Improvement of Contact and Bonding Performance of Mg₂Si/Mg₂SiNi₃ Thermoelectric Joints by Optimizing the Concentration Gradient of Mg

Abstract

Mg₂SiNi_3, is a topologically densely packed intermetallic compound (TCP-IMC) and an excellent diffusion barrier material between nickel and Mg₂Si-based thermoelectric material. However, even a little migration of the Mg atom from Mg₂Si to Mg₂SiNi₃ under the action of a driven force promotes the formation of an Mg deficiency region on the Mg₂Si side, which destroys the balance of point defects and leads to performance deterioration. In this work, by adjusting the chemical potential of Mg across the Mg_xSi₁₅Ni₅₀/Mg₂Si (x = 36, 50, 130) interface, the migration of Mg in thermoelectric material has been suppressed effectively. The results indicate that the contact performance and service stability of the Mg₁₃₀Si₁₅Ni₅₀/Mg₂Si interface has been improved by 50% compared to that of Mg₃₆Si₁₅Ni₅₀/Mg₂Si, which remains quite well after annealing at 400 ℃ for 480 h. TCP-IMC barrier with proper composition is a promising design for the manufacture of TEG interface to ensure its consistency in service especially under a large temperature difference.

Introduction

As a new kind of clean energy conversion technique, thermoelectric power generators (TEG) can directly convert waste heat into electric energy and improve the efficiency of energy usage based on Seebeck effect. They have been applied in a wide range of fields including the recovery and utilization of industrial waste heat and deep-space exploration.^1,2 A large number of materials with zT > 1 have been reported, e.g., nanostructured PbTe, CoSb₃ -based skutterudites, half-Heuslers, SnSe, Cu₂Se-based materials, and Mg₂Si-based solid solutions, which have potential prospect for use in the mid-temperature region between 500 K and 800 K.³

In recent years, the performance of thermoelectric materials has continued to make breakthroughs, which is determined by the dimensionless figure of merit,⁴ zT = S²σT/k,^5,6,7 where σ is the electrical conductivity, S is the Seebeck coefficient, k is the thermal conductivity and T is the absolute temperature. When the temperature difference reaches 350 K, the theoretical value of thermoelectric conversion efficiency of thermoelectric materials with average zT between 1 and 2 is about 10%.⁸ However, these high values never translate adequately into the performance of devices, mainly due to challenges in making stable contact between thermoelectric material and electrodes. To realize a stable contact between thermoelectric material and electrodes, a barrier layer is usually applied. However, due to the inevitable mutual diffusion reaction at the interface of the barrier and thermoelectric material, there is still a long way to improve the interface contact by designing proper barrier materials. Furthermore, it must be pointed out that the interfacial reaction, usually accompanied by atomic diffusion across the interface, could be more critical for the magnesium-based TEG working in the mid-temperature region. It has been proved that a deficiency of Mg may deteriorate the performance of the thermoelectric materials^4,9 and contact to the electrode^10,11 as well. So, it is necessary to inhibit the diffusion of elements between thermoelectric material and electrodes by adjusting the chemical composition gradient across the interface.

For thermoelectric devices, in addition to the performance of the thermoelectric material itself, contact resistance and bonding strength of the interface between thermoelectric materials and electrodes also have great influence on the efficiency, stability and life of thermoelectric devices.^12,13 In the cases of waste heat harvesting from the vehicle transport and furnace operation, the thermal interfacial resistance of the interface of the diffusion barrier materials and thermoelectric materials are not critical.¹⁴ So, as shown in Eq. 1,¹⁰ contact resistance is one of the main key factors that reduces the effective $\left\langle {ZT} \right\rangle_{{\text{D}}}$ of thermoelectric devices.

$$ \left\langle {ZT} \right\rangle_{D} = \left( {L/\left( {L + 2R_{c} \sigma } \right)} \right)\left\langle {zT} \right\rangle_{M} $$

