32

EXPERIMENTAL TECHNIQUES

The material in this chapter describing the microwave measurements can also be
found in the article: Instrumentation for millimeter-wave magnetoelectrodynamic
investigations of low-dimensional conductors and superconductors, M. M. Mola, S. Hill,
P. Goy, and M. Gross, Review of Scientific Instruments 71, 186, 2000 [20].

Millimeter-wave measurements

Introduction
Microwave measurement techniques have been used in experimental condensed
matter physics for the better part of the last century. Throughout that time, significant
advances have been made within the field, providing ever more sensitive and ingenious
experiments [21]. The significance of the millimeter and sub-millimeter wave spectral
range is due to the characteristic energy scales of a variety of phenomena which occur
within many of the novel materials pertinent to contemporary physics.

For example

these techniques have been used to probe electron paramagnetic resonance in magnetic
systems [22], cyclotron like studies in novel conducting materials [2], and a slew of
phenomena in the superconducting state including: determination of the penetration depth
[23], energy gap [12], and analysis of collective electronic modes [24]. Due to the varied
uses and power of such a technique, one would think high frequency electron magnetic
resonance (and non-resonance) has become one of the more widely used techniques
available to date. However, as will be shown, microwave measurements are difficult to

33
perform, particularly in the presence of large magnetic fields, which restricts the number
of researchers presently working in this frequency range.
The reasons for this difficulty can be attributed to several factors. The first is a lack of
commercial spectrometers available for this spectral range. Those that are available are
quite expensive, often difficult to use, work over a narrow frequency range, and posses
either low power or unstable output. The second is that spectroscopy at these high
frequencies is complex, particularly in the case of metals, where the high reflectivity
requires extreme sensitivity to small changes in the complex conductivity. The inclusion
of large magnetic fields (> 1 tesla) complicates matters considerably, as the typical bore
size of most magnets [including those at the National High Magnetic Field Laboratory
(NHMFL) and Montana State University (MSU)] severely confines the space with which
to couple the microwave radiation from the spectrometer (source and detector) to the
sample. Although these problems prove difficult, they are by no means impossible to
overcome, and a method will be described here, by which measurement sensitivity is
maximized, while maintaining a broadband spectral range (16 200 GHz) for
investigation of novel materials.

Technical Description

The MVNA. The Millimeter wave Vector Network Analyzer (MVNA) [9, 25, 26]
acts as source and detector, for phase and amplitude sensitive transmission experiments.
The MVNA 8-350, which employs purely solid state electronics, allows measurements
over an extended frequency range (8-350 GHz) through harmonic multiplication of a
tunable centimeter source S1, which provides a nominally flat output power in the range

34
of F1 = 8-18 GHz. The mm-wave signal is extracted from a Schottky diode (harmonic
generator HG) which has been optimized to produce the desired harmonic N of the
sweepable cm source, i.e., Fmm = N F1. Detection is then achieved by mixing the mmwave signal Fmm with the signal from a second cm source S2, at a second Schottky diode
(harmonic mixer HM).

The beat frequency Fbeat, which preserves the phase and

amplitude information of the mm-wave signal relative to the local oscillator S2, is then
sent to a heterodyne vector receiver (VR).
The low noise floor of the MVNA is achieved by defining the frequency difference
between the two cm sources, S1 and S2, using a main oscillator (see figure 17). If one
then uses the same harmonic rank on the source side (HG), and on the detection side
(HM), the phase noise associated with the cm sources cancels in the beat signal Fbeat,
which is sent to the VR. Thus, the phase reference of the VR can be taken directly from
the main oscillator [27].

Figure 17. Schematic diagram of the MVNA, which acts as source and detector
for the millimeter-wave measurements.

With the noise characteristics of the analyzer optimized, its dynamic range is limited
only by the harmonic conversion efficiency of the Schottky diodes. For experiments

35
conducted in this thesis, four pairs of Schottky diodes were used, operating in the K
(~ 16-36 GHz, N = 2 and 3), V (~ 45-70 GHz, N = 3 and 4), W (70-110 GHz, N = 5 and
6), and D bands (~ 110-200 GHz, N = 7-12). The V-band diodes are nominally flat
broadband, while the W and D band diodes are mechanically tunable and require
optimization each time the source frequency is changed. Operating in this mode, it is
possible to perform bench-top tests up to 350 GHz. Table 2 lists the optimum dynamic
ranges achieved at MSU (MVNA 8-350-1-2) in each frequency band, for each harmonic
up to 200 GHz (N=12). Similar analyzers are in use at several high magnetic field
laboratories around the world; in particular, the national magnet labs in the US and The
Netherlands have MVNA 8-350 analyzers (with ESA options) which are available to
external users. The MVNA at the NHMFL, which has been used for some of the studies
discussed in this thesis, additionally operates in the Q (30-50 GHz, N=3), and X (8-18
GHz, N= 1) bands.

Table 2. Dynamic range achieved for various frequencies in each microwave band.
Band, Harmonic
V, N = 3
V, N= 4
V, N= 4
W, N = 5
W, N = 5
W, N = 6
W, N = 6
D, N = 8
D, N = 9
D, N = 10
D, N = 12

Frequency (GHz)
48
60
70
79
88
99
108
119
135
157
186

Dynamic Range (dB)

>128
>128
126.4
105.7
103
94
83.8
83
76.8
65.3
57.8

36
Although the cm sources are phase locked to each other, their absolute frequencies
must also be stabilized. The frequency precision and stability provided by the MVNA 8350 is not adequate for narrow band cavity perturbation measurements when the
bandwidth of the cavity is less than about 100 MHz. For this reason, it is common to
phase lock one of the sources (i.e. both) to a quartz standard. To this end, EIP 575 source
locking frequency counters [28] are used, which provide both the stability and precision
necessary for the measurements described in this thesis. One other mode of operation
involves phase locking the source S1 directly to the high Q cavity resonator used in the
experiment.

The counter is also useful in this case for recording changes in the

dispersion within the cavity (resonance frequency).

Comparisons between the two

frequency locking techniques are discussed in detail shortly.

The dynamic ranges listed in table 2 represent ideal values for the MVNA 8-350
assuming no insertion losses between HG an HM. Due to the considerable size of a
typical magnet cryostat and, in particular, the dimension separating the magnetic field
center and the top of the cryostat (~ 1.5 m), a considerable insertion loss is unavoidable.
Furthermore, it is essential to keep a reasonable distance (~ 2 m ) between the MVNA
and the magnet, since the cm sources (YIG oscillators) are sensitive to stray magnetic
fields of more than a few Gauss. The Schottky diodes (HG an HM) on the other hand,
are not field sensitive and may be placed closer to the experiment. Thus the cm and beat
frequencies are propagated between the MVNA and the Schottky diodes through flexible
low loss coaxial cables, which introduce a combined insertion loss of 4-5 dB.
Meanwhile, the mm-wave signal is propagated from HG to the cavity within the magnet
cryostat, and back to HM, using a pair of rigid waveguides. These waveguides account

37
for a major part of the insertion loss of the system. Minimizing these losses is the subject
of further sections of this thesis. A schematic of the experimental arrangement is shown
in figure 18.

