Self-Nano-Structuring in SrTiO 3 : A Novel Strategy for Enhancement of Thermoelectric Response in Oxides

: Nano-structuring is recognised as an efficient route for enhancing thermoelectric response. Here we report a new synthesis strategy for nanostructuring oxide ceramics and demonstrate its effectiveness on an important n-type thermoelectric SrTiO 3 . Ceramics of Sr 0.9 La 0.1 TiO 3 with additions of B 2 O 3 were synthesized by the mixed oxide route. Samples were sintered in air followed by annealing in a reducing atmosphere. Crystallographic data from X-ray and electron diffraction showed Pm (cid:885)(cid:3364) m cubic symmetry for all the samples. High resolution transmission electron microscopy (HRTEM) showed the formation of a core-shell type structure within the grains for the annealed ceramics. The cores contain nanosize features comprising pairs of nano-size voids and particles; the feature sizes depend on annealing time. Atomic-resolution, high-angle annular-dark-field imaging and electron energy loss spectroscopy in the scanning transmission electron microscopy (STEM-HAADF-EELS) showed the particles to be rich in Ti and the areas around the voids to contain high concentrations of Ti 3+ . Additionally, dislocations were observed, with significantly higher densities in the shell areas. The observed dislocations are combined (100) and (110) edge dislocations. The major impact of the core-shell type microstructures, with nano-size inclusions, is the reduction of the thermal conductivity. Sr 0.9 La 0.1 TiO 3 ceramics containing grain boundary shells of size  1 µm and inclusions in the core of 60 to 80 nm exhibit a peak power factor of 1600  W/m.K 2 at 540 K; at 1000 K they exhibit a low thermal conductivity (2.75 W/m.K) and a power factor of 1050  W/m.K 2 leading to a high of ZT of 0.39 ± 0.03. This is the highest ZT reported so far for Sr 0.9 La 0.1 TiO 3 based-compositions. This nanostructuring strategy should be readily applicable to other functional oxides.

thermal conductivity. Over the last two decades, improvements of the ZT value have been achieved in traditional thermoelectric materials and to some extent in newly developed oxide thermoelectrics by micro-and nano-structuring of the materials 1,[4][5][6][7] Part of the driving force for nanostructuring has been the need to decouple the closely related material properties S,  and k and thereby increase the thermoelectric figure of merit, ZT beyond that in normal bulk materials [8][9][10][11][12] .
For many nanostructured bulk materials, the enhancement in the value of ZT can be attributed to a significant reduction in the lattice thermal conductivity as a result of changes in structure and local chemistry, which give rise to a high density of phonon-scattering interfaces 13 . However, the introduction of such interfaces can be counterproductive, as it will generally result in a significant reduction of the electrical conductivity 2 . An alternative approach for the reduction of thermal conductivity is by the fabrication of nanocomposite structures, for instance by compaction of the main material with nanosize inclusions, or by in-situ nano-inclusion formation by means of precipitation [14][15][16] . Here, the main role of nanoinclusions is to generate additional phonon scattering centres without severely reducing the bulk electrical conductivity. The size, shape and volume fraction of the nanoparticles have a significant influence on the scattering efficiency, 17 . However, it is reported that inclusions could increase electrical conductivity while reducing thermal conductivity 18 . The general approach for nanostructuring thermoelectrics involves the introduction of nanosized features such as voids, particles with round or faceted morphology, atomic scale platelets dispersed within the grains or the grain boundaries of the bulk material to reduce thermal conductivity.
In traditional thermoelectrics, significant ZT improvements have been achieved in many systems; in half-Heusler alloys by adding nano-ZrO 2 19 , in Bi 2 Te 3 by adding nano-SiC particles 20 , and in Yb 0.2 Co 4 Sb 12+y by dispersing in situ partially oxidized Yb 2 O 3 nanoparticles 21 It is also has been reported that more significant enhancements can be achieved by embedding metal or conductive nanoparticles into the matrix: examples include lead and antimony in PbTe 22 antimony in Yb y Co 4 Sb 12 23 .
Metal oxides have great potential as thermoelectric materials for high temperature applications, with SrTiO 3 being considered one of the most promising n-type oxides 2,24 . It crystallizes in a simple cubic Pm m perovskite structure, has a high S value of 600 µV/K and is normally utilised as an insulator. However, by doping the Sr sites with lanthanides or the Ti sites with Nb, it becomes an n-type semiconductor 25,26 . The thermoelectric power factor, S 2 , of La-doped single-crystal strontium titanate is comparable with that of traditional thermoelectric materials such as Bi 2 Te 3 , 2 . Despite these excellent electronic transport properties, the lattice thermal conductivity of SrTiO 3 is comparatively high, limiting its usability for practical applications. Various strategies have been employed to reduce the thermal conductivity of SrTiO 3 ceramics including controlling the grain size 27 simultaneously increases the Seebeck coefficient and the electrical conductivity 38 .
Here, we present a novel strategy for self-nanostructuring in oxide ceramics and demonstrate its effectiveness for optimising the transport properties of strontium titanate based thermoelectrics. Through the use of different sintering atmospheres, a distinct core-shell grain microstructure is achieved. In turn, the core of the formed grains is itself nano-structured, comprising a matrix surrounding uniformly distributed pairs of nano-size precipitates and voids. Atomic level imaging and chemical characterization, using scanning transmission electron microscopy (STEM), high-angle annular dark field (HAADF), and electron energy loss spectroscopy (EELS) has been employed to investigate the microstructural changes that occur in Sr 0.9 Ln 0.1 TiO 3 upon annealing in a reducing atmosphere, with a particular focus on the formation of these core-shell type structures containing pairs of voids and inclusions and how they affect the thermoelectric response of the material. This approach to controlling the nanostructure of thermoelectrics could find much wider applicability in the routes to synthesise future target materials and achieving higher performance.

