Effects of doping on transport properties in Cu-Bi-Se-based thermoelectric materials

Article abstract of DOI:10.1021/ic5014945

The thermoelectric properties of Zn-, In-, and I-doped Cu_1.7Bi_4.7Se₈ pavonite homologues were investigated in the temperature range from 300 to 560 K. On the basis of the comprehensive structural analysis using Rietveld refinement of synchrotron radiation diffraction for Cu_x+yBi_5-ySe₈ compounds with the inherently disordered crystallographic sites, we demonstrate a doping strategy that provides a simultaneous control for enhanced electronic transport properties by the optimization of carrier concentration and exceptionally low lattice thermal conductivity by the formation of point defects. Substituted Zn or In ions on Cu site was found to be an effective phonon scattering center as well as an electron donor, while doping on Bi site showed a moderate effect for phonon scattering. In addition, we achieved largely enhanced power factor in small amount of In doping on Cu site by increased electrical conductivity and moderately decreased Seebeck coefficient. Coupled with a low lattice thermal conductivity originated from intensified point defect phonon scattering by substituted In ions with host Cu ions, a thermoelectric figure of merit ZT of 0.24 at 560 K for Cu_1.6915In_0.0085Bi_4.7Se₈ was achieved, yielding 30% enhancement compared with that of a pristine Cu_1.7Bi_4.7Se₈ at the same temperature.

Full text of DOI:10.1021/ic5014945

Article
pubs.acs.org/IC
Effects of Doping on Transport Properties in Cu−Bi−Se-Based
Thermoelectric Materials
Jae-Yeol Hwang,†,◊ Hyeon A. Mun,‡,◊ Sang Il Kim,§ Ki Moon Lee,§ Jungeun Kim,∥,⊥ Kyu Hyoung Lee,*,#
and Sung Wng Kim*,†,‡
†Center for Integrated Nanostructure Physics, Institute for Basic Science (IBS), Sungkyunkwan University, Suwon 440-746, Korea
‡Department of Energy Science, Sungkyunkwan University, Suwon 440-746, Korea
§Materials R&D Center, Samsung Advanced Institute of Technology, Samsung Electronics, Suwon 443-370, Korea
∥Japan Synchrotron Radiation Research Institute, Hyogo 679-5198, Japan
⊥RIKEN SPring-8 Center, Kouto, Sayo-cho, Hyogo 679-5148, Japan
#Department of Nano Applied Engineering, Kangwon National University, Chuncheon 200-701, Korea
S
* Supporting Information
ABSTRACT: The thermoelectric properties of Zn-, In-, and I-
doped Cu1.7Bi4.7Se8 pavonite homologues were investigated in the
temperature range from 300 to 560 K. On the basis of the
comprehensive structural analysis using Rietveld refinement of
synchrotron radiation diffraction for Cux+yBi5−ySe8 compounds
with the inherently disordered crystallographic sites, we
demonstrate a doping strategy that provides a simultaneous
control for enhanced electronic transport properties by the
optimization of carrier concentration and exceptionally low lattice thermal conductivity by the formation of point defects.
Substituted Zn or In ions on Cu site was found to be an effective phonon scattering center as well as an electron donor, while
doping on Bi site showed a moderate effect for phonon scattering. In addition, we achieved largely enhanced power factor in
small amount of In doping on Cu site by increased electrical conductivity and moderately decreased Seebeck coefficient. Coupled
with a low lattice thermal conductivity originated from intensified point defect phonon scattering by substituted In ions with host
Cu ions, a thermoelectric figure of merit ZT of 0.24 at 560 K for Cu1.6915In0.0085Bi4.7Se8 was achieved, yielding 30% enhancement
compared with that of a pristine Cu1.7Bi4.7Se8 at the same temperature.
carrier scattering mechanisms, the optimization of nc by a precise
control is the prerequisite requirement for maximizing PF.
Further, the substitutional doping reduces the κlat due to the
point defect phonon scattering originated from the mass
difference between host atoms and dopants.4,5
However, the accurate prediction of the changes of electronic
and thermal transport properties in complexly structured
materials associated with the substitutional doping are
complicated because there are difficulties in clarifying the role
of structural factors (such as atomic and ionic disorders,
structural distortion, and/or various crystallographic sites of
constituent atoms) on the TE properties. Thus, the establish-
ment of doping strategies is a vital issue in complexly structured
TE materials for enhancing the ZT value, and the understanding
of electronic structures as well as lattice dynamics based on the
structural analysis is the key to conceive the strategies.
