PHONON STRUCTURE OF IMPURITY-RELATED OPTICAL SPECTRA. ..

2029

by F. Seitz and D. Turnbull (Academic, New York, 1958), 23Although Hughes's curves are densitometer traces,
p. 353. the experimental procedures followed were such that the
~VJ. T. Ritter, J. Chem. Phys. ~53 3461 (1970). results should be proportional to the absorption coeffi-
Handbook of Mathematical Eunctions saith Formulas, cient, or nearly so tA. E. Hughes (private communica-
G~aPhs, and Mathematical Tr. bles, 3rd ed. , Natl. Bur. tion)]. In any case, the quantitative significance of the
Std. Appl. Math. Ser. 55, edited by M. Abramowitz and experimental results is not of major importance for our
I. A. Stegun (U. S. GPO, Washington, D. C. , 1965), primary purpose, which is to illustrate our computation-
pp. 928-935. al approach.
Eormulae and Tables for Statistical Sbrk, edited by 4A. M. Karo and J. B,. Hardy, Phys. Rev. 141, 696
C. R. Rao, S. K. Mitra, and A. Matthai (Statistical (1966).
Publishing Society, Calcutta, India, 1966). G. Raunio, L. Almqvist, and R. Stedman, Phys. Hev.
G. F. Imbusch. W. M. Yen, A. L. Schawlow, D. E. 178, 1496 (1969); J. R. D. Copley, R. W. Macpherson,
McCumber, and M. D. Sturge, Phys. Rev. 133, A1029 and T. Timusk, ibid. 182, 965 (1969).
(1964) . 26J. E. Ralph and M. G. Townsend, J. Phys. C 3 8
2~A. E. Hughes, Proc. Phys. Soc. (London) 87, 535 (1970).
(1966). C. C. Klick, D. A. Patterson, and R. S. Knox, Phys.
22C. B. Pierce, Phys. Rev. 148, 797 (1966). Rev. 133, A1717 (1964).

PHYSICAL REVIEW B VOLUME 4, NUMBER 6 15 SE PTE MBER 1971

Thermal Conductivity and Specific Heat of Noncrystalline Solids
R. C. Keller and R. O. Pohl
Laboratory of Atomic and Solid State Physics, Cornell University, Ithaca, Nets York 14850
(Received 28 May 1971)
The thermal conductivity of vitreous SiO&, Se, and silica- and germania-based glasses has
been measured between 0. 05 and 100'K. Comparison with earlier work on noncrystalline
solids shows that they all have the same conductivity within a factor of 5 over the entire tbm-
perature range investigated, with the same characteristic plateau around 10'K, and that their
-
conductivity varies as T", n 1. 8, below T = 1'K. Furthermore, the average phonon mean
free path is large by comparison with the phonon wavelength, about 10 4 cm at 2 'K and de-
creasing as T 4 at larger T, suggesting a Rayleigh-type scattering mechanism. Such a mean
free path can be quantitatively explained by approximating the glassy structure with that of a
crystal in which every atom is displaced from its lattice site. Then every atom scatters like
an interstitial atom, or-even simpler — like one that is missing at its regular lattice site, with
a scattering cross section determined by the missing mass (isotopic defect). The specific
heat of amorphous Si02, Ge02, and Se has been found to vary as AT +BT3 between 0. 1 and
1'K, with A=10 erg/g 'K~ to within a factor of 2. This departure from the Debye specific
heat may be characteristic of the glassy state, as all earlier measurements of other glasses
[polystyrene, glycerol, Lucite (PMMA)] indicate a similar anomaly. Its origin is not clear.
Impurities or surface effects through adsorbed gases are unlikely because of the many sam-
ples and experimental techniques used in different laboratories. We have tried to attribute
the anomaly to low-lying electronic states, motional states of ions, trapped atoms or large
groups of atoms, or one-dimensional vibrations within a three-dimensional solid, so far with-
out success. At the present time, the only independent evidence for these excitations appears
to be in the low-temperature thermal conductivity at T &1'K described above.

I. INTRODUCTION specific heat decreases. The conductivity of crys-

tals depends very much on the material, and in ad-
The thermal conductivity of noncrystalline di- dition any disturbance of the lattice periodicity

talline ones. '

electric solids differs markedly from that of crys-
As an example, Fig. 1 shows the
lowers the conductivity in a sometimes very char-
acteristic way, as shown in Fig. 1, for example.
conductivities of crystalline and vitreous silicon In noncrystalline solids, the conductivity is several
dioxide. In crystals, the conductivity increases orders of magnitude smaller than in crystals, it
with decreasing temperature because the anhar- decreases monotonically with decreasing tempera-
monic umklapp processes become less frequent, ture, and furthermore it is independent of the chem-
and hence the phonon mean free path increases. ical composition: In Fig. 2, the conductivity of
Eventually, the mean free path becomes compara- vitreous silica~' 7 is compared with that of silica-
ble to the sample dimensions, and the conductivity based glasses containing large amounts of other
goes through a maximum and then decreases as the oxides. ' The conductivities are practically
 R, C. ZE LLER AND R. G. PQHL

/". IO
Io"
O. -Quartz
~o II C Axis

IO-'

S
Io

l
IO
I.o IO IOO Iooo
Temperature, 'K

FIG. 2. Thermal conductivity of vitreous silica and

IO-5
silica-based glasses. Vitreous silica: A. (Ref. 5), 8
(Ref. 2), C (Ref. 6), and D (Ref. 7). Borosilicate glass-
es, such as Pyrex (approximate composition by weight:
IOS I I I l l I l I I I
80% SiO, , 13% a, O, , 4% N, O, 2% A.l, O, , 0. 4% K,O, 0. 2%
.I I IO IOO Li2O): E (Ref. 9), E (Bef. 8), 6 (Ref. 2), and (Ref. 8
Temperature, K
13). Crown glass (unspecified composition): (Ref. 10). I
FIG. 1. Thermal conductivity of crystalline and of vit- "Soft" soda-lime glass (approximate composition: 70%
reous Si02 and of crystalline KC1: CN (ncm-=4. 9x10~9 SiO2, 15% Na2O, and 10% Cao): X (Refs. 11 and 12). A
-3 summary of the chemical composition of glasses can be
cm ). Sample dimensions; 5&5~40 mm, sandblasted
surfaces. Heatflow parallel to g axis in the quartz sam- found in Ref. 6.
ple. The data above 100'K for quartz after Eucken (Ref.
1) and for vitreous silica after Ratcliffe (Ref. 6). Note
that an impurity concentration of order 0. 1% in the KCl phonon mean free path / is defined by
crystal results in a conductivity smaller than that of the
glass (the conductivity of pure KCl is similar to that of
quartz). In KCl: CN, the low conductivity is caused by
resonance scattering of the phonons by quasirotational where E is the thermal conductivity, C„ the specific
states with energies close to 1.6 and 18 cm ~ (0. 13 &10 heat of the plane lattice waves, and v the Debye sound
and 1.4 &10 3 eV) (Ref. 4). velocity, /at1 'K is found tobe of order 10 lcm in
glasses, i. e. , 100 times the wavelength of the domi-

identical. Even noncrystalline solids with com-

pletely different chemical compositions, such as l I i

s 14-1v varnishes 18 Se 12 and Geoa 19 all Io

have very similar conductivities (Fig. 3). I

The only exceytions appear to be glass ceram- e

ics ' such Rs Pyroceram, Rnd tile partially cl'ys-

tallizing yolymers, such as nylon and Tef lone'4'~'33 Vit
and polyethylene. ' Their conductivities are consid-
lo

erably smaller below 1 'K than those of the noncrys-

talline solids, they depend on the sample, and they 'aC"
usually vary more rayidly with temperature. The
dependence of the conductivity on the crystallinity cl IO
E
of polyethylene has been studied by Salinger. 3 al