(1)

where L and $\sigma$ are the length and electrical conductivity of the thermoelectric leg, respectively, R_c is the contact resistivity, and $\left\langle {zT} \right\rangle_{M}$ is the average zT of the thermoelectric material between T_h and T_c. Usually, the ratio of contact resistance to total leg resistance of less than 10% is acceptable. For a typical device, R_c should be much less than L/2 $\sigma$ or 10² μΩ·cm².¹⁰

Mg₂Si-based thermoelectric material, as a very promising thermoelectric material with a zT up to 1.45,^15,16,17 has been attracting more attention of scientists committing to its popularization and application as a medium-temperature TEG. As a defective compound, its transport behavior is determined by the concentration of intrinsic point defects, which is closely related to the stoichiometric ratio of elements.^18,19 The mutual diffusion and reaction occurring across the interface destroys the balance of point defects and causes the increase of interface resistance and deterioration of efficiency and even failure due to mechanical damage.¹⁰ In our recent work, it has been shown that Mg₂SiNi₃ is an excellent diffusion barrier between Cu electrodes and Mg₂Si, which not only effectively blocks the diffusion of Cu from the electrode to Mg₂Si thermoelectric material, but also ensures the maximum shear strength (28.29 MPa).²⁰ It is especially worth mentioning that the contact resistance between the intrinsic Mg₂Si and Mg₂SiNi₃ is pretty low compared to that reported in the literature.²⁰ However, the influence of the chemical potential of Mg in the interface of Mg₂SiNi₃/Mg₂Si on the dynamic equilibrium of point defects and thermoelectric transport must be further discussed and clarified.

Figure 1 illustrates the complex crystal structure of the Mg₂SiNi₃ material. Mg₂SiNi₃ is a reaction product between Mg₂Si and Ni, with $R\overline{3}$ m space group structure, which belongs to the μ phase in the topologically densely packed intermetallic compound (TCP-IMC).²¹ In the space unit, the coordination numbers of Mg, Si, and Ni are 15,12,12, respectively, and the packing ratio is as high as 0.85. It is very dense and stable in air even at 300 °C.²² The introduction of Mg₂SiNi₃ to form TE/TCP-IMC interface is thought to improve the contact and stability of the interface by suppressing the chemical potential of diffusion and reaction in both thermodynamics and dynamics. Additionally, the dual-bond characteristics of metallic bond and covalent bond is helpful to decrease interface residual stress due to mismatch as well. In this work, Mg₂SiNi₃ is used as a barrier to prepare Mg₂Si/Mg_xSi₁₅Ni₅₀ (x=36,50,150) thermoelectric joints to discover how the composition gradient affects the interface diffusion between Mg₂Si and Mg₂SiNi₃. The contact resistance, strength and thermal stability of these joints were studied in detail.

Fig. 1

The crystal structure of the Mg₂SiNi₃ material.

Experimental Procedure

Magnesium powder (≤50 μm, 99.5% pure, Aladdin, Shanghai), silicon powder (≤45 μm, 99.5% pure, Aladdin, Shanghai) and nickel powder (≤45 μm, 99.5% pure, Kermel, Tianjin) were ball mixed for 30 min in stoichiometric ratios of Mg:Si=1.8:1, Mg:Si=2:1 and Mg:Si:Ni=x:15:50 (x=36, 50, 130), respectively, in a high-speed ball mill (QM-3B) with a ball-to-powder mass ratio of 10:1 to obtain Mg_1.8Si, Mg₂Si and Mg₂SiNi₃.

The milled powders of thermoelectric materials were then cold-pressed inside a graphite die to obtain a green pellet (a diameter of 20 mm). The sample was sintered in a spark plasma sintering apparatus (SPS-630Lx) under sintering temperature 750 °C, uniaxial pressure of 40 MPa and vacuum of 20 Pa. The milled powders of thermoelectric and intermediate materials were then cold-pressed inside a graphite die to obtain a pellet (a diameter of 20 mm) with a symmetrical structure of Mg_xSi₁₅Ni₅₀/Mg₂Si (x=36, 50, 130).

