ZHANG Chao-Hua WANG Yi-Bo LI Jun-Qin LIU Fu-Sheng

(College of Materials Science and Engineering, Shenzhen Key Laboratory of Special Functional Materials, Shenzhen Engineering Laboratory for Advanced Technology of Ceramics, Guangdong Research Center for Interfacial Engineering of Functional Materials, Institute of Deep Underground Sciences and Green Energy, Shenzhen University, Shenzhen 518060, China)

ABSTRACT The widespread applications of thermoelectric (TE) materials in power generation and solid-state cooling require improving their TE figure of merit (ZT) significantly. Recently, GeTe-based alloys have shown great promise as mid-temperature TE materials with superhigh TE performance, mostly due to their relatively high-degeneracy band structures and low lattice thermal conductivity. In this perspective, we review the most recent progress of the GeTe-based TE alloys from the view of phase and defect engineering. These two strategies are the most widely-used and efficient approaches in GeTe-based alloys to optimize the transport properties of electrons and phonons for high ZT. The phase transition from rhombohedral to cubic structure is believed to improve the band convergence of GeTe-based alloys for higher electrical performance. Typical defects in GeTe-based alloys include the point defects from Ge vacancies and substitutional dopants, linear and planar defects from Ge vacancies. The defect engineering of GeTe-based alloys is important not only for optimizing the carrier density but also for tuning the band structure and phonon-scattering processes. The summarized strategies in this review can also be used as a reference for guiding the further development of GeTe-based alloys and also other TE materials.

Keywords: GeTe, thermoelectric, phase, defect, Ge vacancies; DOI: 10.14102/j.cnki.0254-5861.2011-2850

1 INTRODUCTION

Enabling direct heat-to-electricity energy conversion, thermoelectric (TE) materials can meet the global demand for clean energy[1-4]. The performance of a TE material is usually judged by the dimensionless figure of merit[1],ZT=S2σT/κ=S2σT/(+) =T(PF/κ), whereT,S,σ,κ,PF,andare absolute thermodynamic temperature, Seebeck density (n) is the most important interconnecting parameters amongS,σand, and thereforenusually needs to be firstly optimized to get compromisedS(n),σ(n) and(n) for achieving highZT[2]. Theis generallyn-inde- pendent that is less connected withS,σand, which makes the reduction ofas one of the most efficient strategies for obtaining highZT.However, reducingby intro- ducing various phonon scattering processes can also depress the carrier mobility (μ), which therefore suppresses the value ofσaccording to the equation:σ=neμ. Therefore, the ratio ofμ/is also an important criterion for judging the coefficient, electrical conductivity, thermal conductivity, power factor, electronic thermal conductivity and lattice thermal conductivity, respectively. Therefore, high-perfor- mance TE materials require highSandσbut lowκto get highZTvalue. However, these TE parameters are adverse interdependence[1,5], hindering the achievement of high enoughZTfor the widespread use of TE materials. Carrier efficiency of different strategies for highZT.

Besides the carrier-density optimization, various band-en- gineering strategies have also been widely proved to further elevate thePF[6,7], such as band convergence, band alignment, resonant distortion of density of states (DOS), minority carrier blocking and quantum confinement. On the other hand, various strategies of phonon engineering have been adopted to reduce the[8-10], such as enhancing the intrinsic phonon-phonon scattering from complex unit cell, weak chemical bonds and strong lattice anharmonicity, and introducing extra phonon scattering from different defects like point defects, linear defects and planar defects. Having highPFand lowthen can result in highZT. Note that TE devices have to work between the hot and cold ends, and thus TE materials with high-averageZTover the entire working temperature range are more required in practical TE applications compared with those with a highZTspike[1]. With continuous efforts from band and phonon engineering, many progresses have been achieved in lots of TE materials with different working temperature ranges, such as V2VI3compounds (mostlyp-type Bi2-xSbxTe3and n-type Bi2Te3-xSex)[11-13], IV-VI compounds (like PbTe[14], GeTe[15,16], SnTe[17]and SnSe[18]based alloys), SiGe alloys[19], half- Heusler alloys (likep-type NbFeSb and n-type ZrNiSn)[20], complex oxides[21], filled skutterudites (like CoSb3-based alloys)[22]and organic conducting polymers[23].

