Secondary Phase Interaction at Interfaces of High-Strength Brazed Joints made using Liquid Phase Sintered Alumina Ceramics and Ag-Cu-Ti Braze Alloys

Alumina-to-alumina brazed joints were formed using 96.0 and 99.7 wt.% Al2O3 ceramics using 150 µm thick Ticusil® (68.8Ag-26.7Cu-4.7 wt.% Ti) braze preforms. Brazing was conducted in a vacuum of 1 × 10−5 mbar at 850 °C for 10 minutes. Joint strengths were evaluated using four-point bend testing and were compared to the monolithic flexural strengths of standard alumina test bars according to ASTM C1161-13. Brazed joints made using 96.0 wt.% Al2O3 consistently outperformed brazed joints made using 99.7 wt.% Al2O3, despite similarities in both the flexural strengths of the standard alumina test bars and the microstructures of brazed joints. Secondary phase interaction led to the formation of Ti5Si3 reaction products at locations where the triple pocket grain boundaries of the 96.0 wt.% Al2O3 surface intersected the Ti-rich reaction layers. It is proposed that due to this interaction, brazed joints made using 96.0 wt.% Al2O3, which were relatively cost-effective to produce, achieved higher strengths than brazed joints made using 99.7 wt.% Al2O3.

joints made using Ag-Cu-Ti braze alloys, can also influence the chemical activity of Ti and its diffusivity towards the joint interfaces 8,22,23 .
The strengths of alumina-to-alumina brazed joints made using Ag-Cu-Ti braze alloys have previously been correlated to the thickness of the reaction layer, which can provide a good indication of chemical bonding at the joint interface 24,25 . Since the reaction layer is usually composed of a brittle phase, however, if excessively thick, degradation in joint strength can also occur. The thickness of the Ag-Cu braze interlayer (assuming complete diffusion of Ti to the joint interfaces) has also been found to influence the strength of alumina-to-alumina brazed joints [26][27][28] . The braze interlayer is the only ductile constituent in the alumina/Ag-Cu-Ti system which can plastically deform to accommodate thermally induced residual stresses. Hence, the residual stress state of a brazed joint can strongly affect the maximum stress that a brazed joint may withstand during mechanical testing and in-service.
Recently, excess Ti in the braze interlayer in the form of isolated Cu 4 Ti and Cu 4 Ti 3 compounds was found to significantly improve the strength of alumina-to-alumina brazed joints 28 . This resulted from a controlled increase in the Ticusil braze preform thickness, which led to corresponding increases in both the reaction layer and braze interlayer thicknesses. In the same study, Kassam et al. (2015) 28 found that joints made using 96.0 wt.% Al 2 O 3 consistently outperformed those joints made using 99.7 wt.% Al 2 O 3 , despite similarities in both the flexural strengths of standard alumina test bars and the microstructures of the brazed joints.
In the literature, most studies have opted to use solid-state sintered alumina ceramics, composed of at least 99.0 wt.% Al 2 O 3 when performing brazing experiments using Ag-Cu-Ti braze alloys (Fig. 1). Less than ~20% of studies have selected liquid phase sintered alumina ceramics, whilst in other studies the alumina purity has not been reported. The superior mechanical properties of solid-state sintered alumina ceramics in comparison to liquid phase sintered alumina ceramics 29 may have influenced these selections. The effect of secondary phase elements, present in liquid phase sintered alumina ceramics, on the interfacial chemistry and resulting joint strength, could not be found in the literature. This formed the basis of the investigation in this study.

Figure 1.
A summary of the grades of alumina selected in wetting experiments and in the formation of alumina-to-alumina brazed joints, using Ag-Cu-Ti braze alloys  . Underline denotes wetting experiments as opposed to the formation of alumina-to-alumina brazed joints. +denotes alumina ceramics were processed in-house, according to the following: Cho et al. (1991) 19

Results and Discussion
In this study, the liquid phase sintered 96.0 wt.% Al 2 O 3 selected was found to contain 3.2 wt.% SiO 2 as the main secondary phase, as well as 0.55 wt.% MgO and 0.06 wt.% CaO. The solid-state sintered 99.7 wt.% Al 2 O 3 selected was found to contain 0.30 wt.% SiO 2 , 0.03 wt.% MgO and 0.02 wt.% CaO (Fig. 2). This phase analysis was performed using electron probe microanalysis, with average values based on 10 measurements.
