Investigation of Phase Segregation in p-Type Bi0.5Sb1.5Te3 Thermoelectric Alloys by In Situ Melt Spinning to Determine Possible Carrier Filtering Effect

One means of enhancing the performance of thermoelectric materials is to generate secondary nanoprecipitates of metallic or semiconducting properties in a thermoelectric matrix, to form proper band bending and, in turn, to induce a low-energy carrier filtering effect. However, forming nanocomposites is challenging, and proper band bending relationships with secondary phases are largely unknown. Herein, we investigate the in situ phase segregation behavior during melt spinning with various metal elements, including Ti, V, Nb, Mo, W, Ni, Pd, and Cu, in p-type Bi0.5Sb1.5Te3 (BST) thermoelectric alloys. The results showed that various metal chalcogenides were formed, which were related to the added metal elements as secondary phases. The electrical conductivity, Seebeck coefficient, and thermal conductivity of the BST composite with various secondary phases were measured and compared with those of pristine BST alloys. Possible band alignments with the secondary phases are introduced, which could be utilized for further investigation of a possible carrier filtering effect when forming nanocomposites.


Introduction
Thermoelectric technology has attracted attention for its use in solid-state cooling and energy harvesting because it can convert heat directly into electricity. The energy conversion efficiency of thermoelectric materials is limited by the dimensionless figure of merit, zT = [S 2 ·σ/(κ ele + κ latt )] × T, where S is the Seebeck coefficient, σ is the electrical conductivity, κ ele is the electronic thermal conductivity, κ latt is the lattice thermal conductivity, and T is the absolute temperature [1][2][3][4]. Accordingly, a high zT value can be achieved by increasing S 2 ·σ and reducing the thermal conductivities (κ ele and κ latt ). However, these thermoelectric parameters are generally interdependent. Therefore, based on a comprehensive analysis of the fundamental mechanisms, thermoelectric materials should be manipulated to achieve optimal thermoelectric properties. In recent years, many approaches have been improved using zT values. Control of the carrier concentration, resonance doping, band engineering, and carrier filtering effects have been suggested for enhancing S 2 ·σ [5][6][7][8]. However, other strategies exist for reducing thermal conductivity. These include inducing point defects, dislocation arrays, or nanostructures by increasing phonon scattering [9][10][11][12].
Of these approaches, carrier energy filtering can effectively improve zT by increasing S and S 2 ·σ. This type of filtering is achieved by energy barriers at heterointerfaces arising from band bending between the thermoelectric matrix and secondary phases [13][14][15], which induce strong energy dependence on the carrier relaxation time. When proper phase segregation is introduced in thermoelectric materials, the carrier energy filtering effect can be achieved, thereby enhancing the thermoelectric performance through low-energy carrier scattering by potential heights formed at heterointerfaces [13,14]. In addition, phonon scattering can be strengthened by the segregated phases to reduce κ latt .
Experimental evidence of S enhancement by the carrier filtering effect has been reported with various thermoelectric nanocomposites. Dou et al. reported an improvement in S of approximately 20%, as compared with that of the Bi 0.5 Sb 1.5 Te 3 matrix, which originated from the energy filtering of carriers [15]. Even more noticeable enhancements in S were observed in Sb/SbTe nanocomposites by Zhang et al. [16]. Fan et al. showed that the formation of nano-inclusions through melt spinning could lead to favorable conditions for thermoelectric applications [17]. Recently, Jiang et al. reported noticeable maximum zT values of 1.56 at 400 K by inducing PbSe nanocomposites with suppressed lattice and bipolar thermal conductivities that effectively inhibit minor charge carriers [18].
In this study, we investigated in situ phase segregation behavior during melt spinning with various metal elements, including Ti, V, Nb, Mo, W, Ni, Pd, and Cu, in p-type Bi 0.5 Sb 1.5 Te 3 (BST) thermoelectric alloys, which could be utilized for further investigation of a possible carrier filtering effect. The possible band alignments with secondary phases are presented with their measured thermoelectric properties.

