Thermoelectric effect can be used to directly convert heat into electrical power. The efficiency of a thermoelectric material depends on a dimensionless figure of merit defined as^[1–3] , where Z, T, S, σ, and κ are the figure of merit, absolute temperature, Seebeck coefficient, electrical conductivity, and thermal conductivity (which is composed of lattice part and electron part , respectively. Increasing the power factor and decreasing the thermal conductivity κ simultaneously lead to a higher ZT. However, these three parameters are coupled, while increasing in S often results in the suppression of σ and . One effective approach to enhance ZT is to lower the total thermal conductivity by decreasing the phonon component while maintaining the electrical conductivity.^{[4, 5]}

Metal oxides have drawn great attention as promising thermoelectric materials due to their non-toxicity, abundance, and stability. Among them, SrTiO is regarded as one of the most promising n-type materials for thermoelectric applications. The wide band-gap perovskite oxide can be easily doped at cation Sr and Ti sites or with oxygen vacancies to form an n-type semiconductor.^{[6, 7]} The power factor of La doped SrTiO single crystal reaches 36 W/cmK at room temperature,^[7] which is comparable to the typical thermoelectric materials like SnSe,^[8] and likely due to the significant contribution from the Ti d band.^[9] Despite the excellent electronic properties, the lattice thermal conductivity of SrTiO is comparatively high, which results in a relatively lower ZT. Various methods, such as nanostructuring,^[10] cation doping,^[11] and introduction of Sr site^[12] or oxygen vacancies,^{[13, 14]} are proposed to improve the ZT of SrTiO . As a common post-treatment method,^[15–19] thermal annealing in reducing atmosphere or in vacuum can generate oxygen vacancies in metal oxides and will affect their electrical and thermal properties.

In this work, the thermal transport properties (including thermal diffusivity, heat capacity, thermal conductivity) of SrTiO are tuned by thermal annealing and measured by the laser flash method and differential scanning calorimetry. Then the structural, vibrational, and electronic properties of SrTiO are studied by scanning electron microscopy (SEM), Fourier transform infrared spectroscopy (FTIR), and near-edge x-ray absorption fine structure (NEXAFS), respectively. We also perform first-principles calculations to compare with the experiments. A phonon scattering model is proposed to explain the reduction of thermal conductivity after annealing treatments.

Commercially available single-crystalline substrates of SrTiO ( ) were obtained from HEFEI KEJING Materials technology Co., LTD. The SrTiO substrates were annealed at various temperatures (from 1073 K to 1473 K) either in Ar atmosphere for 24 h or under vacuum ( Pa) for 4 h. The morphology of the samples was characterized by SEM using LEO 1530. The thermal diffusivity of the samples was measured by the laser flash method (LFA)^[20] using a NETZSCH LFA457 MicroFlash apparatus with the temperature ranging from 373 K to 773 K in a nitrogen atmosphere. The specific heats of the samples were characterized by differential scanning calorimetry (DSC) using a NETZSCH DSC 404 F3 Pegasus apparatus. The thermal conductivity was calculated according to the equation

(1)

FTIR measurements were performed by Bruker VERTEX 70v spectrometer in the far infrared region. The NEXAFS was measured at synchrotron radiation beamline 24A1 in the NSRRC. The O K-edge and Ti L-edge NEXAFS data were collected in the total electron yield (TEY) mode. The photon energy was calibrated by the binding energy of Au 4f. First principle calculations have been carried out utilizing the WIEN2k code^[21] with the method of full potential linearized augmented plane wave plus local orbital (FP-LAPW+lo).^[22] For the ground-state calculations, the exchange–correlation interaction was described by the Perdew–Burke–Ernzerhof generalized gradient approximation (PBE-GGA).^[23] The muffin-tin radii were selected to be 2.5 a.u. for Sr, 1.87 a.u. for Ti, and 1.69 a.u. for O. The self-consistency was defined when the total energy converges less than 1 mRy and the charge converges less than 0.0005e.

