3216 Inorg. Chem.

2010, 49, 3216–3223

DOI: 10.1021/ic902072v

Correlation between the Structural Distortions and Thermoelectric Characteristics

in La1-xAxCoO3 (A = Ca and Sr)
Yang Wang,†,‡ Yu Sui,*,‡,§ Peng Ren,† Lan Wang,† Xianjie Wang,‡ Wenhui Su,‡ and Hong Jin Fan*,†
†
Division of Physics and Applied Physics, School of Physical and Mathematical Sciences, Nanyang
Technological University, 21 Nanyang Link, 637371 Singapore, Singapore, ‡Center for Condensed Matter
Science and Technology (CCMST), Department of Physics, Harbin Institute of Technology, Harbin 150001,
People’s Republic of China, and §International Center for Materials Physics, Academia Sinica, Shenyang
110015, People’s Republic of China
Received October 20, 2009

Detailed structures and thermoelectric (TE) properties are investigated for the perovskite La1-xCaxCoO3 and
La1-xSrxCoO3 with 0 e x e 0.3. The monoclinic crystal structures for all samples are refined from powder X-ray
diffraction, which reveals that the La1-xSrxCoO3 series has smaller global distortions but larger local distortions. The
different structural distortions strongly influence the transport and TE properties in terms of resistivity, thermopower,
thermal conductivity, bandwidth, and effective mass as well as electronic correlation. These results unambiguously
demonstrate a close correlation between the structural distortions and TE characteristics in this family. Furthermore, a
new approach is suggested to improve the TE performance by controlling the structure factors in such strongly
correlated oxide systems. The effects of the spin-state transition of Co ions on the TE properties are also discussed.

Introduction (e.g., cobaltites, titanates, and manganites) has aroused

Recently, as one type of energy conversion technology, renewed interest.3-8
thermoelectric (TE) energy conversion has received increas- Among these metal oxides, perovskite-type cobaltites
ing attention. TE materials are solid-state energy converters R1-xAxCoO3 (R denotes a rare earth, and A denotes an
of which the combination of thermal and electrical properties alkaline earth) exhibit rich physical properties, such as
allows them to directly convert waste heat into electricity or complex spin-state transitions and spin glass behavior.9,10
electrical power into cooling and heating.1,2 The advantages For example, for the La1-xCa/SrxCoO3 system, the parent
of TE devices include solid-state operation, vast scalability, LaCoO3 has a nonmagnetic insulating-like ground state; with
zero emissions, no maintenance, and long lifetime. Usually, increasing temperature, a paramagnetic insulating state gra-
the efficiency of TE materials is described by a dimensionless dually develops above ∼100 K and an insulator-metal
figure of merit ZT = S2T/Fκ, where S, F, and κ are thermo- transition occurs around 500 K.10 The spin-state transition
power, resistivity, and thermal conductivity, respectively. of Co3þ in LaCoO3 at around 100 K remains controversial,
For practical applications, ZT > 1 is required. Among TE but the low-spin (LS; t62g) to intermediate-spin (IS; t52ge1g)
materials, metal oxides are more suitable for high-temperature scenario is a consensus and is consistent with many experi-
TE applications than metal alloys because of the structural mental results.11 By divalent Ca2þ or Sr2þ doping, the
and chemical stabilities, oxidation resistance, and low cost of nonmagnetic ground state is strongly suppressed, and the
metal oxides. Although the performances of earlier TE oxides LS-IS transition is shifted to lower temperatures, which
are very far from application criterion, the recent discovery suggests that doping stabilizes the IS state.12 Usually, light
of a large TE response in some 3d transition-metal oxides
(7) Flahaut, D.; Mihara, T.; Funahashi, R.; Nabeshima, N.; Lee, K.;
Ohta, H.; Koumoto, K. J. Appl. Phys. 2006, 100, 084911.
*To whom correspondence should be addressed. E-mail: fanhj@ (8) Bocher, L.; Aguirre, M. H.; Logvinovich, D.; Shkabko, A.; Robert,
ntu.edu.sg (H.J.F.), suiyu@hit.edu.cn (Y.S.). R.; Trottmann, M.; Weidenkaff, A. Inorg. Chem. 2008, 47, 8077.
(1) Bell, L. E. Science 2008, 321, 1457. (9) Takami, T.; Zhou, J. S.; Goodenough, J. B.; Ikuta, H. Phys. Rev. B
(2) Snyder, G. J.; Toberer, E. S. Nat. Mater. 2008, 7, 105. 2007, 76, 144116.
(3) Terasaki, I.; Sasago, Y.; Uchinokura, K. Phys. Rev. B 1997, 56, (10) Kriener, M.; Zobel, C.; Reichl, A.; Baier, J.; Cwik, M.; Berggold, K.;
R12685. Kierspel, H.; Zabara, O.; Freimuth, A.; Lorenz, T. Phys. Rev. B 2004, 69,
(4) Ohta, H.; Sugiura, K.; Koumoto, K. Inorg. Chem. 2008, 47, 8429. 094417.
(5) Androulakis, J.; Migiakis, P.; Giapintzakis, J. Appl. Phys. Lett. 2004, (11) Kobayashi, Y.; Fujiwara, N.; Murata, S.; Asai, K.; Yasuoka, H.
84, 1099. Phys. Rev. B 2000, 62, 410.
(6) Masset, A. C.; Michel, C.; Maignan, A.; Hervieu, M.; Toulemonde, (12) Tokura, Y.; Okimoto, Y.; Yamaguchi, S.; Taniguchi, H.; Kimura,
O.; Studer, F.; Raveau, B.; Hejtmanek, J. Phys. Rev. B 2000, 62, 166. T.; Takagi, H. Phys. Rev. B 1998, 58, R1699.