Experimental Section
To prepare a set of samples of Bi 0.5 Te 1.5 Se 3 (M) 0.1 (M = Ti, V, Nb, Mo, W, Ni, Pd, and Cu), all high-purity elements (Bi (99.999%, 5 N plus), Te (99.999%, 5 N plus), Se (99.999%, 5 N plus), and metal elements) were stoichiometrically synthesized by subsequent conventional melting and quenching techniques. The synthesized samples were blended using a ballmilling process (8000D, SPEX SamplePrep, Metuchen, NJ, USA) for 5 min. We conducted rapid solidification through melt spinning (Cu wheel rotation, 3600 rpm). The molten ingot was sprayed under a pressure of 0.03 MPa in an argon atmosphere. Using an agate mortar, the ribbons from the melt-spinning process were pulverized. Finally, the powders were sintered at 430°C by spark plasma sintering (SPS) for 5 min under a pressure of 50 MPa.
To analyze the crystalline phases of the samples, X-ray diffraction (XRD, D8 Discover, Bruker, Billerica, MA, USA) was performed at room temperature. Then, the temperaturedependent σ and S parameters were measured simultaneously over the temperature range between room temperature and 480 K using a ZEM-3 measurement system (Advanced-RIKO, Yokohama, Japan) perpendicular to the SPS pressing direction. The κ values were also computed from the theoretical density (ρ s ), heat capacity (C p ), and thermal diffusivity (D) in the same direction (κ = ρ s × C p × D). Then, the diffusivities λ were measured by the laser flash method (LFA 467, Netzsch, Wittelsbacherstraße, Germany). Figure 1 shows the XRD patterns of the experimental samples of Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = Ti, V, Nb, Mo, W, Ni, Pd, and Cu). Each diffraction peak commonly showed its own matrix phase (Bi 0.5 Sb 1.5 Te 3 , JCPDs PDF #49-1713) with the secondary phases, which was related to the added metals. The observed secondary phases were TiTe 2 , VTe 2 , NbTe 2 , MoTe 2 , W, NiTe 2 , PdTe 2 , and Cu 4 Te 3 for various added metal elements (Ti, V, Nb, Mo, W, Ni, Pd, and Cu, respectively). Most secondary phases were formed as dichalcogenides, whereas the addition of Cu caused Cu 4 Te 3 to form. The addition of W did not lead to the formation of compounds.

Band Bending at Heterointerfaces
The band alignment schematics at the heterointerfaces between BST and the secondary phases are shown in Figures 2 and 3. Figure 2 shows the energy bands of the Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = Nb, Ni, W, Ti, and V) samples, whose secondary phases (NbTe 2 , NiTe 2 , W, TiTe 2 , and VTe 2 ) were metallic. The work functions of NbTe 2 , NiTe 2 , W, TiTe 2 , and VTe 2 were 4.62, 4.44, 4.5, 4.86, and 4.92 eV, respectively [19][20][21]. Possible carrier filtering barriers were formed in Bi 0.5 Sb 1.5 Te 3 Ni 0.1 , Bi 0.5 Sb 1.5 Te 3 Nb 0.1 , and Bi 0.5 Sb 1.5 Te 3 W 0.1 , with NbTe 2 , NiTe 2 , and TiTe 2 for hole transport, respectively. Their energy barrier heights were 0.08, 0.26, and 0.20 eV for NbTe 2 , NiTe 2 , and TiTe 2 , respectively. For Bi 0.5 Sb 1.5 Te 3 V 0.1 and Bi 0.5 Sb 1.5 Te 3 Ti 0.1 , no energy barrier was expected with the secondary phases of VTe 2 and TiTe 2 , respectively.   samples, whose secondary phases (PdTe 2 , MoTe 2 , and Cu 4 Te 3 ) were semiconducting. The band gap (E g ), Fermi level (E f ), and electron affinity (χ) of PdTe 2 , MoTe 2 , and Cu 4 Te 3 were taken from the literature [19,[22][23][24]. The χ of Bi 0.5 Sb 1.5 Te 3 is 4.50 eV and the E g is 0.2 eV [25]. Given the band structure of PdTe 2 , the band diagram of Bi 0.5 Sb 1.5 Te 3 Pd 0.1 is presented in Figure 3a. A possible filtering barrier of 0.04 eV in Bi 0.5 Sb 1.5 Te 3 Pd 0.1 is shown. In the case of Bi 0.5 Sb 1.5 Te 3 Mo 0.1 , because of the relatively wide E g as compared to that of BST, an expected band diagram is given in Figure 3b. It formed a hole barrier of 0.26 eV, whereas the electron filtering barrier reached 0.43 eV. In the case of Bi 0.5 Sb 1.5 Te 3 Cu 0.1 , no band data were available for Cu 4 Te 3 . Because a quantitative illustration of the band diagram was unavailable, the illustration is shown with no quantitative values. Table 1 lists the work functions of E g and χ for the segregated phases.

