THERMAL COKDUCTIVITY OF LIQUIDS

3017

Thermal Condilnctivity of Liquids

by Sheng Hsien Lin, Henry Eyring, and Walter J. Davis

Department of Chemistry, University of Utah, Salt Lake City, Utah (Received J u n e 1 , 1965)

The significant structure theory of liquids is applied to calculate the thermal conductivities of liquids. Pressure and temperature effects on thermal conductivities are discussed. For the three substances, nitrogen, argon, and methane, the discrepancy between the calculated and experimental results is less than 4y0.

I. Introduction Application of the method of significant structure theory to both thermodynamical and transport properties of liquids has been made with good The purpose of this paper is to examine the thermal conductivity of liquids based on the significant structure model, ie. K1
=

of change due to scattering processes indicated on the right-hand side. Since the exact solution of eq. 2 is i~npossible,~ approximation is made by putting6 an

(V,/V)K,

+ (1 - V5V)Kg

(1)

N=Nfl+n (3) Where N Q is the equilibrium distribution, Nfl = - l)-. It is apparent that if the temperature gradient is small, then n << NO, so dn/dT << dNO/dT thus
C ; i T dNo,fdT = b N / d t

where K I is the contribution to the liquid thermal conductivity, K , is the Contribution by the solidlike or lattice vibrational degrees OF freedom, and IClz is the contribution by the gas-like or fluidized degrees of freedoin. For the calculation of K,, we use the phonon theory first developed by Peierls4 6 and then modified by Kleniene6 and C a l l a ~ a y . Since the con-. ~ tribution of K , to the liquid thermal conductivity irr sinall, generally less than 5% except in the critical region, we use the idleal gas formulas for the calculation of K,. Three substances-argon, nitrogen, andl methane-are chosen for calculating the liquid therim I conductivity. Pressure and temperature effects on the thermal conductivity are also discussed.

(4)

In t e r m of phonons, the heat current is just the energy flux due to the transport of phonons each of which + carries energy hw, so if we let f(k)dk represent the number of nornial modes in the wave-number interval and dc, we obtain for the heat flux in the direc+ tion of V T

11. Thermal Conductivity of Solids When a temperature gradient is applied to a solid, the equilibrium state will be disturbed, and the average distribution of phonons should be determined from the Boltzmann equation
- C,

where the suinmation is over different polarizations. + Since N,O(Ic) == N,O(-k), N,(lc) in eq. 5 can be re+ placed by n,(Ic),i . e . , only the deviation n3 contributes to the heat flux, because the integration in eq. 5 over
(1) G. Blomgren, A . m . S . Y . Acad. Sci., 79, 781 (1959). (2) C. M. Carlson, Theory of Fused Salts and Molten Metals. Doctoral Dissertation, University of Utah, 1960.

grad T d N / d T

blV/bt

(2)

where N is the number of phonons per unit volume and C, the group velocity of solids. The left-hand side of eq. 2 represents the rate of change of N due to the transport of phonons, and this is balanced by the rate

(3) E. J. Fuller, Significant Liquid Structures, Doctoral Dissertation, University of Utah, 1960. (4) R. Peierls, A n n . Phys., 3, 1055 (1929). (5) R. Peierls, Quantum Theory of Solids, Oxford Cniversity

Preas, London, 1955. (6) P.G. Klemens, Proc. R o y . SOC.(London), A208, 108 (1951). (7) J. Callaway, P h y s . Rei.., 113, 1046 (1959).

Volwme 68, S u m b e r 10 October, 1,904

3018

S. H. LIN, H. EYRIKG, W.J. DAVIS AND

the equilibrium distribution N t vanishes. This is generally the case for the transport phenomena in nonequilibrium statistical mechanics. Next we assume that the discontinuous processes tend to obliterate exponentially any deviations from equilibrium, L e . dN,'dt
=

