A pulsed phase-sensitive technique for acoustical measurements

Vladimir F. Kozhevnikov
University of Utah, Salt Lake City, Utah 84112
Dmitrii I. Arnold
Russian Research Center Kurchatov Institute, Moscow 123182, Russia
Matthew E. Briggs
University of Utah, Salt Lake City, Utah 84112
Salim P. Naurzakov
Russian Research Center, Kurchatov Institute, Moscow 123182, Russia
John M. Viner and P. Craig Taylor
University of Utah, Salt Lake City, Utah 84112
Received 5 January 1999; accepted for publication 8 September 1999
The pulsed phase-sensitive PPS technique for measurements of sound velocity and attenuation in
uids and solids is reviewed. With this technique, which uses a cycle-overlap principle, a time delay
is measured between any two acoustical pulses transmitted through a sample or reected from its
boundaries. A current realization of the technique allows one to resolve the time-delay variation
down to 0.1 ns. Thus at relative measurements with the PPS technique, precise sound velocity data
can be obtained for samples of small thickness about 1 mm. The technique is versatile and can also
be used for accurate absolute measurements. The technique is most advantageous for studies of
samples with high attenuation, in particular near phase transitions and at high temperatures. The
technique also allows one to measure sound attenuation with moderate accuracy. An application of
the technique that employs narrow-band radio frequency rf bursts for relative measurements of the
phase sound velocity is considered in detail. The technique is applied to studies of liquid alkali
metals and mercury at temperatures up to 2100 K and pressures up to 200 MPa. As a verication
of the capabilities of the technique, new results are presented on sound attenuation in mercury in the
metalnonmetal transition region. A table for sound velocity in mercury at temperatures from 550
to 1900 K and pressures up to 190 MPa is presented in the Appendix. 1999 Acoustical Society
of America. S0001-49669906412-7
PACS numbers: 43.58.Vb, 43.35.Yb, 43.35.Bf SLE
INTRODUCTION
The well-known power of physical acoustics to study
the properties of matter has resulted in the development of
numerous techniques for measurements of sound velocity
and attenuation in uids and solids. Some of the techniques
have found important industrial applications. The appropriate
theory, principal methods, and corresponding experimental
details can be found in comprehensive reviews and mono-
graphs, such as McSkimin,
1
Truell et al.,
2
Papadakis,
3
Kolesnikov,
4
Trusler,
5
and Cantrell and Yost.
6
The pulsed phase-sensitive PPS technique was origi-
nally developed for accurate measurements of sound velocity
in liquid metals at high temperatures (T2500 K and high
pressures ( P300 MPa.
7
It will be seen, however, that the
capabilities of the PPS technique are not restricted to this
eld.
The main difculties with experiments at such condi-
tions are well known. They are caused by small sample sizes,
utilization of rather long and stationary buffer rods, and large
sound attenuation in the sample and buffer rods. Although
some techniques, such as pulse-echo overlap,
8,9
pulse
superposition,
10
and phase comparison,
11
allow very high ac-
curacy for the time-delay measurements, and some of them
are adaptable for use with the buffer rods,
9,12,13
the high-
temperature and high-pressure conditions are not favorable
for the application of these techniques. Other accurate tech-
niques, such as the pulse-echo
2,14
and pulse interferometric
techniques,
1,15,16
are also inappropriate for measurements at
simultaneously high temperatures and pressures. Recently,
M. Hensel
17
conducted measurements of sound velocity in
expanded mercury with the pulse-echo technique. Reliable
data were obtained at temperatures below 1500 K.
Only a pulse transmission-echo technique
18
facilitates
the measurements at both high temperatures and high pres-
sures. This technique accommodates the buffer rods and uti-
lizes a pulse transmitted through the sample in a single pass.
Therefore this technique allows one to reach the highest tem-
peratures where most samples exhibit the greatest attenua-
tion. This pulse transmissionecho technique was used by
Suzuki et al.
19
in the pioneering high-temperature, high-
pressure study of sound velocity in mercury.
A schematic diagram of the cell used in this technique is
shown in Fig. 1. The sample space is the gap between the
buffer rods. Three time delay intervals are measured: the
delay for the echoes in each of the rods and the delay for the
signal transmitted through the entire cell. The difference be-
tween the time required for the signal to travel through the
cell, and the average of the echo times, represents the time
required for the signal to traverse the sample. Similarly, the
3424 3424 J. Acoust. Soc. Am. 106 (6), December 1999 0001-4966/99/106(6)/3424/10/$15.00 1999 Acoustical Society of America
difference between the relative levels in decibels of the trans-
mitted signal and the average of the echo signals yields at-
tenuation due to the sample, provided that corrections are
made to account for the samplebuffer rod interfaces. The
velocity and coefcient of attenuation are then easily com-
puted when the sample length is known.
A disadvantage of the original version of the
transmissionecho technique is that the accuracy of the time-
delay measurements is determined by the duration of the
leading edge of the rectied rf pulses. The leading edge al-
ways occurs over a number of rf periods depending on the
resonance properties of the transducer with corresponding
electrical circuit, and the acoustical characteristics of the
buffer rod.
5
An uncertainty of 10%20% is a reasonable
estimate of the error in the speed measurements with this
technique for samples of about 1-mm length.
18
The PPS
technique combines the advantages of the transmissionecho
method with the high accuracy inherent in the phase com-
parison techniques.
The principles of the PPS technique and corresponding
experimental apparati are described in the next section. Ap-
plications of the technique are presented in Sec. II. Section
III describes possible cell designs for acoustical measure-
ments in liquids at high temperatures, and Sec. IV presents
the summary. An Appendix contains tabulated data obtained
for sound velocity in mercury at sub- and supercritical tem-
peratures and pressures.
I. THE TECHNIQUE
Below we describe the technique for using the cell sche-
matically shown in Fig. 1 with moderately narrow-band rf
bursts, i.e., for phase sound velocity measurements. In such a
case, the use of transducers of different resonance frequen-
cies allows one to study the frequency dependence of the
acoustical properties. However, these conditions are not
mandatory. The technique is applicable without buffer rods,
as well as with either one or two transducers. In addition,
broadband pulses can be employed to measure the group
velocity of sound.
A. Basic principles of the time-delay measurements
For high-temperature measurements, an ultrasonic probe
pulse is introduced into the sample through a buffer rod. The
sound propagation times in the buffer rods including the
small delays in the transducerrod bonds
7
,
1
and
2
, are
unknown. However, these times can be eliminated provided
that two time delays, t
1
and t
2
, between a transmitted
and reected signal are measured, by applying the probe
pulse rst to one and then to the other transducer.
t
1

