Sb Alloying for Engineering High‐Thermoelectric zT of CuGaTe2

Decades of studies on thermoelectric materials have enabled the design of high‐performance materials based on basic materials properties, such as bandgap engineering. In general, bandgap energies correspond to the temperature at which the peak thermoelectric performance occurs. For instance, CuGaTe2 with a relatively wide bandgap of 1.2 eV has its peak zT > 1 at > 900 K. On the other hand, the zT is usually very low (<0.1) for this material at room temperature. This severely limits its average zT and hence overall performance. In this study, a phase diagram‐guided Sb alloying strategy to improve the low‐temperature zT of CuGaTe2 is used, by leveraging on the solubility limits to control the formation of the microstructural defects. The addition of Sb simultaneously improves the electrical conductivity and decreases the lattice thermal conductivity. For a low‐temperature range of 300–623 K, this Sb‐alloying strategy enables the achievement of a record high average zT of 0.33. The strategy developed in this study targets the improvement of the low‐temperature range of CuGaTe2, which is rarely focused on for wide‐bandgap ABX2 compounds, opening up more opportunities for holistic performance improvements, potentially enabling ultrahigh‐performance thermoelectrics over a wide temperature range.


Introduction
Exponential growth in energy utilization, driven by advancements in electronic devices, has created problems in thermal management and waste heat.Thermoelectrics, enabling the interconversion of waste heat and electricity, has seen a resurgence in the recent few years in tandem with technological demands to tackle waste heat.At the device level, the efficiency of thermoelectric devices relies on two main factors: temperature gradient and materials performance zT.The zT represents a dimensionless figure of merit and can be defined as zT = S 2 σT/(κ L þ κ e ), where S, σ, κ L , and κ e represent Seebeck coefficient, electrical conductivity, lattice, and electronic thermal conductivity, respectively.Overall, in order to achieve high-performance devices, high zT over a wide temperature range is desirable, in addition to other factors such as compatibility factors, electrical, and thermal impedance matching. [1]Besides generating electricity from a thermal gradient, thermoelectric devices can also be used in reverse as solid-state cooling devices for on-chip thermal management of electronic devices and localized personal cooling for medical applications. [2]n recent years, thanks to the advancement in understanding the fundamentals of physical properties of materials, high zT has been reported in numerous material systems across different temperature ranges. [3]For achieving high zT, various physical strategies such as nanostructuring, band convergence, and scattering mechanisms engineering are commonly investigated. [4]In addition, to reduce lattice thermal conductivity, strategies such as phonon scattering, large lattice anharmonicity, complex crystal structure, and defect engineering via point Decades of studies on thermoelectric materials have enabled the design of highperformance materials based on basic materials properties, such as bandgap engineering.In general, bandgap energies correspond to the temperature at which the peak thermoelectric performance occurs.For instance, CuGaTe 2 with a relatively wide bandgap of 1.2 eV has its peak zT > 1 at > 900 K. On the other hand, the zT is usually very low (<0.1) for this material at room temperature.This severely limits its average zT and hence overall performance.In this study, a phase diagram-guided Sb alloying strategy to improve the low-temperature zT of CuGaTe 2 is used, by leveraging on the solubility limits to control the formation of the microstructural defects.The addition of Sb simultaneously improves the electrical conductivity and decreases the lattice thermal conductivity.For a lowtemperature range of 300-623 K, this Sb-alloying strategy enables the achievement of a record high average zT of 0.33.The strategy developed in this study targets the improvement of the low-temperature range of CuGaTe 2 , which is rarely focused on for wide-bandgap ABX 2 compounds, opening up more opportunities for holistic performance improvements, potentially enabling ultrahigh-performance thermoelectrics over a wide temperature range.4g,5] In terms of the temperature at which the peak zT occurs, thermoelectrics are generally categorized into low-, medium-, and high-temperature materials.For instance, at low temperatures (300-450 