Tunable Electrical Conductivity and Simultaneously Enhanced Thermoelectric and Mechanical Properties in n‐type Bi2Te3

Abstract The recent growing energy crisis draws considerable attention to high‐performance thermoelectric materials. n‐type bismuth telluride is still irreplaceable at near room temperature for commercial application, and therefore, is worthy of further investigation. In this work, nanostructured Bi2Te3 polycrystalline materials with highly enhanced thermoelectric properties are obtained by alkali metal Na solid solution. Na is chosen as the cation site dopant for n‐type polycrystalline Bi2Te3. Na enters the Bi site, introducing holes in the Bi2Te3 matrix and rendering the electrical conductivity tunable from 300 to 1800 Scm–1. The solid solution limit of Na in Bi2Te3 exceeds 0.3 wt%. Owing to the effective solid solution, the Fermi level of Bi2Te3 is properly regulated, leading to an improved Seebeck coefficient. In addition, the scattering of both charge carriers and phonons is modulated, which ensured a high‐power factor and low lattice thermal conductivity. Benefitting from the synergistic optimization of both electrical and thermal transport properties, a maximum figure of merit (ZT) of 1.03 is achieved at 303 K when the doping content is 0.25 wt%, which is 70% higher than that of the pristine sample. This work disclosed an effective strategy for enhancing the performance of n‐type bismuth telluride‐based alloy materials.


Introduction
The current low efficiency of energy utilization not only causes energy waste but also indirectly contributes to global climate issues. [1] Thermoelectric (TE) materials can realize the direct conversion between thermal energy and electrical energy in an environmentally friendly way, thus causing extensive concern in recent years. [2][3][4] Thermoelectric properties can be evaluated by the dimensionless figure of merit (ZT), which is defined as ZT = S 2 T/ , where S is the Seebeck coefficient, is the electrical conductivity, is the total thermal conductivity, and T is the absolute temperature. [5] From this equation, the high ZT value requires the coexistence of high electrical transport properties (S and ) and low thermal conductivity ( ). However, it is difficult to decouple S, , and because of the inner relationship among them. [6,7] In the past two decades, rapid progress has been achieved for thermoelectric materials due to the exploration As the only commercial thermoelectric material with high near-roomtemperature performance and excellent processability, bismuth telluride was intensively investigated during this period. [8][9][10] Due to the target of practical applications, the majority of research has focused on polycrystal bismuth telluride based on powder metallurgy. [9] The TE performance of polycrystal Bi 2 Te 3 was thereby enhanced by many optimization strategies, such as energy filtering, [11][12][13] band engineering, [14][15][16] texturing, [8,[17][18][19] and defect engineering. [10,[20][21][22] In particular, an outstanding ZT value of 1.86 was achieved in a p-type Bi 2 Te 3 alloy via a liquid phase sintering strategy. [20] In contrast, although it has also been attempted multiple times, the n-type Bi 2 Te 3 still behaves with much lower TE performance, particularly at near room temperature. One of the reasons for this outcome is that the weakened anisotropy by the polycrystalline process is more harmful to n-type Bi 2 Te 3 , which can be resolved by texturing to some extent. [8] However, because the repeatedly hot-forging process for texturing is time-consuming and strenuous, [17][18][19] other reasons for the inferior TE performance still deserve attention. It was reported that mechanical deformation can intensively produce a donor-like effect in Bi 2 Te 3 . [21,[23][24][25]  The significantly increased n-type charge carriers would thereby deviate the carrier concentration to the optimum value. [26] It is reasonable that reducing the carrier concentration would be available for powder-processed, n-type Bi 2 Te 3 , which is generally realized by alloying or doping. [9,10] For n-type Bi 2 Te 3 that is traditionally alloyed with Se to synergistically optimize the point defects and band structure, Pan et al. further regulated the Se content in powder-processed (Bi, Se) 2 Te 3 to reduce the carrier concentration. [26] However, although the peak ZT value was increased, the increased Se also enlarged the bandgap that moved the peak ZT to a high temperature. In addition, doping with the element of +1 or +2 valance in the cation site was theoretically effective in reducing the carrier concentration. However, it is conventionally considered that doping at the cation site can easily affect the conduction band structure, [27] which is adverse to maintaining the high transport properties. Therefore, only a few reports about n-type Bi 2 Te 3 have focused on cation site doping. However, this strategy is still worthy of further attention.
