Atomic-resolution Interfacial Microstructure and Thermo-electro-magnetic Energy Conversion Performance of Gd/Bi_0.5Sb_1.5Te₃ Composites

2.1 Phase Constituent and Microstructure

The diffraction peak at around 32° for the mixed powders of Gd and BST is identified as the diffraction peak of the $\left(10\overline{1}1\right)$ plane in hexagonal Gd. However, it is worth noting that no diffraction peaks of Gd were observed in all the Gd/BST composites. This abnormal phenomenon should be caused by the interfacial reaction between Gd and BST. In order to investigate the interfacial reaction between Gd and BST, a microstructural analysis was conducted with EPMA at the micron scale. Figure 2 clearly shows that the interfacial reaction layer consists of Gd and Te without the trace of Bi and Sb. This phenomenon is consistent with previous findings reported in the literature.^[43] Moreover, it was observed that the average thickness (t_avg) of interfacial reaction layers gradually increases as the T_S increases, increasing from 7.8 to 13.8 μm (Figure 2b). These results suggest that the interfacial reaction range between Gd and BST matrix can be effectively controlled by tuning the T_S.

To further investigate the Gd/BST interfacial structures, the microstructures of the heterogeneous interfaces were characterized using scanning transmission electron microscopy (STEM) at the atomic scale. Figure 3a presents that a typical interfacial reaction layer was approximately 500 nm in thickness formed at the interface between Gd (white contrast) and BST matrix (black contrast). The elemental mappings (Figure 3) and energy-dispersive X-ray spectroscopy (EDS) line-scan (Figure 3f) of Te, Bi, Sb, and Gd clearly demonstrate that the interfacial reaction layer consists of Te and Gd at the nanometer scale. Transmission electron microscopy (TEM) image of the Gd/BST interface (Figure 3g) reveals that the reaction layer is composed of many nanocrystals. The corresponding selected area electron diffraction (SAED) pattern can be indexed to the GdTe structure with space group $\mathrm{Fm}\overline{3}\mathrm{m}$ (Figure 3h). The Gd:Te ratio from EDS result is close to 1:1, further confirming that the reaction layer is GdTe. Using atomic-resolution high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM), BST grain along the [2$\overline{1}\overline{1}$0] zone axis, GdTe nanoparticles (NPs), and Gd along the [0001] zone axis can be identified (as shown in Figure 3i,j). Furthermore, the interface between BST and GdTe layer appears relatively smooth, while the interface between Gd and GdTe layer is more irregular. These features indicate that the interfacial reaction is conducted by the diffusion of Te atoms into Gd during the sintering process. The resulted interfacial microstructures should have a large impact on the electrical and thermal transport properties by enhancing the carrier scattering and phonon scattering.^[44] In addition, Figure 3i shows a quite amount of BST quintuple layered (QL) structures and Bi (Sb) bilayers (BL) near the Gd/BST interface. It has been reported that Bi (Sb) bilayers generally appear in a Bi (Sb)-rich environment.^[45] Herein, the Bi (Sb)-rich environment near the Gd/BST interface is evidently caused by the interfacial reaction between Gd and diffused Te from BST. The evolution mechanism of the Bi (Sb) bilayer structure is illustrated in Figure 4. Weakly bonded Te atoms in Van der Waals gaps tend to diffuse and react with Gd and eventually leave a number of Te vacancies (${\mathrm{V}}_{\mathrm{Te}}^{\bullet \bullet }$) in BST QL. The Bi (Sb) atoms then occupy these Te vacancies and form ${\mathrm{Bi}}_{\mathrm{Te}}^{\prime }\left({\mathrm{Sb}}_{\mathrm{Te}}^{\prime }\right)$ cation antisite defects, which possess the lowest formation energy in a Te-poor environment. When the number of ${\mathrm{Bi}}_{\mathrm{Te}}^{\prime }\left({\mathrm{Sb}}_{\mathrm{Te}}^{\prime }\right)$ antisite defects exceeds the maximum limit, they transform into Bi (Sb) bilayers, which are filled in the van der Waals gaps of BST QLs. These characteristics can be observed in the atomic-resolution HAADF-STEM image of BST as shown in Figure 4. These p-type ${\mathrm{Bi}}_{\mathrm{Te}}^{\prime }\left({\mathrm{Sb}}_{\mathrm{Te}}^{\prime }\right)$ antisite defects and n-type Bi (Sb) bilayers should have impact on the electrical and thermal transport properties by influencing carrier concentration and enhancing phonon scattering.^[45]

2.2 Magnetocaloric Properties

MC cooling is realized by the cyclic magnetization and demagnetization of MC materials based on the MC effect. When the MC materials are magnetized, magnetic moments align in parallel, resulting in a reduction of magnetic entropy (S). This causes the materials to heat up due to an increased lattice entropy under adiabatic conditions. When the MC materials are demagnetized, MC materials absorb heat from the surrounding cooling media under isothermal conditions and MC cooling occurs. Consequently, the magnetic entropy change (ΔS) is a crucial parameter for the MC materials. The maximum ΔS (ΔS_M) is typically observed near the T_c as there is a dramatic change in magnetic order.^[46] Magnetocaloric properties of Gd and Gd/BST composites are depicted in Figure 5. The M-T thermomagnetic curve suggests that the T_c of Gd is approximately 293 K (Figure 5a). The magnetic phase transition can be determined using the Arrott plot (M²-H/M curves). Based on the Banerjee criterion, a negatively sloped or S-shaped M²-H/M curve near T_c indicates a first-order magnetic phase transition, while a positive slope suggests a second-order magnetic phase transition.^[47] In the case of Gd, the Arrott plot displays a positive slope (Figure 5b), which confirms the characteristic of a second-order magnetic phase transition.

