Thermodyn

ABSTRACT
THERMODYNAMICS
Miroslav Holeček
PLZEŇ - NYMBURK 2006

This publication is supported by the Czech Ministry of Educa-
tion, Youth and Sports, the project MSM 4977751303,
and by the Grant Agency of the Czech Republic, the projects
401/02/0579 and 106/05/0219.
c Miroslav Holeček, 2006

ISBN 80-239-6075-X
TO MY PARENTS
Preface
I still remember my first reading of a thermodynamic textbook

(written by my teacher, J. Kvasnica, [16]). I cannot forget first
impressions, especially a big surprise that thermodynamics is
not only about heat engines. An amazing generality of the
subject did fascinate me. On the other hand, the formulation
of thermodynamics liked rather clumsily. Certain definitions
did not seem to be able to avoid some circularity and the main
law of the whole subject, the second law, was introduced by
formulations resembling more some engineering rules than a
fundamental law of nature.
These feelings acted to me as a challenge to search for better
formulation of thermodynamics. Of course, many people had
similar feelings and some of them took up this challenge in full
seriousness. It is especially the school of the so-called rational
thermodynamics, whose main idea is to lay a rational axiomatic
foundation of thermodynamics that allows studying the subject
by a rigorous mathematical language. However, the choice of
its basic axiomatic schemas and the logic of whole construction
does not seem to me to catch precisely the core of thermody-
namics. It looks like that something, what is very important
for classical thermodynamics, is missing in this approach.
ii
I realized it very clearly when reading an interesting work

by E. H. Lieb and J. Yngvason [21]. It is an axiomatic founda-
tion of classical equilibrium thermodynamics but the logic of its
construction is, in my opinion, essentially different from other
attempts. It has influenced considerable my thinking about
foundations of thermodynamics. After several years, I started
to be more and more convinced that thermodynamics cannot
be fully formulated in the spirit of Newtonian physics. Espe-
cially, the idea of an inert spatial background, existing by itself,
seems to contradict the logic of thermodynamics. The Lieb’s
and Yngvason’s approach outlines the way of escape from the
Newton’s heritage.
This book is an attempt to show it. The problem is, how-
ever, that I am still very far to have a comprehensive physical
theory formulated fully in the thermodynamic spirit. There-
fore, this book only outlines important ideas and explains their
meaning in the context of such a possible theory. Since mo-
tivations come especially from thermodynamics, the sketched
scheme of a physical theory not using the idea of spatial back-
ground is called the abstract thermodynamics. The abstract
thermodynamics, however, is not intended to be a concrete
physical theory. It is rather a scheme of a (general) physical
description that is not formed in the Newtonian spirit.
The book could be hardly written without the support and
encouragement of my friends and colleagues, Josef Rosenberg,
Jaroslav Šesták, Josef Voldřich, and without many people who
helped me to enter the field where physics and mathematics
uncover their philosophical foundations, especially Jiřı́ Fiala,
Ivan M. Havel, Zdeněk Neubauer and Petr Vopěnka. I am also
indebted to Jana Čepicková and Michal V. Hanzelı́n for their
help and advice in preparation of the book.
Contents
Preface i
Introduction 1
List of symbols 15
1 Physical systems 21
1.1 Two levels of description . . . . . . . . . . . . . . 22
1.2 Physical systems and subsystems . . . . . . . . . 25
2 States and manipulations 29

2.1 States of physical systems . . . . . . . . . . . . . 31
2.2 Manipulations with physical systems . . . . . . . 34
2.3 Processes . . . . . . . . . . . . . . . . . . . . . . 40
2.4 The surrounding . . . . . . . . . . . . . . . . . . 43
2.5 Global variables . . . . . . . . . . . . . . . . . . . 46
2.6 Temperature and equilibrium . . . . . . . . . . . 49
3 Systems as sets of parts 53

3.1 Structure . . . . . . . . . . . . . . . . . . . . . . 56
3.2 Parts as subsystems and other parts . . . . . . . 62
iv CONTENTS
3.3 Local variables . . . . . . . . . . . . . . . . . . . 67

3.4 Process variables . . . . . . . . . . . . . . . . . . 72
3.5 State space . . . . . . . . . . . . . . . . . . . . . 73
3.6 The continuity rule . . . . . . . . . . . . . . . . . 76
3.7 Averaging and scale . . . . . . . . . . . . . . . . 81
4 Special structures 87
4.1 Simple systems . . . . . . . . . . . . . . . . . . . 88
4.2 Thermomechanical systems . . . . . . . . . . . . 93
4.3 Continuum . . . . . . . . . . . . . . . . . . . . . 100
4.4 State space of a continuum model . . . . . . . . . 106
5 Space 113
5.1 Space as a surrounding . . . . . . . . . . . . . . . 116
5.2 Symmetries of space . . . . . . . . . . . . . . . . 123
5.3 Special symmetries . . . . . . . . . . . . . . . . . 125
5.4 Newton space . . . . . . . . . . . . . . . . . . . . 128
5.5 Complete decomposition of . . . . . . . . . . . 133
5.6 Mach space . . . . . . . . . . . . . . . . . . . . . 135
5.7 Adiabatic space . . . . . . . . . . . . . . . . . . . 139
6 Accessibility 141
6.1 Relative motion . . . . . . . . . . . . . . . . . . . 145
6.2 Definition of ‘accessibility’ . . . . . . . . . . . . . 148
6.3 Time arrow . . . . . . . . . . . . . . . . . . . . . 151
6.4 Adiabatic accessibility . . . . . . . . . . . . . . . 153
6.5 Thermodynamics and entropy . . . . . . . . . . . 155
Bibliography 163
Index 168
Introduction
Isaac Newton laid foundations of modern physics. The main

reason of his triumph is the choice of a surprisingly fruitful
way of description of matter and its interactions. An essential
point of the Newton’s scheme is a strict difference between ob-
jects and empty space between them. In Newton’s viewpoint,
space is something external what remains always similar and
immovable [29]. It is identified with a ‘container’ holding mate-
rial objects and having an independent, absolute existence. It
should be emphasized that the Newton’s approach represents
the crucial change of the viewpoint comparing to the conception
of his predecessor, the famous philosopher and mathematician
R. Descartes [6]. For him, matter was not located in something
extending but formed this extension by itself. This omnipresent
extension is expressed by the concept of “res extensa”. From
the Cartesian point of view, space is thus merely a conceptual
abstraction losing any meaning without bodies.
The Newton’s concept of space as some independent entity
of its own is a little metaphysical because there is no observ-
able evidence preferring the idea of an absolute space filled by
matter more than the idea of the extension formed by matter
2 INTRODUCTION
itself1 . We, however, cannot ignore the extreme usefulness of

Newton’s space in development of physics and remarkable pros-
perity of the Newton’s programme. The reason consists also in
the fact that the hypothesis about existence of an indifferent,
unchangeable background allows formulating physical problems
with extremely simplified interactions.
Namely, when studying an arbitrary physical system (a
molecule, a pendulum on the Earth or a galaxy somewhere
in the Universe) we need to define clearly the influence of sur-
rounding matter. There is a huge number of extremely com-
plicated objects of various natures in the Universe and thus
the possibility of studying some of them needs a very simple
assumption about influence of others. If we had no such an as-
sumption, the physics would not be possible, because we would
never know which objects should be taken into account.
The idea of empty space enables us to simplify the situation.
Namely, we can imagine studied objects as if they were only in
empty space without influence of anything else. Due to the
assumption of existence of such an indifferent ‘container’ this
consideration has a good operational meaning: we may ever
imagine that the other matter is far enough not to influence
the studied system. The system is thus only in space. Let us
realize that such a consideration would not be possible when
accepting the scheme in which there is only extending matter –
then the system cannot be separated from surrounding matter
because there is nothing else than the “surrounding matter”.
1
The origin of the concept of absolute space lies more in theology than in
physical considerations. Newton was influenced by the doctrine of Henry
Moore (H. More, Enchiridium metaphysicum, in Opera omnia, London
(1679), Vol II.) who proposed the conception of space as an actual mani-
festation of God. Since God has to exist everywhere, space must exist even
where there are no bodies [18], p. 205.
INTRODUCTION 3
Modern physics, however, brings many objections against

the conception of absolute space as an inert arena on which
other physical phenomena take place. In the general theory
of relativity, for example, there is no background space (and
time) – matter fields are not in space-time there are rather on
‘each others’2 . The absence of a really empty space (vacuum)
in the quantum-field physics is also interesting from this point
of view. Such examples seem to show us that we are going back
to a modification of the Cartesian description of the world.
Nevertheless, physics is rather Newtonian than Cartesian.
Though there are many suspicions that the idea of an inert spa-
tial background is problematic or even mistaken, we still believe
that it is a good auxiliary hypothesis that works perfectly in
an essential majority of physical situations. Physical problems
in which the Newtonian conception collapses are indeed very
interesting and appealing for theoretical physics but rather ex-
otic and marginal for physics of everyday use. The physics of
macroscopic objects (especially at human spatial and tempo-
ral scales) is formulated in the Newtonian spirit without any
doubts. There are no indicia of disputableness of the Newtonian
approach in this sort of problems. Really?
Namely, the only reason we fully believe to an inert spatial
background in everyday physics is that we are not in a position
of having a principally competitive approach. Any concepts or
thought schemata of physical investigations of macroscopic ob-
jects are thus “grafted” into the Newtonian scheme and hence
we are more and more insisted on its eminent usefulness. Pos-
sible indicia of dubiousness of the schema of description is thus
2
Such considerations appear when thinking about a theory connecting
the general relativity with quantum physics. Namely, such a theory - called
the quantum gravity - could hardly have a fixed spatial background [35].
4 INTRODUCTION
difficult to be noticed. We have to attend to of rather concep-

tual discrepancies than technical problems to find an incorrect-
ness of the Newtonian way of descriptions. A possible case of
such a discrepancy (of the Newtonian conceptual scheme with a
suitable way of description of macroscopic objects) is presented
in thermodynamics.
Thermodynamics is a peculiar physical discipline3 . It came
into being from highly practical problems concerning the effi-
ciency and optimal designing of heat machines used in indus-
try. The crucial work by Sadi Carnot, Réflexions sur la Puis-
sance Motrice de Feu, et sur les Machines Propres a Développer
cette Puissance4 [3], published in 1824 (at Carnot’s own ex-
penses), however revealed some universal features touching an
arbitrarily constructed machine. Carnot’s considerations indi-
cated clearly to a universal law of nature that is now known
as the second law of thermodynamics. The way of Carnot’s
argumentation, however, is very unusual in comparison with
typical schemas of the Newtonian physics. Although his consid-
erations were essentially elucidated in subsequent works (done
by E. Clapeyron, W. Thomson alias Lord Kelvin, R. Clausius
and others – the story is briefly resumed in [15]), an element of
something strange has remained in thermodynamics till now.
First of all, what is the essence of the second law? What
does it exactly claim? Its verbal formulations resemble more
engineering rules about impossibility of constructing machines
realizing certain processes than statements of modern physics.
The way of translation of these statements into precise physical
formulations consists in existence of a full-valued physical quan-
3
Compare language and approach to the subject in [2, 9, 23, 39, 40].
4
Reflections on the Motive Force of Fire and on the Machines Fitted to
Develop that Power [24], see also comments at [15], pp. 67-72.
INTRODUCTION 5
tity, the entropy. It was introduced by R. Clausius (in 1865)

who noticed that Carnot’s argumentation may be expressed
mathematically as I
δQ
= 0, (1)
T
where δQ is the exchanged infinitesimal heat, T the absolute
temperature and the integration is done over one cycle of a
reversible process. It implies that the quantity S, defined by
δQ
dS ≡ , (2)
T
does not change after a reversible return to initial state. Clau-
sius proposed to call S the entropy (from the Greek word τ%oπη,
change, transformation). It means that the entropy is a state
variable. This quantity thus represents similar characterization
of physical systems as, say, momentum or energy, and gives a
linkage to the way of thinking in the Newtonian spirit.
The problem, however, is that the definition of entropy is
related to the concept of heat via (2). But what is ‘heat’ from
the viewpoint of Newtonian physics? It cannot be identified
with a kind of molecular motion. It is rather an “energy trans-
fer to the hidden molecular modes” as observed by H. B. Callen
([2], p. 8). The essence of ‘heat’ is thus an energetic “communi-
cation” between macroscopic and microscopic scales. However,
the concept of ‘macroscopic scale’ is difficult to be defined. Nev-
ertheless, we can avoid the problem by defining the heat via the
familiar thermodynamic relation5 ,
dE = −δW + δQ, (3)

5
See e.g. [23], p. 26.
6 INTRODUCTION
where dE is an infinitesimal change of the energy of the system

and δW the performed work. These quantities have an evident
relation to Newtonian physics – the energy as well as the work
are defined in mechanics. In thermodynamics, however, the
concept of ‘work’ is not so simple because its mechanical defin-
ition as a ‘force times distance’ is insufficient. The question is
which forces and which distances. Namely, if we calculated the
sum of ‘forces times distances’ for all molecules of the system
the work would always equal the change of the whole energy of
the system and no ‘heat’ would appear in (3), see Fig. 1.
Figure 1: What is work? If we calculated a sum of all ‘forces times

distances’ including interatomic forces we would obtain a full energy
exchange of the system with its surrounding. The work, however, is
the ‘force times distance’ concerning only the movement of the piston.
Let us illustrate this on the familiar thermodynamic sys-

tem consisting of a gas in a box that includes a movable piston.
INTRODUCTION 7
The piston is connected through a rope with a weight in grav-

itational field as illustrated in Fig. 1. The work is defined by
the displacement of the piston. However, if this physical sys-
tem is understood as a conglomerate of molecules of the gas
and molecules of the box and piston, then the ‘displacement of
the piston’ is only a special molecular movement (among many
many others). Why this movement has a fundamental mean-
ing to define the decomposition of the energy into work and
heat? And how it could be generally defined? It is difficult to
answer such questions from the point of view of the system it-
self in a different way than speaking about some “macroscopic
modes” of the movement (see e.g. [2], Section 1.2) and their
eminent relevance for thermodynamics. Thereby we again face
the problem with an explanation of the word ‘macroscopic’.
Let us try to look at the system from the point of view
of the other physical system in its surrounding – the weight
that interacts with the system through a connecting rope (see
Fig. 1). When displacing the weight, the piston begins to move.
The weight thus defines what is meant by that special “mode”
of the motion. It is enabled by a special arrangement that
allows manipulating with the system by another system. The
thermodynamic work then may be defined as a form of energy
communication between the system and its surrounding which
arises in special manipulations with the system6 . The problem,
however, is how these special manipulations should be defined.
In the pioneering work by R. Giles [9], that is attempting to
6
S. Kauffman [14], a biologist, (when realizing in the discussion with a
physician P. W. Anderson that the concept of work is far to be clear, p. 96)
specifies work as “constrained release of energy” whereas by ‘constrains’
means external organization of energy transfer by various pistons, wheels,
paddles and so on. These constrains thus realize various manipulations
with the system.
8 INTRODUCTION
formulate an axiomatic foundation of thermodynamics, the au-

thor introduces a class of “mechanical systems” that serve as
suitable surroundings allowing to define the adiabatic process
([9], Chapter 5), i.e., the process when the only energy trans-
fer is realized by the work. Recently, the Giles’ work has been
successfully crowned by proposing a perfect axiomatic system
of equilibrium thermodynamics by E. H. Lieb and J. Yngva-
son [21]. Though the authors are getting near the Giles’ work
at many points, they avoid any attempts to define adiabatic
processes (and work) in this way. Instead, they introduce the
concept of adiabatic accessibility (in the spirit of Planck’s for-
mulation of the second law) by using only a very special “me-
chanical system” in the surrounding of studied thermodynamic
system – a weight in gravitational field7 ([21], p. 17). We will
present it in Chapter 6.
Thermodynamics thus seems as not being able to avoid the
presence of something (a mechanical device, the weight in a
gravitational field, etc.) in the surrounding of the studied sys-
tem. If the system were only alone, the concept of heat, work or
adiabatic accessibility would become superfluous or even non-
sensical. In the Newtonian physics, ‘alone’ means to be only
in space (so that influence of other bodies may be neglected).
The (hypothetical) possibility for any system to be studied as
‘being alone’ is hence the leading idea of the Newton’s pro-
gram: it allows simplifying considerable external interactions
and studying inner dynamics of the system8 .
7
The other place where the concept of work is used, is the passage when
the authors define simple systems. The ‘work’ is represented there only by
a sort of special coordinates without a detail explanation ([21], p. 38).
8
For example, when deriving the Kepler’s laws of motion of planets
around the Sun, we imagine one planet (e.g. the Earth) and the Sun as a
physical system which is alone in empty space.
INTRODUCTION 9
Thus thermodynamics presents a small but remarkable con-

ceptual controversy with the Newtonian scheme because its ba-
sic concepts cannot be principally defined when using only sys-
tems in empty space. It may be a source of many problems
with understanding the entropy. Namely, we are tempted to
translate any concept into the Newtonian scheme of thinking9 .
Hence the entropy is expected to be assigned to the system
‘alone in empty space’ (as any Newtonian quantity) in spite of
its immediate relation to the concept of heat or adiabatic ac-
cessibility (which, however, cannot be meaningfully explained
for systems being alone in empty space).
When studying thermodynamics, one has to notice its strik-
ing difference from other physical disciplines. Its methods, con-
cepts and ways of defining physical problems are strange in
comparison with other branches of classical physics, such as
mechanics or electromagnetic theory. For example, a typical
object of the thermodynamic study is a gas in a box equipped
with a piston (as seen in Fig. 1). But what do we actually
study? Are these the properties of the gas itself or those of the
box, piston and all accompanying machinery (as the rope and
weight in Fig. 1)? Many attempts have been made to “purify”
thermodynamics from the baggage of various pistons, barriers,
weights, etc., and formulate it similarly as, say, the theory of
electromagnetic field10 . Roughly speaking, statements of such
formulations of thermodynamics are only about gas itself. Sur-
prisingly, despite an enormous effort of its representatives to
9
Recall problems with understanding quantum mechanics, where such
a translation is impossible or highly problematic.
10
Such a “pure” thermodynamics is sometime called the rational thermo-
dynamics. Its representatives are B. D. Coleman, W. A. Day, M. E. Gurtin,
W. Noll, D. R. Owen, J. Serrin, M. Šilhavý, C. A. Truesdell, and many
others, see e.g. [5, 10, 31, 32, 40, 41, 42, 43].
10 INTRODUCTION
convince others about the rightness of their formulation of ther-

modynamics (see e.g. [42]), this approach has never forced out
the classical conception (compare the comment in [21], p. 10).
In my opinion, the fact that thermodynamics defies (whatever
it means) to be completely formulated in the Newtonian spirit
is worth noticing. As if its strange formulation and certain
inaptitude (in comparison with elegant formulations of other
disciplines of classical physics) indicated the presence of some
barriers. The barriers that close physics in a hardly visible
prison. The prison that looks like empty space.
That is why we try to concentrate especially on those fea-
tures of thermodynamics that are unusual and strange. This
book is not about thermodynamics, it is only motivated by its
special way of description. Typical features of this description
(especially those that do not appear in other physical theories)
are separated out and used in formation of a very abstract for-
mulation of a general physical description of arbitrary physical
systems. We call it the abstract thermodynamics to empha-
size its similarity with the description introduced by classical
thermodynamics. The main property of this abstract scheme
of description is the absence of any spatial background as well
as a presupposed temporal scheme. Instead of ‘space’ we have
only another physical system, called the surrounding of the sys-
tem, and the temporal scheme of standard physics is replaced
by an abstract concept of change. The change is just a couple
of states, whereas the concept of state has similar meaning as
that in standard physics11 . The crucial (and essentially ther-
11
Notice that in thermodynamics (as well as in quantum mechanics),
the concept of state plays much more important role than in newtonian
mechanics. Surprisingly, the concept of state seems to be rather Cartesian
than Newtonian as noticed, e.g., by L. Kvasz [17].
INTRODUCTION 11
modynamic) point is that any change is understood as a result

of certain manipulation with the system. Any manipulation is
caused by an action in the surrounding of the system12 , which
represents the only causality in our approach. The surround-
ing, however, is only a special kind of the system, hence this
“action” is nothing but a change defined on the surrounding.
All these concepts are introduced and studied in Chapter 2.
In the Newtonian scheme, the idea of physical systems in
empty space allows not only simplifying external interactions,
but also offers a general way of characterization of states of
the system. States may be characterized by current distribu-
tions of physical quantities in space. Since Newton’s space is
a universal fixed background, such a description of the state is
uniquely defined. In our approach, however, no such a universal
background exists. That is why we introduce, in Chapter 3, a
fixed set of some parts of the system. Individual parts may not
be subsystems of the system – their role is only to characterize
states of the system by a suitable mathematical description.
The set of parts is thus an auxiliary tool of description of the
system and may be chosen arbitrarily. Any choice of this set
is called the structure of the system. In Chapter 4, important
structures of equilibrium thermodynamics (simple systems) and
continuum thermomechanics (discrete systems) are introduced.
The concept of the continuum is then studied as a special struc-
ture of description of various macroscopic systems.
What about the Newton’s absolute space? In our approach,
it becomes only a special kind of surrounding, called simply
the Newton space, and therefore its properties may be studied
without a baggage of unclear metaphysical assumptions. It
is done in Chapter 5. Here we obtain an interesting result.
12
Whereas a ‘manipulation’ may also mean to “do nothing”.
12 INTRODUCTION
Namely, special properties of this surrounding (especially its

symmetries) imply that it causes a strange manipulation with
a physical system that is surrounded by it. There is only one
manipulation caused by the Newton space that, however, gives
an infinite number of changes. In other words, any state of the
system may change into an infinite number of different states.
It resembles the time evolution of the system when letting alone
in empty space. We can conclude that the Newtonian ‘time’ is
a result of symmetries of the Newton space.
However, the detail analysis of the problem shows that this
“time interpretation” is only one possibility from many others.
The way in which the Newton space “yields” the (Newtonian)
time, indicates the presence of a more fundamental structure
that may occur in various “modes” of realization. One of them
is very similar to the Newtonian space and time but the ‘time’
has no meaning of an absolute entity flowing independently of
anything else. It is rather defined by a set of clocks that are
inherent part of the structure. Another realization is space
with a weight in a gravitational field that does not include any
‘clocks’ and the concept of time is missing. Maybe, there are
other interesting realizations but we do not search for them.
The general structure yielding all those realizations is called
the Mach space in honor of the famous physicist and philoso-
pher Ernst Mach. Namely, his objections against the Newton’s
absolute space seem to outline a channel towards this structure
(Section 5.6).
In Chapter 6, we outline some features of a physical descrip-
tion that is not defined on a given spatial background. We go
back to the Cartesian conception for which space is only a re-
lation between bodies. This approach, however, brings serious
problems because of its utter relativity. In Newtonian physics,
INTRODUCTION 13
space includes fixed places (spatial points) allowing to define

motion as going from one point of space to another point of
space. In Cartesian physics, on the other hand, everything is
in a relative motion towards anything else and no fixed ”plat-
form” exists. How can ‘motion’ be defined then? The problem
with the definition of motion was one of reasons of the failure
of the Cartesian physics.
That is why we start by analyzing the problem of definition
of motion. Being motivated by the thermodynamic concept of
‘adiabatic accessibility’, we show that the essence of this con-
cept offers an escape from the devastating relativity. Namely,
motion may be defined even if there is no spatial background,
but not at any situation. The idea consists in restricting only to
situations when surrounding bodies behave like spatial places,
i.e., they are fixed and remain in the same state. It is the core
of the concept of accessibility. The accessibility replaces, in
a way, the concept of ‘dynamics’ if no assumption about an
(inert) spatial background is done.
This accessibility is motivated by the immobility and inert-
ness of the Newton space. When replacing the Newton space
by the Mach one, we obtain a more general accessibility called
the weak accessibility. The weak accessibility has an under-
standable meaning in physics. If the realization of the Mach
space with clocks is used, for instance, this kind of accessibility
defines the time arrow (only transitions, when the datum on
clocks increases, are accessible). The realization with a weight
in a gravitational field then defines the adiabatic accessibility
that has a direct relation to the entropy. A relation between
‘entropy’ and ‘time arrow’ then appears in a surprising context.
At this point, we join the many times mentioned work by Lieb
and Yngvason [21], and the book declines to its end (Fig. 2).
14 INTRODUCTION
The book does not develop the abstract thermodynamics

as a comprehensive axiomatic structure. The text describes by
no means a complete physical theory. It rather tries to express
some ideas of a possible physical description that essentially
differs from the standard one. It could be globally charac-
terized as a formulation of physics without a spatiotemporal
“background”. Contrary to the general relativity or quantum
gravity, the motivation for such a description and main ideas
of its construction come from thermodynamics.
Figure 2: The diagram outlining main connections of our approach.