Considering that cation site doping may affect the band structure, [27] pristine Bi 2 Te 3 without band engineering by Se was selected as the research object. The material exhibited strong ntype properties due to the donor-like effect introduced by highenergy ball milling. Sodium doping was subsequently conducted on this n-type Bi 2 Te 3 because of its +1 valance in the cation site. In addition, the similar ionic radius of Na + (Na + : 102 pm) to Bi 3+ (Na + : 102 pm, Bi 3+ : 103 pm) would make it more stable in the cation site, unlike Cu + (73 pm), which would move to the Van de Waals gap of the Bi 2 Te 3 matrix. [28,29] The results show that a small amount of Na may not affect the band structure but can effectively reduce the carrier concentration. The modified defects can also synergistically optimize both charge carrier and phonon transport. The Na content of 0.25 wt% was determined to significantly increase the peak ZT value to 1.03 and move it closer to room temperature. The result of this work affords a significant guide for cation site doping in n-type bismuth telluride.

Results and Discussion
The X-ray diffraction patterns for the crushed powders of all samples are plotted in Figure 1a. All samples matched the standard PDF card (PDF#89-4302). No peaks for the impurity were detected even after doping with Na. The right side of Figure 1a demonstrates the magnified view of the (015) peaks. Note that the peak gradually shifted to a lower angle with an increase in Na content for x ≤ 0.2 and then shifted to a higher angle when the Na content further increased. These shifting peaks indicate that the lattice first expanded and then shrank with an increase in Na content, which can be understood by the behavior of the doped Na. The cation substitution of Bi 3+ by Na + could be expressed by: This reaction is the main reason of the donor concentration reduction. The introduction of Na is equivalent to making the sample Te-deficient. Under this condition, anion vacancies V ⋅⋅⋅ Te can be generated and rapidly occupied by Bi to form antisite defects Bi′ Te , thereby inflating the lattice, which happened in Zn doping cation sites as well. [30] These changes can be more precisely reflected by the refined lattice parameter, as shown in Figure 1b, which also reveals more details about the generated Bi′ Te . Lattice parameter c exhibited obvious expansion compared with the lattice parameters a and b. [31] It is well known that Bi 2 Te 3 consists of the quintuple-layer unit -Te(1)-Bi-Te(2)-Bi-Te(1)-, which interconnects by van der Waals forces. As explored by Hashibon et al., [32] the antisite defects Bi′ Te in Bi 2 Te 3 are more inclined to be generated at Te(1) sites. Therefore, the increased Bi′ Te at the Te(1) site with added Na may enlarge the van der Waals gap between the two Te(1) layers. When the Na content increases to a certain degree (x = 0.25), the formation energy of V ··· Te becomes lower than that of Bi′ Te due to the excessive Te deficiency. The generated V ··· Te then shrank the lattice and decreased the density. [9] Figure 2a shows the low-magnification transmission electron microscope (TEM) images of the Bi 2 Te 3 +0.25 wt% Na sample, and strip-like grains and pores are observed. Figure 2b shows the high-resolution, TEM image corresponding to the red rectangle marked in Figure 2a   lattice distortion around the dislocations. The existence of dislocations is in favor of scattering mid-frequency phonons to deteriorate the lattice thermal conductivity as well as strengthening the mechanical performance of materials. Figure 2f shows the high-resolution, TEM image corresponding to the red rectangle marked in Figure 2e, showing a grain boundary (marked with a white-dotted line). Two corresponding FFT images are given in Figure 2f1,f2), which correspond to the blue and red rectangles marked in Figure 2f, respectively, indicating that the precipitate is Na 2 TeO 4 with the direction of [- 3 1-2]. This result shows agreement with the EDS mapping shown in Figure S1, Supporting Information. The Na 2 TeO 4 nanoprecipitates may be beneficial to the low lattice thermal conductivity of Bi 2 Te 3 bulk materials.
