In this research, a fundamental study is conducted to identify the materials and develop the processes for producing barrier/bonding composite on Bi2Te3 for high temperature thermoelectric applications. The composite must meet four basic requirements: (a) prevent inter-diffusion between the electrode material, for our design, silver(Ag) and Bi2Te3, (b) bond well to Bi2Te3, (c) bond well to Ag electrode, and (d) do not themselves diffuse into Bi2Te3. The composites investigated include palladium (Pd), nickel/gold (Ni/Au), Ag, and titanium/gold (Ti/Au). After annealing at 250 °C for 200 h, only the Ti/Au design meets all four requirements. The thickness of Ti and Au, respectively, is only 100 nm. Other than meeting these four requirements, the Ti/Au layers exhibit excellent step coverage on the rough Bi2Te3 surface even after the annealing process.
1 Introduction
Bismuth telluride (Bi2Te3) is a narrow band gap semiconductor with trigonal unit cell. It was first investigated as a thermoelectric (TE) material in 1950s [1‐3]. The TE power generator uses Seebeck effect to convert temperature gradient ΔT into electrical power. The voltage produced is given by V = αΔT, where α is Seebeck coefficient. Conversely, the TE cooler uses the Peltier effect to convert electrical power into temperature gradient by transporting heat from cold side to hot side. The heat transported is given by Qp = πI, where I is the current and π is the Peltier coefficient. Bi2Te3 is so far the most popular TE material. Alloying with Sb2Te3 and Bi2Se3 can control carrier concentrations alongside a reduction in lattice thermal conductivity [4]. It can also improve the figure-of-merit, Z = α2 (σT/λ), where σ is the electrical conductivity and λ is its thermal conductivity. Bismuth-based alloys with antimony, tellurium, and selenium are referred to as low-temperature materials with Z around 2 ~ 4 × 10−3 K−1 and can be used at temperature up to 450 K [5]. For temperature from 450 to 850 K, the most common thermoelectric material is lead telluride. At even higher temperature, silicon germanium alloys can operate up to 1,300 K [4, 5]. Z of lead telluride and silicon germanium alloy are below 2 × 10−3 K−1 [5]. Most TE devices on the market are made of Bi2Te3-based alloys.
A TE cooler or generator, portrayed in Fig. 1, is constructed of an array of p-type and n-type TE columns bonded between two ceramic plates [6]. These columns are connected electrically in series to form a chain of p-n junctions but thermally in parallel. Solder is employed to bond the TE columns to the ceramic plates. TE modules on the market have a maximum operating temperature of 80 °C on the hot side [7]. This constraint prohibits numerous applications and economic opportunities where higher operation temperature is required. An example is to cool the transistor in the front-end amplifier of an outdoor wireless communications base station to reduce its noise. Another example is to cool the front-end photodiode of a military laser range-finding receiver to increase its sensitivity. The 80 °C limit is imposed by severe diffusion of solder material into the TE materials rather than by TE materials themselves. The melting points of TE materials Bi2Te3, Sb2Te3, and Bi2Se3 are 585 °C [8], 618 °C [9], and 706 °C [10], respectively. Therefore, to extend the operating temperature to 150 °C, solder diffusion at high temperature must be stopped. Accordingly, high temperature bonding/barrier layers must be investigated and identified.
Fig. 1
The structure of a TE module
×
Anzeige
Several groups had studied diffusion barrier for Bi2Te3 and copper (Cu) interconnect. Few showed positive results. In one study, Ni was used as a diffusion-barrier for tin (Sn) in both (Bi,Sb)2Te3 and Bi2(Te,Se)3 [11]. Ni diffuses into Bi2(Te,Se)3 by several microns during the soldering process. Ni was also found to diffuse into Bi2(Te,Se)3 and form a nickel telluride interfacial region after heat treatment at 200 °C [12]. In another study, Ni-7% vanadium (V), Pd and platinum (Pt) films were used but failed to prevent inter-diffusion between Cu and Bi2Te3 after annealing at 200 °C for a few hours [13]. It is clear that several research groups had searched for diffusion barrier on Bi2Te3-based compounds. So far, the best barrier design can sustain 200 °C only for a few hours.