List of symbols
Some mathematical symbols
[a, b] an ordered couple of objects a and b

{x, P (x)} means the set of objects having property P
a∈A means ‘a is an element of the set A’
a 6∈ A means ‘a is not an element of the set A’
A⊂B means ‘the set A is the subset of the set B’
A∩B the intersection of sets A and B
A∪B the union of sets A and B
A\B the complement of the set B in the set A
Pow(X) the set of all subsets of the set X
f : A→B the mapping from the set A to the set B
f −1 : B → A the inverse mapping of the mapping f
f (A) ⊂ B the range of values of the mapping f
N set of natural numbers 1, 2, 3, . . .
R set of real numbers
R+ set of positive real numbers,
R0,+ set of nonnegative real numbers,
hr1 , r2 i the closed interval r1 ≤ x ≤ r2 (r1 , r2 , x ∈ R)
(r1 , r2 ) the open interval r1 < x < r2 (r1 , r2 , x ∈ R)
16 LIST OF SYMBOLS
Ed d-dimensional Euclidean space

x, x0 (∈ Ed ) points of Euclidean space
B, BG , Bq Banach spaces (complete normed vector spaces)
||b|| the norm of a vector from a Banach space
Q, Q−1 , Q> a matrix, its inversion and transposition
A σ-algebra
AX σ-algebra that includes all sets from X
µ,
R µV various measures
X f dµ the integral of the function f over the set X
Symbols of concepts that have no mathematical description
S, S0 , Si (physical) systems
sub
S0 ⊂ S means that S0 is a subsystem of S
(S1 , S2 ) the composition of systems S1 and S2
Ssur the surrounding as a physical system
Sets assigned to physical systems and derived concepts
Ξ(S) , Ξ the set of representative subsystems of S

Σ(S), Σ the state set of the system S
fsub the mapping between Σ(S) and Σ(S0 )
s (∈ Σ) a state (element of the state set)
ssur (∈ Σsur ) a state of the surrounding
M(S), M the set of manipulations with S
α, β, γ (∈ M) manipulations
id the identity manipulation
(α, β) the parallel composition of manipulations
LIST OF SYMBOLS 17
π the process (a subset of Σ)

πτ1 ,τ2 the parametrization of the process π
Π(π, s) the tangent of the process π at the state s
F the mapping defining the surrounding
G the global variable
DG the definition domain of the global variable
Qg the set of global variables
T the global variable ‘generalized temperature’
feq the mapping defining the equilibrium
Symbols describing parts and structures
P(S), P the set of parts

p, p0 (∈ P) parts
Dcomp the set of couples of disjunct parts
Comp the mapping defining the composition
(P, Comp) the structure
(p1 , p2 ) the composition of parts p1 and p2
p0 the total part
sp
p0 ⊂ p means that p0 is a subpart of p
Patom the set of all atomic parts from P
Psub parts corresponding to subsystems
Fs the mapping defining the set Psub
Pch externally defined parts
q̂, V̂ , m̂, b̂ various (local) variables
qs the mapping corresponding to the variable q̂
Dq,s the definition domain of qs
Q the set of local variables
18 LIST OF SYMBOLS
q̂ q̂ 0 absolute continuity of variables

q̂ P the process variable
qs,Π the mapping corresponding to q̂ P
Dq,s,Π the definition domain of qs,Π
Γ(p) the state space of the part p
Γ the state space of atomic structures
Γf ull (p1 , . . . , pn ) the full state space of n parts
A the averaging mapping
Sc the mapping defining the scale
l the scale (a length)
P(l) (= Sc(l)) l-space
Symbols concerning special structures
v(p) the scaling parameter of extensive quantities

Σeq the state space of equilibrium composition
Γ(r) the state space of a scaled copy
rX (∈ Γ(r) ) a point in Γ(r)
σ, ∂σ a material sample and its boundary
Ê energy as the (local) variable
V̂ volume as the (local) variable
b̂q the flux of q̂
δ̂ q the supply of q̂
U the set of elements modeling parts
u (∈ U) an element of this set
Ms the mapping modeling parts as sets
Xs the model of a part at state s
Xs0 the model of the total part at state s
LIST OF SYMBOLS 19
Υ the index set

ξ (∈ Υ) an element of index set (a “label”)
ηs the configuration at state s
C the set of configurations
µV a measure modeling the variable V̂
λqs a measure modeling the variable q̂
Symbols concerning space
E the physical system called ‘space’

Cd the set of (d-dimensional) coordinate systems
a, b (∈ Cd ) coordinate systems
fE,a the mapping defining spatial points
qsa a spatial field
ψ, ψ̃ fields as points of the state space
G a symmetry operation
Ty translation
RQ rotation
Or scaling
EN ewt the Newton space
s0 the state of Newton space
the Newton manipulation
< the set of generalized lengths
` (∈ <) a generalized length
` mappings decomposing
fL the mapping assigned ` to couples of states
EM the Mach space
M` mappings caused by the Mach space
ĉ the clock variable
20 LIST OF SYMBOLS
Symbols concerning accessibility and entropy
N
≺ the relation of accessibility
≺ the relation of weak accessibility
E energy as a global variable
t time as a global variable
S entropy as a global variable
w the “vector” of work variables
T
∼ the relation of thermal equilibrium
Chapter 1
Physical systems
If we study any problem we use a language. The more precisely

and unambiguously the problem should be defined the more ac-
curate language has to be used. The physics is an exact science
and its ambition is to be precise as much as possible. Therefore
the definite language of mathematics is the basement of phys-
ical thinking. Physics, however, is not a part of mathematics.
Any physical concepts have a relation to reality around us and
cannot be defined only within a framework of a mathematical
theory. For example, the electromagnetic field cannot be fully
identified with a set of vector functions on a suitably chosen
geometric space. The field is a physical object that has to be
also defined by a measuring procedure, what reflects its rela-
tion to other entities of reality1 . The fact that studied physical
object cannot be fully identified with its description within a
mathematical structure is expressed here by introducing the
concept of ‘physical system’. It means a part of reality that is
studied from physical point of view in a systematic approach.
1
In quantum mechanics, these questions play the crucial role.
22 CHAPTER 1. PHYSICAL SYSTEMS
This concept has no mathematical definition. To say what we

mean by a concrete physical system is possible only in common
language of everyday experience.
We try to differ strictly between the concept of physical
system, that means an entity of physical reality, and its de-
scription by a suitable mathematical language. Namely, if we
identify this physical concept with its description in mathemat-
ics we may accustom to its mathematical definition so firmly
that we forget the way by which it has been transformed from
the reality to mathematics. Then there is a danger of various
misunderstandings and paradoxes. That is why we use two
levels of the description: a nonformalized language of every-
day experience and a strict mathematical language (in which
physical theories are formulated).
1.1 Two levels of description

There are many various objects and material structures around
us. Some of them are simply visible by naked eyes, others
need more or less sophisticated device to be detectable in a
way. They may be very large (such as the Earth or the whole
galaxy) or extremely tiny (as atoms or elementary particles).
In physics, we look at these objects and structures from special
point of view whose purpose is to understand their physical
properties, find objective laws of their behavior and possibly use
them in some technological applications. This special viewpoint
is called the physical description. In this description, structures
and objects become physical systems.
The main task of any physical description is to find useable
mathematical tools and structures enabling us to speak about
physical systems in a correct mathematical language. There are
1.1. TWO LEVELS OF DESCRIPTION 23
many steps by which we shift from an intuitively defined ma-

terial structure to its mathematical description. Some of them
are evident and clearly explainable but there are also steps
that are very far to be obvious. These steps are very important
since they often hide crucial events in the history of physical
thinking. The assumption of ‘empty space’ in which physical
objects ‘are’ is a very good example (whereas its role in mathe-
matical description of arbitrary systems is colossal). Therefore,
we explain now in which way the mathematical description is
understood in our approach.
We differ between two levels of description. The first level
includes only some primitive concepts as ‘physical system’ or
‘subsystem’ and their simple relations (e.g. ‘composition’ of
physical systems). The physical system, however, cannot be
characterized as a mathematical structure. It may be described,
if necessary, only in common language. Macroscopic systems
of our everyday experience may be described in great details at
this level (a ‘car’ consists of ...), other systems may be described
only very roughly (an ‘atom’ in a glass of water), or not at all
(a ‘quark’). Sometime, a subsystem of the system is clearly
describable (a ‘wheel’ is a subsystem of ‘car’), sometime it is
quite problematic. Nevertheless, the description at this level is
only intuitive and cannot form an essence of a mathematical
description.
The whole mathematics can be formulated in the language
of the set theory. That is why we define the second level of the
description by introducing several sets for each physical system.
It is the set of states (whereas the ‘state’ itself is an undefined,
primitive concept2 ), the set of manipulations (allowing to de-
2
In the set-theoretical language, the set of states thus corresponds to a
set of urelements, i.e. elements that are not sets.
scribe any external influence) and the set of parts, which forms
a basement of the description of any system by physical vari-
ables. The parts and manipulations (defining the concept of
‘change’) generalize the standard spatiotemporal description in
which systems are understood as being in empty space with an
(absolute) flowing time.
These two levels of description cannot be mixed together.
At the second one, all concepts and relations are describable in
the framework of set theory (allowing to define various func-
tions, functional spaces, etc.). The first, however, is only in-
tuitive. If we speak about a physical system S we cannot do
much more than to denote it and speak, if necessary, about
some of its subsystems. If we speak about a property P of a
physical system S we denote it simply as P (S). It means that
this property may be (at least partially and approximately) de-
scribed in the common language of the first level. This property
may be a mathematical object (set, function) what means that
this object is assigned to this system (and roughly described in
common language).
A typical example is the state set, Σ, assigned to a concrete
system S, i.e. Σ(S). We can describe various states of the
system in the common language to demarcate roughly its state
set, but it cannot be defined within this language. On the other
hand, the state set is a primitive concept of our approach and
its (approximate) characterization can be done only in common
language. In this language, the state is more or less identified
with some configuration, i.e. with a concrete arrangement of
the structure3 . This description cannot be strict – we put it
more precisely into a mathematical language (by defining vari-
3
Possibly endowed with vectors determining current velocities of indi-
vidual parts.
1.2. PHYSICAL SYSTEMS AND SUBSYSTEMS 25
ous sets and their relations) but this step is impossible without
a previous pre-understanding done in the common language.
If we study systems being unattainable to human experience
(atoms, elementary particles, black holes, the Universe, etc.)
the concept of state becomes more abstract and its description
in the common language is only very rough and approximate.
Nevertheless, any physical description, whatever abstract,
carries a trace of understanding it in the language of everyday
experience. We speak about an atom as consisting of electrons
occupying various configurations around the nucleus though we
know that it is a very inaccurate picture. The choice of physi-
cal variables is also motivated to be close to our understanding
of the state as a mechanical configuration as much as possible.
That is why we introduce the concept of ‘part’ and describe
systems by specifying them as they are build up from individ-
ual parts. Parts, however, need not be some material “pieces”
– they may be chosen very abstractedly and artificially (e.g.
parts as geometrical points or infinitesimal pieces of matter play
important role in classical physics though they can be hardly
understood as something approachable by our experience).
1.2 Physical systems and subsystems

If we study an object or structure by means of a physical de-
scription we call it the physical system or simply the system.
It means that the material object or structure is studied by a
systematic physical approach. It may be a box with gas, a
car, a culture of living cells, a galaxy or a hurricane but also a
molecule or an emitted radiation interacting with surrounding
matter. There is a vast number of examples. Physical systems
will be denoted by S and derived symbols.
Some parts of the studied object may be studied as phys-

ical systems, too. We say that the physical system S0 is a
sub
subsystem of S and write it by means of a relation ⊂ , i.e.
sub
S0 ⊂ S,
that has following simple properties, namely

sub
S ⊂ S (1.1)
and
sub sub
S ⊂ S0 ∧ S0 ⊂ S ⇒ S = S0 . (1.2)
This relation, however, does not only mean that ‘something’ is
included in ‘something larger’. As well as the concept of physi-
cal system means also a special viewpoint from which material
structures are studied, the fact that one physical system is un-
derstood as a subsystem of the other means that they both may
be studied from the same viewpoint. A specification of several
subsystems of a system (in common language) thus determines
much better the way in which we do understand the system.
Therefore, we introduce the concept of a representative
set of subsystems Ξ(S) that includes the system S and such
its subsystems which may be studied from the same viewpoint
as the system alone. This set is thus only fuzzily defined4 be-
cause usually there is no complete list of all “representative”
subsystems5 . From this point of view, it is describable only in
common language. Nevertheless, we will use this concept also
4
Excepting some special cases, e.g. if the system is formed by only a
few mechanical parts that form a clearly given set subsystems.
5
The ‘representative set’ is thus rather a semiset in the spirit of the
definition of P. Vopěnka [45], see also Section 3.6.
1.2. PHYSICAL SYSTEMS AND SUBSYSTEMS 27
in the abstract level of description where it means just a suit-

able set of subsystems (whose elements more or less coincide
with “representative” subsystems)6 .
Another important concept comes from the possibility of
understanding two systems as one. We introduce the compo-
sition of two systems: the composition of a couple of systems
S1 and S2 is a system S uniquely defined by the systems S1
and S2 . It is denoted as
S = (S1 , S2 )
and we accept that
(S1 , S2 ) = (S2 , S1 ), (S, S) = S. (1.3)
The composition is related to the concept of subsystem by the
following relations:
sub sub
S1 ⊂ (S1 , S2 ), S2 ⊂ (S1 , S2 ) (1.4)
and
sub
S0 ⊂ S ⇒ (S0 , S) = S. (1.5)
The composition of two systems resembles the union of two
sets and the subsystem looks like subset. Nevertheless, physical
systems cannot be understood as sets (what would be their el-
ements?). The conflict is evident when we try to define a coun-
terpart to the set intersection (∩) or complement (\). Namely,
we can have two systems that have a common subsystem. A
counterpart of ‘intersection’ then would be the “largest” com-
mon subsystem. It, however, cannot be defined because we
have no list of all subsystems.
6
Subsystems that are not representative exist, to be sure, but they have
no relation to a chosen form of description of the studied system.
The concept of physical system is the basement of any phys-

ical description. It names the thing which is intended to be de-
scribed. This ‘thing’, however, may by undefined or only very
fuzzily defined. What means, for example, a hurricane or a liv-
ing cell as a physical system? Any definition of such systems
means more or less artificial separation of some part of reality
from everything else. In our approach, such questions play a
very important role. We will study them at various levels of
description. First, we define the ‘surrounding’ as another phys-
ical system that has a deep relation to the studied system. For
example, the surrounding air is the surrounding of the hurri-
cane, other cells and an extracellular space is the surrounding
of the cell. Though there is no clearly defined “border” between
the system and its surrounding an interplay between these two
“parts of reality” plays the crucial role.
Second, we define the concept of ‘parts’ of the system.
These parts are defined “pieces” that model the system in a
way. For example, the living cell may be understood as con-
sisting of several parts – the membrane, intercellular liquids,
cytoskeleton networks, etc. The system usually cannot be iden-
tified with a set of all its parts (the cell is not strictly defined by
its membrane and the above described components) but parts
allow describing the system by suitable mathematical tools.
Especially, by assigning various variables to individual parts.
Chapter 2
States and
manipulations
Any physical description of a real material structure means

a “drastic” simplification of many complications and complex
features of the studied object. For example, a piece of metal
with an extreme rough, grainy polycrystal structure is repre-
sented by a smoothly deformable “continuum”, the Earth is
understood as a “mass point” in the celestial mechanics, and
so on.
Physical simplifications, however, do not originate only in
neglecting details at essentially smaller scales than that of the
chosen description but are also connected with a highly sim-
plified influence of surroundings. A molecule, for example, is
studied as being “alone” in the universe. While the first kind
of simplification is well understood and seriously studied by
physicists who try to estimate errors or corrections caused by
neglecting details, the second is understood differently. We al-
ways describe a pure case without an influence of surrounding
30 CHAPTER 2. STATES AND MANIPULATIONS
and then add (if necessary) some defined forces of interactions

as external conditions. We do it although we know that no such
“pure cases” exist. This thinking has roots in the Newton’s
program of physical description of nature and it is embodied
in such concepts as ‘empty space’ (that is what remains if all
external fields and objects are removed).
Nevertheless, an exception exists. It is the physical disci-
pline called ‘thermodynamics’. In the thermodynamic descrip-
tion, the surrounding is never neglected or identified only with
some empty space1 . For instance, the Kelvin’s formulation of
the second law may be paraphrased as follows [21]:
No process is possible, the sole result of which is that a body is
cooled and work is done.
The meaning of the “sole result” or that “work is done” can
be explained only when there is a communication between the
system and its surrounding. Especially the concept of ‘work’
indicates an evident interconnection of a thermodynamic sys-
tem with its surrounding. The work is important when defining
the heat and, substantially, the concept of crucial importance,
the entropy. It seems that thermodynamics cannot completely
avoid some concepts describing a ‘communication’ of the sys-
tem with its surrounding2 .
In this book, the thermodynamic viewpoint is studied with
a full respect. We supply to the concept of ‘state of the sys-
1
Surprisingly, similar features are also presented in the Copenhagen
formulation of quantum mechanics where the measuring device plays also
the role of an external ‘thing’ that is nothing but a certain kind of a
surrounding.
2
In modern formulations of thermodynamics (e.g. [21]), the ‘work’ is
replaced by the concept of ‘adiabatic accessibility’ that also cannot be
defined without supposing existence of some objects outside the system.
2.1. STATES OF PHYSICAL SYSTEMS 31
tem’ (that is crucial in any physical description) the concept

of ‘manipulation with the system’. The ‘manipulation’ is any
influence coming from the surrounding of the system. The ques-
tions how it can be simplified or whether it may be fully ne-
glected are not self-evident assumptions. They must be seri-
ously studied without a baggage of metaphysical ideas about a
fixed spatiotemporal background. An interplay between ‘states’
and ‘manipulations’ gives a rich structured description of phys-
ical systems that allows us to abandon several prejudices of
newtonian physics.
2.1 States of physical systems

The ‘state’ is probably the most important concept in any phys-
ical description. Starting with description of any physical sys-
tem we speak about its “states”, changes of “states”, suitable
“description of states” and possible experimental “determina-
tion of states”. On the other hand, to say what we exactly mean
by the “state of the system” can bring us to terrible problems.
Remark: The problem of explanation of the concept of state is that
usually the state alone is confused with the description of the state.
However, the first is more philosophical concept which may be un-
derstood as a fact (attitude, viewpoint) that at any time moment the
system has its own identity connected to a set of properties, qualities,
specific interactions with surrounding, etc. This identity is called the
state. At this level, the question if two systems can be at the same
state or if one system can reach the same state at a latter moment is
problematic and belongs to philosophical considerations (remember
Heraclitus: “You cannot step twice into the same river”). There is
no physical definition of “identity” or “state alone”. For example,
we could say that two systems are in the same state if any mea-
surement done on both systems gives the same results. But what
means “any measurement”? What is, for example, a complete set
of measurements saying us whether two glasses of water are in the
same state? In description of simple, microscopic systems such a set
of all measurements could be determined but, surprisingly, we obtain
that measurements performed, say, on two same atoms (in the “same
state”) may give different results as implied by quantum mechanics.
On the other hand, the description of state is clearly defined phys-
ical question. We simply look for a suitable set of quantities that
characterizes nothing else than the state of the system. But it does
not explain the concept of state more than an identity of the system
connected to a “set of properties, qualities, specific interactions with
surrounding, etc.” So we must be careful. We should be aware that
the concept of state is rooted in those philosophical considerations
and is based on acceptance of some “identity” or “interchangeability”
of the system with another one but our effort should be concentrated
into the problem of searching for the best physical description of this
(undefined but accepted) identity.
To avoid many conceptual problems with explaining the mean-
ing of statements like a ‘system is in a state’, we will understand
the ‘state’ as a primitive, undefined concept. Let S be a phys-
ical system. We assume that there is a set Σ, called the state
set, that is assigned to this physical system, i.e. Σ(S). Any
element s ∈ Σ is called the state (of the system S).
At the intuitive level of the description, the state set is
explained by a potentiality of the studied object to occur at
various configurations (whereas individual configurations are
related mutually by the “identity” of studied object). Recall
that ‘physical system’ is not a material structure but rather
a special viewpoint from which we do study such a structure.
The state set is thus not absolutely given but it depends on that
2.1. STATES OF PHYSICAL SYSTEMS 33
what we want to study. If, for example, we study only various

movements of a piece of chock the state set is considerable
smaller then that arising when we also admit that the chock
may be broken or heated, etc. At the level of the description
in common language, the state set is thus a property of the
systematic physical study what is the meaning of the expression
“Σ(S)” at that level.
Nevertheless, can we specify at the abstract level of descrip-
tion whether, for example, the state of a glass of water means
a ‘macroscopic arrangement’ or a ‘conglomerate of coordinates’
of all molecules? It depends on that how we understand the
system, i.e. from which point of view we look at the studied
material object (e.g. if it is understood as a conglomerate of
molecules or as a macroscopic body). It has been said that
a specification of some representative subsystems helps us to
determine better the way in which we understand the system.
An individual molecule, for instance, is not a representative
subsystem of a macroscopically studied object.
On the other hand, states of the state set should express
possibilities that are important from the chosen viewpoint. For
example, if the body is studied as a macroscopic object then
states differing only by coordinates of a molecule are not impor-
tant possibilities and should not be understood as two states.
That is why we relate the state set with the representative set
of subsystems3 , Ξ(S), by the following rule:
For every S0 ∈ Ξ(S), there is a surjective mapping fsub ,
fsub : Σ(S) → Σ(S0 ), (2.1)
uniquely determined by the set Ξ(S) and the subsystem S0 .
3
Here, the representative set of subsystems is understood as a well de-
fined set because we use it in the fully formalized level of description.
The meaning of this rule is obvious: for any couple of different

states of S0 , say s01 and s02 , the states s1 and s2 defined by
s01 = fsub (s1 ) and s02 = fsub (s2 ) are different, i.e. the state
of S is “sensitive” on changes in S0 . Hence if we understand
that individual molecules are subsystems of the water then the
state of water means a special configuration of all molecules.
The state of any representative subsystem is thus determined
by the state of the system.
Though a more detail structure of states from the state set is
not introduced a state set may be constructed from other state
sets so that its elements are derived from elements of them. An
example is the state set of a subsystem whose elements may be
understood as being constructed from elements of the system
by the mapping fsub . Another case is given by the following
definition:
We say that systems S1 and S2 are mutually isolated if the
state set of its composition, Σ((S1 , S2 )), is a set of all couples
of states from Σ(S1 ) and Σ(S2 ), namely
Σ((S1 , S2 )) = Σ(S1 ) × Σ(S2 ). (2.2)
Every state of the composition of mutually isolated systems
is thus simply given by states of the both systems. On the
other hand, any couple of states may realize a state of the
composition. That means that mutually isolated systems do
not influence each other, in the sense that the state of one of
them is independent of the state of the other.
2.2 Manipulations with physical systems

The common physical study emphasizes the search for laws
of inner dynamics that govern the system behavior when it is
2.2. MANIPULATIONS WITH PHYSICAL SYSTEMS 35
let alone. In thermodynamics, however, studied systems are

mostly not let “alone”. We rather study various manipula-
tions with systems to find how effectively they may ‘produce’
work or how special changes of states can be realized. It is not
only due to technological applications of thermodynamics, we
cannot avoid some ‘manipulations with the system’ even in def-
initions of basic concepts, especially the ‘adiabatic accessibility’
that is deeply related to the concept of entropy (see Chapter 6).
That is why we define the ‘manipulation’ as another primitive
concept. Its meaning, however, is much more general than that
used in the standard thermodynamic description.
We assign to the studied physical system S a set M(S)
or simply M (if there is no problem with misunderstanding).
This set is called the set of manipulations and its elements
are called manipulations (individual manipulations will be
denoted by small Greek letters). They are defined as follows:
Any manipulation α ∈ M(Σ) is a subset of Σ(S) × Σ(S), i.e.
α ⊂ Σ(S) × Σ(S),
so that the set M(S) fulfills following conditions:
(i) The identity manipulation, id = {[s, s], s ∈ Σ}, be-

longs to M(S).
(ii) ∅ ∈ M(S).
(iii) For every couple of manipulations α, β ∈ M(S) there is

a manipulation γ ∈ M(S) so that whenever [s1 , s2 ] ∈ α
and [s2 , s3 ] ∈ β then [s1 , s3 ] ∈ γ.
The ordered couple of states [s, s0 ] ∈ α is called the change

under the manipulation α. For simplicity, the change will be
sometime denoted as s1 → s2 (without a specification of the

manipulation). The second condition describes the fact that if
there is a change s1 → s2 and the change s2 → s3 there exists
a manipulation γ that realizes the change s1 → s3 . Generally,
there may exist many such manipulations. Each of them is
called the composition of the manipulations α and β.
We introduce two useful auxiliary concepts. The set α|1 ,
α|1 ≡ {s ∈ Σ, ∃s0 ∈ Σ : [s, s0 ] ∈ α},
and the set α(s) for every s ∈ α|1 as
α(s) ≡ {s0 ∈ Σ : [s, s0 ] ∈ α}.
The meaning of manipulations is obvious: any ordered couple

[s1 , s2 ] ∈ α denotes a possible change of the state under the
manipulation α. It is worth stressing that it may not be defined
for all states, i.e. α|1 may differ from Σ. Namely, a concrete
influence on the system may be applied only in special states.
For example, the manipulation defined by “putting the body
between two pistons with a given distance ∆x and pushing it
by a prescribed force” is possible only at states in which the
corresponding size of the body has the dimension ∆x.
On the other hand, we can have more “results” of the same
manipulation under the system at the same state, i.e. α(s)
may have more elements then the only one. The reasons may
be various. For instance, the manipulation is “roughly” defined
and there are various finer details that “decide” which state will
be eventually realized. The other possibility is that there are
some inner bifurcations in the system allowing the realization
of more states (such as in the problem of buckling, see Fig. 2.1).
Figure 2.1: The manipulation defined by applying a given load on

a bar may cause different deformations of the bar, e.g. s → s1 or
s → s2 , (some of them may not be stable, however).
We say that the manipulation α is perfect if α is a mapping,

i.e.
α : α|1 → Σ.
To describe the “roughness” of the manipulation (i.e. how much

it differs from a mapping) we define the following concept:
We say that the manipulation α ∈ M is a parallel compo-
sition of manipulations β, γ ∈ M if α|1 = β|1 = γ|1 and
[s, s1 ] ∈ β and [s, s2 ] ∈ γ imply that [s, s1 ], [s, s2 ] ∈ α. We
denote
α = (β, γ).
Obviously, the parallel composition is commutative, i.e. (β, γ) =

(γ, β). The definition of parallel composition of an arbitrary
amount of manipulations β, γ, . . . (finite, infinite or even un-

countable) is obvious. We denote again α = (β, γ, . . .).
A manipulation α which gives multiple valued results (i.e.
changes one state into two or more different states) may be
only roughly defined and we look for finer defined ones (see
Fig. 2.2). That is, we search for perfect manipulations whose
parallel composition gives the ambiguous manipulation α. It
is a very important activity in experimental physics because
we ever try to obtain a unique result of an experiment (i.e. a
special manipulation with the system). In theoretical physics,
we also prefer a description of the problem (i.e. the choice of
suitable variables) in which transformations of states into other
states is unique as much as possible. That is why we define:
The manipulation α has a complete decomposition if there
is a set of perfect manipulations α(ι) so that
α = (. . . , α(ι), . . .),
where ι is from a linearly ordered set of indices so that individ-

ual items in (. . .) are ordered according to these indices.
To find a complete decomposition of a manipulation is a very
important physical problem which will play the crucial role in
this book. It concerns the problem of space that will be studied
in Chapter 5.
Remark: The identity mapping does not mean that “nothing has
happened”. The change [s, s] ∈ id means only that the ‘initial’ and
‘final’ states are the same. On the other hand, we cannot define more
than this condition because there is no “finer structure” underlaid the
change. We have no ‘time parameter’ that allows us to identify any
change with a part of time evolution (i.e. an infinite succession of
states parameterized by a time parameter).
Figure 2.2: The manipulation characterized as a “man’s hammer

blow” is very roughly defined: there are thousands of ways by which
a man can hit with a hammer. A complete decomposition of this
manipulation would be, say, a set of precise device that can impact
by a strictly defined way.
The concept of manipulation is quite general. It includes also

the situation when the system is “let alone”. Namely, it means
also a preparation of some special external conditions, namely
those that minimize interactions with surroundings. In our ap-
proach, however, there is no implicit assumption about such a
possibility. It is worth stressing that the study of a system ‘be-
ing alone without any external influence’ is a pure idealization
that can never be fully guaranteed. It is deeply related with
the Newtonian concept of ‘empty space’. We will study these
interesting questions in Chapters 5,6.
Manipulations are various external influences with the sys-
tem. If we manipulate with the system we manipulate with
each subsystem, too. The manipulation α with a system S
is represented by a set of couples [s, s0 ], where s, s0 ∈ Σ(S).
If S0 is a subsystem of S then the mapping fsub , given by
(2.1), defines the manipulation α0 with S0 as the set of cou-
ples [fsub (s), fsub (s0 )]. Nevertheless, the set of manipulations
with a subsystem may not be completely determined by the set
of manipulations with the system. Let us imagine, for exam-
ple, a molecule inside a macroscopic body. Understanding the
body as a system S we can have a broad set of manipulations
with this body, M(S), that may describe various macroscopic
influences on the body (defined by forces on its boundary, exter-
nal fields, heating, etc.). These manipulations, however, may
not describe manipulations with the molecule that is defined
by completely different processes and actions. If the molecule
is understood as a subsystem of S (i.e. the states of S are
sensitive on states of the molecule) then there is an evident
disproportion between the definition of the state set of S (it is
defined very finely) and the definition of the set of manipula-
tions with S which is defined very low. That is why we define:
Let Ξ be the representative set of subsystems and S0 ∈ Ξ.
We say that M(S) is consistent with Σ(S) if there are no
other manipulations in M(S0 ) than those represented by cou-
ples [fsub (s), fsub (s0 )] where the set of couples [s, s0 ] represents
a manipulation from M(S). The consistence of states and ma-
nipulations thus defines a surjective mapping g,
g : M(S) → M(S0 ), (2.3)
so that
g(α) = {[fsub (s), fsub (s0 )], [s, s0 ] ∈ α}. (2.4)
2.3 Processes
There is no idea of flowing time in which changes happen. We
have no temporal determination underlying the concept of ma-
2.3. PROCESSES 41
nipulations. If [s1 , s2 ] ∈ α then we may say that the state

s1 is “earlier” than the state s2 but it has no deeper meaning
than a formal description of the ordering of two states. It has
by no means an absolute validity. The physical system is not
“immersed” in an independent temporal background. We even
do not suppose an external flow of time. If something like the
“time flow” is described it has to be derived from the structure
of manipulations and not conversely (see Chapter 5).
This comment is very important in understanding the con-
cept of ‘process’. In thermodynamics, we speak often about
various processes because many concepts (like heat) may be
defined only in dependence on some processes. Also in our
abstract thermodynamic schema, processes play an important
role. We also accept the existence of various processes as con-
tinual successions of states corresponding to a continual change
of states of the system. We, however, do not explain these suc-
cessions as states in a continually flowing time. Our definition
sounds as follows:
The process π is a subset of states, π ⊂ Σ, so that there exists
a one-to-one mapping
πτ1 ,τ2 : hτ1 , τ2 i → Σ (2.5)
where hτ1 , τ2 i is a closed real interval with τ1 < τ2 ,
πτ1 ,τ2 (hτ1 , τ2 i) = π,
and for any couple τ, τ 0 ∈ hτ1 , τ2 i, τ < τ 0 , there is a manipula-

tion α ∈ M so that [πτ1 ,τ2 (τ ), πτ1 ,τ2 (τ 0 )] ∈ α.
The mapping πτ1 ,τ2 is called the parameterization of the
process. The set of possible parameterizations is restricted as
follows. If πτ1 ,τ2 and πτ10 ,τ20 are two parameterizations then the
mapping πτ−1 1 ,τ2