K), Bi 2 Te 3 and Mg 3 Sb 2 -based compounds have been the leading materials with a peak zT > 1. [6] For low-temperature applications, inorganic thermoelectric materials can also be fabricated into flexible thin-film devices for wearable body heat harvesting applications, [5h,7] while organic semiconductor materials can also be easily made into such devices due to their inherent flexibility. [8]n addition, at medium temperatures (450-900 K), GeTe, AgSbTe 2 , PbTe, SnS, and SnSe-based compounds have been reported with remarkable zT of >2. [9]At the high-temperature range (>900 K), Si-based materials such as SiGe and Pr 3 Te 4 lead the performance chart. [10]he ABX 2 family of compounds, where "A" is a monovalent metal cation, "B" is a trivalent metal cation, and "X" is a divalent chalcogenide anion, is an increasingly popular class of thermoelectric materials, where its earliest known use is for alloying with GeTe to form (GeTe) 0.85 (AgSbTe 2 ) 0.15 , which has been used since the 1950s for constructing radioisotope thermoelectric generators for deep space power source applications.Similar strategies of alloying GeTe with other ABX 2 materials like CuBiSe 2 , [9l] AgInSe 2 , [11] LiBiTe 2 , and LiSbTe 2 still remain popular in recent research.3e] Another popular subset of the ABX 2 materials is the copper diamondoid compound, where A = Cu, B = Ga or In, X = Se or Te, which usually have intrinsically higher roomtemperature lattice thermal conductivities than most traditional TE materials, with values such as 7.4, 5.4, and 4.6 W m À1 K À1 for CuGaTe 2 , CuInTe 2 , and CuInSe 2 .It is interesting to note that the lattice thermal conductivities of their Ag counterparts are significantly lower, with values such as %1.8-1.9W m À1 K À1 for Ag(In,Ga)Te 2 , and 0.86 W m À1 K À1 for AgInSe 2 , which is likely due to the low-lying optical phonon modes from the vibrations from the more weakly bound Ag atoms.Although CuGaTe 2 and most diamondoid compounds have higher room-temperature lattice thermal conductivities than AgSbTe 2 , their generally higher carrier mobilities also imbue them with promising charge transport properties.Moreover, their much wider bandgaps immunize them from intrinsic excitation and saturation of zT values at high temperatures, potentially allowing them to operate at a much wider temperature range.Nevertheless, in most materials, there is a drastic drop in zT values outside of the temperature range at which peak zT occurs.For instance, as a member of the ABX 2 family of compounds, CuGaTe 2 suffers from very-low room-temperature zT (< 0.1).This is despite the record-highest peak zT of 1.64 at 873 K being reported for that material system, through the alloying of isoelectronic elements In and Ag, together with inducing the formation of a complex nanosized strain domain structure.
In this work, we leverage on Sb alloying to improve the lowtemperature zT of CuGaTe 2 .Despite its limited solubility, the addition of Sb into CuGaTe 2 lowers its lattice thermal conductivity and at the same time drastically increases the electrical conductivity and power factor.The extra lone pair at the valence 5s subshell of Sb 3þ compared to a filled valence shell of Ga 3þ is hypothesized to be the underlying reason behind the lowered lattice thermal conductivity, in addition to point defect scattering and weaker bonding due to lattice expansion.Moreover, it is evident from the weighted mobility trend that due to the lower electronegativity difference between Sb-Te compared to Ga-Te, the introduction of Sb also resulted in higher mobility.This is consistent with the intuition that lower electronegativity differences lead to lighter effective mass and higher mobility.As a result, a high average zT of 0.33 is achieved between 300 and 623 K, which is the highest value ever achieved for this compound via Ga site alloying.The strategy reported here can be extended to design high-performance thermoelectric at low temperatures in other ABX 2 family of compounds, especially those with wide bandgaps.

Results and Discussions
Figure 1 shows the comparison of the zT achieved in this work versus the literatures so far.As evident from the temperature trend of zT in Figure 1a, majority of the literature reports are closer to the exponential increase of zT at high temperature.In contrast, the zT in this work shows a more gradual and linear increase from low to medium temperatures.As a result, the average zT over the low-to-medium temperature range (300-623 K) in this work is more than 2 times higher compared to other literatures, as shown in Figure 1b.