INTRODUCTION
■
Exploring semiconducting materials with complex structure has
been widely demonstrated as a promising route to develop new
materials with a high thermoelectric (TE) figure of merit, ZT =
S2σT/κ (where S, σ, κ, and T denote the Seebeck coefficient,
electrical conductivity, thermal conductivity, and absolute
temperature), since a low lattice thermal conductivity (κlat)
originates in the structural complexity.1,2 On the basis of this
strategy, we found a new n-type pavonite homologue
Cux+yBi5−ySe8 (1.2 ≤ x ≤ 1.5, 0.1 ≤ y ≤ 0.4) TE compound
with an intrinsically low κlat (0.41−0.55 W m−1 K−1 at 300 K)
derived from an inherently disordered structure, which is
highlighted with the interstitial Cu and substitutional Cu on Bi
sites followed by the displacement of Se sites.3 Maximum ZT
value of 0.27 at 560 K was obtained through compositional
tuning for adjusting the site occupancy and carrier concentration
(nc); however, the moderate power factor (PF, S2σ) prevents the
achievement of a higher TE performance.3 A substitutional
doping approach has been the most general way to improve the
PF by optimization of nc. Because the carrier transport properties
such as Hall mobility (μH), carrier relaxation time (τ), and
effective mass (m*) are largely affected by the change of nc
through substitutional doping associated with the change of
Here we report the doping effects on the electronic and
thermal transport properties of pavonite homologue
Cux+yBi5−ySe8 compounds. A systematic doping strategy by the
comprehensive structural analysis was provided to clarify the
Received: June 23, 2014
© XXXX American Chemical Society
A
dx.doi.org/10.1021/ic5014945 | Inorg. Chem. XXXX, XXX, XXX−XXX

Inorganic Chemistry
Article
Figure 1. (a) Crystal structure and (b) the (0 0 1) view of Cux+yBi5−ySe8. All distinguishable atomic sites for Cu, Bi, and Se in the unit cell were visualized
in (c), (d), and (e), respectively.
calculated from measurements taken separately: sample density (ρs),
heat capacity (Cp), and thermal diffusivity (λ) measured under vacuum
by laser-flash method (TC-9000, ULVAC, Japan), in which Cp was used
as the constant value of 0.225 J g−1 K−1 estimated from the Dulong−
Petit fitting using low-temperature Cp data (see calculation in
Supporting Information, S1). All measured data, which were acquired
at the same dimension and configuration, are obtained within the
experimental error of ∼5%.
Synchrotron X-ray Diffractions and Rietveld Refinements.
Room-temperature synchrotron powder diffraction measurements were
carried out at the beamline BL02B2 at SPring-8 (Japan) using a large
Debye−Sherrer camera with an image plate detector. The incident X-ray
wavelength (λ = 0.354 97 Å) with horizontal polarization was
determined by calibration on a standard CeO2 sample (a = 5.411 102
Å). The data were acquired from 2θ = 1.5° to 2θ = 50° with a step size of
0.01°. Owing to the small diameter of the capillary (0.1 mm) and the
high energy of the incoming X-rays, absorption and extinction are
negligible. For the Rietveld refinement, FULLPROF suite was used, and
structure factors were extracted after successive refinements with
different structural models at the condition of convergence with the best
pattern match.6
uncertainties relating to their transport properties. First, the
structural characteristics of each atomic site for Cu1.3Bi4.9Se8,
Cu1.7Bi4.7Se8, and Cu1.9Bi4.6Se8 were analyzed using Rietveld
refinement of synchrotron radiation diffractions. Then, we
selected Cu1.7Bi4.7Se8 as a material for demonstration of doping
effects and designed the compositions with doping elements for
both substitutional and interstitial sites by considering the results
of structural analysis including valence state, site occupancy,
charge distribution, and bonding characteristics. It was
demonstrated that an extremely low κlat value of 0.32 W m−1
K−1 was achieved by Zn doping on Cu site, and the largest ZT of
0.24 at 560 K was obtained in Cu1.6915In0.0085Bi4.7Se8 due to the
synergetic effects of increase in PF by tuning nc and decrease in
κlat by intensified point defect phonon scattering on interstitial
sites.
EXPERIMENTAL SECTION
■
Sample Preparation. Crystal ingots of Cux+yBi5−ySe8 ((x, y) = (1.2,
0.1), (1.4, 0.3), and (1.5, 0.4)), Zn- and In-doped Cu1.7Bi4.7Se8
(Cu1.7−uMuBi4.7Se8 and Cu1.7Bi4.7−vMvSe8, M = Zn and In, u and v =
0.0085 and 0.025, respectively), and I-doped Cu1.7Bi4.7Se8 compounds
(Cu1.7Bi4.7Se8−wIw, w = 0.08 and 0.16) were fabricated through the
conventional melting technique by use of high-purity elemental Cu
(99.999%, CERAC), Bi (99.999%, 5N Plus), Se (99.999%, 5N Plus), Zn
(99.999%), In (99.999%), and I (99.999%) as starting materials. The
stoichiometric mixtures of the elements were loaded into a quartz tube
of 14 mm in diameter. The tube was vacuum-sealed, and the mixed
contents were melted in a furnace for 10 h at 1273 K; then they were
water-quenched. The ingots were ground using ball mill, and compacted
bulk samples of 10 mm in diameter and 13 mm in thickness were
prepared using spark plasma sintering (SPS) technique under dynamic
vacuum and with the application of 50 MPa of uniaxial pressure at 663 K.