The presence of Si at the interfaces of the alumina-to-alumina brazed joints made using 96.0 wt.% Al 2 O 3 was observed, using SEM-EDX techniques, at locations where the triple pocket grain boundary regions of the alumina surface intersected with the Ti-rich reaction layers; beneath a γ -TiO layer on the alumina side of the joint interface (Fig. 3). These locations, characterised using backscattered electron imaging, showed the formation of a relatively darker phase in comparison to the Ti 3 (Cu+Al) 3 O layer which formed on the braze side of the joint interface. These results appeared to show that Ti and/or Cu from the braze alloy had penetrated approximately 300 to 500 nm deep into the triple pocket grain boundary regions of the faying surfaces of 96.0 wt.% Al 2 O 3 , where these elements had interacted with the relatively lower atomic number elements Si and/or Mg. This secondary phase interaction was observed at every location where the triple pocket grain boundary regions of the 96.0 wt.% Al 2 O 3 surface intersected with the Ti-rich reaction layers (Fig. 4). No such observation was made at the interfaces of the alumina-to-alumina brazed joints made using 99.7 wt.% Al 2 O 3 .
Using an in-situ lift out method performed using focussed ion beam milling, TEM specimens were prepared from locations of the interfaces of joints made using 96.0 wt.% Al 2 O 3 , where secondary phase interaction had been observed. STEM-EDX analysis indicated that the secondary phase interaction had led to the formation of Ti 5 Si 3 , with the composition of 67.4Ti-26.9Si-5.7O wt.% (Fig. 5). These Ti 5 Si 3 compounds were observed to have formed between a triple pocket secondary phase grain, with composition 25.6Si-10.0Al-7.0Mg-57.4O wt.%, and the Ti-rich reaction layers; γ -TiO and Ti 3 (Cu+Al) 3 O. The resulting Ti 5 Si 3 compounds, which appeared to have formed following the partial consumption of both the secondary phase grains and the Ti-rich reaction layers, were dispersed in an area that was approximately 1.0 µm wide. At the centre of this interaction region, several Ag particles that were 30 to 60 nm in diameter were also observed ( Fig. 6).
Electron diffraction was used to confirm the hexagonal Ti 5 Si 3 phase of the P63/mcm (193) space group with lattice parameters a = b = 7.42 Å and c = 5.15 Å (Fig. 7). This was consistent with the selected area diffraction patterns of the Ti 5 Si 3 phase, which have been reported elsewhere in the literature 30 .
A second TEM specimen from a similar interfacial region of another brazed joint made using 96.0 wt.% Al 2 O 3 showed a similar result i.e. secondary phase interaction. Again, the formation of Ti 5 Si 3 , with composition 74.1Ti-23.3Si-2.6O wt.% was observed at locations where the triple pocket grain boundary regions of the 96.0 wt.% Al 2 O 3 surface intersected with the Ti-rich reaction layers (Fig. 8). At the centre of the 0.5 µm wide interaction region, small Ag particles were again observed.  The results confirmed the formation of Ti 5 Si 3 at the interfaces of brazed joints made using 96.0 wt.% Al 2 O 3 and Ticusil. This may have formed due to the presence of SiO 2 at the triple pocket grain boundary regions of the 96.0 wt.% Al 2 O 3 surface reacting with Ti, following the diffusion of Ti to the joint interfaces. A possible reaction mechanism is described in equations (1) and (2). It is postulated that following the reduction of SiO 2 by Ti, the liberation of Si and its subsequent reaction with Ti led to the formation of Ti 5 Si 3 . These results were similar to the formation of Ti 5 Si 3 reported in a study by Kang and Selvarian 31 , in which a 3 µm thick Ti coating was deposited onto the surface of a 95.0 wt.% Al 2 O 3 ceramic before the coated alumina specimen was heated in a vacuum atmosphere to 980 °C for 10 minutes, simulating a typical brazing experiment.

