Roles of Cu in the Enhanced Thermoelectric Properties in Bi0.5Sb1.5Te3

Recently, Cu-containing p-type Bi0.5Sb1.5Te3 materials have shown high thermoelectric performances and promising prospects for practical application in low-grade waste heat recovery. However, the position of Cu in Bi0.5Sb1.5Te3 is controversial, and the roles of Cu in the enhancement of thermoelectric performance are still not clear. In this study, via defects analysis and stability test, the possibility of Cu intercalation in p-type Bi0.5Sb1.5Te3 materials has been excluded, and the position of Cu is identified as doping at the Sb sites. Additionally, the effects of Cu dopants on the electrical and thermal transport properties have been systematically investigated. Besides introducing additional holes, Cu dopants can also significantly enhance the carrier mobility by decreasing the Debye screen length and weakening the interaction between carriers and phonons. Meanwhile, the Cu dopants interrupt the periodicity of lattice vibration and bring stronger anharmonicity, leading to extremely low lattice thermal conductivity. Combining the suppression on the intrinsic excitation, a high thermoelectric performance—with a maximum thermoelectric figure of merit of around 1.4 at 430 K—has been achieved in Cu0.005Bi0.5Sb1.495Te3, which is 70% higher than the Bi0.5Sb1.5Te3 matrix.

lattice thermal conductivities and consequently lead to excellent TE performances. However, the peak zTs of these alloys usually appear around room temperature because of their small band gaps (less than 0.2 eV) [5]. At high temperature, the zTs are severely degraded due to the enhanced κ and quickly decreased S caused by the thermal-activated bipolar effect [6,7]. Thus, the applications of these materials for harvesting waste heat as TE power generators are greatly limited. Some effective strategies have been adopted to suppress this thermal-activated bipolar effect. Among them, introducing a tiny amount of Cu into p-type Bi 2 Te 3 -Sb 2 Te 3 -based alloys can effectively decrease the minor carrier (electron) density and significantly weaken the negative effects of minor carriers on the TE performance [8]. High zTs, about 1.4 at 400-500 K, have been successfully obtained in Cu 0.005 Bi 0.5 Sb 1.495 Te 3 . Furthermore, the TE modules based on the developed Cu 0.005 Bi 0.5 Sb 1.495 Te 3 material show a high energy conversion efficiency, about 6% under a temperature gradient of 217 K, superior to the modules using undoped materials.
In spite of the much enhanced TE performance, the roles of Cu in the electrical and thermal transports in p-type Bi 2 Te 3 -Sb 2 Te 3 -based alloys are still not completely clarified. Especially, the occupancy site of Cu atoms is a matter of ongoing debate. Currently, there are two different standpoints. One is that the Cu atoms are in the van der Waals gaps between the two adjacent Te layers in the crystal structure as intercalated ions to introduce excess electrons to the system. As early as 1970, McCarthy et al. [9] found that the external electric field could drive Cu ions into the van der Waals gap. In 2011, Liu et al. [10] proposed the presence of Cu ions in the van der Waals gap of Bi 2 Te 2.7 Se 0.3 via the characterization and analysis of electrical transport properties and lattice parameters. In 2014, theoretical calculations showed that the formation energy of Cu intercalation in p-type Bi 2 Te 3 is favorable [11]. Another point is that the Cu atoms prefer to enter the Sb sites due to the small electronegativity and radius difference between Cu and Sb [12,13]. Considering that the valence state of Cu ions is +2, its substitution in the Sb 3+ site is expected to introduce additional holes to the system, which is completely contrary to the first standpoint. Nevertheless, the substitution of Cu in the Sb sites is well supported by the enhanced σ and lowered S values obtained in experiments.
In this work, we carried out a further study on the occupancy sites of Cu atoms in p-type Bi 0.5 Sb 1.5 Te 3 materials. Investigations on the phase compositions, microstructures, electrical and thermal transports, and material stabilities show that the Cu atoms are doped at the Sb sites rather than entering the van der Waals gap in Bi 0.5 Sb 1.5 Te 3 . Then, the roles of Cu atoms on the TE properties were systematically analyzed.

