Low thermal conductivity of Al-doped ZnO with layered and correlated grains

Abstract

Bulk Al-doped zinc oxide (ZnO) with a novel self-assembled layered and correlated grain structure is found to exhibit sharp reduction in thermal conductivity. The microstructure consists of a two-dimensional layered network of oriented grains, which interconnect in the third dimension through inter-planar contacts, and grains are embedded with nano-precipitates. The contact represents anisotropic connectivity of voids trapped between the grain layers. The effects of the synthesis atmosphere and Al doping concentration upon the formation of porous correlated grains are explained by taking into account the contributions of the vapor transport mechanism for grain growth under vacuum. The inhomogeneous density distribution with spring-back effect due to the uniaxial compaction leads to the anisotropic grain growth. Compared with traditional dense ZnO, Al-doped ZnO with layered microstructure exhibits a 52% decrease in the thermal conductivity across layers (3.0 W m⁻¹ K⁻¹ at 573 K) while maintaining the magnitude of electrical conductivity (1000 S cm⁻¹). The resultant power factor 4.78 × 10⁻⁴ W m⁻¹ K⁻² at 423 K and figure of merit of 0.14 × 10⁻³ K⁻¹ at 572 K is higher in comparison to the normal grain structured material.

---

Introduction

Thermoelectric materials for energy harvesting applications require high conversion efficiency and stability at elevated temperatures. The conversion efficiency is dependent upon the ratio of electrical to thermal conductance in the thermoelectric elements. Thus, there has been focus on reducing the lattice thermal conductivity (κ_lattice) for improving the performance of thermoelectric materials.^1,2 Doped ZnO is a promising n-type thermoelectric material with acceptable Seebeck coefficient.³ It is cheap, non-toxic, and utilizes abundantly available elements, Zn and O. Further, bulk ZnO has excellent high temperature durability in air up to 1000 °C. However, the simple wurtzite crystal structure and light component elements result in its high thermal conductivity of the order of 100 W m⁻¹ K⁻¹ at room temperature.⁴ Hence reduction of thermal conductivity (κ) has been the primary focus towards improving the thermoelectric performance of ZnO.

Significant improvement in reducing the thermal conductivity of thermoelectric alloys has been achieved by synthesizing nanostructured materials.^1,2 When the scale of microstructural features becomes closer to the phonon mean free path then phonon scattering can be enhanced leading to a sizeable reduction in κ_lattice. The center for phonon scattering can be point defects (especially for heavy elements), grain boundaries, nano-scale inclusions and pores/voids.² A recent study has indicated that phonons with long mean free path could be scattered by developing mesoscale architectures.⁵ Both nano- and micro-scale inclusions coupled with grain boundaries have been used for promoting the phonon scattering in bulk ZnO.⁶ Nano-inclusions have been found to have the pronounced effect towards reducing the thermal conductivity.^6,7 Another method proposed in literature is based on the combustible nano-sized polymer particles as a void forming agent (VFA). Nano-sized voids introduced in the ZnO matrix by using combustible nanoparticles have been found to reduce the thermal conductivity by 16–25%;⁸ however, those samples displayed severe deterioration of electrical property. Similar degradation of electrical conductivity by void forming had been noticed in SiGe and Si systems as well.^9,10 A porous networked structure provides the option for lowering κ if degradation of the electrical conductivity can be avoided.

In this study, we provide the thermoelectric characteristics of ZnO ceramics with self-assembly layered and correlated grains. The matrix was embedded with nanoscale precipitates. The interspace (long pores) between the layered grain network exhibited ordered orientation in 2D cross-section. This microstructure was found to reduce the thermal conductivity by 52% as compared to that of conventional ZnO with nano-precipitates. Unlike the porous ZnO synthesized by utilizing forming agent, SiGe, and Si, the ZnO in this work did not exhibit electrical deterioration. A characterization was conducted to investigate the thermoelectric performance of ZnO and to identify the parameters controlling the formation of microstructure with layered and correlated grain structure.

