Thermoelectric properties of in-situ plasma spray synthesized sub-stoichiometry TiO2−x

The thermoelectric properties of sub-stoichiometric TiO2−x deposits produced by cascaded-plasma spray process are investigated from room-temperature to 750 K. Sub-stoichiometric TiO2−x deposits are formed through in-situ reaction of the TiO1.9 within the high temperature plasma flame and manipulated through introduction of varying amounts of hydrogen in the plasma. Although the TiO2−x particles experience reduction within plasma, it can also re-oxidize through interaction with the surrounding ambient atmosphere, resulting in a complex interplay between process conditions and stoichiometry. The deposits predominantly contain rutile phase with presence of Magneli phases especially under significantly reducing plasma conditions. The resultant deposits show sensitivity to thermoelectric properties and under certain optimal conditions repeatedly show Seebeck coefficients reaching values of −230 μV K−1 at temperatures of 750 K while providing an electrical conductivity of 5.48 × 103 S m−1, relatively low thermal conductivity in the range of 1.5 to 2 W m−1 K−1 resulting in power factor of 2.9 μW cm−1 K−2. The resultant maximum thermoelectric figure of merit value reached 0.132 under these optimal conditions. The results point to a potential pathway for a large-scale fabrication of low-cost oxide based thermoelectric with potential applicability at moderate to high temperatures.

. Therefore, oxygen deficiencies and corresponding crystal lattice defects can be easily induced with concomitant implications on electrical properties. In particular, sub-stoichiometric TiO 2−x has been explored as a potential thermoelectric material 19 . Lu et al. have demonstrated that TiO 1.95 prepared by oxidizing and reducing from the Ti metal reveals S of − 800 μ V K −1 and σ of 55 S m −1 at 460 K 20 . Tsuyomoto et al. have reported that TiO 1.94 with an orthorhombic crystal structure prepared by reducing the anatase in H 2 , exhibits a peak S of − 518 μ V K −1 and σ of 1.9 × 10 3 S m −1 at 343 K 21 . They showed that both S and σ increased with temperature resulting in power factors (PF) of ~510 μ W m −1 K −2 at 343 K and ~35 μ W m −1 K −2 at 460 K, respectively. Furthermore, a single crystal of rutile TiO 2 reduced to TiO 2−x via annealing in H 2 elevated temperature at 1053 K shows exceptionally high S of up to 60000 μ V K −1 at cryogenic temperatures of 10 K 22 , and interpreted the results preferably due to the phonon drag effect of the holes. Evidently, the introduction of oxygen vacancies in the TiO 2 crystal lead to increase in carrier concentration resulting in the enhancement of σ. However, an excessive oxidation of TiO 2 at elevated temperature lowered σ and S, resulting in low PF and ZT. Thus, obtaining a fabrication process for TiO 2−x should be chosen to maximize the high PF while maintaining the low thermal conductivity.
Sharma et al. 23 observed that plasma-sprayed TiO 2 deposits behave as n-type semiconductor with reasonable conductivities (1.76 × 10 3 S cm −1 ) up to about 823 K, after which rapid re-oxidation occurs with concomitant increases in resistance. This was attributed to oxygen loss occurring during the spray process and subsequent trapping of the vacancy state through rapid solidification. Subsequently, Colmenares-Angulo et al. 24 examined process-induced stoichiometric changes in plasma sprayed TiO 2 deposits allowing tunability in electrical conductivity. There have been significant efforts to use thermal spray processes to obtain high anatase content TiO 2 for potential applications such as for large area photo catalysts for decomposition of hydrocarbons and nitrogen oxides 25 where TiO 2 is extensively used for variety of industrial applications particularly in the form of films and coatings including optical, electrical, photocatalytic and tribological applications 26 .
