongravitational origin. These are absent, however,
in the ideal case of purely tidal relative motion. To illustrate this ideal situation, let
us assume that φα̂β̂ = 0, so that the reference path x̄ μ is a timelike geodesic and the
orthonormal tetrad frame is parallel transported along the fiducial geodesic world
line, i.e. Dλμ α̂ /dτ = 0. The geodesic equation of motion of a free test particle in the
corresponding Fermi coordinates relative to the reference test mass that is fixed at the
spatial origin of Fermi coordinates can be expressed in terms of relative separation
X as
d 2 X î
+ R0̂î 0̂ ĵ X ĵ + 2 Rî k̂ ĵ 0̂ V k̂ X ĵ
dT 2
2 ˆ ˆ

+ 3R0̂k̂ ĵ 0̂ V î V k̂ + Rî k̂ ĵ lˆV k̂ V l + R0̂k̂ ĵ lˆV î V k̂ V l X ĵ + O(|X|2 ) = 0 .
3
(13)
This geodesic deviation equation is a generalized Jacobi equation [10] in which the
rate of geodesic separation (i.e. the relative velocity of the test particle) V = dX/dT
is in general arbitrary; however, |V| < 1 at X = 0. It is clear from Eq. (13) that all
of the curvature components in Eq. (12) can be measured from a careful study of
the motion of the test masses in the congruence relative to the fiducial observer.
Neglecting terms in the relative velocity V, Eq. (13) reduces to the Jacobi equation,
d 2 X î
+ Kî ĵ X ĵ + O(|X|2 ) = 0 , (14)
dT 2
which is the relativistic analog of the Newtonian tidal equation given by Eq. (1), and
Kî ĵ = R0̂î 0̂ ĵ . (15)
This symmetric relativistic tidal matrix is traceless in Ricci-flat regions of spacetime

and reduces in the nonrelativistic limit to the Newtonian tidal matrix (2).
The relativistic tidal matrix is thus determined by the projection of the Riemann
curvature tensor upon the parallel-transported tetrad frame of the fiducial geodesic
observer. The local spatial frame of the fiducial observer is unique up to a constant
spatial rotation corresponding to the choice of the initial orthonormal triad in Eq. (3).
The freedom in the choice of the initial local triad implies that the form of the tidal
matrix is unique up to a constant spatial rotation.
The Jacobi equation can be used to study the influence of a gravitational field on
the relative motion of nearby test masses in general relativity [19–22]. Einstein’s field
equations locally relate the energy-momentum tensor of matter to the Ricci tensor.
148 B. Mashhoon
At any event in spacetime, the Riemann curvature tensor can be decomposed into a
matter part and a part that is independent of matter; that is,
1
Rμνρσ = Cμνρσ + gμ[ρ Rσ]ν − gν[ρ Rσ]μ − (gμρ gνσ − gμσ gνρ ) R , (16)
6
where Cμνρσ is the traceless Weyl curvature tensor that represents the “free” grav-
itational field. At any point on the manifold, the Riemann tensor has in general 20
independent components, whereas the Ricci tensor has 10 independent components.
Beyond any point on the spacetime manifold, the two parts of the curvature tensor are
connected to each other via the Bianchi identity Rμν[ρσ;δ] = 0. Introducing decom-
position (16) into the Jacobi equation and employing a canonical null tetrad frame,
Szekeres has shown via the Petrov classification that the behavior of the free part of
the gravitational field can be described in terms of the superposition of a transverse
wave component, a longitudinal component and a Coulomb component [20]. The
matter part has been treated in [22]. Some of the basic astrophysical applications of
Eq. (14) have been studied in [9, 23, 24].
The Gravity Probe B (“GP-B”) experiment has recently measured the exterior
gravitomagnetic field of the Earth [25]. The gravitomagnetic field of a rotating mass
contributes to the spacetime curvature and can thus influence the relative tidal motion
of nearby test masses. In 1980, Braginsky and Polnarev [26] proposed an experiment
to measure such an effect in a space platform in orbit around the Earth, since they
claimed that such an approach could circumvent many of the difficulties associated
with the GP-B experiment. However, in 1982, Mashhoon and Theiss [27, 28] showed
that to measure the relativistic rotation-dependent tidal acceleration in a space plat-
form, the local gyroscopes that would fix the local spatial frame carried by the space
platform must satisfy the same performance criteria as in the GP-B experiment.
The achievements of the GP-B could possibly be integrated with Paik’s super-
conducting gravity gradiometer [29] in future space experiments in order to measure
the tidal influence of the gravitomagnetic field using an orbiting platform [30, 31].
We will consider the prediction of GR for the nature of the tidal matrix in such
experiments in Sect. 5.
4 Special Tidal Directions
Let us return to the main focus of relativistic gravity gradiometry, namely, the deter-
mination of the Riemann curvature tensor projected on the tetrad frame of the fiducial
observer as in Eq. (12). Taking advantage of the symmetries of the Riemann tensor,
this quantity can be represented by a 6 × 6 matrix R = (R I J ), where the indices I
and J range over the set (01, 02, 03, 23, 31, 12). Thus we can write

E B
R= , (17)
B† S
where E and S are symmetric 3 × 3 matrices and B is traceless. The tidal matrix E
represents the “electric” components of the curvature tensor as measured by the fidu-
cial observer, whereas B and S represent its “magnetic” and “spatial” components,
respectively. Imagine next an observer that is boosted with speed β in a given direc-
tion with respect to the fiducial observer at the same event in spacetime. Let R be
the Riemann curvature tensor as measured by the boosted observer. It turns out that
under the boost the elements of E, B and S in the direction parallel to the direction of
the boost are not affected, whereas those perpendicular to the direction of the boost
are enhanced by γ 2 , where γ = (1 − β 2 )−1/2 is the Lorentz factor; moreover, the
mixed elements are enhanced by a factor of γ. This circumstance is reminiscent of
the behavior of the electromagnetic field under a boost: The components of the elec-
tric field (E) and magnetic field (B) parallel to the direction of the boost remain the
same as before, while those perpendicular to the direction of the boost are enhanced
by a factor of γ.
In this way the strength of the gravitational field can be augmented by a factor of
γ 2 ; alternatively, one can say that the radius of curvature of spacetime measured by the
boosted observer is Lorentz contracted [32, 33]. In Ricci-flat regions of spacetime,
Eq. (17) simplifies, since S = −E, E is traceless and B is symmetric. Hence, the
Weyl curvature tensor with 10 independent components is completely determined
by its “electric” and “magnetic” components that are symmetric and traceless 3 × 3
matrices.
These results imply that a gravity gradiometer would in general measure extremely
strong tidal forces when it moves very fast (β → 1). However, along certain excep-
tional directions in space, such as the radial direction in the exterior Schwarzschild
spacetime, tidal forces remain finite as β → 1 [32, 33]. Along such a special tidal
direction, the corresponding world line of the boosted observer approaches a null
direction in the local null cone as β → 1. In this way, special tidal directions are
associated with certain tidally nondestructive null directions in spacetime. The sig-
nificance of these null directions can be further elucidated via the invariant Petrov
classification of gravitational fields.
The Petrov classification involves the Weyl curvature tensor and provides an
invariant characterization of a gravitational field. This can be accomplished, for
instance, in terms of the principal null directions of the Weyl tensor. A vector k,
kα k α = 0, which satisfies the condition
k[α Cμ]νρ[σ kβ] k ν k ρ = 0 (18)
is a principal null direction of the Weyl tensor. In a gravitational field, at least one
and at most four such null vectors exist at each event in spacetime [34, 35].
The basic mathematical connection between the special tidal directions and the
principal null directions of the Weyl tensor has been established by Beem and
Parker [36] and Hall and Hossack [37]. It turns out that in general a nondestruc-
tive null direction at a point p in spacetime is a principal null direction of the Weyl
tensor at p; moreover, it is a repeated principal null direction of the Weyl tensor at p
150 B. Mashhoon
if and only if it is a Ricci eigendirection at p. A vector N μ is a Ricci eigendirection

at p if
Rμν N ν = σ Nμ (19)
for a real number σ at p. This means that in a Ricci-flat spacetime, or more generally
when
Rμν = gμν (20)
for a real number , a special tidal direction at p corresponds to a repeated principal

null direction of the Weyl tensor at p. A vector N μ is a repeated principal null vector
of the Weyl tensor at p if Nα N α = 0 and
Cμνρσ N ν N σ = λ Nμ Nρ (21)
at p for a real number λ. There are at least zero and at most two such directions at
each event in Ricci-flat spacetimes.
Let us assume that the Weyl tensor vanishes at p, then a nondestructive null
direction at p exists if and only if it is a Ricci eigendirection at p. In this case, one
can have 0, 1, 2 or ∞ nondestructive null directions at p [37]. For instance, there are
no special tidal directions in any of the standard Friedmann–Lemaître–Robertson–
Walker cosmological models. However, every direction is a special tidal direction
in a spacetime of constant nonzero curvature, namely, de Sitter (or anti-de Sitter)
universe.
The behavior of the measured components of the Riemann curvature tensor under
boosts along special tidal directions can be determined based on the results given in
Ref. [33]. Let us consider, in particular, the Kerr gravitational field, which is of type
D in the Petrov classification. The Weyl tensor at each point in this spacetime has two
repeated principal null directions; therefore, there are two special tidal directions at
each event. For example, along the axis of rotation, the outgoing and ingoing radial
directions are the special tidal directions, see Ref. [14] for an extended treatment. In
general, along the special tidal directions in Kerr spacetime, the curvature remains
invariant under boosts (R = R); in fact, the “electric” and “magnetic” components
of the curvature can be made “parallel” such that the super-Poynting vector
Pî = −î ĵ k̂ (EB) ĵ k̂ (22)
vanishes. An analogous situation is encountered in the case of the electromagnetic

field in an inertial frame in Minkowski spacetime. If the electromagnetic field is not
null, so that the invariants E 2 − B 2 and E · B do not simultaneously vanish, then a
boost with velocity v along the Poynting vector, i.e.
v E×B
= 2 , (23)
1+v 2 E + B2
renders the electric and magnetic fields parallel in the boosted frame. In the new iner-
tial frame, the Poynting vector vanishes and any boost along the common direction
of the fields leaves them invariant. The analogy between the electromagnetic field
and algebraically special gravitational fields of types D and N has been treated in
Ref. [33].
5 Tidal Matrix Around a Rotating Mass
To get some idea regarding the form of the relativistic tidal matrix, it is instructive
to consider first the tidal field along stable circular orbits in the equatorial plane of
the Kerr spacetime. The exterior Kerr metric can be expressed as [38]
2 2Mr
ds 2 = −dt 2 + dr + dθ2 + (r 2 + a 2 ) sin2 θ dϕ2 + (dt − a sin2 θ dϕ)2 ,

(24)
where M is the mass of the gravitational source, a = J/M is the specific angular
momentum of the source, (t, r, θ, ϕ) are the standard Boyer–Lindquist coordinates
and
= r 2 + a 2 cos2 θ , = r 2 − 2Mr + a 2 . (25)
The Kerr metric contains dimensionless gravitoelectric and gravitomagnetic poten-

tials U = G M/(c2 r ) and V = G J/(c3 r 2 ), which correspond to the mass and angular
momentum of the source, respectively. For instance, in the case of the Earth, we have
U⊕ ≈ 6 × 10−10 and V⊕ ≈ 4 × 10−16 .
We are interested in the tidal matrix along the circular equatorial trajectory of a
fiducial test mass that follows a future-directed timelike geodesic world line about
the Kerr source. The circular orbit has a fixed radial coordinate r0 and orbital fre-
quency [38]
dϕ ω0
= , (26)
dτ (1 − 3 r0 + 2aω0 )1/2
M
where the Keplerian frequency ω0 is given by
M
ω02 = . (27)
r03
The circular geodesic orbit is such that at proper time τ = 0, the azimuthal coordinate
vanishes (i.e. ϕ = 0). Moreover, at this event, the initial directions of the orthonormal
triad λμ î , i = 1, 2, 3, point along the spherical polar coordinate directions. The spatial
triad then undergoes parallel propagation along the circular orbit. The resulting radial
and tangential components of the spatial frame, namely, λμ 1̂ and λμ 3̂ , respectively,
turn out to be periodic in τ with period 2π/ω0 . The difference between the orbital
frequency (26) and the Keplerian frequency ω0 leads to a combination of prograde
152 B. Mashhoon
geodetic and retrograde gravitomagnetic precessions of these frame components with

respect to static inertial observers at spatial infinity in the asymptotically flat Kerr
spacetime [39].
The tidal matrix is obtained as a certain symmetric and traceless projection of the
Riemann curvature tensor evaluated along the orbit. The nonzero components of the
tidal matrix consist of constant terms proportional to ω02 plus terms that are periodic
in τ with frequency 2 ω0 and can be expressed as [39]
K1̂1̂ = ω02 [1 − 3γ02 cos2 (ω0 τ )] , (28)

3
K1̂3̂ = K3̂1̂ = − ω02 γ02 sin(2 ω0 τ ) , (29)
2
K2̂2̂ = ω02 (3γ02 − 2) , (30)
K3̂3̂ = ω02 [1 − 3γ02 sin2 (ω0 τ )] , (31)
where γ0 is given by
1/2
r02 − 2Mr0 + a 2
γ0 = . (32)
r02 − 3Mr0 + 2r02 aω0
More generally, the tidal matrix for arbitrary timelike geodesics of Kerr spacetime
has been calculated by Marck [40].
Let us next consider the tidal field along a tilted spherical orbit of fixed radial
coordinate r0 about a slowly rotating spherical mass. The exterior gravitational field
is represented by the Kerr metric linearized in the angular momentum parameter a or,
equivalently, the Schwarzschild metric plus the Thirring–Lense term. The symmetric
and traceless tidal matrix can be obtained from [39]
K1̂1̂ = ω02 [1 − 3 2 cos2 (ω0 τ )] ,

K1̂2̂ = K2̂1̂ = ω02 cos(ω0 τ ) ,
3
K1̂3̂ = K3̂1̂ = − ω02 2 sin(2 ω0 τ ) ,
2
K2̂2̂ = ω02 (3 2 − 2) ,
K2̂3̂ = K3̂2̂ = ω02 sin(ω0 τ ) ,
K3̂3̂ = ω02 [1 − 3 2 sin2 (ω0 τ )] , (33)
where and are given by

1/2
1 − 2 rM0 a ω0 cos α
:= 1− (34)
1 − 3 rM0 1 − 3 rM0
and 1/2
J 1 − 2 rM0 (1 + 2 rM0 )
:= −3 sin α sin η . (35)
M r02 ω0 1 − 3 rM0
Here, the angle α denotes the inclination of the orbit with respect to the equatorial
plane and η,
ω0
η := ωτ + η0 , ω= , (36)
(1 − 3 rM0 )1/2
is the angular position of the reference test mass in the orbital plane measured from
the line of the ascending node and η0 is a constant angle. For α = 0, the spherical
orbit under consideration turns into the circular equatorial orbit, = 0, reduces
at the linear order in a to γ0 and the tidal matrix agrees to first order in a with our
previous results for the equatorial circular orbit in Kerr spacetime.
6 Beat Effect
The off-diagonal terms K1̂2̂ = K2̂1̂ and K2̂3̂ = K3̂2̂ in the tidal matrix (33) represent
the beat phenomenon first pointed out in Ref. [27]. The beat effect involves frequen-
cies ω and ω0 with a beat frequency ω F := ω − ω0 . This is the frequency of the
gravitoelectric (geodetic) Fokker precession of an ideal test gyro following a circular
orbit about a spherical mass M. The tidal terms under consideration here that involve
have dominant amplitudes that are proportional to the angular momentum J and
are independent of the speed of light c.
In the work of Mashhoon and Theiss [27, 41–44], the resonance effect involving
ω and ω0 appeared in the calculation of the parallel-transported frame along the
tilted spherical orbit about a rotating mass. It resulted in a small divisor phenomenon
involving ω F . For a near-Earth orbit, the Fokker period 2π/ω F is about 105 years;
therefore, in practice the Mashhoon-Theiss effect shows up as a secular term in the
corresponding off-diagonal elements of the tidal matrix with amplitude
9 GJ 2
ω τ sin α , (37)
2 c2 r03 0
which is consistent with the first post-Newtonian gravitomagnetic precession of the

spatial frame [44, 45]. The possibility of measuring the Mashhoon-Theiss effect via
neutron interferometry [46] has been discussed by Anandan [47].
In the first post-Newtonian approximation, the motion of an ideal test gyro of spin
S in orbit about a central rotating mass can be written as [25]
dS
= (ω ge + ω gm ) × S , (38)
dτ
154 B. Mashhoon
where
3 GM G
ω ge = , ω gm = [3 (J · x) x − J r 2 ] , (39)
2 c2 r 3 c2 r 5
|x| = r and = x × v is the specific angular momentum of the gyro orbit. Here, ω ge
is the (gravitoelectric) geodetic precession frequency of the gyroscope, while ω gm is
its gravitomagnetic precession frequency. For the Earth, these precession frequencies
have been directly measured via GP-B [25], which involved four superconducting
gyroscopes and a telescope that were launched on 20 April 2004 into a polar Earth
orbit of radius 642 km aboard a drag-free satellite.
During a satellite gradiometry experiment over a period of time τ , we expect that
the spatial frame of the gradiometer would accumulate geodetic and gravitomagnetic
precession angles that are of order G M ω0 τ /(c2 r0 ) and G J τ /(c2 r03 ), respectively.
From the comparison of these angles with Eq. (37), it is clear that only the post-
Newtonian gravitomagnetic secular term survives in the calculation of the projection
of the Riemann tensor onto the tetrad frame of the gradiometer for the case of the
tilted spherical orbit. A recent detailed discussion of the beat effect is contained in
Ref. [39], which should be consulted for a more complete treatment of relativistic
gravity gradiometry in Kerr spacetime.
The results presented in the last two sections may be considered surprising and
contrary to expectations. That is, it may appear on the basis of Eq. (15) that the main
post-Newtonian terms in (Kî ĵ ) can be obtained intuitively by combining Newtonian
tides with the post-Newtonian motion of the spatial frame of the fiducial observer.
However, in practice the projection of the Riemann tensor onto the frame of the fidu-
cial observer involves detailed calculations in which the symmetries of the Riemann
tensor need to be carefully taken into account.
Finally, the results presented here can be used to find the main relativistic effects
in the motion of the Moon. Consider the nearly circular orbit of the Earth-Moon
system about the Sun. In the Fermi normal coordinate system established along this
orbit, the solar tidal acceleration −Kî ĵ X ĵ is a small perturbation on the dynamics
of the Earth-Moon system. Here Kî ĵ is essentially given by Eq. (33) and we recall
that the ecliptic has a small inclination of α ≈ 0.1 with respect to the equatorial
plane of the Sun. In this way, the main relativistic tidal effects in the motion of the
Moon relative to the Earth caused by the gravitational field of the Sun have been
determined [48–50].
7 Post-Schwarzschild Approximation
The exterior vacuum field of a spherically symmetry mass can be uniquely described
by the Schwarzschild spacetime. Small deviations from spherical symmetry can
then be treated in the post-Schwarzschild approximation scheme. This method was
employed by Mashhoon and Theiss in their investigation of the relativistic tidal
matrix for a gradiometer in orbit about a rotating mass [27, 41–44]. Thus in the
first post-Schwarzschild approximation, the angular momentum of the central body

is considered to first order while its mass is taken into account to all orders. The
post-Schwarzschild method can be extended to include the quadrupole and higher
moments of the central body. Indeed, the effect of oblateness, treated as a first-order
static deformation of the source, has been investigated by Dietmar Theiss for a gravity
gradiometer on a circular geodesic orbit of small inclination about a central oblate
body [51].
8 Discussion
Gravity gradiometry in GR involves the measurement of a certain projection of the

Riemannian curvature tensor of spacetime upon the orthonormal tetrad frame of an
observer. In a satellite gravity gradiometry experiment in Earth orbit, the mass M⊕ ,
angular momentum J⊕ , quadrupole moment Q ⊕ and higher moments of the Earth
will all contribute to the result of the experiment. For an inclined spherical geodesic
orbit about a slowly rotating mass, Eq. (33) gives the relativistic tidal matrix to all
orders in the mass of the source M and to linear order in its angular momentum J . The
result contains the beat phenomenon first pointed out by Mashhoon and Theiss [27,
41–44]. The corresponding influence of the quadrupole moment of the source has
been studied by Theiss [51].
References
1. H.J. Paik, Superconducting tensor gravity gradiometer for satellite geodesy and inertial navi-
gation. J. Astronaut. Sci. 29, 1 (1981)
2. H.A. Chan, M.V. Moody, H.J. Paik, Null test of the gravitational inverse square law. Phys. Rev.
Lett. 49, 1745 (1982)
3. H.A. Chan, M.V. Moody, H.J. Paik, Superconducting gravity gradiometer for sensitive gravity
measurements. II. Experiment. Phys. Rev. D 35, 3572 (1987)
4. M.J. Snadden, J.M. McGuirk, P. Bouyer, K.G. Haritos, M.A. Kasevich, Measurement of the
Earth’s gravity gradient with an atom interferometer-based gravity gradiometer. Phys. Rev.
Lett. 81, 971 (1998)
5. J.M. McGuirk, G.T. Foster, J.B. Fixler, M.J. Snadden, M.A. Kasevich, Sensitive absolute-
gravity gradiometry using atom interferometry. Phys. Rev. A 65, 033608 (2002)
6. T. Levi-Civita, Sur l’écart géodésique. Math. Ann. 97, 291 (1926)
7. J.L. Synge, On the geometry of dynamics. Phil. Trans. R. Soc. Lond. A 226, 31–106 (1926)
9. B. Mashhoon, Tidal radiation. Astrophys. J. 216, 591–609 (1977)
10. C. Chicone, B. Mashhoon, The generalized Jacobi equation. Class. Quantum Gravity 19, 4231–
4248 (2002)
11. C. Chicone, B. Mashhoon, Ultrarelativistic motion: Inertial and tidal effects in Fermi coordi-
nates. Class. Quantum Gravity 22, 195–205 (2005)
12. C. Chicone, B. Mashhoon, Explicit Fermi coordinates and tidal dynamics in de Sitter and Gödel
spacetimes. Phys. Rev. D 74, 064019 (2006)
156 B. Mashhoon
13. D. Bini, L. Lusanna, B. Mashhoon, Limitations of radar coordinates. Int. J. Mod. Phys. D 14,
1413–1429 (2005)
14. B. Mashhoon, J.C. McClune, Relativistic tidal impulse. Mon. Not. R. Astron. Soc. 262, 881–
888 (1993)
15. C. Chicone, B. Mashhoon, B. Punsly, Relativistic motion of spinning particles in a gravitational
field. Phys. Lett. A 343, 1–7 (2005)
16. C. Chicone, B. Mashhoon, Tidal acceleration of ultrarelativistic particles. Astron. Astrophys.
437, L39–L42 (2005)
17. C. Chicone, B. Mashhoon, B. Punsly, Dynamics of relativistic flows. Int. J. Mod. Phys. D 13,
945–959 (2004)
18. D. Bini, C. Chicone, B. Mashhoon, Relativistic tidal acceleration of astrophysical jets. Phys.
Rev. D 95, 104029 (2017)
19. F.A.E. Pirani, On the physical significance of the Riemann tensor. Acta Phys. Pol. 15, 389–405
(1956)
20. P. Szekeres, The gravitational compass. J. Math. Phys. (N.Y.) 6, 1387 (1965)
21. J. Podolský, R. Švarc, Interpreting spacetimes of any dimension using geodesic deviation. Phys.
Rev. D 85, 044057 (2012)
22. R.F. Crade, G.S. Hall, The deviation of timelike geodesics in space-time. Phys. Lett. A 85,
313–315 (1981)
23. B. Mashhoon, Tidal gravitational radiation. Astrophys. J. 185, 83–86 (1973)
24. B. Mashhoon, On tidal phenomena in a strong gravitational field. Astrophys. J. 197, 705–716
(1975)
25. C.W.F. Everitt et al., Gravity probe B: final results of a space experiment to test general relativity.
Phys. Rev. Lett. 106, 221101 (2011)
26. V.B. Braginsky, A.G. Polnarev, Relativistic spin-quadrupole gravitational effect. JETP Lett.
31, 415–418 (1980)
27. B. Mashhoon, D.S. Theiss, Relativistic tidal forces and the possibility of measuring them. Phys.
Rev. Lett. 49, 1542 (1982)
28. B. Mashhoon, D.S. Theiss, Erratum: relativistic tidal forces and the possibility of measuring
them. Phys. Rev. Lett. 49, 1960 (1982)
29. H.J. Paik, Detection of the gravitomagnetic field using an orbiting superconducting gravity
gradiometer: principles and experimental considerations. Gen. Relativ. Gravit. 40, 907–919
(2008)
30. H.J. Paik, B. Mashhoon, C.M. Will, Detection of the gravitomagnetic field using an orbit-
ing superconducting gravity gradiometer, in Experimental Gravitational Physics, ed. by P.F.
Michelson, Hu. En-ke, G. Pizzella (World Scientific, Singapore, 1988), pp. 229–244
31. B. Mashhoon, H.J. Paik, C.M. Will, Detection of the gravitomagnetic field using an orbiting
superconducting gravity gradiometer. Theoretical principles. Phys. Rev. D 39, 2825 (1989)
32. B. Mashhoon, Wave propagation in a gravitational field. Phys. Lett. A 122, 299–304 (1987)
33. B. Mashhoon, On the strength of a gravitational field. Phys. Lett. A 163, 7–14 (1992)
34. H. Stephani, D. Kramer, M. MacCallum, C. Hoenselaers, E. Herlt, Exact Solutions of Einstein’s
Field Equations, 2nd edn. (Cambridge University Press, Cambridge, 2003)
35. J.B. Griffiths, J. Podolský, Exact Space-Times in Einstein’s General Relativity (Cambridge
University Press, Cambridge, 2009)
36. J.K. Beem, P.E. Parker, Sectional curvature and tidal accelerations. J. Math. Phys. (N.Y.) 31,
819–827 (1990)
37. G.S. Hall, A.D. Hossack, Some remarks on sectional curvature and tidal accelerations. J. Math.
Phys. (N.Y.) 34, 5897–5899 (1993)
38. S. Chandrasekhar, The Mathematical Theory of Black Holes (Clarendon, Oxford, 1983)
39. D. Bini, B. Mashhoon, Relativistic gravity gradiometry. Phys. Rev. D 94, 124009 (2016)
40. J.-A. Marck, Solution to the equations of parallel transport in Kerr geometry; tidal tensor. Proc.
R. Soc. Lond. A 385, 431 (1983)
41. B. Mashhoon, On a new gravitational effect of a rotating mass. Gen. Relativ. Gravit. 16, 311
(1984)
42. D.S. Theiss, Ph.D. thesis, University of Cologne, Köln, Germany (1984)
43. D.S. Theiss, A general relativistic effect of a rotating spherical mass and the possibility of
measuring it in a space experiment. Phys. Lett. A 109, 19–22 (1985)
44. B. Mashhoon, Gravitational effects of rotating masses. Found. Phys. (Bergmann Festschrift),
15, 497 (1985)
45. C.A. Blockley, G.E. Stedman, Gravitomagnetic effects along polar geodesics about a slowly
rotating spherical mass in the PPN formalism. Phys. Lett. A 147, 161–164 (1990)
46. H. Rauch, S.A. Werner, Neutron Interferometry, 2nd edn. (Oxford University Press, Oxford,
2015)
47. J. Anandan, Curvature effects in interferometry. Phys. Rev. D 30, 1615–1624 (1984)
48. B. Mashhoon, D.S. Theiss, Gravitational influence of the rotation of the sun on the earth-moon
system. Phys. Lett. A 115, 333–337 (1986)
49. B. Mashhoon, D.S. Theiss, Relativistic lunar theory. Nuovo Cim. B 106, 545–571 (1991)
50. B. Mashhoon, D.S. Theiss, Relativistic effects in the motion of the Moon. Lect. Notes Phys.
562, 310–316 (2001)
51. D.S. Theiss, A new gravitational effect of a deformed mass. Phys. Lett. A 109, 23–27 (1985)
Reference-Ellipsoid and Normal Gravity
Field in Post-Newtonian Geodesy
Sergei Kopeikin
Abstract Modern geodesy is undergoing a crucial transformation from the New-

tonian paradigm to the Einstein theory of general relativity. This is motivated by
advances in developing quantum geodetic sensors including gravimeters and gradi-
entometers, atomic clocks and fiber optics for making ultra-precise measurements of
geoid and multipolar structure of Earth’s gravitational field. At the same time, Very
Long Baseline Interferometry, Satellite Laser Ranging and Global Navigation Satel-
lite System have achieved an unprecedented level of accuracy in measuring spatial
coordinates of reference points of the International Terrestrial Reference Frame and
the world height system. The main geodetic reference standard to which gravimetric
measurements of Earth’s gravitational field are referred, is called normal gravity field
which is represented in the Newtonian gravity by the field of a uniformly rotating,
homogeneous Maclaurin ellipsoid having mass and quadrupole momentum equal to
the total mass and (tide-free) quadrupole moment of the gravitational field of Earth.
The present chapter extends the concept of the normal gravity field from the New-
tonian theory to the realm of general relativity. We focus on the calculation of the
post-Newtonian approximation of the normal field that would be sufficiently precise
for near-future practical applications. We show that in general relativity the level
surface of homogeneous and uniformly rotating fluid is no longer described by the
Maclaurin ellipsoid in the most general case but represents an axisymmetric spheroid
of the fourth order (PN spheroid) with respect to the geodetic Cartesian coordinates.
At the same time, admitting post-Newtonian inhomogeneity of mass density in the
form of concentric elliptical shells allows us to preserve the level surface of the fluid
as an exact ellipsoid of rotation. We parametrize the mass density distribution and the
level equipotential surface with two parameters which are intrinsically connected to
the existence of the residual gauge freedom, and derive the post-Newtonian normal
gravity field of the rotating spheroid both inside and outside of the rotating fluid body.
The normal gravity field is given, similarly to the Newtonian gravity, in a closed form
by a finite number of the ellipsoidal harmonics. We employ transformation from the
S. Kopeikin (B)
Department of Physics and Astronomy, University of Missouri,
322 Physics Bldg, Columbia, MO 65211, USA
e-mail: kopeikins@missouri.edu
URL: https://physics.missouri.edu/people/kopeikin

160 S. Kopeikin
ellipsoidal to spherical coordinates to deduce a more conventional post-Newtonian

multipolar expansion of scalar and vector gravitational potentials of the rotating
spheroid. We compare these expansions with that of the normal gravity field gen-
erated by the Kerr metric and demonstrate that the Kerr metric has a fairly limited
application in relativistic geodesy as it does not match the normal gravity field of the
Maclaurin ellipsoid already in the Newtonian limit. We derive the post-Newtonian
generalization of the Somigliana formula for the normal gravity field measured on
the surface of the rotating PN spheroid and employed in practical work for measur-
ing the Earth gravitational field anomalies. Finally, we discuss the possible choice
of the gauge-dependent parameters of the normal gravity field model for practical
applications and compare it with the existing EGM2008 model of gravitational field.
1 Introduction
This chapter reviews the results of our previous studies of the relativistic geoid, ref-
erence ellipsoid and equipotential surface of rotating fluid body which we conducted
over decades and published in a number of articles [1–4] and in textbook [5].
1.1 Earth’s Gravity Field in the Newtonian Theory
Gravitational field of the Earth has a complicated spatial structure that is also subject
to short and long temporal variations [6–8]. Studying this structure and its time evolu-
tion is a primary goal of many scientific disciplines such as fundamental astronomy,
celestial mechanics, geodesy, gravimetry, etc. The principal component of the Earth’s
gravity field is well approximated by radially-isotropic field that can be thought as
being generated by either a point-like massive particle located at the geocenter or
by a massive sphere (or a shell) having a spherically-symmetric distribution of mass
inside it. According to the Newtonian gravity law the spheres (shells) of different size
and/or of different spherically-symmetric stratifications of the mass density generate
the same radially-isotropic gravitational field under condition that the masses of the
spheres (shells) are equal. The same statement holds in general relativity where it is
known under the name of Birkhoff’s theorem [9]. The radially-isotropic component
of the Earth’s gravity field is often called a monopole as it is characterized by a single
parameter - the Earth’s mass, M. Generally speaking, the total mass M of the Earth
is not constant because of the loss of hydrogen and helium from atmosphere, gradual
cooling of the Earth’s core and mantle, energy loss due to tidal friction, the dust accre-
tion from an outer space, etc. Nonetheless, the temporal change of the Earth’s overall
mass is minuscule, Ṁ/M ≤ 10−15 [https://en.wikipedia.org/wiki/Earth_mass], and
can be neglected in most cases. Thus, in the present chapter we consider the Earth’s
mass, M, as constant. The time variability of mass does not affect the radial isotropy
of the monopole field. It only changes its strength.
Reference-Ellipsoid and Normal Gravity Field in Post-Newtonian Geodesy 161
Monopole approximation is a good textbook example for discussion in under-

graduate physics courses but it is insufficient for real scientific applications because
Earths figure is not spherically-symmetric causing noticeable deviations from the
radial isotropy of the gravity field. These deviations are taken into account by apply-
ing the next approximation in the multipolar expansion of the gravity field called
ellipsoidal [6, 10]. This is achieved by modelling the distribution of Earth’s matter
as a rotating bi-axial ellipsoid with its center of mass located at the geocenter and
the minor axis coinciding with the Earth’s polar principal axis of inertia. Moreover,
gravitational potential on the surface of the rotating ellipsoid is equated to the value
of the gravitational potential, W0 , measured on geoid that is the equipotential surface
which coincides with the undisturbed level of the world ocean [6, 8]. It is the ellip-
soidal approximation which is called the normal gravity field and the corresponding
ellipsoid of revolution is known as a (global) reference ellipsoid [8, chapter 4.2.1].
In the Newtonian theory of gravity the normal gravity field is uniquely specified
in the exterior domain to ellipsoid by four parameters: the geocentric gravitational
constant, G M, the semi-major axis of the reference ellipsoid, a, its flattening, f ,
and the nominal value of the Earth’s rotational velocity, ω which are considered
as fundamental geodetic constants [11, Table 1.1]. The normal field is used as a
reference in description of the actual gravity field potential, W , of the Earth which
can be represented as a linear superposition of the normal gravity field potential, U
and a disturbing potential T , that is [8, 10]
W =U +T . (1)
Notice that on the surface of rotating Earth there is also a centrifugal force besides
the force of gravity. The potential of the centrifugal force is considered as a well-
defined quantity which can be easily calculated at each point of space. Therefore,
although the potential is small compared with U , it is not included to the pertur-
bation T but considered as a part of the normal gravity field potential U that consists
of the gravitational potential V of a non-rotating Earth, and the centrifugal potential
,
U = V +. (2)
The disturbing potential, T , includes all high-order harmonics in the multipolar

expansion of the gravitational field of the Earth associated with the, so-called, anoma-
lies in the distribution of the mass density. The multipolar harmonics are functions of
spatial coordinates (and time) that can be expressed in various mathematical forms.
For example, the multipolar harmonics expressed in spherical coordinates are known
as the gravity potential coefficients [6, 8] representing the, so-called, gravity distur-
bances or anomalies.
The gravity anomalies of the disturbing potential T have been consistently mea-
sured for a long time with a gradually growing accuracy. Originally, their measure-
ments were limited to rather small, local regions of the Earth’s surface and were
conducted by means of gravimeters giving access to the absolute value of the grav-
ity force and the deflection of vertical (plumb line) at the measuring point. The
162 S. Kopeikin
gravimetric ground-based measurements are indispensable for regional studies of

the gravity field anomalies but they remain sparse and insufficient to build a com-
prehensive model of the global gravity field which is the primary task of geodesy.
Advancement in constructing the global model of the gravity field was achieved with
the help of the dedicated geodetic space missions like LAGEOS and satellite laser
ranging techniques [12]. More recently, further progress have been spawned with the
advent of space gravity gradiometers – GRACE and GOCE [13]. The overall set of
measurements of the gravity field anomalies has been processed and summarized in
2008 in the form of the Earth Gravitational Model (EGM2008) that has been built
and publicly released by the National Geospatial-Intelligence Agency (NGA) EGM
Development Team in 2012 [14, 15]. The disturbing gravitational potential T of
this model is complete up to all spherical harmonics of the degree and order 2159,
and contains some additional potential coefficients up to degree 2190. Full access to
the model’s coefficients is provided on website of NGA [http://earth-info.nga.mil/
GandG/wgs84/gravitymod/egm2008/index.html].
1.2 General Relativity in Geodesy. Why do We Need It?
Positions of reference points (geodetic stations) on the Earths surface can now be
determined with precision at the level of few millimeters and their variation over
time at the level of 1 mm/year, or even better [16]. Continuous geodetic observa-
tions become more and more fundamental for many Earth-science applications at the
global and local levels like large scale and local Earth-crust deformation; global tec-
tonic motion; redistribution of geophysical fluids on or near Earths surface including
ocean, atmosphere, cryosphere, and the terrestrial hydrosphere; monitoring of the
mean sea level and its variability for evaluation its impact on global warming, and
many others [8]. All these important applications depend fundamentally on the avail-
ability and accuracy of the global International Terrestrial Reference System (ITRS).
In addition to the above-mentioned geoscience applications, the ITRS – through its
realization by an International Terrestrial Reference Frame (ITRF), is an indispens-
able reference needed to ensure the integrity of Global Navigation Satellite System
(GNSS), such as GPS, GLONASS, Galileo, and clock’s synchronization [11].
It is believed that the requirements of geoscience to measurement precision,
including the most stringent one – the mean sea level variability, are to reach the
availability of the reference frame that will be reliable, stable and accessible at the
positional accuracy down to 1 mm, and stability of 0.1 mm/year [17]. It is crucial
to reach this accuracy from both scientific and practical points of view as economy
and safety of modern society is extremely vulnerable to even small changes in sea
level [18]. Stability of the reference frame means that no discontinuity or drift should
occur in its time evolution, especially for its defining physical parameters, namely
the origin and the scale. Unfortunately, the current level of reference frame accu-
racy (based on the latest ITRF realization) is about ten times worse than the science
requirement [19]. New technological methods and theoretical models are required
to fill up this gap.
Relativistic effects of gravitational field of Earth have a fractional order of 1
ppb or about 1 cm on a geospatial scale. Albeit small, they depend globally on the
geographic position of observer and produce a systematic bias in the height mea-
surements, unless properly taken into account. It is, therefore, mandatory to switch
from the Newtonian paradigm to general theory of relativity in order to thoroughly
accommodate relativistic effects to geodesy. Nowadays, it is commonly accepted that
a network of high precision clocks and their comparison will be able to significantly
contribute to the solution of the problem of stability and accuracy of a new generation
ITRF through very precise determination of height differences and relative velocities
of clocks participating in the network [20]. In addition to the highly precise geomet-
rical coordinates of the ITRF (such as ellipsoidal heights), clock measurements will
help to consistently provide physical heights at the reference points of observatories
[21–23]. Since clocks, according to general relativity, directly measure the differ-
ence of gravitational potential this opens up a fundamentally new conceptual basis
for physical geodesy, that is, an unambiguous geoid determination and realization of
a new global dynamic reference system.
For thorough theoretical description of clock’s behaviour in gravitational field,
one has to take into account all special and general relativistic effects like gravi-
tational red shift, Doppler effect, gravitational (Shapiro) time delay, Sagnac effect,
and even Lense-Thirring effect which appears as gravitomagnetic clock effect [24,
25]. All these effects depend on clocks relative motion and strength of gravitational
field. Based on this we can solve the inverse problem to model the mass density and
height variations affecting the clock measurement, e.g., related to the solid Earth
tides. Gravitational effects associated with Earth’s rotation and tides limit the metro-
logical network of ground-based atomic clocks at fractional level 10−16 [26], and
must be accurately calculated and subtracted from clock’ readings. Further interest-
ing details in developing theoretical and technological approaches to solving this
problem can be found in the presentations of participants of ISSI-Bern workshop on
spacetime metrology, clocks and relativistic geodesy [http://www.issibern.ch/teams/
spacetimemetrology/], and in a review article [20].
This section would be incomplete without mentioning the other branches of mod-
ern geodesy which are tightly connected with the experimental gravitational physics
and fully based on the mathematical apparatus of general relativity. This includes
Very Long Baseline Interferometry (VLBI) that is used as a main tool of the Interna-
tional Earth Rotation Service (IERS) for monitoring precession, nutation and wobble
(polar motion) as well as for producing the International Celestial and Terrestrial Ref-
erence Frames (ICRF and ITRF respectively). VLBI requires taking into account a
stunning number of relativistic effects which are outlined in corresponding papers
and recorded in IAU resolutions (see, e.g. [11, 27, 28]). Motion of geodetic satel-
lites must take into account a significant number of relativistic effects as well, like
geodetic precession, Lense-Thirring effect, relativistic quadrupole, relativistic tidal
effects, etc. General-relativistic model of relativistic effects in the orbital motion of
geodetic and navigation spacecraft has been worked out by Brumberg and Kopeikin
164 S. Kopeikin
[29, 30] (c.f. [31]). It was numerically analysed in a number of recent papers [32–35]
studying a feasibility of observing various relativistic effects. A particular attention
has been recently paid to experimental measurement of the Lense-Thirring effect in
the orbital motion of LAGEOS and LARES satellites [36–39]. This experimental
study is especially important for relativistic astrophysics because the Lense-Thirring
effect is considered as a main driving mechanism for the enormous release of energy
in quasars and active galactic nuclei caused by accretion of matter on a central,
supermassive Kerr black hole [40].
One of the important relativistic problem in geodesy is the description of normal
gravity field of Earth represented as the international reference ellipsoid that is used
as a reference for geodetic and gravimetric measurements. The goal of the present
chapter is to provide the reader with a solution of this problem.
1.3 The Normal Gravity Field in Classic Geodesy

and in General Relativity
In classic geodesy the normal gravity field of the Earth is generated by a rigidly
rotating bi-axial ellipsoid which is made of a perfect (non-viscous) fluid of uniform
density, ρc which value is determined from the known total mass and volume of
the Earth. In the Newtonian theory this is the only possible distribution of mass
density because any other mass distribution of rotating fluid yields the shape of the
body being different from the bi-axial ellipsoid [10, 41]. Relativistic geodesy is an
advanced branch of physical geodesy that is based on Einstein’s general relativity
which supersedes the Newtonian theory of gravity. General relativistic approach
requires reconsidering the concept of the normal gravity field by taking into account
the curvature of spacetime manifold and other post-Newtonian effects caused by
Earth’s mass.
General theory of relativity replaces a single gravitational potential, V , with ten
potentials which are components of the metric tensor gαβ , where, here and anywhere
else, the Greek indexes α, β, γ, . . . take values from the set {0, 1, 2, 3}. General rela-
tivity modifies gravitational field equation of the Newtonian theory correspondingly.
More specifically, instead of a single Poisson equation for the scalar potential V ,
general relativity introduces ten partial differential equations of the second order for
the metric tensor components. These equations are known as Einstein’s equations [5]
1 8πG
Rαβ − gαβ R = 4 Tαβ , (3)
2 c
where Rαβ is the Ricci tensor, R = g αβ Rαβ is the Ricci scalar, Tαβ is the tensor
of energy-momentum of matter which is the source of gravitational field, G is the
universal gravitational constant, and c is the fundamental speed of the Minkowski
spacetime that is equal to the speed of light in vacuum or to the speed of propagation
of weak gravitational waves.
The left part of (3) is called the Einstein tensor which is a hyperbolic differential
operator of the second order in partial derivatives applied to the metric tensor gαβ
[42]. The Einstein tensor is non-linear and, for this reason, Einstein’s equations (3)
cannot be solved exactly in the most physical situations of practical importance. In
order to circumvent this difficulty researchers resort to iterative approximations to
solve the Einstein equations. One of the most elaborated iterative schemes is called
the post-Newtonian approximations (PNA) which basics are discussed in Sect. 2 in
more detail. In many astrophysical applications (especially in gravitational-wave
astronomy) one needs to make several post-Newtonian approximations for calcu-
lating observable effects [43]. For the purposes of relativistic geodesy and celestial
mechanics of the solar system the first post-Newtonian approximation is usually
sufficient [5] though there are indications that one may soon need a second PN
approximation [44] and exact, axially-symmetric solutions of general relativity [2,
45, 46].
The problem of determination of a figure of rotating fluid body is formidably
difficult already in the Newtonian theory [41]. It becomes even more complicated in
general relativity because of non-linearity of the Einstein equations. Geophysics is
interested in finding distribution of mass density inside the Earth to understand better
its thermal behaviour and seismological response. The interior structure of the Earth is
also important for the International Earth Rotation Service (IERS) to account properly
for free core nutation (FCN) in calculation of polar wobble and tidal variations of
the Earth’s rotational velocity affected by the elasticity of the Earth’s interior [11].
Geodesy does not require precise distribution of mass density inside the Earth as it
basically needs to know the surface of equal geopotential (geoid) and the gravity
anomalies in the domain being exterior to geoid. Geoid’s reconstruction from the
gravity anomalies utilizes the normal gravity field for solving the integral equations
of the Stokes -Molodensky problem [8]. As a rule, the most simple, homogeneous
distribution of mass density inside the Earth is used to model the normal gravity
field. Attempts to operate with more realistic distributions of mass inside the Earth
led to the models of the normal gravity field which turned out to be too complicated
for practical computations and were abandoned.
We emphasize that the internal density distribution and the surface of the rotating
fluid body taken for modelling the normal gravity field must be consistent with the
laws of the theory of gravitation. In the Newtonian theory the surface of the uniformly
rotating homogeneous fluid is a bi-axial ellipsoid of revolution - the Maclaurin ellip-
soid [41]. More realistic, non-homogeneous distribution of mass of the rotating fluid
does not allow it to be the ellipsoid of revolution yielding more complicated figure
having a spheroidal surface [8, 10]. Such models have less practical significance in
geodesy because of a more complicated structure of the normal gravity field.
One would think that modeling the normal gravity field in relativistic geodesy
could be achieved by finding an exact solution of the Einstein equations which New-
tonian limit corresponds to the homogeneous Maclaurin ellipsoid. Unfortunately,
the exact solutions of general relativity describing gravity field of a single body
consisting of homogeneous, incompressible fluid are currently known only for
spherically-symmetric, non-rotating configurations [47–49]. There is a certain
166 S. Kopeikin
progress in understanding the general relativistic structure of rotating fluid configura-

tions [50–52] but whether a rigidly rotating fluid body can be made of incompressible,
homogeneous fluid, is not yet known. In the post-Newtonian approximation of gen-
eral relativity it was found that a rigidly-rotating body consisting of a perfect fluid
with homogeneous distribution of mass density inside it, can exist but it is not a
bi-axial ellipsoid [53–56]. On the other hand, by assuming that the distribution of
mass density has a post-Newtonian ellipsoidal component in addition to the constant
density ρc , we can chose the parameters of the density distribution such that the
figure of the rigidly rotating fluid will remain exactly ellipsoidal in the first (and
higher-order) post-Newtonian approximations [3]. Thus, we have to make a decision
what type of the post-Newtonian distribution of mass and the figure of the rotating
fluid are to be used in relativistic geodesy. Depending on the choice we shall have
slightly different relativistic descriptions of the normal gravitational field outside
rotating spheroid. We will proceed by assuming that the surface of the rotating fluid
has a small post-Newtonian spheroidal deviation and that the distribution of the fluid
density is almost homogeneous with a small post-Newtonian correction taken in the
form of a homeomorphic ellipsoidal distribution.
Thus, the normal gravity field in relativistic geodesy is a solution of the
Einstein field equations with matter consisting of a uniformly rotating, perfect fluid
of nearly constant density occupying a spheroidal volume. We notice that a number
of researchers solved the Einstein equations to find out gravitational field of uni-
formly rotating bi-axial ellipsoid of constant density [57–59]. Their solutions are
not directly applicable in relativistic geodesy for the shape of a uniformly rotating
and incompressible homogeneous fluid is not an ellipsoid. Moreover, the authors of
the papers [57–59] were mostly interested in astrophysical applications and never
approached the problem from the geodetic point of view.
The chapter is organized as follows. Section 2 explains the post-Newtonian
approximations in general relativity. We pay a special attention to various coor-
dinate systems used in relativistic geodesy and transformations between them as
well as to the Green functions used for solving the Einstein equations. Section 3
discusses the model of matter distribution used as a source of normal gravitational
field and geometric shape of the post-Newtonian reference spheroid used for inte-
gration of the Newtonian and post-Newtonian gravitational potentials. Section 4 is
devoted to a comprehensive calculation of the Newtonian gravitational potential.
Section 5 presents details of the calculation of the post-Newtonian scalar and vec-
tor gravitational potentials. In Sect. 6 we discuss relativistic multipole expansion of
gravitational field of rotating spheroid which includes both mass and spin multipole
moments. Section 7 gives a full description of the normal gravity field of rotating
spheroid in relativistic geodesy including definitions of equipotential surfaces, the
gravity field potential, the figure of equilibrium of the rigidly rotating fluid, and
the Somigliana formula for the normal gravity force on the surface of the rotating
spheroid. Section 8 calculates the normal gravity field of the Kerr metric and com-
pares it with that of a rigidly rotating spheroid made out of the ideal fluid. It proves
that the Kerr metric is unsuitable for purposes of relativistic geodesy due to the
peculiar structure of its multipole expansion.
1.4 Mathematical Symbols and Notations
The following notations are used throughout the present chapter:

• the spherical coordinates are denoted {R, , },
• the ellipsoidal coordinates are denoted {σ, θ, φ},
• the Greek indices α, β, . . . run from 0 to 3,
• the Roman indices i, j, . . . run from 1 to 3,
• repeated Greek indices mean Einstein’s summation from 0 to 3,
• repeated Roman indices mean Einstein’s summation from 1 to 3,
• the unit matrix (also known as the Kronecker symbol) is denoted by δi j = δ i j ,
• the fully antisymmetric symbol Levi-Civita is denoted as εi jk = εi jk with ε123 =
+1,
• the bold letters a = a i , b = bi , etc., denote spatial 3-dimensional vectors,
• a dot between two spatial vectors, for example a · b = a 1 b1 + a 2 b2 + a 3 b3 =
δi j a i b j , means the Euclidean dot product,
• the cross between two vectors, for example (a × b)i ≡ εi jk a j bk , means the
Euclidean cross product,
• we use a shorthand notation for partial derivatives ∂α = ∂/∂x α ,
• covariant derivative with respect to a coordinate x α is denoted as ∇α ,
• the Minkowski (flat) space-time metric ηαβ = diag(−1, +1, +1, +1),
• gαβ is the physical spacetime metric,
• the Greek indices are raised and lowered with the metric ηαβ ,
• the Roman indices are raised and lowered with the Kronecker symbol δ ij ,
• G is the universal gravitational constant,
• c is the fundamental speed of the Minkowski space,
• ω is a constant rotational velocity of rigidly rotating matter,
• ρ is a mass density distribution of matter,
• ρc is a constant central density of matter,
• a is a semi-major axis of the Maclaurin ellipsoid of revolution,
• b is a semi-minor axis of the Maclaurin ellipsoid of revolution,
• f is the geometric flattening: f ≡ (a − b)/a, √
• is the first eccentricity of the Maclaurin ellipsoid: ≡ a 2 − b2 /a√= 2 f − f 2 ,
• κ is the second eccentricity of the Maclaurin ellipsoid: κ ≡ a 2 − b2 /b =
/(1 − f ),√
• α ≡ a = a 2 − b2 ,
• r ≡ R/α is a dimensionless spherical radial coordinate,
• κ ≡ πGρc a 2 /c2 is a dimensionless parameter characterizing the strength of grav-
itational field on the surface of the field-generating body.
Other notations are explained in the text as they appear.
168 S. Kopeikin
2 Post-Newtonian Approximations
2.1 Harmonic Coordinates and the Metric Tensor
Discussion of relativistic geodesy starts from the construction of the spacetime man-
ifold for the case of a rigidly rotating fluid body having the same mass as the mass
of the Earth. We shall employ Einstein’s general relativity to build such a manifold
though some other alternative theories of gravity discussed, for example in textbook
[60], can be used as well. Einstein’s gravitational field equations (3) represent a
system of ten non-linear differential equations in partial derivatives for ten compo-
nents of the (symmetric) metric tensor, gαβ , which represents gravitational potentials
generalizing the Newtonian gravitational potential V . Because the equations are dif-
ficult to solve exactly due to their non-linearity, we resort for their solution to the
post-Newtonian approximations (PNA) [60, 61].
The PNA are the most effective in case of slowly-moving matter having a weak
gravitational field. This is exactly the situation in the solar system which makes PNA
highly appropriate for constructing relativistic theory of reference frames [27], and
for relativistic celestial mechanics, astrometry and geodesy [5, 62, 63]. The PNA are
based on the assumption that solution of the Einstein equations for the metric tensor
can be presented in the form of a Taylor expansion of the metric tensor with respect
to the inverse powers of the fundamental speed, c, that is equal to the speed of light
in vacuum and the speed of weak gravitational waves in general relativity.
Exact mathematical formulation of a set of basic axioms required for doing the
post-Newtonian expansion was given by Rendall [64]. Practically, it requires having
several small parameters characterizing the source of gravity which is often is an
isolated astronomical system comprised of extended bodies. The parameters are: εi ∼
vi /c, εe ∼ ve /c, and ηi ∼ Ui /c2 , ηe ∼ Ue /c2 , where vi is a characteristic velocity of
motion of matter inside the body, ve is a characteristic velocity of the relative motion
of the bodies with respect to each other, Ui is the internal gravitational potential of
each body, and Ue is the external gravitational potential between the bodies. If one
denotes a characteristic radius of a body as and a characteristic distance between the
bodies as R, the estimates of the internal and external gravitational potentials will be,
Ui G M/ and Ue G M/R, where M is a characteristic mass of the body. Due to
the virial theorem of the Newtonian gravity theory [53] the small parameters are not
independent. Specifically, one has εi2 ∼ ηi and ε2e ∼ ηe . Hence, parameters, εi and εe ,
characterizing the slow motion of matter, are sufficient in doing the iterative solution
of the Einstein equations by the post-Newtonian approximations. Because within the
solar system these parameters do not significantly differ from each other, we shall not
distinguish between them. Quite often we shall use notation, κ ≡ πGρc a 2 /c2 ∼ ηi ,
to mark the powers of the fundamental speed c in the post-Newtonian terms.
We assume that physical spacetime within the solar system has the metric ten-
sor denoted gαβ . This spacetime is well-approximated in case of the slow-motion
and weak-field post-Newtonian approximation, by a background manifold which is
the Minkowski spacetime having the metric tensor denoted ηαβ = diag(−1, 1, 1, 1).
Einstein’s equations admit a gauge freedom associated with the arbitrariness in choos-
ing coordinate charts covering the spacetime manifold. The gauge freedom is used
to simplify the structure of Einstein’s equations. The most convenient choice is asso-
ciated with the harmonic coordinates x α = (x 0 , x i ), where x 0 = ct, and t is the
coordinate time. The class of the harmonic coordinates is used by the International
Astronomical Union for description of the relativistic coordinates systems and for
the data reduction [11, 27] as well as in relativistic geodesy [4, 65] The harmonic
coordinates are defined by imposing the de Donder gauge condition on the metric
tensor [66, 67], √
∂α −gg αβ = 0 . (4)
Imposing the harmonic gauge greatly simplifies the Einstein equation (3) and allows
us to solve them by the post-Newtonian iterations.
Because gravitational field of the solar system is weak and motion of matter is
slow, we can solve Einstein’s equations by post-Newtonian approximations. In fact,
the first post-Newtonian approximation of general relativity is fully sufficient for the
purposes of relativistic geodesy. We focus in this chapter on calculation of the normal
gravitational field of the Earth generated by uniformly rotating ideal (perfect) fluid.
Under these assumptions the spacetime interval has the following form [5]
ds 2 = g00 (t, x)c2 dt 2 + 2g0i (t, x)cdtd x i + gi j (t, x)d x i d x j , (5)
where the post-Newtonian expressions for the metric tensor components read

2V (t, x) 2V 2 (t, x) 1
g00 (t, x) = −1 + − +O 6 , (6a)
c2 c4 c

4V i (t, x) 1
g0i (t, x) = − +O 5 , (6b)
c3 c

2V (t, x) 1
gi j (t, x) = δi j 1 + +O 4 . (6c)
c2 c
Herein, the scalar potential V = V (t, x) and a (gravitomagnetic) vector potential

V i = V i (t, x) are functions of time and spatial coordinates satisfying the Poisson
equations,

1 3p
V = −4πGρ 1 + 2 2v 2 + 2V + + , (7)
c ρ
V i = −4πGρv i , (8)
with ρ = ρ(t, x) being the mass density, p = p(t, x) and v i = v i (t, x) – pressure
and velocity of matter respectively, and = (t, x) is the specific internal energy of
matter per unit mass. We emphasize that ρ is the local mass density of baryons per unit
√
of invariant (3-dimensional) volume element dV = −gu 0 d 3 x, where u 0 = dt/dτ
170 S. Kopeikin
√
is the time component of the 4-velocity of matter’s particle, where dτ = −ds 2 /c
is the proper time of the particle.1 The local mass density, ρ, relates in the post-
√
Newtonian approximation to the invariant mass density ρ∗ = −gu 0 ρ, which post-
Newtonian expression is given by [5]

∗ ρ 1 2 1
ρ =ρ+ 2 v + 3V +O 4 . (9)
c 2 c
The internal energy, , is related to pressure, p, and the local density, ρ, through the
thermodynamic equation (the law of conservation of energy)

1
d + pd =0, (10)
ρ
and the equation of state, p = p(ρ).

We shall further assume that the background matter rotates rigidly around fixed
z axis with a constant angular velocity ω. This makes the background spacetime
stationary with the background metric being independent of time. In the stationary
spacetime, the mass density ρ∗ obeys the exact, steady-state equation of continuity

∂i ρ∗ v i = 0 . (11)
The velocity of the rigidly rotating fluid is a linear function of spatial coordinates,
v i = εi jk ω j x k , (12)
where ω i = (0, 0, ω) is a constant angular velocity. Replacing velocity v i in (11)

with (12), and differentiating yield,
v i ∂i ρ = 0 , (13)
which is equivalent to dρ/dt = 0, and means that the linear velocity v i of the fluid
is tangent to the surfaces of constant density ρ.
2.2 Ellipsoidal and Spherical Coordinates
Equipotential surfaces of gravitational field produced by a rigidly rotating fluid body

are closely approximated by biaxial ellipsoids. Therefore, it sounds reasonable to
solve Einstein’s equations in the oblate ellipsoidal coordinates. These coordinates
are well known and widely used in geodesy [6, 8, 10]. In order to introduce these
1 Theminus sign in definition of the proper time appears because ds 2 < 0 due to the choice of the
metric signature shown in (6a)–(6c).
coordinates let us consider a point P in space that is characterized by three Cartesian

(harmonic) coordinates x = (x, y, z). We choose the origin of the coordinates at the
center of mass of the rotating body with z-axis coinciding with the direction of the
vector of the angular rotation, ω = (ω i ), and x and y axes lying in the equatorial
plane.
In the Newtonian theory the surface of rotating, homogeneous fluid takes the
shape of an oblate ellipsoid of revolution
√ (Maclaurin ellipsoid) with a semi-major,
a, and a semi-minor axis, b = a 1 − 2 , where the constant parameter
√
a 2 − b2
= , (14)
a
is called the first eccentricity [8], and 0 ≤ ≤ 1. The oblate ellipsoidal coordinates
associated with the ellipsoid of revolution, are defined by a set of surfaces of confocal
ellipsoids and hyperboloids being orthogonal to each other (see [https://en.wikipedia.
org/wiki/Oblate_spheroidal_coordinates]). It means that the focal points of all the
ellipsoids and hyperboloids coincide, and the distance of the focal points from the
origin of the coordinates is given by the distance α = a.
In order to connect the Cartesian coordinates, (x, y, z), of the point P to the oblate
ellipsoidal coordinates, (σ, θ, φ), we pass through P the surface of the ellipsoid
which is confocal with the Maclaurin ellipsoid formed by the rotating homogeneous
fluid. Geodetic definition of the transformation from the Cartesian to the ellipsoidal
coordinates used in geodesy, is given, for example, in [10, eq. 1-103], and reads

x = α 1 + σ 2 sin θ cos φ , (15a)

y = α 1 + σ 2 sin θ sin φ , (15b)
z = ασ cos θ , (15c)
where the√radial coordinate σ ≥ 0, 0 ≤ θ ≤ π, 0 ≤ φ ≤ 2π, and the constant parame-

ter α ≡ a 2 − b2 = a. The interior domain, V, of the ellipsoidal coordinate system
is separated from the exterior domain, (Vext ), by the surface S of the Maclaurin ellip-
soid. The interior domain is determined by conditions 0 ≤ σ ≤ 1/κ, and the exterior
domain has σ > 1/κ respectively where the constant
√
a 2 − b2
κ≡√ = , (16)
1 − 2 b
is called the second eccentricity [8], and we notice that 0 ≤ κ ≤ ∞. In terms of the
second eccentricity, the focal parameter α = bκ.
It is worth noticing that Eq. (15) looks similar to that used for definition of the
Boyer–Lindquist coordinates which have been used in astrophysical studies of the
Kerr black hole that is an exact axisymmetric solution of vacuum Einstein’s equation
[68, chapter 17]. Nonetheless, the oblate ellipsoidal coordinates, {σ, θ, φ} that we
172 S. Kopeikin
use in this chapter don’t coincide with the Boyer–Lindquist coordinates which are
connected to the original, non-harmonic, Kerr coordinates.
The volume of integration in the ellipsoidal coordinates is

d 3 x = α3 σ 2 + cos2 θ dσd , (17)
where d = sin θdθdφ is the infinitesimal element of the solid angle in the direction
of the unit vector
x̂ = sin θ î cos φ + ĵ sin φ + k̂ cos θ , (18)
where ( î, ĵ , k̂) are the unit vectors along the axes of the Cartesian coordinates
(x, y, z) respectively. Notice that the unit vector x̂ is different from the unit vec-
tor of the external normal n̂ to the surface S that is given by
√
1 − 2 sin θ î cos φ + ĵ sin φ + k̂ cos θ

n̂ = . (19)
1 − 2 sin2 θ
We also introduce the standard spherical coordinates, {R, , } related to the

harmonic coordinates, x α = {x, y, z}, by the relations
x = R sin cos ,
y = R sin sin ,
z = R cos . (20)
In what follows, it will be more convenient to use a dimensionless radial coordinate

r by definition: r ≡ R/α so that α2 r 2 = x 2 + y 2 + z 2 . The volume of integration in
the spherical coordinates is
d 3 x = α3r 2 dr dO , (21)
where dO = sin dd is the infinitesimal element of the solid angle in the spher-
ical coordinates in the direction of the unit vector

X̂ = sin î cos + ĵ sin + k̂ cos . (22)
Comparing (15) and (20) we can find out a transformation between the oblate
elliptical coordinates, (σ, θ, φ), and the spherical coordinates, (r, , ), given by
relations,

1 + σ 2 sin θ = r sin , σ cos θ = r cos , φ=. (23)
The radial elliptical coordinate, σ and the radial spherical coordinate, r , are interre-
lated
r 2 = σ 2 + sin2 θ . (24)
Solving (23) and (24) we get a direct transformation between the elliptical and
spherical coordinates in explicit form [6, equation (20.24)]
⎛ ⎞

2
r − 1 ⎝1 + 1 + 4r 2 cos 2 ⎠,
σ=
2 (r 2 − 1)2
r cos
cos θ = ⎛ ⎞. (25)

2
r − 1 ⎝ 4r cos ⎠
2 2
1+ 1+ 2
2 (r − 1)2
The approximate form of the relations (25) for relatively large values of the radial
coordinate, r 1, reads

sin2 sin2
σr− + ... , cos θ = cos 1 + + ... . (26)
2r 2r
2.3 Green’s Function of the Poisson Equation

in the Ellipsoidal Coordinates
The Einstein equations (7), (8) represent the Poisson equations with the known right-
hand side. The most straightforward solution of these equations can be achieved with
the technique of the Green function G(x, x ) that satisfies the Poisson equation,
G(x, x ) = −4πδ (3) (x − x ) , (27)
where, = ∂x2 + ∂ y2 + ∂z2 , is the Laplace operator, and δ (3) (x − x ) is the Dirac
delta-function in the harmonic coordinates {x, y, z}. We need the Green function in
the oblate ellipsoidal coordinates, {σ, θ, φ}. In these coordinates the Laplace operator
reads

1 ∂2 ∂ ∂2
≡ 2 2 (1 + σ 2 ) 2 + 2σ + 2
α σ + cos θ 2 ∂σ ∂σ ∂θ

∂ σ + cos θ ∂ 2
2 2
+ cot θ + . (28)
∂θ (1 + σ 2 ) sin2 θ ∂φ2
After substituting this form of the operator to the left side of (27), and applying
a standard procedure of finding a Green function [69], we get the Green function,
G(x, x ), in the ellipsoidal coordinates. It is represented in the form of expansion
with respect to the ellipsoidal harmonics [70, 71]
174 S. Kopeikin
Table 1 The modified associated Legendre functions.

p0 (σ) = 1 q0 (σ) = arccotσ
p1 (σ) = σ q1 (σ) = − p1 (σ)q0 (σ) + 1
3 1
p2 (σ) = σ 2 + q2 (σ) = p2 (σ)q0 (σ) − 23 σ
2 2
σ 2
5σ 2 2
p3 (σ) = 5σ + 3 q3 (σ) = − p3 (σ)q0 (σ) + +
2 2 3
35 4 15 2 3
p4 (σ) = σ + σ + q4 (σ) = p4 (σ)q0 (σ) − 35
8 σ 3 − 55 σ
24
8 4 8 σ
p11 (σ) = 1 + σ 2 q11 (σ) = p11 (σ)q0 (σ) − √
1 + σ2
3
σ 13 + 15σ 2
p31 (σ) = 1 + σ 2 1 + 5σ 2 q31 (σ) = p31 (σ)q0 (σ) − √
2 2 1 + σ2
1
G(x, x ) =
|x − x |
⎧
∞ m= (−|m|)!
⎪ ∗
⎨ α =0 m=− (+|m|)! q|m| σ p|m| (σ) Ym ( x̂ )Ym ( x̂) : (σ ≤ σ )
1
=
⎪
⎩ ∞ m=
1
α =0
(−|m|)!
m=− (+|m|)!
∗
p|m| σ q|m| (σ) Ym ( x̂ )Ym ( x̂) : (σ ≤ σ).
(29)
Here, pm (u) and qm (u) are the modified (real-valued) associated Legendre func-
tions of a real argument u, that are related to the associated Legendre functions of
an imaginary argument, Pm (iu) and Q m (iu), by the following definition2
(−1)m
Pm (iu) = i n pm (u) , Q m (iu) = qm (u) , (30)
i +1
where i is the imaginary unit, i 2 = −1. In case, when the index m = 0 we shall use
notations, p (u) ≡ p0 (u), and, q (u) ≡ q0 (u). We shall also use special notation
for the associated Legendre functions taken on the surface of ellipsoid of rotation
having a fixed radial coordinate σ = 1/κ. More specifically, we shall simply omit
the argument of the surface functions, for example, we shall denote p ≡ p (1/κ)
and q ≡ q (1/κ). Several modified associated Legendre functions which are ubiq-
uitously used in the present chapter, are shown in Table 1.
Functions Ym ( x̂) in (29) are the standard spherical harmonics3
Ym ( x̂) ≡ Cm P|m| (cos θ)eimφ , (31)
2 We remind that the associated Legendre functions of the imaginary argument, z = x + i y, are
defined for all z except at a cut line along the real axis, −1 ≤ x ≤ 1. The associated Legendre
functions of a real argument are defined only on the cut line, −1 ≤ x ≤ 1 [69, Section 12.10].
3 Definition of the associated Legendre polynomials adopted in the present chapter follows [72,
Sec. 8.81]. It differs by a factor (−1)m from the definition of the associated Legendre polynomials
adopted in the book [10].
where Pm (cos θ) are the associated Legendre polynomials, and the normalization
coefficient
( − |m|)!
Cm ≡ (2 + 1) . (32)
( + |m|)!
∗
The spherical harmonics are complex, Ym ( x̂) = Y,−m ( x̂), and form an orthonormal
basis in the Hilbert space, that is for |m| ≤ , |m | ≤ the integral over a unit sphere
S 2,
∗
Ym ( x̂)Y m ( x̂)d = 4πδ δmm , (33)
S2
where δm = diag(1, 1, . . . , 1) is a unit matrix (the Kronecker symbol).

√ when there is no dependence on the angle φ, the index m = 0 in (31) and
In case
C0 = 2 + 1. Then, Green’s function (29) takes on a more simple form,
1
G(x, x ) =
|x − x |
⎧ ∞
⎪
⎪ 1
⎪
⎪ (2 + 1)q σ p (σ) P (cos θ )P (cos θ) : (σ ≤ σ )
⎪
⎨ α =0
⎪
=
⎪
⎪ ∞
⎪
⎪ 1
⎪
⎪ (2 + 1) p σ q (σ) P (cos θ )P (cos θ) : (σ ≤ σ) ,
⎩ α
=0
(34)
where the Legendre polynomials P (cos θ) are normalized such that
π
2
P (cos θ)Pm (cos θ) sin θdθ = δm . (35)
2 + 1
0
In what follows the following expressions are used for connecting different Leg-
endre polynomials between themselves and with the trigonometric functions,
1
cos2 θ = [1 + 2P2 (cos θ)] ,
3
2
sin2 θ = [1 − P2 (cos θ)] , (36)
3
2
sin θ P11 (cos θ) = − [1 − P2 (cos θ)] ,
3
12
sin θ P31 (cos θ) = − [P2 (cos θ) − P4 (cos θ)] , (37)
7
2
1 + σ 2 q11 (σ) = [q0 (σ) + q2 (σ)] ,
3
176 S. Kopeikin
12
1 + σ 2 q31 (σ) = [q2 (σ) + q4 (σ)] , (38)
7
in order to make transformations of integrands in the process of calculation of grav-
itational potentials.
3 Mathematical Model of Matter Distribution and

Geometry of the Post-Newtonian Reference Spheroid
3.1 Modeling Matter Distribution
Gravitational field of a stationary-rotating matter is fully described by the particular

solutions of the Einstein equations (7), (8) for the metric tensor (6) which include
solution of the Poisson-type equation for scalar potential
1
V (x) = VN (x) + 2 V pN (x) , (39)
c
ρ(x ) 3
VN (x) = G d x , (40)
|x − x |
V

ρ(x ) 2 3 p(x ) 3
V pN (x) = G 2v (x ) + 2V (x ) + (x ) + d x , (41)
|x − x | ρ(x )
V
and that for a vector (often called gravitomagnetic [73–75]) potential

ρ(x )v i (x ) 3
V i (x) = G d x , (42)
|x − x |
V
where the field point has harmonic coordinates x. In order to calculate the above
integrals we have to know the distribution of mass density ρ, pressure p, velocity v i ,
and the internal energy density of the fluid , as well as the boundary of the volume
V occupied by the fluid.
Real Earth is near equilibrium shape. Small measurable changes in shape of the
equipotential surface are from post-glacial viscous rebound, elastic adjustments to
the shifting mass from melting glaciers, plate tectonics, and other long-wavelength
geoid variations [76]. These factors are important in studying the problem of the
dynamic Earth. However, our goal in the present chapter is more pragmatic and
relates to the study of relativistic corrections in Earth’s gravity field. Therefore, we
shall neglect the dynamic changes in the distribution of masses and Earth’s shape.
According to previous studies [1, 54] the surfaces of the equal gravity potential,
density, pressure, and the internal energy coincide both in the Newtonian and the
post-Newtonian approximations so that in order to find out their shape it is enough to

find out the surfaces of the equal potential. In what follows, we shall follow the model
of the normal gravity field in classic geodesy and assume that the density of the fluid
is almost uniform with a small post-Newtonian deviation from the homogeneity.
ρ(x) = ρc [1 + κF(x)] , (43)
where ρc is a constant density at the center of the spheroid, and F(x) is a homothetic
function of ellipsoidal distribution with respect to its center,

q1 x 2 + y2 z2
F(x) = A 2 + , (44)
κ a2 b2
q1 ≡ q1 (1/κ), and the constant parameter A = A() is kept arbitrary in the course
of the calculations that follow. We shall find the equations constraining the value
of the parameter A later. Such type of the density distribution has been chosen
because it is consistent with the distribution of pressure, at least in the post-Newtonian
approximation (see below). The ratio q1 /κ 2 was introduced to (44) explicitly to
make the subsequent formulas look less cumbersome. We notice that the choice of
the distribution (44) allows us to handle calculations analytically in a closed form
without series expansion while other assumptions on the mass distribution would
lead to analytical results that are more complicated than the results given in this
chapter.
Distribution (44) in the ellipsoidal coordinates takes on the following form
F(x) = Aq1 (1 − 2 )R(σ, θ) , (45)
where the function

R(σ, θ) ≡ 1 + σ 2 sin2 θ + 1 + κ 2 σ 2 cos2 θ . (46)
It is worth noticing that the ellipsoidal distribution of density (45) means that the
surfaces of constant density are not the same as the surfaces of constant value of the
radial coordinate σ. The density ρ remains dependent on the angular coordinate θ
everywhere inside the ellipsoid except at its surface, where σ = 1/κ with the post-
Newtonian accuracy, and R(κ −1 , θ) = −2 . We also draw attention of the reader that
in the limiting case of vanishing oblateness, κ → 0, the post-Newtonian correction
to the density is not singular because lim κ −2 q1 = 1/3.
κ→0
Distribution of pressure inside the rotating homogeneous fluid is obtained by
integrating the law of the hydrostatic equilibrium. Pressure enters calculations of
the integrals characterizing the gravitational field, only at the post-Newtonian terms.
Hence, it is sufficient to know its distribution to the Newtonian approximation which
is easily obtained by solving the equation of hydrostatic equilibrium [41, 77]
178 S. Kopeikin
q1
p(x) = 2πGρ2c a 2 1 − 2 R(σ, θ) , (47)
κ2
where we have denoted q1 ≡ q1 (1/κ) once again. The internal energy is also required
for calculation of the integrals only in terms of the post-Newtonian order of magni-
tude. In this approximation the internal energy can be considered as constant,
(x) = 0 , (48)
in correspondence with the thermodynamic equation (10) solved for the constant
density, ρ = ρc . From now on we incorporate the constant thermodynamic energy to
the central density and will not show 0 explicitly in our calculations. Because the
fluid rotates uniformly in accordance with the law (12), we have for the distribution
of the velocity squared,

v 2 (x) = ω 2 α2 1 + σ 2 sin2 θ . (49)
3.2 Post-Newtonian Reference Spheroid
All integrals are calculated over a (yet unknown) volume occupied by the rotating
fluid. The surface of the rotating, self-gravitating fluid is a surface of vanishing
pressure that coincides with the surface of an equal gravitational potential [1, 53,
54]. In classical geodesy the reference figure for calculation of geoid’s undulation is
the Maclaurin ellipsoid which is a surface of the second order formed by a rigidly
rotating fluid of constant density ρ. Maclaurin’s ellipsoid is described by a polynomial
[41]
x 2 + y2 z2
+ =1, (50)
a2 b2
where a and b are semi-major and semi-minor axes of the ellipsoid. We also assume
a > b, and define the eccentricity of the Maclaurin ellipsoid as
√
a 2 − b2
e≡ . (51)
a
Physically, the ellipsoidal shape of rotating, self-gravitating fluid is formed because
the Newtonian gravity potential is a scalar function represented by a polynomial of
the second order with respect to the Cartesian spatial coordinates, and the differential
Euler equation defining the equilibrium of the gravity and pressure is of the first order
partial different equation which leads to the quadratic (w.r.t. the coordinates) equation
of the level surface [41].
We shall demonstrate in the following sections that in the post-Newtonian approx-
imation the gravity potential, W , of the rotating fluid is a polynomial of the
fourth order as was first noticed by Chandrasekhar [53]. Hence, the post-Newtonian
equation of the level surface of a rigidly-rotating fluid is expected to be a surface of

the fourth order. We shall assume that the surface remains axisymmetric in the post-
Newtonian approximation and dubbed the body with such a surface as PN-spheroid.
Let coordinates x i = {x, y, z} denote a point on the surface of the PN-spheroid
with the axis of symmetry directed along the rotational axis and with the origin
located at its post-Newtonian center of mass.4 Let the rotational axis coincide with
the direction of z axis. Then, the most general equation of the PN-spheroid is
x 2 + y2 z2
+ = 1 + κQ(x) , (52)
a2 b2
where κ ≡ πGρa 2 /c2 , and function

2 2
x 2 + y2 z2 x 2 + y2 z4 x + y2 z2
Q(x) ≡ K1 + K2 + B1 + B2 4 + B3 ,
a2 b2 a2 b a 2 b2
(53)
and K1 , K2 , B1 , B2 , B3 are arbitrary numerical coefficients.
Each cross-section of the PN-spheroid being orthogonal to the rotational axis,
represents a circle. The equatorial cross-section has an equatorial radius, σ = re ,
being determined from (52) by the condition z = 0. It yields

1
re = a 1 + κ (K1 + B1 ) . (54)
2
The meridional cross-section of the PN-spheroid is no longer an ellipse (as it was

in case of the Maclaurin ellipsoid) but a curve of the fourth order. Nonetheless, we
can define the polar radius, z = r p , of the PN-spheroid by the condition, x = y = 0.
Equation (52) yields
1
r p = b 1 + κ (K2 + B2 ) . (55)
2
The equatorial and polar radii of the PN-spheroid should be used in the post-
Newtonian approximation instead of the equatorial and polar radii of the Maclaurin
reference-ellipsoid for calculation of observable physical effects like the normal grav-
ity force. We characterize the ‘oblateness’ of the PN-spheroid by the post-Newtonian
eccentricity
re2 − r 2p
≡ . (56)
re
It differs from the eccentricity (51) of the Maclaurin ellipsoid by relativistic correction
4 Post-newtonian definitions of mass, center of mass, and other multipole moments can be found,
for example, in [5].
180 S. Kopeikin
Fig. 1 Meridional cross-section of the PN-spheroid (a red curve in on-line version) versus the
Maclaurin ellipsoid (a blue curve in on-line version). The top panel represents the most general
case with arbitrary values of the PN-spheroid shape parameters K1 , K2 , B1 , B2 when the equatorial,
re , and polar, r p , radii of the PN-spheroid differ from the semi-major, a, and semi-minor, b, axes of
the Maclaurin ellipsoid, re = a, r p = b. The bottom panel shows the most important physical case
of B1 = K1 , B2 = K2 when the equatorial and polar radii of the PN-spheroid and the Maclaurin
ellipsoid are equal. The angle ϕ is the geographic latitude (−90◦ ≤ ϕ ≤ 90◦ ), and the angle θ is
a complementary angle (co-latitude) used for calculation of integrals in appendix of the present
chapter (0 ≤ θ ≤ π). In general, when B1 = K1 , and/or B2 = K2 , the maximal radial difference
(the ’height’ difference) between the surface of the PN-spheroid and that of the Maclaurin ellipsoid
can amount to a relatively large value of several centimeters, and even more. In case of B1 = K1 ,
B2 = K2 the radial undulation between the two surfaces is defined by the parameter B3 ≡ B, and
it does not exceed one centimeter
1 − e2
=e−κ (K2 − K1 ) + (B2 − B1 ) . (57)
2e
PN-spheroid versus the Maclaurin ellipsoid is visualized in Fig. 1.
Theoretical formalism for calculating the post-Newtonian level surface can be
worked out in arbitrary coordinates. For mathematical and historic reasons the most
convenient are harmonic coordinates which are also used by the IAU [27] and IERS
[11]. The class of the harmonic coordinates is selected by the gauge condition (4).
Different harmonic coordinates are interrelated by coordinate transformations which

are not violating the gauge condition (4). This freedom is known as a residual gauge
(or coordinate) freedom. The field equations (7), (8) and their solutions are form-
invariant with respect to the residual gauge transformations.
The residual gauge freedom is described by a post-Newtonian coordinate trans-
formation,
x α = x α + κξ α (x) , (58)
where functions, ξ α , obey the Laplace equation,
ξ α = 0 . (59)
Solution of the Laplace equation which is convergent at the origin of the coordi-
nate system, is given in terms of the harmonic polynomials which are selected by
the conditions imposed by the statement of the problem. In our case, the problem
is to determine the shape of the PN-spheroid which has the surface described by
the polynomial of the fourth order (62) with yet unknown coefficients B1 , B2 , B3 .
The form of the Eq. (62) does not change (in the post-Newtonian approximation) if
the functions ξ α in (58) are polynomials of the third order. It is straightforward to
show that the admissible harmonic polynomials of the third order have the following
form
x 2
ξ 1 = hx + p 2
σ − 4z 2 , (60a)
a
y
ξ = hy + p 2 σ 2 − 4z 2 ,
2
(60b)
a
z
ξ 3 = kz + q 2 3σ 2 − 2z 2 , (60c)
b
where h, k, p and q are arbitrary constant parameters. Polynomials (60a)–(60c) rep-
resent solutions of the Laplace equation (59). We choose ξ 0 = 0 because we consider
stationary spacetime so all functions are time-independent.
Coordinate transformation (58) with ξ i taken from (60a)–(60c) does not violate
the harmonic gauge condition (4) but it changes the numerical post-Newtonian coef-
ficients in the mathematical form of Eqs. (52) and (53)
K1 → K1 + 2h , (61a)
K2 → K2 + 2k , (61b)
B1 → B1 + 2 p , (61c)
B2 → B2 − 4q , (61d)
b2 a2
B3 → B3 − 8 p 2 + 6q 2 , (61e)
a b
Thus, it makes evident that only one out of the five coefficients K1 , K2 , B1 , B2 , B3
is algebraically independent while the four others can be chosen arbitrary. To sim-
182 S. Kopeikin
plify our calculations and eliminate the gauge-dependent terms from mathematical
equations we decide to fix the numerical values of four parameters K1 , K2 , B1 , B2 .
The constant B3 is left free. It is fixed by the condition of a hydrostatic equilibrium
of the rotating fluid body (see Sect. 7.3).
One of the most simple and attractive choice of fixing the residual gauge freedom
is K1 = K2 = B1 = B2 = 0. This choice of the residual gauge has been employed
in our papers [3, 4]. It is particularly useful for conducting calculations in the Carte-
sian coordinates. With such a choice of the coordinates the polar radius r p = b, the
equatorial radius re = a, the eccentricity of the PN-spheroid = e, and function
σ2 z 2
Q(x) ≡ B3 . (62)
a 2 b2
In the ellipsoidal coordinates it is more convenient to define the surface of the
rotating fluid by the following equation,

1 ω2 a2
σs = 1 + B 2 2 P2 (cos θ) , (63)
κ c
where B = B() is a constant arbitrary parameter which possible numerical value will
be discussed below at the end of Sect. 6.1 and in Sect. 7.3. The reason for picking up
equation of the surface of PN-spheroid in the form of (63) is a matter of mathematical
convenience. It is worth making two remarks. First, the appearance of 2 , in the
denominator in the right side of (63) does not lead to divergence as the angular
velocity of rotation ω 2 ∼ 2 . Second, Eq. (63) corresponds to the following choice
of parameters in (53):
4q2 16q2 4q2

B1 = B2 = 0, K1 = − B, K2 = B, B3 = B, (64)
κ3 κ3 κ3
where parameter B is the same as in (63), q2 ≡ q2 (1/κ), and the angular velocity ω is
related to q2 by means of the Maclaurin relationship (86). Parameterization (64) does
not allow us to keep the equatorial, re , and polar, r p , axes of the PN-spheroid as well
as its eccentricity equal to the parameters of the reference Maclaurin ellipsoid (see
Eqs. (54)–(57)). However, parameterization (64) simplifies calculation of integrals
in the equatorial coordinates and will be used throughout this chapter.
With the definition (63) of the upper limit of the integration with respect to the
radial coordinate σ, the volume integral from an arbitrary function
F< (σ, θ, σ , θ ) if σ ≤ σ ,
F(x, x ) = (65)
F> (σ, θ, σ , θ ) if σ ≥ σ .
can be calculated with sufficient accuracy as a sum of two terms:

(1) in case when the radial coordinate σ of the field point x is taken inside the body
σ π
3
F(x, x )d x = 2πα 3
F< (σ, θ, σ , θ )(σ 2 + cos2 θ )dσ dθ
V 0 0
1/κπ
+2πα3 F> (σ, θ, σ , θ )(σ 2 + cos2 θ )dσ dθ
σ 0
π
ω2 a4 b
+2πB 2 F> σ, θ, κ −1 , θ κ −2 + cos2 θ P2 (cos θ )dθ , (66)
c
0
(2) in case when the radial coordinate σ of the field point x is taken outside the body
1/κπ
F(x, x )d 3 x = 2πα3 F< (σ, θ, σ , θ )(σ 2 + cos2 θ )dσ dθ
V 0 0
π
ω2 a4 b
+2πB F< σ, θ, κ −1 , θ κ −2 + cos2 θ P2 (cos θ )dθ . (67)
c2
0
The very last integral in the right hand side of (66) and (67) is of the post-Newtonian
order of magnitude and will be treated as a post-Newtonian correction to the New-
tonian gravitational potential.
4 Newtonian Gravitational Potential
The Newtonian gravitational potential VN is given by (40) where the density dis-
tribution, ρ(x), is defined in (43) and the integration is performed over the volume
bounded by the radial coordinate σs in (63). The integral can be split in three parts:
VN = VN [ρc , S] + VN [δρ, S] + VN [ρc , δS] , (68)
where VN [ρc , S] denotes contribution from the constant density ρc , VN [δρ, S] is

contribution from the variation δρ ≡ c−2 F(x) of the density given by (45), and
VN [ρc , δS] represents contribution from the fraction of the constant density ρc
enclosed in the part of the volume lying between the real boundary and that of
the Maclaurin ellipsoid. The integrals VN [ρc , S] and VN [δρ, S] are volume integrals
taken over the Maclaurin ellipsoid with the fixed value of the radial coordinate on
its boundary, σ = 1/κ. The term VN [ρc , δS] comes from the very last integrals in
184 S. Kopeikin
(66) and (67). Below we provide specific details of calculations of the three terms
entering the right hand side of (68).
4.1 Integral Contribution to the Newtonian Potential from

Constant Density
Contribution from the constant density ρc to the Newtonian potential is given by

integral
d3x
VN [ρc , S] = Gρc , (69)
|x − x |
V
which can be calculated by making use of the Green function (29). Depending on
position of the field point x in space we distinguish the internal and external solutions.
4.1.1 The Internal Solution
The internal solution is valid for the field point x with the radial coordinate, 0 ≤ σ ≤
1/κ. With the help of the Green function (34) it reads,
∞

VN [ρc , S] = 2πGρc α2 (2 + 1)q (σ) P (cos θ) ×
=0
σ π
2
dσ σ + cos2 θ p σ P (cos θ ) sin θ dθ
0 0
∞

+2πGρc α 2
(2 + 1) p (σ) P (cos θ) ×
=0
1/κ π

2
dσ σ + cos2 θ q σ P (cos θ ) sin θ dθ . (70)
σ 0
We, first, integrate with respect to the angular variable θ and, then, with respect to
the radial coordinate σ. We also use the relation
2
σ 2 + cos2 θ = [ p2 (σ) + P2 (cos θ)] , (71)
3
that allows us to operate with the Legendre polynomials instead of the trigonometric
functions, and use the condition of orthogonality (35). Then, after substituting (71)
to (70) and making use of the normalization condition (35), we obtain
VN [ρc , S] = V0 (σ) + V2 (σ)P2 (cos θ) , (72)
where
⎡ ⎤
σ 1/κ
8π
V0 (σ) = Gρc α2 ⎣q0 (σ) dσ p2 (σ ) + dσ p2 (σ )q0 (σ )⎦ , (73)
3
0 σ
⎡ ⎤
σ 1/κ
8π
V2 (σ) = Gρc α2 ⎣q2 (σ) dσ p2 (σ ) + p2 (σ) dσ q2 (σ )⎦ . (74)
3
0 σ
Calculation of the integrals yields
2πGρc α2
V0 (σ) = 3
κ 1 − κ 2 σ 2 + 2 1 + κ 2 arctan κ , (75)
3κ
2πGρc α2
V2 (σ) = − 3
κ + κ 3 + 2κ 2 σ 2 − 1 + 3σ 2 1 + κ 2 arctan κ . (76)
3κ
Adding up the two expressions according to (72) and making the inverse transfor-
mation from the ellipsoidal to Cartesian coordinates, we get

1 + κ2 z2
VN [ρc , S] = πGρc α 2 2
arctan κ − 2
κ3 α2
2
x + y − 2z
2 2
1+κ 2
+ 1− arctan κ . (77)
α2 κ 2 κ
This expression for the Newtonian potential VN [ρc ] inside the Maclaurin ellipsoid
is well-known from the classic theory of figures of rotating fluid bodies [41, 78]
(see also [4, eq. 50]). It is straightforward to check by direct differentiation that (77)
satisfies the Poisson equation VN [ρc ] = −4πGρc , in accordance with (69).
4.1.2 The External Solution
Making use of the Green function (34) we get for the field point x with the radial
ellipsoidal coordinate, σ > κ −1 , the following external solution for the Newtonian
potential of the homogeneous Maclaurin ellipsoid,
∞

VN [ρc , S] = 2πGρc α2 (2 + 1)q (σ) P (cos θ) ×
=0
1/κ π

2
dσ σ + cos2 θ p σ P (cos θ ) sin θ dθ . (78)
0 0
186 S. Kopeikin
We again integrate over the angular variable θ and, then, with respect to the radial
variable σ. It yields,
4πGρc α2
VN [ρc , S] = 3
1 + κ 2 [q0 (σ) + q2 (σ)P2 (cos θ)] . (79)
3κ
In the asymptotic regime at spatial infinity, when the radial coordinate σ is very large,
the Legendre functions have the following asymptotic behavior

1 1 2 1
q0 (σ) = + O 3 , q2 (σ) = 3
+O 5 , (80)
r r 15r r
so that the asymptotic expression for the Newtonian external gravitational potential
at large distances from the body, is

Gm N 1
VN [ρc , S] = +O 3 , (81)
r r
where the notation m N ≡ M N /α, and M N is the Newtonian mass of the Maclaurin
ellipsoid
4πρc α3 1 + κ 2 4πρc a 2 α 4πρc a 2 b
MN = = = . (82)
3 κ 3 3 κ 3
Therefore, expression (79) can be simplified to

VN [ρc , S] = Gm N q0 (σ) + q2 (σ)P2 (cos θ) . (83)
It is worth noticing that on the surface of the rotating body the two expressions for
the internal and external gravitational potential, (72) and (83) match smoothly for
the gravitational potential is a continuous function. It is also useful to remark that for
a fixed value of the fluid’s density, ρc , the normalized mass, m N , decreases inversely
proportional to the eccentricity: m N ∼ κ −1 ∼ −1 .
The surface of the Maclaurin ellipsoid is equipotential, and it is defined by equation
1 2
v + VN [ρc , S] = W0 = const. , (84)
2
where v 2 is defined in (49). After taking into account (36) and (83), equation of the
ellipsoid reads,
ω 2 α2
W0 = 2
1 + κ 2 [1 − P2 (cos θ)] + Gm N q0 + q2 P2 (cos θ) , (85)
3κ
where we have introduced shorthand notations q0 ≡ q0 (1/κ) and q2 ≡ q2 (1/κ).
Since the left hand side of (85) is constant, the right hand side of it must be constant
as well. It yields two relationships
4πGρc
ω2 = q2 , (86)
κ
1
W0 = ω 2 a 2 + Gm N q0 . (87)
3
Equations (86), (87) can be recast to yet another form,
3Gm N
ω2 = q2 , (88)
a2
W0 = Gm N [q0 + q2 ] . (89)
Equation (86) yields relation between the angular velocity of rotation of the homo-
geneous fluid and oblateness of the Maclaurin ellipsoid while (87) or, equivalently,
(89) defines the gravity potential on its surface. For small values of the second eccen-
tricity κ 1, when the deviation of the ellipsoid from a sphere is very small, we can
expand the Legendre function q2 in the Taylor series, which yields the asymptotic
expression for the angular velocity

8 6
ω2 = Gπρc κ 2 1 − κ 2 + O κ 6 , (κ 1) . (90)
15 7
On the other hand, when the ellipsoid has a disk-like shape, we have κ 1, and the
asymptotic expression of the angular velocity takes on another form,

Gπρc 2 1
ω =
2
π− +O , (κ 1) . (91)
κ κ κ3
Equation (86) tells us that the angular velocity of rotation of the Maclaurin ellipsoid,
ω = ω(κ), considered as a function of the eccentricity, κ, has a maximum which
is reached for κ 2.52931 [78]. The maximal value of the angular velocity of the
Maclaurin ellipsoid at this point is ω 2 0.45πGρc [77] .
4.2 Integral Contribution to the Newtonian Potential

from the Density Inhomogeneity
Contribution from the non-homogeneous part of the mass density to the Newtonian
potential is given by the integral

Gρc F(x )d 3 x
VN [δρc , S] = , (92)
c2 |x − x |
V
188 S. Kopeikin
where function F(x) is given in (43)–(45). Making use of (45) in the integral (92),
brings it to the form,
πG 2 ρ2c α2 b2 q1
VN [δρc , S] = A I1 (x) , (93)
c2
where we have introduced a notation

1 R(σ , θ )d 3 x
I1 (x) ≡ . (94)
α2 |x − x |
V
The integral (94) is calculated with making use of the Green function (34). We
consider the internal, (σ ≤ 1/κ), and external, (σ ≥ 1/κ), solutions separately.
Making use of the Green’s functions (34) we get,

∞

I1 (σ, θ) = 2π (2 + 1)q (σ) P (cos θ)
=0
σ π

2
× dσ σ + cos2 θ R(σ , θ ) p σ P (cos θ )dθ
0 0
∞

+2π (2 + 1) p (σ) P (cos θ)
=0
1/κ π

2
× dσ σ + cos2 θ R(σ , θ )q σ P (cos θ )dθ . (95)
σ 0
We, first, integrate with respect to the angular variable θ and, then, with respect to
the radial variable σ. The integrand of the above integral is
2 2
σ + cos2 θ R(σ, θ) = 5 − 3κ 2 p2 (σ) + 4 3 + κ 2 p4 (σ)
105
2
+ 5 − 3κ 2 + 8κ 2 p4 (σ) P2 (cos θ)
105
8
− 3 + κ 2 − 2κ 2 p2 (σ) P4 (cos θ) . (96)
105
After integration over the angular variable θ, the integral can be represented as a
linear combination of several terms,
I1 (σ, θ) = I10 (σ) + I12 (σ)P2 (cos θ) + I14 (σ)P4 (cos θ) , (97)
where each part corresponds to its own Legendre polynomial,

⎧
σ
8π ⎨
I10 (σ) = q0 (σ) dσ 5 − 3κ 2 p2 (σ ) + 4 3 + κ 2 p4 (σ )
105 ⎩
0
⎫
1/κ ⎬

+ dσ 5 − 3κ 2 p2 (σ ) + 4 3 + κ 2 p4 (σ ) q0 (σ ) , (98)
⎭
σ
⎧
σ
8π ⎨
I12 (σ) = q2 (σ) dσ 5 − 3κ 2 + 8κ 2 p4 (σ ) p2 (σ )
105 ⎩
0
⎫
1/κ ⎬

+ p2 (σ) dσ 5 − 3κ 2 + 8κ 2 p4 (σ ) q2 (σ ) , (99)
⎭
σ
⎧
σ
32π ⎨
I14 (σ) = − q4 (σ) dσ 3 + κ 2 − 2κ 2 p2 (σ ) p4 (σ )
105 ⎩
0
⎫
1/κ ⎬

+ p4 (σ) dσ 3 + κ 2 − 2κ 2 p2 (σ ) q4 (σ ) . (100)
⎭
σ
Calculation of the integrals reveals

π
I10 (σ) = κ 1 − κ 2 σ 2 3 + 5κ 2 + 3κ 2 σ 2 + κ 4 σ 2
15κ 5

2 2
+ 12 1 + κ arctan κ , (101)

2π
I12 (σ) = − 3 + 5κ 2 + σ 2 9 + 15κ 2 + 2κ 4
21κ 4

2 2

2 arctan κ
+ 2κ σ − 3 1 + κ
6 4
1 + 3σ , (102)
κ

π
I14 (σ) = 9 + 15κ 2 + 6σ 2 15 + 25κ 2 + 8κ 4
105κ 4

+ σ 4 105 + 175κ 2 + 56κ 4 − 8κ 6

2 2

4 arctan κ
−3 1+κ 3 + 30σ + 35σ
2
. (103)
κ
190 S. Kopeikin
The external solution is obtained by making use of the Green function (34)
∞

1/κ π
2
I1 (σ, θ) = 2π (2 + 1)q (σ)P (cos θ) dσ σ + cos2 θ
=0 0 0

×R(σ , θ ) p σ P (cos θ )dθ .