The objective of this research is to design and develop barrier/bonding composite that can sustain annealing at 250 °C for 200 h. The composite is fabricated between Bi2Te3-based compound and Ag electrode. One surface of the composite must bond to Bi2Te3-compound and the other surface must bond to Ag. Furthermore, it must prevent Ag atoms from diffusing into Bi2Te3 and prevent Bi2Te3 from diffusing into Ag electrode. Ag rather than Cu is chosen as the electrode material because it has the highest electrical conductivity and thermal conductivity among metals. To identify the right bonding/barrier materials and processes, we went through a systematic study, starting with Pd, followed by Ni/Au, Ag, and Ti/Au. Some preliminary results were presented in [14]. In this paper, we report more complete analyses and results. Additional information obtained includes: extensive 250 °C annealing data of Ti/Au barrier/bonding layer from 10 to 200 h, Ag surface morphology and adhesion on different barrier layers, SEM/EDX analyses on inter-diffused regions, and discussion on compound formation in inter-diffused regions.
In what follows, experimental design and procedure are presented. Experimental results are reported and discussed. A short summary is then given. For convenience, all chips made of Bi2Te3 and related alloys are referred to as Bi2Te3 chips.
2 Experimental design and procedure
Hot-pressed p- and n-type Bi2Te3 ingots were acquired from a commercial vendor. They are usually manufactured starting with bismuth antimony, selenium and tellurium shot of 99.999% purity. An alloying process converts the pure elements into compound of designed composition. Powdered alloys are extruded at a temperature between 400 and 500 °C [15]. They are then cut into 8mm × 8mm × 1 mm chips. The composition of n-type alloy is (Bi0.95Sb0.05)2 (Te0.95Se0.05)3 and that of p-type alloy is (Bi0.2Sb0.8)2Te3. The room temperature figures of merit are up to 3.3 × 10−3 K−1 for p-type alloys and 2.85 × 10−3 K−1 for n-type alloys [15]. The coefficient of thermal expansion (CTE) of Bi2Te3 is 16.4 ppm/ °C. The Bi2Te3 chips are thoroughly cleaned and baked at 120 °C for 30 min for dehydration.
Anzeige
Figure 2 portrays the structure of the barrier/bonding layers on Bi2Te3 chips with Ag as the electrode. The Ag electrode can be connected to the ceramic plate by soldering or other bonding means. The Ag layer is fabricated by an electroplating process. A thickness range of 10–20 μm is used. To search for bonding/barrier layers, we start with Pd because it has long been used in commercial TE cooling modules. Ni/Au, Ag, and Ti/Au are then studied. To produce a few microns of Pd, Ni/Au, and Ag layers, respectively, on Bi2Te3 chips, electroplating processes are employed. For depositing Ti films, we are not able to find a plating solution. Thus, the Ti/Au layers are deposited by electron beam thermal evaporation in one vacuum cycle in 1 × 10−6 torr vacuum. The Ti layer is expected to bond to Bi2Te3 chips while the Au layer protects the Ti layer against oxidation.
Fig. 2
Structure of barrier/bonding layers on a Bi2Te3 chip
×
The samples fabricated are cut in cross section and polished for examinations by optical microscope, scanning electron microscope (SEM) and EDX analysis. Some samples are annealed at 250 °C for 200 h prior to the cutting and polishing processes.
3 Experimental results and discussion
3.1 Pd barrier/bonding layer
The Bi2Te3 chips were coated with 1 μm Pd by electroplating process. Figure 3a and b show the SEM images of Pd surface on Bi2Te3 at low and high magnification, respectively. The surface of Pd is smooth and bonds well to Bi2Te3. The Pd grain size is less than 100 nm. The samples were then plated with 10 μm Ag and annealed at 250 °C for 10 h. After annealing, the Ag layer still adheres to the Pd layer which also bonds to the Bi2Te3 chip. To further study whether the Pd layer can sustain stresses induced by CTE mismatch while bonding on alumina substrate, the Bi2Te3/Pd/Ag samples were plated with a layer of 5 μm In and a thin Ag cap layer. The Bi2Te3 chips were bonded to alumina substrates. The bonded structure was cut in cross section and polished for evaluations. Figure 4a exhibits a cross section SEM image. It is observed that a gap exists between Pd and Bi2Te3. This is the breakage interface due to CTE mismatch between Bi2Te3 and alumina. The Ag layer on Bi2Te3 was well bonded to the Ag layer on the alumina substrate. On Fig. 4a, it is also found that Pd and Ag atoms have diffused into Bi2Te3 by as deep as 5 μm during the annealing process and prior to the bonding. Figure 4b and c display the EDX spectra at locations A and B marked on Fig. 4a. Table 1 provides quantitative EDX data. Location A is in a darker region on the back-scattered electron (BSE) image. The EDX system detects Bi, Te, Ag and Pd. Table 1 shows 54 at.% Ag and 5 at.% Pd. Significant Ag has diffused through Pd into Bi2Te3. This makes the Ag diffused region darker on the BSE image because Ag has a smaller atomic number than Bi and Te. The BSE image suggests that Ag has diffused to 5 μm in depth after annealing at 250 °C for 10 h. At location B, which is 7 μm away from the interface, 7 at.% Ag is detected. The small Ag amount does not make the region appear darker on the BSE image. As a result of Ag diffusion, Pd is not a good barrier layer on Bi2Te3 to block Ag.
