Role of filler in thermal transport
4.2 Complex phonon modes
4.2.2 Point defect scattering
Figure 4.1 Different mechanisms in reducing lattice thermal conductivity between resonant scattering and avoided crossing on the phonon spectra. a) The resonant scattering model targets phonons near π€π€0. b) BvK phonon dispersions for a stiff framework (m1, k1) and loosely bound guest atoms (m2, k2). Increased k2 stiffness results in increased coupling (extent of avoided crossing) between the framework and guest modes. c) The avoided crossing reduces ππππ2(π€π€) in the vicinity of π€π€0. d) π π π π (π€π€)for an empty BvK framework, using Umklapp and boundary scattering terms (curve A). Including resonant scattering reduces π π π π (π€π€) near π€π€0 (curve B). If instead the effect of coupling on ππππ2(π€π€) is accounted for, a similar reduction is observed (curve C). Reproduced with copyright
permission obtained Β© Royal Society of Chemistry [2011]
(http://pubs.rsc.org/en/Content/ArticleLanding/2011/JM/C1JM11754H#!divAbstract).
From Figure 4.1 we can see that both the resonant scattering effect (reduction of scattering time ππ(π€π€)) and the effect of avoided crossing on the phonon spectra (reduction of group velocity ππππ2(π€π€)) will result in the similar reduction of lattice thermal conductivity. It is thus difficult to distinguish between the two effects. Temperature dependent thermal conductivity measurements are not enough but rather frequency-dependent measurements of ππππ2(π€π€) and ππ(π€π€) are required to unravel the intertwined effects on π π πΏπΏ.
lattice thermal conductivity values (as shown below from a Klemens model): the larger the scattering parameter, the lower the lattice thermal conductivity.
According to Callaway and Klemens94-96, if we consider a combined scattering mechanism including Umklapp scattering for the pure crystal without disorder, and extra point defect scattering process for the crystal with disorder, then the ratio of π π πΏπΏ of the crystal with disorder to that without disorder, π π πΏπΏππ is:
π π πΏπΏ
π π πΏπΏππ=arctan (π’π’)
π’π’ (Eq. 4.4) π’π’2= ππβππ2πππ·π·πΊπΊ
π π 2 π π πΏπΏπππ€π€πππππππ‘π‘ (Eq. 4.5) where π’π’, β, and π€π€πππππππ‘π‘ are the disorder scaling parameter, the Planck constant, and the experimental disorder scattering parameter, respectively. Thus, π€π€πππππππ‘π‘ can be derived from π π πΏπΏ measurements using Eq. 4.4 and 4.5. It can then be compared to calculated values, as detailed in the following.
Ξ can be separated into two components:
π€π€=π€π€ππ+ π€π€ππ (Eq. 4.6) where the scattering parameters π€π€ππ and π€π€ππ represent mass and strain field fluctuations due to the introduction of point defects (e.g., R into the void), respectively.
(Eq. 4.7) where ππ is an empirical fitting parameter related to Gruneisen parameter and elastic properties. ππ is the lattice constant of pure alloy whereas Ξππ is the difference in lattice constant between the alloy with point defects and the pure alloy.
When there is more than one constituent element in the compound (as in the case of skutterudites RxCo4Sb12), there is some discrepancy regarding the definition of π€π€ππ.
According to Klemens:
(Eq. 4.8) where βMis the mass difference between two species on the same site and Mis the molar mass of the compound.
In the case of skutterudite compound, with Yb filling, we are comparing YbxCo4Sb12 to unfilled Co4Sb12, β =M MYb=173.04 and
4 12 (1 ) 4 12 173.04* 4*58.93 12*121.76
YbCo Sb Co Sb
M=xM + βx M = x+ +
So
2 2
173.04
(1 ) (1 )
173.04 * 4 *58.93 12 *121.76
M
x x M x x
M x
β ο£Ά  ο£Ά
Ξ = β ο£¬ο£ ο£·ο£Έ = β ο£¬ο£ + + ο£·ο£Έ (Eq. 4.9)
Later, Yang97 proposed a different way of calculating π€π€ππwhich weighs the influence of mass fluctuation on each site by its degeneracy. To be more specific, it takes into account not only the number of atoms on a specific site but also the mass contrast between the specific site and the rest of the sites. For example, the chemical composition of a material can be expressed as A1c1A2c2A3c3A4c4β¦Ancn, where the Ai are crystallographic sublattices in the structure and the ci are the relative degeneracies of the respective sites. In this context, the skutterudite compound YbxCo4Sb12 has n=3, A1=Yb, A2=Co, A3=Sb, and c1=1, c2=4, c3=12. In general there will be several different types of atoms that occupy each sublattice, and the kth atom of the ith sublattice has mass πππ‘π‘ππ, radius πππ‘π‘ππ, and fractional occupation πππ‘π‘ππ. The average mass and radius of atoms on the ith sublattice are:
k k
i i i
k
M =
β
f M (Eq. 4.10)k k
i i i
k
r =
β
f r (Eq. 4.11) The mass fluctuation scattering parameter is then given by:2
, 1
1 n
i
i M i
i
M n
i i
c M M
c
=
=
 
 ο£· Ξ
ο£ ο£Έ
Π=  
 
ο£ ο£Έ
β
β
(Eq. 4.12)
where the mass fluctuation scattering parameter for the ith sublattice is:
2
, 1
k
k i
M i i
k i
f M M
 