Results and Discussion
Figure 1a,c shows the powder XRD patterns for nominal Cu-doped Cu x Bi 0.5 Sb 1.5-x Te 3 and nominal Cu-intercalating Cu x Bi 0.5 Sb 1.5 Te 3 samples. All the peaks are indexed with the standard Bi 0.5 Sb 1.5 Te 3 phase (PDF#49-1713). No obvious impurity phase is detected. We further refined the lattice parameters of these Bi 2 Te 3 -Sb 2 Te 3 -based alloys by using the Rietveld method. The results are shown in Figure 1b,d, respectively. If the Cu atoms go into the van der Waals gaps as intercalated ions, the lattice parameter a should be scarcely changed, but lattice parameter c should be greatly increased [10]. In contrast, if the Cu atoms occupy the Sb sites, both a and c should be decreased because the atomic radius of Cu (135 pm) is smaller than that of Sb atoms (145 pm). As shown in Figure 1b, the monotonous reduction in a and c with increasing Cu content strongly suggests that Cu atoms enter the Sb sites in the Cu x Bi 0.5 Sb 1.5-x Te 3 samples. Similar lattice contractions are also observed in nominal Cu-intercalating Cu x Bi 0.5 Sb 1.5 Te 3 samples (see Figure 1d). Thus, the Cu atoms must be more energy favorable at Sb sites than at the van der Waals gaps in these p-type Bi 2 Te 3 -Sb 2 Te 3 -based alloys. The occupying position of Cu atoms in Bi0.5Sb1.5Te3 samples can also be identified by transmission electronic microscope (TEM) characterization. Figure 2a,b displays the regular (009) plane images for Bi0.5Sb1.5Te3 and Cu0.005Bi0.5Sb1.495Te3, respectively. If Cu atoms go to the van der Waals gaps, the (00l) plane distance will be expanded, which is the scenario observed in Cu-intercalated n-type Bi2Te2.7Se0.3 by Liu et al. [10]. If Cu atoms go to the Sb sites, the (00l) plane distance will shrink. Thus, we compared the distance of 29 fringes in the (009) plane images for both Bi0.5Sb1.5Te3 and Cu0.005Bi0.5Sb1.495Te3. The total distance of these 29 fringes in Cu0.005Bi0.5Sb1.495Te3 is only 9.22 nm, which is quite smaller than that in Bi0.5Sb1.5Te3 (9.40 nm). This proves that Cu atoms enter the Sb sites rather than the van der Waals gaps in the CuxBi0.5Sb1.5-xTe3 samples, which is consistent with the contracted lattice parameters obtained by the above XRD refinement. Interestingly, these Cu atoms at Sb sites induce significant alternations on the microstructures. Figure 2c shows the low-magnification TEM image of the undoped Bi0.5Sb1.5Te3 matrix. The electron diffraction pattern obtained from the circumvented area demonstrates a typical single crystal-like character. This phenomenon is quite different from that in Cu0.005Bi0.5Sb1.495Te3. As shown in Figure 2d, the electron diffraction pattern obtained from the circumvented area in Cu0.005Bi0.5Sb1.495Te3 shows typical polycrystalline rings, which could be indexed with standard planes in Bi0.5Sb1.5Te3 system. These polycrystalline rings indicate that there are numerous nanograins inside this area. The high-magnification TEM image shown in Figure 2e further proves that the Cu-doped material is indeed composed of nanoscale grains with coherent grain boundaries. The creation of numerous nanograins should be caused by the doped Cu atoms at the Sb sites (see the signal of Cu elements detected by energy dispersive spectrometer in Figure 2f). The discrepancy in atomic radius between Cu 2+ (135 pm) and Sb 3+ (145 pm), as well as their different electronegativities and valence states, induce the formation of numerous lattice distortions (see Figure 2g,h). Once the lattice distortions exceed the permitted level, dislocations and grain boundaries will be created to release the strain energy, leading to the formation of numerous nanograins, as shown in Figure 2e. These lattice defects will strengthen the scattering to lowfrequency heat-carrying phonons, which is expected to greatly lower the lattice thermal conductivity. Theoretically, various kinds of defects-such as antisite defect, dislocation, and/or vacancies-may exist in Bi2Te3-Sb2Te3-based materials [14]. However, in this study, we used a long annealing process time to try to obtain samples near the equilibrium state. In such a case, the formation of various The occupying position of Cu atoms in Bi 0.5 Sb 1.5 Te 3 samples can also be identified by transmission electronic microscope (TEM) characterization. Figure 2a,b displays the regular (009) plane images for Bi 0.5 Sb 1.5 Te 3 and Cu 0.005 Bi 0.5 Sb 1.495 Te 3 , respectively. If Cu atoms go to the van der Waals gaps, the (00l) plane distance will be expanded, which is the scenario observed in Cu-intercalated n-type Bi 2 Te 2.7 Se 0.3 by Liu et al. [10]. If Cu atoms go to the Sb sites, the (00l) plane distance will shrink. Thus, we compared the distance of 29 fringes in the (009) plane images for both Bi 0.5 Sb 1.5 Te 3 and Cu 0.005 Bi 0.5 Sb 1.495 Te 3 . The total distance of these 29 fringes