In this study, the thermoelectric properties of plasma sprayed TiO 2−x are examined through variations in-situ reduction within the process. Plasma spray offers a range of conditions through not only the melt state of the material is affected, but also its stoichiometry associated with complex reactions between the high temperatures and its surrounding environments comprising initially of an active (reducing in presence of H 2 and increase Seebeck coefficient) plasma and subsequently oxygen at the location of deposition 27,28 . The deposit itself comprises of layered "splats", some porosity and cracking which can offer potential benefits (reduced thermal conductivity and increased Seebeck coefficient) and deleterious (reduced electrical conductivity) in the formed deposit. Through a systematic series of processing experiments coupled with specimen fabrication, the thermoelectric properties of the specimens were evaluated and optimized with promising initial results. An exhaustive parametric study using both conventional and cascaded plasma torches considering two types of powders; a near stoichiometric (TiO 1.9 ) and highly reduced (TiO 1.7 ). For brevity only cascaded plasma spray results with the TiO 1.9 powder is reported as only these conditions showed promising results. Of importance is the uniform thermal fields of cascaded plasmas that reduces variabilities in both particle temperature and plasma-particle interactions 29 . Plasma spray is a flexible and scalable manufacturing processing technique capable of operating in an ambient environment and allows assembly of mesoscale multilayers both as blanket deposit and patterned structures with metals (as potential conductors), ceramics (as dielectric interlayers) and functional oxides [30][31][32] . In addition, the process allows for conformal deposition directly onto components potentially enabling direct fabrication of useful thermoelectric devices for exploiting applications including cooling and refrigeration 33 , energy harvesting from heat 34 , solar thermoelectric generators and radioisotope thermoelectric generators 35 , and sensors 36 .

Results and Discussion
As noted in the methods section (end of paper), a nominally stoichiometric TiO 2 industrially produced feedstock powder (measured to be TiO 1.90 ) was used to produce plasma spray deposits. This powder is a plasma spray grade with particle sizes ranging from 8.6 to 28.9 μ m. As noted in the introduction, as part of initial screening study, over 18 different iterations of spray conditions were attempted considering both conventional and cascaded plasma spray torches as well as high velocity combustion thermal spray systems. The cascaded plasma spray torches with its uniform thermal field yielded more homogeneous deposits with significant results and as such only data from this study is reported in the paper. In this case, the H 2 content of the plasma was systematically varied to control the oxygen loss and thus stoichiometry and phase of the deposit. Three significant process conditions (Deposit A, B, and C) comprising of differences in plasma spray state and H 2 content of the plasma gas are discussed in detail to and were analyzed for their phase, stoichiometry, microstructure and thermoelectric properties. This integrated strategy not only allows highlighting the significant finding but also establish the process-property relationships and to identify the contributing aspects that enhance the thermoelectric properties.
Phase study and microstructure of the deposits. Figure 1 shows the XRD patterns of the starting feedstock powder and three deposits produced under different hydrogen content within the plasma indicating that diffraction peaks from the feedstock powder diminishes and broadens for as-sprayed deposits as they showed broad diffuse reflection in the 2θ positions near by the rutile peaks. The starting feedstock powder is primarily composed of rutile and Magneli phases (Ti n O 2n−1 where n = 5, 6,7,8,9,10). The sprayed deposits retained the majority of the rutile phase along some of the Magneli phases (n = 5, 6) and trace amounts of anatase. Among the deposits, deposits A and B (with 0 and 3 L min −1 of H 2 ) show Magneli phases with n = 5, 6 while deposit C (sprayed with 9 L min −1 of H 2 ) shows Magneli phases with n = 4, 5 and 6. The increase in Magneli phases especially Ti 4 O 7 is resultant from both the high enthalpy of the plasma at this condition and more reducing environment of the plasma.