The relative densities of the resulting consolidated samples were found
to range from 7.22 to 7.48 g cm−3, which are more than 95% of the
theoretical value.
RESULTS AND DISCUSSION
■
The Structural Characteristics of Atomic Sites in
Cux+yBi5−ySe8. Figure 1a,b shows the crystal structure of
monoclinic Cux+yBi5−ySe8 with the space group C2/m, which
can be described as a sandwich structure of “accreting” and “non-
accreting” slabs (namely, a nonaccreting slab is sandwiched
between two accreting slabs along the c-axis). One accreting slab
centered at c = 0 has two trivalent octahedral sites of a central Bi1
and a marginal Bi2, and Bi2 is connected with the other adjacent
slab. It is noteworthy that the nonstoichiometric structural
formula of Cux+yBi5−ySe8 compounds is caused by partial
substitution of Bi sites by Cu atoms.3 Two trivalent Bi sites are
partially substituted by divalent Cu (Cu1 for Bi1 and Cu2 for Bi2
sites), and the neighboring Se atom position is split into two
positions of Se3 and Se4. Minor Cu6 is also positioned in this
slab. The non-accreting slab centered at c = 1/2 is composed of
Electronic and Thermal Transport Properties. The electronic
transport properties including σ and S were measured from 300 to 560 K
using an ULVAC ZEM-3 system. The κ values (κ = ρsCpλ) were
B
dx.doi.org/10.1021/ic5014945 | Inorg. Chem. XXXX, XXX, XXX−XXX

Inorganic Chemistry
Article
fully occupied octahedral divalent Bi3 and complex occupation
by interstitial Cu atoms (Cu3, Cu4, and Cu5) surrounding the
random distribution of Se5 atoms. The 14 atomic sites of Cu, Bi,
and Se in a unit cell are summarized in Supporting Information
(Table S1). For the engineering of electronic and thermal
transport properties in this complex structure by doping
approach, the fundamentals of the structural characteristics for
each atomic site including atomic position, bonding, and charge
valence should be clarified.
To address this issue, the Rietveld refinements of high-
resolution synchrotron powder diffractions were performed for
three compositions of Cu1.3Bi4.9Se8, Cu1.7Bi4.7Se8, and
Cu1.9Bi4.6Se8 as shown in Figure 2. For a high accuracy of
less than unity. Refinement results are summarized in Supporting
Information (Tables S1 and S2). Note that the occupancies of
the interstitial Cu sites (Cu3−Cu6) considerably increased with
increasing Cu content in Cux+yBi5−ySe8, whereas the occupancies
of substitutional Cu sites were almost same with altering Cu/Bi
ratio. This indicates that Cu ions are favorable to occupy the
interstitial atomic sites rather than the substitutional ones in this
structure. Total occupancy ratio (Occint/Occsub) between
interstitial (Occint) and substitutional (Occsub) Cu sites were
11.26, 15.04, and 16.92 for Cu1.3Bi4.9Se8, Cu1.7Bi4.7Se8, and
Cu1.9Bi4.6Se8, respectively. Both lattice constants and the value of
β angle were slightly changed maintaining the similar volume
with varying Cu/Bi ratio. Interestingly, the lattice parameter of b
decreased with increasing Cu content, while the value of β angle
increased. This implies that the lattice distortion and atomic
disordering are relevant to the configuration of Cu ions (that is,
the occupancy of interstitial Cu sites). Thus, a large structural
change can be induced by the impurity doping on the interstitial
Cu sites together, leading to the expectation of a dramatic change
in TE transport properties. The simulated charge density
distributions of the (2 0 0) planes based on the structure
refinements are represented in the insets of Figure 2. The charge
densities of four interstitial Cu ions at c = 1/2 are overlapping
each other along the b-axis, and the degree of charge density
overlapping between the interstitial Cu ions depends on the
occupancy of interstitial Cu, resulting in the change in the lattice
parameter of b. This is correlated with the change in electronic
transport properties, since the interstitial Cu sites along b-axis
become a conduction path, leading to an improvement of
electron transfer across the basal plane.3 Additionally, interstitial
Cu sites distributed in a zigzag pattern along b-axis can effectively
scatter (or block) the heat-carrying phonon between alternating
slabs. Indeed, the highest σ and the lowest κ values were observed
in Cu1.9Bi4.6Se8 with highest Cu content.3 This strongly suggests
that the interpretation of structural characteristics can predict the
changes of electronic and thermal transport properties by doping
at interstitial sites.