Experimental:
Ceramics of Sr 0.9 La 0.1 TiO 3 were produced by the standard mixed oxide route.

X-ray Diffraction
XRD spectra of L10 samples sintered in air and then annealed at 1350C for different times are shown in Figure 1a. For convenience, these samples will be referred to as 0h, 12h, 24h,

Scanning Electron Microscopy
To evaluate the sample microstructures after sintering and annealing SEM characterization was initially performed. Figure

Transmission Electron Microscopy.
To provide further insight into the details and nature of the core-shell type structure, conventional TEM and advanced electron microscopy using atomically resolved STEM and EELS was performed. TEM data for the shell and core areas of a L10 sample annealed for 24 h in Ar -5%H 2 are presented in Figure 4. The presence of a high density of dislocations in the shell area is the main feature of the microstructure in Figure 4a. Some dislocations are also visible in the core area of the sample. Dislocations are common lattice imperfections in un-doped single crystal and polycrystalline SrTiO 3 prepared in air [47][48][49] . Some of the nanosized inclusions observed in the SEM microstructure analysis can also be seen in the core area in this micrograph (arrowed, blue in colour). Higher magnification TEM images of the core area for the sample, Figure 4(b, c), show a uniform distribution of inclusions, with sizes ranging from 20 nm to 80 nm. The macroscopic morphology of these inclusions can be broadly described as a 'peanut' or 'dumbbell' shape: closer inspection shows that the inclusions actually comprise two 'sub-structures' (see inset Figure 4b) images acquired from such particles and the surrounding matrix, shows that the particles and the matrix of the core may differ in structure (Figure 4c). Furthermore, as noted above, much lower densities of dislocations were observed within the cores, but when present, the dislocations were frequently observed to be connecting sets of particle-void pairs. More detailed analysis of the secondary phase was performed using precession electron diffraction tomography, allowing the complete reciprocal lattice of the crystal structure to be determined. The reciprocal lattice projections were found to be a superstructure of the parent perovskite structure of SrTiO 3 and projections of the crystal structure along one of the perovskite <100> directions and along one of the cubic <110> directions are shown in Figure   5a and 5b respectively. The superlattice reflections between the parent perovskite reflections indicate a much larger unit cell with a volume potentially as large as 1800 Å 3 rather than the perovskite cell volume of 64 Å 3 . The precise atomic arrangement that gives rise to the structure is still under investigation but the high degree of coherency between this phase and the parent perovskite structure is clear from the reciprocal lattice reconstructions, hence the strong directionality of the nanosized features in Figure 3. The small size of these phases, the relatively weak superstructure reflection intensity and the high degree of coherency suggest that this phase would not be easily identified in the conventional XRD analysis.