2
2
1
, 1
t
2

1
2
2
. 2
The sound propagation time through the sample and
the velocity c are
t
1
t
2
/2, 3
cL/, 4
where L is the length of the sample.
The intervals t are measured by delaying the sweep of
the oscilloscope with respect to a basic synchronization pulse
that periodically initiates the coherent rf probe pulses. The
choice of the pulse repetition rate is determined by the time
required for the echoes in the buffer rods to be attenuated
before the next pulse is applied. The sweep of the oscillo-
scope is delayed for different times for alternate pulses ap-
plied to the two oscilloscope channels. For example, if the
two delays are 2
1
and
1

2
, then the reected and
transmitted pulses are displayed, respectively. In this way
one may achieve a coincidence of the phases of the corre-
sponding rf cycles for these two pulses the cycle overlap or
the cycle-for-cycle matching. To avoid undesirable interfer-
ence effects, the time delay should be set using the rf cycles
close to the onset of the rst transmitted and reected sound
pulses where the magnitudes of the oscillations are nearly
constant and there is no overlap of the waves reected from
the sample boundaries.
The block and time diagrams for the apparatus at the
University of Utah are shown in Figs. 2 and 3, respectively.
A MATEC 5100 gating modulator with plug-in is used to
produce coherent rf pulses synchronously with the basic syn-
chronization pulses shown on line 1 of Fig. 3. A Wavetek
3006 generator with frequency instability less than 0.2
ppm/hr is used as a continuous wave cw source for the
gating modulator. For accurate measurements of the fre-
quency dependence of the acoustical properties, a counter to
control the frequency of the source is added. The reected
line 2 and transmitted line 3 rf pulses are amplied by the
broadband MATEC 625 receivers, and then applied to the
inputs of a dual-channel one-beam oscilloscope Tektronix-
475A, working in the alternating signal mode with external
triggering.
A calibrated time-delay generator EG&G Instruments,
Inc., Model 9650A with accuracy 0.3 ns310
7
FIG. 1. Schematic diagram of a cell for acoustical measurements at high
temperatures.
FIG. 2. Block diagram of the PPS apparatus at the University of Utah. TDG
is the time delay generator; RPG is the generator of the coherent radio-
frequency pulses; S-C is the synchronizer-commutator; OSC is the oscillo-
scope. The numbers correspond to the timing diagrams of Fig. 3.
3425 3425 J. Acoust. Soc. Am., Vol. 106, No. 6, December 1999 Kozhevnikov et al.: Pulsed phase-sensitive technique
(delay), and jitter 50 ps10
8
(delay) produces syn-
chronization pulses A line 4 and B line 7 delayed with
respect to the pulse line 1 by times close to 2
1
and
1

