Lattice thermal conductivity of nanostructured thermoelectric materials based on

PbTe
Yee Kan Koh, C. J. Vineis, S. D. Calawa, M. P. Walsh, and David G. Cahill

Citation: Applied Physics Letters 94, 153101 (2009); doi: 10.1063/1.3117228

View online: http://dx.doi.org/10.1063/1.3117228
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/94/15?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

Improved thermoelectric properties of PbTe0.5Se0.5
AIP Conf. Proc. 1447, 1027 (2012); 10.1063/1.4710356

Cross-plane thermal conductivity of a PbSnSe/PbSe superlattice material

Appl. Phys. Lett. 99, 041903 (2011); 10.1063/1.3615797

In-plane thermal and thermoelectric properties of misfit-layered [ ( PbSe ) 0.99 ] x ( WSe 2 ) x superlattice thin
films
Appl. Phys. Lett. 96, 181908 (2010); 10.1063/1.3428577

PbTe nanocomposites synthesized from PbTe nanocrystals

Appl. Phys. Lett. 90, 222112 (2007); 10.1063/1.2745218

Enhanced thermoelectric performance in PbTe-based superlattice structures from reduction of lattice thermal
conductivity
Appl. Phys. Lett. 87, 023105 (2005); 10.1063/1.1992662

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.126.153.227 On: Sun, 08 Feb 2015 20:24:03
 APPLIED PHYSICS LETTERS 94, 153101 共2009兲

Lattice thermal conductivity of nanostructured thermoelectric materials

based on PbTe
Yee Kan Koh,1,a兲 C. J. Vineis,2,3 S. D. Calawa,2 M. P. Walsh,2 and David G. Cahill1
1
Department of Materials Science and Engineering and Materials Research Laboratory, University of Illinois,
Urbana, Illinois 61801, USA
2
Lincoln Laboratory, Massachusetts Institute of Technology, Lexington, Massachusetts 02420, USA
3
Wakonda Technologies, Woburn, Massachusetts 01801, USA
共Received 23 February 2009; accepted 19 March 2009; published online 14 April 2009兲
We report the through-thickness lattice thermal conductivity ⌳l of 共PbTe兲1−x / 共PbSe兲x nanodot
superlattices 共NDSLs兲 over a wide range of periods 5 nmⱕ h ⱕ 50 nm, compositions 0.15ⱕ x
ⱕ 0.25, growth temperatures 550 K ⱕ Tg ⱕ 620 K, and growth rates 1 ␮m h−1 ⱕ R ⱕ 4 ␮m h−1. All
of our measurements approach ⌳l of bulk homogenous PbTe1−xSex alloys with the same average
composition. For 5 nmⱕ h ⱕ 50 nm, ⌳l is independent of h; a result we attribute to short mean-free
paths of phonons in PbTe and small acoustic impedance mismatch between PbTe/PbSe. We alloyed
the PbTe layers of four NDSLs with SnTe up to a mole fraction y = 18%; ⌳l is reduced by ⬍25%.
© 2009 American Institute of Physics. 关DOI: 10.1063/1.3117228兴