3
The formation of Ti 5 Si 3 has previously been reported to occur at the interfaces of joints made using both SiC 32,33 and Si 3 N 4 32,34-41 ceramics and Ti-containing active braze alloys. In these systems, a Ti 5 Si 3 layer can typically form as a result of either free Si on the ceramic surface, or from Si that is liberated following the reduction of  the SiC or Si 3 N 4 ceramic surface by Ti. In both of these systems, preventing the formation of a Ti 5 Si 3 layer on the ceramic side of the joint interface (which can act as a diffusion barrier to Ti), whilst also preventing the diffusion of Si into the braze interlayer (which can lead to brittle reaction products), have been reported to help minimise any degradation in joint strength 41,42 . The formation of a Ti 5 Si 3 layer in these systems, therefore, has typically been found to adversely affect joint strength. In this study, however, Ti 5 Si 3 was not observed as a continuous layer at the joint interface but instead as 0.5 to 1.0 µm wide interaction regions. The formation of Ti 5 Si 3 was observed only at locations where the triple pocket grain boundary regions of the 96.0 wt.% Al 2 O 3 surface intersected with the Ti-rich reaction layers (Fig. 9). At these locations, the Ti 5 Si 3 compounds did not appear to act as a diffusion barrier to Ti. Further, Si was not found to have diffused into the braze interlayer.
In addition to the γ -TiO and Ti 3 (Cu+Al) 3  The bulk microstructures of both of sets of joints, made using 150 µm thick Ticusil braze preforms, were also similar. The braze interlayer in these joints consisted of Ag-rich and Cu-rich phases in a eutectic-like distribution, as well as several Cu-Ti phases in central regions. These Cu-Ti phases formed as a result of excess solid Ti in the braze interlayer, following the formation of stable γ -TiO and Ti 3 (Cu+Al) 3 O layers at the joint interfaces 28 .
The only significant difference between the brazed joints made using these two grades of alumina, which were both ground in the same way according to ASTM C1161-13, was the presence of Ti 5 Si 3 at the interfaces of joints made using 96.0 wt.% Al 2 O 3 . This may have, therefore, provided the improvement in joint strength observed, which could be explained based on two possible mechanisms.
The regular spatial distribution of the Ti 5 Si 3 phase at locations of the interface where the triple pocket grain boundary regions of the 96.0 wt.% Al 2 O 3 surface intersected with the Ti-rich reaction layers may have led to an improvement in the gradation of the coefficient of thermal expansion (CTE) across the joint interface. The Ti 3 (Cu+Al) 3 O layer, with a CTE value of 15.2 × 10 −6 . °C −1 , introduces a gradual transition in the CTE across the joint interface between alumina, which has a relatively lower CTE value of 8.1 to 8.5 × 10 −6 . °C −1 , and the braze interlayer, in which the Ag-rich and Cu-rich phases have CTE values of 19.2 × 10 −6 . °C −1 and 22.0 × 10 −6 . °C −1 , respectively. The CTE of a Ti 5 Si 3 crystal lattice exhibits high anisotropy, with CTE values of 5.9 × 10 −6 . °C −1 along the a-axis and 16.9 × 10 −6 . °C −1 along the c-axis 43 . The formation of Ti 5 Si 3 reaction products may have made the residual stress distribution at the joint interface relatively complex. Secondary phase interaction may have further graded the CTE transition across the joint interface leading to improvements in both the residual stress state of the joints and the joint strength (Fig. 10).
A second potential mechanism may be that the regular spatial distribution of Ti 5 Si 3 compounds at locations of the interface where the triple pocket grain boundary regions of the 96.0 wt.% Al 2 O 3 surface intersected with the Ti-rich reaction layers provided a nanostructured interlocking mechanism. This may be similar to the way in which ~20 nm thick Ti 5 Si 3 dendrites were suggested to improve the strengths of glass-to-Ti-6Al-4V joints 44 . Further, the formation of these Ti 5 Si 3 phases may have provided additional chemical bonding at the interfaces of 96.0 wt.% Al 2 O 3 braze joints, enabling relatively higher levels of both, thermally induced residual stresses and applied stresses, to be transferred into the relatively ductile matrix of the braze interlayer. Secondary phase interaction, therefore, may have provided a nanostructured interlocking mechanism which led to the improvement in joint strength observed.