(104)
The result is
1/κ
8π
I1 (σ, θ) = q0 (σ) dσ 5 − 3κ 2 p2 (σ ) + 4 3 + κ 2 p4 (σ )
105
0
1/κ
8π
+ q2 (σ)P2 (cos θ) dσ 5 − 3κ 2 + 8κ 2 p4 (σ ) p2 (σ )
105
0
1/κ
32π
− q4 (σ)P4 (cos θ) dσ 3 + κ 2 − 2κ 2 p2 (σ ) p4 (σ ) . (105)
105
0
After performing the integrals it results in

2
4π 1 + κ 2
I1 (σ, θ) = [7q0 (σ) + 5q2 (σ)P2 (cos θ) − 2q4 (σ)P4 (cos θ)] . (106)
35 κ5
4.3 Integral Contribution to the Newtonian Potential from

the Difference Between the Volumes of PN Spheroid and
Maclaurin Ellipsoid
Contribution VN [ρc , δS] from the spheroidal deviation of the shape of the rotating
fluid from the Maclaurin ellipsoid is given by Eqs. (66), (67), where function F is
proportional to the Green function (34). More specifically,
4π
VN [ρc , δS] = BGρc ω 2 α4 I2 (σ, θ) , (107)
c2
where
⎧ 1 ∞
⎪
⎪ =0 (2 + 1)q p (σ)P (cos θ)
⎪
⎪
2
(π −2
⎪
⎪ × κ + cos2 θ P2 (cos θ )P (cos θ )dθ : (σ ≤ 1/κ)
⎪
⎪
⎨ 0
I2 (σ, θ) = (108)
⎪
⎪ 1 ∞
⎪
⎪
⎪ 2 =0 (2 + 1) p q (σ)P (cos θ)
⎪
⎪ (π −2
⎪
⎩ × κ + cos2 θ P2 (cos θ )P (cos θ )dθ : (σ ≥ 1/κ)
0
Calculation of the integral in (108) is performed with the help of (35) yielding
π
−2
κ + cos2 θ P2 (cos θ )P (cos θ )dθ
0

2 2 1 11 12
= δ0 + + δ2 + δ4 . (109)
2 + 1 15 κ2 21 35
It yields, ⎧ 2q0 1
⎪
⎪ + κ 2 + 11 q2 p2 (σ)P2 (cos θ)
⎪
⎪
15 21
⎨ + 12q
35
4
p 4 (σ)P4 (cos θ) : (σ ≤ 1/κ)
I2 (σ, θ) = (110)
⎪
⎪
⎪
⎪
2
q (σ) + κ12 + 11 p2 q2 (σ)P2 (cos θ)
⎩ 15 0 21
12 p4
+ 35 q4 (σ)P4 (cos θ) : (σ ≥ 1/κ)
where q0 ≡ q0 (1/κ), q2 ≡ q2 (1/κ), q4 ≡ q4 (1/κ), p2 ≡ p2 (1/κ), p4 ≡ p4 (1/κ).
5 Post-Newtonian Potentials
5.1 Scalar Potential V pN
The post-Newtonian correction (41) to the Newtonian gravity potential obeys the
Poisson equation
V pN (x) = −4πGρ pN (x) , (111)
where
ρ pN (x) ≡ ρc 2v 2 (x) + 2VN (x) + (x) + 3 p(x) , (112)
and the functions entering the right hand side of (112) are defined by Eqs. (47)–(49)
and (72). Fock had proved (see [66, Eq. 73.26]) that for any (including a homoge-
neous) distribution of mass density the following equality holds
192 S. Kopeikin

1 2
ρ(x) v (x) + VN (x) = p(x) + ρ(x)(x) . (113)
2
It can be used in order to re-write (112) as follows

ρ pN (x) ≡ ρc v 2 (x) + 3(x) + 5 p(x) . (114)
This allows to eliminate the Newtonian gravitational potential VN from the calcula-
tion of the post-Newtonian gravitational potential V pN by solving (111).
After making use of expressions (47)–(49) and including the constant term 3 =
30 , to the constant density (re-normalizing the central density ρc ) we get in the
elliptical coordinates,

ρ pN (x) = ρc ω 2 α2 1 + σ 2 sin2 θ − 10πGρ2c b2 q1 R(σ, θ) . (115)
Integrating (111) directly with the help of the Green function (34), yields
q1
V pN (x) = −10πG 2 ρ20 α4 I1 (σ, θ) + Gρc ω 2 α4 I3 (σ, θ) , (116)
κ2
where the integral I1 (σ, θ) has been calculated in Sect. 4.2, and I3 (σ, θ) is the integral
from function (1 + σ 2 ) sin2 θ to the post-Newtonian gravitational potential V pN . The
integral I3 (σ, θ) is performed as follows.
5.2 Integral from the Source (1 + σ 2 ) sin2 θ
The explicit form of the integral I3 (σ, θ) is as follows,

1 (1 + σ 2 ) sin2 θ d 3 x
I3 (σ, θ) = , (117)
α2 |x − x |
V
where
2
σ + cos2 θ (1 + σ 2 ) sin2 θ =
)
16
p2 (σ) + p4 (σ) + [1 − p4 (σ)] P2 (cos θ) − [1 + p2 (σ)] P4 (cos θ) .
105
(118)
Substituting (118) to (117) and integrating with respect to the angular variables we
get the internal and external solutions.
Making use of the Green’s functions (34) the internal solution of (117) takes on the
following form
∞
σ π

2
I3 (σ, θ) = 2π (2 + 1)q (σ)P (cos θ) dσ σ + cos2 θ
=0 0 0

×(1 + σ 2 ) sin2 θ p σ P (cos θ )dθ
∞
1/κ π
2
+2π (2 + 1) p (σ)P (cos θ) dσ σ + cos2 θ
=0 σ 0

×(1 + σ ) sin θ q σ P (cos θ )dθ ,
2 2
(119)
which can be represented as a linear combination of the Legendre polynomials,
I3 (σ, θ) = I20 (σ) + I22 (σ)P2 (cos θ) + I24 (σ)P4 (cos θ) , (120)
where the coefficients are functions of the radial coordinate σ,
σ
64π
I20 (σ) = q0 (σ) dσ p2 (σ ) + p4 (σ )
105
0
1/κ )

+ dσ p2 (σ ) + p4 (σ ) q0 (σ ) , (121a)
σ
σ
64π
I22 (σ) = q2 (σ) dσ 1 − p4 (σ ) p2 (σ )
105
0
1/κ )

+ p2 (σ) dσ 1 − p4 (σ ) q2 (σ ) , (121b)
σ
σ
64π
I24 (σ) = − q4 (σ) dσ 1 + p2 (σ ) p4 (σ )
105
0
1/κ )

+ p4 (σ) dσ 1 + p2 (σ ) q4 (σ ) . (121c)
σ
Calculation of the integrals in (121) yields

194 S. Kopeikin

2π
2 2 arctan κ
I20 (σ) = 1 + 2κ 2
− 2κ σ
4 2
− κ σ
4 4
+ 4 1 + κ , (122)
15κ 4 κ

2π
I22 (σ) = 6
3 + 4κ 2 − κ 4 + 9 + 12κ 2 − 3κ 4 − 2κ 6 σ 2 + 2κ 6 σ 4
21κ

2 arctan κ
− 1 + κ 2 3 − κ 2 1 + 3σ 2 , (123)
κ

π
I24 (σ) = 6
15 + 34κ 2 + 23κ 4 + 150 + 340κ 2 + 230κ 4 + 32κ 6 σ 2
70κ

1190 2 805 4 128 6 4
+ 175 + κ + κ + κ σ
3 3 3

2 arctan κ
− 1 + κ 2 5 + 3κ 2 3 + 30σ 2 + 35σ 4 . (124)
κ
The external solution of (117) is obtained by making use of the Green function (34)
∞

I3 (σ, θ) = 2π (2 + 1)q (σ)P (cos θ)
=0
1/κ π

2
× dσ σ + cos2 θ (1 + σ 2 ) sin2 θ p σ P (cos θ )dθ ,
0 0
(125)
which is reduced after implementing (118) and integrating over the angular variable
θ to
1/κ
64π
I3 (σ, θ) = q0 (σ) dσ p2 (σ ) + p4 (σ )
105
0
1/κ
64π
+ q2 (σ)P2 (cos θ) dσ 1 − p4 (σ ) p2 (σ )
105
0
1/κ
64π
− q4 (σ)P4 (cos θ) dσ 1 + p2 (σ ) p4 (σ ) . (126)
105
0
Performing the integrals over the radial variable yields the external solution
2
4π 1 + κ 2 2 3 − κ2
I3 (σ, θ) = q0 (σ) − q2 (σ)P2 (cos θ)
κ 5 15 21κ 2

5 + 3κ 2
− q 4 (σ)P 4 (cos θ) . (127)
35κ 2
5.3 Vector Potential V i
Vector potential V i is defined above by Eq. (42). As we need it only in the post-
Newtonian approximation, the density of the fluid entering (42) can be treated as
constant ρc . Each element of a rigidly rotating fluid has velocity, v i (x) = εi jk ω j x k ,
where εi jk is the Levi-Civita symbol, so that (42) can be written as follows
V i (x) = εi jk k̂ j Dk (x) , (128)
where k̂ i = ω i /ω is the unit vector along z-axis which coincides with the direction
of the angular velocity vector, ω i , and the Cartesian vector Dk = {D x , D y , D z } is
given by k 3
x d x
D (x) = Gωρc
k
. (129)
|x − x |
V
We denote, D+ ≡ D x + iD y . In the ellipsoidal coordinates one has5

√ 2
+ 1 + σ sin θ eiφ d 3 x
D (x) = Gωρc α . (130)
|x − x |
V
Because the angular velocity, ω i = (0, 0, ω), the vector potential V i = (V x , V y , V z )

has V z = 0. The remaining two components of the vector potential can be combined
together
V + = V x + i V y = iD+ . (131)
Equation (131) reveals that calculation of the vector potential is reduced to calculation
of the integral in the right hand side of (130) which depends on the point of integration
and is separated into the internal and external solutions. We discuss these solutions
below.
5 Notice that D z = 0 but we don’t need this component for calculating V + .

196 S. Kopeikin
The internal solution is obtained for the field points located inside the volume occu-
pied by the rotating fluid. Making use of the Green function (29) we have
∞ m=+
( − |m|)!
D+
= q|m| (σ) Ym ( x̂)
Gωρc α3 =0 m=−
( + |m|)!
σ

∗
dσ d σ 2 + cos2 θ 1 + σ 2 sin θ eiφ p|m| σ Ym ( x̂ )
0 S2
∞ m=+
( − |m|)!
+ p|m| (σ) Ym ( x̂)
=0 m=−
( + |m|)!
1/κ

∗
dσ d σ 2 + cos2 θ 1 + σ 2 sin θ eiφ q|m| σ Ym ( x̂ ) ,
σ S2
(132)
where, d = sin θ dθ dφ , is the element of the solid angle, and we integrate in

(132) over the unit sphere. We expand functions under the sign of integrals in terms
of the spherical harmonics

1 + σ 2 σ 2 + cos2 θ sin θeiφ =

2 Y31 ( x̂) 1 Y11 ( x̂)
− 1 + σ2 + σ2 + , (133)
15 C31 5 C11
and perform calculations of (132) with the help of the orthogonality relation (33).
Finally, we obtain the internal solution of the potential (131) in the following form
√
πGωρc α3 1 + σ 2
D+ = 12(1 + σ 2 κ 4 )P11 (cos θ)
30κ 4

+ 8σ 2 κ 4 + (1 + 5σ 2 )(3 + 5κ 2 ) P31 (cos θ)
arctan κ )
− 3(1 + κ ) 4P11 (cos θ) + (1 + 5σ )P31 (cos θ)
2 2 2
eiφ . (134)
κ
The external solution of (131) is obtained for the field points lying outside the volume
occupied by the rotating fluid. Making use of the Green functions (29), we obtain
∞ m=+
( − |m|)! 1/κ
D+
= q|m| (σ) Ym ( x̂) dσ
Gωρc α3 =0 m=−
( + |m|)!
0

∗
× d σ 2 + cos2 θ 1 + σ 2 sin θ eiφ p|m| σ Ym ( x̂ ) . (135)
S2
After making use of (133) and integrating over the angular variables, we get
⎡
1/κ
1
D = −2πGωρc α ⎣ q31 (σ) P31 (cos θ)
+ 3
1 + σ 2 p31 (σ )dσ
45
0
⎤
1/κ
1
+ q11 (σ) P11 (cos θ) 1 + σ 2 σ 2 + p11 (σ )dσ ⎦ eiφ . (136)
5
0
Integration with respect to the radial coordinate σ yields
D+ = Deiφ , (137)
where

(1 + κ 2 )2 1
D = −2πGωρc α3 q 11 (σ) P11 (cos θ) + q 31 (σ) P31 (cos θ) . (138)
5κ 5 6
Similar result has been obtained in [58, 59].

We notice that (138) can be transformed to yet another (differential) form

3S d 1 d P3 (cos θ)
D = − 2 1 + σ sin θ
2 q1 (σ) + q3 (σ) , (139)
4α dσ 6 d cos θ
where S = 2Ma 2 ω/5 is the angular momentum of the rotating spheroid.
6 Relativistic Multipole Moments of a Uniformly Rotating

Spheroid
Expansion of gravitational field of an extended massive body to multipoles is ubiq-

uitously used in celestial mechanics and geodesy in order to study the distribution of
the matter inside the Earth and other planets of the solar system [6, 8, 10]. General
relativity brings about several complications making the multipolar decomposition
of gravitational fields more difficult. First, all multipole moments of the gravitational
field should include the relativistic corrections to their definition. Second, besides the
198 S. Kopeikin
multipole moments of a single gravitational potential V of the Newtonian theory, one

has to include the multipole moments of all components of the metric tensor. There
is a vast literature devoted to clarification of various aspects of the multipolar decom-
position of relativistic gravitational fields but it goes beyond the scope of the present
chapter (see , for example, [5, 79–84]). We need two types of the post-Newtonian
multipole moments which appear in the decomposition of the scalar potential V , and
the vector potential V i . The scalar and vector multipoles are more commonly known
as mass and spin multipole moments [81] following the names of the leading terms
in the multipolar decompositions of the potentials V and V i respectively.
6.1 Mass Multipole Moments
Mass multipole moments of the external gravitational field of the rotating spheroid
are defined by expanding the scalar potential V entering g00 component of the metric
tensor (6a), in the asymptotic series for a large radial distance r in the spherical coor-
dinates. The scalar potential V takes into account the post-Newtonian contributions
from the internal energy , pressure p, the kinetic energy of rotation and the internal
gravitational energy as well as the spheroidal shape of matter distribution and its
inhomogeneity,
1
V = VN [ρc , S] + VN [δρc , S] + VN [ρc , δS] + V pN , (140)
c2
where the terms standing in the right hand side of this formula have been provided
above in Sects. 4 and 5. As we consider the multipolar expansion of V outside the
body, we need only the external solutions which are
VN [ρc , S] = Gm N [q0 (σ) + q2 (σ)P2 (cos θ)] , (141)

9A G 2 m 2N q1 5 2
VN [δρc , S] = q0 (σ) + q2 (σ)P2 (cos θ) − q4 (σ)P4 (cos θ) ,
20 c2 κ 7 7
(142)

2B Gm N 2 2 15 11 9 3
VN [ρc , δS] = ω a q0 (σ) + + + q2 (σ)P2 (cos θ)
5 c2 2 42 7κ 2 2κ 4

9 30 35
+ 3 + 2 + 4 q4 (σ)P4 (cos θ) , (143)
28 κ κ

2 5 3 − κ2
V pN = Gm N ω 2 a 2 q0 (σ) − q2 (σ)P2 (cos θ)
5 2 7κ 2

3 5 + 3κ 2
− q 4 (σ)P 4 (cos θ)
2 7κ 2

9 q1 5 2
− Gm N q0 (σ) + q2 (σ)P2 (cos θ) − q4 (σ)P4 (cos θ) ,(144)
2κ 7 7
where m N = M N /α, and the constant Newtonian mass, M N , is given by (82).

After replacing (141)–(144) in (140) and reducing similar terms, the scalar poten-
tial takes on the following form,
1
V = E0 q0 (σ) + E2 q2 (σ)P2 (cos θ) + E4 q4 (σ)P4 (cos θ) , (145)
c2
where the constant numerical coefficients are
)
1 9q1
E0 ≡ Gm N 1 + (A − 10) Gm N + 4(B + 1)ω a
2 2
, (146)
10c2 2κ

1 9q1 11 1 3
E2 ≡ Gm N 1 + 2 (A − 10) Gm N + 3B + 2
+ ω2 a2
c 28κ 42 7κ 2κ 4
)
1 3
+ 1− 2 ω a 2 2
, (147)
7 κ

9Gm N q1 30 35
E4 ≡ (10 − A) Gm N + B 3 + 2 + 4 ω 2 a 2
70 κ κ κ
)
2 5
− 3 + 2 ω2 a2 . (148)
3 κ
It is remarkable that the three constants, E0 , E2 , E4 are interrelated. Indeed, by direct

inspection of (146)–(148), we obtain
E4
E2 = E0 + . (149)
c2
Therefore, Eq. (145) takes on a more simple form,
1
V = E0 [q0 (σ) + q2 (σ)P2 (cos θ)] + E4 [q2 (σ)P2 (cos θ) + q4 (σ)P4 (cos θ)] .
c2
(150)
The scalar potential (150) is given in the ellipsoidal coordinates in terms of the
ellipsoidal harmonics which are the modified Legendre functions q0 (σ), q2 (σ) and
q4 (σ). The advantage in using the ellipsoidal harmonics is that it allows us to represent
the post-Newtonian scalar potential V with a finite number of a few terms only. The
residual terms in (150) are of the post-post-Newtonian order of magnitude (∼1/c4 )
which are systematically neglected.
In spite of the finite form of the expansion (150) in terms of the ellipsoidal func-
tions it is a more common practice to discuss the multipolar structure of external
gravitational field of an isolated body in terms of spherical coordinates (20). Mass
multipole moments of the gravitational field are defined in general relativity similarly
to the Newtonian gravity as coefficients in the expansion of scalar potential V with
respect to the spherical harmonics [80]. For axially-symmetric body the spherical
multipolar expansion of the scalar potential reads as follows [10, 80],
200 S. Kopeikin
* +
GM ∞ a
2
V = 1− J2 P2 (cos ) , (151)
R =1
R
where M is the relativistic mass, and J2 are the relativistic multipole moments of the
gravitational field that are defined (in terms of the spherical coordinates) by integrals
over the body’s volume

1
M= ρ(x) + 2 ρ pN (x) R 2 d RdO , (152)
c
V

2 1
J2 = − ρ(x) + ρ pN (x) R 2+2 P2 (cos )d RdO , (n ≥ 1) ,
Ma 2 c2
V
(153)
with dO ≡ sin dd is the infinitesimal element of the solid angle in the spherical
coordinates.
In order to read the multipole moments of the potential V out of (145) we have
to transform (145) to spherical coordinates. This can be achieved with the help of
the auxiliary formulas representing expansions of the ellipsoidal harmonics in series
with respect to the spherical harmonics. Exact transformations between ellipsoidal
and spherical harmonic expansions have been derived by Jekeli [85] for numerical
computations. However, Jekeli’s transformation lacks a convenient analytic form
and are not suitable for our purposes. Therefore, below we present a general idea
of calculation of the series expansion of the ellipsoidal harmonics in terms of the
spherical harmonics.6
The ellipsoidal harmonics are solutions of the Laplace equation and are repre-
sented by the products of the modified Legendre functions qm (σ) or pm (σ) with
the associated Legendre polynomials Pm (cos θ). We are interested in the expansion
of the th ellipsoidal harmonic q (σ)P (cos θ) in series of the spherical harmonics
which are also solutions of the Laplace equation. The most general expansion of this
type reads
∞
An Pn (cos )
q (σ)P (cos θ) = , (154)
n=
r n+1
where An are the numerical coefficients depending on n. As both sides of (154) are
analytic harmonic functions, they are identical at any value of the coordinates. In
order to calculate the numerical coefficients An , it is instructive to take the point with
θ = 0. At this point, we also have = 0, while σ = r , so that the expansion (154)
is reduced to
, ∞
, An
q (σ), = n+1
, (155)
σ=r
n=
r
6 Our method is partially overlapping with a similar development given in [10, Section 2.9].
which means that the coefficients An are simply the coefficients of the asymptotic
expansion of the modified Legendre function q (r ) for large values of its argument.
These coefficients are found by writing down the modified Legendre function q (σ)
in the left side of (155) in terms of the hypergeometric function 2 F1 (see [86, Eq.
VI-56b ])

+1 3 1
, √ F
2 1 1 + ; ; + ; −
, π ( + 1) 2 2 2 r2
q (σ), = +1 , (156)
σ=r 2 3 r +1
+
2
and equating the coefficients of the expansion to An in the right side of (155).
After applying the above procedure, we get for the first several elliptic harmonics
the following series,
∞
(−1) P2 (cos )
q0 (σ) = + , (157)
=0
2 + 1 r 2+1
∞
2(−1) P2 (cos )
q2 (σ)P2 (cos θ) = − , (158)
=1
(2 + 1)(2 + 3) r 2+1
∞
4( − 1)(−1) P2 (cos )
q4 (σ)P4 (cos θ) = + . (159)
=2
(2 + 1)(2 + 3)(2 + 5) r 2+1
Replacing expansions (157)–(159) in (150), reducing terms of the same power in

1/r 2+1 , and comparing the terms of the expansion obtained with similar terms in
(151), we conclude that
G M = E0 , (160)

3(−1)+1 2 14 E4
J2 = 1− 2 , (n ≥ 1) (161)
(2 + 1)(2 + 3) 3c 2 + 5 E0
where the second term in the square brackets yields the post-Newtonian correction
to the Newtonian multipole moments of the Maclaurin ellipsoid which are defined
as the coefficient standing in front of the square brackets in (161).
It is convenient from practical point of view to express the relativistic multipole
dyn
moments (161) in terms of the dynamical form factor J2 of an extended body
with an arbitrary internal distribution of mass density. The dynamical form factor is
expressed in terms of the difference between the polar, C and equatorial, A, moments
of inertia,
dyn C−A
J2 = . (162)
Ma 2
202 S. Kopeikin
We follow the technique developed by Heiskanen and Moritz [10, Section 2.9]
according to which the quadrupole moment of the homogeneous ellipsoid must be
exactly equal to the dynamical form factor
dyn
J2 = J2 . (163)
It is rather straightforward to prove that in terms of the dynamical form factor equation
(161) reads
* +
3(−1)+1 2
dyn
J2 4 − 1 E4
J2 = 1 − + 5 2 + 2 , (164)
(2 + 1)(2 + 3) 3c 2 + 5 E0
where C and A are the principal moments of inertia. Equation (164) looks quite
different from (161) but the difference is illusory since the coefficient in the square
brackets of (164) is identically equal to the corresponding term in (161) because for
the model of the (almost) homogeneous spheroid accepted in the present chapter, the
ratio
C−A 2 2 E4
= 1 − , (165)
Ma 2 5 3c2 E0
that can be easily checked by direct calculation of the integrals defining the moments
of inertia. Equation (164) is a relativistic generalization of the result obtained previ-
ously by Heiskanen and Moritz [10, Equation 2-92].
Post-Newtonian equation (164) allows to calculate the multipole moments of the
normal gravity field at any order as soon as the other parameters of the spheroid
are defined. As a particular example we adopt the model of GRS80 international
ellipsoid that is characterized by the following parameters (see [8, Section 4.3] and
[11, Table 1.2]):
G M = 398600.5 × 109 m3 s−2 ,

a = 6378137 m ,
= 1082.63 × 10−6 ,
dyn
J2
ω = 7.292115 × 10−5 rad s−1 ,
1/ f = 298.257222101 .
Corresponding (derived) values for the first and second eccentricities of GRS80 are
= 0.08181919104282 ,
κ = 0.08209443815192 .
The value of the ratio E4 /E0 can be calculated on the basis of Eq. (209) which is
derived below in Sect. 7.3 from the condition of the hydrostatic equilibrium. Making
use of the numerical values of the parameters of GRS80 model, we get
E4
= −5.58671 × 10−7 (B − 0.0023445) , (166)
c2 E0
where B is the parameter defining in our spheroidal model a small deviation from the
shape of the Maclaurin ellipsoid. The value of this parameter can be chosen arbitrary
in the range being compatible with the limitations imposed by the post-Newtonian
approximation, say, −10 B 10. Calculating the scalar multipole moments in
our spheroidal model by means of (164) yields for the case of pure ellipsoid,
J4 = −2.37091222014 × 10−6 , J6 = +6.08347064201 × 10−9 ,

J8 = −1.42681406953 × 10−11 , (B = 0) (167)
and for spheroid
J4 = −2.37091158429 × 10−6 , J6 = +6.08346483750 × 10−9 ,

J8 = −1.42680988487 × 10−11 , (B = 1) . (168)
These values can be compared with the corresponding values of the GSR80 geodetic
model [8, Equation 4.77c] which does not take into account relativistic corrections,
J4GRS80 = −2.37091221865 × 10−6 , J6GRS80 = +6.08347062840 × 10−9 ,

J8GRS80 = −1.42681405972 × 10−11 (169)
One can see that accounting for relativistic corrections gives slightly different numer-
ical values of the multipole moments for different models of the normal gravity field
generated by rotating spheroid. Significance of these deviations for practical appli-
cations in geodynamics is a matter of future theoretical and experimental studies.
Nonetheless, already now we can state that general relativity changes the classic
model of the normal gravity field. Hence, the separation of the observed value of the
field into the normal gravity and its perturbation differs from the Newtonian theory
and has certain consequences for interpretation of the gravity field anomalies.
6.2 Spin Multipole Moments
The spin multipole moments are defined as coefficients in the expansion of vector
potential V i with respect to vector spherical harmonics [80]

∞ m=+

V i (r, , ) = E m (r )Y E,m
i
(, ) + B m (r )Y B,m
i
(, )
=0 m=−

+R m (r )Y R,m
i
(, ) , (170)
204 S. Kopeikin
where E m , B m , R m are the spin multipole moments depending on the radial coor-
i i i
dinate r , and Y E,m , Y B,m , Y R,m are the Cartesian components of the three vector
spherical harmonics, Y E,m , Y B,m , Y R,m . The harmonics Y E,m and Y R,m are of
“electric-type” parity (−1) , while Y B,m have “magnetic-type” parity (−1)+1 [80].
Only the “magnetic-type” harmonics present in the expansion of the vector potential
in case of an axially-symmetric gravitational field [58], hence, we don’t consider the
“electric-type” harmonics below.
The “magnetic-type” harmonics are defined as follows [69]
LYm (, )
Y B,m (, ) = i √ , (171)
( + 1)
where L = −i x × ∇ is the operator of the angular momentum, the cross ‘×’ denotes
the Euclidean product of vectors, and ∇ is the gradient operator. The Cartesian com-
ponents (L x , L y , L z ) of the vectorial operator of the angular momentum L expressed
in terms of the spherical coordinates, are [69, Exercise 2.5.14]

∂ ∂
L x = i cos cot + sin , (172a)
∂ ∂

∂ ∂
L y = i sin cot − cos , (172b)
∂ ∂
∂
L z = −i . (172c)
∂
We have found in Sect. 5.3 that all of the non-vanishing components of the vector
potential V i are included to the potential V + defined in (131). This potential is
proportional to the components of the vector spherical harmonics, Y +,m ∼ L + Ym
where the action of the operator L + on the standard spherical harmonics is as follows
[69, Exercise 12.6.7]

L + Ym (, ) = ( − m)( + m + 1)Y,m+1 (, ) , (173)
which tells us that V + ∼ Y,m+1 . On the other hand, due to the fact that the angular
coordinates of the ellipsoidal and spherical coordinates coincide, = φ, and V + =
iDeiφ as follows from (131) and (137), we conclude that the multipolar expansion
(170) of V + with respect to the spherical harmonics contains only the spherical
harmonics with m = 1, that is V + = iD+ ∼ Y1 ∼ P1 eiφ . This can be seen directly
after applying the Green function in spherical coordinates and taking into account the
rotational symmetry with respect to the angle which yields Eq. (137) with function
D having the following form
* +
S2+1 a
2
∞
GS
D= sin + P2+1,1 (cos ) , (174)
2R 2 =1
2 + 1 R
where

S ≡ S1 = ω ρ(x)R 4 sin2 d RdO , (175)
V

1 ω
S2+1 = ρ(x)R 2+4 sin P2+1,1 ()d RdO, ( > 1) (176)
+ 1 Sa 2
V
are the absolute value of the angular momentum (spin) of the rotating spheroid and
the spin multipole moments of the higher order, and V is the volume bounded by
the surface of the Maclaurin ellipsoid. It is sufficient to perform calculation of the
integral in (175) with the constant value of density, ρ(x) = ρc . It yields
2
S= Ma 2 ω , (177)
5
which coincides with the result obtained in textbooks on classic mechanics.
Calculation of the spin multipoles S2+1 can be also performed directly but it will
be more instructive to find them out from the Taylor expansion of function D given in
the ellipsoidal coordinates by Eq. (138). It is convenient to write down this equation
by replacing the central density ρc with the total mass, M, as follows

3 (1 + κ 2 ) 1
D = − G Mω q11 (σ) P11 (cos θ) + q31 (σ) P31 (cos θ) . (178)
10 κ2 6
In order to calculate the spin multipole moments, we have to transform (178) from
the ellipsoidal to spherical harmonics. For we have in (178) the ellipsoidal harmonics
qm (σ)Pm (cos θ) with the index m = 1, and the odd index = 2k + 1, we have to
apply a slightly different approach to get the transformation formula as compared
with that employed in the previous Sect. 6.1. More specifically, because both the
ellipsoidal and spherical harmonics are solutions of the Laplace equation, we have
∞
Bn P2n+1,1 (cos )
q2+1,1 (σ)P2+1,1 (cos θ) = , (179)
n=
r 2n+2
where Bn are the numerical coefficients depending on n. As both sides of (179) are
analytic harmonic functions, they are identical at any value of the coordinates. In
order to calculate the numerical coefficients Bn , we take the point with θ = π/2.
√ point we also have = π/2 and cos = 0, while the radial coordinate,
At this
σ = r 2 − 1. The Legendre polynomials
3
(n + )
P2n+1,1 (0) = (−1)n+1 √
2 2 = (−1)n+1 (2n + 1)!! (180)
π (n + 1) 2n n!
206 S. Kopeikin
so that the expansion (179) is reduced to

, ∞
(2 + 1)!! , (2n + 1)!! Bn
(−1)+1 q (σ) , = (−1)n+1 , (181)

2 !
2+1,1 , √ 2n n! r 2n+2
σ= r 2 −1 n=
which means that the coefficients Bn are simply the √ coefficients of the asymptotic
expansion of the modified Legendre function q1 ( r 2 − 1) for large values of its
argument, r 1. These coefficients are found by writing down the modified Leg-
endre function in the left side of (181) in terms of the hypergeometric function (see
[86, Eq. VI-57b ])

3 1 5 1
, √ 2 F1 + ; + ; 2 + ; 2
, π (2 + 3) 2 2 2 r
q2+1,1 (σ),, √ = 2+2 2+2
.
σ= r −1
2 5 r
2
2 +
2
(182)
Comparing the coefficients of the expansion of the right side of (182) with the
numerical coefficients in the right side of (181), we can find coefficients Bn .
After applying this procedure, we get for the first two ellipsoidal harmonics the
following series expansions,
∞
(−1) P2+1,1 (cos )
q11 (σ)P11 (cos θ) = 2 , (183)
=0
(2 + 1)(2 + 3) r 2+2
∞
(−1) P2+1,1 (cos )
q31 (σ)P31 (cos θ) = −24 . (184)
=1
(2 + 1)(2 + 3)(2 + 5) r 2+2
Replacing these expansions to (178) and reducing similar terms, we obtain

* +
G Ma 2 ω ∞
(−1) 2 a
2
D= sin − 15 P2+1,1 (cos ) .
5R 2 =1
(2 + 1)(2 + 3)(2 + 5) R
(185)
Comparing expansion (185) with (174), we conclude that the coefficients of the
multipolar expansion of the vector potential in (174) are
15(−1)+1 2
S2+1 = , ( ≥ 1). (186)
(2 + 3)(2 + 5)
The spin multipole moments, S2+1 , are uniquely related to the mass multipole
moments, J2 , of a homogeneous and uniformly rotating Maclaurin ellipsoid as
follows
2 + 1
S2+1 = 5 J2 . (187)
2 + 5
Relations (186) and (187) have been also derived by Teyssandier [58] by making use
of a different mathematical technique.
It is convenient to give, yet another form of the expansion (185) in terms of the
derivatives of the Legendre polynomials. To this end, we employ the relation [72,
Eq. 8.752-1]
m/2 d m P (u)
Pm (u) = (−1)m 1 − u 2 , (188)
du m
which allows us to recast (185) to the following form
⎡ ⎤
∞
a
2 d P
G Ma 2 ω sin ⎣ (−1) 2 2+1 (cos ) ⎦
D= 1 + 15 .
5R 2 (2 + 1)(2 + 3)(2 + 5) R d cos
=1
(189)
After accounting for relations (131), (137), the vector potential V i can be written
down explicitly in a vector form
⎡ ⎤
∞ a
2 d P
G (S × x)i ⎣ (−1) 2 2+1 (cos ) ⎦
Vi = 1 + 15 ,
2 R 3 (2 + 1)(2 + 3)(2 + 5) R d cos
=1
(190)
where the angular momentum vector S = {0, 0, S}, and S is defined in (177). We
notice that a similar expansion formula given by Soffel and Frutos [44, Eq. 23] for
the vector potential has a typo and should be corrected in accordance with (190).
7 Relativistic Normal Gravity Field
7.1 Equipotential Surface
Harmonic coordinates introduced in Sect. 2.1 represent an inertial reference frame

in space which is used to describe the motion of probe masses (satellites) and
light (radio) signals in metric (6a)–(6c). Let us consider a continuous ensemble
of observers rotating rigidly in space with respect to z axis of the inertial reference
frame with the angular velocity of the rotating spheroid, ω i . Each observer moves
with respect to the inertial reference frame along a world line x i ≡ {x(t), y(t), z(t)}
such that z(t) and x 2 (t) + y 2 (t) remain constant. We assume that each observer
carries out a clock measuring its own proper time τ = τ (t) where t = x 0 /c is the
coordinate time of the harmonic coordinates x α introduced in Sect. 2.1.
The proper time of the clock is defined by equation −c2 dτ 2 = ds 2 where the
interval ds is calculated along the world line of the clock [63, 87]. In terms of the
metric tensor (5) the interval dτ of the proper time reads,
208 S. Kopeikin
1/2
2 1
dτ = −g00 (t, x) − g0i (t, x)v − 2 gi j (t, x)v v
i i j
dt , (191)
c c
where, x = {x i (t)} is taken on the world line of the clock, and v i = d x i /dt =
(ω × x)i is a constant linear velocity of the clock with respect to the inertial ref-
erence frame. The ensemble of the observers is static with respect to the rotating
spheroid and represents a realization of a rigidly rotating reference frame extending
to the outer space outside the spheroid. It should be understood that the rigidly rotat-
ing observers are generally not in a free fall except of those which are at the radial
distance corresponding to the orbit of geostationary satellites. The rotating reference
frame is local - it does not go to a spatial infinity and is limited by the radial distance
at which the linear velocity equates to the speed of light, v ≤ c, that is |x| ≤ c/ω.
For the Earth this distance does not exceed 27.5 AU - a bit less than the radius of
Neptune’s orbit.
After replacing the metric (6a)–(6c) in (191) and extracting the root square, we
get the fundamental time delay equation in the post-Newtonian approximation [20]
dτ W
= 1 − 2 + O c−6 , (192)
dt c
where the time-independent function, W is given by

1 1 1 4 3 2 1
W = v2 + V + 2 v + v V − 4v i V i − V 2 . (193)
2 c 8 2 2
Function W is the post-Newtonian potential of the normal gravity field taken at the
point of localization of the clock [1, 5].
The equipotential surface is defined by the condition of the constant rate of clock’s
proper time with respect to the coordinate time, that is [2, 20, 45]
dτ
= W = const . (194)
dt
In case of a stationary spacetime generated by a rigidly rotating body through Ein-
stein’s equations, the equipotential surface is orthogonal at each point to the direction
of the gravity force (the plumb line) [1, 2, 45, 46]. Inside the rotating fluid the equipo-
tential surface also coincides with the levels of equal density - ρ, pressure - p, and
thermodynamic energy - [1, 5].
7.2 Normal Gravity Field Potential
The post-Newtonian potential, W , of the normal gravity field inside the rigidly rotat-
ing fluid body has been derived in detail in our previous publications [3, 4], and its
derivation corresponds to the internal solution for the potentials worked out in the
previous sections. The present chapter focuses on the structure of the normal gravity
field outside the rotating spheroid which is described by the external solutions of the
metric tensor coefficients discussed above.
The external solution of the scalar potential V entering (193) has been given in
(145). Vector (gravitomagnetic) potential outside the body is given by (128) or, more
explicitly,
V x = − sin φD , V y = cos φD , Vz = 0 , (195)
where D is given in (178). It is straightforward to show that the Euclidean dot product,
v i V i = v x V x + v y V y , entering the post-Newtonian part of W , is

v i V i = ωα 1 + σ 2 D sin θ , (196)
or, by making use of (38), (37)

2 5 54
vV =
i i
Gm N ω a q0 (σ) + q2 (σ) − q0 (σ) − q2 (σ) − q4 (σ)
2 2
15 49 49
)
54
× P2 (cos θ) − q2 (σ) + q4 (σ) P4 (cos θ) . (197)
49
The rest of the terms entering expression (193) for the normal gravity potential W
are
2
v 2 = ω 2 α2 1 + σ 2 sin2 θ = ω 2 α2 1 + σ 2 [1 − P2 (cos θ)] , (198)
3
2 2
v 4 = ω 4 α4 1 + σ 2 sin4 θ = 8ω 4 α4 1 + σ 2

1 2 1
× − P2 (cos θ) + P4 (cos θ) , (199)
15 21 35
2 1
v 2 VN = Gm N ω 2 α2 1 + σ 2 q0 (σ) − q2 (σ)
3 5
)
5 18
− q0 (σ) − q2 (σ) P2 (cos θ) − q2 (σ)P4 (cos θ) , (200)
7 35

1 1
VN2 = G 2 m 2N q02 (σ) + q22 (σ) + 2q2 (σ) q0 (σ) + q2 (σ) P2 (cos θ)
5 7
)
18 2
+ q2 (σ)P4 (cos θ) . (201)
35
Summing up all terms in (145) we can reduce it to a polynomial
1
W (σ, θ) = W0 (σ) + W2 (σ)P2 (cos θ) + W4 (σ)P4 (cos θ) , (202)
c2
210 S. Kopeikin
which coefficients are functions of the radial coordinate σ,
1 2 2
W0 (σ) = ω α (1 + σ 2 ) + Gmq0 (σ)
3 )
1 2 2 1 2 2 1
+ 2 ω α (1 + σ ) 2
ω α (1 + σ ) + Gm q0 (σ) − q2 (σ)
2
c 15 5
)
Gm 8 2 2 1 1 2
− 2 ω a q0 (σ) + q2 (σ) + Gm q0 (σ) + q2 (σ) , (203)
2
c 15 2 5
1 2 2
W2 (σ) = − ω α (1 + σ 2 ) + Gmq2 (σ)
3
1 2 2 2
+ 2 E4 q2 (σ) − ω 2 α2 (1 + σ 2 ) ω α (1 + σ 2 )
c 21
)
5
+Gm q0 (σ) − q2 (σ)
7

Gm 8 2 2 5 54
+ 2 ω a q0 (σ) − q2 (σ) − q4 (σ)
c 15 49 49
)
1
−Gmq2 (σ) q0 (σ) + q2 (σ) , (204)
7
1
W4 (σ) = E4 q4 (σ) + ω 2 α2 (1 + σ 2 ) ω 2 α2 (1 + σ 2 ) − 18Gmq2 (σ)
35
)
9 16
− Gm Gmq22 (σ) − ω 2 a 2 q2 (σ) + q4 (σ) , (205)
35 7
and we have denoted, m ≡ M/α, where M is the relativistic mass (152) that is related
to the Newtonian mass m N through Eqs. (160) and (146).
7.3 The Figure of Equilibrium
The surface of a rotating fluid body is defined by the boundary condition of vanishing
pressure, p = 0. This surface coincides with the level of the constant gravitational
potential [2, 5] that is defined by the condition,
W(σs , θ) = W0 = const., (206)
for the value of the radial coordinate σs = σs (θ) defined above in (63). After expand-
ing the left hand side of (206) around the constant value of the radial coordinate 1/κ,
the equation of the level surface takes on the following form
2 3
γ2 ω 2 a 3 2 ω a
W̄0 − B + W̄2 − γ0 + γ2 B P2 (cos θ)
5κc2 7 κc2

1 18 γ2 ω 2 a 3
+ 2 W̄4 − B P4 (cos θ) = W0 , (207)
c 35 κ
where γ0 and γ2 are the components of the gravity force defined below in (219), and
W̄0 ≡ W0 (κ −1 ), W̄2 ≡ W2 (κ −1 ), W̄4 ≡ W4 (κ −1 ).
Because the potential is constant on the level surface the left hand side of (207)
cannot depend on the angle θ which means that the coefficients in front of the Leg-
endre polynomials, P2 (cos θ) and P4 (cos θ), must vanish. Equating,
18 γ2 ω 2 a 3
W̄4 − B=0, (208)
35 κ
yields
108 G 2 m 2 54 2 2
q4 E4 + q1 q2 B = − G m q2 (q2 + 8q4 ) , (209)
35 κ 245
where E4 has been defined in (148), q1 = q1 (1/κ), q2 ≡ q2 (1/κ), q4 ≡ q4 (1/κ),
and we have made use of (88). Equation (209) determines the coefficient B in the
equation of the spheroidal surface (63) of the rotating fluid body as a function of the
coefficient A defining the deviation of the internal density of the fluid, ρ, from the
uniform distribution by Eqs. (43), (44).
Equating, 2 3
2 ω a
W̄2 − γ0 + γ2 B=0, (210)
7 κc2
yields a relationship generalizing the Maclaurin equation (86) to the post-Newtonian

approximation,

4 Gm
ω a = 3Gmq2 1 −
2 2
20q2 + 49q0 + 36q4
245 c2

q2 3G 2 m 2
+ 2 E4 − 21 + 11κ q1 B ,
2
(211)
c 7κ 3
where q0 , q1 , q2 and q4 have the same meaning as in (209) above, and the constant
coefficient B relates to E4 by means of (209). In its own turn the coefficient E4 is
given by Eq. (148).
Finally, the constant value of the gravity potential on the surface of the rotating
ellipsoid is,
γ2 ω 2 a 3
W0 ≡ W̄0 − B, (212)
5κc2
or, more explicitly,
212 S. Kopeikin

G2m2 7 17 2 12
W0 = Gm(q0 + q2 ) − q0 + q0 q2 + q2 + q1 q2 B .
2
(213)
2c2 5 5 κ
In the small eccentricity approximation the value of the gravity potential on the level
surface is

G2 M 2 13 8B 4
W0 = Gm(q0 + q2 ) − 1+ + κ +O κ
2
. (214)
2a 2 c2 25 15
The condition of the hydrostatic equilibrium (206) imposes a constraint on the

linear combination of the constant parameters E4 and B through (209) which estab-
lishes the correspondence between the constants A and B defining the distribution
of mass density (44) inside the rotating body and the shape of its surface (63). There
are not any other limitations on these parameters. Hence, one of them can be chosen
arbitrary. One choice is to accept A = 0 that is to admit that the density of the fluid
is homogeneous at any order of the post-Newtonian approximations. This makes the
figure of the equilibrium of the rotating fluid deviate from the ellipsoid of revolution.
This choice was made, for example, in papers [4, 54, 56, 88–90] that consider the
corresponding figures of the post-Newtonian rotating homogeneous spheroids with
emphasis on astrophysical applications. On the other hand, one can postulate the
equipotential surface to be exactly the Maclaurin ellipsoid at any post-Newtonian
approximation which is achieved by choosing the parameter B = 0. Such an ellip-
soidal figure of equilibrium of a rotating fluid body has a non-homogeneous distri-
bution of mass density so that the parameter A = 0. This case has been considered
in our paper [3]. There is also a possibility to choose the constant parameter B in
such a way that the post-Newtonian formula (211) connecting the angular velocity of
rotation, ω, with the geometric parameters of the figure of equilibrium, will formally
coincide with the classic Maclaurin relationship (88). This case deserves a special
attention but we shall not dwell upon it over here as it relates to the problem of con-
sistency of a set of astronomical constants which is a prerogative of the International
Astronomical Union [http://asa.usno.navy.mil/SecK/Constants.html].
The gravity potential W0 is defined by formula (213). The value of W0 is the
gravity potential on the surface of geoid, that is currently chosen as a defining
constant7 without taking into account the post-Newtonian contribution as follows,
W0 = 62636856.0 ± 0.5 m2 s−2 [11, Table 1.1]. The fractional uncertainty in the
IERS Convention 2010 value of W0 is δW0 /W0 8 × 10−9 . This uncertainty is sig-
nificantly less than that of the global geometric reference system and is no longer
accurate enough to match the operational precision of VLBI and SLR measurements
as well as satellite laser altimetry of ocean’s surface which have reached the level
of one millimeter. This accuracy is at the level of the post-Newtonian correction
to the Newtonian value of W0 (the first term in a right hand side of (213)) that
can be easily evaluated on the basis of the approximate formula (214), and is about
L G = W0 /c2 = 6.969290134 × 10−10 that determines the difference

7 It is equivalent to a constant
between TT and TCG time scales (see [20, 27] or [5, Appendix C.2, Resolution B1.9.]).
G M⊕ /2a⊕ c2 3.5 × 10−10 or, in terms of height, 2.2 mm on the Earth surface. This
is within the operational precision of current geodetic techniques. We suggest that
the significance of the post-Newtonian correction to the defining constant W0 should
be thoroughly discussed by corresponding IAU/IAGG working groups developing
a new generation of the system of the geopotential-based geodetic constants (see
discussion in [8, Section 4.3]).
7.4 The Somigliana Formula
The Somigliana formula in classic geodesy gives the value of the normal gravity γ i
on the reference ellipsoid [6, 8, 10]. Vector of the normal gravity is perpendicular to
the equipotential surface, and is calculated in the post-Newtonian approximation in
accordance with equation [2, 5]

∂ 1
γ i = −c2 i j log 1 − W , (215)
∂x j c2
where the matrix operator

1 1 i j 1
i j = δ i j − v v + δ i j VN +O , (216)
c2 2 c4
defines transformation to the inertial frame of a local observer being at rest with
respect to the rotating frame of reference.
We are looking for the normal component, γn = n̂ i γ i , of the vector γ i in the
direction of the plumb line that is given by the unit vector n̂ defined in (19). A
particular interest represents the value of γn taken on the surface of the ellipsoid which
corresponds to the classical derivation of the formula of Somigliana [8, 10]. After
taking the partial derivative in (215), and making use of the ellipsoidal coordinates,
it reads
* 1/2 +
1 1 2 2 1 + σ2 ∂W
γn = − 1 + 2 ω α (1 + σ ) sin θ
2 2
, (217)
α 2c σ 2 + cos2 θ ∂σ
σ=σs
or, more explicitly

1/2
1 1 + σs 2
γn = 1 + 2 ω 2 a 2 sin2 θ
2c σs 2 + cos2 θ

1
γ0 (σ) + γ2 (σ)P2 (cos θ) + 2 γ4 (σ)P4 (cos θ) , (218)
c σ=σs
214 S. Kopeikin
where we have introduced the following notations for the partial derivatives of the
components of the normal gravity potential
1 ∂W0 1 ∂W2 1 ∂W4

γ0 (σ) ≡ − , γ2 (σ) ≡ − , γ4 (σ) ≡ − , (219)
α ∂σ α ∂σ α ∂σ
and the radial coordinate, σs = σs (θ), in accordance with (63). It means that γ0 (σs ),
γ2 (σs ), and γ4 (σs ) depend on the angular coordinate θ.
In what follows, it is more convenient to expand all functions entering the right
hand side of (218) into the Taylor series expanded around the value of the radial
coordinate σ = κ −1 . This brings (218) to the following form,
1/2
1 + κ2 1 1 − 3 cos2 θ
γn = 1 + 2 ω 2 a 2 sin2 θ 1 + B
1 + κ 2 cos2 θ 2c 1 + κ 2 cos2 θ

1
γ̄0 + γ̄2 P2 (cos θ) + 2 γ̄4 P4 (cos θ) , (220)
c
where now we have

1 ω 2 a 2 ∂γ2
γ̄0 = γ0 + B , (221)
5 52 c2 ∂σ σ=κ −1

ω 2 a 2 ∂γ0 2 ∂γ2
γ̄2 = γ2 + 2 2 + B , (222)
c ∂σ 7 ∂σ σ=κ −1

18 ω 2 a 2 ∂γ2
γ̄4 = γ4 + B . (223)
35 2 c2 ∂σ σ=κ −1
According to Somigliana [10] it is more convenient to write down the normal

gravity force γn in terms of two constants which are the values of the normal gravity
force taken at two particular positions on the surface: (1) a point a on the equator
with θ = π/2, and (2) a point b at the pole with θ = 0. At these points (218) takes
on the following forms,

ω2 a2 γ̄2 3γ̄4
γa = 1 + κ21 + (B + 1) γ̄0 − + 2 , (224)
2c2 2 8c
γ̄4
γb = γ̄0 + γ̄2 + 2 . (225)
c
Solving these equations with respect to γ̄0 and γ̄2 we get the post-Newtonian gener-
alization of the theorem of Pizzetti [8, 78],

2b ω2 a2 1 7
γ̄0 = 1− 2
γa + γb − γ̄4 , (226)
3a 2c 3 12c2
and the theorem of Clairaut [10],


2b ω2 a2 2 5
γ̄2 = − 1− γa + γb − γ̄4 . (227)
3a 2c2 3 12c2
Equations (226) and (227) coincide with the corresponding post-Newtonian formu-
lations of Pizzetti and Clairaut theorems given in our paper [3, Sections 11 and
12] after making transformation of the parameters to the new gauge defined by the
parametrization (63) of the shape of the rotating spheroid.8
Theorems Pizzetti and Clairaut are used to derive the formula of Somigliana
describing the magnitude of the normal gravity field vector in terms of the forces
of gravity measured at equator and at pole [10]. This is achieved by replacing (226)
and (227) back into (220) and expanding it with respect to 1/c2 . It results in the
post-Newtonian generalization of the formula of Somigliana,
aγb cos2 θ + bγa sin2 θ 1 4ω 2 a 2 (aγb − bγa ) − 35aγ4

γn = + 2
sin2 2θ
a cos θ + b sin θ
2 2 2 2 32c a cos θ + b sin θ
2 2 2 2
ω 2 b 3b(aγb − bγa ) sin2 θ − a(aγa + 2bγb ) 2

+B 2 3/2 sin 2θ . (228)
8c a 2 cos2 θ + b2 sin2 θ
The first term in the right side of (228) is the canonical formula of Somigliana used
ubiquitously in classic geodesy,9 and the second and third terms being proportional
to 1/c2 , are the explicit post-Newtonian corrections. It is worth noticing that the post-
Newtonian corrections to the Somigliana formula (228) are also included implicitly
to the canonical (first) term through the values of the normal gravity force at equator,
γa , and at the pole, γb , as follows from (221)–(225).
It is instructive to compare the results of the post-Newtonian formalism of the pre-
vious sections with the post-Newtonian approximations of axially-symmetric exact
solutions of the Einstein equations. There are plenty of the known solutions (see, e.g.,
[91]) and some of their aspects have been analyzed in the application to relativistic
geodesy in [44, 45]. Below, we focus on the post-Newtonian approximation of the
Kerr metric and comment on its practical usefulness in geodesy.
8 Normal Gravity Field of the Kerr Metric
The Kerr metric is an exact, axisymmetric, stationary solution of the Einstein equa-
tions found by Roy Kerr [91]. The Kerr metric is a vacuum solution representing
rotating black hole. It is often assumed in relativistic mechanics that the Kerr metric
can be used to describe the external gravitational field of rotating extended body as
8 For more details about the gauge transformations of the post-Newtonian spheroid the reader is
referred to [3, Section 4].
9 One should notice that in classic geodesy the Somigliana formula is usually expressed in terms
of the geographic latitude on ellipsoid that is related to the ellipsoidal angle θ by θ = β − π/2,
and, a tan β = b tan , [10, Eq. 2-77].
216 S. Kopeikin
well. However, the exact internal solution generating the external Kerr metric has
not yet been found, and may not exist [92]. The internal Kerr solution having been
recently found by Hernandez-Pastora and Herrera [93] requires further study for
verification. Nonetheless, it is instructive to investigate the geometric properties of
the Kerr metric from the point of view of its application in relativistic geodesy. Pre-
liminary study of this problem has been performed in [44, 45]. The present chapter
focuses on the comparison of the multipolar structure of gravitational field of the Kerr
metric with that of gravitational field of rotating spheroid. Because we employed the
harmonic coordinates for description of the normal gravitational field, we will need
the Kerr metric expressed in the harmonic coordinates defined by the condition (4).
8.1 Harmonic and Ellipsoidal Coordinates
The Kerr metric in harmonic coordinates, x α = {x, y, z}, has been derived in [94, 95].
We introduce the Kerr ellipsoidal coordinates, {ς, ϑ, ϕ}, connected to the harmonic
coordinates x α by equations

x = αK 1 + ς 2 sin ϑ cos ϕ , (229a)

y = αK 1 + ς 2 sin ϑ sin ϕ , (229b)
z = αK ς cos ϑ , (229c)
which look similar but not equal to the ellipsoidal coordinates (15) for uniformly
rotating fluid body. The difference between the two types of the ellipsoidal coor-
dinates is due to the fact that the geometric meaning of the parameter α in (15) is
different from that of the Kerr parameter αK which is equal (by definition of the Kerr
geometry) to the ratio of the angular momentum, S, to mass, M, of the rotating body:
αK = S/Mc. Dimension of the Kerr parameter αK is the same (length) as one of the
parameter α = a which is used in definition (15) of the ellipsoidal coordinates in
classic geodesy. Nonetheless, the two parameters, α and αK , have different numerical
values in the most general case making the two types of the ellipsoidal coordinates
related to each other by transformation

αK 1 + ς 2 sin ϑ = α 1 + σ 2 sin θ , αK ς cos ϑ = ασ cos θ , ϕ=φ.
(230)
We can establish a connection between two parameters, αK and α, by making
use of relationship, S = I ω, that expresses the angular momentum S of rotating
extended body with its rotational moment of inertia, I , and the angular velocity of
rotation, ω. The moment of inertia can be expressed in terms of mass of the body
and its equatorial radius, I = λMa 2 , where λ is a dimensionless integral parameter
that depends on the distribution of matter inside the rotating body and is determined
by an equation of state. For example, in case of a homogeneous ellipsoid, λ = 2/5
[77, 96].
Let us now assume that the Kerr metric is generated by an extended, rigidly rotating
ellipsoid with some non-homogeneous distribution of mass density. Homogeneous
distribution of density is excluded as the multipolar expansion of gravitational field of
the Kerr metric does not coincide with that of the homogeneous ellipsoid of rotation
as we show below. Then, in accordance with definition of the Kerr parameter, we
should accept αK = S/Mc = aK , where K = λωa/c is the effective oblateness of
the extended body that supposedly generates the Kerr metric (231). Thus, parameter
αK can be equal to α = a, if and only if, = K . This condition imposes a certain
limitation on the oblateness of rotating extended body which, on the other hand,
is also a function of the rotational angular velocity and the average density, ρ, of
the body, = (ρ, ω), as follows from the condition of hydrostatic equilibrium of
the body’s matter in the rotating frame.10 It means that for a given total mass M
and oblateness of extended rotating body, its angular velocity cannot be adjusted
independently of the other parameters to make αK = α. This is because the other
parameters of the body like density, ρ, or a semi-major axis, a, must be appropriately
chosen to maintain the condition of hydrostatic equilibrium in the rotating frame.
We proceed by assuming that αK = α. At the same time, we postulate that the total
mass M and angular momentum S of the Kerr metric exactly coincide with the total
mass M and angular momentum S of rotating body. This can be always done as these
parameters are defined in terms of conserved integrals given at the spatial infinity of
asymptotically-flat spacetime [42].
8.2 Post-Newtonian Approximation of the Kerr Metric
In the ellipsoidal coordinates (229) the exterior solution of Einstein’s equations for
the Kerr metric reads [94, 95]

dς 2
ds 2 = − c2 dt 2 + αK
2
(ς + μK )2 + cos2 ϑ + dϑ2 (231)
1 + ς 2 − μ2K
2
2μK (ς + μK ) αK μ2K sin2 ϑdς
+ − α K sin 2
ϑdφ + cdt
(ς + μK )2 + cos2 ϑ (1 + ς 2 − μ2K )(1 + ς 2 )
2
μ2K dς
+ αK
2
sin2 ϑ (ς + μK )2 + 1 − dφ ,
(1 + ς 2 − μ2K )(1 + ς 2 )
where the Kerr mass parameter μK ≡ Gm K /c2 , and m K ≡ M/αK . The interior solu-
tion for the Kerr metric is still unknown despite of numerous attempts to find it out
[50, 82, 91, 97, 98]. Recently, a certain progress has been made by Hernandez-
Pastora and Herrera [93] who used a model of a viscous, anisotropic distribution of
mass density inside rotating body.
10 For example, in case of a rigidly rotating homogeneous perfect fluid the relation, = (ρ, ω), is
simply given by the Maclaurin formula (86).

218 S. Kopeikin
The post-Newtonian approximation of the Kerr metric (231) in the ellipsoidal

coordinates is

2μK ς 2 ς − cos ϑ
2 2
4αK μK ς sin2 ϑ
ds 2 = −1 + 2 − 2μ c2 dt 2 − 2 cdtdφ
ς + cos ϑ
2 (ς + cos ϑ)
K 2 2 2 ς + cos2 ϑ
2
2 ς + cos ϑ + 2μK ς
2 2 2μK ς
+ αK dς 2
+ 1 + ς 2
sin ϑ 1 + dφ2
1 + ς2 ς 2 + cos2 ϑ

+ ς 2 + cos2 ϑ + 2μK ς dϑ2 . (232)
Harmonic coordinates are asymptotically Cartesian with the Euclidean metric δi j at

spatial infinity. Components of the Euclidean metric in the ellipsoidal coordinates
can be obtained directly from the coordinate transformation (229), and the result
reads
2
ς + cos2 ϑ 2 2 2 2
δi j d x d x = αK
i j 2
dς + 1 + ς sin ϑdφ + ς + cos ϑ dϑ .
2 2 2
1 + ς2
(233)
Comparing (233) with (232) allows us to recast the spacetime metric (232) to the
following form

2VK 2V 2 1 2G 2 m 2 cos2 ϑ
ds 2 = −1 + 2 − 4K + 4 2 K 2 2 c2 dt 2
c c c (ς + cos ϑ)
i
8VK 2VK
− 2 dtd x + 1 + 2 δi j d x i d x j ,
i
(234)
c c
where
Gm K ς
VK = , (235)
ς 2 + cos2 ϑ
1 (S × x)i VK
VKi = , (236)
2 1 + ς 2 αK 3
mK
and S = {0, 0, S} is a vector of the total angular momentum (spin) of the body
directed along the z axis of the harmonic coordinates which coincides with the
direction of the rotational axis, S = αK Mc. Potentials VK and VKi are harmonic
functions and satisfy the Laplace equation
VK = 0 , VKi = 0 , (237)
where the Laplace operator in the ellipsoidal coordinates is defined in (28) after a
corresponding replacement σ → ς.
The very last term in time-time component of the metric (234) is, actually, of the 2-
nd post-Newtonian order of magnitude, and can be dropped out in the post-Newtonian
approximation when comparing with the post-Newtonian metric (6) of extended
body. The thing is that the metric (234) is written in dimensionless coordinates,
while the parameter αK = S/Mc, that has been used to make them dimensionless,
includes a relativistic factor of 1/c which is the main parameter of the post-Newtonian
expansions. When we go back to the dimensional coordinates and also use the angular
momentum, S and mass M, instead of m = M/αK , we have
1 2G 2 m 2K cos2 ϑ 1 2G 2 S 2 cos2 ϑ
= , (238)
c4 (ς 2 + cos2 ϑ)2 c6 R 2 + α2 cos 2ϑ 2
K
2
where R 2 = x 2 + y 2 + z 2 = αK2
ς + sin2 ϑ , and S = αK Mc is the angular
momentum of the rotating body. The right hand side of (238) is apparently of the order
of 1/c6 which is the post-post-Newtonian term not entering the first post-Newtonian
approximation [5].
8.3 Normal Gravity Field Potential of the Kerr Metric
The normal gravity field represented by the Kerr metric can be expressed in a closed
form similarly to the normal gravity field of uniformly rotating perfect fluid. Potential
W of the normal field in case of the Kerr metric is defined by formula (193) where
we have to use VK and VKi for scalar and vector gravitational potentials. We have
1 2 2 Gm K ς
W = ω αK 1 + ς 2 sin2 ϑ + 2
2 ς + cos2 ϑ

1 1 4 4 2 3 1
+ 2 ω αK 1 + ς 2 sin4 ϑ + (1 + ς 2 ) − 2λ 1 + 2
c 8 2 κ
)
Gm K ς 1 G mKς 2 2 2
× 2 ω 2 αK
2
sin2 ϑ − , (239)
ς + cos2 ϑ 2 ς 2 + cos2 ϑ 2
where λ is the parameter introduced above in Sect. 8.1, to connect the moment of
inertia of a rotating body with its mass and the equatorial radius.
8.4 Multipolar Expansion of Scalar Potential
Analytic comparison of the Kerr metric with an external solution of uniformly rotat-
ing ellipsoid is achieved by comparing the multipolar expansions of the correspond-
ing gravitational potentials. The multipolar expansion of the Kerr metric is obtained
220 S. Kopeikin
by applying the partial fraction decomposition,

ς 1 1 1
=− + , (240)
ς 2 + cos2 ϑ 2i iς + cos ϑ iς − cos ϑ
and a generating function for the Legendre functions of the 2-nd type [72, formula
8.791]
∞
1
= (2 + 1)P (v)Q (u) . (241)
u−v =0
It allows us to represent the Newtonian potential VK of the Kerr metric in the following
form
∞

VK = Gm K (−1) (4 + 1)q2 (ς)P2 (cos ϑ)
=0

= m K q0 (ς) − 5q2 (ς)P2 (cos ϑ) + 9q4 (ς)P4 (cos ϑ) + . . . . (242)
The post-Newtonian terms in time-time component of the metric (234) can be

written as a partial derivative of the Newtonian potential
G 2 m 2K cos2 ϑ ∂VK
− VK2 + = mK . (243)
(ς 2 + cos2 ϑ)2 ∂ς
Derivative of the Legendre function q (ς) is [72, formula 8.832], [71, formula (78)]
dq (ς) 1+

=− q+1 (ς) + ςq (ς) . (244)
dς 1+ς 2
It yields for the partial derivative of the Newtonian potential

∞
∂VK Gm
=− (−1) (4 + 1)(2 + 1) q2+1 (ς) + ςq2 (ς) P2 (cos ϑ) .
∂ς 1 + ς =0
2
(245)
For analytical comparison of the Kerr metric with the post-Newtonian metric gen-
erated by extended rotating body it is sufficient to compare the multipolar expansion
of potential V of the metric of the rotating body and that of VK of the Kerr metric, and
to deduce the correspondence between the parameters of the expansions [82]. Any
kind of a multipolar expansion performed in either ellipsoidal or spherical coordi-
nates can be used. Nonetheless, in theoretical practice, the comparison of multipolar
expansions of gravitational fields is usually done in spherical coordinates. We follow
this practice and introduce the spherical coordinates {, , }
x = αK sin cos , (246a)

y = αK sin sin , (246b)
z = αK cos , (246c)
where the angular coordinates , are the same as in (15) and the radial coordinate
is related to R by a scale transformation, R = αK so that, = (α/αK )r .
To get the multipolar expansion of potential VK we use the following expansion
∞
ς P2 (cos )
= (−1) , (247)
ς 2 + cos2 ϑ =0
2+1
that can be easily confirmed by direct inspection for the point with angular coordinate,
θ = = 0, when ς = , and applying expansion of the elementary function
∞
(−1)

= . (248)
2 + 1 =0
2+1
The expansion (247) can be re-written in the form of a multipolar expansion of a

scalar potential,
* +
GM ∞
2
K a
VK = 1− J2 P2 (cos ) , (249)
R =1
R
where the mass multipole moments of the Kerr metric

2
λωa
K
J2 = (−1)+1 2
K = (−1)
+1
, ( ≥ 1) (250)
c
It should be compared with the multipolar expansion (151) of the potential V of

rotating homogeneous spheroid where the multipole moments J2n are defined in
(161). Comparing VK and V and assuming that each of the potentials is generated by
a corresponding axially-symmetric body, one can see that monopole terms (∼1/R)
in (249) and (151) match perfectly so that the total mass of the Kerr metric can,
indeed, be equated to the post-Newtonian mass of rotating spheroid, as it has been
postulated above. The second order (quadrupole) moments can be matched as well.
Indeed, we are allowed to equate J2K = J2 , under condition

K = √ , ( = 1) . (251)
5
This result means that the Kerr metric can imitate gravitational field of a uniformly
rotating Maclaurin ellipsoid in the quadrupole approximation. However, as soon as
the condition (251) is satisfied, the multipole moments of the higher order (octupole,
222 S. Kopeikin
decapole, etc.) of multipole expansions of two potentials, V and VK , cannot be

matched as follows from comparison of two expressions – (161) and (250). It means
K
that the mass multipole moments of the Kerr metric, J2 , are, in general, different
from the multipole moments J2 of rotating ellipsoid for ≥ 2.
8.5 Multipolar Expansion of Vector Potential
Gravitomagnetic vector-potential, VKi of

the Kerr metric
is given in (236). It has only
y
two non-vanishing components, VKi = VKx , VK , 0 , which can be combined together
in the complex potential, VK+ ≡ VKx + i VK , where i is the imaginary unit, c.f. (131).
y
The explicit form of the potential is
VK+ = iDK eiφ , (252)
where
c sin ϑ c sin ϑ Gm K ς
DK = √ VK = √ . (253)
2 1+ς 2 2 1 + ς ς + cos2 ϑ
2 2
We make use of the expansion (242) for VK , and equation
1
sin ϑP (cos ϑ) = P2−1,1 (cos ϑ) − P2+1,1 (cos ϑ) , (254)
4 + 1
that is given in [72, formula 8.733-4], to bring (253) to the following form
∞
Gm K c q2+2 (ς) + q2 (ς)
DK = − (−1) √ P2+1,1 (cos ϑ) . (255)
2 =0 1 + ς2
We are now use [72, formula 8.734-5]

(2 + 1)(2 + 2)
1 + ς 2 q2+1,1 (ς) = q2 (ς) + q2+2 (ς) , (256)
4 + 3
to obtain the expansion of DK in the final form

∞
Gm K c (−1) (4 + 3)
DK = − q2+1,1 (ς)P2+1,1 (cos ϑ)
4 =0 ( + 1)(2 + 1)

3m K c 7
=− q11 (ς)P11 (cos ϑ) − q31 (ς)P31 (cos ϑ)
4 18

11
+ q51 (ς)P51 (cos ϑ) + . . . . (257)
45
Expansion of function DK in spherical harmonics reads

∞
Gm K c (−1) P2+1,1 (cos )
DK = − . (258)
2 =0 2 + 1 2+2
Thisformula can be checked by taking a point with coordinate, ϑ = = π/2, where

ς = 2 − 1 in accordance with (229) and (246). At this value of the angular coor-
dinate expression (253) yields,
∞
Gm K c 1 Gm K c (2 − 1)!! 1
DK = = , (259)
2 2 − 1 2 =0 2 ! 2+2
but this is exactly the same formula as (258) for = π/2 with a special value of the
polynomial P2+1,1 (0) taken from (180).
Coming back to a physical domain of the dimensional coordinates, and expressing
the Legendre polynomial, P2+1,1 in terms of the first derivative with respect to its
argument, we obtain
* +
G S sin ∞
(−1) 2 a
2 d P
2+1 (cos )
DK = 1+ K
, (260)
2 R2 =1
2 + 1 R d cos
where S = αK Mc is a spin of the body generating the Kerr metric. Comparison of the
multipole expansion (260) with (174) allows us to read out spin multipole moments,
K
S2+1 , of the Kerr metric,
K
S2+1 = (−1)+1 2
K , ( ≥ 1) . (261)
K
As one can see, the spin multipole moments, S2+1 , of the Kerr metric given in (261)
are significantly different from those, S2n+1 , of the homogeneous rotating ellipsoid
given in (186). Relation between spin and mass multipole moments of the Kerr
metric,
K
S2+1 = J2K
, (262)
does not coincide with similar relation (187) between spin and mass multipole
moments of a rotating spheroid made of homogeneous fluid. We also notice that
relation (262) corresponds to a well-known relation between Geroch-Hansen mass
and spin moments of the Kerr black hole [99, Chapter 14].
Replacing (260) to vector potential of the Kerr metric yields
* +
G (S × x)i ∞
(−1) 2 a
2 d P
2+1 (cos )
VKi = 1+ K
. (263)
2 R3 =1
2 + 1 R d cos
224 S. Kopeikin
We can compare the multipole expansion (263) of vector potential of the Kerr metric
with similar expansion (190) given for rotating spheroid. We notice that the first terms
of these expansions match exactly each other so that the angular momentum, S, of
the body generating the Kerr geometry can be equated to the angular momentum of
the rotating spheroid. As we have already equated the mass of the body generating
gravitational field of the Kerr metric to that of the spheroid, we have to conclude that
equating the angular momenta of the two bodies also requires imposing a limitation,
αK = α, or, in other words, K = . However, this equation is not compatible with
the matching condition (251) for the quadrupoles. It means that the Kerr metric is
compatible with the normal gravity field of rotating ellipsoid merely in the mass-
monopole spin-dipole approximation. This approximation is not suitable for geodetic
applications. We conclude that the Kerr metric should not be used for the purposes
of relativistic geodesy.
Acknowledgements I thank Physikzentrum Bad Honnef for hospitality and Wilhelm and Else
Heraeus Stiftung for providing generous travel support to deliver a talk at 609 WE-Heraeus-
Seminar “Relativistic Geodesy: Foundations and Applications” (13.03. - 19.03.2016). This work
contributes to the research project “Spacetime Metrology, Clocks and Relativistic Geodesy” [http://
www.issibern.ch/teams/spacetimemetrology/] sponsored by the International Space Science Insti-
tute (ISSI) in Bern, Switzerland.
References
1. S.M. Kopejkin, Relativistic manifestations of gravitational fields in gravimetry and geodesy.