The complete lack of sensitivity to composition in

the noncrystalline solids would be understandable if 1

I.o
I I

IO
I

loo
the conductivity were as low as it could possibly be, Temperature, K
corresponding to a diffusion of the vibrational en-
FIG. 3. Comparison of the thermal conductivity of vit-
ergy from one atom to the neighboring ones. But reous silica with that of the polymer PMMA (Befs. 14-17)
this is not the case. A crystal containing as few sobd varnish (GE 7031) (Ref. 18), vitreous selenium
as 0. 1% of foreign ions can have a conductivity lower (Ref. 12), and vitreous Geo2 (germania) (Ref. 19). The
than that of the glasses (see Fig. i). If an average curve for silica is a composite curve after Fig. 2.
 THERMAL CONDUCTIVITY AND SPECIFIC HEAT OF. . .
nant lattice vibrations or 10 intermolecular spacings.
Qnly Rt the highest temperatures, where glasses Rnd
crystals have approximately equal conduetivities,
is l of the order of the intermolecular spacing,
i. e. , 10 7 em. 2' We will return to this point later
in Fig. 9. Consequently, the problem has to be
stated as follows: Why do all noncrystalline solids
have the same (high) conductivity at low tempera-
tures, and why do Rll the curves have the same
characteristic shape with a plateau in the neighbor-
hood of 10 'K'P This question was the starting point
for the present investigation.
Two models have been suggested to explain the
conductivity of glasses. In the first model,
Klemens proposed tllat the scRtterlng 18 cRused
by spatial fluctuations in the sound velocity and by -7 l l I

a certain amount of long-range order and that this 2 5 lo 20 IOO

affects the transverse phonons more than the longi- Temperature, K

tudinal ones. At temperatures below the plateau FIG. 4. Specific heat of vitleous Si02 and CFy'stal
(10 'K), the heat is mainly carried by longitudinal quartz, plotted as C„T vs T. A: I. R. Vitreosil (Ref.
29) B. vitreous silica (after Refs. 30-32); C &-quartz
phonons, which are uncoupled fxom the transverse {after Refs. 29, 33, and 34). The bvo short dashed lines
yhonons. At around 10 'K, the onset of the anhar- indicate the specific heats calculated from elastic mea-
monic interaction couples the phonons in the dif- surements: For vitreous silica, C„, = 8. 07 & 10 ? (W sec
&

ferent branches and hence lowers the mean free g ~'K 4)T3, calculated from v~„, 1 =4. 1&& 105 em sec (Ref.
,
path of the longltudinRl phonons, which cRu8es the (
25). For crystal quartz Co e& =5. 5+ 10? see g ~ K ~)T
plateau. Eventually, the conductivity increases calculated from v~~ e& =4. 4&&105 cm sec ~ (Ref. 35). See
Rgain becRuse the speclflc heat lncreRses. As Rn also Table I. The dashed curve shows CvT 3 for quartz
in the Debye approximation for a Debye temperature of
alternative model, Klemens" proposed that in non-
552'K (based on the number density of cggoms in the
crystalline solids the heat carrying plane-wave quartz).
phonons are resonantly scattered by local. ized
phonons in a way similar to the resonant scatter-
ing found in crystals containing substitutional model has also been applied to Se" Rnd, GeQ2, and "
molecules with low-f requency quasi. rotational de- it has been concluded that the frequencies needed
grees of freedom (see Fig. 1) or in crystals con- to explain the phonon scattering do agree with the
taining defects causing resonant modes. 28 Such a frequencies of these excess modes.
scattering, in particular, could explain the plateau Both the structure scattering and the localized
occurring in the temperature region in which most mode scattering do not give a plausible reason for
of the heat is carried by yhonons whose frequencies a particularly important experimental fact, namely,
are close to those of the spatially localized modes. that all noncrystalline solids have practically iden-
These modes should be observable in the specific tical thermal conductivities. This fact suggests
heat of glasses. It is known that the low-tempera; that the scattering must have a very simple origin,
ture specific heat of the same substance in the extremely independent of structural details or the

phase. ' This has been observed in SiO2, 29 "

glassy phase is larger than that in the crystalline vibrational spectrum of the solids.
We end our introduction by pointing out a moat
BQ 36
Se 3'?~39 glyeerO1 31' 32v40 Rnd yOlyethylene 41 peculiar fact demonstrated in Fig. 4: At the lowest
As an example, we show in Fig. 4 the specific temperatures investigated, there is no sign that the
heats of SiQ2 in the vitreous and in the crystalline specific heat of the vitreous silica approaches that
phase (n-quartz) plotted as C„T svs T. The spe- computed from elastic measurement:8 in the Debye
cific heats of the two phases of SiQ2 calculated in approximation. Instead, in both samples seems to
the Debye limit from elastic measurements are be an indication of an upturn beginning below 2 'K.
very similar. Therefore, the difference between This behavior appears not to be restricted to SiQ2,
the vitreous and crystalline specific heat has often as shown in Fig. 5 for the polymers polystyrene
been referred to as an excess specific heat and has (PS) and polymethylmethacrylate (PMMA), 4' for
been associated with some extra modes character- glycerol, 46 and for vitreous GeQ2. 7
istic for the glassy phase. Dreyfus, Fernandes, The yuxyose of this yaper is threefold:
Rnd Maynard 2 have suggested that these modes (a) To present an extension of thermal conduc-
could be the localized modes responsible for the tivity measurements on noncrystalline solids to dif-
resonance scattering in the glassy phase. This ferent materials and to even lower temperatures,
2032 H. C. ZE LLE H AND R. O. POHL
IO-'I I 1
foil glued to the sample or to an evaporated gold
film. Through the proper choice of the various
PS
components of this heat leak we obtained the sam-
I

5C
ple to bath time constants shown in Fig. 6. Heat
0 PMMA
1 pulse duration was varied between 0. 5 and 2 sec.
~ IO5
41
The time constants for the thermal equilibration
I/J

between heater and thermometer and the sam-

ple was always found to be less than 1 sec, the
time constant of the Wheatstone bridge used to mea-
Germania
sure the temperature. The time constant 7 for
IO6 thermal equilibrium within the sample was esti-
.I .5 I. 2 5 IO 20
mated from the heat diffusion equation to be
Temperature ( K)