To investigate the underlying reason behind the drastic increase in average zT, the solubility of Sb and the effect of Sb alloying on the lattice parameters have to be investigated.As shown in Figure 2a and summarized in Table S1, Supporting Information, the introduction of Sb results in the appearance of minor peaks belonging to Sb 2 Te 3 (space group of P3m1, PDF#02-089-6185) while the dominant crystalline phase is the tetragonal CuGaTe 2 (space group of I42d, PDF#00-047-1454).As the Sb fraction increases, the Sb 2 Te 3 peaks get stronger.Furthermore, at very high Sb fraction (x = 0.7 and above), peaks belonging to Cu 2 Te (space group P6/mmm, PDF#04-002-0652) are observed while the original CuGaTe 2 peaks slowly diminish.This points to a phase decomposition of CuGa 1Àx Sb x Te 2 into mainly Sb 2 Te 3 and Cu 2 Te phases, leaving behind a small fraction of CuGaTe 2 .
In order to understand the solid solubility of Sb, it is useful to analyze the lattice parameters as a function of Sb fraction, as shown in Figure 2b.Upon increasing the Sb fraction, the lattice parameters of the main CuGaTe 2 phase increase.This could be attributed to the limited doping of Ga site with Sb brought about by larger cationic radii of Sb 3þ (76 pm) compared to Ga 3þ (62 pm) as well as the extra lone pair.It is evident from the refinement results that increasing the Sb fraction does not result in much higher doping values.Even at Sb fraction of 0.1, not all Sb goes into Ga sites (Table S1, Supporting Information), which means that the solid solubility of Sb into the CuGaTe 2 system is below 10%.
Microscopically, the phase decomposition inferred from Figure 2 and Table S1, Supporting Information, can also be observed in the SEM-EDS elemental Mapping image shown in Figure 3.At Sb fraction of 0.1 shown in Figure 3a, small regions of Sb-rich precipitates can be observed.With increasing Sb content from 0.1 to 0.5 in Figure 3b, clear secondary phases belonging to Sb-rich regions (corresponds to Sb 2 Te 3 ) can be observed.On the other hand, the distribution of Cu-rich and  Ga-rich regions still coincides, pointing to the remaining CuGaTe 2 phase.Furthermore, at high Sb fraction of 0.7 in Figure 3c, besides the Sb-rich regions, the Ga-and Cu-rich regions no longer coincide.At this composition, Cu-and Sb-rich regions seem to dominate, consistent with the X-ray diffraction (XRD) patterns in Figure 2a.
The underlying reason behind phase decomposition at high Sb fraction can be traced back to the phase diagram of Cu 2 Te-Sb 2 Te 3 , as shown in Figure 4a.The absence of vertical (intermetallic) line shows that CuSbTe 2 is not a stable phase and will always decompose into its constituent phases of Cu 2 Te and Sb 2 Te 3 .This observation is consistent with the fact that at high Sb fractions, only very small regions of CuGaTe 2 are observed in the midst of Cu 2 Te and Sb 2 Te 3 , as shown by the Cu:Ga:Te elemental quantification ratio of %25:25:50 in Figure 4b.
The investigation of the effect of Sb alloying in electronic transport properties is presented in Figure 5. Pristine CuGaTe 2 is a relatively wide-bandgap thermoelectric, with bandgap of %1.24 eV, and its crystal structure shows a tetragonal distortion. [12]Any dopants introduced into the cation site of CuGaTe 2 could cause variations in Cu─Te and Ga─Te bond length and strength, which result in changes in the crystal structure and transport properties. [13]In terms of defect formation energy, existing literature reports have shown that the lowest energy defect is the Cu site vacancy.This inherently results in slightly p-type nature in pristine CuGaTe 2 sample, [14g,15] where the hole concentration (n H ) of the pristine Sb 0 sample in this work is %2 Â 10 18 cm À3 , as shown in Figure 5a.Beyond pristine CuGaTe 2 , the introduction of Sb into the system beyond the solubility limit causes the formation of Sb 2 Te 3 sary phase, which is known to have a much higher hole concentration of 6.5 Â 10 19 cm À3 . [16]Consequently, this results in anion deficiency in the remaining CuGaTe 2 system, which energetically further enhances the concentration of Cu vacancies.As a result of these combined effects, Sb alloying generally increases the hole concentration by about two orders of magnitude, as compared to the pristine Sb 0 sample.According to a comprehensive review by Snyder et al., zT for most thermoelectric materials usually optimize in the carrier concentration range of 10 19 -10 21 cm À3 . [17]Since the hole concentration of the pristine Sb 0 sample in this work is smaller than the optimal range, Sb alloying generally helps to bring the hole concentrations of the samples to be within the optimal range.Contrary to the expectation that doping a pristine compound will reduce the hole mobility, the hole mobilities of the Sb-alloyed samples are generally much higher than that of the pristine Sb 0 sample.This can be explained by the