Direct conversion of thermal and electrical energy by efit from nanostructuring by as much as was previously
solid-state thermoelectric devices has attracted the attention thought; in fact, ⌳l of NDSLs is similar to ⌳l of homoge-
of scientists and engineers since the advent of the semicon- neous PbTe1−xSex alloys4 with the same average composi-
ductor era in the 1950s. The dimensionless figure of merit for tion. Using data for the in-plane power factor S2␴ and the
thermoelectric energy conversion1 is ZT = S2␴T / ⌳, where S through-thickness thermal conductivity ⌳, we calculate a
is the Seebeck coefficient, ␴ is the electrical conductivity, T maximum ZT of 0.6 at 300 K, only 30% larger than well-
is the absolute temperature, and ⌳ is the total thermal con- optimized n-type PbTe 共ZT ⬇ 0.45兲.
ductivity; ZT ⬇ 1 for current commercial thermoelectric de- We grew NDSLs by molecular beam epitaxy3 on single-
vices. Nanostructured semiconductors have been widely in- crystal 共111兲 BaF2 or CaF2 substrates. For CaF2 substrates, a
vestigated as a route for increasing ZT through possible 0.5 ␮m PbSe buffer layer was grown before the deposition
enhancement of the power factor S2␴ or the reduction of the of the NDSLs to improve the morphology. The total thick-
thermal conductivity of the lattice ⌳l = ⌳ − ⌳e, where ⌳e is ness of the NDSL films is 1 – 10 ␮m. The growth tempera-
the thermal conductivity of the electronic system. In most ture, measured by a pyrometer, was 548–623 K and the
cases, reductions of ⌳l by nanostructuring are thought to be growth rate was 1 – 4 ␮m h−1. The NDSLs were intention-
more significant than enhancements of S2␴. ally doped using Bi2Te3 共n-type兲 or Na2Te 共p-type兲.
Nanodot superlattices 共NDSLs兲1,2 are an important class We grew three kinds of superlattices: 共1兲 NDSLs with a
of nanostructured semiconductors. NDSLs are epitaxially constant superlattice period in the range of 5–50 nm, 共2兲
grown thin films that contain nanoscale inclusions of a sec- alternating-period NDSLs with two different periods 共6.8 and
ond phase, for example, PbSe embedded in PbTe. The nano- 9 nm, 12 and 16 nm, and 15 and 19 nm兲 alternating through
scale inclusions are produced by the Stranski–Krastinov the thickness of the film, and 共3兲 a multiple-period NDSL
growth mode of the PbSe layer. 共The term “nanodot super- with a sequence of six periods 共11.5, 15.3, 19.2, 23, 19.2, and
lattices” is sometimes used interchangeably with the term 15.3 nm兲 that repeat through the film. In four additional
“quantum dot superlattices” but we use “nanodots” to signify samples, the PbTe layers were alloyed with SnTe to a maxi-
that band offsets are probably not sufficient to create carrier mum mole fraction of 18%. To calibrate our system, we mea-
confinement near room temperature.兲 In 2002, Harman et al.1 sured the beam-equivalent-pressures of PbTe, PbSe and SnTe
reported ZT = 1.6 at T = 300 K for PbTe/PbSe NDSLs; ZT and the respective film thicknesses by cross-section Nomar-
was determined by measuring the maximum temperature ski microscopy and Fourier transform infrared spectroscopy.
change produced by a thermoelectric device where one leg of We measure the total thermal conductivity ⌳ in the
the devices was the NDSL and the other leg of the device through-thickness direction by time-domain thermoreflec-
was a gold wire. This factor of 3 enhancement in ZT com- tance.5–7 Prior to the measurements, the samples are coated
pared to the best homogeneous PbTe-based alloys was attrib- with ⬇100 nm thick Al films by magnetron sputter deposi-
uted to the effects of the embedded nanodots that increased tion. The 1 / e2 radii of the pump and probe beam at the
S2␴ and significantly decreased ⌳l. Recently however, sample surface are 15 ␮m. The total laser power incident on
studies3 of the electrical properties of a large number of the sample is ⬍12 mW, creating steady-state temperature
共PbTe兲1−x / 共PbSe兲x NDSLs have shown that the power factor rises of ⬍10 K. As carriers might be depleted near the sur-
S2␴ of PbTe/PbSe NDSLs is actually reduced by ⬇30% face of the samples, we use a modulation frequency of 1
compared to bulk PbTe. In what follows, we show that the MHz to reduce the sensitivity of our measurements to the
lattice thermal conductivity ⌳l of NDSLs also does not ben- thermal conductivity near the surface. The surface morphol-
ogy of the NDSL samples shows significant roughness; thus,
a兲
Electronic mail: [email protected]. we employ a newly developed two-tint approach8 to reject