FIG. 5. Specific heat of several amorphous materials r=C„md yZ, (2)

plotted as C„/T vs T: PS and PMMA (after Ref. 45), where C„m and E are the heat capacity and thermal
glassy glycerol (after Refs. 31, 32, 40 and 46), and vit-
conductivity of the sample, respectively, and d is
reous german1a (after Ref. 47). Yhe dashed lines rep-
resent the low-temperature specific heat calculated from a characteristic length, typically a diameter of the
the sound velocities. There is some experimental uncer- sample. For the noncrystalline samples used in
tainty in the data on PMMA, indicated by the two short this investigation, this v was less than 0. 1 sec at
dashed lines. the lowest temperatures, and increased to approx-
imately 1 sec at 1 K. For the crystals, v was
even shorter. The largest source of error was
in order to further demonstrate the striking inde- random building vibrations transmitted to the sam-
pendence of the material. ple, causing random temperature fluctuations and
(b) To suggest a model which can quantitatively hence uncertainties in the determination of the
explain the thermal conductivity in the temperature sample temperature while the sample temperature
range from a few degree Kelvin on up, and to test returned to that of the bath following a heat pulse.
this model on three different substances. These fluctuations caused a maximum error of 20%
(c) To show that the specific heat of noncrystal- in the specific heat below 0. 1 'K, but were negli-
line solids below 1 'K does not follow the expected gible above 0. 2'K. As a test of our technique, we
Debye T law, but instead approaches a linear vari- measured the specific heat C„of single crystal
ation with temperature. This striking behavior Rl)81 and 8101 (Flg. 7). Our results al'e ill good
seems to be characteristic for noncrystalline sol- agreement with earlier data (RbBr, Rollefson
ids, and to be independent of accidental impurities. et al. , ~ T&0. 2'K; quartz, White and Birch, 39
This anomaly may also be responsible for the ther-
mal conductivity below 1 'K. Its origin is yet un-
known. 200

II. EXPERIMENTAL METHODS

IOO—
Thermal conductivity was measured with the
steady-state method in cryostats designed to span
the temperature range 0. 05-100 K, Speclflc heat 50—
U
was measured with the transient heating technique S1
Q
described earlier. ' For this, the sample was in
thermal contact with the heat sink through a heat
leak. After a heat pulse has been applied to the
sample through an electrical heater (25-I1-diam
nichrome, 50 cm long, glued to the sample with IO I I I

0.05 O. I 0.2 0,5 I. O

GE 7031 varnish), the initial temperature increase Temperature, K
of the samples decayed exponentially through the
heat leak. The time constant T» of this cooling had FIG. 6. Some of the experimental time constants v»
to be large compared to the duration of the pulse with which the temperature of the sample returned to that
and to any time constant associated with thermal of the bath after the sample had been heated for a short
time ( 1 see) w'ith a known energy. Excitations which
equilibrium within the sample and between sample
have relaxation times larger than v'» will remain unnoticed
and heater or thermometer (a piece of a 220-0 in this kind of experiment. We did not find that our re-
Speer resistor). The heat leak consisted of a 50- sults depended on the ~,b chosen. We conclude that all
p-thick copper w1re o1 a short piece of 1-mm-diam time constants associated with thermal equilibration with-
graphite rod fastened with Epoxy to a thin copper in the sample were much shorter than ~» (see text).
 THERMAL CONDUCTIVITY AND SPEGIFIC HEAT OF. . .
ity of the samples. The anomaly has been attri-
(0 buted to tunneling impurities, ' present even in the
purest crystals in concentrations below the 1-ppm
/ level. A similar anomaly in A1203 occurring
I
0 around 10 'K has been explained with low-energy
spin states of impurities. ' Although we have not
/
P
I
yet investigated whether the anomaly in Rbar and
in quartz is indeed caused by similar impurities,
we assume that it will be absent in perfect crystals.
In any event, this anomaly is over ten times smaller
/ than the specific heat we will be concerned with in
V -7
glasses.
u»
CL
to
l0
0
j 0
~Quartz
Thermal conductivity was measured on natural
/
quartz (Valpey Corp. ) on pieces of vitreous silica
(GE 201) and Pyrex (Corning VV40) and on a piece
of selenium stringsawed from the specific-heat
IO boule described below. Sample dimensions were
I I I I I
approximately 5x 5x40 mm; the surfaces were
5.0 sandblasted. Also measured was a piece of alu-
Temperature, 'K
minogermanate (Corning QV54) with the following
FIG. V. Specific heat of RbBr and quartz. The solid composition (by weight): GeOz 50/0, A40» 25%,
ines» Cv, RbBr 9 V~ 10 (W sec g F )T and Cv, quarts
=5. 7&10 ' (W'seeg 'K )T agree with those calculated Cao 15%, Bao 5%, and Zn 5%.'
from elastic measurements, and in the temperature range
The specific-heat samples used are listed in
of overlap, also with earlier specific-heat measurements Table I. The quartz was natural material, Spec-
(BbBr, T&0. 2 K, Ref. 50, quartz, T&1.4 K, Ref. 29). trosil 8 was synthetic silica, which according to
its manufacturer contained a total of about 0. 1 ppm
of metallic impurities and 0. 1% weight of hydroxyl
T&1. 4 'K). At the lowest temperatures, below (see below, caption of Fig. 13). A chemical analy-
0. 15 'K in Rbar and below 0. 4 'K in quartz, C„ sis of the Pyrex samples5 xevealed iron concentra-
ceases to vary as T'. For alkali halides, such an
" "
anomaly is known (NaF, KC1, NaBr, RbC1") " tions of 2. Sx19'a cm 3 (100 ppm) in the Pyrex VV40
and 2. Qx10'» cm» (12 ppm) in the Pyrex QVOO. The
and has been shown to decrease with increasing pur- selenium sample was outgassed and melted in a

TABLE I. Specific-heat samples used in this investigation. Also listed is room-temperature mass density, molecu-
lar weight, and the O'K Debye sound velocity ez as calculated from elastic measurements, except for Se {see footnotef).

Temp. Mass
Mass range density Molec. Vg
(g) ('K) (g cm"3) weight (10~ cmsec ~)

-quartz
G'. Valpey 44. 0 0. 8 —2. 0 2. 6 60. 09
0'-quartz Valpey 122. 9 0. 1-0.9
Speetrosil B Thermal American 29. 0 0. 3-2. 0 60. 09
Spectrosil B Thermal American 110.0 0. 13-0.8
Pyrex 7740 " Corning 43. 0 0. 3-2. 0 3. 65'
Pyrex 9700" Corning 46. 0 0. 3-2. 0
Vitr. Se Xerox' 53. 6 0. 1-1.5 4. 25 78. 96 1.54'
Vitr. Ge02 Cornell 57. 3 0. 08-1.8 3.6 104. 6 2. 6~
81Br (cryst. ) Cornell" 39.7 0. 08-1.2 3.43' 165.4 1.59'
~After Ref. 35. 'Sample kindly supplied by Dr. Mark Myers, Xerox
Approximate composition (by weight): Si02 80. 5%, Research Laboratory.
f Since elastic data were unavailable,
8203 12. 9/0, Na20 3. 8%» Al203 2. 2%» K20 0. 4%» and the value listed
Li20 0. 2%. was calculated from the specific heat of the crystalline
After %. P. Mason, in American Institute of Physics phase, see I. C. Lasjaunias, Compt. Rend. B269, 763
Handbook (Mcoraw-Hill, Neer York, 1963), pp. 3-88. 0.969).
Extrapolated to 0 'K. gAfter Ref. 47.
Approximate composition (by weight): Si02 80%» "Same sample as the one studied in Ref. 50.
8203 13%, Na20 5%, and Al203 2%. This sample was After J.
T. Lewis, A. Lehoezky, and, C. V. Briscoe,
kindly supplied by Dr. A. A. Erikson, Corning Research Phys. Bev. 161, 8VV (196V). Mass density quoted is the
Laboratory, one at 4'K.
2034 R. C. ZELLER AND R. O. POHL
quartz tube and quenched in air after melting. ~ The 10 I
l
I I l
I

germania sample was melted at 1250 C in vacuum

in a Pt crucible, kept at that temperature for 18 h ttluminagermanate~
in oxygen at 1 atm, rapidly cooled to 600 'C and ~/
then slowly to room temperature. The sample Ger
10
showed an absorption with a maximum absorption
constant k = 15 cm ' at 2450 A. This defect is asso-
ciated with an oxygen deficiency (F 'center). "
Using Smakula's equation and assuming an oscillator ~/

strength = 1,' we determined a defect concentra-

f ::iO4- Selenium~ /(
tion of 1x10'g cm (5 ppm).
The simplest way to analyze thermal conductivity /9
data is in the dominant-phonon approximation. In
this approximation, one assumes that the heat is I05- 4 0/ Pyrex
carried exclusively by those phonons which con- —Vitreous Silica
tribute most to the specific heat at that temperature,
i. e. , those phonons for which dC„/der is a maximum.
In the Debye approximation, the angular frequency
of the dominant phonons is 5. 65x10" radsec '
-6 / I I I