in Cu 0.005 Bi 0.5 Sb 1.495 Te 3 is only 9.22 nm, which is quite smaller than that in Bi 0.5 Sb 1.5 Te 3 (9.40 nm). This proves that Cu atoms enter the Sb sites rather than the van der Waals gaps in the Cu x Bi 0.5 Sb 1.5-x Te 3 samples, which is consistent with the contracted lattice parameters obtained by the above XRD refinement. Interestingly, these Cu atoms at Sb sites induce significant alternations on the microstructures. Figure 2c shows the low-magnification TEM image of the undoped Bi 0.5 Sb 1.5 Te 3 matrix. The electron diffraction pattern obtained from the circumvented area demonstrates a typical single crystal-like character. This phenomenon is quite different from that in Cu 0.005 Bi 0.5 Sb 1.495 Te 3 . As shown in Figure 2d, the electron diffraction pattern obtained from the circumvented area in Cu 0.005 Bi 0.5 Sb 1.495 Te 3 shows typical polycrystalline rings, which could be indexed with standard planes in Bi 0.5 Sb 1.5 Te 3 system. These polycrystalline rings indicate that there are numerous nanograins inside this area. The high-magnification TEM image shown in Figure 2e further proves that the Cu-doped material is indeed composed of nanoscale grains with coherent grain boundaries. The creation of numerous nanograins should be caused by the doped Cu atoms at the Sb sites (see the signal of Cu elements detected by energy dispersive spectrometer in Figure 2f). The discrepancy in atomic radius between Cu 2+ (135 pm) and Sb 3+ (145 pm), as well as their different electronegativities and valence states, induce the formation of numerous lattice distortions (see Figure 2g,h). Once the lattice distortions exceed the permitted level, dislocations and grain boundaries will be created to release the strain energy, leading to the formation of numerous nanograins, as shown in Figure 2e. These lattice defects will strengthen the scattering to low-frequency heat-carrying phonons, which is expected to greatly lower the lattice thermal conductivity. Theoretically, various kinds of defects-such as antisite defect, dislocation, and/or vacancies-may exist in Bi 2 Te 3 -Sb 2 Te 3 -based  [14]. However, in this study, we used a long annealing process time to try to obtain samples near the equilibrium state. In such a case, the formation of various defects is greatly suppressed. All our data show that the Cu atoms are mainly doped at the Sb sites and promote nanocrystallization. defects is greatly suppressed. All our data show that the Cu atoms are mainly doped at the Sb sites and promote nanocrystallization. The bismuth telluride-based materials possess a typical layered crystal structure, and thus the sintered polycrystalline samples usually show oriented plane alignment and anisotropic TE transport properties [15]. In this manuscript, we will just show the electrical and thermal transports along the direction parallel to the sintering press because this direction shows better TE properties. Figure 3a-c show the electrical transport properties for all CuxBi0.5Sb1.5-xTe3 samples. With increasing Cu content, the σ values are gradually increased. At 300 K, the σ for Cu0.005Bi0.5Sb1.495Te3 is about 13.7 × 10 4 Sm −1 , almost 3.5 times larger than that in the matrix compound. Correspondingly, the S values near room temperature are greatly reduced in Cu-containing samples. Moreover, in the CuxBi0.5Sb1.5-xTe3 system, the peak temperature of S is shifted to higher temperatures for the sample with greater Cu content. In p-type Bi0.5Sb1.5Te3 materials, if Cu atoms substitute the Sb atoms, the hole concentration should be enhanced because Cu has fewer electrons in the outermost orbital than Sb (see the dashed line in Figure 3c). This speculation is consistent with the measurement results (square symbols) presented in Figure 3c, which shows that the Hall hole concentration increases when increasing the Cu-doping content. For comparison, we also show the S, σ, and data for the nominal Cu-intercalated CuxBi0.5Sb1.5Te3 samples in Figure 3a-c. Furthermore, the measured electrical transports in CuxBi0.5Sb1.5Te3 go against the expectation that the Cu atoms enter the van der Waals gap and introduce additional electrons (0.65 electrons per Cu atom) to reduce and σ (see the dotted line in Figure 3c) [9]. Thus, the presented electrical transport properties further prove that Cu atoms are doped at the Sb sites in p-type Bi0.5Sb1.5Te3. These Cu-dopants can suppress the intrinsic excitation and shift its occurring temperature to high temperatures, thereby leading to the improved S values in Cu-containing samples above 500 K (see Figure 3b). The bismuth telluride-based materials possess a typical layered crystal structure, and thus the sintered polycrystalline samples usually show oriented plane alignment and anisotropic TE transport properties [15]. In this manuscript, we will just show the electrical