Brazing of transparent polycrystalline Al2O3 and GH99 with the assistance of (Cu, Ni) solid solution cladding layer

In this study, pure Cu foil was rstly vacuum cladding on the GH99 alloy (GH99) surface to prepare a (Cu, Ni) solid solution layer. By varying the cladding temperatures, (Cu, Ni) solid solution layers with different Ni contents were achieved. The vacuum cladding process was then followed by vacuum brazing of the Cu-coated GH99 to transparent polycrystalline Al 2 O 3 (TPA). Typical microstructure of the TPA/Cu-cladding GH99 brazed joint was characterized. The effects of different cladding temperatures on microstructural evolution and mechanical response of the brazed joints were discussed. By varying the cladding temperature, different thickness of the reaction layer at the braze ller/TPA interfaces can be achieved, which shows a strong correlation with the mechanical performance of the brazed joint. The maximum shear strength of the brazed joint reached 103 MPa when the cladding temperature was 1105 ºC. Compared with the directly brazed joint, shear strength was improved by 472%.


Introduction
Transparent ceramics possess combined properties of ceramics and transparent materials. In addition to sound transmission of light and wave, transparent ceramics also have high thermal and corrosion resistance, high strength and hardness, and excellent chemical stability [1 ~ 3]. Good mechanical properties have made transparent ceramics distinctive candidates for many optical applications.
Transparent polycrystalline Al 2 O 3 (TPA) is a traditional transparent ceramic [4], which has been showing broad application potentials in transparent armors, electromagnetic windows, envelopes of high-pressure halide lamps, etc. [1]. Fabrication of these products needs a reliable joining of TPA to metal components.
GH99 is a nickel-based superalloy with high strength and excellent oxidation resistance. Joining of transparent Al 2 O 3 to GH99 is especially signi cant for fabricating visors or protective windows.
Because most ceramics have high melting points, brazing is the most suitable method to join ceramics [5 7]. Up to now, there are only a few researches on the brazing of transparent ceramics. S. Gambaro et al. [8,9] brazed transparent YAG to Ti6Al4V with various llers, including AgCuTi, AgCu, and Ag. They held that a thin and continuous metal-ceramic layer containing Ti formed in contact with the YAG, which ensured a sound interfacial adhesion and joining property. Liu et al. [10] successfully brazed transparent alumina and TiAl alloy by using an AgCuTi ller. Their results showed that a continuous reaction layer with a thickness of 1 µm formed at the transparent Al 2 O 3 side. The reaction layer was composed of (Cu, Al) 3 Ti 3 O single phase. This interfacial reaction layer was lack of Ti-O compound, which was different from the interfacial reaction layers on traditional alumina [11,12]. Transparent alumina had also been brazed with Ti-based alloy by Li et al. [13] and Benedetti et al. [14]. But detailed discussions about the interfacial reactions of the transparent alumina were not given by them.
Researches on the brazing of normal Al 2 O 3 ceramics can provide us some references. R. Arroyave et al. [15] brazed Al 2 O 3 ceramic with dissimilar metal substrates. They investigated the nature of the thermochemical interactions between the metal substrates (Fe and Ni) and Cu-Ti ller and found that both metal substrates decreased the chemical activity of Ti in the Cu-Ti ller. Such phenomenons are widely explained as follows [15,16]: interreaction between metal element M and Ti decreases the chemical activity of Ti. Then chemical reactions between Al 2 O 3 and Ti are weakened. A similar conclusion had also been given by Valette C. et al. [17] They suggested that in the alumina/AgCuTi/CuNi system, Cu and Ni in the CuNi substrate could severely dissolve into the brazing seam during the brazing process. A strong Ni-Ti interaction would lead to Ti being trapped as Ni 3 Ti compounds, which would dramatically decrease the activity of the reaction between Ti and alumina. The weak interaction between Ti and alumina led to the formation of an extremely thin reaction layer [17].