Manuscr. Geod. 16, 301–312 (1991)
2. S.M. Kopeikin, E.M. Mazurova, A.P. Karpik, Towards an exact relativistic theory of Earth’s
geoid undulation. Phys. Lett. A 379, 1555–1562 (2015)
3. S.M. Kopeikin, Reference ellipsoid and geoid in chronometric geodesy. Front. Astron. Space
Sci. 3(5), 5 (2016)
4. S. Kopeikin, W. Han, E. Mazurova, Post-Newtonian reference ellipsoid for relativistic geodesy.
Phys. Rev. D 93(4), 044069 (2016)
5. S. Kopeikin, M. Efroimsky, G. Kaplan, Relativistic Celestial Mechanics of the Solar System
(Wiley, Berlin, 2011), xxxii+860 pp
6. P. Vaníček, E.J. Krakiwsky, Geodesy, the Concepts, 2nd edn. (Amsterdam, North Holland,
1986), xv+697 pp
7. B. Hofmann-Wellenhof, H. Moritz, Physical Geodesy (Springer, Berlin, 2006)
8. W. Torge, J. Müller, Geodesy, 4th edn. (De Gruyter, Berlin, 2012), 433 pp
9. L.D. Landau, E.M. Lifshitz, The Classical Theory of Fields (Pergamon Press, Oxford, 1975),
xiii+402 pp
10. W.A. Heiskanen, H. Moritz, Physical Geodesy (W. H. Freeman, San Francisco, 1967), 364 pp
11. G. Petit, B. Luzum, IERS conventions. IERS Tech. Note 36, 179 pp. (2010)
12. K. Sośnica, D. Thaller, A. Jäggi, R. Dach, G. Beutler, Sensitivity of Lageos orbits to global
gravity field models. Artif. Satell. 47, 47–65 (2012)
13. P.L. Bender, R.S. Nerem, J.M. Wahr, Possible future use of laser gravity gradiometers. Space
Sci. Rev. 108, 385–392 (2003)
14. N.K. Pavlis, S.A. Holmes, S.C. Kenyon, J.K. Factor, The development and evaluation of the
Earth gravitational model 2008 (EGM2008). J. Geophys. Res. Solid Earth 117, B04406 (2012)
15. N.K. Pavlis, S.A. Holmes, S.C. Kenyon, J.K. Factor, Correction to “the development and
evaluation of the Earth gravitational model 2008 (EGM2008)”. J. Geophys. Res. Solid Earth
118, 2633–2633 (2013)
16. Fu Lee-Lueng, On the decadal trend of global mean sea level and its implication on ocean heat
content change. Front. Mar. Sci. 3, 37 (2016)
17. H.-P. Plag, M. Pearlman (eds.), Global Geodetic Observing System (Springer, Dordrecht, 2009),
322 pp
18. L.-L. Fu, B.J. Haines, The challenges in long-term altimetry calibration for addressing the
problem of global sea level change. Adv. Space Res. 51, 1284–1300 (2013)
19. Z. Altamimi, P. Rebischung, L. Métivier, X. Collilieux, ITRF2014: a new release of the inter-
national Terrestrial Reference Frame modeling nonlinear station motions. J. Geophys. Res.
Solid Earth 121, 6109–6131 (2016)
20. J. Müller, D. Dirkx, S.M. Kopeikin, G. Lion, I. Panet, G. Petit, P.N.A.M. Visser, High perfor-
mance clocks and gravity field determination. Space Sci. Rev. 214(5), 1–31 (2018)
21. R. Bondarescu, M. Bondarescu, G. Hetényi, L. Boschi, P. Jetzer, J. Balakrishna, Geophysical
applicability of atomic clocks: direct continental geoid mapping. Geophys. J. Int. 191, 78–82
(2012)
22. E. Mai, J. Müller, General remarks on the potential use of atomic clocks in relativistic geodesy.
ZFV - Zeitschrift fur Geodasie, Geoinformation und Landmanagement 138(4), 257–266 (2013)
23. E. Mai, Time, atomic clocks, and relativistic geodesy. Report No 124, Deutsche Geodátische
Kommission der Bayerischen Akademie der Wissenschaften (DGK) (2014), 128 pp., http://
dgk.badw.de/fileadmin/docs/a-124.pdf
24. E. Hackmann, C. Lämmerzahl, Generalized gravitomagnetic clock effect. Phys. Rev. D 90(4),
044059 (2014)
25. J.M. Cohen, B. Mashhoon, Standard clocks, interferometry, and gravitomagnetism. Phys. Lett.
A 181, 353–358 (1993)
26. V.F. Fateev, S.M. Kopeikin, S.L. Pasynok, Effect of irregularities in the earth’s rotation on
relativistic shifts in frequency and time of earthbound atomic clocks. Meas. Tech. 58, 647–654
(2015)
27. M. Soffel, S.A. Klioner, G. Petit, P. Wolf, S.M. Kopeikin, P. Bretagnon, V.A. Brumberg, N. Cap-
itaine, T. Damour, T. Fukushima, B. Guinot, T.-Y. Huang, L. Lindegren, C. Ma, K. Nordtvedt,
J.C. Ries, P.K. Seidelmann, D. Vokrouhlický, C.M. Will, C. Xu, The IAU 2000 resolutions
for astrometry, celestial mechanics, and metrology in the relativistic framework: explanatory
supplement. Astron. J. (USA) 126, 2687–2706 (2003)
28. M. Soffel, S. Kopeikin, W.-B. Han, Advanced relativistic VLBI model for geodesy. J. Geod.
91(7), 783–801 (2017)
29. V.A. Brumberg, S.M. Kopeikin, Relativistic equations of motion of the earth’s satellite in the
geocentric frame of reference. Kinematika i Fizika Nebesnykh Tel 5, 3–8 (1989)
30. V.A. Brumberg, S.M. Kopejkin, Relativistic reference systems and motion of test bodies in the
vicinity of the Earth. Nuovo Cim. B Ser. 103, 63–98 (1989)
31. T. Damour, M. Soffel, C. Xu, General-relativistic celestial mechanics. IV. Theory of satellite
motion. Phys. Rev. D 49, 618–635 (1994)
32. A. San Miguel, Numerical integration of relativistic equations of motion for Earth satellites.
Celest. Mech. Dyn. Astron. 103, 17–30 (2009)
33. U. Kostić, M. Horvat, A. Gomboc, Relativistic positioning system in perturbed spacetime.
Class. Quantum Gravity 32(21), 215004 (2015)
34. K.-M. Roh, B.-K. Choi, The effects of the IERS conventions (2010) on high precision orbit
propagation. J. Astron. Space Sci. 31, 41–50 (2014)
35. K.-M. Roh, S.M. Kopeikin, J.-H. Cho, Numerical simulation of the post-Newtonian equations
of motion for the near Earth satellite with an application to the LARES satellite. Adv. Space
Res. 58, 2255–2268 (2016)
36. I. Ciufolini, E.C. Pavlis, A confirmation of the general relativistic prediction of the Lense-
Thirring effect. Nature 431, 958–960 (2004)
37. I. Ciufolini, Dragging of inertial frames. Nature 449, 41–47 (2007)
226 S. Kopeikin
38. I. Ciufolini, E.C. Pavlis, A. Paolozzi, J. Ries, R. Koenig, R. Matzner, G. Sindoni, K.H. Neu-
mayer, Phenomenology of the lense-thirring effect in the solar system: measurement of frame-
dragging with laser ranged satellites. New Astron. 17, 341–346 (2012)
39. V.G. Gurzadyan, I. Ciufolini, A. Paolozzi, A.L. Kashin, H.G. Khachatryan, S. Mirzoyan, G.
Sindoni, Satellites testing general relativity: residuals versus perturbations. Int. J. Mod. Phys.
D 26, 1741020 (2017)
40. K.S. Thorne, R.D. Blandford, Black holes and the origin of radio sources, in Extragalactic
Radio Sources. IAU Symposium, vol. 97, ed. by D.S. Heeschen, C.M. Wade (1982), pp. 255–
262
41. S. Chandrasekhar, Ellipsoidal Figures of Equilibrium (Yale University Press, New Haven,
1969), ix+252 pp
42. A.N. Petrov, S.M. Kopeikin, R.R. Lompay, B. Tekin, Metric Theories of Gravity: Perturbations
and Conservation Laws (De Gruyter, Berlin, 2017), xxiv+597 pp
43. L. Blanchet, Gravitational radiation from post-Newtonian sources and inspiralling compact
binaries. Living Rev. Relativ. 17, 2 (2014)
44. M. Soffel, F. Frutos, On the usefulness of relativistic space-times for the description of the
Earth’s gravitational field. J. Geod. 90(12), 1345–1357 (2016)
45. D. Philipp, V. Perlick, D. Puetzfeld, E. Hackmann, C. Lämmerzahl, Definition of the relativistic
geoid in terms of isochronometric surfaces. Phys. Rev. D 95(10), 104037 (2017)
46. M. Oltean, R.J. Epp, P.L. McGrath, R.B. Mann, Geoids in general relativity: geoid quasilocal
frames. Class. Quantum Gravity 33(10), 105001 (2016)
47. K.D. Krori, P. Borgohain, Uniform-density cold neutron stars in general relativity. J. Phys.
Math. Gen. 8, 512–520 (1975)
48. J. Ponce de León, Fluid spheres of uniform density in general relativity. J. Math. Phys. 27,
271–276 (1986)
49. L. Lindblom, Static uniform-density stars must be spherical in general relativity. J. Math. Phys.
29, 436–439 (1988)
50. J.N. Islam, Rotating Fields in General Relativity (Cambridge University Press, Cambridge,
1985), 127 pp
51. E. Gourgoulhon, An introduction to the theory of rotating relativistic stars. Lectures Given at
the Compstar 2010 School (Caen, 8–16 Feb 2010) (2010)
52. J.L. Friedman, N. Stergioulas, Rotating Relativistic Stars (Cambridge University Press, Cam-
bridge, 2013), 438 pp
53. S. Chandrasekhar, The post-Newtonian effects of general relativity on the equilibrium of uni-
formly rotating bodies. I. The Maclaurin spheroids and the virial theorem. Astrophys. J. 142,
1513–1518 (1965)
54. S. Chandrasekhar, The post-Newtonian effects of general relativity on the equilibrium of uni-
formly rotating bodies. II. The deformed figures of the Maclaurin spheroids. Astrophys. J. 147,
334–352 (1967)
55. N.P. Bondarenko, K.A. Pyragas, On the equilibrium figures of an ideal rotating liquid in the
post-Newtonian approximation of general relativity. II: Maclaurin’s P-ellipsoid. Astrophys.
Space Sci. 27, 453–466 (1974)
56. D. Petroff, Post-Newtonian Maclaurin spheroids to arbitrary order. Phys. Rev. D 68(10), 104029
(2003)
57. G.L. Clark, The gravitational field of a rotating nearly spherical body. Philos. Mag. 39(297),
747–778 (1948)
58. P. Teyssandier, Rotating stratified ellipsoids of revolution and their effects on the dragging of
inertial frames. Phys. Rev. D 18, 1037–1046 (1978)
59. H. Cheng, G.-X. Song, C. Huang, The internal and external metrics of a rotating ellipsoid under
post-Newtonianian approximation. Chin. Astron. Astrophys. 31, 192–204 (2007)
60. C.M. Will, Theory and Experiment in Gravitational Physics (Cambridge University Press,
Cambridge, UK, 1993), xi+396 pp
61. S. Kopeikin, I. Vlasov, Parametrized post-Newtonian theory of reference frames, multipolar
expansions and equations of motion in the N-body problem. Phys. Rep. 400, 209–318 (2004)
62. M.H. Soffel, Relativity in Astrometry, Celestial Mechanics and Geodesy (Springer, Berlin,
1989), xiv+208 pp
63. V.A. Brumberg, Essential Relativistic Celestial Mechanics (Adam Hilger, Bristol, 1991), x+263
pp
64. A.D. Rendall, Convergent and divergent perturbation series and the post-Minkowskian approx-
imation scheme. Class. Quantum Gravity 7(5), 803–812 (1990)
65. J. Müller, M. Soffel, S.A. Klioner, Geodesy and relativity. J. Geod. 82, 133–145 (2008)
66. V.A. Fock, The Theory of Space, Time and Gravitation, 2nd edn. (Macmillan, New York,
1964); Translated from the Russian by N. Kemmer, xii+448 pp
67. S. Weinberg, Gravitation and Cosmology: Principles and Applications of the General Theory
of Relativity (Wiley, New York, 1972)
68. A.P. Lightman, W.H. Press, R.H. Price, S.A. Teukolsky, Problem Book in Relativity and Grav-
itation (Princeton University Press, Princeton, 1975), xiv+603 pp
69. G.B. Arfken, H.J. Weber, Mathematical Methods for Physicists, 4th edn. (Academic Press, San
Diego, 1995), xviii+1029 pp
70. E.W. Hobson, The Theory of Spherical and Elliptical Harmonics (Cambridge University Press,
Cambridge, 1931), vi+500 pp
71. V. Pohánka, Gravitational field of the homogeneous rotational ellipsoidal body: a simple deriva-
tion and applications. Contrib. Geophys. Geod. 41, 117–157 (2011)
72. I.S. Gradshteyn, I.M. Ryzhik, in Table of Integrals, Series and Products, 4th edn., ed. by Y.V.
Geronimus, M.Y. Tseytlin (Academic Press, New York, 1965); First appeared in 1942 as MT15
in the Mathematical tables series of the National Bureau of Standards
73. I. Ciufolini, J.A. Wheeler, Gravitation and Inertia (Princeton University Press, Princeton,
1995), 512 pp
74. S.M. Kopeikin, Gravitomagnetism and the speed of gravity. Int. J. Mod. Phys. D 15, 305–320
(2006)
75. S.M. Kopeikin, The gravitomagnetic influence on Earth-orbiting spacecrafts and on the lunar
orbit, in General Relativity and John Archibald Wheeler, vol. 367, Astrophysics and Space
Science Library, ed. by I. Ciufolini, R.A.A. Matzner (Springer, Berlin, 2010)
76. B.H. Hager, M.A. Richards, Long-wavelength variations in Earth’s geoid - physical models
and dynamical implications. Philos. Trans. R. Soc. Lond. Ser. A 328, 309–327 (1989)
77. J.-L. Tassoul, Theory of Rotating Stars (Princeton University Press, Princeton, 1979), xvi+508
pp
78. P. Pizzetti, Principii della teoria meccanica della figura dei pianeti (E. Spoerri, Pisa, 1913),
xiii+251 pp
79. R.O. Hansen, Multipole moments of stationary space-times. J. Math. Phys. 15, 46–52 (1974)
80. K.S. Thorne, Multipole expansions of gravitational radiation. Rev. Mod. Phys. 52, 299–339
(1980)
81. L. Blanchet, T. Damour, Radiative gravitational fields in general relativity. I - general structure
of the field outside the source. Philos. Trans. R. Soc. Lond. Ser. A 320, 379–430 (1986)
82. H. Quevedo, Multipole moments in general relativity - static and stationary vacuum solutions.
Fortschritte der Physik 38, 733–840 (1990)
83. T. Damour, B.R. Iyer, Multipole analysis for electromagnetism and linearized gravity with
irreducible Cartesian tensors. Phys. Rev. D 43, 3259–3272 (1991)
84. L. Blanchet, On the multipole expansion of the gravitational field. Class. Quantum Gravity 15,
1971–1999 (1998)
85. C. Jekeli, The exact transformation between ellipsoidal and spherical harmonic expansions.
Manuscr. Geod. 13, 106–113 (1988)
86. C. Snow, Hypergeometric and Legendre Functions with Applications to Integral Equations of
Potential Theory, 2nd edn. (US Government Printing Office, Washington, 1952), xi+427 pp
87. M. Soffel, R. Langhans, Space-Time Reference Systems (Springer, Berlin, 2013), xiv+314 pp
88. S. Chandrasekhar, J.C. Miller, On slowly rotating homogeneous masses in general relativity.
Mon. Not. Roy. Astron. Soc. 167, 63–80 (1974)
228 S. Kopeikin
89. J.M. Bardeen, A reexamination of the post-Newtonian Maclaurin spheroids. Astrophys. J. 167,
425 (1971)
90. R. Meinel, M. Ansorg, A. Kleinwächter, G. Neugebauer, D. Petroff, Relativistic Figures of
Equilibrium (Cambridge University Press, Cambridge, 2008), p. ix+218 pp
91. H. Stephani, D. Kramer, M. MacCallum, C. Hoenselaers, E. Herlt, Exact Solutions of Einstein’s
Field Equations, 2nd edn. (Cambridge University Press, Cambridge, 2003), xxix+701 pp
92. W.C. Hernandez, Material sources for the Kerr metric. Phys. Rev. 159, 1070–1072 (1967)
93. J.L. Hernandez-Pastora, L. Herrera, Interior solution for the Kerr metric. Phys. Rev. D 95(2),
024003 (2017)
94. C. Jiang, W. Lin, Harmonic metric for Kerr black hole and its post-Newtonian approximation.
Gen. Relativ. Gravit. 46, 1671 (2014)
95. W. Lin, C. Jiang, Exact and unique metric for Kerr-Newman black hole in harmonic coordinates.
Phys. Rev. D 89(8), 087502 (2014)
96. H. Essén, The physics of rotational flattening and the point core model. Int. J. Geosci. 5,
555–570 (2014)
97. A. Krasinski, Ellipsoidal space-times, sources for the Kerr metric. Ann. Phys. 112, 22–40
(1978)
98. T. Wolf, G. Neugebauer, About the non-existence of perfect fluid bodies with the Kerr metric
outside. Class. Quantum Gravity 9, L37–L42 (1992)
99. C. Bambi, Black Holes: A Laboratory for Testing Strong Gravity (Springer, Singapore, 2017),
xv+340 pp
Anholonomity in Pre-and Relativistic
Geodesy
Erik W. Grafarend
Abstract I was invited to speak about anholonomity or the problem to find co-
ordinate reference systems which are differentiable. In general non-differentiable
functions like (pseudo) orthonormal reference systems are differentiable forms be-
ing not classical functions. These differentiable forms are the basis of Elie Cartan’s
“exterior calculus”. Geodetic examples are extensively reviewed in the context of
the pre-and relativistic Geodesy.
1 Motivation: Anholonomity and the Axisymmetric

Gravity Field
Geometric Geodesy as well as Physical Geodesy in spacetime are the fundament

of Geodetic Science. From the beginning “Relativistic Positioning” as well as the
“Somigliana–Pizzetti Gravity Field”- since 1930 the legal official IAG reference
gravity field (International Association of Geodesy) - now the “Kerr metrical rela-
tivistic Gravity Field” - my proposal to the IAG, Member of the International Union
of the Geodesy and Geophysics of ICSU - in spacetime played a dominant role: it
balances the gravitational field and the rotational field in a unique way. It takes into
account that the planet Earth and all other terrestrial planets/moons rotate.
From the birth of Geodetic Science, anholonomity of Geodetic Reference Frames
was a central topic. It is the problem of integration and differentiation of special
geodetic differentiable forms which generalize the notion of functions. In school
we learn differentiation and integration, but not differential forms, special func-
tions which cannot be integrated in a closed form. We have learned that mixed sec-
ond order differentials of functions commute, for instance, f i j = f ji = 0, namely
f 12 : ∂ 2 f /∂ x 1 ∂ x 2 , f 21 =: ∂ 2 f /∂ x 2 ∂ x 1 . It might be a surprise for Geodesists and
Physicists that this conditions is not fulfilled, in general. Indeed there are many
E. W. Grafarend (B)
Department of Geodesy and Geoinformatics, Faculty of Aerospace Engineering and Geodesy,
Faculty of Mathematics and Physics, Geschwister Scholl Strasse 24 D, 70174 Stuttgart, Germany
e-mail: grafarend@gis.uni-stuttgart.de

230 E. W. Grafarend
geodetic functions where this condition does not hold. For those functions -a type of
generalized functions- we note the characteristic property f i j − f ji = ωi j = 0.
A first geodetic example is the scalar-valued gravity potential W , the sum of the
gravitational potential U and the centrifugal potential, also called rotational potential
V . It relates to the differential dW of the gravity potential W called holonomic po-
tential to the height displacement dH - γ and the modulus of the gravity vector in
2
- γ
terms of dW = 2 dH . We have used here the notation of M. Planck document-
ing that “d”- is not a “perfect differential”, also called anholonomic coordinate. Such
a notion is plausible since dH - γ = 0 in a closed loop, also called ring integral. The
argument is that in a local geodetic reference frame we have to refer to the terrestrial
gravity field, a deeper argument we bring here to your attention.
“Imperfect differentials” gave raise to the notion of differential forms. Our exam-
ple refers to the differential 1-form. The modulus of the Cartesian coordinates 2
(2 norm) is called “Frobenius” integrating factor due to dH - γ = 0. Integrability
is a key property of classical functions. But p-forms are extensions towards an-
holonomity. Elie Cartan based his theory of “exterior calculus” on non-integrability.
Here we start with a review of anholonomity: before and after relativity. In Chap. 2
we begin with the geodetic inventor of generalized geodetic functions: F. R. Helmert.
Being a Geodesist, he had studied two years in Mathematics and Physics where he
took lectures at the University of Leipzig of F. G. Frobenius and H. Grassmann.
It was there that he learned about “integrability” and “differentiable forms”: he
showed that the gravity potential of rigid bodies is integrable, but geodetic heights
are not, in general. “Geodätische Kote” was his term used for potential heights.
In Chap. 2 we therefore review anholonomity, before and after Relativity. The two
reports of the Ohio State University (OSU), Department of Geodetic Science (1972,
1973) of the author build up the basis and applied to three-dimensional Geodesy
and gravity gradients. The highlights are (i) the indicator diagrams and (ii) Cartan’s
pseudo-torsion or anholonomity forms in the “natural reference system”, τi j k .
The mathematician J. Zund in his latest book [1] realized the importance of our an-
holonomity formulae. The five characteristic curvature parameters {k1 , k2 , t1 , κ1 , κ2 }
of Hotine-Marussi’s three-dimensional Geodesy were the highlights of contemporary
geodetic science. A detailed computation of the Frobenius matrix of integrating fac-
tors in terms of the five arbitrary curvature parameters was demonstrated for “natural
coordinates” holonomic { ,
, W } (astronomical longitude, astronomical latitude,
potential of gravity) and anholonomic Cartesian coordinates {X, Y, Z }. A special sec-
tion is concerned with the curvature (first curvature) and torsion (second curvature)
of field lines of the gravity field and its orthonormal “plumbline”. We document the
special role of the Marussi gauge emphasizing the central role of gauge theory, in gen-
eral. The duality between vertical and horizontal fields equipped with an orthonormal
(in Relativity “pseudo-orthonormal”) metric δi j leads to the second order differential
equation of the plumbline manifold in Marussi gauge. Basic is the proof that spher-
ical symmetric gravity fields lead to zero curvature/torsion or zero anholonomity or
“holonomic coordinates”, but not non-spherical, for instance axisymmetric gravity
field. Alternatively, a departure from non-spherical symmetry, conventionally called
“anormal potential”, “gravity disturbance”, “vertical deflections”- standard geodetic
Anholonomity in Pre-and Relativistic Geodesy 231
terms - lead to anholonomity as already stated by P. Teunissen [2]. Or standard spher-

ical /ellipsoidal series expansions to the order/degree 180/180, 360/360, or 720/720
up to 7200/7200 illustrate perfectly “anholonomity”. Chapter 3 introduces “Real null
frames and coframes” within General Relativity” in order to prove anholonomity.
We take reference to Blagojevic, M., Garecki, J., Hehl, F. W. and Obukhov, Y.N. [3].
More specifically, in pseudo-Riemannian spacetime we define a non-unique pseudo-
orthonormal reference frame indexed α, β, . . . , ∈ {1, 2, 3, 4}. We choose a timelike
4-leg by e4 and {e1 , e2 , e3 } spacelike 1, 2, 3-legs. Such a popular frame of refer-
ence is pseudo-orthonormal, but unfortunately anholonomic, in general. Its metric is
given by g(eα , eβ ) = diag(1, 1, 1, −1), namely locally Minkowski. With orthonor-
mal frames of reference we have previously experienced anholonomity: Please, study
the subject of “Null frames” called {l, m, n, m}.
First, surface geometry as a two-dimensional Riemann manifold we in-
tent to orthonormalize. We arrive at a non-unique orthogonal reference
frame, a special Cartan frame of reference. We have shown this when
we analyze an ellipsoidal frame of reference in E. Grafarend and F. W.
Krumm [4].
Second, in analyzing the Euler Kinematical equations as well as the
dynamical Euler equation for rigid bodies and the dynamical Euler-
Louisville equations in terms of Euler or Cartan angle we found the well
known anholonomity in terms of a Frobenius matrix. We take reference
to E. Grafarend and W. Kühnel [5].
Chapter 4 is specifying the notion Killing vectors of symmetry, namely for the
sphere (3 Killing vectors) and for the ellipsoid of revolution (1 Killing vector). It was
needed to understand spherical symmetry versus ellipsoidal symmetry, in particular
for the geodetic Somigliana–Pizzetti reference field.
Of particular importance is Chap. 5: we study the influence of local vertical nets
which cause indeed anholonomity due to the reference of the physical local vertical.
At the end we analyze the famous object of anholonomity in terrestrial networks,
namely an example of Cartan’s exterior calculus.
We conclude in Chap. 6 with special comments in the special role of anholonomity
for Geodesy, namely on the irregular boundary of the planet Earth. Our final highlight
is the literature list of geodetic contributions of the topic anholonomity, by no means a
forgotten subject. Finally we recommend to the International Association of Geodesy
to adopt the axisymmetric Kerr metric or its linear approximation Lense-Thirring for a
rotating - gravitating Earth, namely replacing the axisymmetric Somigliana–Pizzetti
reference field to include Relativistic Geodesy.
2 Anholonomity: Before and After Relativity
First, we study geodetic anholonomity as being established by Friedrich Robert

Helmert. He is the real founder of Physical Geodesy, for instance of gravimetry.
232 E. W. Grafarend
His concept of physical heights in terms of the gravity potential W , the sum of the
gravitational potential U and the centrifugal potential V , as part of the scalar-valued
gravity field established holonomic coordinates.
What are holonomic heights, better: holonomic height dif-
ferences, what are holonomic coordinates in contrast to
anholonomic ones? Such questions we will answer!
It is worth to study the arguments of F.R. Helmert why he transformed geometric
heights or height differences called “anholonomic” or not integrable in a closed loop
to “holonomic heights” or their differences in terms of potential numbers. Metrical
heights are constructed by dividing by a constant value, for instance the Global Mean
Value of Gravity (a proposal by S. Heitz) of by the external Somigliana–Pizzetti
gravity field, the standard gravity field known to sub-nano-Gal accuracy according
to A. Ardalan and E. Grafarend [6], also called “orthometric height” (F. Sanso and
P. Vanicek [7].
Born 31 July 1843 in Freiberg/Sachsen/F.R. Helmert attended in his home-
town the “Bürgerschule”, at the age of 14 he joined the “St. Annen Schule”.
in Dresden. The started his university studies at the age 16 years in joining the
“Polytechnical School” nowadays called “Technical University Dresden”. For
his Ph.D. studies he continued working in Dresden with A. Nagel as his super-
visor. As a benefit of his excellent studies a stipend was offered to him: he chose
to study two years at the University of Leipzig, namely to study physics and
mathematics. He attended the lectures of F.G. Frobenius and read H. Grassmann
on the topic of integrable differential forms, nowadays called “Cartan calculus”.
He finished his Ph.D. studies in Dresden: “Studium über rationelle Vermessun-
gen im Gebiet der Höheren Geodäsie”. In the year 1869 he was appointed as the
“observer” at the Hamburg Observatory. His first publication of the year 1874
dealt with “Vermessung und rechnerische Ausgleichung eines Sternhaufens”.
At the age of 29 years he was appointed “full professor” at the “Polytechnis-
che Schule” nowadays called “Aachen University”. It was in Aachen where he
wrote the legendary monumental works:
(i) Ausgleichungsrechnung nach der Methode der kleinsten Quadrate mit Anwen-
dungen auf die Geodäsie und die Theorie, 1872. (This basic work saw many
reprints within the 20th century.)
(ii) Die Mathematischen und Physikalischen Theorien der Höheren Geodäsie, 1880
and 1884, two volumes. (This basic works was often reprinted and translated in
the 20th century.)
F. R. Helmert was from the year 1886 up to his death in the year 1917 Director
of the “Königlich Preussischens Geodätischens Institut” which was founded
in the year 1870 in Berlin from 1890 onward up to this day located at the
Telegraphenberg in the city of Potsdam . At the same time he was appointed
Full Professor to the Chair of Geodetic Sciences at the “Friedrich-Wilhelms-
Universität” in Berlin. In the year 1900 he was elected to ordinary member at
the “Prussian Academy of Sciences”, a colleague of Albert Einstein in Berlin.
A lot of effects in Physical Geodesy bear his name:

• Helmert ellipsoid transformation
• Helmert projection
• Helmert level ellipsoid
• Helmert heights
• Helmert deflection of vertical
• Polar height variation
• Geodynamics
Here we specialize on “holonomic versus anholonomic heights” or the key ques-
tion:
Why is Geodesy part of Physics?
During his studies in Leipzig he became acquainted with “integrating factors” and
the celebrated Frobenius Lemma: Frobenius was teaching in Leipzig at that time.
Example:
- γ - γ
dW = − dH = 0 versus dH = 0
The notion “dH”- was introduced by M. Planck: the ring integral of geometric
-
heights dH is not zero, illustrated by “-” and by “dH” combined to dH. - We also say
“loop integral”. In contrast, the ring integral of the differential of “potential heights”
- the closed loop integral-is zero. W is called “the Gauss-Green potential”, later on
subject to “Cartan calculus” or “differential forms”.
In a co-rotating reference system-a system rotating with the Earth- we enjoy four
types of forces:
(i) Gravitational forces:

“conservative” derived from a scalar potential
(ii) Centrifugal force:
“conservative” derived from a scalar potential
(iii) Euler force:
“produced by angular momentum”, non-conservative
• responsible for “Polar Motion”(POM) and “Length-of-Day” variation
(LOD), or
• Precession-Nutation, vector valued force
(iv) Coriolis force:
“produced currents at Sea”, non-conservative
Summary
“Conservative”: field equations

234 E. W. Grafarend
(i) grad W = gradU + gradV

(ii) div grad W = −4π Gρ + 22
“Non-conservative”: field equations
r ot = 2
“for rigid bodies, similar to
Maxwell equations of a deforming
body”
What are our subjects?
At first we review the contents of my OSU reports (The Ohio State University
Columbus/Ohio/USA) from the year 1972 and 1973 of 81 pages and 126 pages on
anholonomity. Second, we shortly review my contributions on the subject of field
lines of gravity, their curvature and torsion, the Lagrange and the Hamilton equations
of the plumbline. Finally, third, we present examples of anholonomity.
2.1 Two Reports form the Ohio State University (OSU):

1972–1973
There are not too many books on the subject of Differential Geomathematics, Dif-
ferential Geometry, Differential Topology and Theoretical Physics where you study
“holonomity” and “anholonomity”. We advice the interested reader to consult J. A.
Schouten (1954: “Ricci Calculus”) [8] with an index on “anholonomity”) as well
as R. Cushman et al. [9] : “Geometry on non-holonomically constraint systems”,
World Scientific, New Jersey. My favorite is S. Sternberg [10]: read Chap. 6 on Car-
tan calculations in semi-Riemann geometry, in particular “frame fields and co-frame
fields”, as well as Chap. 15 on “The Frobenius theorem”, in particular pages 312–313
on “A dual formulation of the Frobenius theorem”. In this remarkable book you find
also a review of
(i) Relativity
(ii) Higgs et al.
A photo of F.G. Frobenius (1849–1917) is available on page
303 of the book by S. Sternberg [10].
My first report of the Department of Geodetic Science, No. 174, The Ohio
State University, Columbus/Ohio/USA, 81 pages, of the year 1972 treats “Three-
Dimensional Geodesy and Gravity Gradients” based on the Frobenius integration the-
orem. The anholonomity or non-integrability is due to measurements in the “natural”
or astronomic reference system. The characteristic Cartan object of anholonomity

depends on either the gradients of the terrestrial gravity field or their gradients. The
specific integration factors are determined, namely six integrating factors in the as-
tronomic reference frame or nine integrating factors in the geodesic reference frame.
They depend either on the gradients of the standard gravity field or on the Euclidean
norm of the gravity field/vertical deflections of the disturbing gravity field.
At this time, I expressed my opinion that with the availability
of gravity gradients would open a new era of three-dimensional
geodesy, namely developed by Antonio Marussi.
It makes sense to have a more detailed look into the subject of the first Ohio State
University Report (OSU report).
After Sect. 1, introduction and summary, Sect. 2 dealt with “The Forbenius In-
tegration Theorem”: review of the literature, H. Grassmann and exterior algebra,
forms, differential forms, exterior forms, E. Cartan or exterior analysis, H. Poincare
Lemma, G. Stokes Lemma G. Frobenius integration theorem, Pfaffian forms, Pfaffian
equation, integrating factors, examples, Sect. 3 with “Differential Geometry in the
Cartan Symbolism”: review of literature, vector, tensor, object, dual basis, holonomic
and anholonomic coordinate system, E. Cartan’s object of anholonomity, G. Ricci
rotation coefficients, G. Ricci, T. Levi-Civita, J. A. Schouten: parallel displacement,
structure equations, torsion and curvature forms, integrability conditions, E. Cartan’s
theorems de conservations de la courbure et de la torsion as well a Sect. 4: “The As-
tronomic Coordinate System and The A. Marussi Integration Theorem”: Review of
literature, astronomic reference system, geocentric and local astronomic coordinates,
E. Cartan’s object of anholonomity and the G. Frobenius theorem, A. Marussi inte-
gration theorem, gravity gradients. Sect. 5 concentrates on “The Geodetic Coordinate
System”: geodetic coordinate system, E. Cartan’s object of anholonomity and the
G. Frobenius Theorem applied to geodetic coordinate systems, integration theorem,
gravity gradients. The important transformation of one reference system to another
is dealt with in Sect. 6: “Transformation Between Astronomic and Geodetic Coordi-
nate System”: Transformation matrices, E. Cartan’s object of anholonomity, vertical
deflection vector, astronomical and geodetic azimuth, astronomical and geodetic
vertical angles, generalized Laplace condition, integration theorem. Finally, Sect. 7
enjoys “The Geodetic Integration Theorem and the Concept of Boundary Value
Problems”: G. Frobenius integration theorem applied to geodesy, integrating factors,
gravity gradients, geodetic boundary value problem, free boundary value problems,
pseudo-differential operators. Finally, Sect. 8 lists 141 references, namely for rela-
tivists: F. Hehl [11] as well as (1970) [12]. Finally the joint work of F. Hehl and
E. Kröner [13], had to acknowledge! E. Kröner had been quoted by 4 publications,
partially in books. (Both of them were my supervisors in my studies on Theoretical
Physics at Clausthal Techn. University.)
My second report of the Department of Geodetic Science, No. 202, The Ohio
State University, Columbus/Ohio/USA, 126 pages, of the year 1973 treats “Gravity
Gradients and Three-Dimensional Net Adjustments without Ellipsoidal Reference”,
the practical applications of the Marussi transformation form “natural” non-unique
236 E. W. Grafarend
coordinates into the unique coordinates astronomical longitude and latitude, and the
gravity potential by means of gravity gradients as well as the Euclidean norm of
the gravity vector as integrating factors. If we tolerate point errors of about ±10 m,
the r.m.s. error of gravity gradients for terrestrial data is found to be about ±1 E
(“Eötvös”) for point distances of about 15 km and pole free latitudes. Analogous
values are ±1 m and ±0.1 E, ±10 cm and ±0.01 E approximately. A detailed error
analysis for the Marussi transformation is given, completed by an example for a three-
dimensional net adjustment based on the Frobenius theorem and available terrestrial
gravity gradient information.
Here, we review only the first sections of the OSU report number 202, namely
§2: Geodesy, general consideration

and
§3: Geodesy being properly posed,
only indicator diagrams, integrability and its five parameters {k1 , k2 , t1 , κ1 , κ2 } as

integrating factors.
The First Example:

In the beginning or 2d-Riemann geometry Geodesy was two-dimensional surface
geometry being embedded into three dimensional Euclidean space. Gravity was a
new item to Geodesy due to anholonomity! The Science of Geodesy introduces a
new concept into geometry, that of gravitation: it is absent from classical geometry.
Our sense of gravitation furnishes a qualitative measure and a quantitative measure,
albeit fortuitous to a certain extent, it is given by any balance. In I. Newton’s theory,
gravity is described by a potential W as the sum of (i) gravitational potential (U) and
of (ii) centrifugal potential (V) as the scalar potential part of gravity, the gradient of
which, taken negative, it is the acceleration imparted to a small test-body.
The gravitational potential is a property or parameter-of-state. it is associated with
the behavior of the body at the instant under consideration, or else, it is measured
with reference to the instantaneous indication of balance or gravimeter.
In order to give a rigorous mathematical definition of the gravitational “property”
or “parameter-of-state”, it is necessary to consider an example, with two independent
variables x1 , y1 , which must be measurable properties or characteristics of the system,
we can write
(i) dW = X d x + Y dy subject to (2.1)
∂W ∂W
(ii) X = and Y = (2.2)
∂x ∂y
Evidently, we then have the conditions
∂X ∂Y ∂2W ∂2W
(iii) = or = (2.3)
∂y ∂x ∂ y∂ x ∂ x∂ y
which is the necessary and sufficient condition for the expressions X d x + Y dy to

be perfect differential. It is equivalent to the statement that W is a perfect property.
The same condition can be written in integral form

(iv) dW = 0 (2.4)
for any closed path in the x, y-plane. Denoting the two-dimensional vector which is
defined by its components X and Y any symbol, we can apply G. Stokes theorem to
the scalar product of a two dimensional field, obtaining

∂Y ∂X
(v) dW = X d x + Y dy = − d xd y = 0 (2.5)
∂x ∂y

Since the rotor or curl ∂Y
∂x
− ∂∂Xy vanished, it is concluded that the statement dW is,
in fact, equivalent to the assumption that W is a property.
The condition for a perfect differential with n-independent variable is the van-
ishing of the n-dimensional operator “rot” or “curl” and can be represented by
n(n − 1) ÷ 2 equation of the above form. A differential of the type X d x + Y dy
is known as J. Pfaff’s differential, in general. When there are two independent vari-
able it is always possible to transform the expression X d x + Y dy into a perfect
differential by dividing it by a denominator, even it it was not one originally. With
three independent variables x, y, z and a vector differential form, the theory is more
complex. Certain requirement of integrability impose certain conditions on the vector
components. They will be found by the G. Frobenius theorem.
The integral dW = 0 for any closed path can be represented graphically in a
plane if the gradient gradW is taken along the height line. dW = −d H including
the pair acceleration = gradW and geometric height H - a mechanical and a
physical - is used in the indicator diagram, introduced by J. Watt for thermodynamics.
The path of the “L.N. Carnot” cycle 1, 2, 3 and 4 consists of four individuals. In
the left diagram of Fig. 1 12 = 34 , g23 = 41 , H12 = −H34 and H23 = −H41 must
hold, thus 12 H12 + 23 H23 + 34 H34 + 41 H41 results zero. The paths 23 and 41
are isogravitational in the central diagram of Fig. 1. In addition H12 = −H34 must
be provided. In the right diagram of Fig. 1 the path 12 and 34 are isophysical. But
23 = 41 has to be assumed, otherwise the cycle is nonintegrable. Now it is quite
obvious that d H γ = 0.
Fig.
1 Indicator
diagram for
- γ =0
d W = dH
238 E. W. Grafarend
Fig.
- 2γ Indicator
diagram for
dH = −1 d W = 0 1
Γ
is the well known integrating factor for height. In the next paragraph we prove the
non-integrability for the position coordinates also. We keep in mind that the indicator
diagram for an integrable pair of coordinates is perfectly symmetric (Fig. 2).
Geodesy: Properly posed.
Geodetic instruments (theodolites, leveling instruments, electronic distance mea-

surement equipment, satellite cameras) operate in special coordinate systems which
we will shortly review. We will find that it is impossible to construct a unique coordi-
nate system if we measure only directions, angles, and distances. The nonuniqueness
is caused by the Earth’s gravity field, influencing the levels of the instruments, and
the astronomic orientation of the coordinate axes; Geodesy is “an improperly posed
problem”.
In order to make geodesy properly posed we have to apply modern techniques
of differential geometry. Why is differential geometry such an important tool? In
general, the main idea of differential geometry is to apply the tools of analysis to
the solution of geometrical problems. An important device in analysis is to study
geometrical objects by studying their “infinitesimal parts”. Thus a curve is studied
by examining its tangent vectors, a function by studying its differentials, and so
on. The crucial advantage of the passing to the “infinitesimal” is that it linearizes
everything. Thus a curve is “infinitesimally” a straight line when we look at the
tangent, every differentiable map is “infinitesimally” linear when we look at its
Jacobian matrix,. etc. Since we will be interested in studying such objects as curves,
surfaces,. and in general, higher dimensional “differentiable objects”, we first study
their “infinitesimally” analogues.
From an advanced geometric point of view the “geodetic” situation is typical for
a nonsymmetric E. Cartan world with non vanishing pseudo-torsion. The originally
improperly posed problem of Geodesy (non uniqueness of “natural” coordinates) is
regulated by the G. Frobenius integration theorem via the A. Marussi tensor of gravity
gradients. By the method of integrating factors we have a bijective transformation
from the non-unique natural coordinate system into the unique coordinate system of
Physical Geodesy.
On the way of regularization we will lose some pure geometric interpretations of
geodetic coordinate systems because one of three unique coordinates is the Earth’s
gravity potential at its surface. But
this fact is typical for regularization methods.
The question whether or not dx is integrable is crucial. We refer to “natural”
geodetic coordinates in a local astronomic triad, oriented by the orthonormal basis
system e• at any point of the Earth’s surface. The 3-axis is the antidirection of the
Fig. 3 Local astronomic

triad
local gravity vector , the 1-axis and the 2-axis are directed towards astronomic
East and astronomic North. “•” reminds to the astronomic directions. Obviously the
3-axis coincides with the vertical axis of a perfectly leveled theodolite. Well known
are the astronomic azimuth, the vertical angle and the “infinitesimally” Euclidean
distance between two surface points as local polar coordinates. Any misclosure of
a ring polygon dx (vector differential one-form) will decide about uniqueness or
nonuniqueness of these “natural” coordinates. The coordinates are taken along the
local basis vectors e• .
•
dx = eW −E ωW −E + e S−N ω S−N + eV ωV (2.6)
dx = e1 ω1 + e2 ω2 + e3 ω3 (2.7)
eW −E represents the East basis vector, e S−N the North basis vector, and eV the
vertical basis vector (Fig. 3).
•
Orthonormality give (ei , e j ) = δi j , ωi = (dx, ei ). (, ) indicates scalar product.
Within the differential one-form dx we apply “G. Stokes’identity”, due to

dx = (e, ω) = d(e, w) = (e, τ ) = 0 (2.8)

The integral (e, τ ) has to be taken over the surface included by the ringpolygon of the
curve C. τ is the E. Cartan pseudo-torsion, often called the object of anholonomity,
and responsible for the misclosure of the closed path relative to the natural triad
(Fig. 4).
In the kernel-index G. Ricci calculus G. Stokes’ Theorem reads
j
dxi = eij ω j = d f mn em enk ∂ [ j ek]
i
(2.9)
240 E. W. Grafarend
Fig. 4 Line integration
where we have not separated “dead” and “living” tensor indices. [] are the per-
mutation brackets, e.g. [a, b] = ab − ba; ∂i the analytical form of grad; d f mn the
antisymmetric surface element.
j
I proved in (1972) that the pseudo-torsion form τ.mn i
:= em enk ∂ [ j ek]
i
(we apply the
Einstein summation convention over repeated indices, running 1,2,3) does not vanish
relative to the natural basis and is due to Box 1. In order to compute the object of
anholonomity (nonintegrabilitly, nonuniqueness) τ := (e• , (e• , [grad, e• ])) we have
to differentiate the components of the triad and to “strangle” twice by the same basis
vectors. The specialist in differential geometry can speculate that the pseudo-torsion
form should depend on curvature and torsion of the surface. Thus we are led to the five
characteristic quantities k1 , k2 , t1 , κ1 and κ2 which we are going to explain carefully.
•
, resp. refer to the astronomic longitude, latitude resp. The sign = emphasizes
that the right hand side equation holds only in the reference “•”.
Box 1: E. Cartan’s pseudo-torsion in the natural reference system. Grafarend

representation
⎡ ⎤
0 −k1 tan
k1
∗
τi j 1. = 21 ⎣ k1 tan
0 −κ1 tan
+ t1 ⎦
−k1 κ1 tan
− t1 0
⎡ ⎤ ⎡ ⎤
0 −t1 tan
κ1 tan
0 0 −κ1
∗ ∗
τi j 2. = 21 ⎣ t1 tan
0 k2 ⎦ τi j 3. = 1 ⎣ 0 0 −κ2 ⎦
2
−κ1 tan
−k2 0 κ1 κ2 0
End Box 1
κ1 and κ2 are the projections of the gradient grad lng in the East and North
direction. We denote gradW = where W is the real gravity potential at the
Earth’s surface (Fig. 5). The East and North components of −1 grad are
responsible for the misclosure of the third component, the heights. If and only if
const., the heights are integrable. = const. leads to κ1 = 0 and κ2 = 0. Or we
say that the vertical component of the pseudo-torsion form.
The projections κ1 , κ2 and t1 .
κ1 and κ2 are the normal curvatures in the East and North directions of any equipo-
tential surface W passing through the observation point. t1 is the geodetic torsion
Fig. 5 The projections

κ1 , κ2
κ
κ
Fig. 6 The projections

k1 , k2
of the equipotential surface in the East direction. Let us transform the vertical basis
vector e3• in the horizontal plane by grad e3• . The projections of grad e3• onto the bases
e1• and e2• are the components κ1 and κ2 (Fig. 6).
(e1• , grad e3• ) = κ1 e1• (2.10)
(e2• , grad e3• ) = κ2 e2• (2.11)
The procedure of normalization leads to
e1• , (e1• , grad e3• ) = k1 (e1• , e1• ) = κ1 (2.12)
e2• , (e2• , grad e3• ) = k2 (e2• , e2• ) = κ2 (2.13)

242 E. W. Grafarend
The relation to the pseudo-torsion form of the type (e• , (e• , grad e• )) and the
projection κ1 and κ2 of the type (e• , (e• , grad e• )) is quite obvious 1 ÷ κ1 and 1 ÷ κ2
can be interpreted as the curvature radii in the East and North direction. The geodetic
torsion found by
(e1• , (e2• , grad e3• )) = t1
we got a better inside in the structure of the E. Cartan pseudo-torsion if we look upon
gravity gradients.
Gravity gradients
We use the symbol ∇i called the covariant derivative. By definition the tensor of
gravity gradients is
∇∇W = M, ∇i ∇ j W = Mi j
The symbol M is applied in honor of A. Marussi [14] who first noticed the
•
central role of gravity gradients for Geodesy. Introducing −∇W = = e3 the
parameters of the tensor of gravity gradients depend only on ∇ and ∇e3• .
• •
−∇ ∇W = ∇( e3 ) = (∇ )e3 + ∇e3
•
(e1• , (e1• , M)) = − k1 (2.14)
•
(e2• , (e2• , M)) = − k2 (2.15)
•
(e1• , (e2• , M)) = − t1 (2.16)
•
(e1• , (e3• , M)) = − κ1 (2.17)
•
(e2• , (e3• , M)) = − κ2 (2.18)
With the definitions of the five projections k1 , k2 , κ1 , κ2 and t1 the relation to the
tensor M of gravity gradients are quite obvious.
If we recall the coordinates along the local natural triad x, y, z the five character-
istic curvature parameters are given in the form
• 1 ∂2W 1
k1 = − = − Wx x (2.19)
∂ x 2
• 1 ∂2W 1
k2 = − = − W yy (2.20)
∂ y 2
• 1 ∂2W 1
t1 = − = − Wx y (2.21)
∂ x∂ y
• 1 ∂2W 1
κ1 = − = − Wx z (2.22)
∂ x∂z
• 1 ∂2W 1
κ2 = − = − W yz (2.23)
∂ y∂z
Thus the misclosure of a ringpolygon is due to

•
d x = + d x d y −1 Wx x tan
− d x dz −1 Wx x
(2.24)
− dy dz(+ −1 Wx z tan
+ −1 Wx y ) = 0
•
dy = + d x d y −1 Wx y
+ dy dz −1 Wx z tan
(2.25)
− dy dz −1 W yy = 0
•
dz = − dy dz −1 Wx z
(2.26)
− dy dz −1 W yz = 0
•
We had discussed the role of the horizontal gradient of Wz = g related to the third
component z, the height H . Here we have the influence of the other components of
the gravity gradient tensor also. Their non-null causes the misclosure of the position
coordinates x and y. Thus d x, dy, dz are imperfect differentials in the sense of J.
Pfaff. The integration of coordinates between two surface points is path dependent.
Therefore we cannot find unique coordinate differences between these two points.
The way of regularization of the a priori improperly posed problem (violation of
uniqueness) is this. We have to find a coordinate system in which the pseudo-torsion
form τ vanishes. This is a necessary and sufficient condition for applying the G.
Frobenius Theorem in order to find integrating factors and a transformation from
the a priori non-unique system to a unique one. A. Marussi [14] introduced for this
reason a non-orthogonal triad by
∂x ∂x ∂x
e1 = ∂
, e2 = ∂
, e3 = ∂W
,
to which we refer with the symbol e•• . Of course, this is by definition a gradient
field, an integrable system. The second derivative permute, for instance
• •
∂
e1 − ∂ e2 = (∂
∂ − ∂ ∂
)x = 0 (2.27)
Thus a properly posed “geodetic” problem consists of coordinates ,

, W which
have perfect differential d , d
, dW

d = 0, d
= 0, dW = 0 (2.28)
244 E. W. Grafarend
Fig. 7 Local regularized

triad
Unique “state variables” are the astronomic longitude, the astronomic latitude, and
the geopotential (Fig. 7).
What are the connections between the nonunique natural coordinates which we
measure in and the unique ones? That is the question how to calculate d , d
, , dW
out of d x, dy, dz, or what is the G. Frobenius matrix, of integrating factors. From
my report (1972d, p. 38) we take the transformation
⎡ ⎤
−k1 sec
−t1 0
e• = ⎣ −t1 sec
−k2 0⎦ (2.29)
−κ1 sec
−κ2
•
d = −k1 sec
d x − t1 sec
dy + κ1 sec
dz (2.30)
•
d
= −t1 d x − k2 dy + κ2 dz (2.31)
•
dW = dz (2.32)
where we have used the inverse of the G. Frobenius matrix of rank 3. The derived
transformation formulas are the most important ones of Physical Geodesy. They
explain why natural coordinate differentials are non-integrable and non-unique. The
gravity field of the Earth is involved in the transformation matrix. That explains
why geodesy has to be Physical Geodesy. Furthermore we can analyze when the
influence of gravity is excluded. This is the situation if and only if the whole tensor
of gravity gradients vanishes, emphasizing the central role of gravity gradients. In
other words, Gravity gradients are responsible for the non-uniqueness of natural
coordinates which we determine by measuring distances, vertical and horizontal
angles. Finally we rewrite the transformation in terms of gravity gradients explicitly,
• 1 1 1
d = + Wx x sec
d x + Wx y sec
dy + Wx z sec
dz (2.33)

• 1 1 1
d
= + Wx y d x + W yy dy + W yz dz (2.34)

•
dW = −dz (2.35)
firstly derived by A. Marussi [14]. What is the useful result for practice?
Firstly, we can uniquely integrate the coordinate differentials between two points
since d , d
, dW are perfect differentials. For instance the coordinate difference
between two points P and Q read
Q Q Q Q
d = Q − P = −1 Wx x sec
dx + −1 Wx y sec
dy + −1 Wx z sec
dz
P P P P
(2.36)
Q Q
d
=
Q −
P = −1 Wx y dx + −1 W yy dy + −1 W yz dz (2.37)
P P
Q Q
d W = WQ − WP = dz (2.38)
P P
We repeat that
Q Q Q
dx = x Q − x P , dy = y Q − y P , dz = z Q − z P (2.39)
P P P
Secondly, instead of measuring astronomic longitudes and latitudes, for a suf-

ficient accuracy nearly impossible on oceans, we can calculate longitude and lati-
tude differences if we measure distances, horizontal and vertical angles and gravity
gradients. To get a better understanding for d x, dy, dz let us introduce local polar
coordinates that is the azimuth A, the vertical angle B and the distance s (Fig. 8).
•
x = s sin A sin B
•
y = s cos A sin B
•
z = s cos B
It is well known that longitude, latitude, azimuth and vertical angles are not
independent.
d A = d sin
+ cot B(d
− d cos
cos A) (2.40)
246 E. W. Grafarend
Fig. 8 Azimuth A, vertical

angle B, distance s
d B = −d cos
sin A − d
cos A (2.41)
If ds is not the geometrical path, but the optical one, we have to replace ds by n
d s where n is the actual refractive index of the observation line.
Indicator Diagrams:

The indicator diagram for dW = dz = 0 was given in Fig. 9. Now we are
going to represent graphically the same diagrams for d = 0 and d
= 0. These
diagrams are multidimensional because d and d
are functions of dx, dy, dz. The
adjoint pairs of variables
−1 Wx x sec
, x; −1 Wx y sec
, y; −1 Wx z sec
, z;
−1 W yx , x; −1 W yy , y; −1 W yz , z;
, z
connect one physical and one mechanical variable. Their indicator diagram, is six-
dimensional, six-dimensional, and two-dimensional resp. We would like to mention
the fact that the missing Wzz inside the integrating factors is not astonishing. At
the Earth’s surface the Poisson-Laplace equation div grad W = tr M = 2ω2 holds,
explaining that Wx x and W yy are necessary and sufficient to describe Wzz = −(Wx x +
W yy ) + 2ω2 . ω represents the mean rotation speed of the Earth. Therefore only the
five independent components Wx x , Wx y , Wx z , W yy and W yz build the G. Frobenius
matrix of integrating factors.
In order to get some insight into the multidimensional indicator diagrams let us
assume Wx y = Wx z = W yz = 0. This means vanishing off-diagonal elements in the
A. Marussi tensor of gravity gradients or vanishing torsion t1 = 0 and projections
κ1 = κ2 = 0. In this case = const. and the E. Cartan pseudo-torsion results zero.
Thus the height component dz = dh is integrable and unique. Again this fact em-
phasizes the value of our knowledge about the E. Cartan pseudo-torsion of surfaces.
What is the geodetic interpretation of this result? Only k1 = 0 and k2 = 0 is a poly-
gon on an isogravitational surface (not an equipotential surface) with different vertical
axes; the East and North axes are principal curvature directions. Within the indicator
diagram (κ1 = κ2 = t1 = 0)(Wx x sec
)12 κ12 + (Wx x sec
)23 x23 + (Wx x sec
)34
x34 + (Wx x sec
)41 x41 = 0, (W yy )12 y12 + (W yy )23 y23 + (W yy )34 y34 + (W yy )41 y41
= 0 and H12 + H34 = 0 have to be fulfilled, if we have a polygon net of four points.
The Euclidean norm = const. The constrained equations are satisfied by the
sufficient restrictions
(Wx x sec
)12 = (Wx x sec
)34 , (Wx x sec
)23 = (Wx x sec
)41 (2.42)
(W yy )12 = (W yy )34 , (W yy )23 = (W yy )23 = (W yy )41 (2.43)
x12 = −x34 , x23 = −x41 , x42 = −y34 , y23 = −y41 . (2.44)
Of special interest is the same indicator diagram with some further restrictions:
(a) the paths 23 and 41 are isogradient, or (b) the paths 12 and 34 are isocoordinate
(Figs. 10 and 11).
We have to provide in the case (a) x12 = −x34 and in the case (b) (W yy )23 =
(W yy )41 . Another interpretation is this: In the case a) the point 2 and 3 or 1 and 4 lie
on the same gravity gradient—Fig. 9—Wx x sec
, W yy resp. But in the case (b) the
points 1 and 2 or 3 and 4 have identical position coordinates.
Finally we mention that there are a lot of other examples for “L. N. Carnot” cycles
within the indicator diagram. Special cases are k1 = 0, t1 = k1 tan
in which x is
integrable, y and z not, and κ1 = t1 = κ2 = 0 in which y is integrable, x and z not.
A singular case is
= π/2 where the E. Cartan pseudo-torsion and G. Frobenius
matrix of integrating factors are unbounded. But this is well known because we
Fig.
9 Indicator diagram for
= 0, dφ =
d
0, d W = 0, κ1 , κ2 , t1 =
0, k1 = 0, k2 = 0
Fig.10 Indicator diagram

for d = 0, dφ =
0, d W = 0, hypothesis
κ1 , κ2 , t1 = 0, k1 = 0, k2 =
0 isogradient
Fig.11 Indicator diagram

for d = 0, dφ =
0, d W = 0, hypothesis
κ1 , κ2 , t1 = 0, k1 = 0, k2 =
0 isocoordinate
248 E. W. Grafarend
cannot find East and North at the Earth’s poles, that is, there is no singularity-free
coordinate system referring on any star-shaped surface.
A shortest version of these first and second OSU reports

appeared also in E. Grafarend [15]. A more detailed
version appeared in P. Defrise and E. Grafarend [16].
2.2 Field Lines of Gravity, Their Curvature and Torsion, the

Lagrange and Hamilton Equations of the Plumbline
The length of the gravitational field lines of the orthogonal trajectories of a family
of gravity equipotential surfaces of the plumbline between a terrestrial topographic
point and a point on a reference equipotential surface like the Geoid - also known
as the orthometric height - plays a central role in Satellite Geodesy as well as in
Physical Geodesy. As soon as we determine the geometry of the Earth pointwise by
means of a satellite GPS we are left with the problem of converting ellipsoidal heights
(geometric heights) into orthometric heights (physical heights). For the computation
of the plumbline we derive its three differential equations of first order as well as
the three geodesic equations of second order. The three differential equations of
second order take the form of a Newton differential equation when we introduce
the parameter time via the Marussi gauge on a conformally flat three dimensional
Riemann manifold and the generalized force field, the gradient of the super potential,
namely the modulus of gravity squared and taken half. In particular, we compute
curvature and torsion of the plumbline and prove their functional relationship to
the second and third derivatives of the gravity potential. For a spherically symmetric
gravity field, curvature and torsion of the plumbline are zero, the plumbline is straight.
Finally we derive the three Lagrangian as well as the six Hamiltonian differential
equations of the plumbline, in particular in their star form with respect to Marussi
gauge.
With the advent of artificial satellites, in particular the satellite Global Positioning
System, high precision geometric positioning of points of the surface of the Earth
has been developed. An unsolved key problem is the transformation of heights in
geometry space, namely the ellipsoidal heights, into heights in gravity space, namely
the orthometric heights/the length of the plumbline with respect to the Geoid. The
field lines of gravity / the orthogonal trajectories of a family of gravity equipotential
surfaces/the plumblines are derived from a set of first order differential equations as
soon as we balance the horizontal/tangential field of the plumbline with the vertical
field/normal field of an equipotential surface of gravity as described by Caputo [17],
in particular with an ellipsoidal gravity field of reference, for instance.
Section two accordingly focuses on a setup of the differential equations of
first and second order of the plumbline with special reference to the transfor-
mation from the parameter arc length s to the dynamic time parameter t
according to the celebrated Marussi Gauge ([18–20]). The arc length squared
ds 2 = grad w2 (d x 2 + dy 2 + dz 2 ) has been represented in terms of conformal

coordinates/isometric coordinates (e.g., Caputo, [17]) with the modulus of gravity
squared γ 2 = grad w2 = λ2 as the factor of conformality squared λ2 . In par-
ticular we succeed in proving that the second order differential equations of the
plumbline establish a geodesic in a three-dimensional Riemann manifold {M3 , gkl },
notably in the form of a Newton dynamical equation if the matrix gkl of the metric
is conformally flat, gkl = λ2 (x)δkl as restricted to the Marussi Gauge. In order
to determine the departure of the plumbline from a straight line, we compute its
curvature and torsion on the basis of the Frenet derivational equations. We aim at
the proof that the curvature of the plumbline is a functional of the second derivatives
of the gravity potential, its torsion of the third derivatives of the gravity potential,
while straight if the gravity field was spherically symmetric. Section two is devoted
to establishing the three Lagrangean differential equations of second order as well
as the six Hamiltonian differential equations of first order, in particular in Marussi
Gauge. In contrast to Moritz [21] we succeed in constructing non-degenerate star
Lagrangeans and star Hamiltonians.
Curvature and torsion of the field lines of the gravity field, the plumbline
At the beginning let us set up the differential equations of the plumbline/the or-
thogonal trajectory with respect to a family of equipotential surfaces by means of
Box 2. The quality between horizontal and vertical fields establishes the first or-
der differential equations of the plumbline: The normalized tangent vector of the
plumbline is identical to the normalized surface vector of an equipotential surface
pointwise. The normalized surface vector of an equipotential surface agrees with
the negative gravity vector, the gradient of the gravity vector. As soon as we dif-
ferentiate the identity of the horizontal field of the plumbline and the vertical field
of an equipotential surface once more, we arrive at the second order differential
equation of a plumbline of inhomogeneous type. The inhomogeneity is generated
by the quadrupole moment in gravity space. As soon as we introduce the parameter
t in order to replace the curve arc length S via the Marussi gauge (Marussi 1979
[18], 1985 [19]) we are led to the first order differential eqs. and the second order
differential eqs. of a plumbline/orthogonal trajectory of a family of equipotential
surfaces. With respect to the Marussi gauge, the second order differential equations
of a plumbline in {R3 , gkl } coincide with the second order differential equations of a
geodesic in Newton form in the Marussi manifold {M3 , γ 2 (x)δkl } with γ 2 (x) as the
factor of conformality. Gravity squared taken half operates as a potential, accord-
ing to a proposal by Chandrasekhar et al. called superpotential: The gradient of the
superpotential γ 2 /2 operates as the force field balanced by the acceleration vector
x .
Indeed we have to explain better the duality between a curve in {R 3 , δkl }, where
the Kronecker δkl , relates to the canonical metric in a three-dimensional Euclidean
space, and a curve in {M3 , γ 2 (x)δkl }.{M3 , γ 2 (x)δkl } is an abbreviated notation for a
three-dimensional Riemannian space parameterized by three conformal coordinates
/ isometric coordinates, whose canonical metric is given by the product of the factor
250 E. W. Grafarend
of conformality γ 2 , the modulus of gravity squared, and the Kronecker δkl ,. Such a
Riemann manifold {M3 , γ 2 (x)δkl } will be called a Marussi manifold. Indeed there
are many Marussi manifolds dependent on the various representations of the gravity
field of the Earth. While the differential equations of second order generate a curve in
the chart {R3 , δkl }, at the same time this curve can be considered as a geodesic in the
Riemann manifold in terms of special coordinates, the ones of conformal/isometric
type. This conception of the curve as a geodesic will become clearer in the next
chapter. We should mention that the duality described earlier has already been applied
by Goenner et al. [22] in order to interpret Newton mechanics as geodesic flow on a
Maupertuis’ manifold.
Secondly we are going to derive the Frenet equations of the plumbline, a curve in
{R3 , δkl }, a three-dimensional Euclidean space completely covered by one chart of
Cartesian coordinates {x 1 , x 2 , x 3 }. As outlined by means of Box 3, we establish by
the Frenet frame {normalized tangent vector, normalized normal vector, normalized
binormal vector} called {f1 , f2 , f3 } (x), which is subject to the coupling to the gravity
field, thanks to the first order differential equations of the plumbline in Marussi
gauge. The derivational equations of the Frenet frame are built on the celebrated
anti-symmetric -matrix which contains as structure elements the curvature κ and
the torsion τ , also called first and second curvature of the plumbline. A straight-
forward computation of curvature and torsion of the plumbline subject to the coupling
of the gravity field. Here we took advantage of the cross product identity a × b =
a2 − b2 − a|b
2 where a indicates the Euclidean norm of the vector a as
well as a|b
the Euclidean scalar product/inner product of two vectors a and b.
Obviously the curvature κ of the plumbline is proportional to gravity gradients or
second derivatives of the gravity potential. This can be seen by means of as soon as
we apply, namely grad γ 2 /2, (∂k γ 2 )/2 = γ1 ∂k γ1 + γ2 ∂k γ2 + γ3 ∂k γ3 = ∂1 w∂k ∂1 w +
∂2 w∂k ∂2 w + ∂3 w∂k ∂3 w.
Box 2
Duality between horizontal and vertical fields in {R 3 , δi j } equipped with a
Euclidean matrix δi j
Normalized tangent vector of the plumbline is identical to the surface normal
vector of an equipotential surface 1st order differential equations
dx grad w dxk
=− ∼ = − ∂k w/ δlm ∂l w∂m w (2.45)
dS grad w dS
second order differential equations
d2xk γ,l
+ 3 (γ 2 δ kl − γ k γ l ) = 0 (2.46)
dS 2 γ
Marussi gauge

x = grad w , dx = x −1 x (2.47)
dS
fist order differential equation of the plumbline in Marussi gauge
x = −grad w (2.48)
second order differential equation of the plumbline in Marussi gauge
xk = (−∂l γ k )x l = (∂l γ k )∂l w = (∂l γ k )γ l (2.49)
xk − 21 ∂k γ 2 = 0 (2.50)
End Box 2
In contrast, the torsion τ of the plumbline is proportional to the second deriva-
tives of the gravity vector or the third derivatives of the gravity potential. Such a
result is motivated by the identity γ,i2j /2 = γk,i j γk + γ(k,i) γ(k, j) = (∂i ∂ j ∂k w)∂k w +
(∂i ∂k w)(∂i ∂k w). Finally as a corollary we report the result that curvature and torsion
of the plumbline amount to zero if the gravity field has spherical symmetry. Or we
may say that κ0 = 0, τ0 = 0, if the gravity field w0 (x, y, z) = w0 (r ) depends only
on the radial coordinate r . This result gave the motivation for a decomposition of
curvature and torsion κ = κ0 + δκ, τ = τ0 + δτ subject to κ0 = τ0 = 0 in terms of
the gravity field γ = γ0 + δγ the normal gravity field γ0 = γ0 (r ) and the disturbing
gravity field {δγ (λ, ϕ, r )} which depends on the lateral variation of lengthy spherical
coordinates. Since the representations δκ, δτ are lengthy, we drop them here.
Box 3
Curvature and torsion of the plumbline in {R3 , δi j } equipped with a Euclidean
metric δ. The Frenet frame as the natural triad of the plumbline
x
f 1 := (2.51)
x

x − x | f 1 f 1
f 2 := (2.52)
x − x | f 1 f 1

x − x | f 1 f 1 − x | f 2 f 2

f 3 := (2.53)
x − x | f 1 f 1 − x | f 2 f 2
subject to
(i)xi = −γ i , (ii)xi = 21 γ,i2 , (iii)xi = − 21 γ,i2j γ j (2.54)

252 E. W. Grafarend
The derivational equations of the Frenet frame
f 1 = κ S f 2 (2.55)
f 2
= −κ S f 1 + τ S f 3
(2.56)
f 3 = −τ S f 2 (2.57)
⎡ ⎤ ⎡ ⎤⎡ ⎤
f 1 0 ω12 0 f1
⎣ f ⎦ = ⎣ −ω12 0 ω23 ⎦ ⎣ f 2 ⎦ (2.58)
2
f 3 0 −ω23 0 f3
⎡ ⎤ ⎡ ⎤⎡ ⎤
f 1 0 κ S 0 f1
⎣ f ⎦ = ⎣ −κ S 0 τ S ⎦ ⎣ f 2 ⎦ (2.59)
2
f 3 0 −τ S 0 f3
Curvature κ and torsion τ

x × x x |x × x
κ= 3 , τ= , (2.60)
x x x 2
2 2
κ= 1
γ 2 gradγ 2 − γ |gradγ 2
√ 2 2 2 2 2 22
2γ 3
(2.61)
(γ1 +γ2 +γ3 )(γ,1 +γ,1 +γ,1 ) −(γ1 γ,12 +γ2 γ,22 +γ3 γ,32 )2
= 2(γ12 +γ22 +γ32 )3/2
γ,12 j γ j (γ2 γ,32 − γ3 γ,22 ) + γ,22 j γ j (γ3 γ,12 − γ1 γ,32 ) + γ,32 j γ j (γ1 γ,22 − γ2 γ,12 )
τ=
γ 2 δ kl γ,k2 γ,l2 − (δ kl γk γ,l2 )2
(2.62)
Corollary
κ0 = 0, τ0 = 0, if w0 (x, y, z) = w0 (r ) (2.63)
End Box 3
Instead, thirdly, we compute by means of Box 4 the plumbline in a spherically
symmetric gravity field subject to Marussi gauge. Let us depart from the first order
differential equations of a plumbline subject to Marussi gauge. Indeed by means
of we restrict the gravity field to be spherically symmetric: the gravity potential
w(x, y, z) = f (r ) has been chosen to be a function of the radial coordinate only. The
appropriate coordinate system in which to solve the first order differential equations is
the spherical coordinate system {λ, ϕ, r }. We have used the forward transformations
Cartesian coordinates into spherical coordinates in order to represent the first
order differential equations of the plumbline subject to Marussi gauge in spherical
coordinates, to prove λ = 0, ϕ = 0 and r = − f (r ), a result collected in the
corollary. Finally as an example we have chosen the potential and the gravity field of
a homogeneous, massive sphere in the inner zone A and the outer zone B in order to
solve the ordinary differential equation of the radial component of the plumbline. The
function r = r0 exp gm/R s (t− t0 ) is a representation of the solution of r = − f (r )
in case 1, R > r , while r = 3 r03 + 3 gm(t − t0 ) in case 2, R < r . For R = r , both
solutions agree with each other. gm denotes the product of the gravitational constant
g and the mass m of the homogeneous, massive sphere.
Box 4
Computation of a plumbline in a spherical symmetric gravity field. Marussi gauge.
⎡
x = − ∂w
∂x
x = −grad w ⇔ ⎣ y = − ∂w

∂y (2.64)
z = − ∂w
∂z
spherically symmetric gravity field
w(x, y, z) = f (r ) subject to r 2 = x 2 + y 2 + z 2 (2.65)
dw ∂r xk df
∂k w = = f (r ) , f (r ) := (2.66)
dr ∂ x k r dr
forward transformation: Cartesian coordinates into spherical coordinates

λ = ar ctan(y/x) + − 21 sgn(y) − 21 sgn(y)sgn(x) + 1 π, λ ∈ {R|0 ≤ λ < 2π }
φ = ar ctan( √ 2z 2 ), φ ∈ {R| − π/2 < φ < +π/2}
x +y
(2.67)
representation of the first order differential equation of the plumbline in spherical
coordinates d tan λ = x x+y dλ = −yd x+xdy
2 2
2 x2
1 r2
d tan φ = ((x 2
+ y 2
)dz − zxd x − zydy) = dφ (2.68)
(x 2 + y 2 )3/2 x 2 + y2
⎡
x = − f (x) rx
xk
= − f (r ) ⇔ ⎣ y = − f (r ) ry

xk (2.69)
r
z = − f (r ) rz
254 E. W. Grafarend
1
λ = (−yx + x y ) (2.70)
x2 + y2
1 1
φ = (−zx x − zyy + (x 2 + y 2 )z ) (2.71)
(x + y ) 2
2 2 1/2
1
r = (x x + yy + zz ) (2.72)
r
1
λ = (y f (r )x/r − x f (r )y/r ) = 0 (2.73)
x 2 + y2
1 1
φ = (zx f (r )x/r + zy f (r )y/r − (x 2 + y 2 ) f (r )z/r ) (2.74)
(x + y ) 2
2 2 1/2

r 2 r = −(x 2 f (r ) + y 2 f (r ) + z 2 f (r )) = −r 2 f (r ) (2.75)
Corollary

λ = 0, φ = 0, r = − f (r ), i f w(x, y, z) = f (r ) (2.76)
Example: massive sphere

2

gm
3 − Rr 2 ∀0 ≤ r < R : zone A
w0 (r ) = 2R
gm
(2.77)
r
∀R ≤ r < ∞ : zone B
or in terms of the Heaviside function H (R, r )

gm r2 gm
w0 (r ) = H (R − r ) 3 − 2 + H (r − R) (2.78)
2R R r
gm
r ∀0 ≤ r < R : zone A
grad w0 (r ) = −er R3
gm (2.79)
r2
∀R ≤ r < ∞ : zone B
or
gm gm
grad w0 (r ) = −er H (R − r ) 3
r − er H (r − R) 2 (2.80)
R r
gm k gm
− ∂k w0 = H (R − r ) 3
x + H (r − R) 3 x k = x k (2.81)
R r
Corollary
gm gm
λ = 0, φ = 0, r r = r 2 (H (r − r )) 3
+ H (r − R) 3 (2.82)
R r
if w(x, y, z) = w0 (r )
Case1: R > r : r = gmR3

r ⇒ drr = gm
R3
dt ⇒
gm
⇒ ln r − ln r0 = R 3 (t − t0 ) ⇒
⇒ lnlnrr0 = gm
R3
(t − t0 ) ⇒
⇒ r0 = exp R 3 (t − t0 ) ⇒
r gm
r = r0 exp gm
R3
(t − t0 ) (2.83)
Case2: r > R : r = gm
r2
⇒ r 2 dr = gmdt ⇒
3 r03
⇒ r3 − 3 = gm(t − t0 ) ⇒
⇒ 13 r 3 = 13 r03 + gm(t − t0 ) ⇒

r= r0 + 3gm(t − t0 )
3 3
(2.84)
End Box 4
Finally we illustrate by the solution of the first order differential eqs. of the
plumbline subject to Marussi gauge, namely the bundle of straight lines with the
mass center as the focal point for a spherically symmetric gravity field. For a more
realistic gravity field in the crust of the Earth, Svensson in Grafarend [23] has com-
puted a sample plumbline in the Alpes by a Runge–Kutta numerical computation
of the solution of the first order differential equation of the plumbline in Marussi
gauge and a gravity field given by a set of homogeneous massive spheres around
the plumbline representing the local gravity field. Reference is made to Fig. 12.
In addition, by Fig. 13 we illustrate the computation of a realistic plumbline at the
Fig. 12 A set of plumblines

for a spherically symmetric
gravity field γ0 ; straight lines
with the mass center of the
Earth as a focal point
256 E. W. Grafarend
Fig. 13 Computation of a realistic plumbline at a mountain point in the Alpes according to Svensson
in Grafarend [23]
Swiss high mountain point Jungfraujoch performed by Hunziker [24]. An alternative

procedure for the gravitational field in the crust is outlined in Engels and Grafarend
[25], Engels et al. [26] and in Grafarend et al. [27]. Finally we refer to Grossman
[28] and [29] for the focal point of plumblines (Fig. 14).
3 Real Null Frames and Coframes in General Relativity,

Anholonomity, GPS Coordinates
Orthonormal reference frames, in general, are a typical example of the anholonomic

reference, here in Geodesy, Astronomy, Astrometry and Space Science: they are
taken reference to spatial triads with respect to Newton time. For instance, various
frames of reference are presented as in I.I. Mueller et al. ([30, 31]), E. Grafarend
([32, 33]), M. Fujimoto and E. Grafarend [34].
With respect to General Relativity we introduce real null frames and coframes
in order to satisfy the needs of modern Positioning Systems, namely GPS and LPS:
global versus local, based on light and radar signals.
Fig. 14 Computation of a realistic plumbline at the mountain point Jungfraujoch at the Swiss
Alpes by Hunziker (1960, p. 151), departure from a straight line projection onto the reference Fig.
13 mm/30.8 mm in the North/West direction
More specifically, in pseudo-Riemannian spacetime we introduce

at each point in spacetime a local pseudo-orthonormal frame of
reference eα (x) with respect to α, β, · · · ∈ {1, 2, 3, 4}. We chose
a timelike 4-leg e4 , where as {e1 , e2 , e3 } are spacelike. specifically,
this frame of reference is pseudo-orthonormal, but unfortunately an-
holonomic. Its metric is given by g(eα , eβ ) = diag(1, 1, 1, −1), the
local Minkowski metric.
Two complex half null and two real half null reference frames, in total null, tetrads
play now a dominant role, in particular to study gravitational waves, advanced an
retarded ones:
two real null vectors e2 , e3
and
two complex conjugate null vectors e1 , e4
Frames consisting of four null vectors
At each point of the four-dimensional spacetime characterized by coordinates x i

subject to i, j, · · · ∈ {1, 2, 3, 4} we associate a four-dimensional tangent space. For
linear independent vectors eα constitute a basis, also called a frame. Dual to this
258 E. W. Grafarend
frame is the coframe ωβ which consists of 4 covectors, also called one-forms. Let us
decompose frame and coframe with respect to local frames
β
eα = f αi ∂i ∼ ωβ = f j d x j (3.1)
β
The matrices f αi and f j are called Frobenius matrices of integrating factors.
Duality is defined by
f iα f αj = δi f iα f βi = δβα
j
and (3.2)
We refer to the review of Blagojevic, M., Garecki, J., Hehl, F.W. and Obukhov,
Y.N. [3]: The matrices f i .α are called frame components or tetrad components.
Conventionally, in Relativity vectors in an orthonormal or pseudo-orthonormal
frames are labeled {1, 2, 3, 4} underlining the fundamental difference between e4 ,
the negative length g44 = g(e4 , e4 ) = −1 and ea , a ∈ {1, 2, 3} which have positive
lengths gaa = g(ea , ea ) = +1. We call a vector timelike if g(u, u) < 0, spacelike if
g(u, u) > 0 and null of lightlike if g(u, u) = 0. The components of the metric tensor
with respect to the pseudo-orthonormal frame eα are represented by
⎡ ⎤
1 0 0 0
∗ ⎢0 1 0 0⎥
gαβ := ⎢ ⎣0 0 1
⎥ (3.3)
0⎦
0 0 0 −1
The star equal sign indicates that this equation holds only in this specific frame,
the pseudo-orthonormal one.
Null frames
We begin with a pseudo-orthonormal reference frame eα whose metric will take

the standard form
⎡ ⎤
1 0 0 0
⎢0 1 0 0⎥
gαβ := ⎢
⎣0 0 1
⎥ (3.4)
0⎦
0 0 0 −1
We build an alternative frame of reference which is well suited to investigate grav-

itational waves, so-called null coordinates, more specifically advanced and retarded
one. For instance, define
l := √1 (e1
2
+ e2 ), n := √1 (e1
2
− e2 )
(3.5)
m := √1 (e3
2
+ ie4 ), m := √1 (e3
2
− ie4 )
m, m have been introduced by E. T: Newman and R. Penrose, see R. Penrose and

W. Rindler [35]. i is the imaginary unit, the over bar means complex conjugation
subject to i 2 = −1.
Let us analyze the new frame (l, n, m, m) := eα as a functional relation of the old
frame (e1 , e2 , e3 , e4 , ) := eα for spacetime. Both frames of references, eα , respec-

tively eα are elements of the set {1, 2, 3, 4}. We note the condition of null frames in
terms of metric coefficients
g(l, l) = g(n, n) = g(m, m) = g(m, m) = 0 (3.6)
We note also the transformation of the metric in terms of pseudo-orthonormal

frames of reference of type
⎡ ⎤⎡ ⎤
1 0 0 0 ω1
2 ∗
⎢0 1 0 0⎥ ⎢ ω2 ⎥
ds = [ω1 , ω2 , ω3 , ω4 ] ⎢
⎣0
⎥⎢ ⎥
0 1 0 ⎦ ⎣ ω3 ⎦
0 0 0 −1 ω4
⎡ ⎤⎡ ⎤ (3.7)
0 0 0 0 ω1
⎢1 0⎥ ⎢ ⎥
∗∗ ⎢
= [ω1 , ω2 , ω3 , ω4 ] ⎣
0 0 ⎥ ⎢ ω2 ⎥
0 0 1 −1 ⎦ ⎣ ω3 ⎦
0 0 −1 0 ω4
“*” as well as “**” indicate that these identities hold only in this special frame of
reference.
Transformation of the Metric for Null Frames of Reference

Let us study the transformation of the metric for these special frames of reference of
type null frame.

∗ ∗∗
ds 2 = gμν (u, u)ωμ ων = gμ ν (u u )ωμ ων (3.8)
μ
gμν (u, u) jμ jνν ωμ ων = gμ ν (u (u), u (u) ωμ ων (3.9)
jμμ and jνν are Jacobi matrices of first derivatives.

gμν u(u ), u(u ) jμμ jνν = gμ ν (u , u )

(3.10)
Proof
half-null frame “l, π ”
Let us begin with the first case: for all μ, ν ∈ {1, 2} we assume gμν = diag(1, 1)
and g(l, l) = g(n, n) = 0 or g1 1 = g2 2 = 0 as well as g(l, n) = g(n, l) or g1 2 =
g2 1 a result which we call “symmetry of the metric”.
260 E. W. Grafarend
Problem
Given gμν = diag(1, 1) and g(l, l) = g(n, n) = 0 or g1 1 = g2 2 = 0.
Find g(l, n) = g(n, l) or g1 2 = g2 1
Since g(l, l) = g(n, n) or g1 1 = g2 2 = 0 which holds by definition of the half-
null frame we are left with the problem to construct g(l, n) = g(n, l) or g1 2 = g2 1 .
Please, note
l := √1 (e1 + e2 ),
2
n := √1 (e1 − e2 )
2

ω1 := √1 (ωe1 + ω2 ),
2
ω2 := √1 (ωe1 − ω2 )
2
versus
e1 = √1 (l
2
+ n), e2 =√1 (l − n)
2

ω1 = √1 (ω
2
1
+ ω2 ), ω2 = √12 (ω1 − ω2 )
or

(ω1 )2 = 21 (ω1 )2 + 2ω1 ω2 + (ω2 )2

(ω2 )2 = 21 (ω1 )2 + 2ω1 ω2 + (ω2 )2

g(l, l) = g(n, n) = 0 ⇐⇒ (ω1 )2 = (ω2 )2 = 0

(ω ) + (ω2 )2 = 2ω1 ω2 (ω1 )2 = ω1 ω2 , (ω2 )2 = ω1 ω2
1 2
g1 2 = gln = +1 = gnl = g2 1 (3.11)
At this end, we summarize: The first half-null frame accounts to

0 +1
gμ ν = for all μ , ν ∈ {1, 2} (3.12)
+1 0
Proof
half-null frame “m, m”
Let us continue with the second case: for all μ, ν ∈ {1, 2} we define gμν =
diag(+1, −1) and g(m, m) = g(m, m) = 0 or g3 3 = g4 4 = 0 as well as
g(m, m) = g(m, m) or g3 4 = g4 3 a result which we call “symmetry of the pseudo-
metric”.
Problem
Given gμν = diag(+1, −1) and g(m, m) = g(m, m) = 0 or
g3 3 = g4 4 = 0
Find g(m, m) = g(m, m) or g3 4 = g4 3
Since g(m, m) = g(m, m) = 0 or g3 3 = g4 4 = 0 by definition of the half-null
frame we are left with the problem to construct g(m, m) = g(m, m) or g3 4 = g4 3 .
Please, note
m := √1 (e3 + ie4 ),
2
m := √1 (e3 − ie4 )
2

ω3 := √1 (ωe3 + iω4 ),
2
ω4 := √1 (ωe3 − iω4 )
2
versus
e3 = √1 (m
2
+ im), e4 = √1 (m
2
− im)
and

ω3 = √1 (ω3
2
+ ω2 ), ω4 = √1 (ω3
2
− iω4 )
or

(ω3 )2 − (ω4 )2 = 21 (ω3 ) + iω4 (ω3 + i(ω4 )+

+ 21 (iω3 ) + ω4 (iω3 + (ω4 )

g(m, m) = g(m, m) = 0 ⇐⇒ (ω3 )2 = (ω4 )2 = 0

(ω3 )2 − (ω4 )2 = −2ω3 ω4
g3 4 = gmm = −1 = gm,m = g4 3 (3.13)
Finally, we summarize: The second half-null frame accounts to

0 −1
gμ ν = for all μ ν ∈ {3, 4} (3.14)
−1 0
Here, i is the imaginary part and over-bar means complex conjugation. The total
transformation leads to the Lorentz metric in a Newnan–Penrose null frame
eα := (l, n, m, m) :
in detail
⎡ ⎤
0 0 0 0
⎢1 0 0 0⎥
gμ ν =⎢
⎣0
⎥ (3.15)
0 1 −1 ⎦
0 0 −1 0
Such a frame is very well suited for investigating gravitating waves as well as
electromagnetic waves. M. Blagojevic, J. Garecki, F.W. Hehl and Yu.N. Obukhov
[3] developed alternatively Real Null Coframes to which we refer. Here we adopted
the title Real null coframes in General Relativity and GPS type coordinates from M.
Blagojevic et al. [3].
262 E. W. Grafarend
4 Examples
Our first example relates to Killing vectors of symmetry, in particular of the ellipsoid-
of-revolution and the sphere. Second, we study anholonomity, namely the influence
of the local vertical in computing geodetic networks in the form of three dimensional
Geodesy. Our example discusses the misclosure of a triangular network observed
by scaled distance measurements. The local vertical in gravity space depends on
astronomic longitude and astronomic latitude. In detail, we present the analysis of a
triangular network in the range of (i) 25 m and (ii) 500 m. The structure of this network
causes miscloser in the range of (i) 1 mm and (ii) 300 mm. Anholonomity influences
these misclosers. Third, we analyze anholonomity as the problem of integrability or
differential forms. The tool in treating problematic orthonormal frames of reference
is “exterior calculus”, here introduced of analyzing the 2-sphere with respect to two
frames of reference.
4.1 Killing Vectors of Symmetry
We begin with two questions:
First question
Let a transformation group act on the coordinate transformation of a surface-
of-revolution. What are the transformation groups which leaves the first differential
invariant ds 2 of a surface-of-revolution equivariant or form invariant? Answer 1: The
transformation group which leaves the first differential invariant ds 2 , also called “arc
length”, equivariant is the one-dimensional rotation group R3 (longitude), a rotation
“around the 3-axis” of the ambient space {R3 , δi j }. The 3-axis establishes the Killing
vector of symmetry.
The proof
Our proof for the “Answer 1” is outlined in Box 5. First, we present a parameter
representation of a surface-of-rotation defined by {u, v} in a equatorial frame -of-
reference and defined by {u ∗ , v ∗ } in a rotated equatorial frame-of-reference. Second,
we follow the action of the rotation group R3 () ∈ S O(2). Third, we generate the
forward and backward transformations
{e1 , e2 , e3 |0} → {e1∗ , e2∗ , e3∗ |0} as well as
{e1∗ , e2∗ , e3∗ |0} → {e1 , e2 , e3 |0} of
orthogonal base vectors which span the three-dimensional Euclidean ambient space.
Fourth, we fill in the backward transformation of bases into the first parameter
representation of the surface-of-revolution and compare with the second one. In
this way, we find the “Kartenwechsel”, [change form one chart to another chart]
{u ∗ = u − , v ∗ = v}. Fifth, we compute the first differential invariant ds 2 of the

surface-of-revolution, namely the matrix of the metric G = diag[ f 2 , f 2 + g 2 ].
{ f, g} are representations of the parameters describing the surface-of-revolution
given by the formula (4.1) and (4.2) in Box 5! “Cha-Cha-Cha” leads us via the
Jacobi map J to the second representation d ∗2 of the first differential invariant which
turns out to be equivariant or form invarinat. Indeed, we have shown that under the
action of the rotation group ds 2 = ds ∗2 . Sixth, we identity e3 as the Killing vector
of the symmetry of the surface of revolution.
Box 5: Surface of revolution, Killing vector of symmetry, equivariance of the

arc length under the action of the special orthogonal group SO(2)
Surface-of-revolution parametrized in an equatorial frame of reference:
x(u, v) = e1 F(v) cos u + e2 f (v) sin u + e3 g(v) (4.1)
Surface-of-revolution parametrized in a rotated equatorial frame of reference:
x(u ∗ , v ∗ ) = e1∗ F(v ∗ ) cos u ∗ + e2∗ f (v ∗ ) sin u ∗ + e3∗ g(v ∗ ) (4.2)
Action of the special orthogonal group S O(2):

R3 () ∈ S O(2) := {R3 ∈ R3×3 |R3∗ R3 = I3 , |R3 | = 1},
⎡ ⎤ ⎡ ⎤ ⎡ ⎤⎡ ⎤
e1∗ e1 cos sin 0 e1
⎣ e2∗ ⎦ = R3 () ⎣ e2 ⎦ = ⎣ − sin cos 0 ⎦ ⎣ e2 ⎦
e3∗ e3 0 0 1 e3
⎡ ⎤ ⎡ ⎤ ⎡ ⎤⎡ ⎤ (4.3)
e1 e1∗ cos − sin 0 e1∗
⎣ e2 ⎦ = R3∗ () ⎣ e2∗ ⎦ = ⎣ sin cos 0 ⎦ ⎣ e2∗ ⎦
e3 e3∗ 0 0 1 e3∗
e1 = e1∗ cos − e2∗ sin , e2 = e2∗ sin + e2∗ cos , e3 = e3∗
Coordinate transformation
x(u, v) = f (v)e1∗ (cos cos u + sin sin u) + f (v)e2∗ (− sin cos u +

cos sin u) + e3∗ g(v)
x(u, v) = f (v)e1∗ cos u − + f (v)e2∗ sin u − e3∗ g(v) (4.4)
v = v ∗ , x(u, v) = x(u ∗ , v ∗ )
cos u ∗ = cos(u − ), sin u ∗ = sin(u − ), tan u ∗ = tan(u − ) (4.5)
u∗ = u − ω
du
Arc length (first differential invariant): ds = 2
[du, dv]Jx∗ Jx
dv
264 E. W. Grafarend
⎡ ⎤ ⎡ ⎤
Du x Dv x

− f sin u f cos u 2
⎣ ⎦ ⎣
⎦ ∗ f 0
Jx = Du y Dv y = f cos u f sin u , G := Jx Jx =
0 f 2+g2
Du z Dv z 0 g
(4.6)
Killing vector of symmetry (rotational axis):
1st version: 2nd version:

ds 2 = f 2 du 2 + ( f 2 + g 2 )dv 2 ds ∗2 = f ∗2 du ∗2 + ( f ∗ 2 + g ∗ 2 )dv ∗2
u ∗ = u − , v ∗ = v ⇔ du ∗2 = du 2 , dv ∗2 = dv 2 (4.7)

ds 2 = f 2 du 2 + ( f 2
+ g 2 )dv 2 = f 2 du ∗2 + ( f 2
+ g 2 )dv ∗2 = ds ∗2
⎡ ⎤ ⎡ ⎤
0 0
e3 = [e1 , e2 , e3 ] ⎣ 0 ⎦ ∼ ⎣ 0 ⎦ (4.8)
1 1
End Box 5
Second question
Let a transformation group act on the coordinate transformation of the sphere. Or

we may say, we make a coordinate transformation. What are the transformation
groups, the coordinate transformations which leave the first differential invariant
ds 2 of a sphere equivariant or form invariant Answer 2: The transformation group
which leaves the first differential invariant ds 2 , also called “arc length”, equivariant
is the three-dimensional rotation group R( α, β, γ ), a subsequent rotation “around
the 1-axis, the 2-axis and the 3-axis” of the ambient space {R3 , δi j }. The three-axes
establish three Killing vector of symmetry.
Box 6: Sphere, Killing vector of symmetry, equivariance of the arc length

under the action of the special orthogonal group SO(3)
Sphere parametrized in an equatorial frame of reference:
x(u, v) = e1 cos v cos u + e2 cos v sin u + e3 sin v (4.9)
Sphere parametrized in an oblique frame of reference:
x(u ∗ , v ∗ ) = e1∗ cos v ∗ cos u ∗ + e2∗ cos v ∗ sin u ∗ + e3∗ sin v ∗ (4.10)
Action of the special orthogonal group S O(3):
R(α, β, γ ) ∈ S O(3) := {R ∈ S O(3)|R ∗ R = I3 , |R| = 1},

⎡ ⎤ ⎡ ⎤
e1∗ e1
⎣ e2∗ ⎦ = R1 (α)R2 (β)R3 (γ ) ⎣ e2 ⎦ (4.11)
e3∗ e3
⎡ ⎤ ⎡ ⎤
e1 e1∗
⎣ e2 ⎦ = R3∗ (α)R2∗ (β)R1∗ (γ ) ⎣ e2∗ ⎦ (4.12)
e3 e3∗
e1 = e1∗ (cos γ cos β) − e2∗ (sin γ cos α + cos γ sin β sin α + e3∗ (sin γ sin α +
cos γ sin β cos α),
e2 = e1∗ (sin γ cos β) + e2∗ (cos γ cos α + sin γ sin β sin α + e3∗ (− cos γ sin α +
sin γ sin β cos α),
e3 = e1∗ (− sin β) + e2∗ (cos β sin α) + e3∗ (cos β cos α).
Coordinate transformations:
x(u, v) = x(u ∗ , v ∗ )
⇔
e1∗ f 1 (α, β, γ |u, v) + e2∗ f 2 (α, β, γ |u, v) + e3∗ f 3 (α, β, γ |u, v) (4.13)
= e1∗ cos v ∗ cos u ∗ + e2∗ cos v ∗ sin u ∗ + e3∗ sin v ∗ ,

cos v ∗ cos u ∗ = f 1 (α, β, γ |u, v)
= cos γ cos β cos v cos u + sin γ cos β cos v sin u − sin β sin v,
cos v ∗ sin u ∗ = f 2 (α, β, γ |u, v)
−(sin γ cos α + cos γ sin β sin α) cos v cos u+
+ (cos γ cos α + sin γ sin β sin α) cos v sin u + cos β sin α sin v, (4.14)
sin v ∗ = f 3 (α, β, γ |u, v)

(sin γ sin α + cos γ sin β cos α) cos v cos u−
−(cos γ sin α + sin γ sin β cos α) cos v sin u + cos β cos α sin v,
f2
tan u ∗ = , sin v ∗ = f 3 (4.15)
f1
Arc length (first differential invariant):

2 ∗
r cos v 0 du du
ds 2 = [du, dv] (“diffeomorphism”), =
0 r2 dv dv ∗

du Du u ∗ Dv u ∗
J , J := ,
dv Du v ∗ Dv v ∗
d tan u ∗ = (1 + tan2 u ∗ )du ∗ ⇒ du ∗ = cos2 u ∗ d tan u ∗ ,

d sin v ∗ = cos v ∗ dv ∗ ⇒ dv ∗ = √ 1 2 ∗ a sin v∗, (4.16)
1−sin v
ds 2 = r 2 cos2 vdu 2 + r 2 dv 2 = r 2 cos2 v ∗ du ∗2 + r 2 dv ∗2 = ds ∗2
266 E. W. Grafarend
Killing vector of symmetry (rotation axis):

⎡ ⎤
1
1- axis of symmetry: e1 ∼ ⎣ 0 ⎦ ,
⎡0⎤
0
2- axis of symmetry: e1 ∼ ⎣ 1 ⎦ , (4.17)
⎡0⎤
0
3- axis of symmetry: e1 ∼ ⎣ 0 ⎦ .
1
End Box 6
The proof
First, we present a parameter representation of the sphere defined by {u, v} in
a equatorial frame-of-reference and defined by {u ∗ , v ∗ } in a oblique frame-of-
reference generated by the three-dimensional orthogonal group S O(3). Second,
the action of the transformation group S O(3) is parametrized by Cardan an-
gles {α, β, γ }, namely a rotation R1 (α) by around the 1 axis, a rotation R2 (β)
around the 2-axis, and a rotation R3 (γ ) around the 3-axis. Third, we transform
forward and backward the orthonormal system of base vectors {e1 , e2 , e3 |0} and
{e1∗ , e2∗ , e3∗ |0}, which span the three-dimensional Euclidean space, the ambient
space of the sphere Sr2 . {e1 , e2 , e3 |0} establish the conventional equatorial frame-
of-reference, but {e1∗ , e2∗ , e3∗ |0} at the origin the meta-equatorial reference frame.
Fourth, the backward transformation is substituted into the parameters representation
of the placement vector e1r cos v cos u + e2 r cos v sin u + e3r sin v ∈ Sr2 , such that,
e1∗ f 1 (α, β, γ |u, v) + e2∗ f 2 (α, β, γ |u, v) + e3∗ f 3 (α, β, γ |u, v) is a materialization
of the “Kartenwechsel”. In this way, we are led to tan α ∗ = f 2 / f 1 and sin v ∗ = f 3
, both functions of the parameters {α, β, γ } ∈ S O(3) of longitude u and the lat-
itude v. Fifth, as soon as we substitute “Cha-Cha-Cha”, namely the diffeomor-
phism {du, dv} → {du ∗ , dv ∗ } by means of the Jacobi matrix J in the first dif-
ferential invariant ds ∗2 , namely the matrix of the metric G = diag[r 2 cos2 v, r ],
we are led to the first representation ds 2 of the first differential invariant of
form-invariant: ds 2 = r 2 cos2 vdu 2 + r 2 dv 2 = r 2 cos2 v ∗ du ∗2 + r 2 dv ∗2 . Indeed we
have shown that under the action of the three-dimensional rotation group, namely
R(α, β, γ ) = R1 (α)R2 (β)R3 (γ ), ds 2 = ds ∗2 . Sixth, we identify the three Killing
vectors {e1 , e2 , e3 } or [1, 0, 0], [0, 1, 0] and [0, 0, 1], respectively, the symmetry of
the sphere Sr2 . For more details refer to two volume book E. Grafarend, R.J. You R.
Syffus (2012)
5 The Influence of the Local Vertical on the Analysis of

Geodetic Networks Due to Anholonomity
The systematic error in geodetic coordinate computations of local networks - also

called “engineering networks”- usually separated in horizontal and vertical parts and
which is caused by neglected changes of the local vertical are estimated. We docu-
ment that in local network of approximately 25 cm the systematic errors amount to
about a millimeters but in local network of approximately 500 m extension about 10–
30 m. Accordingly for the computation of precise geodetic networks, the influence
of the local gravity field, caused by variations of the local physical vertical-measured
by vertical deflections of gravity disturbances-cannot be neglected.
Why is the influence of the terrestrial gravity potential and its gravity field so
small?
It is worth while to study the gravity field and the shape of the figure of the Earth
more accurate.
Model Assumptions
First assumption
The figure of the Earth is assumed to be described to be the sphere of constant
radius r0 , the radius of its boundary.
Second assumption
The mass distribution of the Earth is assumed to be homogenous.
Inversion w0 −→ r0
We depart form the assumption that geodetic measurements are able to yet infor-
mation of the potential values on the spherical surface. The terrestrial potential w0
with respect to the simple Earth model could be easily converted into the Earth r0 ,
namely
w0 = gmr0−1 −→ r0 = gmw0−1
Inversion γ0 −→ r0
We depart form the assumption that geodetic measurements of type gravimetry γ0

are able to get information of the gravity values on the spherical surface. Terrestrial
gravity values γ0 , the length of the gravity vector, could be easily inverted to the
Earth radius, especially with respect to the simple Earth model, namely for γ0 > 0:
√
γ0 = w0 = gmr0−2 −→ r0 = gmγ0−2

Inversion γ0 = w0 −→ r0
268 E. W. Grafarend
We depart from the assumption that geodetic measurements of type gravity gradi-
ent/of type first derivative of gravity or of type second derivative of the potential, are
able to get information of relative gravity data on the spherical surface. The terrestrial

values γ0 = w0 with respect to the simple Earth model could be easily inverted in
the Earth radius, namely
√ −1/3
γ0 = w0 = 2gmr0−3 −→ r0 = 3 2gmγ0
We have seen that for a simple Earth model (i) homogeneous mass distribution,
(ii) spherical Earth - there is no integrability problem. But the real Earth has a
heterogeneous, inhomogeneous, latitude and longitude dependent mass distribution
and the figure and the gravity filed is neither spherical nor ellipsoidal, but irregular!
There is the argument for small anholonomity.
Typical for surveying problems in local networks, for instance “deformation net-
works”, is the adjustment in horizontal as well as vertical parts. Mainly planar geode-
tic nets are treated exclusively in the surveying literature. The influence of change in
the local vertical is neglected. But the local vertical is the reference direction in po-
sitioning instruments especially in LPS (“Local Positing Systems”): Local networks
are typically anholonomic since the local vertical due to irregular mass distribution
varies form point to point, not being reflected by planar networks.
5.1 The Local Geodetic Reference Systems (LPS)
At first we take reference to the local moving reference frame systematically called
“Horizontal Reference Frame”

(HRF)
as well as the space-fixed
“Equatorial Reference Frame”

(ERF)
We want to document that the change of local vertical is of central importance.

Let us assume that we reduce in the frame of a peripheral model or geodetic mea-
surements, namely
• instrumental errors
• central error
• null point systematic errors
• refractional effects
• relativistic effects of first order.
By means of modern electronic measurement instruments in a local position-

ing system of type theodolite and distance measurement equipment we are able to
determine the spherical coordinates of type
Fig. 15 Triangular network {Pα , Pβ , Pγ |0}, placement vector at the origin O local vertical ,
E3 (Pα ), E3 (Pβ ), E3 (Pγ ), α , β , γ local gravity vectors
• horizontal direction Hαβ , Hβγ Hγ α ,

• vertical direction (the complement of the zenith distance) Vαβ , Vβγ Vγ α ,
• distance Sαβ , Sβγ Sγ α ,
between target points and station points {Pα , Pβ , Pγ }, in particular the relative po-
sition vectors X αβ := X β − X α , X βγ := X γ − X β , and X γ α = X α − X γ . Figure
15 illustrations the triangular network {Pα , Pβ , Pγ |0} with respect to vectors at the
origin 0 and the local verticals E3 (Pα ), E3 (Pβ ) and E3 (Pγ ) in a closed loop as well
as local gravity vectors. It is a common practice to represent the relative position
vector in an orthogonal local horizontal frame of reference.
Consult Fig. 16: Commutative diagram where we illustrate the moving horizontal
reference frames, namely moving relative to the fixed equatorial reference frame
“freely transported to the point {Pα , Pβ , Pγ }”.
fixed moving
orthonormal orthonormal
reference reference
frame frame
{F•1 , F•2 , F•3 } {E∗1 , E∗2 , E∗3 }
We have denoted the fixed equatorial frame of reference by {F•1 , F•2 , F•3 } as well
as the moving horizontal frame of reference by {E∗1 , E∗2 , E∗3 }.
Question
270 E. W. Grafarend
Fig. 16 Commutative
diagram: moving horizontal
reference systems Eα versus
fixed equatorial reference
system
What is a “fixed equatorial frame of reference”?