and thermal transports along the direction parallel to the sintering press because this direction shows better TE properties. Figure 3a-c shows the electrical transport properties for all Cu x Bi 0.5 Sb 1.5-x Te 3 samples. With increasing Cu content, the σ values are gradually increased. At 300 K, the σ for Cu 0.005 Bi 0.5 Sb 1.495 Te 3 is about 13.7 × 10 4 Sm −1 , almost 3.5 times larger than that in the matrix compound. Correspondingly, the S values near room temperature are greatly reduced in Cu-containing samples. Moreover, in the Cu x Bi 0.5 Sb 1.5-x Te 3 system, the peak temperature of S is shifted to higher temperatures for the sample with greater Cu content. In p-type Bi 0.5 Sb 1.5 Te 3 materials, if Cu atoms substitute the Sb atoms, the hole concentration should be enhanced because Cu has fewer electrons in the outermost orbital than Sb (see the dashed line in Figure 3c). This speculation is consistent with the measurement results (square symbols) presented in Figure 3c, which shows that the Hall hole concentration p H increases when increasing the Cu-doping content. For comparison, we also show the S, σ, and p H data for the nominal Cu-intercalated Cu x Bi 0.5 Sb 1.5 Te 3 samples in Figure 3a-c. Furthermore, the measured electrical transports in Cu x Bi 0.5 Sb 1.5 Te 3 go against the expectation that the Cu atoms enter the van der Waals gap and introduce additional electrons (0.65 electrons per Cu atom) to reduce p H and σ (see the dotted line in Figure 3c) [9]. Thus, the presented electrical transport properties further prove that Cu atoms are doped at the Sb sites in p-type Bi 0.5 Sb 1.5 Te 3 . These Cu-dopants can suppress the intrinsic excitation and shift its occurring temperature to high temperatures, thereby leading to the improved S values in Cu-containing samples above 500 K (see Figure 3b). In bismuth telluride-based materials, the intercalated Cu atoms in the van der Waals gap usually form very weak interactions with the nearby Te atoms because the Te atoms do not have extra free electrons. McCarthy et al. [9] found that the external electrical field could easily drive the Cu ions into the van der Waals gap of p-type Bi2Te3. However, these Cu ions will quickly diffuse out from the material when the external electrical field is removed. Interestingly, in the present CuxBi0.5Sb1.5-xTe3 samples, the doped Cu atoms at the Sb sites have quite high stability under external electric field. We have stressed an external current of 24 Acm −2 on the Cu0.005Bi0.5Sb1.495Te3 sample at 500 K to check its stability. If Cu atoms are intercalated in the van der Waals gap, such a large current density will definitely drive the Cu ions out from the van der Waals gap and then change its resistivity. This method has been successfully applied to characterize the Cu-ion migration behavior in Cu2S ionic conductor [16]. As shown in Figure 3d, under a current density of 24 Acm −2 , the resistance of the Cu0.005Bi0.5Sb1.495Te3 sample is scarcely changed, proving that the Cu atoms have high stability under an external electric field. The Cu atoms occupying Sb sites could form strong chemical bonds with surrounding Te atoms, resulting in the observed high stability.
The tiny amount of Cu dopants in Bi0.5Sb1.5Te3 scarcely changes the band structure. In principle, the density-of-states effective mass m* that reflects band structure can be simulated by the single parabolic (SPB) model as shown below [17]: In bismuth telluride-based materials, the intercalated Cu atoms in the van der Waals gap usually form very weak interactions with the nearby Te atoms because the Te atoms do not have extra free electrons. McCarthy et al. [9] found that the external electrical field could easily drive the Cu ions into the van der Waals gap of p-type Bi 2 Te 3 . However, these Cu ions will quickly diffuse out from the material when the external electrical field is removed. Interestingly, in the present Cu x Bi 0.5 Sb 1.5-x Te 3 samples, the doped Cu atoms at the Sb sites have quite high stability under external electric field. We have stressed an external current of 24 Acm −2 on the Cu 0.005 Bi 0.5 Sb 1.495 Te 3 sample at 500 K to check its stability. If Cu atoms are intercalated in the van der Waals gap, such a large current density will definitely drive the Cu ions out from the van der Waals gap and then change its resistivity. This method has been successfully applied to characterize the Cu-ion migration behavior in Cu 2 S ionic conductor [16]. As shown in Figure 3d, under a current density of 24 Acm −2 , the resistance of the Cu 0.005 Bi 0.5 Sb 1.495 Te 3 sample is scarcely changed, proving that the Cu atoms have high stability under an external electric field. The Cu atoms occupying Sb sites could form strong chemical bonds with surrounding Te atoms, resulting in the observed high stability.