Figure 18. The mm-wave experimental apparatus for laboratory fields. The
dewar used at the NHMFL is slightly modified to fit into the bore of the resistive
magnets.

38
In addition to its use as a source and detector for solid state spectroscopy, the network
analysis capability of the MVNA is crucial during the development stages of a
measurement system. The frequency sweeping capability can be used to determine the
precise location of impedance mismatches and/or microwave leaks from the system, and
for the characterization and optimization of different resonator designs.

The Sample Probe. The primary goal of this particular experimental setup is to be
able to conduct sensitive cavity perturbation measurements, at low temperatures and high
magnetic fields over a broad frequency range. Here, an experimental scheme will be
described - the sample probe - for coupling the MVNA to various enclosed cylindrical
and rectangular resonator configurations, which can easily be inserted into a high field
magnet cryostat. The sample probe refers to the passive microwave hardware which has
been developed, and is compatible with, both laboratory and NHMFL magnet systems. It
consists of the following components: two long waveguides for propagating mm waves
from the HG, into the magnet cryostat (incidence waveguide), and back to the HM
(transmission waveguide); a demountable cavity, of which two standard types have been
developed; a coupling between the cavity and the incidence and transmission
waveguides; a vacuum tube to isolate the probe from the surrounding liquid cryogens;
and electronics for controlling the sample temperature, and for magnetic field
modulation. Figure 18 shows a schematic of the probe situated within the laboratory
superconducting magnet cryostat.
Two key experimental objectives must be established: (1) it should be possible to
make measurements at several well separated frequencies without the need to interfere

39
with the sample probe, or to remove it from the magnet cryostat, and (2) the sensitivity,
dynamic range, and mechanical stability of the measurement system should be optimized
as far as possible. The flexibility in choice of frequency is important for experiments at
the NHMFL, where magnet time is often limited. There are also other advantages to this,
e.g., by not having to interfere with the sample, it is possible to ensure that it sits in an
identical electromagnetic field distribution for a given series of cavity modes. The
second goal is fairly self-explanatory; nevertheless, high sensitivity comes at the expense
of some flexibility. To this end, a rigid construction has been chosen, i.e., no in situ
rotation of either the cavity or the sample within the cavity is possible. Such mechanisms
tend to result in radiation leakage from the cavity and diminishing the sensitivity.
The success of recent measurements owes as much to the reductions in the insertion
losses associated with the sample probe, as it does to the optimization of the cavity
designs. In this section, the technical details of how to propagate mm-wave radiation
from the Schottky diodes into a low temperature cryostat (pumped 4He) within the bore
of a high field magnet system where the resonant cavity containing the sample is situated,
is discussed. As illustrated in figure 18, the laboratory setup utilizes a separate inner
cryostat to control the temperature of the experiment. This cryostat, which sits within the
bore of the superconducting magnet, draws liquid helium from the main reservoir into its
tail, which is thermally isolated from the main helium reservoir, thus allowing good
control over the sample temperature.
The microwave probe is based on V-band rectangular waveguides, which are
terminated at their source/detector ends with UG385/U flanges [29] for convenient
connection to the Schottky diodes. V-band waveguide cuts off below ~ 45GHz, which is

40
sufficiently low for most of the intended applications. However, in the rare cases where
low frequencies are required, a second probe of similar design has been manufactured
from X-band, rectangular waveguide. This second probe allows for measurements in
both rectangular as well as cylindrical cavities in the frequency range of 8 - 40 GHz.
Note that, due to the small bore of the high field magnets at the NHMFL, this second
probe is not compatible with these magnets and is, therefore, used only within the magnet
at MSU.
Commonly used low loss, high conductivity, waveguide materials such as copper or
silver are not ideal for the entire length of the sample probe. Unacceptable heat flow
down the waveguides and into the cryostat leads to excessive liquid helium boil off and
limits the ability to cool the cavity/sample to pumped 4He temperatures (~1.5 K). For
cryogenic purposes, stainless steel (SS) waveguide [29, 30] offers an attractive
alternative: it has both a low thermal conductivity [31] and it is possible to use thinner
walled waveguide material [32]. Unfortunately, the microwave losses in SS waveguide
are severe, as illustrated in figure 19, which shows a comparison between the insertion
losses for SS and Ag, as measured using the frequency sweeping capability of the
MVNA; note that, in this range, the losses of SS exceed 10 dB/m, whereas the loss in Ag
is less than 2 dB/m [33]. The solution, therefore, is to construct composite waveguides
from highly conducting waveguide sections isolated by SS sections.

Although this

necessitates several joints in the waveguide assembly, these do not significantly impair
the functionality of the waveguide.

41

Power loss (dB/m)

-2
-4

S ilv er

-6
-8
-10
-12

S tain less steel

-14
-16

52

54

56

58

60

62

64

F req u en cy (G H z)

Figure 19. A comparison between the insertion losses for stainless steel and silver
V band waveguides at room temperature.

Placement of the SS waveguide sections is crucial to the cryogenic performance of the

system, as illustrated in figure 18. The largest temperature gradient occurs in the upper
part of the magnet dewar, where the radiation baffles are situated. It is in this region that
SS is first used. Short SS sections are again used at the lower end of the probe in order to
thermally isolate the cavity/sample from the long Cu/Ag sections, which run from just
below the baffles, through the liquid 4He reservoir, and into the bore of the magnet. The
cavity and sample are, therefore, more-or-less isolated from the main 4.2 K helium
reservoir, enabling control over the temperature of the cavity/sample from about 1.35 K,
up to 40 or 50 K. The lower SS waveguide sections have been gold plated [34] for
reasons which are discussed in the following sections.
An unforeseen benefit of the composite waveguide construction, which greatly
improves the mechanical stability of the probe, is its insensitivity to the liquid helium

42
level in the magnet cryostat. Assuming that the Cu/Ag sections maintain a fairly uniform
temperature (~ 4.2 K) over their entire length, due to their high thermal conductivities,
the temperature gradient in a single SS section does not depend on the liquid helium
level. In contrast, the thermal gradient in a single SS section immersed in liquid helium
would depend on the level of the liquid, causing the waveguide to expand slightly over
time, thereby affecting the microwave phase stability of the system.
The SS and Cu/Ag rectangular waveguide sections, which have different outer
dimensions [32], are coupled together using specially machined clamps, which screw
tightly around the waveguides. The joints are staggered so as to minimize crosstalk
between the waveguides due to any microwave leakage from the joints. Thus, although
both waveguides pass through each clamp, only one of the waveguides is joined at each
clamp. This construction is extremely rugged, yet it easily permits modification in the
overall probe length, and in the placement of SS sections.

