Journal of Alloys and Compounds 588 (2014) 562–567

Contents lists available at ScienceDirect

Journal of Alloys and Compounds

journal homepage: www.elsevier .com/locate / ja lcom

Enhanced thermoelectric performance of highly dense and fine-grained(Sr1�xGdx)TiO3�d ceramics synthesized by sol–gel process and sparkplasma sintering

0925-8388/$ - see front matter � 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.jallcom.2013.11.116

⇑ Corresponding author. Tel.: +86 0551 65592750; fax: +86 0551 65591434.E-mail address: xyqin@issp.ac.cn (X. Qin).

Liangliang Li, Yongfei Liu, Xiaoying Qin ⇑, Di Li, Jian Zhang, Chunjun Song, Ling WangKey Laboratory of Materials Physics, Institute of Solid State Physics, Chinese Academy of Sciences, 230031 Hefei, PR China

a r t i c l e i n f o

Article history:Received 26 July 2013Received in revised form 16 November 2013Accepted 18 November 2013Available online 26 November 2013

Keywords:Oxide materialsSol–gel processesThermoelectric

a b s t r a c t

Highly dense and fine-grained (Sr1�xGdx)TiO3�d ceramics (x = 0.05–0.12) were fabricated by a sol–gel pro-cess plus spark plasma sintering, and their thermoelectric properties were investigated from �300 K to1100 K. The results show that the electrical resistivity of the (Sr1�xGdx)TiO3�d ceramics decreases greatlywhile maintaining large absolute value of the Seebeck coefficient |S| upon Gd doping. As a result, a largepower factor (PF) greater than 10 lW cm�1 K�2 is obtained in (Sr0.9Gd0.1)TiO3�d in the temperature rangefrom 400 K to 1000 K, and peak PF reaches 16 lW cm�1 K�2 at �570 K. Simultaneously, the lattice ther-mal conductivity jL reduces remarkably (e.g. jL decreases from 2.7 to 2.13 W m�1 K�1 at �1000 K).Owing to the large PF and low jL, a record value of figure of merit ZT = 0.37 is achieved at 1006 K for(Sr0.9Gd0.1)TiO3�d, indicating that appropriate combinations of fine-grained microstructure with bothhigh densities and proper doping as well as oxygen deficiency are essential to the effective enhancementof thermoelectric performance of STO-based ceramics.

� 2013 Elsevier B.V. All rights reserved.

1. Introduction

Thermoelectric (TE) materials have attracted much interest dueto their potential applications in the conversion between thermaland electrical energy [1]. Generally, the efficiency of a TE materialis evaluated by the dimensionless figure of merit, defined asZT = S2T/qj, where S, q, j and T are the Seebeck coefficient, electri-cal resistivity, total thermal conductivity and absolute tempera-ture, respectively. It is clear that a relatively large power factorPF (S2/q) with a low value of j is desirable to achieve a better TEperformance. Conventional heavy-metal-based solid materialshave higher ZT values, including Bi0.5Sb1.5Te3 (ZT = 1.39 at 390 K)[2], (Bi0.001Pb0.999Te)0.88(PbS)0.12 (ZT = 1.20 at 573 K) [3], etc. How-ever, the low thermal decomposition temperature with toxic heavymetals restricts the applicabilities at high temperatures. With re-spect to high-temperature TE materials, considerable efforts havebeen made on metal oxides because of their lower cost, excellentthermal and chemical stabilities [4–8]. Among bulk oxides, p-typeTE materials, such as Ca2.7Ag0.3Co4O9/Ag-10 wt.% (ZT = 0.5 at1000 K) [6], Bi0.875Ba0.125CuSeO (ZT = 1.1 at 923 K) [7], are wellknown to exhibit a rather better performance. Nevertheless, theTE properties of n-type bulk oxides such as electron-doped

CaMnO3 and ZnO are still low as compared to those of the p-typebulk oxides [4,8].