The tiny amount of Cu dopants in Bi 0.5 Sb 1.5 Te 3 scarcely changes the band structure. In principle, the density-of-states effective mass m* that reflects band structure can be simulated by the single parabolic (SPB) model as shown below [17]: Here, k B is the Boltzmann constant, e is the elementary charge, η is the reduced Fermi level, and λ is the scattering parameter and taken as −1/2 for electron-phonon scattering [18]. F j is the Fermi integrals calculated by Equation (2). In Equation (3), p is the hole concentration and taken as the measured p H value, and h is the Planck constant. Figure 4a presents the theoretical Pisarenko plot at 300 K by assuming m* = 1.3m e (m e is the inertia mass of electron). The experimental S data for Cu-doped Bi 0.5 Sb 1.5 Te 3 samples, Cd-doped samples, and some other Bi 0.5 Sb 1.5 Te 3 samples previously reported will fall on this calculated line [19][20][21][22][23], suggesting that all samples possess similar m*. This is reasonable, considering the very tiny amount of Cu-or Cd-doping content used in the present study. These dopants just shift the Fermi level downward without obviously changing the band structure near the Fermi level.
Here, kB is the Boltzmann constant, e is the elementary charge, η is the reduced Fermi level, and λ is the scattering parameter and taken as −1/2 for electron-phonon scattering [18]. Fj is the Fermi integrals calculated by Equation (2). In Equation (3), p is the hole concentration and taken as the measured pH value, and h is the Planck constant. Figure 4a presents the theoretical Pisarenko plot at 300 K by assuming m* = 1.3me (me is the inertia mass of electron). The experimental S data for Cu-doped Bi0.5Sb1.5Te3 samples, Cd-doped samples, and some other Bi0.5Sb1.5Te3 samples previously reported will fall on this calculated line [19][20][21][22][23], suggesting that all samples possess similar m*. This is reasonable, considering the very tiny amount of Cu-or Cd-doping content used in the present study. These dopants just shift the Fermi level downward without obviously changing the band structure near the Fermi level. The existence of numerous nanograins in Cu-doped samples is expected to strengthen the carrier scattering and reduce the carrier mobility ( ). Surprisingly, Figure 4b shows that the Cu-doped samples possess much higher than the undoped Bi0.5Sb1.5Te3. At room temperature, for Cu0.005Bi0.5Sb1.495Te3 is 182.6 cm 2 V −1 S −1 , which is about 70% higher than that of the undoped Bi0.5Sb1.5Te3. Especially, such enhanced values are observed over the entire measured temperature range. These enhanced carrier mobilities might benefit from the weakened ionized impurity scattering and acoustic phonon scattering. Generally, the total mobility μ, including these two scattering mechanisms, satisfy the Matthiessen's rule [24]: The existence of numerous nanograins in Cu-doped samples is expected to strengthen the carrier scattering and reduce the carrier mobility (µ H ). Surprisingly, Figure 4b shows that the Cu-doped samples possess much higher µ H than the undoped Bi 0.5 Sb 1.5 Te 3 . At room temperature, µ H for Cu 0.005 Bi 0.5 Sb 1.495 Te 3 is 182.6 cm 2 V −1 S −1 , which is about 70% higher than that of the undoped Bi 0.5 Sb 1.5 Te 3 . Especially, such enhanced µ H values are observed over the entire measured temperature range. These enhanced carrier mobilities might benefit from the weakened ionized impurity scattering and acoustic phonon scattering. Generally, the total mobility µ, including these two scattering mechanisms, satisfy the Matthiessen's rule [24]: where µ I and µ AC is the mobility for ionized impurity scattering and acoustic phonon scattering, respectively. Even in the Bi 0.5 Sb 1.5 Te 3 , the ionized impurity scattering still exists, because there are numerous intrinsic antisite defects (Sb Te ) or vacancies in the material due to their low formation energies [14]. Generally, the corresponding mobility (µ I ) of ionized impurity scattering could be expressed as [25]: Herein, ε is the dielectric constant and taken as 75 [26]. The relation between density-of-states effective mass m * and single valley effective mass m * s satisfies m * = N 2/3 v m * s , where N v is valence band degeneracy and taken as 6 for Bi 2 Te 3 -based materials [5,27]. N I is the ionized impurity concentration. Dimensionless f (L D ) in Equation 5 can be expressed as: where is reduced Planck constant. L D is Debye screen length, with the formula: The shorter L D means a smaller Coulomb potential around ionized impurity and thereby a weaker effect on the carrier transports [25]. Based on the Equations (5)-(7), µ I is simultaneously determined by the ionized impurity concentration N I and Debye screen length L D . As shown in Figure 4c, the L D value in the Bi 0.5 Sb 1.5 Te 3 sample is around 21.9 Å at 300 K. However, in the Cu x Bi 0.5 Sb 1.5-x Te 3 samples, the L D values are reduced to 12.3-15.2 Å, depending on the hole concentrations. Thus, although Cu-doping slightly increases N I , the reduced L D still leads to the weakened ionized impurity scattering to the carriers.