processing methods used in this work, where a rather low sintering temperature of 380 °C was used because the Sb 0.5 sample was observed to begin melting when sintered above 380 °C.In order to limit the process variables in this study, the same sintering temperature was used for all samples, making the only process being the variable Sb content.For the low sintering temperature of 380 °C used in this work, which is insufficient to increase the density of pristine CuGaTe 2 powders to >95%, Sb alloying helps to improve the densification of CuGaTe 2 by inducing the formation of Sb 2 Te 3 sary phases, which act as a sintering aid as Sb 2 Te 3 has a much lower sintering temperature, due to its significantly lower melting point of 620 °C, as compared to the CuGaTe 2 main phase (m.p. = 867 °C). [18]herefore, the Sb-alloyed samples all have densities of >95%, which are higher than those of the pristine Sb 0 sample, which implies that they have improved contact between the polycrystalline grains.Moreover, the significantly higher hole mobility of pristine Sb 2 Te 3 (μ H % 240 cm 2 V À1 s À1 ), [16] as compared to pristine CuGaTe 2 (μ H % 100 cm 2 V À1 s À1 ), [19] will also facilitate better electrical transport in the Sb-alloyed samples, resulting in the higher hole mobilities observed in Figure 5a, as well as the significantly higher electrical conductivities observed in Figure 5b.However, for the Sb 0.7 to Sb 1 samples, the hole mobilities were reduced due to the presence of additional Cu 2 Te precipitates (Figure 2a), which are known to have low hole mobilities (μ H % 12 cm 2 V À1 s À1 ) and very high hole concentrations% (n H % 2.5 Â 10 21 cm À3 ). [20]This causes the free carriers in the Sb 0.7 to Sb 1 samples to be scattered not only by the low-mobility Cu 2 Te precipitates, but also by the enhanced carrier-carrier scattering from the significant increase in hole concentration.showing the absence of stable CuSbTe 2 phase. [22]b) EDS mapping of CuGa 0.1 Sb 0.9 Te 2 showing minor regions of CuGaTe 2 phase remaining.Inset shows the EDS elemental quantification of the selected region marked by the white borders.
As a result of the significant increase in hole concentration upon Sb alloying, the Seebeck coefficient for all Sb-alloyed samples decreases.Figure 5c shows that the Seebeck coefficient for Sb 0.1 sample decreases drastically, reaching around 160 μV K À1 at 623 K, within the range for optimal power factor.In addition, the trend of Seebeck coefficient for all Sb-alloyed samples shows degenerate behavior (increasing Seebeck vs temperature).The Sb 0.1 samples also show significantly improved weighted mobilities over the entire temperature range as shown in Figure 5d.Not surprisingly, the high intrinsic electronic transport properties of Sb 0.1 samples also result in its highest power factor compared to other samples, as shown in Figure 6a.Remarkably, such a level of power factor is also higher than most other literatures on this compound at the same temperature range. [14]inally, the room-temperature total thermal conductivities (κ) of the samples generally increased with Sb content, primarily due to the introduction of excess free holes, which facilitates heat conduction via the flow of charge carriers.This effect can be isolated in the form of the electrical contributions to the thermal conductivities (κ e ), which are plotted in Figure S4b, Supporting Information.While the introduction of Sb drastically increases the electrical conductivity by around 2 orders of magnitude, the total thermal conductivity does not increase significantly, especially for Sb 0.1 sample (Figure 6b).Another noticeable trend for the total thermal conductivities (κ) is the abrupt change at %623 K, for the Sb 0.7-Sb 1 samples.In order to investigate further, differential scanning calorimetry (DSC) was performed, where the heat flow was recorded while the samples were heated from %350 to 675 K and cooled back to 350 K.The results of the Sb 0.7-Sb 1 samples are plotted in Figure S3, Supporting Information, where two intense endothermic peaks were observed in the range of 600-625 K, for all three samples when heated.Intense exothermic peaks were also observed in the same temperature range, for all three samples when cooled, indicating a reversible process like a phase transition.Since the Sb 0.7-Sb 1 samples are the only ones containing significant amounts of Cu 2 Te secondary phase in their XRD patterns (Figure 2a and Table S1, Supporting Information), the abrupt change in total thermal conductivities is most likely due to the Cu 2 Te secondary phase, which is known to have two intense phase transition peaks (β !γ and γ !δ phase transitions) around the temperature range of 600-625 K. [20] For the Sb 0.1-Sb 0.5 samples, which do not have any observable Cu 2 Te secondary phase peaks in their XRD patterns, there are no noticeable changes in their thermal conductivity trends.