SPS is conducted in vacuum by heating to 630 °C at a speed of 100 °C/min under 10 MPa followed by a 2-min dwell, and then rapidly heating to 750 °C under 40 MPa and staying for another 15 min. During cooling, the temperature was first decreased to 400 °C at a speed of 12.5 °C/min, and then decreased to room temperature at a speed of 10 °C/min. All handling was performed under a protective argon atmosphere.

A thermomechanical analyzer (DIL 402SU, NETZSCH) was used to measure the CTEs of Mg_xSi₁₅Ni₅₀(x=36,50,150). The microstructure and chemical composition of thermoelectric joints before and after aging were determined with scanning electron microscopy (SEM, JEOL JSM 6300) and energy dispersive spectroscopy (EDS, HKL). Contact resistance of samples was measured with the four-probe method and the samples for this evaluation were 3×3×δ mm in dimension, with δ being the thickness of the sample, typically 4-5 mm. The phase purity of the thermoelectric materials was confirmed by X-ray diffraction (DX-2700). The Seebeck coefficient and resistivity were performed in a Namicro3 system. A Hall effect test system (CH-100) was used to measure the carrier concentration. The shear strength of samples was tested by a microcomputer-controlled electronic universal testing machine (DNS200) with a head movement rate of 0.5 mm min⁻¹. The samples used for this evaluation were 7×7×δ mm in dimension (δ = 4-5 mm). The joints samples were heat-treated at 400 °C for 120, 240, 360 and 480 h under vacuum of 10⁻⁴ Pa to evaluate the thermal stability.

Results and Discussion

The Presence and the Influence of Mg-Depletion Layer

As shown in Fig. 2a, the interface of the Mg₃₆Si₁₅Ni₅₀/Mg₂Si thermoelectric joint is a sandwich-like structure. The gray part on the right is the diffusion barrier Mg₂SiNi₃ phase, and the dark layer on the left is the Mg₂Si thermoelectric material. No voids or cracks were observed in the joints. Figure 2b shows a uniform and dense diffusion layer formed in the Mg₂Si/Mg₃₆Si₁₅Ni₅₀ interface with a thickness of about 15 μm. There are two different phases in the diffusion layer: dark phase and dispersed particles. Combined with EDS analysis (see Fig. 2c) and Mg-Ni-Si ternary phase diagram,²³ it is shown that the atomic ratios of Mg, Si, and Ni at different site are very different, the phase component of matrix of the diffusion layer is basically a mixture of Mg₂Si and Mg, and the second phase particles is NiSi₂ phase.²⁴ EDS analysis result (see Fig. 2d) indicates that a Mg-depletion region (MDR) exists next to the thermoelectric material with a thickness of 7 μm, which is detrimental to the thermoelectric properties of Mg₂Si (discussed below). The formation of MDR is mainly due to faster diffusion velocity of Mg than Si. It is generally accepted that substitutional migration of atoms in intermetallic compounds occurs by virtue of vacancy-type defects. In such an antifluorite-type structure ($Fm\overline{3}m$), sublattice of Si in Mg₂Si can also be viewed as simple cubic with one Si atom of every two missing, so Mg diffuses faster than Si element and is easier to lose.²⁵ So it would be helpful to suppress the migration of Mg by increasing the percentage of Mg in the barrier and decreasing its ingredient gradient as well. Figure 3a shows the variation of lattice parameters of Mg_xSi₁₅Ni₅₀ (36≤x≤130). Linear variation follows the Vegard’s law up to x=45, subsequently saturating between x=45 and x=50. Partially squeezed out liquid from the interface may be observed when x≥55, which is Mg (see Fig. 3b).

Fig. 2

Mg₃₆Si₁₅Ni₅₀/Mg₂Si thermoelectric joint interface micromorphology (a), (b) micromorphology (c) EDS point analysis results (d) EDS line analysis results of line ab in (b).