To date, electron-doped SrTiO3 (STO) is still considered as apromising n-type TE oxide in high temperature range. Because ofthe large Seebeck coefficient even in a heavily carrier doped state,the PF of this material was reported to be as large as 13–15lW cm–1 K–2 at 1000 K [9]. Many methods have been investigatedto fabricate the electron-doped STO polycrystalline oxides, such assolid-state reaction (SSR) [10–14], hydrothermal method [15],combustion synthesis (CS) [16] and sol–gel method with postspark plasma sintering (SPS) [17]. The molecular-scale mixing ofcations in the sol–gel process can prepare oxide at a relativelylower sintering temperature (1203 K), and the rapid densificationof SPS would benefit the fabrication of highly dense compoundsin a short time with ultra-fine grain sizes (�300 nm) [17].The j for (Sr0.92La0.08)TiO3 was decreased to the minimum of1.19 W m–1 K–1 at 773 K due to the enhanced phonon scatteringat the grain boundaries [17]. However, due to the poor electricalconductivity, the maximum ZT (0.08 at 679 K) was significantlylower than that (0.37 at 1045 K) of the high dense sample (97.7%theoretical density) prepared by CS with post-SPS [16]. Obviously,one should focus on not only reduction in j but also maintaining ahigh PF so as to improve the TE performance. This means thatoptimal combinations of grain sizes (volume fraction of grainboundaries) and densifications should be realized and properelement doping with additional donor-type oxygen vacancies

L. Li et al. / Journal of Alloys and Compounds 588 (2014) 562–567 563

should be introduced (to provide additional itinerant carriers tothe conduction band) in order to improve the thermoelectric per-formance of the STO-based oxides [11]. In this study, we combinedthe sol–gel process with SPS to prepare highly dense and fine-grained Gd-doped STO ceramics, (Sr1�xGdx)TiO3�d (here d is theoxygen deficiency), with an emphsis on the effects of Gd dopingon thermoelectric properties of STO system. Since the substitutionof Gd3+ for Sr2+ will increase effectively electron concentration, andlarger mass of Gd than that of La could effectively lower the j ofSTO due to larger mass fluctuation effect [18], the proper Gd-dopedcompounds (Sr1�xGdx)TiO3�d with fine-grained microstructures areexpected to possess high TE performances.

Fig. 1. TG–DSC curves of the (Sr0.95Gd0.05)TiO3 zero gel.

Fig. 2. XRD patterns of (a) the (Sr0.95Gd0.05)TiO3 before and after SPS and (b) theenlarged (110) peaks: a1, a2, and a3 correspond to the sintered powders in air, theannealed powders (before SPS) and the ceramic sample (after SPS).

2. Experimental procedures

Powders of (Sr1�xGdx)TiO3 (x = 0, 0.05, 0.08, 0.10, 0.12) were prepared by a cit-rate sol–gel method. The mixture of Sr(NO3)2 (99.5%), Gd(NO3)3�6H2O (99.99%, alad-din, China) were dissolved in proper de-ionized water, and appropriate molar ratioof citric acid monohydrated (99.5%, Sinopharm Chemical Reagent Co., Ltd, China)was then added to this solution without any pH adjustment. Subsequently, the tet-rabutyl titanate (99%, Sinopharm Chemical Reagent Co., Ltd, China), alcohol and eth-ylene glycol were blended in the 1:20:1 M ratio, then the mixture was slowlydropped into the above solution with continuously stirring at 343 K until a viscoussol was formed (Sr:Gd:Ti in the desired ratios, and the total metal cations: citric acidmolar ratio = 1:2.5–3). The obtained sol was first aged at 343 K for one day to form acomplete viscous gel, and then dried at 453 K for 6 h to promote polymerization.After removing excess solvents, a deep brown and porous solid mass (zero gel)was formed. Then the (Sr1�xGdx)TiO3 powders were obtained by calcining the zerogel at 1223 K in air for 6 h.

In order to introduce more oxygen vacancies, the synthesized (Sr1�xGdx)TiO3

powders were annealed at 1473 K for 2 h in a reducing atmosphere (flow rate:300 ml/min) of mixture of Ar and H2 gases (5 vol.% hydrogen in argon). Finally,the high-density ceramic samples (Sr1�xGdx)TiO3�d were sintered by SPS for5 min at 1648 K under a pressure of 40 MPa with a heating/cooling rate of100 K/min in an Ar flow.