Therefore, the same

waveguide assembly can easily be reconfigured for use at other magnet facilities such as
the NHMFL.
Finally the entire sample probe fits tightly inside a 0.750 inch outer diameter (0.020
inch wall) vacuum jacket.

The integrity of the vacuum is maintained across the

waveguides at their room temperature ends using Mylar windows clamped between
standard flanges (UG385/U) which have been modified to hold rubber O-rings. Again,
the vacuum seals on the incident and transmission waveguides are offset, and then
wrapped with steel wool to minimize crosstalk.

A hermetically sealed connector

provides 19 electrical feedthroughs for thermometry, field modulation coils, and for
powering a heater.

43
The basic principle behind the cavity perturbation technique is relatively simple, and
involves measuring changes in the complex cavity parameters (i.e. resonance frequency
fo and Q value) upon insertion of a sample [35]. Relating these changes to the complex
electrodynamic properties of the sample can be a formidable task. Therefore, the reader
should refer to a series of three articles [36, 37, 38] dealing with this analysis for
precisely the types of materials which are studied in this thesis, i.e., highly anisotropic
crystalline conductors.

However, in order to reliably apply such an analysis, it is

essential that one first consider how each element in the sample probe effects ones
ability to extract the relevant information from the cavity mounted at the end of the
probe. This is a formidable task in itself, which is complicated considerably by the
restricted access into the bore of a typical high field magnet cryostat, and by the need for
the probe to accommodate cavities resonating over a broad frequency range.
The task of optimizing the probe design is best tackled by considering an equivalent
ac circuit (figure 20). Each component of the probe may be modeled as a self contained
LRC circuit which is inductively coupled to the next. The Schottky diodes, HG and HM,
attach to the upper ends of the incidence and transmission waveguides, which have
impedance ZI and ZT, respectively. The coupling between HG (HM) and the incidence
(transmission) waveguide is modeled as a coupling mutual inductance mHG (mHM). The
cavity, which is mounted at the lower end of the two waveguides, and has impedance ZC,
is coupled to each of the waveguides through coupling mutual input (min) and output
(mout) inductances.

Ideally, the incidence and transmission waveguides should be

coupled solely through the cavity. However, a leak signal between the incidence and

44
transmission waveguides, in parallel with the cavity, is inevitable; this is modeled as a
direct coupling mutual inductance ml between the two waveguides.

Figure 20. An equivalent LRC circuit of the microwave components of the

sample probe. The coupling between various components is modeled as a mutual
inductance.

It is changes in ZC, caused by the insertion of a sample into the cavity, which one
would like to measure in a cavity perturbation experiment. Since the MVNA measures
phase and amplitude, it is important to see how ZC affects both of these parameters.
Before considering the probe as a whole, i.e., the entire circuit in figure 20, one must first
examine the cavity by itself. Variations in the amplitude and phase of a wave transmitted
through the cavity, as the frequency f is swept across a resonance, are given by

A 2 (f ) =

(2.1)

2(f f o )
(f ) = arctan
,

(2.2)

1 + [2 (f f o ) ]

45
where fo is the center frequency and is the full width at half maximum (FWHM) of the
resonance. These expressions have been normalized so that A = 1 and = 0 at resonance.
As f is swept from 0 to , A and f sweep out a circle in the complex plane, where the real
and imaginary amplitudes are given by A1(f) = Acos and A2(f) = Asin, respectively, or
A 1 (f ) =
A 2 (f ) =

1 + [2 (f f o ) ]

2 (f f o )

1 + [2 (f f o ) ]

(2.3)

(2.4)

the more familiar expressions for a Lorentzian. Such a circle is shown in figure 21a, for a
resonance at 60 GHz, with a Q of 5000. Each point in the figure corresponds to a
different frequency, and the frequency interval between each point is 800 kHz. Points
closest to the origin correspond to f << 60 GHz and f >> 60 GHz. The resonance
frequency, fo, lies along the real axis, = 0, where spacing between points is greatest, i.e.,
when the phase rotates the most rapidly with frequency. Clockwise rotation around the
circle represents increasing frequency. The corresponding changes in A(f) and (f) and
A1(f) and A2(f) are shown in figures 21b and 21c respectively.
In a locked frequency experiment, changes in the dispersion (Im{ZC}) cause the cavity
to go off of resonance, i.e., the point corresponding to fo in figure 21a will move away
from the real axis. Provided that this effect is weak (a perturbation), the dominant result
is a change in the phase of the signal transmitted through the cavity, and no appreciable
change in the amplitude. Conversely, changes in dissipation (Re{ZC}) within the cavity
will result in a reduction in the amplitude of the signal transmitted through the cavity and
hence, a reduction in the diameter of the circle in figure 21a. Dissipation alone does not

46
move fo away from the real axis and, therefore, does not effect the phase of the wave.
However, dispersion can affect both the amplitude and the phase of the wave if fo moves
appreciably away from the real axis. For this reason, it is often desirable to conduct a
phase locked experiment, which completely decouples these two effects. With a phase
lock, the cavity stays on resonance [on the real axis in figure 21a], the dispersion affects
the resonance frequency only, which one can measure with a frequency counter.
Meanwhile, changes in dissipation again effect the amplitude of the transmitted signal
only.
90
120

Phase

a)

150

30

1
180
0.0
0.5

1.0

1.0

210

330
240

270

fo

300

100

0.5

50

0.0

-0.5

b)

Phase

Normalized amplitude

60

-50

1.0
0.5

A 1 (f)

c)

0.0
-0.5
59.95

A 2 (f)
60.00

60.05

Frequency (GH z)
Figure 21. Simulation of the complex parameters of a wave transmitted through a
cavity, with a resonance at 60 GHz, and Q = 5000. a. Linear amplitude and phase
versus frequency in the complex plane. b. Linear amplitude and phase versus
frequency, and c. A1 and A2 linear amplitudes versus frequency.

47
Unfortunately, each of the additional circuit elements required to link the MVNA to
the cavity, and/or improper coupling between these components, has the potential to
seriously distort the simple relationships between dissipation, dispersion, amplitude, and
phase discussed above. Ideally, the sample probe should be passive, low loss, insensitive
to temperature and magnetic field and, with the exception of the cavity, should have a
flat, broadband frequency response. In practice, this is never actually possible to achieve.
Nevertheless, by conducting a thorough characterization and optimization of each
element in the microwave circuit (figure 18), it is possible to minimize these instrumental
effects to negligible levels.