◦ πτ10 ,τ20 is a monotone, continuous function from
hτ1 , τ2 i onto hτ1 , τ2 i. It implies that if πτ−1
0 0
1 ,τ2
(s) is an inner point
of the interval then it is an inner point in any parameterization.
The state s ∈ π is said to be inner state of the process if
πτ−1
1 ,τ2
(s) is an inner point of the interval hτ1 , τ2 i.
The parameter τ is not a time. It is an arbitrary real parameter
that models the continual ordering of states. Any process may
be always approximated by an arbitrary large set of manipula-
tions and corresponding (finite) succession of changes. Namely,
if πτ1 ,τ2 is a parameterization of a process then for any division
of the interval hτ1 , τ2 i onto intervals hτ i , τ i+1 i, i = 0, . . . , N and
τ 0 = τ1 and τ N = τ2 , we have a set of manipulations α1 , . . . , αN
so that their composition forms the change from πτ1 ,τ2 (τ1 ) to
πτ1 ,τ2 (τ2 ).
Nevertheless, we cannot say that the process is a “continu-
ous change” because no such concept has been introduced. It
corresponds to everyday observations: we perceive only changes
of a state to another state – the feeling of a continuity arises
when we notice that any such change can be decomposed onto
finer ones.
The process is always defined for a finite real interval. We
would like to characterize processes that are defined on arbitrar-
ily small intervals (that allows to define some concepts assigned
to processes more locally). We do it as follows:
Let s be an inner state of the process π. The class of processes
Π(π, s) is called the tangent of the process π at s if for any
π 0 ∈ Π(π, s) the state s is an inner point of the process π 0 and
the intersection π ∩ π 0 is a process, too.
The tangent is thus a class of processes that are identical at s
and its arbitrarily small vicinity.
2.4. THE SURROUNDING 43
2.4 The surrounding
The origin of any manipulation with the system lies in a sur-

rounding of the system. However, how the ‘surrounding’ can
be defined physically? The most natural idea is to identify the
surrounding with another physical system, denoted by Ssur .
The question is how such a system should be specified. If we
identify it with something like a ‘rest of the Universe’ we get a
highly complicated system without a clear definition and spec-
ification. We want rather to work with a surrounding whose
structure is as simple as possible.
That is why we define the surrounding as a sufficient “source”
of all manipulations with the system. The word ‘source’, how-
ever, has to be defined with the help of concepts characterizing
any system (i.e. the system Ssur , too). How is a manipulation
with the system realized? First, we prepare specific external
conditions (e.g. prepare a measuring device interacting with
the system into a special state or protect the system against
an external influence by a barrier, etc.). These specific exter-
nal conditions may be characterized as a special state ssur of
the system Ssur (i.e. ssur ∈ Σ(Ssur )). Afterwards we manipu-
late with the device (e.g. only by keeping the protection) that
causes a manipulation with the system. The surroundings of
S thus may be characterized as the system Ssur so that any
manipulation with S corresponds to an (initial) state of Ssur
and a subsequent manipulation from M(Ssur ). The complete
definition sounds as follows:
The surrounding of a system S is another system Ssur which
is not a subsystem of S as well as S is not a subsystem of Ssur .
The set of states and manipulations of Ssur is related to those
of S so that for every couple [ssur , αsur ], where ssur ∈ αsur |1 ,
αsur ∈ M(Ssur ), there exists a uniquely given manipulation

from M(S). Moreover the mapping F,
F : Σ(Ssur ) × M(Ssur ) → M(S), (2.6)
defined by this assigning is surjective, i.e. for every manipula-

tion α ∈ M(S) there exists at least one couple [ssur , αsur ] so
that α = F([ssur , αsur ]).
That means that for every manipulation from M(S) (i.e. a
manipulation with the system S) there exists a state ssur ∈
Σ(Ssur ) and a manipulation from M(Ssur ) (i.e. a manip-
ulation with its surrounding) that realizes the manipulation
with the system. We say that the manipulation α ∈ M(S)
is caused by Ssur . This word reflects the (only) causality
accepted in our approach.
Any couple of different manipulations with S corresponds
to different elements of the set Σ(Ssur ) × M(Ssur ). Manipula-
tions with the surrounding are thus fair sensitive to cause any
manipulation with the system. On the other hand, the cou-
ple [ssur , αsur ] may not be defined uniquely (i.e. there may be
more “initial” states of the surroundings and more appropriate
manipulations with the surroundings). It may mean that the
surroundings of the system is needlessly “large”. Maybe, it in-
cludes less states and manipulations to realize any manipulation
with the system. The following Lemma is useful in this context:
Lemma: Let M(Ssur ) be consistent with Σ(Ssur ). If a sub-

system S0sur of Ssur is a surrounding of S then Ssur is the
surrounding of S, too.
sub
Proof : The consistence means that for each subsystem S0sur ⊂ Ssur
there is a surjective mapping g : M(Ssur ) → M(S0sur ) defined by
2.4. THE SURROUNDING 45
(2.3). The assigning described by the function F is then given as
F([ssur , αsur ]) = F0 ([fsub (ssur ), g(αsur )]), (2.7)
where [ssur , αsur ] is an arbitrary element of Σ(Ssur )×M(Ssur ), fsub

is the function defined by (2.1) and F0 is the mapping (2.6) defined
for S0sur .
The surrounding S0sur introduced in the Lemma is said to be
the simpler surrounding than Ssur .
The surrounding is a source of all manipulations with the
system. This may be, however, a source of misunderstand-
ings. When we completely change the surrounding (e.g. when
putting a living cell from an organisms into a test tube) the new
surrounding may not be a source of the same manipulations as
before. However, we cannot forget that the concept of physi-
cal system itself is not defined absolutely – it depends also on
surrounding (the cell in organisms is different physical system
then the same cell in the test tube). Nevertheless, to explain
physically what means a “change of the surrounding” while the
system “remains the same”, we may assume, for example, that
the state set of a “new” physical system, say Snew , (formed by
a new surrounding) remains the same as that of the original
physical system S, i.e.
Σ(Snew ) = Σ(S). (2.8)
The set of manipulations, however, may be different, i.e.
M(Snew ) 6= M(S). (2.9)
Remark: Notice that the state set is the concept that helps us to fix a
“material identity” of the system. The change of surrounding, defined
by (2.8), thus cannot “destroy” the system in the meaning that some
states “vanish” from the state set or some new ones “appear”. It is ex-
actly that what is supposed in experimental physics – an experiment
means a change of a “natural” surrounding of the studied system.
Nevertheless, we believe that the system can be in same states as be-
fore. Notice also that realizations of experiments is usually intended
to enlarge the set of manipulations, i.e. M(Snew ) ⊃ M(S) in (2.9).
2.5 Global variables

The core of any physical description lies in assigning various
quantities to the system. When varying the state of system
(i.e. s ∈ Σ changes to s0 ∈ Σ) we obtain various values of
these quantities and thus they are usually called ‘variables’. We
define special variables that are assigned to the whole system.
The global variable G is a mapping
G : D G → BG , (2.10)
where BG is a Banach space (of this variable) and DG ⊂ Σ.

The Banach space is a suitable structure for defining values of
variables because it is a complete normed vector space4 . Usu-
ally, B = Rn . Notice the important feature of the mapping
(2.10): there may be states for which the global variable is not
assigned to the system (e.g. ‘equilibrium temperature’ is not
defined at non-equilibrium states as illustrated in Fig. 2.3).
Let us assign to the system S a set of global variables
Qg (S) (or Qg ) so that for any global variable G ∈ Qg (S) the
set of states DG is not empty (nevertheless the set Qg (S) itself
may be empty). Global variables included in Qg characterize
4
That means that its elements have a size (a nonnegative real number)
called the norm. The norm of an element b is denoted by ||b||, [8].
2.5. GLOBAL VARIABLES 47
Figure 2.3: A global variable G is a mapping from an abstract state

set Σ to a corresponding Banach space BG . An important feature of
the presented approach is that a variable may not be assigned to all
states – its definition domain may be only a “small” subset of Σ.
the system as a whole. Their values (or changes of these values

when the state of the system is changed) should be determined
somehow. This determination is done by various experiments
that means a form of “communication” of the system with its
surrounding. The value of a global variable (or its change)
should be in principle detectable by the surrounding5 . We in-
troduce two important classes of variables whose values have a
clear relation to a suitable class of surroundings.
The first class are global variables whose values only “flow
over” the system and its surrounding and their total values
remain the same. These variables correspond to some conser-
vative properties, such as energy, momentum, angular momen-
5
Recall that a change of the surrounding may change the physical sys-
tem, too. We look for such a surrounding that determines the system for
which a corresponding global variable is definable.
tum, mass, electric charge, etc. It motivates us to the following

definition:
We say that the surrounding of a system S has the conser-
vative property with respect to the global variable G if this
variable is defined both for the system and the surrounding,
and for every ssur ∈ Σsur , αsur ∈ Msur and every couple
[s, s0 ] ∈ F(ssur , αsur ) the condition
G(s) + G(ssur ) = G(s0 ) + G(s0sur ) (2.11)
holds whenever [ssur , s0sur ] ∈ αsur .

Notice that the conservative property is not a property of the
global variable G alone but rather its relation to a concrete
surrounding. It may seem a little surprising because we un-
derstand, say, the momentum as a conservative variable in an
absolute meaning. However, if a fixed massive wall is part of
the surrounding then the momentum of the system is not a
conservative variable6 .
The second class of variables having a clear relation to spe-
cial surroundings is called the ‘generalized temperature’. In
special conditions, values of these variables only “copy” their
values on the surrounding and thus may be simply determined.
These “special conditions” are called the ‘equilibrium’ and play
a very important role in physics. Let us study them in next sec-
tion.
6
If, for example, the system is a microscopic particle and the surround-
ing is formed by macroscopic objects (walls, etc.). We may assert that
there is a change of the momentum of the wall after the impact of the
particle though, but this change is not definable on the macroscopic sur-
rounding. We concede that this kind of thinking is not familiar to standard
physics, where such concepts as ‘macroscopic surrounding’ are not taken
into account quite seriously.
2.6. TEMPERATURE AND EQUILIBRIUM 49
2.6 Temperature and equilibrium

The temperature is a very important physical quantity that is
deeply related to the second law of thermodynamics. Namely,
its basic property is a “tendency” to be the same over the whole
system. Be motivated by this property, we define the concept
of generalized temperature as follows:
Let S be a system and Ξ the representative set of its subsystems.
The global variable T is called the generalized temperature if
it belongs into the set of global variables of every subsystem from
Ξ, and has the following property: For every s ∈ Σ and every
subsystem S0 ∈ Ξ holds that if s ∈ DT then fsub (s) ∈ DT0 , where
DT0 ⊂ Σ(S0 ) is the definition domain of T on the subsystem,
and fsub is defined by (2.1). Then the relation
T (s) = T (fsub (s)) (2.12)
is always valid.
The meaning of the definition is obvious. The generalized tem-
perature of a system is defined only if it is defined for every
subsystem from the class of subsystems Ξ of the system and its
value equals to values of this global variable on all subsystems
from Ξ.
The equilibrium temperature defined in thermodynamics is
the best case of such a global variable. Its high importance is
evident: even a special law – the zeroth law of thermodynamics
– is formulated to define this variable (see Section 6.5).
Remark: The generalized temperature may not be the thermody-
namic temperature. It may be any quantity whose values on the
system always match with those on its (representative) subsystems.
A nontrivial example is ‘time’ if its values (data measured by a clock)
are understood as values of a global variable (why not, we will show
this possibility in Section 6.3). In this case, the definition domain

covers all states of the system and the condition (2.12) is trivially
valid.
The thermodynamic temperature is defined in equilibrium. Let
us explain this concept and show its relation to the generalized
temperature. If we want to specify some physical or chemical
properties of the water in a vessel we simply take a small sample
of water and do corresponding measuring that determine these
properties. Afterwards, we have information about the water in
the vessel. However, we have obtained properties of this sample
only and not those concerning the water that remains in the
vessel. We simply believe that the water in the sample is the
same as the surrounding water. The similar situation happens
when we measure the temperature by a mercury thermometer.
The datum of the thermometer is nothing but the temperature
of the mercury. We believe again that it corresponds to the
temperature of the surrounding air.
Such belief relies on a very important physical assumption.
Namely, that there is some equilibrium between the thing (phys-
ical system) whose properties we really measure (a sample of
the water, the mercury within the thermometer, etc.) and its
surrounding. The equilibrium then guarantees that some physi-
cal properties are the same, e.g. temperature, density, pressure,
chemical composition, etc.
The concept of equilibrium is thus some relation between
the system and its surrounding. It plays a very important role
in thermodynamics. In our abstract approach, we will specify
a basic feature of this relation, namely the fact that any state
of the surrounding defines a concrete state of the system.
2.6. TEMPERATURE AND EQUILIBRIUM 51
The surrounding Ssur is said to be in equilibrium with the

system S if there is a one-to-one mapping,
feq : Σsur → Σ,
so that the mapping (2.6) is then defined by the prescription

M
F([ssur , αsur ]) = feq (αsur ), (2.13)
M : M
where the mapping feq sur → M is defined by the condi-
tion
[ssur , s0sur ] ∈ αsur ⇔ [feq (ssur ), feq (s0sur )] ∈ feq

M
(αsur ).
(2.14)
Moreover, if there is a generalized temperature T defined for
Ssur then the same global variable is defined for S and
T (ssur ) = T (feq (ssur )) (2.15)
for every state ssur ∈ Σsur .

The function feq represents the above-mentioned “coupling”
of the surrounding with the system. This coupling forms an
extremely simple structure of manipulations with the system –
the system copies changes of the surrounding.
Figure 2.4: The picture may serve as a summary of main concepts

introduced in this chapter. We choose a part of reality (e.g. a pin) an
study it as a physical system, S. A “material identity” of the system
is characterized by its state set, Σ(S) (we understand what means the
‘falling pin’, ‘laying pin’, ‘pin at a different position’, etc.). However,
individual states, per se, are not defined (they may be only described
by various variables – see the next chapter). The dynamics of the
studied system is defined by various changes – couples of states. As
such, they are a property of the system. All possible changes form
the set of manipulations, M(S). The source of all manipulation
are various bodies around the system (other pins, a rolling ball, the
Earth, etc.). Namely, these bodies may cause various changes of
states. All such bodies form another part of reality, a physical system
called the surrounding, Ssur .
Chapter 3
Systems as sets of parts
If we decide to study a thing as a physical system we choose

a special viewpoint from which the thing is thought of. The
basement of our thinking is an idea of some set of states that
means various “modes of being” of the studied thing. The state
set is a field of potentialities of its occurrence: different states
means different “configurations” of the same thing. But how
may this “configuration” be precisely defined and described?
That is, how should the concept of state be connected with its
description?
In classical physics, the state is characterized by a current
distribution of some properties in space. At a moment, the
body has a concrete distribution of the mass density or elec-
tric charges, velocities of individual parts localized at different
spatial points, special spatial configurations of various fields,
and so on. A different state is then characterized by different
distributions of all these properties. This way of understanding
depends essentially on the idea of a fixed space. The distribu-
tion of properties means the distribution in space.
54 CHAPTER 3. SYSTEMS AS SETS OF PARTS
In our approach, however, we try to avoid the concept of

space from many reasons. That is why we have to introduce
a compensation of this fixed background by something what
may also serve as a fixed reference but what is not absolute
and unique and, in particular, is not so encumbered by many
metaphysical assumptions. We introduce some ‘parts’ that are
fixed building blocks of detail description of states of a system.
For any physical system, a fixed set of such parts is introduced.
This set, in a way, replaces space – states are not characterized
by distributions of properties in space but by properties on indi-
vidual parts. The property, as usual in physics, is described by
a variable whose values are points of a mathematical structure
(usually a set of real or complex numbers). These variables, also
called ‘local variables’, differ from the global variables defined
in Chapter 2. While global variables are supposed to repre-
sent some fundamental features of studied systems including
some relations to their surroundings (conservative properties,
equilibrium, etc.), the (local) variables characterize quantita-
tively individual states of the system with respect to a chosen
description. They are defined in dependence on the way of de-
scription (a choice of the set of parts) and their main purpose
is to characterize the state set as well as possible1 .
What do we mean by ‘parts’ ? They may be formed by vari-
ous pieces of matter and identified with a chosen subsystems of
the system. For example, individual bodies of a many-body sys-
1
In physics, the state of the system is often fully identified with a set
of values of all variables that is represented by a point in the so-called
state space. Here we try to avoid such an identification because then two
different levels of the description are mixed together. Namely, the same
system with the same state set may be described differently by choosing
different sets of parts. It influences the choice of variables and we may
obtain different state spaces.
55
tem may be understood as ‘parts’2 . Alternatively, they may be

artificially chosen as pure auxiliary concepts having no relation
to natural “pieces” of the system. An extreme example of such
parts are ‘spatial points’ – parts originated from the concept of
space as will be explained in Chapter 5. Another example are
various ‘boundaries’ used in continuum thermomechanics as it
is described in Chapter 4.
Parts may be composed together to form a larger part. The
idea of ‘composition’ comes from macroscopic systems that con-
sist of a whole hierarchy of various material parts, that are
composed from smaller parts, etc. Nevertheless, its meaning is
essentially broader. It is a very important operation in equi-
librium thermodynamics, continuum thermomechanics and in
such fields of physics where the description of a system at vari-
ous scales (‘scale’ means a minimal or typical size of the descrip-
tion) is systematically studied3 . These questions are deeply
related to a possibility of a natural definition of the concept
of ‘averaging’ that plays the key role in physics of structures
including a huge number of “microscopic” degrees of freedom.
The understanding of a physical systems as a set of parts
is always a way of description. There is no formal reason to be
restricted only on macroscopic systems which may be naturally
2
It should be emphasized that if parts are chosen to be individual par-
ticles of a many-particle system we do not need the concept of space in
which they ‘are’. Namely, their positions (as well as velocities and other
spatiotemporal characteristics) may be understood as variables assigned
to them in dependence on concrete states.
3
The same system is described differently at various scales. The way of
transition between these descriptions plays important role and is studied
by various methods. The most famous is the so-called renormalization
group approach that was systematically introduced by K.G. Wilson for
applications in statistical physics and the quantum field theory [46, 47].
understood as actual sets of parts. If we find a sensible way in

which, say, an atom may be understood as a set of parts in our
way of description then there is no formal reason to refuse the
developed conceptual framework. It opens an appealing ques-
tion if some concepts introduced and used only for macroscopic
objects may have a sensible meaning also for microscopic ones.
3.1 Structure
Physical systems introduced in Chapters 1,2 are determined
only by their global properties. States from Σ are states of the
whole system, global variables are assigned to the whole sys-
tem and the manipulation causes the change with the whole
system, too. Only the presence of subsystems indicates that
there is some inner structure but there is no idea how subsys-
tems should be put together to form the system. Moreover,
there is no clearly defined set of all subsystems (of a system).
In a standard physical approach a studied system is localized
in space by occupying a region (whereas the region may consist
only from several points as it is in newtonian mechanics of mass
points). Hence a ‘substructure’ of the studied system is simply
given by “data” obtained from various subparts of this region.
Since the idea of ‘space’ in which the system is located is
not introduced in our approach, we choose a ‘structure of de-
scription’ of the system by introducing a formal set, P, whose
elements are various parts. Certain couples of parts (‘disjunct’
parts) may be combined together to form a ‘larger’ part. Thus
larger and larger parts arise till a total (‘largest’) part p0 ∈ P,
including any other parts, is made up. Such a part, however,
may not exist. Parts and their compositions are understood in
an abstract level – there is no spatiotemporal characterization
3.1. STRUCTURE 57
of the concept of part and composition. We emphasize again

that the structure of parts means a scheme of description of
a physical system which is based on some primitive concepts
such as ‘states’, ‘parts’, ‘composition’, etc. Nevertheless, we
can imagine many other ways of description underlying other
aspects of the studied physical structure.
Let S be a physical system. We introduce an abstract set
P(S) called the set of parts of this system (the specification
of the system will be omitted if there is no problem with misun-
derstanding). The basic operation is the composition of two
parts to form the one part. This operation may be defined as
a mapping,
Comp : Dcomp → P, (3.1)
where Dcomp ⊂ P × P. Parts p1 , p2 ∈ P forming the couple
belonging to the domain of Comp, i.e. [p1 , p2 ] ∈ Dcomp , are
said to be disjunct4 . The composition of parts p1 , p2 is a part
that is denoted as (p1 , p2 ). The following rules describe the
basic properties of the composition operation:
(i) [p1 , p2 ] ∈ Dcomp ⇒ [p2 , p1 ] ∈ Dcomp and

(p1 , p2 ) = (p2 , p1 )
(ii) [p1 , p2 ], [(p1 , p2 ), p3 ] ∈ Dcomp ⇒ [p2 , p3 ], [p1 , (p2 , p3 )] ∈

Dcomp and ((p1 , p2 ), p3 ) = (p1 , (p2 , p3 )).
The second property allows defining the composition of more

mutually disjunct parts, namely
(p1 , p2 , p3 ) ≡ ((p1 , p2 ), p3 ), etc.

4
There is an important difference between the composition of systems
and the composition of parts. The first may be defined for an arbitrary
couple of systems while the later only for disjunct parts.
The set of parts endowed with the mapping Comp is called

the structure of the system S. The structure thus may be
understood as a couple (P, Comp).
We say that the part p0 is a subpart of the part p if there is a
part p̃ ∈ P so that p = (p0 , p̃). Then we write
sp
p0 ⊂ p.
sp
Notice that p0 ⊂ p means that there is another part, say p00 ,
Figure 3.1: Parts resemble a suitable chosen “bricks” that may be

put together to model the structure of the system as much as possible.
A physicist describing a system as a set of parts thus resembles a child
trying to find a suitable composition of puzzle blocks.
that, put together with p0 , gives p (part is not a subpart of

itself!). Parts thus resemble bricks that are put together to
form the system. Let us define several useful concepts:
The part p0 is called the total part if for every p ∈ P, p 6= p0 ,
sp
p ⊂ p0 .
3.1. STRUCTURE 59
The part p is said to be atomic if there are no parts that are

subparts of p. The set of all atomic parts from P is denoted as
Patom .
If the part includes a subpart then it is said to be composed.
The structure is said to be atomic if any p ∈ P is a composi-
tion of parts from Patom .
The structure is said to be finitely atomic if it is atomic and
Patom = {p1 , . . . , pN }.
The structure that does not include any atomic parts, i.e. Patom =
∅, is called the non-atomic structure.
In non-atomic structures there are only composed parts. It
means that any part p ∈ P includes an infinite succession of
subparts p1 , p2 , . . . so that
sp sp sp sp
. . . ⊂ pi+1 ⊂ pi ⊂ . . . ⊂ p. (3.2)
In atomic structures, every part from P is a composition of

finite number of parts from Patom because the composition is
defined only for a finite set of parts. The situation when parts
are composed from an infinite number of atomic parts cannot
be formalized in our approach5 . Instead we define:
The structure is said to be weakly atomic if every composed
part from P includes a part from Patom as its subpart.
Though weakly atomic structures include an “underlay” set of
atomic parts, they have some features of non-atomic ones as
shown in the following Lemma.
Lemma: If the structure is weakly atomic, but not atomic,
then there is a part p ∈ P that includes an infinite succession
of composed parts, p1 , p2 , p3 , . . ., fulfilling (3.2).
5
We would like to avoid the use of a form of transfinite induction, etc.
Proof : Let us take an arbitrary composed part p ∈ P. If struc-

ture is weakly atomic then there is p0 ∈ Patom so that p = (p0 , p1 ).
If the part p1 is composed then there is an atomic part p01 so that
p1 = (p01 , p2 ). If p2 is composed we go further and obtain a chain of
sp sp sp
parts pi ⊂ . . . ⊂ p1 ⊂ p till the last part of the composition is atomic
too. This situation, however, cannot happen for every composed part
of P because then every part would be a composition of atomic parts.
Therefore there is a part for which the chain is an infinite succession
fulfilling (3.2).
Let us present several examples.
Structure of mass points: We have a set of N mass points
(Newtonian particles). Individual particles form atomic parts
p1 , . . . , pN and the structure is finitely atomic. This structure is
the most often used structure in newtonian mechanics, statisti-
cal physics and so on. Parts arising by compositions of atomic
parts (i.e. consisting of more mass points) play important role
when various interactions of particles are described. Special
variables describing these interactions (e.g. an interaction en-
ergy) are assigned to couples (or triplets or larger collections)
of particles.
Structure of equilibrium cells: The structure is formed by
n atomic parts – simple thermodynamic systems in equilibrium
(see Section 4.1). It is used in equilibrium thermodynamics. It
is also finitely atomic. The fact that compositions of two atomic
parts may form another part that is described by same means as
atomic ones is crucial – see the beginning of the Callen’s book
[2] where the operation of removing of some inner constrains
(e.g. a barrier) explains main features of thermodynamics. The
removing of constrains is nothing but a formation an equilib-
rium cell as a composition of several atomic parts.
3.1. STRUCTURE 61
Structure of spatial points: Let P be identical with d-

dimensional Euclidean space, Ed , i.e. parts are points of Ed .
Since points have no subparts the structure consists only of
atomic parts and is atomic (but not finitely atomic). This
structure plays crucial role when describing the system by spa-
tial fields (see Chapter 5).
Structure of sets of spatial point: If the previous structure
is enriched by various sets (finite or infinite) of spatial points
it becomes weakly atomic. This structure is seldom used in
physics.
Structure of regions: We understand the ‘region’ as an arbi-
trary measurable subset of points of Ed so that its d-dimensional
Lebesgue measure (volume) is nonzero. Let us have a region Ω
and form the set of all subsets of Ω that are regions, too. A re-
gion Ω1 is a subpart of a region Ω2 if Ω1 is a proper subset of Ω2 .
This structure has no atomic parts because any region includes
a proper subset that is a region, too. This structure plays em-
inent role when modeling continuum systems (see Chapter 4).
The structure (parts and their compositions) may also model

the fact that the one physical structure has usually an infinite
number of levels of description; whereas the ‘level of descrip-
tion’ means a scale under which all minute details are neglected
or, say more precisely, averaged. The composition operation
thus represents the passage to less accurate descriptions of the
system. But why do we try to model cascades of more and
more inaccurate descriptions? The reasons are two, at least.
First, each physical description of a concrete system begins
with neglecting or smoothing out some details (inner structure,
microscopic degrees of freedom, some processes) at lower scales.
Even a description at a deep microscopic level has to neglect
something (quarks, fluctuations of a space-time foam or some-

thing else). Therefore we are ever somewhere “between” a more
detail or more inaccurate description and the understanding of
steps ‘up’ or ‘down’ is of the crucial importance. Second, we are
not able to model complex systems at an arbitrarily detail level
and the search for a “compromise” description, which models
appropriately behavior of the systems, plays the key role in
applied physics. The systematic way of such a search means
the understanding of the passage to less accurate description
at larger scales.
Atomic structures play important role in physical thinking.
They model the situation in which there is a scale that has an
eminent meaning. Processes under this scale have small or neg-
ligible influence on the global behavior of the studied system.
The description at this scale thus seems to be the fundamental
one and usually means a conscious or unconscious suppressing
of other descriptions. The discovery of the atomic structure of
matter is a good example – we believe that there is a possibility
to explain all macroscopic behavior at this fundamental level.
Nevertheless, we have no method how to do it actually and that
is why we eventually construct more inaccurate (but realistic)
descriptions at higher scales.
3.2 Parts as subsystems and other parts

Introducing the concept of parts and their compositions is mo-
tivated by the mechanical understanding of the system as a
structure consisting of various parts (and the state as a configu-
ration of these parts). Physics follows this thought line not only
in mechanics and thus physical variables are assigned to indi-
vidual parts of the system, whatever is their concrete meaning.
3.2. PARTS AS SUBSYSTEMS AND OTHER PARTS 63
This procedure enables us to “catch” the system by sophisti-

cated mathematical tools. A typical example is the approach
used in continuum theories, where a decomposition of the sys-
tem into a set of infinitesimal cells allows us to describe states of
the system by a suitable chosen set of spatial functions. The set
of parts defines the description of the system and the ‘system’
(described only in the common language) thus becomes a math-
ematically definable structure (see Fig. 3.2). Parts themselves,
Figure 3.2: The system is illustrated by a fuzzy region to express its

only “verbal” definition by a phrase like “let us study a hurricane”
or “we describe a living cell”, etc. From the viewpoint of a precise
mathematical language, the ‘system’ is not clearly described and de-
fined. Therefore a structure of parts is introduced to “cover” them
as well as possible to form a definable structure.
however, may or may not be identified with certain subsystems

of the system. For example, a part in the mentioned continuum
description (an infinitesimal cell) can be hardly identified with
a macroscopic subsystem (in continuum description the system
as well as any representative subsystem has to be macroscopic).
Any subsystem is a full-valued system which implies that it

has its state set. On the other hand, the state set for a part
is not generally defined (contrary to the so-called state space
which is defined for individual parts, see Section 3.5). We only
assign values of some variables to parts in dependence on the
state of the whole system. It may be a source of various misun-
derstandings. Namely, we need not suppose that the part is in
a (physically defined) state because the “state of part” is not
introduced and have no sense for parts that are not subsystems.
We divide the set of parts into two subsets,
P = Psub ∪ Pch . (3.3)
The first set, Psub , may be identified with a subset of the set
Ξ (a set of representative subsystems of the studied physical
system) and some supplementary parts. The second set, Pch ,
includes some other parts, whose definition will be concretized
lately.
First, let us define the part Psub . The idea is as follows.
There is a set Ξ0 , Ξ0 ⊂ Ξ, and a one-to-one mapping Fs from
Ξ0 into a subset of P,
Fs : Ξ0 → P, (3.4)
so that for every S1 , S2 ∈ Ξ0 the following relations are valid.