What is a “moving horizontal frame of reference”?
Answer
Our point of view is “operational”! We start from observations: It is a custom
to begin with the “Earth rotation” as all planets rotate: unfortunately the polar axis
is moving, the effects are called polar motion (POM) and Length of day (LOD).
Traditionally, we define the Mean Rotation Axis by averaging over time, for instance
over a year, over ten years or a decades or centuries. There are daily data available
from the International Earth Rotation Service, in particular for POM and LOD, also
for time. Various excitation mechanisms are studied, worth mentioning “atmospheric
tides”, tides of the solid part as well as the fluid part (sea tides), also the core-mantle
coupling and Earth Quakes.
The background of the other base vectors F•1 and F•2 in the equatorial plane,
orthonormal to the base vector F•3 is also fascinating: it takes reference to the Green-
wich zero meridian fixed to date it. In the center of time and space we define the
origin of the fixed equatorial triad or 3-leg. A parallel transparent to the center of the
Earth, we fix the Greenwich base vector F•1 as well as the orthogonal base vector F•2
orthonormal to F•1 and F•2
Answer
We have only discuses up to now the “fixed equatorial frame of reference”
{F•1 , F•2 , F•3 |0}. What is the definition of the “moving horizontal frame of reference”
{F∗1 , F∗2 , F∗3 |0} or {E∗1 , E∗2 , E∗3 |P}? A theodolite and a leveling instrument measure
the local physical vertical as well as local physical horizontal plane. With respect
to the so called “orientation unknown” the local vertical as well as the local hori-
zontal plane at the station point Pα we measure the modulus of the gravity vector
= grad W by gravimetry as well as astronomic latitude φ and astronomic
longitude , the spherical coordinate of the gravity vector. The moving horizontal
frame of reference at placement Pα is defined by {F∗1 , F∗2 , F∗3 |0} or {E∗1 , E∗2 , E3 |P}∗
in particular
• F∗1 orthogonal to the vector in the horizontal plane “orienta-
tion unknown”
• F∗2 orthogonal to the vector in the horizontal plane, orthog-
onal to F∗1 and F∗3
• F∗3 = E∗3 = (Pα ) ÷ α (Pα )
• E∗1 orthogonal to the vector in the horizontal plane, define
azimuth and vertical angle
• E∗2 orthogonal to the vector define azimuth and vertical angle
• E∗3 = F∗3 = (Pα ) ÷ α (Pα )
The various definitions of reference frames {E∗1 , E∗2 , E∗3 |P} or {F∗1 , F∗2 , F∗3 |P} as
well as {E•1 , E•2 , E•3 |0} or {F•1 , F•2 , F•3 |0} become more obvious when we study the
observations from a station point Pα to a target point Pβ , for instance. We introduce a
new index, also called “a dummy index”, for instance, Pαα versus Pβα . It is indicating
that both reference frames P at the points Pα and Pβ are the same! How to do this?
We apply parallel transport from Pα and Pβ of the reference frame attached to P! In
the Euclidean space in terms of three-dimensional Geodesy we think in terms of the
Euclidean Axiom 5 of the embedding space. Concretely we connect the orthonormal
reference frame {F∗1 , F∗2 , F∗3 |P} freely transported from O −→ P compared to the
orthonormal reference frame {F∗1 , F∗2 , F∗3 |P}
[F∗1 , F∗2 , F∗3 |P] = [F•1 , F•2 , F•3 |0]R

E ( , φ, ) (5.1)
denotes the “orientation unknown” typically for the chosen theodolite ori-
gin in the horizontal plane. The 3 parameters ( ,
, ) with in R( ,
, ) :=
R3 ()R2 ( π2 −
)R3 ( ) refer to Euler angles astronomical longitude , the com-
pliment of astronomical latitude
and the “orientation unknown” . Let us use the
transformation {F•1 , F•2 , F•3 |0} to {E•1 , E•2 , E•3 |0} reminding us the definition of dw
dt
,
namely the dot.
[F•1 , F•2 , F•3 ] = [E•1 , E•2 , E•3 ]R

E () (5.2)
a transformation caused by the orientation unknown. For local measurement tasks

we apply the south-North orientation of the local horizontal reference system
[E•1 , E•2 , E•3 ]. is the sign of the transposition. In contrast, for global measure-
ment problems the Greenwich orientation of the equatorial frame of reference is
advantages, namely orthogonal to the rotation axis of the Earth.
The global fixed reference frame F• produces holonomic, but the local moving
reference frame E∗ anholonomic coordinates, also called differential forms or leg
calculus (Fig. 17).
272 E. W. Grafarend
Fig. 17 Local reference

system
{E∗1 , E∗2 , E∗3 |P}, {F∗1 , F∗2 , F∗3 |P}
and global reference system
{F•1 , F•2 , F•3 |0}, angles
( ,
, )
With respect to the moving horizontal reference system we are able now to rep-
resent the relative vector “station point to target point” on the surface of the Planet
Earth
Xαβ = E1· X αβ + E2· Yαβ + E3· Z αβ (5.3)
X αβ = Sαβ cos Aαβ cos Bαβ

Yαβ = Sαβ sin Aαβ cos Bαβ (5.4)
Z αβ = Sαβ sin Bαβ
Aαβ = Hαβ + α = Arctan Yαβ ÷ Yαβ

Bαβ = Vαβ = Arctan Z αβ ÷ 2
X αβ + Yαβ
2
(5.5)

Sαβ = X αβ2
+ Yαβ
2
+ Z αβ
2
Aαβ denotes the southern azimuth form the station point X α to the target point
X β . The horizontal direction is abbreviated by Hαβ , namely form the point X α or
Pα to the point X β . The orientation are unknown at the point X α or Pα is called α .
The vertical direction directed upwards was denoted by the letter Bαβ . Finally, Sαβ
abbreviated the oblique distance. Figures 18 and 19 illustrate the relative position
vector in the moving frame of reference in the horizon. Please, note X βα = X αβ
because of different frames of reference.
Fig. 18 Relative position

vector X αβ in the horizontal
frame of reference Pα

vector X βα in the horizontal
frame of reference Pβ :
X βα = X αβ
5.2 Parallel Transport of Reference Frames
Due to the parallel transport of frames of reference, namely E∗α (X β ) := Eβ∗ and
E∗ (X α ) : E∗ , the reference frames enjoy the inequality Eα∗ := Eβ∗ , in consequence
Aαβ = Aβα − π, Bαβ = −Bβα (5.6)
as well as the anholonomic condition

⎡ β ⎤ ⎡ α ⎤
⎡ ⎤ ⎡ ⎤ X ∗ ∗
X βα X αβ ⎢ βα ⎥ ⎢ Xα αβ ⎥
⎢ β ⎥
− ⎣ Yβα ⎦ = ⎣ Yαβ ⎦ , − ⎢ Y ∗ ⎥ = ⎢ ∗ ⎥ (5.7)
Z βα E∗ (X Z αβ E∗ (X ) ⎣ βα ⎦ ⎣ Yααβ ⎦
β ∗
β)
Z∗
α
Z αβ
βα
Now we are in a position to interpret the commutative diagram in Figs. 18 and 19

better. The fixed frame of reference F• is always on top, but the moving frames of
reference not:
α β α
E∗
→ E∗
= R ( α ,
α , 0)RTE ( β ,
β , 0)
E∗ E
α γ α
E∗
→ E∗
= R ( α ,
α , 0)RTE ( γ ,
γ , 0)
E∗ E
274 E. W. Grafarend

vector (Pα , Pβ , Pγ ), moving
reference frames at the points
E3∗ (X α ), E3∗ (X β ), E3∗ (X γ )
The Heitz notation places the indicies {α, β, γ } on top of the kernel symbol. This
notation is needed to identify the reference frame attached to the point where we
measure.
Taylor expansions
We use the following Taylor expansions:
β = α + βα ,
β =
α +
βα
γ = α + γ α ,
γ =
α +
γ α
as well as

•
cos β = cos( α + βα ) = cos α cos βα − sin α sin βα = cos α − βα sin α
•
sin β = sin( α + βα ) = sin α cos βα − cos α sin βα = sin α − βα cos α ,

•
cos γ = cos( α + γ α ) = cos α cos γ α − sin α sin γ α = cos α − γ α sin α
•
sin γ = sin( α + γ α ) = sin α cos γ α − cos α sin γ α = sin α − γ α cos α ,
These are transformation elements of the astronomical longitude, a similar result

we find for astronomical latitude. The left-over transformations apply now for the
matrix products, for instance
⎡ ⎤
1 − sin
α −
•
R E ( α , α , 0)R E ( γ , γ , 0) = ⎣ sin
α
T
1 cos
α ⎦ ,

− cos
α 1
namely the sum of (i) a unit matrix I and of (ii) antisymmetric matrix (Fig. 20)
⎡ ⎤
0 − sin
α −
A := ⎣ sin
α 0 cos
α ⎦ = −AT .

− cos
α 0
Box 7: Transformation of rectangular and curvilinear coordinate differences

in two reference systems of the gravity space, Case one
rectangular coordinates in two moving reference systems

α −→ β
⎡ α ⎤ ⎡ β ⎤
∗ ∗
X αβ ⎢ X βα ⎥
α α α ⎢ α ⎥ α α α ⎢ β∗ ⎥
X αβ = −X βα = E 1∗ , E 2∗ , E 3∗ ⎢ ∗ ⎥
⎣ Yαβ ⎦ = − E 1∗ , E 2∗ , E 3∗ ⎢ ⎥=
α ⎣ Yβα ⎦
∗ β
Z αβ ∗
Z βα
⎡
β ⎤
∗
X
⎢ βα ⎥
α α α ⎢ β ⎥
= E 1∗ , E 2∗ , E 3∗ (I + A) ⎢ Y ∗ ⎥ (5.8)
⎣ βα ⎦
β
∗
Z βα
β −→ α
⎡ β ⎤ ⎡ α ⎤
∗ ∗
⎢ X βα ⎥ X αβ
⎢ β∗ ⎥ ⎢ α ⎥
⎢ Y ⎥ = R E ( β ,
β , 0)RTE ( α ,
α , 0) ⎢ ∗ ⎥
⎣ Yαβ ⎦ =
⎣ βα ⎦ α
β ∗
Z∗βα
Z αβ
⎡ α ⎤ ⎡ α ⎤
∗ ∗
X αβ X αβ
⎢ α ⎥ ⎢ α ⎥
= (I + A ) ⎢ ∗ ⎥ ⎢ ∗ ⎥
⎣ Yαβ ⎦ = (I − A) ⎣ Yαβ ⎦ (5.9)
α α
∗ ∗
Z αβ Z αβ
curvilinear coordinates in two moving reference systems

-horizontal direction, vertical direction-
⎡
β
∗
β
∗
⎢ A βα = H βα + = Arctan Y ÷ X
⎢
β
βα βα
⎢ β β
⎢ Bβα := Vβα = Arctan β Z βα ∗2
÷ X ∗2
+ Y ∗2 (5.10)
⎢ βα βα
⎣ β β β
∗2 ∗2 ∗2
Sβα = X βα + Yβα + Z βα
α α α
. ∗
− sin
α X αβ ∗
+Yαβ ∗
− cos
α Z αβ
Aβα = Hβα + β = Arctan ∗
α
∗
α
∗
α
X αβ + sin
α Yαβ +
Z αβ
. ∗

α X αβ ∗
− cos
α α Yαβ ∗
− α Z αβ
Bβα := Vβα = Arctan
α α α α α α (5.11)
∗2 ∗2 ∗ ∗ ∗ ∗
X αβ +Yαβ −2 cos
α Yαβ Z αβ +2
X αβ Z αβ

α α α
∗2 ∗2 ∗2
Sβα = Sαβ = X αβ + Yαβ + Z αβ
276 E. W. Grafarend
End Box 7
Box 8: Transformation of rectangular and curvilinear coordinate differences
in two reference systems of the gravity space, Case two
rectangular coordinates in two moving reference systems

β −→ α
⎡ β
X βγ = Sβγ cos Aβγ cos Bβγ
⎢β
⎢ (5.12)
⎣ Y βγ = Sβγ sin Aβγ cos Bβγ
β
Z βγ = Sβγ sin Bβγ
α −→ β
⎡ α β β β
X βγ = X βγ − αβ sin
α Y βγ −
αβ Z βγ
⎢α β β
⎢ (5.13)
⎣ Y βγ = αβ sin
α X βγ − αβ cos
α Z βγ
α β β β
Z βγ =
αβ X βγ − αβ cos
α Y βγ + Z βγ
curvilinear coordinates in two moving reference systems

-horizontal direction, vertical direction-
αβ := β − α
αβ :=
β −
α
(Taylor expansion)
⎡
β β
⎢ A βα = H βα + = Arctan Y βα ÷ X βα
β
⎢ β β β
⎢
⎢ Bβα = Vβα = Arctan Z βγ 2
÷ 2
X βγ + Yβγ2 (5.14)
⎢
⎣ β β β
Sβα = X βγ 2
+ Yβγ2
+ Z βγ
2
α α α
. − αβ sin
α X βγ + Y βγ − αβ cos
α Z βγ
Aβγ = Hβγ + β = Arctan α α α
X βγ + αβ sin
α Y βγ +
αβ Z βγ
α α α
. −
αβ X βγ + αβ cos
α Y βγ + Z βγ
Bβγ = Vβγ = Arctan
α α (5.15)
α α α α
2
X βγ +Yβγ
2
−2 αβ cos
α Y βγ Z βγ +2
αβ X βγ Z βγ

α α α
Sβα = 2
X βγ + Yβγ
2
+ Z βγ
2
⎛ ⎞
α α α α α
. Y βγ ⎜ X βγ Z βγ ⎟ Y βγ Z βγ
Aβγ = Hβγ + = Arctan α −⎜
⎝sin
α + α
⎟
α cosα ⎠ αβ − α α
αβ
β
X βγ 2 + Y2
X βγ 2 + Y2
X βγ
βγ βγ
(5.16)
α α α
. Z βγ Y βγ X βγ
Bβγ = Vβγ Arctan + cos
α αβ −
αβ (5.17)
α α α α α α
2 + Y2
X βγ X 2 + Y2 X 2 + Y2
βγ βγ βγ βγ βγ
End Box 8
Obviously, the Euclidean distance is independent of the choice of the reference
frame. The results is in contrast to the analysis of the observed directions of type
horizontal and vertical. The observational equations for horizontal and vertical direc-
tions contain three orientation angles, indeed the classical orientation unknown β at
the station point Pβ , or Pβ as well as the longitudinal and latitudinal difference αβ ,
αβ or αγ ,
αγ between the horizontal reference system of the station point and
those points which we chose as reference points in a triangular network. In addition,
we have modeled the coordinate differences between station point and target point.
Two numerical examples
At the beginning of the section, we discuss two Earth models:

(i) If the Earth would be spherical or a sphere of constant radius, the
result at the first pages w0 = gm/r0 are applicable: r = gm/w0 .
indeed we would have no anholonomity.
(ii) If the Earth would be ellipsoidal or a rotational symmetric 2-axis
ellipsoidal of type Somigliana–Pizzetti of radius (a0 , b0 ) (semi-
major radius, semi-minor radius), the result of the Chaps. 2–5 are
applicable: we would have no anholonomity.
Obviously the real irregular Earth figure is more complex due to the heteroge-
neous, inhomogeneous mass distributions and the complex Earth rotation, namely
observed Length-of-Day variations (LOD) and Polar Motion (POM). For our first
example, we start with a triangular geodetic network between the point {Pα , Pβ , Pγ }
in the horizontal reference frame of the point Pα . The three holonomity conditions of
Box 9 express the loop condition for our 25 m local network to zero. In addition, we
assume vertical deflection variation of the order or 1 , 05 , 2.5 for longitude and
latitude of an approximate latitude
α = 48.783◦ .
278 E. W. Grafarend
Fig. 21 Triangular network

calculated in the horizontal
frame of reference at the
point Pα
Box 9: 25 m local network

α α α
X αβ = +30 m, X βγ = +50 m, X γ α = −80 m,
α α α
Y αβ = +30 m, Y βγ = −50 m, Y γ α = +20 m,
α α α
Z αβ = +05 m, Z βγ = +15 m, Z γ α = −20 m,
αβ = 1“ ∼ 4.85 · 10−6 R AD

αβ = −0.5“ ∼ − 2.42 · 10−6 R AD
αγ = −1“ ∼ − 4.85 · 10−6 R AD

αγ = −2.5“ ∼ −12.12 · 10−6 R AD

α = −48.788◦
End Box 9
Box 10: 25 m local network, detailed computation

⎡ ⎤
⎡ ⎤ ⎡ ⎤ α
X βγ 1 + αβ sin
α +
αβ X
⎢ α βγ ⎥
⎣ Yβγ ⎦ = ⎣ − αβ 1 − αβ cos
α ⎦ = ⎢ ⎥
⎣ Y βγ ⎦ =
Z βγ −
αβ + αβ cos
α 1 α
⎡ ⎤⎡ Z βγ⎤
1 +3.65 · 10 −2.42 · 10−6
−6
+50 m
⎣ −3.65 · 10−6 1 −3.19 · 10−6 ⎦ ⎣ −50 m ⎦
−6 −6
+2.42 · 10 +3.19 · 10 1 +15 m
⎡ ⎤ ⎡α ⎤
γ ⎡ ⎤
X
⎢γ ⎥γα 1 + αγ sin
α +
αγ ⎢ Xα γ α ⎥
⎢ ⎥ = ⎣ − αγ −
⎦ = ⎢ ⎥
⎣ Yγα ⎦ 1 αγ cos α ⎣ Yγα ⎦ =
γ −
αγ + αγ cos
α 1 α
Z⎡γ α ⎤⎡ Z γ α⎤
1 −3.65 · 10−6 −12.12 · 10−6 −80 m
⎣ +3.65 · 10−6 1 +3.19 · 10−6 ⎦ ⎣ +20 m ⎦
−6 −6
+12.12 · 10 −3.19 · 10 1 −20 m
End Box 10
Box 11: 25 m local network, misclosures
β γ
X βγ = +50 m − 0.22 mm, X γ α = −80 m + 0.17 mm
β γ
Y βγ = −50m − 0.23 mm, Y γ α = +20 m − 0.36 mm
β γ
Z βγ = +15 m − 0.04 mm, Z γ α = −20 m − 1.03 mm
α β
X αβ + X βγ +γ X γ α = 0 : −0.22 mm + 0.17 mm = −0.05 mm
α β
Y αβ + Y βγ +γ Yγ α = 0 : −0.23 mm − 0.35 mm = −0.59 mm
α β
Z αβ + Z βγ +γ Z γ α = 0 : −0.04 mm − 1.03 mm = −1.07 mm
End Box 11
Have a look at the triangular network of the Fig. 21. Box 10 reviews the transfor-
mation of the frame of reference at the point P β and P γ to the frame of reference
at out datum point P α . In addition, Box 11 summarizes the misclosures of the order
of under 1mm. For our second example, we start again with a triangular geodetic
network between the points {Pα , Pβ , Pγ } in the horizontal frame at the point Pα . The
three holonomity conditions for our 500 m local network to zero, see Box 12. In ad-
dition, we assumed vertical deflection variation of the order of 2 − 5 for longitude
and latitude of approximate latitude
α = 48.782◦
Box 12: 500 m local network
α α α
X αβ = +500m, X βγ = +800m, X γ α = −1300m,
α α α
Y αβ = +500m, Y βγ = −800m, Y γ α = +300m,
α α α
Z αβ = +50m, Z βγ = +150m, Z γ α = −200m,
αβ = 25“ ∼ 12.12 × 10−5 R AD

αβ = −15“ ∼ − 7.27 × 10−5 R AD
αγ = −15“ ∼ − 7.27 × 10−5 R AD

αγ = −45“ ∼ − 2.18 × 10−4 R AD

α = 48.788◦
280 E. W. Grafarend
End Box 12
Box 13: 500 m local network, detailed computation

⎡ β ⎤ ⎡ ⎤
⎡ ⎤ α
⎢β ⎥
X βγ 1 + αβ sin
α +
αβ ⎢ X βγ ⎥
⎢ ⎥ = ⎣ − αβ 1 − cos
⎦ = ⎢ Yα ⎥ =
⎣ Y βγ ⎦ αβ α ⎣ βγ ⎦
β −
αβ + αβ cos
α 1 α
Z⎡βγ
Z βγ
⎤⎡ ⎤
1 +9.12 · 10−5 −7.27 · 10−5 +800 m
⎣ −9.12 · 10−5 1 −7.99 · 10−5 ⎦ ⎣ −800 m ⎦
−5 −5
+7.27 · 10 +7.99 · 10 1 +150 m
⎡γ ⎤ ⎡α ⎤
⎡ ⎤
X
⎢γ ⎥γα 1 + αγ sin
α +
αγ ⎢ Xα γ α ⎥
⎢ ⎥ = ⎣ − αγ −
⎦ = ⎢ ⎥
⎣ Yγα ⎦ 1 αγ cos α ⎣ Yγα ⎦ =
γ −
αγ + αγ cos
α 1 α
Z⎡γ α ⎤⎡ Z γ α⎤
1 −5.47 · 10−5 −2.18 · 10−4 −1300 m
⎣ +5.47 · 10−5 1 +4.79 · 10−5 ⎦ ⎣ +300 m ⎦
−4 −5
+2.18 · 10 −4.79 · 10 1 −200 m
End Box 13
Box 14: 500 m local network, misclosures

β γ
X βγ = +800 m − 83.9 mm, X γ α = −1300 m + 27.2 mm
β γ
Y βγ = −800 m − 84.9 mm, Y γ α = +300 m − 80.7 mm
β γ
Z βγ = +150 m − 5.8 mm, Z γ α = −200 m − 297.8 mm
α β γ
X αβ + X βγ + X γ α = 0 : −83.9 mm + 27.2 mm = −56.7 mm
α β γ
Y αβ + Y βγ + Y γ α = 0 : −84.9 mm − 80.7 mm = −165.6 mm
α β γ
Z αβ + Z βγ + Z γ α = 0 : −5.8 mm − 297.8 mm = −303.6 mm
End Box 14
Box 13 reviews the transformation of the frame of reference at the points P β and
P to the frame of reference at the datum point P α . In addition, Box 14 summarizes
γ
the misclosure of the order 250 mm. The results speak for themselves. For observation
in the range of kilometers (3rd order networks) these effects cannot be neglected.
5.3 The Object of Anholonomity in Terrestrial Networks,

Cartan Exterior Calculus, an Example
Our example documents the anholonomity of a spherical coordinate system once we

use an orthonormal frame of reference, a Cartan 2-leg, namely an three-dimensional
Euclidean space and its embedding into a two-dimensional Riemann space. It is the
embedding of 2-space within Euclidean geometry. Our point of view in extrinsic.
Box 15: Gauss surface geometry, Cartan surface geometry orthonormal

frame of reference, example of the sphere
x(u, v) = r cos u cos ve1 + r sin u cos ve2 + r sin ve3

∂x
g1 = ∂u = r cos v(−e1 sin u + e2 cos u)
∂x
g2 = ∂v = −r sin v cos ue1 − r sin v sin ue2 + r cos ve3
g3 = v(e1 cos u + e 2 sin u) + e3 sin v
cos
g1 = r cos v, g2 = r, g3 = 1, g1 |g2
= 0
c1 := gg11 , c2 := gg22 , c3 := g3
End Box 15:
Box 16: Cartan’s first structure equations, displacement vector of the surface
of the sphere Gaussian frame of reference versus Cartan frame of reference,
integrability
“directional equations of first kind”

∂x
dx = du + ∂x dv = c1r cos vdu + c2 r dv = σ 1 c1 + σ 2 c2
1 ∂v
∂u
σ r cos v 0 du ab du
= =
σ2 0 r dv cd dv
−1 1 −1
1
du ab σ (r cos v) 0 σ
= =
dv cd σ2 0 r −1 σ2
du = (r cos v)−1 σ 1 , dv = r −1 σ 2 ↔ σ 1 = r cos vdu, σ 2 = r dv
integrability?
∂σ 1
− ∂σ = −r sin v = 0
2
∂v1 ∂u
σ du
=: σ , du :=
σ2 dv
if σ = A du , then dC σ = d A ∧ du
End Box 16
First, we span the tangent space-here denoted {g1 , g2 }- as well as the normal space-
here denoted {g 3 } in honor
of
C.F. Gauss (Mathematician, Geodesist, Physicist). The
L 2 − nor ms g1 and g1 as well as the unweighted scalar product < g1 , g2 >, in
addition g3 = 1 prove orthogonality of the vectors of the tangent space elements,
but miss normality. That is done by defining the Cartan 2-frame denoted by {e1 , e2 }
which is orthonormal. We have collected the details in Box 15.
282 E. W. Grafarend
Second, in Box 16 we present Cartan’s “derivational equations of the first kind”.

Once we denote longitude by the letter u and latitude by the letter v we are in a position
the base {w , w } by the transformation
1 2
to design an orthonormal 2-leg by introducing

ab
(du, dv) −→ (ω1 , ω2 ) by the matrix. .{ω1 , ω2 } are the elements of the Cartan
cd
2-leg. Here the matrix elements are differential factors (a b) = (a, 0), (c, d) = (0, d)
−1
a b (r cos v)−1 0
and a = r cos v, d = r or =
c d 0 r −1
ω1 = r cos vdu, ω2 = r dv
The differential forms {ω1 , ω2 } the question:

are the differential forms integrable?
Can we interchange the differentiation order?
∂ω1 ∂ω2
− = −r sin v = 0 for v = 0 (5.18)
∂v ∂u
The result
1 is no. Such a result is typical for anholonomity.
Let us collect the differential
ω du
form is the column vector, in contrast the column vector du as well as
ω2 dv
the matrix
ab
(5.19)
cd
We can summarize the result in the Cartan derivative or differential:
if ω = A du, then dω = dA ∧ du (5.20)
Box 17: Cartan’s first structure equation 1-differential forms, exterior

calculus Cartan-derivative
“1-differential form”
σ 1 = a du + b dv = aα du α
σ 2 = c du + d dv = bα du α
“exterior” or E. Cartan-derivative
2
1 α α β
dC σ α = σβ ∧ σγ
2! βγ
βγ σ ∧ σ γ
β,γ =1 β<γ
!anti-symmetry!
du α ∧ du β = −du β ∧ du α
σ α = aβα du β ∼ σ = A du
three-index-symbol
2
dC σ α = σβ ∧ σγ
β,γ
dσC1 = 21 112 σ 1 ∧ σ 2 + 21 121 σ 2 ∧ σ 1 = 21 (112 − 121 )σ 1 ∧ σ 2 = 112 σ 1 ∧ σ 2
dσC2 = 21 212 σ 1 ∧ σ 2 + 21 221 σ 2 ∧ σ 1 = 21 (212 − 221 )σ 1 ∧ σ 2 = 212 σ 1 ∧ σ 2

σ 1 ∧ σ 2 = −σ 2 ∧ σ 1
comparison of coefficients
112 = −121 = r1 tan v
212 = −221 = 0
assumption : dC du = 0
proof : dC du = ( ∂du − ∂dv )dv ∧ du = 0
∂v ∂v
−r sin vdv ∧ du
dC σ = d A ∧ du =
0
1
− r tan vσ 2 ∧ σ 1
dC σ = = 0
0
“summary”
2
dσCα = αβγ σ β ∧ σ γ
β,γ =1
α β
= βγ σ ∧ σ γ
β<γ
End Box 17
Finally, we dig into the definition of the Cartan derivative. We ask what is the
measure of anholonomity or what is the three-index-symbol of anholonomity. Please,
for a fast information consult Box 17 at this end. In the case of spherical coordinates
of type longitude and latitude, we can represent the Cartan derivative by dC ω =
[− r1 tan v w 2 ∧ w 1 ]
6 Final Comments
At this end, we want to illustrate the intimate relation between Geodesy and Physics:
Our target is the anholonomity problem in the classical and relativistic
gravity field including numerical examples. It applies for relativistic Real
Null Frames and coframes, applying for instance for GPS. It is a measure
how well the assumed physical and geometric field fits a sphere or an
ellipsoid or an irregular boundary of the planet Earth.
We come back to the recommendation of the International Union of Geodesy
and Geophysics (IUGG) to adopt the International Reference Ellipsoid as early as
1924. The International Gravity Formula in terms of the axisymmetric, non-harmonic
Somigliana–Pizzetti gravity field was accepted as early as 1930 and recommended
by the International Association of Geodesy (IAG).
Ever since geodesists did hesitate to use it, but instead the spherical symmetric for
99% the spherical symmetric Newton’s gravitational field which is harmonic outside
the spherical Earth of constant radius. Series expansions to degree/order 7200 are
284 E. W. Grafarend
nowadays a standard. Even at our conference on Relativistic Geodesy two speakers

declared in public that Geodesy does not need Relativity.
As everyone can see it: the Planet Earth as all terrestrial planets rotates!
We recommend therefore to the International Association of Geodesy (IAG) to
adopt the axisymmetric Kerr metric or its linear approximation officially. Perhaps
our detailed report helps.
Recommendation to the IAG
Adopt the axisymmetric Kerr metric or its linear approximation

Lense-Thirring for a rotating-gravitating Earth
We are relatively certain that only few Geodesists as well as Physicists understand
the topic anholonomity. Simply speaking, it is the problem to decide what is the
correct reference figure in Geometry Space or in Gravity Space. For instance, if the
Earth were not rotating and has a radial symmetric gravity field, namely of potential
type gm/r0 , the Earth could not have an anholonomic coordinate system. But for a
real Earth it would be a central question if the mass distribution of the Earth or any
other planet would depend on longitude and latitude, not only radial. The figure of
the Earth would be best described by a dependence of radial, longitudinal as well
as latitudinal coordinates. Such an Earth model would be free of anholonomity. At
least, the celebrated axisymmetric Somigliana–Pizzetti and a series development in
terms of degree/order 3600 or 7200 of its perturbed gravity field, of course, in terms
of ellipsoidal harmonics, could solve many topographic problems. In particular, the
disturbances are the proper key for solving global, regional, nearly local geodetic
problems. It has to be realized that the disturbance potential, vertical deflections and
the modules of the second derivatives of the anomalous gravity field and its higher
derivatives as well as higher derivatives cause the effect of anholonomity: surveying
engineers as well as the common geodesists are not familiar with the concept of
holonomic and anholonomic coordinates, of the leg-calculus or “differential forms”
applied to geodetic sciences. It is for this community that we list important geodetic
contributions on this subject:
A. Marrussi [14, 18, 19, 36–42], B. Chovitz [43–45], E. Doukakis [46], E.
Grafarend [15, 47–55], E. Grafarend and W. Kuehnel [5], E. Livieratos [56],
F. Bocchio [57–63], F. Sanso [64], F. Sanso, and P. Yanicek [7], G. Ricci [65,
66], G. Ricci and T. Levi-Civita [67], H. Flanders [68], H. Weyl, [69, 70], J.
A. Schouten [8, 71], J. D. Zund [1, 72–78], J. D. Zund and W. Moor, [79], J. D.
Zund and J. M. Wilkes [80]. J. M. Wilkes, and J. D. Zund [81], M. Blagojevic
,J. Garecki,F.W. Hehl, and Y.N. Obukhov [3], M. Fujimoto and E. Grafarend
[34], M. Hotine [20, 82–86], N. Grossman [28, 29, 87, 88], P. Defrise [89,
90], P. Defrise and E. Grafarend [16, 91], P. Holota and Z. Nadenik [92], P.
Pizzetti [93–95], P. Stäckel [96, 97], P. Vanicek [98], S. Roberts [99, 100],
Teunissen [2], V. Schwarze [101], W. Neutsch [102], Y. Georgiadou and E.
Livieratos [103], Z. Nadenik [104],
Dedication
Our contribution is dedicated to late Josef D. Zund, mathematician, in par-

ticular for the book “Foundations of Differential Geodesy”, Springer Verlag,
Berlin-Heidelberg-New York 1994, namely for introducing to geodesists the
notion of the general Leg Calculus in terms of anholonomic coordinates, bet-
ter called differential forms in 2-,3-, and 4-dimensions, the Hotine-Marussi
approach, so-called “Three Dimensional Geodesy”, 5 basic forms. To read
his chapter VII, The “Fundamental Theorem of Differential Geodesy”, in par-
ticular the celebrated Pizzetti sphere, the Pizzetti inequality, the bifurcation
process for coupling gravitation and rotation as well as the algebraic theory
of the Marussi tensor of gravity gradients in chapter VIII, is a special high-
light for geodesists and physicists. His chapter X on the hierarchy of geodetic
reference systems is magic.
Acknowledgements C. Lämmerzahl and D. Pützfeld, /Bremen/ invited me to speak about an-

holonomity in the context of Relativistic Geodesy within the WE-Heraeus Seminar. Special thanks
go to D. Pützfeld, F.W. Hehl/Cologne/ and H. Quevedo/Mexico City/ for their helpful comments.
In addition, I am grateful for the support of J. Müller (Hanover) on the International Reference
Ellipsoid and to S. M. Kopeikin (Columbia/Missouri) on studying Relativistic Equilibrium Figures
and the relativistic theory of the Geoid. Last, but not least, I am grateful to M.A. Javaid(Stuttgart)
for his expert typing.
References
1. J.D. Zund, Foundations of Differential Geodesy (Springer, Berlin, 1994)

2. P.J.G. Teunissen, Anholonomity when using the development method for the reduction of
observations to the reference ellipsoid. Bull. Geod. 56, 356–363 (1982)
3. M. Blagojevic, F.W. Hehl, J. Garecki, Y.N. Obukhov, Real null coframes in general relativity
and GPS coordinates. Phys. Rev. 65(4), 044018 (2002)
4. E.W. Grafarend, F.W. Krumm, Map Projections: Cartographic Infromation Systems (Springer,
Berlin, 2006)
5. E.W. Grafarend, W. Kuehnel, A minimum atlas for the rotation group SO(3). J. Geomath.
2(2011), 113–122 (2011)
6. E.W. Grafarend, A.A. Ardalan, Ellipsoidal geoidal undulations (ellipsoidal Bruns formula):
case study. J. Geod. 75, 544–552 (2001)
7. F. Sanso, P. Vanicek, The orthometric height and the holonomity problem. J. Geod. 80, 225–
232 (2006)
8. J.A. Schouten, Ricci-Calculus, 2nd edn. (Springer, Heidelberg, 1954), p. 127
9. R. Cushmann, H. Duistermaat, J. Sniatycki, Geometry of Nonholonomically Constrained
Systems (World Scientific, Singapore, 2010), p. 404
10. S. Sternberg, Curvature in Mathematical and Physics (Dover Publication, Mineola, 2012)
11. F. Hehl, Der Spin und Torsion in der Allgemeinen Relativitätstheorie. Abh. Braunschweigis-
che Wiss. Ges. 18, 98 (1966)
12. F. Hehl, Der Spin und Torsion in der Allgemeinen Relativitätstheorie. Universität Clausthal,
Habilitationsschrift Techn (1970)
13. F. Heh, E. Kröner, Über den Spin in der Allgemeinen Relativitätstheorie. Z. Phys. 187, 478
(1965)
286 E. W. Grafarend
14. A. Marussi, Fondements de geometrie differentielle absolue du champ potential terrestre.

Bull. Geod. 14, 411–439 (1949)
15. E.W. Grafarend, Three dimensional geodesy I: the holonomity problem. Z. Vermesssungswe-
sen 100, 269–281 (1975)
16. P. Defrise, E.W. Grafarend, Torsion and anholonomity of geodetic frames. Bollettino di Geode-
sia e Scienze Affini 35, 153–160 (1976)
17. M. Caputo, The Gravity Field of the Earth (Academic, New York, 1967)
18. A. Marussi, Natural reference systems and their reciprocals in geodesy, Technical report, Publ.
T.J, Kukkamaki 70th Birthday, Publ. Finnish Geodetic Institute, Nr. 89, Helsinki (1979)
19. A. Marussi, Intrinsic Geodesy (trans: Reilly WI) (Springer, Berlin, 1985)
20. M. Hotine, in Differential Geodesy (Edited with commentary by J. D. Zund) (Springer, Berlin,
1991)
21. H. Moritz, The Hamiltonian structure of refraction and the gravity field. Manuscripta Geod.
20, 52–60 (1994)
22. H. Goenner, E.W. Grafarend, R.J. You, Newton mechanics as geodesic flow on maupertuis’
manifold; the local isometric embedding into flat spaces. Manuscripta Geod. 19, 339–345
(1994)
23. E.W. Grafarend, Differential geometry of the gravity field. Manuscripta Geod. 11, 29–37
(1986)
24. E. Hunziker, Lotlinienkrünunung und Projektion eines Punktes oder einer Strecke auf das
Geoid, Schweiz. Z Vermessung. Kulturtechnik und Photogrammetrie 58, 144–152 (1960)
25. J. Engels, E. Grafarend, The gravitational field of topographic/isostatic masses and the hy-
pothesis of mass condensation. Surv. Geophys. 140, 495–524 (1993)
26. J. Engels, E. Grafarend, P. Sorcik, The gravitational field of topographic-isostatic masses and
the hypothesis of mass condensation II - the topographic-isostatic Geoid. Sun Geophys. 17,
41–66 (1996)
27. E.W. Grafarend, J. Engels, P. Sorcik, The gravitational field of topographic/isostatic masses
and the hypothesis of mass condensation. Part I and II, Technical report, Department of
Geodesy, Stuttgart (1995)
28. N. Grossman, Holonomic measurables in geodesy. J. Geophys. Res. 79, 689–694 (1974)
29. N. Grossman, The pull-back operation in physical geodesy and the behaviour of plumblines.
Manuscripta Geod. 3, 55–105 (1978)
30. I.I. Mueller, E. Grafarend, H.B. Papo, B. Richter, Investigations on the hierarchy of reference
frames in geodesy and geodynamic, Technical Report 289, Department of Geodetic Science,
The Ohio State University (1979)
31. I.I. Mueller, E. Grafarend, H.B. Papo, B. Richter, Concepts for reference frames in geodesy
and geodynamics: the reference directions. Bull. Geod. 53(289), 195–213 (1979)
32. E.W. Grafarend, Spacetime geodesy. Boll, di Geodesia e Scienze Affini 38, 551–589 (1975)
33. E.W. Grafarend, Die Beobachtungsgleichungen der dreidimen-sionalen Geodasie im
Geometrie- and Schwereraum, ein Beitrag zur operationellen Geodasie. Zeitschrift für Ver-
messungswesen 106, 411–429 (1981)
34. M. Fujimoto, E. Grafarend, Spacetime coordinates in the geodetic reference frame. in Rel-
ativity in Celestial Mechanics and Astronomy, ed. by J. Kovalevsky, V.A. Brumberg (IAU,
1986), pp. 269–276
35. R. Penrose, W. Rindler, Spinors and Space-Time, vol. 1 (Cambridge University Press, Cam-
bridge, 1987)
36. A. Marussi, Les principes de la geodesie intrinseque. Bull. Geod. 19, 68–76 (1951)
37. A. Marussi, Fondamenti di geodesia intrinseca. Publicazioni della Commissione Geodetica
Italiana, Ser. HI 7, 1–47 (1951)
38. A. Marussi, Su alcune propriety integrali delle rappresen-tazioni conformi di superfici su su-
perfici, rendiconti della classe di scienze fisiche, matematiche e naturali. Accademia Nazionale
dei Lincei (Roma), Ser. VIII 10, 307–310 (1951)
39. A. Marussi, La coordination des system geodesiques. Bull. Geod. 43, 16–19 (1957)
40. A. Marussi, Dalla geodesia classica alla geodesia in tre dimensioni. Bollettini di Geodesia e
Scienze Affini, anno XVIII, 485–495 (1959)
41. A. Marussi, The tidal field of a planet and the related intrinsic reference systems. Geophys.
J. R. Astron. Soc. 56, 409–417 (1979)
42. A. Marussi, Intrinsic geodesy (a revised and edited version of his 1952 lectures by J.D. Zund),
Technical report, Report No. 390, Department of Geodetic Science and Surveying, The Ohio
State University, Columbus (1988)
43. B.H. Chovitz. Hotine’s mathematical geodesy, in IV Symposium on Mathematical Geodesy
(1969), pp. 159–172
44. B.H. Chovitz, Generalized three-dimensional conformal transformations. Bull. Geod. 104,
159–163 (1972)
45. B.H. Chovitz, The influence of Hotine’s mathematical geodesy. Bollettino di Geodesia e
Scienze Affini, anno XLI, 57–64 (1982)
46. E. Doukakis, Remark on time and reference frames. Bull. Geod. 81 (1978)
47. E.W. Grafarend, The object of anholonomity and a generalized Riemannian geometry for
geodesy. Bollettino di Geofisica Teorica ed Applicata 13, 241–253 (1971)
48. E.W. Grafarend, Three dimensional geodesy and gravity gradients, Technical report, Ohio
State University, report no. 174, Columbus, Ohio, USA (1972)
49. E.W. Grafarend, Le theoreme de conservation de la courbure et la torsion or attempts at a
unified theory of geodesy. Bull. Geod. 109, 237–260 (1973)
50. E.W. Grafarend, Gravity gradients and three dimensional net adjustment without ellipsoidal
reference, Technical report, The Ohio State University, Report No. 202, Columbus (1973)
51. E.W. Grafarend, in Cartan frames and a foundation of Physical Geodesy, Methoden and
Verfahren der Mathematischen Physik, Bd 12, ed. by B. Brosowski, E. Martensen. BI-Verlag,
Mathematical Geodesy, Mannheim (1975), pp. 179–208
52. E.W. Grafarend, Threedimensional geodesy iii: refraction. Bollettino di Geodesia e Scienze
Affini 35, 153–160 (1976)
53. E.W. Grafarend, Geodasie - Gausssche oder Cartansche Flaähengeometrie? Allgemeine
Vermessungs-Nachrichten 4, 139–150 (1977)
54. E.W. Grafarend, Der Einfluss der Lotrichtung auf lokale geodätische Netze. Z. Vermes-
sungswesen 112, 413–424 (1987)
55. E.W. Grafarend, Tensor algebra, linear algebra, multi-linear algebra, Technical report, 344
references, Department of Geodesy and Geoinformatics, Stuttgart University Stuttgart (2004)
56. E. Livieratos, On the geodetic singularity problem. Manuscripta Geod. 1, 269–292 (1976)
57. F. Bocchio, Su alcune applicazioni di interesse geode-tico delle connessioni non simmetriche,
Rendiconti della classe di scienze Fisiche, matematiche e naturali. Accademia Nazionale dei
lincei (Roma) Ser. VIII 48, 343–351 (1970)
58. F. Bocchio, From differential geodesy to differential geophysics. Geophys. J. R. Astron. Soc.
39, 1–10 (1974)
59. F. Bocchio, The holonomity problem in geophysics. Bollettino di Geodesia e Scienze Affini,
anno XXXIV, 453–459 (1975)
60. F. Bocchio, Some of Marussi’s contributions in the field of two-dimensional and three di-
mensional representation. Bollettino di Geodesia e Scienze Affini, anno XXXVII, 441–450
(1978)
61. F. Bocchio, An inverse geodetic singularity problem. Geophys. J. R. Astron. Soc. 67, 181–187
(1981)
62. F. Bocchio, Geodetic singularities in the gravity field of a non-homogenious planet. Geophys.
J. R. Astron. Soc. 68, 643–652 (1982)
63. F. Bocchio, Geodetic singularities, reviews of geophysics and space physics, in Advances in
Geodesy, vol. 20, ed. by R.H. Rapp, E.W. Grafarend (American Geophysical Union, Wash-
ington, 1982), pp. 399–409
64. F. Sanso, The geodetic boundary value problem in gravity space, Memorie Scienze Fisiche
(Academia Nationale dei Lincei, Roma, 1997)
65. G. Ricci, Lezioni sulla teoria delle superficie (F. Drucker, Verona-Padova, 1989)
288 E. W. Grafarend
66. G. Ricci, Opere, in Edizioni Cremonese, 2 (1956/1957)

67. G. Ricci, T. Levi-Civita, Methodes de calcul differentiel absolu et leurs applications. Math.
Ann. 54, 125–201 (1901)
68. H. Flanders, in Differential Forms with Applications to the Physical Sciences. (Academic,
New York, 1963)
69. H. Weyl, Raum-Zeit-Materie: Vorlesungen überiber allgemeine Relativitätstheorie, 5th edn.
(Springer, Berlin, 1918)
70. H. Weyl, Gruppentheorie and Quantenmechanik (Verlag S. Hirzel, Leipzig, 1928)
71. J.A. Schouten, Tensor Analysis for Physicists (Clarendon, Oxford, 1951)
72. J.D. Zund, Tensorial methods in classical differential geometry - i: basic principles. Tensor
NS 47, 74–82 (1988)
73. J.D. Zund, Tensorial methods in classical differential geometry - i: basic surface tensors.
Tensor NS 47, 83–92 (1988)
74. J.D. Zund, Differential geodesy of the Eötvös torsion balance. Manuscripta Geod. 14, 13–18
(1989)
75. J.D. Zund, The assertion of hotine on the integrabil-ity conditions in his general coordinate
system. Manuscripta Geod. 15, 373–382 (1990)
76. J.D. Zund, An essay on the mathematical foundations of the Marussi-Hotine approach to
geodesy. Bollettino di Geodesia e Scienze Affini anno XLIX, 113–179 (1990)
77. J.D. Zund, The Hotine problem in differential geodesy. Manuscripta Geod. 15, 373–382 (1990)
78. J.D. Zund, The mathematical foundations of the hotine-marussi approach to geodesy. Bollet-
tino di Geodesia e Scienze Affini anno LI, 125–138 (1992)
79. J.D. Zund, W. Moore, Hotine’s conjecture in differential geodesy. Bull. Goodesique 61, 209–
222 (1987)
80. J.D. Zund, J.M. Wilkes, The significance and generalization of two transformation formulas
in Hotine’s mathematical geodesy. Bollettino di Geodesia e Scienze Affini anno XLVII, 77–85
(1998)
81. J.M. Wilkes, J.D. Zund, Group-theoretical approach to the Schwarzschild solution. Am. J.
Phys. 50, 25–27 (1982)
82. M. Hotine, Metrical properties of the Earth’s gravitational field, report to i.a.g, Technical
report, Toronto Assembly 33-64 of Hotine (1957)
83. M. Hotine, Geodesic coordinate systems, Technical report, Venice Symposium, 65-89 of
Hotine (1957)
84. M. Hotine, A primer on non-classical geodesy, Technical report, Venice Symposium, 91–130
(1959)
85. M. Hotine, The orthomorphic projection of the spheroid. Emp. Surv. Rev. 8, 300–311 (1967)
86. M. Hotine, Mathematical geodesy, U.S. Department of Commerce Washington, D.C. (1969)
87. N. Grossman, Is the geoid a trapped surface? Bollettino di Geodesiae Scienze Affini anno
XXXIV, 173–183 (1978)
88. N. Grossman, The nature of space near the Earth. Bollettino di Geodesia e Scienze Affini
anno XXXV, 413–424 (1979)
89. P. Defrise, Meteorologie et geometrie differentiae. Institute Royal Moteorologique de Bel-
gique, Bruxelles, A 91 (1975)
90. P. Defrise, Sur des applications de la geometrie differentielle en Meteorologie et en Goodesie.
Bollettino di Geodesia e Scienze Affini 37, 185–196 (1978)
91. P. Defrise, E.W. Grafarend, Torsion and Anholonomity of geodetic frames. Bollettino di
Geodesia e Scienze Affini 35, 81–92 (1976)
92. P. Holota, Z. Nadenik, Les formes differentielles exterieures dans la Geodesie ii: Courbure
moyenne. Studia geoph. at geod. 15, 106–112 (1971)
93. P. Pizzetti, Un principio fondamentale nello studio delle studio delle superfici di livello ter-
restri. Rendiconti della Reale Accademia dei Lincei (Roma) 10, 35–39 (1901)
94. P. Pizzetti. Höher Geodäsie, Enzyklopädie der Mathematischen Wissenschaften Band VI.
Geodasie und Geophysik, 125–239, B.G. Teubner, Leipzig, Erster Teil, 1906
95. P. Pizzetti, Principii della teoria mecannica della figura dei pianeti (E. Spoerri, Pisa, 1913)
96. P. Stäckel, Über die integration der Hamilton-Jacobischen Differentialgleichung mittelst Sep-
aration der Variablen (Habilitationsschrift, Halle, 1891)
97. P. Stäckel, Über die Bewegung eines Punktes in einer n-fachen Mannigfaltigkeit. Math. Ann.
42, 537–563 (1893)
98. P. Vanicek, To the problem of holonomity of height systems in: Letter to the Editor, vol 36.
The Canadian Surveos (1982), pp. 122–123
99. S. Roberts, On the parallel surfaces of conicoids and conics. Proc. Lond. Math. Soc. 1(4),
57–91 (1872)
100. S. Roberts, On parallel surfaces. Proc. Lond. Math. Soc. 1(4), 218–235 (1973)
101. V. Schwarze, Satellitengeodätische Positionierung in der Relativistischen Raum-Zeit. Ph.D.
Thesis, Deutsche Geodätische Kommission, Bayerische Akademie der Wissenschaften,
München (1995)
102. W. Neutsch, Koordinaten-Theorie und Anwendungen (Spektrum Akademischer Varlag, Hei-
delberg, 1995)
103. Y. Georgiadou, E. Livieratos, Anholonomity of the reciprocal natural frame in the normal
gravity field of the Earth. Geophys. J. R. Astron. Soc 67, 177–179 (1981)
104. Z. Nadenik, Les forms differentielles exterieures dans la Geodesie i: coubure de Gauss. Studia
Geoph. et Geod. 15, 1–6 (1971)
Epistemic Relativity: An Experimental
Approach to Physics
Bartolomé Coll
Abstract The recent concept of relativistic positioning system (RPS) has opened
the possibility of making Relativity the general standard frame in which to state any
physical problem, theoretical or experimental. Because the velocity of propagation
of the information is finite, epistemic relativity proposes to integrate the physicist
as a truly component of every physical problem, taking into account explicitly what
information, when and where, the physicist is able to know. This leads naturally to the
concept of relativistic stereometric system (RSS), allowing to measure the intrinsic
properties of physical systems. Together, RPSs and RSSs complete the notion of
laboratory in general relativity, allowing to perform experiments in finite regions
of any space-time. Epistemic relativity incites the development of relativity in new
open directions: advanced studies in RPSs and RSSs, intrinsic characterization of
gravitational fields, composition laws for them, construction of a finite-differential
geometry adapted to RPSs and RSSs, covariant approximation methods, etc. Some
of these directions are sketched here, and some open problems are posed.
1 Introduction
As everyone knows, General Relativity is a theory of the gravitational field but, also
and primarily, a theory of the space-time. During the last century, it has been largely
proved experimentally that the gravitational field, as well as the space-time, is better
described by it than by Newtonian theory.
Nevertheless, by natural reasons, the environments in which most of the experi-
ments have taken place have been either static (e.g. Earth surface), or stationary (e.g.
I want the organizers of this seminar, Dirk Puetzfeld and Claus Lämmerzahl, to know how much I
appreciated their inviting me to talk about my ideas on this subject. It is also a pleasure to thanks
the Wilhelm und Else Heraeus-Stiftung for their kind hospitality.
B. Coll (B)
Relativistic Positioning Systems, Department of Astronomy & Astrophysics,
University of Valencia, c/ Dr. Moliner 50, 46100 Burjassot, Valencia, Spain
e-mail: bartolome.coll@uv.es

292 B. Coll
aircrafts or satellites in circular orbits), or periodic (e.g. satellites in elliptic orbits), or

quasi-periodic with almost-periodic or constant non-periodic parts (e.g. Solar system
ephemeris or radiation by black hole accretion disks). These environments allow to
trivialize some important specificities of the relativistic space-time, especially some
related to the finite velocity of propagation of the information.
Thus, for example, in these particular environments, the deterministic charac-
ter of Einstein equations seems to be predictive, like the (Laplace) determinism of
Newtonian theory. But out of these environments, because Einstein equations can be
associated to a hyperbolic system1 with influence domain2 attached to the velocity of
light, the initial data on an initial instant cannot be physically known but at or after the
cusp of the influence domain, so that Einstein determinism is generically retrodic-
tive,3 not predictive. Nevertheless, most of the problems related to initial conditions
in relativity, even for generic environments, have been thought as predictive ones, as
if someone at the initial instant were able to know the initial data. This manner of
thinking, strongly attached to the evolutive Newtonian point of view4 but physically
inadmissible in relativistic generic environments, masks the concepts that we should
develop in order to be able to describe physically generic environments in relativity.
The aim of epistemic relativity is to give us such ability.5 Section 2 is devoted to
specify its basic ingredients. The first of these ingredients is constituted by relativistic
positioning systems,6 of which auto-locating systems and autonomous positioning
systems are the more interesting ones. They are the object of Sect. 3. The second
basic ingredient is constituted by relativistic stereometric systems,7 which allow
to detect intrinsic properties of physical systems. Their notion and first physical
results are presented in Sect. 4. By the way, epistemic relativity plays a heuristic
role giving new meanings to already known subjects or opening new ones. Section 5
shows how the intrinsic characterization of metrics may help epistemic relativity to
1 The differential operator associated to the n(n + 1)/2 Einstein equations is not hyperbolic, but
degenerate. Nevertheless, because n of these equations (constraint equations) are involutive with
respect to the other n(n − 1)/2 ones (evolution equations), one can supplement this last system with
n suitable additional equations (coordinate conditions). It is the differential operator of this new
system of n(n + 1)/2 equations for the metric coefficients which may be made hyperbolic with a
suitable choice of coordinate conditions (for example, with harmonic ones).
2 The influence domain for an initial instant (local spatial hypersurface) is the domain where the
solution exists and is unique for every initial (Cauchy) data on this initial instant.
3 Retrodictive, as opposed to predictive, means here that one cannot but verify afterwards that the
physical quantities measured in the influence domain agree with the initial data received at or after
the cusp of this domain.
4 In Newtonian physics, the contents of the three-dimensional space at every instant is supposed
known or knowable at that instant, so that the evolutive equations describe the dependence in time
of this three-dimensional contents.
5 Epistemic relativity was first presented at the GraviMAS FEST workshop, in honor of Lluís Mas,
Mallorca, Spain, 2008, http://www.uib.es/depart/dfs/GRG/GraviMAS_FEST/. See also [1].

6 For the genesis of the concept of relativistic positioning systems, see for example [1].
7 They were relativistic stereometric systems which, joined with relativistic positioning systems,
suggested the idea of epistemic relativity. This is why relativistic stereometry and epistemic relativity
were conceived conjointly.
Epistemic Relativity: An Experimental Approach to Physics 293
identify gravitational fields and, finally, Sect. 6 develops a finite-differential geometry

in which distance function and metric ought to play symmetric roles.
I believe that in epistemic theory, as an incipient theory, what is interesting are
the basic concepts. This is why I have restricted the bibliography to them.8
2 Epistemic Relativity
Being the best theory of the space-time in which all the physical phenomena take
place, relativity ought to be able to describe any physical experiment in terms of its
proper, relativistic, concepts, regardless of the quantitative evaluation of the experi-
ment for which, in many cases, Newtonian calculations could suffice.
Also, being the best theory of the gravitational field, relativity ought to propose
experiences and methods of measurement of general gravitational fields.
But today, such descriptions or proposals are conspicuous by their absence. In
fact, we do not know how to do them. Thus it seems evident that relativity needs to
develop a proper experimental approach to the physical world.
2.1 The Notion of Epistemic Relativity
Fortunately, we already know the conceptual basic ingredients for such a develop-
ment. Before to describe them, the idea of a ‘relativistic experimental approach’
needs to be more explicit.
In relativity, a large number of scientific works analyze physical and geometrical
properties of the space-time, but
• do not integrate the physicist as a part of them, and
• forget implicitly that
– information is energy,
– neither the density of energy, nor its velocity of propagation can be infinite in
relativity.
Many of these properties of the space-time may be analyzed by a geometer on his

desk, but to be known by an experimental physicist,9 they would require qualities
8 Some subjects related to epistemic relativity have been the object of interesting developments (e.g.
positioning systems, relativistic geometric optics, intrinsic characterization of non-vacuum metrics,

Regge calculus) not directly related to the basis of this theory. For this reason the corresponding
bibliography is absent here.
9 In all this text, the worlds ‘physicist’, ‘observer’ or ‘user’ denote any person or device able to receive
the pertinent information, to record and to analyze it and to perform the actions and computations
needed for the problem in question. For short, we shall refer to any of them as ‘it’.
294 B. Coll
of an omniscient10 god ! For these reasons, we shall say that such works belong to
ontic relativity.11
Ontic relativity is necessary to develop a relativistic experimental approach (and
even sufficient in static, stationary, periodic or almost periodic gravitational fields!),
but it is also generically insufficient.
Let us remember that the statement of any problem in general relativity implies:
• to describe it directly in the space-time, by means of intrinsic relativistic objects, i.e.
by means of worldlines and worldtubes (the analogs to the histories or evolutions
of the instantaneous objects of Newtonian theory),
• to banish generically the use of any physical extended present (due to high precision
clocks or regions of non-constant inertial or gravitational fields).
The works in relativity that, in addition:
• integrate the physicist as an element of the problem considered,
• concern physical properties that the physicist can measure, and
• take into account explicitly what information, when and where, the physicist is
able to know,
will be considered as characterizing epistemic relativity.12
Paradoxically, until now a very few number of relativistic problems have been
solved explicitly under these assumptions.
2.2 The Ingredients of Epistemic Relativity
The main objective of epistemic relativity is to provide the physicist with the knowl-
edge and protocols necessary to make relativistic gravimetry (chronometry in arbi-
trary directions) in any unknown space-time environment (Fig. 1).
This is the first and unavoidable step to develop experimental relativity as the
natural scientific approach to the physical world.
The basis of relativistic gravimetry are Einstein equations, and these equations
need a precise mathematical model of the space-time. The adequacy of this mathe-
matical model and the physical space-time that it describes implies a one-to-one cor-
respondence between the points of the model and the physical events they describe.
And because in the mathematical model points are precisely identified (by their
coordinates), we need to know how to locate the events in the physical space-time.
What else do we need in epistemic relativity? In fact, what we need is to transform
the space-time region of physical interest in a (finite) laboratory.
But, what is a laboratory in relativity? In fact, a reflection on this matter shows
that any laboratory, regardless of the specificity of its measurement devices, has to
provide us with:
10 Remember that relativity is retrodictive.

11 From the Greek ‘ontos’, being, with the meaning of ‘what is’ as opposed to ‘how one knows it’.
12 From the Greek ‘episteme’, knowledge, with the meaning of ‘how we obtain it’.
Fig. 1 Epistemic relativity

wants to provide the
physicist with the baggage
and protocols necessary to
make gravimetry in any
unknown region of arbitrary
space-times
• a precise physical location of the significant parts of the system, and

• a precise physical description of its intrinsic properties.
The devices able to carry out these two tasks are called relativistic positioning
systems and relativistic stereometric systems respectively. Thus:
A finite laboratory in relativity is a space-time region endowed with:
• a relativistic positioning system and
• a relativistic stereometric system.
The first task of epistemic relativity is to transform the finite regions of interest
of the space-time in laboratories or, equivalently, to construct these two relativistic
systems.
These two relativistic systems are sufficient to describe any experiment in their
region of space-time. Remember that, as already mentioned, we refer to physical
systems as relativistic objects, i.e. as worldlines or worldtubes so that they naturally
contain the changes of state that the Newtonian physical systems at every instant
may suffer during an experiment (they already are the histories or evolutions of
Newtonian systems). In the same point of view, the physical properties of relativistic
objects cannot but be the history of the physical properties of the Newtonian objects
at every instant of its constituent events. It is thus clear that the two above functions
of locating and detecting are effectively able to describe any physical experiment in
the space-time domain in question.
3 Relativistic Positioning Systems
3.1 What is to Locate
In general, to locate an object is to identify the place it takes up. To locate with
accuracy objects of any size (avoiding vagueness such as near of, inside of, etc.),
we need to assign a proper name to every event of the space in question. A locating
system is a system of assignation of proper names to the events of a space.
In particular, because the space-time is a continuum, the assignation of a proper
name to every one of its events has to be done with numbers. And because it is a
296 B. Coll
four-dimensional continuum, every proper name has to be done with four numbers.
So that, specifically, in (a region of) the space-time, a locating system is a system of
assignation of four numbers to every one of its events.
In physics, the devices for assigning four numbers to every event may be con-
structed with different materials and be based in different physical properties, so
that their specific construction and protocols of use give raise to different locating
systems.
The set of tetrads of numbers generated in a region by a locating system consti-
tutes (a physical realization of) a coordinate system. On the other hand, in topological
manifolds, a coordinate system in a region R may be extracted from a local chart (R,
ϕ) by means of the coordinate functions associated to ϕ, where ϕ is a homeomor-
phism from R to the associated linear space. We see that, mathematically, locating
systems are represented by homeomorphisms ϕ. Thus, a locating system in a finite
region R of the space-time, may be seen as the physical construction of a local chart
or, conversely, a local chart may be seen as a mathematical idealization of a locating
system.13
3.2 Properties of Locating Systems
According to the characteristics of the assignation system of coordinates, a locating

system may be active if its assignation system only operates for events that emit
signals of presence, or passive if it operates for all events of the region irrespective
of their emission state; and it may be immediate if the values of the coordinates of
every event are obtained without delay, or retarded otherwise. Also, according to its
use, a locating system is said real if its assignation system starts one for all in all
the region, and it is said virtual if it is used case by case to obtain only the specific
coordinates of particular events.
Another classification takes into account the function allocated to the locating
system. Two locating systems are of particular importance for us: reference sys-
tems, which allow one observer to know the coordinates of the events of the region
and positioning systems, which allow every event of the region to know its proper
coordinates (Fig. 2).
It is then evident that real locating systems may be immediate and passive, mean-
while virtual locating systems are necessarily active and retarded. And, taking into
account that in relativity information propagates at finite velocity, it result that all
reference systems are retarded.
13 It is to be noted, because frequently forgotten, that the local charts on a differentiable manifold may
be structural, i.e. belonging to the atlas defining its differentiable structure, or not. Only in the first
case, the region R has to be an open set. Nonstructural local charts may be of lesser differentiable
class than structural ones. In particular, they may be simply continuous, although in this case, the
natural frame being absent, one may be lead to complete the local chart with an independent field
of vector tetrads in order to construct a basis for the tensor algebra.
Fig. 2 A reference system allows one observer to know the coordinates of the events of the region.
A positioning system allows every event of the region to know its proper coordinates
We can conclude that, when it is possible to construct them, real positioning

systems are the locating systems that offer the best physical performances.
3.3 Relativistic Positioning Systems
We have just seen that, among all locating systems, the real positioning systems
offer the performances of being passive and immediate. These performances are
still insufficient for us, because the main objective of epistemic relativity is to make
relativistic gravimetry in an unknown space-time environment.
Because the space-time environment is unknown, the locating system has also
to be generic, that is to say, able to be constructed in regions of any space-time.14
And because the main objective is to make relativistic gravimetry, it has also to be
(gravitation-) free, i.e. able to be constructed without the previous knowledge of the
space-time metric.
Relativistic positioning systems are the real, generic and free positioning systems
of the space-time.
The construction of relativistic positioning systems is unexpectedly simple. A
clock may be seen as a continuous generator of numbers, namely the time that it
displays at every instant. Broadcasting this time by means of an electromagnetic
signal, every event receiving the signal will know this number at the instant of recep-
tion. And because the proper name of every event of the space-time consists of four
numbers (its coordinates), four clocks arbitrarily distributed will provide every event
reached by the four signals with such numbers. It is thus easy to see that: Four clocks
broadcasting their times constitute a relativistic positioning system15 (see [3]). The
14 It is understood that such regions have the appropriate accessibility characteristics to make
gravimetry.
15 Generically. Some distributions of clocks may associate same coordinates to different events,
whatever the region considered (see [2]).

298 B. Coll
coordinates generated by relativistic positioning systems are called emission coor-

dinates [4].
Note: I have chosen to introduce the well-known notion of coordinate system in
this unusual form because the traditional presentations induce to give too importance
to ingredients that are in fact dispensable or of secondary interest, but wrongly make
many people to feel them necessary to understand physically a coordinate system.
Thus, ingredients like origin, coordinate lines or synchronization of a coordinate
system are respectively irrelevant, of secondary interest or inexistent16 for relativistic
positioning systems.
It is worthwhile to remark that relativistic positioning systems are a particular
class of positioning systems but that, although similar in some aspects to positioning
systems like GNSSs (global navigation satellite systems), differ clearly from them,
whatever be the relativistic formalism with which these last ones are analyzed. These
GNSS are technical objects, wonder technical objects, with a lot of technical, social
and even scientific applications. But they are not scientific objects. The reason is that
they have been thought, calculated and constructed using consciously a defective
theory, the Newtonian one, to construct a Newtonian space in the region between the
Earth surface and the constellation of satellites, in which the International Atomic
Time on the Earth surface is imposed on all the region as a Newtonian absolute
time. The “relativistic effects”17 are not used to improve the approximate Newtonian
calculations and to obtain more precise values of the physical times involved, but to
deform the values of these physical times so as to mimic an unphysical Newtonian
time everywhere (the calculated “relativistic effects” are subtracted to the physical
values, not added to the Newtonian calculations of them!).
The above comment is not at all a criticism of GNSSs: as technical objects, they
have to be useful to us and give us what we ask them to give. For relational (technical,
social) convenience, it is the case, in particular, of a universal time for all of us, be we
on the Earth surface or at 15.000 km height, at rest or at whatever velocity. But, I think
that it would be better to have, first of all, a scientific object, a RGNSS (Relativistic
GNSS) able to give us our proper physical time, value of the gravitational field or
local distance, whatever our position and velocity with respect to the Earth surface.
The reason is that, starting from such a scientific object, a simple software would
be able also to mimic GNSSs with their absolute time everywhere. The (theoretical)
question is not so much to improve GNSSs, but instead to construct cutting-edge
scientific RGNSSs.
There are no specific coordinates associated to GNSSs.18 Nevertheless, emission
coordinates are an important tool in the study of relativistic positioning systems and
in particular in RGNSS. Among others, they suggest new questions, provide new
16 Or fourfold. Generically there does not exist a privileged synchronization of the form time =
constant in relativistic positioning systems, but emission coordinates being constituted by four times
τ A , one could say that there exist four synchronizations τ A = constant.
17 Euphemism for “Newtonian defaults”.
18 The pseudo-ranges are only considered as parameters able to calculate conventional coordinates
on the Earth surface (WGS84, ITRS or others). The appellation ‘GPS coordinates’ refers to these
conventional coordinates provided by the GPS.
Fig. 3 An auto-locating
system allows any user to
know its proper trajectory in
emission coordinates, but
also the trajectories of the
clocks of the system (here a
two-dimensional scheme)
results and allow the constellation of satellites of a RGNSS to generate autonomous

positioning systems for the Earth.
3.4 Auto-Locating Systems
As an intermediate step, auto-locating systems are those relativistic positioning sys-

tems whose clocks are endowed with a transponder broadcasting the times they
receive from the other clocks of the system (see [3, 5] and references therein).
So, every event P of the region receives the sixteen times τ A and τ AB A, B =
1, . . . , 4, A = B. The four (τ A ) are the emission coordinates of P, meanwhile every
one of the four sets A of four times (τ A , τ AB ) are the emission coordinates of the
clocks A of the system (see Fig. 3). Thus, any user that records these data is able to
know its trajectory as well as the trajectories of the clocks in the grid of emission
coordinates.19
An echo-interval for the clock A due to the clock B is the segment of world line of
A from the instant of emission of a broadcast signal by A to the instant that it receives
that signal echoed by B. In two dimensions, it is already proved a very interesting
theorem for auto-locating systems (see [5]):
Theorem 3.4.1 In a flat two-dimensional space-time, consider a user of an auto-

locating system that knows its trajectory and those of the two clocks in the grid. If
it also knows the acceleration of one of the clocks during one echo-interval, then it
knows its acceleration and those of the two clocks along all their known trajectories
(Fig. 4).
19 Thegrid G of emission coordinates is the Cartesian product of the segments [τ A ] of the times
broadcast by the relativistic positioning system: G ≡ [τ 1 ] × [τ 2 ] × [τ 3 ] × [τ 4 ].
300 B. Coll
Fig. 4 In absence of a
gravitational field, a user of
an auto-locating system is
able to know its trajectory
and that of the clocks in the
grid of emission coordinates.
If it also knows the
acceleration of one of the
clocks during one
echo-interval AB, then it
knows its acceleration and
those of the two clocks along
all their trajectories
This theorem is very interesting because of the following fact: the algorithm that
proves it in absence of a gravitational field may be as well applied in presence of it.
Of course, in this case it will provide a wrong acceleration along all the three tra-
jectories excepting on the starting echo-interval of the clocks. But the importance of
this fact is that the difference between these wrong accelerations (calculated as if the
gravitational field were absent) and the true accelerations (measured by accelerom-
eters along the world-lines of clocks and user in the gravitational field), is all the
dynamical effect that the gravitational field produces on the auto-locating system and
the user. Thus, it is a complete, although relative (to the choice of echo-interval),
gravimetric measure.
Unfortunately, this fundamental Theorem 3.4.1 is only know in two-dimensional
space-times. It is of crucial importance for the theory of relativistic positioning
systems and for all its near-future applications (Earth surface, RNGSS, Deep Space
Navigation) to have its four-dimensional generalization. But this problem remains
open. As in the two-dimensional case, the starting point is
• the explicit knowledge of the coordinate transformation between the emission
coordinates of four arbitrarily accelerated clocks and inertial coordinates, and
• the explicit knowledge of the metric in emission coordinates.
Fortunately, the solution to these problems is already known. If the world-lines
γ A (τ A ), A = 1, . . . , 4, of four emitters with respect to inertial coordinates are known,
we can evaluate the quantities
1
ea ≡ γa − γ4 , a ≡ (ea )2 , χ ≡ ∗(e1 ∧ e2 ∧ e3 )
2
1
H ≡ ∗(1 e2 ∧ e3 + 2 e3 ∧ e1 + 3 e1 ∧ e2 ) , y∗ ≡ i(ξ)H
ξ.χ
where ξ is an arbitrary transversal vector, ξ.χ = 0, and a = 1, 2, 3. Then, the answer

to the first item is (see [2] and [6]):
Theorem 3.4.2 Suppose known the world-lines x = γ A (τ A ) of the four emitters of

a positioning system with respect to an inertial coordinate system {x}, and let {τ A }
be their emission coordinates. In term of the quantities χ and y∗ evaluated from the
γ A (τ A )’s, the coordinate transformation x = κ(τ A ) between emission and inertial
coordinates is given by:
y2 χ
x ≡ κ(τ R ) = γ4 + y∗ + ∗
(y∗ . χ) + ε̂ (y∗ . χ)2 − y∗2 χ2
where ε̂ is the orientation of the position system with respect to the user.
And the answer to the second item is:
Theorem 3.4.3 In terms of the above transformation x = κ(τ R ) and the world-lines
x = γ A (τ A ), the contravariant components of the metric, g AB (τ R ), in the emission
coordinates {τ R } are given by:
AB
g AB (τ R ) = ,
μ A μB
where
1
AB ≡ (γ A − γ B )2 , μ A ≡ (κ − γ A ) . γ̇ A .
2
3.5 Autonomous Positioning Systems
Auto-locating systems contain all what a coordinate system needs in order to be

drawn in a space in which a field of light-cones is given.20 But if the light-cones
are specifically related to a geometry (here the gravitational field represented by
a metric), and we want to identify individually the coordinate system making in it
gravimetry (measure of the metric), we need additional information. This information
has to be also broadcast by the clocks and consists of:
20 The field of light-cones defines all the possible coordinate hypersurfaces of the coordinate systems
associated to all the possible relativistic positioning systems. Accordingly, the particular trajectories
of the clocks select the specific ones for the auto-locating system and, consequently, its region of
validity or domain of definition in the grid.
302 B. Coll
Fig. 5 An observer,
determined by its unit
velocity, projects its past
light-cone on its celestial
sphere
• dynamical data of the satellites (acceleration, gradiometry),

• observational data from them (e.g. particular masses, directions of reference
quasars or pulsars, etc.), and
• gravitational knowledge of the coordinate region (theoretical, experimental or
mixed).
This information constitutes the autonomous data. Autonomous positioning systems
are auto-locating systems that broadcast autonomous data.
Autonomous positioning systems are, generically, the best locating systems we are
able to construct. Their notion was already proposed in the first paper on positioning
systems (see [3]).
4 Relativistic Stereometric Systems
4.1 The Notion
We already know that, to transform a finite region of the space-time in a laboratory,

we need to endow it with, on one hand, a relativistic positioning system, to locate in
the best form the significant parts of the physical system in question, and on the other
hand, a relativistic stereometric system, to obtain their intrinsic physical properties.
Similarly to a relativistic positioning system, that consists essentially of four
clocks broadcasting their times,21 a relativistic stereometric system consists of a
system of four observers receiving the signals broadcast from a physical system.
In relativity, the word ‘observer’ is used in different contexts with different mean-
ings. Here an observer means an eye able to record and to analyze the input, and
an eye22 is a small (physically local, mathematically differential) object, determined
by its space-time worldline, able to project its past light-cone on its celestial sphere
(Fig. 5).
21 The time broadcasted by every clock may be any convenient time, not necessarily its proper time,
although proper time simplifies in general the theoretical analysis.

22 More precisely, a 4π-wide hypergon eye or 4π steradian eye.
Fig. 6 Four events that are

seen in a circle on the
celestial sphere of an
observer are also seen in a
circle on the celestial sphere
of any other observer at the
same event, whatever its
relative velocity
Fig. 7 Relativistic
stereometric systems are the
causal duals of relativistic
positioning systems
There exists interesting papers on relativistic vision but they are very different in
strategy, starting hypothesis and definition of ‘eye’, and their number is very small.
It would be worthwhile to analyze and classify them, and specially to select those
related to hypergon eyes.
But almost none of them consider the invariants of the observed figures. The
study of these invariants is important for positioning as well as for stereometry. For
example, any set of four events in the space-time (four stars in the sky or four material
points of a system), that are seen on a circle of the celestial sphere of an observer,
will be seen also on a circle, by any other observer at the same event, whatever be
their relative velocity at the event (Fig. 6). In addition, the relative velocity of such
two observers may be obtained by comparison of the relative positions on the circle
of the four points as seen by them.23
From the point of view of the space-time geometry, relativistic stereometric sys-
tems are the causal duals of relativistic positioning systems: the four observers (resp.
four clocks) of the relativistic stereometric system (resp. relativistic positioning sys-
tem) receive (resp. broadcast) the signals emitted by the event to be analyzed (resp.
the signals received by the event to be located). As a consequence, every geomet-
ric result on one of these systems generates another geometric result for the other
system, by a simple change of time orientation (Fig. 7).
Relativistic stereometric systems are conceived to obtain the intrinsic properties
of physical systems but, in their basic structure, they are also locating systems. As
locating systems, the coordinates of every event are the (four) times of reception by the
(four) observers of the signal emitted by the physical system at this event. According
to the properties considered in Sect. 3.2, relativistic stereometric systems are active,
23 In principle, a suitable choice of stars, in e.g. Hipparcos star catalogue, could help a spacecraft in
the Solar system to estimate its velocity with respect to the Barycentric Celestial Reference System.
304 B. Coll
Fig. 8 A two-dimensional
relativistic stereometric
system measuring the proper
frequency f of a material
point
retarded and virtual locating systems, in contrast with the passive, immediate and
real properties of relativistic positioning systems. As we have seen in Sect. 3.2, the
active, retarded and virtual properties of relativistic stereometric systems are not
the best as locating systems but, nevertheless, they are well adapted to the principal
function of relativistic stereometric systems, because the intrinsic properties of a
physical system cannot be obtained but by active, retarded and virtual ways. Thus,
relativistic stereometric systems not only describe the intrinsic properties of events
but also locate them.24
4.2 First Results in Relativistic Stereometry
Every observer of a relativistic stereometric system receives a (relativistic) perspec-

tive of the properties broadcast by the events. These four observer-dependent per-
spectives have to be combined in order to obtain the observer-invariant (proper or
intrinsic) properties that generate them.
To show how to proceed, we shall present here the simplest case of relativistic
stereometry, that of a material point broadcasting a signal in a flat two-dimensional
space-time. In spite of its simplicity: (i) it leads to a simple but interesting extension
of the unidirectional Doppler formula and (ii) it shows the difference between an
ontic law and an epistemic one (Fig. 8).
24 Of course, by broadcasting their proper times, a relativistic stereometric system may also act as
a relativistic positioning system. But because of their dual ways of working, I believe it is clearer,
for the moment, to study them separately.
Thus, let C, C1 and C2 be respectively the worldlines of a material point and of

the two observers of the relativistic stereometric system in Minkowski space M2 .
Suppose that at a particular instant25 C broadcasts a signal of proper frequency f
that is received by C1 and C2 at the Doppler frequencies f 1 and f 2 respectively. Let
v1 and v2 be the scalar relative velocities of C with respect to C1 and C2 evaluated
by them at the instants of reception of the frequencies f 1 and f 2 respectively. As it
is well known, f 1 and f 2 are related to f by the Doppler expressions:

1 − v1 1 − v2
f1 = f , f2 = f . (1)
1 + v1 1 + v2
Denoting by v12 the relative scalar velocity of C1 and C2 at the instants of reception
of the Doppler frequencies, one can prove:
Theorem 4.2.1 In terms of the relative scalar velocity v12 of the observers C1 and
C2 of a relativistic stereometric system at the instant of reception of the Doppler
frequencies f 1 and f 2 from a material point C, the proper frequency f of the material
point at the instant of emission is given by:

1 + v12
f 2 = f1 f2 . (2)
1 − v12
Of course, if the material point is dragged by one of the observers of the system,
say f = f 2 , expression (2) coincides with the first of (1).
The Doppler frequencies also allow to know the relative velocities of the material
point C at the instant of emission of the signal with respect to the observers C1 and C2
at the instants of its reception. By elimination of f from every one of the expressions
(1) and (2), one has:
Theorem 4.2.2 The relative velocities v1 and v2 of the material point C at the instant
of emission of the signal with respect to the observers C1 and C2 at the instants of
its reception are given by:
√ √
f 2 1 + v12 − f 1 1 − v12
v1 = √ √ , (3)
f 2 1 + v12 + f 1 1 − v12
√ √
f 1 1 + v12 − f 2 1 − v12
v2 = √ √ . (4)
f 1 1 + v12 + f 2 1 − v12
The above results depend on the Doppler frequencies f 1 and f 2 but also on the
quantity v12 , the scalar velocity of one of the observers of the relativistic stereometric
system with respect to the other at the instants at which they receive the Doppler
25 Thereis no matter here what instant-identifier is used: a clock associate to the point, measuring
any time, non-necessarily proper, a flash, or any other pertinent one.
306 B. Coll
Fig. 9 The observer C2 has

been chosen as the
‘physicist’. From the instant
τ12 on, it may know all the
quantities of the problem
frequencies. This quantity cannot be measured directly, but has to be calculated in

terms of other physical quantities. This reason is sufficient to show that the above
two theorems, in their present form, do not constitute results of epistemic relativity.
In fact, these two theorems do not fulfill any of the three conditions of Sect. 2.1
characterizing epistemic relativity. For that, we must specify (i) what physicist we
have chosen to make the experiment, (ii) when and where it will know the quantities
needed to solve the problem and (iii) how can it know or measure these quantities.
In the present case, we make the following choices:
• the physicist is one of the observers of the relativistic stereometric system, say C2 ,
as shown in Fig. 9,
• it will be informed of the pertinent quantities at the instant τ12 of reception of the
information coming from the observer C1 ,
• at that instant τ12 , it already knows the quantity f 2 measured and recorded by
itself, it is being informed of the quantity f 1 , and then may know the quantity v12
by computation.
At the instant of reception of the signal f 1 , the observer C1 , in addition to send
its measure, must send also its reflection or its proper frequency so as to allow C2
to know its relative velocity at its instant τ12 . On the other hand, C2 is supposed to
know its proper acceleration, so that it is able to know its relative velocity between
its instant of reception of the signal f 2 and the instant τ12 . The knowledge of these
two velocities, allow it to calculate the velocity v12 of the theorems. For simplicity,
suppose here that C2 is geodesic. Then, the relative frequency between the proper
time of C1 at the instant of reception of the frequency f 1 and the proper time of C2
at any instant is constant. Denoting it by ν12 and inverting the Doppler formula we
have:
1 − ν12
2
v12 = . (5)
1 + ν12
2
Now if, for short, we call an epistemic theorem an ‘epistem’, the above two
theorems in this geodesic case become respectively:
Epistem 1 In terms of the frequencies f 1 and f 2 received by a relativistic stereo-
metric system {C1 , C2 } from a material point C and of the relative frequency ν12
of the proper time of the observer C1 with respect to the geodesic observer C2 , the
proper frequency f of the material point C is given by:
f1 f2
f2 = . (6)
ν12
Epistem 2 In terms of the frequencies f 1 and f 2 received by a relativistic stereo-

metric system {C1 , C2 } from a material point emitter C and of the relative frequency
ν12 of the proper time of the observer C1 with respect to the geodesic observer C2 ,
the relative velocities v1 and v2 of the material point C with respect to the observers
C1 and C2 at the instants of reception of the signals f 1 and f 2 are given by:
f 2 − f 1 ν12 f 1 − f 2 ν12
v1 = , v2 = . (7)
f 2 + f 1 ν12 f 1 + f 2 ν12
These results are very simple, but show roughly the way of working with stere-
ometric systems in epistemic relativity. Of course, the problem of determining the
proper distances to its neighboring elements of a material point cannot be considered
in a two-dimensional space-time, because there the celestial sphere of an observer
reduces to two opposite points. It is evident that this work has to be extended and
generalized in three and four dimensions. It remains an open problem.
5 Intrinsic Characterization of Gravitational Fields
In epistemic relativity, events are located in emission coordinates generated by rela-

tivistic positioning systems.
A complete set of gravimetric measurements (whatever be the methods) will lead
to the experimental values of the components of the metric in emission coordinates.
It rest to identify this metric or, equivalently, its sources. For example, suppose
we suspect that it corresponds either to a Kerr or to a Schwarzschild gravitational
field, but that it can be neither one nor the other. How can we discern between these
three possibilities?
The most part of the known gravitational fields (exact or approximate solutions
to Einstein equations), and in particular Kerr and Schwarzschild, are known in very
particular coordinate systems which have no simple relation with emission coordi-
nates.
308 B. Coll
For example, for the Schwarzschild case, the direct procedure would be
• to calculate in Schwarzschild coordinates the equations of the field of light-cones,
• to model in these coordinates the world-lines of the clocks of the positioning
system, parameterized with their proper time,
• to select with both results the four families of light-cones emitted by the clocks
or, in other words, to obtain the transformation from emission coordinates to the
Schwarzschild ones,
• to solve for the inverse transformation,
• to transform the Schwarzschild metric components in emission metric components,
• to compare them with the experimental values obtained in these emission coordi-
nates.
Every one of these items presents non-negligible difficulties and can introduce
specific uncertainties. It results in a hard and long procedure.
For this reason, to compare the experimental value of the gravitational field
obtained in emission coordinates with a given solution of Einstein equations, it may
be better to characterize intrinsically the solution and to check if the experimental
data verify this characterization.
We understand here by intrinsic characterization of a metric, a set of necessary and
sufficient local conditions on some of the concomitants of the metric that characterize
it, its sources an their position, regardless of the coordinate system.
The idea is that an observer, from the sole measure of the gravitational field and
its variations in a local region, is able to know the masses that produce it as well as
its positions with respect to it.
The intrinsic characterization of individual metrics is, apart from some trivial
cases, a relatively recent problem. I would like to draw attention on the pertinence
not only of its development for all exact solutions of Einstein equations of physical
interest, but also of its extension to the usual approximate solutions in experimental
applications. By the reason above indicated, I believe this characterization to be the
simplest and shortest way to identify gravitational fields in epistemic relativity.
The idea to characterize individual metrics intrinsically originates some years ago
as part of a general project of IDEAL solution of problems, somehow similar in spirit
to that of classic Greek mathematicians but of vivid actuality for the problems we
were concerned, where the acronym stands for
• Intrinsic (depending only of the concepts mentioned in the statement of the prob-
lem),
• Deductive (not involving inductive or inferential methods or arguments),
• Explicit (expressing the elements of the solution non implicitly) and
• Algorithmic (giving the solution as a flow chart with a finite number of steps).
The first solution intentionally obtained under the IDEAL spirit was that of the
point particle in Newtonian gravity ([7]). I cite it here because the solution to its
relativistic analog, the Schwarzschild gravitational field, in contrast with an extended
opinion but like many other generic problems,26 admits simpler expressions than that
of Newtonian physics.
The intrinsic characterization of the Schwarzschild gravitational field is due to J.
Ferrando and J.A. Sáez ([8]). Denoting by Riem the Riemann tensor, by tr and i(.)
the trace operator and the interior product respectively, and by ∧ the exterior product
(considering the metric g as a double 1-form), they introduce the two scalars:
13
1 2 1
σ≡ tr Riem 3 , α≡ g(dσ, dσ) + 2σ , (8)
12 9 σ2
allowing to construct the two tensors:

1 1
S≡− Riem + σg ∧ g , Q ≡ i 2 (dσ)S . (9)
3σ 2
and, with the help of an arbitrary direction x, generate a third, direction-dependent

scalar θ:
θ(x) ≡ 2Q(x, x) + tr Q . (10)
These quantities may be evaluated for any gravitational field, and involve the gradient
of a function of Riem, a third variation of the metric.
Their first result is:
Theorem 5.1 A vacuum metric g, Ric(g) = 0, is the Schwarzschild metric if, and
only if, the three scalars σ, α and θ are strictly positive,
σ > 0 , α > 0 , θ(x) > 0 , (11)
and the two following tensors vanish:
i 2 (dσ) ∗Riem = 0 , S 2 + S = 0 , (12)
where ∗ is the Hodge duality operator.