0003-6951/2009/94共15兲/153101/3/$25.00
This article 94,is153101-1
is copyrighted as indicated in the article. Reuse of AIP content © 2009 American InstituteDownloaded
subject to the terms at: http://scitation.aip.org/termsconditions. of Physics to IP:
130.126.153.227 On: Sun, 08 Feb 2015 20:24:03
 153101-2 Koh et al. Appl. Phys. Lett. 94, 153101 共2009兲

diffusely scattered light from the pump beam. We analyze the (a) 2 (b) 2

Λl (W m K )

Λl (W m K )
-1
data following procedures described in Ref. 9, taking into

-1
-1
account changing of the radius of the pump beam at different

-1
1 1
relative delay time between the pump and probe pulses.10
We determine the concentration and in-plane mobility of
0 0
the charge carriers from Hall measurements using a van der 0.0 0.1 0.2 0.3 2 10 100
x h (nm)
Pauw geometry. The in-plane Seebeck coefficients are deter-
mined from simultaneous voltage and temperature measure- (c) 2 (d) 2

Λl (W m K )
Λl (W m K )

-1
-1
ments at two small pressed In contacts on the samples as the

-1
samples are sandwiched between a hot rod and a cold rod.3

-1
1 1
The mobility and Seebeck coefficients of the NDSLs are
similar to values reported in Ref. 3 and are not reported here. 0 0
The electronic contribution ⌳e to thermal conductivity is cal- 500 550 600 650 0 1 2 3 4 5
T (K) -1
culated from the measured electrical conductivity using the g
R (μm hr )
Wiedemann–Franz law and reduced Lorenz numbers3 for FIG. 1. Through-thickness lattice thermal conductivity ⌳l of single-period
nondegenerate electrons or holes in PbTe; ⌳l = ⌳ − ⌳e. To test 共circles兲, alternating-period 共up triangles兲, and multiple-period 共squares兲
the validity of this approach, we measured a highly n-doped 共PbTe兲1−x / 共PbSe兲x NDSLs plotted as a function of 共a兲 mole fraction x of
PbSe; 共b兲 superlattice period h; 共c兲 growth temperature Tg; and 共d兲 growth
共1.6⫻ 1019 cm−3兲 molecular beam epitaxy 共MBE兲-grown
rate R. Solid symbols are for n-doped samples and open symbols are for
PbTe epitaxial film and found ⌳ = 3.5, ⌳e = 1.1, and ⌳l p-doped samples. Data points with error bars are from this study; data points
= 2.4 W m−1 K−1; close to ⌳l = 2.0 W m−1 K−1 of bulk11 without error bars are drawn from the literature. If not otherwise specified,
PbTe and ⌳l = 2.5 W m−1 K−1 of epitaxial12 PbTe. x ⬇ 0.16, h ⬇ 15 nm, 573 K ⬍ Tg ⬍ 603 K, and R ⬇ 2 ␮m h−1. In part 共a兲,
5 nm⬍ h ⬍ 30 nm; and the in-plane ⌳l of PbTe/ PbTe0.8Se0.2 superlattices
We emphasize that our measurements of the total ther-
共diamonds兲 共Ref. 12兲 and cross-plane ⌳l of a PbTe0.7Se0.3 / PbTe0.9Se0.1 su-
mal conductivity ⌳ are in the through-thickness direction perlattice 共down triangle兲 共Ref. 13兲 are included for comparison. The dashed
while the electrical conductivity ␴ is measured in the in- and solid lines in 共a兲 are the measured 共Ref. 4兲 and calculated 关using ⌫ from
plane direction. PbTe and PbSe have cubic crystal structures; Eq. 共1兲兴 ⌳l of bulk PbTe1−xSex alloys, respectively. The dashed lines in 共b兲,
therefore, ⌳ and ␴ reduce to scalars for homogenous crys- 共c兲, and 共d兲 are ⌳l of an MBE-grown PbTe0.85Se0.15 thin film and are in-
cluded as a baseline for comparisons.
tals. For a nanodot superlattice, some degree of anisotropy of
⌳ and ␴ might be created by the layering of the nanodots
within the PbTe matrix. Transmission electron microscopy phonons in PbTe and the relatively small difference in acous-
shows that the nanodots do not, in most cases, form well- tic impedance between PbTe and PbSe; for example, for lon-