pubs.acs.org/IC Published on Web 03/05/2010 r 2010 American Chemical Society

Article Inorganic Chemistry, Vol. 49, No. 7, 2010 3217
doping can lead to spin glass behavior, whereas heavier
doping causes ferromagnetic ordering.13,14 Particularly, the
La1-xCa/SrxCoO3 family shows good TE characteristics for
reasons that are not fully understood; the spin degeneracy of
Co ions may be a possible source.15 It has been reported that
the ZT value of La1-xSrxCoO3 ceramics can reach 0.18 at
300 K.5 Nevertheless, because of the relatively high κ in single
crystals, ZT of La1-xSrxCoO3 single crystals is much
smaller than that of ceramic samples.16 To optimize the TE
properties of this cobalt oxide family, several investigations
have been conducted.17-23 However, most of this research
concerns only the TE results but pays little attention to the
factors determining the TE response in this family. Systema-
tic and detailed studies are still scanty.
In addition, the crystal structure of the La1-xAxCoO3
system also remains controversial. Although the crystal
structure of LaCoO3 has been interpreted to have rhombo-
hedral R3c symmetry for a long time,24,25 this space group is Figure 1. XRD patterns and Rietveld refinement results for La0.7Ca0.3-
incompatible with cooperative Jahn-Teller (JT) distortion CoO3 at room temperature. The experimental data are shown as dots; the
global fitting profile and the difference curve are shown as solid lines; the
because Co3þ (IS) is strongly JT-active. Recent high-resolution calculated reflection positions are indicated by stick marks. The inset
X-ray studies revealed that LaCoO3 has monoclinic distor- shows the SEM micrograph of La0.7Ca0.3CoO3.
tion (space group I2/a) above 90 K,9,26 compatible with
JT distortion. Considering the strong coupling between the under a high pressure of 3 GPa with a special steel die, and
structure and physical properties, in this study, we focus on then sintered in air at 1473 K for 24 h. The pellets were slowly
cooled down to room temperature in a furnace.
the effects of structural distortions on the TE characteristics
Powder XRD data at room temperature were collected
of Ca- and Sr-doped LaCoO3 and discuss the different TE using an XRD diffractometer (D8 Advanced) with Cu KR
properties in these two systems. By step-scanned X-ray (λ = 0.154 06 nm) radiation. The data were collected from 10

diffraction (XRD) data, good refinement results based on up to 140
in 2θ with 0.02
steps and a counting time of 30 s
space group I2/a were obtained. Then we found that La1-x- per step. The O content in the samples was determined by
CaxCoO3 and La1-xSrxCoO3 do not have identical global iodometic titration. The results show that the oxygen stoichio-
and local distortions, and the different structural distortions metry for all of the samples is equal to 3.00, except for x = 0.3
induced by Ca and Sr doping are responsible for the different samples, whose oxygen stoichiometry is ∼2.98-2.99. The
TE responses. microstructure was observed with a JEOL JSM-6700F scan-
ning electron microscope. Temperature dependences of re-
sistivity, thermopower, thermal conductivity, and the Hall
Experimental Section coefficient were all measured by using the commercial Quan-
Polycrystalline samples of La1-xCaxCoO3 and La1-xSrx- tum Design physical property measurement system. Disk-
CoO3 with 0 e x e 0.3 were prepared by a standard solid-state shaped samples were cut into bars in dimensions of 10 mm 
reaction. Reagent-grade La2O3, Co3O4, CaCO3, and SrCO3 1 mm  1 mm. Resistivity and thermopower measurements
powders were mixed in stoichiometric quantities and calcined were performed by a standard four-probe method, whereas
in air at 1273 K for 24 h. Then the mixture was ground, the Hall coefficient was measured using a five-probe method,
pressed into disk-shaped pellets, and sintered in air at 1373 K with copper wire as leads. Thermal conductivity and specific
for 24 h. Finally, in order to get compact ceramic samples, the heat were measured using a steady-state technique in a closed-
products were repeatedly ground, cold-pressed into pellets cycle refrigerator pumped down to 10-5 Torr.

