Chapter 4 P-type PbSe with Na doping
4.3 Transport Property Modeling
77
Figure 4.5. zT as function of temperature for p-type PbSe.
Figure 4.6. a) Seebeck coefficient and b) Hall mobility as function of Hall carrier density for both p-type and n-type PbSe at 300 K. Data from literature are also included.
On the other side it is seen that the mobility of n-type and p-type samples are very different, with the p-types much lower in mobility than the n-types. The same effective masses for both, the elastic constant Cl are further found independent of the type of doping by measuring the speed of sound, which turned out to be identical (vl ≈ 3200 m/s, vt ≈ 1700 m/s). Thus the difference in mobility indicates a difference in the deformation potential coefficient Ξ, which for the L conduction band was found 25 eV, and for the L valence band it is determined to be 35 eV.
It is necessary to note that the effective mass determination from Seebeck coefficient is subject to a fair amount of uncertainty due to factors such as uncertainty in Seebeck measurement and Hall measurement, as well as the scattering mechanism(s) used. By comparing data from Caltech as well as most of published data on PbSe, as shown in Figure 4.6 a), the majority of results has suggested a same effective mass for the L conduction band and valence band. Bear in mind a possible small difference beyond the resolution of this method, the difference in mobility seen in Figure 4.6 b) could be possibly in part due to such difference as well. However, the difference seen is about a factor of 2 for the acoustic phonon scattering dominated regime, which would require a 30%
difference in effective mass to explain and would be certainly detectable in the Pisarenko relation.
Thus, it is possible that the conclusion about identical effective masses for the L conduction band and valence band is inaccurate, and the difference in Ξ is not as significant, but the general finding of Ξ being larger in the L valence band than that in the L conduction band should be solid.
So far, it is also known that the bands at L are non-parabolic Kane bands, and the effective mass for the conduction band increases with temperature with dlnm*/dlnT = 0.5, this should be the same for
p type
Caltech, Na (Se) Wang, Ag n type
Caltech, Br (Pb) Prokofeva, Cl
Androulakis, Cl Vinogradova, Na
a
p type
Caltech, Na (Se) Allgaier, Se Smirnov, Se Schliting, Se Wang, Ag n type
Caltech, Br (Pb) Chernik, Cl Prokofeva, Cl Androulakis, Cl Smirnov, Pb
b
79 the L valence band. On the temperature dependence of band gap, our most recent study142 (Figure 4.7) has provided the most accurate estimate of:
Eg(eV)=Eg,0+3!10"4T Equation 4.2
The zero temperature band gap Eg,0 is 0.17 eV taken from low temperature measurement results.
The Σ valence band is assumed to be parabolic, with effective mass independent of temperature and is isotropic for each valley.
Figure 4.7. Temperature dependent band gap of PbSe and PbS measured at Caltech and re- interpreted from literature. Image taken from Appl. Phys. Lett 103, 262109, 2013.
The expression of transport parameters in a multi-band system given by Putley143 are used here:
n= ni
i=Lc,Lv,!v
"
Equation 4.3! = !i i=Lc,Lv,!v
"
Equation 4.4RH= RH,i!i2
i=Lc,Lv,!v
"
!ii=Lc,Lv,!v
"
#
$%% &
'((= eAiniµi2
i=Lc,Lv,!v
"
niµii=Lc,Lv,!v
"
#
$%% &
'(( Equation 4.5
nH =1 /eRH Equation 4.6
µH =! /enH =!RH = e2Ainiµi 2
i=Lc,
"
Lv,!v Equation 4.7S= Si!i
i=Lc,Lv,!v
"
!ii=Lc,Lv,!v
"
Equation 4.8!"#
!$#
!%#
&$ '
( )* ( )+ ( ), ( )- ( ).
!/0#
&$ '/
12 34 " 567 / 8
&$ 9/
( * . ( , ( ( - . ( : ( ( ; . (
T<!=#
( )* . ( )+ . ( ), . ( )- . ( ). .
E
g86 3/ %> <! / 0 #
&$ '<?" 5 > / %@
&$ '<A6 $ B2 1
&$ '<C2 1 / B
&$ '/ <?" 5 > / %@
&$ '/ <A6 $ B2 1
&$ '/ <D/ E F " > G 2 E "
( * . ( , ( ( - . ( : ( ( ; . (
T<!=#
( )+ ( ( )+ . ( ), ( ( ), . ( )- (
E
g86 3/ %> <! / 0 #
?" 5 > / %@
HI 9" J $ / 3 <* K : : A6 $ B2 1 <* K . + '" " G F " 1 <* K : :
! = !L+ Li"iT
i=Lc,Lv,!v
"
+T Si2"ii=Lc,Lv,!v
"
# Si"ii=Lc,Lv,!v
"
$
%&& ' ())
2
"i i=Lc,Lv,!v
"
* + ,,
- . //
Equation 4.9
The results are shown in Figure 4.8. With the contribution from Σ valence band taken into account, the transport properties of p-type PbSe can be well modeled throughout the temperature range 300 – 800 K. The deformation potential coefficient of L valence band is set at 35 eV while to model the 800 K result it is allowed to increase by 10 %, just as for the L conduction band. Interestingly it doesn’t require additional adjustment, which one would expect if the inter-valley scattering gets intensified at higher temperatures. The parameters determined for the Σ valence band are: the effective mass for each valley mb* = mI* = 0.8 me; the deformation potential 28 eV; and its maximum is separated from that of L valence band by:
!E=0.32"2.2#10"4T Equation 4.10
Figure 4.8. a) Seebeck coefficient and b) Hall mobility as function of Hall carrier density for p- type PbSe at different temperatures. Solid curves calculated with multi-band model. Literature
data included (not marked) in 300 K mobility.