defined layers. This fact, combined with the observed lack of gitudinal acoustic modes in the 共100兲 direction, the imped-
⌳l dependence on superlattice period, leads us to the conclu- ance mismatch is only 1.1. 共Measured in the same way, the
sion that anisotropy in ⌳ and ␴ are insignificant compared to acoustic mismatch of AlAs/GaAs and Si/Ge are 1.2 and 1.3,
the uncertainties associated with calculating the Lorenz respectively.兲 Using the Debye–Callaway model described
number. below, we find that half of the heat is carried by phonons
We summarize the effects of period h, composition x, with mean-free-path ⬍19 nm in PbTe, compared to ⬍300
growth temperature Tg, and growth rate R on ⌳l of and ⬍140 nm in Si and GaAs. In other words, heat carrying
共PbTe兲1−x / 共PbSe兲x NDSLs in Fig. 1; ⌳l varies with the mole phonons in PbTe are already strongly scattered by anhar-
fraction of PbSe x but we have not been able to identify any monic phonon-phonon interactions and additional scattering
systematic variations with superlattice period or growth con- of phonons by nanodots is relatively small.
ditions. Data for PbTe-based planar superlattices12,13 are in- We attempted to decrease the lattice thermal conductiv-
cluded in Fig. 1共a兲 for comparison. To determine ⌳l of the ity further by alloying the PbTe layers of NDSLs with SnTe.
samples studied in Refs. 12 and 13, we calculated ⌳e using At a composition of 18% SnTe, the reduction in ⌳l is 25%;
the reduced Lorenz numbers of Ref. 3. 共We disregarded one a similar reduction is found in the homogenous
previous study that reported a strong dependence14 of ⌳l of Pb1−mSnmTe1−nSen alloys17,18 with similar compositions, see
PbTe/ PbTe0.75Se0.25 planar superlattices on superlattice pe- Fig. 2.
riod; we concluded that the measurements reported in Ref. To provide a baseline for comparisons and to help guide
14 were unreliable because the authors found unreasonably future work, we calculate the thermal conductivity of a ho-
high values of ⌳l for PbTe and PbTe0.85Se0.15 of 8.3 and mogeneous PbTe-based alloy using a model that includes
3.6 W m−1 K−1, respectively.兲 We find that ⌳l of both PbTe/ phonon scattering by point defects. The cross section for
PbSe NDSLs and planar superlattices are similar to ⌳l of phonon scattering by point defects is characterized by the
4
bulk homogeneous PbTe1−xSex alloys with the same average parameter ⌫ that has contributions19 from variations in the
composition. atomic masses 共⌫mass兲, and variations in the length and
We note that our ⌳l values in Fig. 1 are significantly strength of the bonds 共⌫bond兲; ⌫ = ⌫mass + ⌫bond. We follow
higher than the estimation by Harman et al.1 and a lower- Ref. 19 and derive an expression of ⌫ for an A1−mBmC1−nDn
bound measurement15 of total thermal conductivity 共⌳ quaternary alloy with m Ⰶ 1 and n Ⰶ 1,
= 0.85⫾ 0.13 W / m K at ⬃360 K兲 taken on a 95 ␮m thick
NDSL film metallized on both sides, using an apparatus
based on one-dimensional heat flow in vacuum.16 Reasons
for the discrepancy are unclear to the authors.
We attribute the lack of dependence of ⌳l on superlattice
⌫mass =
2
冉 M AB
冊
m共1 − m兲 M A − M B 2 n共1 − n兲 M C − M D 2
+
2
冉 M CD
冊 ,

period
This article to the short
is copyrighted mean-free-paths
as indicated for of
in the article. Reuse heat carrying
AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded共1a兲 to IP:
130.126.153.227 On: Sun, 08 Feb 2015 20:24:03
 153101-3 Koh et al. Appl. Phys. Lett. 94, 153101 共2009兲