These studies can give us some enlightenment that the activity of Ti has a signi cant effect on the interfacial reaction degrees. Namely, the type and thickness of an interfacial reaction layer can be controlled by varying the activity of Ti. The mechanical strength of a brazed joint can thus be regulated because the type and thickness of the reaction layer at the ceramic side have been widely agreed to be a critical point in the ceramic/metal brazed joints. In view to this, our investigations aim at preparing (Cu, Ni) solid solution (named as (Cu, Ni) for convenience) layers with different Ni content on the surface of GH99. By varying the Ni content in the (Cu, Ni) layer, the chemical activity of Ti in the brazing seam is expected to be adjusted. In our study, pure Cu was cladding on GH99 at different temperatures in vacuum. During cladding, Ni would diffuse from GH99 to the Cu melt. Because Ni has an increased solid solubility in Cu with the increase of temperature, (Cu, Ni) solid solution layers with different Ni content were achieved. The cladding process was followed by brazing the cladded GH99 with TPA. The relationship between microstructure and mechanical properties of the brazed joint was studied in detail.

Experimental Procedures
TPA and GH99 used in our experiments were commercially obtained. The transparent Al 2 O 3 was polycrystalline, with an average grain size of 20 µm. GH99 was an aging strengthed Ni-based superalloy. The chemical composition of GH99 is listed in Table 1. Cr, Co, W, Al, and Mo are the main alloying elements in the GH99. Pure Cu foils with a thickness of 200 µm were used to prepare the (Cu, Ni) cladding layers on the GH99 plates. Before the cladding process, GH99 was wire cut to 16 mm × 8 mm × 2.5 mm plates. The Cu foils were cut to the same size as the GH99 plates. Surfaces of the GH99 plates and Cu foils were polished by abrasive papers up to grid 1000. Then a Cu foil was carefully put on a GH99 plate and was sent into a vacuum furnace with a vacuum degree of 5 × 10 − 3 Pa. The samples were heated to different peak temperatures (varying from 1090 ºC to 1150 ºC).
It should be noted that the melting point of Cu is 1083 ºC. Therefore, the Cu foils could totally melt at the peak temperatures. The samples were held at the peak temperatures for 1 min in order to make sure that the actual temperatures reach the set temperatures. Then the samples were furnace cooled down. The cladding process is illustrated in Fig. 1 (a).
In the brazing process, TPA was rst cut to 5 mm × 4 mm × 3 mm blocks. The 5 mm × 4 mm surface was chosen as the brazing surface. AgCuTi foils and all the brazing surfaces of the TPA and cladded GH99 were polished up to grid 400. The cladded GH99 was then assembled with a TPA block and an AgCuTi foil, as shown in Fig. 1 (b). The assemblies were then sent into the vacuum furnace and were heated at 860 ºC for 10 min under a vacuum of 5 × 10 − 3 Pa.
Cross-sections of the (Cu, Ni) cladding layers and the brazed joints were characterized by an Empyrean Xray diffraction (XRD), and a Quanta 200FEG scanning electron microscope (SEM) equipped with an energy dispersive X-ray spectroscopy (EDS). To analyze the interfacial products adjacent to the TPA, the focus ion beam (FIB) technique was conducted by an EFI HELIOS NanoLab 600i focused ion/electron double beam electron microscopy at the interface between the TPA and the braze. Detailed FIB sample preparation processes have been depicted by Xu et al. [18]. Then the FIB sample was observed by a Talos F200X eld emission transmission electron microscopy (TEM). Shear tests were conducted by an Instron-5569 electronic universal testing machine, with a head movement speed of 0.5 mm/min. The shear test process is illustrated in Fig. 1 (c). Each shear strength was determined by average the data from three shear samples.