(13) Sathe, V. G.; Pimpale, A. V.; Siruguri, V.; Paranjpe, S. K. J. Phys.: Results and Discussion
Condens. Matter 1996, 8, 3889.
(14) Wu, J.; Leighton, C. Phys. Rev. B 2003, 67, 174408. Doping-Induced Structural Distortions. The XRD pat-
(15) Koshibae, W.; Tsutsui, K.; Maekawa, S. Phys. Rev. B 2000, 62, 6869. tern and scanning electron microscopy (SEM) micro-
(16) Berggold, K.; Kriener, M.; Zobel, C.; Reichl, A.; Reuther, M.; graph for La0.7Ca0.3CoO3 are shown in Figure 1. Other
M€uller, R.; Freimuth, A.; Lorenz, T. Phys. Rev. B 2005, 72, 155116.
(17) Robert, R.; Bocher, L.; Trottmann, M.; Reller, A.; Weidenkaff, A. samples have similar XRD patterns and microstructure.
J. Solid State Chem. 2006, 179, 3839. The SEM micrographs indicate that these ceramics fab-
(18) He, T.; Chen, T. G.; Calvarese, T. G.; Subramanian, M. A. Solid ricated by a cold high-pressure method are quite dense,
State Sci. 2006, 8, 467.
(19) Hebert, S.; Flahaut, D.; Martin, C.; Lemonnier, S.; Noudem, J.; without obvious pores. The typical grain size is ∼3-5 μm.
Goupil, C.; Maignan, A.; Hejtmanek, J. Prog. Solid State Chem. 2007, 35, The XRD data confirm the single-phase nature of the
457. synthesized compounds. The structural parameters, as
(20) Zhou, A. J.; Zhu, T. J.; Zhao, X. B. Mater. Sci. Eng., B 2006, 128, 174.
(21) Zhang, X.; Li, X. M.; Chen, T. L.; Chen, L. D. J. Cryst. Growth 2006, summarized in Table 1, were determined by the Rietveld
286, 1. refinement method, using the profile analysis program
(22) Liu, Y.; Qin, X. Y.; Xin, H. X.; Zhang, J.; Li, H. J.; Wang, Y. F. Fullprof. Herein the reliability factors Rwp, RB, and RF
J. Appl. Phys. 2007, 101, 083709. mean weighed profile, Bragg factor, and crystallographic
(23) Iwasaki, K.; Ito, T.; Nagasaki, T.; Arita, Y.; Yoshino, M.; Matsui, T.
J. Solid State Chem. 2008, 181, 3145. factor, respectively. The Rietveld refinement of all of the
(24) Thornton, G.; Tofield, B. C.; Hewat, A. W. J. Solid State Chem. samples was carried out by considering a monoclinic
1986, 61, 301. structure described by space group I2/a. As mentioned
(25) Radaelli, P. G.; Cheong, S. W. Phys. Rev. B 2002, 66, 094408.
(26) Maris, G.; Ren, Y.; Volotchaev, V.; Zobel, C.; Lorenz, T.; Palstra, already, most neutron diffraction or XRD experiments of
T. T. M. Phys. Rev. B 2003, 67, 224423. LaCoO3 have been interpreted in rhombohedral space
3218 Inorganic Chemistry, Vol. 49, No. 7, 2010 Wang et al.
Table 1. Refined Cell Parameters, Atomic Positions, Isotropic Thermal Factor Biso, Fitting Factors (Rwp, RB, and RF), A-Site Average Ionic Radii ÆrAæ, A-Site Cation Size
Variance σ2, and Tolerance Factor t for All of the Samples

La1-xCaxCoO3 La1-xSrxCoO3

LaCoO3 x = 0.05 x = 0.1 x = 0.2 x = 0.3 x = 0.05 x = 0.1 x = 0.2 x = 0.3

a (Å) 5.3682(3) 5.3671(4) 5.3633(3) 5.3595(5) 5.3522(3) 5.3698(4) 5.3716(4) 5.3881(3) 5.4018(5)
b (Å) 5.4326(4) 5.4321(4) 5.4315(5) 5.4304(3) 5.4289(2) 5.4335(3) 5.4343(4) 5.4365(4) 5.4396(4)
c (Å) 7.6386(5) 7.6376(3) 7.6357(4) 7.6329(5) 7.6298(4) 7.6421(4) 7.6469(5) 7.6543(5) 7.6604(5)
β (deg) 90.9818(7) 90.9537(8) 90.9263(7) 90.8735(9) 90.8116(8) 90.8856(7) 90.8124(8) 90.6715(8) 90.4643(9)
yLa/Ca/Sr 0.24911(14) 0.24934(10) 0.24951(15) 0.24976(12) 0.24985(13) 0.24942(13) 0.24975(17) 0.24993(12) 0.25046(16)
Biso(La/Ca/Sr) (Å2) 0.18(2) 0.18(3) 0.19(3) 0.18(2) 0.17(3) 0.17(2) 0.17(2) 0.16(2) 0.16(3)
Biso(Co) (Å2) 0.39(3) 0.40(3) 0.41(3) 0.40(4) 0.42(5) 0.41(4) 0.39(4) 0.40(4) 0.42(5)
yO(1) -0.29973(16) -0.29635(15) -0.29372(18) -0.28841(23) -0.28882(17) -0.29134(17) -0.28879(21) -0.28365(18) -0.28517(20)
Biso[O(1)] (Å2) 0.52(3) 0.54(4) 0.53(3) 0.54(3) 0.52(4) 0.55(3) 0.55(4) 0.56(4) 0.58(4)
xO(2) 0.02129(13) 0.02115(12) 0.02101(15) 0.02058(13) 0.01934(18) 0.02104(14) 0.02087(12) 0.02036(15) 0.0187(16)
yO(2) 0.03604(14) 0.03634(15) 0.03628(12) 0.03673(11) 0.03725(15) 0.03659(13) 0.03713(14) 0.03766(12) 0.03772(13)
zO(2) 0.22865(24) 0.22891(19) 0.22902(21) 0.22905(17) 0.22897(22) 0.22885(23) 0.22893(18) 0.23011(17) 0.23163(22)
Biso[O(2)] (Å2) 0.65(3) 0.63(4) 0.62(4) 0.63(3) 0.62(4) 0.64(3) 0.64(4) 0.63(4) 0.63(3)
Rwp (%) 6.22 6.12 6.03 5.87 5.96 5.95 6.13 5.83 5.92
RB (%) 3.36 3.34 3.78 3.55 3.47 3.44 3.61 3.82 3.71
RF (%) 2.58 2.61 2.72 2.57 2.62 2.69 2.58 2.66 2.59
ÆrAæ (Å) 1.2160 1.2142 1.2124 1.2088 1.2052 1.2207 1.2254 1.2348 1.2442
σ2 (Å2) 0 0.00006 0.00012 0.00021 0.00027 0.00042 0.00079 0.00141 0.00186
t 0.92029 0.92136 0.92270 0.92513 0.92757 0.92365 0.92729 0.93435 0.94146