Fig. 3
SEM images of 1 μm Pd plated on Bi2Te3 at a low magnification and b high magnification
Fig. 4
SEM and EDX evaluations of Bi2Te3/Pd (1 μm)/Ag (10 μm) bonded to alumina/Ag (60 μm)/In (5 μm)/Ag (100 nm) structure to produce an Ag-In joint. a Cross section SEM image. The alumina substrate is not shown here. The Bi2Te3/Pd (1 μm)/Ag (10 μm) sample was annealed at 250 °C for 10 h. The Bi2Te3 chip broke off along the Bi2Te3-Pd interface after the bonding due to CTE mismatch with alumina substrate. Significant Pd and Ag have diffused into Bi2Te3 after the annealing and prior to bonding. b EDX spectra at location A, c EDX spectra at location B. EDX quantitative data are presented in Table 1
Table 1
EDX quantitative data at locations A and B marked in Fig. 4a
Location
At.%
Bi
Te
Pd
Ag
A
8
31
5
54
B
14
78
1
7
The accelerating voltage is 20 kV
×
×
3.2 Ni/Au barrier/bonding layers
Bi2Te3 chips were plated with 5 μm Ni, followed by 1 μm Au and 10 μm Ag. Au layer works as an adhesion layer between Ni and Ag. Figure 5a, b, c, and d display SEM images on the Au surface of Bi2Te3/Ni (5 μm)/Au (1 μm) and on the Ag surface of Bi2Te3/Ni (5 μm)/Au(1 μm)/Ag (10 μm) at low and high magnification, respectively. These images show that the surface of Au and Ag layers are flat and smooth. The surface roughness is less than 1 μm and the grain sizes of electroplated Au and Ag are both smaller than 50 nm. All the electroplated layers are well bonded to each other. Subsequently, the Bi2Te3/Ni/Au/Ag samples were annealed at 250 °C for 10 h. Figure 6 displays the cross section SEM image of Bi2Te3/Ni (5 μm)/Au (2 μm)/Ag (10 μm) sample after annealing at 250 °C for 10 h. Severe inter-diffusions among Bi2Te3/Ni, Ni/Au and Ni/Au/Ag layers are observed. Table 2 presents EDX quantitative data at five locations, A, B, C, D, and E, marked on Fig. 6. The accelerating voltage is 20 keV. These EDX data show inter-diffusions between Bi2Te3 and Ag/Au/Ni. At location A, large amount of Ag and Au are detected in the original Ni layer, indicating that Ag and Cu atoms have diffused into the Ni layer during annealing. At locations B and C, which are right above and below the breaking interface, respectively, a large quantity of Ni and Te are detected as well as a small amount of Bi. This suggests that Ni and Bi2Te3 have inter-diffused, forming NiTe layer [16, 17]. Since the atomic ratio between Ni and Te is around 0.7 ~ 1:1, this layer is probably NiTe compound. Location D and E are basically Bi2Te3. The BSE image clearly shows that the breakage incurs inside the NiTe compound. Apparently, NiTe compound is weak and breaks due to stress induced by CTE mismatch in the multilayer structure. Thus, the Ni/Au composite is not a good choice for bonding/barrier layer.