Here, we review the most recent progresses in the GeTe-based TE materials from the view of phase and defect engineering. We first summarize the crystal structures of GeTe and their phase transition between R-GeTe and C-GeTe. The approaches to tune the crystal structure and phase- transition behavior are summarized, and the corresponding impact of phase engineering on band structures is also discussed. The defects in the GeTe-based alloys generally include the intrinsic defects of Ge vacancies and extrinsic defects from substitutional doping or alloying with foreign atoms, whose impacts on thermal and electrical transport are fully discussed. Moreover, the TE properties of GeTe-based alloys prepared by different doping and alloying strategies are compared with each other for discussion. Finally, conclusions and outlook are provided as a direction for the further enhancement of TE performance of GeTe-based alloys as well as other TE materials.

2 CRYSTAL STRUCTURES AND PHASE ENGINEERING

Fig. 1. Crystal structure of (a) rhombohedral GeTe (R-GeTe) in (2 × 2 × 2) supercell, and (b) cubic GeTe (C-GeTe) in (1 × 1 × 2) supercell. The primitive cell of (c) R-GeTe and (d) C-GeTe

As shown in Fig. 2a, the band structures of GeTe show great dependence on the phase structure[31]. In the high- symmetry cubic phase, the C-GeTe has two dominant valence bands for electronic transport[67], which are the light band ofLand heavy band of ∑ with degenerate band valleys of 4 and 12, respectively. By decreasing the symmetry from cubic phase to rhombohedral phase, the 4Lpockets can generally split into 3L+ 1Z and also 12∑ pockets into 6∑ + 6η[31]. J. Li et al. demonstrated that slight symmetry reduction of C-GeTe could also result in high valley degeneracy for higher TE performance[31]. Generally, the high-symmetry cubic structure is believed to be more favorable for achieving high degenerate band valleys[15], which is the key for obtaining highZT. Therefore, many efforts have been put into converting the R-GeTe to C-GeTe by various doping and alloying approaches. Z. Zhenget al.have shown that the GeTe can be gradually changed from rhombohedral to cubic phase by Mn doping[54], as shown in Fig. 2b～2c. The phase-transition temperature can be reduced to 338 K that is nearly close to room temperature (Fig. 2c)[54]. Z. Liu et al. have also observed similar phenomenon of phase transition by co-doping of Bi and Mn elements (Fig. 2d)[47], which displays that the phase-transition temperature can be suppressed below 300 K. However, the Mn doping can also greatly suppress the carrier mobility, which somehow hinders the enhancement ofZTvalues[47,54]. Y. Gelbsteinet al.have done systemic work on the TE study of GeTe-Bi2Te3based alloys[26,33,44], which also revealed the rhombohedral-to-cubic phase transition of GeTe-based alloys by increasing the amount of Bi2Te3[44]. As shown in Fig. 2e～2f, our work on (Ge0.87Pb0.13Te)1-x(Bi2Te3)xalso clearly demonstrated the phase transition[35]. Generally speaking, most doping and alloying approaches can more or less shift the phase transition point of GeTe-based alloys from rhombohedral to cubic phase, such as doping or alloying with Pb[29], Sb[68], Bi[45], AgSbSe2[63]and AgBiSe2[69]. Therefore, the phase engineering of GeTe is significant for tuning the TE performance and has been widely discussed in most studies of GeTe-based TE alloys.

Fig. 2. Phase engineering of GeTe. (a) Evolution of crystal structures, Fermi surface, and the dominant transporting valence band of GeTe. (b) XRD patterns of Ge1-xMnxTe. (c) Phase-transition temperature as a function of Mn content in Ge1-xMnxTe. (d) Lattice parameter and interaxial angle dependence on Mn content in Ge0.96-xMnxBi0.04Te. (e) XRD and (f) lattice parameter and interaxial angle of (Ge0.87Pb0.13Te)1-x(Bi2Te3)x. Figures are reprinted from ref.[31] for (a), ref.[54] for (b and c), ref. [47] for (d) and ref. [35] for (e～f)

3 GE-VACANCY ENGINEERING

Due to the high hole-carrier concentration induced by the frequently observed Ge vacancies, many studies of GeTe- based alloys have taken Ge vacancies as negative effects for improving theZT[29,37,40,47,57,66,71]and have also demonstrated various strategies for suppressing the high-density Ge vacancies. J. F. Donget al. reported that adding excessive Ge in pristine GeTe could effectively reduce the Ge vacancies, which results in the reduction of carrier density and increase of carrier mobility, thus obtaining a peakZTof 1.6 at ～650 K for pristine GeTe[40]. J. Liet al. reported that alloying with PbSe could increase the formation energy of Ge vacancies by increasing the size of cations and decreasing the size of anions, leading to suppressed Ge vacancies[37]. K. S. Bayikadiet al.demonstrated that the density of Ge vacancies can also be suppressed by proper heat treatment, resulting in a peakZTof ～1.37 in pristine GeTe[71].