(a). Divalent Substitutional Cu (Cu1 and Cu2) and
Interstitial Cu Site (Cu3, Cu4, Cu5, and Cu6). Cu sites in
Cux+yBi5−ySe8 can be categorized into two types in terms of
coordination number and site occupancy shared with adjacent Bi
sites (Figure 1c). Substitutional Cu sites (Cu1 and Cu2, azure)
share the occupancy with the nearest neighbor Bi1 or Bi2 sites.
Otherwise, the others are interstitial Cu sites (Cu3−Cu6, blue).
Divalent Cu ions are distributed at two substitutional sites of
trivalent Bi sites (Bi1(2a) and Bi2(4i)) being substituted in the
accreting slab and at four interstitial sites (4i) in nonaccreting
slab. Substitutional Cu sites are very close to the neighboring Bi
sites (Cu1−Bi1 (0.54 Å), Cu2−Bi2 (1.10 Å)), while the
interatomic distances for Cu1−Cu6 (4.0 Å), Cu2−Cu6 (2.98
Å), and Cu1−Cu2 (3.77 Å) are too long. Interstitial Cu sites are
statistically distributed in a zigzag pattern around c = 1/2 along b-
axis. In nonaccreting slab, Cu−Cu bonding along b-axis
consisting of Cu3−Cu5 (1.77 Å), Cu5−Cu3 (1.77 Å), and
Cu3−Cu3 (1.25 Å) connections in a diamond shape can form an
electrical conduction path. In accreting slab, zigzag-shaped
bonding between Cu6−Cu6 (2.32 Å) chains can also make
another electrical conduction path along b-axis. Thus, the
electron conduction might be more favorable in a path parallel to
b-axis, suggesting the anisotropy in electrical and electronic
thermal conductivities. These directional atomic distributions
and bondings between interstitial Cu ions can also induce the
Figure 2. Refinement results of Rietveld analyses for (a) Cu1.3Bi4.9Se8,
(b) Cu1.7Bi4.7Se8, and (c) Cu1.9Bi4.6Se8 pavonite homologues.
Synchrotron XRD patterns (black line), calculated patterns (red circle),
calculated Bragg peak positions (green) for primary (up) and secondary
(down: Bi2Se3) phases, and residual, which is the difference between
measured and calculated patterns (blue). Measured range is from 1.5° to
50° in 2θ. (inset) The charge densities on the (2 0 0) planes for each
compound.
refinements between experimental and calculated diffraction
patterns, thermal displacement parameters were refined first
under the fixed occupancies followed by the refinement of
occupancy for each atom considering three structural con-
straints: (i) the divalent Bi site (Bi3) is fully occupied, (ii) the
total occupancies of Se is 4, and (iii) the sum of occupancy for
substitutional Bi (Bi1 and Bi2) and Cu (Cu1 and Cu2) sites is
C
dx.doi.org/10.1021/ic5014945 | Inorg. Chem. XXXX, XXX, XXX−XXX

Inorganic Chemistry
Article
effective changes in electron conduction and phonon scattering
by doping.
(b). Trivalent Substitutional (Bi1 and Bi2) and Divalent Bi
Sites (Bi3). Trivalent and divalent Bi sites coexist in the
Cux+yBi5−ySe8 structure, as shown in Figure 1d. In accreting
slab, Bi1 has a nearly regular octahedral coordination, and the
marginal Bi2 site is asymmetric. Both trivalent Bi sites are
noncoincident with the divalent substitutional Cu ions but share
the site occupancy, suggesting the possibility of intensified
phonon scattering by increase in effective ionic radius. On the
other hand, half of the octahedral divalent Bi3 site in
nonaccreting slab possesses the most asymmetric coordination
with Bi3−Se5 (2.71 or 3.02 Å) and Bi3−Se2 (2.93 Å). The
occupancy of Bi3 site is not changed since this site is not
substituted by Cu. There is no Bi−Bi bonding (Bi dimer)
observed in this structure.
(c). Divalent Se Site. As shown in Figure 1e, Se atoms are
entirely distributed in the unit cell of Cux+yBi5−ySe8 by taking the
space in the vicinity between Cu and Bi atoms. The only
exception is the Se3−Se4 pair with the distance of 0.555 Å that
has a bonding to the central Bi1 octahedron. Because of these
unique atomic configurations, Se sites can provide effective
phonon scattering centers. Structure refinement revealed that the
coordination of Se ions could be altered by the change in the
configuration of interstitial Cu ions. In particular, the atomic
arrangements on each lattice plane along each axis (a-, b-, and c-
axis) are different, resulting in highly asymmetric complex
disorders in the unit cell.