Experimental procedure

Doped ZnO powders were synthesized using zinc nitrate hydrate [Zn(NO₃)₂·6H₂O] and aluminum nitrate hydrate [Al(NO₃)₃·9H₂O] (1–3 at%) with oxalic acid [(COOH)₂·2H₂O] as a precursor and ethanol as a solvent. The mixture, viscous white gel, was stored for 24 h for uniformity and then dried at 80 °C for 12 h. All the dry gel products were subsequently calcined at 600 °C for 2 h with 10 °C min⁻¹ heating and cooling rate under nitrogen flow. The powder size distribution was measured using Particle size analyzer (Horiba Partica LA-950 Laser Diffraction). Doped powders were pelletized using uniaxial press into pellets with 15 mm diameter and 1.5 mm thickness. Pellets were sintered at 1200 °C for 5 h in sealed quartz tubes under the following vacuum conditions: 10⁻¹, 10⁻², 10⁻³, and 10⁻⁵ Torr. For studying microstructure formation, pellets were also sintered at 800–1100 °C, under 10⁻⁵ Torr, air and nitrogen. The phases and microstructures of the starting powders and sintered samples were examined by X-ray diffractometer (XRD, PANalytical X'Pert, CuKα; Philips, Almelo, The Netherlands) and scanning electron microscopy (SEM, LEO (Zeiss) 1550 field-emission). The Oxford INCA Energy E2H X-ray energy dispersive spectrometer (EDS) system with silicon-drifted detector was utilized for local elemental analysis. The thermal conductivity was determined from thermal diffusivity measurement using Xenon flash technique in a custom measurement system (Xenon Flash Bulb System, EG&G Electro-optics Model LS-187). The electrical resistivity was measured by the Van der Pauw method (Stanford Research Systems SR830 Lock in Amplifier and Keithley 2000 Multimeter) and by voltage–current method using Precision Premier II tester (Radiant Technologies) for sample with high resistivity. Seebeck coefficient was determined by voltage difference per temperature difference in an argon environment.

Results and discussion

Formation of layered-and-correlated-grain ZnO

The morphology of as-synthesized Al-doped ZnO powders is shown in Fig. 1a–c, which is distinct from that of undoped ZnO powders (long bar shaped with dimensions on the order of micrometer scale, Fig. S1a†). The size distribution of the synthesized powders is in the narrow range of 82–92 nm (Fig. S1b†). Based on the XRD results (Fig. 1d), all the synthesized powders exhibit only hexagonal wurtzite-type ZnO phase, indicating that Al element stays in the lattice of ZnO nanoparticle. All bulk samples were sintered using ZnO–x%Al (x = 1, 2, 3) nano-powders as starting material.

Fig. 1 SEM micrographs of ZnO nanoparticles synthesized by sol–gel process (a) ZnO–1%Al, (b) ZnO–2%Al, (c) ZnO–3%Al; and (d) XRD patterns of synthesized ZnO–x%Al (x = 0, 1, 2, 3) nanoparticles.

Bulk ZnO samples (1200 °C 5 hours, under 10⁻⁵ Torr condition) exhibit varying porous structures consisting of 5–10 μm grains interspersed by nano-precipitates, as shown in Fig. 2a–c. The Al concentration coupled with vacuum condition is essential ingredient in the formation of the microstructure. The grains of ZnO–1%Al (Fig. 2a) exhibit random orientation but it has higher density in comparison to the samples with higher Al doping. The ZnO–2%Al ceramic shows a self-assembled correlated grain structure arranged in layers (Fig. 2b). The ZnO–3%Al ceramic also exhibits the layered and correlated grain structure with large amount of nano-precipitates on grain surface. Due to the oriented grain growth in 2% and 3% Al-doped samples, the cross-sectional long interspaces are found to distribute themselves parallel to the sample surface. The samples synthesized under the high vacuum condition (10⁻⁵ Torr) exhibit slight better-organized correlated grain structure (inter-connection of grains in 2D and inter-connection of layers in 3D) than that under relative low vacuum condition, such as 10⁻² Torr shown in Fig. S2.† XRD results (Fig. 2d) for bulk ZnO–x%Al (x = 1, 2, 3) indicate the formation of a hexagonal wurtzite type ZnO phase and a small fraction of the gahnite phase ZnAl₂O₄, referred as second phase precipitates.^7,11,12 As seen in zoomed-in XRD spectrum in Fig. 2d, when the Al doping increases so does the intensity of XRD peaks from the gahnite phase. This change indicates that with the increase of Al doping amount more ZnAl₂O₄ precipitates have formed due to the reaction of ZnO and Al, which is consistent with the SEM images. EDS analysis for ZnO–3%Al (10⁻⁵ Torr) confirms that the Al concentration in precipitates is higher than that in the ZnO grain (Fig. 2e).