.I I.O lO

(3 cm ' in the wave-number measure) at 1 'K and is Temperature, K

proportional to the temperature. Thus, if C„and v FIG. 8. Thermal conductivity of vitreous silica, Pyrex
are known, l (va, ) can be determined from the ex- 7740, and selenium between 0. 06 and 100'K and of alumin-
periment with the help of Eq. (1). A simple esti- ogermanate glass between 1 and 100'K. For comparison
mate of the error to be expected if this method is included is the conductivity of vitreous Ge02 (Fig. 3 and
used to determine the frequency dependence of the Ref. 19). Our data on vitreous silica and on Pyrex 7740
agree closely with those obtained by Herman (Ref. 2).
phonon mean free path was made as follows: The
The data on Se are twice as large as those reported in
phonon scattering rates 7 „', (&o) resulting from iso- Ref. 12. As far as we know, this is the only case where
topic mixtures in crystals are known both from the conductivity of a noncrystalline solid has been found
theory and experiment. s %Ye analyzed the thermal to be sample dependent. This difference may result from
conductivity data obtained on Ge crystals of dif- a difference in sample preparation: White et g$. (Ref. 12)
ferent isotopic compositions' with the dominant- quenched their sample in water; our sample was cooled in
phonon approximation and obtained a relaxation air.
rate g „', (or~, ) which varied as the fourth power of
frequency and which differed from the theoretically
predicted rateg6 r, '„(~)by less than a factor of 2. possibly introduced by not using the vitreous spe-
Another test for the validity of the dominant-phonon cific heats will be discussed below, but it is negli-
approximation, involving phonon resonance scat- gible for the present discussion, which deals with
tering, was performed by Narayanamurti and Pohl. " temperatures above 1'K. The sound velocities
used were the ones listed in Table I. At high tem-
III. EXPERIMENTAL RESULTS AND DISCUSSION peratures, the mean free paths for crystalline
and vitreous materials alike approach the order of
A. Thermal Conductivity
a few angstrom, i. e. , the interatomic separation,
The thermal conductivities of vitreous and crys- as is to be expected. As the temperature decreases,
talline SiO~ measured in this investigation have been the mean free path increases very rapidly (almost
compared in Fig. 1. In Fig. 8 we compare the con- exponentially) in the crystal. The mean free paths
ductivities of several vitreous materials. As we in the vitreous samples also increase, although not
pointed out in the introduction, all noncrystalline as rapidly, and with no sign of a resonance scatter-
materials have practically identical conductivities. ing process. In the region around 10 'K, the mean
This is confirmed by our measurements. Our data, free path in these samples varies approximately as
which extend to lower temperatures, show that the the fourth power of the temperature, or in the dom-
conductivities vary as T" with n-1. 8, below 1'K. inant-phonon approximation,
The mean free paths, determined for quartz, vi- l ~ etta, m (3)
treous silica, germania, ' and selenium with the
((d aom) r

help of Eq. (1) in the dominant-phonon approxima- reminding us of a Rayleigh scattering process.
tion, are compared in Fig. 9. The specific heat Static point defects, such as foreign atoms, vacan-
used for this analysis was that of the crystal phase, cies, interstitial atoms, or isotopes in crystal
except for GeO„ for which the specific heat is lattices, cause such phonon scattering.
available only for the glassy phase. 7 The error %e attempt to explain the observed mean free
 THERMAL CONDUCTIVITV AND SPEeIFIC HEAT OF. ..

OOMI
IO
I I
IO
~ ~
IO
I ~
IO I A bers to perform thermal conductivity experiments.
i
On the other hand, it is known from numerous
~ ~
slop
P OOM IOIO 10 IO Io' sec-'
studies that the phonon scattering in crystals con-
I I /
I I
I
I I I
I

taining substitutional foreign atoms is usually quite

IO well described by the mass mismatch alone between
the impurity and the host atom, even if charge
IO compensation in the case of ionic crystals requires
E the incorporation of a vacancy or an interstitial
CP

ion together with the impurity. We refer to the

IO
4
0
Q
work by Walker and co-workers~ and also to a
OP
CD
review on doped alkali haljdes. For an order-of-
tL IO magnitude estimate of the average scattering rate
c by the single intexstitial atom and the vacancy asso-
C
IO ciated with it, we therefore assume that this rate is
O
0 close to that caused by the missing mass alone (the
Q. mass of the displaced atom). With that assumption,
IO
we have reduced the calculation to that of the isotope
effect. The missing mass causes an isotopic Ray-
IO leigh scattering with the scattering rate~s
cat tering
-8
1. '= 0, ra)'/4vn'
IO I I I I I I I I I I
.I I IO IOO 500
Temperature, K Here, Ao is the volume of the individual atoms, v
the sound velocity, and I' measures the mass mis-
FIG. 9. Average phonon mean free path )(T) after Eq.
(1) for quartz and for vitreous SiO2, Ge02, and Se. Spe- match and the concentration of defects:
cific heats used for the evaluation of l(T) divas that of crys-
tal quartz (Refs. 25, 29, and 30), vitreous GeO2 (Ref. 47), I"=Q,f, (1- m, /m)'.
extrapolated below 2 'K vrith the help of the Debye model
(Fig. 5), and crystal Se (Ref. 39). Sound velocities used f, is the fraction of the ith defect, m, its mass, and
are listed in Table I. The dashed lines are theoretical m is the average mass. In our case, f, = 1 and m,
values for the phonon mean free paths calculated from our =0, since the atom is missing. Hence we have
"isotopic" scattering model. At the top of the figure are I'= 1 and
indicated the phonon @wavelengths and frequencies for SiO2
(both vitreous and crystalline} in the dominant-phonon ap- l =4m v /Aolo . (6)
proximation.
For SiO~ and GeO~ the volume Qo chosen was that
of the molecules (3. V6x10 '
and 4. 95x10 cm', "
paths in the noncrystalline systems with the follow- respectively), since these molecules can be re-
ing model: We view these solids as crystals, in garded at low temperatures as the rigid units of
which every atom (or molecule) is displaced from which the solid is formed. For Se, the atomic vol-
its crystal lattice site by a random amount and in a ume 2. 92x10 cm was used. The mean free
random direction. It sits somewhere on an inter- paths obtained from Eq. (6) have been plotted in
stitial site. Hence, the solid consists of as many Fig. 9, in the dominant-phonon approximation, and
interstitial atoms as there are atoms in the solid. agree very well with the experimental values in the
Because of the random displacement, the total tempexature region in which the mean free path
scattering rate of these intexstitial atoms will be varies as T
given by the sum of the scattering rates of the in- One must not take this amazingly good agreement
dividual atoms. This is similar to the phonon between our model calculation and the experiment
scattering by the random mass fluctuations in an too seriouslv. What we believe to have shown is
isotopically mixed crystal: In the most general that the agreement is better than to within an order
case, no atom has a mass equal to the average mass of magnitude. This caution seems advised for the
and the different isotopes are distributed randomly following three reasons: (a) The dominant-phonon
over the lattice sites. In this case, too, the total approximation can introduce an error of about a
scattering rate is equal to the sum of the individual factor of 2, as mentioned in Sec. II. (b) In using
scattering rates. In contrast to the isotopic scat- Ell. (1) to de'tel'111111e 'tile 111eall fl'ee patll, olle must
tering, however, the magnitude of the scattering by use the specific heat of the plane-wave phonons.
individual interstitial atoms or vacancies is not Haman scattering and far-infrared absorption mea-
known. This ignorance is caused largely by the surements have been interpreted as indicating the
fact that it is difficult to produce interstitials or
vacancies in controlled and sufficiently large num- group velocity, "
existence of low-frequency modes of probably low
"the excess modes referred to
2036 R. C. ZE LLE R AND R. O. POHL
To summarize, we have obtained the following
remarkable result: In the temperature range above
IO 5 'K (15 cm '), the average phonon mean free path
decreases as the fourth power of the temperature,
and its magnitude is close to that predicted by the
IO
"isotopic" model. Above 20 K the mean free path
I