group R3c.24,25 However, Maris et al. recently found the increases with doping, which also suggests a decrease of
monoclinic distortion for LaCoO3 by performing a high- global distortions. Furthermore, Sr-doped samples show
angular-resolution scan using synchrotron radiation, and lower β and larger t, indicating that they have lower global
they analyzed their XRD data based on I2/a symmetry, distortions than Ca-doped samples, although P they have
which was well consistent with the observed JT distor- larger size mismatching σ2 (defined by σ2 = yiri2 - ÆrAæ2,
tion.26 Also, Takami et al. refined their XRD spectra of where ri is the ionic radius, yi is the fractional occupancy of
La1-xSrxCoO3 (0 e x e 0.3) in both space groups R3c and the ith atoms of the A site, and ÆrAæ is the average A-site
I2/a and found that the R factors of the fits were smaller radius).
when the space group I2/a was assumed.9 Hence, we also Figure 2 shows variations of the bond length dCo-O and
refined our XRD data on the basis of I2/a monoclinic the bond angle θCo-O-Co as a function of the doping level
perovskite symmetry, with a ∼ b ∼ 21/2ap and c ∼ 2ap x determined by structure analysis. In I2/a symmetry,
(where ap is the lattice parameter of an ideal cubic there are two different Co-O-Co bond angles, both
perovskite structure). As can be seen in Figure 1, all of of which increase as the Ca/Sr content increases. The
the reflections can be indexed by a monoclinic perovskite increasing θCo-O-Co consists of alterations of β and t,
structure and good agreement is obtained between the meaning reductions of the global structure distortion.
observed and calculated XRD patterns. The refined Also, θCo-O-Co of Sr-doped samples is closer to an ideal
quality factors (Rwp, RB, and RF) of all samples are small, 180
than that of Ca-doped samples, corresponding to the
ensuring the high reliability of the refinement. lighter global distortion in the La1-xSrxCoO3 series. In
From Table 1, one can find that the lattice parameters R3c symmetry, all of the Co-O bond lengths are iden-
a, b, and c decrease slightly with an increase in the Ca tical, so JT distortion cannot exist. However, unlike space
content, whereas Sr doping causes an increase in all of the group R3c, there are three unequal Co-O bond lengths in
a-, b-, and c-axis lengths. This should be attributed to the space group I2/a, one Co-O(1) and two Co-O(2). As can
variation of the A-site radius because the ionic radius of been seen from Figure 2a, all samples show a clear
La3þ (1.216 Å) is much smaller than that of Sr2þ (1.31 Å) difference in the bond lengths, indicating that these
but only slightly larger than that of Ca2þ (1.18 Å). (Here samples are JT-distorted. With Ca or Sr doping, Co-O(1)
we use the nine-coordinated A-site ionic radii because the exhibits only a quite weak x dependence, while one
coordination number in orthorhombic or monoclinic Co-O(2) increases but the other Co-O(2) decreases as
distorted perovskites is usually nine.) It should be noted x increases. This behavior suggests that the local CoO6
that, while Sr doping brings Co4þ ions into the octahedral-site distortion (or JT distortion) is enhanced
Co3þ matrix, the ionic radii of Co4þ (0.53 Å) and Co3þ by doping. In addition, Sr doping induces a more notable
(0.545 Å) are very close. Therefore, the increase in the separation between the two Co-O(2) bonds, which in-
lattice parameters in La1-xSrxCoO3 is mainly due to the dicates that the La1-xSrxCoO3 series has more evident
size difference between Sr2þ and La3þ. For these two local CoO6 distortion than La1-xCaxCoO3.
dopings, the monoclinic angle β systematically decreases In order to probe the local distortion in this system,
with an increase in the Ca/Sr content, implying that a higher variations of the JT distortion (calculated by dividing the
symmetry (orthorhombic or cubic) is approached. The difference between the two Co-O(2) bond lengths by
tolerance factor t, defined as t = (rA þ rO)/21/2(rB þ rO),27 their average) along with CoO6 octahedral distortion are
was also calculated for all of the samples. t gradually plotted in Figure 3. As we know, CoO6 octahedra in
LaCoO3 with the perovskite structure are not rigid. The
(27) Goldschmidt, V. M. Naturwissenschaften 1926, 14, 477. cooperative tilts and rotations of the CoO6 octahedra are
Article Inorganic Chemistry, Vol. 49, No. 7, 2010 3219

Figure 2. (a) Bond length dCo-O and (b) bond angle θCo-O-Co as a function of doping level x for La1-xCaxCoO3 and La1-xSrxCoO3.