The fractured surfaces of samples with different Na contents were characterized by scanning electron microscope (SEM) and are shown in Figure S2, Supporting Information. All samples exhibited intergranular fracture with good crystallinity. The grains showed a lamellar morphology with an average size of ≈5 μm. These results indicate that adding Na would not deteriorate the quality of the sintering process. In addition, the grains were all randomly oriented for samples with different Na contents, which suggested that adding Na was not assisted in the orientated alignment of grains. All samples contain some micropores at the grain boundaries, while no obvious changes in the number of micropores were observed. However, the density of the sample gradually increased with an increase in Na content, as shown in Figure  S2f, Supporting Information. The increased density might be attributed to the increased antisite defect amount, which was consistent with the theoretical density calculated by Miller et al. from the antisite point model. [9,33] Figure 3 shows the temperature-dependent, electrical transport properties for the samples with different Na contents. The specific values of the electrical transport properties at 323 K are exhibited in Table 1. The generally decreased with an increase in Na content, as shown in Figure 3a. The value was 1710 S cm -1 at 323 K for the sample without Na. When the Na content increased to x = 0.25, the value decreased to 683 S cm -1 at 323 K. The Na content monotonically decreased with an increase in temperature when x ≤ 0.2, indicating the property of degenerate semiconductors. However, the decline gradually slowed with an increase in Na content and even increased at high temperatures. The Seebeck coefficient (S) overall increased with an increase in Na content, especially in the low-temperature range, as shown in Figure 3b. The S-value at 323 K increased ≈88% from 118 μV K -1 to 222 μV K -1 when x increased from 0 to 0.25, the optimized carrier concentration is the main reason. [34] The results showed a trend of first increasing and then decreasing for x ≤ 0.2, of which the peak gradually shifted to a lower temperature with an increase in Na content. When x > 0.2, the peak shifted out of the measuring temperature range and caused S to show a monotonically decreasing trend. The variation in bandgap (E g ) can also be reflected by the shifted S peaks according to the Goldsmid-Sharp band gap estimation: [35] E g = 2e|S| max T max (2) where |S| max is the absolute value of the maximum S and T max is the temperature at which |S| max is observed. The calculated E g values are shown in Table 1. Note that E g gradually shrank with an increase in Na content. The reason for the decreased S with an increase in temperature was the intrinsic excitation of narrow bandgap semiconductors, which have been widely reported in multiple studies. [9,36] The more easily occurring intrinsic excitation caused by the narrowed band gap moved the S peak forward, which also explains the change in against the Na content. The change in and S correspondingly contributes to the variation in the power factor (PF), as shown in Figure 3c. The PF overall increased when x ≤ 0.2. When x = 0.25, the PF still demonstrates a significant increase at temperatures lower than 398 K, while it considerably decreases in the high-temperature range. A high PF value of 3377 μW m −1 K −2 was obtained at 303 K for the x = 0.25 sample. A further increase in the Na amount significantly decreased the PF. Therefore, the proper regulation of Na content can optimize the electrical transport properties. The intrinsic reasons for the variation in the electric transport properties can be clarified by the Hall carrier concentration (n H ) and Hall mobility (μ H ) shown in Figure 3d. Note that n H decreased with an increase in Na content. As known by the single parabolic band (SPB) model, S can be expressed as follows [3,37] : where k B is Boltzmann's constant,h is the reduced Planck constant, m * d is the density-of-state (DOS) effective mass, n is the carrier concentration, = E F /(k B T) is the reduced Fermi level, and s is the scattering factor, which is −1/2 for acoustic-phonon scattering. F j ( ) is the Fermi-Dirac distribution function, which can be expressed as follows: where = E/(k B T) is the reduced energy. According to Equations (1) and (2), the Pisarenko curve at 323 K is plotted and shown in Figure 3e, which illustrates that the effective mass (m * d ) matches 0.85 m 0 when x ≤ 0.25. The decreased carrier concentration via the movement of E F would be the main reason for the increased S. Although this decrease would also cause a decrease in , the increased carrier mobility would compensate for the decrement. The μ H increased with an increase in Na content when x ≤ 0.25, which may be attributed to the reduced interactional scattering by charge carriers. The sharply decreased μ H when x = 0.3 may be attributed to the excessively charged point defect that changed the scattering mechanism or perhaps the change band structured by Na doping. This speculation can be supported by the large deviation in the point to the fitting curve for the x = 0.3 sample.
Based on the fitted m * d of 0.85 m 0 , the theoretical PF can also be predicted by the single parabolic band (SPB) model. In the case of acoustic-phonon scattering, the relation between carrier mobility and the nondegenerate limits of drift mobility (μ 0 ) can be described as: Therefore, according to Equation (1) and = neμ, the curve for the carrier concentration dependence of the theoretical PF at 303 K is plotted in Figure 3f with the data points of the experimental values. Note that the points were shifted closer to the peak of the curve when x increased to 0.2. Further increasing x would move the points far from the peak. This result more clearly illustrated the role of Na doping in optimizing the carrier concentration. In addition, the point for the x = 0 sample is far below the curve, showing a much lower μ 0 . The increased μ 0 when increasing x to 0.2 indicates that the microstructure and the defect composition of the sample became more favorable for charge carrier transport.