-(N

- N")/r =

-TL/T

(6)

where I , = rjC,, the relaxation path length. and C,, = (NC&)/V,, the heat capacity of solids per unit volume for the j t h polarization. Equation 12 has been used by Kittelll to explain the thermal conductivity of glasses at' low temperatures. If no distinction is made between polarizations, eq. 12 can be written as

where r is an over-all relaxation time and is assumed to be dependent on w and T . For an isotropic solid, we may use spherical coordinates in k-space with the axis of symmetry in the direction of temperature graf dient. I we let e be the angle between 12 and grad T , eq. 4 becomes, after substitution of eq. 6 into eq. 4
n
=

TC, cos 0 dT/dZ X dNo/dT

where R = N k ,the gas law constant, and C, is the sound velocity of solids. As with Klemens,Gwe assume that each scattering process can be represented by a relaxation time, i e . , 1 / ~ = l / r ~and existence of the following processes ,
1

(rC, hw)Cph COS 0 d T i d Z

(7)

where C p h is the specific heat of phonons. Combining eq. 7 and 5 and integrating the angular parts of eq. 5, we obtain

Q
or

4./3

dT/dZ
3

Sr,(k)C,,(k)'C,h~(k)f(k) dk

(8)
(8a)l

K,

4 ~ / 3 Sr,(k)C,,(k)2Cph,(k)f(k) dk
3

is assumed: (1) normal three-phonon processes where the relaxation timeg is taken to be proportional t o ( w * T ) - ~(;2 ) the Uinklapp process with a relaxation time proportional to (e-"/TozT) where a is a constant -l characteristic of the vibrational spectrum of the material; (3) lattice imperfection scatteringG"' whose relaxation time is independent of temperature and proportional to m-*. The other processes such as size effect, etc., are neglected because we are interested only in high temperature regions. Thus we can write
r-l

Ao4

Because of lack of knowledge of crystal vibrational spectra, it is impossible to integrate eq. 8a exactly. If Debye's model is used, then we have 4nf(k) dk
=
3 -

+ BTw2

(14)

4 . v2dv C,

= __-

471.02ddw c3(a. 3 , )

(9)

where A o 4 represents the scattering by lattice imperfections, and the term B T d includes the normal and Umklapp processes. Combining eq. 14 and 13 we obtain

where v is the frequency. If no distinction is made between longitudinal and transverse phonon^,^ eq. 8a becomes

111. Thermal Conductivity of Gases

Various formulas for the thermal conducfivity of gases I<, are derived from kinetic theory.12 The general results are usually expressed as

where is the over-all relaxation time, and omax the Debye cut-off frequency. Equation 10 has been used to explain the thermal conductivity of solids a t low temperature s ~ c c e s s f u l l y . ~ - ~ ~ From the specific heat theory of solids, we know that in the high temperature region the Einstein model is a good approximation. With this approximation, we have 4 ~ f ( k dk )
=

K,

eqKCve

(16)

where E is a pure number, 7, the viscosity, and C,, the specific heat a t constant volume. As we already mentioned, the contribution of gas thermal conductivity to the liquid thermal conductivity is generally small For our present purpose, therefore, we may use the
(8) B. A. Agrawal and G. S. Verma, Phgs. Rez., 126, 25 (1962). (9) J. Callaway and H. C. von Baeyer, ibid., 120, 1149 (1960). (10) P., Carruthers, Rev. M o d . Phys., 33, 92 (1961). (11) C. Kittel, Phgs. Rec:., 7 5 , 972 (1949). (12) L. B. Loeh. "Kinetic Theory of Gases." McGraw-Hill Book Co., Inc., New York, N. I-., 1927, Chapter G .

N ' 1 1 , 6 ( ~-

we)

do

(11)

where w e is the Einstein characteristic frequency, N is Avogadro's number, and S(o - w,) Dirac's &function. Substituting eq. 11into eq. 8,l \$-eobtain
The Journal of Physical Chemistry

THERMAL CONDUCTIVITY

OF

LIQUIDS

3019

ideal gas equations for E, vn, and CIp, namely, E = (9y 5j/4, where y is the ratio of specifc heat at constant pressure to that at constant volume, and vS = mE/ (34271.~79, where m is the mass of a molecule, ~7 the molecular diameter and the average velocity, E = ( 8 h - T / ~ 1 n ) ~ For the temperature and pressure ~. ranges of concern, K g is less than or around 5 X cal./cm. deg. for all three substances. For the rigorous treatment of K,, one should refer to the literature. l 3 , l 4

r
I

IV. Thermal Conductivity of Liquids Substituting eq. 15 and 16 into eq. 1, we obtain the equation for the liquiid thermal conductivity as
K1 =