The temperature dependences of thermal diffusivity and specific heat of the annealed samples (treated under the vacuum pressure of Pa for 4 h) where compared to those of the as-received sample, as shown in Fig. 1. The results were used to calculate the thermal conductivity of the samples (shown in Fig. 1). The annealed samples show a dramatic decrease in thermal conductivity with increasing annealing temperature. For example, the thermal conductivity of the SrTiO sample annealed at 1473 K is 3.26 W/mK at 773 K, which is 32% lower than that of the as-received sample. According to the study of Muta,^[13] the electrical conductivity of a SrTiO sample annealed under vacuum is about m at 373 K and gradually decreases to m at 773 K. According to the Widemann–Franz law (where is the Lorentz number and equals to /K for a degenerated semiconductor^[11]), the electronic contribution to the thermal conductivity is in the range of 0.001 W/mK to 0.01 W/mK, which is negligible comparing to the measured thermal conductivity (3–8 W/mK in the temperature range of 300–800 K). Thus, the electronic thermal conductivity can be ignored in this work, and the large suppression of thermal transport is mainly due to the decrease of the lattice thermal conductivity.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. (color online) Thermal conductivity of the annealed SrTiO samples, compared with that of as-received SrTiO sample.

In order to investigate the potential mechanism for the decrease of the lattice thermal conductivity, FTIR was employed to characterize the as-received SrTiO sample as well as the annealed (in Ar atmosphere at 1073 K, 1173 K, and 1273 K for 24 h) samples, as shown in Fig. 2. The absorption bands consist of two features (labeled as A and B) in the range of 590–640 cm , which originate from the Ti–O stretching vibration and assigned to the transverse optic (TO) phonon mode.^[24] The existence of the two features in this range is likely due to the slight difference in Ti–O bond structures (e.g., near or away from defects). For the spectra of the as-received samples, the asymmetric shape of the two features suggests that more than one component exists in each feature, which leads to an FWHM of approximately 10 cm . With increasing annealing temperature, the shapes of the two features become symmetric, which results in a drop of FWHM of both features (around 5 cm . This is consistent with an earlier report, and attributes to relaxing of the strain and reordering of the SrTiO crystal.^[25] Such development in the spectral shape also implies that the Ti–O band structures gradually segregate into two distinct categories, which is consistent with the observed evolution of samples’ morphology (emerging of linear dislocations in scanning microscopic images, as shown in Appendix Fig. A1). The center of feature A remains stable (around 621 ± 1 cm , while the center of feature B increases from 606 cm to 610 cm with increasing annealing temperature. This suggests that feature A is more likely associated with Ti–O bonds within perfect lattice structure (without oxygen vacancies), and feature B is linked with Ti–O bonds near defects (e.g., oxygen vacancies). Furthermore, the intensity ratio (B/A) gradually increases with increasing annealing temperature, from 0.89 (as-received) to 1.2 (annealed at 1273 K), which indicates that the concentration of defects rises with increasing annealing temperature. This suggests that the annealing treatment improves the crystallinity of the samples in some regions, but leads to segregation of defects in other regions.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. (color online) FTIR spectra of the annealed SrTiO samples, compared with that of as-received SrTiO sample. For easy comparison, the spectra were shifted vertically.

The crystallinity of a material correlates with its electronic structure. Hence, the NEXAFS spectra were measured for the annealed samples (treated at 1473 K under the vacuum pressure of Pa for 4 h), and compared with that of the as-received sample, in order to gain more insight into the samples’ properties after annealing. The O K-edge spectra shown in Fig. 3(a) reflect the transitions from the O 1s core level to the unoccupied O 2p (and potentially hybridized Ti 3d) conduction band. For comparison, the calculated partial densities of states (PDOS) for oxygen and titanium are presented in Fig. 3(a) as well. The peak A at 531.0 eV is likely originated from the hybridization of O p _x and p _y states with Ti state. The peak B (533.4 eV) is related to the hybridization of O 2p and Ti e states.^[26] The FWHMs of the peaks are slightly enlarged after annealing, which indicates an increase of the oxygen vacancies.^[27] This is likely due to the reduced symmetry caused by the presence of the oxygen vacancies.^[28] Meanwhile, the intensity ratio between peaks A and B is reduced after annealing. Similar results have been reported for an oxygen-deficient system.^[29] Such change can be associated with the filling of the low energy band (hence, reducing the density of the unoccupied states) by the excess electrons due to the oxygen vacancies. To further confirm the presence of the oxygen vacancies, the Ti -edge NEXAFS spectrum of the annealed sample is shown in Fig. 3(b) as well as that of the as-received sample. The peaks C at 457.1 eV and D at 459.4 eV are originated from the transitions from the Ti 2p core level to the unoccupied Ti 3d bands (t and , respectively). A slight broadening of the peaks is observed in the spectra of the annealed samples. This is likely contributed by the presence of Ti state.^[29] In the oxygen-deficient SrTiO system, each oxygen vacancy provides two electrons to the Ti d band, leading to a reduction of the Ti from to partial state.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (color online) (a) O K-edge and (b) Ti L-edge NEXAFS spectra of the annealed SrTiO sample, compared with as-received SrTiO sample. The calculated O 2p and Ti 3d PDOS of SrTiO is shown in panel (a) as reference.