approaches the intermolecular separation. The

-5
I
IO
characteristic plateau of the conductivity of non-
Glass crystalline solids also finds a very simple explan-
g IO ation: The phonon mean free path decreases with
increasing temperature, and finally reaches a con-
I
U
stant value. The specific heat, on the other hand,
IO
V
O increases with increasing temperature, and finally
O
reaches a constant value. The plateau occurs in
~IO-
CL

the temperature region in which the product of

mean free path and specific heat happens to be
temperature independent. According to this pic-
ture, the plateau has nothing to do with a resonance
IO
dip, and its origin is actually similar to that of the
P I I I
maximum in the conductivity curve of a crystal.
O. I I 0 IO IOO 500
Temperature, K Below 5 'K, the mean free path changes less
rapidly than as T 4. Below 1 'K in vitreous silica
FIG. 10. Specific heat of vitreous and of crystalline
Si02. Our data (open circles) agree well with earlier data
and Pyrex it varies as T ' . The same tempera-
in the temperature region of overlap. For references see ture dependence is found in vitreous Se below 0. 3 K,
Fig. 4. The dashed line below 0. 4 oK marks the specific and even the vitreous GeO~ data appear to approach
heat computed for crystalline Si02 from elastic measure- the same temperature dependence (between 0. 5 and
ments. 1. 2'K, the mean free path varies as T '4). In
their work on the polymers PS, polyvenylacetate,
and PMMA, Choy et al. found ~0( T-&.4~0. i in the
in the Introduction. In our analysis, we have, temperature interval 0. 4 —1.0 'K. '~ This behavior
therefore, taken as C„ the specific heat of the crys- cannot be explained with our model. It appears as
talline phase, where available, namely, for SiO if a different scattering mechanism becomes dom-
Si02 9' ' 4 and Se, ~ 3 although we are by no means inant in this temperature range. Indications for this
convinced that this is the proper choice. In GeO~, scattering have been reported before for silica-
however, only the vitreous phase has been mea- based glasses and some polymers'": As explana-
sured. 7'6 The lack of knowledge of the appropriate tion, Chang and Jones' and Anderson et al. in-
specific heat introduced a possible error estimated voked phonon scattering by some internal boundaries
not to exceed a factor of 5, the maximum difference of 10 3-cm spacing. Independent evidence for such
between the specific heats of vitreous and crystal- boundaries has never been reported as far as we
line phase, for SiQ2, and Se. (c) Finally, we should know. Also, what kind of boundary scattering would
ask the fundamental question whether it is at all one have to assume in order to explain such a tem-
reasonable to describe the lattice vibrations car- perature dependence of the conductivity 7 These ques-
rying the heat in glasses with plane elastic waves tions must remain unanswered. The specific heat
having a Debye density of states in the temperature of glasses, to be discussed in Sec. III(B), shows an
and frequency range under consideration. Equa- anomalous behavior below 1 'K. Conceivably, a
tion (4) is based on this assumption. With the pres- connection exists between the low-temperature
ently available experimental information this ques- anomaly and the phonon scattering at these tempera-
tion cannot be answered. On the other hand, almost tures. We will return to this point.
any fluctuation from uniform density will produce B. Specific Heat
a scattering proportional (b p/p)2, similar in form
to Eq. (4). We have measured the specific heat of vitreous and
One of the great attractions of the isotopic model crystalline Si02, and of vitreous GeO~ and Se, below
must not be forgotten among all the legitimate con- 1. 5 'K (see Figs. 10—12). Our data show a specific-
cerns, and that is its simplicity. It seems to be heat anomaly below 1 'K in all three vitreous solids.
invevitable that any theory attempting to explain This anomaly is proportional to T in Si02 and GeO2,
the experimental observations will have to be about and, as far as one can tell from the data, also in
as simple as the isotopic model in order to be Se. The first suspicion is that this anomaly is
equally useful for all noncrystalline solids. caused by some impurities. It has been shown by
 THERMAL CONDUCTIVITY AND SPECIFIC HEAT OF. . .
j I l ~ l
j l
C. Possible Origins of Specific-Heat Anomaly

%e begin by comparing the low-temperature spe-

j0 cific heat of vitreous SiO2 with that of two samples
of Pyrex glass (Fig. 15). Pyrex, too, shows an
anomaly, but it does depend on the sample, and it
has the shape expected for a Schottky specific-heat
0
jO
GeOz anomaly. It was first observed by Fisher et al.
'
. Glass in Pyrex 7740 (see uppermost solid curve in Fig.
15). In a magnetic field, the anomaly shifted to
higher temperatures. It was suggested that the
anomaly was caused by the spins of iron impurities.
This was confirmed by our results: %e found that
the anomaly increased with increasing iron concen-
tration. Chemical analysis revealed 100 ppm of
/I iron in Pyrex '7740, and 12 ppm in Pyrex 9V00.
From the entropy associated with the anomaly, we
jO
II
0 I determined a concentration (assuming a two-level
o
system) of 200 ppm and 50 ppm for the two glasses,
in reasonable agreement with the analytic data. In
vitreous silica Hornung et al. s' found no change of
o&
I
j07 ~j I t l I I I l
O. j j.o jO 50 the specific-heat anomaly in a magnetic field of
'K
Temperature, 90 'kG, which speaks against electronic states in
FIG. 11. Specific heat of vitreous Ge02. Our data (open this material.
circles) join on smoothly to the data by Wycherley (Ref. 66) Motional states of ions are another possible cause
above 1.4 K. The latter data, in turn, agree we11 with for the anomaly, since such states are quite insen-
earlier data by Antoniou and Morrison (Ref. 47 and Fig. sitive to magnetic fields because of the large ionic
5) except between 2 and 5'K, where Antoniou's data are
up to 20% smaller than %ycherley's (see Ref. 66}. The
mass. 8'69 Since motional states of iona should be
dashed line is the Debye specific heat calculated from the
sound velocities in vitreous Ge02. I I I g i
j l j

Rollefson that the motional states of F ions in

NaBr, associated with the tunneling motion of the
fluorine ion in the bromine vacancy, can cause a
linear specific-heat anomaly below 1 K. This I
jo

anomaly was found to scale with the fluorine concen- 0

tration, and was practically absent in undoped crys- Vitreou

tals. In contrast, the anomaly in glasses seems to jO

be quite independent of the sample, as shown in Fig.