In order to separate the effect of increased κ e , the lattice contributions to the thermal conductivities (κ L ) are plotted in Figure 6c.Since the κ L values of the identified secondary phases are lower than that of the Sb 0 sample, where κ L of Sb 2 Te 3 and Cu 2 Te are %2 and 0.46 W m À1 K À1 respectively, [16] they are expected to lower the overall κ L of the Sb-alloyed samples.As shown in Figure 6c, while κ L of the Sb 0.1 sample follows the expected trend of being slightly lower than that of the pristine Sb 0 sample across the entire temperature range, the higher-doped Sb 0.3-Sb 0.7 samples deviate from the expected trend by generally having higher κ L values.Referring to the same explanation regarding the higher hole mobilities of the Sb-alloyed samples, this may be due to the rather low sintering temperature of 380 °C, which is insufficient to increase the density of pristine CuGaTe 2 powders to >95%.However, the existence of Sb 2 Te 3 sary phases in the Sb-alloyed samples can function as a sintering aid due to its lower melting point as compared to the CuGaTe 2 main phase.Therefore, the Sb-alloyed samples all have densities >95%, which are higher than those of the pristine Sb 0 sample, which can partially account for the lower κ L values of the pristine Sb 0 sample.The lower density of the pristine Sb 0 sample also means there will be more pores to scatter phonons.In addition, from the calculated static lattice strain from Rietveld refinement of the XRD patterns, it was revealed that the static lattice strain of the CuGaTe 2 main phase gradually decreases from the Sb 0-Sb 0.3 samples, as shown in Figure 2c.This is because for the same sintering temperature, the fact that it is easier for the samples with higher Sb content to be densified implies that they have a higher material mobility during the sintering process, making it easier for dislocations to be annihilated in samples with higher Sb content, thus making it easier for the static lattice strain to be relaxed.Besides dislocation lines, static lattice strain can also come from the strain fluctuations induced by point defects, as well as the strain between grains aligned with a rotation angle which are induced by planar defects (e.g., grain boundaries and interfaces).5c] Despite the Sb 0.1 sample having a lower static lattice strain than the pristine Sb 0 sample, its κ L is still slightly lower due to the presence of some phonon scattering mechanisms, such as the well-distributed Sb 2 Te 3 precipitates (1-8 μm) shown in the Sb elemental map image of Figure 3a, as well as the Sb Ga Â substitutional point defects inferred from the increase in lattice parameters shown in Figure 2b.The Sb 2 Te 3 precipitates can provide more interfaces and the strain from the lattice parameter mismatch with the surrounding CuGaTe 2 main phase may also induce the formation of additional dislocations around the precipitate.In addition, due to the size and mass difference between the Sb 3þ ions (radius = 76 pm, mass = 121.76g mol À1 ) and Ga 3þ ions (radius = 62 pm, mass = 69.72 g mol À1 ), mass and strain field fluctuations are caused.The extra 5s 2 subshell lone pair of the Sb 3þ ion in the Ga site will also result in a different stereochemistry and bonding character, as compared to the filled valence shell of Ga 3þ , resulting in the loss of registry within the host lattice, as well as the changes in bond strain and bond anharmonicity, which also increase the Umklapp phonon-phonon scattering. [21]Therefore, the addition of defects with different length scales, namely the atomic scale for point defects and the nanoor microscale for Sb 2 Te 3 precipitates, has been known to be helpful for phonon scattering across a wider range of phonon frequencies.