In addition to the improved electrical transport properties, the thermal transport properties were also optimized by Na doping, as shown in Figure 4. generally decreased with an increase in Na content (≤0.25 wt%), especially near room temperature. was 1.97 W m −1 K −1 for the undoped sample at 323 K. After adding 0.25 wt% Na, this value decreased by nearly half to 1.02 W m −1 K −1 . The decrease in was contributed by both electronic transport and phonon transport. As shown in Figure 4b, because the electronic thermal conductivity ( e ) is strongly correlated with , it also decreased with an increase in Na content. As previously analyzed, the main reason should be the decreased carrier concentration. However, the decreased carrier concentration and narrowed E g also produced an enhanced bipolar effect when the temperature increased. As a result, the more Na that was added, the higher the bipolar thermal conductivity ( bipolar ) at high temperatures (Figure 4d). The x = 0.3 sample even shows a significant bipolar at 323 K, which differed from the other four samples. Therefore, the bandgap of the x = 0.3 sample should be reduced. The changed band structure conformed to the result of the Pisarenko curve shown in Figure 3e. The enhanced bipolar effect thereby caused an increase in at high temperatures, which was detrimental to the TE performance.
The lattice thermal conductivity ( L ) also significantly decreased with an increase in Na content at the whole measuring temperature, as shown in Figure 4d. Even the sample with a small amount of Na would show a sudden decrease in L . As known by the scattering mechanism of phonons, [38] phonon scattering by point defects occurs as a relaxation time of PD ≈ −4 , which focuses on high-frequency phonons. Therefore, the reduced L contributed by the point defect scattering of phonons should be more significant in the high-temperature range. The decrease in the whole temperature range indicated other reasons that must explain these findings. As exhibited in Table 1, the transverse sound velocity (v T ) of the samples monotonically decreased with an increase in x, indicating that the bond stiffness might be weakened by Na doping. This lattice softening might explain the reduced L . However, note that the decrement of v T did not match L and was a reasonable inference that L might be intensively scattered by other effects focused on low-or mid-frequency.
The representative scattering effect on low-or mid-frequency phonons was caused by grain boundary ( B ≈ 0 ) and dislocation ( DC ≈ -3 for dislocation cores and DS ≈ -1 for dislocation strain). Nevertheless, the similar grain size shown in Figure 2 indicates that grain boundary scattering was not a factor. Therefore, the introduction of dislocation, which reduced L near room temperature, might be a factor. The point defects introduced by Na doping were quite helpful for the formation of line defects such as dislocations. Many studies have reported that dislocations can significantly reduce the L of Bi 2 Te 3 -based materials near room temperature. [20,22,39] The internal strain induced by the dislocations could also have an important role in decreasing sound velocity (refer to Table 1). [22] Therefore, it is highly possible that phonon scattering and lattice softening by dislocations significantly reduce L after adding Na. Despite the increased bipolar , the -L substantially decreased due to the reduced L , especially at near room temperature, as shown in Figure 4e. This value was lower than the reported values decreased by adding excessive Te (green curve in Figure 4e). [40] Benefitting from the synergistic optimization of the electrical and thermal transport properties, the total TE performance was significantly improved. As shown in Figure 5, the peak ZT gradually increased with an increase in Na content when Na was added at less than 0.25 wt%. In addition, the temperature for the peak gradually shifted to a low temperature, which becomes more appropriate for bismuth telluride application scenarios. The maximum ZT ≈1.03 was achieved at 303 K for the sample with 0.25 wt% Na, which was more than 70% higher than the sample without Na. The samples with Na contents of 0.1 wt% and 0.2 wt%, although not the best performing, demonstrated an overall increase in the whole measured temperature range. Generally, an increase in the ZT value caused a significant increase in the average ZT value (ZT ave ), which can be obtained by the following formula:   [41][42][43][44][45][46] The electric current dependence of the c) output voltage, output power, and d) TE energy conversion efficiency for the sample with 0.25 wt% Na.