The MVNA performs a pivotal role in this hardware

development process.
A leak wave bypassing the cavity directly through to the transmission waveguide has
two adverse effects. First it diminishes the useful dynamic range - ideally 100% of the
signal reaching the detector should pass through the cavity. Second, if the leak amplitude
is comparable to the amplitude of the signal passing through the cavity, the resonance
may become severely distorted, making it extremely difficult to distinguish between
dissipative and dispersive effects within the cavity. As illustrated in figure 22, a leak
wave adds a complex vector to the signal transmitted through the cavity. By minimizing
the leak, one can control its amplitude. However, it is not possible to control the phase of
the leak wave. Consequently, big leaks lead to an arbitrary vector translation of the circle
in figure 21a. This is a pure translation, i.e., the same vector translates each point on the
circle. Hence, the line joining the resonance frequency fo and the f = 0 and points,
remains parallel to the real axis. As a result, the transmitted amplitude on resonance is
not necessarily the maximum amplitude; indeed, it can take on any value from zero to

48
one plus the leak amplitude. This is illustrated in figure 22 for an arbitrary translation of
the circle in figure 21a, together with the corresponding variation in amplitude and phase,
plotted versus frequency. It is apparent from this figure that the phase of the leak signal
is entirely responsible for the way in which the dissipation and dispersion affect the phase
and amplitude of the signal transmitted through the cavity. Thus, an appreciable leak
signal is intolerable, and one must take every step possible to reduce the leak in the probe
to at least 20 dB below the typical signal transmitted through the cavity on resonance, as
discussed above.
90
60

120

Phase

a)

150

0.0

Phase

Amplitude

1.0

270

fo

180
1
0.5

1.0
0.8
0.6
0.4
0.2
360

30

210

0
330

240

270

300

fo
b)
c)

180
90
0
59.95

60.00

60.05

Frequency (GHz)
Figure 22. The effect of a leak wave on the ideal resonance in figure 21. a. The
resonance frequency, fo, no longer lies on the real axis, thus, fo no longer
corresponds to the maximum amplitude, resulting in a distortion of the Lorentzian
line shapes in b. and c.

49
Standing waves in the waveguides are unavoidable and, without proper attention, can
cause considerable problems, especially when operating in the phase-locked mode in
which the incident mm-wave frequency is locked to the cavity resonance frequency.
Changes in this frequency will result in changes in the phase and amplitude of the mmwave incident upon the cavity. Thus, it becomes impossible to distinguish between the
intrinsic cavity response and the spurious effects due to the standing waves.
Furthermore, the phase is no longer truly locked to the cavity resonance under these
circumstances, but rather to the coupled response of the entire circuit in figure 20.
Standing waves should not be ignored altogether in the frequency locked mode either,
particularly when the cavity is well coupled to the waveguides.

Under these

circumstances, changes in ZC influence the impedance matching between the cavity and
the waveguides (i.e., min and mout) and therefore, affect the standing waves.
There is little to be gained from trying to eliminate the standing waves completely.
This would require precise impedance matching of each component in figure 20, which is
only possible to achieve over a narrow frequency range and would, therefore, defeat the
purpose of the probe, which is intended to work over a fairly broad frequency range.
Instead, one must concentrate on minimizing the influence of the standing waves on a
measurement at any given frequency.

This is achieved by reducing the frequency

bandwidth of the measurement to well below the periodicity of the standing wave pattern,
so that the response of the waveguide is essentially flat over the relevant frequency
interval. In this way, a range of cavities or cavity modes may be utilized, covering an
extremely broad frequency range in comparison to the standing wave periodicity.

50
Meanwhile, each cavity mode samples only a minute portion of the waveguide spectrum,
over which its response is essentially flat.
The fastest standing wave period is governed by the longest dimension of the probe,
which is about 3 m in this case (HG to HM, via the cavity). This gives rise to a standing
wave periodicity of about 100 MHz which, in turn, requires cavity filling factors (volume
of sample/volume of cavity) of less than 10-4, so that the frequency shift in any given
measurement never exceeds about 10-4 of the measurement frequency, i.e., fo < 10 MHz
for f < 100GHz. Cavity filling factors of between 10-5 and 10-4 are typical for the types
of samples studied here, so standing waves do not pose a problem in this measurement
regime. Nevertheless, to compensate for the small filling factors, it is essential to have Q
factors on the order of 104. In order to attain such high Q values, the cavity must
necessarily be coupled weakly to the waveguides, which further reduces problems
associated with standing waves.
Phase instabilities, and/or a strong frequency dependence of the phase reaching the
VR, can lead to a variety of problems. Although the internal sources within the MVNA
are phased locked, additional phase jitter can arise in the mm-wave signal due to
mechanical and/or thermal instabilities in the sample probe. Mechanical vibrations do
not pose any problems either at MSU or at the NHMFL, due to the tight fit between the
sample probe and the SS vacuum jacket. However, thermal stability is essential for
achieving the best results. Due to extremely high cavity Q values, minor temperature
fluctuations can lead to significant phase instabilities, particularly when studying samples
with a strongly temperature dependant electrodynamic response. For this reason, active
temperature control is necessary.

Long-term thermal stability of the entire probe,

51
including the 1.5 m waveguide sections, is also desirable and is best achieved using the
composite waveguides described in the previous section.
A strong frequency dependence of the phase reaching the VR occurs when the
distances between the sources, S1 and S2, and the HM are very different. The signal
originating at S1 has to travel the extra 2 1.5 m into, and out of, the magnet cryostat.
For this reason, it can be beneficial to compensate for the extra distance between S1 and
HM by adding 3 m of coaxial cable between S2 and HM, especially when performing
phase locked or frequency swept measurements, even though this introduces an extra 2-3
dB insertion loss.
Spurious magnetic resonances caused by paramagnetic contamination of either the
cavity, or the waveguides close to the magnetic field center, will give rise to both sharp
and broad instrumental features [electron paramagnetic resonances (EPR)] in the
transmission versus magnetic field response of the system. These spurious resonances
can be hard to distinguish from the genuine response of the sample within the cavity, and
should be eliminated so that the magnetic field response of the unloaded probe (i.e., with
no sample) is as flat as possible. The SS waveguide sections cause the most severe
problems, which is likely due to the presence of small traces of Fe3+ (rust) at the surface
of the metal. For this reason the SS sections, which couple directly to the cavity, have
been gold platted. This not only reduces the insertion loss due to these sections, but also
completely eliminates a broad instrumental resonance, leaving an extremely flat response
(A < 0.05 dB), as shown in figure 23. A number of other measures have been taken to
avoid contaminating the lower end of the probe (i.e., closer to the field center) with
paramagnetic impurities. These include clamping all components together rather than

52
using adhesives or solders, and avoiding the use of cutting oils when machining the

Change in Amplitude (dB - offset)

components of the electrolytic Oxygen-free high-conductivity (OFHC) Cu resonators.

0 .2
0 .1
0 .0
-0 .1
A u p lated [C M I]

-0 .2

A u p lated (M S U )
-0 .3

U n p lated
0

M agn etic field (tesla)

Figure 23. Comparison between the empty cavity response of the sample probe
using Au plated and unplated SS sections of waveguide for the lower end of the
probe. Custom Microwave Inc. (CMI) [34] professionally plated the sections,
currently used in the probe.