As the H 2 content in the plasma increases, not only the phase structure and crystallinity are affected, but there is deviation from stoichiometry due to the oxygen loss. Deposit A produced without any H 2 in the plasma indicated a small gain in stoichiometry (TiO 1.93 ) perhaps associated with interaction of the hot particle with oxygen in the environment at the deposition location 37,38 . Deposit C sprayed with 9 L min −1 of H 2 showed significant loss of oxygen resulting in a stoichiometry of TiO 1.83 . Deposit B sprayed with 3 L min −1 of H 2 in the plasma maintains the stoichiometry, but this again may be associated with competition of oxygen loss in the active portion of the plasma and regaining the oxygen at the deposition location due to comingling with atmospheric oxygen dampening the next changing in stoichiometry. Deposit B* sprayed with 6 L min −1 of H 2 (included in Table 1) followed this trend with a stoichiometry of TiO 1.86 confirming this operative mechanism. Figure 2(a-c) presents the representative cross-sectional SEM micrographs of the as-sprayed deposits A, B and C processed under 0, 3, and 9 L min −1 of H 2 conditions, respectively. The image of Fig. 2(a) shows deposit A with some porosity and micro-cracks within the lamellae. Since relatively lower temperatures are involved in the formation of deposit A, it consists of thinner splat layers with fewer and substantially narrower interlamellar pores. Figure 2(b) shows cross-section of denser deposit B with numerous micro-cracks. The image of Fig. 2(c) is similar to 2(b), but the deposit C now comprises of larger connected macro-cracks. The deposit C sprayed at higher plasma energy and particle temperature ( Table 1) will lead to greater quenching stresses for the solidifying particle as well as higher substrate temperature (530 K for 9 L min −1 of H 2 whereas 425 K for 0 L min −1 of H 2 ) consequently leading to greater propensity of micro and macro-crack formation. Deposit B had relatively lower particle and substrate temperature resulting in reduced propensity for micro-cracking. This micro-cracking is attributed to the large quenching stresses associated with the rapid solidification of the depositing molten particles. The particles are exposed to the high thermal energy resulting in melting followed by rapid solidification which induces volume change and temperature change resulting in build-up of large tensile stress and which are relieved through cracking 39,40 . Deposit A at low thermal energy condition (0 L min −1 of H 2 ) shows the lowest density (3.925 g cm −1 ) compared to those sprayed at higher temperature, and exhibits a somewhat porous microstructure. Deposit B and C show increase in density from 4.138 to 4.218 g cm −1 , but also greater content of micro-cracks.
In situ beam curvature measurements were conducted to further quantify these microstructural observations via extracting the deposition stress and elastic modulus of the sprayed deposits following principles described in the literature [41][42][43][44] . This technique provides a volume averaged information about the microstructural integrity and interlamellar bonding and as such provides reliable information of regarding the nature of the deposits. Despite the higher enthalpy processing condition, deposit C showed lower modulus (139 GPa) and lower evolving stress (24 MPa) compared to the deposit A whose values were 160 GPa for modulus and 90 MPa for evolving stress, respectively. Typically, H 2 imparts higher enthalpy to the plasma and thus higher particle temperature and this should generally produce higher density deposits. The lower evolving stress and lower modulus in deposition C confirm the microstructural observations in terms of observation of substantial micro-and macro-cracking at the higher H 2 conditions. These cracks also lower the deposit integrity captured by the lower modulus of deposit C.  