In the circumstances suggested by Fig. 10 this result assures the experimental
physicist, whatever the coordinates used, that it is immersed in the field of a spheri-
cally symmetric mass. But what mass and where? The answer is also given in [8]:
Theorem 5.2 The mass m and the radial coordinate r of a Schwarz-schild metric
are given by:
m = σα− 2 , r = α− 2 .
3 1
(13)
In fact, any other intrinsic characteristic of Schwarzschild metric may be obtained

in an IDEAL way. For example, differentiating the second of relations (13) one
26 Formulation of Maxwell equations, Cauchy problem for the permanence of electromagnetic
waves, shock, detonation and deflagration waves in hydrodynamics, and some others.
310 B. Coll
Fig. 10 An observer that measures the gravitational field in a local space is able to know the masses
that produce them as well as their positions
obtains the intrinsic expression in terms of the invariant α of the radial codirection
−dr of the mass, allowing the physicist to know its direction, distance and value.
Also, on the event horizon r = 2m one would find α = 2σ and, as (8) shows, also
g(dσ, dσ) would vanish. Furthermore, as proven also in [8], the time-like direction
of static evolution of the gravitational field, i.e. its integrable Killing vector field ξ,
is given by:
Q(x)
ξ = −σ − 3 √
4
. (14)
Q(x, x)
These results already allow us to appreciate the interest of local intrinsic character-
izations of gravitational fields. But they are far from being only the characterization
of the simplest gravitational field. The scalars σ, α and θ, in (8) and (10), and the
tensors S and Q, in (9), may be evaluated in any gravitational field and their com-
parison with relations (11) and (12) will give us an information that we would study
in detail, in particular if the gravitational field in question may be well described as
a Schwarzschild perturbation.
At present, the intrinsic characterization of some other classes of vacuum exact
solutions of Einstein equations are known, in particular Kerr metric (see [9] and
references therein). We need to extend these results and, specially, develop these
techniques for approximate solutions.
The above expressions, simple for a theoretical physicist, may seem complicated
for an experimental one. They need to be broken down in measurable terms so as to
conceive the devices and procedures to measure them. But there is no doubt that the
questions that local intrinsic characterizations of metrics allow to decipher, deserve
a deeper study.
Fig. 11 A constellation C of
satellized clocks around a
mass M, every tetrad of them
working as an auto-locating
system, recording the data set
of all the signals emitted
and received between them
6 Finite-Differential Geometry
As it is well known, the mathematical substratum of general relativity is Lorentzian

differential geometry. It thus follows that differential geometric methods are unavoid-
able in relativity. But there exists many situations in which these differential methods
appear to be manifestly insufficient.
Among these situations there are specific epistemic ones, involving the physical
description in real time and by arbitrary observers of general variable gravitational
fields, but also situations much more near of our usual ones in Earth’s relativity
(Fig. 11).
Thus, in a vacuum space-time, consider a mass M, not necessarily spherical,
and a constellation C of satellized clocks around M and such that every tetrad of
them works as an auto-locating system. Let be the data set of all the signals that
every clock emits and receives both, as an element of all the auto-locating systems
at which it belongs and as a user of the auto-locating systems constituted by all the
other clocks.
Suppose we know the mass M and the world-lines of the constellation C. We can
model the system as follows. Start from the space-time metric around M, integrate
the geodesic equations and specify them for every one of the clocks so as to model
the whole constellation C. Then, integrate the equations of the light-cones, either
starting from the general solution of the null geodesic equations, or integrating the
geodesic distance function and considering its vanishing, so as to model any signal
able to be emitted or received by any of the clocks of the constellation C. With these
two modeled ingredients, we are able to predict any of the data of the set received
or emitted by any of the clocks in terms respectively of the data emitted or received
by all the other clocks.
The physical interest of such a mathematical model is not, however, very intense.
On the contrary, the following scenario could be the prelude for a fine gravimetry
of the Earth ... if we were able to develop the adequate mathematical instrument
(Fig. 12).
312 B. Coll
Fig. 12 When we know the

constellation of satellites C
and the data set , how to
obtain the mass M?
Suppose now that we know the constellation C and the data set , and that we
want to obtain the mass M. At present the only way we know to tackle the problem is
the above model, but for this scenario it is heavy and unadapted to the starting point,
the data set . Indeed, consists of time-like distances (geodesic time intervals
between any two events on the world-line of every clock) and null distances (light
links between every pair of clocks), that is to say, of values of the distance function
between pairs of events causally separated. For this reason, any direct method to
model this scenario would start with the obtainment, from the data set , of the
metric distance function by means of a suitable interpolation method. This is part of
the adequate mathematical instrument that we would be able to develop for obtaining
the mass M. But it is still insufficient.
The objective of finite-differential geometry is to study finite and interchangeable
versions of the ingredients of differential geometry (metric, connection, curvature).
Because for general relativity the basic ingredient is the metric g, and that its finite
version, the distance function, already poses problems, we shall consider it here.27
It is well known that the finite version of the metric g is the distance function
D(x, y), or its half-square (x, y), the Synge’s world-function,
1
(x, y) ≡ D(x, y)2 , (15)
2
given by
1
1 dγ dγ
(x, y) = g , dλ , (16)
2 0 dλ dλ
and verifying its fundamental equations:
g αβ ∂α ∂β = 2 , g ab ∂a ∂b = 2 , (17)
27 This notion of finite-differential geometry was first presented as part of a lesson at the International
School on Relativistic Coordinates, Reference and Positioning Systems, Salamanca, 2005. The
mathematical results also appeared in [1].
where Greek indices correspond to coordinates at the point x and Latin ones at the
point y.
Mathematically, distance spaces, i.e. manifolds endowed with a distance function,
are well known,28 but their link with differential geometry have not been sufficiently
explored.
An important obstruction for the interchangeability between the differential and
the finite concept is that the most part of distance functions are not geodesic distance
functions of any metric.
It is known that, when a distance function D(x, y) is the geodesic distance function
of a metric, this metric may be obtained as minus the limit y → x of the mixed first
derivatives of the half square (Synge’s world function) of the distance function.29
If this limit is applied to an arbitrary distance function, it may give rise to a zero,
degenerate or regular metric. But, even when this metric is regular, the distance
function D(x, y) will not be generically the geodesic distance function of it.
Suppose, in our physical case, that the interpolated distance function from the data
set is at least of differentiability class 2. We can obtain a metric from the above
limiting process, but this metric will depend strongly30 on the interpolation method
used, for which we have no control. For this and other applications, to discern if a
distance function is a geodesic distance function of a metric, it would be convenient
to have an IDEAL31 criterion involving solely the distance function, without any
limiting process.
I solved this problem some years ago. Let us introduce the following quantities
of the first and second derivatives of the distance function D(x, y):
α
Vmn ≡ αλμν Dλ Dmμ Dnν , V aα ≡ amn αλμν D Dλ Dmμ Dnν , (18)
as well as the quantity:

V α ≡ Vmn
α
x ym zn , (19)
where x , y m , z n are arbitrary independent directions. Define the two scalars

ρ
≡ Dλ V λ , ≡ r mn Vmn Dr Dρ , (20)
and form the two quantities:
Vα V aα
Dα ≡ , D aα ≡ 3 . (21)

Then, we have:
28 The concept is due to Fréchet. Haussdorff named them ‘metric spaces’ (‘metrischer Raum’) but
in our context it is better to call them ‘distance spaces’.

29 I am grateful to Abraham Harte for a pertinent observation on this fact.
30 At points out of the data of .
31 See Sect. 5 for this notion.
314 B. Coll
Theorem 6.1 (Structure theorem for geodesic distance functions) The necessary
and sufficient condition for a distance function D(x, y) to be the geodesic distance
function of a metric, is that its derivatives verify:

D ρ Dabcρ + D ρ D mσ (Damρ − D n Dmnρ Da )Dbcσ = 0 (22)
abc
where the subscripts denote partial derivatives and D a and D aα are the quantities
just defined.
In our physical case, these are the constraints to be imposed directly to any inter-
polated distance function on the data set . Once these equations verified at the
suitable degree of precision, one has to extract the metric from this distance function
without any limiting process. I solved this problem together with the above one. The
result is:
Theorem 6.2 (Metric of a geodesic distance function) In terms of the partial deriva-
tives Dα , Daα and Dabα of a geodesic distance function, the contravariant compo-
nents g αβ of the metric at the point x, are given by
g αβ = D α D β + D aα D bβ Dabγ D γ , (23)
where D α and D aα are the quantities above defined.
Coming back to our physical case, (23) would give us the metric in the region of the
constellation. Then, the intrinsic characterization method of Sect. 5, once generalized
to perturbations of Schwarzschild gravitational field, would give us the mass M that
satellizes the constellation C and generates the data set .
Theorems 6.1 and 6.2 give to the distance function D(x, y) the interchangeable
character with its differential homologue, the metric g(x), that our finite-differential
geometry requires.32 But finite-differential geometry has to be developed in many
other directions. Among them, perhaps the more urgent ones are the development of
methods of interpolation and approximation of distance functions.
Acknowledgements This work has been supported by the Spanish “Ministerio de Economía y
Competitividad”, MICINN-FEDER project FIS2015-64552-P.
32 On a metric manifold a distance function is generically local (normal geodesic domain), so that
a (global) metric cannot be generically interchanged by a sole distance function, but by a suitable
atlas of normal geodesic domains with distance functions submitted, in the intersection of charts,
to compatibility conditions. In our case, nevertheless, the constellation C is contained clearly in a
normal geodesic domain. Anyway, what we want here is to show the interest of the concept.
References
1. B. Coll, Relativistic positioning systems: perspectives and prospects. Acta Futura 7, 35–47
(2013)
2. B. Coll, J.J. Ferrando, J.A. Morales-Lladosa, Positioning systems in Minkowski space-time:
from emission to inertial coordinates. Class. Quantum Grav. 27, 065013 (2010)
3. B. Coll, Elements for a theory of relativistic coordinate systems. Formal and physical aspects, in
Proceedings of the Spanish Relativity Meeting 2000 on Reference Frames and Gravitomagnetism
(Valladolid, Spain) (World Scientific, Singapore, 2001), pp. 53–65
4. B. Coll, J.M. Pozo, Relativistic positioning systems: the emission coordinates. Class. Quantum
Grav. 23, 7395 (2006)
5. B. Coll, J.J. Ferrando, J.A. Morales-Lladosa, Positioning in a flat two-dimensional space-time:
the delay master equation. Phys. Rev. D 82, 084038 (2010)
6. B. Coll, J.J. Ferrando, J.A. Morales-Lladosa, Positioning in Minkowski space-times: Bifurcation
problem and observational data. Phys. Rev. D 86, 084036 (2012)
7. B. Coll, J.J. Ferrando, The Newtonian point particle, in Proceedings Spanish Relativity Meeting
1997, Palma de Mallorca, Spain (Pub. Univ. Illes Balears, 1997), pp. 184–190
8. J.J. Ferrando, J.A. Sáez, An intrinsic characterization of the Schwarzschild metric. Class. Quan-
tum Grav. 15, 1323 (1998)
9. J.J. Ferrando, J.A. Sáez, An intrinsic characterization of the Kerr metric. Class. Quantum Grav.
26, 075013 (2009)
Use of Geodesy and Geophysics
Measurements to Probe the Gravitational
Interaction
Aurélien Hees, Adrien Bourgoin, Pacome Delva,

Christophe Le Poncin-Lafitte and Peter Wolf
Abstract Despite its extraordinary successes, the theory of General Relativity is

likely not the ultimate theory of the gravitational interaction. Indeed, General Rel-
ativity as such is a classical theory and is therefore incomplete since it does not
include any quantum effects. Moreover, most physicists agree that GR and the Stan-
dard Model are only effective field theories that are low-energy approximation of a
more fundamental and more general theory that would provide a unified description
of all the fundamental interactions. On the observational side, Dark Matter and Dark
Energy are required to explain most of astrophysical and cosmological observations
and very few is known and this two Dark components, which is sometimes interpreted
as an hint that our theory of gravitation is incomplete. For these reasons, General
Relativity is confronted to an increasing number of measurements, searching for
deviations in more and more frameworks that extend General Relativity. Amongst
all the measurements used to search for and to constrain deviations from General
Relativity, a observations developed in the context of geodesy and geophysics are
playing an important role like for example atomic clocks comparison, gravimetry
measurements, satellite and lunar laser ranging, very long baseline interferometry,
etc …In this communication, we present briefly each of these geodesic/geophysics
measurements and show how they have recently been used to constrain extensions
of General Relativity, model of Dark Matter or of Dark Energy.
A. Hees (B) · P. Delva · C. Le Poncin-Lafitte · P. Wolf

SYRTE, Observatoire de Paris, Université PSL, CNRS, Sorbonne Université, LNE,
61 avenue de l’Observatoire, 75014 Paris, France
e-mail: aurelien.hees@obspm.fr
A. Bourgoin
Dipartimento di Ingegneria Industriale, University of Bologna,
via fontanelle 40, Bologna, Italy

318 A. Hees et al.
1 Introduction
The current paradigm to describe the gravitational interaction is the General theory
of Relativity (GR) published by Einstein in 1915. Since 1915, this theory has been
extremely successful making several predictions that were all verified observation-
ally. Nowadays, the domain of application of GR is extremely large: (i) at small scales
like for clocks comparison, time transfer, construction of time references, (ii) at the
scale of the Solar System where GR needs to be taken into account to describe the
motion of planets, of the Moon and of satellites, (iii) at the galactic scale to describe
the galactic dynamics and also the motion of stars orbiting our Galactic Center, and
(iv) at the cosmological scales to describe the behavior of the Universe and to explain
large scales observations. In spite of the overwhelming success of these theory in
describing much of the macroscopic observed universe, a number of open issues,
both theoretical and experimental remain.
First of all, on the theoretical side, GR is a classical field theory while the other
interactions from Nature are described within the framework of the Standard Model
of particles physics, which is based on a quantum field theory. As such, GR is
fundamentally incomplete since it does not include any quantum effects. Moreover,
most physicists agree that GR and the Standard Model are only effective field theories
that are low-energy approximation of a more fundamental and more general theory
that would provide a unified description of all the fundamental interactions. Most
attempts to derive such a theory that would unify the Standard Model with GR leads
to deviations either in the Standard Model, either in the gravitational sector, either
in both. These deviations occur at a level that is in general not predicted and not
predictable.
On the observational side, several galactic observations cannot be explained within
the frameworks of GR and of the Standard Model. In the most adopted paradigm, a
new type of cold matter called Dark Matter (DM) is introduced. So far, DM has not
been directly detected despite a large effort like e.g. with particles accelerator. The
nature of this elusive type of matter is therefore still completely unknown. Note that
while the introduction of new particles to explain DM correspond to a modification
of the Standard Model, it has also been suggested that DM could be the result of a
modification from GR (this is the case for the MOND theory for example, see e.g.
[1]). In addition, for some model of DM, the occupation number of the corresponding
field will be relatively high and the particle will actually behave as a classical field.
This is the case for some model of ultra light bosonic DM. In that case, DM can also
be interpreted as a classical modification from GR.
In addition to DM, large scale observations like the temperature fluctuations of the
cosmic microwave background, the redshift-distance measurements from Supernova
Ia or baryonic acoustic oscillations have shown that our Universe undergoes a phase of
accelerated expansion that is attributed to Dark Energy (DE). In the simplest scenario,
DE is attributed to a cosmological constant whose fine-tuned value is completely
unexplained. A lot of modifications of GR have been suggested to provide a more
Use of Geodesy and Geophysics Measurements to Probe the Gravitational Interaction 319
elegant solution to DE by introducing at least a new dynamical field that would cause
our Universe to accelerate (see e.g. [2]).
All these considerations have pushed the scientific community to confront GR to
a wide number of experiments searching for tiny deviations that would be the hint
of a new physics and that would guide the construction of new models of unification
theory, of DM and of DE. So far, despite measurements that are more and more
accurate, Einstein’s theory has passed all the observational tests with flying color (see
e.g. [3]). As already mentioned above, none of the alternative theory of gravitation
developed is making a clear prediction for the actual level at which a deviation
will become detectable. This property makes the search for new physics particularly
difficult but also very exciting. These last decades, searches for deviations from GR
have become a very active field of research. In particular, outstanding performance
provided by the development of technology developed for geodesy or geophysics
have been used to develop new high-accuracy fundamental physics experiments.
The goal of this communication is to give several successful examples of mea-
surements and experiments that have been developed in the context of geodesy and
geophysics and that have been used in order to search for new physics. The range
of measurements that will be considered is relatively large. Amongst other, we will
consider several laboratory experiments like the comparison of very accurate atomic
clocks or atomic gravimetry. Best present day clocks reach relative uncertainties in
the low 1018 range and progress is rapid with no obvious hard limit in sight. Amongst
the applications of the comparison of high stable clocks is geodesy and the measure
of the gravitational potential (see e.g. [4]) the experimental study of fundamental
physics, and in particular the two fundamental theories mentioned above (GR and
standard model). Gravimeters on the other hand are mainly used to measure the grav-
itational acceleration for geophysics purposes but have also found an application in
testing GR.
At larger scales, GNSS satellites but also satellite laser ranging have been widely
used in geodesy. In particular, they are used to derive high-precision models of Earth’s
gravity field. Both these techniques have also been used successfully to test the theory
of GR with an impressive accuracy. In addition, Very Long Baseline Interferometry,
used in particular to monitor the Earth’s rotation has also been successfully used to
probe the deflection of electromagnetic signals from quasar by the Sun.
Section 2 from this communication is devoted to the description of the different
measurements and experiments that have been developed in the context of geodesy
or geophysics and that have found an application in fundamental physics. Section 3
will present the theoretical background of GR, several alternative frameworks that
allow scientists to search for deviation from GR and how measurements made using
geodesic techniques have led to constraints on these alternative frameworks. We will
also focus on the theoretical implications of such constraints. It is impossible to
review all the alternative theories of gravitation and their constraints. In this com-
munication, we will therefore select a couple of them that are very important and/or
quite recent to illustrate this interplay between geodesy/geophysics and fundamental
320 A. Hees et al.
physics. We hope this will motivate more interactions between these two communi-
ties to develop more and more accurate experiments that would potentially lead to
the discovery of a tiny deviation from GR that would be the sign of new physics.
2 Brief Description of Several Geodetic/Geophysical

Measurements used for Fundamental Physics
The goal of this section is to give a brief overview of several different measurements
that have been developed in the context of geodesy or geophysics and that have found
some application in fundamental physics.
2.1 Gravimetry
Gravimetry is the measurement of the local gravitational (plus inertial due to the
Earth’s rotation) acceleration g experienced by an observer on the Earth. More gen-
erally, accelerometry measures the local acceleration experienced by an observer,
whatever its origin (gravitational or inertial). Simply speaking a gravimetric mea-
surement consists of “dropping” an object and measuring its acceleration using a
ruler and a clock. In modern gravimetry the ruler is replaced by a laser whose wave-
length is ultimately referenced to an atomic clock and to the SI definition of the meter
and the second, thus providing an absolute measurement of local acceleration is SI
units.
Classical absolute gravimeters use falling corner cubes and laser interferome-
try to determine the local acceleration [5]. More recently these instruments have
become rivaled by gravimeters based on atom interferometry, that use falling atoms
in quantum superpositions with lasers as the read-out of their acceleration [6]. Best
absolute gravimeters reach uncertainties in the low 10−8 m/s2 . Additionally relative
gravimeters and accelerometers measure time-variations of local acceleration, but
not its absolute value, using in general electrostatic devices and readout (possibly
superconducting). Such relative accelerometers are widely used is space missions
like GRACE, GOCE or MICROSCOPE.
Gravimeters and accelerometers, whether relative or absolute, have been widely
used in geodesy and geophysics (on ground and in space) for many decades. They
provide the bulk of data that is used for the determination of the geopotential and
numerous other applications. As described below, they have also been widely used
in fundamental physics, in particular for tests of the universality of free fall, which
is one of the foundations of general relativity.
2.2 Atomic Clock Comparison
Quite generally, an atomic clock is a device that delivers an electromagnetic signal

(typically in the microwave or optical domain) whose frequency ν is locked to the
unperturbed transition frequency between two quantum states of an atom or ion
2π(E e − E g )
ν= , (1)

where E e and E g are the exited and ground state energies of the atom/ion, and
is the reduced Planck constant. As all atoms/ions of a given isotope are identical,
an atomic clock is thus a universal time piece, delivering the same proper time
up to technical imperfections. The best atomic clocks are presently approaching
impressive uncertainties of one part in 1018 [7–11] in fractional frequency which
makes them some of the most accurate measurement devices ever built. They can be
compared inside the same laboratory or over large distances using the exchange of
electromagnetic signals in free space and via satellites, or through cables or optical
fibers. The realization of such links, especially over large distances and with an
uncertainty compatible with the best clocks is a major challenge today and a subject
of active research [12].
Because of their exceptional stability and accuracy, and the fact that they are
the basic building block of any space-time measurement, atomic clocks have been
used in fundamental physics for many decades and some of those applications will be
discussed in detail below. But more recently they have started to also be studied in the
context of what has become known as “chronometric geodesy”, the determination of
geopotential differences between two distant locations using the comparison of two
clocks [4, 13, 14]. Their measured frequency difference then provides a measurement
of the geopotential difference at the two locations via the gravitational redshift of
general relativity. A fractional uncertainty of 1 × 10−18 on such a measurement
corresponds to an uncertainty of 0.1 m2 /s2 in geopotential difference or 1 cm in
height, which is comparable to the best present geopotential models obtained from
gravimetry and satellite geodesy.
2.3 Satellite Laser Ranging
In 1964, NASA carried out the first laser ranging to a near-Earth satellite. Since that
time, ranging precision has improved by a factor of a thousand from a few meters
to a few millimeters, and more satellites equipped with corner cubes have been
launched. During the subsequent decades, the first objectives of dedicated Satellite
Laser Ranging (SLR) experiments (for example, STARLETTE, STELLA, LAGEOS)
were to produce high accuracy orbitography of those satellites and, consequently, to
study in details the Earth gravity field, i.e. the solid Earth. SLR technique is also used
322 A. Hees et al.
in altimeter missions mapping the ocean surface, for mapping changes in continental
ice masses, for topography.
With a measurement accuracy better than the part-per-billion level, GR must be
considered by adding additional perturbations to the orbit dynamics, corrections to
the light-time computation and more fundamental aspects of the definition of the
geocentric reference frame. While these effects are significant, they are generally
not large enough to provide competitive tests of GR with those available from lunar
laser ranging and other Solar System tests. An important exception, however, is
the relativistic prediction of the Lense-Thirring orbit precession, i.e. the effect of
frame-dragging on the satellite orbit due to the spinning Earth mass. Using the two
LAGEOS satellites, [15, 16], Ciufolini and collaborators were able to detect at the
level of 10% this effect. In 2012, a new mission, completely passive with 92 corner
cubes, called LARES has been launched to improve this determination[17].
2.4 Lunar Laser Ranging
On August, 20th 1969, after ranging to the lunar retro-reflector placed during the
Apollo 11 mission, the first LLR echo was detected at the McDonald Observatory in
Texas. Currently, there are five stations spread over the world which have realized
laser shots on five lunar retro-reflectors. Among these stations four are still operating:
Mc Donald Observatory in Texas, Observatoire de la Côte d’Azur in France, Apache
point Observatory in New Mexico and Matera in Italy while one on Maui, Hawaii
has stopped lunar ranging since 1990. Concerning the lunar retro-reflectors three are
located at sites of the Apollo missions 11, 14 and 15 and two are French-built array
operating on the Soviet roving vehicle Lunakhod 1 and 2.
LLR is used to conduct high precision measurements of the light travel time of
short laser pulses emitted at time t1 by a LLR station, reflected by a lunar retro-
reflector and finally received at time t3 at a station receiver. The data are presented as
normal points which combine time series of measured light travel time of photons,
averaged over several minutes to achieve a higher signal-to-noise ratio measurement
of the lunar range at some characteristic epoch. Each normal-point is characterized
by one emission time (t1 in universal time coordinate – UTC), one time delay (tc
in international atomic time – TAI) and some additional observational parameters as
laser wavelength, atmospheric temperature and pressure etc.
LLR measurements are used to produce the Lunar ephemeris but also provide a
unique opportunity to study the Moon’s rotation, the Moon’s tidal acceleration, the
lunar rotational dissipation, etc [18]. In addition, LLR measurements have turn the
Earth-Moon system into a laboratory to study fundamental physics and to conduct
tests of the gravitation theory.
2.5 Interplanetary Spacecraft Tracking and Planetary

Ephemerides
Modeling and understanding the motion of planets and minor bodies in the Solar
System has been a long active field of research. This is important to understand the
formation of our Solar System [REF NEEDED?], to successfully develop interplan-
etary space-missions, to predict an hypothetical “rendez-vous” between the Earth
and an asteroid, to detect and quantify the presence of Dark Matter in our Solar Sys-
tem [19], etc…In addition, the analysis of the motion of the planet Mercury around
the Sun was historically the first evidence in favor of GR with the explanation of
the famous advance of the perihelion in 1915. From there, planetary ephemerides
have always been a very powerful tool to constrain GR and alternative theories of
gravitation.
Currently, three groups in the world are producing planetary ephemerides: the
NASA Jet Propulsion Laboratory with the DE ephemerides [20–26], the French
INPOP (Intégrateur Numérique Planétaire de l’Observatoire de Paris) ephemerides
[27–32] and the Russian EPM ephemerides [33–37]. These analyses use an impres-
sive number of different observations to produce high accurate planetary and asteroid
trajectories. The observations used to produce ephemerides comprise radioscience
observations of spacecraft that orbited around Mercury, Venus, Mars and Saturn,
flyby tracking of spacecraft close to Mercury, Jupiter, Uranus and Neptune and opti-
cal observations of all planets.
2.6 Very Long Baseline Interferometry
VLBI is a geometric technique measuring the time difference in the arrival of a

radio wavefront, emitted by a distant quasar, between at least two Earth-based radio-
telescopes. VLBI observations are done daily since 1979 and the database contains
nowadays almost 6000 24 h sessions, corresponding to 10 millions group-delay obser-
vations, with a present precision of a few picoseconds. One of the principal goals of
VLBI observations is the kinematical monitoring of Earth rotation with respect to a
global inertial frame realized by a set of defining quasars, the International Celestial
Reference Frame [38], as defined by the International Astronomical Union [39]. The
International VLBI Service for Geodesy and Astrometry (IVS) organizes sessions
of observation, storage of data and distribution of products, in particular the Earth
Orientation parameters. Because of this precision, VLBI is also a very interesting
tool to test gravitation in the Solar System. Indeed, the gravitational fields of the
Sun and the planets are responsible of relativistic effects on the quasar light beam
through the propagation of the signal to the observing station and VLBI is able to
detect these effects very accurately.
324 A. Hees et al.
3 Theoretical Background and Tests of General Relativity
The classical theory of General Relativity provides a geometrical description of the

gravitational interaction. It is based on two fundamental principles: (i) the Einstein
Equivalence Principle (EEP) and (ii) the Einstein field equations that can be derived
from the Einstein–Hilbert action. All GR extensions or alternative theories of gravi-
tation will break at least one of these principles.
3.1 The Einstein Equivalence Principle
The first part of GR, the EEP, published in 1911 by Einstein [40], gives gravitation
a geometric nature. This principle implies that gravity can be identified to space-
time geometry which is described mathematically by a symmetric order 2 tensor, the
space-time metric gμν . More precisely, the EEP implies that there exists only one
space-time metric to which all matter minimally couples to (see [41]). In practice,
this means that the equations of motion for matter can be derived from the action of
the Standard Model of particles in which the Minkowski metric ημν is replaced by
the space-time metric gμν

√
Smat = d 4 x −g Lmat , gμν , (2)

where g is the determinant of the space-time metric and Lmat , gμν is the standard
Lagrangian for matter depending on the matter fields .
From a theoretical point of view, this part of Einstein theory allows one to derive the
effects of gravitation from the space-time curvature. In particular, this implies that test
bodies follow geodesic from this space-time. Furthermore, ideal clocks will measure
the quadratic invariant of the space-time metric dτ 2 = −gμν d x μ d x ν . Similarly, the
propagation of electromagnetic waves is governed by Maxwell equations in which
standard derivatives are replaced by covariant derivatives [42]. At the geometric optic
approximation, this implies that light rays are described by null geodesics.
From a phenomenological point of view, three aspects of the EEP can be tested
(see [3, 43]): (i) the Universality of Free Fall (UFF), (ii) the Local Lorentz Invariance
(LLI) and (iii) the Local Position Invariance (LPI). Furthermore, the Schiff conjecture
stipulates that any complete, self-consistent theory of gravity that embodies the UFF
necessarily embodies the two other parts of the EEP: the LPI and the LLI (see e.g.
[3]). Nevertheless, no complete proof of this conjecture has been proposed and some
counter-examples are known (see the discussion in Sect. 2.1.1 of [3]).
3.2 Einstein Field Equations
The EEP allows one to derive the effects of gravitation from the space-time metric
gμν . In order to have a complete theory, one needs to give a prescription to determine
the form of this metric. In GR, the form of the metric is determined by solving the
Einstein field equations
1 8πG
Rμν − gμν R + gμν = 4 Tμν , (3)
2 c
where Rμν and R are the Ricci tensor and scalar curvature from the metric, G the
gravitational constant, c the speed of light in vacuum, Tμν the stress-energy tensor and
the cosmological constant. These equations characterize the dynamic of space-time
geometry essentially described in the left part of the equation and the matter/energy
content in this space-time essentially described by the stress-energy tensor. In other
words, this set of equations describes how space-time is curved by the presence of
matter and energy.
The Einstein field equations can be derived by means of a variational principle
from the Einstein–Hilbert action (see [44])

c4 √
Sgrav = d 4 x −g (R − 2) . (4)
16πG
The total action which describes completely the GR theory is the sum Sgrav + Smat
which makes GR an extremely simple theory from a conceptual point of view. It
is also interesting to mention that due to the Lovelock theorem, the Einstein field
equations are the only second order equations that can be derived from a least-
principle action based on the space-time metric (and its derivatives) in a 4 dimensional
Riemannian space (see [45, 46]). Following this theorem, alternative (metric) theories
of gravitation will always imply one of the following (see [2]):
• The existence of new fields in addition to (or instead of) the space-time metric.
• The existence of higher order derivatives of the metric in the field equations.
• To work in a space-time with higher dimension than 4.
• To give up locality.
If the EEP specifies how different types of mass-energy react to gravitation, the
Einstein field equations govern how gravitation is generated by matter and energy. A
modification of the Einstein–Hilbert action will lead to different field equations that
can be reflected in differences in space-time geometry. From an experimental point of
view, it is of prime importance to search for metric deviations that would be produced
by any deviations from GR. The class of metric theories of gravitation is huge and it is
particularly difficult to analyze observational data in a general framework. So far, two
frameworks have been widely used to analyze data: the parametrized post-Newtonian
(PPN) formalism (see [3, 43]) and the fifth force formalism (see [47–51]). As we
326 A. Hees et al.
shall see in Sect. 4.2, these two formalisms have been very precisely constrained with
observations at Solar System scales.
Nevertheless, it is useful to analyze experimental results in new extended frame-
works. Indeed, even if observations lie very close to GR when analyzed within the
PPN or fifth force framework, this does not mean that this has to be true in any
other framework. The existing frameworks indeed cover a limited set of alternative
theories of gravity (see [52, 53]). For example, a formalism aiming at considering
violation of Lorentz invariance in the gravity sector has been develop within the SME
framework (see [54–57]). In this formalism, an expansion at the level of the action
is performed which naturally leads to a post-Newtonian metric that differs from the
PPN one. Other examples of theories not entering the PPN or fifth force frameworks
are given by the MOND (MOdified Newtonian Dynamics) phenomenology which
produces a quadrupolar deviation from the Newton potential in the Solar System
[58, 59], by the post-Einsteinian Gravity (PEG, see [60, 61]), by the parametrized
post-Newtonian-Vainshteinian formalism [62], by Horndeski’s gravity [63], by mas-
sive tensor-scalar theories, … Currently, only few data analysis are exploring these
formalisms or theories.
4 Frameworks to Search for New Physics and Constraints

on Their Parameters
The task of confronting each of the modified gravitational theory to observations

is huge. In practice, several frameworks have been used to analyze observations in
order to search for new physics in the gravitational sector. In this section, we will
present several of these frameworks and show how geodesy or geophysics measure-
ments developed in Sect. 2 have been used to constrain their parameters. The list
of the frameworks presented in this communication is obviously not exhaustive but
demonstrates clearly the impact of geodesic measurements in fundamental physics.
The different formalisms presented in this section are motivated at different level
from a theoretical perspective: some of them are complete theory characterized by
a full action with new fields in addition to the standard space-time metric, some are
phenomenology developed at the level of the action (the SME for example), some
are phenomenology developed at the level of field equations (MOND for example),
some are phenomenology developed at the level of the action (the PPN formalism
is a good example) and some are only a phenomenological parametrization at the
level of the observable (the redshift test or the UFF measurement are good example).
They all have advantages and drawbacks but are aiming at facilitating the search for
new physics with observations.
4.1 Tests of the Einstein Equivalence Principle
4.1.1 The Universality of Free Fall
The UFF states that the motion of a test body is independent of its composition. It
has been constrained by various experiments. Precise tests of the UFF compare the
free fall accelerations, a1 and a2 , of two different test bodies 1 and 2 falling in the
gravitational field sourced by a body S. A succinct phenomenological parametrization
for the test of the UFF takes the form (see [3, 43]).

a a1 − a2 mP mP
η= =2 ≈ − , (5)
a S;1−2 a1 + a2 mA 1 mA 2
where m P and m A are the passive and active masses of each body. This empirical
formulation is quite generic and it is interesting to test the UFF by using different
types of test bodies. The most thoroughly tested and known version of this principle
is attributed to Galileo Galilei and consider two macroscopic test bodies of different
compositions falling in the same gravitational field. This version is currently tested
at the impressive level of 10−13 with torsion balances [51, 64–66]. In addition,
LLR observations also provide a wonderful tool to test the UFF by searching for a
differential acceleration between the Moon and the Earth with respect to the Sun.
Formally, search for a breaking of the UFF using extended objects like the Earth
and the Moon tests a combination of the Einstein Equivalence Principle and of the
Strong Equivalence Principle (see e.g. [67]). Such a violation of the UFF will lead
to a polarization of the Moon’s orbit as notice by Nordtvedt [68]. Analyzes of LLR
observations described in Sect. 2.4 has also provided a constraint on the UFF at the
level of 10−13 on η [69, 70]. Very recently, new infrared observations from the Grasse
station [71] has allowed to measure the range to the Moon during the new and the
full Moon periods, when the signature from a breaking of the UFF is maximal. This
lead to the best LLR constraint on the UFF between the Moon and the Earth (in the
field of the Sun) given by: η = (−3.8±7.1)×10−14 [67].
Finally, very recently, the first results of the MICROSCOPE space mission has
been released. The MICROSCOPE project consists in performing a test of the UFF
between two macroscopic bodies in space. The main advantage of performing this test
in space comes from the fact that the free fall time baseline can be increased signifi-
cantly compared to what is achievable on Earth. An analysis of the first dataset from
this mission gave η = (−1 ± 9 (stat) ± 9 (syst)) × 10−15 [72]. This result, which
has been interpreted for different theoretical models (see [73, 74]), is expected to be
improved by roughly an order of magnitude with the full data of the mission.
More recently, the UFF has been tested by comparing the acceleration measured
by a macroscopic mass with respect to the acceleration measured by a microscopic,
quantum system and also by comparing the accelerations measured by two differ-
ent microscopic quantum systems. This has been achieved using atom interferom-
etry which provided a measurement of the local gravitational acceleration g (see
328 A. Hees et al.
Sect. 2.1). The current constraints on the UFF between macroscopic and micro-
scopic systems are at the level of 10−9 [75]. Similar constraints using two different
types of atom interferometers have provided constraints at the level of 10−9 as well
[76–78]. Furthermore, some modified theories of gravitation predict a violation of
the UFF between bodies with different spins (see as examples [79–84], and refer-
ences therein). Therefore, the UFF has also been tested by considering atoms with
different spins and the related constraints are at the level of 10−7 [85]. One can
wonder if anti-matter falls any differently than standard matter. Tests of the UFF
using anti-hydrogen atoms are currently on-going at CERN with the AEgIS [86] and
GBAR [87] experiments.
Predictions of a violation of the UFF between a black hole and standard matter
has also been recently predicted and tested for a very particular case of Galileon
theory by using astrophysical observations [88, 89].
Finally, the question of the validity of the UFF in the Dark sector remains com-
pletely open. Since dark matter has not been directly detected so far, tests of the EEP
in the dark sector are model dependent. A lot of models introduced a long range
interaction in the dark sector usually leading to a fifth force experienced either by
standard matter or by Dark Matter (see [90–93]). This hypothetical fifth force has
been constrained using various astrophysical observations (see for examples [92–98],
and references therein). More recently, a model of Dark Matter coupled to a scalar
field has been proposed by Damour et al. [99] and predicts a violation of the EEP
between Dark Matter and standard matter. Initially introduced as a model leading to
a time variation of the gravitational constant G, it has been shown that the violation
of the EEP in the Dark sector can naturally produce Dark Energy [100–102]. This
violation of the EEP has recently been constrained using galactic observations by
Mohapi et al. [103].
The UFF is certainly the first observational principle upon which the gravitation
theories (Newton’s theory and later GR) have been built. As described in this section,
testing the UFF can be performed using a wide various of test bodies and take various
forms. In particular, observations that are used for geodesic purposes (LLR, atomic
interferometry) gives some of the best current constraints. Considering that this
principle is at the heart of the gravitation theory, testing its experimental validity
is crucial and several projects aimed at improving the current searches for a UFF
violations.
4.1.2 Local Lorentz Invariance
is a feature of relativity stating that the outcome of any local non gravitational exper-
iment is independent of the velocity and orientation of the apparatus (see e.g. [3]).
While completely integrated into special relativity, the effort to test LLI has recently
increased. Indeed, it is often believed that models of quantum gravity will produce
violations of Lorentz invariance. These models usually introduce a fundamental
length (the Planck length). Since this length is not Lorentz invariant, in most scenar-
ios, a breaking of Lorentz symmetry at some level is produced (see the discussion in
Sect. 2.1.2 from [3] or [104]). Several formalisms have been used to test Lorentz sym-
metry. The simplest one predicts anisotropies in the speed of light, often parametrized
by the c2 -formalism (see e.g. [3]). Such a violation requires the existence of a privi-
leged reference frame.
A violation of the LLI would produce a shift in the energy levels of a particle,
which depends on the orientation of the quantization axis and the quantum number
of the state. The “clock anisotropy” experiments verify that this shift is null. The first
experiments of this type were done by [105] and [106]. The cooling of atoms and
trapped ions then permitted to limit collision effects, which further increased the accu-
racy of the experiments [107–109]. Other methods exist to test the LLI. For example,
Michelson–Morley type experiments and its numerous variants (see e.g. [110] for a
review): let’s cite the famous [111] experiment, which used a Fabry–Perot interfer-
ometer, and the experiments comparing the frequencies of electromagnetic cavities
with each other, or with atomic clocks [112, 113].
Another widely used formalism to test Lorentz Invariance (LI) is called the
Robertson-Mansouri-Sexl (RMS) framework [114–117] in which a modification of
the kinematic Lorentz transformation is parametrized by functions which depend
on the relative velocity w between a preferred frame – usually taken as the cosmic
microwave background – and the observer frame. The three classical LI tests are
the Michelson–Morley, Kennedy–Thorndike, and Ives–Stillwell experiments (see
e.g. [114]); they are second-order tests as the LI violating signal depends on w2 /c2 ,
where w = |w| and c is the velocity of light in vacuum [116]. With the advent of
heavy-ion storage rings, Ives–Stillwell type experiments gave until recently the best
constraint on α, the time dilation parameter of the RMS parametrization. A limit of
|α| 8.4 × 10−8 was found using 7 Li+ ions prepared in a storage ring to 6.4% and
3.0% of the speed of light [118]. The experiment described in [119] uses 7 Li+ ions
confined at a velocity of 33.8% of the speed of light. When neglecting higher order
RMS parameters, the constraint on the LI violating parameter is |α| 2.0 × 10−8.
First-order tests in w/c are based on the comparison of clocks [115, 120]. Compar-
ing atomic clocks on-board GPS satellites with ground atomic clocks, [121] obtained
the constraint |α| 10−6 .
A recent test within the RMS framework is based on comparison of four optical lat-
tice clocks using Sr atoms, two located at SYRTE, Observatoire de Paris, France [122,
123], one at PTB, Braunschweig, Germany [124, 125], and one at NPL, Teddington,
UK [126]. These clocks are connected by two fibre links, one running from SYRTE
to PTB operated in June 2015 [127], and one from SYRTE to NPL operated in June
2016. In a simplified set-up, an optical clock comparison using a phase noise com-
pensated fibre link can be described as a two-way frequency transfer between two
observers A and B (see e.g. [128–132]). By searching for a daily variation of the
frequency difference these four strontium optical lattice clocks, [133] improve upon
all previous tests of RMS time dilation parameter with |α| 1.1 × 10−8 .
More recently, a very wide formalism has been developed to consider hypothet-
ical violation of Lorentz symmetry in all fields of physics. Contrarily to the RMS
framework, this formalism is dynamical. This framework has been named Standard
Model Extension (SME) and contains an impressive number of parameters encod-
330 A. Hees et al.
ing hypothetical deviations from the Standard Model of particles or from GR (see
e.g [134–137], and references therein). A number of these SME coefficients have
been constrained by various experiments (for a review of the current constraints
on SME parameters, see [138]). It is worth mentioning that in the matter-sectors,
atomic clocks and gravimetry has been used to constrain some SME coefficients (see
as examples [113, 139, 139–142]). The SME formalism has also been developed in
the gravitational sector to model deviations from Einstein’s theory at the level of the
gravitational part of the action. This part of the SME formalism will be developed
in Sect. 4.2.4.
4.1.3 Local Position Invariance
Local Position Invariance (LPI) stipulates that the outcome of any local non-
gravitational experiment is independent of the space-time position of the freely-
falling reference frame in which it is performed [3]. This principle is mainly tested
by two types of experiments: (i) search for variations in the constants of Nature and
(ii) redshift tests.
The question of the constancy of the constants of Nature was first addressed by
Dirac. This question is driven by the principle of reason: what could be the reasons
behind the specific values of the constants of Physics? (see the discussion in Sect. 2
of Damour [143]). This argument led to many developments of new theories where
the constants of physics become dynamical entities. In parallel, many observational
investigations try and search for any space/time evolution of the constants of Physics
[144].
Amongst all the observations performed, atomic clocks have an important role
leading to currently some of the best constraints currently available. In particular,
linear drifts in the evolution of the fine structure constant α, in the ratio μ between the
mass of the electron and the mass of the proton and in the ratio between the mass of the
light quarks (up and down) and the quantum chromodynamics (QCD) energy scale
3 . Several groups in the world have pursued effort to constrain such hypothetical
linear drifts: at SYRTE, Observatoire de Paris [145], NIST [146], Berkeley [147],
NPL [148], PTB [149], …The current constraints on a linear drift variation of the
three constants (fine structure constant, ratio between the mass of the electron and
the mass of the proton and ration between the mass of the light quarks and the QCD
energy scale) are at the level of 10−16 per year. Note that astrophysical observations
also constrain variations of the constants of Nature [144] and furthermore, these
observables can be related to other cosmological observables like the evolution of
the cosmic microwave background temperature, spectral distortion of the cosmic
microwave background and violation of the cosmic distance-duality relation (see for
example [150–152]).
In addition to temporal variations of the constants of Nature, one can search for
spatial variations. Regarding this, atomic clocks have also been widely used to search
for a variation of the constants of Nature with respect to the gravitational potential of
the Sun. The idea is to compare two clocks working on different atomic transitions
and therefore sensitive differently to the different constants of Physics, located at the
same place and search for periodic variations in their frequencies comparison. Several
groups in the world have been measuring this effect which is now constrained at the
level of 10−6 : the SYRTE [145], the USNO [153], Berkeley [147], NIST [154],…
The second way to test the LPI is to measure the gravitational redshift. The
gravitational redshift is a consequence of the EEP predicted by A. Einstein, 1911.
It was observed for the first time by Pound and Rebka in 1959 [155]. A simple and
convenient formalism to test the gravitational redshift is to introduce a new parameter
αred defined through [3]
ν U
= (1 + αred ) 2 (6)
ν c
where ν is the difference between the observed frequency of the same signal mea-
sured in different locations in the gravitational potential U . The parameter αred van-
ishes when the EEP is valid. The best constraint on αred is at the level of 10−4 and
has been obtained in 1976 by comparing the frequency of two clocks: one onboard
a rocket and the other one on Earth [156]. An improvement on this constraint is
expected in the near future by using observations of GNSS satellites Galileo V and
VI. These two satellites were launched on August, 30th 2014. Because of a technical
problem, the launcher brought them on a wrong, elliptic orbit which makes them dif-
ficult to use for GNSS purposed. Nevertheless, the eccentricity of the orbit induces
a periodic modulation of the gravitational redshift, which combined with the good
stability of recent GNSS clocks can be used to test the gravitational redshift to a
very good level of accuracy [157]. Contrary to the GP-A experiment, it is possible to
integrate the signal on a long duration, therefore improving the statistics. A thorough
study of the statistical and systematic uncertainties has shown that a constraint on
αred at the level of 10−5 can be achieved with a year of data [157]. The main limitation
for this experiment comes from systematic effects that have the same signature than
the redshift signal. These systematics are mainly due to orbital mismodelling and in
particular due to mismodelling of the solar radiation pressure. Nevertheless, Delva
et al. [157] have shown that a year of data leads to a decorrelation between the two
signals.
On the mid-term, the Atomic Clock Ensemble in Space (ACES) project [158]
that aims at comparing atomic clocks on the international space station (ISS) with
atomic clocks on Earth will provide a test of the redshift test at the level of 10−6 . The
ACES payload is expected to be launched on board of the ISS in 2018 for a duration
between 18 months up to 3 years. The payload includes the first cold atom clock in
space, PHARAO, consisting of a Cs clock and of a H-maser. Another key element is
the microwave link (MWL) that uses radio in a two-way configuration to compare
space to ground clocks at unprecedented stability and accuracy. The test of the EEP
is only one of the scientific objective of ACES. In addition, the measurement of
the gravitational redshift can be used to measure gravitational potential differences
between different clock locations, which is a new type of geodetic measurements
using clocks called chronometric geodesy [4, 159–162].
332 A. Hees et al.
In the longer term, several future space missions have been proposed to further
improve the test of the gravitational redshift. One example is the SpaceTime Explorer
and QUantum Equivalence Space Test (STE-QUEST) space mission [163], a M-
class mission proposal that was pre-selected by the European Space Agency in 2010
together with four other missions for the cosmic vision M3 launch opportunity. It
carries out tests of different aspects of the Einstein Equivalence Principle using atomic
clocks, matter wave interferometry and long distance time/frequency links, providing
fascinating science at the interface between quantum mechanics and gravitation that
cannot be achieved, at that level of precision, in ground experiments. STE-QUEST
will test the UFF at a level of 10−15 , two orders of magnitude better than the best
present result using macroscopic test-objects, and at least 6 orders better than present
UFF tests in the quantum regime. It will also improve best present results on redshift
tests (in the field of the Earth, Sun, and Moon) by up to 4 orders of magnitude.
Finally, let us mention that all these redshift tests are performed in a low gravi-
tational field in the Solar System. A “strong” field version of this test was recently
performed by measuring the gravitational redshift from the star S0-2 orbiting the
supermassive black hole in our Galactic Center, Sagittarius A* [98, 164–166].
4.1.4 Models of Ultralight Dark Matter
Motivated by the unsuccessful searches for a Dark Matter particle at high energy,
models of light scalar DM have recently gained a lot of attention in the scientific
community (see e.g. [167–189] and references therein). In those models, a light
scalar field is introduced in addition to the standard space-time metric and to the
standard model fields. Such scalar fields are also ubiquitous in theories with more
than 4 dimensions, and in particular in string theory with the dilaton and the moduli
fields [190–194].
Such models have been shown to produce nice galactic and cosmological predic-
tions for very low masses of the scalar field ranging from 10−24 to 10−22 eV [169, 173,
174, 182–184, 186, 187, 189]. Because of the high occupation numbers in galactic
halos, the scalar field can be treated as a classical field for masses eV [169, 181]
and this model of DM is actually a particular tensor-scalar modification of GR which
can break the Einstein Equivalence Principle through the coupling between the scalar
field and standard matter. The full action for this model writes

c3 4 √ μν
c2 m 2ϕ
S= d x −g R − 2g ∂μ ϕ∂ν ϕ − 2 2 ϕ + Smat gμν , , ϕ ,

16πG
(7)
where ϕ is the scalar field of mass m ϕ and Smat is the matter action including the
coupling between the scalar field and standard matter. A widely used parametrization
of the interaction between regular matter and the scalar field is introduced in [195]
and is given in terms of the following Lagrangian
( j)
( j) de dg( j) β3 A 2 ( j) (1)
Lint =ϕ j
F −
2
F − dm i + γm i dg m i ψ̄i ψi . (8)
4μ0 2g3 i=e,u,d
The superscripts ( j) indicate the type of coupling considered: a coupling linear in ϕ or

a coupling proportional to ϕ2 . In this Lagrangian, Fμν is the standard electromagnetic
Faraday tensor, μ0 the magnetic permeability, Fμν A
the gluon strength tensor, g3 the
QCD gauge coupling, β3 the β function for the running of g3 , m j the mass of the
fermions (electron and light quarks), γm j the anomalous dimension giving the energy
running of the masses of the QCD coupled fermions and ψ j the fermion spinors. The
constants d (i)
j characterize the interaction between the scalar field ϕ and the different
matter sectors.
This interaction Lagrangian leads to the following effective dependency of five
constants of Nature

αEM (ϕ) = α 1 + de(i) ϕi , (9a)

m j (ϕ) = m j 1 + dm(i)j ϕi for j = e, u, d (9b)

3 (ϕ) = 3 1 + dg(i) ϕi , (9c)
where αEM is the electromagnetic fine structure constant, m j are the fermions (elec-
tron and quarks up, down and strange) masses, 3 is the QCD mass scale 3 and
the superscripts (i) indicate the type of coupling considered (linear for i = 1 and
quadratic for i = 2).
At the cosmological level, this scalar field will oscillate at a frequency ω that is
directly related to its mass m ϕ through ω = c2 m ϕ /
ϕ(t) = ϕ0 cos (ωt + δ) . (10)
In addition, at the cosmological level, it can be shown that this scalar field can be
interpreted as a perfect fluid whose mean pressure vanishes, making it a good DM
candidate. Under the assumption that the DM is made completely by one mode of
this scalar field, the amplitude of the scalar oscillations are directly determined by
the DM energy density through
c6 m 2ϕ ϕ20
ρϕ = . (11)
4πG2 2
In addition, the coupling of the scalar field with the standard model fields will
exhibit signatures that can be searched for using different types of measurements.
These signatures depend highly on the type of coupling considered.
Linear coupling: When the coupling between standard matter and the scalar field is
linear in the Lagrangian from Eq. (8), the scalar field will be the sum of two distinct
contributions: (i) a static Yukawa contribution, which is characteristic of massive
334 A. Hees et al.
gauge boson and which is independent of the identification of the scalar field to dark
matter and (ii) an oscillating contribution which is identified as dark matter.
The Yukawa interaction is a generic feature appearing in most theories that intro-
duce a new massive gauge boson. Measurements of the UFF are particularly sensitive
to such a modification of the 2-body interaction. The best current measurements of
the UFF on Earth have been done by the Eöt-Wash laboratory [65, 66]. This consists
of a measurement of the differential acceleration between two test masses at the
Earth’s surface. Two types of pairs of tests masses have been used: (i) Be versus Ti
and (ii) Be versus Al. For each of these pairs of test masses, a violation of the UFF in
the field of the Earth, in the field of the Sun and in the field generated by the galactic
dark matter distribution has been searched for. It is possible to reinterpret these con-
straints in terms of the coefficients that appear in the Lagrangian from Eq. (8) [196].
Note that in order to probe the UFF at very short distances, the Eöt-Wash group also
performed an experiment where they made a body of Uranium rotate around the test
masses to search for a violation of the UFF in the gravitational field of that body
[64]. This measurement also has the advantage of being sensitive to different linear
combinations of the matter-scalar coupling coefficients. For this particular experi-
ment, the test masses are made of Cu and Pb and the 238 U source is located 10.2 cm
from the test masses. Figure 1 shows the upper limit on the various coefficients that
parametrize the interaction between the scalar field and the standard model fields.
In addition to the Yukawa 2-body interaction, if the scalar field is identified as
DM, it will oscillate and these oscillations will be reflected in the time evolution of
the constants of Nature because of the scalar/matter coupling. This is a signature
of the Einstein Equivalence Principle that can be searched by comparing atomic
transitions frequencies using atomic clocks as described in Sect. 2.2. Recently, this
model has been constrained by three different teams: (i) using Dysprosium (Dy)
atomic transitions at Standford [198], (ii) using the dual Cs/Rb atomic fountain
from SYRTE [197]. These three searches constrain the scalar/matter linear coupling
constant for a range of scalar mass m ϕ between 10−24 and 10−12 eV as shown on
Fig. 1.
The clocks are most powerful for low scalar field masses while UFF experiments
become more interesting at larger masses. The UFF measurements start to deter when
the interaction length scale of the Yukawa interaction
λϕ ∼ 1/m ϕ is of the same order of the distance between the body that generate grav-
itation and the test masses. That is why, although the measurements from MICRO-
SCOPE are one order of magnitude better than the ones on Earth (which is reflected
for low masses), for larger masses, the corresponding upper limit falls quickly because
of the altitude of the satellite while Earth experiments are more powerful to probe
short Yukawa interaction length.
Note that the so-called natural couplings — usually defined as coupling of the
order of unity — are excluded for scalar field masses m ϕ up to ∼10−5 eV for de(1) ,
up to ∼10−4 eV for dm̂(1) − dg(1) and up to 10−5 eV for dm(1)e − dg(1) .
Quadratic coupling: The phenomenology arising in the case of a quadratic coupling
between the scalar field and the standard model fields is quite different from the one
Fig. 1 Upper limit (at 95% confidence level) on the various scalar/matter coupling coefficients in
the case of a linear coupling between matter and the scalar field. The SYRTE Cs/Rb analysis is from
[197], the Dy analysis is presented in [198], the UFF measurement around Earth between Be and
Ti is from [65], the UFF measurement between Cu and Pb in the gravitational field of a 238 U body
is from [64], MICROSCOPE’s result is presented in [72, 73]. The constraints derived from clock
measurements assumed that the scalar field comprises all local DM while the UFF constraints do
not rely on this assumption
that arises in the case of the linear coupling. In particular, it can be shown that no
Yukawa interaction is present in that case, making it a big difference with respect to
the linear interaction [199]. There is still an oscillatory mode that can be identified
as DM but the amplitude of the oscillation will be impacted by the presence of
other bodies and is now dependent on the distance to bodies, leading to a very rich
phenomenology. In particular, the mode of the scalar field that can be identified as
DM writes
m ϕ c2 G MA
ϕ(2) (t, x) = ϕ0 cos t + δ 1 − s (2) , (12)
A
c2 r
where s (2)
A depends on the scalar/matter couplings and on the compactness of the
body A. It has been shown that the case of large couplings or large compactness
336 A. Hees et al.
leads to non-linear interesting behavior. For positive couplings, there is a screening

mechanism which will reduce strongly the amplitude of the oscillations, making this
scalar field harder to detect. On the other hand, large negative values of the couplings
constants enhance strongly the amplitude of the oscillations, making it easier to con-
strain. This phenomenon is closely related to the spontaneous scalarization described
in [200, 201].
There are several observable consequences implied from this scalar field profile.
First of all, regarding violation of the UFF, this coupling will produce two distinct
consequences: (i) it will lead to a regular UFF violation whose η parameter will
depend on the location in the gravitational field and (ii) a violation of the UFF that
will oscillate with time. Both the amplitude of these violations will depend on the
scalar charge s (2)
A of the central body.
Similar consequences will arise for the clocks observables and two effects will be
present: (i) a time-independent but location-dependent effect and (ii) an oscillating
effect whose amplitude is location-dependent. Both the amplitude of these violations
will depend on the scalar charge s (2) A of the central body.
The phenomenology in the quadratic coupling is way richer than in the linear
coupling. Let us mention a couple of interesting features. First of all, for large com-
pactness or coupling, the scalar field will tend to vanish at the surface of the central
body, making it impossible to detect with clocks. This is characterized by the pres-
ence of vertical asymptotes in Fig. 2. This behavior favored clocks experiments in
space with respect to clocks comparison on Earth. On the other hand, violation of
the UFF are sensitive to the gradient of the scalar field and does not suffer from the
same divergences although experiments performed in space are also favored.
On Fig. 2 is presented the current constraints on the quadratic coupling constant
de(2) . Similar figures for the other coupling coefficients can be found in [196].
It is worth mentioning that, contrary to the linear coupling case, values corre-
sponding to so-called “natural couplings” (i.e. di(2) of the order of unity) are either
not constrained at all, or only very marginally constrained for extremely small DM
masses. This leaves a lot of space for so-called “natural” models to exist in the context
of quadratic couplings.
This type of search for Dark Matter is a remarkable example of interdisciplinary
use of experiments built for metrology purposes that end up to be useful in a totally
different context: to search for Dark Matter.
4.1.5 Cosmological Implications of Laboratory Measurements
While the previous section is devoted to searches of Dark Matter with laboratory mea-
surements, these have also some implications at the cosmological level. For example,
a coupling between a scalar field and the electromagnetic part of the Lagrangian such
that the electromagnetic part of the action writes

1 √
SEM = − d 4 x −g 1 − de(1) ϕ Fμν F μν + q p Aμ d x μ , (13)
4μ0
Fig. 2 Upper and lower (MRA) limits (at 95% confidence level) on the scalar/electromagnetic cou-
(2)
pling coefficients de in the case of a quadratic coupling between matter and the scalar field. Similar
figures for the other coupling coefficients can be found in [196]/ The constraints have been derived
using the following measurements: the SYRTE Cs/Rb data from [197], the Dy measurements from
[198], the UFF measurement around Earth between Be and Ti from [65] and the MICROSCOPE’s
result presented in [72]. Note that the dashed line is not an actual constraint but an estimate of the
potential sensitivity that would be obtained by searching for an oscillating violation of the UFF
within MICROSCOPE data
where q p is the electric charge of a particle interacting with the EM field. As men-
tioned in the previous section, this type of coupling has implications on atomic clocks
and on the universality of free fall. In addition to these effects, four cosmological
observables are modified (with respect to GR) and are intimately related to each
other in this class of theories (see [151]): (i) temporal variation of the fine struc-
ture constant, (ii) violation of the distance-duality relation, (iii) modification of the
evolution of the Cosmic Microwave Background (CMB) temperature and (iv) CMB
spectral distortions. It is worth to insist on the fact that the derivation relies only on
the matter part of the action and not on the gravitational part. This means that our
results apply to a very wide class of gravitation theories. In a Friedman–Lemaître–
Robertson–Walker space-time, the expressions of the four observables are given by
the following expressions:
• temporal variation of the fine structure constant. A straightforward identification
in the action leads to
α α(z) − α0
= = de(1) (ϕ(z) − ϕ0 ) (14)
α α0
where z is the redshift and the subscripts 0 refer to z = 0.

• violation of the cosmic distance-duality relation. The optic geometric limit of the
modified Maxwell equations shows that photons propagate on null geodesics but
338 A. Hees et al.
their number is not conserved due to an exchange with the scalar field (see [202]).
Therefore, the expression of the angular diameter distance (D A ) is the same as
in GR but this leads to a modification of the distance-luminosity expression (D L )
(see also [202]) and hence to a violation of the cosmic distance-duality relation:
D L (z) de(1)
η(z) = = 1 + (ϕ(z) − ϕ0 ) . (15)
D A (z)(1 + z)2 2
• modification of the evolution of the CMB temperature. Considering the CMB as

a gaz of photons described by a distribution function solution of a relativistic
Boltzman equation and using the geometric optic approximation of the modified
Maxwell equations leads to a modification of the CMB temperature evolution:

T (z) = T0 (1 + z) 1 + 0.12de(1) (ϕ(z) − ϕ0 ) . (16)
• spectral distortion of the CMB. Using the same approach as the one sketched in the
last item, one can show that the evolution of the CMB radiation leads to deviations
from its black body spectrum. When this deviation is parametrized by a chemical
potential μ, one can show that its expression at current epoch is given by
μ = 0.47de(1) [ϕ(z CMB ) − ϕ0 ] . (17)
To summarize, a coupling between the scalar field and EM implies that the four
observables are intimately linked to each other through the relations

α T (z)
de(1) [ϕ(z) − ϕ0 ] = = 2 [η(z) − 1] = 8.33 −1 , (18a)
α T0 (1 + z)

α(z CMB ) T (z CMB )
μ = 0.47 = 0.94 [η(z CMB ) − 1] = 3.92 − 1 . (18b)
α T0 (1 + z CMB )
These relations hold for all theories whose matter part of the action can be cast in the
form of the action (13). This class of theories is very large and includes all metric
theories. These relations imply also that if a deviation of the EEP is observed around
Earth, it is likely to have cosmological counterparts as well, which would be an
excellent check to confirm any observed deviations. On the other hand, it also shows
that local measurements can have cosmological interpretation.
As an example, several parametrizations of the distance-duality relationship have
been considered in the literature: η(z) = η0 , η(z) = 1 + η1 z, η(z) = 1 + η2 1+z z
,
η(z) = 1 + η3 ln(1 + z) and for the evolution of the CMB temperature T (z) =
(1 + z)1−β . Assuming that the theory of gravitation is described by the multiplicative
coupling introduced in Eq. (13) (which is a large class of theories including GR),
we can use the relations from Eq. (18) to transform observational constraints on one
type of observations into constraints on another type. As an example, we use the con-
straint from optical clocks from [146] (α̇/α = (1.6 ± 2.3) × 10−17 yr−1 to constrain
the cosmic distance-duality parameters: η1 = η2 = η3 = (1 ± 1.4) × 10−7 and the

evolution of the CMB temperature β = −0.3 ± 0.3 (see [151]). These estimations
are a factor 5 better than similar estimations based on measurements of variations of α
on astrophysical distances and order of magnitudes better than direct measurements
of the cosmic distance-duality. This illustrates very clearly the interplay between
small scales measurements and cosmological interpretation.
4.1.6 Violations of the Equivalence Principle and Dark Energy
The action from Eq. (7) with a vanishing potential (i.e. a massless scalar field) has
also been studied in the context of the acceleration of the cosmic expansion (see
e.g. [203, 204] and references therein). In this case, it is useful to work with the
cosmological energy density and pressure associated with the scalar field and it can be
ϕ2
shown that the Dark Energy equation of state is given by (see[204]) wϕ = −1 + 23 ϕ
8πGρ
where ϕ = 3H 2 c2ϕ . This equation allows one to derive an expression that makes the
connection between the temporal evolution of α to cosmological variables [205]

α̇ (1) 3
= −d H ϕ0 (1 + wϕ0 ) . (19)
α 0 e 0
2
This relation is very useful in order to combine data in a global analysis. For example,
[205] uses atomic clocks constraints on α and cosmological observations (Super-
novae Ia and Hubble parameter measurement) in order to constrain the parameters
for this Dark Energy model. In particular, it is shown that one type of observations
only would not constrain this model at all but that it’s really the combination of local
measurements with astrophysical observations that is useful. It is also remarkable
to see in that context the importance of local measurements that have been devel-
oped in the context of metrology and geodesy and to appreciate the impact of such
measurements into a completely different field: cosmology.
4.1.7 The Matter-Gravity Sector of the Standard Model Extension

(SME)
The SME framework is a wide framework that aims at parametrizing all possible
violation of Lorentz symmetry in all sectors of physics. A part of this framework
aims at parametrizing violations of the EEP by modifying the matter action as

SMatter = dλ c −m −(gμν + 2cμν )u μ u ν − (aeff )μ u μ , (20)
where the particle’s worldline tangent is u μ = d x μ /dλ [55]. The fields cμν and (aeff )μ
are new dynamical fields whose kinematic terms are unspecified. In the linearized
340 A. Hees et al.
gravity limit, the motion of test masses depend only on the background value (or
vacuum expectation value) of these fields (the background values will be denoted by
a bar). The coefficients c̄μν and (āeff )μ are species dependent, therefore leading to
the three facets of the EEP.
The c̄μν coefficients have been widely constrained by different laboratory experi-
ments including atomic clock measurements [139–142, 206] and optical cavity [207,
208].
On the opposite, the (āeff )μ components are less constrained. The temporal com-
ponent is currently only constrained through redshift test [141]. Superconducting
gravimeters have recently been used to measure the spatial components [209] at the
level of 10−5−−−6 GeV. The idea is to search for periodic variation in the gravita-
tional acceleration. The main difficulty from such measurements come from tidal
effects that produce similar signatures and that needs to be removed carefully. A
recent reanalysis of LLR observations have provided constraints two orders of mag-
nitude more stringent [210] and are currently the best constraints on these parameters.
Note that a reanalysis of the MICROSCOPE measurements within the context of the
SME would provide an improvement on these constraints [211]. Such an analysis is
currently on-going.
4.1.8 Conclusion
In conclusion, the EEP is an essential part of Einstein’s theory that is shared by a

number of alternative theories of gravitation. All the theories of gravitation satisfying
the EEP are called metric theories. In these theories, it is sufficient to know the space-
time metric (to which matter is minimally coupled) to infer all effects produced
by gravitation (motion of bodies, light propagation, behavior of clocks, ...). It is
interesting to mention that the equivalence principle is a generalization of the fact that
all bodies seem to fall with the same acceleration in a gravitational field. Nevertheless,
as mentioned by T. Damour [143]: “Despite its name, the equivalence principle is
not one of the basic principles of physics. There is nothing taboo about having
an observable violation of the EP. In contrast, one can argue (notably on the basis
of the central message of Einsteins theory of general relativity) that the historical
tendency of physics is to discard any, a priori given, absolute structure (principle of
absence of absolute structures).” It is therefore highly important to pursue our quest
to test the various facets of this principle. Several on-going projects or proposals
aim at improving our current searches for deviations from the EEP in the standard
phenomenological framework used to test the UFF, LLI and LPI.
In all the different kind of tests, experiments to measure geodetic/geophysical
properties like clocks, gravimeter, LLR have an important role regarding tests of
the Einstein Equivalence Principle. It is also amazing that such measurements have
implications in several different fields related to the search for Dark Matter and even
at the cosmological level with the search for Dark Energy.
4.2 Tests of the Einstein Field Equations
There exist several frameworks to test the form of the space-time metric, the second
building block of General Relativity. The two most famous ones are the parametrized
post-Newtonian formalism (PPN) and the fifth force formalism. Nevertheless, these
two formalisms do not cover all the extensions from GR (see e.g. [60, 61]) and it
is useful to analyze experimental results in new extended frameworks. Indeed, even
if observations lie very close to GR when analyzed within the PPN or fifth force
framework, this does not mean that this has to be true in any other framework. In
addition, a framework that would allow a direct comparison between local tests of
GR and cosmological observation has not been developed so far and still needs to
be developed.
In this section, we will review recent constraints obtained for different frameworks
used to probe the form of space-time geometry.
4.2.1 The Parametrized Post-Newtonian Formalism
The PPN formalism is providing an interface between theoretical developments and

data analyses. It is fully described in [43]. In this formalism, a phenomenological
expansion at the level of the space-time metric is performed by introducing 10 dimen-
sionless parameters. From an observational point of view, these parameters can be
constrained regardless of any considerations about the hypothetical underlying the-
ory. From a theoretical point of view, if the post-Newtonian metric of a theory can be
matched to the PPN metric, it can automatically be constrained by all observations
used to constrain the PPN parameters.
In the simplest case, only two PPN parameters are considered and the spacetime
metric for a spherically symmetric configuration is given by

φN φ2 φN
ds 2 = − 1 + 2 2 + 2β N4 + . . . c2 dt 2 + 1 − 2γ 2 + . . . d x 2 , (21)
c c c
where φ N is the Newtonian potential and γ and β are the PPN parameters. These
parameters take the value of 1 in GR and my deviate from 1 in alternative theories.
The PPN formalism was historically the first one developed and therefore was
extensively constrained by several different types of observations. The best constraint
on the γ parameter was obtained by measuring precisely the Shapiro time delay with
the Cassini spacecraft when it was cruising between Jupiter and Saturn in 2003 and
is given by (see [212])
γ − 1 = (2.1 ± 2.3) × 10−5 . (22)
This constraint has been confirmed by measuring the light deflection around the Sun
using Very Long Baseline Interferometry (see [213, 214]), by using radioscience
observations of spacecraft orbiting Mars (see [215]) or Mercury (see [31]).
342 A. Hees et al.
The β PPN parameter is essentially constrained through orbital dynamics. Lunar

Laser Ranging analysis provides a constraint at the level of 104 (see [67, 69, 70]).
Furthermore, planetary ephemerides analysis currently provides constraints at the
level of 10−5 (see [31, 32, 35, 37, 215, 216]). The other PPN parameters are also
very well constrained. An updated list of constraints can be found in [3].
4.2.2 The Fifth Force Formalism
In the fifth force formalism, a Yukawa-type deviation from Newtonian gravity is

considered. The deviation from the Newtonian potential is given by
GM
φ=− 1 + αe−r/λ , (23)
r
where α is the strength of the interaction and λ a characteristic length scale. The idea
is to constrain the couple of parameters (α, λ). An impressive number of experiments
and observations have also been used to constrain this formalism at various scales:
at Solar System scales by using planetary ephemerides analyses [215], Lunar Laser
Ranging observations [217], ranging measurements of LAGEOS satellite around
Earth [50]. At lower distances, several geophysical observations have been used to
constrain a hypothetical fifth force. A short distances, torsion balances provide the
best constraints. For λ ≤ µm, it becomes extremely hard to constrain the strength
of the interaction because of several quantum noise sources (like e.g. the Casimir
effect). The current limit on the Yukawa parameters are summarized in Fig. 3. In
conclusion, the fifth force formalism is very well constrained except at very low and
very large distances where deviations can still be searched for.
4.2.3 A Temporal Evolution of the Gravitational Constant G
As soon as a new field is introduced in addition to the standard space-time metric to

describe the gravitational interaction, it is possible that the gravitational constant G
will become space-time dependent. This is always the case when a coupling between
the scalar field and the scalar curvature is introduced like e.g. in Brans–Dicke theory
[218–220]. This feature also appears in several model of Dark Matter [99] or of Dark
Energy [101, 202].
A very interesting aspect is related to the fact that a temporal evolution of G might
still be present even when screening mechanisms are acting. Screening mechanisms
are non-linear effects that produce a strong reduction of the deviations from General
Relativity in a specific regime. Several mechanisms are know like e.g. the chameleon
mechanism [221–223] where the deviation is reduced in region of high matter density,
the symmetron mechanism [224, 225] where a scalar field is screened through a
symmetry restoring mechanism in region of high density or the Vanshtein mechanism
[226, 227] appearing in massive gravity and in Galileons where deviations are hidden
Fig. 3 95% upper confidence limit on the strength |α| of the Yukawa interaction as a function of
the interaction length scale λ. Several geodesic techniques have provided the best current limit on
the Yukawa interaction. The red curves are limits that are expected from the Gaia mission. The
purple curve is a limit obtained around the supermassive black hole from our Galactic Center [98]
below a certain length scale or the k-mouflage mechanism [228] which is due to
non-linearity in the scalar field kinetic term in the action. While these mechanisms
strongly reduced the deviations from GR in the static limit, allowing these theories
to pass the standard Solar System tests of gravity, it has been shown that in some
cases, the temporal variation of the gravitational constant will not be screened and
will be a perfect signature to search for and to constrain such theories [229]. That
argument shows that constraining an hypothetical Ġ is highly important.
So far, only linear variation of the gravitational constant has been searched for. The
best measurements of Ġ/G are at the level of 10−13 yr−1 from Lunar Laser Ranging
[230] and from planetary ephemerides [32, 215], both these important measurements
relying on techniques developed for geodesy.
4.2.4 The Gravitational Sector of the Standard Model Extension (SME)
As mentioned previously, the SME is a very wide framework aiming at considering

systematically Lorentz violations in any sector of Physics. In the pure gravity sector,
SME can be interpreted as a metric extension of GR. Any action-based model that
breaks local Lorentz symmetry either explicitly or spontaneously can be matched
to a subset of the SME coefficients. Therefore, constraints on SME coefficients can
directly constrain these models. Matches between various toy models and coeffi-
cients in the SME have been achieved for models that produce effective s̄ μν , c̄μν ,
āμ , and other coefficients. This includes vector and tensor field models of sponta-
neous Lorentz-symmetry breaking [54, 55, 231–234], models of quantum gravity
344 A. Hees et al.
[235, 236] and noncommutative quantum field theory [237]. Furthermore, Lorentz
violations may also arise in the context of string field theory models [238].
The SME framework is built by considering an expansion at the level of the action
by introducing a series of terms that violate Lorentz symmetry [54]. At the lowest
order, this action can be written as
c3
L = LEH + h μν s̄ αβ Gαμνβ + . . . , (24)
32πG
where LEH is the standard Einstein–Hilbert term, Gαμνβ is the double dual of the Ein-
stein tensor linearized in h μν (with h μν = gμν − ημν ). The Lorentz-violating effects
in this expression are controlled by the 9 independent coefficients in the traceless and
dimensionless s̄ μν [54]. These coefficients are treated as constants in asymptotically
flat cartesian coordinates. The ellipses represent additional terms in a series includ-
ing terms that break CPT symmetry for gravity; such terms are detailed elsewhere
[239–242] and are part of the so-called nonminimal SME expansion. Note that the
process by which one arrives at the effective quadratic Lagrangian (24) is consistent
with the assumption of the spontaneous breaking of local Lorentz symmetry.
Several measurements have been used to search for a breaking of the Lorentz
invariance and to constrain the SME parameters. Amongst them, measurements ded-
icated to geodesy and geophysics are the most important ones like e.g.: atomic inter-
ferometry [243, 244], gravimetry [209, 245], planetary ephemerides [246], very long
baseline interferometry [247], lunar laser ranging [210, 248, 249]. Table 1 summa-
rizes the order of magnitude of the constraints on the SME coefficients from the pure
gravity sector that have been obtained using the techniques mentioned in Sect. 2.
More informations about these constraints and about other constraints can be found
in the reviews [56, 57].
Table 1 Order of magnitude of the constraints on the pure gravitational SME coefficients derived
from measurements that are related to geodesy and geophysics
Experiments s̄ T T s̄ T J s̄ J K
Atom interferometry [244] – ∼10−5 ∼ 10−9
Gravimeters [209] – ∼10−7 ∼ 10−9
Gravimeters [245] – – ∼10−10
Planetary ephemerides [246] – ∼10−8 ∼ 10−10
Lunar laser ranging [249] – ∼10−9 ∼ 10−11
VLBI [247] 10−4 – –
4.2.5 Modified Newtonian Dynamics (MOND)
The MOND phenomenology was introduced in 1983 by M. Milgrom [250–252]

in order to explain observations at the galactic scale (in particular galactic rotation
curves) without the introduction of Dark Matter. The main idea behind the MOND
phenomenology is to replace the standard Newton acceleration g N valid in “high”
√
gravitational field by a0 g N for regions of the Universe where the gravitational
field is very low. More precisely, the MOND phenomenology depends on a MOND
interpolating function ν and on a MOND acceleration scale a0 . The interpolating
function is making the transition between the Newtonian regime and the MONDian
regime appearing in low gravitational fields. While developed initially as a purely
phenomenological Newtonian model, the MOND phenomenology has later been
extended as relativistic theories. A review of different relativistic MOND theories
can be found in [253] Bruneton and Esposito-Farse or in [1]. A review of the different
interpolating functions used in the literature can also be found in [1].
Recently, it has been shown that contrarily to what was previously expected, the
MOND phenomenology would produce detectable effects in the Solar System. This
effect is due to the non-linearity of the MOND phenomenology which implies that
the external galactic gravitational field will play a role in this theory. This External
Field Effect was first discovered by [58] and was studied later in [59]. The main
contribution from this External Field Effect in the Solar System takes the form of a
quadrupolar deviation from the Newton potential

GM Q2 i j 1
φN = − − x x ei e j − δi j , (25)
r 2 3
where the unit vector ei points toward the galactic center and Q 2 is a parameter
that depends directly on the MOND interpolating function ν and on the MOND
acceleration scale a0 . Values of this parameter for different interpolating functions
are given in [58, 59, 254].
The interesting point is that this modification of gravitation does not enter the
fifth force or the PPN formalisms and it can be used efficiently to detect or constrain
the MOND phenomenology. A recent analysis based on the Saturn ephemeris (see
Sect. 2.5) obtained mainly from the radio tracking data of the Cassini spacecraft
(between 2004 and 2013) led to a constraint on Q 2 given by [26]
Q 2 = (3 ± 3) × 10−27 s−2 . (26)
This constraint on MOND theory is the best currently available on Solar System
scales. When combined with galactic rotation curves observations, this analysis nar-
rows down the possible interpolating functions allowed to only one class of function
as shown in [254].
As such, this local constraint does not constraint the MOND theory. What is
interesting is to combine this local constraint with astrophysical measurements of
galactic rotation curves. Indeed, several classes of interpolating function ν have
346 A. Hees et al.
been used to explain the behavior of galactic rotation curves. Using observations
of 27 galactic rotation curves (using data from [255]), fitting them with different
MOND phenomenology and combining this analysis from the Solar System leads to
a very stringent constraint on the MOND phenomenology. Indeed, all but one class of
interpolating function is excluded by this combined analysis [254]. This result shows
the power of combining multi-scales measurements and once again how geodesic
measurements can help to improve our understanding of fundamental physics, even
at a totally different scale in our Universe.
4.2.6 Conclusion
In conclusion, any modification of GR will at least lead to a modification of the

form of the space-time metric. This will be induced by new fields in addition to the
regular space-time metric, by the consideration of higher dimensions, the inclusion
of derivatives of the space-time metric of the order higher than 2, …Measuring the
curvature of space-time around different body in different location in space and time
in our Universe is therefore one of the best way to search for new physics beyond GR.
Two formalisms have been widely used so far to search for deviations from GR at
the level of the space-time metric: the PPN formalism and the fifth force formalism.
In this section, we have shown that geodesic measurements are very important to
constrain these formalisms. In addition, not every alternative theory of gravitation
enters these two frameworks. In the previous sections, we give examples of other
modifications from GR that do not enter the PPN or fifth force frameworks. It is
important to reanalyze existing measurements in the context of these GR extensions
and in that context, geodetic measurements are also very important.
5 Conclusion and Outlook
In this communication, we highlight how experiments and measurements that have

been developed in the context of geodesy and geophysics have found very interesting
applications in fundamental physics. These two fields are very highly active field of
research that thrives on technological developments, in particular of clocks, long
distance radio and optical links, and inertial sensors (accelerometers, gyroscopes). It
is impossible to describe all the constraints on the modified theory of gravitation that
have been derived using such measurements and in this paper, we focused only on
several important and/or recent ones. This shows the vitality of the field and many
more exciting results are expected in the near and more distant future, which will
hopefully pave the way towards the new physics beyond general relativity and the
standard model of particle physics. And it is likely that measurements or experiments
developed in the context of geodesy will continue to play a major role in that process.
References
1. B. Famaey, S.S. McGaugh, Modified Newtonian dynamics (MOND): observational phe-

nomenology and relativistic. Living Rev Relativ 15, 10 (2012)
2. T. Clifton, P.G. Ferreira, A. Padilla, C. Skordis, Modified gravity and cosmology. Phys. Rep.
513, 1–189 (2012)
3. C.M. Will, The confrontation between general relativity and experiment. Living Rev. Relativ.
17, 4 (2014)
4. G. Lion, I. Panet, P. Wolf, C. Guerlin, S. Bize, P. Delva, Determination of a high spatial res-
olution geopotential model using atomic clock comparisons. J. Geodesy 91, 597–611 (2017)
5. Microg Lacoste. FG5-X and FGL absolute gravity meters
6. μ QUANS. Absolute quantum gravimeter
7. C.W. Chou, D.B. Hume, J.C.J. Koelemeij, D.J. Wineland, T. Rosenband, Frequency compar-
ison of two high-accuracy Al+ optical clocks. Phys. Rev. Lett. 104(7), 070802 (2010)
8. K. Beloy, N. Hinkley, N.B. Phillips, J.A. Sherman, M. Schioppo, J. Lehman, A. Feldman,
L.M. Hanssen, C.W. Oates, A.D. Ludlow, Atomic clock with 1×10−18 room-temperature
blackbody stark uncertainty. Phys. Rev. Lett. 113(26), 260801 (2014)
9. T.L. Nicholson, S.L. Campbell, R.B. Hutson, G.E. Marti, B.J. Bloom, R.L. McNally, W.
Zhang, M.D. Barrett, M.S. Safronova, G.F. Strouse, W.L. Tew, J. Ye, Systematic evaluation
of an atomic clock at 2×10−18 total uncertainty. Nat. Commun. 6, 6896 (2015)
10. I. Ushijima, M. Takamoto, M. Das, T. Ohkubo, H. Katori, Cryogenic optical lattice clocks.
Nat. Photon. 9, 185–189 (2015)
11. N. Huntemann, C. Sanner, B. Lipphardt, C. Tamm, E. Peik, Single-ion atomic clock with
3×10−18 systematic uncertainty. Phys. Rev. Lett. 116(6), 063001 (2016)
12. Peter Wolf. Viewpoint: Next generation clock networks. Physical Review: Viewpoint, Physics
9, 51, May 11, 2016 2016
13. H. Denker, L. Timmen, C. Voigt, S. Weyers, E. Peik, H.S. Margolis, P. Delva, P. Wolf, G.
Petit, Geodetic methods to determine the relativistic redshift at the level of 10 (-18) - 18 in
the context of international timescales: a review and practical results. J. Geodesy 92, 487–516
(2018)
14. T.E. Mehlstäubler, G. Grosche, C. Lisdat, P.O. Schmidt, H. Denker, Atomic clocks for geodesy.
Rep. Prog. Phys. 81(6), 064401 (2018)
15. I. Ciufolini, E. Pavlis, F. Chieppa, E. Fernandes-Vieira, J. Perez-Mercader, Test of general
relativity and measurement of the lense-thirring effect with two earth satellites. Science 279,
2100 (1998)
17. I. Ciufolini, A. Paolozzi, E.C. Pavlis, R. Koenig, J. Ries, V. Gurzadyan, R. Matzner, R. Penrose,
G. Sindoni, C. Paris, H. Khachatryan, S. Mirzoyan, A test of general relativity using the
LARES and LAGEOS satellites and a GRACE Earth gravity model. Measurement of Earth’s
dragging of inertial frames. Eur. Phys. J. C 76, 120 (2016)
18. J.O. Dickey, P.L. Bender, J.E. Faller, X.X. Newhall, R.L. Ricklefs, J.G. Ries, P.J. Shelus, C.
Veillet, A.L. Whipple, J.R. Wiant, J.G. Williams, C.F. Yoder, Lunar laser ranging: a continuing
legacy of the apollo program. Science 265, 482–490 (1994)
19. N.P. Pitjev, E.V. Pitjeva, Constraints on dark matter in the solar system. Astron. Lett. 39,
141–149 (2013)
20. E.M. Standish, The JPL planetary ephemerides. Celest. Mech. 26, 181–186 (1982)
21. X.X. Newhall, E.M. Standish, J.G. Williams, DE 102 - A numerically integrated ephemeris
of the moon and planets spanning forty-four centuries. A&A 125, 150–167 (1983)
22. E.M. Standish Jr., The observational basis for JPL’s DE 200, the planetary ephemerides of
the Astronomical Almanac. A&A 233, 252–271 (1990)
23. E.M. Standish, Testing alternate gravitational theories, in IAU Symposium, ed. by S.A. Klioner,
P.K. Seidelmann, M.H. Soffel, vol. 261 (2010), pp. 179–182
348 A. Hees et al.
24. E.M. Standish, J.G. Williams, Orbital ephemerides of the Sun, Moon, and Planets, in Explana-
tory Supplement to the Astronomical Almanac, ed. by S.E. Urban, P.K. Seidelmann, 3rd edn.
(Univeristy Science Books, 2012), pp. 305–346
25. W.M. Folkner, J.G. Williams, D.H. Boggs, R. Park, P. Kuchynka, The planetary and lunar
ephemeris DE 430 and DE431. IPN Prog. Rep. 42(196) (2014)
26. A. Hees, W.M. Folkner, R.A. Jacobson, R.S. Park, Constraints on modified Newtonian dynam-
ics theories from radio tracking data of the Cassini spacecraft. Phys. Rev. D 89(10), 102002
(2014)
27. A. Fienga, H. Manche, J. Laskar, M. Gastineau, INPOP06: a new numerical planetary
ephemeris. A&A 477, 315–327 (2008)
28. A. Fienga, J. Laskar, T. Morley, H. Manche, P. Kuchynka, C. Le Poncin-Lafitte, F. Budnik, M.
Gastineau, L. Somenzi, INPOP08, a 4-D planetary ephemeris: from asteroid and time-scale
computations to ESA Mars Express and Venus Express contributions. A&A 507, 1675–1686
(2009)
29. A. Fienga, J. Laskar, P. Kuchynka, C. Le Poncin-Lafitte, H. Manche, M. Gastineau, Gravity
tests with INPOP planetary ephemerides, in IAU Symposium, ed. by S.A. Klioner, P.K. Sei-
delmann, M.H. Soffel, vol. 261 (2010), pp. 159–169
30. A. Fienga, J. Laskar, P. Kuchynka, H. Manche, G. Desvignes, M. Gastineau, I. Cognard, G.
Theureau, The INPOP10a planetary ephemeris and its applications in fundamental physics.
Celest. Mech. Dyn. Astrono. 111(3), 363–385 (2011)
31. A.K. Verma, A. Fienga, J. Laskar, H. Manche, M. Gastineau, Use of MESSENGER radio-
science data to improve planetary ephemeris and to test general relativity. A&A 561, A115
(2014)
32. A. Fienga, J. Laskar, P. Exertier, H. Manche, M. Gastineau, Numerical estimation of the
sensitivity of INPOP planetary ephemerides to general relativity parameters. Celest. Mech.
Dyn. Astron. 123, 325–349 (2015)
33. E.V. Pitjeva, High-precision ephemerides of planets EPM and determination of some astro-
nomical constants. Solar Syst. Res. 39, 176–186 (2005)
34. E.V. Pitjeva, EPM ephemerides and relativity, in Proceedings of IAU Symposium 261, ed. by
S.A. Klioner, P.K. Seidelmann, M.H. Soffel (2010), pp. 170–178
35. E.V. Pitjeva, N.P. Pitjev, Relativistic effects and dark matter in the Solar system from obser-
vations of planets and spacecraft. MNRAS 432, 3431–3437 (2013)
36. E.V. Pitjeva, Updated IAA RAS planetary ephemerides-EPM2011 and their use in scientific
research. Solar Syst. Res. 47, 386–402 (2013)
37. E.V. Pitjeva, N.P. Pitjev, Development of planetary ephemerides EPM and their applications.
Celest. Mech. Dyn. Astron. 119, 237–256 (2014)
38. A.L. Fey, D. Gordon, C.S. Jacobs, C. Ma, R.A. Gaume, E.F. Arias, G. Bianco, D.A. Boboltz,
S. Böckmann, S. Bolotin, P. Charlot, A. Collioud, G. Engelhardt, J. Gipson, A.-M. Gontier, R.
Heinkelmann, S. Kurdubov, S. Lambert, S. Lytvyn, D.S. MacMillan, Z. Malkin, A. Nothnagel,
R. Ojha, E. Skurikhina, J. Sokolova, J. Souchay, O.J. Sovers, V. Tesmer, O. Titov, G. Wang,
V. Zharov, The second realization of the international celestial reference frame by very long
baseline interferometry. AJ 150, 58 (2015)
39. M. Soffel, S.A. Klioner, G. Petit, P. Wolf, S.M. Kopeikin, P. Bretagnon, V.A. Brumberg,
N. Capitaine, T. Damour, T. Fukushima, B. Guinot, T.-Y. Huang, L. Lindegren, C. Ma, K.
Nordtvedt, J.C. Ries, P.K. Seidelmann, D. Vokrouhlický, C.M. Will, C. Xu, The IAU 2000
resolutions for astrometry, celestial mechanics, and metrology in the relativistic framework:
explanatory supplement. Astron. J. 126, 2687–2706 (2003)
40. A. Einstein, Über den Einfluß der Schwerkraft auf die Ausbreitung des Lichtes. Annalen der
Physik 340, 898–908 (1911). Traduction anglaise dans [256]
41. K.S. Thorne, C.M. Will, Theoretical frameworks for testing relativistic gravity I. Foundations.
ApJ 163, 595 (1971)
42. C.W. Misner, K.S. Thorne, J.A. Wheeler, Gravitation. Physics Series (W. H. Freeman, 1973)
43. C.M. Will, Theory and Experiment in Gravitational Physics (1993)
44. A. Einstein, Die Grundlage der allgemeinen Relativitätstheorie. Annalen der Physik 354,
769–822 (1916). Traduction anglaise dans [256]
45. D. Lovelock, The Einstein tensor and its generalizations. J. Math. Phys. 12, 498–501 (1971)
46. D. Lovelock, The Four-Dimensionality of Space and the Einstein Tensor. J. Math. Phys. 13,
874–876 (1972)
47. E. Fischbach, D. Sudarsky, A. Szafer, C. Talmadge, S.H. Aronson, Reanalysis of the Eotvos
experiment. Phys. Rev. Lett. 56, 3–6 (1986)
48. C. Talmadge, J.-P. Berthias, R.W. Hellings, E.M. Standish, Model-independent constraints on
possible modifications of Newtonian gravity. Phys. Rev. Lett. 61, 1159–1162 (1988)
49. E. Fischbach, C. Talmadge, Six years of the fifth force. Nature 356, 207–215 (1992)
50. E. Fischbach, C.L. Talmadge, The Search for Non-Newtonian Gravity, Aip-Press Series
(Springer, Berlin, 1999)
51. E.G. Adelberger, J.H. Gundlach, B.R. Heckel, S. Hoedl, S. Schlamminger, Torsion balance
experiments: a low-energy frontier of particle physics. Prog. Part. Nucl. Phys. 62, 102–134
(2009)
52. A. Hees, B. Lamine, S. Reynaud, M.-T. Jaekel, C. Le Poncin-Lafitte, V. Lainey, A. Füzfa, J.-M.
Courty, V. Dehant, P. Wolf, Radioscience simulations in general relativity and in alternative
theories of gravity. Class. Quantum Gravity 29(23), 235027 (2012)
53. A. Hees, W. Folkner, R. Jacobson, R. Park, B. Lamine, C. Le Poncin-Lafitte, P. Wolf, Tests
of gravitation at Solar System scales beyond the PPN formalism, in Journées 2013 Systèmes
de référence spatio-temporels ed. by N. Capitaine (2014), pp. 241–244
54. Q.G. Bailey, V.A. Kostelecký, Signals for Lorentz violation in post-Newtonian gravity. Phys.
Rev. D 74(4), 045001 (2006)
55. V.A. Kostelecký, J.D. Tasson, Matter-gravity couplings and Lorentz violation. Phys. Rev. D
83(1), 016013 (2011)
56. A. Hees, Q. Bailey, A. Bourgoin, H. Pihan-Le Bars, C. Guerlin, C. Le Poncin-Lafitte, Tests
of Lorentz symmetry in the gravitational sector. Universe 2, 30 (2016)
57. J.D. Tasson, The standard-model extension and gravitational tests. Symmetry 8, 111 (2016)
58. M. Milgrom, MOND effects in the inner Solar system. MNRAS 399, 474–486 (2009)
59. L. Blanchet, J. Novak, External field effect of modified Newtonian dynamics in the Solar
system. MNRAS 412, 2530–2542 (2011)
60. M.-T. Jaekel, S. Reynaud, Post-Einsteinian tests of linearized gravitation. Class. Quantum
Gravity 22, 2135–2157 (2005)
61. M.-T. Jaekel, S. Reynaud, Post-Einsteinian tests of gravitation. Class. Quantum Gravity 23,
777–798 (2006)
62. A. Avilez-Lopez, A. Padilla, P.M. Saffin, C. Skordis, The parametrized post-Newtonian-
vainshteinian formalism. J. Cosmol. Astropart. Phys. 6, 044 (2015)
63. M. Hohmann, Parametrized post-Newtonian limit of Horndeski’s gravity theory. Phys. Rev.
D 92(6), 064019 (2015)
64. G.L. Smith, C.D. Hoyle, J.H. Gundlach, E.G. Adelberger, B.R. Heckel, H.E. Swanson, Short-
range tests of the equivalence principle. Phys. Rev. D 61(2), 022001 (1999)
65. S. Schlamminger, K.-Y. Choi, T.A. Wagner, J.H. Gundlach, E.G. Adelberger, Test of the
equivalence principle using a rotating torsion balance. Phys. Rev. Lett. 100(4), 041101 (2008)
66. T.A. Wagner, S. Schlamminger, J.H. Gundlach, E.G. Adelberger, Torsion-balance tests of the
weak equivalence principle. Class. Quantum Gravity 29(18), 184002 (2012)
67. V. Viswanathan, A. Fienga, O. Minazzoli, L. Bernus, J. Laskar, M. Gastineau, The new lunar
ephemeris INPOP17a and its application to fundamental physics. MNRAS 476, 1877–1888
(2018)
68. K. Nordtvedt, Testing relativity with laser ranging to the Moon. Phys. Rev. 170, 1186–1187
(1968)
69. J.G. Williams, S.G. Turyshev, D.H. Boggs, Lunar laser ranging tests of the equivalence prin-
ciple with the Earth and Moon. Inte. J. Mod. Phys. D 18, 1129–1175 (2009)
70. J.G. Williams, S.G. Turyshev, D. Boggs, Lunar laser ranging tests of the equivalence principle.
350 A. Hees et al.
71. C. Courde, J.M. Torre, E. Samain, G. Martinot-Lagarde, M. Aimar, D. Albanese, P. Exertier,