80

120

160

I 200

O K

RC, (huekT)ZPekT V - V , -~ ErlgCVg V(A + BIT) ( e h w e / h T V

Figure 1. Temperature dependence of K1 of argon: , experimental values; A, calculated values; curve 1. P = 48 atm., curve 2, P = 120 atm.

(17) where A = Arne4and B = Brne2. The first term on the right-hand side cif eq. 17 represents the contribution to thermal conduction due to the vibration of molecules sitting near equilibrium position and the second term, the contribution due to the random motion of molecules. For kw,/kT < 1, we may put x 2 e x / ( e x- 1j2 = 1 ; this is generally the case for the liquid therrnal conductivity. Furthermore, if the pressure is not very large, C, is constant, and it is convenient to absorb C, and R into constaints A and B. The temperature range considered in this paper is from the normal melting point to the critical temperature for the three substances. For argon, we use P := 12 atm., T = 106.1K., K , = 2.44 X cal./deg. cm., and P = 12 atni., T = 93.6OK., K1 = 2.81 X cal./deg. with C, = 1310 im./sec.l6 to calculate the parameters A and B. They are 2.50 ;< and 5.52 X for A and B , respectively, using we = 7.85 X The corresponding values obtained for nitrogen and methane are A = 1.82 >< lO-*O, B = 5.64 X and A = 1.60 X B == 1.77 X respectively. Below the critical ternperatures, the liquid thermal conductivity is nearly a linear function of temperature at constant pressure as shown in Fig. 1 and 2. To consider the pressure effects on the liquid thermal conductivity, we USE C, = ( y , / p , / ? ~ ) ~ to estimate the dependence of C, on the pressure, where y, is the ratio of specific heats of solids, p,, the density of solids, and PT, the isothermal compressibility coefficient of solids. C, does not vary very much with pressure, so we have to consider the pressure dependence of C, only a t very high pressures. Because of being limited by the availability of data for @I, only two values of Kl under the pressure of 8 0 0 K , / c n ~have been calculated for argon. ~ The data used for the calculation are P = 0, PT = 6.2

4,00r

80

120 T
OK

I60

Figure 2. Temperature dependence of Kl of nitrogen: , experimental values; A, calculated values; curve 1, P = 33.5 a t m . ; curve 2, P = 134 atm.

X ~ n i . ~ / K and P = 8 0 0 K , / ~ m . ~ , = 4.36 x ,, PT 10-j c n ~ . ~ / K , .V ~ 28.89 ~ i n . ~ / m o l e T = 119.6 ~= at OK.; V = 27.54 ~ r n . ~ / m o l at ~ = 102.2 OK. The el T results obtained by using eq. 17 are K 1 = 3.30 X cal./deg. cni. at P = 8 0 0 K , / ~ n i . ~T = 119.6OI<., , cal./deg. cm. at P = 800K,/ and K1 = 3.91 X cni.2, T = 10%.2OK.,in comparison with the extrapolated experiiiiental values of Rice.Ig K 1 = 3.40 X
(13) J. 0. Hirschfelder, C. F. Curtiss, and R. B. Bird, Molecular Theory of Gases and Liquids, John Wilzy and Sons, Inc., New York, N. Y . , 1954. (14) S.Chapman and T. G. Cowling, The Mathematical Theory of Non-uniform Gases, Cambridge University Press, London, 1953. (15) H. Ziebland and J . Burton, Brit. J . A p p l . Phys., 9, 52 (1958). (16) J. W. Stewart, J . Phys. Chem. Solids, 1, 146 (1956). 117) E. J. Fuller, T. Ree, and H. Eyring, Proc. S a t l . Acad Sci. U . 8 , 45, 1594 (1959). (18) P. W. Bridgman, Proc. Am. Acad. A r k Sci., 70, 1 (1935). (19) L. D. Ikenberry and S. A. Rice, J . Chem. Phys., 39, 1961 (1963).