Fig. 2 Cross-sectional SEM micrographs of ZnO bulk pellets after sintering under 10⁻⁵ Torr (a) ZnO–1%Al, (b) ZnO–2%Al, (c) ZnO–3%Al with the high-magnification inset image showing the ZnAl₂O₄ phase (the brighter particles are ZnAl₂O₄ precipitates). The grain layers are parallel with the pellet surface. (d) XRD from sintered ZnO–x%Al (x = 1, 2, 3%) bulk pellets sintering under 10⁻⁵ Torr with the inset of a zoomed-in vision of the data from the dotted box where black dots indicate peaks from ZnAl₂O₄. The XRD data are normalized. (e) EDS area scanning of ZnO–3%Al sintering under 10⁻⁵ Torr (top square is on grain surface and bottom square is on cross-section of grain).

Mechanism of structure development

Next, we provide the understanding behind the formation of the layered and correlated grains. To investigate the effect of the synthesis atmospheres, experiments were conducted under three different conditions: 10⁻⁵ Torr, air and nitrogen, on the composition ZnO–2%Al, which exhibits the optimum orientation of layers under vacuum sintering condition. To understand the grain growth in different atmospheres, varying synthesis temperatures (800 to 1100 °C) were carried out. According to the microstructure evolution tracing in Fig. 3, the sintering atmosphere plays a critical role in the formation of the correlated grain structure. Under 10⁻⁵ Torr condition, at 800 °C, the necking between nanosize grains starts to occur preferring horizontal direction (parallel to pellet surface). At 900 °C, grain growth occurs in conjunction with the precipitation. With increasing sintering temperature, the grains connect to each other into the layers resulting in the formation of the voids with well-defined interspacing. At 1100 °C, most layers consist of single layer of grains. With the same composition, pellets sintered in air and nitrogen do not exhibit correlated grain structure. Pore smoothing and shrinking are observed at 1000 °C in both air and N₂ conditions. At 1100 °C, most pores are isolated in the corner of the grain boundaries. By comparing the microstructure in N₂ and air, there is no measurable effect from the solely nitrogen (low oxygen partial pressure) in an inert atmosphere.

Fig. 3 SEM micrographs from a cross-section of ZnO–2%Al bulk pellets after sintering at 800, 900, 1000, 1100 °C under 10⁻⁵ Torr, air atmosphere, and nitrogen. Insets are higher magnification images. The surface of pellets is parallel with horizontal direction.

The average porosity values of ZnO–2%Al after sintering under vacuum 10⁻⁵ Torr and air at varying temperatures are displayed in Fig. 4a. With higher synthesis temperature, the samples sintered in vacuum show lower porosity first and the porosity remains around 25% above 1000 °C sintering temperature. The samples sintered in air show continuous densification with increase in sintering temperature. Fig. 4b shows that the grain size of both groups exhibits increasing trend with higher sintering temperature but the one in vacuum condition displays larger grain size expansion. According to the inset of Fig. 4b, the pore sizes of samples in air slightly increase and the shrinkage occurs above 1000 °C. The pore size of samples synthesized in vacuum, however, exhibited growth together with grain size. One of the mechanisms of grain growth at a relatively high temperature has been attributed to physical sintering which is controlled by the vaporization rate of ZnO.¹³ Above 900 °C, ZnO particles was observed on the inside surface of quartz tube by EDS (Fig. S3 and Table S1†), indicating that zinc oxide gas filled the tube during sintering. The vacuum condition in a sealed tube together with large nanoparticle surface area enhances the sublimation of ZnO whereby vapor transport plays a dominating role during the sintering. Increasing the sintering temperature enhances the vapor transport, which simultaneously reduces the driving force for densification due to its coarsening effect.¹⁴ Therefore, coarsening by the vapor transport mechanism increases the grain size accompanied by the pore growth as well, leading to the correlated grain structure with relatively high porosity. Interior of the layer with cluster of grains, another mechanism referred to as curvature-driven boundary migration,¹⁴ induces the densification in each layer at temperatures of 1100 °C. Curvature-driven boundary migration also promotes the motion of pores trapped between the layers.