Thermogravimetric (TG) and differential scanning calorimetry (DSC) analyses ofthe zero gel were performed simultaneously in air by a SDTQ600 instrument with aheating rate of 10 K/min from room temperature to 1273 K. The phase structures ofthe samples were investigated by X-ray diffraction (XRD, Philips X’pert PRO) withCu Ka radiation, and the lattice parameters were calculated from the XRD patternsby using a Si standard for calibration. The oxygen deficiency content d in the sin-tered samples was estimated by measuring the weight increase after heating at1373 K in oxygen flow for 24 h. The (Sr1�xGdx)TiO3�d bulk samples were polishedand thermal-etched in vacuum at 1373 K for 10 min, and the surface morphologieswere observed by field emission scanning electron microscope (FESEM, Sirion 200FEI). The q and S for the samples were measured simultaneously using a ZEM-3equipment (Ulvac Riko, Inc.) in helium atmosphere. Thermal diffusivity D and spe-cific capacity Cp were measured by a laser-flash method (Netzsch, LFA 457) and dif-ferential scanning calorimetry (DSC, Perkin–Elmer), respectively. Archimedesmethod was used to measure the density k of bulk samples and the total thermalconductivity j was calculated from the formula j = DCpk.

3. Results and discussion

The TG-DSC curves of the 5 at.% Gd-doped STO zero gel areshown in Fig. 1. In consideration of the zero weight loss above1223 K from TG curve, the (Sr0.95Gd0.05)TiO3 powders were ob-tained by calcining the zero gel at 1223 K in air for 6 h. Fig. 2ashows the XRD patterns of the (Sr0.95Gd0.05)TiO3 powders sinteredin air (a1), annealed powders (before SPS, a2) and ceramicsample (after SPS, a3), respectively. It can be seen that all of thediffraction peaks for the samples can be indexed cubic STO (JCPDSNo. 35-0734), and no detectable impurity peaks are observed inXRD patterns, which indicates that not only the powders obtainedat 1223 K, but also the samples sintered by SPS are pure phase ofSTO. The results show that the combination of the sol–gel methodand SPS can lower the sintering temperature so that fine grain sizeswith high densification can be obtained (for details see the textbelow), which is superior to the conventional solid-state reaction(SSR) method (T > 1733 K) to prepare the STO-based oxides[10–12].

The introduction of oxygen vacancies could significantly im-prove the electronic transport properties of the Gd-doped STOmaterials by increasing the carrier concentration. The lattice oxideions are removed by the reduction of the samples, which can be ex-pressed by the following equation:

O�O ! V ��O þ 2e0 þ 12

O2 ð1Þ

Meanwhile, a corresponding amount of Ti4+ is believed to bereduced to Ti3+ with larger ionic radius to keep the charge equilib-rium [19], which explains why the diffraction peaks of the reducedsamples shift slightly to the lower diffraction angles, as shown inFig. 2b

The room-temperature XRD patterns of the (Sr1�xGdx)TiO3�d

ceramics with different Gd contents x are shown in Fig. 3. It canbe seen that all the (Sr1�xGdx)TiO3�d samples have the same singlephase. However, the lattice constant a decreases as the Gd contentincreases due to the substitution of Gd3+ for Sr2+, as listed in Ta-ble 1, for the ionic radius of Gd3+ is slightly smaller than that ofSr2+ [20].The relative densities of all samples after SPS are 98.6%,99.0%, 99.1%, 98.7% and 97.8% for x = 0, 0.05, 0.08, 0.10, and 0.12,respectively, indicating that highly dense bulk samples are suc-cessfully fabricated.

FESEM micrographs of the thermal-etched surfaces for(Sr1�xGdx)TiO3�d samples are shown in Fig. 4. There are only afew pores can be detected in the surfaces for all doped samples,

Fig. 3. XRD patterns of (Sr1�xGdx)TiO3�d ceramics with x = 0, 0.08, 0.10, and 0.12.