2
, respectively. These synchronization pulses drive a
home-made synchronizer-commutator constructed from
NAND elements from microcircuits 74LSOON which pro-
duces a sequence of pairs of synchronization pulses A and B
line 10 that trigger the oscilloscope sweep. These pulses are
unambiguously synchronized with the corresponding oscillo-
scope channels. Specically, pulse A appears only when the
rst channel is active, and Bonly when the second channel
is active. The synchronization is achieved with the aid of a
gate pulse line 5. This gate pulse is an internal pulse from
the oscilloscope, which switches its display between the
channels at the end of each sweep in the alternating mode.
Thus on the oscilloscope screen an operator sees two
pulses, each of which can be independently shifted on the
horizontal axis by variation of the corresponding delay of
the delay generator, and amplied or attenuated on the ver-
tical axis. This feature of the PPS technique distinguishes it
from the other methods for time-delay measurements. The
sweep rates are limited only by the oscilloscope. When the
two traces are matched cycle-for-cycle, the difference be-
tween the time delays of the two channels between A and B
pulses in Fig. 3 of the delay generator yields the interval t
see Sec. I C for additional elucidation.
The PPS technique can be realized with various
schemes. A principal element of all these schemes is the
synchronizer-commutator that provides the sequence of
pulses, shown on line 10 of Fig. 3, to synchronize the corre-
sponding channels of the oscilloscope.
A diagram of the apparatus at the Kurchatov Institute is
shown in Fig. 4. In this apparatus two one-channel delay
generators, I18 Russian standard, with accuracy 0.5 ns
0.510
7
(delay), and jitter 0.4 ns10
8
(delay) are
used. These are connected in series. The delayed output of
the rst generator TDG1 drives the second generator TDG2,
which allows TDG2 to read t directly. The generator for
the coherent rf pulses is home made courtesy of E. Grodz-
inskii. An oscilloscope C 170 with a dual-channel
amplier 401101 is used. The timing diagrams for this
apparatus are the same as in Fig. 3.
A third diagram of the PPS apparatus, in which the two
time delays are provided by a single one-channel delay gen-
erator with the use of its internal clock signal, is described in
Ref. 7.
The process of measuring sound velocity is partly auto-
mated. The monitoring of the time-delay generator, the re-
cording of original data, and the analysis of these data are
performed using a computer. However, the superposition of
the rf pulses on the oscilloscope screen is performed by
hand. A realization of the PPS technique that employs a digi-
tal oscilloscope or other digital device to control the cycle
matching should allow one to completely automate the pro-
cess and should provide an additional increase in the sensi-
tivity of the apparatus.
B. Sensitivity and accuracy of the technique
The sensitivity of the technique, or the resolution of the
smallest phase difference between the cycles of the transmit-
ted and reected rf pulses or other pulses chosen for the
measurements is limited by the jitter in the delay generator
and the scope trigger. These jitters result in an apparent trace
thickness on the oscilloscope screen. The apparati in Salt
Lake City and Moscow allow one to distinguish the variation
of t down to 0.1 and 0.5 ns, respectively.
The sensitivity represents a lower bound on the uncer-
tainty of the time measurements, which also depends on
other factors of a particular experiment. After necessary cor-
rections of systematic errors such as, for example, a phase
shift caused by reections from the two surfaces of the bond
between the sample and the transducer for the measurements
in solids,
20
and a possible difference in time delays for the
different oscilloscope channels, the accuracy is determined
from the dispersion of the individual results obtained at a
given experimental point. At high temperatures the disper-
sion is mostly determined by the temperature instability of
the sample. For example, in the experiments with
mercury
2124
where the probe pulse was switched several
times between the transducers for each experimental point,
the dispersion of the measured t was 1 ns; this value
FIG. 3. Timing diagrams for the apparati shown in Figs. 2 and 4. See text
for details.
FIG. 4. Block diagram of the PPS apparatus at the Kurchatov Institute. The
labels are dened in the caption to Fig. 2. The numbers correspond to the
timing diagram of Fig. 3.
3426 3426 J. Acoust. Soc. Am., Vol. 106, No. 6, December 1999 Kozhevnikov et al.: Pulsed phase-sensitive technique
represents the uncertainty of the time measurements in this
work. Drift of the delay pulses, which should also be taken
into account to estimate uncertainty in the data, is much
smaller than 1 ns.
The high sensitivity in the time-delay measurements al-
lows accurate relative data on sound velocity to be obtained
with samples of short lengths. The lower limit for the sample
length L is determined by the duration of the leading edge of
the rf pulses and by the sound velocity c in the sample: L
NTc/2, where N is number of rf periods T in the leading
edge. Currently the minimum length of the samples used for
the measurements with the PPS technique is about 0.7 mm.
24
C. Relative measurements of sound velocity
In principle, Eqs. 14 describe the procedure for
making the absolute measurements of sound velocity. How-
ever, problems associated with selection of the correspond-
ing cycles for the chosen rf pulses, as well as delays in the
electrical circuits, prevent a simple realization of this ap-
proach. Correct selection of the corresponding rf cycles can
be accomplished by employing different path lengths, i.e., by
measurements with different echoes,
1,3,20
or with samples of
different lengths.
15,25
These problems are obviated by mak-
ing relative measurements. For relative measurements a
change of the propagation time
r
is measured as a
function of variation in the environment with respect to a
reference point. The parameter
r
is an effective time delay
measured at this point. The sound velocity c
r
, and therefore
the true value of the time delay in the reference point
r
0
L
r
/c
r
(L
r
is the length of the sample at the temperature of
the reference point, must be known from the literature or
from another experiment. Note that the absolute measure-
ments at the reference point can also be performed with the
PPS apparatus. This point is usually chosen at more or less
normal conditions, when sound attenuation is not large and
the high uniformity of a temperature eld is readily achiev-
able over relatively large samples with length on the order
of 1 cm, for example. Probably the best way to make abso-
lute measurements at the reference point is to use the PPS
apparatus in one of the classical modes, which have been
specically developed for precise measurements under such
conditions see Sec. I F below.
Representative oscilloscope pictures with reected and
transmitted signals for the reference point which is often the
starting point of the experiment are shown in Fig. 5. There
are no delays at either of the delay generator channels for the
signals shown in Fig. 5a zero delay is set for A and B
synchronization pulses in the diagram of Fig. 3. In Fig. 5b
the time delays of the synchronization pulses are set so as to
achieve an approximate alignment of the transmitted and re-
ected rf pulses. Finally, one chooses the rf cycles which
will be used in all the further measurements 19th cycles in
Fig. 5b and displays the region around them on an ex-
panded scale 10. Measurements are made using this ex-
panded scale. One corrects the delay of the B synchroniza-
tion pulse to achieve the cycle-for-cycle matching with
desirable resolution approximately 0.1 ns for the signals in
Fig. 5c, and the interval t
1
is determined as described in
Sec. I A. The interval t
2
is measured in the same way, and

r
is computed using Eq. 3. The measured time
r
dif-
fers from the true sound propagation time
r
0
L
r
/c
r
. This
difference d
r
0