Electronic Transport Properties (σ, S, and S 2 ·σ)
The temperature dependences of σ for Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = Ti, V, Nb, W, and Ni) are shown in Figure 4a. The σ value of the pristine BST sample was 767 S/cm at room temperature and decreased to 448 S/cm with increasing temperature. For the Bi 0.5 Sb 1.5 Te 3 Ti 0.1 and Bi 0.5 Sb 1.5 Te 3 V 0.1 samples, which did not form energy barriers at the heterointerfaces (Figure 2), the decreasing slope of σ with increasing temperature was much lower than that of the pristine BST, whereas the σ of Bi 0.5 Sb 1.5 Te 3 Ti 0.1 and Bi 0.5 Sb 1.5 Te 3 V 0.1 generally decreased and increased, respectively, as compared with that of the pristine BST. Note that these two samples did not form adequate energy barriers for hole carrier filtering ( Figure 2  S for the Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = Ti, V, Nb, W, and Ni) samples is given as a temperaturedependent function in Figure 4b. The S values of all the samples were suppressed as compared with that of the pristine BST. The BST sample had a peak S magnitude of 209 µV/K at 360 K and decreased to 171 µV/K with increasing temperature (at 480 K). At room temperature, S decreased to 194, 142, 117, 99, and 63 µV/K for the W-, Ni-, V-, Nband Ti-added samples, respectively. Figure 4c shows the temperature dependence of S 2 ·σ (power factor) for the Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = Ti, V, Nb, W, and Ni) samples. Bi 0.5 Sb 1.5 Te 3 W 0.1 showed very similar power factor values to the pristine BST sample over the entire temperature range. The addition of W did not form the telluride, which generally only affects the electric transport of the BST matrix. Otherwise, the power factors decreased to 2.34, 1.97, 1.08, and 0.19 mW/mK 2 for the Ni-, Nb-, V-, and Ti-added samples. With the addition of Ti and V, which did not form energy barriers at the heterointerfaces with metallic TiTe 2 and VTe 2 , σ and S decreased simultaneously, and the power factor was then reduced considerably. With the addition of Ni and Nb, which did form proper energy barriers at the heterointerfaces with metallic NiTe 2 and NbTe 2 , σ increased significantly, whereas S decreased. As a result, the power factors were moderately reduced. For the Nb-added samples, the power factors at high temperatures of 440 and 480 K were higher than that of the pristine BST. For Ni-and Nb-added samples, further experiments with smaller additions of metal (Bi 0.5 Sb 1.5 Te 3 (M) x (M = Ni and Nb, x ≤ 0.01) were conducted to investigate the possible carrier filtering effect [26]. With a small addition of x = 0.01, power factor enhancements were observed with an increase in the effective mass, suggesting that a possible carrier filtering effect occurred.   Figure 5c. The power factor of Bi 0.5 Sb 1.5 Te 3 Pd 0.1 decreased slightly as compared with that of the pristine BST. For the Pd-and Cu-added samples, the power factors decreased further, to 2.42 and 2.23 mW/mK 2 , respectively, at room temperature, and greater values were observed at higher temperatures above 400 K.