Besides the ionized impurity scattering, the acoustic phonon scattering to the carriers is also weakened in the Cu-doped samples. As shown in Figure 4b, the acoustic phonon scattering (µ H ∝ T −3/2 ) is the dominant scattering mechanism for all Cu x Bi 0.5 Sb 1.5-x Te 3 samples around room temperature. The mobility (µ AC ) for acoustic phonon scattering in the SPB model can be written as [27]: where ρ is density, ν is the longitudinal velocity of sound and taken as 2884 m/s [28], and E AC is the deformation potential, which is generally used to evaluate the intensity of electron-phonon interaction. ψ(η) in Equation (8) is a monotonous decreasing function with reduced Fermi level (η) or hole concentration. It can be expressed as: Herein, the increased hole concentration by Cu doping reduces ψ(η) and thus carrier mobility with the magnitude within 9%-25% as compared with undoped matrix. However, by using the µ H data at 300 K, Equations (8) and (9) (see Figure 4d) show that Cu-doping reduces the deformation Materials 2017, 10, 251 8 of 12 potential E AC from 18.7 eV to 14.2 eV, which could enhance mobility by more than 70%. Thus, these two effects show that the interaction between carriers and phonons is greatly lessened in Cu-doped samples. Combining the weakened ionized impurity scattering mentioned above, the enhanced carrier mobilities, observed in Figure 4b, in Cu-doped samples can be well explained. As shown in Figure 4c-d, the reduced L D and E AC can also account for the abnormal enhanced carrier mobilities in the Cd-doped Cd x Bi 0.5 Sb 1.5-x Te 3 system. Figure 5a displays the temperature dependence of total thermal conductivity κ for Cu-doped Cu x Bi 0.5 Sb 1.5-x Te 3 samples. The κ includes three parts, termed as the carrier thermal conductivity (κ e ), lattice thermal conductivity (κ L ), and bipolar thermal conductivity (κ b ). κ e is calculated according to Wiedemann-Franz law (κ e = LσT, where L is the Lorenz number and taken as 1.6 × 10 -8 V 2 K -2 ) [29]. In this system, the substitution of Cu at the Sb sites significantly improves the electrical conductivity by simultaneously enhancing p H and µ H . Thus, the κ e in Cu-doped samples is much higher than that in Bi 0.5 Sb 1.5 Te 3 matrix. Then, although the κ L values in Cu-doped samples are reduced, the total κ near room temperature is enhanced. Because Bi 0.5 Sb 1.5 Te 3 has a very narrow band gap, around 0.2 eV, electrons are easily excited from the valence band to conduction band to form hole-electron pairs. The recombination of these hole-electron pairs releases extra heat and induces the remarkable κ b in the Cu-free Bi 0.5 Sb 1.5 Te 3 sample above 350 K. Doping Cu into the system effectively suppresses the κ b via reducing minor carrier concentration (electrons) [8], which can account for the observed low κ values at high temperature in Cu-doped Bi 0.5 Sb 1.5 Te 3 samples.  Figure 5a displays the temperature dependence of total thermal conductivity κ for Cu-doped CuxBi0.5Sb1.5-xTe3 samples. The κ includes three parts, termed as the carrier thermal conductivity (κe), lattice thermal conductivity ( ), and bipolar thermal conductivity ( ). κe is calculated according to Wiedemann-Franz law (κe = LσT, where L is the Lorenz number and taken as 1.6  10 -8 V 2 K -2 ) [29]. In this system, the substitution of Cu at the Sb sites significantly improves the electrical conductivity by simultaneously enhancing and . Thus, the κe in Cu-doped samples is much higher than that in Bi0.5Sb1.5Te3 matrix. Then, although the values in Cu-doped samples are reduced, the total κ near room temperature is enhanced. Because Bi0.5Sb1.5Te3 has a very narrow band gap, around 0.2 eV, electrons are easily excited from the valence band to conduction band to form hole-electron pairs. The recombination of these hole-electron pairs releases extra heat and induces the remarkable in