13, where C„T ' is plotted against Ta for several +
U 4
samples of vitreous SiO~ of different origins, and jO

also in Fig. 14 for GeO~ and Se. The anomalies are 4p

represented by the following expressions: 4P

CL
(0 jO" 5

Si02: 4 C„= (1 x 10 s W sec g ' 'K 3) T, f g'~T

Ge02: DC„=(1.5xl0 ~Wsecg ''K2) T, joe-

Se: EC„=(0.7x 10 6Wsecg ''K 2) T.

jo-y l I i l
.j j.0 jO j00 300
Note that these anomalies are comparable in magni-
Temperature, K
tude to the electronic specific heat of copper, which
is C„=(1.2xl0 ~Wsecg ''K ) T. The most FIG. 12 Specific heat of vitreous and crystalline Se.
Our data (open circles) join on smoothly to the data by
amazing fact about our anomalies is their apparent Lasjaunias (Ref. 39) above 1.4'K. High-temperature data
independence of the substance and accidental im- after Anderson {Ref. 37) and Desorbo (Ref. 38}. The
purities. This suggests a common origin, con- dashed lines are interpolations, drawn for clarity. Calcu-
ceivably the amorphous state itself. In Sec. III C, lations of the Debye specific heat from elastic data are not
we will discuss this further. available.
2038 R. C. ZE LLE R AND R. O. POHL
concentration could be much larger than 5x10'~ cm 3.
Specific Heat —Vitreous In any event, an upper limit of the dipole moment
SIOUX
P = 0. 2 D = 0. 04e A mould have to be assumed in
order to explain the dielectric results and the sye-
cific-heat anomaly with the motion of the same
atoms. Although this is a very small dipole mo-
IO 2
ment (polarizable defects in alkali halides have
moments ranging between 0. 3 and 6 D'~) it is
~O
N
not out of the range of what might be exyected in a
S
4P Ql solid with predominantly covalent bonding. Hence,
Zell
atomic motions cannot be ruled out with certainty
at this point. Dielectric measurements with higher
IO— ~IO erg
gm K~
precision are under way. Fiory's observations do,
however, contain strong evidence against 1ow-lying
states of electrons.
Excitations seen in specific heat below 1 'K have
energies corresponding to less than 1 cm ' in the
I. wave-number measure. Although no evidence for
I.O
polarizable states was found in the kilocycle range,
one might hope to see them in far-infrared optical
FIG. 13. Specific heat of five samples of vitreous sil- absorption measurements. Sievers" searched for
ica, plotted as C„T ' vs T . The samples and their im- this absorption in vitreous silica and in plexiglass
purities, as far as known, are Flubacher et gE. (Ref. 30, (PMMA). At helium temperatures, silica was
optical-quality Amersil, crushed tubing), White and Birch transparent (absorption constant 0 & 0. 04 cm ') f rom
(Ref. 29 and private communication, I. R. Uitreosil). ' to the longest wavelengths studied, cor-
This material has more metallic impurities than Spectro- v =4 cm
sil B, but only about 3-ppm (by weight) hydroxyl f W. C. responding to v=1. 5 cm . This is consistent with
Dumbaugh and P. C. Schultz, EecycloPedkr Of Chemical an extrayolation to longer wavelength of the optical
TecknoEogy, edited by%'. Kirk-Othmer (Wiley, New York, absorption measured by Stolen at p&10 cm ' and
1969), Vol. 18, p. 73j, Hornung ef gE. (Ref. 66, solid a temperature 7=8'7'K. In PMMA, Sievers found
cylinder of Amersil silica), Zeller (this work, Spectrosil
B). According to its manufacturer Spectrosil B contains
less than 0. 1 ppm of Fe and Ca, less than 0. 04 ppm of
IOOO,
Na, Al, and B, less than 0. 004 ppm of Ga, K, Mn, and
P, and less than 0. 0002 ppm of As, Sb, and Cu. It does,
however, contain 0. 1 . % hydroxyl (catalog, Thermal
American Fused Quartz Co. , Montville, J.). In spite ¹

of their different origins, the samples appear to have sim-

ilar low-temperature anomalies.
50

20
expected to influence the electric yolarizability,
Fiory measured the dielectric constant of a piece
'.
e IO-
of vitreous silica cut off from one of our specific-heat Cl 5
samples, at temperatures above 0. 1'K, in the fre- I
I-
quency range 102-10 sec '. The dielectric constant 20- b
20
decreased with decreasing temperature, and was IO— / —IO
/
temperature independent below 0. 5 K to within the /
/
experimental resolution. With the assumption that /~a T~
/
the yolarizable ions have dipole moments of order
1 debye (uncorrected for local fields), Fiory's
measurements lead to an upper limit of 10'6 cm ' .OI
I

IO

for their concentration. From the entropy associ-

ated with the specific heat, a lower limit of about
n=5x10" cm ' can be determined. For this analy- FIG. 14. Specific heat of crystalline and vitreous Se
sis it was assumed that the linear syecific-heat and of vitreous Ge02, plotted as C„T vs T .
Note the
displaced abscissas for the two substances. The specifie-
anomaly did not extend beyond approximately 1 'K, heat anomalies are approximately equal for both samples,
since above this temperature the rapid rise of the and approach a linear temperature dependence. Indica-
specific heat does not allow indentification of a possible tions for similar anomalies were presented earlier, for
extension of the linear term any more. Hence, the PS, PMMA, and glycerol (see Fig. 5).
 THERMAL CONDUCTIVITY AND SPECIFIC HEAT OF. . . 2039

IO silica-based glasses (see Fig. 2) and also the

aluminogermanate glass (Fig. 8) to the list of ma-
terials with a linear specific-heat anomaly.
In summary, all noncrystalline solids studied to
4 7740 date have a low-temperature specific-heat anomaly
Pyiex
0
I