Characteristics of Elemental Doping for Each Atomic
Site and Electronic Transport Properties. (a). Cu Site
Doping. In the previous report,3 it was assumed that interstitial
Cu ions acts as a single donor and that nc could be controlled by
variation of Cu/Bi ratio; however, nc values were varied within
the narrow range from 4.56 × 1019 cm−3 (Cu1.6Bi4.75Se8) to 7.86
× 1019 cm−3 (Cu1.5Bi4.8Se8) for all allowed Cu/Bi ratios. Because
S and σ values are determined by the following correlations:
Figure 3. Powder X-ray diffraction patterns of (a) Cu site, (b) Bi site,
and (c) Se site modified Cu−Bi−Se structures. Asterisk (*) and circle
●
( ) represent Bi2Se3 and CuI secondary phases, respectively.
2/3
2
⎛
⎜
⎝
⎞
⎟
⎠
8πkB
3eh2
η
S =
m*T
3n
(1)
c
where kB is the Boltzmann constant, e is the electron charge, h is
the Planck constant, η is the reduced Fermi energy, and σ =
nceμH, the control of nc is the most important and straightforward
way to manipulate electronic transport. From the result of
Rietveld refinement, in case of impurity doping to Cu sites, it is
expected that the dopants are preferentially substituted into the
interstitial Cu sites, not into the substitutional ones. To examine
the effect of dopants on the nc, we selected Zn (2+) and In (3+)
as doping elements with different charge valences. Figure 3
shows the powder X-ray diffraction (XRD) patterns for pristine
and doped Cu1.7Bi4.7Se8 compounds. All patterns are indexed
with monoclinic C2/m space group as a major phase, and Bi2Se3
as a minor phase was also observed. Additional CuI secondary
Figure 4. Magnified XRD patterns of the (1 1 4) reflections for (a)
Cu1.7−uMuBi4.7Se8, (b) Cu1.7Bi4.7−vMvSe8, and (c) Cu1.7Bi4.7Se8−wIw.
temperature dependences of σ and S for pristine and Cu site
doped Cu1.7−uMuBi4.7Se8 (M = Zn and In, u = 0.0085 and 0.0255,
respectively) compounds. Cu site doped Cu1.7−uMuBi4.7Se8
compounds showed the dramatic changes in both σ and S
values, indicating the increase in nc. In spite of the valence
difference between Zn2+ and In3+ dopants, the induced changes
in σ and S were similar (Figure 5a,d), which implies that the
majority of dopants are at interstitial Cu sites rather than
substitutional ones and the electronegativity mismatch provokes
the change of electronic transport properties due to the reduced
electron negativity (1.65 for Zn and 1.78 for In) compared to Cu
phase was only observed in Cu1.7Bi4.7Se7.84 0.16. As shown in
I
Figure 4, the peak shifts of (1 1 4) reflections to the lower angle
are clearly observed in the XRD patterns. The uniformly
distributed chemical compositions are also confirmed by energy-
dispersive X-ray spectroscopy (EDS) analysis for all doped
samples (see Supporting Information, Figure S1). These verify
that the dopant ions are well-substituted into the host matrix of
Cu−Bi−Se pavonite homologue resulting in the changes of
structural parameters such as interplanar distance, interplanar
angle, atomic coordinate, and so on. Figure 5a shows the
D
dx.doi.org/10.1021/ic5014945 | Inorg. Chem. XXXX, XXX, XXX−XXX

Inorganic Chemistry
Article
locate nearby interstitial Cu ions and the atomic distribution of
Cu induces the rearrangement of Se, the nc increases by I doping,
leading to the increase in σ and the decrease in S values. However,
σ for 2 atom % I-doped compound (Cu1.7Bi4.7Se7.84 0.16)
I
decreased. This is related to the formation of secondary phase
CuI as shown in Figure 3c, which can reduce μH and σ by
impurity scattering in spite of the increase of nc.
In summary, we found that σ was drastically changed by the
modification of Cu or Se sites as shown in Figure 5a−c. On the
other hand, the region of intrinsic conduction also shifts to higher
temperatures with the increase in doping level. This is a typical
behavior of degenerated semiconductors induced by the increase
in nc since the increased extrinsic carriers suppress the onset
temperature of intrinsic conduction.
Thermal Transport Properties and Thermoelectric
Performance. Figure 6 shows the temperature dependences
Figure 5. (a−c) Electrical conductivities (σ), (d−f) Seebeck coefficients
(S), (g−i) power factors (PF) as a function of temperature for
Cu1.7−uMuBi4.7Se8, Cu1.7Bi4.7−vMvSe8, Cu1.7Bi4.7Se8−wIw, respectively.