FIG. 3. Compilation of the lattice thermal conductivity ⌳l of

FIG. 2. Through thickness lattice thermal conductivity ⌳l of p-doped
共PbTe兲1−x / 共PbSe兲x NDSLs 共open up triangles兲 and 共Pb1−ySnyTe兲0.85 /
共Pb1−ySnyTe兲0.85 / 共PbSe兲0.15 NDSLs 共open circles兲, compared to ⌳l of bulk
Pb1−mSnmTe⬃0.87Se⬃0.13 alloys 共m = 0.85y兲 from Ref. 17 共left triangles兲 and 共PbSe兲0.15 NDSLs 共open circles兲 measured in this work with comparison to
Ref. 18 共right triangles兲. The superlattice periods are h = 18 nm. The dashed PbTe/ PbTe0.8Se0.2 superlattices 共diamonds, Ref. 12兲 and PbTe0.7Se0.3 /
PbTe0.9Se0.1 superlattices 共down triangle, Ref. 13兲; bulk PbTe1−xSex 共solid
line is ⌳l of an MBE-grown PbTe0.85Se0.15 thin film and is included as a
baseline for comparisons. The solid line is the lattice thermal conductivity of up triangles, Ref. 4兲, Pb1−mSnmTe 共solid circles, Refs. 17 and 22兲,
the Pb1−zSnzTe0.85Se0.15 alloys calculated using the Debye-Callaway model, Pb1−mGemTe 共left triangles, Refs. 17 and 23兲 and PbTe1−nSn 共squares, Ref.
17兲 alloys. ⌫ is the impurity scattering cross section discussed in text, see
using ⌫ from Eq. 共1兲.
Eq. 共1兲. The solid line is the calculations of the Debye–Callaway model for

冉 冊
a hypothetical PbTe solid with the corresponding impurity scattering cross
␦AC − ␦BC 2 section ⌫.
⌫bond = 2␧m共1 − m兲 + 2␧n共1 − n兲
␦ABC

