CHINESE JOURNAL OF PHYSICS

VOL. 18, NO. 2

SUMMER, 1980

Iteration Method for the Calculation of Phonon-Limited Thermal Conductivity of Normal Metals
W. C. C H A N (@$$i~.)
Departmenr of Phyks Tamkang College of Arrs and Sciences Tamsui, Taipei Hsien, Taiwan 2S1, China

(Received February 29, 1978) A simple linear Boltzmann equation for the calculation of phonon-limited thermal conductivity of normal metals has been obtained. It has been solved with the generalized relaxation time approximation. At high temperature, it gives the Wiedemann-Franz law. At low temperature, it gives a Tz dependence for the thermal resistivity.

1. INTRODUCTION
NTIL now, thermal conductivity of nornian metals has been calculated most frequently by the But there are always doubts about the use of the first trial function, especially in the low temperature region(2y3). Another approach is io solve the complex Boltzmann equation numerically(4t5). It is intent here to reduce the complex Boltzmann equation into a simple form and then solve it algebraically.

u.J use of the variational method) .

2. GENERAJ, CONSIDERATIONS AND THE BOLTZMANN EQUATION The electrical and electronic thermal currents in a metals are given, respectively, by(l*a) J,-aE--usVT, J, - SUTE + LTTVT, (1) (2)

where G, s and LTT are the electric conductivity, thermalpower and thermalelectric coefficient respectively. For thermal conductivity (K) measurements, J,=O, and so K== -(LTT+s%T). Since &T/n. is of the order lo- or less for most metals, one can take approximately K== - Ly*. (4)

(3)

The linearized Boltzmann equation with zero external electric and magnetic fields can be written)

. ..-_____ .._-- ~~~ (1) See, for example, J. M. Zimam, Electrons and Phonons (Oxford U. P. Oxford, England, 1960). Chap.
7 and Chap. 9.

(2) E. II. Sondheimer, Proc. Roy. Sot. A203, 75 (1950).

(3) (4) (5) (6) (7) J. W. Ekin, Phys. Rev. BG, 371 (1972). P. G. Klemens, Aust. J. Phys. 7, 64 (19S4). C. 17. Leavens, J. Phys. F: Metal Phys. 7, 163 (1977). * Ii. S. Newrock and B. W. Maxficld, Phys. Rev. B7, 1233 (1973): F. J. Blatt, Physics of Electronic Conduction in Solids (New York, blcGraw-Hill, 1968) p. 183. 61

62

ITERATION METHOD FOR THE CALCULATION

1 gj(k, k) 1 I $ l&/,(e)

[l--fo(E+~~dl (E+hwqj-ri)}
(5)

+ (analogous terms involving emission processes),

where gj(k, kf) is the matrix element for the scattering of an electron from k to kf through the absorption or emission of a phonon with wave vector q, polarization vector e,,j and energy Aw,j, nr and &(E) are the equilibrium distribution functions of the phonons and electrons respectively, 0 is the Fermi energy and R is the mean displacement travelled by the electrons between collisions. All integrals performed over the energy surface S tr are subjected to the constraints E-E+F2O,j for absorption processes and e=E-/?ti,j for emission processes. Just for the sake of convenience, let us define two new sets of functions Ij(k) and I;(k) such that

I,(k) =

J/

gl(k, k) I3 R -$

(6)
(7)

I;(k) - *I gj(k, k) jz R $; .I

In order to eliminate the rapidly ,.,;;:+ing functions (af0/8e) and fo(e)[l--f,(e)], we integrate both sides of Eq. (5) with respect to E. \sslme v and all I,(k), I;(k) are smooth functions of e in comparing with (afo/&) and fO(e)[l --/o(E)] and expand them by Taylor series in terms of (E-T) up to first order. Then, we have,

*(l/3) (zkJ) 2 @v/L%>

(8)

J Ij(k>fdt) Cl -fo(~+h~qj)l
S(kBT)
F,(X)

(E-VW

= . Jpj(k,)+$m (E-_i]f~(t)[l-f~(~+fiO,jl(E--O)dt

J I;(k) f&J [l-~(~+h~,j)l(~+h~,j-~)~~

and for emission processes, we have

Ij(k,)+(k,T)3 Fz(x)(aIj/ac)

(9)

= (kBT)Z [-F,(r)] I;@,) + (k,TY [F,(x) +xFdx)l -FE

(10)

J Ij(k)f~(~)[1--f~(~-h~,j)l('-_77)"
I'I;(k)fo(~) [l-fo(~-h~aj)l
where x represents hWqj/k,<T and , I

= (kBT) [ - F,(X)] e- Ii + (kB7) F,(X) e-I -a,; (~-h@qj-V)d~

+xF~(x)] e-I q>

(11)