where T h and T c are the temperatures of the hot side and cold side, respectively, which are 323 and 398 K, respectively. Once a trace amount of Na was doped, the ZT ave significantly increased when x ≤ 0.25. A high ZT ave greater than 0.9 can be obtained when x = 0.2 and 0.25. The increase in ZT ave indicates the potential for practical application. A large range of Na contents means higher fault tolerance in industrial production. Thus, adding Na is confirmed to be a promising strategy for enhancing the TE performance of n-type polycrystal Bi 2 Te 3 materials. Figure 6 displays the maximum ZT value and minimum with the corresponding temperatures of the sample with 0.25 wt% Na in this work and pure Bi 2 Te 3 prepared by different means. Doping with 0.25 wt% Na decreased the minimum to the level of single crystals and was even similar to that of the sample prepared by ball milling. The maximum ZT value for the sample with 0.25 wt% Na was much higher than that of pure Bi 2 Te 3 , regardless of how it was prepared. The corresponding temperatures for the maximum ZT value and minimum for the sample www.advancedsciencenews.com www.advancedscience.com with 0.25 wt% Na were also maintained at 298 K as single crystals, which ensured a potential high efficiency at near temperature. Besides, the peak ZT near 300K for Bi 2 Te 3 -0.25 wt% Na specimen is comparable to that of other Bi 2 Te 3 -based composites as shown in Figure 6b. Figure 6c,d exhibit the output voltage, output power, and efficiency increased with an increase in temperature differences for a single leg sample with copper electrode and Ni transition layer. An appreciable efficiency of 4.2% was obtained with a temperature difference of only 176 K. The temperature difference of 78 K can also cause a high efficiency of 2.3%. The high efficiency obtained at such a low-temperature difference reflected the high TE performance in the low-temperature range, which is comparable to other Bi 2 Te 3 -based TE modules in the literature. [45,47] The 3D contour map of mechanical properties for the pure Bi 2 Te 3 sample and x = 0.25 sample were investigated by the nanoindentation method (nanoindentation force was 5 mN), as demonstrated in Figure 7. The hardness and Young's modulus showed large differences in the variety of positions for both samples, displaying mountain-like fluctuations in the map. A possible reason for the position dependence of mechanical properties might be the diverse orientation of the laminar grains. More notably, the hardness and Young's modulus of the sample with 0.25 wt% Na content were generally higher than those of the pure Bi 2 Te 3 sample. The average hardness and Young's modulus of the pure Bi 2 Te 3 sample were 1.29 and 39.26 GPa, respectively; after doping with 0.25 wt% Na, these values increased to 1.59 and 47.51 GPa, respectively. The hardness and Young's modulus are positively correlated with the strength of chemical bonds. However, the reduced v T in Table 1 reflected that the strength of chemical bonds was reduced after doping Na, indicating that there should be another reason for the enhanced mechanical properties. [48][49] It is clear to observe that the density of all samples shows an increment with adding Na (Figure S3e, Supporting Information). The high carrier concentration of pure sample shows that the defects of V Te were generated widely in matrix, this kind of evaporation could introduce pores into grains to decrease density. By introducing Na, the escaped Te in melting or sintering process could be trapped in bulk sample and reacted with extra Na element forming the compound Na 2 TeO 4 which could fill some of the pores located at grain boundaries. Compared with the pores and normal grain boundaries, the increased relative density and complex semi-coherent interfaces always increase the strength of bulk materials. Finally, the IFFT images www.advancedsciencenews.com www.advancedscience.com (Figure 2c,d) show that there are lots of dislocation cores inside of the grains, which also introduce obvious lattice distortion and strain to increase the ability of dislocation strengthening.

Conclusions
In summary, the TE performance of n-type Bi 2 Te 3 was significantly improved via Na doping. The defects modulated by this strategy synergistically optimized the electrical and thermal transport properties. Specifically, the properly moved Fermi level significantly improved the Seebeck coefficient. The correspondingly decreased carrier concentration not only weakened the interactional scattering between two charge carriers but also decreased the electronic thermal conductivity. The phonon scattering was overall enhanced in the whole temperature range, which may be attributed to the synergistic scattering effect via the point defects and the potentially introduced dislocations. A maximum ZT value of 1.03 at 303 K and a ZT ave of >0.9 for the range of 323-398 K were achieved with a Na content of 0.25 wt%. This improved peak ZT value greater than 70% compared with undoped n-type Bi 2 Te 3 reveals the great potential of Na for n-type Bi 2 Te 3 . The high TE energy conversion efficiency of 4.2% for a low-temperature difference of 176 K and the enhanced mechanical properties revealed the promising application perspective for Na doping.

Experimental Section
Sample Synthesis: The sample was synthesized by the melting method combined with spark plasma sintering. The experimental details were shown in the supporting information.
Characterization: The phase composition of the materials was analyzed by X-ray diffraction. The Seebeck coefficient and electrical conductivity were measured by ZEM-3. The thermal diffusivity D was measured by LFA 457. The microstructure of the samples was observed by scanning electron microscopy and transmission electron microscopy. The output power and efficiency of a single leg sample were measured by a thermoelectric conversion efficiency tester. The Vickers hardness and Young's modulus of the sample were measured by the nanoindentation method. The experimental details are shown in the supporting information.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.