r
tnT/2, where t represents a time
delay in the electrical circuits; T is the period of sound os-
cillation, and n is an integer, which counts the mismatch of
the rf cycles chosen for the measurements; and the coef-
cient 1/2 comes from Eq. 3.
All further measurements must be conducted using the
same rf cycles. That is, the experiment must be monitored in
such a way, which allows one to identify these cycles on the
oscilloscope screen. In practice, this identication is readily
achieved if the variation in the sound propagation time be-
tween neighboring experimental points is limited to two or
three periods of the rf oscillations. In regions where the
sound velocity varies rapidly, such as near a critical point, it
is better to keep the time variation to within one period.
While running the experiment, in regions with weak sound
dispersion one can check for a lost cycle by turning the mag-
nier off and looking at the onsets of the pulses Fig. 5b.
If there are no phase errors, then the coincidence of the lead-
ing edges remains nearly the same as it was in the beginning.
It should be stressed that for relative measurements the strict
correspondence of the cycles or exact alignment of the lead-
ing edges is not of primary importance. It is important to
perform all measurements with the same cycles chosen at the
reference point.
The sound velocity is calculated from the formula
cL/
r
0
L
0
1 TT
0
/ d, 5
where L
0
is the sample length at temperature T
0
ordinarily
T
0
is room temperature; is the linear thermal expansion
FIG. 5. Oscilloscope traces of the ultrasonic pulses for a brass cell lled
with water. The length of each rod was 85 mm and the diameter was 14 mm.
The length of the sample was about 2 mm. The frequency of the sound was
7.7 MHz. The upper traces represent the reected signal, and the lower
traces, the transmitted one. No matching network was used. The sweep rates
are 10, 0.5, and 0.05 s/div for the traces in a, b, and c, respectively.
The sound reections within the sample are seen in the lower trace of a.
3427 3427 J. Acoust. Soc. Am., Vol. 106, No. 6, December 1999 Kozhevnikov et al.: Pulsed phase-sensitive technique
coefcient of the cell material relative to T
0
; T is the tem-
perature of the sample. The time is the effective propaga-
tion time computed using Eq. 3. For determination of the
thermal expansion of the sample, the principal parts of the
cell see Sec. I D below must be made from the same ma-
terial.
In rst-order phase transitions, like the liquidgas tran-
sition, the speed of sound changes discontinuously, and it is
essentially impossible to follow the phases of the signals.
After such a transition a new coincidence of the phases must
be established. The denominator in Eq. 5, (
r
0
), will
now differ from the true value of the propagation time by
mT/2, where m is another integer. The matching can be cor-
rected in different ways. One way
21,22
is to obtain data by
passing around the critical point. Using this method one
nds
21,22
that m is 0, or 1.
D. Sample length
Correct measurement of the sample length is obviously
an important component of the acoustical measurements.
While it has been discussed for solids see Ref. 2, for ex-
ample, we will briey consider this issue for uids.
Commonly, a ringlike spacer with at and parallel ends
connes the sample length. An elastic element of the cell
spring presses the rods to the spacer, insuring that the
spacer and sample lengths are identical see Ref. 12, for ex-
ample. The thermal expansion of the sample is then deter-
mined by knowledge of the thermal expansion of the spacer.
In the experiments at high temperature and high pres-
sure, the buffer rods are usually hermetically sealed in the
cell body. Such construction signicantly complicates the
use of an elastic element in the cell. The use of a spacer is
possible provided that the cell components such as the cell
body, the buffer rods, and the spacer that contribute to the
thermal expansion are made strictly from the same material
including the crystalline orientation if single crystal buffer
rods are employed. However, the spacer is not recom-
mended for highly attenuating samples, because parasitic
signals through the spacer are not negligible any more. In
such a case the requirement of the identical materials allows
one to remove the spacer and to calculate the sample thermal
expansion as it is done in Eq. 5. The length L
0
can be
determined from calibration measurements of the sound ve-
locity versus temperature with a sample for which the tem-
perature dependence of the sound velocity is well known; or
from direct measurements on the outer ends of the buffer
rods.
For example, our experiments with mercury
2124
have
been carried out with spacer-less cells the cell design is
described in Sec. III below. The change of position of the
inserted buffer rod with respect to the state when it is in
contact with the other rod was measured with a micrometer.
In different experiments the length L
0
varied from 0.7 to 2.6
mm. The uncertainty in the length data was within 57
m. The sources of this uncertainty are nonparallelism in
the ends of the buffer rods and in the axes of the inserted
buffer rod and the cell body. The sound velocity data ob-
tained with the different sample lengths and the cells made
of different materials molybdenum and niobium agree well
with one another as well as with the data on mercury of
Spetzler and Meyer
12
obtained by the pulse superposition
technique with a spacer of 8 mm-length. A value of 10 m
was taken as an upper limit on the uncertainty for the sample
length at room temperature. Such an uncertainty in the
sample length was sufcient for the purposes of the
experiments;
2124
however, it can be improved if necessary.
E. Uncertainties in the sound velocity data
Equation 5 yields the following formula for an esti-
mate of the fractional uncertainty c of the sound velocity
data at the relative measurements,
cc/cc
r
1/
r