The cavities. To date, most efforts have focused on the use of the enclosed cylindrical
resonators for cavity perturbation measurements. However, rectangular resonators have
also been used at very low frequencies (16-40 GHz) in the K band probe. There are
several reasons to use such simple, high symmetry resonators. To begin with, the simple
design concept, in which the cavities are assembled from relatively few easily machined
components (see figures 24 and 25), enables one to fabricate a range of cavities and,
therefore, switch frequency range/coverage with relative ease. More importantly, relative
to rectangular cavities, it is strait forward to achieve extremely high Q values (>104) for
the TE01n (n = 1, 2, 3) modes of cylindrical resonators. The reason for this has to do
with the fact that no ac currents flow between the end and the sidewalls of a cylindrical
cavity excited in the TE01n modes. Consequently, joints at these ideal locations do not

53
diminish the cavity Q factor. These modes, also posses electromagnetic field geometries
which are highly desirable for the measurements described in this thesis.

Figure 24. Schematic diagram of the axial cavity construction.

For the most part, axial cavities of the type shown in figure 24, i.e., ones in which the
cavity axis is coincident with the axis of the superconducting solenoid are used. This is
by far the most versatile design, although it does have some limitations, especially when
it is necessary to tilt the sample with respect to the applied dc magnetic field. The cavity
is constructed with OFHC electrolytic copper in three pieces: a blank end plate, a
coupling plate, and a cylindrical barrel. This assembly bolts onto the under side of a
copper housing which clamps around the incidence and transmission waveguides. These
bolts also provide the pressure for reproducibly clamping the cavity assembly tightly
together. No adhesives or solders are used. The cavity may be disconnected from the
waveguides, in a matter of seconds for easy sample insertion/removal. The housing,

54
which clamps around the waveguides is also easily interchanged with a separate housing
for a transverse cylindrical cavity (see below).
Coupling between the waveguides and the cavity is achieved by means of small
circular apertures in the thin coupling plates. Since these plates terminate the waveguides
and the cavity, there are no transverse microwave electric (EAC) fields at the locations of
the apertures. Thus, it is the microwave magnetic (HAC) fields in the waveguides and the
cavity, which should be matched. For the TE01n modes, the HAC fields flow radially at
the cavity ends and, therefore, the HAC field in the waveguides should do so also. For this
reason, the incoming waveguides are oriented with their shortest edges closest together
(see figures 24 and 25) [39].
One can control the degree of coupling between the waveguides and the cavity (min
and mout) by means of the dimensions of the coupling apertures (diameter and thickness).
There is an important trade off here between: (i) strong coupling (large apertures), which
ensures good power through put from the source to the detector and, hence, a large
dynamic range, and (ii) weak coupling (small apertures), which limits the radiation losses
from the cavity, resulting in a higher cavity Q values and increased sensitivity, at the
expense of some dynamic range.

It has been found empirically that the optimum

coupling apertures should be small for this setup, with a diameter ~ mm/4.25. It is also
necessary for the coupling plate to be very thin (~ mm/20), since the signal is obviously
attenuated as it passes through the apertures, which are way below cutoff. The relatively
weak coupling to the cavity makes it all the more important to reduce any/all other losses
in the sample probe and, hence, preserve a reasonable dynamic range, i.e., without taking

55
steps to reduce losses in the waveguides, one would be forced to increase the coupling,
resulting in reduced sensitivity.
The resonance frequency of a particular cavity is determined by the length and inner
diameter of the cylindrical barrel. Because of the simplicity of machining this section, it
is no great task to construct a large number of cavities, providing many TE01n modes
covering a wide range of frequencies. However, because of the wavelength dependence
of the coupling between the waveguides and the cavity, it is necessary to construct
several different coupling plates - roughly one for each frequency band, i.e., V, W, D, etc.
Because the inner diameters of the cavities generally get smaller at higher frequencies,
the positioning of the apertures is also critical. In all cases, the input and output coupling
apertures are located diametrically opposite each other, and at the same radial distance
from the axis of the cavity. In order to minimize the number of these coupling plates, the
cavity diameters have been limited to four standard sizes. For V-band (cavity diameter =
9.52 mm), the coupling apertures are located in the optimum positions, both with respect
to the cavity and the waveguides, i.e., in the middle of the waveguides and halfway
between the cavity axis and its perimeter. For the higher frequency bands, a compromise
between these positions is made.
A clear gap of 0.020 inches is maintained between the incidence and transmission
waveguides, which terminate flush with the under side of the cavity housing. The
coupling plate is machined with a 0.020 inch wide and 0.030 inch high ridge running
perpendicular to the line joining the coupling apertures, and intersecting the mid point
between them (see figure 24). This ridge locates inside a matching groove on the under
side of the cavity housing and, therefore, in between the incidence and transmission

56
waveguides. It has been found that it is this joint between the cavity and the waveguides
which is most susceptible to microwave leaks, and that the ridge in the coupling plate
dramatically reduces the leak amplitude.

The leak may be reduced further still by

applying a small amount of indium into the groove before assembling the cavity.
For a perfectly cylindrical cavity, the TE01n modes are degenerate with the TM11n
modes. It is a trivial task to lift this degeneracy without diminishing the Q values of the
TE01n modes. One achieves this by drilling a small indent in the center of the blank end
plate where essentially no currents flow in the TE01n modes. This hole may also be used
to hold a dielectric pillar for mounting a sample on the axis above the cavity end plate
(see figure 26b). Consequently, the TM modes are shifted to lower frequencies by up to
200 MHz; they are also weaker and have lower Q values than the TE modes.

Figure 25. Schematic diagram of the transverse cavity construction.

A transverse cavity optimized to work in the V-band frequency range has also been
developed, and is shown in figure 25. Although the construction of this cavity is a little

57
more complex than the axial cavities, it offers a major advantage for experiments in
which it is necessary to tilt the sample relative to the applied dc magnetic field. The
major difference between this cavity and the axial one is that the mm waves are coupled
through apertures drilled directly through the sidewall of the cylindrical barrel, i.e., a
separate coupling plate is not employed. This requires removing material from one side
of the cylindrical barrel so that the sidewall of the cavity is sufficiently thin in the vicinity
of the coupling apertures. This flat surface also facilitates attachment to the underside of
a specially designed cavity housing, which again clamps to the incidence and
transmission waveguides. The cavity end plates simply bolt onto either end of the cavity
barrel. One of these end plates has machined slots at the bolt circle radius so that it may
be rotated about the cavity axis.
As with the axial cavity, a ridge/groove arrangement between the waveguides and the
cavity is implemented in order to minimize any microwave leakage at this position. The
locations of the coupling apertures in the sidewall of the cavity have been optimized for
exciting the TE012 mode, i.e., at h and h from the ends of the cavity, where h is the
cavity height. Nevertheless, these positions provide good coupling to the TE011 mode
and to a lesser extent, the TE013 mode.
Using one or the other of the two cavity types, it is possible to subject a sample to
virtually any combination of EAC and HAC field polarizations, relative to the applied DC
magnetic field Bo orientation, e.g., at an HAC node with EAC||Bo, or at an EAC node with
HACBo, etc. However, it turns out that for studies of highly anisotropic conductors, two
convenient locations are sufficient for most experiments utilizing the TE01n modes [35,
37]. In the end plate configuration, the sample is placed on the blank cavity end plate,

58
exactly half way between the cavity axis and its perimeter, as shown in figure 26a. In the
dielectric pillar configuration, the sample is mounted atop a thin dielectric pillar
(usually quartz or sapphire, diameter = 0.71 mm) on the axis of the cavity, as shown in
figure 26b. In both of these configurations, the sample sits in an H-field antinode for the
TE011 mode, the polarization of which is radial for the endplate configuration and axial
for the dielectric pillar configuration.