Volume 68, Yumber 10

October, 156.4

3020

S.H. LIN, H. EYRISG, W. J. DAVIS AXD

3.00
rs

1 -,
I .
I

Table I : Results of Calculation"

I
b

Temp., OK.

Pressure, atm.

K I (calod.) X 104, Ki (exptl.) X 104, cal./deg. cm. cal./deg. om.

Argon' 93.3 93.3 93.3 48 72 120 48 48 48 19.3 72 48 48d 774 774 2.87 2.89 2.96 2.68 1.84 2.1c 1.89 2.01 1.54 0.914 3.91 3.30 2.90 2.93 3.04 2,70 1 .793 2.11 1.88 1.99 1.48 3.80" 3 , 4OC 3.14 2.97 2.90 2.60 2.17 2.25 2,46 1,75 3.26 4,946 5.06 4.096 3.176 3.284

2.00
I

'
I
I

100
130 120 125.7 125.7 140 150.7d 102,2 119.6 87.7 87.7 87.7 98.3 105.8 105,8 105.8 116.1 126, Id 80 99.0 99.0 125.6 150.3 150.3

_-_

Figure 3. Pressure effert on K1 of argon: experimental results; A, calculated values; curve 1, T = 93.3"K.; curve 2, T = 106.lOK.
~

3.00

A' A

.-

P atm
Figure 4. Pressure effect on Ki of nitrogen: ---, experimental results; A, calculated values; curve 1, T = 87.7"K.; curve 2, T = 105.6"K.

Nitrogenb 134 3.18 67 2.93 33.5 2.88 67.0 2.53 33.5 2.20 50.3 2.24 100.5 2,50 33.5 1.83 33, 5d 1.04 33.5 3.22 1.6 50.5 50.1 25.9 49.7 Methane" 4.97 5.01 4.05 3.26 3.30

cal./deg. cm., K1 = 3.80 X cal./deg. cm., respectively. The liquid thermal conductivity us. pressure for pressures below 140 atin. are shown in Fig. 3 and 4. It can be seen that for small ranges of pressure, K I is linear with respect to pressure. Sonie of the calculated results of Kl for the three substances are tabulated in Table I, in comparison with the experimental values. The discrepancy, as can be seen from the table, is less than 4y0. In concluding this paper, it is interesting to notice that Ks's calculated from eq. 15 by using the values of A and B obtained from liquids are in good agreement with the experimental values of K,; for example, a t P = 180 atm., T = 83.6'K., K , (calcd.) = 4.29 X cal./deg. cm., K , (exptl.) = 4.27 x cal./deg. cni.,15and a t T = 77OK., K , (calcd.) = 4.22 X cal./deg. cm., K , (exptl.) = 4.20 x 1 0 - ~ cal./deg. cm. If this is the case, we may assert that the structure of the solid part of liquids is not much different from that of solids at high temperatures, but in view of the limited amount

a V , = 22.5 ml./mole for argon, 26.14 ml./mole for nitrogen, Experiment'al data of Ki and 30.90 ml./mole for methane. are from H. Ziebland and J. Burton, Brit. J . A p p l . Phys., 9, 52, (1958). Experimental data of K I are from L. D. Ikenberry and S. A. Rice, J . Chem. Phys., 39, 1961 (1963). The wittical point. The data used to calculate A and B are P = 8.4 atm., T = 88.2"K., K I = 2.79 X 1 0 - 4 , a n d P = 27.6atm., T = 80.9"K., K l = 3.18 X lo-* for nitrogen, and P = 25.3 atm., T = 99"K., K I = 4.999 X and P = 24 atrn., T = 125.6"K.,Ki = 4.016 X for methane.

'

of data of K , at high temperatures (2' > eo, Debye's characteristic temperature) which are available, it seems that further justification is necessary.
Acknowledgment. Acknowledgment is made to the Army Ordnance under Contract DA-ilRO(D)-31124-G298 for financial support of this research.

The Journal of Physical Chemistry