2

2
L
0
1
r
/
1

2
TT
0

2
TT
2

1/2
, 6
where c is the absolute error of the velocity c, c
r
, is the
uncertainty in the reference point, ()/(
r
) is
the uncertainty in the time-delay measurements () is
the absolute error of the measurements, L
0
is the un-
certainty in the sample length at room temperature, and is
the uncertainty of the literature data on the thermal expan-
sion coefcient of the cell material, and T is the uncertainty
in the sample temperature T. The last two terms in Eq. 6
are ordinarily signicantly smaller than the rst three and
can be omitted.
26b
For example, in the experiments with mercury,
22
the
data on sound velocity at the reference point were taken from
the literature. The uncertainty in this value is on the order of
10
3
. The uncertainty of the sample length measured at
room temperature is also of the same order. The length L
2 mm. The error () is 1 ns, and the term in Eq. 6
is on the order of 10
4
, which is much smaller than the other
main contributions. The last two terms in Eq. 6 are on the
order of 10
5
. Therefore, the uncertainty in the sound ve-
locity c at low temperatures not far from the reference
point is dominated by the uncertainty of the literature data
for c
r
. On the other hand, at high temperatures (T1500
K the uncertainty is dominated by the uncertainty in the
length of the sample. Thus the uncertainty in the time mea-
surements essentially does not contribute to the error in the
sound velocity data,
22
which is less than 0.4% Tabulated data
on sound velocity in mercury obtained with the PPS tech-
nique are presented in the Appendix.
F. A test
The technique was rst tested with water. An essential
advantage of water as a testing substance is a relatively at
maximum in sound velocity near 74 C. This feature allows
one to make accurate measurements of sound velocity near
the maximum even with a relatively large uncertainty in the
sample temperature.
The sound velocity in pure water at standard pressure is
known with very high precision; disagreement of the data of
different groups
2729
is about 0.5 m/s or 0.03%. Reproduc-
ibility of the data obtained in each of these works is much
3428 3428 J. Acoust. Soc. Am., Vol. 106, No. 6, December 1999 Kozhevnikov et al.: Pulsed phase-sensitive technique
better. What is more important for the relative measurements
is that the change of sound velocity in these works, for ex-
ample between points at 74 and 20 C, is in fact identical:
72.8 m/s in Refs. 27 and 28, and 72.7 m/s in Ref. 29.
Experiments with water using the PPS technique have
shown explicitly the principal importance of the use of the
same material for all the parts of the cell. Application of
nonidentical materials resulted in signicant disagreement
between the measured and literature data even over less than
a 100-deg temperature interval.
30
Results of two test runs with distilled water are shown in
Fig. 6. In Refs. 30 and 31 one can nd results of other test
experiments with water performed with cells made of differ-
ent materials and with spacers of different length. A solid
curve represents the data of Ref. 28. The relative measure-
ments were performed with respect to a reference point at
room temperature, in which sound velocity was taken from
Ref. 28. In this work the measurements were performed us-
ing an acoustic interferometer with about a 90-mm change of
path length. The authors estimate of the relative accuracy is
on the order of 10
5
. In our tests we used a horizontally
oriented stainless steel cell consisting of a cell body, two
inserted buffer rods, and a spacer of 5 mm length. The un-
certainty of the spacer length was 2 m. A tungsten/
rhenium thermocouple was installed in a groove of the cell
body near the middle of the sample space. A total length of
the cell assembly was 365 mm. A cylindrical heater of about
80-mm length surrounded the cell; it was centered with re-
spect to the sample space.
To insure that the sample was under steady-state condi-
tions, the delays t
1
and t
2
were measured several times
for each experimental point. The temperature was assumed
to be stable, if successive time delays were reproduced with
an error not greater than 1 ns. This value therefore represents
the maximum error in the time measurements ().
In Fig. 6 a horizontal bar denotes the uncertainty in the
temperature measurements 3 C. The uncertainty in the
time in Eq. 6 (310
4
) is the principal contribution to
the uncertainty of the data on sound velocity. Thus c in
these measurements is about or less than 0.03%.
The data obtained agree well with precise literature data.
The precision of the sound velocity measurements can be
estimated from comparison of the results obtained near the
maximum with literature data used for the reference point.
The maximum sound velocity in Ref. 28 equals 1555.147
m/s; we obtained 1555.0 and 1555.1 m/s. Thus the discrep-
ancy between the reference data and the data obtained in
these tests with the PPS technique is close to the discrepancy
between the precise literature data from different groups
28,29
obtained with signicantly larger samples.
It should be pointed out, however, that precise absolute
measurements are very complicated, and it is a great advan-
tage to make them with long samples.
G. PPS and other techniques
The ability to manipulate the signals independently in
both the vertical and horizontal axes of the oscilloscope
screen makes the PPS technique very versatile. In fact, many
of the known pulse techniques can be realized as particular
modes of operation of the PPS apparatus. Several examples
are given below.
The pulse superposition technique
1
can be realized if the
repetition rate of the probe pulse is increased to achieve an
interference of the echoes in the sample.
By removing the buffer rods, one can realize a quasi
phase-comparison technique
11
without real interference of
the sound waves by setting the delay of channel for the
reected pulse to zero line 4 in Fig. 3. A transmitted pulse
this can be a reected pulse as well is then delayed to
overlap with the probe pulse on the oscilloscope screen. Note
that this mode is virtually the same as a transmission variant
of the pulse echo-overlap technique.
8
The double pass technique
25
provides one of the afore-
mentioned ways to select the corresponding cycles and ex-
clude t, and therefore obtain absolute data on sound veloc-
ity. The probe pulse is split for two samples of different
lengths L
1
and L
2
. The delays are measured with the PPS
apparatus for three sample lengths: L
1
, L
2
, and L
1
L
2
.
The echo-overlap PEO technique
3,8,9,20
can be realized
by switching the oscilloscope to one channel operation mode
and removing the inverter from the synchronizer-
commutator NAND element 8 in Fig. 2. The delays of the
synchronization pulses A and B lines 4 and 7 of Fig. 3,
respectively are adjusted so as to achieve the rf cycle match-
ing between any two echoes from the sequence of the pulses
reected from the boundaries of the sample. In order to dis-
play both echoes, the sweep duration must be less than the
interval between the B and A pulses. As was mentioned, this
mode of operation of the PPS apparatus also allows one to
perform absolute measurements of sound velocity, and can
be used, in particular, to obtain the data for the reference
point. It should be noted that in making absolute measure-
ments, especially with small samples, one must be careful
with possible delays in the electrical circuits. In particular, a
difference in the lengths of the cables connecting the outputs
of the delay generator with synchronizer-commutator may
result in a systematic error.
The PEO and PPS techniques are similar in accuracy
because they are both based on the rf-cycle overlap principle.
FIG. 6. Sound velocity in water. Solid curve is depicted on data of Del
Grosso and Mader Ref. 28; squares and circles are experimental points
measured in the runs with the PPS technique.
3429 3429 J. Acoust. Soc. Am., Vol. 106, No. 6, December 1999 Kozhevnikov et al.: Pulsed phase-sensitive technique
However, the PPS technique is easier to perform and is ap-
plicable to a wider variety of physical and technical prob-
lems.
H. Sound attenuation
Simultaneously with the phase measurements, one can
measure the amplitude of the rf pulses, and thus obtain data
on sound attenuation.
18
For completeness, we describe this
procedure. If the amplitude of the probe signal is A
0
, in the
plane-wave approximation the amplitudes of the signals re-
ected at the rst (A
1
) and at the second (A
2
) buffer rods
and that transmitted through the cell (A
t
) can be written as
A
t
K
t
A
0
t
1
T
1S
t
s
T
S2
t
2
,
A
1
K
1
A
0
t
1
2
R
1S
, 7
A
2
K
2
A
0
t
2
2
R
2S
,
where t
s
, t
1
, t
2
are attenuations of the sound waves in the
sample, the rst buffer rod, and the second buffer rod, re-
spectively. T
1S
and T
S2
are the acoustical transmission coef-
cients through the corresponding rod/sample boundaries.
R
1S
and R
2S
are the acoustical reection coefcients. The
transmission and reection coefcients are determined by the
acoustical impedances of the sample and the buffer rods ac-
cording to well-known formulas.
1
The coefcients K
t
, K
1
,
and K
2
depend on the geometrical factors of the cell, such as
the nonparallel congurations of the rod ends, diffraction of
the sound waves, and losses in the buffer-rod/transducer
boundaries. These coefcients also depend on the tempera-
ture distribution in the buffer rods.
From Eq. 7, the sound attenuation t
s
can be written
t
s
KA
t
R
1S
R
2S