Table 1List of lattice parameters, samples densities, relative densities, d and the estimatedpowers a for (Sr1�xGdx)TiO3�d (x = 0–0.12) ceramics.

x a (Å) Density (g/cm3) Relative density d a

0 3.9049 5.05 98.6% 0.0170.05 3.9030 5.17 99.0% 0.032 1.840.08 3.9018 5.23 99.1% 0.025 1.970.10 3.9010 5.26 98.7% 0.035 2.100.12 3.8995 5.24 97.8% 0.033 2.08

564 L. Li et al. / Journal of Alloys and Compounds 588 (2014) 562–567

while some larger pores appear in the non-doped sample (Fig. 4a).Obviously, dense microstructures are obtained for Gd doped-STOsamples, which are consistent with the high relative densities (Ta-ble 1). This almost full dense microstructure would be beneficial toelevating carrier mobility as well as power factor due to the lessen-ing of carrier scattering by the reduced pores [21]. As one can seefrom Fig. 4a that the grain size of non-doped SrTiO3-d is inhomoge-neous, and most of grain sizes range from 5 to 20 lm with anabnormal grain even larger than 20 lm. In comparison, homoge-neous fine-grained microstructures are achieved in Gd-dopedSrTiO3-d samples, as shown in Fig. 4b–e, which could attribute tothe refinement effect of rare earth addition [22]. Although the grainsize increases gradually with increasing in Gd-doping content, theaverage grain sizes (d � 0.66, 0.71, and 0.86 lm for x = 0.05, 0.08,and 0.10, respectively) are less than 1 lm for the samples withx 6 0.10. For the doped sample with x = 0.12, the grain size becomeinhomogeneous, and a few large grains d > �4–5 lm are observed;but, even in this sample there are a great number of grains lessthan 1 lm, resulting in average grain size of around 2 lm. As onecan expect, the fine-grained microstructure with a large volumefraction of grain boundaries would be beneficial to enhancing thephonon scattering to suppress the thermal conductivity.

The temperature dependence of the electrical resistivity (q) of(Sr1�xGdx)TiO3�d samples is shown in Fig. 5a. One can see thatthe q of non-doped sample SrTiO3�d increases monotonically withincreasing temperature in the whole temperature investigatedhere. In contrast, all the doped samples exhibit semiconductor-likebehavior (i.e. dq/dT < 0) at low temperatures (300–450 K) andmetallic behavior (i.e. dq/dT > 0) at T > �450 K. Similar behaviorwas also observed in the Gd-doped STO prepared by SSR [10]. Gen-erally, there are two main viewpoints to explain this transitionfrom semiconductor to metallic behavior: One thinks that it couldbe ascribed to the scattering of potential barriers at grain bound-aries [23]; the other believes this phenomenon could arise from atemperature-dependent polaron-type hopping process between

Ti3+ and Ti4+ via the bridging oxygen atoms [24]. However, thistransition behavior is only observed in our doped (polycrystalline)samples, which suggests that this phenomenon could be related tothe interplay or synergistic effects of impurity (dopant) with grainboundaries.

Moreover, it can be seen from Fig. 5a that with increasing Gdcontent, the q of STO decrease monotonically at T > �500 K. For in-stance, the q of (Sr1�xGdx)TiO3�d decreases from 47.8 � 10–3 to1.3 � 10–3 X cm at round 510 K as x increases from 0 to 0.12, indi-cating that the resistivity can be decreased effectively by Gd dop-ing due to substitution of Gd3+ for Sr2+. It should be pointed outhere that oxygen vacancy is an important factor that can influencetransport properties of STO. For example, the SrTiO3 sample withno oxygen vacancy is an insulator [25], while oxygen deficientSrTiO3-d is conducting. Specially, the resistivity of SrTiO3�d withd = 0.017 show metallic behavior and ranges from �24 to120 � 10–3 X cm over the measuring temperature range, as shownin Fig. 5a. It is known that oxygen vacancies act as a kind of donor,as expressed in formula (1) that indicates that a lack of O2– canyield two free electrons. According to the one-band model [26],the estimated carrier concentration from oxygen vacancies byusing the measured vales of d (0.013–0.035) can reach up to�5–12 � 1020 cm–3, which indicates that a degenerate state canoccur even in non-doped SrTiO3�d sample [27], as observed inFig. 5a. In addition, the dependences of q on temperature atT > �750 K have a similar trend for all doped samples and can beexpressed as q / Ta. By best fitting the experimental data to thisrelation, one obtains values of power a for the samples with differ-ent Gd contents, as listed in Table 1. One can see that a varies inthe range 1.8 to 2.1, implying that the dominant transport mecha-nism comes from acoustic phonon scattering at high temperature.