Conclusions
In the alumina/Ag-Cu-Ti joining system a wide range of alumina ceramics have been selected and these alumina ceramics have predominantly been solid-state sintered. The presence of Si at the interfaces of brazed joints made using liquid phase sintered alumina ceramics and Ag-Cu-Ti braze alloys has previously been observed using  8,16,28 . The formation and positioning of Ti-Si phases at the joint interface and their effect on joint strength, however, has not previously been reported in the literature. It is hereby reported that alumina-to-alumina brazed joints made using liquid phase sintered 96.0 wt.% Al 2 O 3 ceramics can achieve superior joint performance in comparison to brazed joints made using solid-state sintered 99.7 wt.% Al 2 O 3 ceramics, formed under the same brazing conditions. The formation and regular spatial distribution of Ti 5 Si 3 at locations where the triple pocket grain boundary regions of a 96.0 wt.% Al 2 O 3 surface intersects with the Ti-rich reaction layers may, (i) benefit the residual stress distribution at the joint interface by enabling an improved gradation in the coefficient of thermal expansion across the joint interface, and (ii) provide a nanostructured interlocking mechanism through additional chemical bonding, enabling greater load transfer to the relatively ductile braze interlayer. It is proposed, therefore, that due to the presence of a Si-rich secondary phase, liquid phase sintered alumina ceramics, which are relatively cheaper to manufacture than solid-state sintered alumina ceramics, can be more suitable for achieving high-strength brazed joints. Alumina-to-alumina brazed joints, whereby short test bars were brazed to themselves, were prepared by first arranging short test bars in a butt-joint configurations. Several of these joint assemblies, aligned vertically, were supported in a bespoke stainless steel fixture, designed to allow uniform heating of each joint interface during brazing. The brazing fixture and fittings were spray coated using the boron nitride based LOCTITE SF 7900 also known as AERODAG CERAMSHIELD. This was to act as a stop-off, protect the brazing fixture during brazing, and prevent any undesirable effects such as the alumina test bars reacting with, or becoming clamped in, the stainless steel brazing fixture.  For each joint assembly, a braze preform was placed between the faying surfaces of two short test bars and five joints were produced in each brazing cycle. The braze preforms had dimensions of 7 × 5 mm and were mechanically punched from the 150 µm thick Ticusil braze foil. No additional load other than the self-weight of the upper short test bar in each joint assembly was applied. Prior to each brazing cycle, the short test bars, braze preforms, and the brazing fixture and fittings, were all ultrasonically cleaned in acetone for 15 minutes.
Brazing was performed in a Vacuum Generation Ltd vacuum furnace, at a pressure of 1 × 10 −5 mbar. Thermocouples, placed in the brazing fixture, were used to control the furnace temperature. A peak brazing temperature of 850 °C and a brazing time of 10 minutes followed a 10 minute isothermal soak at 750 °C. This enabled the temperatures of the brazing fixture and the joint assemblies to homogenise before the peak brazing temperature was reached. The heating and cooling rates were 10 °C/min and 5 °C/min, respectively. Alumina crucibles containing Ti granules were used as getters to reduce the partial pressure of oxygen in the vacuum furnace during brazing.
Both the standard test bars and brazed joints were mechanically tested using four-point bend testing, which was conducted at ambient room temperature according to the ASTM C1161-13 45 .
Polished sections of both, the standard alumina test bars and cross-sections of the brazed joints, were first studied using a Zeiss ΣIGMA ™ scanning electron microscope (SEM). Electron probe microanalysis (EPMA) was performed using a Cameca SX-100 with five wavelength dispersive spectroscopy detectors and an FEI Nova 600 Nanolab Dualbeam FIB/FEG-SEM was used to prepare thin lamella specimens for transmission electron microscopy (TEM) analysis using the in-situ 'lift-out' method. These specimens were studied using a JEOL 2100 field emission gun TEM with scanning-TEM (STEM) capabilities set to an operating voltage of 200 kV. TEM electron micrograph and electron diffraction micrograph images were acquired using a Gatan Orius CCD camera, while STEM images were acquired using Jeol bright-field and/or dark-field detectors and Gatan Digiscan image acquisition software. Image processing of micrographs was carried out using Gatan Microscopy Suite. Energy dispersive X-ray (EDX) analysis and STEM-EDX mapping was carried out using a Thermo Scientific Noran 7 silicon drift detector EDX system coupled with NSS software.
Data Availability. All metadata pertaining to this work can be accessed via the following link: https://doi. org/10.17633/rd.brunel.5812995.