(1.90), resulting in the delocalization of electron density around
interstitial Cu site. Note that the PF values of
Cu1.6915In0.0085Bi4.7Se8 were higher than those of other
compositions over the whole measured temperature range
(Figure 5g). This is originated from relatively moderate decrease
in S (from −144 to −85 μV K−1 at room temperature).
Figure 6. Total thermal conductivity (κtot) and lattice thermal
conductivity (κlat) as a function of temperature for (a) Cu1.7−uMuBi4.7Se8,
(b) Cu1.7Bi4.7−vMvSe8, (c) Cu1.7Bi4.7Se8−wIw, respectively.
(b). Bi Site Doping. On the basis of the results of Rietveld
refinements and the above experiments, we found that nc could
be significantly changed by doping into Cu sites. However, it is
conceived that the modification of Bi sites has a moderate effect
on electronic transport parameters compared to that of Cu sites.
Indeed, the overall changes in σ, S, and PF of Bi site doped
Cu1.7Bi4.7−vMvSe8 compounds are much smaller than those of Cu
site doped Cu1.7−uMuBi4.7Se8. Figure 5b,e shows the temperature
dependence of σ and S for pristine and Bi site doped
Cu1.7Bi4.7−vMvSe8 (M = Zn and In, v = 0.0085 and 0.0255,
respectively) compounds, exhibiting smaller changes than those
of Cu site doped compounds. Further, the shift of onset
temperatures for bipolar conduction was relatively lower than
those of Cu site doped compounds. This result suggests that the
only small increase in nc was achieved by Zn or In doping on Bi
sites. Additionally, the difference in electronegativity between
dopants and Bi (electronegativity value of Bi is 2.02) hardly
affects the electronic transport due to the delocalized electron
density around Bi sites. PF values of Zn- and In-doped
compounds were slightly lower than that of Cu1.7Bi4.7Se8, despite
increased σ as shown in Figure 5h. We believe that the
electronegativity mismatched doping is a more effective way to
control both electronic and thermal transport than conventional
doping scheme based on the charge imbalance in the complex
structures.
of total (κtot), lattice (κlat), and electronic thermal conductivity
(κele). We consider that the extremely low thermal conductivities
in our samples are an intrinsic property originating from their
structural complexities and that the TE contributions by the
minor secondary phases (such as, Bi2Se3 and CuI) are
insignificant, since the values of room temperature thermal
conductivities for Bi2Se3 and CuI are 3.07 and 1.68 W m−1 K−1,
respectively,7 which are much higher values than that of the
matrix. We calculated κlat using the following equation: κtot = κlat
+ κele, where the κele is estimated from the Wiedemann−Franz
law, κele = L0Tσ. The Lorenz number, L0, was obtained using
following equation:
2
⎛
⎞
⎟
⎟
⎠
7
2
3
2
5
2
3
2
⎡
⎢
⎢
⎣
⎤
⎥
⎥
⎦
2
r +
(
r +
(
F
r+5/2(η)
r +
(
r +
(
Fr+3/2(η)
⎛
⎜
⎝
⎞ ⎜
)
)
)
)
⎟
kB
e
⎟
⎜
L0 =
−
⎠
⎜
Fr+1/2(η)
Fr+1/2(η)
⎝
(2)
where r is the scattering parameter, Fn(η) is the Fermi integral of
order n, Fn(η) = ∫ 0∞(xn/(1 + e(x−η)))dx, and η is calculated from
the value of S, respectively. The value of r, which was derived
from the temperature dependence of μH, was ∼1.5 in the Cu−
Bi−Se-based materials, suggesting that the main scattering
mechanism was mixed by the interaction between acoustic and
optical phonon scatterings of the lattice atoms and/or structural
(c). Se Site Doping. Figure 5c,f shows the temperature
dependence of σ and S for Cu1.7Bi4.7Se8−wIw (w = 0.08 and 0.16).
In case of 1 atom % I-doped composition (Cu1.7Bi4.7Se7.92 0.08
I
), σ
largely increases, especially at low temperatures. Since the Se ions
E
dx.doi.org/10.1021/ic5014945 | Inorg. Chem. XXXX, XXX, XXX−XXX

Inorganic Chemistry
Article
defects (point defects).8,9 Calculated L0 was nearly constant in all
compounds and was in the range of (2.48−2.58) × 10−8 V2 K−2.