The Seebeck coefficients (S) as a function of temperature and Gdcontent are shown in Fig. 5b. In the whole temperature range, thenegative values of S confirm that the major charge carriers in allthe samples are electrons. The increase in q and |S| with tempera-ture for the SrTiO3�d shows the metallic like conductivity. At thetemperatures T < 800 K, the |S| of the samples with x > 0.05 in-creases approximate linearly with increasing temperature (exceptfor the sample with x = 0.05, for which the corresponding temper-ature is around 600 K), showing a degenerate semiconductorbehaviour, in which Seebeck coefficients can be expressed as (Mottformula) [28]:

S ¼ p2k2BT

3e@ lnrðEÞ

@E

� �����EF

¼ p2k2BT

3negðeÞ þ p2k2

BT3e

@ ln lðEÞ@E

� �����EF

ð2Þ

with r = 1/q = nel, where kB, e, g(e), l(E), r, EF and n are Boltzmannconstant, electronic charge, density of states, mobility, electricalconductivity, Fermi level and carrier concentration, respectively.One can see from Fig. 5b that as x increases from 0.05 to 0.12, theS decreases monotonically from 149 to 77 lV K–1 at room tempera-ture and from 278 to 196 lV K–1 at about 1000 K, respectively.Obviously, this decrease of S with increasing Gd content shouldbe mainly ascribed to the increase in carrier concentration n,which can be deduced from Eq. (2). This is consistent with thedecreased electrical resistivity as x increases, as discussed above.One notices that at much higher temperatures (at T > �600 K forx = 0.05 and at T > �800 K for x > 0.05), the temperature depen-dence of S deviates from linear relation. A similar behavior wasreported in La-doped epitaxial STO thin films [29], and highlyreduced n-type STO [27]. Lee et al. [27] explained that suchanomalies in temperature behaviour of S could be due to the heter-ogeneous oxygen non-stoichiometry. However, the oxygen-vacancyimpurity band and a narrow conduction band around the Fermienergy also influence the S at high temperatures [11].

Fig. 4. FESEM images of the thermal-etched surfaces for (Sr1�xGdx)TiO3�d ceramics for (a) x = 0, (b) x = 0.05, (c) x = 0.08, (d) x = 0.10 and (e) x = 0.12.

L. Li et al. / Journal of Alloys and Compounds 588 (2014) 562–567 565

The calculated power factors (PF) are shown in Fig. 5c. In thecase of SrTiO3�d, the PF decreases with temperature due to theincreased q, however, the PFs of all doped samples firstlyincrease with increasing temperature until they reach a maximumat a certain critical temperature, and then decrease with furtherincreasing temperature. The peak value of PF increases firstwith x from 0.05 to 0.10, and then it decreases with furtherincrease in x, and the PF for (Sr1�xGdx)TiO3�d with x > 0.05 is largerthan 10 lW cm�1 K�2 in the high temperature range from 500to 1000 K. A maximum PF = 16 lW cm�1 K�2 is achieved in(Sr0.9Gd0.1)TiO3�d at �570 K, which is larger than that previouslyreported for (La0.1�xDy0.1+xSr0.8)TiO3 (13.2 lW cm�1 K�2 at�600 K) [13], and (Sr1�xPrx)TiO3 (�14 lW cm–1 K–2 at 670 K) syn-thesized by SSR [14]. Clearly, in addition to the optimized carrierconcentration arising from Gd substitution for Sr, the high PF valuecan be ascribed to the dense samples with appropriate mean grainsizes, which provides a passage for carriers with high mobility,such that the electrical conductivities remain comparatively high.

Fig. 6a shows the temperature dependence of the total thermalconductivity (jtotal) for (Sr1�xGdx)TiO3�d ceramic samples. jtotal forall samples decreases with temperature. In general, jtotal includesthe lattice thermal conductivity jL and the electronic thermal con-ductivity je, where je can be estimated by Wiedemann–Franz law:je = L0T/q, with Lorentz number L0 = 2.44 � 10�8 V2 K�2. AtT > �500 K, the addition of Gd leads to increased je due to the ra-pid decrease in q (inset of Fig. 6b), however, the je values are smallas compared to jtotal. The SrTiO3�d sample shows the largest jtotal