Namely, if S1 6= S2 then
sub sp
S1 ⊂ S2 ⇔ Fs (S1 ) ⊂ Fs (S2 ), (3.5)
and if p1 , p2 ∈ Fs (Ξ0 ) then
[p1 , p2 ] ∈ Dcomp ⇒ (p1 , p2 ) = Fs (Fs−1 (p1 ), Fs−1 (p2 )) (3.6)

3.2. PARTS AS SUBSYSTEMS AND OTHER PARTS 65
(notice that (Fs−1 (p1 ), Fs−1 (p2 )) is a composition of subsys-

tems). The relation (3.5) implies that there must be a part
p ∈ P so that Fs (S2 ) = (Fs (S1 ), p) (see the definition of sub-
parts). However, no subsystem corresponding to the part p
may be included in Ξ0 . That is why we complete the set Fs (Ξ0 )
by adding all such parts. It forms the set Psub . That means
that if p ∈ Psub then either S0 ∈ Ξ0 exists so that p = Fs (S0 )
or there exist S1 , S2 ∈ Ξ0 so that Fs (S2 ) = (Fs (S1 ), p). In the
second case we will say that p is a complement part.
If Pch = ∅, parts from P model individual subsystems or
complement parts. Contrary to subsystems alone, the concept
of part has features of spatial determination – if p1 and p2 are
disjunct (i.e. the couple [p1 , p2 ] belongs to Dcomp ) then they
may be understood to be spatially separated. The relation
(3.6) then transfers this property onto subsystems (which are
assigned to some parts).
Parts from Psub are defined by a set of subsystems. It means
that they have a natural physical meaning (parts correspond-
ing to subsystems have clear physical meaning, complementary
parts are in fact defined by couples of subsystems). Parts from
Pch , if any, do not have such a natural status. They may be
said to be chosen artificially for purposes of the description.
Let us explain it in the case when the total part, p0 , belongs
to Pch , i.e. if it cannot be identified with the system alone by
the relation Fs (S) = p0 . Imagine, for example, a closed, firmly
chosen container including a piece of a special substance, say a
liquid. The liquid can be in a vast number of states. For ex-
ample, it can occupy various parts of the container, change its
phase and become a gas expanding to the whole container, etc.
Though the system is the liquid, we usually use the description
whose variables are defined rather for the container alone than
for the liquid, e.g. by understanding its volume as a variable.

In our language, the container defines the total part p0 which
is not identical with the system alone.
Remark: There are some puzzling questions whose origin lies in iden-
tification of the system with a defined total part. For instance, the
entropy of ideal gases depends on their volumes (which has an effect
if two gases are mixed together). But if we identify the system - ideal
gas - with a collection of its molecules, the dependence of the entropy
on the volume is puzzling – is the entropy a property of empty space
between molecules?6 In thermodynamics, the difference between the
system (as a material “holder” of physical properties) and externally
defined parts is important and has to be clearly defined. Several
problems connected with the definition of entropy lie in this differ-
ence.
We generalize this example by postulating that parts from

Pch mean something externally defined, chosen for purposes of
the most suitable description. The external definition may be
realized by various pistons, walls, special boundaries or barriers
controlling (and defining) some parts of the system exogenously.
All such “machinery” namely should be understood more as
belonging to the surrounding than to the system alone. That
is why we accept that the set of parts Pch is a property of the
surrounding, i.e.
Pch (Ssur ),
what means that these parts are describable by concepts con-
cerning the surrounding. It may seem strange if the artificially
defined parts from Pch are only mentally delimitated by some
thought barriers somewhere inside the system. However, as we
6
An evident manifestation of this puzzle is the so called Gibbs paradox.
3.3. LOCAL VARIABLES 67
will see later, the possibility of doing that is supported by the

idea of (empty) space. And space alone may be understood as a
very special kind of surrounding. It will be shown in Chapter 5.
3.3 Local variables

Global variables express some basic quantitative characteriza-
tion of individual states of the system. They are introduced
on a rather abstract level of the description because their def-
inition relies only on the idea of state. When introducing the
structure of the system – a set of parts that are chosen to de-
fine the description of the system – a new sort of variables
may be introduced. We call them the ‘local variables’ or only
‘variables’.
The set of parts, P, is fixed, i.e. it does not depend on the
state of the system (it is an important property of the descrip-
tion using space as a fixed background – parts only generalize
this conception). The dependence of the structure of system on
its state is then expressed by various values of these variables
that are assigned to individual parts.
Remark: Imagine a physical system formed by a water (in a vessel)
and a piece of floating ice. The ice is a clearly definable subsystem
of the system. The state set of the system includes also states when
there is no ice (because it has melted). At these states, this sub-
system becomes indefinable (we may identify it with a collection of
molecules dispersed somewhere in the water but it may be hardly
understood as a good definition of a subsystem). Nevertheless, the
piece of ice may be defined as a part from P. There are various vari-
ables assigned to this part (mass, momentum, energy, temperature,
etc.). At states when there is no ice, this part exists, but values of
these variables correspond to the way in which we understand this
part. E.g. they are zero or equal those of the water, and so on.
This approach seems rather strange and artificial, but notice that
we do the same when working with spatial background (space exists
permanently only values of physical variables at its individual points
may change and sometime have some special values, e.g. zero mass
density, etc.).
We define the ordered set Q called the set of variables. An
element q̂ ∈ Q, called the variable or local variable, defines
for every state of the system s ∈ Σ the mapping qs ,
qs : Dq,s → Bq , (3.7)
where Bq is a Banach space whose elements are values of the
variable q̂ (e.g. Bq = Rn ). The domain of the mapping, Dq,s ,
is a subset of P.
The question how large is the domain Dq,s plays the crucial
role. Namely, the fact that the composition of parts forms
another part is thus connected with the question whether a
variable can or cannot be defined for this composition (i.e. if
the composition of parts belongs into the domain, too). It is of
the crucial importance. The repeated composition is a passage
to larger scales. Variables which “survive” this passage are
those which play crucial role in the macroscopic description of
matter.
The question, if a (local) variable may be identified with
a global one, is nontrivial. For example, the global variable of
the crucial importance is the (global) energy. We believe that it
may be assigned to any state of an arbitrary system (at least in
classical physics) and we are able to measure its changes during
various processes. On the other hand, we are able to define the
(local) energy of various parts of the description (e.g. as the
sum of kinetic and potential energy of a particle). Nevertheless,
the definition of a concrete form of the local energy may not

be in agreement with results of those measurements. Then we
have to search for a better form of local energy and (maybe) a
better choice of a set of parts.
The variable q̂ may be formally enlarged to a global variable
G ≡ G(q̂) if the structure includes the total part p0 . Namely,
the total part is defined uniquely that allows defining uniquely
the mapping (2.10) by choosing DG = {s ∈ Σ, p0 ∈ Dq,s } and
defining
G(q̂)(s) = qs (p0 ). (3.8)
The essential problem, however, is to find a local variable cor-
responding to a given global one. Namely, global variables ex-
press some fundamental characterization of systems and our
choice of local variables should correspond with that. We could
define formally the mapping (3.7) for Dq,s = {p0 } (if there is
a total part) by the prescription (3.8) but the variable has to
be defined also for smaller parts (as guaranteed by the continu-
ity rule formulated afterwards). The search for local variables
corresponding to global ones is a very important problem of
physics.
The way in which concrete values of a variable are assigned
to individual parts is a complex physical problem which may
be studied only in concrete cases. Generally, a different way of
this assignment is expected if the part is from Psub or Pch . In
the first case, the variable is assigned to a subsystem and thus
it may (but may not) correspond to a global variable defined
on the subsystem (e.g. the energy of subsystem in continuum
thermomechanics). If the part is from Pch it is defined exter-
nally by means of a surrounding and thus values of a variable
defined on this part should be also defined by means of this
surrounding (we will discuss it in more detail in Chapter 4).
Remark: One concrete physical body (e.g. a piece of chalk) may be

described by an infinite number of various choices of a set of parts and
a set of global and local variables. We can choose an extremely rough
description representing the system by one part (e.g. when studying
it as a rigid body) or a very fine description in which the system is
understood as an infinite set of parts (in a continuum description).
The choice of a set of global and local variables depends on the set
of states (e.g. when supposing only pure mechanical displacements
in space or if the system may change temperature, may be broken to
several pieces and so on), but neither Qg nor Q strictly depend on Σ.
Namely, the chosen set of variables expresses (as well as a splitting
the system into parts) a degree of simplification of a real structure.
We can accept a broad variety of states but describe them by a small
set of variables. On the other hand, a rich group of variables may
show itself as superfluous because there are no meaningful changes of
some of them on the variety of studied states.
Some special kinds of variables are very important to be

defined:
The variable q̂ is said to have the averaged property if for
any couple of disjunct parts p1 , p2 the following implication is
true in any state s, namely
p1 , p2 ∈ Dq,s ⇒ (p1 , p2 ) ∈ Dq,s . (3.9)
Variables with averaging property thus “survive an arbitrary
long trip” to higher scales and play an eminent role in de-
scription of macroscopic systems. A very important class of
variables with averaging property is formed by the so-called
additive variables. They are defined as follows:
The additive variable q̂ is a variable with averaging property
(i.e. it obeys (3.9)) that fulfills moreover the following relation
for every state s ∈ Σ,
qs (p1 , p2 ) = qs (p1 ) + qs (p2 ), (3.10)
where we omit the double parenthesis and write simply (p1 , p2 )

in place of ((p1 , p2 )).
Now we define the relation of two special variables that is very
important when parts are modeled by a set structure:
Let values of a variable q̂ 0 be only nonnegative real numbers
(i.e. for every state and every part, qs0 (p) ≥ 0). We say that q̂
is absolutely continuous with respect to the variable q̂ 0 if for
every state s ∈ Σ the following statement is valid: For every
> 0 exists δ > 0 so that for every part p ∈ Dq,s ∩ Dq0 ,s for
which qs0 (p) < δ holds that kqs (p)k < where k.k is the norm
in Bq . We denote
q̂ q̂ 0 .
The meaning of absolute continuity is obvious7 . If q̂ q̂ 0 then

the value of q̂ at an arbitrary part tends to zero whenever the
value q̂ 0 tends to zero. Additive variables play important role for
systems whose parts are modeled by some sets. Such modeling
(studied in Chapter 4) allows then to define various measures
on parts and additive variables are good candidates for such
“measures”. The concept of absolute continuity is then very
important when comparing various measures mutually.
7
The concept of absolute continuity originates in the measure theory
where it is defined as a comparison of different measures, e.g. [36], see
Chapter 4.
3.4 Process variables

There is no problem of assigning a variable to a process. Let
π be a process. For a concrete parameterization πτ1 ,τ2 , every
variable q̂ and every part p ∈ Dq,s where s ∈ π, we may define
q(τ ) ≡ qs(τ ) , (3.11)
where s(τ ) = πτ1 ,τ2 (τ ) and τ1 ≤ τ ≤ τ2 . q(τ ) is thus a mapping
q(τ ) : hτ1 , τ2 i → Bq . (3.12)
It models an “evolution” of the local variable q̂ in the parameter
τ which means nothing but values of the variable q̂ on states
from the set π ⊂ Σ.
On the other hand, there are also variables that cannot be
assigned to individual states alone but only to certain processes.
A typical example is the heat. The (exchanged) heat is defined
only in a processes and not at individual states (we have no
‘content of heat’ in the body). Nevertheless, there is a possibil-
ity of defining slightly modified state variables that correspond
in a way to such variables. The idea is as follows. For a state s
and any process π that includes s as its inner state, there is the
tangent Π(s, π). It may be understood as the process π in an
“infinite vicinity” of s. If we assign variable to the tangent we
mean that it depends only on what “happens” (“during” the
process π) at s and no at other states.
The variable q̂ P is called the process variable if it defines for
a state s ∈ Σ and a tangent Π at s the mapping
qs,Π : Dq,s,Π → Bq . (3.13)
The domain Dq,s,Π is, similarly as for ordinary (local) variables,
a part of P. The concept of process variable is more general
3.5. STATE SPACE 73
then that of (local) variable. Namely, any (local) variable may

be understood as a process variable whose domain Dq,s,Π , as
well as the mapping qs,Π , do not depend on Π.
3.5 State space

The state space of a part p, denoted by Γ(p), is defined as
Γ(p) = {[qs (p)]Q , s ∈ Σ} , (3.14)
where [qs (p)]Q is an ordered set (e.g. a vector) of values qs (p) of

all variables8 from Q so that p ∈ Dq,s . The definition is obvious
– we simply form a set of all values of all variables which may
be assigned to a part of the system in all states.
Values of any variable on the part p then may be under-
stood as values assigned to points of the state space Γ(p) by
the following prescription,
q(Xp ) = qs (p), (3.15)
where Xp = [qs (p)]Q and (3.15) thus means a choice of corre-

sponding “coordinates” of the “vector” [qs (p)]Q .
Let the structure of the system be finitely atomic. Then
the state space of the system Γ is defined as
Γ = Γ(p1 ) × . . . × Γ(pN ), (3.16)
where {p1 , . . . , pN } = Patom . Every point X ∈ Γ thus repre-

sents values of all variables on all atomic parts of the system.
8
Values of process variables may be included into the definition of state
space, too. But only if there is a unique process – for example velocities
defined by the time evolution.
If the structure is atomic but not finitely atomic, then there

is a straightforward possibility of generalizing the concept of
state space by replacing N -tuples of values from state spaces
of atomic parts by a mapping assigning to individual atomic
parts these values. Let us suppose that the state space of all
atomic parts (3.14) is the same, i.e. there is a given set Γ0 so
that for every p ∈ Patom
Γ(p) = Γ0 . (3.17)
Then the state space Γ may be introduced as a set of mappings,
Γ so that any its element f ,
f : Patom → Γ0 , (3.18)
represents values of all variables on all atomic parts.
In a standard physical description, the state space is a ba-
sic concept that plays the role of our state set. Namely, if we
assume that every couple of different states, s1 , s2 ∈ Σ, is rep-
resented in Γ by a couple of different points, X1 , X2 ∈ Γ, then
the state space may play the role of the state set. This as-
sumption, however, depends on the choice of a sufficiently rich
set of variables (points of state space are defined by values of
variables). It motivates us to the following definition:
We say that the choice of the set of local variables, Q, on the
system with atomic structure is complete with respect to a state
set Σ if every couple of different states from Σ is represented
by different points of the state space Γ.
The completeness of the set of local variables is a very impor-
tant assumption. It allows us to characterize an abstract state
set by a mathematical structure (for example the set RkN where
k is the number of components of the “vector” [qs (p)]Q or a spe-
cial class of mappings). That is why we always want to have a
3.5. STATE SPACE 75
complete set of variables. Its determination, however, may be

a very complicated problem. It means to find a sufficient set of
variables that characterize completely all states of the system.
The search for such a set is a full-valued physical problem.
Why is the state space defined only for atomic structures?
Namely, these structures have a clearly defined set of some “ba-
sic” parts so that any part of the structure is a composition of
these parts. Hence, if we have information about these parts
(values of all variables) then we believe that we have a complete
information about the whole system. This belief is substantial
and have the essential influence on the whole physics. In our
approach, however, there cannot be excluded cases when some
important information about the state of the system “emerges”
on some larger parts only. That is why we introduce an impor-
tant rule that excludes such a situation. It will be studied in
the next section.
But what happens if the structure is not atomic? Then
there are infinite “chains” of subparts of subparts (3.2) and no
clearly defined “basic” parts exist. The state space may be in-
troduced only for concrete parts but not for a whole system.
That is why we search for a more sophisticated description in
which an atomic structure is introduced. Since there are no
some “smallest” subsystems, we construct atomic parts artifi-
cially, i.e. atomic parts are included in Pch . This procedure is
deeply related to the concept of space, where these additional
atomic parts are spatial points. We will study it in detail in
Chapter 5. Another possibility of constructing an artificial set
of atomic parts makes use of a representation of parts of the
structure on a system of sets. This procedure, however, may
not work perfectly because a given set of variables may not be
complete. It is studied in Chapter 4.
3.6 The continuity rule

For the time being, we have not embodied in our abstract
schema the fact that the description of the system on a finer
set of parts means a better information about the state of the
system. The information about any state of the system is given
only in a form of the set of values of all variables on individual
parts. If we have two disjunct parts, p1 , p2 ∈ P, and its com-
position p3 = (p1 , p2 ), we have at a state s values of a variable
q̂ on those parts, qs (p1 ), qs (p2 ) and qs (p3 ) (if they are defined).
It is obvious that information about the composition (p1 , p2 )
should be better if we have information about its subparts, p1
and p2 , individually. It is expressed also by the fact that the
set of numbers “{qs (p1 ), qs (p2 )}” is larger than “{qs (p3 )}”. It
remains valid also if the value qs (p3 ) is not defined.
But the problem is if the value of q̂ is defined for the com-
position p3 , but it is not for its subparts p1 and p2 . Therefore
we will always suppose that the system of variables and their
assigning to individual parts is chosen to be rich enough so that
the following statement is always guaranteed, namely
∀q̂ ∀s : (p1 , p2 ) ∈ Dq,s ⇒ p1 ∈ Dq,s ∨ p2 ∈ Dq,s . (3.19)
It ensures that the composition of parts is described by the

same or less number of values than there is the total number
of values describing its individual parts. Moreover, it means
that some ‘new’ variable cannot immediately arise if we form a
composition of two parts and hence it guarantees a continuity
of the description when forming compositions of parts. We call
the statement (3.19) the continuity rule.
Remark: To fulfill this rule, we have to choose suitable variables. For
example, let us have the structure of mass particles. If we introduce
3.6. THE CONTINUITY RULE 77
the interaction energy between two particles as a variable, the rule

is not fulfilled because this variable is not defined on single parts
(individual particles). That is why we have to choose, instead, the
energy (including both kinetic, potential and interaction energy) as
the suitable variable.
This rule, however, seems to contradict the physical de-

scription of macroscopic systems. Imagine, for instance, an
atomic structure whose atomic parts are individual molecules
of a gas in a box. The temperature cannot be assigned to indi-
vidual molecules, but it emerges as a suitable variable for some
‘macroscopic’ parts, i.e. compositions of atomic parts includ-
ing, say, 1010 or more molecules. Nevertheless, the rule remains
valid because there is really no division of this part onto two
subparts, so that the temperature is not defined for the both
parts (at least one part has to be macroscopic!)9 .
The fact that there is a variable defined for huge composi-
tions of atomic parts but not for small ones, is not a contra-
diction with the rule alone, but with our thinking that uses a
form of induction principle. Namely, we think as follows. The
temperature is not defined for any atomic part (molecule). It
implies (due to the continuity rule) that it cannot be defined for
the composition of two atomic parts. Now, if the temperature
is not defined for a composition of n atomic parts then (due to
(3.19) again) it cannot be defined for its composition with an
arbitrary atomic part (to form the composition of n + 1 atomic
9
If we have two parts with defined temperature then their composition
may not have defined the temperature too (e.g. if the parts are not in
mutual thermal equilibrium). But the situation, when the temperature of
a part is defined while that of its macroscopic subpart not, contradicts the
concept of temperature. See the definition of generalized temperature in
Section 2.6.
parts). As a result, the temperature cannot be defined for any

composition.
The problem with the induction consists in the fact that it
is obvious for small collections of objects and we automatically
uses it for arbitrarily large collections. This, however, is not so
obvious and may lead to paradoxical results (as the problem
with the interpretation of the continuity rule for large collec-
tions of objects).
Remark: The induction principle is formalized by the axioms of Peano
arithmetic. It is embodied in the induction axiom,
[φ(1) ∧ ∀n : φ(n) ⇒ φ(n0 )] ⇒ ∀n : φ(n),
where φ(n) is a property and n0 is a (uniquely defined) successor of n.
The quantifier ∀n means all states which can be reached by a stepwise
procedure n → n0 . However, the pure Peano arithmetic (PA) cannot
tell us how far can we get by this procedure! For example, by using
the induction axiom only in the framework of PA, we cannot formally
deduce that the stepwise procedure exceeds a very large number, say,
1020 .
We need a model of PA. In several last decades, the results in logic
and the set theory have shown that there is not only the standard
model of PA, picturing the set of familiar natural numbers N, but
also some nontrivial nonstandard models. All nonstandard models
have the structure schematically figured as the line fuzzily divided
into the parts [44] (see Fig. 3.3) The whole line models the structure
of PA, but it does also the shorter line which is formed by standard
natural numbers N. Thus this (essentially shorter!) part itself has to
go to infinity. Consequently, there is something beyond the infinity!
In the well-known Robinson’s nonstandard arithmetic [34], the part
IN consists of some infinite natural numbers obeying all arithmetic
laws but being larger than any natural number.
There is a very nontrivial interpretation of the nonstandard model
of PA that is relevant to our problem with the continuity rule. It has
been found out by P. Vopěnka [45] who realized that our ability to
see or grasp enormous collections of objects as a whole (for example
3.6. THE CONTINUITY RULE 79
Figure 3.3: Structure of a nonstandard model – if we “look” at any

piece of the part IN, we “see” a linear succession of numbers without
any beginning
the table consisting of a huge numbers of atoms) does not automat-

ically mean that we are really able to go through these objects by a
stepwise procedure. Otherwise speaking, he sensed that the shorter
Figure 3.4: Standard model (S), Robinson’s non-standard model

(Rob) and the Vopěnka’s model (Vop) whose “standard” part (FN)
is essentially smaller than the standard natural numbers.
part of the nonstandard model, denoted him by FN, may not reach to
an absolute infinity but only to a horizon fuzzily limiting our ability
to continue each stepwise procedure. Because the way to the horizon
never finishes (we ever may do an other step) he called this phe-
nomenon the natural infinity. Thus in the Vopěnka’s approach, large
numbers, such as 1020 , may be understood to be infinite and not ac-

cessible by any step-by-step process and, of course, by an induction
proof! The comparison among the standard model and Robinson’s
and Vopěnka’s nonstandard ones are illustrated in Fig. 3.4.
Figure 3.5: A semiset as a fuzzily defined part of a (large) set. It

may be defined, for instance, as a group of leaves on a given tree
that are not green. If the set represents a macroscopic collection
of molecules then the semiset may be interpreted as the “largest”
collection of molecules that does not form a macroscopic sample.
While Robinson’s model is created by a very complicated proce-

dure in the framework of the standard set theory, Vopěnka embodied
the idea of natural infinity directly to a new set theory called the al-
ternative set theory where the nonstandardness is implemented from
the beginning by using the concept of semisets [37, 45]. The semisets
are parts of sets which are not sets alone. They model the horizon
included in large collection of objects (they correspond to the part
FN). This approach thus allows us to formalize such concepts as a
“macroscopic collection of molecules”. Namely, any macroscopic set
of molecules includes a semiset. The horizon included within semisets
3.7. AVERAGING AND SCALE 81
then prevents the use of the induction principe to “carry” a property

(e.g. that the temperature cannot be defined) of small collections to
the whole set including a semiset. See Fig. 3.5.
3.7 Averaging and scale

Let us have a set of n mutually disjunct parts, p1 , . . . , pn . Each
part has its own state space, Γ(pi ), i = 1, . . . , n. Denoting their
composition as p, i.e.
p = (p1 , . . . , pn ),
we can construct the state space of this part, Γ(p), as given in

(3.14). On the other hand, due to the continuity rule (3.19),
the information about states of all parts pi gives a better de-
scription of p than a set of values of variables assigned to p, i.e.
a point in Γ(p).
It motivates us to define the full state space of p with
respect to p1 , . . . , pn as done when defining the state space of
an atomic structure (see (3.16):
Γf ull (p1 , . . . , pn ) = Γ(p1 ) × . . . × Γ(pn ). (3.20)
The averaging then may be defined as a passage from the full

state space to the state space Γ(p), i.e. as a mapping
A : Γf ull (p1 , . . . , pn ) → Γ(p). (3.21)
The averaging thus means a “loss of information” due to a

change of description of individual parts to their composition.
It is the crucial concept in any physical theory of large (e.g.
macroscopic) systems. Let us employ the concept of ‘scale’
Figure 3.6: The system looks differently at various scales. When

zooming out, various details vanish and some new features may
emerge. It corresponds to the description of the structure by more
roughly defined parts – compositions of parts defined at smaller
scales. Individual l-spaces are formed by parts defined at the same
scale.
in our abstract description10 . To do that we introduce the
10
The scale is understood as a characteristic (or smallest) length taken
into account in a physical model. In the last century, this concept started
to play a very important role due to various reasons: the presence of diver-
gences in quantum field theories (the ‘cutoff’ removing these divergences
corresponds to this length – see [19, 33]), the impossibility of managing
microscopic models of systems with many degrees of freedom (e.g. large
fluctuations in critical phenomena have been described by employing the
concept of scale in the works of K.G. Wilson and others [47]), the so-called
”size-scaled” effects in macroscopic physics (e.g. the failure of materials
mapping
Sc : R+ → Pow(P), (3.22)
where Pow(P) is a set of all subsets of P. The subset P(l) ≡
Sc(l) is called the l-space if
(i) l 6= l0 ⇔ P(l) ∩ P(l0 ) = ∅ ,
(ii) if p1 , . . . , pn ∈ P(l) then there is a scale l0 > l depending

on n so that (p1 , . . . , pn ) ∈ P(l0 )
(see Fig. 3.6). The l-space may be empty because the empty
set belongs to Pow(P), too. Moreover, if the system has the
total part p0 and it belongs into P(l0 ) then for any l > l0 the
subset P(l) = ∅. Let us present the defined concepts on two
simple examples.
Structure of mass points. The set of parts consists of three
parts, P = {p1 , p2 , p3 }, where p3 = (p1 , p2 ) and Comp =
{[p1 , p2 ], [p2 , p1 ]}. We suppose that there are two vector vari-
ables assigned to every part at any state, namely [qi , pi ], i =
1, 2, 3 where the first is a position and the second a momen-
tum. This structure represents two mass points with constant
masses, m1 and m2 . If such description of a real structure is
realistic or not, depends on the set of studied states, Σ, which
represents a physical “background” of the problem.
and structures [4]) and many others. Nevertheless, the scale is usually un-
derstood as only helpful, “artificial” concept enabling us to construct rea-
sonable models describing nature at chosen levels and formulate relations
between these descriptions. An exception is the approach of L. Nottale
[30] who develops the concept of scale relativity being motivated by the
Einstein’s theory of relativity. Another example of grasping the scale as a
more fundamental concept is presented by I.M. Havel [11] who introduces
the concept of ‘scale axis’ (see Fig. 3.6, [12]).
A full description of this system (the maximal information

about the structure in this description) is given by the double
vector [q1 , p1 , q2 , p2 ]. The demand that composition of two
parts (i.e. the third part) is described by the same set of vari-
ables, [q3 , p3 ], means an averaging defined on the structure.
Namely, the step
[q1 , p1 , q2 , p2 ] → [q3 , p3 ] (3.23)
expresses the essence of averaging. The state set Σ is supposed

to include all positions and momenta of the parts p1 and p2 so
that the state space of each mass point is defined as
Γ1 = Γ2 = {[qi , pi ]} ⊂ R6 (3.24)
and
Γf ull = Γ1 × Γ2 . (3.25)
The averaging (3.23) thus can be expressed as
Γf ull → Γ3 , (3.26)
where Γ3 is the state space of the part p3 . We may suppose

that Γ3 = Γ1 = Γ2 , too. The state spaces Γf ull and Γ3 describe
the same system but at different scales because we may define
two l-spaces: P(l1 ) = {p1 , p2 } and P(l2 ) = {p3 } where l1 < l2 .
The concrete form of the averaged value p3 is simple to be
determined, namely
p3 = p1 + p2 . (3.27)
The reason is that the momentum is an additive quantity that

describes the global behavior of any mechanical system. The
averaged position may be, in principle, defined by various man-

ners, e.g. q3 = q1 + q2 , but it would be a rather awkward
definition. The suitable one defines the averaged position as a
position of the center of gravity of the system, i.e.
m1 q1 + m2 q2
q3 = . (3.28)
m1 + m2
Atomic structure of geometric cells.TWe choose an atomic S

structure with P = {Ω1 , Ω2 , Ω} where Ω1 Ω2 = ∅ and Ω1 Ω2 =
Ω. We postulate that Ω = (Ω1 , Ω2 ). The state set Σ is chosen
in such way that each of the parts Ω1 and Ω2 alone is in thermal
equilibrium at every state from Σ. We have the only variable
- the thermodynamic temperature T̂ whose values are positive
real numbers. The state space of the first and second part is
thus
Γ1 = Γ2 = Γ = R+ . (3.29)
The temperature T̂ of the composed part Ω is not defined if the

parts are not in a mutual thermal equilibrium. The definition
of local variables and their values may be done as follows:
1. We simply “ban” the definition of the temperature on

the part Ω = (Ω1 , Ω2 ) whenever parts are not mutually
in thermal equilibrium. The mapping Ts is then defined
on the part Ω only if s ∈ Σeq where Σeq ⊂ Σ are states
in which the temperatures of the parts Ω1 and Ω2 are the
same. Then
Ts (Ω) = Ts (Ω1 ) = Ts (Ω2 ). (3.30)
If s 6∈ Σeq then DT,s = {Ω1 , Ω2 }.