Fig. 5
SEM images on the surfaces of Bi2Te3/Ni (5 μm)/Au (1 μm) at a low magnification and b high magnification, and of Bi2Te3/Ni (5 μm)/Au (1 μm)/Ag (10 μm) at c low magnification and d high magnification
Fig. 6
Cross section SEM image of Bi2Te3/Ni (5 μm)/Au (2 μm)/Ag (10 μm) sample after annealing at 250 °C for 10 h. EDX data at locations A, B, C, D, and E are presented in Table 2. Severe inter-diffusions among layers are observed. A compound layer NiTe is formed between Ni layer and Bi2Te3. NiTe seems to be mechanically weak. The multilayer structure eventually broke up within NiTe layer due to CTE mismatch
Table 2
EDX quantitative data at location A, B, C, D, and E marked in Fig. 6
Location
At.%
Bi
Te
Ni
Au
Ag
A
4
2
13
39
42
B
6
45
46
2
1
C
7
52
38
1
1
D
62
31
5
1
1
E
54
42
2
1
1
The accelerating voltage is 20 kV
×
×
3.3 Ag barrier/bonding layer
This is the simplest design because Ag itself is the electrode. In experiment, 10 μm Ag was electroplated on Bi2Te3 chips, followed by annealing at 250 °C for 10 h. Figure 7a and b show the surface of Ag layer on Bi2Te3 before annealing at low and high magnification, respectively. The Ag layer bonds well to Bi2Te3 and its grain size is less than 50 nm. Figure 8 exhibits the cross section SEM image a Bi2Te3/Ag (10 μm) sample after annealing. Severe inter-diffusion between Ag and Bi2Te3 is observed. An Ag2Te compound layer is formed based on EDX analysis and Ag-Bi-Te phase diagram [18]. This layer has some voids and partly breaks off the Bi2Te3 chip. It is clear that Ag is not a good barrier/bonding layer for Bi2Te3.
Fig. 7
SEM images on the surfaces of Bi2Te3/Ag (10 μm) at a low magnification and b high magnification
Fig. 8
SEM image on the cross section of Bi2Te3/Ag (10 μm) after annealing at 250 °C for 10 h. An Ag2Te compound layer is formed between Bi2Te3 and Ag. It has some voids and partly breaks off the Bi2Te3 chip
×
×
3.4 Ti/Au barrier/bonding layers
Bi2Te3 were coated with 100 nm Ti and 100 nm Au by an E-beam thermal evaporator and then plated with 20 μm Ag. Figure 9a and b are the SEM images on Au surface of a Bi2Te3/Ti/Au sample at low and high magnification, respectively. The Ti (100 nm)/Au (100 nm) layers are well deposited on the Bi2Te3 chip with good uniformity. Figure 9c and d show the Ag surface of a Bi2Te3/Ti/Au/Ag sample at low and high magnification, respectively. The Ag morphology is different from the Ag plated on other barrier layers. The grains of Ag layer plated on Ti/Au are coarser. The grain size is about 100 nm. Without annealing, a few samples were cut and polished for examinations. Figure 10a and b display the cross section SEM images of a Bi2Te3/Ti/Au/Ag sample at low and high magnification, respectively. It is observed that the Ti layer bonds well Bi2Te3, the Au layer bonds well to the Ti layer and the Ag layer bonds well to the Au layer. Figure 10b also demonstrates that the Ti/Au layers have excellent step coverage over the rough Bi2Te3 surface. The Ti/Au layers cover all the steps, jags, and protuberances. Moreover, no Ag or Au atoms diffuse through the Ti layer and get into Bi2Te3. Also, no Bi2Te3 diffuses through the Ti layer and gets into the Ag layer either.
Fig. 9
SEM images on the Au surfaces of Bi2Te3/Ti (100 nm)/Au (100 nm) at a low magnification and b high magnification, and on the Ag surfaces of Bi2Te3/Ti (100 nm)/Au (100 nm)/Ag (10 μm) at c low magnification and d high magnification
Fig. 10
Cross section SEM images of Bi2Te3/Ti (100 nm)/Au (100 nm)/Ag (20 μm) at a low magnification and b high magnification without annealing. The Ti/Au layers exhibits excellent step coverage on the rough Bi2Te3 surface. The Ti/Au layers are continuous and cover the entire Bi2Te3 surface with any breakage or pinhole
×
×
To further evaluate the barrier/bonding functions of Ti/Au layers, Bi2Te3 chips with Ti (100 nm)/Au (100 nm)/Ag (20 μm) were annealed at 250 °C for 50 h. Figure 11a and b exhibit the cross section SEM images of a sample after annealing at low and high magnification, respectively. These BSE images indicate that there is no interdiffusion among Ag, Ti/Au, and Bi2Te3 chips. In the Bi2Te3 region, no other element traces show up on the contrast of the image. This means that the integrity of Bi2Te3 chips is well preserved after the annealing. Moreover, the Ti/Au layers still maintain exceptional step coverage over the rough Bi2Te3 surface. They also adhere to the Ag electrode and the Bi2Te3 chip well. Next, the annealing time was extended to 200 h. Figure 12a exhibits the cross section SEM image of a sample which was annealed at 250 °C for 200 h. No interdiffused region is observed. The thin Ti/Au layers on the SEM image do not appear as sharp as those of the sample annealed for 50 h. But even so, the Ti/Au layers are still continuous and complete. To investigate whether Ag atoms diffuse through the Ti/Au layers and get into Bi2Te3 region, EDX analysis was performed on the cross section exhibited in Fig. 12a. Figure 12b, c, d, and e present the EDX spectra at an accelerating voltage 20 keV at the locations 5, 10, 20, and 100 μm away from the Ti/Au interface. The Sb, Te, and Bi peaks are clearly detected. Sb is there because this is a p-type material. The EDX data are provided in Table 3. At all four locations, the compositions of Sb, Te, and Bi are 30 ~ 34 at.%, 52 ~ 54 at.%, and 8 ~ 10 at.%, respectively. On the EDX spectra, Ag peak is barely seen. The 1 ~ 2 at.% Ag shown in Table 3 is within the error bar of the EDX system. It is thus clear that no Ag atoms diffuse through Ti/Au layers and get into Bi2Te3. The Ti/Au layers still work functionally as barrier and adhesion layers. Ti/Au themselves do not diffuse and spread outwards into Ag and Bi2Te region.