in the whole measuring temperature. However, significant reduc-

tion in jtotal is observed in doped samples (Fig. 6a), indicating thatthe decrease of jtotal mainly comes from their reduction in jL,which dominates the jtotal. The reduction in jL as x increase from0 to 0.05 should be due to the significant decrease of grain sizesand/or the larger difference between Gd and Sr ion of both the io-nic radius and the mass [18]. jL decreases obviously with increas-ing x value (except for the heavily doped sample with x = 0.12 atT < �400 K), although the larger grain sizes are observed in thesamples with x = 0.12, which suggests that doping with Gd onthe Sr site enable to increase the phonon scattering by the massdifference. As a result, the jL reduces from 2.7 Wm�1 K�1 forx = 0.05–2.13 Wm�1 K�1 for x = 0.12 at �1000 K and the lowestjtotal, 2.8 Wm�1 K�1 at 1023 K, was obtained in (Sr0.88Gd0.12)TiO3�d

sample.It should be pointed out that the thermal conductivity of our

samples are larger than that of (Sr0.9Gd0.1)TiO3 prepared by SSR[10], which can be attributed to the few pores in our samples (rel-ative density P97.8% in our samples, while relative density is83.6% for (Sr0.9Gd0.1)TiO3 prepared by SSR). According to effectivemedium theory [21], effective thermal conductivity jeff is relatedto the pore volume fraction U, as jeff = jh(2 � 2U)/(2 + U), wherejh is the thermal conductivity of the host (without pores). This for-mula shows that the thermal conductivity of a material is signifi-cantly affected by the pores, i.e., the smaller the density k, thelower the thermal conductivity j. Grain boundaries can scatterphonons and the thermal conductivity can be reduced due to theincrease in the number of grain boundaries (or decrease in grainsize). The effects of grain size d (or volume fraction of interfaces)

Fig. 5. Temperature dependences of (a) electrical resistivity, (b) Seebeck coefficient,and (c) power factor for (Sr1�xGdx)TiO3�d ceramics.

Fig. 6. Temperature dependences of the (a) total thermal conductivity, (b) latticethermal conductivity and electronic thermal conductivity (inset) for (Sr1�xGdx)-TiO3�d ceramics.

Fig. 7. Temperature dependence of the figure of merit for (Sr1�xGdx)TiO3�d

ceramics. Data for (Sr0.9Gd0.1)TiO3, (Sr0.9Pr0.1)TiO3, (Sr0.92La0.08)TiO3, andSr(Ti0.8Nb0.2)O3 from Refs. [10,14,16,32] are taken for comparison.

566 L. Li et al. / Journal of Alloys and Compounds 588 (2014) 562–567

on the jL of non-doped STO have been investigated experimentallyby Wang et al. [30] and they found that as the grain size decreasefrom 20 lm to 50 nm, the jL decreases from 8 to 5.3 Wm�1 K�1 atroom temperature. Moreover, Yang et al. [31] theoretically investi-gated the effects of grain size on lattice thermal conductivity innanocrystalline material and found that the effective value jL canbe estimated as:

jL ¼j0

1þ j0Rkd

ð3Þ

where j0 is the thermal conductivity of bulk or single crystal and Rk

is the interfacial thermal resistance (4.98 � 10–9 m2 W K–1 at 300 Kand 3.41 � 10–9 m2 W K–1 at 1000 K for the nanogained STO [30]).For the coarse-grained sample, jL � j0. However, the jL for a nano-grained sample (especially the grain sizes are smaller than 100 nm)is reduced obviously. Formula (3) indicates that the smaller the

grain size d, the smaller the effective value jL for the nanograinedsamples. Hence, fine-gained microstructure is beneficial to loweringthermal conductivity.