Fig. 4 (a) Porosity of ZnO–2%Al bulk pellets after sintering in vacuum and air at varying synthesis temperatures. Quantitative microscopy was applied to determine the porosity. (b) Grain size of ZnO–2%Al sintered in vacuum and air at varying synthesis temperatures. Inset shows the pore dimension (the large variation of pore size of samples in vacuum is because of pore dimension anisotropy).

The peak intensity of second phase ZnAl₂O₄ increases with synthesis temperature based on the XRD results (Fig. S4†). ZnO vaporization results in Al enrichment on the surface increasing the formation of ZnAl₂O₄ second phase. Further, the second phase formed at higher temperature (1000 °C) in comparison to the samples synthesized in vacuum, which indicates that the vacuum condition helps in the formation of ZnAl₂O₄. We believe that the precipitates rich on the surface prevent grains belonging to upper and lower layers from forming boundaries and favour growth of interspace between layers as shown in Fig. S4c.† This explains why when the Al concentration is relatively low (1 at%), the ZnO develops microstructure close to a normal polycrystalline (Fig. 2a). Thus, the vacuum condition coupled with Al concentration leads to the formation of the correlated grain structure.

The remaining important question is why the network of grains tends to grow horizontally to develop layered structure. The second phase precipitates or pores considered in many of the prior studies have symmetric structure as the shape evolution is driven by minimization of the interface energy. The asymmetric precipitate structures have been found in Sb₂Te₃–PbTe system, and energetics analysis indicated that lamellar, ribbon-like, and needle-like precipitates are residual of the interfacial energy minimization criterion.²³ The layered and correlated grain structure in ZnO formation follows the energy minimization criterion as well. In our work, there is no texturing and no obvious anisotropic crystal growth, by the XRD in Fig. 2d. The horizontal direction of grain growth is probably due to the external driving forces. Thick pellets (15 mm diameter, 8 mm thickness) were pressed, diced into cubical shape, and placed in a quartz tube with different direction as shown in the schematic of Fig. 5a. According to the SEM images shown in Fig. 5a, the grain layer orientation correlates with the direction of the applied pressure during uniaxial pressing. The function of the residual stress during and after unloading might be the reason for a layered network. Different pressing pressures were applied to see the effect on the microstructure (Fig. 5b–e). Using low pressure (<10 MPa), grain growth is not confined in a specific direction clearly. Under higher pressures (100–300 MPa), the correlated grains exhibit layered structure formation. At higher pressures, the powder compact displays a stronger spring-back effect¹⁵ during unloading. Since the spring-back is much greater in the axial direction than in the radial direction,¹⁶ powder density in the radial direction is larger than that in the axial direction. On the other hand, the stress in the axial (pressing) direction is much larger than that in the radial direction. So due to Le Chatelier principle,¹⁷ the ZnO sublimation in axial direction is stronger. Therefore grain boundary tends to form and develop perpendicular to that of the press direction.

Fig. 5 (a) Schematic diagram of the axial pressing and SEM micrographs from a cross-section of ZnO–2%Al bulk pellets with 90 degree placement during sintering. The color gradient shows the compaction pressure gradient in the powder compact due to wall friction effects. All other samples in this study had a thickness/diameter ratio of 0.13, thus the pressure gradient is negligible. (b)–(e) SEM images from a cross-section of ZnO–2%Al ceramic pressed under different pressures. The surface of pellets is parallel with horizontal direction.

In short, vapor transport occurring through ZnO sublimation and increased Al solubility resulting through the depleted Zn concentration on particle surface, coupled with the external factors of vacuum sintering condition and uniaxial pressing, cause the formation of the layered and correlated grain structure.