Results And Discussion
3.1 Microstructure of the (Cu, Ni) cladding layer  Table 2, the cladding layer is mainly composed of (Cu, Ni) solid solutions (see Point A) and Cr(s.s) (see Point B). The Cr(s.s) only distribute in the grain boundaries of the (Cu, Ni). At the interface between the cladding layer and GH99, an interfacial diffusion layer can be seen. The interfacial layer is enlarged in the inset of Fig. 2 (a). Alloying elements of the GH99, such as W and Co were largely accommodated in this layer.
To clarify the formation mechanism of the (Cu, Ni) and Cr (s.s) phases in the cladding layer, binary phase diagrams of Cu-Cr and Cu-Ni are exhibited in Fig. 2 (b) and (c). According to the Cu-Ni phase diagram, Ni can completely dissolve both in the solid Cu and the liquid Cu. In addition, Cr also exhibits high solubility in Cu liquid. A high solubility will induce a signi cant dissolution of a solid metal into a molten metal, which has been discussed by R. Arroyave [15]. So it can be expected that when in contact with the Cu liquid, Ni and Cr in GH99 would largely dissolve to the Cu melt. In addition, it can be seen from the Cu-Cr binary diagram that Cu and Cr can form a eutectic liquid phase at 1076.6 °C (see Fig. 2 (b)). The solidus of Cu-Cr binary alloy is lower than the Cu-Ni alloy (above 1084.87 °C), seen Fig. 2 (c). Hence during the solidifying process, (Cu, Ni) solidi ed rstly, followed by the solidi cation of the Cu-Cr binary alloy. During the eutectic solidi cation of the Cu-Cr system, the Cu-Cr system was seperated to a Cu-rich solid solution and a Cr-rich solid solution. Therefore, the Cr-rich solid solution is observed to distribute in the grain boundaries of the (Cu, Ni), as shown in Fig. 2 (a). Table 2 EDS chemical analysis (at.%) of different positions in Fig. 2  The Ni atoms of the (Cu, Ni) were dissolved from the GH99 to the cladding layer. Hence the Ni content in the (Cu, Ni) could be affected by the degree of the dissolution of Ni. Figure  GH99 brazed joint is revealed in Fig. 3. Figure 3 (a) gives the integral microstructure of the brazed joint. TPA and the (Cu, Ni) cladding layer were tightly joined, and the joint was free of voids or cracks. The brazing seam is further magni ed in Fig. 3 (b). Three phases comprise the brazing seam. EDS results of the three phases are listed in Table 3. The white (Point A) and the light grey (Point B) phases mainly contain Ag and Cu elements, respectively, which indicates that they are Ag-based and Cu-based solid solutions. Additionally, it is notable that some small dark grey blocks appear near the (Cu, Ni) cladding layer (point C), which are speculated as Cu-Ni-Ti ternary compound according to EDS analysis. The Cu-Ni-Ti ternary compound is further veri ed to be CuNi 2 Ti intermetallic by XRD analysis, see Fig. 4. The formation of CuNi 2 Ti intermetallic has also been found in the research of A. A. A. Atiyah et al. [19], in whose research they prepared NiTiCu alloys by powder metallurgy at 850 ºC. The interface between the TPA and the braze is further magni ed in Fig. 3 (c). It can be seen that a thin interfacial layer was generated. In order to analyze the crystal structure of the reaction layer, the joint was immersed into a nitric acid solution to etch the braze ller off and expose the reaction layer on the TPA. Then the reaction layer was characterized by XRD analysis. XRD results in Fig. 4 Fig. 5 (a). Figure 5  has been largely reported to be a typical reaction product in the AgCuTi/Al 2 O 3 brazing system [11,12].