2. We change the set of variables to a more suitable one.

For example, we use the fact that the inner energy of
the system in thermal equilibrium, Es , is the function of
the temperature, Es = E(Ts ) (if no other variables are
taken into account). The fact that the inner energy is
an additive variable allows us to define its value on Ω
in any state. Namely, Es (Ω) = Es (Ω1 ) + Es (Ω2 ) (see
Section 6.5).
Let us discuss another possibility consisting in defining the tem-

perature of the part Ω at any state from Σ as an averaged
concept. That means, we suppose that Ts (Ω) is defined by an
averaging procedure, A : Γ × Γ → Γ,
Ts (Ω) = A(Ts (Ω1 ), Ts (Ω2 )), (3.31)
that projects values from the full state space onto the state
space of the part Ω. For example, by the prescription
V1 Ts (Ω1 ) + V2 Ts (Ω2 )
Ts (Ω) = , (3.32)
V1 + V2
where Vi are volumes of the parts Ωi .
The variable T̂ , however, cannot be understood as a global
variable because it does not obey the definition of generalized
temperature. Namely, if we define a global variable T by the
relation (3.8), i.e. T (s) = Ts (Ω), then T cannot be the gener-
alized temperature because its value generally differs from its
values on subsystems (if Ωi represents subsystems of Ω). Nev-
ertheless, when T̂ is understood as a local variable only then
there is no problem with understanding it as an averaged con-
cept. Any definition of such a ‘non-equilibrium temperature’,
however, needs a careful physical explanation [25, 32].
Chapter 4
Special structures
In this chapter, several structures are introduced and studied

in detail. Chosen examples illustrate two main schemas of de-
scription used in thermodynamics and continuum thermome-
chanics, but not solve concrete physical problems. Especially,
we concentrate into choice of parts, introduction of variables
and construction of the state space.
We start by the structure used in equilibrium thermody-
namics. Its importance is crucial because it is the only schema
of description for which the variable ‘entropy’ has been rigor-
ously introduced. Second example is the structure used when
describing various macroscopic processes (thermal, mechanical,
hydrodynamical or electromagnetical) in macroscopic media.
The discipline studying such processes and their combinations
is called the continuum thermomechanics. We outline its meth-
ods of description and especially the way of formulating con-
tinuum description by the use of non-atomic structures.
88 CHAPTER 4. SPECIAL STRUCTURES
4.1 Simple systems
Look back at the consideration presented in Section 2.6. We

have noticed there an important assumption allowing to inter-
pret results of many measuring procedures that are done on
other systems than those whose properties should be deter-
mined. Let us remind ourselves the experiment when a sample
of water is taken from a vessel and its physical and chemical
properties are determined. The assumption of an equilibrium
between the sample and remaining water enables us to assign
the measured data to the water what remains in the vessel.
In Section 2.6, we have formally defined what means that
the system and its surrounding are in equilibrium. But what is
the system and what its surrounding in the experiment with a
sample of water? The water in the vessel before the measure-
ment is a physical system, say S0 . When taking a sample, we
form two new physical systems, Ssample and S00 , representing
the sample and remaining water respectively. What is a rela-
tion of Ssample and S0 ? First of all, S00 is not a surrounding of
Ssample because the sample of water loses any contact with the
remaining water (no manipulation with the water in the vessel
can influence the sample afterwards it was displaced).
The water in the vessel is a surrounding of the sample of
water only so long as the sample is within the water. But what
does the ‘sample of water’ mean until it is defined by taking it
into a test tube? To determine a moment when the sample is
both defined and in equilibrium contact with surrounding wa-
ter, we assume that a closed wall separating the sample from
surrounding water suddenly appears somewhere within the wa-
ter, see Fig. 4.1. At this moment, the water outside the wall
plays the role of surrounding of the sample. Moreover, this sur-
4.1. SIMPLE SYSTEMS 89
rounding is in equilibrium with the sample in the spirit of our

definition (the state of the surrounding water defines the state
of the sample).
Figure 4.1: A sample of water is taken off to form a system defined

by walls separating it from surrounding water.
The situation may be described in terms of ‘parts’. Namely,

the existence of a subsystem Ssample of the water in the vessel
that lately forms the sample of water may be supposed before
than it is defined by the wall (e.g. as a conglomerate of cer-
tain molecules). But this subsystem has no clear definition (we
cannot write a “list” of all molecules belonging to it). We may
describe this situation as the fact that there is no part from Psub
defined for this system. That is, though the system Ssample ex-
ists, there is no part corresponding to this system (any part has
to be defined ). Only the presence of the wall defines clearly a
part – the water inside the space defined by the wall. This part
is thus defined externally and has all features of a part from
Pch . Writing more formally, P(Ssample ) = Pch = {p}, because
there is only one part defined by the wall. It motivates us to
the following definition:
The system S is called the simple unit if the following condi-

tions are valid:
(i) Its surrounding, Ssur , is in equilibrium with S,
(ii) the system of parts includes only one part, p, that belongs
to Pch , i.e.
P(S) = Pch = {p}, (4.1)
(iii) for each variable q̂ ∈ Q, there is a global variable Gq ∈

Qg (Ssur ) so that for every part p ∈ P,
−1
qs (p) = (v(p))kq Gq (feq (s)), (4.2)
where v(p) is a positive real number determined by the

part p, kq = 0 or kq = 1, and feq is the mapping defining
the equilibrium situation (see Section 2.6).
The variables for which kq = 0 are said to be intensive

and those for which kq = 1 are said to be extensive. The
global variable Gq corresponding an extensive variable q̂ is usu-
ally called the density of variable q̂.
If S0 , S are simple units defined by the same surrounding then
S0 is called the scaled copy of S.
The meaning of this concept is clear from the definition of sim-
ple unit. The both systems have the same set of variables
defined by the common surrounding. The only difference be-
tween values of variables on S and S0 is thus in the value of v
as implied by (4.2). All intensive variables thus have the same
values on the both systems. Let us denote
v(p0 )
r≡ .
v(p)
4.1. SIMPLE SYSTEMS 91
If kq = 1 then
qs (p) = rqs (p), (4.3)
i.e. values of extensive variables are r-times scaled. If Γ(p)
is the state space of a simple unit then we usually denote by
Γ(r) (p) the state space of its scaled copy. If Xp ∈ Γ we denote
the corresponding state of the scaled copy as rXp (we take over
the nomenclature from [21]).
Let us form a composition of n simple units, S1 , . . . , Sn ,
that are pairwise mutually isolated. It means that the state set
of this composition is
Σ ≡ Σ((S1 , . . . , Sn )) = Σ(S1 ) × . . . × Σ(Sn ). (4.4)
The structure of this composition may be understood as being

formed by n atomic parts, .
Patom = {p1 , . . . , pn },
and their various compositions. We suppose that there is a

generalized temperature T defined for every surrounding of in-
dividual simple units, S1 , . . . , Sn . This global variable defines
an intensive local variable T̂ on the system S = (S1 , . . . , Sn ),
called the temperature, by the prescription (4.2) with Gq ≡ T
and kT = 0. Hence it is defined at every state on an arbitrary
part pi ∈ P.
The temperature, however, may not be defined on their
compositions (see the example at the end of Section 3.7). That
is why we define a subset of Σ,
Σeq ⊂ Σ,
so that at every state s ∈ Σeq all simple units have the same
value of the temperature.
The system S = (S1 , . . . , Sn ) whose state set equals Σeq is

called the simple system.
For simple systems, the temperature has the averaged property
because it is defined at any composition, Ts ((pi , pj )) = Ts (pi ) =
Ts (pj ), compare (3.30).
The scaled copy of a simple system is then defined as a
composition of scaled copies of individual simple units. Since
parts of individual simple units are atomic, the state space of
a simple system is given by (3.16), i.e.
Γ = Γ(p1 ) × . . . × Γ(pn ). (4.5)
We denote by Γ(r) the state space of the scaled copy and rX ∈

Γ(r) a point corresponding to X ∈ Γ.
Variables of the simple system may be always chosen so
that the temperature T̂ is among them. The other variables
are called work variables because their values may be always
defined by a mechanical (or electric or magnetic) manipulation
with the surrounding. Such a kind of manipulation is usually
connected with a performed work, see Fig. 4.2. The work vari-
ables are deeply connected with the definition of parts p1 , . . . pn
of the simple system. For example, a typical work variable is
the volume that defines the size of (externally defined) parts pi
(see [21], p. 38).
Sometime, an extensive real variable Ê, called the energy1 ,
is chosen instead of the temperature (see again the end of Sec-
tion 3.7). The relation of the energy and temperature is studied
in equilibrium thermodynamics. We outline the essence of this
relation in Section 6.5.
1
The energy as a global variable is defined at the end of Chapter 5.
4.2. THERMOMECHANICAL SYSTEMS 93
Figure 4.2: When taking off a sample of gas, we have to guarantee

the same conditions as if it were in the vessel with the gas. It means
to exert a pressure on walls. Then any change of the volume V of the
sample is connected with a work, δW = pdV .
4.2 Thermomechanical systems

The discipline describing thermal, mechanical, hydrodynamical
and electromagnetical processes in macroscopic media is called
the continuum thermomechanics (see e.g. [13, 22, 23, 26,
27, 40, 43]). The important feature of the thermomechanical
way of description is a rich use of many macroscopic subsys-
tems that may be thought as parts of the system. The most
important step, however, is an introduction of parts that are
not material physical systems – boundaries of subsystems. It
has an origin in classical thermodynamics where systems are al-
ways in a container (e.g. a box with gas endowed by a piston).
The container does not play only the role of an external “lim-
iter”, there are various variables that are assigned rather to the
container than to the system alone. For example, the volume
or the heat flow. Boundaries in continuum thermomechanics
have, however, a more general meaning.
To define the structure of a thermomechanical system we

return to the thought experiment in which a simple unit is
formed by taking a sample of water from a water reservoir.
The sample has all properties defined by the surrounding wa-
ter because of the assumption of equilibrium. The only “free
parameter” is the size of the sample – it occurs in the relation
(4.2) as a “scaling parameter” v(p). This size is, however, also
defined externally by a wall forming the sample. Namely, the
surrounding is not only the surrounding water but also the wall
because it has to be present at the moment when the sample
as a physical system arises2 .
In this section, we study the situation when only this aspect
of an external definition of a sample plays the role because its
surrounding is not in equilibrium with it. The part p ∈ Pch
represents the boundary delimiting the sample but a physical
description of the sample cannot be done by assigning vari-
ables only to this boundary. Namely, there is no direct relation
between states (and properties) of the surrounding and states
(and properties) of the sample. Therefore, the description of
such a sample cannot be done by the use of only this part.
That is why we add a part σ ∈ Psub which represents the
sample as a physical system Ssample . Using (3.4), we may write
σ = Fs (Ssample ). (4.6)
The part p may be understood as a boundary of σ and therefore

it is called the oriented boundary3 and denoted by ∂σ. Thus
2
Walls are supposed to be inert so that properties of the water are not
influenced by them.
3
“Oriented” means that the boundary defines always a uniquely given
interior and exterior.
we obtain the structure
P = {σ, ∂σ}. (4.7)
The system of this structure is called the discrete system.

Remark: It may seem strange that we add the part ∂σ into the struc-
ture while σ represents the whole system (a sample). The structure,
however, is chosen to provide the best description of the system by
a set of variables assigned to its parts. The part ∂σ is crucial in de-
scription of discrete systems. In fact, it replaces partially the “lost”
equilibrium because variables assigned to this part are definable on
the surrounding.
Figure 4.3: The discrete system need not be defined by a given

piece of mass. The mass may flow through the boundary and its
value “inside” then plays the role of a variable of the system.
In what follows, we will study discrete systems that are vari-

ous macroscopic bodies described in common language by their
arrangement, phase, chemical constitution of various compo-
nents and so on. In this concrete meaning, they are also called
Schottky systems (see e.g. [23, 26]). Each discrete system
has a state set which corresponds to various states that are
taken into account in the study of the body (the body may
or may not change the phase or chemical constitution, mass,
momentum, energy, etc.), see Fig. 4.3. We also accept various
manipulations with the body that are caused by a chosen sur-
rounding of the discrete system. The set of manipulations, M,
is supposed to be rich enough so that it forms a large set of
processes.
Variables of a discrete system split into two categories. The
group of variables defined only on the part σ and the group
defined only on the part ∂σ. Variables from the first group are
called the c-variables. Typical c-variables are mass, energy,
momentum, etc. Variables from the second group are called the
b-variables (they are defined only on oriented boundaries).
Within this category, special process variables play an impor-
tant role. They describe an exchange of a physical property
with surrounding. A typical example is the heat. Another ex-
ample of an eminent importance is the stress that describes the
exchange of momentum between the system and surrounding.
This kind of b-variables will be called the process b-variables.
The great advantage of simple systems (defined in the last
section) consists in a full definability of their states externally.
Namely, the equilibrium guarantees that we always know ‘what
is inside’. It is not the case of discrete systems. Values of c-
variables are not definable externally (see Fig. 4.4). Therefore,
the crucial role plays a special relation between certain c-variables
and process b-variables, whose values are definable externally:
A c-variable q̂ is said to have the balance property if for any
process π there is such a class of parameterizations that if πτ1 ,τ2
belongs to this class then the derivative,
d
qs(τ ) (σ) ≡ q̇s(τ ) (σ), (4.8)
dτ
Figure 4.4: Where is a treasure? From an external point of view,

some properties of the system are difficult to be determined.
exists for every4 τ ∈ (τ1 , τ2 ), and there is a process b-variable

b̂q and c-variable δ̂ q so that the equality
q̇s(τ ) (σ) = bqs(τ ) (∂σ) + δs(τ

q
) (σ) (4.9)
is valid for every τ ∈ (τ1 , τ2 ).

The process b-variable b̂q is called the flux of the variable q̂.
The c-variable δ̂ q represents either an external supply of q̂ that
is not connected with processes on the boundary (e.g. a volume
force, like gravity or centrifugal force, is an external supply of
the momentum) or the production of q (see e.g. [26]). The
equality (4.9) presents in a general form the famous balance
equality.
4
This condition may be weaker by assuming only existence of the deriv-
ative almost everywhere within the interval (τ1 , τ2 ), see [40].
A typical example of a variable with balance property is the

mass m̂. It has the balance property so that
δsm (σ) = 0. (4.10)
Remark: Variables with zero supply and production (such as the

mass) have an origin in global variables having conservative proper-
ties. It is worth noticing a remarkable relation between simple and
discrete systems. Variables assigned to simple systems are defined
(measured) on an external surrounding by the global variables de-
fined via equilibrium. Several variables assigned to discrete systems
are defined (measured) on an external surrounding by the global vari-
ables defined via conservative properties of the equilibrium.
If a c-variable q̂ has the balance property and δ̂ q is always zero
or it is an externally defined supply (e.g. defined by an external
field) then any change of its value during any process is exter-
nally controllable. The problem is if a quantity has a nonzero
production.
Let us have a macroscopic body that is understood as a
physical system S. The main idea of the thermomechanical
description of the body consists in imaging it as a composition
of an uncountable number of discrete systems. The description
is thus based on a special non-atomic structure. The structure
is defined as follows: If p ∈ Psub then either there is a discrete
system, so that p = σ, or there are n disjunct discrete systems
and p = (σ1 , . . . , σn ). For every p ∈ Psub , there is exactly one
element Bd(p) ∈ Pch so that if p = σ then Bd(σ) = ∂σ. These
elements are called oriented boundaries5 . If p ∈ Pch then
5
Oriented boundaries are defined externally by the use of a suitable
discrete system (or a composition of discrete systems). This system is
somewhere “inside” the studied body but, whenever we understand it as
it is either an oriented boundary or its subpart. The thermo-

mechanical system is described by a non-atomic structure of
parts from Psub (discrete systems and their compositions) and
Pch (oriented boundaries and their subparts).
Variables of a thermomechanical system arise as a natural
enlarging of variables defined for discrete systems. The set
of c-variables remains the same, only their properties may be
characterized in more detail. For example, the fact that there
is no only one part in Psub allows us to specify if they have
averaged properties or are additive.
Concerning the b-variables, the situation is slightly differ-
ent. Namely, we have b-variables that are defined only on ori-
ented boundaries (and not on their subparts). Thanks to the
unique relation between oriented boundaries and parts from
Psub , the value of such a variable on an oriented boundary p0
can be assigned also to the corresponding part p = Bd−1 (p0 ).
A typical example of such a variable is ‘volume’.
Let us imagine, for a while, an oriented boundary ∂σ of
a discrete system as a wall of a closed container that is occu-
pied by a piece of the matter (it does not matter how “fully”,
see Fig. 4.3). The volume of the space inside the container
is determined by this boundary and not by the substance in-
side. Therefore the volume has to be understood as a b-variable
V̂ whose definition domain covers all oriented boundaries. It
points out its origin in thermodynamics – the volume is defined
by an external box and not by matter inside. On the other
a discrete system, the resting part of the body belongs to the surrounding
of this discrete system. The boundary is thus no real interface but it is
rather defined by a mental construction. It resembles the composition of
simple units – surroundings that define individual simple units are not
really present but we ever may add them mentally.
hand, the volume behaves as an additive c-variable. Namely, if

disjunct parts σ1 , σ2 have volumes V1 and V2 , respectively, then
the volume of their composition is V1 + V2 . Since the volume
is b-variable we have to express its additivity as
Vs (Bd((σ1 , σ2 ))) = Vs (Bd(σ1 )) + Vs (Bd(σ2 )). (4.11)
This kind of b-variables may be understood as c-variables

as well (the volume is usually understood as a c-variable and
not a b-variable!). Notice, however, a great difference between
b-variables that are understood as c-variables (as volume) and
“pure” c-variables (such as energy). The value of b-variable on
a concrete discrete system is determined by the situation on its
boundary. Hence its value is fully determined by an external
procedure definable on the surrounding of this discrete system.
For example, the exchanging heat or the stress on the boundary
are determined from outside. Typical c-variables, however, de-
scribe properties that cannot be determined at a state by only
analyzing a “momentary situation” of the surrounding6 .
4.3 Continuum
Parts defining the structure of a physical system are basic ob-
jects, building blocks, having no further detail specification (the
detail structure of a part is given only by its subparts, i.e. the
way by which they are composed from “smaller” ones). The
only specification is given by various variables assigning to indi-
vidual parts (notice that we do not specify the part correspond-
ing to a ‘mass point’, for instance, otherwise then by values of
6
A current energy of a body, for example, cannot be determined from
a current state of the surrounding, see Fig 4.4.
4.3. CONTINUUM 101
variables assigned to it: mass, position, momentum, etc.). The

problem is, however, if the structure is non-atomic. Then there
are no basic blocks that enable us to describe the whole system
by variables assigned to them. A typical non-atomic structure
is a set of parts forming continuum.
The continuum means that smaller and smaller parts may
be always chosen when forming the set of parts. The relation
(3.2) expresses a typical situation of the structure describing
the continuum. An example of the continuum is the struc-
ture of thermomechanical system whose parts from Psub form
continuum because there are no smallest macroscopic pieces of
matter. When modeling boundaries, it is also suitable to as-
sume that subparts of any part from ∂Psub form the continuum.
The crucial idea consists in modeling parts of a non-atomic
structure by some sets7 . These sets are formed by collections of
some abstract elements whose representative is denoted by the
symbol u. These elements, however, have no physical meaning
and serve only to form various sets of the class. Let U be
a “universe” of these elements, i.e. a set including all these
elements.
Now, let us model parts from P as sets formed by elements
of U. For every state s, let us choose a set of subsets of U, i.e.
the set
Xs ⊂ Pow(U),
(Pow(U) is the set of all subsets of U) and postulate the exis-
tence of a one-to-one mapping,
Ms : P → Xs ,
that fulfills the following rules:
7
Our approach thus differs from majority of others defining the contin-
uum directly as a set of subsets of a set, see e.g. [10].
sp
(i) p0 ⊂ p ⇔ Ms (p0 ) ⊂ Ms (p) ,
(ii) [p, p0 ] ∈ Dcomp implies Ms (p) ∩ Ms (p0 ) = ∅.

The set Xs = Ms (p) is said to be the model of p at s and
{Ms , s ∈ Σ} the model of P. We suppose that the structure
includes the total part p0 , and denote
Xs0 ≡ Ms (p0 ).
If we model a non-atomic structure then any one-element subset

of U cannot be model of a part (the property (i) implies that
the part M−1
s ({u}) then would be atomic). Therefore we claim
∀s ∈ Σ, ∀u ∈ U : {u} 6∈ Xs . (4.12)
The model of a structure describing continuum needs an addi-

tional assumption. Namely, there is no minimal size of pieces
of a continuum structure – every piece may be always divided
into smaller ones. It implies that if different elements u1 , u2 be-
long to a set modeling a part then there is another part whose
model includes u1 but not u2 , i.e. it separates the element u1
from u2 . That is, we assume that for every s and every couple
of different elements, u1 , u2 ∈ Xs0 ,
∃p ∈ P : u1 ∈ Ms (p) ∧ u2 6∈ Ms (p). (4.13)
Elements of U thus play the role of an auxiliary scaffold, by

which the continuum description is build up, but whose con-
crete elements have no physical meaning. The model of P
for which (4.12) and (4.13) are valid is called the continuum
model of the non-atomic structure P.
Elements of U alone have no physical meaning (as models
of some parts) but the condition (4.13) strongly restricts an
4.3. CONTINUUM 103
arbitrariness in their assigning to parts at various states. We

describe it by introducing the concept of configuration. Let Υ
be an index set so that it may be mapped on the whole U.
We define a set of mappings, C, that includes for every state s
the mapping ηs ,
ηs : Υ → U,
so that for every ξ ∈ Υ and every p ∈ P the condition
ηs (ξ) ∈ Ms (p) ⇒ ηs0 (ξ) ∈ Ms0 (p) (4.14)
holds for any s, s0 ∈ Σ. Mappings from C are called configu-
rations.
The definition of configurations is sketched at Fig. 4.5.
Figure 4.5: Parts are modeled as subsets of a set U. A configura-

tion means a mapping from an index set Υ into U, which keeps the
“identity” of parts at individual states.
Obviously, we may choose an infinite amount of different

set of configurations, C, for one continuum model. A nontrivial
question is, however, how big is this “liberty” if we fix the con-
figuration for one state, say s0 . Namely, other configurations
have to be related to ηs0 by the relation (4.14). The answer is
surprisingly simple:
Lemma 1: If the mapping Ms is a continuum model then the
choice of the configuration for one state from Σ defines uniquely
a configuration for any state from s.
Proof : The state for which the configuration is fixed is denoted by
s0 . Let there be two different sets C1 and C2 . It implies that there is
a state s ∈ Σ so that there are two different configurations, ηs1 ∈ C1
and ηs2 ∈ C2 . It means that there is ξ ∈ Υ so that ηs1 (ξ) 6= ηs2 (ξ).
Hence we have two different elements of U, namely u1 ≡ ηs1 (ξ) and
u2 ≡ ηs2 (ξ). u1 , u2 ∈ Xs0 and thus (4.12) implies that there is p ∈ P
so that u1 ∈ Ms (p) and u2 6∈ Ms (p). Next, we continue by following
consideration. First, we use (4.14) for C1 by identifying s0 → s0 and
obtain that ηs0 (ξ) ∈ Ms0 (p). Second, we use (4.14) for C2 but in
opposite direction, i.e. by identifying s → s0 and s0 → s. Using the
fact that at s0 C1 merges with C2 , we get that u2 ≡ ηs (ξ) ∈ Ms (p)
which is a contradiction.
The set of configurations is thus fixed by a choice of the config-
uration at one state. Moreover, a choice of this configuration
as a one-to-one mapping (it may be always done by a suitable
choice of the index set) guarantees that all configurations are
one-to-one mappings:
Lemma 2: If the mapping Ms realizes a continuum model and
the configuration ηs0 at a state s0 is a one-to-one mapping then
the configuration at every state is a one-to-one mapping.
Proof : Let us suppose that there is a state s, so that there are
two different elements ξ1 , ξ2 ∈ Υ for which ηs (ξ1 ) = ηs (ξ2 ). At s0 ,
however, ηs0 (ξ1 ) 6= ηs0 (ξ2 ) that implies that there are two different
4.3. CONTINUUM 105
elements u1 = ηs (ξ1 ) and u2 = ηs (ξ2 ). Consequently, there is a part

p ∈ P whose model at s0 includes u1 but does not include u2 . The
same consideration as in the previous proof shows that it contradicts
to the definition (4.14).
The couple [Υ, ηs0 ] where ηs0 is a one-to-one mapping is called
the referential configuration.
In what follows, let us suppose that the referential configura-
tion is firmly chosen. For every part p ∈ P, we may define a
subset Z(p) of Υ by the prescription
Z(p) ≡ {ξ ∈ Υ, ηs0 (ξ) ∈ Ms0 (p)}. (4.15)
If we have a continuum model then every state defines a uniquely
given configuration,
s → ηs ∈ C.
It does not mean, however, that the set C fully characterizes
the state set Σ because there may be different states from Σ
that are represented by the same configuration. Nevertheless,
we may define a class of variables for which the specification of
a configuration is sufficient.
We say that a variable q̂ is configuration dependent if the
value of this variable at any state s and any part p ∈ Dq,s
depends only on the configuration ηs on Z(p), i.e.

qs (p) = F ηs (ξ)|ξ∈Z(p) . (4.16)
If only configuration dependent variables are used in the de-
scription of a thermomechanical system, we speak about con-
tinuum mechanics.
Let π be a process realized by a set of manipulations with
the studied thermomechanical system. We choose a parame-
terization πτ1 ,τ2 and suppose that we have a continuum model
of the studied system. Then the process in the parameteri-

zation πτ1 ,τ2 may be understood as a continuous succession of
configurations ηs(τ ) . The mapping
χ : < τ1 , τ2 >→ C,
defined by the prescription
χ(τ ) = ηs(τ ) , (4.17)
is called the motion. The set of elements from U that have in

the process the same index ξ is called the trajectory during
the process. It is worth reminding that both the concept of
motion and trajectory are defined on an abstract set U, and no
assumption about a location of the system in space is necessary.
4.4 State space of a continuum model

The concept of state space of the whole system uses essentially
the occurrence of some “smallest” basic parts for which values
of a complete set of variables are given at every state. These
parts are clearly defined if the structure is atomic. If it is non-
atomic, there are no such parts. Nevertheless, when forming
continuum models we define the structure by means of elements
of a set U. Although these elements are not models of parts
from P, their large collections form any part of P and the choice
of a referential configuration determines uniquely elements of
U that have the same “label” ξ at various states, namely
us (ξ) = ηs (ξ). (4.18)
May the set Ω0 ≡ Z(p0 ) ⊂ Υ, that includes all “labels”, be

identified with a set of some atomic parts? The problem is
4.4. STATE SPACE OF A CONTINUUM MODEL 107
that variables are defined only on parts from P and thus they
are not defined on individual elements of U.
In continuum mechanics, where only configuration depen-
dent variables are used, we may enlarge the set of parts P to
P 0 = P ∪ Patom , with
Patom = Ω0 , (4.19)
and define a special variable û by the prescription (4.18)8 . Since

the value of every variable is derivable from values of ηs (ξ) by
(4.16), we may choose
Q = {û} (4.20)
that forms a complete set of local variables. The state space
(3.18) is then identical with the set of all configurations, namely
Γ = C. (4.21)
If there is a variable that is not configuration dependent (e.g.

inner energy if thermal phenomena are taken into account), the
situation looks hopelessly. Namely, values of such a variable
cannot be assigned to atomic parts (4.19) and the knowledge
of a concrete configuration cannot determine values of this vari-
able for individual parts from P.
Nevertheless, if this variable is additive, there is a possibility
of defining its values for a suitable amount of parts from Patom ,
so that this information is sufficient for determining the value
of the variable for every part from P. To show it, we have to
introduce firstly some important mathematical concepts.
The so called σ-algebra A is a set of subsets of a given set X 0
so that
8
“Values” of this variable are elements of an abstract set. But U usually
forms a Banach space – the Euclidean space Ed .
(i) X ∈ A implies that X 0 \ X ∈ A,