STEM-HAADF-EELS
In order to provide further insight into the details of the nano size features and dislocations, atomically resolved STEM imaging and EELS measurements were performed. Firstly, the particle-void nature of the inclusions was independently confirmed by low loss EELS measurements of embedded nanoinclusions, demonstrating a drop in relative thickness (see Figure S3) consistent with the presence of a nano-void. The chemistry of the particle part of the inclusion can be further elucidated by looking at the near edge fine structure differences of the Ti L 2,3 and O K EELS edges from a particle/void and matrix area presented in Figure 7a. Figure 7b shows a Ti L 2,3 spectrum extracted from the Sr 0.9 La 0.1 TiO 3 matrix: the spectrum shows the characteristic white line shape of the edge, with the additional splitting of the L 3 and L 2 peaks into e g , t 2g sub-components, corresponding to the Ti +4 octahedral coordination, as expected from SrTiO 3 type compounds 50,51 . The matrix spectrum is plotted against a Ti L 2,3 spectrum extracted from the particle part of the inclusion. It can be readily seen that the e g -t 2g splitting of the L 3 and L 2 peaks is far less pronounced, while the onset of the edge is slightly shifted to lower energies, indicating that the valence of Ti in the particle is reduced towards Ti +3 . Similarly, the near edge fine  As mentioned earlier, dislocations were observed in the core and shell areas of the sample.
The nature of the dislocations was analysed using HAADF images by Fourier filtering to enhance the visibility of the location of the dislocation core (as presented in Figure 8a-c). The application of a Fourier filter on the (100) and (011)

Thermoelectric Response
The air-sintered samples are, as expected, insulators and exhibit low electrical conductivity, consistent with the transport properties of other air-sintered strontium-titanate-based thermoelectrics 32 . Thus, the discussion will focus on the transport properties of samples annealed under reducing conditions. However, the discussion of thermal properties includes the air sintered sample as a reference baseline.
The transport properties of the samples are presented in Figure 9. Although it is difficult to determine the exact structure and chemistry of the nano inclusions, the STEM-EELS data point towards a Ti rich structure, still containing Sr, and exhibiting high level of oxygen deficiency. This phase could be akin to a reduced ternary titanate or a Sr-doped TiO x Magneli-type phase 53 . The electrical conductivity of TiO x -based Magneli phases is high, ranging from 800 S/cm to 300 S/cm at temperatures of 450 K to 1000 K, depending on the oxygen deficiency level 54,55 . Therefore, particles of a similar nature within the microstructure would have a positive effect on the overall electrical conductivity. The pore components of the nano inclusions will, by comparison, have low electrical conductivity. However, since there is no significant reduction in the overall macroscopic density of the annealed samples, it would appear that the effect of voids on the overall electrical conductivity is minimal. Another microstructural feature that will influence electrical conductivity is the presence of dislocations. The full charactorization of the types of dislocations present in these materials is beyond the scope of this study and requires further exploration. However, an atomistic simulation study of the role of <100>{011} edge dislocations, similar to that highlighted in Figure 8, on the defect chemistry and oxide ion transport properties of SrTiO 3 found that oxygen vacancies close to the dislocation core have lower formation energies 56 . Thus dislocations facilitate the removal of oxygen from the structure and aid the formation Ti 3+ ions; in turn these are beneficial to the enhancement of the electrical conductivity 56 . Additionally, dislocations (as structural defects) can contribute to phonon scattering and help to reduce thermal conductivity 36 .
The absolute value of Seebeck coefficients range from 85 to 240 V/K in the temperature range 300 K to 1000 K and reduces systematically with increasing annealing time (Fig. 9b).
As discussed above, this decrease is mainly attributed to an increase in carrier concentration,  To further understand the electrical conduction in the nanostructured samples, the carrier concentrations and mobility were calculated from the electrical conductivity and Seebeck coefficients. The carrier concentrations for the sample are calculated by the modified Hiekes' where is the number of available sites for carrier per unit volume of the unit cell. For the cubic perovskite structure A is 1 and the V is the volume of the unit cell obtained from the refinement of the XRD data shown in Figure 1b. The factor e/k is the ratio of the electronic charge (e) to the Boltzmann constant (k) and is approximately 0.011587 . S is the Seebeck coefficient as shown in Figure 9b.
The modified Hiekes' equation is based on the assumption that only one electron is permitted on a given site and both degeneracies of spin and orbital are negligible 60 . The attribution of vibrational entropy part is also assumed to be 0. Therefore, the calculated values of carrier concentration are expected to be higher than the real values. The carrier mobility can be calculated from the carrier concentration and electrical conductivity ( ) using the following equation: (2) The calculated values for carrier concentration and charge mobility are shown in Table 1  In spite of the high power factor, it is the high thermal conductivity of strontium titanate that limits its use for thermoelectric applications 2 ; a reduction of thermal conductivity is essential for further improvement of the thermoelectric response of this material. The total thermal conductivity of the air sintered and the annealed samples are presented in Figure 9d. The air sintered samples show the expected very high thermal conductivity, typically 7.8 W m -1 K -1 at 350 K to 5.0 W m -1 K -1 at 1000 K, being slightly lower than that for un-doped polycrystalline, air-sintered strontium titanate 32   To explore the relationship between the size of the void-particle nanostructures and thermal conductivity, the size of the longest void-particle pairs in each of the samples (evaluated from the BSE-SEM images), and the lattice thermal conductivity at 1015 K were plotted against the annealing time spent in the reducing atmosphere ( Figure 11). It can be seen that there is a direct correlation between lattice thermal conductivity and void-particle size; the smaller the size of the nano-sized features the lower the lattice thermal conductivity. This trend is valid for all the measurement temperatures for thermal conductivity. However, we do not have reliable data for the density of void-particle pairs in the different samples.