Thermoelectric performance

In prior investigations, several control parameters that include sintering atmosphere, Al doping amount, sintering temperature, and pressing pressure have been explored to quantify their influence on the evolution of layered and correlated grain morphology. Sintering atmosphere in conjunction with pressure is an important factor towards tuning the ZnO microstructure and has governing effect on the formation of the layered-correlated grain structure. Its influence on thermoelectric property is discussed in detail in this section. ZnO–2%Al ceramics under 10⁻⁵ Torr are found to exhibit the microstructure where ZnO grains exhibit well-developed correlated grain structure: 2D network of grains forming layers which are connected in 3D through planar contacts. The thermal conductivity for ZnO–2%Al pellets sintered under different vacuum conditions is shown in Fig. 6a. The data clearly indicates that all the samples with correlated-grain structure exhibit lower κ than contrast sample. It has been reported that nanosize precipitates distributed uniformly in ZnO with 2 at% Al result in low κ of 7.5 and 5.7 W m⁻¹ K⁻¹ at room temperature and 300 °C respectively.⁶ We use this result as a reference point in this study. The contrast data is shown as the brown color plot in Fig. 6a and the microstructure of sample is shown in inset of Fig. 6a.⁶ The magnitude of thermal conductivity decreases slightly as the vacuum level increases. The lowest κ is achieved for the Al-doped ZnO ceramics synthesized under 10⁻⁵ Torr vacuum with magnitude of 5.2–3.0 W m⁻¹ K⁻¹ in the measurement temperature range of 323 K to 573 K. These magnitudes are 30% and 52% lower than the contrast sample at 323 K and 573 K respectively. This κ value at 323 K is 700% lower in comparison to ZnO–2%Al without correlated grain structure.³

Fig. 6 Temperature dependence of (a) thermal conductivity, (b) lattice thermal conductivity, and (c) electrical conductivity for ZnO–2%Al with layered and correlated grains. The contrast sample is the identically compacted and dense ZnO–2%Al sintered under air (insert image is from contrast sample⁶).

For conventional ZnO ceramics, more than 90% of the thermal conductivity is associated with lattice thermal conductivity (κ_lattice) where contributions arise from phonon transport and the rest is related to the electron motion. The κ_lattice is shown in Fig. 6b, by subtracting κ_electron (calculated by Wiedemann–Franz law) from total κ. For ZnO–2%Al (10⁻⁵ Torr), the electronic thermal conductivity (κ_electron) is calculated to be 0.86 to 1.14 W m⁻¹ K⁻¹ in the temperature range of 323–573 K. The phonon transport impediment in layered correlated grains with greatly reduces the proportion of κ_lattice in κ. The plot of the lattice thermal conductivity vs. 1/T shows linear relation, indicating that Umklapp processes dominate in the phonon scattering.¹⁸ Because of the low electrical conductivity and near zero electronic thermal conductivity for the contrast ZnO–2%Al sample, its thermal conductivity value can be assumed to be equal to the lattice thermal conductivity value. Thereby, κ_lattice of ZnO–2%Al with a layered correlated grain structure are 58% and 32% lower that of the contrast ZnO–2%Al sample at 323 K and 573 K. Since both samples in this study and the contrast sample contain nano-precipitates, we attribute the reduction in κ mainly to the layered and correlated grain microstructure.

Fig. 6c shows that ZnO–2%Al with correlated grain structure has electrical conductivity (σ) of 10³ S cm⁻¹ similar to that of thermoelectric alloys, which is greatly enhanced when compared to the ZnO–2%Al ceramics sintered in air. Sintering under vacuum condition creates large amount of intrinsic defects, which enhance the electron concentration. The samples sintered under different vacuum conditions exhibit similar magnitude of electrical conductivities with a slightly reducing trend with increasing vacuum level. The electrical resistivity of compared ZnO–2%Al is as high as 10⁵ Ω cm, due to the formation of ZnAl₂O₄ and Zn vacancy in the present of oxygen.^24,25 The Zn²⁺ vacancies compensate for the free electrons in ZnO.