⫻ 冉 ␦AC − ␦AD
␦ACD
冊 2
, 共1b兲
U.S. Department of Energy under Grant No. DEFG02-
91ER45439. The MIT Lincoln Laboratory portion of this
work was sponsored by ONR under Air Force Contract No.
where M i is the mass of atom i, M AB = mM B + 共1 − m兲M A, FA8721-05-0002.
M CD = nM D + 共1 − n兲M C, ␦ij is the lattice constant of unper- 1
T. C. Harman, P. J. Taylor, M. P. Walsh, and B. E. LaForge, Science 297,
turbed lattice ij, ␦ABC = m␦BC + 共1 − m兲␦AC, ␦ACD = n␦AD + 共1 2229 共2002兲.
− n兲␦AC, and ␧ is a constant depending on the elastic proper- 2
T. C. Harman, P. J. Taylor, D. L. Spears, and M. P. Walsh, J. Electron.
ties of the matrix; ␧ = 65 for PbTe.17 We derive ⌫mass Mater. 29, L1 共2000兲.
3
= 0.010 and ⌫bond = 0.046 for PbTe0.85Se0.15; ⌫mass = 0.024 and C. J. Vineis, T. C. Harman, S. D. Calawa, M. P. Walsh, R. E. Reeder, R.
Singh, and A. Shakouri, Phys. Rev. B 77, 235202 共2008兲.
⌫bond = 0.053 for Pb0.85Sn0.15Te0.85Se0.15. This calculation 4
A. V. Ioffe and A. F. Ioffe, Izv. Akad. Nauk SSSR, Ser. Fiz. 20, 65 共1956兲.
suggests that phonon scattering by point defects in most 5
C. A. Paddock and G. L. Eesley, J. Appl. Phys. 60, 285 共1986兲.
PbTe alloys is controlled by variations in bond strengths and 6
D. A. Young, C. Thomsen, H. T. Grahn, H. J. Maris, and J. Tauc, in
lengths rather than variations in the atomic masses. We con- Phonon Scattering in Condensed Matter, edited by A. C. Anderson and J.
struct a Debye–Callaway model similar to what we devel- P. Wolfe 共Springer, Berlin, 1986兲, p. 49.
7
D. G. Cahill, W. K. Ford, K. E. Goodson, G. D. Mahan, A. Majumdar, H.
oped previously for III-V materials.10,20 We fix the relative J. Maris, R. Merlin, and S. R. Phillpot, J. Appl. Phys. 93, 793 共2003兲.
anharmonic scattering strengths of umklapp and normal pro- 8
K. Kang, Y. K. Koh, C. Chiritescu, X. Zheng, and D. G. Cahill, Rev. Sci.
cesses, and obtain the absolute strengths from a fit to ⌳l of Instrum. 79, 114901 共2008兲.
bulk PbTe.11 We use Eq. 共1兲 to calculate ⌫. We include only
9
D. G. Cahill, Rev. Sci. Instrum. 75, 5119 共2004兲.
10
Debye phonons with frequencies below the frequency of Y. K. Koh and D. G. Cahill, Phys. Rev. B 76, 075207 共2007兲.
11
E. D. Devyatkova and I. A. Smirnov, Sov. Phys. Solid State 3, 1666
phonons at Brillouin zone boundaries; ⬃10% of the avail- 共1962兲.
able phonon modes are accounted for by this approach. We 12
H. Beyer, J. Nurnus, H. Böttner, A. Lambrecht, T. Roch, and G. Bauer,
estimate the thermal conductivity of the unaccounted Appl. Phys. Lett. 80, 1216 共2002兲.
phonons from the minimum thermal conductivity,21 ⌳min
13
D. G. Cahill, A. Bullen, and S.-M. Lee, High Temp. - High Press. 32, 135
共2000兲.
⬇ 0.36 W m−1 K−1. We add ⌳min to the calculations of the 14
J. C. Caylor, K. Coonley, J. Stuart, T. Colpitts, and R. Venkatasubrama-
Debye–Callaway model. The results are plotted as solid lines nian, Appl. Phys. Lett. 87, 023105 共2005兲.
in Figs. 1共a兲, 2, and 3. The agreement between the calcula- 15
P. Mayer, “High-density thermoelectric power generation and nanoscale
tions and the measurements are satisfactory for a broad range thermal metrology,” Ph.D. thesis, Massachusetts Institute of Technology,
of ⌫, see Fig. 3. 16
2007.
P. Mayer and R. J. Ram, 24th International Conference on Thermoelec-
In conclusion, PbSe nanodots do not reduce ⌳l of PbTe- trics, 19–23 June 2005 共unpublished兲, pp. 280–283.
based NDSLs below the alloy limit. The ⌳l of PbTe-based 17
G. T. Alekseeva, B. A. Efimova, L. M. Ostrovskaya, O. S. Serebryanni-
NDSLs is similar to ⌳l of homogenous alloys with the same kova, and M. I. Tsypin, Sov. Phys. Semicond. 4, 1122 共1971兲.
18
average composition and can be readily estimated using a E. A. Gurieva, P. P. Konstantinov, L. V. Prokof’eva, D. A. PshenaŽ-
simple model taking into account scattering of phonons by Severin, M. I. Fedorov, and Yu. I. Ravich, Semiconductors 40, 763
共2006兲.
point defects. Our work provides guidelines for future work 19
B. Abeles, Phys. Rev. 131, 1906 共1963兲.
on nanostructured thermoelectrics based on PbTe. 20
Y. K. Koh, Y. Cao, D. G. Cahill, and D. Jena, Adv. Funct. Mater. 19, 610
The work at the University of Illinois 共UIUC兲 was sup- 共2009兲.
21
D. G. Cahill and R. O. Pohl, Annu. Rev. Phys. Chem. 39, 93 共1988兲.
ported by ONR 共Grant No. N00014-07-1-0190兲. Sample 22
M. Orihashi, Y. Noda, L. Chen, and T. Hirai, Mater. Trans. JIM 41, 1196
characterization used the facilities of the Center of Mi- 共2000兲.
23
croanalysis of Materials which is partially supported by the I. Kudman, Metall. Trans. 2, 163 共1971兲.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.126.153.227 On: Sun, 08 Feb 2015 20:24:03