of order 10 ~ (W sec g 'K ~)T or show some indication

io'— D
for it. The excitations associated with this anomaly
0
h
o so far have not been influenced by dc magnetic
-49700 Pyrex
~ / fields and by low- and high-frequency electric
O
l fields. There is the possibility that they can scat-
X b Crysta line I
ter phonons. If these excitations are connected with
O
IO
o~ $jc~ atomic motion, for instance of the kind known for
P Vitreous 0 tunneling defects, they must have very small elec-
t/) SiOp~ gO
0
0 tric dipole moments. Conceivably, large numbers
0
p I/ /
0
of atoms form glassy domains, tied together in the
/
I I /
/
solid with weak forces, and cause excitations in the
1 'E range. The linear specific-heat anomaly could
I /9
// /
I//0 I
IO
I
// i I also suggest one-dimensional chains. Under the
ABC l
arbitrary assumption that the sound velocity along
Temperature, K
these chains equals that in the bulk, it follows that
FIG. 15. Specific heat of SiO& and of Pyrex. Data about 10' chains per cm are needed in order to
points were obtained during present investigation; sam-
explain the magnitude of the observed anomaly. '
ples are listed in Table I. The solid lines are crystal
Si02 (after Ref. 29); top curve: 7740 Pyrex (after Ref. 8);
Because of the low frequencies involved in thermal
solid line just below: vitreous silica (after Ref. 30). The excitations below 1 'K, each chain would have to be
dashed lines are the specific heats calculated from elastic longer than 103 A (see the scale for Xd, in Fig. 9).
(see Table I). A. : pyrex, C„=(1.14x10 6 Wsecg How such linear chains, essentially decoupled from
'data
K ) T; 8: vitreous Si02, C„= (8. 07 x 10-7W sec g 'K ) T; the rest of the solid, can exist in the glass is not
C: crystal Si02, C„=(5.5x10 Wsecg K )T . understood.
One aspect, however, is quite clear: Any model
aiming at an understanding of this anomaly has to
k &0. 04cm ' between 1.5 and 15cm '.
Hence, these be extremely simple in order to be equally applic-
measurements also failed to give evidence for able to a large number, if not to all, noncrystal-
polarizable states. line solids. Herein lies the great attraction of the
At this point, the only other measurements in problem.
which the anomaly may have been observed are the
thermal conductivity measurements described ACKNOWLEDGMENTS
above. Below 1 'K, the conductivity of all noncrys-
talline solids was found to vary with a temperature The first impulse to study heat transfer in glasses
dependence close to T, or T'eto to be ex- T", came from conversations with Dr. K. Dransfeld
act. Assume that the low-temperature specific while on sabbatical leave at Cornell and with Dr. J.
heat of these solids contains a T' background whose W. Wilkins. We want to thank them and also our
magnitude is determined by the sound velocities in colleagues R. B. Stephens, Dr. A. T. Fiory,
the Debye approximation, and that these Debye Dr. J. A. Krumhansl, Dr. ¹ W. Ashcroft, and
phonons are the only excitations which can carry Dr. A. J. Sievers for their help and stimulating
heat. The mean free paths below 1 'K obtained with conversations. We also thank Dr. G. K. White
this assumption, vary approximately as T ' (see and Dr. J. A. Birch for their permission to show
Fig. 9). If the excitations associated with the their unpublished low-temperature data on vitreous
linear specific-heat anomaly act as scatterers for silica (Fig. 18), and Dr. K. E. Wycherley and
the plane-wave phonons, then it would be conceiv- Dr. A. J. Leadbetter for informing us about their
able that the linear increase of the specific heat work on GeO~. We are very grateful to Dr. W.
with increasing temperature would be reflected by Dumbaugh, Dr. A. A. Erikson, and Dr. B. Wedding
a decrease of the phonon mean free path varying of the Corning Research Labortory and to Dr. Mark
as the inverse of the temperature or close to it. Myers from the Xerox Research Laboratory for
With this picture, we conclude that a linear specific- some of the samples used in this work. We also
heat anomaly should also exist in all other noncrys- wish to thank Dr. P. J. Vergano of the Owens-
talline solids whose conductivity has been found to Qlinios Corp. for his advice regarding the prepar-
vary close to T below 1'K. With the help of this ation of the GeO2 sample, and G. Schmidt of the
argument we can add tentatively Pyrex and other Crystal Growing Laboratory of Cornell's Materials
2040 R. C. ZE LLEH AND H. O. POHCY

Science Center for actually producing it. Dr. J. ials Science Center performed the spectrochemical
Both of the Analytic Facility of the Cornell Mater- analyses on the Pyrex.

*Work supported by the U. S. Atomic Energy Commis- Center Report No. 1453 (unpublished).
sion under Contract No. At(30-1)-2391, Technical Report 5Charles Kittel, Phys. Rev. ~75 972 (1949).
Additional support was 26P. G. Klemens, Proc. Roy. Soc. (London) A208, 108
No. NYO-2391-130 (unpublished).
received from the Advanced Research Projects Agency (1951).
2~P. G. Klemens, in Ref. 20, p. 162.
through the facilities of the Materials Science Center at
Cornell University, Material Science Center Report No.
~
R. 0, Pohl, in Localized Excitations in Solids, edited
1556 (unpublished). This work is part of the M. S. thesis by IL F. Wallis (Plenum, New York, 1968), p. 434.
of R. C. Zeller submitted to the Cornell Graduate School 296. K. White and J.
A. Birch, Phys. Chem. Glasses
(unpublished). We also refer to this thesis for a review 6, 85 (1965).
of the work up to 1970 on the thermal conductivity, spe-
3
P. Flubacher, A. J. Leadbetter, J. A.. Morrison,
cific heat, elastic properties, and light scattering and and B. P. Stoicheff, J. Phys. Chem. Solids ~12 53(1959).
absorption performed on noncrystalline solids, Materials
3
F. Simon, Ann. Physik 68, 241 (1922).
Science Center Report No. 1453 (unpublished). F. Simon and F. Lange, Z. Physik 38, 227 (1926).
)Present address: Department of Physics, Northamp- E. F. Westrum, Jr. , in Proceedings of the
I'ourth
ton County Area Community College, Bethlehem, Pa. International Congress on Glass, Paris, 1956 (Chaix,
18017. Paris, 1956), p. 396.
A. Eucken, Ann. Physik 34, 185 (1911}.
3
B. Wietzel, Z. Anorg. Allgem. Chem. 116, 71 (1921).
. , ( 3~0. L. Anderson, J. Phys. Chem. Solids 12, 41 (1959).
y
38ee Charles Kittel, Introduction to Solid State Physics, E. F. Westrum, Jr. , in Thermodynamics and Trans-
3rd. ed. (Wiley, New York, 1966), Chap. 6. Port Properties of Gases, Liquids and Solids (McGraw-
Wo Do Sewal d and V. Narayanamurtl, Phys e Reve Hill, New York, 1959), p. 275.
463 (1966). YC. T. Anderson, J. Am. Chem. Soc. 59, I036 (1937).
5R. A. . Fisher, G. E. Brodale, E. W. Hornung, and W. Desorbo, J. Chem. Phys. 21, 1144 (1953).
W. F. Giauque, Rev. Sci. Instr. 40, 365 (1969). 39Jean-Claude Lasjaunias, Compt. Rend. 269, B763
6E. H. Ratcliffe, Glass Technol. 4 113 (1963). (1969).
H. E. Seeman, Ph. D. thesis (Cornell University, 4oG. E. Gibson and W. F. Giauque, J. Am. Chem. Soe.
1927) (unpublished). 45, 93 (1923).
8R. A. Fisher, G. E. Brodale, E. W. Hornung, and 4~J. E. Tucker and W. Reese, J. Chem. Phys. 46,
W. F. Giauque, Rev. Sci. Instr. 39, 108 (1968). 1388 (1967).
A. C. Anderson, W. Reese, and J. C. Nheatley, Rev. B. Dreyfus, C. Fernandes, and R. Maynard, Phys.
¹

Sci. Instr. 34 1386 (1963). Letters A26, 647 (1968).