Thermoelectric Properties. The electrical conductivity, σ, Seebeck coefficient, S, and power factor, PF, for the deposits are summarized in Fig. 3. The electrical conductivities for TiO 2−x deposits respond similar to a semiconductor following same temperature-dependent trend as reported in the literature 20,45 . However, for the deposit A sprayed without H 2 in the plume with a resultant stoichiometry of TiO 1.93 can observe an increase in slope at 620 to 730 K following which there is a small decrease. The first change in slope above 620 K is attributed to the thermal excitation of carriers, while the second is related to the phase transition from space group I4 1 /amd to the P4 2 /mnm 24 . The temperature dependence of the electrical conductivity for the deposit B was relatively weak and the value of the σ is significantly lower than that of the deposit A because of the micro-cracks within its microstructure. Deposit C shows electrical conductivity similar to that of deposit A despite presence of macro-cracks in the deposit, due to the significant presence of Ti 4 O 7 phase. This confirms the greater conductivities of the deposits associated oxygen loss with larger deviation from microstructure and phase. The Seebeck coefficients for the deposit A, B, and C were all negative, ranging from − 73.5 μ V K −1 to − 246 μ V K −1 , and were measured − 230, − 139, − 100 μ V K −1 at 750 K, respectively, suggesting that the deposits are n-type conductors. Deposit A comprising of larger fraction of the rutile phase exhibited the largest Seebeck coefficient. As He et al. 17 has reported, stoichiometry close to TiO 2 resulted in an increase in the Seebeck coefficient where bulk rutile had Seebeck coefficient of − 440 μ V K −1 . The power factor increased with increasing measurement temperature for all of the deposits and the highest power factor of 2.91 μ W cm −1 K −2 at 750 K was achieved with deposit A. The difference in the Seebeck coefficient is attributed to the presence of oxygen defects which is inversely proportional to the carrier density whereas electrical conductivity depends upon combination of oxygen vacancy, microstructure and crystallinity 23 . Increase in hydrogen content of the plasma can lead to greater oxygen loss and thus further phase reduction (Ti 4 O 7 ) with associated increase in electrical conductivity and reduction in Seebeck coefficient as well. However, microstructural integrity can also contribute to the measure of electrical conductivity. Although the deposit A has a lower oxygen deficiency, the deposit does not have significant micro and macro-cracking. The deposit C has a relatively high conductivity due to presence of Ti 4 O 7 phase 46,47 despite the presence of macro-cracks. The deposit B has neither the extent of phase nor the microstructural benefit resulting in low electrical conductivity. The results point to the operative mechanisms and a processing approach to optimize the desired features. Thermal Conductivity. The total thermal conductivity for the TiO 2−x deposits ranged from 1.36 W m −1 K −1 to 2.17 W m −1 K −1 depending on the stoichiometry of the deposits and the temperature at which it was measured as shown in Fig. 4(a). The thermal conductivity of all three deposits decreased from room temperature to 450 K, but, surprisingly the behavior of the thermal conductivity changes above 450 K depending from the deposit because the difference in stoichiometry has effect on the phonon scattering 17,48 . A lowest total thermal conductivity value of 1.36 W m −1 K −1 at 650 K was measured from deposit A and the thermal conductivity value increased to 1.65 W m −1 K −1 at 750 K, which is relatively low compared to the literature 17,20,24,45,[48][49][50] . Figure 4(b) shows the ratio of lattice thermal conductivity over total thermal conductivity of the TiO 2−x deposits. The κ l was calculated by subtracting the κ e from the total thermal conductivity. The electronic contribution was estimated by using the Wiedemann-Franz law 51 and shows to be at most 0.1 W m −1 K −1 at 750 K in deposit A among the two other deposits and the ratio of κ l to κ tot are shown in Fig. 4(b). This ratio indicates that the total thermal conductivity is dominated by phonon transport. These results suggest that the extrinsic microstructural features such as density, micro-and macro-cracks, and porosity contributes higher than the intrinsic non-stoichiometry like defects and vacancies of the material 27,52 .