While the Sb 0.3 and Sb 0.5 samples have similarly high relative densities of >99.5%, since the Sb 0.5 sample has a higher proportion of the Sb 2 Te 3 phase, which has a lower room temperature κ L than the CuGaTe 2 main phase, the Sb 0.5 sample also has a lower κ L near room temperature.However, the Sb 0.7 sample has a higher κ L near room temperature due to the larger average crystallite size of the Sb 2 Te 3 majority phase.However, for the Sb 0.9 and Sb 1 samples, the κ L values are extremely low due to the overwhelming majority of Sb 2 Te 3 phases, where such tetradymite phases have band structures that can deviate significantly from the SPB model, causing an underestimation in the Lorenz numbers (plotted in Figure S4a, Supporting Information, used to calculate the κ e values (see Supporting Information for equations).Subtracting lower κ e values from the same κ values would lead to an overestimation of the κ L values.
Consequently, the combination of high power factor and low lattice thermal conductivity results in higher zT in the Sb 0.1 sample over a wide temperature range compared to other samples, as shown in Figure 6d.Besides having the highest peak zT at 623 K, the Sb 0.1 sample also has the highest near-roomtemperature zT of %0.1.Other than the lower κ L , one of the major contributing factors for its high zT near room temperature is its well-optimized hole concentration of %7.27 Â 10 19 cm À3 , as compared to the pristine Sb 0 sample with a very low hole concentration of %10 18 cm À3 and the other samples with high hole concentrations >10 20 cm À3 .Such a carrier concentration range of 10 19 -10 20 cm À3 , which the Sb 0.1 sample falls under, is also usually found in optimized compositions of Bi 2 Te 3 alloys with record-breaking performances close to room temperature.In order to gauge the potential for further improvement, the theoretical minimum lattice thermal conductivity (κ min ) of CuGaTe 2 was calculated using the Debye-Cahill model (see Supporting Information for details) to be %0.47-0.49W m À1 K À1 across the entire temperature range and plotted as a dashed line in Figure 6c.Since the κ L of the best performing Sb 0.1 sample is %1.26W m À1 K À1 at 623 K, which is almost triple the value of the calculated amorphous limit, there is a lot more room for further improvements for the Sb 0.1 sample, which already has a high average zT of 0.33.In order to increase the structural disorder to be closer to that of an amorphous solid, entropy engineering can be employed to introduce a large amount of substitutional dopants, together with their accompanying defects.While the Sb 2 Te 3 precipitates in the Sb 0.1 sample are already quite well dispersed, they are mostly still in the microscale size range.To make them even more finely dispersed (nanoscale size range), such that they can scatter phonons more effectively, it is important to improve the homogeneity of the molten mixture during the furnace heating step.This may be possible with the use of a rocking furnace.

Conclusion
In summary, this work demonstrates a strategy based on isovalent alloying to simultaneously improve electronic and thermal transport properties in CuGaTe 2 .Essentially, Sb was conceptually chosen due to its larger ionic radii and the extra lone pair it possesses.Through varying Sb amount, it was found that Sb0.1 achieved the highest average zT of 0.33 across a broad working temperatures of 300-623 K.This indicates the possibilities of attaining stable, yet high, power conversion from waste heat of fluctuating temperature conditions.In addition, with the defect energetics knowledge in mind, the tendency of Sb alloying to form Sb 2 Te 3 sary phase was leveraged to drastically increase the electrical conductivity.The Sb alloying strategy also has practical benefits as it allows the material to be densified at a significantly reduced sintering temperature, which would save on production costs.Moving forward, similar strategy described in this work can be applied to other ABX 2 systems to achieve confluence of good electronic and thermal transport properties.For device applications, adding secondary phase nanostructures in the material by solid-state mixing methods can be considered, to serve as additional phonon scattering sites to push the lattice thermal conductivity closer to the theoretical limit or to improve the mechanical properties of these rather brittle materials.Further studies can also be conducted on the suitable barrier or metallization layer materials for CuGaTe 2 -based materials, in order to further advance the studies on their device applications.