Ξ =  β ο£·
ο£ ο£Έ
β
(Eq. 4.13) and the average atomic mass of the compound is:1
1 n
i i
i n
i i
c M M
c
=
=
=  
 
ο£ ο£Έ
β
β
(Eq. 4.14)
In the case of skutterudite compound YbxCo4Sb12, there is no mass fluctuation on sites A2 (Co) and A3 (Sb).
2 ( ) 58.93
M =M Co = ,
3 ( ) 121.76 M =M Sb = .
Consequently ΞM,2 = ΞM,3=
0
For site A1 (Yb/void), M1= f M1Yb 1Yb+ f1voidM1void =173.04*x
2 2
1 1
,1 1 1
1 1
2 2
2
1 1
173.04 0
1 (1 ) 1
173.04 * 173.04 *
( 1) 1 1
(1 ) (1 )(1 )
Yb void
Yb void
M
M M
f f
M M
x x
x x
x x x
x x
x x x
   
Ξ =  β ο£· +  β ο£·
ο£ ο£Έ ο£ ο£Έ
   
= ο£¬ο£ β ο£·ο£Έ + β ο£¬ο£ β ο£·ο£Έ
β β β
= + β = β + =
1 1 2 3
1
4 12 173.04 * 4 *58.93 12 *121.76
1 4 12 17
n
i i
i n
i i
c M M M M x
M
c
=
=
+ + + +
= = =
  + +
 
ο£ ο£Έ
β
β
2 2
1
, ,1
1
1
2
2
17
173.04 *
173.04 * 4 *58.93 12 *121.76 17 1
17 173.04
17 (1 )
173.04 * 4 *58.93 12 *121.76
n i
i M i M
i
M n
i i
M M
c
M M
c
x x
x x x x x
=
=
   
Ξ Ξ
   
ο£ ο£Έ ο£ ο£Έ
Ξ = =
 
 
ο£ ο£Έ
 
 
 + + 
 
ο£ ο£Έ β
=
 
= ο£¬ο£ + + ο£·ο£Έ β
β
β
(Eq. 4.15)
Now if we compare Eq. 4.15 to Klemensβs result in Eq. 4.9, we can see that there is clearly a difference of 17 times which equals to the total number of atoms per compound formula.
If we modify Yangβ formula such that the degeneracy of each site is still taken into account but with a less complicated form, we get
π€π€ππ =π₯π₯(1β π₯π₯)ππ οΏ½π₯π₯πππποΏ½οΏ½2 (Eq. 4.16) This formula is quite similar to Klemensβs formula (Eq. 4.8) except with an extra term f and a different definition for πποΏ½, which stand for the fraction of substituted lattice sites and the average atomic mass per number of atoms in the compound (rather than the previous molar mass for the compound) respectively. ΞM still holds for the mass difference between the host and impurity atom.
In the case of skutterudite compound, β =M MYb=173.04 and
[
YbCo Sb4 12 (1 ) Co Sb4 12 173.04* 4*58.93 12*121.76 /17]
M= xM + βx M = x+ +
So
2 2
2
1 173.04
(1 ) (1 )
173.04 * 4 *58.93 12 *121.76 17
17 173.04
17 (1 )
173.04 * 4 *58.93 12 *121.76
M
x x f M x x
M x
x x
x
 
 
β ο£Ά
Ξ = β ο£¬ο£ ο£·ο£Έ = β  + + ο£·ο£·
ο£ ο£Έ
 
= β ο£¬ο£ + + ο£·ο£Έ
(Eq. 4.17)
Compare Eq. 4.9, Eq. 4.15 and Eq.17, we get
(Eq. 4.18) Let us take Cu2ZnGeSexS4-x for another example. Using Klemens model, we have:
A = Cu2ZnGeSe4
B = Cu2ZnGeS4
Cu2ZnGeSexS4-x = (x/4)*Cu2ZnGeSe4 + (1-x/4)*Cu2ZnGeS4
( )
( )
2 2
,
2
(1 ) (1 ) 4
4 4 4 4 * (4 ) * 2 *
(1 ) 1
4 4 4
Se S
M Klemens
Se S Zn Ge Cu
Se S
M M
x x M x x
M x M x M M M M
M M
x x
M
β
 
β ο£Ά
Ξ = β ο£¬ο£ ο£·ο£Έ = β ο£¬ο£ + β + + + ο£·ο£Έ
β
 
= β  ο£·
ο£ ο£Έ
(Eq. 4.19)
(Eq. 4.20)
(Eq. 4.21)
Compare Eq. 4.19, Eq. 4.20, and Eq. 4.21, and we get
(Eq. 4.22)
Thus the modified equation for calculating gives the same result as Yangβs formula, yet it is easier to implement. From now on, we will only compare Yangβs formula and Klemensβs formula.
In a more general case, for compound A1c1A2c2A3c3A4c4β¦Ancn, the relationship between the calculated scattering parameters from Klemensβs formula (Eq. 4.8) and Yangβs formula (Eq. 4.12) will be:
(Eq. 4.23)
with the ith element being substituted and Ai position having a Ci degeneracy.
The difference between Yangβs formula and Klemensβ formula is that Yangβs formula takes the degeneracy of each site as the weighing factor in calculating scattering parameter, which makes more sense because doping on the Co site (CoxNi1-xSb3) should result in a smaller scattering parameter from doping on the Sb site (Co(SbxTe1-x)3) even when the doping content x is the same.
Since is always larger than