Fig. 11
Cross section back-scattered SEM images of Bi2Te3/Ti (100 nm)/Au (100 nm)/Ag (20 μm) sample annealed at 250 °C for 50 h: a high magnification and b low magnification. The thin Ti/Au layers show exceptional step coverage over the rough Bi2Te3 surface even after the annealing
Fig. 12
SEM and EDX evaluation of a Bi2Te3/Ti (100 nm)/Au (100 nm)/Ag (20 μm) sample after annealing at 250 °C for 200 h. a Cross section SEM, b, c, d, and e EDX spectra at locations 5, 10, 20 and 100 μm below the Ti/Au interface with an accelerating voltage 20 keV. EDX quantitative data are provided in Table 3. Again, the Ti/Au layers still keep the excellent step coverage on the rough Bi2Te3 surface after the annealing process. No Ag is detected in the Bi2Te3 chip
Table 3
EDX quantitative data of a Bi2Te3/Ti (100 nm)/Au (100 nm)/Ag (10 μm) sample at locations in Bi2Te3 at depth 5, 10, 20, and 100 μm, respectively, from the Ti/Au interface
Location
At.%
Sb
Te
Bi
Ag
d = 5 μm
31
54
9
1
d = 10 μm
34
52
9
1
d = 20 μm
30
54
10
2
d = 100 μm
32
54
9
1
The sample was annealed at 250 °C for 200 h. The accelerating voltage is 20 keV
×
×
The results presented above illustrate that Ti/Au layers meet the four requirements of barrier/bonding composite design even after annealing at 250 °C for 200 h: (a) prevent Ag and Bi2Te3 from inter-diffusion, (b) bond well to Ag, (c) bond well to Bi2Te3, and (d) do not themselves diffuse into Bi2Te3 and Ag.
4 Summary
We have demonstrated that Ti (100 nm)/Au (100 nm) layers deposited on Bi2Te3 chips by E-beam thermal evaporation performs well as bonding/barrier composite between Bi2Te3 chips and Ag electrode. After annealing at 250 °C for 200 h, there is no sign of inter-diffusion between Bi2Te3 chips and Ag electrode. The Ti/Au layers still bond well to Bi2Te3 on one side and Ag on the other. The Ti/Au atoms themselves do not diffuse into Bi2Te3 either. Furthermore, the Ti/Au layers provide excellent step coverage on the Bi2Te3 surface with a few microns of surface roughness. The coverage is continuous and complete.
Other bonding/barrier composite designs that were experimented with include Pd, Ni/Au, and Ag, respectively. After annealing at 250 °C for 10 h, they diffuse into Bi2Te3 to produce alloys or form compounds. The resulting compounds are usually weak and break more easily than other layers in the structure.
Acknowledgments
This project was supported by Sandia National Laboratories. Wen P. Lin was partly supported by Materials and Manufacturing Technology fellowship at the University of California, Irvine. Sandia National Laboratories is a multi-program laboratory operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000.
Open Access
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
Open AccessThis is an open access article distributed under the terms of the Creative Commons Attribution Noncommercial License (https://creativecommons.org/licenses/by-nc/2.0), which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