Figure 3. x dependence of (a) JT distortion and (b) Q and φ for La1-xCaxCoO3 and La1-xSrxCoO3.

around all three axes, which change the cubic lattice Effects of Doping on Electric Transport. All of the
parameters from ap  ap  ap to a ∼ 21/2ap, b ∼ 21/2ap, La1-xCaxCoO3 samples exhibit semiconducting-like be-
and c ∼ 2ap and cause the bending of Co-O-Co bond havior (i.e., dF/dT < 0), and F decreases observably with
angles and the splitting of Co-O bond lengths. The bond- increasing x (see Figure 4). The reduction of F with doping
length splitting can be described by two degenerate can be attributed to an increase in the carrier concentra-
vibrational modes, Q2 and Q3, determined by Q2 = tion n as well as the mobility μ. By Hall coefficient
dCo-O(2)1 - dCo-O(2)2 and Q3 = (2dCo-O(1) - dCo-O(2)1 - measurements, the evaluated n and μ monotonically
dCo-O(2)2)/31/2. In addition to JT distortion, two distor- increase with the Ca/Sr content (see Table 2). In this
tion parameters Q = (Q22 þ Q32)1/2 and φ = tan-1(Q3/ family, the substitution of Ca2þ or Sr2þ for La3þ will yield
Q2) can directly reflect the local CoO6 distortion in this large numbers of hole carriers based on the valence
system. Figure 3 clearly indicates that the local distortions equilibrium, which can facilitate the carrier transport
(including the degree of JT distortion, Q, and φ) increase and thus lower F. Moreover, the increase in n is consider-
with doping, although the global distortion decreases as able for lightly doped samples, whereas it becomes gentle
mentioned above. The enhanced local distortion can be for heavily doped samples, suggesting that the doped
well correlated with the behavior of the spin-state transi- carrier is nearly saturated. In contrast, the enhancement
tion with Ca/Sr doping.9 The temperature of the spin- of μ becomes noticeable with gradual doping, which
state transition, which is ∼100 K for undoped LaCoO3, implies that μ plays a more important role in F for heavily
decreases with the substitution of Ca2þ or Sr2þ for La3þ. doped samples. La1-xSrxCoO3 has similar F-T curves
Consequently, doping stabilizes the IS state, so the po- except that La0.7Sr0.3CoO3 shows a metallic-like behavior
pulation of the JT-active IS state increases, resulting in an (i.e., dF/dT > 0; see the inset of Figure 4b). Overall, the
increase in the CoO6 local distortion. It should be noted temperature and doping level dependences of F agree well
that these enhanced local distortions cannot be revealed with previous studies.10,23,28 In addition, herein the F data
by the usual structural parameters (monoclinic angle, are very close to those of reported single-crystal samples
(usually, the metallic-like transport of La0.7Sr0.3CoO3 can
bond angle, and tolerance factor). Additionally, com-
only be observed in single-crystal samples).10,29 Hence, it
pared with La1-xCaxCoO3, the La1-xSrxCoO3 series has
implies that in our highly compact ceramics, which were
smaller global distortion but larger local distortion be-
fabricated by the cold high-pressure method, the grain
cause it shows higher degree of JT distortion, Q, and φ. In
the following, we will see that these doping-induced
(28) Onose, Y.; Tokura, Y. Phys. Rev. B 2006, 73, 174421.
structural distortions strongly influence the TE charac- (29) Kobayashi, Y.; Murata, S.; Asai, K.; Tranquada, J. M.; Shirane, G.;
teristics in this system. Kohn, K. J. Phys. Soc. Jpn. 1999, 68, 1011.
3220 Inorganic Chemistry, Vol. 49, No. 7, 2010 Wang et al.

Figure 4. (a) Temperature dependence of resistivity F for La1-xCaxCoO3. (b) Room-temperature resistivity as a function of x for La1-xCaxCoO3 and
La1-xSrxCoO3. The inset of part b presents F-T curves of La0.8Sr0.2CoO3 and La0.7Sr0.3CoO3.

Table 2. Room-Temperature Carrier Concentration n, Carrier Mobility μ, Effective Mass m*, Relaxation Time τ, and Bandwidth W as Well as F, S, κ, and ZT at 300 K
W F300 K S300 K κ300 K
x n (1021 cm-3) μ (cm2 V-1 s-1) m* (me) τ (10-16 s) (Å-3.5) (mΩ cm) (μV K-1) (W m-1 K-1) ZT300 K
LaCoO3 0 0.0126 0.026 12.66 1.87 0.009 06 10858 560.2 1.88 0.0005
La1-xCaxCoO3 0.05 3.44 0.039 9.31 2.07 0.010 62 119.2 280.3 1.98 0.0099
0.1 5.32 0.181 6.63 6.83 0.013 78 6.51 168.8 2.32 0.0566
0.2 6.24 0.963 3.09 16.92 0.067 79 1.04 74.2 2.91 0.0346
0.3 6.76 1.21 1.43 9.84 0.096 35 0.77 25.5 3.29 0.0077
La1-xSrxCoO3 0.05 3.52 0.067 8.51 3.24 0.011 83 68.5 249.9 1.95 0.0141
0.1 5.61 0.313 5.27 9.38 0.019 48 3.58 159.7 2.20 0.0972
0.2 6.47 1.20 2.78 18.97 0.076 99 0.81 62.4 2.75 0.0524
0.3 7.02 1.51 1.32 11.34 0.099 44 0.59 21.1 3.16 0.0072