0
S (infinite or not) {Xk } of subsets of X ,
(ii) given a sequence
Xk ∈ A, then Xk ∈ A,
(iii) X 0 ∈ A.
If f is a mapping from X 0 into a topological space T then we

say that f is measurable if for every open set Z ⊂ T holds
f −1 (Z) ∈ A.
Any complex-valued function µ defined on a σ-algebra A is a
(complex) measure if
(i) µ(∅) = 0 .
S P
(ii) µ ( Xk ) = µ(Xk ) whenever {Xk } is a sequence (infi-
nite or not) of pairwise disjoint subsets of A, i.e. Xi ∩
Xj = ∅ for i 6= j.
We say that a function is defined almost everywhere on X 0

if for its definition domain X ⊂ X 0 holds that µ(X 0 \ X) = 0.
These definitions allow us to define the concept of an abstract
integral of measurable functions on any σ-algebra. If µ is a non-
negative measure on a σ-algebra A, X ∈ A and f is a complex-
valued measurable function, we denote by
Z
f dµ
X
the integral of the function f over the set X (a precise definition

of the integral is not quite simple and is not presented here,
see e.g. [36]). The integral is a (finite or infinite) complex
number that is defined uniquely by the function f , the set X
and the measure µ. We denote by L1 (µ) the set of all complex

measurable functions on X 0 for which
Z
|f | dµ < ∞. (4.22)
X0
Let us go back to the study of a non-atomic structure rep-

resented by a continuum model. We suppose that there is a
variable V̂ whose values are positive real numbers and its de-
finition domain covers the whole set P at every state. If P is
a set Psub of a thermomechanical system then the volume is a
good candidate for such a variable.
Remark: The set P is not understood here as representing the whole
structure. When studying a thermomechanical system, for example,
it may represent the set Psub or a set of all subparts of an oriented
boundary ∂σ (including ∂σ itself). Each such a substructure then
may have different variables V̂ (e.g. the volume for elements of Psub
and the surface for ∂σ and its subparts).
Let us define for every s ∈ Σ a σ-algebra AXs on Xs0 so that
Xs ⊂ AXs , (4.23)
and choose a nonnegative measure µV on AXs that coincides

with V̂ in the following sense. If Xs ∈ Xs then
µV (Xs ) = Vs (M−1
s (Xs )). (4.24)
The measure µV is thus fixed only on sets that are models of

parts. Consequently, measures of these sets are nonzero (be-
cause Vs is always a positive number). What about the other
sets in AXs , i.e. sets that are not models of parts? May some
of them have a nonzero measure, too? The important question
is how wide is the group of such sets because the larger it is the
higher is the arbitrariness in modeling (realize that the value of

µV (X) may be chosen arbitrary if X belongs into this group).
Therefore we define:
We say that the model of P is scale complete if the σ-algebra
AXs may be chosen in such way that the condition,
X ∈ AXs \ Xs ⇒ µV (X) = 0, (4.25)
is valid for every state s.
Remark: The expression “scale complete” is chosen from the follow-
ing reason. If AXs \ Xs includes some sets with a nonzero measure µV
it may be interpreted as the absence of parts at certain size scales.
Notice, that the scale completeness may not be defined only for con-
tinuum models. It is not clear, however, if an arbitrary continuum
model may be supposed to be scale complete.
Let q̂ be a variable whose values are complex numbers9 and
q̂ V̂ (4.26)
(i.e. q̂ is absolutely continuous with respect to V̂). Let us define
a complex measure λqs on AXs corresponding to this variable by
the prescription: If Xs ∈ Xs then
λqs (Xs ) = qs (M−1
s (Xs )). (4.27)
If X ∈ AXs \ Xs , we define its value arbitrary but in such
a way that λqs is absolutely continuous10 with respect to µV .
The Radon-Nikodym theorem [36] then says that there exists
a function φqs ∈ L1 (µV ),
φqs : Xs0 → Bq , (4.28)
9
This variable may be a component of a vector or tensor variable.
10
The absolute continuity of one measure with respect to another is
defined in the same way as for variables [36].
so that for every state s and every X ∈ AXs

Z
q
λs (X) = φqs (u)dµV . (4.29)
X
Especially, for sets from Xs (i.e. sets that are models of parts),
we can write Z
qs (p) = φqs (u)dµV . (4.30)
Ms (p)
The function φqs (u) is uniquely determined by the measure λqs

almost everywhere on Xs0 . That means that if we have two such
functions then they may differ only on such a set X 0 ⊂ Xs0
whose measure µV (X 0 ) = 0. Such sets, however, cannot be
models of parts from P because any part from P has a nonzero
value of the variable V̂.
Nevertheless, the arbitrariness in the function φqs may be es-
sentially higher if the model of the structure is not scale com-
plete. Then there are sets in AXs \ Xs whose measure µV is
nonzero. On these sets, the measure λqs (which represents the
variable q̂) may be chosen arbitrarily, which implies an addi-
tional arbitrariness in the choice of φqs (u) at a set .
Whatever the case, the set {φqs s ∈ Σ} determines values of
the variable q̂ on all parts from P at arbitrary states. It de-
pends on u but when using (4.18), we may construct a function
φ̃qs that depends on elements of the index set Υ, i.e.
φ̃qs (ξ) ≡ φqs (us (ξ)). (4.31)
Now, we may understand φ̃qs (ξ) as values of a variable on atomic

parts from Patom = Ω0 . This variable may not be defined on
all atomic parts (since φqs also may not be defined everywhere)
but it is defined on a sufficient amount of these parts so that
the value of variable q̂ is determined for any part at every state.

When denoting
Lq ≡ {ϕqs , s ∈ Σ},
we may enlarge the state space (4.21) into the state space
Γ0 = C ∪ Lq1 ∪ Lq2 ∪ . . . , (4.32)
where Lqi are additive variables that are absolutely continuous

with respect to V̂.
However, if configuration dependent variables and additive
variables do not form a complete set of variables then the state
space Γ0 does not represent fully the state set. It may happen
when there is some essential variable, that is neither configu-
ration dependent nor additive, and that cannot be replaced by
some other configuration dependent or additive variables.
Chapter 5
Space
Isn’t space the machine that causes all those size-related miracles?
V. Šedivý1
What is space? There are two main views about the nature
of space in western thinking. Space is understood either as an
entity of its own or as a relation between bodies ([35], p. 38).
Though the second viewpoint was dominant in western culture
(from Aristotle to Descartes, see [18]), physics, eventually, an-
chored its foundations in the idea of independent space which
exists by itself. It was done by I. Newton in his “Principia”,
[29]. This construction has secured physical thinking from an
extreme relativity yielding by the idea of space which is just
a relation between bodies (more in Chapter 6). The Newton’s
conception of absolute space has shown itself as extraordinary
successful. But what is space in the conceptual framework of
our approach?
1
[38], p. 222.
114 CHAPTER 5. SPACE
In this chapter, let us try to study space in the Newtonian

spirit, i.e., as if it were a special entity. However, we have intro-
duced no other entities than ‘physical systems’. If space is an
entity then it has to be a special physical system. Newtonian
space is a ‘container’ holding material objects and having an
independent, absolute existence. Thus it may be understood as
an absolutely indifferent surrounding. Therefore, we introduce
a special class of systems that are studied only as surroundings.
We postulate properties of these systems to model Newton’s
space as well as possible. Especially the way in which New-
tonian space saved physics from devastating relativity, plays
eminent role in our “modeling”. Namely, space yields absolute
places 2 – points of space. In our scheme, spatial points are un-
derstood as special parts of the description of a physical system.
They are defined by its surrounding (space) in the spirit of the
definition in Section 3.2. To describe points concretely, we use a
suitable set of auxiliary physical systems called the coordinate
systems. These systems also enable us to define symmetries of
space as a physical indiscernibleness of descriptions in different
coordinate systems.
The most important point is the relation of the system mod-
eling space and an (arbitrary) system which is in this space.
Namely, the first is the surroundings of the latter, and we know
that surrounding is a source of all manipulations with the sys-
2
Z. Neubauer gives an extraordinary deep contemplation about the role
of ‘place’ (in a general sense) in the whole western thinking. He shows, for
example, that the alphabet, as alphabetical ordering, is an order that allows
constitution of a place as such. He emphasizes that “Aristotle claimed that
if there truly were something like that (a place as such), that is if it existed
in itself, it would be something so amazing, frightening and powerful, with
so sovereign power that it would precede everything and be superior to
everything - it would be omnipotent”, [28].
115
tem. What does it mean? Any manipulation is a set of changes.

Any change of the system, which is in space only, is a manifes-
tation of its inner dynamics. Hence, in our scheme, the inner
dynamics of the system is caused by space. Though it may
look rather strange, it only reflects the general line of thinking
motivated by thermodynamics that never understands bodies
as being “alone”. A ‘communication’ between the body and its
surroundings is always present, whatever its meaning.
The important property of Newton’s space is its immobility.
It is modeled in our approach by the surrounding, called the
Newton space, that has only one state. It implies that it may
cause only one manipulation. The inner dynamics, however,
includes many changes of every state (in dependence on time
interval during which the system is observed). It means that
the manipulation caused by the Newton space is very roughly
defined. The crucial result of this chapter is the proof that
such a “roughness” is a result of properties of the Newton space
(especially its symmetries). Moreover, the proof outlines why
the inner dynamics has a form of an infinite amount of resulting
states without an assumption about a passage of time. Thus
‘time’ may be understood, roughly speaking, as a mathematical
result of properties of the Newton space.
As emphasized in Section 2.2, physicists try to construct
highly sophisticated experimental device to be able to manip-
ulate with studied physical systems very finely. They do not
wish to get many various results when acting on the system by
a defined action. From this point of view, the Newton space
is a very bad ‘device’ resembling a “man’s hammer blow” in
Fig. 2.23 . What about to improve this ‘device’ ? As well as an
3
At this point, our concept of space approaches very closely to ideas of
V. Šedivý [38], who compares space with a ‘machine’ (see the motto).
undefined hammer blow may be replaced by a precise device

that can impact by a strictly defined way, we look for a more
“sophisticated” space. It means to find a complete decomposi-
tion of the manipulation caused by the Newton space. When
doing that, we obtain a modified space – the surrounding – that
causes changes parameterized by a length parameter. We call
it the Mach space and show its correspondence with ideas of
Ernest Mach based on the criticism of the Newton’s absolute
space.
5.1 Space as a surrounding

In our conception, there is nothing else than various physical
systems characterized by their states. States may be changed
as results of various manipulations with these systems. There
is no spatial “background” in which systems are located and
no independent flow of time in which manipulations obtain a
temporal determination. If we want to speak about ‘space’
and ‘time’, we must use only concepts like ‘system’, ‘state’,
‘manipulation’, ‘change’, and so on.
Space in a common understanding is something in which
any physical system is sited. This ‘something’ is not a part of
the system – space is always an external entity. But we have
no other entities than physical systems in our approach. That
is why we introduce a special physical system called simply the
space and denoted as E. In what follows, we define properties
of this system so that they model main features of space as an
independent entity. These properties do not specify this system
uniquely – there are many different spaces. Afterwards, we will
specify some important types of spaces. Then we will use the
symbol E with indices.
5.1. SPACE AS A SURROUNDING 117
The first property should model the idea that physical sys-
tems are “in space”. The idea of “being in space” is deeply re-
lated with the main property of space, namely, that it includes
fixed places – spatial points. Physical systems may occupy
these places by various manners while places themselves re-
main always the same. In our approach, only parts introduced
in Chapter 3 are understood as fixed (firmly chosen) elements
of the description. Recall that a special set of parts, Pch , has
been introduced to model the situation when these fixed ele-
ments are defined by the surrounding (e.g., as an external wall
that defines a container with a gas).
Space may be characterized as a “tool” defining special kind
of parts – spatial points. The main idea of our approach is that
the set of parts Pch , defined by E, includes also parts that
may be identified with spatial points. We, however, have no
possibility of determining a concrete point in space otherwise
than by means of another object located in space. Namely,
this object (if it has some suitable properties) may serve as
a referential frame that enables us to describe points by their
coordinates. It motivates us to the following definition:
Let Ssur be an arbitrary surrounding of S. We say that Ssur
defines spatial points for S if there is an integer d and a set
of physical systems, Cd = {a, b, . . .}, so that the composition of
each of them with Ssur (i.e. (Ssur , a), etc.) is the surrounding
of S, too, and each of these surroundings defines sets of parts,
Pch ((Ssur , a)) ⊂ P(S), Pch ((Ssur , b)) ⊂ P(S), etc., so that
for every a ∈ Cd there is a uniquely given one-to-one mapping,
fE,a : Ed → Pch ((Ssur , a)),
that assigns to every point x ∈ Ed an atomic part fE,a (x) called

the spatial point.
Systems {a, b, . . .} are called the d-dimensional coordinate

systems. The first property of E thus may be postulated as
follows:
(S1) If E is a surrounding of a physical system S, then E de-
fines spatial points for S.
If we have a concrete d-dimensional coordinate system a we
denote
Pch (a) ≡ fE,a (Ed ) ⊂ Pch ((Ssur , a)).
Remark: The identity of points in different coordinate systems is
expressed by the relation
fE,a (x) = fE,b (x0 ). (5.1)
It allows constructing various transformation laws interpreting the

description in one coordinate system into another. The set of coordi-
nate systems, Cd , is usually supposed to define the same set of spatial
points, i.e. Pch (a) = Pch (b).
If there is a maximal d for which Cd exists, we call E the d-

dimensional space. We say that the system S is in (d-
dimensional) space if there is a (d-dimensional) space E that
is the surrounding of S.
Let us have a system S in d-dimensional space and a con-

crete d-dimensional coordinate system a defining the subset of
parts Pch (a) ⊂ Pch whose elements are spatial points. The
variable q̂ ∈ Q is called the field in a if for any state s ∈ Σ(S),
Dq,s = Pch (a). (5.2)
The mapping qs , given by (3.7), defines the spatial field,
qsa = qs ◦ fE,a , (5.3)
that assigns to every point x ∈ Ed the value(s) qsa (x) from Bq ,

i.e.
qsa : Ed → Bq .
We can have various coordinate systems within Cd so that one
variable is the field in one of them but it is not in another. That
is why we choose a “reasonable” class Cd defined as follows:
For any couple of coordinate systems a, b ∈ Cd , for each variable
q̂ ∈ Q and any states s, sn ∈ Σ, n ∈ N, the following conditions
hold:
(i) if q̂ is a field in a then it is a field in b,
(ii) if qsa is a constant function then qsb is a constant function,
(iii) if qsa is a continuous function then qsb is a continuous func-
tion, and
(iv) if for every x, qsan (x) → qsa (x) then qsbn (x) → qsb (x),
where “→” means the convergence in Bq .
In what follows, we will use a firmly chosen reasonable class Cd
of coordinate systems and study only systems whose set of vari-
ables, Q, includes only fields (in an arbitrary coordinate system
from Cd ). It allows us to define the convergence of states from
Σ:
An infinite succession of states sn converges to the state s,
sn → s,
if qsan (x) → qsa (x) for each q̂ ∈ Q at every x ∈ Ed (the coordi-

nate system a may be chosen arbitrarily from Cd ).
Remark: The convergence is point-like (it is defined at individual
spatial points) which means that behavior of series of fields repre-
senting states sn at very distant points (|x| → ∞) does not play an
essential role. Imagine, for example, the sequence of fields qsan (x) so
that qsan (x) = qsa (x) within a succession of balls with a fixed center
x0 and increasing radius, Rn → ∞. This sequence converges to qsa
even if values of every field qsan outside the ball with the radius Rn
are chosen arbitrarily.
This convergence enables us to characterize when a manipula-
tion is not sensitive on behavior of fields at infinity:
Let S be in space. We say that the state s ∈ Σ(S) is stable in
the manipulation α ∈ M(S) if for any succession of couples,
[sn , s0n ] ∈ α, holds that whenever sn → s ∈ Σ then there exists
a state s0 ∈ Σ so that s0n → s0 , too.
Let us suppose that the structure of parts is atomic and

Pch (a) = Patom for every a ∈ Cd . The state space of every part
from Pch (a) is supposed to be the same (and independent on
a), namely Γ(P a ) ≡ Γ0 , and
Γ0 = B ≡ Bq1 × . . . × Bqn , (5.4)
where Bqi are Banach spaces of individual variables. Follow-

ing (3.18), let us define the set of mappings Γa that represents
various possibilities of assigning values of all variables to all
spatial points4 . Since spatial points are defined uniquely by
the chosen coordinate system, the set Γa may be understood
as a set of mappings from Ed to B. An element of Γa will
be denoted simply by ψ (with omitting the specification of the
coordinate system whenever there is no menace of misunder-
standing). That means that ψ ∈ Γa is a mapping,
ψ : Ed → B, (5.5)
4
These possibilities are thus characterized by a form of mappings in Γa .
They may be continuous, differentiable, integrable, etc.
defined at every point x by the prescription,
ψ(x) = [qsa (x)]q̂∈Q . (5.6)
We will always suppose that the set of variables Q (i.e. the set
of fields) is complete with respect to the state space Σ of an
arbitrary system under study. That means that any concrete
ψ represents one state from Σ.
The set of parts Pch (a) consists of an uncountable num-
ber of spatial points. Every state s is represented by a set of
spatial fields, ψ, whose values are defined at these points. We
understand these fields as “distributions of physical properties
in space”. Points, however, are not some separated parts –
they form an infinitely dense continuum. That means that
we should assume some “correlations” of values of fields at
“nearby” points. For example, the continuity of fields ψ at
a point x means that values at points “infinitesimally near” x
cannot differ arbitrarily from ψ(x). These correlations, how-
ever, are not only properties of individual states alone but also
of their changes. If ψ changes to ψ̃ (a set of fields with different
values) the correlations at x mean that ψ̃(x) depends also on
values of ψ at some circumjacent points, i.e. ψ(x0 ), and not
only on ψ(x). It motivates us to the following definition:
Let α ∈ M be a manipulation with S caused by the surrounding
E. We say that α is the trivial manipulation if there is a
function f : B → B so that for any couple ψ, ψ̃ ∈ Γa , which
represents states s, s̃, the condition
[s, s̃] ∈ α ⇒ ψ̃(x) = f (ψ(x)), (5.7)
holds at every point where all fields in ψ are continuous.

The trivial change thus represents situations when values of
changed fields at x, i.e. ψ̃(x), depend only on values of fields ψ

at x. It means that there is no influence of fields at surround-
ing points. The definition concerns only points at which all
fields are continuous (at points of discontinuity, there is even
no “initial” correlation of values of fields at near points). If all
changes were trivial, spatial points would not be “cohesive” and
space would disintegrate into an infinite number of incoherent
elements. Therefore, we accept the following property of E:
(S2) Let S be in space E. Then there exists a manipulation

with S, caused by E, that is not trivial.
This property says more than that the relation (5.7) is not valid.
Namely, if all states of the system S are represented by fields
that are discontinuous at every point, then any manipulation
is trivial. Therefore the property (S2) claims also that any
physical system in space has a state at which all representing
fields are continuous at least at one point.
In our approach, space is represented by a physical system,
E, though we do not introduce its “material” properties. Its
nonmaterial character underlines the fact that it extends into
infinity (in the sense that there are successions of spatial points,
represented by coordinates xn , so that |xn | → ∞). The space
E is thus a little strange “entity” without a substantial specifi-
cation.
How could space be approached by a material substance?
Newton’s space, for example, may be modeled as an infinite,
uniform substance (remember the idea of ether). But no in-
finite substance exists. To guarantee that a sufficiently large
substance behaves like an infinite one, we accept this property
of E:
5.2. SYMMETRIES OF SPACE 123
(S3) If state s models the space E then it is stable in any ma-

nipulation caused by E.
State s is an infinite (hypothetical) state of a material system

that is in E (e.g. an uniform substance). The stability means
that this state may be approximated by a succession of states
that are not infinite (they differ “at infinity”). The meaning
of the fact that a state models the space will be defined in the
next section by (5.10).
5.2 Symmetries of space

To characterize more various spaces, we define the crucial prop-
erty of any space – its symmetries. We do it by using the con-
cept of coordinate system. It is a material (or geometrical)
structure that is not a subsystem of E. The composition of a
coordinate system, say a, and E is a new physical system, (E, a).
If the composition is done with another coordinate system b,
b 6= a, we obtain another physical system, (E, b). Though the
systems (E, a) and (E, b) are different, they may be physically
indistinguishable.
A model of such a situation is an infinite homogeneous envi-
ronment that represents E, and a system of three perpendicular
infinite bars endowed with a sufficient set of scale lines. Let the
system of bars be in a fixed position in the environment. Then
it represents a coordinate system a because whenever a material
object is in the environment, triplets of scale lines on individ-
ual bars determine localized parts of the system. If the same
system of bars is fixed in another position in the environment,
the same triplet of scale lines determines another part of the
system. Thus the bars in another position represent another
coordinate system b. If we have only the environment and one

of the coordinate systems, we cannot distinguish whether it is
a or b. This fact represents a symmetry of the environment.
If, for example, the environment is not homogeneous (i.e. it
includes a salient place), the coordinate systems a and b can
be simply distinguished because their positions may be deter-
mined with respect to this place. That is why we define:
Coordinate systems a and b are said to be equivalent if there
is no property (e.g. a number) having a source either in E or in
the coordinate systems alone which can distinguish them. For
example, there is no function f so that
f (E, a) 6= f (E, b). (5.8)
A typical example of equivalent coordinate systems are the
same but translated coordinate systems in the Euclidean space.
On the other hand, the orthogonal and non-orthogonal ones
are not equivalent because f (E, a) from the relation (5.8) may
represent the angle between two axis. If there is a salient (lo-
calized) place (point) P in E then the translated coordinate
systems may not be equivalent because f (E, a) may be chosen
as the distance of the origin of coordinate system a to P .
Let G be a mapping which transforms states of the system to
other states, i.e.
G : Σ → Σ.
We say that G is the symmetry operation of E if there are
equivalent coordinate systems a and b, and G maps any state s
to the state G(s) so that for each q̂ ∈ Q,
qsa = qG(s)
b
. (5.9)
The concept of symmetry operation allows us to define states
that model the space E:
5.3. SPECIAL SYMMETRIES 125
The state s of a system in E is said to model E if for any

symmetry operation G of E,
G(s) = s. (5.10)
Symmetry operations characterize symmetries of E. We

suppose that the domain of each symmetry operation is the
whole state set Σ. Let α ∈ M(S) be an arbitrary manipula-
tion with S caused by the surrounding E. Symmetries of space
have to influence a concrete form of the manipulation α be-
cause it is fully determined by this space. Namely, any couple
[s, s0 ] ∈ α has to respect all symmetries of the space. It may be
formulated as the following statement:
Proposition: If G is a symmetry operation of E then
[s, s0 ] ∈ α ⇒ [G(s), G(s0 )] ∈ α. (5.11)
The condition (5.11) is justified by the fact that the states s

and G(s) are equivalent from a physical point of view: s in
(E, a) is indistinguishable from G(s) in (E, b) and the couples
(E, a), (E, b) alone cannot be differed anyhow.
Let a system S be in space E. There is a deep relation
between symmetries of the space E and a possibility of find-
ing global variables of S so that E has conservative properties
with respect to these variables. Namely, the condition (5.11)
allows defining such variables. We, however, do not study these
interesting questions here.
5.3 Special symmetries

Let us define for every y ∈ Ed the mapping
Ty : Σ → Σ
as follows: Ty (s) = s0 if and only if for each q̂ ∈ Q and every

a ∈ Cd ,
qsa0 (x) = qsa (x − y). (5.12)
If Ty is the symmetry operation then the equality (5.9) defines
when coordinate systems a and b are equivalent. It has to be
valid at every x ∈ Ed . Hence, (5.12) implies that
qsa (x) = qTb y (s) (x) = qsb (x − y), (5.13)
and thus the coordinate system b is translated by the vector

y with respect to a. The symmetry operation Ty is called the
translation. If Ty is the symmetry operation for every y ∈
Ed then the surrounding E is said to be homogeneous. It
expresses the fact that there is no salient place within it. The
homogeneity of space plays very important role in physics.
Similarly, we can define the symmetry operation called ro-
tation. The definition is analogous to the previous one. Let Q
be a properly orthogonal matrix, i.e.
Q−1 = Q> , det Q = 1. (5.14)
For every matrix Q fulfilling (5.14), the mapping
RQ : Σ → Σ
is defined as follows: RQ (s) = s0 if and only if for every q̂ ∈ Q

and every a ∈ Cd ,
qsa0 (x) = qsa (Q x). (5.15)
If RQ is the symmetry operation then it means that the coor-
dinate systems a and b, fulfilling
qsa (x) = qsb (Q x), (5.16)

5.3. SPECIAL SYMMETRIES 127
are equivalent. The symmetry operation RQ is called the ro-

tation. If RQ is the symmetry operation for every Q the sur-
rounding E is said to be isotropic. It means that there is no
salient “direction” in E.
The last symmetry studied in this section arises when there
is no salient length in E. Let us realize that coordinates of
points x express lengths in E (coordinate systems endow E with
a metric). Therefore, if a new length unit is used we obtain a
new inscription by “x”, i.e., we get a new coordinate system.
Let us introduce the mapping Or : Σ → Σ for any positive real
number r ∈ R+ as follows: Or (s) = s0 if and only if
qsa0 (x) = qsa (r−1 x). (5.17)
There may be length units among dimensions of values from qsa .