Figure 26. a. The endplate configuration: the sample experiences a transverse

HAC. b. The pillar configuration: the sample experiences an axial HAC.
For the end plate configuration, all TE01n modes have radial H-field antinodes at the
sample location. This is not the case for the dielectric pillar configuration. For example,
if the sample is mounted precisely midway between the cavity endplates, the even n
modes have both E- and H-field nodes at this location. Careful forethought as to the
positioning of the sample can rectify this problem to a certain extent.

However,

whenever positioning the sample away from the midpoint of the cavity, sensitivity is
compromised. Indeed, even the end plate position is appreciably less sensitive than the
cavity midpoint [37].
To understand how it is that the end plate and dielectric pillar locations within a
TE01n cavity can be sufficient for studying anisotropic systems, consider the

59
electrodynamics of a quasi-two-dimensional (Q2D) conductor.

Using a quasi-static

approximation, consider the Faraday field, EF, resulting from the time varying magnetic
field, HAC, i.e., an oscillatory EF field which curls around the polarization of the HAC
field. In an isotropic conductor, this Faraday field would induce circulating currents in a
plane perpendicular to the HAC field. Indeed, this is approximately what happens if the
HAC field is polarized perpendicular to the highly conducting planes of a Q2D conductor,
as illustrated in figure 27a. However, because of the high in-plane conductivity, these
induced currents are damped within the interior of the sample, i.e., currents only circulate
within the skin layer at the edge of the sample. Consequently, the in-plane complex
surface impedance, ZS = RS + iXS, governs the electrodynamic response of the sample in
this situation, where RS and XS are the surface resistance and reactance, respectively [36,
37, 38]. If, instead, the HAC field is polarized parallel to the highly conducting layers, EF
will induce both in-plane and interlayer currents. As before, the in-plane currents will
only flow at the edges of the sample, whereas the interlayer current flow throughout the
bulk of the sample, due to the poor conductivity in this direction. Consequently, the
complex interlayer conductivity dominates the electrodynamic response of the sample
under these conditions, as depicted in 27b [36, 37, 38].

a.

b.

Figure 27. For Q2D samples with the low conductivity axis aligned with the axis
of the cavity (indicated with an arrow), a. an axial HAC will drive currents within
the highly conducting planes, b. while a transverse HAC will drive current both in
the planes as well as along the least conductive direction.

60
Of major interest for studies of low-dimensional conductors is the possibility of
rotating the applied DC magnetic field Bo with respect to the sample. Since the cavities
are rigidly connected to the sample probe, the orientation of Bo cannot be adjusted
relative to the cavity, i.e., Bo is fixed parallel to the cavity axis for the axial cavity and
perpendicular to the axis for the transverse cavity. Thus, the axial cavity essentially
limits the investigation to only two orientations of Bo relative to the sample, namely
parallel or perpendicular to the Q2D layers, which we denote B|| or B, respectively. One
could tilt the sample away from either of these geometries within the cavity. However,
this would result in a misalignment of the HAC field polarization relative to the sample.
This is the main reason for developing the transverse cavity. Because of the cylindrical
symmetry of the TE01n modes, one can rotate the sample about the cavity axis by means
of rotating the cavity end plate to which it is attached, without effecting the polarization
of HAC relative to the sample. This works for both the end plate and dielectric pillar
configurations as illustrated in figure 26.
Being constructed from a block of high conductivity copper, the cavity makes an
excellent heat reservoir for controlling the temperature of the sample, which is kept in
good thermal contact with the cavity at all times, either by directly attaching it to the
cavity end plate, or by mounting it on a dielectric (quartz or sapphire) pillar. Silicone
grease is used to hold the sample in place. A coil of high resistance wire is used as a
heater. This coil is wound around a pillar which screws into the cavity housing and is,
therefore, easily interchangeable between various cavity geometries.

Low-pressure

helium gas is admitted into the 0.750 inch vacuum jacket for exchanging heat between
the sample probe and the surrounding pumped liquid helium cryostat. Thus, the cavity

61
and the sample can accurately and controllably be maintained in any temperature range
from 1.35 K up to 50 K. The temperature is stabilized according to a calibrated Cernox
resistance thermometer embedded in the walls of the copper cavity. Cernox thermometers
have negligible magnetoresistance above 4.2 K, and one can correct for a weak
magnetoresistance below 4.2 K.

Laboratory magnetic fields are generated using a

standard commercial superconducting solenoid and power supply. The magnet routinely
operates to 8 T, and will go to 9 T at pumped liquid helium temperatures. At the
NHMFL, the strongest magnetic fields are produced in axial Bitter-type, water cooled
resistive magnets, powered by 20 MW supplies. The highest field used in these studies
was 33 T. Details of the NHMFL magnets are published elsewhere [40].

Tests
Figure 28 shows a frequency sweep across the TE011 mode of an axial cavity (cavity
A, height= 6.7 mm, diameter= 9.52 mm, see table 3). The cavity is loaded with a sample
of -(ET)2Cu(NCS)2 (approximate dimensions: 0.5 0.5 0.2 mm3, see chapters three
and four) in the endplate configuration, and the temperature is 4.2 K. In the upper panel,
linear amplitude versus phase is plotted. The data points form a perfect circle passing
through the origin, indicating a negligible leak vector. The frequency interval between
each point is approximately 250 kHz and the solid line is a fit to the data. In the lower
panel, both the phase and linear amplitude (normalized) are plotted versus frequency.
This resonance is perfectly symmetric due to the fact that the leak amplitude is 34.5 dB
below the amplitude on resonance. The loaded Q value of the cavity is 19,000 and, thus,
the resonance width (2.34 MHz) is considerably less than the standing wave period,

62
which is on the order of 100 MHz. The absolute value of the phase returned by the VR is
arbitrary, which is why the phase on resonance is 64o rather than 0o. In any subsequent
experiment, we would null the phase on resonance, and interpret changes in the complex
parameters of the signal returned to the VR according to the procedure described earlier.
It should be noted that these data are about as good as one could expect were the cavity
mounted on the bench top and the HG and HM connected directly to the cavity, i.e., the
influence of the intervening waveguides has been completely eliminated.