boundaries have a negligible influence on the carrier The calculated W is listed in Table 2. W monotonously
conduction. increases with x, corresponding to decreasing F. Because
The differences in room-temperature resistivity bet- for the Ca-doped series the deviation from cubic symme-
ween La1-xCaxCoO3 and La1-xSrxCoO3 are shown in try is more pronounced compared with that of the Sr-
Figure 4b. For the same doping level, La1-xSrxCoO3 doped samples, the former has a stronger deviation of the
exhibits lower F (this F distinction actually exists not only Co-O-Co angle from 180
, and correspondingly the
at room temperature but also in the whole measured bandwidth (i.e., hopping integral) is smaller than that of
temperature range). From the Hall coefficient measure- La1-xSrxCoO3. As a consequence, the La1-xCaxCoO3
ments, the Ca-doped series has comparable n with the Sr- series shows higher F in the same doping level and remains
doped series but has much smaller μ (especially for lightly semiconducting-like even for x = 0.3. For La0.7Sr0.3-
doped samples); see Table 2. Considering the close cou- CoO3, when the bandwidth increases to approach the
pling between the crystalline structure and transport narrow-band limit of itinerant electrons, the compound
properties in such a strongly correlated narrow-band shows a metallic F-T curve; namely, a doping-induced
system, we have to take into account the effective band- metal-semiconductor transition occurs. In summary,
width determined by structure distortions. The electronic this result suggests that structural variation by doping
structure revealed by photoemission spectroscopy also has a strong effect on the electrical transport; higher
implies that the band structure that mainly depends on symmetry and lower global distortion (i.e., La1-xSrx-
the valence of Co ions has a close correlation with the CoO3 system) are advantageous to the enhancement of
transport characteristics in this system.30-32 On the basis electric conduction.
of the refinement results, the bandwidth W of this family Figure 5a presents thermopower S versus temperature
can be estimated by of La1-xCaxCoO3. Similar temperature and x depend-
cos½ðπ -θ Co -O -Co Þ=2 ences were observed in La1-xSrxCoO3. The signs of S for
W 7=2
ð1Þ all of the samples are positive, indicating the nature of the
d Co -O hole carrier in this system, as is consistent with the Hall
coefficient measurements. It has been reported that un-
where d Co-O and θhCo-O-Co are the average Co-O bond doped LaCoO3 can show both positive and negative S16,34
length and Co-O-Co bond angle, respectively.33 because S of pure LaCoO3 is extremely sensitive to the O
content. The tiny deviation in the O content in
(30) Pandey, S. K.; Kumar, A.; Banik, S.; Shukla, A. K.; Barman, S. R.;
Pimpale, A. V. Phys. Rev. B 2008, 77, 113104. LaCoO3þδ, which causes a small concentration 2δ of hole
(31) Pandey, S. K.; Kumar, A.; Patil, S.; Medicherla, V. R. R.; Singh, or electron doping, is below the accuracy of oxygen
R. S.; Maiti, K.; Prabhakaran, D.; Boothroyd, A. T.; Pimpale, A. V. Phys. determination. Herein the positive S of our pure LaCoO3
Rev. B 2008, 77, 045123. indicates that there should be a very slight excess of
(32) Saitoh, T.; Mizokawa, T.; Fujimori, A.; Abbate, M.; Takeda, Y.;
Takano, M. Phys. Rev. B 1997, 56, 1290.
(33) Medarde, M.; Mesot, J.; Lacorre, P.; Rosenkranz, S.; Fischer, P.; (34) Robert, R.; Bocher, L.; Trottmann, M.; Reller, A.; Weidenkaff, A.
Gobrecht, K. Phys. Rev. B 1995, 52, 9248. J. Solid State Chem. 2006, 179, 3893.
Article Inorganic Chemistry, Vol. 49, No. 7, 2010 3221

Figure 5. (a) Temperature dependence of thermopower S for La1-xCaxCoO3. (b) Room-temperature thermopower as a function of x of La1-xCaxCoO3
and La1-xSrxCoO3 together with the theory curve calculated by eq 5.