The κtot values increased by doping at Cu and Bi sites due to the
increase of the κele value. However, the addition of a small
amount of iodine decreases κtot due to the suppression of κlat by
impurity scattering. Interestingly, all κlat values of Cu site doped
compounds (Cu1.7−uMuBi4.7Se8) were significantly suppressed
compared to those of Bi site doped Cu1.7Bi4.7−vMvSe8. This
indicates that the atomic arrangements at interstitial Cu sites are
highly related to thermal transport properties and that the doping
at interstitial Cu sites is effective to reduce phonon propagation,
resulting in the decrease of κlat. As shown in Figure 6, κlat was
decreased by the modification of each atomic site except for the
compositions of Cu1.7Bi4.7−vZnvSe8 (v = 0.0085 and 0.0255),
mainly due to the mass difference between host ions (Cu, Bi, and
Se) and dopants (Zn, In, and I). Room-temperature values of κtot,
ranging from 0.52 to 0.69 W m−1 K−1, were found for all
compositions. These κtot values are much lower than those of
state-of-the-art low κlat systems such as Zn4Sb3 (κtot ≈ 1.05 W
m−1 K−1 at 300 K)10 and Cu2−xSe (κtot ≈ 1 W m−1 K−1 at 400
K).11
structural and TE characteristics were observed in complex
structured β-Zn4Sb3 and Cu2−xSe. Cargnoni et al. demonstrated
that Zn interstitial atoms played a fundamental role as electron
suppliers to enhance the value of S, as well as to suppress the κlat
value of β-Zn4Sb3.22 Liu et al. showed with first-principle
calculations that dopants occupied preferentially the Zn
interstitial (or vacancy) sites rather than substitutional ones in
β-Zn4Sb3 due to the relatively low formation energy.23 It was also
reported that Cu2−xSe showed p-type conduction with a high S
and low κlat due to the ionic and liquid-like behaviors of randomly
distributed interstitial Cu ions, respectively.11 Therefore,
considering such correlations between the common structural
features and TE properties in three different TE materials with
low κlat, we surmise that the interstitial site ions play a major role
to affect TE properties including both thermal and electronic
transport characteristics. From σ, S, and κ values, we calculated
the ZT values.
Figure 8 shows the temperature dependence of the ZT for
pristine and Zn-, and In-doped Cu1.7Bi4.7Se8 (Cu1.7−uMuBi4.7Se8
To clarify the role of doping on different atomic sites affecting
the thermal transport, we modeled the temperature dependence
of the κlat for pristine, interstitial site-, and substitutional site-
doped samples using the Debye−Callaway model12−21 (see
Supporting Information, S4).
As shown in Figure 7, the calculated κlat is well-matched with
experimental data, and impurity scattering prefactor (I) has a
Figure 8. Dimensionless thermoelectric figure of merit (ZT) as a
function of temperature for (a) Cu1.7−uMuBi4.7Se8, (b) Cu1.7Bi4.7−vMvSe8,
(c) Cu1.7Bi4.7Se8−wIw, respectively.
(Figure 8a) and Cu1.7Bi4.7−vMvSe8 (Figure 8b), M = Zn and In, u
and v = 0.0085 and 0.0255, respectively), and I-doped
Cu1.7Bi4.7Se8 compounds (Cu1.7Bi4.7Se8−wIw (Figure 7c), w =
0.08 and 0.16). By small amount of In substitution on Cu site, ZT
increased owing to the due to the synergetic effects of increase in
PF and decrease in κlat, and the largest ZT value (0.24 at 560 K)
was observed for Cu1.6915In0.0085Bi4.7Se8, yielding an enhance-
ment of greater than 30% compared with that of a pristine
Cu1.7Bi4.7Se8.
Figure 7. Experimental and calculated lattice thermal conductivity as a
function of temperature for Cu1 . 6 9 1 5 Zn0 . 0 0 8 5 Bi4 . 7 Se8 ,
Cu1.7Bi4.6745In0.0255Se8, and Cu1.7Bi4.7Se8. In the parentheses, U (10−18
s K−1), B (107 s−1), and I (10−40 s3) denote scattering prefactors for
Umklapp, boundary, and impurity scatterings, respectively.
CONCLUSIONS
■
strong influence on the κlat, indicating that the suppression of κlat
is mainly due to the point defect scattering induced by doping. In
particular, significant reduction (∼34%) in κlat is observed in the
case of the modification of Cu site with Zn ion, decreasing from
0.56 W m−1 K−1 for undoped sample to 0.37 W m−1 K−1 for
doped compound at 300 K. This is an unexpected result because
the mass and the size differences between Cu and isoelectronic
Zn are only 2.7% and 5%, respectively. This might be relevant to
the preferential occupations of dopants at statistically distributed
interstitial Cu sites as well as to the fractional occupancies of
those sites in this system. Similar effects of interstitial atoms on
In this article, we discussed how the electronic and thermal
transport characteristics can be manipulated through a variety of
strategies involving structural features of each atomic site in
complex structured Cu1.7Bi4.7Se8 pavonite homologues. Synchro-
tron X-ray diffractions and successive Rietveld refinements
revealed that this complex structure showed coexistence of
substitutional and interstitial Cu sites with fractional occupancies
as well as randomly distributed multiple disorders. It is verified
that the structural features, including atomic bonding and
configurations, are strongly correlated with both electronic and
thermal transport properties in this system. In particular, we
F
dx.doi.org/10.1021/ic5014945 | Inorg. Chem. XXXX, XXX, XXX−XXX

Inorganic Chemistry
Article
(20) Wang, H.; LaLonde, A. D.; Pei, Y.; Snyder, G. J. Adv. Funct. Mater.