Phase
90
60

120

30

150

0
120

0.4

90

0.9

0.6
0.80.6

60

0.3

30

0.0

44.440

44.445

44.450

Phase (degrees)

Normalized amplitude

180
0.21.2

44.455

Frequency (GHz)

Figure 28. A real TE011 resonance for a loaded axial cavity. The upper panel
shows the resonance circle in the complex plane. The squares are raw data and
the line is a fit. The lower panel shows the linear amplitude and phase variation as
a function of frequency.

Next, consider the influence of the nearby TM111 mode on the measurements made
on the TE011 resonance frequency. Figure 29 shows two such resonances obtained at
liquid helium temperature (cavity A): the main part of the figure plots linear amplitude
versus frequency, and the inset shows the circles in the complex plane obtained for each

63
of the resonances. The TM111 mode has been shifted 230 MHz below the TE011 mode,
which corresponds to almost 100 times the width of the TE011 mode [(TE011) = 2.42
MHz] and about 40 times the width of the TM111 mode [(TM111) = 5.73 MHz]. The
resonance amplitude of the TM111 mode is about 60% of the TE011 resonance
amplitude. However, more importantly, the power of the TM111 signal (obtained by
extrapolation of a Lorentzian fit) is 44.5 dB below the power of the TE011 signal when
the TE011 mode is at resonance. Thus, for all intents and purposes, one can rule out any
interference between these two modes. Even if there were a slight mixing, both modes
have HAC fields perpendicular to the applied DC field Bo for the endplate configuration,
and the TM111 mode has an HAC field node at the center of the cavity where the sample
is usually placed in the dielectric pillar configuration.

This has been verified

experimentally using electron paramagnetic resonance standards.

1.0

Normalized amplitude

Phase 90
120

TE011
60

0.8
150

0.6

30

TM111
180

0.0
0.2

0.4

0.4
0.6

210

330

0.8

0.2
0.0

1.0

44.2

240

44.3

270

300

44.4

Frequency (GHz)
Figure 29. The cavity construction lifts the degeneracy of the TE01n and TM11n
modes. The main portion of the figure shows the two resonances separated by
230 MHz. The inset shows the corresponding circles in the complex plane.

64
Q values as high as 25,000 have been obtained at liquid helium temperatures for the
loaded axial cavities excited in the TE011 modes. In general, higher n (higher fo) TE01n
modes have reduced Q values. In addition, the shorter wavelengths associated with the
higher frequencies slightly increase the leak amplitude relative to the signal transmitted
through the cavity. These facts, together with the diminished dynamic range of the
spectrometer (see table 2) at higher frequencies, make it harder to observe TE01n (n>1)
resonances of comparable quality to the TE011 modes. Nevertheless, the data in figure
28 far exceed the criteria discussed previously for making successful cavity perturbation
measurements. Consequently, one is able to make reliable measurements at frequencies
up to 200 GHz.

90
6000

a)

120

b)

54.0 GHz
60

103.5 GHz

1000

61.2
30 GHz

150

60

1500

58.8 GHz

5000
4000

90
120

150

30

78.0 GHz

3000

500
2000

88.7 GHz

1000

180
180

180

0 180
1000

500

2000
3000
4000

210

1000

330

5000
6000

210

330

1500

240

240

300

300
270

270

90

c) 120

600

400

60

150

30

200

180

0 180

200

126.9 GHz
400

210 155.9 GHz

600

169.1 GHz

330

240

300
270

Figure 30. Various cavity resonances plotted as circles in the complex plane,
obtained in: a. V band, b. W band, and c. D band. The points are raw data,
while the solid lines are fits.

65
Table 3. Resonance Parameters for different modes obtained in various cavities. The
first column lists the frequency band and the MVNA harmonic. The second column lists
the cavity (letter) and the endplate used (number), as well as the mode (if known);
cavities A, B, and C have dimensions (lengthdiameter) 0.260 in. 0.375 in., 0.344 in.
0.354 in., and 0.218 in. 0.222 in., respectively, cavity T is the transverse cavity (0.400
in. 0.305 in.), and endplates 1, 2, and 3 have coupling hole diameters of 0.052, 0.033,
and 0.024 in., respectively. The next three columns list the resonance frequencies (fo), Q
values, and dynamic ranges (signal-to-noise) for each mode. The final column lists the
contrast in dB between the transmitted amplitude on resonance [A(fo)] and the leak
amplitude (Al). The dynamic range is somewhat lower at 45 GHz than in the main part
of the V and W bands due to the fact that the V band waveguide is so close to cutoff at
these low frequencies.
Band/Harmonic
V3
V3
V3
V3
V4
V4
V4
V4
W-5
W-5
W-5
W-6
W-6
D-8
D9
D 10
D 12
V4
V4

Cavity/mode
A1 TE011
A1 TM111
B1 TE011
B1 TM111
A1 TE012
A1 TE212
A1 TE312
A1 TE412
A2 TE213
A2 TE013
A2
A2 TE014
A2
C3 TE014
C3
C3
C3
T TE012
T - TE013

fo(GHz)
44.450
44.219
44.414
44.265
58.754
53.951
61.247
68.906
73.924
78.044
88.678
98.715
103.532
127.233
145.227
155.815
187.868
56.110
65.078

Q
19000
7700
24900
7400
13900
10400
10000
4800
8800
7900
23000
5300
4350
13600
7900
5800
5150
15500
14000

S/N (dB)
76
74.6
81.4
80.3
85.4
87.8
90.4
92.4
84.5
89.0
90.0
83.0
84.5
45.8
48.5
41.2
31.0
81.9
74.6

A(fo) Al (dB)
34.5
19.0
26.0
13.8
27.0
40.4
24.5
20.8
25.0
23.0
22.0
16.5
27.0
16.5
17.5
21.0
4.0
18.0
13.0

Above about 130 GHz, there is less confidence in the mode assignment of the
resonances. However, by following the frequency dependence of data containing distinct
features, which depend strongly on the polarization of the AC fields within the cavity,
one is able to characterize and use the axial cavities up to 230 GHz. Figure 30 shows
several higher n TE01n axial cavity modes, together with selective higher frequency

66
resonances. Table 3 lists the frequencies, Q values, dynamic ranges and amplitude on
resonance minus the leak amplitude associated with these modes, as well as some
parameters for the transverse cavity (all at 4.2 K) [20]. These figures clearly demonstrate
the potential of the system for cavity perturbation measurements. It should also be noted
from figure 30 and table 3, that many of the resonances were obtained using a single
resonator (cavity A), which was one of the main objectives of this system from the outset.
Finally, figure 31 illustrates the importance of using active temperature stabilization.
The upper panel shows magnetic resonance data taken at the base temperature of the
cryostat (1.4 K), while the upper panel shows the same data obtained using active
temperature stabilization at 1.8 K. A clear drift in the unlocked temperature data is

Amplitude (dB)

observed throughout the course of the up and down sweeps of the magnetic field.