The reduction of thermal conductivity for the Al-doped ZnO is ascribed to the correlated grain structure characterized by the presence of 2D network of grains forming layers. Due to the extremely low thermal conductivity of air (0.057 W m⁻¹ K⁻¹), interspacing between grain layers can act as a good thermal insulator. For solids with pores having dimensions exceeding the phonon mean free path of ZnO (about 30 nm), the effective thermal conductivity can be expressed as κ = κ₀Φ(p), where κ₀ is the thermal conductivity of the corresponding dense material, Φ is the factor determined by the porosity p. Several estimations and models for Φ have been proposed in literatures,^19–21 and the shape and orientation distribution of voids influences Φ resulting in anisotropic thermal transport properties.²⁰ The two dimensional models of pore structures are displayed in Fig. 7, and the correlated-grain structure can be simply represented by the schematic model of Fig. 7c. The long interspacing between the grain layers in ZnO is approximated by an ellipsoidal shaped void. According to an estimation performed by Braginsky²¹ by taking the aspect ratio of the ellipsoidal pore as 1 : 4, the factor of porosity Φ was found to be 1 − (1 + J)p, when there is no thermal transportation in pores. The magnitude of J equals 0.5 for thermal conductivity of pore structure shown in case (a) and (b) and 0.1 and 0.8 along the x-axis, y-axis in case (c) respectively. The effective thermal conductivity exhibits the largest reduction along the y-axis in structure shown in Fig. 7c, for which porosity factor Φ was calculated to be ∼0.45 with 30% porosity, based on the Braginsky's model.²¹ This magnitude is in good agreement with the experimental value for the ZnO with layered and correlated grains, with 58% reduction of κ_lattice. Thus, by controlling the interconnection between the grain layers, one can modulate the morphology of the void and thus optimize the magnitude of thermal conductivity.

Fig. 7 Schematic representation of the two-dimensional models of pore structures: (a) circular pores (b) elliptical pores of random orientation, and (c) elliptical pores of same orientation. The aspect ratio of the ellipsoidal pore is 1 : 4.

ZnO–2%Al synthesized under 10⁻⁵ Torr exhibited the lowest magnitude of thermal conductivity. For this sample, the Seebeck coefficient (α) was measured to be −58 to −72 μV K⁻¹ in the temperature range of 323 to 572 K (Fig. 8a). This results in the room-temperature Power Factor (P. F.) as high as 3.67 × 10⁻⁴ W m⁻¹ K⁻², which is about 20 times larger than that of Al-doped ZnO nanocomposites reported recently.⁷ With increasing measurement temperature, the power factor increases to 4.78 × 10⁻⁴ W m⁻¹ K⁻² at 423 K and then slightly decreases to 4.16 × 10⁻⁴ W m⁻¹ K⁻² at 572 K due to the faster σ decrease than α increase. We expect that the power factor could rise again at higher temperature because of the square dependence of α in P. F. and the increasing trend of α reported in literature.^3,7,11 The figure of merit (Z) of ZnO–2%Al (10⁻⁵ Torr) is shown in Fig. 8b as a function of temperature. The room temperature Z value 0.07 × 10⁻³ K⁻¹ is about 3 times larger than that reported by Tsubota.²² The Z value continuous increases with temperature and reaches 0.14 × 10⁻³ K⁻¹ at 572 K, which is also larger than the results reported in the literature.²²

Fig. 8 Temperature dependence of (a) Seebeck coefficient and calculated power factor (b) figure of merit Z for ZnO–2%Al sample synthesized under 10⁻⁵ Torr.

Conclusions

We have synthesized bulk nanostructured ZnO–x%Al ceramics by varying the doping concentration and sintering atmosphere. The self-assembly correlated-grain structure has formed when compacted samples are sintered in vacuum condition and Al additions are 2 at% and 3 at%. Vapor transport mechanism for grain growth dominates under vacuum leading to grain coarsening and therefore correlated grain structure. The vacuum condition also facilitates ZnO sublimation, promoting the formation of Al-rich secondary phase on grain surface, which favors development of correlated grain structure. The layered grain structure is related to density distribution with spring-back effect due to uniaxial pressing leading to anisotropic grain growth. Compared with the dense ZnO–2%Al with nano-precipitates, there are 30% and 52% decrease in the thermal conductivity (323 and 573 K) of ZnO–2%Al (10⁻⁵ Torr) across the grain layers. The κ_lattice has 58% reduction at 323 K, which has a good agreement with the reference model. The electrical conductivity is found to remain at 1000 S cm⁻¹, typical for a thermoelectric alloy. Thus, these correlated-grain structured ZnO ceramic provide enhanced thermoelectric performance, exhibiting 0.14 × 10⁻³ K⁻¹ figure of merit at 572 K.