Also, Ni 2 Ti 4 O has also been detected by C. G. Zhang et al. [20] in the Al 2 O 3 /Kovar brazed joint. In our study, both Cu and Ni elements evenly distribute in the reaction layer (see Fig. 5  As shown in Fig. 6 (a), there is also a 20 nm ultra-thin layer well between the TPA and the Cu 3 Ti 3 O + Ni 2 Ti 4 O layer. This ultra-thin layer is further magni ed in an HRTEM image, see Fig. 6 (b). HRTEM image of the interface between the ultra-thin layer and the TPA is FFT transferred to diffraction patterns, as depicted in the inset of Fig. 6 (b). From the diffraction patterns, two phases can be indexed, which are the hexagonal alumina and the hexagonal Ti 2 O 3 . It also can be found from the diffraction patterns that the alumina and the Ti 2 O 3 phases follow a particular orientation, which is This phenomenon is also shared by R. Voytovych et al. [24]. In their research, Ag-Cu-Ti llers with different Ti content were applied to wet Al 2 O 3 ceramic. They found that as Ti content decreased, the Ti-O compound with a lower degree of Ti would form at the interface.
In addition, speci c orientations between Ag and Al 2 O 3 can be found.
From the HAADF images shown in Fig. 7 (a) and Fig. 8 (a), it can be seen that the grain boundaries of the Cu 3 Ti 3 O + Ni 2 Ti 4 O reaction layer exhibit a relatively bright color. It indicates that heavy metal elements have gathered in the grain boundaries. The elements include Ni, Co, and Cr, which can be con rmed by map distribution images given in Fig. 8 (b)~(d). Grain boundaries are the most common short-circuiting paths for diffusion [26]. Hence grain boundary diffusion can be orders of magnitude faster than diffusion through a grain. 3.3 Effect of cladding temperature on microstructure and mechanical strength of the TPA/GH99 brazed joint The GH99 plates cladded by (Cu, Ni) with different Ni content were brazed with TPA. Figure 9 (a)~(d) are the corresponding images of the joints brazed at 860 ºC for 10 min, with the cladding temperatures varying from 1090 ºC to 1150 ºC. Namely, the Ni content of the (Cu, Ni) layer varied from 7 at.% to 22 at.% in Fig. 9 (a)~(d). During the brazing process, the (Cu, Ni) layer would continually dissolve to the brazing seam, which led to a decrease in the thickness of the (Cu, Ni) layer. After cooling down, it can be seen from Fig. 9 that (Cu, Ni) layers remained for all cases. Accordingly, the (Cu, Ni) layer could be a barrier layer that prevented the Ni from severely diffusing from GH99 to the brazing seam.
Microstructures of the brazing seams are further enlarged, as shown in the insets of Fig. 9 (a)~(d). In the brazing seam, the formation of the CuNi 2 Ti compound is due to the chemical reaction between Ti and the Ni that dissolved from the (Cu, Ni) layer. For the (Cu, Ni) layer with a higher Ni content, a larger amount of Ni would dissolve from the (Cu, Ni) layer to the brazing seam. Therefore, a more massive amount of Ti would be attracted by the dissolved Cu and Ni elements to form CuNi 2 Ti compounds. Hence it can be seen from Fig. 9 (a)~(d) that with the increase of Ni content of the (Cu, Ni) cladding layers, the amount of the CuNi 2 Ti in the brazing seam increased as well.
The TPA/braze interfaces of the brazed joints with different Ni contents in the (Cn, Ni) cladding layers are further enlarged in Fig. 10 (a)~(d). When the Ni content was 7 at.% (the cladding temperature was 1090 °C), the thickness of the reaction layer was about 1 µm. When the Ni content was increased to 10 at.% (the cladding temperature was 1105 °C), the thickness dramatically decreased to 0.6 µm. As discussed above, this phenomenon is mainly induced by the reduction of the activity of Ti. A weak activity of Ti leads to a weak interfacial reaction between Ti and TPA. When the Ni content further increases to 14 at.% (the cladding temperature is 1120 °C), however, the continuity of the reaction layer is destroyed. The formation of the thin and discontinuous reaction layer is caused by a further decrease in the activity of Ti. When the cladding temperature is raised to 1150 °C (the Ni content is increased to 22 at.%), the reaction layer totally disappears, as shown in Fig. 10 (d). Accordingly, the thickness of the interfacial reaction layer could be regulated by varying the activity of the active Ti element.