the Cu-free Bi0.5Sb1.5Te3 sample above 350 K. Doping Cu into the system effectively suppresses the via reducing minor carrier concentration (electrons) [8], which can account for the observed low κ values at high temperature in Cu-doped Bi0.5Sb1.5Te3 samples. Figure 5. Temperature dependences of (a) total thermal conductivity κ (and lattice thermal conductivity ) and (b) bipolar thermal conductivity for CuxBi0.5Sb1.5-xTe3 samples. The data for x = 0 and x = 0.005 are taken from [8]; (c) Low-temperature for Bi0.5Sb1.5Te3 and Cu0.01Bi0.5Sb1.49Te3. The solid lines are the fitting results by Debye model; (d) Contribution of phonon-phonon Umklapp scattering (U), point defects scattering (PD), and grain boundary scattering (GB) to the total for Bi0.5Sb1.5Te3 and Cu0.01Bi0.5Sb1.49Te3 at 300 K. The color-filled area represents the reduction of by enhanced U scattering in Cu0.01Bi0.5Sb1.49Te3.
The very low values approaching the theoretical minimum value (0.31 Wm −1 K −1 ) in Cu-doped Bi0.5Sb1.5Te3 samples are very interesting [30]. We fitted the low-temperature of Bi0.5Sb1.5Te3 and Cu0.01Bi0.5Sb1.49Te3 by the Debye model, in which is expressed by [31]: and the spectral lattice thermal conductivity κs can be expressed by [32]: The very low κ L values approaching the theoretical minimum value (0.31 Wm −1 K −1 ) in Cu-doped Bi 0.5 Sb 1.5 Te 3 samples are very interesting [30]. We fitted the low-temperature κ L of Bi 0.5 Sb 1.5 Te 3 and Cu 0.01 Bi 0.5 Sb 1.49 Te 3 by the Debye model, in which κ L is expressed by [31]: and the spectral lattice thermal conductivity κ s can be expressed by [32]: where average sound speed ν is taken as 2930 m/s [33]. In this equation, x = ω/k B T, and ω is the phonon frequency. The total relaxation time for Bi 2 Te 3 -based materials is given by: where θ D is Debye temperature and taken as 160 K [33]. τ B , τ P , and τ U are the relaxation time for grain boundary scattering (GB), point defect scattering (PD), and phonon-phonon Umklapp scattering (U), respectively [34]. d is the grain size, and P and U are the coefficients to evaluate the intensity of each phonon scattering mechanism. By fitting the low-temperature κ L , these parameters can be obtained (see Table 1). Figure 5d shows the integral area of reduced κ L by each individual scattering mechanism. Because the exotic Cu atoms interrupt the periodicity of lattice vibration (see the TEM results shown in Figure 2) and bring stronger anharmonicity, the integral area of reduced κ L by U process is greatly reduced. Thus, the total κ L in Cu 0.01 Bi 0.5 Sb 1.495 Te 3 is much lower than the Cu-free Bi 0.5 Sb 1.5 Te 3 .  Figure 6 shows the temperature dependence of TE figure of merit zT for Cu-doped Cu x Bi 0.5 Sb 1.5-x Te 3 samples. The undoped Bi 0.5 Sb 1.5 Te 3 matrix displays a maximum zT around 0.8 at 330 K. Due to the combined effects of enhanced hole concentration, improved carrier mobility, reduced lattice thermal conductivity, and bipolar thermal conductivity, the maximum zT is significantly improved to 1.4. In addition, the suppressed intrinsic excitation in Cu-doped Cu x Bi 0.5 Sb 1.5-x Te 3 successfully shifts the peak zT towards high temperatures of 430-470 K. Combining the good stability under electric field, these Cu-doped Cu x Bi 0.5 Sb 1.5-x Te 3 samples show great potential in applications such as low-grade waste heat recovery. combined effects of enhanced hole concentration, improved carrier mobility, reduced lattice thermal conductivity, and bipolar thermal conductivity, the maximum zT is significantly improved to 1.4. In addition, the suppressed intrinsic excitation in Cu-doped CuxBi0.5Sb1.5-xTe3 successfully shifts the peak zT towards high temperatures of 430-470 K. Combining the good stability under electric field, these Cu-doped CuxBi0.5Sb1.5-xTe3 samples show great potential in applications such as low-grade waste heat recovery.