In this study we have only been able to report the thermal conductivity in the through-thickness direction as it is extremely difficult to reliably measure the conductivity in the in-plane direction for these thin deposits. It would be expected that the in-plane conductivity would be somewhat higher due to the lamellar nature of the sprayed material. However, the microstructure of sprayed layers is rather chaotic with successive assemblage of thousands of particles which eliminates some of the microstructural anisotropy. Chi et al. 52 has reported this orientation dependence on thermal conductivity ranging from near equivalent up to 25% lower dependent on materials and process conditions. Recent studies conducted on very thick (25 mm) zirconia coatings allowing for measurements in both orientations confirm this earlier observation. Nevertheless, the process-property trends reported in this work should be consistent.  Fig. 5. From this figure, it is apparent that the results reported here for deposit A are comparable or superior to the other reported data on the thermoelectric properties of TiO 2−x 45,48-50 . It is clear that optimization of high Seebeck coefficient, high electrical conductivity and low thermal conductivity promotes enhanced thermoelectric property, in which the ZT value reached maximum at 0.132 at 750 K. Lastly, condition for deposit A was repeated in six separate processing trials to ensure repeatability with separate fabrication and measurements resulting in a highly repeatable ZT ranging from 0.123 to 0.133. In addition, complete multilayer devices incorporating multiple junction incorporating both series and parallel connections with Ni as a nominal p-type material showed scalable increase in voltage and current.  Interplay among process-microstructure-thermoelectric properties. The results presented in this paper are significant from several points of view. First confirmation on the potential of sub-stoichiometric oxides such as TiO 2−x as alternative thermoelectric materials prepared through and industrial scale, layered manufacturing process. In Fig. 5, the results from this study are benchmarked with those reported in the literature indicating equivalent of better properties suggesting that plasma spray strategies may be of relevance to other oxide thermoelectric systems. Although plasma spray has been contemplated for thermoelectrics for many years, this is the first time that significant results have been achieved. This unusual result is related to non-obvious process-phase-microstructure-property relationships and as such warrants more detailed examination. In this section, we seek to distill these interrelations which will not only clarify the results presented in this paper but provide a framework for future developments for thermoelectric oxides in general and thermal sprayed functional oxides in particular.
As noted, earlier although many iterations of process parameters and initial stoichiometry were examined as part of the larger study and among them, only condition for deposit A showed this relatively high thermoelectric figure of merit. By comparing condition for deposit A with B and C presented in this paper, key attributes that allow for high ZT can be identified: • Consistent with contemporary knowledge, significant retention of the original rutile phase and reduced decomposition are beneficial to retain the high Seebeck coefficient in the TiO 2−x system. Both XRD and TGA results (Table 1) confirm that condition for deposit A retains higher fraction of the rutile phase and devoid of the decomposed Ti 4 O 7 phase. This result is explained based both absence of H 2 and lower average spray plume particle temperatures for deposit A. Lower average temperature indicates that some of the particles are likely unmelted and retained as such in the deposit. Detailed temperature distribution analysis from single particle measurements (not shown in the paper) confirms this to be case. The median temperature through this alternative measurement for deposit A is 2100 K compared to 2200 K for deposit B and 2340 K for deposit C. The absence of H 2 in the plasma not only reduces the particle temperature but also suppresses the propensity for any inflight reduction, and potentially encourages interaction of the hot particle with the oxygen in ambient environment resulting in a small stoichiometry gain in the deposit A compared to deposit B and C. • Under higher particle temperature and H 2 rich conditions, loss of stoichiometry and presence of Ti 4 O 7 Magneli phase are favored which suppress the Seebeck coefficient, but concomitantly offer higher electrical conductivity. • Condition for deposit A is unusual in that it not only retains the Seebeck but also has high electrical conductivity with no significant deviation in thermal conductivity resulting in high ZT. To understand