Based on the FTIR and NEXAFS results, we propose a structure evolution model to explain the thermal conductivity reduction mechanism. First, annealing induces a reordering process for the atoms in some areas. Therefore, within the areas, the atoms line up in a better order, which improves the overall crystal quality. However, this leads to a more disorder/imperfect structure, such as oxygen vacancies, at the regions between the areas. Second, the samples annealed with reduced oxygen pressure may yield a large amount of oxygen vacancies. For example, the sample that has been annealed in the Ar atmosphere for 24 h shows chain-like features (observed on the SEM image, S.I.-1), which is consistent with other reports.^{[30, 31]} Thus, the change in thermal conductivity may result from a competition between the crystallization inside some domains and the increased imperfection scattering in the region between these domains.

To quantitatively evaluate the proposed model, a phenomenological Callaway's model^[32–37] is used to fit the thermal conductivity data (shown in Fig. 1). In this model, the lattice thermal conductivity is given by

(2)

(3)

(4)

(5)

(6)

(7)

Table 1.

Table 1.

Table 1. Fitted scattering parameter and A for annealed sample compared with that for as-received sample. . Sample As received 0.95 1.96 1073 K annealed 0.99 3.65 1273 K annealed 1.32 2.53 1473 K annealed 1.50 2.61

Table 1. Fitted scattering parameter and A for annealed sample compared with that for as-received sample. .

The LFA has been used to investigate the phonon scattering in lead chalcogenide thermoelectric^{[5, 37]} and the thermal characterization of a bridge-link carbon nanotubes array used as a thermal adhesive.^[38] More recently, we used combined SEM, LFA, DSC, and finite element method to investigate the thermal properties of boron nitride, and demonstrated that hexagonal boron nitride laminates exhibit a high thermal conductivity, which is significantly larger than that currently used in thermal management.^[39] With additional characterizing methods (such as FTIR and NEXAFS) included, this work successfully demonstrates the usefulness of the integrated study of several tools to reveal the origin of reduced thermal conductivity in annealed SrTiO . It is clearly shown that the unique advantage of the combined methods can be achieved because the different aspects of functional materials can be investigated by different characterization means. For example, the thermal transport properties such as thermal diffusivity can be characterized by LFA, the heat capacity can be obtained by DSC measurement, and the mechanism of thermal conductivity can be analyzed by Callaway's model. The crystal structure can be observed by SEM, the vibrational properties can be determined by FTIR, and the electronic properties can be revealed by near edge x-ray absorption fine structure and first-principles calculations. The whole picture of the crystal structure of the material and its properties can then be obtained. The combined method, as proposed in this work, is expected to be useful to study other advanced functional materials.

SrTiO samples were annealed at various temperatures either in Ar atmosphere or under vacuum. Based on the experimentally measured thermal diffusivity and specific heat, the thermal conductivity of the samples was calculated, which dramatically decreased with increasing annealing temperature. The FTIR and NEXAFS reveal that the defects (i.e., oxygen vacancies) in the samples increase (especially after being annealed under vacuum) and segregate after annealing, which is consistent with the change of the morphology of the samples and other reported results. A phenomenological model based on Callaway's theory was used to fit the thermal conductivity of the samples. The results confirm that the increasing annealing temperature leads to an increase and segregation of oxygen vacancies, which is responsible for the significant reduction of the thermal conductivity after annealing. This work demonstrates the unique usefulness of the integrated study for revealing the mechanism of thermal conductivity of SrTiO . Therefore, this work suggests a promising means to characterize and optimize the material for thermoelectric applications.