However, the change of values of fields caused by the change of
length unit may be compensated by an appropriate change of
other units. A condition allowing doing that is that dimensions
of individual fields, [qi ], where q̂i ∈ Q, cannot be combined to
give the dimension of length [1]. That is, if for any real numbers
p1 , . . . , p n ,
[q1 ]p1 . . . [qn ]pn 6= length. (5.18)
Then the relation (5.17) expresses a description of the field qsa
in a new coordinate system using the r times smaller or larger
length unit. Now, if there is no salient length in E then there is
no number coming from E or the coordinate system alone which
can distinguish coordinate systems a, a0 using different length
units (the used length unit cannot be compared with anything
else to give a number which is different for individual couples
(E, a), (E, a0 ) alone). We say that E is scale symmetric and
the mapping Or is the corresponding symmetry operation.
5.4 Newton space

In Newton’s “Principia” [29], space is defined as something ex-
ternal what remains always similar and immovable. The immo-
bility of this space may by translated to our language as that
E can be only in one state, say s0 . This state has to have very
special properties because space introduced by Newton includes
no salient places, directions or lengths. Otherwise speaking, it
is a one-state surrounding that is homogeneous, isotropic and
scale symmetric. We will call it the Newton space and denote
this space as EN ewt .
The fact that the Newton space has only one state, s0 , im-
plies that any change of this physical system is the identity,
id = {[s0 , s0 ]}. This space, as a surrounding, thus may caused
only one manipulation with an arbitrary system being in this
space because
Σsur × Msur = {[s0 , id]},
and there is only one-element domain of the function F in the

relation (2.6). Let us call this universal manipulation the New-
ton manipulation and denote it as , i.e.
= F([s0 , id]). (5.19)
Notice that the relation (2.11) implies also that if the Newton
space has the conservative property with respect to a global
variable G then the value of this variable is the same at every
state of the system because ssur = s0sur = s0 .
How does the manipulation act on the state space of a
physical system S for which the Newton space is the surround-
ing? That is, what is the subset of Σ × Σ of various changes?
In Newtonian physics, the system “located in empty space”
5.4. NEWTON SPACE 129
changes its initial state, s, into an infinite set of states para-

meterized by a parameter t. It expresses the inner dynamics
of the system and the parameter t is called the time. In our
scheme, the manipulation thus includes, for any state s ∈ Σ,
an infinite set of couples,
Dyn(s) ≡ {[s, s(t)], t ∈ R0,+ } ⊂ , (5.20)
if we denote s(0) ≡ s.
Is this conclusion given only by empirical observations (that
physical systems have inner dynamics changing their states into
infinite sequences of states) or has it a deeper origin? Surpris-
ingly, the fact that the Newton space manipulates with any
system in this way is a consequence of its symmetries. To show
it, let us proof the following theorem first.
Theorem: The manipulation caused by the Newton space
cannot be perfect for any system S which is in the Newton
space, i.e., there are states s, s1 , s2 ∈ Σ(S), s1 6= s2 , so that
[s, s1 ], [s, s2 ] ∈ .
Proof : Let us suppose that the manipulation is perfect, i.e., it is
a mapping : Σ → Σ so that [s, s̃] ∈ means that
s̃ = (s). (5.21)
Due to the homogeneity and scale symmetry of the Newton space,

(5.21) implies that also
Ty (s̃) = (Ty (s)) (5.22)
for every y ∈ Ed , and
Or (s̃) = (Or (s)) (5.23)
for every r > 0. In the field representation, states become sets of

fields and the mapping becomes a mapping T between them. If we
represent s by ψ and s̃ by ψ̃ we obtain from (5.21) the definition of

T , namely
ψ̃ = T (ψ). (5.24)
Let us have an arbitrary couple of states s, s̃ fulfilling (5.21), i.e.
an arbitrary couple of fields ψ and ψ̃ fulfilling (5.24). Let x0 be an
arbitrary point at which fields ψ are continuous (such a point has to
exists as implied by (S2)). We will study values of fields ψ̃ at this
point, i.e. ψ̃(x0 ), in dependence on ψ. The relation (5.22) guarantees
that if we choose a translated coordinate system, the mapping trans-
forming the fields corresponding to s and s̃ in the new coordinate
system remains the same. That is why we may suppose that x0 = 0
because we can ever choose a coordinate system with the center at
an arbitrary point.
Now, let us choose a coordinate system using a smaller length
unit. The scale symmetry of the Newton space guarantees that the
mapping T remains the same (as implied by (5.23)). Taking smaller
and smaller units we construct a succession of fields, ψn , that con-
verges at arbitrary point x ∈ Ed to ψ(0),
lim ψn (x) = ψ(0). (5.25)
n→∞
That means that the limit of the succession of states sn corresponding

to the fields ψn is a constant field, sc , whose field representation is
ψ0 , defined as
ψ0 (x) ≡ ψ(0). (5.26)
Any state representing by constant fields is a model of EN ewt . Hence
(S3) implies that (sn ) converges to (sc ). In the field representation,
T (ψn ) converges to T (ψ0 ) at every point. On the other hand,
lim (T (ψn ))(0) = ψ̃(0), (5.27)
n→∞
because (5.23) guarantees that the re-scaled field ψ is mapped onto

the re-scaled field ψ̃ and all re-scaled fields have the same value at 0.
As a result, we obtain that
ψ̃(0) = T (ψ0 )(0). (5.28)
5.4. NEWTON SPACE 131
As implied by (5.26), ψ0 is defined only by the value ψ(0). Let us

define a function f : B → B so that
b
f (b) = T (ψcon )(0), (5.29)
b
where ψcon is a set of constant fields so that
b
ψcon (x) = b (5.30)
(notice that the scale invariance guarantees that the state correspond-
b
ing to ψcon has to exists in Σ if there is a state sb ∈ Σ so that there
exists a point x ∈ Ed at which fields representing sb are continuous
and reach the values b). The relation (5.28) then may be written as
ψ̃(0) = f (ψ(0)). (5.31)
The condition (5.31) can be written in the original coordinate system
(because it depends only on values of the fields at the point x0 ) and
we obtain
ψ̃(x0 ) = f (ψ(x0 )). (5.32)
This condition is valid at all points of Ed where fields ψ are contin-
uous. It means that the manipulation is trivial (see the definition
at Section 5.1). Since there is no other manipulation than we have
obtained a contradiction with the property (S2). Therefore the map-
ping T cannot exist.
The core of the proof is the validity of (5.23) implied by

the scale symmetry of EN ewt (and by the assumption (5.18)
which say, for example, that ψ cannot include a field whose
values have the dimension of length). If were dependent on
a length `, i.e. ` , the relation (5.23) could not be in force and
` might be a mapping. An illustrative example is the gaussian
transformation,
(x−x0 )2
Z
ψ̃(x) = π −d/2 −d
` e− `2 ψ(x0 )dd x0 , (5.33)
Ed
that defines a perfect mapping between spatial fields whose do-

main is a broad class of fields ψ (if ψ includes more fields, (5.33)
is defined for each component). Notice that the occurrence of
the length parameter ` is essential because the dimension of
coordinates x0 has to be compensated by a length parameter
(the homogeneity of space does not make possible using x in
place of `)5 .
A length-dependent manipulation ` , however, cannot be
caused by the Newton space because it does not include a
salient length `. The existence of such a manipulation would
break its scale symmetry. The impossibility of occurrence of a
salient length in EN ewt may be also interpreted as the fact that
EN ewt includes all lengths. This expression may seem as rather
unusual, but notice that the same interpretation is done in the
case of homogeneity: EN ewt cannot include a salient place, so
it is supposed to include all places.
This small shift in locution, however, brings a remarkable
result. Namely, if EN ewt includes all lengths then it may caused
(“simultaneously”) all mapping ` parameterized by all length
parameters. We may do it it formally by understanding as an
infinite “vector”,
≡ (. . . , ` , . . .), (5.34)
whose “components” are indexed by all lengths. Individual
mappings, ` , do not obey the condition (5.23) because the
operator Or acts as Or (` (s)) = r` (Or (s)). That is, Or changes
the length ` into the length r` and thus realizes a “shift” of
components in . However, it does not change the “vector” as
a whole! The relation (5.23) thus may be formally understood
5
When trying to “construct” a mapping T without the use of a length
parameter, we either disturb the homogeneity of the problem or obtain a
trivial solution, ψ̃(x) = f (ψ(x)).
5.5. COMPLETE DECOMPOSITION OF 133
as remaining in force for the whole “vector”,

(Or (s)) = Or ((s)).
This interpretation of leads to a “discovery” of time dynamics!
Namely, the set of all lengths is isomorphic to the set of all
positive real numbers. If we identify ` ∼ t ∈ R+ and denote
` (s) ≡ s(t) we have
(s) = (. . . , ` (s), . . .) ≡ {s(t), t ∈ R+ }. (5.35)
When adding the identity change, [s, s], we can write (5.35) in
the form (5.20). Thus the structure of time dynamics seem to
appear as a result of symmetries of the Newton space.
5.5 Complete decomposition of

The one-state Newton space can cause only one manipulation .
Since this space is homogeneous and scale symmetric, cannot
be a mapping. We have uncloaked a possible structure of
as consisting of infinite set of mappings parameterized by all
length parameters. From now, we take this structure as given
(taking into account also its “experimental evidence” in a form
of a permanently observed time evolution).
The “vector” (5.34) resembles a parallel composition of the
manipulation . Since ` are supposed to be mappings, (5.34)
presents indeed a complete decomposition of . In Section 2.2,
we explained this concept as a result of searching for a bet-
ter surrounding (e.g. a device) that can manipulate with the
system much more finely to give only single valued results (see
Fig. 2.2).
Let us try to introduce a “better” surrounding that causes
the manipulations ` . We will suppose that this surrounding is a
space, too, and denote it by EM . This space may be very similar

to the Newton space (e.g. be homogeneous and isotropic) but
it cannot be scale symmetric. Namely, (5.11) cannot be valid
if α ≡ ell and G ≡ Or .
What should be its state set? For any length `, ` has to be
defined via a given mapping F (see (2.6)), as
` = F(ssur , αsur ),
where ssur ∈ Σ(EM ) and αsur ∈ M(EM ). It means that either

the state of EM or the manipulation of EM has to depend on
the parameter `. If states of EM were dependent on length
parameters, then any couple of these states would depend on a
couple of length parameters and hence any manipulation αsur
would depend on the couple (or couples) of length parameters.
That is why we suppose that not states of EM alone but
their couples, [ssur , s0sur ], depend on `. A simple example is the
Newton space equipped with one material particle and a firmly
chosen direction6 . States of this space are given by shifting the
position of the particle along the direction. While no length
can be assigned to individual states (we have no possibility
of defining the position of the particle in space alone7 ), any
couple of different states of EM defines a length as a distance of
the particle in these states. To be able to define a length also
for the couple of identical states and differ shifts in opposite
directions, we introduce the set <, whose elements are called
the generalized lengths. < includes all lengths (i.e. objects
denoted by `), all minus lengths (denoted by −`) and the object
0` called the zero length (0` ≡ ` − `) (< is isomorphic to the
set of real numbers).
6
This space is neither isotropic nor homogeneous.
7
Coordinate systems are not parts of space!
5.6. MACH SPACE 135
Being motivated by this example, we suppose that for any

space EM there is a one-to-one mapping fL ,
fL : Σsur × < → Σsur , (5.36)
so that for any ssur ∈ Σsur the set {fL (ssur , `), ` ∈ <} = Σsur ,
and
fL (ssur , 0` ) = ssur ,
fL (fL (ssur , `1 ), `2 ) = fL (ssur , `1 + `2 ).
(5.37)
It implies that for any couple of different states, [ssur , s0sur ],
the equation fL (ssur , `) = s0sur defines a unique length |`| (a
positive generalized length).
5.6 Mach space

The space with a particle resembles the situation of celestial
observers. For them, the Sun plays the role of such a material
particle. Any couple of states forming a change defines a length
– the distance of two positions of the Sun (measured in an
angle measure multiplied by an arbitrarily chosen length unit).
Changes of the Sun position, however, realize some basic clock
for celestial observers8 . Let us notice also that any clock may
be understood as connecting a length (e.g. the high of sand
column in a sand-glass, the length of a trajectory passed by a
pendulum, etc.) with an arbitrary change9 .
8
Notice that other clocks were consequently derived from the movement
of the Sun.
9
The digital clock that shows only numbers uses the possibility of trans-
forming any length to real numbers. We, however, have no experience with
a “digital time”.
This noticing guides us to a simple model of the space EM .

It is a surrounding having property of space and including spe-
cial subsystems called clocks. Their important feature is the
existence of a variable ĉ, called the clock variable, assign-
ing a length to each clock at any state of EM . Values of clock
variables, however, cannot be determined at individual states.
Instead, the following property is supposed: in any couple of
two different states of the system EM , say sM M
1 , s2 , the clock
variable of an arbitrary couple of clocks C1 , C2 fulfills
csM (C1 ) − csM (C1 ) = csM (C2 ) − csM (C2 ) (5.38)

1 2 1 1
(a clock synchronization). It allows defining the mapping fL as

follows: fL ([sM M
1 , s2 ]) = csM (C) − csM (C) where C stands for
1 2
an arbitrary clock.
Moreover, we assume that the identity manipulation with
EM , denoted as idM , may cause only the identity manipulation
with the studied system. This is a natural idea connected with
any clock: if there is no change on the clock, there is no change
of anything. Formally, it is formulated as the demand that for
every state sM of the surrounding EM holds the condition
F([sM , idM ]) = id, (5.39)
where F is the mapping defined by (2.6) and id the identity

manipulation with the studied system. This surrounding will
be called the clock space. If we suppose that clocks are the
same physical systems, negligibly small and located arbitrarily
“densely”, the clock space is homogeneous and isotropic.
The clock space looks like the Newton space with an ab-
solute flow of time. But there is no ‘flowing time’ in EM . We
have only a set of states of a physical system, EM , whose im-
portant components are various subsystems called clocks. The
5.6. MACH SPACE 137
clocks belong inherently to this space. They cannot be looked

apart from it. It resembles the situation in generally relativistic
physic that shows that there cannot be some physical subsys-
tems (e.g. clocks) that are independent of spacetime. But if
clocks are indispensable physical systems belonging to the sur-
rounding then we cannot exclude some physical influences of
these subsystems on the studied system.
This is a very interesting point. In physics (excepting the
theory of general relativity), clocks – as physical systems – play
a negligible role, and if they are taken into consideration as a
matter of fact, they are always supposed to have no influence
on the system. Namely, the idea of an absolutely flowing time
is so suggestive that there is even no reason to think about a
necessary presence of some material systems called clocks. This
attitude, however, is not supported by our experience. When
going back to our main clock - the Sun in the sky - we cannot
neglect its physical influence. But we can involve the Sun and
the whole solar system into the studied system and look for a
more distant clock with a negligible physical influence.
Such clocks are distant stars. We observe their regular mo-
tion in the sky and any change “here” can be related to their
trajectory on the firmament. Moreover, a direct physical in-
fluence of them may be neglected. Nevertheless, the fact that
distant stars move in the sky is in a way related to some physical
phenomena observable on the Earth. Namely, there are gyro-
scopic effects that may be measured in rotating systems. The
Earth rotates. But what does ‘rotation’ mean? In Newtonian
physics, the rotation is defined towards (absolute) space10 . But,
as noticed by famous physicists and philosopher Ernest Mach,
10
In fact, Newton proved the existence of absolute space by showing an
absolute character of the rotation (in his famous bucket experiment).
the rotation of the Earth is not defined towards an absolute

space but towards distant stars. The fact that we observe mov-
ing stars in the sky is a good explanation of some gyroscopic
effects.
Though there is probably no causal influence of stars on
the Earth, some noticeable relation between ‘clocks’ (stars) and
some effects on the Earth cannot be omitted. We will not open
some interesting questions connected with the problem of the
relativity of rotation motion. The only result accepted by us
is just the noticing that clocks may be interconnected with
the studied system by some physical relations (whatever their
physical meaning)11 . Are then manipulations caused by such
a clock space those that form a complete decomposition of the
Newton manipulation ?
The example with distant stars shows that the answer may
be negative because certain gyroscopic effects cannot be ob-
served in the Newton space. They may be understood as a
“new” (non-Newtonian) influence of the clock space into a sys-
tem which is in this space12 . In this case, the space EM is
characterized only by the existence of the mapping fL and no
by its ”yield” of manipulations ` giving a complete decomposi-
tion of the Newton manipulation . The space EM understood
in such a generalized sense will be called the Mach space.
The Mach space may or may not cause manipulations ` . In
any case it causes manipulations parameterized by all lengths,
11
In the case of distant stars, this connection is principal because distant
objects may work as good clocks only if there is some relative rotation.
Namely, it is the only way in which they can work permanently.
12
Notice also that ‘manipulation’ does not mean only a direct physical
interaction. The manipulation caused by the Newton space means to “let
system alone”, the manipulation caused by the clock space with distant
stars means to “let rotating system alone”.
5.7. ADIABATIC SPACE 139
say (. . . , M M
` . . .). If ` 6= ` then their parallel composition,
M ≡ (. . . , M
` . . .),
defines a generalization of the Newton manipulation . It is

not clear, if it is worth speculating about a generalized of (one
state) Newton space causing the manipulation M instead of
. Nevertheless, such a space would be more natural for celes-
tial observers who can never study physical phenomena in the
Universe without distant stars.
5.7 Adiabatic space

The Mach space is more general than the clock space. Espe-
cially the validity of the condition (5.39) may not be assumed.
As a result, we get much more general situations. An impor-
tant example of Mach space is used in thermodynamics to define
the concept of adiabatic accessibility. This kind of Mach space
presents an extremely difference of manipulations M ` from ` .
Namely, the important feature of this space is that it includes
a weight in a homogeneous gravitational field which is a source
or sink of energy of the system.
This space may be simply realized anywhere on the Earth’s
surface if the studied system is negligible small with respect to
the Earth’s radius13 . Various states of this space are character-
ized by different positions of the weight. That means that any
couple of states defines a plus or minus length (or zero length
if there is no shift of the weight). Notice that not states alone,
13
Though the space E always produces parts forming the Euclidean
space, it does not mean that it goes “through the whole Universe”. For
example, local inertial systems of the theory of general relativity (four-
dimensional Minkowski spaces) are “relatively small”.
but only their couples define the length (we have only one body
in homogeneous field and its position cannot be absolutely de-
termined). We will call this kind of Mach space the adiabatic
space.
The adiabatic space is supposed to have the conservative
property with respect to the global variable called the total en-
ergy. Usually, this variable is defined by the use of an adiabatic
space:
The total energy is a real global variable E so that the adia-
batic space has conservative property with respect to this vari-
able, and the relation,
E(sA A A A
1 ) − E(s2 ) = mg `([s1 , s2 ]), (5.40)
holds for its values, where `([sA A

1 , s2 ]) is the generalized length
assigned to the couple of states [sA A
1 , s2 ] of the adiabatic space,
m is a mass, and g is the intensity of the gravitational field.
The equality (2.11) then allows us to define the value of total
energy (up to a constant) for any system whose surrounding is
the adiabatic space. Namely, (2.11) implies that
E(s2 ) − E(s1 ) = E(sA A

1 ) − E(s2 ). (5.41)
The difference of manipulations ` , arising when the system is

“let alone”, and M
` , caused by the adiabatic space, is without
any doubts: the energy of the system is transmitted from the
space to the system and vice versa. The difference between
the adiabatic and clock space is also evident: there may be
many nontrivial changes of the system while the weight does
not change its position (i.e. the adiabatic space remains in the
same state).
Chapter 6
Accessibility
We need to free ourselves from the prejudices associated with the habit
of thinking of the world as “inhabiting space” and “evolving in time”.
C. Rovelli1
We have defined a special class of surroundings called generally

the space. Each space defines special parts – spatial points –
that have crucial meaning in physics. Namely, spatial points
represent fixed places that allow us to define precise position (or
distribution) of bodies as well as their motion2 . The simplest
1
[35], p. 7.
2
Spatial points in our conception are always defined with respect to a
concrete coordinate system from a class Cd . A group of equivalent coordi-
nate systems defines a certain “relativity” in the description. For instance,
the galilean relativity is obtained by a group of coordinate systems in
“rectilinear and uniform” motion. But this relativity does not annul an
“absolute” character of spatial points. Realize that transformation laws
between various coordinate systems are based on the identity of points
(5.1) defined in different coordinate systems (in special relativity, it is the
identity of points in Minkowski spacetime, called ‘events’).
142 CHAPTER 6. ACCESSIBILITY
kind of such a surrounding is the Newton space which mod-

els very faithfully the idea of an absolute, inert and immobile
empty space. Another example is the Mach space that may be
understood as space with clocks as well as space with a weight
in gravitational field or something else3 .
However, the assumption that the system is in space (what-
ever its concrete kind) is rather a metaphysical construction
than a result of empirical observations. We observe things and
various substances around us but no ‘space’ as independent en-
tity in which are bodies. Remind that the dominant conception
of space in western culture (till Newton) was its understanding
as a relation between bodies (such as “contiguity” or “adja-
cency”, see [35], p. 38).
It is important to realize that the Newton’s predecessor,
R. Descartes, laid foundations of physics (as a program of math-
ematical description of nature4 [6]) without using the concept
of space as an independent entity. Instead, Descartes identifies
space with the extension of bodies, i.e. bodies are not in space
but constitute space by themselves [20]. This extension is called
“res extensa”. A perfect correspondence of the Descartes’ po-
sition with the central result of the general relativistic physics
is startling. To demonstrate it, let us cite the leading physi-
cists in the field of quantum gravity5 , C. Rovelli, ([35], p. 7):
3
The Mach space has also this spatial character (it defines spatial
points) regardless some simple additional objects belonging to it (clocks, a
weight in a gravitation field, etc.).
4
Physics of R. Descartes represents a transitory stage in the mathema-
tisation of nature, connecting Galileo with Newton [17].
5
The main point of the programm of quantum gravity is the formulation
of physics without a given spatiotemporal background. Though the quan-
tum gravity is intended to be a fundamental physical theory, its possible
“applications” concern only extreme short scale structure of spacetime. At
143
“In Newtonian and special relativistic physics, if we take away

the dynamical entities – particles and fields – what remains is
space and time. In general relativistic physics, if we take away
the dynamical entities, nothing remains. The space and time
of Newton and Minkowski are reinterpreted as configurations of
one of the fields, the gravitational field. This implies that phys-
ical entities – particles and fields – are not immersed in space
and moving in time. They do not live on spacetime. They live,
so to say, on one another.”
It is appealing to formulate physics without a given spatial
(and temporal) background. And not only as the method of de-
scription at extreme spatiotemporal scales (where it hardly can
be formulated otherwise) but as a central point of any physical
description. In this chapter, we outline some possible features
of such a (still hypothetical) approach in which the existence
of space as a special entity (surrounding) is not supposed. We
concentrate especially on the question how “inner dynamics” of
the system could be studied. Namely, when studying the inner
dynamics of a system, we have to leave the system alone, which
means that no influence is supposed. But what means that the
system is “let alone” if there is no possibility of “removing” sur-
rounding matter? In Newtonian physics, it is space itself what
remains after removing all bodies and material substances and
physics is then formulated in this space only. However, when
abandoning an independent space, the concept of “inner dy-
namics” becomes problematic.
We have to realize that an irremovable presence of surround-
ing substances (“res extensa”) means a permanent interaction
of the system with something in its surrounding. How does
scales of everyday physics, the quantum gravity should coincide perfectly

with the Newtonian physics.
the Newtonian conception avoid this problem? Namely, why

may we suppose that space does not interact with studied sys-
tems, too? Here is an extraordinary subtle step in the Newton’s
thinking. Namely, Newton does not speculate if and in which
way space affects matter. He demands only that space is not
affected by matter by postulating that space remains always
the same.
This is a very essential point of our contemplation on physics
without a spatial background. Namely, the previous noticing
allows removing various baggage related to the concept of space.
Such as “voidness”, “impalpability”, “ethereality”, etc. Maybe
these attributes are helpful in accepting (but not explaining!)
the exclusiveness of rectilinear and uniform motion in space
(e.g. by imaging motion with vanishing friction), but they are
not essential in the main role that space plays in Newtonian
physics. ‘Space’ could be represented by an arbitrary milieu
that remains always the same. The problem, however, is that
no such immobile, inert matter exists.
That is why we introduce the concept of accessibility that
supersedes the demand of existence of an entity that remains
always the same. The idea comes from the conception of adi-
abatic accessibility defined in thermodynamics. Its essence is
as follows. Let us have a system and its surrounding which
is an arbitrary (material) system (we have no ‘space’ around).
A state s2 of the system is accessible from another state s1 if,
roughly speaking, the change s1 → s2 is caused by a surround-
ing that remains the same during this change. That is, the
surrounding behaves (when causing the change s1 → s2 ) like
space. In the next section, we explain the importance of ac-
cessibility for definition of motion when no space yielding fixed
places (spatial points) is supposed.
6.1. RELATIVE MOTION 145
6.1 Relative motion

The Cartesian physics was not successful. One reason of its
failure, among others, was the problem with definition of mo-
tion 6 . We say that a body moves from a place A to another
place B. But if there is no background space, places can be de-
termined only by a reference to bodies. Motion thus has to be
constituted only by a shifting of parts of res extensa relatively
to each other7 . That implies that an evidence of motion needs
a distinction of parts. But how can parts be distinguished
without specifying their positions, i.e. places where they are
located? It is difficult to avoid some circularity when trying to
define motion by accepting such an utter relativity8 . It became
clear to Newton that the concept of motion could be defined by
concept of place which is exempt from relativity. The idea of
absolute space solves it perfectly because places are absolutely
defined by space. However, it seems to be an exaggeratedly
radical solution.
Notice the way in which we understand motion in every-
day experience. We have no problems to explain that a car
moves from a place A to B. The places A and B are not defined
with respect to some absolute space. They are given parts of a
6
The problem with the definition of motion in Descartes’ conception is
well described by I. Leclerk [18], pp. 192-194. In the next passage, we take
briefly over his argumentation.
7
It is both the Aristotle’s and Descartes’ conception of motion.
8
In general relativistic physics, such a relativity is also accepted by
possibility of formulating the problem in general coordinate systems (e.g.
by defining motion with respect to a cluster of bees). But it is saved
from problems with an “absolute” Cartesian relativity by the existence of
very small referential systems (infinitesimal Minkowski spacetimes [7]) that
behave as “islets with classical spacetime” and form a clue of the whole
conception.
countryside that plays a role of a referential background toward

which the car is moving. But there may be various changes
in the countryside alone during the motion of car. How can
such a movable countryside play the role of a referential system
defining the movement of car? The way of avoiding an endless
relativity is that accepted in everyday life: we do not try to de-
fine the motion in any situation9 . If the majority of referential
parts (including those representing places A and B) changes or
is even destroyed then the motion from A to B is not defined
at all.
The crucial idea is that we cannot define a ‘general motion’
but only concrete motions of concrete bodies with respect to
other concrete bodies. It is enough to define the motion of our
car if the countryside (or its chosen part) behaves like absolute
space during the motion of car. Imagine the countryside and
the car as two physical systems so that the first forms the sur-
rounding of the latter. The “motion of car” may be understood
as a change of one state of the car, s1 , into another state, s2 10 .
If the countryside remains during the motion of car in the same
state, say s0 , then it behaves like a fixed background and may
serve as a perfect referential system allowing to define (and
quantify) the motion of car.
Let us try to formulate it only in the framework of abstract
thermodynamics. In this approach, we cannot say that “during
9
We do not live in res extensa where everything permanently moves
with respect to anything else. Main feature of our “res extensa” is rather
a permanent “supply” of a sufficient amount of things in rest. Thus the
definition of motion is always possible and we do not realize that it is de-
fined, in fact, only in special situations (during an earthquake, say, nobody
takes care to define a movement of a car).
10
We recall that the ‘state’ is a primitive concept of our approach. We
need not define the states of car with a relation to the countryside!
6.1. RELATIVE MOTION 147
a change of states of the system, its surrounding is in the same

state”. Namely, there are no temporal relations of the system
and its surrounding, there are no “simultaneous states”, etc.
The surrounding is defined as a source of all manipulations
with the system (see the definition at Section 2.4). In the spirit
of this definition, the situation when motion may be defined, is
described as follows: The change s1 → s2 is caused by an iden-
tity manipulation with the surrounding. Namely, the identity
manipulation includes only such changes that the initial and
final state of the surrounding are the same. Recall that the
only manipulation with the Newton space is just the identity –
the surrounding thus behaves like the Newton space.
Remark: The formulation may seem rather strange because of the
word ‘caused’. In which way does the countryside cause the move-
ment of the car? Here we again meet the uncommon feature of our
approach. Namely, the causality is understood as a relation of the
system and its surrounding. The fact that the car changes its state
from s1 to s2 means two things. First, that such a change is possible
due to an inner dynamics of the system, second, that it is “allowed”
by the surrounding. In our approach, however, the both aspects are
included in the concept of ‘manipulation’. The only source of ma-
nipulations is the surrounding (in the sense that it may also mean to
“do nothing”).
Notice that the question if the surrounding (the country-

side) is permanently at the same state is nonsensical in our ap-
proach. The manipulation is formed only by couples of states;
no ‘intermediate’ states (of any change) cannot be defined11 .
The crucial question is only whether the change s1 → s2 can
11
We may define, however, the process whose any couple of states
[s(τ1 ), s(τ2 )], where τ1 < τ2 , is caused by the identity manipulation.
actually be caused by an identity manipulation with the sur-

rounding. If yes, we say that the state s2 is accessible from the
state s1 .
The relation of accessibility then gives the restriction un-
der which certain concepts (like the concept of motion12 ) may
be defined. The definition of accessibility, however, depends
on a concrete surrounding whose states occur in the definition
(their couples define the identity manipulation). It is very far
to the conception of ‘inner dynamics’ introduced in physics be-
cause that is supposed to be a property of the system alone,
not depending on a concrete surrounding. The accessibility,
however, seems to depend strongly on a concrete surrounding.
The modification of the definition, that enables us to eliminate
this dependence, will be done in the next section.
6.2 Definition of ‘accessibility’

The way of eliminating a concrete surrounding in the definition
of accessibility is a remarkable step introduced in thermody-
namics13 . Namely, we may eliminate any concrete surrounding
by a weakening of the definition:
We say that the state s2 is accessible from the state s1 if there
exists such a surrounding of the system, that the change s1 → s2
is caused by an identity manipulation with this surrounding.
N
We denote s1 ≺ s2 .
12
Notice an analogy between ‘motion’ in Cartesian physics and ‘work’ in
thermodynamics.
13
The work of E. H. Lieb and J. Yngvason [21] presents clearly this step
when defining the adiabatic accessibility and, especially, when explaining
its role in the axiomatic scheme of thermodynamics.
6.2. DEFINITION OF ‘ACCESSIBILITY’ 149
The letter “N” emphasizes the fact that the accessibility has
the same basic property as the manipulation with the Newton
space. Namely, that it is ‘identity’. Notice that the identity
manipulation with the Newton space includes only one change,
s0 → s0 , because the Newton space has only one state, s0 . In
the definition of accessibility, the identity manipulation may be
formed by an arbitrary couple of same states.
The definition of accessibility requires to study such manip-
ulations with a system that are caused by the following “mech-
anism”. We find a physical system (e.g. a device) that plays
the role of a surrounding of the system. It means that it is a
source of all manipulations with the system14 . The surround-
ing may be quite arbitrary, the system may interact with an
arbitrary amount of objects but, when it reaches the state s2 ,
the surrounding has to be at the same state as when the system
“was” at the state s1 .
Remark: The logic of the definition is following. We “try” various
surroundings till find (or do not find) the “right” one. Any exchange
of surrounding, however, may change the system (see the end of Sec-
tion 2.4). Nevertheless, if a “testing” surrounding “destroyed” either
the state s1 or s2 of the system then it could not be the “right” one.
The situation, when the system is surrounded by the “right”

surrounding realizing the accessibility, corresponds to an arrange-
ment when there is no “record” in the surrounding after the
change from s1 to s2 . Surprisingly, it seems to have a con-
nection with ideas of quantum mechanics: if the state s1 is
pure then the state s2 has to be pure, too, because there is no
entanglement with the surrounding.
14
It also defines these manipulations – see the end of Section 2.4.
The situation, that is very close to that if the system is in

the Newton space, arises when the surrounding has even the
same conservative properties as the Newton space:
We say that the state s2 is conservatively accessible from the
state s1 if there exists such surrounding of the system that the
change s1 → s2 is caused by an identity manipulation with the
surrounding and this surrounding has conservative properties
with respect to the same global variables as the Newton space.
The surrounding that defines the accessibility can influence the
system very heavily without changing its own state. An exam-
ple is a massive elastic wall (a part of the surrounding) that
changes momenta of light particles (being parts of the system).
States differing by directions of momenta of particles thus may
be accessible.
The conservative accessibility is a finer concept. Namely,
the change s1 → s2 has to be realized without a change of global
momenta and other global conservative quantities. There are
no “massive walls” in the surrounding that causes “accessible”
changes of states of the system. It is not so easy to manage
such a change of the state “during which” the surrounding re-
turns to the same state (e.g. an inverse of momenta of only one
particle is not possible). It is an essential aspect of difference
between the accessibility and conservative accessibility. It con-
sists in a possibility or impossibility of occurrence of a scale gap
between the system and the surrounding. The massive wall has
to remain at the same state after reversing the momentum of a
particle (its change of momenta is negligible). In other words,
the description of the wall is done at much higher scales that
that of the system15 .
15
There are some connotations to presence of macroscopic measuring
6.3. TIME ARROW 151
The “dynamics” of changes s1 → s2 that realize the conser-

vative accessibility is described by conservative laws. A very
nontrivial question is whether there is a general formulation of
some additional conditions that guarantees the return of the
surrounding to the initial state. Namely, the fulfillment of the
conservative laws only, does not appear to be sufficient. In
thermomechanics, for instance, there is a set of balance laws
that describe conservative properties of the studied medium
(Section 4.2). However, the balance laws are not sufficient to
describe dynamics of the studied system. A set of the so-called
constitutive laws has to be added. The search for such laws
is the crucial problem of thermomechanics.
6.3 Time arrow

The Newton space is a very nice surrounding because its struc-
ture is as simple as be only possible. On the other hand, manip-
ulations caused by this surrounding are very roughly defined.
It behaves like a very bad measuring device that may give an
infinite number of possible results. That is why we looked for a
better “device” doing the same as the Newton space but manip-
ulating with the system much more gentle, and giving exactly
“targeted” results (it is the meaning of the complete decompo-
sition of a rough manipulation). The Mach space is an example
of such a surrounding. Its structure is more complicated though
(as the structure of any sophisticated device), but it “produces”
an infinite number of manipulations that cause uniquely defined
results.
device in quantum mechanics. We, however, avoid detail discussions con-

cerning quantum mechanics in the entire of the book.
The accessibility means the existence of surrounding that

causes the change by the same “mechanisms” as the Newton
space. Because the set of manipulation with Mach space is
much more larger than a simple identity, we may radically gen-
eralize the definition of accessibility by replacing the Newton
space by the Mach one. The definition is as follows:
We say that the state s2 is weakly accessible from the state s1
if there exists such surrounding of the system that the change
s1 → s2 is caused by such a manipulation α with the surround-
ing, that any couple of states [s1sur , s2sur ] ∈ α defines uniquely a
length ` ∈ < via a mapping fL (5.36). We denote s1 ≺ s2 .
The term “weak” is justified by the fact that the accessibility
implies the weak accessibility. Namely, the accessibility means
that the manipulation α is the identity, which defines the gen-
eralized length ` = 0. The opposite implication, however, is
not valid. The essence of the weak accessibility is the same:
we look for an arbitrary surrounding which causes the change
s1 → s2 so that its state may change into another that, how-
ever, may differ from the original one only by something that
uniquely defines a certain plus, minus or zero length.
If that ‘something’ defining a generalized length is a clock
(located within an arbitrary surrounding) then we call the re-
lation ≺ the weak accessibility in clock realization. It rep-
resents the time arrow in physics. Namely, the weak accessi-
bility then simply means that the state s1 may be transformed
into s2 by an arbitrary manipulation with a surrounding so
that the surrounding is, at the end, at the same state as at the
beginning, excepting the datum on the clock. The fact that
there are states s1 , s2 so that s2 is not accessible from s1 ex-
presses an impossibility of traveling to past. To say it in other
words, everything (not belonging to the system) may be “tided
6.4. ADIABATIC ACCESSIBILITY 153
up” to the original conditions, only the ‘passed time’ cannot be

“passed back”.
It offers an interesting possibility of defining the ‘time’ as a
global variable:
The global variable t is called the time if its values copy the
relation ≺ in the meaning that
t(s1 ) ≤ t(s2 ) ⇔ s1 ≺ s2 , (6.1)
where s1 ≺ s2 means the weak accessibility in clock realization.