Thermal Conductivity (κ tot , κ elec , κ latt )
To further investigate the total thermal conductivity (κ tot ) behavior in Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = Ti, V, Nb, Mo, W, Ni, Pd, and Cu), we determined the κ tot values to be mainly binary parts of thermal conductivity, namely, κ ele and κ latt . They were calculated using the following equation: The κ ele values were calculated using the Wiedemann-Franz equation, as follows: where L is the Lorenz number (calculated as L = 1.5 + exp(−|S|/116)). L and S are treated as units in terms of 10 −8 WΩK −2 and µV/K, respectively [27]. The κ tot and κ latt values for the Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = Ti, V, Nb, Ni, and W) samples, as functions of temperature, are shown in Figure 6a,b, respectively. As shown in Equation (1), we computed the κ latt values by subtracting the κ ele values (which were calculated in advance) from the κ tot values. The κ latt values of the standard BST sample were increased from 0.99 to 1.42 W/mK as the measuring temperature increased. For the Bi 0.5 Sb 1.5 Te 3 Ni 0.1 sample, κ tot and κ latt increased. κ latt was significantly reduced for the Vand Nb-added samples. The addition of W did not form the telluride, which seemed to not affect the thermal conductivity of the BST matrix much. In all the samples that formed tellurides, except for the Nb-and Pd-added samples, some degrees of reduction in κ latt were shown due to the presence of secondary phases, as observed in Figure 6 [28]. However, adding Nb or Pd, which form NbTe 2 and PdTe 2 , respectively, increased the κ latt , or had little effect. At this stage, these different results cannot be elaborated. Further investigation into the possible carrier filtering effects of smaller amounts of Nb-and Pd-added Bi 2 Te 3 -based alloys showed a small degree of reduction in κ latt [26,29].  Figure 7a, the Ti-added samples, which showed a significantly reduced power factor due to the simultaneous reduction of σ and S, exhibited a considerably reduced zT. For the Ni-added sample, a lower zT was observed under all temperatures, which was mainly due to the increased κ tot and κ latt . Therefore, the Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = Ti and Ni) samples showed a lower zT over the entire temperature range. For the Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = V and Nb) samples, the zT at low temperatures decreased, whereas that at a high temperature (480 K) exhibited a slightly higher value as compared with the zT values of the pristine BST. The addition of W did not form any chalcogenides, which seemed to affect the thermal conductivity of the BST matrix. It showed an improvement in zT of approximately 5% as compared with that of the pristine BST. In Figure 7b, the Pd-added samples, which showed a moderately decreased power factor with slightly increased κ tot and κ latt , exhibited a reduced zT over the entire temperature range. In the range of 300-440 K, the Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = Mo, and Cu) samples had lower zT values than that of the pristine BST. However, from 440 K to 480 K, the Bi 0.5 Sb 1.5 Te 3 (M) 0.1 (M = Mo, and Cu) samples had slightly higher zT values than that of the pristine BST.

Conclusions
We investigated the in situ phase segregation behavior during melt spinning with various metal elements, including Ti, V, Nb, Mo, W, Ni, Pd, and Cu, in p-type Bi 0.5 Sb 1.5 Te 3 (BST) thermoelectric alloys. The observed secondary phases were TiTe 2 , VTe 2 , NbTe 2 , MoTe 2 , W, NiTe 2 , PdTe 2 , and Cu 4 Te 3 for various added metal elements (Ti, V, Nb, Mo, W, Ni, Pd, and Cu, respectively). The electrical conductivity, Seebeck coefficient, and thermal conductivity of the BST composite with various secondary phases were measured and compared with those of the pristine BST alloys. The possible band alignments with the secondary phases were introduced, which could be utilized for further investigation of a possible carrier filtering effect when forming nanocomposites.

Conflicts of Interest:
The authors declare no competing financial interest.