this unusual result, one needs to look at the microstructure more critically. As shown in Fig. 2, deposit A does not have significant micro or macro-cracks which can impede carrier flow. But beyond this obvious difference, more nuanced differences are seen through additional SEM analysis of fracture cross-sections and examining single isolated droplets on polished substrates. This is shown in Fig. 6 comparing low and high magnification images of deposit A and C. Low magnification images confirm the presence of both vertical and horizontal cracks in deposit C. The higher magnification (30,000 X) images provide additional insights. Deposit A shows relatively clean fracture surface with good microstructural integrity and reasonable connectivity among the splats. The image also shows presence of unmelted or partially melted particles with somewhat larger grain size. In contrast deposit C shows microspheres which are a few micrometers in diameter which result from particle splashing on impact. This is largely absent in deposit A. Single splat deposition on polished substrates analysis of all three conditions shown in Fig. 7 confirms that condition A results in a deposit with limited splashing. This is attributed to its lower overall particle temperatures which can lead to reduced splashing 53 . These observations indicate that even though deposit A does not contain a larger fraction of decomposed Ti 4 O 7 Magneli phase there are sufficient charge carriers and reduces impedances to provide reasonable electrical conductivity. The nanoscale separations and interfaces that are still present likely contribute to the reduced thermal conductivity. • Higher resolution XRD data collected on deposits A and C (Fig. 8) not only confirms the phase results from Fig. 1, but also suggests lower peak broadening in deposit A associated with greater fraction of unmelted particles and fine debris in the deposit. The diffraction angle for (101) peak shown in Fig. 8 shifted from 36.05° to 35.82° as the H 2 flow changed from 0 to 9 L min −1 . This shift in the 2θ angle can potentially be attributed to differences in residual stresses between the coating observed earlier in the curvature based stress measurement. • Lastly, curvature based evolving stress and modulus data reported are consistent with the microstructural observations indicating that deposit A shows excellent integrity and bonding leading to high electrical conductivity.
In summary, the results and integrated analysis presented in this paper suggest potential for considering plasma spray as both a synthesis and manufacturing tool for thermoelectric oxides. Ongoing activities are focused on building thermoelectric waste heat harvesting devices that integrate p-n junctions, sequencing the layers and junctions to provide series and parallel connections to add voltage while retaining current carrying capability enabling useful energy harvesting systems. As more promising materials are identified, the fabrication concepts can be concurrently contemplated for device engineering.

Conclusions
The thermoelectric properties of cascaded plasma spray synthesized TiO 2−x having oxygen defects and a combination of rutile and Magneli phases were investigated. It was observed that use of H 2 in the plasma forming gas allowed tunability in TiO 2−x stoichiometry and phase content. Deposits produced under high H 2 conditions tended to comprise of rutile phase and reduced Magneli phases with significant oxygen deficiency, while deposits produced without any H 2 in the plasma flame resulted in predominantly rutile phase with stoichiometry closer to TiO 2 . The introduction of oxygen deficiencies with resultant formation of Ti 4 O 7 increases electrical conductivity σ, but diminishes S. The deposits have significantly reduced through-thickness thermal conductivities due to the  presence of large number of pores and interfaces within the defected microstructure of the plasma sprayed deposits. The combined attributes of optimized phase, stoichiometry and microstructure resulted in a TiO 2−x material with reasonable ZT value reaching 0.132 at 750 K for this class of transition metal oxides.
A key benefit of the plasma spray approach is the ability to directly apply these functional oxide deposits onto thermostructural components in either blanket or patterned (either through masking or via direct writing) with the ability to integrate insulator and conductors to produce complete devices. An added embodiment is the ability to accomplish this at relatively low processing temperatures (< 473 K) allowing for multilayer assembly to enable direct fabrication of thermoelectric systems on both planar and conformal waste heat systems. The results point to a pathway for further exploration of transition metal oxide based thermoelectric materials through a flexible and scalable manufacturing process.