Fig. 3

(a) lattice constant of Mg_xSi₁₅Ni₅₀ (36 ≤ x ≤ 130) sample (b) sample with Mg squeezed out during sintering.

Influence of Mg Defect

Because of the high vapor pressure of Mg and high sintering temperature of Mg₂Si, it is difficult to accurately control the stoichiometry of Mg.²⁶ The excess or lack of magnesium in Mg₂Si affects the balance of the intrinsic defects (I^Mg,²⁷ V^{Mg 28} and anti-defects), thus affecting the electrical properties. In order to study the influence of Mg content on the electrical properties of the material, Mg_1.8Si and Mg₂Si were made by SPS. As shown in Fig. 4a, all of diffraction peaks can be indexed to the antifluorite structure (PDF#35-0773; a = b = c = 6.351 Å), with a small amount of MgO. The study on the stability and electronic properties of point defects and multivacancies in Mg₂Si shows that the stability of the defects is strongly dependent on the stoichiometric conditions.²⁹ The V^MgSi, V^Mg2Si and I^Mg are the most stable defects when Mg is rich, while the antisite Si^Mg becomes more stable, and I^Mg becomes less stable when Mg is poor. Mg₂Si is always n-type in experiments, so the most stable defect is I^Mg, which acts as a donor. In stoichiometric conditions, the I^Mg defect becomes less favorable, whereas the V^Mg defect becomes more stable and the multivacancies are still very favorable.⁴ The temperature-dependent electrical properties of Mg_1.8Si and Mg₂Si indicate a non-degenerate feature, as shown in Fig. 4b and c. The higher resistivity of the Mg_1.8Si sample can be ascribed to the formation of V^Mg and loss of the I^Mg due to lack of Mg.³⁰ V^Mg has a significant effect on the electric properties of Mg₂Si. In stoichiometric conditions, the formation energy of V^Mg is as low as 1.54 eV.²⁹ This is one of the reasons that the magnesium atoms diffuse very easily. Our recent work shows that the use of intermetallic compounds as a barrier may block the diffusion of Mg effectively.²⁰

Fig. 4

(a) Room-temperature XRD patterns, (b) Resistivity and (c) Carrier concentration of Mg_1.8Si and Mg₂Si.

Suppressing Mg Deletion Through Ingredient Gradient Design

Suppression of the excessive diffusion of Mg atoms in thermoelectric materials can alleviate the generation of Mg vacancies and improve the interface performance. In order to study the effect of the percentage Mg content in the barrier on the net diffusion of interface elements, thermoelectric joints with different concentration gradients (Mg_xSi₁₅Ni₅₀/Mg₂Si; x=36,50,130) were designed.

Figure 5a and c presents the microstructures of Mg₁₃₀Si₁₅Ni₅₀/Mg₂Si and Mg₅₀Si₁₅Ni₅₀/Mg₂Si joints. A dense and uniform intermediate layer is formed between the thermoelectric material and the barrier layer with a thickness of 10 μm and 15 μm, respectively. The chemical composition of the diffusion layer is Mg:Si:Ni≈1:1:1, which is close to ω-(Mg_0.52Ni_0.48)₇Si₄ phase according to the phase diagram and literature.²⁸ The EDS results show a diffusion layer next to thermoelectric material as thin as 2 to 3 μm, and no MDR is observed, as shown in Fig. 5b and d, which proves that increasing Mg concentration gradient across the interface by increase of Mg content in Mg₂SiNi₃ side can effectively suppress the Mg loss in the Mg₂Si effectively. Furthermore, when x increases from 50 to 130, no obvious effect is observed in either interface but excess squeezed-out liquid Mg is present in the former sample.

Fig. 5

Interface micromorphology and element distribution across the interface of (a), (b) Mg₁₃₀Si₁₅Ni₅₀/Mg₂Si and (c) (d) Mg₅₀Si₁₅Ni₅₀/Mg₂Si.