1/2
T
1S
T
S2

1
A
1
A
2

1/2
, 8
where K is a combination of K
t
, K
1
, and K
2
. The coef-
cient of sound attenuation is found from
ln t
s
/L. 9
The sound attenuation for short uniform samples L
(D
2
/8), where D is the sample diameter, and is the
wavelength of sound
32
is determined mainly by absorption
of sound.
The temperature dependence of the coefcients K in
Eqs. 7 and 8 does not allow one to eliminate them com-
pletely, even for the relative measurements (
r
), where

r
is the coefcient of attenuation at a reference point. Other
obstacles preventing accurate measurements of the attenua-
tion are the impedances of the sample and the hot ends of the
buffer rods. To calculate these impedances, data on density
are required, which are not always available. In these acous-
tical experiments the average sound velocity in the buffer
rods can be readily measured, but special efforts are required
to determine the sound velocity in the rods near the rod/
sample interfaces. These numbers are needed to compute the
impedances. An additional problem for the absorption mea-
surements concerns diffraction effects in the buffer rods,
which must be carefully taken into account when the plane-
wave approximation is not valid.
1,2,32
For these reasons it is
difcult to estimate a priori the accuracy of the sound attenu-
ation data measured in this approach. Nevertheless, even
qualitative data on sound absorption are very important, and
this method provides the easiest way to obtain such informa-
tion for uids at high temperatures and high pressures. In
particular, the amplitude measurements have played a crucial
role in the interpretation of the acoustical anomalies in
mercury
23,24
and have yielded the most accurate estimate of
the critical parameters of this metal.
22
Examples of sound
attenuation computed using Eqs. 79 are available in
Refs. 18 and 33, and in the next section.
I. Limitations of the technique
A necessary condition for the PPS technique to work is
the existence of at least one signal transmitted through the
sample. Because the technique allows one to carry out the
measurements with very short samples, this condition puts
lower limits on the transmission coefcients T
rs
and T
sr
see
Eq. 7, which are determined by the ratio of the sample and
the buffer rod impedances.
1
Since the sound velocities for all
substances do not vary appreciably in the solid, liquid, and
gas phases,
34
the aforementioned condition leads to a limita-
tion on the mass density of the sample. To date, the lower
limit on density in the experiments that employ the PPS tech-
nique with metallic mostly molybdenum cells is about 0.7
g/cm
3
. This value corresponds to approximately double the
critical density for cesium
26
and to one-tenth of the critical
density for mercury.
26b
II. EXPERIMENTAL EXAMPLES
The PPS technique has so far been applied mostly to
liquid metals, such as cesium
26
and mercury,
2124
at high
temperatures up to 2100 K and high pressures up to 190
MPa.
Figure 7 presents experimental data for the sound veloc-
ity in mercury and data for the transmitted signal amplitude
measured at a frequency of 10 MHz with a molybdenum cell
at a pressure of 162 MPa.
The maximum uncertainty of the results on sound veloc-
ity measurements is 0.4%. The temperature was measured
FIG. 7. Sound velocity and transmitted signal amplitude for mercury at 162
MPa, as measured with the PPS technique. Point A corresponds to the
liquidgas transition, E to the prewetting transition, and M to the metal
nonmetal transition. See text for details.
3430 3430 J. Acoust. Soc. Am., Vol. 106, No. 6, December 1999 Kozhevnikov et al.: Pulsed phase-sensitive technique
by two tungstenrhenium thermocouples manufactured from
a single pair of wire coils, which were calibrated according
to the international standard. The thermocouples were in-
stalled near the upper and lower edges of the sample space.
22
Absolute error of the temperature measurements varies from
5 at 1000 K to 10 at 2100 K. Discrepancies in the readings of
the thermocouples made from different parts of the coils lay
within 1 K. Pressure was measured by a Heise manometer
with maximum error 0.2 MPa.
In Fig. 7 point A marks the liquidgas transition; E
denotes a prewetting transition;
23
and M corresponds to a
density close to 9 g/cm
3
, where, according to data on the
electronic properties see Ref. 24 for references, mercury
undergoes a metalnonmetal transition. A phase diagram
for mercury, which was obtained from acoustical measure-
ments using the PPS technique, is available in Refs. 23 and
24. A similar minimum in the amplitude of the transmitted
signal was observed for other isobars at their intersection
with the isochore of 9 g/cm
3
see, for example, Figs. 3 and 4
in Ref. 24. The variation of the attenuation coefcient, as
estimated from Eqs. 79, is shown in Fig. 8. The at,
poorly shaped minimum in the transmitted signal amplitude
of Fig. 7 corresponds to a well-formed maximum in the at-
tenuation curve of Fig. 8. This maximum coincides well with
the metalnonmetal transition. An estimate of the error of
the data on sound attenuation depends on assumptions made
for the temperature distribution along the buffer rods, but for
any reasonable assumptions this is less than 30%.
III. THE CELL FOR HIGH-TEMPERATURE
MEASUREMENTS
In static experiments at high temperatures and high pres-
sures, a cell with a heater is placed in a high-pressure vessel
that is lled with an inert pressure-transferring medium.
7
Un-
der such conditions, two kinds of cells are currently used in
acoustical experiments. These cells contain either two buffer
rods inserted in the cell body
7,17,19,33
or just one inserted rod.
In the latter case, the second rod is a part of the cell
body.
21,22
These designs are schematically shown in Figs. 8