Materials and Methods
High purity elements Bi (99.999%), Sb (99.999%), Te (99.999%), Cu (99.999%), and Cd (99.999%) were used to synthesize the samples according to the stoichiometric composition Cu x Bi 0.5 Sb 1.5-x Te 3 (x = 0.005, 0.01, 0.015,) and Cd x Bi 0.5 Sb 1.5-x Te 3 (x = 0.005, 0.01, 0.015). The samples with the compositions of Cu x Bi 0.5 Sb 1.5 Te 3 (x = 0.005, 0.01, 0.015) were also prepared in this study. These elements were firstly sealed into quartz tubes in vacuum, and then melt at 1373 K for 12 h. After quenching into a water bath, they were placed into furnace again and annealed at 673 K for 5 days. Subsequently, the grown ingots were ground to fine powders in agate mortar by hand. These obtained powders were then consolidated by spark plasma sintering (SPS) to form dense cylinders with a height of about 12 mm, which are used to characterize the TE properties along the directions parallel and perpendicular to the pressing. During the sintering procedures, the powders were filled into graphite die and held at 683 K for 10 min under a pressure of 50 MPa. High densities (>98% of the theoretical density) for all sintered samples were obtained.
X-ray diffractometer (XRD, D8 ADVANCE, Bruker Co. Ltd., Karlsruhe, Germany) was used to analyze phase purity for powder samples. The microstructures of the prepared powder samples were also observed by high-resolution transmission electronic microscope (HRTEM, JEOL 2100F, Tokyo, Japan). The chemical composition was analyzed by equipped energy dispersive spectrometer (EDS, Oxford Instruments plc, Oxfordshire, UK). Electrical conductivity (σ) and Seebeck coefficient (S) above the room temperature were measured by using commercial ZEM-3 (ULVAC Co. Ltd, Kanagawa, Japan) apparatus. The stability under large current density was evaluated by in situ measured resistance in a modified thermal dilatometer (DIL 402c, Netzch Co. Ltd., Selb, Germany). Thermal conductivity (κ) is the product of thermal diffusivity (D), heat capacity (C p ), and density (ρ). The D was measured by commercial instrument (LFA 457, Netzch Co. Ltd., Selb, Germany) based on laser flash method. The C p was measured by using differential scanning calorimetry (DSC 200 F3, Netzch Co. Ltd., Selb, Germany). The ρ was measured by the Archimedes method. The low-temperature Hall coefficient (R H ), σ, and κ were measured in Physical Property Measurement System (PPMS, Quantum Design, San Diego, CA, USA). R H was measured by sweeping magnetic field to 3 T in both positive and negative directions. The Hall carrier concentration is calculated by p H = 1/eR H , and the mobility (µ H ) can be estimated by µ H = σR H .

Conclusions
In this study, the occupancy site of Cu atoms in p-type Bi 0.5 Sb 1.5 Te 3 materials was studied and their roles on TE properties were illustrated. Via systematical characterizations on the lattice parameters, microstructures, electrical transports, and material stabilities, the possibility of Cu intercalation in p-type Bi 0.5 Sb 1.5 Te 3 materials has been excluded and the position of Cu is identified as doping at the Sb sites. These Cu dopants not only increase the hole concentration, but also improve the hole mobility by weakening the acoustic phonon scattering and ionized impurity scattering on carriers. Thus, significantly enhanced electrical conductivities are achieved in Cu-doped Bi 0.5 Sb 1.5 Te 3 materials. Furthermore, the Cu dopants can significantly reduce the lattice thermal conductivity and bipolar thermal conductivity by inducing nanostructural defects and suppressing the intrinsic excitation. Combining the excellent electrical transport properties and low lattice thermal conductivities, the maximum figure of merit zT around 1.4 at 430 K has been achieved in Cu-doped samples, almost 70% higher than undoped Bi 0.5 Sb 1.5 Te 3 matrix.