According to Fick's first law of diffusion, the flux of Mg atom from Mg₂Si to Mg₃₆Si₁₅Ni₅₀ is given by:

$$ J = ( - D_{0} \Delta C)/\Delta x $$

(2)

where D₀ is the diffusion coefficient of Mg in the reaction layer, $\Delta c$ is concentration difference of $c_{{{\text{Mg}}_{2} {\text{Si}}}}$-$c_{{{\text{Mg}}_{36} {\text{Si}}_{15} {\text{Ni}}_{50} }}$, $\Delta x$ is the thickness of reaction layer (15 μm). The per second total number of Mg atom diffusion from Mg₂Si to Mg₃₆Si₁₅Ni₅₀ is calculated by

$$ J_{{{\text{total}}}} = J \cdot S $$

(3)

where S is the cross-section of the interface. The per second amount of Mg atoms diffused from the Mg₃₆Si₁₅Ni₅₀/Mg₂Si interface is calculated by

$$ J_{\sec } = J_{{{\text{total}}}} /C_{{\left( {{\text{Mg}}_{2} {\text{Si}}} \right)}} $$

(4)

where $c_{{{\text{Mg}}_{2} {\text{Si}}}}$ is concentration of Mg in Mg₂Si (3.1 × 10²² at./cm³). In Mg₂Si, the thickness increase per second of the Mg-poor layer is calculated by

$$ \Delta l_{{{\text{Sec}}}} = J_{{{\text{sec}}}} /S $$

(5)

Therefore, the thickness of the Mg-poor layer in Mg₂Si is calculated by

$$ \Delta l = \left( { - D_{0} \Delta c \cdot t} \right)/\left( {\Delta xC_{{\left( {{\text{Mg}}_{2} {\text{Si}}} \right)}} } \right) $$

(6)

$c_{{{\text{Mg}}_{36} {\text{Si}}_{15} {\text{Ni}}_{50} }}$ is the concentration of Mg in Mg₃₆Si₁₅Ni₅₀ (3 × 10²² at./cm³), t is aging time. According to the Mg-poor layer thickness from Figs. 2 and 7, $\Delta l$ is 4 μm. So, the diffusion coefficient of Mg from Mg₂Si to Mg₃₆Si₁₅Ni₅₀ at 400 ℃ is 2.15×10⁻¹² cm²/s. Compared with that²⁵ (2.0 × 10^-10 cm²/s) of Mg in the Mg₂Si at 500 °C , the diffusion coefficient of Mg from Mg₂Si to Mg₃₆Si₁₅Ni₅₀ at 400 ℃ was lower. For the interface of Mg₂Si/Mg₅₀Si₁₅Ni₅₀, $c_{{{\text{Mg}}_{36} {\text{Si}}_{15} {\text{Ni}}_{50} }}$ in Mg₅₀Si₁₅Ni₅₀(4.2 × 10²² at./cm³). The direction of Mg concentration difference is from Mg₅₀Si₁₅Ni₅₀ to Mg₂Si. The results above indicate that increasing of Mg content of Mg₂SiNi₃ side can effectively suppress the Mg diffusion from Mg₂Si to Mg₂SiNi₃.

Interface Performance is Improved by Gradient Design

Contact resistivity has a significant effect on conversion efficiency and power output for thermoelectric modules,³³ which mainly depends on the interface microstructure as well as the electrical properties of the interfacial materials.^{34,35,36,37,38,39,40} Lower contact resistivity (R_C<< L/2σ) for thermoelectric joints is expected in long-term service.