Fig. 3. Ge-vacancy engineering of GeTe. (a) Schematic supercell structure of the C-GeTe with a Ge vacancy. (b) Schematic change of band structures by increasing the amounts of Ge vacancies. (c) Calculated Ge-vacancy dependent band energy depicted in (b). (d) Schematic planar Ge vacancies within grains. (e)～(f) High-resolution TEM images demonstrating planar Ge vacancies. (g) High-density arrays of Ge vacancies inside submicron ferroelectric domains. (h) Schematic of the formation of Ge vacancies in GeTe-based materials by alloying with layered Bi2Te3. (i) Gradually reduction of lattice thermal conductivity by planar Ge vacancies. Figures are reprinted from ref.[70] for (a～c), ref. [49]for (d～f), ref.[43] for (g), and ref.[35] for (h～i)

4 DOPING AND ALLOYING

Table 1. Reported Peak ZTmax, Average ZTav, Seebeck Coefficient S, Electrical Conductivity σ, and Lattice Thermal Conductivity κlat for Typical GeTe-based Thermoelectric Materials

Fig. 4. Recent progress of GeTe-based TE alloys. (a) Temperature-dependent figure of merit (ZT) and (b) average ZT (ZTav) over the entire testing temperature range. The corresponding specific compositions and references of the data with number 1～20 are listed in Table 1

5 SUMMARY AND OUTLOOK

Approaches of Ge-vacancy engineering and defect engineering have been widely proved to be effective for tuning the carrier density, band structures and phonon scattering process, which therefore enhance the TE performance of GeTe-based alloys. By doping or alloying with different elements or compounds, many GeTe-based alloys have achieved peakZTbeyond 2.0 and averageZT＞ 1.2, as shown in Fig. 4 and Table 1. Note that the measuring errors ofZTgenerally can be higher than 10% and even up to 20%[79], which should be taken critically when researchers follow up those studies shown in Table 1. Those highZTachieved in GeTe-based alloys are mostly explained by the mechanisms of optimization of carrier density, rhombohe- dral-to-cubic phase transition, high band degeneracy, band convergence, resonant bonding, enhanced intrinsic phonon- phonon scattering, and phonon scattering from various defects, especially the point defects and planar vacancies. Those developed mechanisms for enhancingZTcan also pave a way to tune the TE properties of other TE materials.

Although theZTvalues of GeTe-based alloys are much higher than many other TE materials, further researches on the GeTe-based materials may still need to be proceeded in the following directions:

(1) Developing compatible n-type GeTe-based alloys for making full GeTe-based devices. Recently, M. Samantaet.al. have realized n-type GeTe-based alloys by alloying with AgBiSe2[69]. Although the maximumZTcan only reach 0.6 at 500 K for the n-type (GeTe)50(AgBiSe2)50, this work proves the possibility of achieving n-type GeTe-based alloys.

(2) Enhancing the low-temperature TE performance of GeTe-based alloys. High-averageZTis more practical for TE applications. Besides further developing new mechanism to boost theZTover 3, enhancing the low-temperatureZTof GeTe-based alloys beyond 1 is also more challengeable. Though we have observed the low-temperatureZT～1 in the quenched GeTe-based alloys[16], the thermal stability is poor and still needs to be improved.

(3) Developing GeTe-based devices. The great promise of GeTe-based alloys for power generation needs to be fulfilled by developing TE devices[68]. Although many progresses have been made in the GeTe-based alloys, the development of GeTe-based devices are quite limited. As an example, we have developed an Al-Si alloy as a suitable diffusion barrier for GeTe-based devices with high interfacial performance[80]. Many scientific problems related to GeTe-based devices are still rarely studied, such as the mechanical strength of devices, contact resistivity, aging stability and device building.