To some extent, getting information about the Σ band from modeling is not convincing because the result is subject to the formulism of the modeling and its input. The bad news is that there hasn’t been a way to directly observe the Σ band. For example for the energy gap measurement, the primary direct transitions happen at much larger probabilities since they do not require phonon participation. They would dominate the absorption spectra making the L–Σ indirect transitions hard to detect. It is even more difficult to get information about effective masses and deformation potential experimentally. To some theorists such a two-‐
band picture is even wrong because the two local maximum are actually from the same eigenstate thus should be regarded as a single band with complex energy contours144.
300 K 450 K 600 K
800 K
Vinogradova, Na:PbSe
a
300 K450 K 600 K
800 K
b
81 Nonetheless, such a simple model has been proven to explain the experiments at different temperatures very well and successfully guided experiments to improve the performance of PbSe.
The advantage of having the Σ band is directly shown in the Seebeck coefficient. Figure 4.9 compares the modeling result on Seebeck coefficient and mobility with and without contribution from Σ band. At 300 K, the influence from the Σ band is evident from modeling, but is arguably observable due to the uncertainty of Seebeck measurements. As temperature increases to 450 K, the difference becomes noticeable and the experiment results clearly suggested the importance of the Σ band. At higher temperatures the Seebeck coefficient is enhanced even at low carrier densities and at 800 K the enhancement is over 30%. On the other hand, the Hall mobility doesn’t change much by the presence of the Σ band, especially below 600 K where almost no difference is seen. The total Hall mobility is proportional to the sum of conductivity from each band weighed by mobility. Thus at lower temperatures due to large band offset σL >> σΣ, in addition µc,L >> µc,Σ because of the big difference in effective mass, as a result the overall Hall mobility is dominated by the L band.
Figure 4.9. a) Seebeck coefficient and b) Hall mobility as function of Hall carrier density modeled with and without the contribution from the Σ band.
Even though the transport properties are well explained by the current model and combination of parameters, it is also important to investigate how much the overall properties would change if the parameters for the Σ band were chosen differently. This result is shown in Figure 4.10. Error bars represent 5% uncertainty in property measurements. The result indicates: changing mb*, Ξ, ΔE0K for sigma band by ±25%, 15%, and 13% leads to no difference in the mobility modeling, whereas the difference seen in Pisarenko relations indicates that all combinations
300 K 450 K 600 K
800 K
dashed: L band only
300 K
450 K 600 K
800 K dashed: L band only
a b
of {mb*, Ξ, ΔE0K} within the range of {0.7±0.2 me, 30±4 eV, 0.34±0.02 eV} would give reasonably good fit to the experiment result. This is the suggested range for parameters of the Σ band.
Figure 4.10. Influence of different parameters for the Σ band on overall transport properties at 300 K and 600 K.
Figure 4.11. zT calculated from 3-‐band model for n-‐type and p-‐type PbSe at 850 K as function of Hall carrier density.
Figure 4.11 plots the zT of both n-type and p-type PbSe calculated at 850 K, based on the 3-band model calculation. It generally predicted the same maximum zT for both types with experimentally suggested ones, although for some unknown reason the Hall carrier densities corresponding to the maximum zT are lower than the optimum carrier density determined experimentally, for both n-type
300 K 600 K
m* = 0.6 m* = 1.0
a b
300 K
600 K
m* = 0.6 m* = 1.0
300 K
600 K
ΞΣ = 24 eV ΞΣ = 32 eV
d
ΞΣ = 24 eV ΞΣ = 32 eV 300 K
600 K
c
300 K 600 K
ΔE = 0.28 eV ΔE = 0.36 eV
e
300 K
600 K
ΔE = 0.28 eV ΔE = 0.36 eV
f
zT
nH n-type PbSe
p-type PbSe
p-type PbSe only L band
83 and p-type. It also shows the zT calculated for the L valence band only, assuming there is no Σ band. Due to its larger deformation potential coefficient, the maximum zT if there is no Σ band would be much smaller around 0.6.