75.5
75.4
75.3
75.2
75.1

Down
Up
Base temperature
No control

75.5

T=1.8 K, controlled

75.4

Up/down

75.3
75.2

Magnetic field (tesla)

Figure 31. Comparison between field sweeps with the temperature unlocked
(upper) and actively controlled (lower); the broad dip is due to impurities in
waveguide, prior to Au plating.

67
By eliminating the influence of all components of the sample probe (aside from the
cavity) on the signal returned to the VR, it is possible to record changes in the complex
cavity parameters in real time, i.e., the vector (either amplitude and phase, or amplitude
and frequency) recorded at the VR is directly related to the impedance of the cavity ZC.
Here, once again, the power of the MVNA is apparent. Using a scalar detection scheme,
it would be necessary to modulate the frequency in order to extract the complex cavity
response. This would inevitably result in a much longer time for recording each data
point. The MVNA essentially returns phase and amplitude information at the detection
frequency of the VR, which is approximately 10kHz. This aspect of the instrument
described here makes it highly suitable for measurements in high magnetic fields, which
can be expensive to run over long periods.
A distinct advantage of conducting fixed frequency optical measurements, as a
function of magnetic field, is that the spectral features which are under investigation may
be expected to change with field, i.e., the magnetic field is the variable that is used to
tune the electronic excitation spectrum of the material under investigation.

If, for

example, this induces a change from an insulating state, to a conducting state, then the
optical response at frequencies comparable to the gap will reflect this change. These
types of behavior will be thoroughly investigated in chapter three using the techniques
described above.

68
Torque magnetization measurements

The cantilever beam torque magnetometer

The mm-wave experimental apparatus discussed in the previous portion of this chapter
was developed explicitly for the work described in this thesis. Consequently, its features
were outlined in great detail [20]. In contrast, the magnetic torque apparatus used for the
experiments discussed in chapter four were developed elsewhere, and will, therefore be
described in much less detail. The interested reader should refer to two references, [41]
and [42], for more explicit details.
The method of cantilever beam torque magnetometry (CBTM) allows for
determination of a samples magnetic moment, through the relation of magnetization,
applied field and the related torque ( = M B). Micro-machining of single silicon
crystals into cantilever beams allows for sensitivity on par with that obtained from force
magnetometry and is, in fact, far superior to magnetometers based on the
superconducting quantum interference device (SQUID), which do not work at all in high
fields [42]. Another feature of the CBTM attractive to those that work in high magnetic
fields is the compact size of such devices. As most experiments require 4He temperatures
or below, the available space for components within the cryostat is usually quite
restrictive. At the lowest temperatures, few magnetometers can be employed, leaving
CBTM as the only viable option for measuring the magnetic moment of the entire
sample, within the confines of 3He/4He dilution refrigerator cryostat [42].
A schematic of the CBTM is shown in figure 32. Here it is seen that the sample sits
on the large face of the magnetometer, which forms one plate of a parallel plate capacitor,

69
while the other face lies stationary. As the magnetic moment of the sample experiences a
torque in the applied magnetic field, it will deflect the cantilever beam, thereby changing
the capacitance of the device. In this geometry, the voltage across the capacitor is
directly proportional to the distance between the plates. Thus, the change in electrostatic
potential can then be measured using an AC capacitance bridge in conjunction with a
lock-in amplifier. By using the capacitance bridge to null the voltage at zero field, one
can measure the change of potential as a function of applied field.

Assuming the

cantilever beam obeys Hooks law for small displacements, the torque experienced by the
beam can be written as = , where is the angular deflection of the beam, and is the
spring constant for the cantilever. In the case of machined silicon single crystals,
Ewt 3
=
, where E is the elasticity modulus, l, w, and t are the length, width, and
12 l
thickness of the cantilever, respectively [42]. Thus for small deflections, the torque is
directly related to the displacement, which can be measured as a change in voltage across
the parallel plates with the lock-in amplifier. Note that the actual amplitude of the
deflection, , is so small that it has no effect on the calibration of the device. Of course,
the measured torque can then be related back to the magnetization using the torque
experienced by a magnetic moment in an applied field: = M B. Hence, by measuring
the change in potential across the parallel plate capacitor as a function of applied field,
one can directly measure the magnetization of a given sample.

70

Figure 32. Schematic diagram of the CBTM. The torque experienced by the
magnetic moment of the sample causes the capacitance of the device to change as
a function of field. This change in capacitance can be measured, and related back
to the magnetization of the sample.

For the layered superconductor studied throughout this thesis, the magnetic moment of
the sample is almost always directed perpendicular to the highly conducting planes. It is
only when the direction of the applied field falls within the conducting planes, that the
magnetization deviates from its usual direction [19]. Therefore, to convert the measured
torque into a magnetization, one must divide the torque by the magnitude of the applied
field and the sine of the angle between the magnetization and the field: M = /Bsin.
Consequently, the angle will be defined as the angle between the applied field and the
direction normal to the highly conducting planes of the crystal. Hence, at = 0o, the field
and the direction normal to the layers are aligned, i.e., the field is normal to the
conducting layers, and at = 90o the field lies parallel to the highly conducting planes.
As the entire CBTM was mounted on a single axis rotator, any angle between 0o and 90o
could be obtained.

71
Temperature Control
As mentioned above, the compact size of the CBTM was a key necessity when
attempting experiments at low temperatures. Temperature control for the measurements
described in chapter four was obtained using an Oxford Instruments top loading 3He/4He
dilution refrigerator in the bore of a 20 T superconducting magnet. The base temperature
for this instrument was 25 mK, with the possibility of raising the temperature to a
maximum of around 300 mK. The temperature was monitored with a ruthenium oxide
resistor, and adjusted by changing the cooling power of the system.
Due to the high sensitivity to small changes in temperature of the dilution refrigerator,
one must ensure slow field sweeps, and plenty of equilibration time after changing the
ambient temperature of the experiment. One must also be aware of thermal drift during a
field sweep. Hence, these measurements are often time consuming and rather difficult.
However, as this temperature range is otherwise impossible to reach, patience must be
used when attempting to study materials in dilution refrigerators. The interested reader
should refer to reference [43] for further details on the workings of dilution refrigerators
and cryogens in general.
Conclusion

It is clear that the two techniques described above will be applicable in very different
regimes of magnetic field and temperature.

To this end, the cavity perturbation

measurements have been preformed in fields of 0 - 8 T at MSU, and 0 - 33 T at the

NHMFL. The limiting variable in these measurements is the temperature, which has only
been varied from 1.5 15 K. The acquisition of a 3He cryogenic system will soon allow

72
for an extended temperature range of ~ 300 mK 300 K. However, this will still not
reach the lowest temperatures obtained in dilution refrigerators.

Thus, the coldest

measurements must be of a DC nature. Fortunately, CBTM measurements have proven

to be extremely sensitive, bridging the gap between zero temperature and 300 mK, in
fields of 0 33 T. Consequently, one can study materials from nearly 0 K up to room
temperature, and in DC magnetic fields anywhere from 0 45 T.