A. Fienga, H. Mariey, G. Metris, H. Viot, V. Viswanathan, Lunar laser ranging in infrared at
the Grasse laser station. A&A 602, A90 (2017)
72. P. Touboul, G. Métris, M. Rodrigues, Y. André, Q. Baghi, J. Bergé, D. Boulanger, S. Bremer,
P. Carle, R. Chhun, B. Christophe, V. Cipolla, T. Damour, P. Danto, H. Dittus, P. Fayet, B.
Foulon, C. Gageant, P.-Y. Guidotti, D. Hagedorn, E. Hardy, P.-A. Huynh, H. Inchauspe, P.
Kayser, S. Lala, C. Lämmerzahl, V. Lebat, P. Leseur, F. Liorzou, M. List, F. Löffler, I. Panet,
B. Pouilloux, P. Prieur, A. Rebray, S. Reynaud, B. Rievers, A. Robert, H. Selig, L. Serron,
T. Sumner, N. Tanguy, P. Visser, MICROSCOPE mission: first results of a space test of the
equivalence principle. Phys. Rev. Lett. 119(23), 231101 (2017)
73. J. Bergé, P. Brax, G. Métris, M. Pernot-Borràs, P. Touboul, J.-P. Uzan, MICROSCOPE mission:
first constraints on the violation of the weak equivalence principle by a light scalar dilaton.
Phys. Rev. Lett. 120(14), 141101 (2018)
74. P. Fayet. MICROSCOPE limits for new long-range forces and implications for unified theories.
ArXiv e-prints (2017)
75. A. Peters, K.Y. Chung, S. Chu, Measurement of gravitational acceleration by dropping atoms.
Nature 400, 849–852 (1999)
76. A. Peters, K.Y. Chung, S. Chu, High-precision gravity measurements using atom interferom-
etry. Metrologia 38, 25–61 (2001)
77. S. Merlet, Q. Bodart, N. Malossi, A. Landragin, F. Pereira Dos Santos, O. Gitlein, L. Timmen,
SHORT COMMUNICATION: comparison between two mobile absolute gravimeters: optical
versus atomic interferometers. Metrologia 47, L9–L11 (2010)
78. L. Zhou, S. Long, B. Tang, X. Chen, F. Gao, W. Peng, W. Duan, J. Zhong, Z. Xiong, J.
Wang, Y. Zhang, M. Zhan, Test of equivalence principle at 1 0−8 level by a dual-species
double-diffraction Raman atom interferometer. Phys. Rev. Lett. 115(1), 013004 (2015)
79. F.W. Hehl, P. von der Heyde, G.D. Kerlick, J.M. Nester, General relativity with spin and
torsion: foundations and prospects. Rev. Mod. Phys. 48, 393–416 (1976)
80. A. Peres, Test of equivalence principle for particles with spin. Phys. Rev. D 18, 2739–2740
(1978)
81. B. Mashhoon, Gravitational couplings of intrinsic spin. Class. Quantum Gravity 17, 2399–
2409 (2000)
82. Y.N. Obukhov, Spin, gravity, and inertia. Phys. Rev. Lett. 86, 192–195 (2001)
83. W.-T. Ni, Searches for the role of spin and polarization in gravity. Rep. Prog. Phys. 73(5),
056901 (2010)
84. W.-T. Ni, Searches for the role of spin and polarization in gravity: a five-year update, in
International Journal of Modern Physics Conference Series, vol. 40 (2016), pp. 1660010–
146
85. M.G. Tarallo, T. Mazzoni, N. Poli, D.V. Sutyrin, X. Zhang, G.M. Tino, Test of einstein
equivalence principle for 0-spin and half-integer-spin atoms: search for spin-gravity coupling
effects. Phys. Rev. Lett. 113(2), 023005 (2014)
86. S. Aghion, O. Ahlén, C. Amsler, A. Ariga, T. Ariga, A. S. Belov, G. Bonomi, P. Bräu-
nig, J. Bremer, R. S. Brusa, L. Cabaret, C. Canali, R. Caravita, F. Castelli, G. Cerchiari,
S. Cialdi, D. Comparat, G. Consolati, J. H. Derking, S. Di Domizio, L. Di Noto, M. Doser,
A. Dudarev, A. Ereditato, R. Ferragut, A. Fontana, P. Genova, M. Giammarchi, A. Gligorova,
S. N. Gninenko, S. Haider, J. Harasimovicz, S. D. Hogan, T. Huse, E. Jordan, L. V. Jørgensen,
T. Kaltenbacher, J. Kawada, A. Kellerbauer, M. Kimura, A. Knecht, D. Krasnický, V. Lago-
marsino, A. Magnani, S. Mariazzi, V. A. Matveev, F. Moia, G. Nebbia, P. Nédélec, M. K.
Oberthaler, N. Pacifico, V. Petráček, C. Pistillo, F. Prelz, M. Prevedelli, C. Regenfus, C. Ric-
cardi, O. Røhne, A. Rotondi, H. Sandaker, P. Scampoli, A. Sosa, J. Storey, M. A. Subieta
Vasquez, M. Špaček, G. Testera, D. Trezzi, R. Vaccarone, C. P. Welsch, and S. Zavatarelli.
Prospects for measuring the gravitational free-fall of antihydrogen with emulsion detectors.
J. Instrum. 8, 8013P (2013)
87. P. Perez, Y. Sacquin, The GBAR experiment: gravitational behaviour of antihydrogen at rest.
88. L. Hui, A. Nicolis, Proposal for an observational test of the vainshtein mechanism. Phys. Rev.
Lett. 109(5), 051304 (2012)
89. J. Sakstein, B. Jain, J.S. Heyl, L. Hui, Tests of gravity theories using supermassive black holes.
ApJl 844, L14 (2017)
90. J.A. Frieman, B.-A. Gradwohl, Dark matter and the equivalence principle. Phys. Rev. Lett.
67, 2926–2929 (1991)
91. B.-A. Gradwohl, J.A. Frieman, Dark matter, long-range forces, and large-scale structure. ApJ
398, 407–424 (1992)
92. S.M. Carroll, S. Mantry, M.J. Ramsey-Musolf, C.W. Stubbs, Dark-matter-induced violation
of the weak equivalence principle. Phys. Rev. Lett. 103(1), 011301 (2009)
93. S.M. Carroll, S. Mantry, M.J. Ramsey-Musolf, Implications of a scalar dark force for terrestrial
experiments. Phys. Rev. D 81(6), 063507 (2010)
94. M. Kesden, M. Kamionkowski, Tidal tails test the equivalence principle in the dark-matter
sector. Phys. Rev. D 74(8), 083007 (2006)
95. M. Kesden, M. Kamionkowski, Galilean equivalence for galactic dark matter. Phys. Rev. Lett.
97(13), 131303 (2006)
96. C.W. Stubbs, Experimental limits on any long range nongravitational interaction between
dark matter and ordinary matter. Phys. Rev. Lett. 70, 119–122 (1993)
97. Y. Bai, J. Salvado, B.A. Stefanek, Cosmological constraints on the gravitational interactions
of matter and dark matter. J. Cosmol. Astropart. Phys. 10, 029 (2015)
98. A. Hees, T. Do, A.M. Ghez, G.D. Martinez, S. Naoz, E.E. Becklin, A. Boehle, S. Chappell,
D. Chu, A. Dehghanfar, K. Kosmo, J.R. Lu, K. Matthews, M.R. Morris, S. Sakai, R. Schödel,
G. Witzel, Testing general relativity with stellar orbits around the supermassive black hole in
our galactic center. Phys. Rev. Lett. 118(21), 211101 (2017)
99. T. Damour, G.W. Gibbons, C. Gundlach, Dark matter, time-varying G, and a dilaton field.
Phys. Rev. Lett. 64, 123–126 (1990)
100. J.-M. Alimi, A. Füzfa, Is dark energy abnormally weighting? Int. J. Mod. Phys. D 16, 2587–
2592 (2007)
101. J.-M. Alimi, A. Füzfa, The abnormally weighting energy hypothesis: the missing link between
dark matter and dark energy. J. Cosmology Astropart. Phys. 9, 14 (2008)
102. A. Füzfa, J.-M. Alimi, Toward a unified description of dark energy and dark matter from the
abnormally weighting energy hypothesis. Phys. Rev. D 75(12), 123007 (2007)
103. N. Mohapi, A. Hees, J. Larena, Test of the equivalence principle in the dark sector on galactic
scales. J. Cosmol. Astropart. Phys. 3, 032 (2016)
104. D. Mattingly, Modern tests of lorentz invariance. Living Rev. Relativ. 8, 5 (2005)
105. V.W. Hughes, H.G. Robinson, V. Beltran-Lopez, Upper limit for the anisotropy of inertial
mass from nuclear resonance experiments. Phys. Rev. Lett. 4, 342–344 (1960)
106. R.W.P. Drever, A search for anisotropy of inertial mass using a free precession technique.
Philos. Mag. 6, 683–687 (1961)
107. T.E. Chupp, R.J. Hoare, R.A. Loveman, E.R. Oteiza, J.M. Richardson, M.E. Wagshul, A.K.
Thompson, Results of a new test of local Lorentz invariance: a search for mass anisotropy in
21 Ne. Phys. Rev. Lett. 63, 1541–1545 (1989)
108. S.K. Lamoreaux, J.P. Jacobs, B.R. Heckel, F.J. Raab, E.N. Fortson, New limits on spatial
anisotropy from optically-pumped sup201Hg and 199 Hg. Phys. Rev. Lett. 57, 3125–3128
(1986)
109. J.D. Prestage, J.J. Bollinger, W.M. Itano, D.J. Wineland, Limits for spatial anisotropy by use
of nuclear-spin-polarized Be-9(+) ions. Phys. Rev. Lett. 54, 2387–2390 (1985)
110. M.P. Haugan, C.M. Will, Modern tests of special relativity. Phys. Today 40, 69–86 (1987)
111. A. Brillet, J.L. Hall, Improved laser test of the isotropy of space. Phys. Rev. Lett. 42, 549–552
(1979)
112. P.L. Stanwix, M.E. Tobar, P. Wolf, C.R. Locke, E.N. Ivanov, Improved test of Lorentz invari-
ance in electrodynamics using rotating cryogenic sapphire oscillators. Phys. Rev. D 74(8),
081101 (2006)
352 A. Hees et al.
113. P. Wolf, S. Bize, A. Clairon, A.N. Luiten, G. Santarelli, M.E. Tobar, Tests of Lorentz invariance
using a microwave resonator. Phys. Rev. Lett. 90(6), 060402 (2003)
114. H.P. Robertson, Postulate versus observation in the special theory of relativity. Rev. Mod.
Phys. 21, 378–382 (1949)
115. R. Mansouri, R.U. Sexl, A test theory of special relativity. I - simultaneity and clock synchro-
nization. II - first order tests. Gen. Relativ. Gravitat. 8, 497–513 (1977)
116. R. Mansouri, R.U. Sexl. A test theory of special relativity: II. first order tests. Gen. Relativ.
Gravit. 8, 515–524 (1977)
117. R. Mansouri, R.U. Sexl, A test theory of special relativity: III Second-order tests. Gen. Relativ.
Gravit. 8, 809–814 (1977)
118. S. Reinhardt, G. Saathoff, H. Buhr, L.A. Carlson, A. Wolf, D. Schwalm, S. Karpuk, C. Novotny,
G. Huber, M. Zimmermann, R. Holzwarth, T. Udem, T.W. Hänsch, G. Gwinner, Test of
relativistic time dilation with fast optical atomic clocks at different velocities. Nat. Phys. 3,
861–864 (2007)
119. B. Botermann, D. Bing, C. Geppert, G. Gwinner, T.W. Hänsch, G. Huber, S. Karpuk, A.
Krieger, T. Kühl, W. Nörtershäuser, C. Novotny, S. Reinhardt, R. Sánchez, D. Schwalm,
T. Stöhlker, A. Wolf, G. Saathoff, Test of time dilation using stored Li+ ions as clocks at
relativistic speed. Phys. Rev. Lett. 113(12), 120405 (2014)
120. C.M. Will, Clock synchronization and isotropy of the one-way speed of light. Phys. Rev. D
45, 403–411 (1992)
121. P. Wolf, G. Petit, Satellite test of special relativity using the global positioning system. Phys.
Rev. A 56, 4405–4409 (1997)
122. J. Lodewyck, S. Bilicki, E. Bookjans, J.-L. Robyr, C. Shi, G. Vallet, R. Le Targat, D. Nicolodi,
Y. Le Coq, J. Guéna, M. Abgrall, P. Rosenbusch, S. Bize, Optical to microwave clock frequency
ratios with a nearly continuous strontium optical lattice clock. Metrologia 53, 1123 (2016)
123. R. Le Targat, L. Lorini, Y. Le Coq, M. Zawada, J. Guéna, M. Abgrall, M. Gurov, P. Rosenbusch,
D.G. Rovera, B. Nagórny, R. Gartman, P.G. Westergaard, M.E. Tobar, M. Lours, G. Santarelli,
A. Clairon, S. Bize, P. Laurent, P. Lemonde, J. Lodewyck, Experimental realization of an
optical second with strontium lattice clocks. Nat. Commun. 4, 2109 (2013)
124. S. Falke, N. Lemke, C. Grebing, B. Lipphardt, S. Weyers, V. Gerginov, N. Huntemann, C.
Hagemann, A. Al-Masoudi, S. Häfner, S. Vogt, U. Sterr, C. Lisdat, A strontium lattice clock
with 3×10−17 inaccuracy and its frequency. New J. Phys. 16(7), 073023 (2014)
125. C. Grebing, A. Al-Masoudi, D. Sören, H. Sebastian, G. Vladislav, W. Stefan, L. Burghard, R.
Fritz, S. Uwe, L. Christian, Realization of a timescale with an accurate optical lattice clock.
Optica 3(6), 563–569 (2016)
126. I.R. Hill, R. Hobson, W. Bowden, E.M. Bridge, S. Donnellan, E.A. Curtis, P. Gill, A low
maintenance Sr optical lattice clock. In Journal of Physics Conference Series, volume 723 of
Journal of Physics Conference Series, page 012019, June 2016
127. C. Lisdat, G. Grosche, N. Quintin, C. Shi, S.M.F. Raupach, C. Grebing, D. Nicolodi, F. Stefani,
A. Al-Masoudi, S. Dörscher, S. Häfner, J.-L. Robyr, N. Chiodo, S. Bilicki, E. Bookjans, A.
Koczwara, S. Koke, A. Kuhl, F. Wiotte, F. Meynadier, E. Camisard, M. Abgrall, M. Lours, T.
Legero, H. Schnatz, U. Sterr, H. Denker, C. Chardonnet, Y. Le Coq, G. Santarelli, A. Amy-
Klein, R. Le Targat, J. Lodewyck, O. Lopez, P.-E. Pottie, A clock network for geodesy and
fundamental science. Nat. Commun. 7, 12443 (2016)
128. P.A. Williams, W.C. Swann, N.R. Newbury, High-stability transfer of an optical frequency
over long fiber-optic links. J. Opt. Soc. Am. B Opt. Phys. 25, 1284 (2008)
129. G. Grosche, O. Terra, K. Predehl, R. Holzwarth, B. Lipphardt, F. Vogt, U. Sterr, H. Schnatz,
Optical frequency transfer via 146 km fiber link with 10ˆ-19 relative accuracy. Opt. Lett. 34,
2270 (2009)
130. G. Grosche, Eavesdropping time and frequency: phase noise cancellation along a time-varying
path, such as an optical fiber. Opt. Lett. 39, 2545 (2014)
131. F. Stefani, O. Lopez, A. Bercy, W.-K. Lee, C. Chardonnet, G. Santarelli, P.-E. Pottie, A.
Amy-Klein, Tackling the limits of optical fiber links. J. Opt. Soc. Am. B Opt. Phys. 32, 787
(2015)
132. J. Geršl, P. Delva, P. Wolf, Relativistic corrections for time and frequency transfer in optical
fibres. Metrologia 52, 552 (2015)
133. P. Delva, J. Lodewyck, S. Bilicki, E. Bookjans, G. Vallet, R. Le Targat, P.-E. Pottie, C. Guerlin,
F. Meynadier, C. Le Poncin-Lafitte, O. Lopez, A. Amy-Klein, W.-K. Lee, N. Quintin, C. Lisdat,
A. Al-Masoudi, S. Dörscher, C. Grebing, G. Grosche, A. Kuhl, S. Raupach, U. Sterr, I.R. Hill,
R. Hobson, W. Bowden, J. Kronjäger, G. Marra, A. Rolland, F.N. Baynes, H.S. Margolis, P.
Gill, Test of special relativity using a fiber network of optical clocks. Phys. Rev. Lett. 118(22),
221102 (2017)
134. D. Colladay, V.A. Kostelecký, CPT violation and the standard model. Phys. Rev. D 55, 6760–
6774 (1997)
135. D. Colladay, V.A. Kostelecký, Lorentz-violating extension of the standard model. Phys. Rev.
D 58(11), 116002 (1998)
136. V.A. Kostelecký, M. Mewes, Signals for Lorentz violation in electrodynamics. Phys. Rev. D
66(5), 056005 (2002)
137. J.D. Tasson, What do we know about Lorentz invariance? Rep. Prog. Phys. 77(6), 062901
(2014)
138. V.A. Kostelecký, N. Russell, Data tables for Lorentz and CPT violation. Rev. Mod. Phys. 83,
11–32 (2011)
139. H. Pihan-Le Bars, C. Guerlin, R.-D. Lasseri, J.-P. Ebran, Q.G. Bailey, S. Bize, E. Khan,
P. Wolf, Lorentz-symmetry test at Planck-scale suppression with nucleons in a spin-polarized
133 Cs cold atom clock. Phys. Rev. D 95(7), 075026 (2017)
140. P. Wolf, F. Chapelet, S. Bize, A. Clairon, Cold atom clock test of Lorentz invariance in the
matter sector. Phys. Rev. Lett. 96(6), 060801 (2006)
141. M.A. Hohensee, S. Chu, A. Peters, H. Müller, Equivalence principle and gravitational redshift.
Phys. Rev. Lett. 106(15), 151102 (2011)
142. M.A. Hohensee, N. Leefer, D. Budker, C. Harabati, V.A. Dzuba, V.V. Flambaum, Limits on
violations of Lorentz symmetry and the Einstein equivalence principle using radio-frequency
spectroscopy of atomic dysprosium. Phys. Rev. Lett. 111(5), 050401 (2013)
143. T. Damour, Theoretical aspects of the equivalence principle. Class. Quantum Gravity 29(18),
184001 (2012)
144. J.-P. Uzan, Varying constants, gravitation and cosmology. Living Rev. Relativ. 14, 2 (2011)
145. J. Guéna, M. Abgrall, D. Rovera, P. Rosenbusch, M.E. Tobar, P. Laurent, A. Clairon, S. Bize,
Improved tests of local position invariance using Rb87 and Cs133 fountains. Phys. Rev. Lett.
109(8), 080801 (2012)
146. T. Rosenband, D.B. Hume, P.O. Schmidt, C.W. Chou, A. Brusch, L. Lorini, W.H. Oskay, R.E.
Drullinger, T.M. Fortier, J.E. Stalnaker, S.A. Diddams, W.C. Swann, N.R. Newbury, W.M.
Itano, D.J. Wineland, J.C. Bergquist, Frequency ratio of Al+ and Hg+ single-ion optical
clocks; metrology at the 17th decimal place. Science 319, 1808– (2008)
147. N. Leefer, C.T.M. Weber, A. Cingöz, J.R. Torgerson, D. Budker, New limits on variation of
the fine-structure constant using atomic dysprosium. Phys. Rev. Lett. 111(6), 060801 (2013)
148. R.M. Godun, P.B.R. Nisbet-Jones, J.M. Jones, S.A. King, L.A.M. Johnson, H.S. Margolis,
K. Szymaniec, S.N. Lea, K. Bongs, P. Gill, Frequency ratio of two optical clock transitions
in 171 Yb+ and constraints on the time variation of fundamental constants. Phys. Rev. Lett.
113(21), 210801 (2014)
149. N. Huntemann, B. Lipphardt, C. Tamm, V. Gerginov, S. Weyers, E. Peik, Improved limit on
a temporal variation of m p /me from comparisons of Yb+ and Cs atomic clocks. Phys. Rev.
Lett. 113(21), 210802 (2014)
150. P. Brax, C. Burrage, A.-C. Davis, G. Gubitosi, Cosmological tests of the disformal coupling
to radiation. J. Cosmol. Astropart. Phys. 11, 1 (2013)
151. A. Hees, O. Minazzoli, J. Larena, Breaking of the equivalence principle in the electromagnetic
sector and its cosmological signatures. Phys. Rev. D 90(12), 124064 (2014)
152. R.F.L. Holanda, K.N.N.O. Barros, Searching for cosmological signatures of the Einstein
equivalence principle breaking. Phys. Rev. D 94(2), 023524 (2016)
354 A. Hees et al.
153. S. Peil, S. Crane, J.L. Hanssen, T.B. Swanson, C.R. Ekstrom, Tests of local position invariance
using continuously running atomic clocks. Phys. Rev. A 87(1), 010102 (2013)
154. N. Ashby, T.P. Heavner, S.R. Jefferts, T.E. Parker, A.G. Radnaev, Y.O. Dudin, Testing local
position invariance with four cesium-fountain primary frequency standards and four NIST
hydrogen masers. Phys. Rev. Lett. 98(7), 070802 (2007)
155. R.V. Pound, G.A. Rebka, Gravitational red-shift in nuclear resonance. Phys. Rev. Lett. 3,
439–441 (1959)
156. R.F.C. Vessot, M.W. Levine, E.M. Mattison, E.L. Blomberg, T.E. Hoffman, G.U. Nystrom,
B.F. Farrel, R. Decher, P.B. Eby, C.R. Baugher, Test of relativistic gravitation with a space-
borne hydrogen maser. Phys. Rev. Lett. 45, 2081–2084 (1980)
157. P. Delva, A. Hees, S. Bertone, E. Richard, P. Wolf, Test of the gravitational redshift with stable
clocks in eccentric orbits: application to Galileo satellites 5 and 6. Class. Quantum Gravity
32(23), 232003 (2015)
158. L. Cacciapuoti, C. Salomon, Atomic clock ensemble in space. J. Phys. Conf. Ser. 327(1),
012049 (2011)
159. M. Soffel, H. Herold, H. Ruder, M. Schneider, Relativistic theory of gravimetric measurements
and definition of thegeoid. Manuscr. Geod. 13, 143–146 (1988)
160. S.M. Kopejkin, Relativistic Manifestations of gravitational fields in gravimetry and geodesy.
Manuscripta Geodaetica 16 (1991)
161. J. Müller, M. Soffel, S.A. Klioner, Geodesy and relativity. J. Geodesy 82, 133–145 (2008)
162. P. Delva, J. Lodewyck, Atomic clocks: new prospects in metrology and geodesy. Acta Futura,
(7), 67–78, 7:67–78, November 2013
163. B. Altschul, Q.G. Bailey, L. Blanchet, K. Bongs, P. Bouyer, L. Cacciapuoti, S. Capozziello,
N. Gaaloul, D. Giulini, J. Hartwig, L. Iess, P. Jetzer, A. Landragin, E. Rasel, S. Reynaud,
S. Schiller, C. Schubert, F. Sorrentino, U. Sterr, J.D. Tasson, G.M. Tino, P. Tuckey, P. Wolf,
Quantum tests of the Einstein equivalence principle with the STE-QUEST space mission.
Adv. Space Res. 55, 501–524 (2015)
164. S. Zucker, T. Alexander, S. Gillessen, F. Eisenhauer, R. Genzel, Probing post-Newtonian
physics near the galactic black hole with stellar redshift measurements. ApJl 639, L21–L24
(2006)
165. M. Grould, F.H. Vincent, T. Paumard, G. Perrin, General relativistic effects on the orbit of the
S2 star with GRAVITY. ArXiv e-prints (2017)
166. Gravity Collaboration, R. Abuter, A. Amorim, N. Anugu, M. Bauböck, M. Benisty, J. P. Berger,
N. Blind, H. Bonnet, W. Brandner, A. Buron, C. Collin, F. Chapron, Y. Clénet, V. Coudé Du
Foresto, P. T. de Zeeuw, C. Deen, F. Delplancke-Ströbele, R. Dembet, J. Dexter, G. Duvert,
A. Eckart, F. Eisenhauer, G. Finger, N. M. Förster Schreiber, P. Fédou, P. Garcia, R. Garcia
Lopez, F. Gao, E. Gendron, R. Genzel, S. Gillessen, P. Gordo, M. Habibi, X. Haubois, M. Haug,
F. Haußmann, T. Henning, S. Hippler, M. Horrobin, Z. Hubert, N. Hubin, A. Jimenez Ros-
ales, L. Jochum, K. Jocou, A. Kaufer, S. Kellner, S. Kendrew, P. Kervella, Y. Kok, M. Kulas,
S. Lacour, V. Lapeyrère, B. Lazareff, J.-B. Le Bouquin, P. Léna, M. Lippa, R. Lenzen,
A. Mérand, E. Müler, U. Neumann, T. Ott, L. Palanca, T. Paumard, L. Pasquini, K. Perraut,
G. Perrin, O. Pfuhl, P. M. Plewa, S. Rabien, A. Ramírez, J. Ramos, C. Rau, G. Rodríguez-Coira,
R.-R. Rohloff, G. Rousset, J. Sanchez-Bermudez, S. Scheithauer, M. Schöller, N. Schuler,
J. Spyromilio, O. Straub, C. Straubmeier, E. Sturm, L. J. Tacconi, K. R. W. Tristram, F. Vincent,
S. von Fellenberg, I. Wank, I. Waisberg, F. Widmann, E. Wieprecht, M. Wiest, E. Wiezorrek,
J. Woillez, S. Yazici, D. Ziegler, and G. Zins. Detection of the gravitational redshift in the
orbit of the star S2 near the Galactic centre massive black hole. A&A 615, L15 (2018)
167. S. Weinberg, A new light boson? Phys. Rev. Lett. 40, 223–226 (1978)
168. J. Preskill, M.B. Wise, F. Wilczek, Cosmology of the invisible axion. Phys. Lett. B 120,
127–132 (1983)
169. W. Hu, R. Barkana, A. Gruzinov, Fuzzy cold dark matter: the wave properties of ultralight
particles. Phys. Rev. Lett. 85, 1158–1161 (2000)
170. F. Piazza, M. Pospelov, Sub-eV scalar dark matter through the super-renormalizable Higgs
portal. Phys. Rev. D 82(4), 043533 (2010)
171. A. Khmelnitsky, V. Rubakov, Pulsar timing signal from ultralight scalar dark matter. J. Cosmol.
Astropart. Phys. 2, 019 (2014)
172. N.K. Porayko, K.A. Postnov, Constraints on ultralight scalar dark matter from pulsar timing.
Phys. Rev. D 90(6), 062008 (2014)
173. H.-Y. Schive, T. Chiueh, T. Broadhurst, Cosmic structure as the quantum interference of a
coherent dark wave. Nat. Phys. 10, 496–499 (2014)
174. J. Beyer, C. Wetterich, Small scale structures in coupled scalar field dark matter. Phys. Lett.
B 738, 418–423 (2014)
175. A. Arvanitaki, J. Huang, K. Van Tilburg, Searching for dilaton dark matter with atomic clocks.
Phys. Rev. D 91(1), 015015 (2015)
176. Y.V. Stadnik, V.V. Flambaum, Searching for dark matter and variation of fundamental con-
stants with laser and maser interferometry. Phys. Rev. Lett. 114(16), 161301 (2015)
177. Y.V. Stadnik, V.V. Flambaum, Can dark matter induce cosmological evolution of the funda-
mental constants of nature? Phys. Rev. Lett. 115(20), 201301 (2015)
178. P.W. Graham, D.E. Kaplan, J. Mardon, S. Rajendran, W.A. Terrano, Dark matter direct detec-
tion with accelerometers. Phys. Rev. D 93(7), 075029 (2016)
179. A. Arvanitaki, S. Dimopoulos, K. Van Tilburg, Sound of dark matter: searching for light
scalars with resonant-mass detectors. Phys. Rev. Lett. 116(3), 031102 (2016)
180. Y.V. Stadnik, V.V. Flambaum, Enhanced effects of variation of the fundamental constants in
laser interferometers and application to dark-matter detection. Phys. Rev. A 93(6), 063630
(2016)
181. Y.V. Stadnik, V.V. Flambaum, Improved limits on interactions of low-mass spin-0 dark matter
from atomic clock spectroscopy. Phys. Rev. A 94(2), 022111 (2016)
182. D.J.E. Marsh, Axion cosmology. Phys. Rep. 643, 1–79 (2016)
183. L.A. Ureña-López, A.X. Gonzalez-Morales, Towards accurate cosmological predictions for
rapidly oscillating scalar fields as dark matter. J. Cosmol. Astropart. Phys. 7, 048 (2016)
184. E. Calabrese, D.N. Spergel, Ultra-light dark matter in ultra-faint dwarf galaxies. MNRAS
460, 4397–4402 (2016)
185. D. Blas, D.L. Nacir, S. Sibiryakov, Ultralight dark matter resonates with binary pulsars. Phys.
Rev. Lett. 118(26), 261102 (2017)
186. T. Bernal, V.H. Robles, T. Matos, Scalar field dark matter in clusters of galaxies. MNRAS
468, 3135–3149 (2017)
187. L. Hui, J.P. Ostriker, S. Tremaine, E. Witten, Ultralight scalars as cosmological dark matter.
Phys. Rev. D 95(4), 043541 (2017)
188. C. Abel, N.J. Ayres, G. Ban, G. Bison, K. Bodek, V. Bondar, M. Daum, M. Fairbairn, V.V.
Flambaum, P. Geltenbort, K. Green, W.C. Griffith, M. van der Grinten, Z.D. Grujić, P.G.
Harris, N. Hild, P. Iaydjiev, S.N. Ivanov, M. Kasprzak, Y. Kermaidic, K. Kirch, H.-C. Koch,
S. Komposch, P.A. Koss, A. Kozela, J. Krempel, B. Lauss, T. Lefort, Y. Lemière, D.J.E.
Marsh, P. Mohanmurthy, A. Mtchedlishvili, M. Musgrave, F.M. Piegsa, G. Pignol, M. Rawlik,
D. Rebreyend, D. Ries, S. Roccia, D. Rozpȩdzik, P. Schmidt-Wellenburg, N. Severijns, D.
Shiers, Y.V. Stadnik, A. Weis, E. Wursten, J. Zejma, G. Zsigmond, Search for axion-like dark
matter through nuclear spin precession in electric and magnetic fields. Phys. Rev. X 7(4),
041034 (2017)
189. T. Bernal, L.M. Fernández-Hernández, T. Matos, M.A. Rodríguez-Meza, Rotation curves of
high-resolution LSB and SPARC galaxies with fuzzy and multistate (ultralight boson) scalar
field dark matter. MNRAS 475, 1447–1468 (2018)
190. M.B. Green, J.H. Schwarz, E. Witten, Superstring Theory (1988)
191. T. Damour, A.M. Polyakov, The string dilation and a least coupling principle. Nucl. Phys. B
423, 532–558 (1994)
192. T. Damour, A.M. Polyakov, String theory and gravity. Gen. Relativ. Gravit. 26, 1171–1176
(1994)
193. M. Gasperini, F. Piazza, G. Veneziano, Quintessence as a runaway dilaton. Phys. Rev. D 65(2),
023508 (2001)
356 A. Hees et al.
194. T. Damour, F. Piazza, G. Veneziano, Runaway dilaton and equivalence principle violations.
Phys. Rev. Lett. 89(8), 081601 (2002)
195. T. Damour, J.F. Donoghue, Equivalence principle violations and couplings of a light dilaton.
Phys. Rev. D 82(8), 084033 (2010)
196. A. Hees, O. Minazzoli, E. Savalle, Y.V. Stadnik, P. Wolf, Violation of the equivalence principle
from light scalar dark matter. Phys. Rev. D 98(6), 064051 (2018)
197. A. Hees, J. Guéna, M. Abgrall, S. Bize, P. Wolf, Searching for an oscillating massive scalar
field as a dark matter candidate using atomic hyperfine frequency comparisons. Phys. Rev.
Lett. 117(6), 061301 (2016)
198. K. Van Tilburg, N. Leefer, L. Bougas, D. Budker, Search for ultralight scalar dark matter with
atomic spectroscopy. Phys. Rev. Lett. 115(1), 011802 (2015)
199. T.A. de Pirey Saint Alby, N. Yunes, Cosmological evolution and Solar System consistency of
massive scalar-tensor gravity. Phys. Rev. D 96(6), 064040 (2017)
200. T. Damour, G. Esposito-Farese, Nonperturbative strong-field effects in tensor-scalar theories
of gravitation. Phys. Rev. Lett. 70, 2220–2223 (1993)
201. T. Damour, G. Esposito-Farèse, Tensor-scalar gravity and binary-pulsar experiments. Phys.
Rev. D 54, 1474–1491 (1996)
202. O. Minazzoli, A. Hees, Late-time cosmology of a scalar-tensor theory with a universal mul-
tiplicative coupling between the scalar field and the matter Lagrangian. Phys. Rev. D 90(2),
023017 (2014)
203. S.M. Carroll, Quintessence and the rest of the world: suppressing long-range interactions.
Phys. Rev. Lett. 81, 3067–3070 (1998)
204. N.J. Nunes, J.E. Lidsey, Reconstructing the dark energy equation of state with varying alpha.
Phys. Rev. D 69(12), 123511 (2004)
205. C.J.A.P. Martins, A.M.M. Pinho, Fine-structure constant constraints on dark energy. Phys.
Rev. D 91(10), 103501 (2015)
206. M.A. Hohensee, H. Müller, R.B. Wiringa, Equivalence principle and bound kinetic energy.
Phys. Rev. Lett. 111(15), 151102 (2013)
207. H. Müller, S. Herrmann, A. Saenz, A. Peters, C. Lämmerzahl, Optical cavity tests of Lorentz
invariance for the electron. Phys. Rev. D 68(11), 116006 (2003)
208. H. Müller, Testing Lorentz invariance by the use of vacuum and matter filled cavity resonators.
Phys. Rev. D 71(4), 045004 (2005)
209. N.A. Flowers, C. Goodge, J.D. Tasson, Superconducting-gravimeter tests of local lorentz
invariance. Phys. Rev. Lett. 119(20), 201101 (2017)
210. A. Bourgoin, C. Le Poncin-Lafitte, A. Hees, S. Bouquillon, G. Francou, M.-C. Angonin,
Lorentz symmetry violations from matter-gravity couplings with lunar laser ranging. Phys.
Rev. Lett. 119(20), 201102 (2017)
211. H. Pihan-Le Bars, C. Guerlin, P. Wolf, Progress on testing Lorentz symmetry with MICRO-
SCOPE. ArXiv e-prints (2017)
212. B. Bertotti, L. Iess, P. Tortora, A test of general relativity using radio links with the Cassini
spacecraft. Nature 425, 374–376 (2003)
213. S.B. Lambert, C. Le Poncin-Lafitte, Determining the relativistic parameter γ using very long
baseline interferometry. A&A 499, 331–335 (2009)
214. S.B. Lambert, C. Le Poncin-Lafitte, Improved determination of γ by VLBI. A&A 529, A70
(2011)
215. A.S. Konopliv, S.W. Asmar, W.M. Folkner, Ö. Karatekin, D.C. Nunes, S.E. Smrekar, C.F.
Yoder, M.T. Zuber, Mars high resolution gravity fields from mro, mars seasonal gravity, and
other dynamical parameters. Icarus 211(1), 401–428 (2011)
216. R.S. Park, W.M. Folkner, A.S. Konopliv, J.G. Williams, D.E. Smith, M.T. Zuber, Precession
of mercury’s perihelion from ranging to the MESSENGER spacecraft. AJ 153, 121 (2017)
217. E.G. Adelberger, B.R. Heckel, A.E. Nelson, Tests of the gravitational inverse-square law.
Ann. Rev. Nucl. Part. Sci. 53, 77–121 (2003)
218. P. Jordan, Formation of the stars and development of the universe. Nature 164, 637–640 (1949)
219. C. Brans, R.H. Dicke, Mach’s principle and a relativistic theory of gravitation. Phys. Rev.
124, 925–935 (1961)
220. T. Damour, G. Esposito-Farese, Tensor-multi-scalar theories of gravitation. Class. Quantum
Gravity 9, 2093–2176 (1992)
221. J. Khoury, A. Weltman, Chameleon cosmology. Phys. Rev. D 69(4), 044026 (2004)
222. J. Khoury, A. Weltman, Chameleon fields: awaiting surprises for tests of gravity in space.
Phys. Rev. Lett. 93(17), 171104 (2004)
223. A. Hees, A. Füzfa, Combined cosmological and solar system constraints on chameleon mech-
anism. Phys. Rev. D 85(10), 103005 (2012)
224. K. Hinterbichler, J. Khoury, Screening long-range forces through local symmetry restoration.
Phys. Rev. Lett. 104(23), 231301 (2010)
225. K. Hinterbichler, J. Khoury, A. Levy, A. Matas, Symmetron cosmology. Phys. Rev. D 84(10),
103521 (2011)
226. A.I. Vainshtein, To the problem of nonvanishing gravitation mass. Phys. Lett. B 39, 393–394
(1972)
227. E. Babichev, C. Deffayet, R. Ziour, The vainshtein mechanism in the decoupling limit of
massive gravity. Jo. High Energy Phys. 5, 98 (2009)
228. E. Babichev, C. Deffayet, R. Ziour, k-MOUFLAGE gravity. Int. J. Mod. Phys. D 18, 2147–
2154 (2009)
229. E. Babichev, C. Deffayet, G. Esposito-Farèse, Constraints on shift-symmetric scalar-tensor
theories with a vainshtein mechanism from bounds on the time variation of G. Phys. Rev.
Lett. 107(25), 251102 (2011)
230. F. Hofmann, J. Müller, Relativistic tests with lunar laser ranging. Class. Quantum Gravity
35(3), 035015 (2018)
231. M.D. Seifert, Vector models of gravitational Lorentz symmetry breaking. Phys. Rev. D 79(12),
124012 (2009)
232. V.A. Kostelecký, R. Potting, Gravity from local Lorentz violation. Gen. Relativ. Gravit. 37,
1675–1679 (2005)
233. V.A. Kostelecký, R. Potting, Gravity from spontaneous Lorentz violation. Phys. Rev. D 79(6),
065018 (2009)
234. B. Altschul, Q.G. Bailey, V.A. Kostelecký, Lorentz violation with an antisymmetric tensor.
Phys. Rev. D 81(6), 065028 (2010)
235. R. Gambini, J. Pullin, Nonstandard optics from quantum space-time. Phys. Rev. D 59(12),
124021 (1999)
236. V.A. Kostelecký, M. Mewes, Electrodynamics with Lorentz-violating operators of arbitrary
dimension. Phys. Rev. D 80(1), 015020 (2009)
237. S.M. Carroll, J.A. Harvey, V.A. Kostelecký, C.D. Lane, T. Okamoto, Noncommutative field
theory and Lorentz violation. Phys. Rev. Lett. 87(14), 141601 (2001)
238. V.A. Kostelecký, R. Lehnert, Stability, causality, and Lorentz and CPT violation. Phys. Rev.
D 63(6), 065008 (2001)
239. Q.G. Bailey, V.A. Kostelecký, R. Xu, Short-range gravity and Lorentz violation. Phys. Rev.
D 91(2), 022006 (2015)
240. V.A. Kostelecký, J.D. Tasson, Constraints on Lorentz violation from gravitational Čerenkov
radiation. Phys. Lett. B 749, 551–559 (2015)
241. V.A. Kostelecký, M. Mewes, Testing local Lorentz invariance with gravitational waves. Phys.
Lett. B 757, 510–514 (2016)
242. Q.G. Bailey, D. Havert, Velocity-dependent inverse cubic force and solar system gravity tests.
Phys. Rev. D 96(6), 064035 (2017)
243. H. Müller, S.-W. Chiow, S. Herrmann, S. Chu, K.-Y. Chung, Atom-interferometry tests of the
isotropy of post-Newtonian gravity. Phys. Rev. Lett. 100(3), 031101 (2008)
244. K.-Y. Chung, S.-W. Chiow, S. Herrmann, S. Chu, H. Müller, Atom interferometry tests of
local Lorentz invariance in gravity and electrodynamics. Phys. Rev. D 80(1), 016002 (2009)
245. C.-G. Shao, Y.-F. Chen, R. Sun, L.-S. Cao, M.-K. Zhou, Z.-K. Hu, C. Yu, H. Müller, Limits
on Lorentz violation in gravity from worldwide superconducting gravimeters. Phys. Rev. D
97(2), 024019 (2018)
358 A. Hees et al.
246. A. Hees, Q.G. Bailey, C. Le Poncin-Lafitte, A. Bourgoin, A. Rivoldini, B. Lamine, F. Mey-

nadier, C. Guerlin, P. Wolf, Testing Lorentz symmetry with planetary orbital dynamics. Phys.
Rev. D 92, 064049 (2015)
247. C. Le Poncin-Lafitte, A. Hees, S. lambert, Lorentz symmetry and very long baseline interfer-
ometry. Phys. Rev. D 94(12), 125030 (2016)
248. J.B.R. Battat, J.F. Chandler, C.W. Stubbs, Testing for Lorentz violation: constraints on
standard-model-extension parameters via lunar laser ranging. Phys. Rev. Lett. 99(24), 241103
(2007)
249. A. Bourgoin, A. Hees, S. Bouquillon, C. Le Poncin-Lafitte, G. Francou, M.-C. Angonin,
Testing Lorentz symmetry with lunar laser ranging. Phys. Rev. Lett. 117(24), 241301 (2016)
250. M. Milgrom, A modification of the Newtonian dynamics as a possible alternative to the hidden
mass hypothesis. ApJ 270, 365–370 (1983)
251. M. Milgrom, A modification of the Newtonian dynamics - implications for galaxy systems.
ApJ 270, 384 (1983)
252. M. Milgrom, A modification of the Newtonian dynamics - implications for galaxies. ApJ 270,
371–389 (1983)
253. J.-P. Bruneton, G. Esposito-Farèse, Field-theoretical formulations of MOND-like gravity.
Phys. Rev. D 76(12), 124012 (2007)
254. A. Hees, B. Famaey, G.W. Angus, G. Gentile, Combined solar system and rotation curve
constraints on MOND. MNRAS 455, 449–461 (2016)
255. R.A. Swaters, R.H. Sanders, S.S. McGaugh, Testing modified Newtonian dynamics with
rotation curves of dwarf and low surface brightness galaxies. ApJ 718, 380–391 (2010)
Operationalization of Basic Relativistic
Measurements
Bruno Hartmann
Abstract We present a novel phenomenological foundation of relativistic physics.

That means, we focus on the observable entities and make no mathematical pre-
assumptions. Like Einstein for relativistic kinematics we start from vivid measure-
ment operations and simple natural principles. Seeking, formulating and refining
operational definitions reveals the physical meaning. We grasp the basic observables
(length, duration, inertial mass, momentum, energy) in a physical way. We define an
order of energy and impulse from a physical comparison. Each step (the construction
of “sufficiently constant” reference devices and of a machinery, which “functions”
for a basic measurement) follows from practical requirements. One can directly count
the tangible measurement units and ultimately derive the fundamental equations (e.g.
the kinetic energy-velocity relation or the mass-energy equivalence).
1 Introduction
In relativistic geodesy physicists develop the measurement practice. One manufac-

tures increasingly precise reference devices, operates with them, and fits the data with
equations of relativity theory. We present a foundation of these basic equations. We
define all mathematical variables and operations strictly from the underlying physical
operations. We review that approach for relativistic kinematics and dynamics.
We all use the terms length, duration, mass or energy in our everyday lives and
come to an intuition about them. The challenge in learning about physics is how
to reconcile the common understanding with the scientific concepts that are more
precise [1]. Among physics education researchers broad consensus shows, that one
“constructs” new knowledge from prior knowledge [2]. Our goal is to help understand
learning as “the refinement of everyday thinking”. The common everyday experi-
ences and language serves as a starting point to understand space and time and energy.
We develop the preconceptions into a physicists view of measurements so that one
can describe these observables more precisely. We reveal a novel aspect of physics,
B. Hartmann (B)
Humboldt University, Berlin, Germany
e-mail: brunohartmannjr@gmail.com

360 B. Hartmann
Fig. 1 a Count unit weights m u on balance scale b count rulers along straight line
the methodological knowledge of a quantification, i.e. of a direct measurement. We

construct precise physical quantities of space, time and energy hands-on and then
derive the basic equations. We develop the theory from an empirical basis and not
retrospectively from non-vivid expressions and formal equations. For the resulting
formalism one can understand scope and limitations.
Before our approach is outlined we specify what we understand as a basic mea-
surement. According to Gerthsen [3] “Physics is a measuring science”. One defines
“measuring a quantity” by the direct or indirect comparison with a measurement
unit. Mostly this is not done directly - one does not use an energymeter - but indi-
rectly by measuring other primary quantities and then using an agreed formula. This
risks to be circular. For physics Gerthsen demands solely operational definitions, that
specify the quantity to be measured, e.g. inertial mass or energy, by its effect in a
measurement process.
We are long familiar with basic measurements as in the case of weights and
lengths. We order the basic observables through a practical comparison, for example
“heavier than” if for example a person called Otto outweighs a reference body on
a balance scale (Fig. 1a). One wants to specify the values also numerically: “how
many times” heavier. Hence a procedure is needed that leads to a quantification. The
procedure for finding these values is the measurement. In a weight measurement
one successively piles up weight units until the pile is sufficiently precise neither
heavier nor lighter than Otto. One manufactures standard reference objects, units
which are all of the same weight, and counts them. If one needs z weight units then
the weight of Otto is equal to z times the weight of the reference object. Formally
we write a measurement result m[O] = z · m u . One also knows the basic observable
length. One defines: Otto is “longer than” a measuring stick, if he towers over the
latter. For the measurement one manufactures equally long measuring sticks and
successively places them along a straight line until they cover Otto (Fig. 1b). If one
needs n sticks, then the length of Otto is equal to n times the length of the reference
stick l[O] = n · lu .
Helmholtz [4] specifies the basic measurement process. Essentially one begins
from counting same objects. The basic observable (weight, length etc.) is ordered
by a practical comparison. A measurement quantifies the basic observable. One
obtains a quantified observable expressed as a number (of reference devices) times
Operationalization of Basic Relativistic Measurements 361
its dimension (standard weight, standard length etc.). A basic measurement requires
knowledge of the method of comparison (of a particular attribute of both bodies),
we write “>”, “=” or “<”, and of the physical concatenation method. Obviously the
way of concatenation (piling weight units, placing meter sticks along straight line,
coupling impulse units etc.) generally depends on the type of measure. One simply
writes the formal symbol “+”. Our physical foundation will specify the underlying
tangible operations. We will describe an approach how space-time in Sect. 2, iner-
tial mass and momentum in Sect. 3, and energy in Sect. 4 can be introduced as an
observable and measurable concept. We will derive the exact mathematical relations
between these physical quantities in the end.
2 Kinematics
First we outline Helmholtz program for measuring relativistic motion [5]. We define
a spatiotemporal order “longer than” by the practical comparison, whether one object
or process covers the other. Without one word of mathematics Alice can manufac-
ture uniform running light clocks “L” and place them literally one after the other
or side by side “L + L” and then count, how many building blocks one needs to
assemble a regular grid which covers the measurement object (Fig. 2a). These are
the elementary measurement operations. From their structure and overlap we derive
mathematical relations, e.g. for two different observers Alice and Bob the formal
Lorentz transformation (Fig. 2b).
2.1 Reference Process
The construction of a basic measurement process for space-time intervals makes

it necessary to define a suitable reference process. Traditionally one manufactures
rulers and clocks. An arbitrary chunk of matter does not function as a measurement
instrument. Lorenzen, Janich [6] explain the actions that lead to these artifacts. How
does one manufacture increasingly precise clocks if one did not have them before?
By testing procedures one can examine the straight form of a constructed ruler and
the uniform running of a clock. Such test norms developed historically from grind-
ing practice, from intuitively controlled actions in everyday technical work. With
Lorenzen, Janich [6] we can presuppose (circularity free and without mathemati-
cal presuppositions) that every observer can produce “straight”, “rigid” rulers and
“uniform running” clocks. Provided the classical measurement practice one gets the
Euclidean metric. We will demonstrate the step to relativistic kinematics.
Einstein did analyze the principles for measuring moving objects. To make any
sense of the concept of time Einstein [7] required an additional process that estab-
lishes a connection between clocks at distant locations. In principle one could use
any type of process. Most favorable for the theory one chooses a process about which
362 B. Hartmann
Fig. 2 a Count regular layout of light clocks L b intrinsic operations by Alice and Bob
we know something certain. For the motion of light this holds much more than for
any other process. By local comparison with classical rulers and clocks the light
propagates in a uniform, isotropic and straight way. The motion of light provides a
universal reference for any intrinsic observer.
Let Alice construct a light clock L with two nearby mirrors in a rigid frame.
Between both mirrors the light constantly oscillates back and forth (vertical sequence
in Fig. 2a). Each tick of her reference device covers the same standard distance sL
and because of the light principle each successive tick takes the same standard time
tL .1 We use the light clock as a new measurement unit. The width of the light clock
represents unit length sL and each tick a unit time tL .
1 One replaces the traditional rulers and clocks by new light clocks because they approximate the
aspired ideal of uniform running more precisely. Contemporary metrology regards the speed of
light c as an invariant natural constant and introduces optical clocks as a frequency standard. The
definition of the standard length sL is based on a given standard duration tL and the universal speed
of light c.
Basic dimensions (unit length, unit time etc.) are “arbitrarily chosen constant reference mea-
sures” [8]. Contemporary metrology refers to the standard duration tCs of a Cesium period and the
speed of light c. On paper one defines the atomic second secSI := 9192631770 · tCs as a multiple
of that standard duration (to match the calendrical second). This defines a standard distance, which
light propagates in one atomic second of flight. Then one defines the standard meter, again on paper,
as a certain fraction of that distance 299792458 · mSI := (c · secSI ). These new SI units are more
practical. Now everyone can reproduce the SI second and meter from invariant natural processes
(Cs-period and speed of light) and fixed numerical factors (to match the traditional prototypes, that
refer to a tropical year and to a platinum-iridium standard bar from the international bureau of
weights and standards, that are not accessible and change in time!).
2.2 Quantification
In the measurement Alice joins together light clock ticks one after another until
the sequence covers e.g. the waiting interval from a starting moment A1 to an end
in moment A (Fig. 2a). Consider a swarm of identically constituted light clocks.
Alice can place ticking light clocks literally side by side. By letting their inner
light rays overlap all light clocks tick synchronized. This connection (dashed line
in Fig. 2a) represents Alice simultaneous straight measurement path AO towards
the distant event O. From solely congruent building blocks Alice assembles a grid
which covers larger objects or processes, e.g. the relative motion of Bob or the
light ray from A1 to O. In both segments of her material model Alice counts the
number n of congruent ticks and the number l of congruent clocks. Alice measures
the spatiotemporal interval A1 O
t A 1 O = n · tL (1)
s A 1 O = l · sL
by the number of measurement units and their dimension (standard duration tL and
width sL ).
By inspection of the regular pattern of light clocks (Fig. 2a) one counts for
the outgoing light ray (t, s) A1 O = (n · tL , 2n · sL ) and for the returning light ray
(t, s)O A2 = (n · tL , −2n · sL ) the same distance units and duration units (n the num-
ber of ticks from A1 to A resp. from A to A2 , 2n the number of clocks sitting side
by side along AO). Both measured values are proportional. In standard light clock
dimensions the velocity of light has the value c(L) := 2n n
= 2. For any given duration
tc the light ray travels a distance
sL
sc = 2 · · tc . (2)
t
L
≡c
For a light clock built so that tL := secSI and (2 · sL ) =: 299792458 · mSI (defini-
tion SI-meter) we express the speed of light in conventional SI-units c = 2 · stLL =
299792458 · sec m
. For the outgoing and the returning light ray we measure the same
velocity cout = cin because our measurement unit, the light clock is build from two-
way light cycles on equal footing.
Alice can express her direct measurement of the spatiotemporal interval A1 O (by
literally covering the object with a layout of measurement units) with the help of
t A1 A = 21 · t A1 A2

(2) 1 c
( t A1 O , s A1 O ) = ·t , ·t (3)
2 A1 A2 2 A1 A2
364 B. Hartmann
in terms of measuring just the round trip waiting times t A1 A2 . That corresponds to the
familiar principle of laser ranging measurements, by sending a light ray in moment
A1 to a distant object O and waiting until the reflection returns in A2 (Fig. 2a). We
derive that basic equation from direct measurements (1) of the outgoing and returning
light ray.
2.3 Lorentz Transformation
We utilize a reference process, which represents two dimensions (the length and
the duration of the light cycle between two mirrors). Alice and Bob manufacture
their own light clocks and place them intrinsically until the measurement object is
covered. Along the measurement paths everyone counts the number of ticks and of
clocks. The relation between these numbers that is the Lorentz transformation.
Let one more observer Bob measure the distant object O. By connecting his light
clocks in an intrinsically timelike resp. spacelike way Bob constructs his straight
simultaneous measurement path to the object BO (dashed line in Fig. 2b). He can
also determine his simultaneity lines from symmetric round trip signaling times. For
the moments of departure B1 and arrival B2 of a light signal reflected from an event
(anywhere on line BO) that is simultaneous to the event B one counts the same
number of light clock ticks t B1 B = t B B2 in Fig. 2b. That corresponds to Einstein’s
convention of synchronizing distant clocks. Alice and Bob conduct the same intrinsic
operations with light clocks. Though the results are not identical. Alice constructs
simultaneity lines AO with different orientation than Bob’s simultaneity lines BO
(dashed lines in Fig. 2b).
Let Alice and Bob measure the same spacetime interval PO. Without restricting
generality both measure a segment of the moving object O, let all start from the joint
moment P. Figure 3 shows the interrelation of their laser ranging measurements:
a the clock ticks until Alice A1 sends out the light signal
b + b the ticks until the reflection from Bob B1 returns in A2
d + d the ticks until the reflection from the object O returns in A3
m the clock ticks until the coinciding light signal passes Bob B1
n + n the ticks until the reflection from the object O returns to Bob B2 .
From his round trip signaling durations m, n Bob measures for the moving object
(B) (3)
tPO = m+n
(B)
sPO =c·n .
Alice measures for the same object from her laser ranging durations a, d
(A)
tPO = a+d
(A)
sPO = c·d .
Fig. 3 coinciding light rays in the laser ranging measurements by Alice and Bob
366 B. Hartmann
With the laser ranging durations a, b Alice also measures the relative motion of Bob
tP(A)
B
= a+b
1
(A)
sP B
= c·b .
1
Now we only need to find the connection between Bob’s light clocks readings
m, n and Alice measurements a, b, d. We find one relation from the similar triangles
P B1 A2 and P B2 A3 (the light rays are pairwise parallel or coincide)
m m+n+n
= .
a+b+b a+d +d
For the second relation we interpret the two triangles P A1 B1 and P B1 A2 as a cali-
bration procedure by which Alice and Bob compare their light clocks L(A) and L(B) :
Let Alice send two light signals at the moments P and A1 towards Bob. He receives
both signals at the moments P and B1 . Vice versa let Bob send two light signals
at the moments P and B1 towards Alice. She receives both signals at the moments
P and A2 . According to the relativity principle both configurations are intrinsically
similar. Alice and Bob determine the same ratio between the duration for receiving
both signals (heard from the other) and the duration of the sending interval (measured
by themselves)
m ! a+b+b
= .
a m
After successive substitution (details in appendix of [5]) we can express Bob’s
spacetime measurements of the moving object (t, s)(B)
PO
in terms of Alice two mea-
surements of the moving object (t, s)(A)
PO
and of the relative motion of Bob (t, s)(A)
PB 1
(B) 1 (A) 1 vB (A)

tPO = · tPO − · · sPO (4)
v 2B v 2B c2
1− c2
1− c2
(B) 1 (A) 1 (A)
sPO =− · v B · tPO + · sPO
v 2B v 2B
1− c2
1− c2

(A) (A)
where Alice determines Bob’s relative velocity from v B := sP t . We derive the
B PB
Lorentz transformation between the spacetime measurements of Alice and Bob. That
is a strictly physical approach to physics where initially mathematics must remain
outside - and then every step where mathematics is introduced requires an extra jus-
tification. In the same way we develop Helmholtz program for direct measurements
of interactions.
Provided the classical measurement practice we get Galilei kinematics. In ana-
lyzing the measuring of moving objects Einstein discovered the additional need to
establish a physical connection between clocks at different locations and speeds. Our
intrinsic operations with light clocks represent the classical metric locally (Euclidean
geometry). For the connection Einstein chose the universal motion of light. By includ-
ing the (local) light principle and the relativity principle we derive Poincare kine-
matics. Our locally regular grid of light clocks can potentially spread out into every
direction. It represents our metric connection between distant measurements.
In local laser ranging practice the accelerations of an object are negligible com-
pared to the enormous speed of light. For larger configurations though effects accu-
mulate for example in the twin paradox [5]. One ought to keep the clocks and rulers
in an inertial reference frame. We conclude the discussion of kinematics with a brief
remark on the principle of inertia that also connects to the following discussion of
dynamics. In practical experimentation one probes the effects of certain sources on a
given object. When those actions become reproducible one can utilize the knowledge
for building and steering machines [9]. We identify the action of external forces on
a given body by changes in its state of motion - relative to other objects that are not
effected by systematic changes of the source. One can examine a given process with
the reference to the “relatively fixed” stars or one can mount the reference instru-
ments to the rigid ground of Earth, whose enormous mass will practically not be
effected by systematic changes of the source. One can also provide local reference
devices on a microscopic scale. In portable electronic devices one can increasingly
find microelectromechanical systems with integrated accelerometers, i.e. an intrin-
sic detector that measures proper acceleration resp. inertial motion. Thus, for all
practical purposes the selection of a (sufficiently) inertial reference is well defined.
Then one can apply the principle of inertia: “Every body with no forces acting on it
remains - as judged from the lab - in a state of rest or of uniform rectilinear motion”
[10].
3 Momentum and Inertial Mass
Our goal is to also define impulse, inertial mass (and later energy) as a basic observ-
able that can be measured directly. Momentum can be seen as a power to overrun
other objects in collisions. We define a direct comparison method. Let two generic
bodies a and b run into each other with initial velocities va resp. vb . According to
Galilei (in [11]) object a has “more momentum than” object b if in an head-on colli-
sion test a overruns b (they collide, stick together and move in a’s initial direction).
We symbolize this case with p [a] > p [b]. In a special case they can come to rest;
then their momentum is the same p [a] = p [b]. We remark that the comparison of
inertial mass is a special case where the initial velocities of the colliding bodies must
be the same. Sommerfeld [12] gives essentially the same definition by the reverse
process: “Impulse means (with regards to direction and magnitude) that kick, which
is capable of generating a given state of motion from the initial state of rest.”
For quantification we want to generate that kick from a certain number k of
congruent standard kicks. In physics didactics one uses the example of a cannon
368 B. Hartmann
which is shooting snowballs at an object. With every shot the velocity of the object
decreases. One counts the momentum units until the object stops. This gives an
intuitive idea of a direct impulse measurement. We define suitable reference objects
for these observables. We provide standard bodies with a standard velocity. Then we
construct such a collision machinery: The generic particle collides with a swarm of
standard momentum carriers. We count how many impulse units one must connect
in order to reproduce the same impact. Then we can quantify the impact of any given
object O as a number k times the impulse of our reference bodies. Formally we
write the measurement result p[O] = k · pu . Similarly we determine the mass and
velocity by independent measurement procedures. A quantitative description of the
collision machinery leads to the basic equation between momentum, inertial mass
and velocity. We derive the momentum-velocity relation.
3.1 Reference Process
We define standard bodies and springs as sufficiently constant reference objects for
“impact” and “capability to do work”. Consider a reservoir of standard bodies with
the same mass (Note that we can test this through head-on collisions.) and standard
springs S with the same capability to do work (We can test this by catapulting stan-
dard bodies). For our measurements in entire mechanics we refer to two elementary
standard processes:
1. Let a particle pair of two standard bodies with the same mass m 1 = m 2 and
standard velocities ±vS in opposing directions compress a spring that stays at
rest (Fig. 4 t1 ).
2. Let the compressed spring that remains at rest catapult two initially resting stan-
dard objects into opposite directions (Fig. 4 t2 ).
Hence standard particles with a standard momentum equal in magnitude but opposite
in direction compress a standard spring. Vice versa, the standard spring S turns
standard particles with a standard mass into standard momentum carriers. The initial
and final state of that inelastic collision (of irrelevant internal structure) are well-
defined.
If Alice couples the compression and decompression of her spring then the particle
pair, which initially flew towards each other, will instantly be catapulted apart (Fig. 4
from t0 to t2 ). Alice generates an eccentric elastic collision between bodies of the
same mass. If Bob drives by with the same horizontal velocity, he will see the same
process as an elastic transversal collision (Fig. 4 Bob). One particle kicks in from
below and rebounds antiparallel. The other particle moves on with the same velocity
v into a slightly deflected direction. The kinematics is well-defined by the symmetry
(collision of two equivalent bodies) and by the relativity principle (the view from a
moving observer). For any given initial velocity v and transversal impact velocity w
one determines, with Feynman’s trick [13], the deflection angle α
Fig. 4 t0 : A spring is compressed by particles moving in opposite directions. t1 : After the process
the compressed spring has capability to do work. t2 : Decompression of the charged spring. Alice
observes a symmetric collision. The same process appears as a transversal kick when Bob drives
by with the same horizontal velocity to the left

α vx2
1− c2
sin = ·w . (5)
2 v
Vice versa for a given angle one can determine the necessary velocity v.
3.2 Assemble Collision Models
From these standard kicks we assemble increasingly complex collision models. In

Fig. 5 we construct a reversion process, where 10 radial kicks with the same impact
reverse the motion of a particle, that comes in from the lower left side with a velocity
v2 . Similarly we construct a reversion process for another particle that comes in from
the opposite side (upper right) with velocity v1 . Let this particle be slower and require
only half the number of standard kicks. We choose the velocity v1 so that after one
kick with the same impact its motion is deflected by the same angle α1 = 2 · α2 as
before after two kicks for the faster particle.
We align the reversion processes in the depicted way so that all the temporary
activated recoil particles in the center can be captured again and recycled. In the net
result only the motion of the three incident particles (one from left and two from
right) is exactly reversed. We determine the relation between their velocities v1 and
v2 from matching the two building blocks, so that the total machinery functions.
Depending on the velocities v1 resp v2 we steer two types of radial kicks from the
outside (highlighted in Fig. 5). To leave no trace in the external reservoir all pairs of
370
B. Hartmann
Fig. 5 Align standard collisions

recoil particles must have the same radial velocities w := · v1 and are aligned at
diametrically opposed locations. Finally, in order to link the collision of one particle
from left and two particles from right we match the deflection angles α1 = 2 · α2 .
This leads, after substitution (5), to a relation between the initial velocities v1 and v2 .
One can generalize the construction for the collision of 1 + z particles in the same
way.
By refinement of the radial kicks one can construct a similar model for the elastic
head-on collision between two generic bodies: Let one standard body with mass m
(from left) and a rigid composite of z standard objects with equal masses m 1 = m 2 =
· · · = m z (from right) collide and rebound off of one another with reversed velocities.
We symbolize only the velocity changes of the standard particles
v(m) = u , v1 = · · · = vz = v ⇒ v(m) = −u , v1 = · · · = vz = −v .
Then in Poincare kinematics one can derive [14], that the initial velocities satisfy the
relation u v
(5)
= −z · . (6)
1 − vc2
2 2
1 − uc2
We do not presuppose how the velocities of two generic objects change in an elas-
tic collision. The trick is to mediate their direct interaction by steering a process
with an external reservoir. The model is built from one elementary collision process
between standard elements which must behave in a symmetrical way. From the geo-
metric layout we derive the relation (6) between the amount of matter and the impact
velocities.
3.3 Quantification
This gedanken experiment allows a direct impulse measurement. From everyday

experience one knows, what is meant by “impact” in a collision. With our collision
test we can specify the concept more precisely. We measure the momentum of particle
a directly with an aggregate of l standard impulse carriers that has the same impact.
In an inelastic head-on collision test this means that neither body overruns the other.
For our version of the collision test, which is reversible, the corresponding criterion
is that we catapult all incoming particles back into the reversed direction. In that
aggregate each impulse carrier represents the same standard impulse p1 . According
to the congruence principle
(direct) (Congr.)
p [a] = p1 + · · · + p1 =: l · p1
we measure “how many times” larger the momentum of particle a is than the impulse
p1 of one reference body.
372 B. Hartmann
In the same way we measure the inertia m of the body a with an equally massive
aggregate of standard elements m i and the latter according to the congruence principle
(direct) (Congr.)
m [a] = m1 + · · · + m1 =: z · m1
by the number z of standard elements and their unit mass m 1 .
3.4 General Collision Law
In the collision model we can count the coinciding numbers of standard elements.
When we build the model in Galilei kinematics we probe one standard particle
with the mass m and the velocity n · v1 . Then we find the same number n of impulse
units. Here we measure the momentum p[m] = n · p1 . For an aggregate of z standard
particles m i , i = 1, . . . , z with the same velocity one can conduct the collision model
z times. Thus we count that z times more impulse units have the same impact. We
measure the momentum p[m 1 , . . . , m z ] = z · n · p1 .
This leads to the general theorem: The particle a with mass m a = z · m 1 (m 1 a
standard mass) and velocity va = n · v1 (v1 the standard velocity) has a momentum
p [a] = z · n · p1 .
The equation between these physical quantities looks more familiar in the numerical
form
p [a] m a va
= · .
p1 m 1 v1
In Galilei kinematics the magnitude of the impulse equals the magnitude of the mass
times the magnitude of the velocity.
In Poincare kinematics we conduct the same direct impulse measurement. We
assemble the collision test model from well-defined radial kicks (Fig. 5). In the
relation (6) between the impact velocities w, vi and the particle number z appear
additional Lorentz terms γ. For a given particle a with the velocity v one counts the
number of impulse units, that generate the same impact. This leads to the relativistic
relation between momentum and velocity [14]
p [a] = {γ · m · v} · p1 .
4 Energy
We define energy as a basic observable which can be measured directly. In the

familiar context of working with machines one can identify the “source” that drives
Fig. 6 Direct measurement of kinetic energy: count compressed standard springs
the process and the “raw material” on which work is done. We regard energy as
a capability of one source or system to do work against another system. One also
knows what is meant if one source has more capability to do work (lift more weights,
grind more corn, generate more heat etc.) than another. Leibniz and Helmholtz give a
more precise physical definition. One source S has more energy than another source
T if, until complete exhaustion, the work of S exceeds the work of T on the same
test system [15]. Formally we write E [S] > E [T ].
We focus on the kinetic form of energy which is associated with slowing down the
motion of a body. When the body stops this form of energy is exhausted. According
to Papadouris, Constantinou [1] (p. 209) we measure the kinetic energy possessed by
a moving object by exploring “the extent of damage that it could cause” by colliding
with certain other objects. We construct a machinery which functions for a direct
measurement, a calorimeter model. The measurement process works as follows: A
calorimeter absorbs a given object, stops it and generates a swarm of standard energy
sources (Fig. 6). We will count how many standard springs S an incoming object with
vinitial can compress before it stops (vend = 0). Finally we derive the kinetic energy-
velocity equation.
4.1 Steering a Calorimeter Model
We illustrate the measurement process with an example. Consider the elastic head-on
collision (3.2) between one fast standard particle with the mass m and an aggregate
of 9 elements with the standard mass m i , i = 1, . . . , 9. For a drive-by observer the
incident particle kicks the resting aggregate into motion and rebounds with reduced
velocity to the left (Fig. 7 t1 ). From those deceleration kicks we build our calorimeter
model. On the left we place again a suitable number of 7 reservoir elements into
the way, such that they get kicked out with the same standard velocity. The incident
particle successively rebounds with reduced velocity (Fig. 7 t1 to t5 ) until it stops
inside the calorimeter. On the left side of the calorimeter we kick 7 + 3 reservoir
elements into motion. On the right side we generate 9 + 5 + 1 impulse carriers with
the same standard velocity. Thus, for a particle with velocity 5 · v we mobilize a total
of 25 initially resting reservoir elements. We kick 10 particle pairs with the same
374 B. Hartmann
standard velocity out of both sides of the calorimeter and 5 single recoil particles.
We symbolize the velocities in the initial and final state
v(m) = 5 · v , v1 = · · · = v25 = 0 ⇒
v(m) = 0 , 10 pairs with vi = v and v j = −v , 5 particles with vk = v .
(7)
When we absorb the same standard particle with a higher velocity, we have to couple
more suitable deceleration kicks, that happen before the five collisions t1 to t5 in
Fig. 7. We mobilize more reservoir elements and create more standard particle pairs
and impulse carriers.
4.2 Quantification
Now we interpret the model from a physical perspective. The direct comparison
methods specify the physical meaning of our reference objects as units for energy
and momentum. The standard impulse carrier m with v(m) = v S represents the unit
of momentum p1 . It also has an energy, namely one half of the energy of the stan-
dard spring, because one particle pair can compress one standard spring Sect. 3.1.
The calorimeter output (7) has the same capability to do work as the incident par-
ticle, because our calorimeter model is reversible. This is plausible because each
building block is essentially an elastic head-on collision and because one can steer
the complete process in both ways.
Hence the kinetic energy E[a] of the incident particle a is completely transformed
into potential energy of the absorber material; therein every spring S carries the same
standard energy. According to the congruence principle
(direct) (Congr.)
E [a] = E[ S
+ ·
· · + S ] =: k · E [S]
k times
we measure “how many times” larger its kinetic energy is than the potential energy
of one standard spring S. We quantify the energy as a certain number k of reference
units times the dimension of the reference objects.
4.3 Basic Equations
In the calorimeter model we can count the coinciding numbers of activated stan-
dard springs, reservoir elements, and velocity units. We will now derive the relation
between these physical quantities. This step leads to the basic dynamical equations.
In the example in Fig. 7 one can count the absorption effect for one standard
particle with the mass m and the velocity n = 5 times the standard velocity v. In
Fig. 7 An incident particle with the mass m comes to rest by a series of collisions with initially
resting elements with standard masses m i on the left and right side of a calorimeter. The deceleration
kicks at t1 to t5 bring particle m with velocity 5 · v to rest. We generate 7 + 3 particle pairs with
vi = v and v j = −v and 2 + 2 + 1 impulse carriers with vk = v
376 B. Hartmann
total we generate 25 standard impulse carriers with a velocity ±v and an energy

1
2
· E [S]. In total one can compress 25 · 21 springs. Thus we measure an energy
E [m] = 25 · 21 · E [S]. When we absorb an aggregate of z standard particles m i ,
i = 1, . . . , z with the corresponding velocities vi = n · v we count z · n 2 · 21 springs
and measure the energy E[m 1 , . . . , m z ] = z · n 2 · 21 · E [S].
This leads to the general theorem: The particle a with mass m a = z · m 1 (m 1 a
standard mass) and velocity va = n · v (v the standard velocity) has a kinetic energy
1
E [a] = · z · n 2 · E [S] .
2
When we express all numerical values for energy k =: E[S]

E[a]
, mass z =: mma1 , velocity
n =: vva1 in the form measure/unit measure we recover the familiar formulation
2
E [a] 1 ma va
= · · .
E [S] 2 m1 v1
For the proof we build the calorimeter in Galilei kinematics. Then for every step of
right and left collisions we can add the extracted energy-momentum carriers and the
successive deceleration. We derive the relation between these physical quantities in
[16].
When we build the same model in Poincare kinematics, then we saw that for the
individual deceleration kicks additional Lorentz terms appear (6). Now we construct
the deceleration-cascade with suitable fragments of standard particles, so that they all
get kicked out again with the same standard velocity on both sides of the calorimeter.
Then we can integrate all fragments of congruent energy and momentum units for
the entire deceleration and thus derive the basic dynamical equations

E kin [a] = m · c2 · (γ − 1) · E [S]
p [a] = {m · γ · v} · p1
and from them all the rest of relativistic dynamics (details see [14]).
5 Discussion
We introduce a novel strictly physical foundation of the measure of energy, momen-

tum and inertial mass without taking equations of motion etc. as a basis. This approach
draws on measurement principles of Galilei, Leibniz and Helmholtz and more recent
developments: Einstein’s relativity principle, light principle and the measurement
theoretical foundation of kinematics as well as the protophysical understanding of
the origin of reference devices and procedures. In the debate on the status of Ein-
stein’s clock postulate, of providing clocks and rods as unstructured entities, to give
empirical meaning to the assertions of the theory, Weyl represented the axiomatic
view (that measurement instruments ought to be regarded as solutions to differential
equations) while Pauli “denied the very possibility of applying a single conceptual
model to the theory and to the measurement instruments that verify it”; Einstein took
a stance in between [17]. The main idea of protophysics [6] is that physics does not at
the same time provide a theory of its measurement units. A theory of measurements
in the foundation of physics therefore also implies a systematization of measurement
conventions and norms, which the instruments must obey [15].
References
1. R. Duit. Teaching and learning of energy in K-12 education, ed. by R.F. Chen, A. Eisenkraft,
D. Fortus, J. Krajcik, K. Neumann, J. Nordine, A. Scheff (Springer, Heidelberg, 2014), pp. 153
2. D. Hammer, Student ressources for learning introductory physics. Am. J. Phys. 68, 52 (2000)
3. C. Gerthsen, H. Vogel, Gerthsen Physik (Springer, Berlin, 1995)
4. H.V. Helmholtz, Zählen und Messen, erkenntnistheoretisch betrachtet. Philosophische Vorträge
und Aufsätze, ed. by H. Hörz, S. Wollgast (Akademie Verlag, Berlin, 1971), pp. 109
5. B. Hartmann, Operationalization of Relativistic Motion (Kinematics) (2012), arXiv:1205.2680
6. P. Janich, Das Maß der Dinge: Protophysik von Raum Zeit und Materie (Suhrkamp, 1997)
7. A. Einstein, Grundzüge der Allgemeinen Relativitätstheorie (Springer, Berlin, 2002)
8. J. Wallot, Grössengleichungen Einheiten und Dimensionen (Johann Ambrosius Barth, Leipzig,
1952)
9. H. Hertz, Einleitung zur Mechanik. Zur Grundlegung der theoretischen Physik, ed. by R.
Rompe, H.-J. Treder (Akademie Verlag, Berlin, 1984), pp. 82
10. R.U. Sexl, H.K. Urbantke, Relativity, Groups, Particles - Special Relativity and Relativistic
Symmetry in Field and Particle Physics (Springer, Berlin, 2001)
11. H. Weyl, Philosophy of Mathematics and Natural Science (Princeton University Press, Prince-
ton, 1949), p. 139
12. A. Sommerfeld, Mechanik (Verlag Harry Deutsch, Thun, 1994), pp. 4
13. R.P. Feynman, R.B. Leighton, M. Sands, The Feynman Lectures on Physics - Mainly Mechanics,
Radiation and Heat (Addison-Wesley Publishing Company, Boston, 1977), pp. 4–1–4–8
14. B. Hartmann, Operationalization of Relativistic Energy-Momentum. Dissertation, Humboldt-
University, urn:nbn:de:kobv:11-100233941 (2015), pp. 115–138
15. O. Schlaudt, Messung als konkrete Handlung - Eine kritische Untersuchung über die Grund-
lagen der Bildung quantitativer Begriffe in den Naturwissenschaften (Verlag Königshausen &
Neumann, Würzburg, 2009)
16. B. Hartmann, Operationalization of Basic Observables in Mechanics (2015). arXiv:1504.03571
17. M. Giovanelli, But one must not legalize the mentioned sin: phenomenological vs. dynamical
treatments of rods and clocks in Einstein’s thought. Stud. Hist. Philos. Mod. Phys. 48, 20–44
(2014)
Can Spacetime Curvature be Used
in Future Navigation Systems?
Hernando Quevedo
Abstract We argue that the curvature generated by a gravitational field can be used to
calculate the corresponding metric which determines the trajectories of freely falling
test particles. To this end, we present a method to compute the metric from a given
curvature tensor. We use Petrov’s classification to handle the structure and properties
of the curvature tensor, and Cartan’s structure equations in an orthonormal tetrad
to investigate the differential equations that relate the curvature with the metric.
The second structure equation is integrated to obtain the explicit expression for
the connection 1−form from which the components of the orthonormal tetrad are
obtained by using the first structure equation. This opens the possibility of using the
curvature of astrophysical objects like the Earth to determine the position of freely
falling satellites that are used in modern navigation systems.
1 Introduction
One of the most important practical applications of general relativity is the Global
Positioning System (GPS), the most advanced navigation system known today. It
consists essentially in a set of artificial satellites freely falling in the gravitational field
of the Earth. To determine the location of any point on the Earth by using the method
of triangulation, it is necessary to know the exact position of several satellites at a
given moment of time. This means that the path of each satellite must be determined
as exact as possible. In fact, due to the accuracy expected from the GPS, specially
H. Quevedo (B)
Instituto de Ciencias Nucleares, Universidad Nacional Autónoma de México,
AP 70543, 04510 Mexico, DF, Mexico
e-mail: quevedo@nucleares.unam.mx
H. Quevedo
Dipartimento di Fisica and ICRANet, Università di Roma “La Sapienza”,
00185 Rome, Italy
H. Quevedo
Department of Theoretical and Nuclear Physics, Kazakh National University,
Almaty 050040, Kazakhstan

380 H. Quevedo
for navigation purposes, it is necessary to take into account relativistic effects for
the determination of the satellites trajectories and the gravitational field of the Earth.
This method is therefore essentially based upon the use of the geodesic equations of
motion for each satellite. Moreover, it is necessary to consider the fact that according
to special and general relativity clocks inside the satellites run differently than clocks
on the Earth surface. Indeed, it is known that not taking relativistic effects into account
would lead to an error in the determination of the position which could grow up to
10 kilometers per day.
The curvature seems to be an alternative way to determine the position of any
point on the surface of the Earth. Indeed, if we could measure the curvature of
the spacetime around the Earth, and from it the corresponding metric, one could
imagine that the determination of the position of the satellites could be carried out
in a different way. Maybe this method could be more efficient and more accurate. To
this end, it is necessary to measure the curvature of spacetime. Several devices have
been proposed for this purpose. The five-point curvature detector [1] consists of four
mirrors and a light source. By measuring the distances between all the components
of the detector, it is possible to determine the curvature. Another method uses a
local orthonormal frame which is Fermi-Walker propagated along a geodesic [2]. A
gyroscope is directed along each vector of the frame so that the relative acceleration
will allow the determination of the curvature components. The gravitational compass
[3] is a tetrahedral arrangement of springs with test particles on each vertex. Using
the geodesic deviation equation, from the strains in the springs it is possible to infer
the components of the curvature. More recently, a generalized geodesic deviation
equation was derived which, when applied to a set of test particles, can be used to
measure the components of the curvature tensor [4].
It seems therefore to be now well established that the curvature can be measured by
using different devices that are within the reach of modern technology. The question
arises whether it is possible to obtain the metric from a given curvature tensor. This
is the problem we will address in this work. In Sect. 2, we study a particular matrix
representation of the curvature tensor which allows us to calculate its eigenvalues in
a particularly simple way. Petrov’s classification is used to represent the curvature
matrix in terms of its eigenvalues. In Sect. 3, we use Cartan’s formalism to derive all
the algebraic and differential equations which must be combined and integrated to
determine the components of the metric from the components of the curvature. As
particular examples, we present the Schwarzschild, Taub-NUT and Kasner metrics
with cosmological constant. All the components of the metric are found explicitly in
terms of the components of the curvature tensor. It turns out that for a given vacuum
solution it is possible to find several generalizations which include the cosmological
constant.
Can Spacetime Curvature be Used in Future Navigation Systems? 381
2 Matrix Representation of the Curvature Tensor
There are several ways to represent and study the properties of the curvature tensor.
Here, we will use a method which is based upon the formalism of differential forms
and the matrix representation of the curvature tensor. The reason is simple. Imagine
an observer in a gravitational field. Locally, the observer can introduce a set of four
vectors ea to perform measurements and experiments. Although it is possible to
choose the direction of each vector arbitrarily, the most natural choice would be to
construct an orthonormal system, i.e., ea ⊗ eb = ηab = diag(+1, −1, −1, −1). Of
course, the observer could also choose a local metric which depends on the point.
Nevertheless, the choice of a constant local metric facilitates the process of carrying
out measurements in space and time. This choice is also in the spirit of the equivalence
principle which states that locally it is always possible to introduce a system in which
the laws of special relativity are valid. The set of vectors ea can be used to introduce
a local frame ϑa by using the orthonormality condition ea ϑb = δab , where is the
internal product. The set of 1-forms ϑa determines a local orthonormal tetrad that is
the starting point for the construction of the formalism of differential forms which
is widely used in general relativity.
There is an additional advantage in choosing a local orthonormal frame. General
relativity is a theory constructed upon the assumption of diffeomorphism invariance,
μ μ
i.e, it is invariant respect to arbitrary changes of coordinates x → x such
μ with
that J = det ∂x ∂x μ
= 0. Once a local orthonormal frame ϑa is chosen, the only

freedom which remains is the transformation ϑa → ϑa = aa ϑa , where aa is a

Lorentz transformation, satisfying the condition aa a b = ηab . This means that the
diffeomorphism invariance reduces locally to the Lorentz invariance, which is easier
to be handled.
In the local orthonormal frame, the line element can be written as
ds 2 = gμν d x μ ⊗ d x ν = ηab ϑa ⊗ ϑb , (1)
with
ϑa = eaμ d x μ . (2)
The components eaμ are called tetrad vectors, and can be used to relate tetrad com-
ponents with coordinate components. For instance, the components of the metric are
given in terms of the tetrad vectors by gμν = eaμ ebν ηab . The exterior derivative of the
local tetrad is given in terms of the connection 1−form ωab as [5]
dηab = ωab + ωba . (3)
Since the local metric is constant, the above expression vanishes, indicating that the
connection 1−form is antisymmetric. Furthermore, the first structure equation
dϑa = −ω ab ∧ dϑb , (4)

382 H. Quevedo
can be used to calculate all the components of the connection 1−form. Finally, the
curvature 2−form is defined as
ab = dω ab + ω ac ∧ ω cb (5)
in terms of a connection. In this differential form representation, the Ricci and Bianchi
identities can be expressed as
ab ∧ ϑb = 0 , dab + ω ac ∧ cb − ac ∧ ω cb = 0 , (6)
respectively.
The curvature 2−form can be decomposed in terms of the canonical basis ϑa ∧ ϑb
as
1
ab = R abcd ϑc ∧ ϑd , (7)
2
where R abcd are the components of the Riemann curvature tensor in the tetrad repre-
sentation.
It is well known that the curvature tensor can be decomposed in terms of its
irreducible parts which are the Weyl tensor [6]
1
Wabcd = Rabcd + 2η[a|[c Rd]|b] + Rηa[d ηc]b , (8)
6
the trace-free Ricci tensor
1
E abcd = 2η[b|[c Rd]|a] − Rηa[d ηc]b , (9)
2
and the curvature scalar
1
Sabcd = − Rηa[d ηc]b , (10)
6
where we use the following convention for the components of the Ricci tensor:
Rab = η cd Rcabd . (11)
Due to the symmetry properties of the components of the curvature tensor, it is

possible to represent it as a (6×6)-matrix by introducing the bivector indices A, B, . . .
which encode the information of two different tetrad indices, i.e., ab → A. We follow
the convention used in [5] which establishes the following correspondence between
tetrad and bivector indices
01 → 1 , 02 → 2 , 03 → 3 , 23 → 4 , 31 → 5 , 12 → 6 . (12)
This correspondence can be applied to all the irreducible components of the Riemann
tensor given in Eqs. (8)–(10). Then, the bivector representation of the Riemann tensor
reads
R AB = W AB + E AB + S AB , (13)
with

M N
W AB = , (14)
N −M

P Q
E AB = , (15)
Q −P

I3 0
S AB = − 12
R
. (16)
0 −I3
Here M, N and P are (3 × 3) real symmetric matrices, whereas Q is antisym-

metric.
We see that the bivector representation of the curvature is in fact given in terms of
the (3×3)-matrices M, N , P, Q and the scalar R, suggesting an equivalent represen-
tation in terms of only (3×3)-matrices. Indeed, since (13) represents the irreducible
pieces of the curvature with respect to the Lorentz group S O(3, 1) and, in turn,
this group is isomorphic to the group S O(3, C), it is possible to introduce a local
complex basis where the curvature is given as a (3×3)-matrix. This is the S O(3, C)-
representation of the Riemann tensor [6, 7]:
R = W + E +S, (17)
W = M + iN , (18)
E = P +iQ , (19)
S = 12
1
R I3 . (20)
In this representation, Einstein’s equations can be written as algebraic equations.

Consider, for instance, a vacuum spacetime for which E = 0 and S = 0. Then, the
vanishing of the Ricci tensor in terms of the components of the Riemann tensor
corresponds to the algebraic condition
Tr(W ) = 0 , W T = W . (21)
In general, from Einstein’s equations in the presence of matter
1
Rab − Rηab + ηab = −κTab , (22)
2
we find that
R = 4 + κT , T = η ab Tab , (23)
and the components of the curvature tensor satisfy the relationships

384 H. Quevedo
κ
η cd Rcabd = κTab + + T ηab . (24)
2
It is then easy to see that the following curvature tensor
1
S= (4 + κT ) diag(1, 1, 1) , (25)
12
⎛ ⎞
T − T00 + 21 T T12 − i T03 T13 + i T02
κ ⎝ 11
E= T12 − i T03 T22 − T00 + 21 T T23 − i T01 ⎠ , (26)
2
T13 − i T02 T23 + i T01 T33 − T00 + 21 T
W arbitrary (3 × 3) − matrix with Tr(W ) = 0 , W T = W , (27)
satisfies Einstein’s equations identically. Thus, we see that the matrix W has only
ten independent components, the matrix E is hermitian with nine independent com-
ponents and the scalar piece S has only one component.
The energy-momentum tensor determines completely only the trace-free Ricci
tensor and the scalar curvature. The Weyl tensor contains in general ten independent
components. However, since the local tetrad ϑa is defined modulo transformations
of the Lorentz group S O(3, 1), we can use the six independent parameters of the
Lorentz group to fix six components of the Weyl tensor. Accordingly, we can use the
eigenvalues of the matrix W to write the four remaining parameters in the form
⎛ ⎞
a1 + ib1
W =⎝ ⎠ .
I
a2 + ib2 (28)
−a1 − a2 − i(b1 + b2 )
In fact, this is the most general case of a Weyl tensor, and corresponds to a type
I curvature tensor in Petrov’s classification. If the eigenvalues of the matrix W are
degenerate, then a2 = a1 = a and b2 = b1 = b and therefore
⎛ ⎞
a + ib
W =⎝ ⎠ ,
D
a + ib (29)
−2a − 2ib
which represents a type D curvature tensor.