L. Li et al. / Journal of Alloys and Compounds 588 (2014) 562–567 567

Fig. 7 shows the temperature dependence of the dimensionlessfigure of merit (ZT) for (Sr1�xGdx)TiO3�d ceramic samples. It can beseen that the ZT values increase steeply with increasing tempera-ture. The ZT values increase with increasing Gd content (exceptfor x = 0.12) specially at T > 500 K, and the largest ZT, 0.37 at1006 K, is achieved in (Sr0.9Gd0.1)TiO3�d, due to its larger PF(�10.7 lW cm�1 K�2 at 1006 K) and relatively smaller thermalconductivity. Clearly, the obtained ZT values of (Sr1�xGdx)TiO3�d

samples with x = 0.08, 0.10, and 0.12 are substantially larger thanthose reported for (Sr0.9Gd0.1)TiO3 [10], (Sr0.9Pr0.1)TiO3 [14], andSr(Ti0.8Nb0.2)O3 [32], prepared by SSR, at T > 500–600 K. Althoughthe ZT value of (Sr0.92La0.08)TiO3, prepared by SPS [16], is largerthan that of our samples at T < �800 K (as shown in Fig. 7) dueto its smaller electrical resistivity (arising from absence of thesemiconducting behavior). At T > 800 K, the ZT values of our(Sr1�xGdx)TiO3�d samples with x = 0.1 and 0.12 are obviously largerthan that of (Sr0.92La0.08)TiO3 (Ref. [16]) due to steeper increase ofZT with increasing temperature for our Gd-doped STO samples.Nevertheless, by comparing ZT of the samples with x = 0.1 andx = 0.12 one can see that further increase in Gd content cannot ele-vate ZT of (Sr1�xGdx)TiO3�d due to large decrease in S, though thesample has a smaller jtotal and a smaller q (as shown by the smaple(Sr0.88Gd0.12)TiO3�d), indicating that proper doping is very impor-tant to optimize the thermoelectric performance of STO. Hence,the appropriate combinations of fine-grained microstructure withhigh densities and proper doping (doping species and its content)as well as oxygen deficiency are essential to the effective enhance-ment of thermoelectric performance of STO-based ceramics. Fur-ther improvement of ZT of STO should be achieved by preparingdual- or even multiple-doped STO samples, for co-doping with dif-ferent rare earth ions could effectively reduce the j of STO [13].

4. Conclusions

(Sr1�xGdx)TiO3�d (x = 0.05–0.12) ceramics were prepared by acitrate sol–gel method with SPS, and their TE properties were mea-sured from room temperature to 1100 K. The results indicate thatthe |S| increases monotonically with increasing temperature;meanwhile, the q and |S| decrease (at T > 450 K for q) as the Gdcontent increases, which should be mainly attributed to the in-crease in electron concentration due to the substitution of Gd3+

for Sr2+. In the high temperature range from 400 K to 1000 K, thePF for (Sr0.9Gd0.1)TiO3�d is higher than 10 lW cm�1 K�2, and thelargest value of 16 lW cm–1 K–2 is obtained at 570 K. Moreover,jL reduces remarkably with increasing Gd content (e.g., it decreasefrom 2.7 W m–1 K–1 for x = 0.05–2.13 W m–1 K–1 for x = 0.12 at�1000 K). As a result, a largest ZT value of 0.37 at 1006 K isachieved in (Sr0.9Gd0.1)TiO3�d sample. Such an improvement in ZTsuggests that this ceramic have potential applications in the cas-cade-type TE modules for mid-to-high temperatures. Present re-sults reveal that appropriate combinations of high densities withboth fine-grained microstructures and proper doping (such as rare

earth Gd doping with x = 0.1) as well as oxygen deficiency areessential to the effective enhancement of thermoelectric perfor-mance of STO-based systems.

Acknowledgment

This work was financial supported by the National NaturalScience Foundation of China (50972146, 10904144, 11174292,11374306 and 51101150).

References

[1] M. Zebarjadi, B.L. Liao, K. Esfarjani, M. Dresselhaus, G. Chen, Adv. Mater. 25(2013) 1577–1582.

[2] H.Y. Li, H.Y. Jing, Y.D. Han, Y.C. Xu, G.Q. Lu, L.Y. Xu, J. Alloys Comp. 576 (2013)369–374.

[3] X.X. Li, J.Q. Li, F.S. Liu, W.Q. Ao, H.T. Li, L.C. Pan, J. Alloys Comp. 547 (2013) 86–90.

[4] K. Koumoto, R. Funahashi, E. Guilmeau, Y. Miyazaki, A. Weidenkaff, Y.F. Wang,C.L. Wan, J. Am. Ceram. Soc. 96 (2013) 1–23.