Experimental Section
Synthesis: Polycrystalline CuGa 1Àx Sb x Te 2 samples (x = 0, 0.1, 0.3, 0.5, 0.7, 0.9, 1) were synthesized using copper powders (Skyspring Nanomaterials, 99.8%), gallium chunks (Shanghai Macklin Biochemical, 99.99%), antimony granules (Shanghai Macklin Biochemical, 99.999%), and tellurium chunks (Loyal Target Advanced Material Technology, 99.99%).For easy naming convention, sample names were referred to as Sb 0, Sb 0.1, and so on, in accordance to its Sb mole fraction in the material system.The elements were weighed according to the composition ratio and sealed in a quartz tube.The samples were heated to 1000 °C within 12 h and kept at the 1000 °C for 12 h in a muffle furnace.The samples were subsequently quenched in ice water.The obtained ingots were ground into powders using an agate mortar and pestle and sintered by spark plasma sintering (SPS) at 380 °C with a pressure of 60 MPa and a holding time of 5 min.
Characterization and Thermoelectric Property Measurements: The roomtemperature crystal structures were investigated by Bruker D8 Advance with a Cu Kα source (1.5418 Å).The data were collected in the 2θ range of 20-80°with the step size of 0.02°.The XRD patterns were refined by Rietveld method using GSASII software.The background was modeled using either Chebyshev or cosine functions.Scanning electron microscopy (SEM, microscope equipped with an energy-dispersive spectroscope [EDS]  for compositional analysis) was used to investigate the structural and chemical properties of the samples.
The samples obtained by SPS sintering were cut into cuboids, the size of which was 10 mm Â 3 mm Â 2 mm.After that, the electrical conductivity and Seebeck coefficient were measured by ZEM-3 (ULVAC Co. Ltd.) apparatus under Helium atmosphere.The sintered sample was cut into square flat pieces with side length of 6 mm and thickness of 2 mm.Thermal diffusivity was measured by laser flash method (NETZSCH LFA 457).Dulong-Petit heat capacity was used to estimate the thermal conductivity (C p = 3k B •N A •n/M W, where "n" is the number of atoms in the chemical formula and к = DρC p , where r is the total thermal conductivity, ρ is the density, and the C p is heat capacity).The usual errors of commercial measurement instruments were AE3%, AE4%, AE1%, and AE3% for electrical conductivity, Seebeck coefficient, density, and thermal diffusivity.Since ZT was determined by a combination of these measurements, the cumulative error for ZT was expected to be AE11%.The Lorenz numbers were calculated using the single-parabolic band (SPB) model and the electronic thermal conductivity was calculated based on Wiedemann-Franz law, as shown in Figure S4, Supporting Information.Charge carrier concentrations and mobilities of the CuGa 1Àx Sb x Te 2 samples were determined from the Hall coefficient measurement using the Van der Pauw method (Bio-Rad Microscience, Hall measurement system HL5500, United States).

Figure 1 .
Figure1.a) zT as the function of temperature and b) average zT over 300-623 K for CuGa 0.9 Sb 0.1 Te 2 compared to literature values.[14]

Figure 2 .
Figure 2. a) XRD patterns of all samples showing phase separation as Sb fraction increases.b) Tetragonal lattice parameters and c) calculated static lattice strain of CuGaTe 2 phase and crystallite size of Sb 2 Te 3 phase in the CuGa 1Àx Sb x Te 2 samples, plotted as a function of Sb fraction (x in CuGa 1Àx Sb x Te 2 ).

Figure 3 .
Figure 3. EDS mapping showing the evolution of microstructures and compositions with increasing Sb content from a) 0.1 to b) 0.5 and to c) 0.7.Scale bar represents 50 μm.

Figure 4 .
Figure 4. a) Phase diagram of Cu 2 Te-Sb 2 Te 3showing the absence of stable CuSbTe 2 phase.[22]b) EDS mapping of CuGa 0.1 Sb 0.9 Te 2 showing minor regions of CuGaTe 2 phase remaining.Inset shows the EDS elemental quantification of the selected region marked by the white borders.

Figure 6 .
Figure 6.a) Temperature-dependent power factor.b) Total thermal conductivity.c) Lattice thermal conductivity, and d) zT.Error bars for zT are plotted based on the reported expected error of AE11%.