Figure 5c shows that the ΔS_M for Gd under 2.5 T is 5.99 J·kg⁻¹·K⁻¹. The RCP is calculated using the equation RCP = ∆S_M·δ_FWHM, where δ_FWHM is the full-width at half-maximum of the ΔS–T curve.^[49] Figure 5d plots the ∆S values of Gd/BST composites and 2%Gd under 2.5 T, while Table 1 summarizes the corresponding MC properties. It is noticeable that the T_c of all Gd/BST composites are approximately 293 K, which is consistent with that of Gd. Furthermore, the ∆S_M exhibits a monotonically decreasing trend with increasing T_S due to the promotion of interfacial reaction of Gd. The ∆S_M of HP713 under 2.5 T is 0.096 J kg⁻¹ K⁻¹, which has a 19% decrease compared to that of HP653. Overall, using the low-temperature high-pressure sintering approach, the MC performance of Gd/BST composites have been well retained.

2.3 Thermoelectric Properties

All samples exhibit positive α (Figure 6e), confirming the p-type conduction of Gd/BST composites. This observation is consistent with the Hall measurement results. The Seebeck coefficient (α) first increases as the temperature rises and then decreases due to intrinsic excitation. At room temperature, the α decreases with T_S, which results from the increase in n.^[53] Figure 6f shows that the PF of the all Gd/BST composites increases steadily with elevating T_S. Benefitting from significantly increased σ, the HP713 sample achieves the highest PF of 4.09 mW K⁻² m⁻¹ at 300 K.

The thermal transport properties and ZT of Gd/BST composites in the temperature range of 300–500 K are displayed in Figure 7. The total thermal conductivity (κ) decreases firstly and then increases with increasing the measuring temperature (Figure 7a). The decrease in κ is attributed to enhanced phonon scattering from lattice vibration, whereas the increase in κ originates from the bipolar thermal effect induced by intrinsic excitation. Figure 7b shows that the carrier thermal conductivity (κ_C), which is proportional to σ by κ_C = LσT, also exhibits an increasing trend with T_S. The trend of κ_L is similar to that of κ, as the contribution of κ_C to κ is small in this work. A higher sintering temperature leads to more ${\mathrm{Bi}}_{\mathrm{Te}}^{\prime }$ antisite defects and heterogeneous interfaces induced by serious interfacial reaction, which can strongly scatter phonons and consequently depress κ_L.^[54] Nonetheless, this effect is counteracted by the reduced phonon scattering stemming from the preferential orientation of {000 l} under the high-temperature sintering condition. Consequently, the κ_L decreases first and then increases as T_S increases. Due to the simultaneous optimization of electron and phonon transport properties, HP693 exhibits a maximum ZT of 1.27 at 300 K, which is 31% higher than HP653 (Figure 7d). In terms of ZT value and ∆S_M, this work is superior to previously similar works.^{[39, 40]} The ZT of La(Fe_0.92Co_0.08)_11.9Si_1.1/Bi_0.3Sb_1.7Te₃ reached 0.87 at room temperature and ∆S_M was only 0.00086 J kg⁻¹ K⁻¹.^[39] The ZT of MnCoGe/Bi_0.37Sb_1.63Te₃ reached 0.68 at room temperature and ∆S_M was 0.0043 J kg⁻¹ K⁻¹.^[40] It demonstrated that the low-temperature high-pressure SPS method used in this work is feasible to obtain both excellent TE and MC properties.

2.4 Cooling Performance

Two single-leg devices with the same dimension of 2.5 mm × 3.2 mm × 11.0 mm were fabricated with HP653 and HP693, which are named as HP653_D and HP693_D, respectively. Figure 8 shows the time-dependence cold-side temperature (T_C) and hot-side temperature (T_H) of two single-leg devices under different operating currents (I). When a current is applied, both T_H and T_C undergo an initially rapid change, followed by a gradual approach to the equilibrium. This phenomenon can be attributed to the lag effect caused by Fourier heat and Joule heat. The temperature difference between T_H and T_C is referred as the working temperature difference (∆T), while the temperature difference between the room temperature T_R (300 K) and T_C is referred as the cooling temperature difference (∆T_C). Both ∆T and ∆T_C are used to evaluate the cooling performance of the single-leg device. When the current is lower than the threshold current (2.0 A for HP653_D and HP693_D), both ∆T and ∆T_C monotonically increase with the increase of current. However, once the current exceeds the threshold, ∆T continues to increase with the current, while ∆T_C gradually decreases. Additionally, when the current exceeds the threshold, ∆T_C increases first and then decreases over time. The increase in ∆T_C originates from the Peltier effect which enables rapid refrigeration in a short time. The subsequent decrease in ∆T_C is attributed to the Fourier heat and Joule heat, which requires more time for heat transport. The time difference between these factors contributes to this phenomenon.^[55] Furthermore, under the threshold current, the ∆T_C of HP693_D stabilizes at 6.4 K, showing an increment of 12% compared to that of HP653_D. Figure 8 illustrates that both ∆T_C and ∆T of HP693_D are greater than those of HP653_D. Therefore, the enhanced ZT value results in enhanced cooling performance. Precise control of T_S effectively preserves both the MC effect and TE performance of composites, providing a viable strategy to fabricate thermo-electromagnetic materials with both excellent TE and MC properties for all-solid-state TE/MC hybrid cooling technology. Such TE/MC hybrid cooling technology might overcome the technique barriers as faced by the TE and MC cooling, such as low efficiency, low frequency, low cooling power density, and large heat loss, showing a large potential in the application field of food storage, distribution warehousing, and other refrigeration circumstances.