[5] S.F. Wang, Z.L. Bai, H.F. Wang, Q. Lu, J.L. Wang, G.S. Fu, J. Alloys Comp. 554(2013) 254–257.

[6] Y. Wang, Y. Sui, J.G. Cheng, X.J. Wang, W.H. Su, J. Alloys Comp. 477 (2009) 817–821.

[7] J. Li, J. Sui, Y. Pei, C. Barreteau, D. Berardan, N. Dragoe, W. Cai, J. He, L-D. Zhao,Energy Environ. Sci. 5 (2012) 8543–8547.

[8] L. Han, N.V. Nong, L.T. Hung, T. Holgate, N. Pryds, M. Ohtaki, S. Linderoth,J. Alloys Comp. 555 (2013) 291–296.

[9] S. Ohta, T. Nomura, H. Ohta, M. Hirano, H. Hosono, K. Koumoto, Appl. Phys. Lett.87 (2005) 092108.

[10] H. Muta, K. Kurosaki, S. Yamanaka, J. Alloys Comp. 350 (2003) 292–295.[11] J. Liu, C.L. Wang, W.B. Su, H.C. Wang, P. Zheng, J.C. Li, J.L. Zhang, L.M. Mei, Appl.

Phys. Lett. 95 (2009) 162110.[12] J. Liu, C.L. Wang, W.B. Su, H.C. Wang, J.C. Li, J.L. Zhang, L.M. Mei, J. Alloys Comp.

492 (2010) L54–L56.[13] H.C. Wang, C.L. Wang, W.B. Su, J.A. Liu, Y. Sun, H. Peng, L.A.M. Mei, J. Am.

Ceram. Soc. 94 (2011) 838–842.[14] A.V. Kovalevsky, A.A. Yaremchenko, S. Populoh, A. Weidenkaff, J.R. Frade,

J. Appl. Phys. 113 (2013) 053704.[15] Y. Wang, H.J. Fan, Scr. Mater. 65 (2011) 190–193.[16] A. Kikuchi, N. Okinaka, T. Akiyama, Scr. Mater. 63 (2010) 407–410.[17] P.P. Shang, B.P. Zhang, J.F. Li, N. Ma, Solid State Sci. 12 (2010) 1341–1346.[18] B. Abeles, Phys. Rev. 131 (1963) 1906–1911.[19] J.T.S. Irvine, P.R. Slater, P.A. Wright, Ionics 2 (1996) 213–216.[20] L. Fang, W. Dong, F.G. Zheng, M.R. Shen, J. Appl. Phys. 112 (2012) 034114.[21] H. Lee, D. Vashaee, D.Z. Wang, M.S. Dresselhaus, Z.F. Ren, G. Chen, J. Appl. Phys.

107 (2010) 094308.[22] D.A. Payne, H.U. Anderson, J. Am. Ceram. Soc. 50 (1967) 491–503.[23] R. Moos, K.H. Hardtl, J. Appl. Phys. 80 (1996) 393–400.[24] X. Li, H.L. Zhao, W. Shen, F. Gao, X.L. Huang, Y. Li, Z.M. Zhu, J. Power Sour. 166

(2007) 47–52.[25] M. Capizzi, A. Frova, Phys. Rev. Lett. 25 (1970) 1298–1302.[26] M. Onoda, I. Goto, J. Phys.: Condens. Matter 21 (2009) 435603.[27] S. Lee, G.Y. Yang, R.H.T. Wilke, S. Trolier-McKinstry, C.A. Randall, Phys. Rev. B

79 (2009) 134110.[28] B. Fisher, L. Patlagan, G.M. Reisner, A. Knizhnik, Phys. Rev. B 61 (2000) 470–

475.[29] S.R.S. Kumar, A.I. Abutaha, M.N. Hedhili, H.N. Alshareef, J. Appl. Phys. 112

(2012) 114104.[30] F. Wang, K. Fujinami, R.Z. Zhang, C.L. Wan, N. Wang, Y.S. Ba, Appl. Phys. Express

3 (2010) 031101.[31] H.S. Yang, G.R. Bai, L.J. Thompson, J.A. Eastman, Acta Mater. 50 (2002) 2309–

2317.[32] S. Ohta, H. Ohta, K. Koumoto, J. Ceram. Soc. Jpn. 114 (2006) 102–105.