Conclusions
We have established a new path towards engineering nanostructures in strontium titanate based thermoelectrics which may be applicable to other oxides. High density Sr 0.9 La 0.1 TiO 3 polycrystalline samples, with regular polygonal grains having a narrow grain size distribution and average size ~ 7 µm, were synthesised through liquid phase sintering by addition of boron and sintering in air atmosphere; they were then annealed in a reducing environment.
This combined heat treatment led to the formation of a complex nano-structure within the grains.
X-ray diffraction confirmed Pm m cubic crystal structure for all the samples. SEM revealed a core-shell type microstructure with the presence of nano-size features within the core area.
HRTEM showed that the nano-size features are in the form of pairs of nano-sized voids and particles; their sizes depending on the annealing time. Atomic-resolution STEM-HAADF-EELS characterization in an aberration-corrected microscope showed that the precipitates are rich in Ti and the areas around the voids contain a high concentration of Ti 3+ . Additionally, a high density of dislocations was observed in the shell areas; dislocations are also present in the core of the nano-structures, but with a lower density. Their presence may enhance electrical conductivity, although this aspect of the structure remains to be explored in more details.
The self-nano-structured Sr 0.9 La 0.1 TiO 3 ceramics showed a high power factor of 1600 W/m.K 2 to 1050 W/m.K 2 at temperatures of 600 K to 1015 K. However, the major impact of nano structuring was the reduction of thermal conductivity. Nano structured Sr 0.9 La 0.1 TiO 3 ceramics with shell size of ~1 micron and inclusions of 60 to 80 nm exhibit a low thermal conductivity of K = 2.75 W/m.K at 1015 K leading to a high of ZT of 0.39±0.03 at this temperature. This is the highest ZT achieved for the highly studied Sr 0.9 La 0.1 TiO 3 composition 15,25,26,34,59,[69][70][71] and for other 10 mole % lanthanide doped SrTiO 3 thermoelectrics 25 ; a summary of published data is provided in Table S1. The study demonstrates a powerful nanostructuring strategy for significantly enhancing the performance of thermoelectric oxides; the approach could find much wider application in providing valuable guidance in the routes to synthesise future target materials.