Acknowledgements

Authors gratefully acknowledge the financial support provided by NSF/DOE Thermoelectrics Partnership and Center for Energy Harvesting Materials and Systems (CEHMS). Authors would like to thank Dr David Clark, Dr Kathy Lu, Dr Ashok Kumar, and Dr Yongke Yan for helpful discussions. Authors would also like to thank Grayson Doucette and Cary Hill for suggestions.

Notes and references

1. J. R. Sootsman, D. Y. Chung and M. G. Kanatzidis, Angew. Chem., Int. Ed., 2009, 48, 8616.
2. C. J. Vineis, A. Shakouri, A. Majumdar and M. G. Kanatzidis, Adv. Mater., 2010, 22, 3970.
3. M. Ohtaki, T. Tsubota, K. Eguchi and H. Arai, J. Appl. Phys., 1996, 79, 1816.
4. U. Özgür, Y. I. Alivov, C. Liu, A. Teke, M. A. Reshchikov, S. Doğan, V. Avrutin, S. J. Cho and H. Morkoç, J. Appl. Phys., 2005, 98, 041301.
5. K. Biswas, J. He, I. D. Blum, C. I. Wu, T. P. Hogan, D. N. Seidman, V. P. Dravid and M. G. Kanatzidis, Nature, 2012, 489, 414.
6. Y. Zhao, Y. K. Yan, A. Kumar, H. Wang, W. D. Porter and S. Priya, J. Appl. Phys., 2012, 112, 034313.
7. P. Jood, R. J. Mehta, Y. Zhang, G. Peleckis, X. Wang, R. W. Siegel, T. Borca-Tasciuc, S. X. Dou and G. Ramanath, Nano Lett., 2011, 11, 4337.
8. M. Ohtaki and R. Hayashi, ICT'06: XXV International Conference on Thermoelectrics, Proceedings, 2006, p. 276.
9. H. Lee, D. Vashaee, D. Z. Wang, M. S. Dresselhaus, Z. F. Ren and G. Chen, J. Appl. Phys., 2010, 107, 094308.
10. J. Boor, D. S. Kim, X. Ao, M. Becker, N. F. Hinsche, I. Mertig, P. Zahn and V. Schmidt, Appl. Phys. A: Mater. Sci. Process., 2012, 107, 789.
11. K. H. Kim, S. H. Shim, K. B. Shim, K. Niihara and J. Hojo, J. Am. Ceram. Soc., 2005, 88, 628.
12. M. Ohtaki, K. Araki and K. Yamamoto, J. Electron. Mater., 2009, 38, 1234.
13. L. F. Norris and G. Parravano, J. Am. Ceram. Soc., 1963, 46, 449.
14. M. N. Rahaman, Ceramic Processing, CRC Press, Taylor& Francis Group, 6000 Broken Sound Parkway NW, Suite 300, 2007.
15. S. J. Glass and K. G. Ewsuk, Compaction Science and Technology, MRS Bull., 1997, 22(12), 24.
16. K. G. Ewsuk, J. G. Arguello, D. N. Bencoe, D. T. Ellerby, S. J. Glass, D. H. Zeuch and J. Anderson, in Improved Ceramics through New Measurements, Processing, and Standards, The American Ceramic Society, 735 Ceramic Place, Westerville, Ohio 43081, United States, 2002.
17. P. Joos and G. Serrien, J. Colloid Interface Sci., 1991, 145, 291.
18. R. Peierls, Bird of Passage: Recollections of a Physicist, Princeton University Press, 1988.
19. P. G. Klemens, High Temp. - High Pressures, 1991, 23, 241.
20. B. Shafiro and M. Kachanov, J. Appl. Phys., 2000, 87, 8561.
21. L. Braginsky, V. Shklover, G. Witz and H. P. Bossmann, Phys. Rev. B, 2007, 75, 094301.
22. T. Tsubota, M. Ohtaki, K. Eguchi and H. Arai, J. Mater. Chem., 1997, 7, 85.
23. Y. Y. Liu, L. Q. Chen and J. Y. Li, Acta Mater., 2014, 65, 308.
24. D. Bérardan, C. Byl and N. Dragoe, J. Am. Ceram. Soc., 2010, 93, 2352.
25. Y. Zhao, A. Kumar, G. A. Khodaparast, A. Eltahir, H. Wang and S. Priya, Energy Harvesting and Systems, 2014 DOI:10.1515/ehs-2013–0024.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4ra01220h

---