In general, all the eigenvalues can depend on the coordinates x μ of the spacetime.
The real part of the eigenvalues a1 and a2 represent the gravitoelectric part of the
curvature, whereas the imaginary part b1 and b2 correspond to the gravitomagnetic
field, i.e., the gravitational field generated by the motion of the source.
3 Integration of Cartan’s Structure Equations
Our aim now is to show that for a given curvature tensor it is possible to integrate
Cartan’s equation in order to compute the components of the metric. To this end, it is
necessary to rewrite Cartan’s equations so that the dependence on the spacetime coor-
dinates becomes explicit. First, let us introduce the components of the anholonomic
connection abc by means of the relationship
ω ab = abc ϑc , (30)
and the condition abc = −bac . Then, from the definition of the connection 1−form,
we obtain
ea[μ,ν] = abc e[νb eμ]c , (31)
which represents a differential equation for the components of the tetrad vectors eaμ .
Here, the square brackets denote antisymmetrization. On the other hand, the exterior
derivative of the curvature 2−form yields
1
a
dab = R bcd,μ eeμ + 2R ab f d fec ϑe ∧ ϑc ∧ ϑd , (32)
2
which together with
1 f
dab = R bed af c ϑe ∧ ϑc ∧ ϑd , (33)
2
leads to the following equation
R ab[cd,|μ| eeμ = R af [cd |b|e] − R ab[cd | f |e] − 2R ab f [c de] .

f f f
(34)
This equation represents an algebraic relationship between the components of the

tetrad vectors eaμ and the components of the connection 1−form abc .
Finally, the components of the curvature tensor can be expressed in terms of the
anholonomic components of the connection as
1 a
R = ab[d,|μ| ecμ + abe e[cd] + ae[c e|b|d] , (35)
2 bcd
which can be considered as a system of partial differential equations for the compo-
nents of the connection with the components of the curvature and the tetrad vectors
as variable coefficients.
To integrate Cartan’s equations we proceed as follows. First, we consider the 20
particular independent equations (35) together with the 18 equations which follow
from Eq. (34). The idea is to obtain from here all the 24 anholonomic components
of the connection abc . Then, this result is used as input to solve the 24 independent
386 H. Quevedo
equations which follow from Eq. (31). This procedure leads to a large number of
equations which are complicated to be handled. They have been analyzed with some
detail in [7]. Here, we will limit ourselves to quoting the some of the final results
obtained previously.
4 Type D Metrics
Consider a type D curvature tensor with eigenvalue a + ib, and suppose that
a = a(x 3 ) , b = b(x 3 ) , (36)
i.e., we assume that the curvature depends on only one spatial coordinate. Further-
more, it is well known that type D spacetimes can have a maximum of four Killing
vector fields. Then, we will consider spacetimes with two Killing vector fields which
can be taken along the coordinates x 0 and x 1 ; consequently,
∂gμν
gμν,0 = gμν,1 = 0 , gμν,0 = . (37)
∂x 0
This means that the only relevant spatial direction should be x 3 . Therefore, we can
use the diffeomorphism invariance of general relativity in order to bring four metric
components into any desired form. We then assume that
g30 = g31 = g32 = 0 , g33 = g33 (x 3 ) . (38)
In terms of the local tetrad, the above assumption implies that

√
ϑ3 = g33 d x 3 = e33̇ d x 3 , (39)
where the dot denotes coordinate indices. As a consequence we have that
dϑ3 = 0 , (40)
which implies that six components of the tetrad vectors vanish, namely,
e03̇ = e13̇ = e23̇ = e30̇ = e31̇ = e32̇ . (41)
This means that we now have a system of only ten components of eaμ that are
unknown. On the other hand, the vanishing of the exterior derivative of ϑ3 implies
that
3[ab] = 0 , (42)
which drastically simplifies the set of differential equations for the components of
the connection. A detailed analysis of the resulting equations shows that it is conve-
nient to consider particular cases which are obtained for different choices of some
components of the connection. In fact, it turns out that the choices
121 = 0 , 123 = 0 (43)
and
121 = 0 , 123 = 0 (44)
lead to completely different solutions which we will analyze in the following sub-
sections.
It is then possible to show that with these simplifying assumptions, we can inte-
grate the set of partial differential equations. Several arbitrary functions arise in the
tetrad vectors which can then be absorbed by means of coordinate transformations.
4.1 Schwarzschild and Taub-NUT Metrics
The particular choice

121 = 0 , 123 = 0 (45)
leads to a compatible set of algebraic and differential equations which allow us to

calculate all the components of the tetrad vectors. We present the final results without
the details of calculations which can be consulted in [7].
Consider, for instance, the following curvature tensor in the S O(3, C) represen-
tation:
M
R = − 3 diag(1, 1, −2) + diag(1, 1, 1) , (46)
r 3
where r = x 3 . Then, the integration of all the differential equations yields

−1/2
2M
e33̇ α − − r2 , e22̇ = r , e12̇ = r F 12̇ , (47)
r 3
1/2 1/2
2M 2M
e0ṁ = C 0ṁ α − − r2 , e02̇ = F 02̇ α − − r2 , (48)
r 3 r 3
where m = 0, 1, α and C 0ṁ are arbitrary real constants and F 02̇ and F 12̇ are non-zero
functions of the coordinate x 2 . It is then possible to find a coordinate system in which
the above tetrad vector components lead to the line element
388 H. Quevedo

2M 2 dr 2
ds = α −
2
− r dt 2 − − r 2 (dθ2 + sin2 θdφ2 ) , (49)
r 3 α − r − 3 r 2
2M
which represents the Schwarzschild-de-Sitter spacetime.

Consider now a curvature tensor with gravitoelectric and gravitomagnetic com-
ponents:
M +iP
R=− diag(1, 1, −2) + diag(1, 1, 1) , (50)
(r + iC)3 3
where P and C are arbitrary real constants. It is then possible to show that the result
of the integration leads to a line element of the form

dr 2 dθ2
ds 2 = 1 (dt + 2C cos θdφ)2 − − (r 2 + C 2 ) 2 sin2 θdφ2 + , (51)
1 2
with −1
P 2
1 = (r + C )
2 2
(r − C 2 ) − 2Mr − (r 2 + C 2 )2 , (52)
C 3
P 4
2 = + C 2 . (53)
C 3
Different choices of the parameters P and C lead to different particular solutions
of Einstein’s equations. For instance, the choice

4 2
P = l 1 − l , C =l (54)
3
corresponds to the Taub-NUT metric with cosmological constant [8], where l is the
NUT parameter. Furthermore, the choice

4
P = kl 1 − l 2 , C = l , k = −1, 0, +1 (55)
3
is known as the Cahen-Defrise spacetime [9].

The Taub-NUT metric is obtained for the choice P = l and C = l with = 0.
It is then possible to obtain several different generalizations which include the cos-
mological constant. In fact, the simplest choice corresponds to P = C = l and the
cosmological constant entering only the scalar part of the curvature. Other general-
izations are obtained by choosing the free parameter P as a polynomial in , for
instance,

C = l , P = l c1 + c2 l 2 + c3 2 l 4 + · · · , (56)
where c1 , c2 , etc. are dimensionless constants. Another example is obtained for the
choice

P = l , C = l c1 + c2 l 2 + c3 2 l 4 + · · · . (57)
All these examples generalize the Taub-NUT metric to include the cosmological
constant. In principle, all of them should represent different physical configurations
since they all differ in the behavior of the Weyl tensor. This opens the possibility of
analyzing anti-de-Sitter spacetimes which are equivalent from the point of view of
the scalar curvature, but different from the point of view of the Weyl curvature.
We conclude that in the particular case analyzed here the method presented above
can be used to generate new solutions of Einstein’s equations with cosmological
constant.
4.2 Generalized Kasner Metrics
Another particular choice of the connection components given by
121 = 0 , 123 = 0 (58)
leads to a set of algebraic and differential equations which can be integrated com-
pletely for a curvature tensor with only gravitomagnetic components, i.e.,

R = a(x 3 ) diag(1, 1, −2) + diag(1, 1, 1) . (59)
3
Indeed, after applying a series of coordinate transformations, the corresponding line
element can be expressed as
|a0 | (a0 )2
ds 2 = dt 2 − dr 2
|3a0 − 2 |2/3 2|a0 |(3a0 − 2 )
2
− 1
(d X 2 + dY 2 ) , (60)
|3a0 − 2 |2/3
with

a0 = a + <0, (61)
2
and the prime represents derivation with respect to r = x 3 . Here we see that the metric
can be calculated immediately from the gravitoelectric component of the curvature
a(r ). Several particular metrics can be written down. We quote only the metric that
follows from the eigenvalue
γ
a = − 2β − , (62)
r 3
where γ and β are real constants. For this case, we obtain
390 H. Quevedo

−1 2(β−1) 2
ds 2 = γ − 6 r 2β r −2β/3 dt 2 − 29 β 2 γ − 6 r 2β r dr
−r 4β/3 (d X 2 + dY 2 ) . (63)
In the limiting case = 0, we obtain for each value of β a particular case of the
Kasner metric [10]. In general, the above line element represents a generalization
of the Kasner space which includes the cosmological constant. We see that in this
particular case we have chosen a curvature eigenvalue which contains the cosmo-
logical constant explicitly. This has been done in order to obtain a simple expression
for the Kasner metric with . However, one can always change in the function a(r )
the term containing the cosmological constant, in order to obtain different solutions.
The simplest spacetime would correspond to the one in which the Weyl tensor does
not depend on the cosmological constant, i.e.,
γ
a=− , (64)
r 2β
for which we obtain the generalized Kasner metric

γ
− r 2β + 2 2β 2 γ 2
ds =
2

3γ
2/3 dt 2

−
γ

2 dr
2
− r 2β + r 2(2β+1) − 2β + 2 − 3γ +
r 2β

2
r
− 3γ
1
2/3

d X + dY 2 . (65)
− r 2β +
This particular choice seems to be more complicated than the solution (63); however,
from a physical point of view it corresponds to the simplest choice in which the Weyl
tensor is not affected by the presence of the cosmological constant.
We see that it is possible to obtain several generalizations of the Kasner metric
with cosmological constant and, in principle, each of them should correspond to a
different physical configuration.
5 Conclusions
In this work, we presented a method to calculate the components of the metric tensor
from the components of the Riemann curvature tensor. We use the formalism of
differential forms and Cartan’s structure equations in order to calculate explicitly
the algebraic and differential equations that relate the components of the local tetrad
vectors with the components of the connection 1−form and the curvature 2−form.
We integrate the differential equations for the case of a type D curvature tensor in
Petrov’s classification which is characterized by only one complex eigenvalue. We
found that for a given curvature eigenvalue, it is possible to obtain different metrics,
depending on some assumptions made for the components of the connection 1−form.
For the computation of explicit examples, we assume that the curvature eigenvalue
depends on only one spatial coordinate. This simplifies the set of differential equa-
tions and allows us to carry out the integration completely. We obtain as concrete
examples two classes of spacetimes. The first class contains the Schwarzschild met-
ric, the Taub-NUT metric and several generalizations which include the cosmological
constant. The second class contains a family of particular Kasner spacetimes with
cosmological constant.
The main result of the present work is that is possible to obtain the metric from
the curvature. Furthermore, we found that for any given vacuum spacetime, we can
apply the procedure presented in this work to obtain different generalizations which
include the cosmological constant. This means that solutions of Einstein’s equations
with cosmological constant are not unique. The main physical difference between
different spacetimes with cosmological constant is reflected in the Weyl tensor which
behaves differently for each metric.
The concrete examples of curvature analyzed in this work involve terms with
gravitoelectric monopole (mass parameter) and gravitomagnetic monopole (NUT
parameter) only. In the case of an astrophysical gravitational source, a more realistic
situation involves higher mass and angular momentum multipole moments. It is then
easy to see that if we consider the Weyl tensor in the form
∞
mn
W =− diag(1, 1, −2) , (66)
n=1
r 2n+1
the integration of the structure equations can be performed in a way similar to the
one used to obtain the Schwarzschild and the Taub-NUT metrics. The explicit metric
components can be computed by using the general formula presented here and in [7].
The resulting metric will contain the parameters m n which correspond to higher mass
multipole moments. In this way, one could generate exact solutions with a prescribed
set of multipoles. In a realistic situation, for instance in the case of the Earth, one
would need only a limited number of moments m n , whose values can be from the
measurement of the curvature components.
To take into account higher gravitomagnetic moments, it will be probably nec-
essary to generalize the method presented here. Indeed, the presence of rotational
moments implies that the curvature must depend on at least two spatial coordinates (a
radial and an angular coordinate). In addition, it will probably necessary to consider
not only type D, but also type I Weyl tensors. In this case, we need to construct
a more general method than the one presented here. However, if we fix the angu-
lar coordinate and consider, for instance, the equatorial plane of the gravitational
source, the curvature will depend only on the radial distance and it will be possible
to consider a Weyl tensor of the form
∞
m n + i jn
W =− diag(1, 1, −2) , (67)
n=1
r 2n+1
392 H. Quevedo
where the parameters jn represent the multipoles of the curvature generated by the
rotation of the source. Of course, this would be only an approximation of a real-
istic compact object since the dependence on the angular coordinate is completely
neglected. However, since in the case of an object like the Earth, the deviations
from spherical symmetry due to rotation are very small, one could expect that this
equatorial plane approximation would lead results with a good degree of accuracy.
We conclude that the method presented in this work can be used, in principle,
to generate particular metrics, describing the gravitational field of realistic compact
objects. It would be interesting to investigate this problem in detail in the case of
the Earth, to study the possibility of developing new navigation systems by using as
input the curvature of the spacetime around our planet.
Acknowledgements This work has been supported by the UNAM-DGAPA-PAPIIT, Grant No.
IN111617.
References

2. F.A.E. Pirani, Acta Phys. Pol. 15, 389 (1956)
3. P. Szekeres, J. Math. Phys. 6, 1387 (1965)
4. D. Puetzfeld, Y.N. Obukhov, Phys. Rev. D 93, 044073 (2016)
5. K.S. Thorne, C.W. Misner, J.A. Wheeler, Gravitation (W. H. Freeman, San Francisco, 1973)
6. R. Debever, J. Geheniau, Bull. Acad. R. Soc. Belg. 42, 114 (1956)
7. H. Quevedo, Gen. Relativ. Gravit. 24, 693 (1992)
8. M. Demiański, Phys. Lett. A 42, 157 (1972)
9. M. Cahen, L. Defrise, Commun. Math. Phys. 11, 16 (1968)
10. M. MacCallum, C. Hoenselaers, H. Stephani, D. Kramer, E. Herlt, Exact Solutions of Einstein’s
Field Equations (Cambridge University Press, Cambridge, 2003)
World-Line Perturbation Theory
Jan-Willem van Holten
Abstract The motion of a compact body in space and time is commonly described
by the world line of a point representing the instantaneous position of the body. In
General Relativity such a world-line formalism is not quite straightforward because
of the strict impossibility to accommodate point masses and rigid bodies. In many sit-
uations of practical interest it can still be made to work using an effective hamiltonian
or energy-momentum tensor for a finite number of collective degrees of freedom of
the compact object. Even so exact solutions of the equations of motion are often not
available. In such cases families of world lines of compact bodies in curved space-
times can be constructed by a perturbative procedure based on generalized geodesic
deviation equations. Examples for simple test masses and for spinning test bodies
are presented.
1 Test Bodies in General Relativity
The newtonian theory of gravity is a theory of instantaneous action at a distance,

which is consistent with the concept of absolute time and absolute simultaneity. This
allows for the existence of rigid bodies. Taking that for granted Newton proved in the
Principia that the orbital motion of a homogeneous spherical rigid body is correctly
represented by that of a point mass located at the center of gravity. Thus he was able
to explain the motion of the moon orbiting the earth. Also in more general situations
the motion of a rigid body can be represented by a single curve: its world line,
identified as the orbit in space and time of the center of gravity, whilst the remaining
kinematical degrees of freedom are restricted to rotational motion of the body about
the center of gravity and specify the orientation of the body at every point of the
world line. In this frame work there is no obstacle to consider the limit of a very
small rigid body of which the gravitational influence on the motions of other bodies
is negligeable. Such a body which probes the gravitational field without disturbing
it is commonly referred to as a test body. It is characterized by a finite number of
J.-W. van Holten (B)

Nikhef, Science Park 105, 1098 XG Amsterdam, The Netherlands
e-mail: v.holten@nikhef.nl

394 J.-W. van Holten
physical degrees of freedom as its motion is completely specified in terms of its

world line plus orientational degrees of freedom.
In General Relativity (GR) the situation is more subtle, as strictly point-like par-
ticles cannot be accommodated in the theory. Most importantly any object of finite
non-zero mass m has an associated Schwarzschild radius
2Gm
ρS = ,
c2
such that if the size of the object shrinks below this scale it becomes a black hole
with a finite surface area equal to A = 4πρ2S for spherical bodies. Thus any massive
body is an extended body, at least in classical GR. In situations where quantum
effects become relevant this may change, but then there are other limitations like the
Compton wave length opposing complete localization of objects.
If a classical object also possesses an internal angular momentum (spin) there
are additional complications. First of all in GR rigid bodies can not exist and there
is no unique, observer-independent center of mass. Indeed as there is no absolute
simultaneity the relative positions of different particles composing the body at any
fixed time depend on the state of motion of the body with respect to the observer.
Some elementary considerations showing this state of affairs are discussed for a
simple two-body system in Appendix A. In addition, in a relativistic context for a
composite system it is more appropriate to discuss the motion of a center of energy
or centroid rather than a center of mass, although such a concept is still observer-
dependent. Actually it has been shown [1–3] that under reasonable assumptions the
world lines of all possible centroids of an object with mass m and spin s fill a time-like
oriented tube of radius
s
ρM = .
mc
The upshot of this discussion is that strictly speaking in GR no unique world line
can be associated with the motion of massive bodies and any particular choice of
representative world line is at least in part a matter of convenience and requires
careful specification.
Nevertheless there are circumstances in which the localization of a compact body
is possible with sufficient accuracy that for practical purposes it may be regarded
as a mass point moving along a world line. Moreover if its mass is small enough
that one can neglect the associated space-time curvature and its influence on other
bodies, one can still regard such an approximate point mass as a test body probing
the space-time geometry in which it moves.
In the context of GR the space-time geometry and the motion of compact bodies
are linked by the Einstein equations1
G μν + 8πG Tμν = 0, (1)
1 Here and in the following we use natural units in which the speed of light c = 1.
World-Line Perturbation Theory 395
where the Einstein tensor G μν is specified by the space-time geometry, and the
energy-momentum tensor Tμν describes the physical degrees of matter. Irrespective
of the precise background geometry the Bianchi identities for the Einstein tensor
require the energy-momentum tensor to be divergence-free:
∇μ T μν = 0. (2)
For a compact body with mass m moving as an approximate point mass along a world
line ξ μ (τ ) parametrized by the proper time τ the effective energy-momentum tensor
is [4]

m
μν
T (x) = √ dτ ξ˙μ ξ˙ν δ 4 (x − ξ(τ )), (3)
−g
with the overdot denoting a derivative w.r.t. proper time.2 Its divergence vanishes if
the world line is a geodesic:

m
∇μ T μν
=√ dτ ξ¨ν + μλν (ξ)ξ˙μ ξ˙λ δ 4 (x − ξ(τ )) = 0. (4)
−g
Neglecting the back reaction of the compact body is allowed if the contributions of
the test body to the geometry of space time are too small to be of interest. As an
illustration take the local background geometry to be that of flat Minkowski space-
time:
(0) (0) ν
gμν = ημν and μλ = 0. (5)
In particular we can then choose to work in the local inertial frame in which the body
is at rest:
(0)
T00 = mδ 3 (x), Ti0(0) = Ti(0)
j =0 i, j = (1, 2, 3). (6)
Taking into account these energy-momentum source terms, the solution of the Ein-
stein equation is modified to first order to read
(1)
gμν = ημν + h μν , (7)
where the correction term satisfies the linearized Einstein equation

h μν − ∂μ ∂ λ h λν − ∂ν ∂ λ h λμ + ∂μ ∂ν h λλ − ημν h λλ − ∂ κ ∂ λ h κλ
(8)
(0)
= −8πG Tμν .
Removing gauge degrees of freedom by the De Donder conditions
2 In
this paper the delta function is defined as a scalar density of weight 1/2 such that
d 4 y f (y) δ 4 (y − x) = f (x).
1
∂ μ h μν = ∂ν h μμ ,
2
the Einstein equation simplifies further to

1
h μν − ημν h λλ = −16πG Tμν
(0)
, (9)
2
which has the solution

2Gm
h i0 = 0, h i j = δi j h 00 , h 00 = . (10)
r
With this correction we obtain the modified line element

(1) μ ν 2Gm 2Gm
gμν d x d x = − 1 − dt + 1 +
2
dr2 . (11)
r r
It follows that the geometry near the test body deviates strongly from flat space on
the scale of its Schwarzschild radius, and it can be considered as a near point-like
object only as long as the external curvature is comparatively small on this scale.
Actually to first order in G the geometry specified by the line element (11) coin-
cides with that of Schwarzschild space-time in isotropic co-ordinates:
2 4
μ ν 2r − Gm Gm
gμν d x d x = − dt + 1 +
2
dr2 . (12)
2r + Gm 2r
In hindsight it is not such a surprise that the test-particle approximation is the limit of
a black-hole geometry, as it is well-known that the standard black-hole space-times
are exact vacuum solutions of the Einstein equations characterized by a finite number
of parameters like mass, spin and electric charge, in which respect they resemble test
bodies. The difference is of course that their action on the space-time geometry has
been fully taken into account to the extent that –in contrast to test particles– their
energy-momentum tensor has been absorbed completely in the space-time curvature,
i.e. it has become part of the Einstein tensor in Eq. (1).
2 The Motion of Test Bodies
Restricting our considerations to situations where the internal degrees of freedom

of a compact body do not take part in its gravitational interaction and its size and
gravitational back reaction can be neglected, its motion can be represented adequately
by a world line which is a time-like geodesic of the external space-time geometry.
This is the simplest case of the test-body approximation. The more elaborate test-
body limit of a compact body with spin will be discussed in Sect. 4 and later.
This description of a test body as an object with a finite numer of degrees of free-
dom to which one can assign at any moment a representative position has considerable
mathematical advantage: to such a body we can associate a finite-dimensional phase
space in which the evolution of the system is described by a simple curve. This orbit
is generated by a hamiltonian H depending on a finite number of phase-space vari-
ables including position and momentum. As by construction it neglects the finite size
and gravitational back reaction of the body, it is of course an effective hamiltonian,
its validity restricted by the test-body limit.
The effective hamiltonian dynamics of a massive test body with no other degrees
of freedom, like spin or charge, and an energy-momentum tensor of the form (1) is
straightforward to construct. The phase space is spanned by the position variables
ξ μ (τ ) and the momentum variables πμ (τ ), with the usual equal proper-time canonical
Poisson brackets
{ξ μ , πν } = δνμ . (13)
The free hamiltonian

1 μν
H= g [ξ] πμ πν (14)
2m
then generates the equations of motion
1
1
ξ˙μ = {ξ μ , H } = g μν πν , π̇μ = πμ , H = ∂μ g νλ πν πλ . (15)
m 2m
By simple algebra these equations can be rewritten in the form
πμ = mgμν ξ˙ν , ξ¨μ + λν ξ˙λ ξ˙ν = 0,

μ
(16)
reproducing the geodesic equation derived earlier from the energy-momentum tensor
(1).
In this hamiltonian frame work it is easy to find constants of motion, simplify-
ing the solution of the geodesic equation. There is at least one universal constant
independent of the specific metric gμν , the hamiltonian itself:
m
H =− ⇔ gμν ξ˙μ ξ˙ν = −1. (17)
2
It establishes the usual relation between proper time and co-ordinate time. Other
constants of motion depend on the symmetries of the background space-time, as
implied by Noether’s theorem. For example a quantity
J [ξ, π] = αμ [ξ] πμ (18)
is a constant of motion if αμ is a Killing vector, representing an isometry of the

metric:
J˙ = {J, H } = 0 ⇔ ∇μ αν + ∇ν αμ = 0. (19)
As an example in static space-times the kinetic energy E of the test body is a constant
of motion:
αμ = (−1, 0, 0, 0) ⇒ E = −πt . (20)
In particular in Minksowki space-time:
dt
E = −m ηtt ξ˙t = m = γm, (21)
dτ
where γ is the time-dilation factor.
In spherically symmetric geometries, such as Minkowski, Schwarzschild and
Freedmann–Lemaitre type cosmological space-times, all three components of angu-
lar momentum are conserved:
J1 = − sin ϕ πθ − cotan θ cos ϕ πϕ ,
J2 = cos ϕ πθ − cotan θ sin ϕ πϕ , (22)
J3 = πϕ .
Here the momenta (πr , πθ , πϕ ) are defined w.r.t. a polar co-ordinate frame (r, θ, ϕ)
such that the hamiltonian is given by

θθ
πϕ2
2m H = −g tt
(t, r )πt2 +g rr
(t, r )πr2 + g (t, r ) πθ2 + . (23)
sin2 θ
Clearly if the metric components gμν are time independent the system is both static
and spherically symmetric, and energy and angular momentum are all conserved.
This holds in particular for Minkowski and Schwarzschild space-time.
The standard metric of Schwarzschild space-time in Droste co-ordinates for an
object with mass M can be obtained from the isotropic co-ordinate representation
(12) by the co-ordinate transformation

GM 2
r̄ = r 1 + , (24)
2r
reproducing the static, spherically symmetric line element

μ ν 2G M d r̄ 2
gμν d x d x = − 1 − dt 2 + + r̄ 2 dθ2 + sin2 θ dϕ2 . (25)
r̄ 1 − r̄
2G M
Replacing r̄ → r we can then write the hamiltonian for a test body in Schwarzschild
space-time as

πt2 2G M 1 πϕ2
2m H = − + 1− πr + 2 πθ +
2 2
, (26)
1 − 2Gr M r r sin2 θ
which is a special instance of a static hamiltonian of the type (23). The conservation
of kinetic energy then holds in the form

E 2G M dt
ε≡ = 1− . (27)
m r dτ
Rotating the co-ordinate system such that the plane of the orbit is at constant θ = π/2,
we have J1 = J2 = πθ = 0 and
J3 dϕ
≡ = r2 . (28)
m dτ
In addition the universal constant of motion H = −m/2 implies that
2
dr 2G M 2
= ε2 − 1 − 1+ 2 . (29)
dτ r r
In particular there are circular orbits r = R = constant for which

dϕ GM 1
= 2 = ,
dτ R R3 1 − 3GRM
(30)
dt ε 1
= = .
dτ 1 − 2GRM 1 − 3GRM
Observe that these equations reproduce Keplers third law for circular orbits:

dϕ GM 4π 2 3
= ⇒ T2 = R ,
dt R3 GM
where T is the orbital period measured by the clock of a distant observer keeping
co-ordinate time t, and the radial co-ordinate R is determined by the orbital circum-
ference through the relation L = 2π R.
3 Geodesic Deviations
In most space-time geometries general exact solutions of the geodesic equation are
difficult to obtain, and when they are available they are often expressed in terms of
non-elementary transcendental functions [5]. However given one particular geodesic
curve it is possible to find approximate solutions for arbitrary nearby geodesics using
the geodesic deviation equation and its higher-order generalizations [6, 7]. This
procedure can also be explained in terms of the Schild ladder construction [8, 9], for
which next-to-leading order corrections can be obtained using the results reviewed
in this section.
In terms of the tangent vector u μ = ξ˙μ the geodesic equation (16) reads
μ
Dτ u μ = u̇ μ + λν u λ u ν = 0. (31)
Now consider a family of world lines
ξ μ (τ , σ) = ξ μ (τ ) + σn μ (τ ), (32)
obtained by a displacement of the geodesic in the direction of a vector field n μ (τ )

scaled by the real parameter σ. The corresponding changes in the world line and its
tangent vector are described covariantly by the vectors
ξ μ = σn μ , u μ = Dτ ξ ν = σ Dτ n μ . (33)
For this displacement to respect the geodesic equation (31) to first order in σ we
require
μ
Dτ u μ = Dτ u μ + [, Dτ ] u μ = σ Dτ2 n μ − Rλνκ u λ u κ n ν = 0. (34)
For example, we know all circular orbits in Schwarzschild space-time in analytical

form. Then we can construct approximate non-circular (eccentric) solutions in the
same plane by adding a geodesic deviation σn μ ; generically this is an oscillating term
as the test body moves periodically closer and farther from the central mass between
periastron and apastron. As the period of this perturbation is in general different from
the period of the circular orbit one starts from, the periastron and apastron will shift
their azimuthal directions each turn. This phenomenon is well-known ever since it
was calculated by Einstein for the planet Mercury in its orbit around the sun. It turns
out that for Mercury the first-order deviation (33) from motion on a circle with the
period of Mercury’s orbit actually already gives full numerical agreement with the
observed periastron shift of 43 arcseconds per century [6].
There is however no obstacle to include second and higher-order contributions in
σ to the displaced geodesics ξ μ (σ, τ ). As creates a covariant displacement in the
direction of n, but n is not necessarily displaced parallel to itself (it is not required
to be a tangent vector field to a family of geodesics crossing the world line ξ μ (τ )),
it follows that in general
2 ξ μ = σn μ ≡ σ 2 m μ = 0. (35)
The geodesic family ξ μ (τ , σ) is then parametrized to second order in σ by
μ 1 2 μ μ
ξ μ (τ , σ) = ξ1 (τ ) + σn μ (τ ) + σ m − λν n λ n ν + · · · (36)
2
Using the properties of the vector field n it is straightforward to generalize the
derivation of Eq. (34) and show that (n μ , m μ , ...) are solutions of a hierarchy of
deviation equations [6]
μ
Dτ2 n μ − Rλνκ u κ u λ n ν = 0,

μ μ μ
Dτ2 m μ − Rλνκ u κ u λ m ν = ∇ρ Rκνλ − ∇κ Rρλν u κ u λ n ρ n ν
(37)
μ
+4Rκνλ u κ n ν Dτ n λ ,
...
We note in passing, that whereas the first-order deviations n μ provide information

about the curvature of space-time through the Riemann tensor Rκνλμ , the second-
order deviations provide further information about the gradient ∇ρ Rκνλμ of the Rie-
mann tensor. Thus with sufficient knowledge of families of geodesics in some domain
of space-time one can reconstruct the Riemann tensor in the whole domain in terms
of a Taylor series from the geodesic deviations w.r.t. a given geodesic [10–12].
The procedure described here has been worked out for Schwarzschild space-time
up to and including the second-order deviations [6, 13, 14]. The results for orbits in
the equatorial plane θ = π/2 are summarized by the parametrized expansions
∞

t (τ , σ) = ξ0t (σ)τ + ξnt (σ) sin nω(σ)τ ,
n=1
∞

r (τ , σ) = ξ0r (σ) + ξnr (σ) cos nω(σ)τ , (38)
n=1
∞

ϕ
ϕ(τ , σ) = ξ0 (σ)τ + ξnϕ (σ) sin nω(σ)τ .
n=1
μ
The coefficients ξn (σ) and the fundamental frequency ω(σ) are computed order by
order in σ by solving the hierarchy of deviation equations (37). Of course the terms
of order σ 0 represent the parameters of the circular parent orbit we encountered in
Eq. (30):

1 ϕ GM 1
ξ0t (0) = , ξ0r (0) = R, ξ0 (0) = 3
. (39)
1− 3G M R 1 − 3G M
R R
All other terms have non-trivial dependence on the expansion parameter σ:
μ μ μ 1 2 μ
ξ0 (σ) = ξ0 (0) + σρ1 + σ ρ2 + · · · ,
2
μ μ 1 2 μ
ξ1 (σ) = σn 1 + σ n2 + · · · , (40)
2
μ 1 2 μ
ξ2 (σ) = σ m2 + · · · ,
2
μ
each ξn (σ) contributing only terms of order σ n and higher, whilst the angular fre-
quency
ω(σ) = ω0 + σω1 + · · · (41)
also depends on the order of approximation, reflecting the anharmonicity of the non-
μ μ μ
circular deviations. The explicit expressions for the coefficients ρn , n n and m n for
n = 1 and n = 2 and the expressions for the frequencies ω0 , ω1 are summarized in
Appendix B for the restricted case ρrn = 0. This restriction implies, that we compare
the non-circular orbits to a fixed circular orbit, although in general one might wish
to adapt the circular reference orbit order by order in σ to improve convergence. The
expressions for the unrestricted case can be found in Ref. [13].
Observe that the deviations are bound and periodic as long as the angular fre-
quency ω is real. For R < 6G M Eq. (77) in the appendix shows that it develops an
imaginary part, indicating exponential run-away behaviour. Thus the circular orbit
with R = 6G M is the innermost stable circular orbit (ISCO) for a simple test body
in Schwarzschild space-time.
Finally from the solutions of the geodesic deviations we can evaluate the constants
of motion ε and for the non-circular orbits by substitution of the solutions for dt/dτ
and dϕ/dτ including first and second order deviations into the expressions (27) and
(28). The resulting constants of motion are then also written as a series expansion in
the deviation parameter σ. Up to and including second order deviations the result is
1 2 1
ε(σ) = ε0 + σε1 + σ ε2 + · · · , (σ) = 0 + σ1 + σ 2 2 + · · · , (42)
2 2
where ε0 and 0 are the energy and angular momentum per unit of mass for circular
orbits:
R − 2G M G M R2
ε0 = √ , 0 = . (43)
R(R − 3G M) R − 3G M
In the restricted case ρr1 = 0 the first-order corrections vanish: ε1 = 1 = 0, but the
second-order corrections do not:
G M R 2 − 9G M R + 6(G M)2
ε2 = − ,
2R 7/2 (R − 3G M)3/2
√ (44)
3 G M (R − 2G M)(R − 7G M)
2 = − .
2R 2 (R − 3G M)3/2
The most important aspect of these expressions is that all dependence on the proper
time has disappeared and therefore they are true constants of motion indeed. This is
a strong consistency check on the results listed in Appendix B.
One of the useful aspects of the perturbative construction of orbits by the method
of geodesic deviations is that they provide explicit and accurate expressions for the
position of the test body as a function of the evolution parameter (proper time).
Most results in the literature describe the time-dependence of the orbits only in an
implicit way. The explicit time-dependent representation of the orbits of a test body
in Schwarzschild space-time is especially useful in the computation of the emission
of gravitational waves by extreme mass ratio (EMR) binaries: systems consisting of
a compact stellar object, like a white dwarf, neutron star or stellar-mass black hole
orbiting a supermassive black hole such as found in the center of many galaxies.
Gravitational wave emission by EMR binaries has been studied in detail in Ref. [14].
4 Spinning Test Bodies
The motion of electrically neutral, non-spinning test bodies is represented by

geodesics in space-time. As soon as new degrees of freedom come into play the
world line of a test body becomes non-geodesic. For example, an electrically charged
particle in the presence of electric or magnetic fields is subject to a Lorentz force
in addition to the effect of curvature. Its world line will differ from that of a similar
neutral particle which follows a geodesic.
Similarly the effect of rotation represented by internal angular momentum (spin)
will change the motion of test body in curved space-time even in the absence of other
fields of force [15, 16]. One can think of this as a form of gravitational spin-orbit
coupling. In this section we discuss the motion of spinning test bodies in more detail.
There is an extended literature on spinning test particles; for a review see e.g.
Ref. [17]. The most common approaches are based on a multipole expansion requiring
a dynamical constraint as an implicit choice of the center of mass fixing the monopole
term. As this choice is not unique we have developed in Refs. [18–20] a different
formalism which does not incorporate any constraints but only requires appropriate
initial conditions to solve the equations of motion. In this formulation the spin degrees
of freedom are described by a real anti-symmetric tensor field μν (τ ), which has six
components defined on the world line. Three of these components can be represented
equivalently by a space-like axial vector, a magnetic-type dipole moment which we
identify with the spin proper. Similarly the other three components are described
equivalently by a space-like real vector, an electric-type dipole moment which in the
gravitational context is interpreted as a mass dipole. Denoting the four-velocity of

the test body by the time-like unit vector u μ = ξ˙μ , we can decompose the spin tensor
μν accordingly as follows
1
μν = − √ εμνκλ u κ Sλ + u μ Z ν − u ν Z μ , (45)
−g
with Sμ and Z μ representing the proper spin and mass dipole components, respec-
tively. The decomposition (45) can be inverted to get
1√
Sμ = −g εμνκλ u ν κλ , Z μ = μν u ν . (46)
2
As mentioned both vectors are space-like by construction:
S · u = 0, Z · u = 0. (47)
Hence each contains only three degrees of freedom, their time components vanishing
in the rest frame u = (1, 0, 0, 0). From these expressions it is inferred directly that
under the group of three-dimensional spatial rotations and parity transformations in
the rest frame Z is a real vector, whereas S actually is an axial pseudo-vector.
To obtain equations of motion for a spinning body we follow the same route as for
a non-rotating body starting from an energy-momentum tensor and rederiving the
results from an effective hamiltonian. Following [19] the energy-momentum tensor
is taken to have support on a world line ξ μ (τ ) with

1
T μν
=m dτ u μ u ν √ δ 4 (x − ξ(τ ))
−g
(48)
1 1
+ ∇λ dτ u μ νλ + u ν μλ √ δ 4 (x − ξ(τ )).
2 −g
Its covariant divergence vanishes: ∇μ T μν = 0, if
μ 1 κλ μ ν
m Dτ u μ = u̇ μ + λν u λ u ν = Rκλ ν u ,
2 (49)
Dτ μν ˙ μν + u λ μ κν + u λ λκν μκ = 0.
= λκ
Alternatively in the hamiltonian formulation we introduce a set of classical Poisson-

Dirac brackets on the particle phase space defined by [18, 20]

1
{ξ μ , πν } = δνμ , πμ , πν = κλ Rκλμν ,
2
{ μν , πλ } = λκμ νκ − λκν μκ , (50)

μν , κλ = g μκ νλ − g μλ νκ − g νκ μλ + g νλ μκ .
These brackets have several important features. First, they define an algebra with
structure functions rather than structure constants. Moreover these structure func-
tions encode all of the usual space-time geometry: the metric, the connection and
the Riemann curvature tensor. In addition the properties of these geometric objects
guarantee that the Jacobi identities for cyclic triple brackets are satisfied by this
Poisson-Dirac algebra without specifying the dynamics of the system. Therefore the
bracket algebra is consistent independently of the choice of hamiltonian. In particular
the free hamiltonian (14):
1 μν
H= g [ξ] πμ πν ,
2m
generates the same equations of motion (49) as derived from the energy-momentum
tensor, providing in addition the identification of πμ with the kinetic momentum (16):
πμ = mgμν u ν .
To solve the equations of motion (49) it is again convenient to first identify constants
of motion. In this case there are 3 universal constants of motion, independent of the
specific geometry: the hamiltonian itself:
m
H =− ,
2
as long as it is proper-time independent; and two spin invariants:
1
I = gμκ gνλ μν κλ = S 2 − Z 2 ,
2
(51)
1√
D= −g εμνκλ μν κλ = S · Z .
8
Clearly I is a real scalar whilst D is a pseudoscalar under three-dimensional spatial
rotations and parity transformations.
In addition there may exist other constants of motion connected with symmetries
of the background space time. The construction in Eqs. (18), (19) based on the pres-
ence of Killing vectors can be generalized immediately to include spin [21, 22], as
follows: a quantity
1
J [ξ, π, ] = αμ [ξ] πμ + βμν [ξ] μν (52)
2
is a constant of motion if αμ is a Killing vector and βμν its gradient:
∇μ αν + ∇ν αμ = 0, βμν = −βνμ = ∇μ αν . (53)
Observe that the symmetric part of βμν vanishes by construction as a result of the
Killing condition; this property also implies the identity
∇λ βμν = Rμνλκ ακ , (54)
which is necessary to show that J is a constant of motion.
5 Spinning Test Bodies in Schwarzschild Space-Time
Applying the general formalism above to the case of Schwarzschild space-time we

first construct the generalization of the constants of motion (20) and (22), to wit the
kinetic energy based on the Killing vector of time-translation invariance:

G M tr 2G M GM
E = −πt − = m 1 − u t − 2 tr , (55)
r2 r r
and the total angular momentum based on the Killing vectors of invariance under
rotations:
J1 = − sin ϕ πθ − cotan θ cos ϕ πϕ
−r sin ϕ r θ − r sin θ cos θ cos ϕ r ϕ + r 2 sin2 θ cos ϕ θϕ ,
J2 = cos ϕ πθ − cotan θ sin ϕ πϕ (56)
+r cos ϕ r θ − r sin θ cos θ sin ϕ r ϕ + r 2 sin2 θ sin ϕ θϕ ,
J3 = πϕ + r sin2 θ r ϕ + r 2 sin θ cos θ θϕ .
As for the spinless test bodies we can orient the co-ordinate system such that the
z-axis coincides with the direction of the total angular momentum:
J = (0.0, J ), J = mr 2 u ϕ + r r ϕ . (57)
The total spin J is now composed of the contributions from orbital angular momen-
tum and from spin in the z-direction. The contributions in the perpendicular directions
must then cancel. This is expressed by the conditions
J
r θ = −mr u θ , θϕ = cotan θ. (58)
r2
As a transverse component of spin must be compensated by a transverse component
of orbital angular momentum, the spin proper can precess only if the orbital angular
momentum precesses as well. Orbits can therefore be strictly planar if and only if
the spin and orbital angular momentum are permanently aligned.
With that restriction it is still possible to find planar and even circular orbits. They
necessarily require all θ-components of the spin tensor to vanish:
tθ = r θ = θϕ = 0, (59)
and therefore D = S · Z = 0 identically. Moreover for circular orbits r = R is a

constant, and u r = u̇ r = 0. This is sufficient to fix the motion of a spinning test
body, in the sense that we can derive equations for the time dilation u t and the
angular velocity u ϕ in terms of the energy per unit of mass ε = E/m and the total
angular momentum per unit of mass η = J/m:

εR 2 1 − u t 2 = R (R − 3G M) u t − R 2 − 3G M R + 3(G M)2 u t 3 ,
(60)
R3uϕ 2 6G M 6(G M)2
η 2G M + R 3 u ϕ 2 = R 3 u ϕ 1 − 1− + .
GM R R2
Finally the solutions also have to satisfy the hamiltonian constraint

2G M
1− u t 2 = 1 + R 2 u ϕ 2 ≥ 1. (61)
R
Therefore of the five parameters (ε, η, R, u t , u ϕ ) characterizing a circular obit only

two are independent. Implicitly by Eqs. (55) and (57) they also determine the non-
vanishing components of the spin tensor; these are related to the velocity components
by
1 tr t 2
u − 1 = u t (R − 3G M)u t 2 − R ,
m
(62)
1 rϕ 3 ϕ2
ϕ R 2 ϕ2
2G M + R u = (R − 2G M)Ru 1 − −3 R u .
m GM
Having a complete parametrization of circular orbits for spinning test bodies it is

straightforward to construct non-circular orbits by the method of world-line devia-
tions, a direct generalization of the geodesic-deviation procedure explained in Sect. 3.
First the deviations (δξ, δu, δ) near a given reference world line can be recombined
in covariant expressions
μ
ξ μ = δξ μ , u μ = δu μ + δξ λ λν u ν ,
(63)
μ
μν = δ μν + δξ λ λκ κν + δξ λ λκν μκ .
Note that there is no a priori relation between the variations ξ μ and μν , but all
variations are linked by the first-order world-line deviation equations
Dτ ξ μ = u μ ,
μ 1 ρσ 1
Dτ u μ − Rνκ λ u κ u λ ξ ν = Rρσμν u ν + ρσ Rρσμν u ν
2m 2m
(64)
1 ρσ
+ ∇λ Rρσμν u ν ξ λ ,
2m
μ
Dτ μν = Rκλσ σν + Rκλσν μσ u κ ξ λ .
Here we will consider in particular planar non-circular orbits. The special conditions
(59) then still apply. The only new degrees of freedom are the radial velocity u r
and the mass dipole moment tϕ which become non-zero. Above we observed that
for circular orbits the parameters ε and η, representing energy and total angular
momentum per unit of mass of the test body, can be varied independently even
for circular orbits, e.g. by adjusting z-component of the spin. Therefore there is
always a circular orbit with the same ε and η as the non-circular orbit we wish to
construct. For simplicity we will take this circular orbit as the reference orbit. Then
the covariant deviation equations (64) for near-circular orbits in Schwarzschild space-
time reduce to a set of homogeneous linear differential equations for the deviations
as functions of proper time τ . Now as the spin-tensor deviations are independent
of the orbital deviations there are actually two independent types of solutions of the
coupled deviation equations. Indeed, the first-order approximation for the world lines
of spinning bodies takes the form
t (τ ) = u t τ + σ+ n t+ sin ω+ (τ − τ+ ) + σ− n t− sin ω− (τ − τ− ),
ϕ ϕ
ϕ(τ ) = u ϕ τ + σ+ n + sin ω+ (τ − τ+ ) + σ− n − sin ω− (τ − τ− ),
r (τ ) = R + σ+ n r+ cos ω+ (τ − τ+ ) + σ− n r− cos ω− (τ − τ− ),
(65)
tr (τ ) = 0tr + σ+ N+tr cos ω+ (τ − τ+ ) + σ− N−tr cos ω− (τ − τ− ),
rϕ rϕ rϕ
r ϕ (τ ) = 0 + σ+ N+ cos ω+ (τ − τ+ ) + σ− N− cos ω− (τ − τ− ),
tϕ tϕ
tϕ (τ ) = σ+ N+ sin ω+ (τ − τ+ ) + σ− N− sin ω− (τ − τ− ),
where σ± are the two independent expansion parameters for the different devia-
tion modes, which are characterized by fundamental angular frequencies ω± . The
μ μν
corresponding amplitudes are denoted by (n ± , N± ) for the orbital and spin-tensor
deviations respectively. The explicit expressions for the frequencies and amplitudes
are collected in Appendix C.
rϕ
Finally the lowest order terms 0tr and 0 for the spin tensor components are the
circular-orbit values satisfying equations (62), and τ± are constants of integration,
one for each type of deviation, which fix their initial values. For a complete derivation
of these results I refer to Ref. [20].
6 Stability of Orbits and the ISCO
In Sect. 3 it was shown that for radial co-ordinates R < 6G M the deviations from
circular orbits show exponential runaway behaviour. Therefore R = 6G M is the
innermost stable circular orbit. In the case of spinning particles we can similarly
investigate the solutions (65) of the equations for deviations from circular orbits
for instabilities and determine the existence of an ISCO for different values of the
z-component of spin per unit of mass
R r ϕ
σ= ,
m
which is the spin-contribution to η.
The stable deviations (65) are characterized by real-valued angular frequencies
ω± . If any one of these frequencies develops an imaginary part runaway behaviour
sets in and the circular orbits become unstable. The frequencies themselves are given
by the expressions (80) in Appendix C:
1
ω±
2
= A ± A2 − 4B ,
2
where A and B represent long expressions in terms of the parameters of the circular
reference orbit. For ω± to be real the square root on the right-hand side must be real,
and the whole expression must be nonnegative as it represents a real square. This
results in the following inequalities
A ≥ 0 and 0 ≤ 4B ≤ A2 . (66)
These conditions are plotted in Fig. 1 as a function of the radial co-ordinate R and
the orbital angular momentum
= R 2 u ϕ = η − σ,
Fig. 1 Region of stability of circular orbits in the R-2 plane
both measured in units of G M. The curves labeled f, g, h represent lines of constant

σ. In particular g is the curve for spinless test bodies (σ = 0); it leaves the region of
stability at R = 6G M, as expected. For retrograde spin values σ < 0 as represented
by the curve f the ISCO is reached earlier, whilst prograde spin values σ > 0 as
represented by the curve h stabilize circular orbits in a range of values R < 6G M.
From these results one can infer the radial co-ordinate R of the ISCO as a function
of the spin parameter σ, as plotted in Fig. 2.
The steep line for spin values σ > 0.55 has been included for completeness; here
the upper limit on B in inequality (66) takes over from the condition B > 0 as the
main stability criterion. However these large spin values are physically unrealistic
as they can only be obtained in cases where the test-body limit is not applicable,
such as binary black holes of comparable mass. Also plotted in Fig. 2 is the curve
obtained by minimizing the orbital angular momentum as a function of R at fixed
spin. Cleary the two curves largely coincide.
Fig. 2 ISCO radius R as function the spin parameter σ. The dashed line represents the values of
R for which the orbital angular momentum reaches its minimum
7 Non-minimal Hamiltonian Dynamics of Spinning Test

Bodies
The motion of test bodies has been modeled so far using the minimal hamiltonian (14).
However, it is not difficult to construct more complicated hamiltonians to model test
bodies with additional interactions such as spin-curvature couplings. As the Dirac-
Poisson brackets (50) are closed and model independent the equations of motion
can be derived in straightforward fashion for any such extended hamiltonian. For
example, one can include Stern–Gerlach type of interactions as discussed in Refs. [18,
20, 23]. In this case the extended test-body hamiltonian is
1 μν κ
H= g [ξ] πμ πν + Rμνκλ [ξ] μν κλ . (67)
2m 4
In terms of the four-velocity u μ = ξ˙μ the corresponding equations of motion read
πμ = mgμν u ν ,
1 κλ μ ν κ κλ ρσ
m Dτ u μ = Rκλ ν u − ∇μ Rκλρσ , (68)
2 4

μ
Dτ μν = −κ ρσ Rρσ λ λν + Rρσν λ μλ .
As in the minimal case these equations can also be derived by requiring the vanishing
of the covariant divergence of a suitable energy-momentum tensor [19]

μν κ 1
T μν = Tmin + ∇κ ∇ λ dτ μλ κν + νλ μκ √ δ 4 (x − ξ(τ ))
4 −g
(69)
κ 1
ρσ μ
+ dτ Rρσλ λν + Rρσλν λμ √ δ 4 (x − ξ(τ )).
4 −g
μν
Here Tmin is the energy-momentum tensor (48) of a spinning test body with minimal
dynamics.
Remarkably all conservation laws for spinning bodies we derived in the minimal
case carry over to the case with Stern–Gerlach interactions. In particular any constant
of motion (52), (53) associated with a Killing vector αμ is also conserved by the
Stern–Gerlach terms in the hamiltonian:
κ

J, Rμνκλ μν κλ = 0. (70)
4
For example, in a static and spherically symmetric background like Schwarzschild
or Reissner–Nordstrøm space-time the kinetic energy E and the angular momentum
3-vector J given by Eqs. (55) and (56) are again conserved.
This form of non-minimal hamiltonian dynamics predicts some interesting effects.
In particular, as the hamiltonian is a constant of motion which by evaluation in
a curvature-free region is seen to be expressed in terms of the inertial mass by
H = −m/2, the hamiltonian constraint gets modified to read
κm
g μν πμ π ν + Rμνκλ μν κλ + m 2 = 0. (71)
2
Then the four-velocity is no longer normalized to be a time-like unit vector; instead
the time-like unit vector tangent to the world line actually is
uμ
nμ = 1/2 . (72)
κ
1+ 2m
Rμνκλ μν κλ
Considering a particle at rest:

2
dt κ
gtt =1+ Rμνκλ μν κλ , (73)
dτ 2m
this is seen to imply that the spin-curvature coupling represents an additional source
of gravitational time-dilation. A similar effect related to spinning particles interacting
with electromagnetic fields was conjectured in Refs. [24, 25].
8 Final Remarks
The motion of test bodies carrying a finite number of relevant degrees of freedom
like momentum, spin or charge can be represented by world lines in space-time to the
extent that we can assign them a well-defined position and that their back reaction
on space-time geometry can be neglected. Convenient position co-ordinates are not
necessarily those of a center of mass (or for that matter a center of charge) in the
local rest frame, as the example of spinning test bodies shows. In that case we find
it preferable to associate the world line of free particles with the line on which the
spin tensor is covariantly constant. The mass dipole moment can then be taken to
represent the effective position of the mass with respect to that world line.
This is also clear from the corresponding energy-momentum tensor which receives
contributions from both the spin proper and the mass dipole. In a next step this can
be used to compute the back reaction of the test body on the space-time geometry as
discussed in the simple example in Sect. 1. In general this procedure also includes
determining the self-force and the gravitational waves emitted by test bodies in the
specific background under discussion [14, 26].
As another application it has been shown in the literature how the motion of test
bodies can be used to reconstruct the geometry of space-time [11]. Simple geodesic
motion of a sufficient number of test bodies allows one to determine the curvature
at a point in space by measuring the geodesic deviations in its neighborhood. By
including higher-order corrections as in Eqs. (36), (37) one could also determine the
derivatives of the curvature to obtain the curvature in a region around the point of
interest. As Eq. (64) show, an alternative method is to measure first-order world-line
deviations of spinning test bodies, which also depend on the gradient of the Riemann
curvature tensor.
Acknowledgements I am indebted to Richard Kerner, Roberto Collistete jr., Gideon Koekoek,

Giuseppe d’Ambrosi, S. Satish Kumar and Jorinde van de Vis for pleasant and informative discus-
sions and collaboration on various aspects of the topics discussed. This work is supported by the
Foundation for Fundamental Research of Matter (FOM) in the Netherlands.
A Observer-Dependence of the Center of Mass in Relativity
To illustrate the observer-dependence of the center of mass of an extended body we

consider a simple example: the motion of two equal test bodies revolving at constant
angular velocity in Minkowski space on a circular orbit around an observer located
in the origin of an inertial frame with cartesian co-ordinates (t, x, y, z). The plane of
the orbit is taken to be the x-y-plane. In Fig. 3 we plot the projection of the world lines
in the x-t-plane, represented by the two widely oscillating curves. At any moment
the two test bodies are at equal distance to the observer and in opposite phase with
respect to the origin. In this frame the center of mass is located at the origin x = 0
and moves in a straight line along the t-axis in space-time.
Fig. 3 The world line of the center of mass of two equal masses (e.g., a binary star system) in
circular orbit with respect to a stationary observer on the axis of orbital angular momentum is
represented by the line x = 0. The world line of the center of mass with respect to an observer in an
inertial frame (t , x ) moving at constant velocity along the x-axis is represented by the oscillating
curve labeled CM
A second observer in another inertial frame (t , x , y , z ) moving with constant

velocity v in the positive x-direction has a different notion of simultaneity, as defined
by the appropriate Lorentz transformation. The lines t = constant are represented
by the dashed slant lines parallel to the x -axis. In the limit of large masses and slow
rotation the center of mass CM with respect to this moving frame is located halfway
between the masses at fixed time t . The world line of CM is now represented
by the single curve oscillating at smaller amplitude around the line x = 0 in the
original frame. In fact for the observer in relative motion CM moves in the negative
x -direction while oscillating around the line x = −vt .
It is obvious that in curved space-time the notion of simultaneity is further com-
plicated because of the non-existence of global inertial frames, resulting in additional
distortions of the world line CM with respect to the world line in the local inertial
frame (t, x, y, z) fixed to the center of rotation.
B Coefficients for Geodesic Deviations in Schwarzschild

Geometry
The coefficients for the deviations of bound equatorial orbits w.r.t. parent circular
orbits have been calculated for Schwarzschild space-time up to second order; with the
restriction ρr1 = ρr2 = 0 explained in the main text one gets the following results [6]:
a. Secular terms:
3G M R + GM
ρt1 = 0, ρt2 = ,
2R (R − 3G M)3/2
5/2
(74)
ϕ ϕ 3 G M (R − 2G M) (R + G M)
ρ1 = 0, ρ2 = .
2R 7/2 R (R − 3G M)3/2
b. First-order periodic terms:

4G M R
n t1 =− , n t = 0,
(R − 2G M) (R − 6G M) 2

2G M (75)
n r1 = , 1− n r2 = 0,
R

ϕ 2 R − 2G M ϕ
n1 = − , n 2 = 0.
R R − 6G M
c. Second order periodic terms:

G M 2R 2 − 15G M R + 14(G M)2
m t2 = ,
R 2 (R − 2G M) (R − 6G M)3/2
1 (R − 2G M) (R − 7G M)
m r2 = − , (76)
R2 R − 6G M
ϕ 1 (R − 2G M) (5R − 32G M)
m2 = .
2R 5/2 (R − 6G M)3/2
d. Angular frequency:

G M R − 6G M
ω0 = , ω1 = 0. (77)
R 3 R − 3G M
ϕ μ
In the non-restricted case with ρr1 = 0 also the coefficients ρt1 , ρ1 , n 2 and ω1 all
become non-zero as well [13].
C Coefficients for Spinning World-Line Deviations

in Schwarzschild Geometry
The first-order planar deviations of circular orbits of spinning particles for constant
energy and total angular momentum in Schwarzschild space-time are expressed con-
veniently in terms of the following combinations of orbital and spin parameters [20]
2(R − G M) t 2ε G Mu ϕ
α= u − , β=− ,
R(R − 2G M) R − 2G M mR
2R − 5G M ϕ G Mη G M(R − 2G M) t
γ= u + 3 ,ζ=− u, (78)
R(R − 2G M) R (R − 2G M) m R4
2(R − 2G M) G Mη
κ=− ε, λ = 2Ru ϕ − ,
R2 R2
and
2(R − 3G M) 2(R − 4G M) t 2G M ϕ
μ=− 3
+ 3
εu + u ϕ 2 + ηu ,
R R R3
(R − G M)(R − 3G M) G Mmη
ν = mu ϕ + 2 ,
R − 2G M R (R − 2G M)
(R − G M)(R − 3G M) m Rε
σ= mu t − ,
G M(R − 2G M) GM
(R 2 − 4G M R + 5(G M)2 ) ϕ t G M(3R − 4G M) mεu ϕ

χ= mu u − mηu t
− .
G M(R − 2G M)2 R 3 (R − 2G M)2 GM
(79)
With these definitions the frequencies of the first-order planar deviations are
1
ω±
2
= A ± A2 − 4B ,
2
A = μ − ακ − βν − γλ − ζσ, (80)
B = β (κχ − μν + γ(λν − κσ)) + ζ (λχ − μσ − α(λν − κσ)) ,
whilst the amplitudes are given by
n t± = λ(βγ − αζ) + β(ω±

2
− μ),
ϕ
n ± = −κ(βγ − αζ) + ζ(ω±
2
− μ), (81)
n r± = ω± (βκ + ζλ),
and

mω± R 2 2G M 2m R GM
N±tr = 1− n t± + 1− u t − ε n r± ,
GM R GM R
rϕ ϕ m 2 ϕ
r (82)
N± = −mω± Rn ± − η + R u n±,
R2
tϕ
N ± = ω±
2
(ω±
2
− μ + ακ + γλ).
References
1. C. Møller, Sur la dynamique des systèmes ayant un moment angulaire interne. Ann. Inst. Henri
Poincaré 11, 251 (1949)
2. C. Møller, The Theory of Relativity (Clarendon Press, Oxford, 1952)
3. L.F. Costa, J. Natário, Center of mass, spin supplementary conditions, and the momentum of
spinning particles. Fund. Theor. Phys. 179, 215 (2015)
4. S. Weinberg, Gravitation and Cosmology (Wiley, New York, 1972)
5. E. Hackmann, Geodesic equations and algebro-geometric methods (2015),
arXiv:1506.00804v1 [gr–qc]
6. R. Kerner, J.W. van Holten, R. Collistete jr., Relativistic epicycles: another approach to geodesic
deviations. Class. Quantum Gravity 18, 4725 (2001)
7. R. Colistete Jr., C. Leygnac, R. Kerner, Higher-order geodesic deviations applied to the Kerr
metric. Class. Quantum Gravity 19, 4573 (2002)
8. J. Ehlers, F.A.E. Pirani, A. Schild, The geometry of free fall and light propagation, in General
Relativity: Papers in Honnor of J.L. Synge, ed. by L. O’Raifeartaigh (Oxford University Press,
Oxford, 1972)
9. C.W. Misner, K.S. Thorne, J.A. Wheeler, Gravitation (Freeman and Co., San Francisco, 1970)
10. P. Szekeres, The gravitational compass. J. Math. Phys. 6, 1387 (1965)
12. D. Philipp, D. Puetzfeld, C. Lämmerzahl, On the applicability of the geodesic deviation equation
in general relativity (2016), arXiv:1604.07173 [gr–qc]
13. G. Koekoek, J.W. van Holten, Epicycles and Poincaré resonances in general relativity. Phys.
Rev. D 83, 064041 (2011)
14. G. Koekoek, J.W. van Holten, Geodesic deviations: modeling extreme mass-ratio systems and
their gravitational waves. Class. Quantum Gravity 28, 225022 (2011)
15. M. Mathison, Neue Mechanik materieller Systeme. Acta Phys. Pol. 6, 163 (1937)
16. W. Tulczyjew, Motion of multipole particles in general relativity theory. Acta Phys. Pol. 18,
393 (1959)
17. J. Steinhoff, Canonical formulation of spin in general relativity, Ph.D. thesis (Jena University)
(2011), arXiv:1106.4203v1 [gr-qc]
18. G. d’Ambrosi, S.S. Kumar, J.W. van Holten, Covariant Hamiltonian spin dynamics in curved
space-time. Phys. Lett. B 743, 478 (2015)
19. J.W. van Holten, Spinning bodies in general relativity. Int. J. Geom. Methods Mod. Phys. 13,
1640002 (2016)
20. G. d’Ambrosi, S.S. Kumar, J.W. van Holten, J. van de Vis, Spinning bodies in curved space-
time. Phys. Rev. D 93, 04451 (2016)
21. J. Ehlers, E. Rudolph, Dynamics of extended bodies in general relativity: center-of-mass
description and quasi-rigidity. Gen. Relativ. Gravit. 8, 197 (1977)
22. R. Ruediger, Conserved quantities of spinning test particles in general relativity. Proc. R. Soc.
Lond. A375, 185 (1981)
23. I. Khriplovich, A. Pomeransky, Equations of motion for spinning relativistic particle in external
fields. Surv. High Energy Phys. 14, 145 (1999)
24. J.W. van Holten, On the electrodynamics of spinning particles. Nucl. Phys. B 356, 3–26 (1991)
25. J.W. van Holten, Relativistic time dilation in an external field. Phys. A 182, 279 (1992)
26. G. d’Ambrosi, J.W. van Holten, Ballistic orbits in Schwarzschild space-time and gravitational
waves from EMR binary mergers. Class. Quantum Gravity 32, 015012 (2015)
On the Applicability of the Geodesic
Deviation Equation in General Relativity
Dennis Philipp, Dirk Puetzfeld and Claus Lämmerzahl
Abstract Within the theory of General Relativity, we study the solution and range of
applicability of the standard geodesic deviation equation in highly symmetric space-
times. In the Schwarzschild spacetime, the solution is used to model satellite orbit
constellations and their deviations around a spherically symmetric Earth model. We
investigate the spatial shape and orbital elements of perturbations of circular refer-
ence curves. In particular, we reconsider the deviation equation in Newtonian gravity
and then determine relativistic effects within the theory of General Relativity by com-
parison. The deviation of nearby satellite orbits, as constructed from exact solutions
of the underlying geodesic equation, is compared to the solution of the geodesic
deviation equation to assess the accuracy of the latter. Furthermore, we comment on
the so-called Shirokov effect in the Schwarzschild spacetime and limitations of the
first order deviation approach.
1 Introduction
Applications in space based geodesy and gravimetry missions require the precise
knowledge of satellite orbits and possible deviations of nearby ones. In such a con-
text, one of the satellites may serve as the reference object and measurements are
performed w.r.t. this master spacecraft. GRACE-FO, the successor of the long-lasting
GRACE mission, aims at measuring the change of the separation between two space-
crafts with some 10 nm accuracy [1, 2]. The change of this distance is then used to
obtain information about the gravitational field of the Earth, i.e. to measure the New-
tonian multipole moments of the gravitational potential, and to deduce information
about the mass distribution and its temporal variations.
D. Philipp (B) · D. Puetzfeld · C. Lämmerzahl

Center of Applied Space Technology and Microgravity (ZARM), University of Bremen,
Bremen, Germany
e-mail: dennis.philipp@zarm.uni-bremen.de
D. Puetzfeld
e-mail: dirk.puetzfeld@zarm.uni-bremen.de
URL: http://puetzfeld.org

420 D. Philipp et al.
In this work, we investigate the geodesic deviation equation in General Relativity

(GR) in the case of highly symmetric spacetimes. Our aim is to develop a measure
for the quality of approximation that the deviation equation provides to model test
bodies and their orbit deviations in different orbital configurations. To achieve this,
we construct general solutions of the geodesic deviation equation and compare them
to exact solutions of the underlying geodesic equation in a Schwarzschild spacetime.
Our analysis allows us to reveal physical and artificial effects of such an approxi-
mative description. In particular, we comment on an effect that was reported for the
first time in 1973 by Shirokov [3].
The structure of the paper is as follows: In Sect. 2, we reconsider the deviation
equation in Newtonian gravity. This is followed by an investigation of the first order
geodesic deviation equation in static, spherically symmetric spacetimes in GR in
Sect. 3. A direct comparison between the solution for the Schwarzschild spacetime,
which we focus on for the rest of this work, and the Newtonian results unveils
relativistic effects. In Sect. 4, we describe the shape of perturbed orbits and the
influence of six free integration parameters on the general solution. These parameters
are connected to orbital elements of the orbit under consideration, and they determine
how it is obtained from a perturbation of the reference curve. We assess the range
of applicability of the deviation equation by comparing its solutions to deviations
constructed “by hand” from exact solutions of the underlying geodesic equation.
We study physical effects such as perigee precession and the redshift due to time
dilation between the reference and perturbed orbit. Building on these results, we
uncover some artificial effects previously reported in the literature. Our conclusions
and a brief outlook on future applications are given in Sect. 5. Appendix A contains
a summary of our notations and conventions.
2 Orbit Deviations in Newtonian Gravity
There are two basic equations that govern Newtonian gravitational physics; the field
equation, also known as Poisson’s equation
U (x) = ∂μ ∂ μ U (x) = 4πGρ(x) , (1a)
and the equation of motion
ẍ μ = −∂ μ U (x) , (1b)
where the (x μ ) = (x, y, z) are Cartesian coordinates and the overdot denotes deriva-
tives w.r.t. the Newtonian absolute time t. Here and in the following, Greek indices
are spatial indices and take values 1, 2, 3. The field Eq. (1a) relates the Newtonian
gravitational potential U to the mass density ρ and introduces Newton’s gravita-
tional constant G as a factor of proportionality. Outside a spherically symmetric
(and static) mass distribution, i.e. in the region where ρ = 0, we obtain as a solution
of the Laplace equation U = 0:
On the Applicability of the Geodesic Deviation Equation in General Relativity 421
U (r ) = −G M/r , (2)
where M is the mass of the central object, obtained by integrating the mass density
over the three-volume of the source, and r is the distance to the center of the gravi-
tating mass. The equation of motion (1b) describes how point particles move in the
gravitational potential given by U .
2.1 Newtonian Deviation Equation
We now recall the derivation of the Newtonian deviation equation, see, e.g., Ref. [4]
and references therein. For a given reference curve Y μ (t) that fulfills the equation
of motion we construct a second curve X μ (t) = Y μ (t) + η μ (t) and introduce the
deviation η. This second curve shall be a solution of the equation of motion as well
(at least up to linear order, as we will see below). Hence, we get
Ẍ μ = Ÿ μ + η̈ μ = −∂ μ U (X ) = −∂ μ U (Y + η) . (3)
For small deviations we linearize the potential around the reference object with
respect to the deviation,
U (X ) = U (Y + η) = U (Y ) + η ν ∂ν U (Y ) + O(η 2 ) . (4)
Thereupon, the first order deviation equation in Newtonian gravity becomes (spatial
indices are raised and lowered with the Kronecker delta δνμ )
η̈ μ = −[∂ μ ∂ν U (Y )] η ν =: K μ ν η ν . (5)
For a homogeneous (∂ν U ≡ 0) or vanishing (U ≡ 0) gravitational potential, the

deviation vector has the simple linear time dependence η μ (t) = Aμ t + B μ , with
constants Aμ and B μ . A non-linear time dependence of the deviation is caused by
second derivatives of the Newtonian gravitational potential, i.e. if K μ ν = 0.
Since we are interested in the deviation for highly symmetric situations, we now
use the potential (2) outside a spherically symmetric mass distribution and introduce
usual spherical coordinates by
(x, y, z) = (r sin ϑ cos ϕ, r sin ϑ sin ϕ, r cos ϑ) . (6)
Due to the symmetry of the situation we can, without loss of generality, restrict the
reference curve to lie within the equatorial plane that is defined by ϑ = π/2. Apply-
ing the coordinate transformation x a → x̃ a from Cartesian to spherical coordinates,
Eq. (5) turns into
η̃¨ ν ∂˜ν x μ + 2η̃˙ ν ∂˜ν ẋ μ + η̃ ν ∂˜ν ẍ μ

= ∂r U (r ) ∂ μ ∂ν r + ∂r2 U (r )(∂ μr )(∂ν r ) η̃ σ ∂˜σ x ν , (7)
where (η̃ μ ) = (ηr , η ϑ , η ϕ ) are the components of the deviation in the new coor-
dinates and (∂˜μ ) = (∂r , ∂ϑ , ∂ϕ ). These are three equations for the three unknown
components of the deviation. All angular terms can be eliminated by appropriate
combinations of these equations and a straightforward but rather lengthy calculation
yields the system of differential equations

Ṙ GM R̈
η̈ ϑ = − η̇ ϑ − + ηϑ , (8a)
R R3 R

2G M
η̈r = ˙2+ ηr + 2 Ṙ ˙ + R¨ η ϕ + 2R
˙ η̇ ϕ , (8b)
R 3

2 Ṙ ϕ 2 ˙ r R̈ GM ¨
ϕ
η̈ = − η̇ − η̇ − ˙
+ 3 − η ϕ − ηr ,
2
(8c)
R R R R R
where the quantities represented by capital letters (R, ) are in general functions of
time t and describe the trajectory of the reference object, along which the system
(8) must be solved. In geodesy, the quantity K μ ν = 0 and the Eq. (5) are known in
the framework of gradiometry. However, we did not find the system of differential
equations (8), describing the Newtonian deviations from a general reference curve,
published elsewhere in this form.
2.2 Deviation from Circular Reference Curves
One particular case is the deviation from a circular reference orbit with constant radius
R. This special situation was already considered by Greenberg [5], who derived (only)
the oscillating solutions. However, in [4] the full solution for this case can be found.
In the following we briefly summarize the results in a form that we will use later
to compare to the relativistic results. The azimuthal motion of the reference orbit is
described by

GM GM
˙ =
⇒ (t) = t =: K t . (9)
R3 R3
The quantity K is the well known Keplerian frequency and leads to the Keplerian
¨ ≡
orbital period 2π/ K . For the circular reference orbit, the conditions R̈ = Ṙ =
0 hold, and the system (8) yields three ordinary second order differential equations
of which the last two are coupled (see Eqs. (29), (33) and (34) in [5] for comparison)
η̈ ϑ = −2K η ϑ , (10a)
ϕ
η̈ = 2R K η̇ +
r
32K η ,
r
(10b)
R η̈ ϕ = −2 K η̇r . (10c)
The first equation for the deviation in the ϑ-direction describes a simple harmonic
oscillation around the reference orbital plane and is decoupled from the remaining
ones. The general real-valued solution is given by
C(5) C(6)
η ϑ (t) = cos K t + sin K t . (11a)
R R
The parameters C(5) and C(6) are the amplitudes of the two fundamental solutions
(normalized to the reference radius R) and the deviation component η ϑ oscillates
with the Keplerian frequency K . In [5], Greenberg derived the oscillating solutions
for the remaining two equations. However, the general solution, cf. [4], is given by
ηr (t) = C(1) + C(2) sin K t + C(3) cos K t , (11b)

3
R η ϕ (t) = 2 C(2) cos K t − C(3) sin K t − K C(1) t + C(4) . (11c)
2
Summarizing the results, the perturbed orbit is described by
r (t) = R + ηr (t) = R + C(1) + C(2) sin K t + C(3) cos K t, (12a)

3 C(4)
ϕ(t) = K t + η ϕ (t) = K 1 − C(1) t +
2R 2R
2
+ C(2) cos K t − C(3) sin K t , (12b)
R
π C(5) C(6)
ϑ(t) = + η ϑ (t) = π/2 + cos K t + sin K t . (12c)
2 R R
Obviously, there are several possibilities to perturb the reference orbit. The parame-
ters C(i) , i = 1 . . . 6 define the initial position and velocity (or the orbital elements)
of the test body that follows the perturbed curve. The meaning of these parameters
and their impact on the perturbed orbit was studied briefly in Ref. [4], and the analysis
will be extended in Sect. 4 in the context of the general relativistic results. Note that
the only frequency appearing in the solution so far is the Keplerian frequency K .
3 Geodesic Deviation in General Relativity
In GR, the equation of motion for structureless test bodies takes the form of the
geodesic equation
d2xa dxb dxc

= − bc
a
(x) . (13)
ds 2 ds ds
See Refs. [6, 7] for reviews of methods to derive this, and higher order equations of
motion, by means of multipolar techniques. Latin indices denote spacetime indices,
taking values 0, 1, 2, 3, and bc a are the connection coefficients of the underlying
spacetime (Christoffel symbols).
As in the Newtonian case, we consider two neighboring curves Y a (s) and X a (s),
both of them are now assumed to be geodesics, and s is the proper time measured
along the curve Y a (s). Choosing Y a (s) as the reference curve, we may introduce, in
a coordinate representation, the deviation η a (s) w.r.t. the neighboring curve X a (s)
as
η a (s) := X a (s) − Y a (s) . (14)
Denoting the normalized four-velocity along the reference curve by Ẏ a := dY a /ds,

it can be shown that the second covariant derivative of the deviation fulfills
D2 η a (s)
= R a bcd (Y ) Ẏ b η c Ẏ d + O(η 2 ) , (15)
ds 2
up to the linear order in the deviation and its first derivative, along the reference
curve. This is the well-known geodesic deviation or Jacobi equation, in which R a bcd
denotes the curvature of spacetime. For more details on its systematic derivation, in
particular its possible generalizations, and an overview of the literature see Ref. [8].
From Eq. (15), we infer that the deviation η a will have a non-linear time dependence
if and only if the spacetime is curved. For vanishing curvature, the deviation can
only grow linearly in time as in the Newtonian situation for K μν = 0. The Newtonian
quantity K μν that measures second derivatives of the gravitational potential is replaced
by the curvature tensor in GR.
In the following, we will focus on the solutions of Eq. (15) in the case of timelike
geodesics that may correspond to satellite orbits around the Earth. In particular, we
are going to assume an orthogonal parametrization, see Sect. III in [8], in which the
deviation is orthogonal to the velocity along the reference curve ηa Ẏ a = 0 – cf. Fig. 1
for a sketch. We will further assume the reference curve Y a to be a circular geodesic
and construct orbits out of its perturbation.
3.1 Deviation Equation in Spherically Symmetric and Static

Spacetimes
In a spherically symmetric and static spacetime that is described by the metric
ds 2 = A(r )dt 2 − B(r )dr 2 − r 2 (dϑ2 + sin2 ϑdϕ2 ) , (16)

Fig. 1 Sketch of the deviation of two nearby geodesics Y a and X a = Y a + η a . Here we depict the
case of the orthogonal correspondence – in which the deviation vector η a is chosen to be orthogonal
to the four-velocity Ẏ a along the reference geodesic
we use spherical coordinates (x a ) = (t, r, ϑ, ϕ) and choose units such that the speed
of light c and Newton’s gravitational constant G are equal to one. The angles ϑ
and ϕ are the usual polar and azimuthal angles as in spherical coordinates and the
radial coordinate r is defined such that spheres at a radius r have area 4πr 2 . In these
coordinates, the reference geodesic shall be represented by (Y a ) = (T, R, , ).
Due to the symmetry of the spacetime we can, without loss of generality, assume that
the reference geodesic is confined to the equatorial plane. Thus, we have ≡ π/2
and ˙ = ¨ = 0. For geodesics in the considered spacetime there exist constants of
motion that correspond to the conservation of energy E and angular momentum L,
see for example [9]. Since the metric (16) does neither depend on the time coordinate
t nor on the angle ϕ, i.e. ∂t and ∂ϕ are Killing vector fields, the constants of motion
are given by
E := A(r ) t˙ = const. , L := r 2 ϕ̇ = const. (17)
The general solution of the first order geodesic deviation equation (15) in the
spacetime (16) was given by Fuchs [10] in terms of first integrals, which remain
to be solved. Unfortunately, this solution is not applicable to the simplest case of
the deviation from a circular reference geodesic. The condition Ṙ = 0 causes sin-
gularities in terms ∼1/ Ṙ that appear in the equations. Shirokov [3] was the first
to derive periodic solutions for the deviation from circular reference geodesics in
Schwarzschild spacetime. In [11, 12] the solution for Schwarzschild spacetime and
circular reference geodesics was given in terms of relativistic epicycles. However,
another possible way to obtain the full solution for circular reference geodesics in the
more general spacetime (16) is to refer the system of differential equations (15) to a
parallel propagated tetrad along the reference curve. The solution of the equations in
this reference system is then projected on the coordinate basis [13]. This method is
of direct relevance for relativistic geodesy, since it allows to describe the deviation
as observed in the comoving local tetrad, i.e. by an observer with an orthonormal
frame who is located at the position of the reference object. We will use the results
of this method here.
The motion along the circular reference geodesic with radius R in the equatorial
plane can be described using the constants of motion E and L from Eq. (17):
L
˙s=
(s) = s =: s , (18a)
R2
E
T (s) = Ṫ s = s. (18b)
A(R)
After some lengthy calculations one arrives at the solution of the deviation equation
using the result of Fuchs [13]
LE
η t (s) = √ f (s) , (19a)
A L 2 + R2
ER
ηr (s) = √ √ g(s) , (19b)
AB L 2 + R 2
C(5) C(6)
η ϑ (s) = cos s + sin s , (19c)
√R R
L 2 + R2
η ϕ (s) = f (s) , (19d)
R2
where the two proper time dependent functions f (s) and g(s) are given by

(k 2 − )
f (s) = (C (2) cos ks − C (3) sin ks) + √ C(1) s + C(4) , (20a)
k2
g(s) = C(1) + C(2) sin ks + C(3) cos ks , (20b)
2
2A A − A 3A
k 2 :=
+ , (20c)
2 AB(2 A − A R) 2 AB R
2 A
:= . (20d)
AB R
The prime denotes derivatives w.r.t. the radial coordinate and the metric functions
A = A(R), B = B(R) are to be evaluated at the reference radius R. Furthermore,
for a circular geodesic the constants of motion (17) can be expressed by
2 A2 R 3 A
E2 = , L2 = . (21)
2 A − A R 2 A − A R
3.2 Deviation Equation in Schwarzschild Spacetime
In GR, the Schwarzschild spacetime serves as the simplest model of an isolated and
spherically symmetric central object and might be used as a first order approximation
of an astrophysical object like the Earth.1 The metric functions in Eq. (16) are then
given by
2m
A(r ) = 1 − , B(r ) = A(r )−1 . (22)
r
The constants of motion E and L as well as the remaining quantities k and are
uniquely defined by the radius R of the circular reference geodesic
4m m(R − 6m)
= , k2 = ,
R3 R 3 (R − 3m)
(R − 2m)2 m R2
E2 = , L2 = . (23)
R(R − 3m) R − 3m
We should mention that the mass of the Earth in the units that we use is m ≈ 0.5 cm.
The parameters C(1,...,6) will be used to model different orbital scenarios. Using the
constants in Eq. (23), we can simplify the solution (19) for the case of Schwarzschild
spacetime. We find2 :
√
mR
η (s) =
t
f (s) , (24a)
R − 2m
ηr (s) = g(s) , (24b)
C(5) C(6)
η ϑ (s) = cos s + sin s , (24c)
R R
f (s)
η ϕ (s) = , (24d)
R
where the two functions f (s) and g(s) are given by
1 Planets do not possess any net charge, therefore we do not consider charged solutions like, for
example, the Reissner–Nordstrøm spacetime.
2 W.r.t. Eq. (19) we have slightly redefined the constant parameters C
(1,...,6) in a way such that
ηr (s) = g(s). This is always possible since all coefficients preceding the functions f (s) and g(s)
in (19) are constant because the reference radius R is constant.

R 3 R − 2m
f (s) = 2 C(2) cos ks − C(3) sin ks − C(1) s + C(4) ,
R − 6m 2 R − 3m
(25a)
g(s) = C(1) + C(2) sin ks + C(3) cos ks . (25b)
Notice that the function f (s) contains, besides periodic and constant parts, a term
that grows linearly with the reference proper time s for C(1) = 0. This contribution
is not bounded and will, thus, limit the validity of the framework since we work with
the first order deviation equation, i.e. the deviation η a is assumed to be small and
only contributions up to first order were considered. We observe that in the general
relativistic solution of the first order deviation equation two distinct frequencies
appear:

m R − 6m R − 6m
k= = K , (26a)
R3 R − 3m R − 3m

m R R
= = K . (26b)
R3 R − 3m R − 3m
In the Newtonian limit these two frequencies coincide and yield the Keplerian fre-
quency K . Figure 2 shows the difference − k between both frequencies for
reference radii that correspond to satellite orbits from 100 km to 3.6 · 104 km above
the surface of the Earth. It is worthwhile to note that the general relativistic solution
Fig. 2 The difference between the frequencies and k, which appear in the solution of the
deviation equation, is shown for reference radii that belong to satellite orbits around the Earth. The
frequency difference is of the order of some 10−12 Hz, which yields a period difference in the range
of 10–30 µs. We consider as the mean Earth radius R⊕ = 6.37 · 103 km
(24), (25) approaches the correct Newtonian limit (11) for c → ∞. Studying the
difference allows to uncover relativistic effects in the following. Observe that the
normalization of the reference four-velocity yields
˙ 2 = (1 − 2m/R)Ṫ 2 − r 2 2 = 1
(1 − 2m/R) Ṫ 2 − R 2 (27)

R
⇒ Ṫ = . (28)
R − 3m
When we parametrize the circular reference orbit by coordinate time we get

−1
d dT m
˜ :=
= Ṫ −1
= = K . (29)
ds ds R3
Hence, Kepler’s third law holds perfectly well for circular orbits in the Schwarzschild
spacetime when the orbit is parametrized by coordinate time.
4 Applicability of the Geodesic Deviation Equation
In this section, we study the applicability of the first order deviation equation (15)
in Schwarzschild spacetime to describe the motion of a test body that is close to
a given circular reference geodesic. Its worldline is determined by a small initial
perturbation of that reference curve, described by the solution (24). In the following,
we investigate the shape of the perturbed orbits as well as physical and artificial
effects, which are present in the solution.
To describe different orbital configurations we have to examine the impact of the
parameters C(1,...,6) on the perturbed orbit. A proper way to do this is to investigate the
impact of each parameter separately since the different effects can be superimposed
in this linearized framework. Here, we extend the brief analysis that was done in
Ref. [4]. The connection between the parameters C(i) and the orbital elements of the
perturbed orbit are summarized in Table 1. Hence, for a specific orbital configuration
that is to be modeled we can determine the parameters that must be taken into account
from the table and describe that satellite configuration within the framework of the
geodesic deviation equation using the solution (24).
4.1 Shape of the Perturbed Orbits
The following sections are named after the geometric shape of the perturbed orbits,
caused by the choice of the respectively considered parameter(s). All orbits that we
discuss in the following are shown in Fig. 3.
Table 1 The initial position and velocity of the test body that follows the perturbed orbit X a .
Choose one parameter that shall be the only non-zero one, then the initial state can be read off from
the table. For combinations of different parameters the effects can be superimposed. We list the
orbital elements such as eccentricity e, semi-major axis a, the ascending node a (longitude), the
inclination i, distance to the perigee d p , distance to the apogee da and the argument of the perigee
ω. For the definitions of these orbital elements see Fig. 5. When two values are given, the orbital
element depends on the sign of the respective parameter
= 0: C(1) C(2) C(3) C(4) C(5) C(6)
r(0) R + C(1) R R + C(3) R R R
ϑ(0) π/2 π/2 π/2 π/2 π/2 + C(5) /R π/2
ϕ(0) 0 δω(2) 0 C(4) /R 0 0
ṙ (0) 0 C(2) k 0 0 0 0
ϑ̇(0) 0 0 0 0 0 C(6) /R
ϕ̇(0) + δω(1) +
k δω(3)
e 0 C(2) /R C(3) /R 0 0 0
a R + C(1) R R R R R
a 0 0 0 0 π/2(2 − sgn C(5) ) π/2(1 + sgn C(6) )
i 0 0 0 0 C(5) /R C(6) /R
dp R + C(1) R − C(2) R − C(3) R R R
da R + C(1) R + C(2) R + C(3) R R R
ω 0 3π /(2k) ; π /k ; 0 0 0 0
π /(2k)
Shape: Circular Elliptical Elliptical Circular Circular, inclined Circular, inclined
4.1.1 Circular Perturbation
If we set all parameters but C(1) equal to zero, the perturbed orbit remains in the
reference orbital plane and has still a circular shape. This perturbed orbit is given by
r = R + ηr = R + C(1) , (30a)
3 m C (1)
t (s) = Ṫ s + η t (s) = Ṫ s − Ṫ s =: Ṫ + δt(1) s, (30b)
2 R − 3m R
3 R − 2m C(1)
ϕ(s) = s + η ϕ (s) = s − s =: ( + δω(1) ) s. (30c)
2 R − 3m R
The reference and the perturbed orbit are shown in Fig. 3 for one reference period
and a chosen reference radius of 5000 km above the surface of the Earth. As one
would expect, a positive radial perturbation C(1) yields a smaller azimuthal frequency,
ϕ̇ = + δω(1) < , as compared to the reference motion. The frequency and
radial perturbations are related via
δω(1) 3 R − 2m C(1)
=− , (31)
2 R − 3m R
Fig. 3 The solid line (black) shows the circular reference orbit, whereas the dashed line (green)
shows the perturbed orbit as calculated with the solution of the deviation equation for: only C(1) = 0
(top, left), only C(2) = 0 (top, middle), only C(3) = 0 (top, right), a combination of both such
that initially r (0) = R, ṙ (0) = 0 (bottom left) and a pendulum orbit as the result of an inclined
perturbation using only C(5) (bottom middle) and C(6) (bottom right). We have marked the respective
positions Yn on the reference orbit and X n on the perturbed orbit for reference proper time values
s = n/4 · 2π/ . We used a reference radius of R = R⊕ + 5000 km and a mean Earth radius
R⊕ = 6.37 · 103 km.eps
such that they are not independent. Since we work with the first order deviation equa-
tion, we have to ensure that the radial perturbation is indeed small, i.e. C(1) /R 1.
This is related to upper bounds for C(1) that need to be chosen in a proper way.
For a satellite orbit of about 104 km above the surface of the Earth, the normalized
perturbation is C(1) /R ≈ 10−7 m−1 C(1) . Hence, the allowed values for C(1) strongly
depend on the chosen reference radius and given upper bounds for the radial pertur-
bation. For various values of the reference radius - ranging from Low Earth Orbits
(LEO) to geostationary ones - and the parameter C(1) , we show the magnitude of the
normalized radial perturbation C(1) /R in Fig. 4. To decide whether the description
of a satellite configuration within the framework of the first order deviation equation
is useful or not, one has to define the reference radius and the maximal radial pertur-
bation for the desired scenario. The value C(1) /R can then be estimated from Fig. 4
and if it fulfills the condition C(1) /R 1 the solution may give a simple and useful
description.
After a full azimuthal period on the reference orbit, s = 2π/ , and the perturbed
orbit is not yet closed since ϕ(2π/ ) = 2π. The deficit angle α is given by
δω(1) R − 2m C(1)
α = ϕ (2π/ ) − 2π = 2π = −3π . (32)
R − 3m R
Fig. 4 The magnitude of the normalized radial perturbation C(i) /R as a function of C(1,2,3) and
the reference radius R. The lines represent surfaces of constant C(i) /R. We use a mean Earth radius
R⊕ = 6.37 · 103 km
This angle may correspond, in principle, to an observable quantity and the relation
can be solved for m explicitly,
R(3C(1) π + αR)
m= , (33)
3(2C(1) π + αR)
to obtain an estimate for the relativistic mass monopole of the central object. Thus,
the mass can be obtained from the measurement of the deficit angle, assuming the
situation can be prepared with initially known reference radius R and radial distance
C(1) between both orbits. Also the deviation of a test object from the center of mass
within a hollow satellite might be used for such a measurement.
4.1.2 Elliptical Perturbation I
The two parameters C(2,3) cause elliptical perturbations in the (reference orbital
plane) if we neglect the influence of all other parameters, i.e. if C(1,4,5,6) = 0. If we
choose to have C(2) = 0, the perturbed orbit is described by

r (s) = R + ηr (s) = R + C(3) cos ks = R + C(3) + O s 2 , (34a)
t (s) = Ṫ s + η (s)
t

2R m
= Ṫ s − C(3) sin ks
R − 2m R − 6m

=: Ṫ s + δt(3) sin ks = (Ṫ + δt(3) k) s + O s 2 (34b)

R C(3)
ϕ(s) = s − 2 sin ks
R − 6m R

:= s + δω(3) sin ks = ( + δω(3) k) s + O s 2 . (34c)
For an elliptically perturbed orbit of this kind, the eccentricity e and the semi major
axis a can be linked to the radial perturbation via
C(3)
e= , a = R. (35a)
R
We can as well calculate the distance to the perigee d p and apogee da that are related
to the radial perturbation
d p = R − C(3) , da = R + C(3) , (35b)
and confirm that the semi major axis is half of the sum of the two distances as it should
be. Using these relations the spatial shape of the perturbed orbit can be represented
in a familiar way as
r (s) = a(1 + e cos ks) , (36a)

a
ϕ(s) = s − 2 e sin ks . (36b)
a − 6m
The eccentricity of the perturbed orbit e = C(3) /R is shown in Fig. 4. If for a specific
satellite mission the maximal allowed eccentricity is given, we can read off upper
bounds for the parameter C(3) from Fig. 4, or, vice versa: we can model an orbit with
a given (small) eccentricity by choosing the necessary value for C(3) .
The difference between the effects of the two parameters C(2,3) is just a phase
difference, i.e. a spatial rotation of π/2 of the perturbed orbit within the reference
orbital plane. For the case that only C(2) = 0, the perturbed orbit is described by

r (s) = R + ηr (s) = R + C(2) sin ks = R + C(2) ks + O s 2 , (37a)
t (s) = Ṫ s + η (s)
t

2R m
= Ṫ s + C(2) cos ks
R − 2m R − 6m

=: Ṫ s + δt(2) cos ks = Ṫ s + δt(2) + O s 2 , (37b)

R C(2)
ϕ(s) = s + 2 cos ks
R − 6m R

=: s + δω(2) cos ks = s + δω(2) + O s 2 . (37c)
Both elliptical orbits are shown in Fig. 3 and the spatial rotation as the difference
between the effects of C(2) and C(3) is obvious. These orbits look closed after one
reference period, but they are not (recall that the frequencies and k are just slightly
different). After one reference period s = 2π/ we get
2πk
r (2π/ ) = R + C(3) cos = r (0) , (38a)

2πk
ϕ(2π/ ) = 2π + δω(2) sin = ϕ(0) + 2π . (38b)

However, the difference between the two frequencies is in the range of some 10−12 Hz
and the periods differ by about 10–25 µs for reference radii in the range from LEO
to geostationary orbits. The radial motion has an actual period of s = 2π/k, i.e. this
amount of reference proper time elapses from one perigee to the next. Hence, we get
r (2π/k) = r (0) , (39a)

2π
ϕ(2π/k) = = ϕ(0) + 2π . (39b)
k
Since the increase in the azimuthal angle differs from 2π after one radial period, the
perigee of the elliptical orbit will precess. This precession is investigated in the next
section in more detail.
4.1.3 Elliptical Perturbation II
Another kind of elliptical orbits can be constructed via the combination with a circular
perturbation. For example, we use a combination of both, C(1) and C(3) , to arrive at
a perturbed orbit that initially fulfills r (0) = R , ϕ(0) = (0) = 0 and ṙ (0) = 0.
Hence, the perturbed orbit is initially as close as possible to the reference orbit - cf.
Fig. 3. To construct it we need to choose C(3) = −C(1) . The perturbed orbit is then
described by
r (s) = R + C(1) (1 − cos ks) , (40a)

t (s) = (Ṫ + δt(1) ) s + δt(3) sin ks , (40b)
ϕ(s) = ( + δω(1) ) s + δω(3) sin ks , (40c)
ϑ = ≡ π/2 . (40d)
For this orbit type we find the eccentricity and semi major axis to be
C(1)
e= , a = R + C(1) ⇒ C(1) = ae , (41)
R + C(1)
and this allows to recast the radial motion, again, in the familiar way
r (s) = a(1 − e cos ks) . (42)
This radial motion has a period of 2π/k and we obtain
r (2π/k) = r (0) , (43a)

2π( + δω(1) )
ϕ(2π/k) = = ϕ(0) + 2π . (43b)
k
Hence, also this elliptical orbit will precess. The precession is studied in the next
section in terms of the perigee advance.
4.1.4 Azimuthal Perturbation
The parameter C(4) causes an offset in the azimuthal motion, i.e. a constant phase
difference between the reference and perturbed orbit. When only C(4) = 0 the per-
turbed orbit is the same as the reference orbit, but the two test bodies are separated
by a constant angle. The perturbed motion is then described by
r = R, (44a)
√
mR
t (s) = Ṫ s + η (s) = Ṫ s +
t
C(4) , (44b)
R − 2m
C(4)
ϕ(s) = s + η ϕ (s) = s + , (44c)
R
where the constant azimuthal separation is determined by the value of C(4) /R.
4.1.5 Inclined Perturbation
The two parameters C(5) , C(6) incline the orbital plane with respect to the reference
plane. If only C(5) = 0 we obtain a circular orbit with radius r = R and azimuthal
motion ϕ(s) = s, but with the polar motion given by
π C(5)
ϑ(s) = + cos s . (45)
2 R
Hence, C(5) /R determines the maximal inclination between the two orbital planes. If
only C(6) = 0 instead, there are just little changes: the cos( s) becomes sin( s)
and the difference between the effects of these two parameters is simply related to a
spatial rotation.
Fig. 5 A sketch of the two orbital planes including the orbital elements of the perturbed orbit X a
4.2 The Orbital Elements
Combining the results of the last section we can link all parameters C(1,...,6) to the
initial position and velocity of the test body that follows the perturbed orbit. The initial
quantities r (0), ϑ(0), ϕ(0) and ṙ (0), ϑ̇(0), ϕ̇(0) are summarized in Table 1 together
with the resulting orbital elements of the perturbed orbit. The orbital elements are
introduced in the sketch shown in Fig. 5.
4.3 Physical Effects
As shown before, the parameters C(2,3) lead to an elliptically perturbed orbit if at

least one of them does not vanish. For such an orbit the perigee will precess and
the orbit is not closed. If either of the parameters C(2,3) = 0 the radial motion has
a period given by s = 2π/k, but the azimuthal oscillation is advanced already. The
difference to a full revolution is then given by

a
ϕ = ϕ(2π/k) − 2π = 2π − 1 = 2π −1 , (46)
k a − 6m
where a is the orbit’s semi-major axis. Figure 6 shows this precession of the perigee
for different reference radii ranging from LEO to geostationary orbits. The result
(46) is the same as shown in Ref. [14] and was also derived in [13] as well.3 Up to
linear order in m/a we obtain the well-known result
6πm
ϕ = + O (m/a)2 , (47)
a
3 Notethe misprint in Eq. (4.14) in [13], where actually the inverse value of the correct result is
shown and we assume this to be simply a typo.
Fig. 6 The perigee shift per orbit in 10−3 arcs for the elliptically perturbed orbit with either C(3) = 0
or C(2) = 0 (left) This orbit is shown in Fig. 3 in the middle of the upper row
that is the first term in Einstein’s result [15]
6πGm 6πm
ϕ = ≈ (1 + e2 + e4 + · · · ) , (48)
a(1 − e )
2 a
for the precession of the perigee in the case of small eccentricities. For the second kind
of an elliptically perturbed orbit that is described by Eq. (40) we obtain according to
Eq. (43)
2π( + δω(1) )
ϕ = − 2π
k

R R − 2m R C(1)
= 2π − 1 − 3π . (49)
R − 6m R − 3m R − 6m R
The first term resembles the previous result for the perigee precession where the
parameter C(1) was set equal to zero and we recover this result in the limit. Up to
linear order in e and m/a the result reads
6πm m
ϕ = 1− 1+ e + O e2 , (m/a)2 , (50)
a 2a
and depends on the eccentricity, whereas in the first case the result was independent
of the perturbation parameters.
It should be mentioned that in the Newtonian solution (12) no perigee precession
is present since there is only one frequency, the Keplerian frequency K , involved.
In Newtonian gravity (at least for a spherically symmetric potential) the Kepler
ellipses are closed. Hence, as it is well-known, the precession of an elliptical orbit
is a relativistic effect. It is recovered in the framework of the first order geodesic
deviation equation due to the appearance of a second frequency in the relativistic
solution.
4.4 Redshift and Time Dilation
The redshift z between two standard clocks that show proper times s and s̃ is
ν d s̃
1+z = = . (51)
ν̃ ds
Using the solution of the first order deviation equation, we can derive a formula for
the redshift between the clocks transported along the reference and deviating orbit
as follows. Along the orbit X a (s), the constant of motion E X related to the energy is

2m dt 2m dt ds
EX = 1 − = 1− . (52)
r (s̃) d s̃ r (s) ds d s̃
Hence, we obtain for the redshift using the solution t (s) and r (s) of the first order
deviation equation

d s̃ 2m Ṫ + η̇ t (s)
z+1= = 1−
ds r (s) EX
√
R mR ˙
+ f (s)
2m R − 3m R − 2m
= 1− , (53)
R + g(s) EX
where the functions f (s) and g(s) are given by Eq. (25) and E X is fixed by the initial
conditions of the deviating orbit, i.e. by the choice of parameters C(i) .