oxygen. The sharp collapse of S at around 120 K in where r is the scattering parameter, ζF is the reduced
LaCoO3 may be correlated to the LS-IS transition, chemical potential, and Fr(ζF) is the Fermi integral. In
which alters the configuration entropy evidently. this model, the relationship among n, μ, m*, and Fr(ζF) is
Although S of undoped LaCoO3 can be either positive given by
or negative, even a very slight doping can eliminate the
influence of an excess (or vacancy) of oxygen and induce  
2kB mT 3=2
the collapse of S.16 As mentioned above, the spin-state n ¼ 4π F1=2 ðζF Þ ð3Þ
transition will be drastically suppressed by doping, so the h2
S profile collapses at lower temperatures and becomes less and
pronounced with an increase in the doping level (see
Figure 5a). eτ
μ ¼ ð4Þ
In cobaltites, the larger thermopower may originate m
from strong electronic correlations and large configura-
where h is Planck’s constant and τ is the relaxation time.
tional entropy.15,35-38 The strong electronic correlation
When eqs 2-4 are combined, m* and τ are calculated and
causes a larger electronic effective mass m* and a narrow
listed in Table 2. With doping, m* decreases monoto-
band with a strong energy-dependent density of states in
nously, whereas τ increases first, reaches maxima at x =
the vicinity of Fermi level EF.39 The large configurational
0.2, and then decreases. The reduction of m* is consistent
entropy is driven by the rich spin and orbital degenera-
with a decrease of the electron specific heat coefficient, as
cies. In both cases, S is expected to be dependent on the
revealed from specific heat measurements. These results
carrier concentration: First, a small shift of EF by doping
also correspond well to the above resistivity analysis.
will significantly alter the magnitude of S because of band
Reasons for the changes in m* and τ are proposed as
narrowing due to strong correlation; second, the degen-
follows. The bandwidth increases with an increase in the
eracies will also vary with the carrier concentration
Ca/Sr content, and as a result, m* lessens gradually.
because of a change in the configuration and population
Accordingly, Ca/Sr doping reduces the electronic corre-
ratio of Co3þ and Co4þ ions. Indeed, as shown in Figure 5,
lation, which also facilitates a decrease in S. Therefore,
S decreases monotonously with doping. The increasing
the variation of S with doping actually originates from
carrier density can partially explain the variation of S
both an increase of the carrier concentration and a
because S  ∼ 1/n in a simple Drude picture,40 but the
decrease of the electron correlation. Because S is propor-
change in n cannot account for the decrease of S quanti-
tional to the electron specific heat in strongly correlated
tatively. In order to understand quantitatively the varia-
systems,37 a decrease in the electronic correlation (i.e., a
tion of S, we consider the effects of the electronic
reduction of the electron specific heat and effective mass)
correlation on S and relate S to m* by applying the
is also responsible for the decrease of S. The increase of τ
Boltzmann transport model,41 in which S is described by
for x = 0-0.2 can be also attributed to an increase of the
" # Co-O-Co bond angle and relaxation of distortions. As
kB ðr þ 5=2ÞFr þ 3=2 ðζF Þ for the decrease of τ for x = 0.3, the enhanced point defect
S ¼ -ζ ð2Þ
e ðr þ 3=2ÞFr þ 1=2 ðζF Þ F scattering induced by the formation of oxygen defects
may be a possible source, similar to the report by Iwasaki
et al.23 Finally, Figure 5b also shows that La1-xSrxCoO3
(35) Koshibae, W.; Maekawa, S. Phys. Rev. Lett. 2001, 87, 236603.
(36) Wang, Y. Y.; Rogado, N. S.; Cava, R. J.; Ong, N. P. Nature 2003, has lower S than La1-xCaxCoO3 at the same x values. It
423, 425. can be understood that a larger bandwidth and lighter
(37) Ando, Y.; Miyamoto, N.; Segawa, K.; Kawata, T.; Terasaki, I. Phys. distortion in La1-xSrxCoO3 cause smaller m* but larger τ.
Rev. B 1999, 60, 10580.
(38) Limelette, P.; Hardy, V.; Auban-Senzier, P.; Jerome, D.; Flahaut, D.; At high enough temperatures, S of doped cobaltites is
Hebert, S.; Fresard, R.; Simon, Ch.; Noudem, J.; Maignan, A. Phys. Rev. B expected to be determined by Heikes formula, namely
2005, 71, 233108.
(39) Singh, D. J. Phys. Rev. B 2000, 61, 13397.     
(40) Aschcroft, N. W.; Mermin, N. D. Solid State Physics; Holt Saunders: kB g3 x
Philadelphia, PA, 1976. S ¼ - ln þ ln ð5Þ
(41) Durczewski, K.; Ausloos, M. Phys. Rev. B 2000, 61, 5303. e g4 1 -x
3222 Inorganic Chemistry, Vol. 49, No. 7, 2010 Wang et al.

Figure 6. (a) Temperature dependence of thermal conductivity κ for La1-xCaxCoO3. (b) Thermal conductivity κ at 300 K as a function of x for
La1-xCaxCoO3 and La1-xSrxCoO3. The insets of part b show x dependence of the total thermal conductivity κ, phonon thermal conductivity κph, and
carrier thermal conductivity κcar at 300 K for La1-xCaxCoO3 and La1-xSrxCoO3, respectively.