Figure 6a shows the change of electrical resistance across the section of Mg_xSi₁₅Ni₅₀/Mg₂Si (x=36,50,130) joints. The linearly increasing resistance on the Mg₂Si side indicates a semiconductor while that in the Mg₂SiNi₃ side indicates metallic property. An abrupt change between Mg₂SiNi₃ and intrinsic Mg₂Si can be observed, as summarized in Fig. 6a. The intrinsic contact resistance is 275.4 μΩ cm², 171.9 μΩ cm², 141.3 μΩ cm² for Mg₃₆Si₁₅Ni₅₀/Mg₂Si, Mg₅₀Si₁₅Ni₅₀/Mg₂Si, and Mg₁₃₀Si₁₅Ni₅₀/Mg₂Si, respectively, which are smaller than 320 μΩ cm² for the Mg₂Si/Ni joint measured by Sakamoto et al.⁴¹ Compared with the Mg₃₆Si₁₅Ni₅₀/Mg₂Si joint, the contact resistance of Mg_xSi₁₅Ni₅₀/Mg₂Si joints (x=50, 130) is reduced by 50%. The higher contact resistivity of the Mg₃₆Si₁₅Ni₅₀/Mg₂Si joint may be ascribed to the formation of MDR due to Mg diffusion.

Fig. 6

(a) Testing curve of contact resistance of the joints (b) Variation of contact resistivity with different aging time for Mg_xSi₁₅Ni₅₀/Mg₂Si (x = 36, 50, 130) joints.

To evaluate the interfacial reliability of the thermoelectric joints at high temperatures, aging tests of Mg_xSi₁₅Ni₅₀/Mg₂Si (x=36, 50, 130) joints at 400 °C for 480 h are carried out. Figure 6b shows the change of contact resistance of Mg_xSi₁₅Ni₅₀/Mg₂Si (x=36, 50, 130) joints after aging. These values are smaller than 16.4 mΩ cm² obtained for the Mg₂Si/Cu joint by Cai et al.⁴² We conjecture that the increase in the contact resistivity with aging time might be ascribed to the loss vitalization of Mg of TE material during aging. Similar results have been reported in the literature.^20,24,43 The contact resistance of Mg₁₃₀Si₁₅Ni₅₀/Mg₂Si and Mg₅₀Si₁₅Ni₅₀/Mg₂Si are smaller than that of Mg₃₆Si₁₅Ni₅₀/Mg₂Si, indicating that the Mg-rich diffusion barrier layer can suppress the loss of Mg, thereby increasing the stability of the joint.

As shown in Fig. 7a, compared with that before aging in Fig. 2b, the phase component of the aged interface Mg₃₆Si₁₅Ni₅₀/Mg₂Si after aging for 240 h is almost unchanged. But the thickness of the diffusion layer and MDR increases, where the thickness of the diffusion layer and MDR is around 20 μm and 11 μm, respectively (Fig. 7b). The increase in the MDR thickness may be ascribed to the diffusion layer which has inconspicuous inhibitory on the diffusion of Mg. Figure 7c and e shows the interface microstructure of Mg₅₀Si₁₅Ni₅₀/Mg₂Si and Mg₁₃₀Si₁₅Ni₅₀/Mg₂Si joints after aging at 400 °C for 240 h, respectively. Compared with that before aging, both the thickness and phase component remain unchanged.

Fig. 7

Microstructure and EDS line scan analysis of Mg_xSi₁₅Ni₅₀/Mg₂Si interface after annealing at 400 °C for 240 h (a), (b) Mg₃₆Si₁₅Ni₅₀/Mg₂Si joint (c), (d) Mg₅₀Si₁₅Ni₅₀/Mg₂Si joint (e) (f) Mg₁₃₀Si₁₅Ni₅₀/Mg₂Si joint.

The CTEs of Ni, Mg_xSi₁₅Ni₅₀(x=36, 50, 150) and Mg₂Si within the temperature range of 150-400 °C are shown in Fig. 8a. The CTE of Ni ranges from 13.2×10^-6 K^-1 to 14.8×10^-6 K^-1 between 150 °C and 400 °C,²⁰ that of Mg₃₆Si₁₅Ni₅₀ ranges from 12.1×10^-6 K^-1 to 11.2×10^-6 K^-1 , and that of Mg₅₀Si₁₅Ni₅₀ ranges from 11.3×10^-6 K^-1 to 13×10^-6 K^-1, which is very close to that of Mg₁₃₀Si₁₅Ni₅₀.