1 2m
z+1= 1−
EX R + C(1) + C(2) sin ks + C(3) cos ks

R
× + λ(1) + λ(2) sin ks + λ(3) cos ks , (54)
R − 3m
where
√ √
3 m(R − 2m) mR
λ(1) =− C(1) , λ(2,3) = −2 √ C(2,3) . (55)
2 R(R − 3m)3/2 (R − 2m) R − 6m
This result yields a comparatively simple model for the redshift between the two
satellites and is accurate as long as the orbital deviation is small.
For two circular orbits with radii R and R + C(1) , we recover the correct result to
first order in C(1) /R. The redshift becomes
C(1) 3m
z= . (56)
R 2(R − m)
Note however, that terms related to Doppler effects are not present here, since we do
not consider signals (light rays) send from one orbit to the other. Hence, the formula
for the redshift contains only the part related to time dilation effects.
4.5 Accuracy of the First Order Deviation Approach
Exact solutions of the geodesic equation in the Schwarzschild spacetime can be

constructed using elliptic functions. To our knowledge, the first work on this is
contained in Refs. [16, 17]. The authors used the Jacobi elliptic functions sn, cn, dn
to solve the equation of motion. A more recent study of exact orbital solutions
in Schwarzschild spacetime (and generalizations for, e.g., Kerr–Newman–deSitter
spacetime) can be found in Refs. [18–20], where the Weierstrass elliptic function ℘
is used.
Note that especially in the solutions of the geodesic equation that involve the
Jacobi elliptic function sn, cn, dn the relation to the solutions (24) of the first order
deviation equation is obvious. Take two such exact solutions and construct the devi-
ation between the two orbits as their difference. Then, choosing one of these orbits to
be the circular reference geodesic, the linearization of the deviation corresponds to
the solution of the first order deviation equation. In Eq. (24) sin and cos terms appear
and these are the linearizations of the Jacobi elliptic functions sn and cn.
4.5.1 Circular Orbits
For a circular orbit in the equatorial plane with radius r = R + C(1) , the exact
azimuthal frequency is given by

m R + C(1)
ωϕ = . (57)
(R + C(1) )3 R + C(1) − 3m
Expanding this result w.r.t. the small quantity C(1) /R 1 in a Taylor series, we find
for the azimuthal frequency

m R 3 R − 2m m R C(1)
ωϕ = −
R3 R − 3m 2 R − 3m R3 R − 3m R

(0) (1)
= =:δω(1) =:δω(1)
2
15 (R − 2m)2 + 4/5m m 2
R C(1)
+
8 (R − 3m) 2 R3 R − 3m R

(2)
=:δω(1)
3
+ O C(1) /R . (58)
(i)
The quantities δω(1) are defined as shown above. The superscript denotes the order
of expansion and the subscript denotes the connection to the radial perturbation C(1) .
The 0th order contribution is given by , the azimuthal frequency for a circular
orbit with radius R - cf. Eq. (26). Restricting ourselves to first order contributions
we can compare the approximation (58) to the solution of the first order deviation
equation (30)
(1)
+ δω(1) ≡ + δω(1) . (59)
Hence, the approximation up to linear order in C(1) /R is exactly the result that
appeared in the solution of the first order deviation equation - cf. Eq. (30). Therefore,
the error that we make using this result is dominated by the second order term
(2) (2)
δω(1) . We show the relative error δω(1) /ωϕ in Fig. 7 and conclude that even for a
2
radial separation of some 10 km between the reference and perturbed orbit this
error is less than 0.1%. Using the expansion in Eq. (58), we estimate the error that
remains at the next order, i.e. using also second order contributions in the frequency
expansion. Then, the error is dominated by the third order term and about two orders
(2)
of magnitude smaller. One can show that the quantity δω(1) /ωϕ , shown in Fig. 7,
is also the dominating error term for the conserved quantity L that corresponds
to the angular momentum of the perturbed orbit, since the frequency and angular
momentum are related by L = r 2 ωϕ .
Using the expansion (58), we can write down the solution of the nth order deviation
equation for a circular perturbation in the reference orbital plane. The perturbed orbit
is then described by
r = R + C(1) , (60a)

(1) (n)
ϕ = + δω(1) + · · · + δω(1) s, (60b)
where
k
(k) 1 d k ωϕ C(1)
δω(1) = . (60c)
k! d(C(1) /R)k R
(2)
Fig. 7 The dominating term δω(1) /ωϕ in the frequency error. This relative error is made using the
solution of the first order deviation equation to describe a circular perturbed orbit with initial radial
distance C(1) to a reference orbit with radius R. Hence, the solution of the first order deviation
equation gives the correct result up to a few parts in one hundred
Table 2 Error in the distance between reference and perturbed orbit after one complete reference
period. We compare the distance as modeled with Eq. (60), up to 4th order, with the distance
constructed from two exact circular orbits with radii and frequencies (R, ) and (R + C(1) , ωϕ ).
The table shows the values for the two different reference radii, 1000 km and 36000 km above
Earth’s surface. Bold marked values correspond to cm accuracy level when using the respective
approximation order
Radial separation Error (m)
C(1) 1st 2nd 3rd 4th
Order
R = R⊕ + 1000 km
10 km 159 0.25 4 · 10−4 6 · 10−5
50 km 3955 31 0.24 0.002
100 km 15700 249 3.8 0.06
150 km 35000 535 19 0.43
R = R⊕ + 36000 km
10 km 27 0.007 10−5 10−5
50 km 690 0.95 0.001 5 · 10−5
100 km 2760 7.6 0.02 5 · 10−5
150 km 6206 26 0.1 0.0015
Table 2 shows the error that is made when modeling the distance between the ref-
erence orbit and the circular perturbed orbit after one complete reference period.
To calculate the error we used the result above and compared it with the distance
as calculated from two exact circular orbits with radii and frequencies (R, ) and
(R + C(1) , ωϕ ).
4.5.2 Elliptical Orbits
To judge the accuracy of the elliptical orbits constructed with the solution (34), we
compare them to exact solutions of the geodesic equation. In both cases we have
to use identical initial conditions, i.e. the constants of motion E and L have to be
the same. The exact orbit can either be constructed as a numerical solution of the
geodesic equation, or by using analytic solutions in terms of elliptic functions.
We choose as initial conditions for eccentricity, distance to the perigee and argu-
ment of the perigee
e = 0.02 , d p = R⊕ + 9672.6 km , ω = 0 . (61)
The relative error in the radial and azimuthal motion is shown in Fig. 8 for ten orbital
periods. The mean error in the radial deviation is positive, whilst the mean phase
error is negative. Lowering the eccentricity by a factor of 10 yields relative errors that
decrease two orders of magnitude. The analysis shows that the errors scale roughly
as e2 , which is the expected behavior since terms that are quadratic in the eccentricity
are neglected in this framework and contribute to second order deviations.
4.5.3 Pendulum Orbits
Pendulum orbit constellations are of special interest for satellite geodesy missions.
Hence, we should give an idea of how accurate the modeling of these constellations
can be done using the solutions (45) of the first order geodesic deviation equation.
Equation (45) describes a circular orbit with unchanged radius that is inclined w.r.t
the reference orbit. Thus, all constants of motion are the same but the orbital planes
differ. In this solution, the amplitude C(6) /R of the ϑ-oscillation gives the inclination
of the perturbed orbit. Since we work with first order perturbations, this amplitude
has to be small. For a reference radius of GRACE-type, 500 km above Earth’s surface,
and an inclination of 1◦ the error is about 100 marcs. For two orbital periods, we
show in Fig. 9 the relative error for an even higher inclination of 5◦ . As can be seen
in the figure, the error in the ϑ-motion is periodic with zero mean value and standard
deviation around 0.1 % for this situation.
Fig. 8 Relative error in the radial and azimuthal motion compared to the exact solution for 10
periods (10 T ). The mean value (solid line) and range of one standard deviation around the mean
(dashed lines) are shown
4.6 Other Effects
4.6.1 The Line of Nodes
In Schwarzschild spacetime, due to symmetry, any orbit is confined to one orbital

plane that is determined by the initial conditions. Conventionally this plane is then
chosen to be the equatorial plane, defined by ϑ = π/2, ϑ̇ = ϑ̈ = 0. Let there be two
different bound orbits in Schwarzschild spacetime that we describe in the coordinate
basis. We define the equatorial plane as the plane of motion for one of them. The
Fig. 9 Relative error in the ϑ-motion that is made using the solution (45) of the first order deviation
equation, which describes a pendulum orbit constellation with an inclination of 5◦ between both
orbital planes. The mean value (solid line) and one standard deviation around the mean (dashed
lines) are shown as well
second orbit is, of course, confined to a plane as well but this orbital plane is inclined
w.r.t. the first one. Define the line of nodes as the spatial intersection of these two
planes, i.e. the connection between the points where the second orbit crosses θ = π/2.
Since each of the two orbits is confined to its orbital plane, the line of nodes remains
unchanged for an arbitrary number of revolutions.
We will now describe these two orbits using the solution (24) of the first order
deviation equation. The circular reference orbit with radius R defines the equatorial
plane. We choose C(2,...,5) ≡ 0. Hence, only the parameters C(1,6) describe the shape
of the perturbed orbit. As our analysis in the previous sections has shown, C(1)
causes a constant radial perturbation and C(6) inclines the perturbed orbital plane.
This perturbed orbit is described by
r = R + C(1) , (62a)
ϕ(s) = ( + δω(1) ) s , (62b)
C(6)
ϑ(s) = π/2 + sin s . (62c)
R
The line of nodes is determined by two successive intersections of the ϑ-motion with
the equatorial plane. This happens for s = nπ/ , n = 0, 1, 2 . . .. For these values
we get
ϑ(s = nπ/ ) = π/2, (63a)

nπδω(1)
ϕ(s = nπ/ ) = nπ + . (63b)

Thus, from one orbit to the next the line of nodes shifts by an amount of
2πδω(1) R − 2m C(1)
n := = −3π . (64)
R − 3m R
For C(1) /R = 1% this corresponds to about −5◦ . Since a precession of the line of
nodes must not happen in Schwarzschild spacetime, we have to carefully analyze this
kind of “effect”. As we have shown in the expansion of the exact azimuthal frequency
(58) for a circular orbit with radius R + C(1) , is its 0th order and δω(1) its 1st
order approximation. The approximation order in the frequency of the ϑ-motion in
(62) is one less than the order in the ϕ-motion, because the amplitude C(6) /R of the
ϑ-motion is already of first order. Hence, when using higher order deviation solutions
up to jth order, the orbit will be given by
r = R + C(1) , (65a)
(1) ( j)
ϕ = ( + δω(1) + · · · + δω(1) ) s , (65b)
(1) ( j−1)
ϑ = π/2 + C(6) sin( + δω(1) + ··· + δω(1) ) s , (65c)
and subsequently the shift of the line of nodes after one orbit is
( j)
2πδω(1)
n = (1) ( j−1)
. (66)
+ δω(1) + · · · + δω(1)
( j)
Since δω(1) is the jth term in the Taylor expansion (58) we notice that n → 0 for
j → ∞. Hence, the shift of the line of nodes vanishes in the limit of infinite accuracy
and is simply an artifact of the linearization/approximation (Fig. 10).
4.6.2 Shirokov’s Effect Revisited
In [3] a new effect in the context of the standard geodesic deviation equation
in Schwarzschild spacetime was reported. This effect was also studied in several
follow-up works [21–25], in particular generalizations to other spacetimes than
Schwarzschild were given and Shirokov’s effect was compared to the influence of the
oblateness of the Earth that causes similar perturbations of the reference geodesic.
In the following we critically reassess the derivation, meaning, and physical mea-
surability of Shirokov’s effect. Shirokov [3] did only consider periodic solutions
of the standard geodesic deviation equation in Schwarzschild spacetime. The solu-
tion that is given in the work [3] is obtained in the framework presented here for
C(1,3,4,5) = 0. Only the parameters C(2,6) are involved. According to the previous
analysis of the shape of the perturbed orbit this corresponds to an elliptical orbit in
an inclined orbital plane. This orbit is given by, cf. Eq. (21) in [3],
Fig. 10 The reference and perturbed orbit for three reference periods. We have marked one endpoint
of the line of nodes X n after n elapsed reference periods. The precession of the perturbed orbital
plane is clearly visible
r (s) = R + ηr (s) = R + C(2) sin ks , (67a)

ϕ
ϕ(s) = s + η (s) = s + δω(2) cos ks , (67b)
C(6)
ϑ(s) = π/2 + η ϑ (s) = π/2 + sin s . (67c)
R
There are some subtleties that need to be handled with care. First of all, having
identified a feature of a physical system within an approximate description, like the
first order deviation equation, does not mean that this corresponds to a “real” effect.
We have pointed out an example for this with the line of nodes precession that is
present in the linearized framework but was shown to decrease with higher order
approximations and to vanish in the end. Exactly this artifact of the line of nodes
shift is present in Shirokov’s solution as well. Since the equatorial plane is intersected
by the perturbed orbit for s = nπ/ , we get a shift of the line of nodes after one
ϑ-period
2πk
ϕ = ϕ(s = 2π/ ) − 2π = δω(2) cos . (68)

This shift was not mentioned in the work by Shirokov, but is indeed present in the
solution. However, Shirokov noticed correctly that in (67) the r , ϕ and θ oscillations
have different frequencies and, thus, different periods
Tr = Tϕ = 2π/k , (69a)
Tϑ = 2π/ . (69b)

Linearized in m/R around the Keplerian value TK = 2π/ K = 2π R 3 /m, we
obtain

3m
Tr = Tϕ = TK 1 − , (70a)
R

3m
Tϑ = TK 1 + . (70b)
R
It is by no means obvious why this linearization should be necessary, but it was used
in [3]. Of course, for the Earth m/R is a very small quantity when R corresponds
to radii above the surface, but the geodesic deviation equation works well even
close to black holes, were m/R might be large [12]. However, Shirokov concludes
that, due to the different periods of ϑ and r oscillations, the distance R (ϑ − π/2)
to the equatorial plane, in which the reference orbit lies, is different from 0 after
(several) radial oscillations and this is a new effect of GR. Shirokov imagines a
satellite that moves on the reference orbit and rotates around the axis perpendicular
to the equatorial plane with its orbital period, i.e. 2π/ . Placed within this satellite
a small test mass shall follow the perturbed orbit and the different frequencies of
oscillations in the ϑ and r direction can be observed.
Since each orbit in Schwarzschild spacetime is confined to its orbital plane, the
ϑ and ϕ frequencies have to be equal for exact orbits. Otherwise the line of nodes
would shift. The difference in these periods that Shirokov discovered is a result of
the approximation. However, the r and ϕ periods can be different, which leads to
the perigee precession for elliptical orbits. Hence, the r and ϑ periods can also be
different for elliptical orbits to allow for a perigee precession within an inclined
orbital plane. This is exactly what the approximate solution (67a) of the first order
deviation equation describes: an elliptical orbit with perigee precession in an inclined
orbital plane. The exact solution for this orbit would describe an elliptical orbit in
an inclined but fixed orbital plane that shows perigee precession within this orbital
plane. The solution (67a) is simply the first order approximation of this situation.
Since the radial and the polar period are different, one would observe the object to be
above or below the equatorial plane after (several) radial periods. This is nothing but
a precessing ellipse in the inclined orbital plane. Hence, Shirokov’s “new” effect is
not new at all, it is the first order approximation of the well-known perigee precession
– discovered by Einstein already in 1916 [15] – in an inclined plane. Furthermore,
as we have shown, this is mixed with the artifact of a precessing line of nodes due to
the linearized framework.
We conclude that Shirokov’s effect is no new effect but the first order description
of the perigee shift for an elliptical orbit in an inclined orbital plane.
5 Conclusions
We have shown how to describe orbits using the solution of the first order deviation
equation for circular reference geodesics. In particular, we employ the Schwarzschild
spacetime as the simplest approximation for the Earth to investigate relativistic
satellite orbits and orbit deviations. We describe the shape of all perturbed orbits
and connect free parameters in the general solution to the orbital elements of the
perturbed orbit. Using this description, one can now apply the solution of the first
order deviation equation to model any orbit that is specified in terms of its orbital ele-
ments. We have uncovered artificial effects that are due to the linearized framework.
For elliptical orbits with small eccentricities the perigee precession was derived as a
purely relativistic effect, which is absent in the Newtonian solution of the deviation
problem. The solution of the first order deviation problem in Schwarzschild space-
time has shown that such an approximate description must be handled with care. The
line of node precession, which is forbidden in Schwarzschild spacetime, will mix in
the context of Kerr spacetime with the Lense-Thirring effect that causes a similar
behavior.
Reconsidering the so-called Shirokov effect we uncovered its origin. Rather than
being a new feature of GR, we identified it as the relict of the approximate description
of a well-known perigee precession.
The comparison of perturbed orbits – based on the solution of the geodesic devi-
ation equation – to exact solutions of the underlying geodesic equation has shown,
that higher order deviation equations should be used to model modern satellite based
geodesy missions.
In (simple) spacetimes, for which analytic solutions of the geodesic equation are
available, one can estimate the accuracy of the approximate description to any order.
In more realistic spacetimes numerical methods will become necessary.
In conclusion, already at the present level of accuracy applications in geodesy
and gravimetry require the use of higher order deviation equations. Such equations
will become indispensable for the description of future high precision satellite and
ground based measurements.
Acknowledgements The present work was supported by the Deutsche Forschungsgemeinschaft

(DFG) through the grant PU 461/1-1 (D.P.), the Sonderforschungsbereich (SFB) 1128 Relativistic
Geodesy and Gravimetry with Quantum Sensors (geo-Q), and the Research Training Group 1620
Models of Gravity. We also acknowledge support by the German Space Agency DLR with funds
provided by the Federal Ministry of Economics and Technology (BMWi) under grant number DLR
50WM1547. The authors would like to thank V. Perlick, J.W. van Holten, and Y.N. Obukhov for
valuable discussions.
A Conventions and Symbols
In the following we summarize our conventions, and collect some frequently used
formulas. A directory of symbols used throughout the text can be found in Table 3.
The signature of the spacetime metric is assumed to be (+, −, −, −). Latin indices
i, j, k, . . . are spacetime indices and take values 0 . . . 3. For an arbitrary k-tensor
Ta1 ...ak , the symmetrization and antisymmetrization are defined by
Table 3 Directory of symbols

Symbol Explanation
Geometrical quantities
gab Metric
A(r ), B(r ) Free metric functions
√
−g Determinant of the metric
δba Kronecker symbol
xa, s Coordinates, proper time
Y a, Xa (Reference, perturbed) curve
ab c Connection
Rabc d Curvature
ηa Deviation vector
Physical quantities
G, U Newtonian gravitational (constant, potential)
R Reference radius
K , TK Keplerian frequency, period
, ωϕ Azimuthal freq. of circ. orbit
k Second freq. in relativistic solution
Tr , Tϕ , Tϑ Frequencies of perturbations
C(i) Perturbation parameters, i = 1 . . . 6
δω(i) , δt(i) (Azimuthal, temporal) deviations, i = 1 . . . 3
R⊕ Mean Earth radius 6.37 · 103 km
ϕ, n Angle of (perigee precession, line of nodes
shift)
E Energy
L Angular momentum
ρ Matter density
M, m Mass of the central object (kg), (m)
f (s), g(s), Abbreviations
Orbital elements
a Semi major axis
e Eccentricity
a Ascending node
i Inclination
d p , da Distance to (perigee, apogee)
ω Argument of the perigee
ν True anomaly
Operators
∂i , ∇i (Partial, covariant) derivative
ds = “˙”
D
Total covariant derivative
1
k!
T(a1 ...ak ) := Tπ {a ...a } , (71)
k! I =1 I 1 k
1
k!
T[a1 ...ak ] := (−1)|π I | Tπ I{a1 ...ak } , (72)
k! I =1
where the sum is taken over all possible permutations (symbolically denoted by
π I{a1 . . . ak }) of its k indices.
The covariant derivative defined by the Riemannian connection is convention-
ally denoted by the nabla or by the semicolon: ∇a = “;a ”. Our conventions for the
Riemann curvature are as follows:
2 Ac1 ...ck d1 ...dl ;[ba] ≡ 2∇[a ∇b] Ac1 ...ck d1 ...dl

k
l
= Rabe ci Ac1 ...e...ck d1 ...dl − Rabd j e Ac1 ...ck d1 ...e...dl . (73)
i=1 j=1
References
1. F. Flechtner, K.-H. Neumayer, C. Dahle, H. Dobslaw, E. Fagiolini, J.-C. Raimondo, A. Guent-

ner, What can be expected from the GRACE-FO laser ranging interferometer for earth science
applications? Surv. Geophys. 37, 453 (2016)
2. B.D. Loomis, R.S. Nerem, S.B. Luthcke, Simulation study of a follow-on gravity mission to
GRACE. J. Geod. 86, 319 (2012)
3. M.F. Shirokov, On one new effect of the Einsteinian theory of gravitation. Gen. Relativ. Gravit.
4, 131 (1973)
4. D. Philipp, V. Perlick, C. Lämmerzahl, K. Deshpande, On geodesic deviation in Schwarzschild
spacetime, in IEEE Metrology for Aerospace (2015), p. 198
5. P.J. Greenberg, The equation of geodesic deviation in Newtonian theory and the oblateness of
the earth. Nuovo Cimento 24B, 272 (1974)
6. W.G. Dixon, The new mechanics of Myron Mathisson and its subsequent development, in
Equations of Motion in Relativistic Gravity, ed. by D. Puetzfeld et al. Fundamental theories of
Physics, vol. 179 (Springer, Berlin, 2015), p. 1
7. Y.N. Obukhov, D. Puetzfeld, Multipolar test body equations of motion in generalized gravity
theories, in Equations of Motion in Relativistic Gravity, ed. by D. Puetzfeld et al. Fundamental
theories of Physics, vol. 179 (Springer, Berlin, 2015), p. 67
9. H. Fuchs, Conservation laws for test particles with internal structure. Annalen der Physik 34,
159 (1977)
10. H. Fuchs, Solutions of the equations of geodesic deviation for static spherical symmetric space-
times. Annalen der Physik 40, 231 (1983)
11. R. Kerner, J.W. van Holten, R. Colistete Jr., Relativistic epicycles: another approach to geodesic
deviations. Class. Quantum Gravity 18, 4725 (2001)
12. G. Koekoek, J.W. van Holten, Epicycles and Poincaré resonances in general relativity. Phys.
Rev. D 83, 064041 (2011)
13. H. Fuchs, Deviation of circular geodesics in static spherically symmetric space-times. Astron.
Nachr. 311, 271 (1990)
14. C. Misner, K. Thorne, J.A. Wheeler, Gravitation (Freeman, San Francisco, 1973)
15. A. Einstein, Die Grundlage der allgemeinen Relativitätstheorie. Annalen der Physik 354, 769
(1916)
16. A.R. Forsyth, Note on the central differential equation in the relativity theory of gravitation.
Proc. R. Soc. Lond. A 97(682), 145 (1920)
17. W.B. Morton, LXI. The forms of planetary orbits on the theory of relativity. Lond. Edinb.
Dublin Philos. Mag. J. Sci. 42(250), 511 (1921)
18. E. Hackmann, C. Lämmerzahl, Complete analytic solution of the geodesic equation in
Schwarzschild (anti-)de Sitter spacetimes. Phys. Rev. Lett. 100, 171101 (2008)
19. E. Hackmann, C. Lämmerzahl, Geodesic equation in Schwarzschild-(anti-)de Sitter space-
times: analytical solutions and applications. Phys. Rev. D 78 (2008)
20. E. Hackmann, V. Kagramanova, J. Kunz, C. Lämmerzahl, Analytic solutions of the geodesic
equation in axially symmetric space-times. Europhys. Lett. 88, 30008 (2009)
21. A. Nduka, On Shirokov’s one new effect of the Einsteinian theory of gravitation. Gen. Relativ.
Gravit. 8, 347 (1977)
22. Yu.S. Vladimirov, R.V. Rodichev, Small oscillations of test bodies in circular orbits in the
Schwarzschild and Kerr metrics (Shirokov’s effect). Sov. Phys. J. 24, 954 (1981)
23. W. Zimdahl, Shirokov effect and gravitationally induced supercurrents. Exp. Tech. Phys. 33,
403 (1985)
24. L. Bergamin, P. Delva, A. Hees, Vibrating systems in Schwarzschild spacetime: toward new
experiments in gravitation? Class. Quantum Gravity 26, 18006 (2009)
25. W. Zimdahl, G.M. Gilberto, Temperature oscillations of a gas in circular geodesic motion in
the Schwarzschild field. Phys. Rev. D 91, 024003 (2015)
Measurement of Frame Dragging with
Geodetic Satellites Based on Gravity
Field Models from CHAMP, GRACE
and Beyond
Rolf König and Ignazio Ciufolini
Abstract The experimental measurement of frame-dragging or the Lense-Thirring

(LT) effect based on Satellite Laser Ranging (SLR) observations to the LAGEOS
satellites was successfully demonstrated with an accuracy of about 10%. Here we
look in detail into the effect of the node drift induced by the time variable part of
the C(2,0) term of the gravity field model describing the flattening of the Earth. We
demonstrate that errors in C(2,0) can effectively be taken care of by analyzing two
satellites for the LT measurement. We also adopt some recent gravity field models
in order to independently repeat and extend the LT experiments so far. The gravity
field models used for this are derived either partly depending on LAGEOS SLR
observations or completely independent from LAGEOS, and based on dedicated
gravity field satellite missions like CHAMP, GRACE and GOCE. It turns out that
from all the gravity field models tested the claimed accuracy of 10% of the LT
measurement can be confirmed.
1 Introduction
Frame-dragging, as named by Albert Einstein [1], is an intriguing phenomenon of

General Relativity (GR) that predicts that the orbit of particles, the direction deter-
mined by gyroscopes, the path followed by photons, and even the time marked by
clocks, all are affected by mass-energy currents such as the rotation of a mass. In
1918 Lense and Thirring [2] published the equations for the frame-dragging effect
on the orbit of a particle around a rotating central body, e.g. for a satellite orbit-
R. König (B)
Helmholtz-Zentrum Potsdam Deutsches GeoForschungsZentrum GFZ, c/o DLR
Oberpfaffenhofen, Wessling 82234, Germany
e-mail: koenigr@gfz-potsdam.de
URL: http://www.gfz-potsdam.de
I. Ciufolini
Dip. Ingegneria dell’Innovazione, Università del Salento, Lecce, Italy
I. Ciufolini
Centro Fermi, Roma, Italy

454 R. König and I. Ciufolini
ing the Earth there is a drift of the node and perigee of its orbit due to the rotation
of the Earth, this is the so-called Lense-Thirring (LT) effect. Frame-dragging plays
a key role in some high-energy astrophysics phenomena, such as the mechanism
of emission of jets from active galactic nuclei and quasars [3]. Frame-dragging of
the particles of the accretion disk around a central rotating black hole can explain
the constant orientation of these jets for periods of time that can reach millions of
years. Frame-dragging enters the dynamics of the coalescence of two black holes
producing gravitation waves as observed by LIGO in 2015 [4]. The idea to measure
frame-dragging by means of two laser-ranged satellites, i.e. tracked by Satellite Laser
Ranging (SLR), with supplementary inclinations was first proposed by [5]. The first
rough observation [6] was based on the EGM96 gravity field model [7] and SLR
tracking to LAGEOS-1 [8] and LAGEOS-2 [9]. Later on [10] presented a break-
ing result where again SLR data to the satellites LAGEOS-1 and LAGEOS-2 were
combined but the analysis was based on the by then newly available gravity field
model EIGEN-GRACE02S [11] from the GRACE space mission [12]. Reference
[10] estimated an accuracy of the recovery of the node drift due to frame-dragging
of about 10%. This accuracy estimate was deduced from a thorough assessment of
all possible modelling errors in Precise Orbit Determination (POD).
These results are based on the orbit determination software GEODYN [13] of
the National Aeronautics and Space Administration NASA of the United States
of America (USA). In order to rule out programming errors, the experiment was
repeated by independent orbit determination software packages, one being EPOS-OC
[14] by the German Research Centre for Geosciences GFZ, Potsdam, Germany, and
one being UTOPIA [15] by the Center for Space Research (CSR) at the University
of Texas at Austin, USA. GFZ used some recent GRACE gravity field models to
analyze LAGEOS-1 and LAGEOS-2 data over the period 2000–2011 with EPOS-
OC [16]. In summary the LT precession is recovered with a mean deviation from the
GR prediction of about 8%. CSR tested a multitude of gravity field models using
UTOPIA and SLR tracking to LAGEOS-1 and -2 over the years 1992–2006 [17].
They conclude that based on GRACE gravity field models the LT effect can be
confirmed at the 8–12% level. So with both software packages the 10% accuracy
claim by [10] can be endorsed.
In the following we describe the satellite missions that are involved in this type
of analysis, on the one hand the SLR tracked satellites used to measure the LT
node drift, on the other hand the satellite missions dedicated to measure the Earth’s
gravity field which is the pre-requisite for POD of the previous. We then have a
glance at the major error source of the LT measurement, the C(2,0) term of the
gravity field models. In particular we look into the effect of its variation in time
and the impact thereof on the LT measurement by analyzing an extended period of
LAGEOS observations with EPOS-OC. Finally we adopt some gravity field models
that are generated either with or without LAGEOS data in order to rule out possible
influences on the LT measurement by the fact LAGEOS being part of the gravity
field model development.
Measurement of Frame Dragging with Geodetic Satellites … 455
2 Satellite Missions
Geodetic satellites are sphere satellites with low area-to-mass ratio that are tracked
by SLR. The most important one being LAGEOS-1 (see Fig. 1), launched in 1976
for geodetic and geophysical applications and his twin, LAGEOS-2, launched in
1992 into a different inclination. In 2012 the LAser RElativity Satellite (LARES)
was launched (see Fig. 2). The LARES mission [18] is particularly designed to mea-
sure frame dragging with the goal to achieve the 1% accuracy in combination with
the LAGEOS satellites. The orbit characteristics of these missions are compiled in
Table 1. Table 1 also shows the area-to-mass ratio of the satellites. The smaller this
number the smaller the disturbing accelerations by drag of the upper atmosphere,
solar radiation pressure and other non-conservative forces. Indeed LARES has the
smallest area-to-mass ratio of all artificial satellites ever send into orbit. In the fol-
lowing we are not going to use LARES data as the mission duration is yet too short
at the time of conducting this experiment. Later on a very first result with LARES is
published by [19] indicating an accuracy of 5%.
Gravity field missions seek for exploring the gravity field of the Earth and its
variations in time. The orbit is lower than that of the geodetic satellites, the lower
the orbit the more sensitive to the gravity signal. However the lower the orbit, the
higher the nuisance perturbations inserted by the non-conservative forces. The first
dedicated mission that immediately led to a quantum leap in the accuracy of the
derived gravity field models is CHAMP (see Fig. 3) in service from 2000 to 2005.
CHAMP was equipped with an on-board two-frequency GPS receiver and a Laser
Retro-Reflector (LRR) for POD, a three-axes accelerometer to measure the nuisance
Fig. 1 The LAGEOS

satellite (courtesy NASA)
Fig. 2 The LARES satellite

(courtesy LARES team)
Table 1 Orbit characteristics of geodetic satellites

Satellite Altitude (km) Eccentricity Inclination (deg) Area-to-Mass
Ratio (m2 /kg)
LAGEOS-1 5,900 0.004 109.8 0.000695
LAGEOS-2 5,800 0.014 52.6 0.000697
LARES 1,440 0.001 69.5 0.000269
forces and a star camera to measure the attitude of the satellite for precise reduction
of the antenna offsets. Already in 2002 the two GRACE twin satellites (see Fig. 4)
were launched. In addition to the equipment as that on CHAMP, the distance between
the two satellites was measured with μm precision by an inter-satellite microwave
link. This allowed another leap in accuracy, but more important, the recovery of
the variations in time of the gravity field which in turn had dramatic impacts on
science in the fields of hydrology and cryology. GRACE was designed for a five
years lifetime but stayed in duty for impressive 15 years. GOCE (see Fig. 5) was
launched in 2010 into the very low altitude of 220 km with the goal of an increased
resolution of the static field. The satellite had ion thrusters to keep the spacecraft
free of nuisance forces in flight direction. The gravity measurements were done by
a three-axes gradiometer, an ensemble of six accelerometers working in differential
mode. The gravity campaign took some 6 months only, but it delivered a geoid with
1 cm accuracy at 100 km resolution. The orbit characteristics of these missions are
compiled in Table 2.
Fig. 3 CHAMP (courtesy

Airbus)
Fig. 4 GRACE (courtesy

NASA)
Fig. 5 GOCE (courtesy

ESA)
Table 2 Orbit characteristics of gravity field missions

Satellite Altitude (km) Eccentricity Inclination (deg)
CHAMP 470 0.004 87.3
GRACE 490 0.002 89.0
GOCE 290 0.001 96.7
As said, the gravity field missions are necessary to establish the gravity field
models that are used in POD of the geodetic satellites for the recovery of the node
observations. CHAMP and GRACE are the important missions to establish the long
to medium wavelength part or the low to medium degree harmonic coefficients of
the gravity field model. GOCE is particularly meant to improve the short wavelength
part or the higher degree harmonic coefficients. For POD of the LAGEOS satellites
just the coefficients up to degree about 20 play a role. From linear theory just the
even degree zonal coefficients influence the node drift, that’s why their errors disturb
the LT node observation. The by far largest contribution is due to C(2,0), the next
one, due to errors in C(4,0), is already smaller by about four orders of magnitude.
3 Measuring Frame Dragging
The idea by [5] to measure the frame dragging effect on the LAGEOS satellites
by observing the node drift with the help of SLR tracking is displayed in Fig. 6.
It shows the two sources acting on the nodes of two LAGEOS-type satellites with
complementary inclinations. The LARES 2 satellite, approved by the Italian Space
Agency (ASI), is under construction for a launch in 2019–2020. It will realize the
idea of two laser-ranged satellites with supplementary inclinations [20].
The major drift of the node of an orbit is induced by the C(2,0) term of the spherical
harmonic expansion of the gravity field. The C(2,0) term is also called J2 term (with
C(2,0) = −J 2) or the quadrupole moment of Earth’s inertia. The C(2,0) node drift
can be written in first order approximation as:
3nC2,0 ae2
˙ C2,0 =
cos(i) (1)
2a 2 (1 − e2 )2
where n is the orbital period, ae is the Earth’s radius, a is the semi-major axis of the
orbit, e its eccentricity, and i its inclination. The LT node drift reads:
2J
˙ LT =
(2)
a 3 (1 − e2 )3/2
where J is the angular momentum of the rotating body (the Earth). The sign of the
C(2,0) drift depends on the inclination of the orbit. So in Fig. 6 the drifts of the nodes
Fig. 6 The nodal drifts due

to C(2, 0) and LT for two
complementary orbits
(adapted from [5])
Table 3 Nodal drifts of the geodetic satellites

Satellite Lense-Thirring (mas/a) C(2,0) (mas/a)
LAGEOS-1 30.7 450,000,000
LAGEOS-2 31.5 −830,000,000
LARES 118.4 −2,240,000,000
of the two complementary orbits point in opposite direction. The LT effect however
does not depend on inclination, so it points for both orbits in the same direction. Once
the node observations of both orbits are combined, the nodal drifts due to C(2,0) will
cancel but the drifts due to LT will add. By this not only the comparatively large
numbers of the C(2,0) drift vanish but also the errors of the C(2,0) term are ruled out.
The drifts induced due to LT and C(2,0) are given for the geodetic satellites in Table 3.
The error of the C(2,0) term of current days gravity field models sizes at about
10−8 , so it is subject to erroneous node drifts at the order of the LT effect. Therefore
measuring the node drift of one satellite alone can not reveal the LT effect. This is
depicted in Figs. 7 and 8 where the node measurements from LAGEOS-1 and -2 are
drawn as points. The analysis, done with EPOS-OC, spans the years 2000–2011 and
is based on the EIGEN-6C gravity field model [21]. Eventually the node observations
are accumulated to yield the node drift drawn as line in each graph. If there was no
error in the C(2,0) term, the line should show the LT drift. However as one can see,
the drifts behave different for the two satellites and are far from the expected LT
signal.
If we now take out the time variable part of C(2,0) in EIGEN-6C, the behaviour
of the node changes completely as can be seen in Fig. 9 for LAGEOS-1 and Fig. 10
for LAGEOS-2. The differences of the drifts in Figs. 7 and 9 and those in Figs. 8 and
10 are purely owned to the time variable part of C(2,0) being modeled or not.
The LT drift can finally be observed when the two node observations of the two
satellites are combined. Figure 11 displays the combined node observations and the
Fig. 7 Node observations and drift from LAGEOS-1, EIGEN-6C, obtained by the use of EPOS-OC
Fig. 8 Node observations and drift from LAGEOS-2, EIGEN-6C, obtained by the use of EPOS-OC
Fig. 9 Node observations and drift from LAGEOS, EIGEN-6C static, obtained by the use of
EPOS-OC
Fig. 10 Node observations and drift from LAGEOS-2, EIGEN-6C static, obtained by the use of
EPOS-OC
Fig. 11 Combined node observations and drift from LAGEOS-1 and LAGEOS-2, EIGEN-6C,
obtained by the use of EPOS-OC
node drift based on EIGEN-6C, Fig. 12 those based on EIGEN-6C static where the
C(2,0) variation in time has been omitted. A regression of the drift lines yields 6.9%
deviation from the predicted combined LT drift of 48.2 mas/a in case of EIGEN-
6C and 3.5% in case of EIGEN-6C static. Therefore it can be concluded that the
errors in C(2,0) can effectively be taken care of if the node observables of the two
LAGEOS satellites are combined. This holds even true for the cases considering the
time variable part of C(2,0) or not. The remaining deviations are well within the error
margin of 10% given by [10].
By measuring the LT node drift this way, the important question arises whether
the usage of SLR observations to the LAGEOS satellites in the generation of the
gravity field model could influence the result. In order to answer this, we repeat the
above analysis with an extended period of LAGEOS-1 and LAGEOS-2 data from
2000 to 2013 and base it on an ensemble of models that either have used LAGEOS
data for their generation or not.
The gravity field models with LAGEOS data are the EIGEN-6C, the EIGEN-
6Sp.34, and the EIGEN-6C3. The “C” in the model name denotes “combined” models
meaning satellite data are combined with surface gravity data. So EIGEN-6C is based
on LAGEOS data plus GRACE data spanning the period January 2003 to June 2009,
and GOCE data from November 2009 to June 2010. Finally the satellite data are com-
bined with DTU2010 global gravity anomalies. The EIGEN-6Sp.34 is the inhouse
naming of the official ESA GOCE gravity field model GO_CONS_GCF_2_DIR_R3
and a static, satellite only solution. Besides on LAGEOS it is based on GRACE data
from January 2003 to June 2009, and on GOCE data from November 2009 to April
2011. The EIGEN-6C3 relies besides on LAGEOS on GRACE data from February
Fig. 12 Combined node observations and drift from LAGEOS-1 and LAGEOS-2, EIGEN-6C
static, obtained by the use of EPOS-OC
Table 4 LT measurements based on gravity field models with LAGEOS

Model LT Measurement (mas/a) Deviation from theory (%)
EIGEN-6C static 49.2 +− 0.2 2.2
EIGEN-6Sp.34 47.3 +− 0.2 1.8
EIGEN-6C3 static 44.8 +− 0.2 7.3
2003 to September 2012, on GOCE data from November 2009 to May 2013, in com-
bination with DTU2010/12 global gravity anomalies and the DTU12 geoid over the
oceans and the EGM2008 geoid over the continents. The LT measurements based
on these models are compiled in Table 4. In mean the deviation from the theoretical
LT drift is about 4%.
Now we repeat this experiment with gravity field models free of LAGEOS data.
These are the EIGEN-GRACE02S, the EIGEN-GRACE03S, and the EIGEN-51C.
The EIGEN-GRACE02S is one of the first GRACE-only models with five months
of data out of the period August 2002 to August 2003. EIGEN-GRACE03S is also
a GRACE-only model with more data out of the time span February 2003 to July
2004. The EIGEN-51C is generated from data of the CHAMP mission and from
GRACE data spanning October 2002 to September 2008 combined with DNSC08
global surface gravity anomalies. The LT measurements based on these models are
compiled in Table 5. In mean the deviation from the theoretical LT drift is about 10%.
In summary, in both trials, measuring the LT effect based on gravity field models
Table 5 LT measurements based on gravity field models without LAGEOS

Model LT Measurement (mas/a) Deviation from theory (%)
EIGEN-GRACE02S static 54.0 +− 0.2 12.1
EIGEN-GRACE03S static 41.2 +− 0.2 14.6
EIGEN-51C 46.3 +− 0.3 4.0
generated with or without LAGEOS data, the accuracy of 10% claimed by [10] is
confirmed.
4 Summary
We repeat the frame dragging measurements as done by [10] with GFZ’s indepen-
dent orbit determination software EPOS-OC applied to a longer analysis period of
LAGEOS-1 and -2 data from 2002 to 2013. Firstly we demonstrate that the LT node
drift analyzing just one satellite is hidden behind the drift induced by errors in the
C(2,0) term of modern gravity field models. Once the node observations of the two
LAGEOS satellites are combined, the LT node drift can clearly be seen. Then we
adopt a suite of gravity field models that either had employed LAGEOS data in their
generation or not. In summary all results indicate that the LT drift can be recovered
with a deviation from the theoretical drift of 10%.
References
1. A. Einstein, Letter to Ernst Mach, 25 June 1913, in Gravitation ed. by C. Misner, K.S. Thorne,
J.A. Wheeler (Freeman, San Francisco, 1973), p. 544
2. J. Lense, H. Thirring, Über den Einfluss der Eigenrotation der Zentralkörper auf die Bewegung
der Planeten und Monde nach der Einsteinschen Gravitationstheorie. Phys. Z. 19, 156 (1918)
3. K.S. Thorne, R.H. Price, D.A. Macdonald, The Membrane Paradigm (Yale University Press,
New Haven, 1986)
4. B.P. Abbott et al., Observation of gravitational waves from a binary black hole merger. Phys.
Rev. Lett. 116, 061102 (2016)
5. I. Ciufolini, Measurement of the Lense-Thirring drag on high-altitude, laser-ranged artificial
satellites. Phys. Rev. Lett. 56, 278–281 (1986)
6. I. Ciufolini, E. Pavlis, F. Chieppa, E. Fernandes-Vieira, J. Perez-Mercader, Test of general
relativity and measurement of the Lense-Thirring effect with two Earth satellites. Science 279,
2100–2103 (1998)
7. F.G. Lemoine, S.C. Kenyon, J.K. Factor, R.G. Trimmer, N.K. Pavlis, D.S. Chinn, C.M.
Cox, S.M. Klosko, S.B. Luthcke, M.H. Torrence, Y.M. Wang, R.G. Williamson, E.C. Pavlis,
R.H. Rapp, and T.R. Olson. The development of the joint NASA GSFC and the National
Imagery and Mapping Agency (NIMA) geopotential model EGM96. NASA Technical Paper
NASA/TP1998206861, Goddard Space Flight Center, Greenbelt, USA (1998)
8. R.L. Spencer, LAGEOS - a geodynamics tool in the making. J. Geol. Educ. 25(2), 38–42 (1977)
9. ILRS. Lageos-1/-2, https://ilrs.cddis.eosdis.nasa.gov/missions/satellite_missions/current_

missions/lag1_general.html. Accessed 12 March 2018
11. C. Reigber, R. Schmidt, F. Flechtner, R. Koenig, U. Meyer, K.-H. Neumayer, P. Schwintzer,
S.Y. Zhu, An Earth gravity field model complete to degree and order 150 from GRACE:
EIGEN-GRACE02S. J. Geodyn. 39(1), 1–10 (2005)
12. B.D. Tapley, S. Bettadpur, M.M. Watkins, Ch. Reigber, The gravity recovery and climate
experiment: mission overview and early results. Geophys. Res. Lett. 31, L09607 (2004)
13. T.V. Martin, W.F. Eddy, D.D. Rowlands, D.E. Pavlis, GEODYN II system description. EG&G
Contractor Report, Lanham, MD (1987)
14. S. Zhu, Ch. Reigber, R. Koenig, Integrated adjustment of CHAMP, GRACE, and GPS data. J.
Geod. 78(1–2), 103–108 (2004)
15. B.E. Schutz, B.D. Tapley, UTOPIA: University of Texas Orbit Processor. Inst. for Advanced
Study in Orbital Mechanics, University of Texas at Austin, IASOM TR 80-1 (1980)
16. R. Koenig, B. Moreno-Monge, G. Michalak, Some aspects and perspectives of measuring
Lense-Thirring with GNSS and geodetic satellites, in Second International LARES Science
Workshop, Accademia dei Lincei, Rome (2012)
17. J.C. Ries, R.J. Eanes, M.M. Watkins, Confirming the frame-dragging effect with satellite laser
ranging, in Proceedings 16th international workshop on laser ranging, http://cddis.gsfc.nasa.
gov/lw16/docs/presentations/sci_3_Ries.pdf. Accessed 30 April 2018
18. I. Ciufolini, A. Paolozzi, E.C. Pavlis, J. Ries, R. Koenig, R. Matzner, G. Sindoni, The LARES
space experiment: LARES orbit, error analysis and satellite structure, in John Archibald
Wheeler and General Relativity, ed. by I. Ciufolini, R. Matzner (Springer, Berlin, 2010), pp.
371–434
19. I. Ciufolini, A. Paolozzi, E.C. Pavlis, R. Koenig, J. Ries, V. Gurzadyan, R. Matzner, R. Penrose,
G. Sindoni, C. Paris, H. Khachatryan, S. Mirzoyan, A test of general relativity using the LARES
and LAGEOS satellites and a GRACE Earth gravity model. Eur. Phys. J. C 76, 120 (2016)
20. I. Ciufolini, A. Paolozzi, E.C. Pavlis, G. Sindoni, R. Koenig, J.C. Ries, R. Matzner,
V. Gurzadyan, R. Penrose, D. Rubincam, C. Paris, A new laser-ranged satellite for General
Relativity and space geodesy: I. An introduction to the LARES2 space experiment. Eur. Phys.
J. Plus 132, 336 (2017)
21. C. Foerste, S.L. Bruinsma, R. Shako, J.C. Marty, F. Flechtner, O. Abrikosov, C. Dahle, J.M.
Lemoine, K.H. Neumayer, R. Biancale, F. Barthelmes, R. Koenig, G. Balmino, EIGEN-6C - A
new combined global gravity field model including GOCE data from the collaboration of GFZ-
Potsdam and GRGS-Toulouse. Geophysical Research Abstracts, vol. 13, EGU2011-3242-2,
EGU General Assembly (2011)
Tests of General Relativity with the
LARES Satellites
Ignazio Ciufolini , Antonio Paolozzi, Erricos C. Pavlis, Richard Matzner,

Rolf König, John Ries, Giampiero Sindoni, Claudio Paris
and Vahe Gurzadyan
Abstract LARES (LAser RElativity Satellite) developed by the Italian Space

Agency (ASI) is a laser-ranged satellite successfully launched in February 2012
by ESA (European Space Agency). A second ASI laser-ranged satellite, LARES 2,
is scheduled for launch by ESA at the end of 2019. Here we describe the main scien-
tific objectives achieved and achievable by LARES and LARES 2, both in General
Relativity and in space geodesy and geodynamics. Among the main tests achieved
by LARES is a 5% test of frame-dragging, a fundamental and intriguing prediction
of General Relativity. The LARES 2 satellite together with the laser-ranged satellite
LAGEOS of NASA, is aimed to provide a 0.2% test of frame-dragging together with
other relevant tests and determinations in fundamental physics, space geodesy and
geodynamics.
I. Ciufolini (B)
Dip. Ingegneria dell’Innovazione, Università del Salento, Lecce, Italy
e-mail: ignazio.ciufolini@gmail.com
I. Ciufolini · C. Paris
Centro Fermi - Museo Storico della Fisica e Centro Studi e Ricerche “Enrico Fermi”,
Rome, Italy
A. Paolozzi · G. Sindoni
Scuola di Ingegneria Aerospaziale, Sapienza Università di Roma, Rome, Italy
E. C. Pavlis
Joint Center for Earth Systems Technology (JCET), University of Maryland,
Baltimore County, MD, USA
R. Matzner
Theory Group, University of Texas at Austin, Austin, TX, USA
R. König
Helmholtz-Zentrum Potsdam Deutsches GeoForschungsZentrum GFZ,
c/o DLR Oberpfaffenhofen, Wessling 82234, Germany
J. Ries
Center for Space Research, University of Texas at Austin, Austin, TX, USA
V. Gurzadyan
Center for Cosmology and Astrophysics, Alikhanian National Laboratory,
Yerevan, Armenia

468 I. Ciufolini et al.
1 Introduction: The LARES and LARES 2 Space Missions
LARES (LAser RElativity Satellite) and LARES 2 are two space missions of the
Italian Space Agency (ASI) aimed at accurate tests of General Relativity (GR) and
fundamental physics and other measurements of space geodesy and geodynamics.
The LARES satellite was launched on 13 February 2012 by the VEGA launcher
of the European Space Agency (ESA) [1, 2]. LARES is tracked by the laser
ranging stations of the International Laser Ranging Service (ILRS) [3]. It is a spherical
satellite, covered with 92 retro-reflectors. The LARES semimajor axis is ∼ = 7820 km,
its orbital eccentricity ∼
= 0.0008, and its orbital inclination ∼
= 69.5◦ . LARES, with
LAGEOS (LAser GEOdynamics Satellite) [4], a NASA satellite launched in 1976,
and LAGEOS 2, of ASI and NASA, launched in 1992, and with the Earth’s gravi-
tational field models by the GRACE mission (2002), of NASA and DLR (German
Aerospace Center) [5], has allowed a test of an intriguing prediction of General
Relativity, known as frame-dragging (see Sect. 2), with accuracy of about 5%. Such
accuracy can be increased with longer periods of Satellite Laser Ranging (SLR)
observation of LARES, LAGEOS and LAGEOS 2 to eventually reach an accuracy
of approximately 2% in testing frame-dragging. The LARES 2 satellite of ASI, to
be launched at the end of 2019-beginning of 2020, is aimed to provide a test of
frame-dragging with about 0.2% accuracy.
2 Frame-Dragging
One of the intriguing predictions of General Relativity is the “dragging of inertial

frames”, or “frame-dragging” [6, 7]. (This is the name that Einstein gave to this
GR phenomenon in a letter to Ernst Mach in 1913 [8].) Frame-dragging has a num-
ber of fundamental applications in high-energy astrophysics, from the dynamics of
coalescing black holes with emission of gravitational waves as observed by LIGO
[9] since 2015, to the dynamics of matter falling into spinning black holes with the
ejection of huge jets in active galactic nuclei and quasars characterized by a constant
direction for emission time scales which may reach millions of years (Bardeen-
Petterson effect) [6, 10, 11].
In GR test-gyroscopes determine the axes of the local non-rotating inertial frames
where the equivalence principle holds, so that the gravitational field is locally unob-
servable and all the laws of physics are the laws of special relativity. A test-gyroscope
may be realized by a small current of mass in a loop, i.e. using a sufficiently small
spinning top. However, GR predicts that a gyroscope has a behavior different from
that of classical Galilei–Newton mechanics. In classical mechanics, a torque-free
gyroscope always points toward the same distant “fixed” star. On the other hand,
in GR a gyroscope is dragged by mass-energy currents, such as a rotating mass,
and therefore its orientation will change with respect to a distant “fixed” star. If we
rotate with respect to any such test-gyroscope, we would feel centrifugal forces, even
Tests of General Relativity with the LARES Satellites 469
though we may not rotate at all with respect to an asymptotic inertial frame, i.e. with
respect to the distant “fixed” stars.
In GR, a current of mass-energy generates an additional contribution to the grav-
itational field, which in weak-field and slow-motion is called gravitomagnetic field
because of its formal analogy with electrodynamics. The gravitomagnetic field exerts
a torque on a gyroscope in a way formally similar to a magnetic field exerting a torque
on a magnetic needle in electrodynamics. Indeed, frame-dragging of a gyroscope has
a formal analogy with the Larmor effect, i.e. the change of orientation of a magnetic
dipole by a magnetic field generated in electrodynamics by an electric current [6, 7,
12].
The Lense-Thirring effect is the frame-dragging perturbation of the orbital ele-
ments of a satellite in the weak gravitational field of a slowly rotating body, published
in 1918 by Lense and Thirring [13, 14]. The rate of change of the nodal longitude
˙ = 3 2J 2 3/2 , where J is the angular momentum of the
of the satellite is given by a (1−e )
rotating body, is the nodal longitude of the satellite, a its semimajor axis and e its
orbital eccentricity.
2.1 Tests of Frame-Dragging
Frame-dragging was observed in 1997–1998 by using the LAGEOS and LAGEOS

2 laser-ranged satellites [15]. It was then measured with accuracy of approximately
10% [16–22] in 2004–2010, using LAGEOS, LAGEOS 2 and the Earth’s gravity
models obtained by the space geodesy mission GRACE [5]. It was measured with
about 5% accuracy in 2016 using LARES and the two LAGEOS satellites (see
below) [1]. In 2011 the NASA space mission Gravity Probe B, launched in 2004,
also reported a test of frame-dragging with accuracy of about 19%. LARES, using
longer periods of SLR observations, should reach a test of frame-dragging with
approximately 2% accuracy. LARES 2 of ASI, to be launched at the end of 2019-
beginning of 2020, together with LAGEOS, is aimed to provide in a few years a test
of frame-dragging with accuracy of a few percent.
3 The LARES Satellite
LARES is a passive, spherical laser-ranged satellite (see Fig. 1). It was designed
to approach as closely as possible an ideal test particle [2] and such objective was
achieved by minimizing its surface-to-mass ratio; reducing the number of its parts;
avoiding any protruding component; using a non-magnetic material; and avoiding
the painting of the satellite surface. LARES is made of a non-magnetic tungsten
alloy [2], it has a total mass of 386.8 kg and a diameter of 36.4 cm. Its final mean
density is 15317 kg/m3 which makes LARES the single known orbiting object in the
solar system with the highest mean density and the artificial satellite with the lowest
Fig. 1 The LARES satellite before launch
surface-to-mass ratio. The satellite was built using one single piece of tungsten alloy,
thus reducing thermal contact impedences and the onset of thermal gradients. Indeed,
temperature gradients on the LAGEOS satellites produce thermal thrust and a tiny
but not negligible orbital perturbation.
4 LARES Results
The measurement of frame-dragging with the two LAGEOS satellites was obtained
in 2004–2010 using the two observables provided by the two rates of the nodal
longitude of LAGEOS and LAGEOS 2 to determine two unknowns (see: method in
[23, 24]): the frame-dragging effect and the uncertainty in the Earth’s quadrupole
moment, J2 (the even zonal harmonic of degree two) [25]. The even zonals are the
spherical harmonics of the Earth’s gravitational potential expansion of even degree
and order zero. The largest secular drifts of the nodal longitude of an artificial satellite
are due to the Earth’s even zonal harmonics and by far the largest node shift is due
to the Earth’s quadrupole moment, J2 [25]. To test frame-dragging we either need to
perfectly determine the Earth’s even zonal harmonics or devise a method to eliminate
the propagation of their uncertainties into the satellites’ nodal rates, which led to the
use of the two LAGEOS satellites to eliminate the uncertainties produced by J2 .
The idea of the LARES space experiment is to use its orbital data together with
those of the two LAGEOS satellites to get three observable quantities, i.e. the nodal
rates of the three satellites [23, 26]. Then these three observables can be used to
solve for the three unknowns: frame-dragging and the two uncertainties in the two
even zonal harmonics of lowest degree, i.e. J2 , and J4 (the spherical harmonics of
degree 2 and 4 and order 0). Using this method the two largest sources of uncertainty
in the nodal drifts of the satellites are eliminated, thus providing an accurate test of
frame-dragging within an uncertainty of a few percent [1].
In [1] we reported our orbital analysis of the laser-ranging data of the LARES,
LAGEOS, and LAGEOS 2 satellites from 26 February 2012 to 6 September 2015.
These laser-ranging data were collected from more than 30 ILRS stations all over the
world. We processed approximately one million normal points of LARES, LAGEOS,
and LAGEOS 2. The laser-ranging normal points were processed using NASA’s
orbital analysis and data reduction software GEODYN II, using the 2016 state-
of-the-art Earth gravity field model, GGM05S [27]. GGM05S is an Earth gravity
model based on approximately 10 years of GRACE data. It describes the Earth’s
spherical harmonics up to degree 180. The models in GEODYN II include Earth
tides, solar radiation pressure, Earth albedo, thermal thrust, lunar, solar and planetary
perturbations, and Earth rotation.
The orbital residuals of a satellite provide a measurement of the orbital per-
turbations which are not included (unmodeled) or are modeled with some errors
(mis-modeled) in the data reduction. They represent the obser ved orbital elements
of the satellite minus the computed ones. In our analysis, by far the largest part
of the residuals of the nodal rates of LARES, LAGEOS and LAGEOS 2 are due
to the errors in the Earth’s even zonal harmonics and to the Lense-Thirring effect
which, in our analysis, was not included in the models of GEODYN II. The theoreti-
cal value of the Lense-Thirring effect predicted by General Relativity is about 118.4
milliarcsec/year on the LARES node, and about 30.7 milliarcsec/year and about 31.5
milliarcsec/year, respectively on the LAGEOS and LAGEOS 2 nodes. In our analysis
we used three observables, i.e. the three nodal rates of LAGEOS, LAGEOS 2 and
LARES, and we were then able to eliminate the uncertainties in their nodal rates due
to the errors in the values of even zonal harmonics J2 and J4 of the GGM05S model.
Our method of analysis was able to eliminate the uncertainties in the nodal rates of
the three satellites due to the contribution of the long and medium period tides to the
harmonics J2 and J4 .
We finally fitted our residuals with a secular trend and with the six largest tidal
signals of LAGEOS, LAGEOS 2, and LARES and found the result: μ = (0.994 ±
0.002) ± 0.05, where μ is our experimental value of frame-dragging normalized to
its GR value (i.e. the theoretical prediction of GR is here 1), 0.002 is the formal
1-sigma error and 0.05 is our conservative estimate of the systematic errors due to
the uncertainties in the Earth gravity field model GGM05S and to the other error
sources.
In Fig. 2 we show our experimental value of frame-dragging obtained by fitting
a secular trend together with the six known periodical terms, corresponding to the
largest tidal signals observed on the satellite’s nodes. In the figure the six fitted peri-
Fig. 2 The solid black line represents the experimental value of frame-dragging obtained by fitting
the cumulative combined nodal residuals (the red dots) of LARES, LAGEOS, and LAGEOS 2 with
a linear trend and six periodical terms. These six fitted periodical terms, corresponding to the six
main tidal perturbations observed in the orbital residuals, are removed in the figure both from the
residuals and from the fitting curve where only the secular trend is shown [1]
odical terms are removed both from the residuals and from the fitting curve that shows
only the secular trend. The six tidal signals were identified both by analytical calcu-
lations [1] of the main tidal perturbations of the nodes of the three satellites and by a
Fourier analysis of the observed residuals. Some periodical signals obtained with the
Fourier analysis of the residuals of the nodes correspond to the main nongravitational
perturbations.
The systematic errors affecting our measurement of frame-dragging with LARES,
LAGEOS, and LAGEOS2 are mainly due to the errors in the even zonal harmonics of
GGM05S with degree strictly larger than four. To evaluate this source of systematic
errors, we tripled the published calibrated errors of each even zonal coefficient of
GGM05S, where a calibrated error includes both the statistical and the estimated
systematic errors. This technique to multiply by a factor two or three the calibrated
errors is a standard method in space geodesy to place an upper bound to the real errors
in the Earth’s spherical harmonics. We then propagated these tripled errors into the
nodal rates of LARES, LAGEOS, and LAGEOS 2 and finally found a systematic error
of about 4% in our test of frame-dragging due to the Earth’s even zonal harmonics
[1, 28, 29].
Other smaller sources of systematic errors are due to long and medium period
tidal perturbations and to nongravitational effects either un-modeled or mis-modeled.
Nevertheless, we included in our analysis the main tidal and nongravitational pertur-
bations, such as the direct radiation pressure from the Sun and the Earth radiation,
i.e. the albedo. However, the systematic errors due to the tidal and nongravitational
perturbations, either mis-modeled or un-modeled, are periodical and their effect on

the combined nodal residuals is quite small as shown by the Fourier analysis of the
post-fit orbital residuals shown in Fig. 2. Furthermore, a number of Monte Carlo
simulations and error analyses (see [28, 30] and references in [1]) have shown that
the systematic error in the LARES test of frame-dragging due to tides, nongravi-
tational effects and other error sources is at the level of about 3%. In conclusion
the total root sum squared (RSS) error in our 2016 LARES test of frame-dragging
using the Earth gravity model GGM05S, by including the systematic errors due to
the Earth’s even zonals, tides and nongravitational perturbations, is approximately
5% of frame-dragging.
However, such 5% test of frame-dragging can be improved using the laser-ranging
observations of LARES, LAGEOS, and LAGEOS 2 to decrease the error due the
periodical un-modeled or mis-modeled perturbations. Indeed, the extension of the
observational time of LARES, LAGEOS, and LAGEOS 2 will improve the modeling
of the tidal and nongravitational perturbations and will reduce the total systematic
error. Since different Earth gravity models lead to slightly different results, generating
a set of solutions with different Earth gravity models and different orbital estimators,
such as EPOSOC and UTOPIA, will provide other robust estimates of the systematic
errors.
5 LARES 2
5.1 LARES 2 Objectives
The LARES 2 space experiment is aimed at a measurement of frame-dragging with

an accuracy of a few parts in a thousand. Indeed, using LARES 2 and LAGEOS,
an accuracy of about 0.2% was estimated with detailed error analyses, Monte Carlo
simulations and covariance analyses [31–34]. Therefore LARES 2 will improve by
about an order of magnitude the best measurements of frame-dragging that could
be achieved by the LARES space mission. Furthermore, LARES 2 will provide
important contributions in Space Geodesy and Geodynamics.
5.2 Estimated Error Budget in the LARES 2 Experiment
The main sources of systematic errors in the LARES 2 experiment are: orbital injec-
tion errors coupled with the uncertainties in the even zonal harmonics; non-zonal
Earth harmonics and tides; solar and Earth radiation pressure (albedo), satellites’
eclipses, thermal drag and errors in the determination of the satellites’ orbital ele-
ments [31–34]. Each of these error sources was estimated to produce an error of
approximately 0.1%, therefore the overall RSS error was estimated to be at the level
Fig. 3 The concept of the LARES 2 space experiment to test frame-dragging. Two laser-ranged
satellites, LARES 2 and LAGEOS, with supplementary inclinations [23, 35–39] (see also: [40])
and the same semimajor axes, will provide the most accurate test of frame-dragging. Indeed, the
classical precessions of the nodes of LARES 2 and LAGEOS will be equal and opposite (and thus
their modeling uncertainty can be eliminated in the data analysis) whereas frame-dragging of their
nodes is equal for the two satellites, both in size and sign
of 0.2%. Other smaller perturbing effects will give a negligible contribution to the
RSS error. Such error budget of the LARES 2 experiment was also supported by a
number of Monte Carlo simulations and covariance analyses [32]. Here we briefly
describe the main error sources, for detailed error analyses we refer to [31–34].
• Orbital injection errors The idea of the LARES 2 experiment is to have two
laser-ranged satellites with supplementary inclinations to eliminate the error due
to the uncertainties in the Earth’s even zonal harmonics and thus to accurately test
frame-dragging of their nodes (see Fig. 3) [35, 36]. LARES 2 will have an inclination
supplementary to that of LAGEOS. However, no launch vehicle can achieve an orbit
for LARES 2 with an inclination exactly supplementary to that of LAGEOS and
with exactly the same semimajor axis. Thus, since the cancellation of even zonal
uncertainties will not be perfect, there will be a systematic error in the measurement
of frame-dragging proportional to the uncertainty in the quadrupole moment times
the deviations of inclination and semimajor axis from the LARES 2 ideal orbital
elements. However, thanks to the space missions GRACE and to the recent GRACE
Follow-On, the knowledge of the Earth gravity field has dramatically improved, so
a 3-sigma orbital inclination injection error of the new launch vehicle VEGA C will
induce a fractional error of only about 10−3 in the test of frame-dragging due to
the non-perfect supplementary inclination of LARES 2 with respect to LAGEOS.

A smaller uncertainty will be induced by the injection error in the semimajor axis
of LARES 2. Therefore, the final orbital injection error will be at the level of about
∼
= 0.1%.
In the following we simply list the other main error sources in the LARES 2 space
experiment to test frame-dragging. Detailed analyses of these error sources can be
found in [31–34].
• Non-zonal Earth’s harmonics and tides The non-zonal Earth spherical har-
monics do not produce secular nodal rates. They do produce though periodical nodal
shifts that may induce a bias in the measurement of frame-dragging [25]. However,
the lowest degree spherical harmonics, the ones relevant to the frame-dragging mea-
surement, are very well known thanks to GRACE and GRACE Follow-On. The Earth
tidal models are also well known and they will further improve in the near future. Fur-
thermore, in [1], we used a new method that allows us to dramatically reduce the error
due to periodic effects with known periods on the orbits of LAGEOS and LAGEOS 2
(tides, non-even zonal harmonics and some nongravitational perturbations.) In con-
clusion, the bias in the test of frame-dragging with LARES 2 and LAGEOS due to
non-even zonals and tides will be at a level of about 0.1%.
• Nongravitational perturbations The nongravitational perturbations that will
affect the accuracy of the LARES 2 measurement of frame-dragging are direct solar
radiation pressure, albedo, satellite eclipses, thermal drag and particle drag [23].
Detailed error analyses are presented in [31, 32, 34], here we just report our estimated
systematic error due to each one of these sources: the error due to the modeling of
solar radiation pressure will be less than 0.1%; the error due to the modeling of albedo
radiation pressure will be about 0.1%; the error due to thermal drag (thrust due to the
anisotropic temperature distribution on the satellite) will be about 0.1%; the modeling
error of satellite eclipses by the Earth will be about 0.1%; the error due to the nodal
shift of LAGEOS and LARES 2 by particle drag will be negligible. In addition we
have considered the measurement errors of the orbital parameters. However, Satellite
Laser Ranging provides the position of the LAGEOS satellites with a precision of
the normal points of less than a millimeter. Such a precision is enough to accurately
measure the shift of the node of LAGEOS due to frame-dragging (almost 2 meters
per year.) Nevertheless, the other orbital elements of LAGEOS and LARES 2 must
also be measured with high accuracy. The semimajor axis and the eccentricity of the
LAGEOS satellites are measured with enough accuracy. However the measurement
error in the inclination of LAGEOS and LARES 2 may introduce an error in the
measurement of frame-dragging (this error in the test of the Lense-Thirring effect
with LAGEOS and LARES 2 will be due to the uncertainty in the measurement
of the inclination and not to the uncertainty in the prediction of its behavior.) The
uncertainty in the determination of the inclination of LAGEOS is mainly due to
errors in correcting for atmospheric refraction at low elevations. However, the effect
of the errors due to refraction corrections in the measurement of the inclination drops
significantly at elevations above 20◦ and most of the laser-ranging observations are at
elevations above 20◦ . Thus, by keeping in our data analysis only those observations
corresponding to elevations above 20◦ , we will keep the error in the measurement of
the inclination to a very small level that, when propagated into the nodes of LARES
2 and LAGEOS, will introduce an error of no more than 0.1% in our test of frame-
dragging.
5.3 Error Budget of the LARES 2 Experiment
In conclusion the error sources in the LARES 2 experiment to test frame-dragging

are: injection error and even zonal harmonics ∼
=0.1%, non-zonal harmonics and tides
∼
=0.1%, albedo ∼ =0.1%, satellite eclipses ∼
=0.1%, thermal drag and satellites eclipses
∼
=0.1%, measurement error of the LARES 2 and LAGEOS orbital parameters ∼ =0.1%
of frame-dragging. Thus, by not including here the errors, listed above, smaller than
0.1% (since they would only affect the next decimal digit of the total RSS error) we
estimated the total RSS error in the LARES 2 test of frame-dragging to be at the level
of ∼
= 0.2% of frame-dragging. This error budget of the LARES 2 space experiment
was confirmed by a number of Monte Carlo simulations and covariance analyses
[32].
6 LARES 2 Orbital Parameters and Satellite

Characteristics
The mass and radius of LARES 2 are chosen to minimize its cross-sectional-to-
mass ratio and thus to minimize its nongravitational perturbations. Its mass will be
about 300 kg and its radius will be about 21.2 cm. LARES 2 must have the same
semimajor axis of LAGEOS but its orbital inclination must be supplementary to that
of LAGEOS. The eccentricity of LARES 2 must be zero in order to minimize the
nongravitational perturbations of its node. The maximum deviations of the orbital
parameters of LARES 2 from these optimal values are determined by the requirement
of a 0.1% error (due to the non-perfect elimination of the even zonal harmonics
uncertainties.) In summary the LARES 2 orbital parameters must be:
• Semimajor axis 12270 km ± 20 km.
• Inclination 70.16◦ ± 0.15◦ (supplementary to that of LAGEOS).
• Orbital eccentricity: between 0 and 0.0025.
7 LARES 2 and Space Geodesy
Satellite Laser Ranging (SLR) provides the most accurate definition of the origin of
the International Terrestrial Reference Frame (ITRF) and is comparable with VLBI
in the definition of its scale. The orbits of the Global Navigation Satellite System
(GNSS) satellites use reference frames based on the ITRF frame and the ITRF will
realize the United Nations (UN), “Global Geodetic Reference Frame (GGRF)”, the
standard for all applications. This is accomplished with the LAGEOS and LAGEOS
2 SLR satellites, designed for cm level accuracy. (Nevertheless today objectives are
1 mm accuracy and 0.1 mm/yr stability.) LARES has provided the third SLR target,
with improved specifications, and has enhanced the previous status. Nevertheless,
even with LARES we do not yet have the ideal “SLR Constellation”, in comparison
with the GNSS one, with a number of LARES-type spacecraft. Indeed, a number of
LARES-like satellites, together with LAGEOS, LAGEOS 2 and LARES, will provide
a large number of SLR observations with a number of SLR targets above the horizon
of each ground station. The forthcoming launch of LARES 2 will give the fourth
satellite of the SLR Constellation. LARES 2 was designed to reach submillimeter
ranging accuracy. With LARES and LARES 2 added to the SLR constellation we
will for instance improve the mean site position and the Earth Orientation Parameters
(Polar Motion and Length-of-Day). Finally, in addition to the improved determination
of the ITRF, and thus indirectly of the GNSS accuracy, the enhancement of the SLR
Constellation with LARES and LARES 2 will improve the estimation of terrestrial
geophysical parameters, such as the secular evolution of low degree zonal harmonics,
the terrestrial tides and the elastic properties of the Earth.
8 Conclusions
ASI’s LARES satellite, launched in February 2012, has already provided tests of
General Relativity and fundamental physics, including a 5% test of frame-dragging, a
fundamental and intriguing prediction of General Relativity. Using a longer period of
SLR observation of LARES, LAGEOS and LAGEOS 2, it will be possible to get a test
of frame-dragging with approximately 2% accuracy with relevant implications for
fundamental physics too. ASI’s LARES 2 satellite is scheduled for launch at the end
of 2019-beginning of 2020. Using LARES 2 and LAGEOS, it will be possible to reach
a 0.2% test of frame-dragging together with other relevant tests and determinations
in fundamental physics, space geodesy and geodynamics.
References
1. I. Ciufolini, A. Paolozzi, E. C. Pavlis, R. Koenig, J. Ries, V. Gurzadyan, R. Matzner, R. Penrose,

G. Sindoni, C. Paris, H. Khachatryan, S. Mirzoyan, A test of general relativity using the LARES
and LAGEOS satellites and a GRACE Earth gravity model. Measurement of Earth’s dragging
of inertial frames. Eur. Phys. J. 76, 120 (2016)
2. A. Paolozzi, I. Ciufolini, LARES successfully launched in orbit: satellite and mission descrip-
tion. Acta Astronautica 91, 313–321 (2013)
3. M.R. Pearlman, J.J. Degnan, J.M. Bosworth, The international laser ranging service. Adv.
Space Res. 30, 135143 (2002)
4. S.C. Cohen, R.W. King, R. Kolenkiewicz, R.D. Rosen, B.E. Schutz (Eds.), LAGEOS scientific
results. J. Geophys. Res. 90(B11), 9215–9438 (1985)
5. B.D. Tapley, S. Bettadpur, M. Watkins, C. Reigber, The gravity recovery and climate experi-
ment: mission overview and early results. Geophys. Res. Lett. 31(9) (2004)
6. K.S. Thorne, R.H. Price, D.A. Macdonald, The Membrane Paradigm (Yale University Press,
NewHaven, 1986)
7. I. Ciufolini, Dragging of inertial frames. Nature 449, 41 (2007)
8. A. Einstein, Gravitation, in Letter to Ernst Mach, 25 June 1913, ed. by C. Misner, K.S. Thorne,
J.A. Wheeler (Freeman, San Francisco, 1973), p. 544
9. B.P. Abbott et al. Binary Black Hole Mergers in the first Advanced LIGO Observing Run.
Phys. Rev. X 6, 041015 (2016)
10. J.M. Bardeen, J.A. Petterson, The Lense-Thirring effect and accretion disks around Kerr black
holes. ApJ 195, L65 (1975)
11. C. Nixon, A. King, Do jets precess... or even move at all? ApL 765, L7 (2013)
12. I. Ciufolini, J.A. Wheeler, Gravitation and Inertia (Princeton University Press, Princeton, 1995)
13. J. Lense, H. Thirring, Über den Einfluss der Eigenrotation der Zentralkörper auf die Bewegung
der Planeten und Monde nach der Einsteinschen Gravitationstheorie. Phys. Z. 19, 156–163
(1918)
14. J. Lense, H. Thirring, English translation of [13] by B. Mashhoon, F.W. Hehl, D.S. Theiss.
Gen. Relativ. Gravit. 16, 711 (1984)
15. I. Ciufolini et al., Test of general relativity and measurement of the Lense-Thirring effect with
two Earth satellites. Science 279(5359), 2100–2103 (1998)
17. I. Ciufolini, E.C. Pavlis, R. Peron, Determination of frame-dragging using Earth gravity models
from CHAMP and GRACE. New Astron. 11, 527–550 (2006)
18. I. Ciufolini, E.C. Pavlis, J. Ries, R. Koenig, G. Sindoni, A. Paolozzi, H. Neumayer, Gravit-
omagnetism and its measurement with laser ranging to the LAGEOS satellites and GRACE
Earth gravity models, in General Relativity and John Archibald Wheeler, vol. 367 (Springer
GmbH, Berlino DEU, 2010), pp. 371434
19. J.C. Ries, R.J. Eanes, M.M. Watkins, Confirming the framedragging effect with satellite laser
ranging, in 16th international workshop on laser ranging, Poznan, Poland, 1317 October 2008
(2008)
20. J. Ries, Relativity in satellite laser ranging, in IAU Symposium 261. Relativity in fundamental
astronomy: dynamics, reference frames, and data analysis (Virginia Beach, VA, USA, 27
April–1 May 2009)
21. R. König, B. Moreno-Monge, G. Michalak, Some aspects and perspectives of measuring Lense-
Thirring with GNSS and geodetic satellites, in Second International LARES Science Workshop,
Accademia dei Lincei, Rome (2012)
22. R. König, I. Ciufolini, Measurement of frame-dragging with geodetic satellites based on gravity
field models from CHAMP, GRACE and beyond, in Current Volume
23. I. Ciufolini, A comprehensive introduction to the LAGEOS gravimetric experiment. Int. J.
Mod. Phys. A 4, 3083–3145 (1989)
24. I. Ciufolini, On a new method to measure the gravitomagnetic field using two orbiting satellites.
Nuovo Cimento A 109, 1709–1720 (1996)
25. W.M. Kaula, Theory of Satellite Geodesy (Blaisdell, Waltham, 1966)
26. I. Ciufolini, On the orbit of the LARES satellite (2006). arXiv:gr-qc/0609081v1
27. J. Ries, S. Bettadpur, R. Eanes, Z. Kang, U. Ko, C. McCullough, P. Nagel, N. Pie, S. Poole,
T. Richter, H. Save, B. Tapley, Development and evaluation of the global gravity model GGM05.
CSR-16-02, Center for Space Research. (The University of Texas at Austin, 2016)
28. I. Ciufolini, A. Paolozzi, E. Pavlis, J. Ries, R. Koenig, R. Matzner, G. Sindoni, The LARES
space experiment: LARES orbit, error analysis and satellite structure, in General Relativity and
John Archibald Wheeler, vol. 367 (Springer-Verlag GmbH, Berlino - DEU, 2010), pp. 467–492
29. I. Ciufolini et al., Reply to “A comment on “A test of general relativity using the LARES and
LAGEOS satellites and a GRACE Earth gravity model, ed. by I. Ciufolini et al.,” Iorio”. EPJ
C 78, 880 (2018)
30. I. Ciufolini, B. Moreno Monge, A. Paolozzi, R. König, G. Sindoni, G. Michalak, Monte Carlo
Simulations of the LARES space experiment to test General Relativity and fundamental physics.
Class. Quantum Gravity 30, 235009 (2013)
31. I. Ciufolini, A. Paolozzi, E.C. Pavlis, G. Sindoni, R. Koenig, J.C. Ries, R. Matzner,
V. Gurzadyan, R. Penrose, D. Rubincam, C. Paris, A new laser-ranged satellite for General
Relativity and space geodesy: I. An introduction to the LARES 2 space experiment. Eur. Phys.
J. Plus 132(8), 336 (2017)
32. I. Ciufolini, E.C. Pavlis, G. Sindoni, J.C. Ries, A. Paolozzi, R. Matzner, R. Koenig, C. Paris, A
new laser-ranged satellite for General Relativity and space geodesy: II. Monte Carlo simulations
and covariance analyses of the LARES 2 experiment. Eur. Phys. J. Plus 132(8), 337 (2017)
33. I. Ciufolini, R. Matzner, V.G. Gurzadyan, R. Penrose, A new laser-ranged satellite for General
Relativity and space geodesy: III. De Sitter effect and the LARES 2 space experiment. Eur.
Phys. J. C 77, 819 (2017)
34. I. Ciufolini, R.A. Matzner, J. Feng, D.P. Rubincam, E.C. Pavlis, J.C. Ries, G. Sindoni,
A. Paolozzi, C. Paris, A new laser-ranged satellite for General Relativity and Space Geodesy
IV. Thermal drag and the LARES 2 space experiment. Eur. Phys. J. Plus 133, 333 (2018)
35. I. Ciufolini, Measurement of the Lense-Thirring drag on high-altitude laser-ranged artificial
satellites. Phys. Rev. Lett. 56, 278–281 (1986)
36. I. Ciufolini, Theory and Experiments in General Relativity and other Metric Theories. Ph.D.
Dissertation, Univ. of Texas, Austin, Pub. Ann Arbor, Michigan, 1984
37. B. Tapley, J.C. Ries, R.J. Eanes, M.M. Watkins, NASA-ASI Study on LAGEOS III. Center for
Space Research Report CSR-89-03, The University of Texas at Austin (1989)
38. I. Ciufolini et al., ASI-NASA Study on LAGEOS III (CNR, Rome, Italy, 1989)
39. J.C. Ries. Simulation of an experiment to measure the Lense-Thirring precession using a
second LAGEOS satellite. Ph.D. Dissertation, Center for Space Research Report CSR-89-05,
The University of Texas at Austin (1989)
40. G.E. Peterson. Estimation of the Lense-Thirring precession using laser-ranged satellites. Ph.D.
Dissertation, The University of Texas at Austin (1997)

Puetz - Relativistic Geodesic

Uploaded by

Copyright:

Available Formats

Puetz - Relativistic Geodesic

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Puetz - Relativistic Geodesic

Uploaded by

Copyright:

Available Formats

Fundamental Theories of Physics 196

More information about this series at http://www.springer.com/series/6001

ISSN 0168-1222 ISSN 2365-6425 (electronic)

Library of Congress Control Number: 2018967737

© Springer Nature Switzerland AG 2019

particular, when it comes to implementation and application of relativistic concepts

Bremen, Germany Dirk Puetzfeld2

Time and Frequency Metrology in the Context

Can Spacetime Curvature be Used in Future Navigation Systems? . . . . 379

• A. Bauch Time and frequency metrology in the context of relativistic geodesy

Andreas Bauch Physikalisch-Technische Bundesanstalt, Braunschweig, Germany

Rolf König Helmholtz-Zentrum Potsdam Deutsches GeoForschungsZentrum

Giampiero Sindoni Scuola di Ingegneria Aerospaziale, Sapienza Università di

© Springer Nature Switzerland AG 2019 1

compared with sufficient accuracy in order to determine the height-difference to the

In the context of relativistic geodesy, and for economy of writing, it is admissible

2 Characterization of Time and Frequency Signals

A frequency standard outputs a (nearly) sinusoidal signal voltage described by

V (t) = [V0 + e(t)] × sin{2π ν0 t + φ(t)}, (1)

10 GHz transition is detected with a 1 Hz line width, whereas, to give orders of

3.1 Commercial Clocks

3.1.1 Rubidium Gas Cell and Miniaturized Frequency Standards

3.1.2 Commercial Caesium Clocks

3.1.3 Hydrogen Masers

in Fig. 1, both types of standards share a remarkably low frequency instability at

3.2 Cold-Atom Fountain Clocks

3.3 Optical Frequency Standards

4 International Time Scales and Their Local Realizations

International Atomic Time (Temps Atomique International, TAI) and Coordinated

Dissemination of UTC by the BIPM happens in the form of time series [U T C −

Fig. 6 Comparison of UTC

5 Satellite-Based Time and Frequency Transfer

5.1 GNSS-Based Time and Frequency Transfer

5 Source of data: BIPM ftp server at ftp://ftp2.bipm.org/pub/tai/timelinks/lkc/, data in monthly files,

Read Me file provided.

5.2 Two-Way Satellite Time and Frequency Transfer

6 Look Ahead and Conclusions

1. A. Bjerhammar, On relativistic geodesy. Bull. Geod. 59, 207 (1985)

Pacome Delva, Heiner Denker and Guillaume Lion

© Springer Nature Switzerland AG 2019 25

developed thanks to commercial caesium clocks [8] as well as laboratory caesium

2 The Relativistic Framework

2.1 Observers and the Spacetime Manifold

The theoretical background of chronometric geodesy is general relativity (GR). In

ds 2 ≡ dx · dx = gαβ dx α dx β (1)

between two neighboring events is invariant under coordinate transformation. Trajec-

where c is a constant velocity characterizing the spacetime, which can be identified

Let us integrate this relation along the worldline C : x α = f α (λ) parameterized by

The parameter λ is usually chosen as the coordinate x 0 = ct in the context of rel-

2.2 Simultaneity and Synchronization

Let us define an observer O with trajectory C parameterized by its proper time τ ,

Fig. 2 Illustration of the

Fig. 4 Slow clock transport

This convention is usually called Einstein synchronization. It is a geometrical con-

which best coincides with the (mean) surface of the oceans.

2.3 Relativistic Reference Systems

4 All IAU Resolutions can be found at http://www.iau.org/administration/resolutions/

2.4 Chronometric Geodesy

PFS. Within this project, a proof-of-principle experiment of chronometric geodesy

ds 2 ≡ dx · dx = gαβ dx α dx β (1)

where g = | g |, ν A is the photon frequency at emission (strong gravitational poten-

where N AB = ( x B (t B ) − x A (t A ))/R AB , G is the gravitational constant, and M E is