and 9. A cell with an inserted, or adjustable, rod always has
a nite space between the rod and the cell body. In Figs. 8
and 9 this space is magnied for clarity. The cells with two
adjustable rods, or two-rod cell Fig. 8, are used both in
horizontal and vertical positions. The one-rod cells Fig.
9 are used vertically. We compare the advantages and dis-
advantages of these two cells.
The approximate symmetry of the temperature distribu-
tion along a horizontal two-rod cell is advantageous in some
situations, but such symmetry does not play a serious role for
acoustical measurements because the contributions of the
rods either cancel each other or can be subtracted indepen-
dent of the temperature distribution in the rods. A disadvan-
tage of the horizontal orientation is an inevitable vertical
temperature gradient over the sample at normal gravity,
which becomes crucially important as one approaches the
liquidgas coexistence curve. The dashed line in Fig. 8
shows an approximate meniscus shape at the boiling point
for a symmetrical longitudinal temperature distribution. The
extreme instability of such a meniscus and the corresponding
strong convection ows lead to poor results in this region. In
particular, this geometry leads to a signicant increase in the
transition width the temperature interval between the last
point in the liquid phase and the rst point in the gas phase.
With the PPS technique we failed to perform measurements
within an interval of about 50 C near the boiling point be-
cause of a smeared picture on the oscilloscope screen. It is
possible that the oscillations in the sound velocity observed
in gaseous mercury near saturation by Yao et al.
33
have the
same origin.
The vertical orientation of the cell with two adjustable
buffer rods provides a radial symmetry of the temperature
distribution over the sample, but it does not improve the
FIG. 10. Schematic diagram of a cell with one inserted buffer rod. The
numerical labels are dened in the caption to Fig. 9. The dashed lines on the
rods denote a thread. See text for details.
FIG. 8. Sound attenuation in liquid mercury at a pressure 162 MPa. The
points labeled A and M are dened in the caption to Fig. 7.
FIG. 9. Schematic diagram of a cell with two inserted buffer rods. 1 is the
transducer; 2 is the sample reservoir; 3 is the heater; 4 is the sample space;
5 is the thermocouple; 6 is the buffer rod. A dashed line depicts an approxi-
mate meniscus shape at boiling temperature. See text for details.
3431 3431 J. Acoust. Soc. Am., Vol. 106, No. 6, December 1999 Kozhevnikov et al.: Pulsed phase-sensitive technique
situation near the liquidgas transition. In this regime, strong
convection ows arise because a second meniscus is formed
above the sample space. In our experiments the transition
width with such a cell was about the same as for the hori-
zontal orientation.
One can take data closer to the coexistence curve using
the vertically oriented one-rod cell Fig. 10. In this case the
transition width is determined by a temperature distribution
over just the sample space. That is, in contrast to the above-
mentioned situations, the peripheral regions do not contrib-
ute. For this geometry one can obtain transition width as
small as 2 K.
22
Note that such a cell allows one to use the
buffer rods with a thread on the side surface see Fig. 10
without inuencing the convection ows near the liquidgas
transition. In a cell with two inserted rods the threaded con-
guration leads to expansion of the temperature interval over
which convection ows exist in the sample space. The
thread surface is used to suppress diffraction effects in the
rods.
1,35
The capability to work near the coexistence of two
phases is important in many respects, in particular for com-
parison with other experimental data.
IV. SUMMARY
The PPS technique combines the advantages of the
transmission/echo technique and the phase-comparison
methods for sound velocity measurements. The sensitivity of
the time-delay measurements is approximately 0.1 ns. Thus
using the PPS technique, precise relative data can be ob-
tained over a broad temperature range for samples of short
length currently about 1 mm, and therefore for samples
with large attenuation. The PPS technique also allows simul-
taneous measurements of sound attenuation. The use of
narrow-band rf pulses provides measurements of the phase
velocity of sound, which allows one to investigate the fre-
quency dependence of the acoustical properties. The tech-
nique is versatile; many of the well-known pulse methods
can be realized with an apparatus designed for the PPS tech-
nique. The PPS technique signicantly expands the applica-
bility of precise phase-comparison methods and opens new
opportunities for studies of the physical properties of matter,
particularly near phase transitions and at high temperatures.
A natural limitation of the technique is poor impedance
matching for the sound waves at the interfaces between the
sample and the buffer rods. A further improvement in the
sensitivity of the technique can be achieved by digitizing the
process of matching the rf cycles.
Note added in proof: The anomaly of sound absorption
in mercury presented in Sec. II has been recently conrmed
by H. Kohno and M. Yao, J. Phys.: Condens. Matter 11,
5399 1999.
ACKNOWLEDGMENTS
The help of Sonny Ko in the construction of the appa-
ratus at the University of Utah is greatly appreciated. This
work was partly supported by a grant from the Research
TABLE AI. Sound velocity in mercury.
Sound velocity m/s at pressures MPa
T,
K
P
sat
,
MPa sat 30 60 80 100 120 140 160 170 190
550 1335 1342 1349 1354 1360 1363 1370 1374 1378 1382
600 1309 1316 1324 1329 1335 1340 1346 1350 1352 1358
650 0.1 1283 1291 1299 1304 1310 1316 1322 1326 1329 1335
700 0.3 1257 1265 1275 1280 1285 1291 1288 1303 1306 1311