The TPA and GH99 were also directly brazed by an AgCuTi foil at 860 ºC for 10 min for a comparison purpose. When TPA and GH99 were directly brazed, the signi cant diffusion of Ni would induce the formation of a mass of intermetallic compounds in the brazing seam, see Fig. 11. EDS results show that the chemical composition of the compound is 76.9 at.% Ni and 23.1 at.% Ti. Therefore, the Ni 3 Ti brittle compounds severely aggregate in the brazing seam, which is terrible for relieving the residual thermal stress of the joint. The severe Ni diffusion would weaken the activity of Ti. Hence the reaction between Ti and TPA would be faint. Accordingly, the interfacial layer was not observed at the TPA/braze interface in the inset of Fig. 11. Because of the large amount of the brittle compounds and the poor interfacial bonding, the directly bonded joints exhibited poor mechanical properties. In our experiments, penetrating cracks were observed in all directly bonded samples.
Plenty of researches has suggested that the degree of interfacial reactions could signi cantly affect the mechanical properties of a brazed joint [27,28]. Dai [29]. Thus it can be inferred that in our brazing system, the TPA/braze interfacial reaction phase also exhibited poor plasticity. A too thick brittle interfacial layer is adverse for the joint to relieve the residual interfacial stress.
Whereas if the interfacial reaction layer is quite thin, the bonding between the TPA and the braze is weak. Hence to achieve a maximum mechanical strength of the brazed joint, there should be an optimized thickness of the interfacial reaction layer. As has been discussed, (Cu, Ni) cladding layers with different Ni content could be prepared by cladding Cu foils on GH99 at different temperatures. The varied Ni content of the (Cu, Ni) layers could regulate the activity of Ti. Thus the thickness of the reaction layer between the TPA and the braze could be controlled via the (Cu, Ni) cladding method. Shear strength of the brazed joints without cladding and with different cladding temperatures are given in Fig. 12. With the increase of the cladding temperature, the shear strength of the brazed joints improves then decreases. When the cladding temperature is 1105 °C, the maximum shear strength of the brazed joint is achieved, which is 103 MPa. Compared with the joint brazed without the assistance of the (Cu, Ni) cladding layer, shear strength was improved by 472%. Variation of the mechanical strength is recognized as a result of the regulation of the thickness of the reaction layers by the (Cu, Ni) cladding layer.

Conclusions
In this study, a (Cu, Ni) layer was prepared by cladding a pure Cu foil on GH99. Then TPA and the cladded GH99 were successfully brazed with the assistant of the (Cu, Ni) cladding layer. Interfacial bonding between the TPA and the AgCuTi braze was studied in detail. The in uence of the Ni content in the (Cu, Ni) layer on the microstructure and mechanical property of the brazed joint was also investigated. Based on the analysis above, the main conclusions were achieved, which are as follows: ( (4) The thickness of the reaction layer corresponds to the strength of the brazed joint. By optimizing the Ni content of the (Cu, Ni) layer, the maximum shear strength of the brazed joint was achieved, which was 103 MPa when the cladding temperature was 1105 ºC. Compared with the TPA/AgCuTi/GH99 directly brazed joint, shear strength was improved by 472%. Sketches of (a) the vacuum cladding process, (b) the brazing assembly, and (c) the shear test assembly   Microstructure of (a) the brazed joint with the assistance of (Cu, Ni) cladding layer, (b) the brazing seam enlarged from Fig. 3 (a), and (c) the TPA/braze interface enlarged from Fig. 3

(b)
Page 17/25 Figure 4 XRD results of (a) the nitric acid solution treated TPA surface and (b) the brazing seam          Shear strengthes of the joints brazed with and without the assistant of the (Cu, Ni) cladding layers.