This definition indicates an interesting connection of motion
and time. While the first may be defined by the use of accessi-
bility motivated by the Newton space, the later is defined by a
special realization of weak accessibility motivated by the Mach
space.
6.4 Adiabatic accessibility

If that ‘something’, defining the length in the weak accessibility,
is a weight in a gravitational field, we obtain the definition of
the so-called adiabatic accessibility known from thermody-
namics. We precede it here in the thermodynamic formulation
taken from [21] (p. 21):
A state s0 is adiabatically accessible from state s, in sym-
bols s ≺ s0 , if it is possible to change the state from s to s0 by
means of an interaction with some device (which may consist
of mechanical and electrical parts as well as auxiliary thermo-
dynamic system) and a weight, in such a way that the device
returns to its initial state at the end of the process whereas the
weight may have changed its position in a gravitational field.
In our language, the ‘device’ represents an arbitrary surround-

ing. Any manipulation realizing the change in the definition is
called the adiabatic manipulation. The crucial importance
Figure 6.1: A typical process illustrating the adiabatic accessibility.

A falling weight rotates a propeller which mixes a fluid, being isolated
by a wall from a heat exchange between the surrounding. At the end
of the process, the fluid is warmer that before and the weight is down.
No other change has happened in the surrounding.
of the adiabatic accessibility is given by the fact that it defines

the concept of entropy as a global variable:
The global variable S is called the entropy if its values copy
the relation ≺ in the meaning that
S(s1 ) ≤ S(s2 ) ⇔ s1 ≺ s2 , (6.2)
where s1 ≺ s2 means the adiabatic accessibility.

There are many question concerning the entropy. For example,
if it can be defined uniquely, if it can be understood as a (local)
variable, and so on. The only situation when we know the posi-
tive answer on these questions, is the case when the system is a
composition of simple systems defined in Chapter 4. Then the
6.5. THERMODYNAMICS AND ENTROPY 155
entropy becomes the entropy of equilibrium thermodynamics.

We outline the way of constructing the thermodynamic entropy
in the next section.
Beforehand, we mention the following noticing. As every-
body knows, entropy has a connection to time arrow. The
presented approach reveals a possible explanation of this con-
nection. When we compare the definitions of time and entropy,
given by relations (6.1) and (6.2), we see that they are nearly
identical. If we omit the specification of a concrete realiza-
tion of weak accessibility, we obtain the same definitions. A
provocative idea intrudes herewith. What if entropy and time
are only different ways of interpretation of the same thing, the
weak accessibility?
6.5 Thermodynamics and entropy

In the work of E. H. Lieb and J. Yngvason [21], it is solved the
problem of a unique construction of the local variable Ŝ that
defines the global variable S (by the use of the rule (3.8)) that
fulfills the condition (6.2). It is done only for special kind of
systems – compositions of simple systems (see Section 4.1).
Let us have N simple systems, S1 , . . . , SN . We will study
the system
S = (S1 , . . . , SN )
formed by their composition. The simple system Si is formed

by ni parts and thus its composition is formed by a finite set of
parts. Hence the structure of the system S is finitely atomic.
Following (3.16), we define the state space of S as
Γ = Γ1 × . . . × ΓN , (6.3)
where Γi is the state space of the i-th simple system. Elements

of Γ will be denoted by X, Y, Z, elements of its scaled copy by
rX, rY, rZ. The set of variables, Q, is supposed to be complete.
It implies that elements of state space may be identified with
states of the system,
X ∼ s. (6.4)
The state of a composition will be denoted by (X, Y ) if X and
Y are states of individual systems. The states X1 , X2 are said
to be comparable if either X1 ≺ X2 or X2 ≺ X1 (the symbol
≺ means only the adiabatic accessibility in this section). If the
states X1 , X2 are in the relation X1 ≺ X2 and X2 ≺ X1 we
A
write X1 ∼ X2 .
The composition of N simple systems is formed by Nc ≡
n1 + . . . + nN atomic parts, pi , (see the definition of simple sys-
tems). We suppose that all these parts are mutually disjunct,
i.e., the total part is defined as the composition of all parts,
p0 = (p1 , . . . , pNc ).
It implies that if we find the entropy as a local variable that

has the averaged property, then it is defined on the part p0 ,
too, and Ss (p0 ) defines the entropy as a global variable S(s).
Being motivated by classical equilibrium thermodynamics,
the wanted entropy is supposed not only to have the averaged
property, but to be an additive variable16 (i.e., it is defined for
any composition as a sum of entropies of individual compo-
nents). Since the entropy is supposed to be defined on simple
16
The additivity is a very strong condition considering the relation
(6.2). Namely, (as mentioned in [21], p. 20) the relation S(X) + S(Y ) ≤
S(X 0 ) + S(Y 0 ) says us exactly which adiabatic processes are allowed in the
compound system, independent of any detailed knowledge of the manner
in which the two systems interact.
systems, it has to be either intensive or extensive. However, it

cannot be intensive because of its additivity. When summing
up previous considerations, we look for an additive variable
whose values at total part define a global variable S fulfilling
(6.2). Otherwise speaking, we have to prove the following the-
orem, called the entropy principle (ibid., p. 19).
Entropy principle: There is a real-valued variable Ŝ defined
at all states on all parts of the composition of simple systems
called entropy such that
• Monotonicity: When X and Y are comparable states then
X ≺ Y if and only if S(X) ≤ S(Y ). (6.5)
• Additivity and extensivity: If X and Y are states of some

(possibly different) systems and if (X, Y ) denotes the cor-
responding state in the composition of the two systems,
then the entropy is additive for these states, i.e.,
S((X, Y )) = S(X) + S(Y ). (6.6)
S is also extensive, i.e., for each r > 0 and each state X

and its scaled copy rX,
S(rX) = rS(X). (6.7)
The crucial problem, however, is how to formalize the concept

of adiabatic accessibility because its essence is that “there ex-
ists a surrounding (device) that causes such a manipulation
with the system that realizes the change from s1 to s2 ”. But
it is an absolutely hopeless task to check through all possible
manipulations with the system. The approach of Lieb and Yn-
gvason avoid this problem by an ingenious idea. Namely, they
do not study all possible surroundings realizing the adiabatic

accessibility, but find logical connections that are supposed to
be implied by the relation of adiabatic accessibility.
The logical connection means, for instance, that the exis-
tence of adiabatic manipulations realizing changes X → Y and
Y → Z implies the existence of an adiabatic manipulation re-
alizing X → Y . These logical relations are then formulated in
the form of axioms. The main idea then consists in finding a
sufficient group of such axioms that implies the validity of the
entropy principle. For illustration, let us present a first group
of these axioms (ibid., p. 21):
A
(A1) X ∼ X.
(A2) X ≺ Y and Y ≺ Z implies X ≺ Z.
(A3) X ≺ X 0 and Y ≺ Y 0 implies (X, Y ) ≺ (X 0 , Y 0 ).
(A4) If X ≺ Y, then rX ≺ rY for all r ∈ R+ .
A
(A5) For 0 < z < 1, X ∼ (zX, (1 − z)X).
(A6) If (X, εZ0 ) ≺ (Y, εZ1 ) holds for a sequence of ε’s tending
to zero and some states Z0 , Z1 then X ≺ Y .
(A7) Assume X and Y are states in the same state space, Γ,
that has a convex structure, i.e. if X, Y ∈ Γ then for any
r ∈ (0, 1) the value rX + (1 − r)Y belongs to Γ too. If
r ∈ (0, 1) then (rX, (1 − r)Y ) ≺ rX + (1 − r)Y .
These axioms are justified by the definition of adiabatic acces-
sibility and common physical notions. The validity of (A3), for
instance, seems to be demonstrated by putting the individual
systems side by side and “construct” an adiabatic manipula-
tion based on those realizing X → X 0 and Y → Y 0 separately.
(A4), on the other hand, means that the possibility of realiza-

tion of an adiabatic manipulation remains in force if the system
is made smaller or larger (the question of a possibility of con-
structing an adequately smaller or larger device, as a “re-scaled
surrounding”, may be relevant here).
The axiom (A7) evokes an idea of “mixing” of two sep-
arated systems at two different states. It says that we can
always obtain, by an adiabatic manipulation, the state that is
its “mixture”. It looks obviously if the both systems are gases
(or liquids). Namely, we may simply put them side by side,
remove a separating barrier and let them to mix mutually17 .
Why may we rely on a spontaneous mixing? It is a typical
physical process that is justified by the second law of ther-
modynamics. At the first glance, it may seem strange that
there is no axiom in the Lieb’s and Yngvason’s scheme called
the “second law”. But we must not forget that the axioms do
not “translate” thermodynamics into an axiomatic system, but
model the relation of adiabatic accessibility as well as possible.
The second law is thus an auxiliary statement justifying more
or less the chosen axioms.
In thermodynamics, the second law expresses an impossi-
bility of realizing certain processes. In the axiomatic scheme,
this impossibility is formulated only for simple systems in the
form of the following axiom (ibid., p. 42):
Irreversibility. For each X ∈ Γsimple there is a point Y ∈
Γsimple such that X ≺ Y but Y ≺ X is not valid.
Remark: When using the identification of states with points of state
space, (6.4), we see that for states s1 ∼ X, s2 ∼ Y , where X, Y occur
in the previous axiom, the relation (6.2) implies that S(s1 ) < S(s2 ).
17
The meaning of (A7) in the case of solids is not so clear (ibid., p. 33).
If we compare it with (6.1), we see that an analogy of the irreversibil-

ity axiom for the weak accessibility in clock realization would be a
statement that there is a “nonzero” flow of time.
If states are adiabatically accessible or not is deeply connected
with the physical quantity called the temperature. We can-
not, for instance, cool down a body only (without a change of
work variables) by an adiabatic manipulation18 . The concept
of temperature is deeply related to a possibility of putting two
systems into thermal contact. It is guaranteed by the following
axiom (ibid., p. 55), where E denotes the value of energy of a
simple system and w values of its work coordinates:
Thermal contact. For any two simple systems with state
spaces Γ1 and Γ2 there is another simple system, the thermal
join of Γ1 and Γ2 , with state space
∆  = {(E, w , w ) : E = E + E , (E , w ) ∈ Γ , (E , w ) ∈
Γ }. Moreover, ((E1 , w1 ), (E2 , w2 )) ≺ (E1 + E2 , w1 , w2 ).
Notice that the last relation expresses another assumption con-
cerning the adiabatic accessibility19 .When splitting systems form-
ing the thermal join, we obtain two systems in mutual thermal
equilibrium, which enables us to define the temperature. It
corresponds to the following axiom (ibid., p. 55):
Thermal splitting. For any point (E, w1 , w2 ) ∈ ∆  there is
at least one pair of states, (E1 , w1 ) ∈ Γ1 , (E2 , w2 ) ∈ Γ2 , with
A
E = E1 + E2 , such that (E, w1 , w2 ) ∼ ((E1 , w1 ), (E2 , w2 )).
18
On the other hand, there is no problem to warm up a body by an
adiabatic manipulation, see Fig. 6.1.
19
In this case, the thermal join is realized without a change of the
total energy, E = E1 + E2 . That means (see (5.41)) that the weight
does not move. Hence it fulfills the stronger condition of the accessibility,
N
((E1 , w1 ), (E2 , w2 )) ≺ (E1 + E2 , w1 , w2 ).
T
Writing the last relation as (E1 , w1 ) ∼ (E2 , w2 ), we may formu-
late the third axiom concerning the temperature (ibid., p. 56):
T T T
Zeroth law. If X ∼ Y and if Y ∼ Z then X ∼ Z.
When adding five more axioms (ibid., pp. 88,89), we obtain
the group of axioms that implies the validity of the entropy
principle, i.e., guarantees the existence of the entropy function
fulfilling (6.2) for a special class of systems.
The logic of the construction open many interesting ques-
tions. For example, if there is a possibility of reformulating it
for other classes of physical systems. Especially, the construc-
tion of the entropy function for thermomechanical systems, in
this spirit, would be very inviting. The question, however,
is whether it is possible. Namely, is the role of equilibrium
(which defines simple systems and their compositions) crucial?
It opens speculations concerning a relation of the entropy and
time. The fact that time, as a global variable possessing prop-
erties of the generalized temperature, is also defined in “equi-
librium” (see Section 2.6), may be relevant in this context. It
would be also interesting to formulate an analogy of previous
axioms for time ordering, i.e., to find axioms that model the
weak accessibility in clock realization.
Bibliography
[1] G. I. Barenblatt, Similarity, Self-Similarity, and Interme-

diate Asymptotics, Consultants Bureau, New York (1979)
[2] H. B. Callen, Thermodynamics and an Introduction to

Thermostatics, Wiley, New York (1985)
[3] S. Carnot, Réflexions sur la Puissance Motrice de Feu,

Bachelier, Paris (1824)
[4] A. Carpinteri (Ed.), Size-Scale Effects in the Failure

Mechanisms of Materials and Structures, E & FN SPON,
London (1996)
[5] W. A. Day, A Commentary on Thermodynamics, Sprin-

ger, New York (1988)
[6] R. Descartes, Principles of philosophy (1644), transl. by

V.R. Miller and R.P. Miller, Reidel, Dordrech (1983)
[7] A. Einstein, Grundlage der allgemeinen Relativitäts-

theorie, Annalen der Physik 49 (1916) 769-822
[8] Encyclopaedia of Mathematics, Reidel, Kluwer Academic

Publishers, Dordecht, Boston, Lancaster, Tokyo (1988)
164 BIBLIOGRAPHY
[9] R. Giles, Mathematical foundations of thermodynamics,

Pergamon, Oxford (1964)
[10] M. E. Gurtin, W.O. Williams, An Axiomatic Founda-

tion for Continuum Thermodynamics, Arch. Rat. Mech.
Anal., 26 (1967) 83-117
[11] I. M. Havel, Scale Dimensions in Nature Int. J. General

Systems 24(3) (1996) 295
[12] M. Holeček, Scale axis, fractal and some thermal phe-

nomena, J. Thermal Anal. Cal. 60 (2000) 1093-1100
[13] S. C. Hunter, Mechanics of Continuous Media, Ellis Hor-

wood, New York (1976)
[14] S. Kauffman, Investigations, Pergamon Press (2000)
[15] D. Kondepudi, I. Prigogine, Modern Thermodynamics,

Wiley, New York (1999)
[16] J. Kvasnica, Termodynamka, SNTL, Praha (1965) (in

Czech)
[17] L. Kvasz, The Mathematisation of Nature and Cartesian

Physics, Philosophia Naturalis, 40 (2003) 157-182
[18] I. Leclerk, The Nature of Physical Existence, Washington

D.C. (1972)
[19] G. P. Lepage, What is renormalization?, preprint Cornel

University, CLNS 89/970 (1989)
[20] H. Lichtenegger, B. Mashhoon, Mach’s Principle,

http://arxiv.org/abs/physics/0407078 (2004)
BIBLIOGRAPHY 165
[21] E. H. Lieb, J. Yngvason The physics and mathematics of

the second law of thermodynamics, Phys. Rep. 310 (1999)
1 - 96
[22] G. A. Maugin, Material Inhomogeneities in Elasticity,

Chapman & Hall, London (1993)
[23] G. A. Maugin, The Thermomechanics of Nonlinear Ir-

reversible Behaviour, World Scientific, Singapore, New
Jersy, London, Hong Kong (1999)
[24] E. Mendoza (ed.) Reflections on the Motive Force of Fire

by Sadi Carnot and other Papers on the Second Law
of Thermodynamics by E. Clapeyron and R. Clausius,
Glouster, MA: Peter Smith (1977)
[25] W. Muschik, Empirical foundation and axiomatic treat-

ment of Non-Equilibrium Temperature, Arch. Rat. Mech.
Anal. 66 (1977) 379-401
[26] W. Muschik, C. Papenfuss, H. Ehrentraut: A sketch

of continuum thermodynamics, J. Non-Newtonian Fluid
Mech. 96 (2001) 255-290
[27] M. N. L. Narasimhan, Principles of Continuum Mechan-

ics, John Willey & Sons, New York (1992)
[28] Z. Neubauer, God’s Power and Computer Reason (man-

uscript), Praha (2006)
[29] I. Newton, The Principia (1687), transl. by I.B.. Cohen

and A. Whitman, University of California Press, Berkley
(1999)
166 BIBLIOGRAPHY
[30] L. Nottale, Fractal Space-Time and Microphysics, To-

wards a Theory of Scale Relativity, World Scientific
(1993)
[31] D. R. Owen, A First Course in the Mathematical Foun-

dations of Thermodynamics, Springer, Heidelberg (1984)
[32] M. Pitteri, On the axiomatic foundations of temperature,

in: C. Truesdell, Rational Thermodynamics, Springer-
Verlag, New York (1984)
[33] J. Polchinski, Renormalization and effective lagrangians,

Nucl. Phys. B 231 (1984) 269-295
[34] A. Robinson, Non-Standard Analysis, Studies in Logic

and the Foundations of Mathematics, North-Holland,
1966
[35] C. Rovelli, Quantum Gravity, http://www.cpt.univ-

mrs.fr/∼rovelli/book.pdf (2003)
[36] W. Rudin, Real and Complex Analysis, McGraw-Hill, Inc.

New York (1974)
[37] A. Sochor, The alternative set theory and its approach

to Cantor’s set theory, in: H.J. Skala, S. Termini, and
E. Trillas (eds.), Aspects of Vagueness, (1984) 161-203
[38] V. Šedivý: Space, in Man Bang, Univers - Omnivers,

o.p.s., Plzeň, (1998), (in Czech), translation to English:
M. Kotišová (2005).
[39] J. Šesták, Science of Heat and Thermophysical Studies, A

Generalized approach to Thermal Analysis, Elsevier, The
Netherlands (2005)
BIBLIOGRAPHY 167
[40] M. Šilhavý, The Mechanics and Thermodynamics of Con-

tinuous Media, Springer, Heidelgerg (1997)
[41] C. A. Truesdell, S. Bharata, The Concepts and Logic of

Classical Thermodynamics as a Theory of Heat Engines,
Springer, Heidelberg (1977)
[42] C. A. Truesdell, The Tragicomical History of Thermody-

namics, Springer, New York (1980) 1822-1854
[43] C. A. Truesdell, Rational Thermodynamics, Springer,

New York (1984)
[44] A. Tzouvaras, Modeling vagueness by nonstandardness,

Fuzzy sets and systems, 94 (1998) 385-396
[45] P. Vopěnka, Mathematics in the Alternative Set Theory,

Teubner Texte, Leipzig, 1979
[46] K. G. Wilson, Renormalization group and critical phe-

nomena Phys. Rev. B 4 No. 9 (1971) 3174-3205
[47] K. G. Wilson, J. Kogut, The renormalization group and

the expansion, Phys. Rep. C 12 (1974) 75-200
Index
σ-algebra, 107 clock variable, 136

d-dimensional space, 118 comparable states, 156
complement part, 65
absolute continuity, 71 complete decomposition, 38
abstract thermodynamics, 10 complete set of local variables,
accessibility, 148 74
additive variable, 70 composed part, 59
adiabatic accessibility, 153 composition, 27, 36
adiabatic manipulation, 154 composition of parts, 57
adiabatic space, 140 configuration, 103
atomic part, 59
configuration dependent vari-
atomic structure, 59
able, 105
averaged property, 70
conservative accessibility, 150
averaging, 81
conservative property, 48
b-variable, 96 consistence of states and ma-
balance property, 96 nipulations, 40
constitutive laws, 151
c-variable, 96 continuity rule, 76
causality, 44 continuum, 101
caused manipulation, 44 continuum mechanics, 105
clock, 136 continuum model, 102
clock realization, 152 continuum thermomechanics, 87,
clock space, 136 93
INDEX 169
convergence of states, 119 local variable, 68

coordinate system, 118
Mach space, 138
definition of spatial points, 117 manipulation, 7, 11, 35
density of variable, 90 mass, 98
discrete system, 95 measurable mapping, 108
disjunct parts, 57 measure, 108
model, 102
energy, 92 model of a part, 102
entropy, 5, 154, 157 model of space, 125
entropy principle, 157 motion, 106
equilibrium, 51 mutually isolated, 34
equivalent coordinate systems,
124 Newton manipulation, 128
extensive variable, 90 Newton space, 128
non-atomic structure, 59
field, 118
finitely atomic structure, 59 oriented boundary, 94, 98
flux, 97
parallel composition, 37
full state space, 81
parameterization of the process,
generalized lengths, 134 41
global variable, 46 perfect manipulation, 37
physical system, 22, 25
heat, 72, 96 process, 41
homogeneous surrounding, 126 process b-variable, 96
process variable, 72
identity manipulation, 35 production, 97
inner state of the process, 42
intensive variable, 90 referential configuration, 105
isotropic surrounding, 127 representative set of subsystems,
26, 33
l-space, 83 res extensa, 142
170 INDEX
rotation, 126, 127 symmetry operation of space,

124
scale, 55, 62, 68, 81 system, 25
scale complete model, 110
scale symmetric space, 127 tangent of the process, 42
scaled copy, 90 temperature, 91, 160
Schottky system, 95 thermal join, 160
thermomechanical system, 99,
second law, 4, 159
101
set of global variables, 46
time, 129, 153
set of manipulations, 35
time arrow, 152
set of parts, 57
total energy, 140
set of variables, 68
total part, 58
simple system, 92
trajectory, 106
simple unit, 90
translation, 126
simpler surroundings, 45
trivial manipulation, 121
space, 116
spatial field, 118 variable, 68
spatial point, 117
stable state, 120 weak accessibility, 152
state, 32 weakly atomic structure, 59
state set, 32 work variables, 92
state space, 54
state space of a part, 73
state space of the system, 73
stress, 96
structure of mass points, 60
structure of the system, 58
subpart, 58
subsystem, 26
supply, 97
surrounding, 43
Miroslav Holeček
ABSTRACT THERMODYNAMICS
Book work and composition in LATEX Michal V. Hanzelı́n.

Cover design Zuzana Lebedová.
Illustrations Lenka Kaplanová.
O.P.S., Máchova 1492, 288 02 Nymburk (Czech republic);
o-p-s@post.cz
Printed in Typos - Digital Print, Plzeň (Czech Republic)
Plzeň ≡ Nymburk 2006.
First printing, 2006
174 pages
ISBN 80-239-6075-X

Thermodyn

Uploaded by

Copyright:

Available Formats

Thermodyn

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

Thermodyn

Uploaded by

Copyright:

Available Formats

ABSTRACT

PLZEŇ - NYMBURK 2006

c Miroslav Holeček, 2006

I still remember my first reading of a thermodynamic textbook

I realized it very clearly when reading an interesting work

2 States and manipulations 29

3 Systems as sets of parts 53

3.3 Local variables . . . . . . . . . . . . . . . . . . . 67

Isaac Newton laid foundations of modern physics. The main

itself1 . We, however, cannot ignore the extreme usefulness of

Modern physics, however, brings many objections against

difficult to be noticed. We have to attend to of rather concep-

tity, the entropy. It was introduced by R. Clausius (in 1865)

dE = −δW + δQ, (3)

where dE is an infinitesimal change of the energy of the system

Figure 1: What is work? If we calculated a sum of all ‘forces times

Let us illustrate this on the familiar thermodynamic sys-

The piston is connected through a rope with a weight in grav-

formulate an axiomatic foundation of thermodynamics, the au-

Thus thermodynamics presents a small but remarkable con-

convince others about the rightness of their formulation of ther-

modynamic) point is that any change is understood as a result

Namely, special properties of this surrounding (especially its

space includes fixed places (spatial points) allowing to define

The book does not develop the abstract thermodynamics

Figure 2: The diagram outlining main connections of our approach.

Some mathematical symbols

[a, b] an ordered couple of objects a and b

Ed d-dimensional Euclidean space

Symbols of concepts that have no mathematical description

Sets assigned to physical systems and derived concepts

Ξ(S) , Ξ the set of representative subsystems of S

π the process (a subset of Σ)

Symbols describing parts and structures

P(S), P the set of parts

q̂  q̂ 0 absolute continuity of variables

Symbols concerning special structures

v(p) the scaling parameter of extensive quantities

Υ the index set

Symbols concerning space

E the physical system called ‘space’

Symbols concerning accessibility and entropy

If we study any problem we use a language. The more precisely

This concept has no mathematical definition. To say what we

1.1 Two levels of description

many steps by which we shift from an intuitively defined ma-

1.2 Physical systems and subsystems

Some parts of the studied object may be studied as phys-

that has following simple properties, namely

in the abstract level of description where it means just a suit-

The concept of physical system is the basement of any phys-

Any physical description of a real material structure means

and then add (if necessary) some defined forces of interactions

tem’ (that is crucial in any physical description) the concept

2.1 States of physical systems

what we want to study. If, for example, we study only various

The meaning of this rule is obvious: for any couple of different

2.2 Manipulations with physical systems

let alone. In thermodynamics, however, studied systems are

q̂ q̂ 0 absolute continuity of variables