= (k,T)2 FL(x) e-s I;(k,) + (kBZ) [F2(x)

(12)

(13) .N WkJ__ ydy (e;+~l)(,+e--s) x(x2+ 7rJ) -3(=1 -e-x) (14)

sr ot

1 I

After putting Eq. (S), Eq. (9), Eq. (IO), Eq. (ll), Eq. (12), Eq. (13),and Eq. (1-l) into Eq. (5). we have 8 , .::, ?n ~~-+$ Jlgj(k,k')1*~(x)[(~-~~)+(~-)2(~+-:-~~)]~~~(19 --zF - ii

Tt -.. (1

W. C. CIIAN where the factor F(x) is given by(a) F(x)=x/(e=- 1)(1-P) (16)

63

The sum the first order terms in (6-q) on the right hand side of Eq. (5) is zero after integrating over E because absorption and emission processes cancel each other exactly. Integrating both sides of Eq. (15) with respect to E once more, one finally obtains the simple Boltzmann equation (17) Here, the integration constant is chosen to be zero because the Boltzmann equation is supposed to have a unique solution. 3. RELAXATION TIME AND THERMAL CONDUCTIVITY CALCULATIONS To solve Eq. (17), we apply the generalized relaxation time method as used by Huntington and Chan(g~lo). The generalized relaxation time T)(k), defined by i,,/v, for a particular direction P, is determined from a series

,th ,m-

b to

idk)-

iz rdk)

(18)

by an iteration scheme based on the assumption that Tp(k) is not a rapidly varying function of k. The leading term is given by rpo(k) - -h- T and successively

Jpas. IgLk,k) IF(x)[ (I--$--) + ($)(I+ ;$)I z:-

(19)

3)

x L.rpn-l(k) - z- pn-l(k)1

-$;-}

(20)

From the microscopic transport equation for the electronic thermal current

,lO)

J,-

EVR

[eE-(E--)V In T] (L?fO/Ge)&/c

(21)

one can easily get the thermal conductivity tensor i=(?kZ,T/3) s, s (vRjv)dS,

. .

(22)

(11)

By combining Eqs. (18), (19), (20) and (22), one can obtain the thermal conductivity tensor up to any order. 4. DISCUSSION AND CONCLUSION We now discuss the principal features of our result. (i) At high temperature, the second term inside the square bracket in Eq. (17) bccomes~negligiblc small and one recovers the Boltamann equation with external electric field on!y. From Eq. (22), one obtains ;=(7r2k,T/3e)b (23)

(13) (14)

This is just the Wiedemann-Franz law.

(5), we have (15)

(8) W.C. Chan, J. Phys. P: Metal Phys. S, X59 (1978). (9) 1l.B. Huntington and W. C. Chan, Phys. Rev. 012, 5423 (1975). (10) W.C. Ghan, Chinese J. Phy:, 16, 24 (1978).

64

ITERATION METHOD FOR THE CALCULATION

(ii) At low temperature, in order to compare the result with the variational method, we change the thermal conductivity tensor into thermal resistivity tensor Iv. Using the first order solution for the generalized relaxation time, one approximately has W,, i(3/nZkZ,T) X;; <rp)-I =(3/n2k2,T) X;; (T;)
6x;; -~~~,~,~~g,(k,k),z~(.)~(,.;:~-) = -,.&Z,T (I-$-)

+ _?_ (25) 2 (

X) ] S; -A?

(24)

where ,? is the velocity factor defined by

.Y==

I,.*,

(vvI4 d&I.%

(25)

and <rp) is the average value of the relaxation time over the Fermi surface. The double integration in Eq. (24) is very similar to the one obtained by the variational method(l) except for the factor inside the square bracket. For normal process, one approximately has (26) and v;/vp= 1 (27)

So, at low temperature, Eq. (24) gives a T2 dependence for the termal resistivity of metals. Since the ?? term has a factor of 3/2 instead of 3, one expects that its coefficient B(z WT-2-AT-3) should be roughly one half of the value BJ obtained by the variational method, in better agreement with Rlemens numerical calculation for pure metals (0.67 &). Finally, we like to point out that in obtaining Eq. (24), the expression for thermal resistivity, we have made many approximations. The most serious approximation is the assumption that the relaxation time is a slowly varying function of k on the Fermi surface. Fortunately, both electrical and thermal resistivity can be written as an average of the relaxation time over the Fermi surface. So, they are rather insensitive to the anisotropy of r)(k). Futhermore, from Eq. (20), one has approximately <rpn(k)> s 0 That is <~Jk)>~<~po(k)> So, Eq. (24) is approximately good to any order. (31) for all n>I (30)

ACKNOWLEDGEMENT
The author wishes to thank Dr. H. B. Huntington for intially suggesting this investigation. Also, this work was sponsored by the National Sciences Council of the Republic of China.