2013, 23, 1586.
(21) Klemens, P. G. Solid State Physics; Academic Press: New York,
found that σ and κ are significantly affected by the modification of
interstitial Cu sites. These indicated that the doping strategy in
complexly structured materials should be established based on
the understanding of structural characteristics for each atomic
site correlated to the physical properties. Furthermore, we
suggest that the complex structure composed of interstitial
atomic sites may be a guideline for the discovery of promising TE
materials.
1958.
(22) Cargnoni, F.; Nishibori, E.; Rabiller, P.; Bertini, L.; Snyder, G. J.;
Christensen, M.; Gatti, C.; Iversen, B. B. Chem.Eur. J. 2004, 10, 3861.
(23) Liu, M.; Qin, X.; Liu, C.; Pan, L.; Xin, H. Phys. Rev. B 2010, 81,
245215.
ASSOCIATED CONTENT
* Supporting Information
■
S
Crystallographic information file (CIF) for the synchrotron X-
ray powder diffraction data, lattice thermal conductivity
calculation, structural parameters calculated from Rietveld
refinements, additional refinement results. This material is
available free of charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Authors
*E-mail: khlee2014@kangwon.ac.kr. (K.H.L.)
*E-mail: kimsungwng@skku.edu. (S.W.K.)
■
Author Contributions
◊These authors contributed equally.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This work was supported by IBS-R011-D1, the National
Research Foundation of Korea (2013R1A1A1008025), the
Human Resources Development program (No.
20124010203270) of the Korea Institute of Energy Technology
Evaluation and Planning (KETEP) grant funded by the Korea
government Ministry of Trade, Industry and Energy.
REFERENCES
(1) Snyder, G. J.; Toberer, E. S. Nat. Mater. 2008, 7, 105.
(2) Wright, D. A. Nature 1958, 181, 834.
■
(3) Cho, J. Y.; Mun, H. A.; Ryu, B.; Kim, S. I.; Hwang, S.; Roh, J. W.;
Yang, D. J.; Shin, W. H.; Lee, S. M.; Choi, S.-M.; Kang, D. J.; Kim, S. W.;
Lee, K. H. J. Mater. Chem. A 2013, 1, 9768.
(4) Goldsmid, H. J. J. Appl. Phys. 1961, 32, 2198.
(5) Yang, J.; Meisner, G. P.; Morelli, D. T.; Uher, C. Phys. Rev. B 2000,
63, 014410.
(6) Rodriguez-Carvajal, J. Phys. B 1993, 192, 55.
(7) Perry, D. L. Handbook of Inorganic Compounds; 2nd ed.; CRC
Press: London, U.K., 2011.
(8) Allgaier, R. S.; Scanlon, W. W. Phys. Rev. 1958, 111, 1029.
(9) Brown, D. M.; Bray, R. Phys. Rev. 1962, 127, 1593.
(10) Snyder, G. J.; Christensen, M.; Nishibori, E.; Caillat, T.; Iversen,
B. B. Nat. Mater. 2004, 3, 458.
(11) Liu, H.; Shi, X.; Xu, F.; Zhang, L.; Zhang, W.; Chen, L.; Li, Q.;
Uher, C.; Day, T.; Snyder, G. J. Nat. Mater. 2012, 11, 422.
(12) Callaway, J.; Von Baeyer, H. C. Phys. Rev. 1960, 120, 1149.
(13) Slack, G. A.; Galginaitis, S. Phys. Rev. 1964, 133, A253.
(14) Glassbrenner, G. A.; Slack, G. A. Phys. Rev. 1964, 134, A1058.
(15) Morelli, D. T.; Heremans, J. P.; Slack, G. A. Phys. Rev. B 2002, 66,
195304.
(16) Abeles, B. Phys. Rev. 1963, 131, 1906.
(17) He, J.; Girard, S.; Kanatzidis, M.; Dravid, V. Adv. Funct. Mater.
2010, 20, 764.
(18) Pei, Y.; Shi, X.; LaLonde, A. D.; Wang, H.; Chen, L.; Snyder, G. J.
Nature 2011, 473, 66.
(19) Wang, H.; Pei, Y.; LaLonde, A. D.; Snyder, G. J. Proc. Natl. Acad.
Sci. U.S.A. 2012, 109, 9705.
G
dx.doi.org/10.1021/ic5014945 | Inorg. Chem. XXXX, XXX, XXX−XXX