750 0.6 1229 1238 1249 1254 1260 1267 1274 1280 1283 1287
800 1.0 1200 1211 1223 1229 1236 1243 1250 1257 1260 1263
850 1.9 1170 1183 1195 1202 1211 1218 1225 1233 1236 1239
900 3.0 1139 1153 1167 1175 1183 1191 1199 1208 1212 1215
950 4.6 1108 1122 1136 1146 1155 1163 1171 1181 1186 1189
1000 6.6 1075 1090 1105 1115 1125 1134 1143 1153 1159 1162
1050 9.4 1041 1055 1073 1084 1094 1104 1114 1124 1130 1134
1100 12.8 1006 1020 1039 1051 1062 1072 1082 1093 1099 1106
1150 17.1 970 982 1002 1016 1028 1038 1050 1061 1067 1076
1200 22.2 933 941 963 979 992 1003 1015 1027 1034 1044
1250 28.3 893 894 923 939 953 966 980 993 1000 1011
1300 35.3 854 285 879 897 913 927 944 957 965 977
1350 43.4 812 291 830 852 869 885 903 920 927 940
1400 52.6 768 296 777 802 822 842 860 878 887 901
1450 63.0 723 302 297 747 771 794 813 833 844 860
1500 74.7 676 307 303 686 714 741 763 786 797 813
1550 87.8 627 312 310 309 648 680 707 734 746 762
1600 102.7 577 317 317 315 610 644 675 689 709
1650 119.5 527 323 323 320 526 568 607 624 649
1700 138.7 476 329 329 328 326 475 528 547 580
1750 160.6 426 335 336 335 334 333 460 502
1800 341 342 342 342 342 342 420
1850 346 347 349 349 350 350 352
1900 352 353 355 356 357 358 359
3432 3432 J. Acoust. Soc. Am., Vol. 106, No. 6, December 1999 Kozhevnikov et al.: Pulsed phase-sensitive technique
Committee of the University of Utah, and by the National
Science Foundation under Grant No. DMR-9704946.
APPENDIX
Smoothed data for sound velocity in mercury measured
with the PPS technique are presented in Table AI. See Ref.
22 for the details. The uncertainty in these data is less than
0.4% see Sec. I E. The data from a handbook
34
are used for
computing the saturated pressure P
sat
below 30 MPa. Critical
parameters for mercury 1764 K and 167 MPa are also de-
termined from the acoustical data. Values of the sound ve-
locity in the abnormal regions of subcritical isobars and near
the critical isochore, where the sample was not uniform be-
cause of wetting,
23,24
are excluded from the table. It is inter-
esting to note that, for mercury vapor, the derivative c/P
at constant temperature changes sign at supercritical tem-
peratures.
1
H. J. McSkimin, in Physical Acoustics. Principles and Methods, edited by
W. P. Mason Academic, New York, 1964, Vol. I/A, Chap. 4, pp. 217
332.
2
R. Truell, C. Elbaum, and B. B. Chick, Ultrasonic Methods in Solid State
Physics Academic, New York, 1969.
3
E. P. Papadakis, in Physical Acoustics. Principles and Methods, edited by
W. P. Mason and R. N. Thurston Academic, New York, 1976, Vol. XII,
Chap. 5, pp. 277374.
4
A. E. Kolesnikov, Ultrasonic Measurements Standard, Moscow, 1970,
in Russian.
5
J. P. M. Trusler, Physical Acoustics and Metrology of Fluids IOP, Adam
Hilger, Bristol, 1991.
6
J. H. Cantrell and W. T. Yost, in Handbook of Acoustics, edited by M. J.
Crocker Wiley, New York, 1998, Chap. 41, pp. 483495.
7
D. I. Arnold, A. M. Gordeenko, P. N. Ermilov, V. F. Kozhevnikov, and S.
P. Naurzakov, Prib. Tekh. Eksp. 5, 167170 1985 English transl.: In-
strum. Exp. Tech. 28, 11731177 1985.
8
J. E. May, IRE Natl. Conv. Rec. 6, 134142 1958.
9
E. Papadakis, J. Appl. Phys. 35, 14741482 1964.
10
H. J. McSkimin, J. Acoust. Soc. Am. 33, 1216 1961.
11
B. M. Abraham, Y. Eckstein, J. B. Ketterson, M. Kuchnir, and P. R.
Roach, Phys. Rev. A 1, 250257 1970.
12
H. A. Spetzler, M. D. Neyer, and Tin Chan, High Temp.-High Press. 7,
481496 1975.
13
J. P. Petitet, R. Tufeu, and B. Le Neindre, Int. J. Thermophys. 4, 3550
1983.
14
B. Chick, G. Anderson, and R. Truell, J. Acoust. Soc. Am. 32, 186193
1960.
15
Yu. S. Trelin, I. N. Vasilev, V. P. Proskurin, and T. A. Tsyganova,
Teploz. Vys. Temp. 4, 364368 1966 English transl.: High Temp. 4,
352355 1966.
16
I. I. Novikov, Yu. S. Trelin, and T. A. Tsyganova, Teploz. Vys. Temp. 7,
1220 1969 English transl.: High Temp. 7, 11401141 1969.
17
M. Hensel, Thesis, Marburg University, 1993.
18
A. W. Nolle and P. W. Sieck, J. Appl. Phys. 23, 888894 1952.
19
K. Suzuki, M. Inutake, and S. Fujiwaka, IPPJ-310, Institute of Plasma
Physics, Nagoya 1977; M. Inutake, K. Suzuki, and S. Fujiwaka, J. Phys.
Paris 40, C7-685C7-686 1979.
20
E. P. Papadakis, J. Acoust. Soc. Am. 42, 10451051 1967.
21
V. F. Kozhevnikov, D. I. Arnold, and S. P. Naurzakov, J. Phys.: Condens.
Matter 6, A249A254 1994.
22
V. F. Kozhevnikov, D. I. Arnold, and S. P. Naurzakov, Fluid Phase Equi-
libria 125, 149157 1996.
23
V. F. Kozhevnikov, D. I. Arnold, S. P. Naurzakov, and E. M. Fisher, Phys.
Rev. Lett. 78, 17351738 1997.
24
V. F. Kozhevnikov, D. I. Arnold, S. P. Naurzakov, and E. M. Fisher, Fluid
Phase Equilibria 150151, 625632 1998.
25
J. R. Cunningham and D. G. Ivey, J. Appl. Phys. 27, 967974 1956.
26
a V. F. Kozhevnikov, Zh. Eksp. Teor. Fiz. 97, 541558 1989 English
transl.: Sov. Phys. JETP 70, 298310 1990; b V. F. Kozhevnikov, S.
P. Naurzakov, and A. P. Senchenkov, J. Mosc. Phys. Soc. 1, 171197
1991.
27
M. Greenspan and C. E. Tschiegg, J. Acoust. Soc. Am. 31, 7576 1959.
28
V. A. Del Grosso and C. W. Mader, J. Acoust. Soc. Am. 52, 14421446
1972.
29
A. A. Alexandrov and M. S. Trakhtenherz, Teploenergetika No. 4, 5760
1973 in Russian.
30
O. I. Girnii, Thesis, Moscow Aviation Institute, Moscow, 1988.
31
V. F. Kozhevnikov, D.Sc. dissertation, Kurchatov Institute of Atomic En-
ergy, 1990.
32
W. P. Mason, in Physical Acoustics. Principles and Methods, edited by
W. P. Mason Academic, New York, 1964, Vol. I/A, Chap. 7, pp. 485
500.
33
M. Yao, K. Okada, T. Aoli, and H. Endo, J. Non-Cryst. Solids 205207,
274277 1996.
34
Handbook of Physical Quantities, edited by I. G. Grigoriev and E. Z.
Meilikhov CRC, New York, 1997.
35
H. J. McSkimin and E. S. Fisher, J. Appl. Phys. 31, 16271639 1960.
3433 3433 J. Acoust. Soc. Am., Vol. 106, No. 6, December 1999 Kozhevnikov et al.: Pulsed phase-sensitive technique