G. K. Chang and R. E. Jones, Phys. Bev. 126, 2055 3J. C. Lasjaunias and R. Maynard, J. Noncryst. Solids
(1962). (to be published).
R. Berman, E. L. Foster, and H. M. Rosenberg, 44J. Blanc, D. Brochter, J. C. Lasjaunias, R. May-
Brit. J. Appl. Phys. ~6 181 (1955). nard, and A. Bibeyron, in Proceedings of the Twelfth
~26. K. White, S. B. Woods, and M. T. Elford, Phys. International Conference on I osv Temperature Physics,
Bev. 112, 111 (1958). Kyoto, 1970, edited by Eizo Kanda {Academic Press of
~~R. W. B. Stevens, Phil Mag. 14, 897 (1932). Japan, Kyoto, Japan, 1971), p. 827.
~4R. Berman, Proc. Roy. Soc. (London) A208, 90 45C. L. Choy, R. G. Hunt, and G. L. Salinger, J.
(1951). Chem, Phys. ~52 3629 (1970).
~5W. Reese, J. Appl. Phys. 37, 864 (1966}. 46R. S. Craig, C. W. Massena, and R. M. Mallya, J.
~6K. Eiermann and K. H. Hellwege, J. Polymer Sci. Appl. Phys. 36, 108 (1965).
~57 99 (1962); and K. Eiermann, Kunststoffe 51, 512 4~A. A. Antoniou and J. A. Morrison, J. Appl. Phys.
(1961), 36, 1873 (1965).
"C. L. Choy, G. L. Salinger, and Y. C. Chiang, J. J. P. Harrison, Hev. Sci. Instr. 39 145 (1968).
49H. S. Carslaw and J. C. Jaeger, Conduction of Heat
Appl. Phys. 41, 597 (1970).
~SJ. H. McTaggart and G. A. Slack, Cryogenics, 9 384 in Solids (Oxford U. P. , Oxford, England, 1948), Sec. 36.
(1969). 50R. J. Hollefson, P. P. Peressini, and C. T. Alex-
Kurt Guckelsberger and Jean-Claude Lasjaunias, ander, (unpublished).
Compt. Rend. 270, B1427 (1970). 5~J. P. Harrison, G. Lombardo, and P. P. Peressini,
20J. H. A. Laudy, in Physics of Non-Crystall&re Solids, J. Phys. Chem. Solids ~29 557 (1968).
edited by J. A. Prins (North-Holland, Amsterdam, 1965). ~2R. J. Rollefson, Ph. D. thesis (Cornell University,
W. Heese and J. E. Tucker, J. Chem. Phys. 43 105 1970) (unpublished); Cornell Materials Science Center Re-
(1965). port No. 1382 (unpublished).
R. L. Powell, W. M. Rogers, and D. O. Coffin, J. For a review, see V. Narayanamurti and R. O. Pohl,
Bes. Natl. Bur. Std. (U. S. ) 59 349 (1957). Rev. Mod. Phys. 42, 201 (1970).
23G, L. Salinger, in Proceedings of the Third Interna- ~4R. Q. Fugate and C. A. Swenson, J. Appl. Phys. 40,
tional Conference on the Physics of Noncrystalline Solids, 3034 (1969).
. edited by R. W. Douglas and B. Ellis (Wiley, New York, 558ample kindly supplied by Dr. W. Dumbaugh, Cornell
to be published). Besearch Center.
4For a review, see R. C. Keller, M. S. thesis (Co»ell 56Performed by Dr. J. Both, Analytical Facility, Cor-
University, 1971) (unpublished), Cornell Materials Science nell Materials Science Center.
 THERMAI CONDUCTIVITY AND -SPECIFIC HEAT OF. . .
5'Sample prepared by Dr. Mark Myers, Xerox Research . H.
64+, Stolen, J. Phys. Chem. Glasses 11 83 (1970).
Laboratory. 8~P. T. T. %'ong and E. Trans. Faraday Soc.
Whalley,
58V. Garino-Canina, J. Appl. Chem. Solids ~20 110 (to be published).
(1961). 66K. E. %ycherley, Ph. D. thesis (School of Chemis-
59P. J. Vergano .
and D. 8, Uhlmann, Phys. Chem. try, University of Bristol, England, 1969) (unpublished).
Glasses 11, 30 (1970). 6~K. W. Hornung, 8,. A.. Fisher, G. E. Brodale, and
60J. Callas'ay, Phys. Hev. 113, 1046 (1959). %'. F. Giauque, J. Chem. Phys. 50, 4878 (1969),
6~T. H. Geballe and G. %. Hull, Phys. Hev. 110, 773 68R. A. Brand, S. A. Letzring, H. S. Sack, and%. %'.
(1958). %'ebb, Bev. Sei. Instr. (to be published).
6
J. A. Harrington and C. T. Walker, Phys. Hev. 81, 69A. T. Fiory, Hev. Sci. Instr. (to be published).
882 (1970); J. W. Sehvrartz and C. T. Walker, Phys. ~OA. T. Fiory (unpublished data).
Rev. 155, 759 (1967). '~A. J. Sievers (unpublished data).
63Marvin Hass, J. Phys. Chem. Solids 31, 415 (1970).

PH YSICA L RE VIE%' B VOLUME 4, NUMBER 6 15 SEPTEMBEg 1971

Volume Dependence of the Gruneisen Parameter of Alkali Halides~

R. W. Roberts and R. Ruppin*
Department of Physics, University of North CaroHna, Chapel Hill, North Carolina 27514
(Received 22 April 1971)
The parameter q = (& lny/8 lnV)~ vrhich describes the volume variation of the Gr0neisen pa-
rameter has been calculated by the following turbo independent methods: (a) a macroscopic cal-
culation using a thermodynamic formula and carefully selected thermodynamic data; (b) a
microscopic calculation using a six-parameter pressure-dependent lattice-dynamical shell
model. The first method has been applied to twelve alkali halides and the second to only six
alkali halides for which experimental data on the second-order pressure dependence of the
elastic constants are available. Typical values of q have been found to be equal to about 1.5.
The low-temperature limit of q has been investigated by using an acoustic-continuum model.
It was found that q increases sharply at lour temperatures.

I. INTRODUCTION sonable approximation that q= 1. The assumption

that y is proportional to the volume (which implies
The parameter q = (s Iny/s ln V)r, which describes q= 1) seems to be the most popular. It has been
the volume dependence of the Gruneisen parameter used in theories of shock-wave propagation' and
(y= VPBr/C„) emerges frequently in discussions of also in various discussions of equations of state. ~ v
the thermoelastic properties of solids. Since no Another approach has been adopted by Rice~ in his
theoretical microscopic calculations of q have been discussion of the alkali metals. By making the as-
performed in the past, except for a recent w'ork on sumption that the Griineisen parameter, as well as
inert-gas crystals, some simple assumptions about the adiabatic bulk modulus, is a function of volume
its numerical value have usually been made. Most only, he was able to derive a simple explicit ex-
often, q appears in relations which also involve the pression for the dependence of y on volume. In
Anderson-Griineisen parameters, which describe terms of the parameter q, his relation is q= @+1.
the temperature dependence of the bulk moduli and A slgniflcRnt feature of the pRx'Rmetex' q ls thRt
are defined by ' it can be calculated directly from lattice-dynamical
models. A calculation of this type, based on a six-
1 8 lnBs parameter shell model, is presented in Sec. ID.
The results Rx'e compared w'1th q values obtRlned
in Sec. II from room-temperature ultrasonic and
(8 lnB) thermodynamic data. The low-temperature behav-
ior of q is discussed in Sec. IV.
Here, Bs and B~ are the adiabatic and the isother-
mal bulk moduli, respectively, and I|I is the coeffi- II. EVALUATION OF q FROM THERMODYNAMIC AND
ULTRASONIC DATA
cient of volume expansion. Some formulas for 6s,
which have been given in the past, were based on The most convenient expression for q in terms
the assumption that q «1, which does not hold fox' of available experimental quantities is
the alkali halides, as will be shown later. Ander-
son, on the other hand, after giving expressions fox' c) in@
q= I+ (I+ TPy)bg —Bg+y+ 7
5s and 5~ which involve q, employed the more rea- 8T y