in which g3 (g4) denotes the number of possible config- which is very sensitive to structural variations. In the LS
urations of the Co3þ (Co4þ) ions.15 In Figure 5b, similar state, the CoO6 octahedra are regular because all six 3d
to Takami et al.’s report,42 we present the measured electrons are in the t2g orbitals, whereas in the IS state,
room-temperature values of S as a function of x along the CoO6 octahedra are expanded because of JT distor-
with the theory curve obtained from eq 5, in which the IS tion. Therefore, when the IS f LS transition occurs, the
states of both Co3þ and Co4þ are adopted according to disappearance of JT distortion can result in an increase
most magnetic experimental results and band calculation.14 in the phonon average free path, and thus κ of LaCoO3
Although this theory assumes a high-temperature limit, will show an evident revival below 100 K. As the
Koshibae et al. pointed out that the room-temperature temperature further decreases, the temperature depen-
S of cobaltites is close to that in the high-temperature dence of κ follows that of specific heat, viz., κ  C  T3
limit,15,43 so that eq 5 can still be used approximately to for T f 0 K, so κ decreases rapidly and reaches zero
describe the role of the configurational entropy in S around finally.
room temperature. On the other hand, a large discrepancy With doping, κ rises gradually, and the low-temperature
between the theory and experiment is observed at high peak disappears. The drastic suppression of κ at low
doping levels (see Figure 5b). Because the high-temperature temperature arises from the enhanced local distortions
limit, where eq 5 is derived, corresponds to the transfer for one thing, and the doping-induced disorder for
integral I , kBT (or bandwidth W , kBT), this condition another. For lightly doped samples (x = 0.05 and 0.1),
is not fulfilled when W (or I) is very large. Therefore, the IS f LS transition occurs at lower temperatures, and
the considerable enhancement of W for heavily doped the disorder (results from the differences between La3þ
samples may violate the assumption of Koshibae’s theory, and Ca2þ/Sr2þ together with Co3þ and Co4þ and
which accounts for the discrepancy. Additionally, the the formed magnetic polarons) is pronounced,16,44 so
Sr-doped series has larger W, so it exhibits a more obvious the low-temperature maximum of κ is already strongly
deviation. suppressed. For heavily doped samples (x = 0.2 and 0.3),
TE Response. Values of thermal conductivity κ of the the IS Co3þ ions become dominant in the whole tempera-
samples as a function of the temperature and doping level ture range, so the low-temperature peak of κ remains
are presented in parts a amd b of Figure 6, respectively. absent.
Overall, κ of all of the samples shows a positive tempera- The increase of κ with doping is actually from the
ture dependence. The undoped LaCoO3 exhibits a notice- contribution by charge carriers. The doping-induced
able peak in κ-T curves at lower temperatures, and the mobile charger carriers, which serve as additional scat-
peak is drastically suppressed by doping. It is noted that terers for phonon, are also expected to transport heat.
the values and temperature dependences of κ of our The insets of Figure 6b show the total thermal conduc-
samples are nearly identical with those of single-crystal tivity κ, phonon thermal conductivity κph, and carrier
samples, suggesting that grain boundaries are not impor- thermal conductivity κcar at room temperature according
tant to the phonon transport.16 In a LaCoO3 insulator, to κ = κph þ κcar, where κcar is estimated by Wiedemann-
the heat is usually carried by acoustic phonons. Optical Franz’s law κcar = L0T/F, with Lorentz constant L0 =
phonons play a minor role because their dispersion is π2kB2/3e2 = 2.45  10-8 W Ω K-2. One can see from
typically much weaker than that of the acoustic these insets that κcar, κ for smaller x but κcar becomes
branches.16 In addition, the carrier thermal conductivity important for larger x, and the increase of the total
is negligible because of the large resistivity. Hence, the thermal conductivity κ with doping mainly results from
low-temperature κ peak in LaCoO3 results from an an increase of κcar. κph of the Sr-doped series remains
increase of the average free path of acoustic phonons, unchanged but that of the Ca-doped series rises
slightly, which is because La1-xSrxCoO3 has larger
(42) Takami, T.; Ikuta, H.; Mizutani, U. Trans. Mater. Res. Soc. Jpn.
2004, 29, 2777.
(43) Korotin, M. A.; Ezhov, S. Y.; Solovyev, I. V.; Anisimov, V. I. Phys. (44) Yamaguchi, S.; Okimoto, Y.; Taniguchi, H.; Tokura, Y. Phys. Rev. B
Rev. B 1996, 54, 5369. 1996, 53, R2926.
Article Inorganic Chemistry, Vol. 49, No. 7, 2010 3223

Figure 7. Room-temperature (a) power factor P and (b) figure of merit ZT as a function of x for La1-xCaxCoO3 and La1-xSrxCoO3.

local distortions that can suppress κph more. This can also response by modulating structural characteristics in this
elucidate why the La1-xSrxCoO3 series has smaller global and other relevant strongly correlated oxide systems.
distortions and lower F (or, say, larger κcar), but
yet they show lower total thermal conductivity κ (see Conclusions
Figure 6b). We have presented a systematic study of the structure and
Using the measured F, S, and κ, we can calculate the TE TE properties of the La1-xCaxCoO3 and La1-xSrxCoO3
power factor P (defined by P = S2/F) and figure of merit series. All of the samples are consistently interpreted with
ZT, as shown in Figure 7 as a function of x. Both P and the space group I2/a, which is characterized by three unequal
ZT exhibit maxima at x = 0.1 in the two series. These Co-O bond lengths and is compatible with cooperative JT
optimal TE values (P ∼ 7.2  10-4 W m-1 K-2 and ZT ∼ distortion. La1-xSrxCoO3 has smaller global distortions but
0.1 at 300 K) are among the highest observed values in larger local distortions than La1-xCaxCoO3; these structural
metal oxides so far. Although the largest ZT is still too features significantly influence the TE characteristics of the
small for technical applications, one can expect the ZT of two systems. As a result, the La1-xSrxCoO3 series exhibits
cobaltites to be further improved because the complex lower F, S, and κ compared with La1-xCaxCoO3. From these
configurations and strong disorder of this system can not observations, a close correlation between the structural
only enhance S but also suppress κ. Our present results distortions and the TE response in this family is obtained.
also clearly reveal the close correlation between the Owing to the different structural distortions, La1-xSrxCoO3
structural distortions and the TE response in this family. shows a better TE performance, and the optimal room-
Because of the distinct global/local distortions between temperature ZT reaches 0.1 at x = 0.1. The correlation
La1-xCaxCoO3 and La1-xSrxCoO3, which biases all of between structure and TEs found in this study may shed light
the TE parameters, the La1-xSrxCoO3 series shows higher on the ongoing search of new TE oxides among strongly
TE response than La1-xCaxCoO3, as seen in Figure 7. correlated systems.
Furthermore, on the basis of our present results, in
addition to the carrier concentration, F and S are also Acknowledgment. This work is supported by start-up
determined by the bandwidth that is controlled by global funding from Nanyang Technological University to
distortions, whereas κ is more dependent on local distor- F.H.J. and the National Natural Science Foundation of
tions. This implies that it is possible to enhance the TE China (Grants 50672019 and 10804024).