Fig. 8

(a) The CTE values of the Mg_xSi₁₅Ni₅₀(x = 36, 50, 150) (b) temperature dependence of resistivity of Mg_xSi₁₅Ni₅₀(36 ≤ x ≤ 130) within the temperature range.

Figure 8b shows that the resistivity decreases with increase of x. when x is greater than 50, it remains almost unchanged, and some liquid magnesium can be observed at the surface of the plunger (Fig. 3b), which indicates that no more magnesium may enter into the lattice. The Mg-Si-Ni ternary phase diagram²³ indicates that the residual Mg in MgxSi₁₅Ni₅₀(x=50, 130) exists as a second phase of Mg₂Ni at the grain boundary, with a CTE of 7.5×10^-6 K^-1 to 12.5×10^-6 K^-1 from 150 ℃ to 350 ℃,²⁰ which is far lower than that of Mg₃₆Si₁₅Ni₅₀.

Mg₂Si has a CTE within the scope of 10.2×10^-6 K^-1 and 18.9×10^-6 K^-1 from 150 °C to 400 °C.²⁰ The previous report showed that a CTE difference less than around 5-6 ×10^-6 K^-1 does not lead to failure of the joint of Mg₂Si and Ni.²⁰ CTE of the barrier may be modified by changing the content of magnesium, which is beneficial to reduce the residual interface stress and improve the bonding strength and stability.

The shearing tests were conducted to evaluate the mechanical properties, as shown in Fig. 9. All of them are about 23 MPa before aging. These values are higher than that of 16.2 MPa and 19 MPa obtained for Mg₂Si/Cu and Mg₂Si/Ni joints by Cai et al⁴² and Tohei et al.⁴⁴ Furthermore, this work shows great advantage after aging. As can be seen from Fig. 9, the strength of the Mg₃₆Si₁₅Ni₅₀/Mg₂Si joint decreases more than the other two, which is mainly due to the phase component of the interface diffusion layer. The uniform and dense ω-phase diffusion layer at the interface of Mg_xSi₁₅Ni₅₀/Mg₂Si (x=50, 130) can maintain high bonding strength during the aging. Additionally, the CTE difference (see Fig. 8) in interfaces is one of the reasons that the bonding strength decreases with aging time.²⁴

Fig. 9

The shear strength of Mg_xSi₁₅Ni₅₀/Mg₂Si (x = 36, 50, 130) joints after aging.

Conclusions

In this work, Mg₂SiNi₃ (TCP-IMC) was selected as the barrier layer to bond with intrinsic Mg₂Si by SPS in one step. Adjusting the proportion of Mg in the barrier to suppress the migration of Mg across the interface thereby alleviates the degradation of the thermoelectric properties of the Mg₂Si thermoelectric material and increases the stability in service as well.

A diffusion layer was located between Mg₂Si and Mg₃₆Si₁₅Ni₅₀. The matrix is basically a mixture of Mg₂Si and Mg, with NiSi₂ precipitates as second phase. In the matrix of Mg_xSi₁₅Ni₅₀/Mg₂Si (x=50, 130), a dense and uniform diffusion layer was observed, which was composed of an ω-phase (Mg:Si:Ni≈1:1:1) based on EDS and phase diagram. A higher chemical potential of Mg in the barrier is helpful to suppress the formation of the Mg-depletion layer in Mg₂Si.

The barrier of Mg_xSi₁₅Ni₅₀ (x=50, 130) effectively improves the interface conductivity and the bonding strength. The contact resistivity in Mg_xSi₁₅Ni₅₀/Mg₂Si (x=50, 130) joints are 171.9 μΩ cm² and 141.3 μΩ cm², which are about 50% of that of Mg₃₆Si₁₅Ni₅₀/Mg₂Si joint. The shear strength of all joints with different Mg content is around 23 MPa. Mg₅₀Si₁₅Ni₅₀/Mg₂Si and Mg₁₃₀Si₁₅Ni₅₀/Mg₂Si joints show better stability both in contact and shear strength after aging at 400 ℃ for different time.