Methods
Deposit preparation and particle diagnostics. Sub-stoichiometric TiO 2−x deposits were fabricated by atmospheric plasma spray process with torch hardware (SinplexPro TM , Oerlikon Metco), varying the hydrogen gas ratio. As noted in the paper, although a much larger range of plasma spray operating conditions including use of both conventional and cascaded plasma sprays were contemplated significant results were only observed for the cascaded plasma attributed to its more uniform plasma flow field. The details of spray conditions are provided in Table 1. An air jet attached to the torch and air knife below the substrate both blowing towards the substrate were employed to cool the deposit during spraying. A nominally stoichiometric spray grade feedstock powder was chosen; TiO 1.9 (Metco 102, − 45 + 11 μ m, fused and crushed, Oerlikon Metco). These materials are used for wear applications and as such stoichiometry is not carefully monitored. The deposits for Seebeck and electrical conductivity measurements were deposited on YSZ (9024, − 75 + 10 μ m, Saint-Gobain) coated stainless steel substrate (as a dielectric) through bar-shaped masks (2.5 mm W × 140 mm L and 4 mm W × 25.4 mm L, respectively). Thermal conductivity deposits were fabricated simultaneously on 12.7 mm dia. graphite rods, which allowed for deposit-substrate delamination post-deposition. Average thickness of deposits was 250 μ m for Seebeck and electrical conductivity measurements, and 0.932, 1.400, and 0.854 mm for thermal conductivity measurement, respectively. All the deposits were prepared on preheated substrates with raster speed and standoff distance of the plasma torch set to 1000 mm s −1 and 100 mm, respectively. In-flight particles' temperature and velocity were monitored by using particle diagnostic sensor (AccuraSpray-G3, Tecnar Automation Lte.)(See Table 1). Deposits were also deposited on the In-situ coating property (ICP) sensor, which allows simultaneous measurement of substrate curvature and substrate temperature 42 . Microstructure and Phase Analysis. Cross-section microstructure of as-sprayed deposits and surface microstructure of splats and fractured deposits were investigated by field-emission scanning electron microscopy (FE-SEM, LEO 1550, Zeiss). X-ray diffraction of the feedstock, and deposits was carried out on X-ray diffractometer (XRD, D8 Discover, Bruker) in vertical Bragg-Brentano geometry (2.5° Soller slits in both primary and secondary beam and 0.5° divergence slit in primary path) with filtered CuKα radiation (Ni β filter in secondary path). Since linear 1D detector was used, beam knife was placed above the samples in order to minimize the detection of air scattering. Additional measurement using High resolution XRD (Ultima III, Rigaku) was performed, with a Cu-Kα radiation source at 40 kV and 44 mA between 2 θ values of 10° and 80°. Thermogravimetric analysis was performed in order to measure stoichiometry of powder, and the deposits using a thermal balance (TG/DSC, STA 449C, Netzsch) while heating TiO 2−x deposits up to 1073 K at 10 K min −1 , holding it for 30 minutes in air to fully oxidize. The weight of the TiO 2−x deposits was approximately 30 mg. Thermoelectric Characterization. Voltage and electrical resistivity of the TiO 2−x deposits were measured using digital multimeter (Keithley 2700). Seebeck Coefficient was calculated using the measured voltage generated from temperature gradient through each end of TiO 2−x deposits, heating one end of the substrate with propane gas torch while cooling the other end with cold air gun (Vortex). Pt/chromel wire was measured as a reference for Seebeck coefficient measurement shown in Fig. 3(b). The deposits were junction with Pt wires using Ag paste for the measurements. Data obtained with our technique showed good match to benchmark literature information on these materials. K-type thermocouples were directly attached on each end of deposit surface to measure temperature. Electrical conductivity was measured by four-point probe technique from room temperature to 750 K. More about this measurement can be found in Han S. et al. 54 . Thermal conductivity was measured using laser flashing method (FLASHLINE TM SYSTEM X-PLATFORM TM , Anter Corporation) from room temperature to 750 K. Of those measurements, density was measured using the Archimedes method. Electrical conductivity and Seebeck coefficient were measured in-plane direction and thermal conductivity was measured in through-thickness direction. This is because it is very difficult to accurately measure in-plane thermal conductivity of sprayed materials as they have a very small dimension in the thickness axis. Our past work 49 has shown that there is about 10-20% discrepancy in the two orientations, but they tend to be self-consistent, that is if the conductivity in one orientation is higher so is the other orientation. This is because plasma sprayed structures are built by thousands of splats and chaotic assembly to some extent normalizes the anisotropy correlation to microstructure.