**1. Introduction**

Nickel–aluminum system is well known and commonly researched reactive system. This is associated with the wide practical application of the AlNi3 and AlNi intermetallic phases, which are used on a large scale in many industrial branches for example, as multilayers, oxidation resistant coatings (aluminizing treatment of nickel alloys), turbine blades (aircraft industry), electronics industry. To ensure the protection against the oxidation, aluminum-rich phases such as Al3Ni, Al3Ni2 and AlNi are the most important ones, however, the greatest mechanical properties are associated with the presence of only two phases: AlNi and AlNi3 [1–3].

β-AlNi and γ-AlNi3 are cubic phases possessing B2 and L12 structure types, respectively. These intermetallics are characterized by high degree of order at elevated temperature and indicate high mechanical strength even at high temperature. Also they possess high thermodynamic stability in wide range of chemical composition and high degree of crystal lattice order with various amounts of defects. β-AlNi phase composition varies in the wide range from 40 to 55 at. % of Al at 700 ◦C. Additionally, within its structure many defects can be formed. Bradley and Taylor [4] have shown that nickel replaces aluminum in its lattice causing the excess of the nickel, however, in Ni-deficient compounds, aluminum does not replace the nickel in its lattice, causing formation of vacancies. It explains high number of structure's defects. As a result of high temperature diffusion processes, γ-AlNi3 intermetallic phase can be created in the reaction between AlNi and Ni. Base structural Ni3Al cell crystallizes in cubic lattice, where aluminum atoms are in the corners and nickel atoms are at the center of cell walls [4–6].

Many works consider the diffusion processes in Ni-Al system, however, as variables, only parameters such as annealing time and temperature of reaction either in solid state or solid/liquid state are analyzed. Additionally, the system is studied in many configurations with respect to the

chemical composition of initial substrates (end members) and various their combinations [3,5,7–11]. Moreover, different experimental procedures have been employed in these studies such as: Ni/Al diffusion couples [3,5,8–10,12], Ni/Al/Ni sandwiches [7,13–15], or more complex assemblies such as multilayers, nanocoatings or aluminized nickel alloys [16–21].

One of the promising joining processes is the diffusion soldering (DS) [13,22–24], which is schematically presented in Figure 1. The main stage of this process is the isothermal solidification, where the reaction of liquid low melting interlayer (LM) and solid substrates (HM) takes place. With increasing of the DS temperature to the appropriate one—required in the process, the low melting component turns into the liquid state and the reaction at the solid/liquid interface (between the high and low melting components) begins. After some time, the first intermetallic phase is formed, then, in the next stage of DS diffusion in solid state between the obtained intermetallic phase and the high-melting substrate takes place. When the whole liquid metal is consumed, reaction proceeds only in the solid state. Phases are created and consumed one by one or simultaneously due to inter-diffusion, being replaced by the phases enriched more and more in element of higher melting point.

**Figure 1.** Schematic of diffusion soldering process [22].

In the Ni/Al/Ni system the sequence of appearance of the intermetallics can be predicted based on the Al-Ni equilibrium phase diagram (Figure 2) [25]. The phases are being created going from the lower melting to higher melting component. What is interesting, there are contradictory literature reports concerning the sequence of the intermetallics formation in Ni/Al system, regarding the order of formation of the high aluminum intermetallic phases such as Al3Ni and Al3Ni2. Two opposite approaches are possible. First one involves situation, when as a primary phase Al3Ni phase precipitates and then Al3Ni2 one is created [13,14,26]. Some modeling results and in-situ experiments indicated on formation of Al3Ni2 phase as the first one and then Al3Ni [27–29]. As the process proceeds, the intermetallic compounds such as: AlNi, Al3Ni5 and AlNi3 are formed [13]. In the cited works the chemical composition was verified mainly by EDS technique in SEM [13,14,26–29]. Additionally, in [26,29] works, the XRD measurements were also conducted to confirm the phases' composition present in the interconnection zones, while in [27,28] for the phase sequence occurrence the mathematical model was proposed.

**Figure 2.** Equilibrium phases diagram Ni–Al with distinction between various types of the AlNi phase [25].

Literature's survey also presents the validity of diffusion studies on Ni substrates of variable grain boundaries. Works of Yu-Chen Tseng et al. [30] based on liquid–solid Sn/Ni system indicated significant differences in the course of the diffusion process conducted with applied commercial coarse-grained Ni substrates and fine-grained Ni layer electroplated (e-Ni) on Ba2Te3. In both type of samples after the annealing, the Ni3Sn4 phase is formed, but for e-Ni substrate, additional Ni3Sn2 phase in the form of the thin film between Ni and Ni3Sn4 appeared. A large number of grain boundaries in e-Ni layer, being fast diffusion channels for Ni atoms towards Ni3Sn4, resulted in the growth of this additional phase, consisting of more nickel. The additional layer suppressed the growth of Ni3Sn4 on e-Ni layer, which is thinner in comparison to the one, formed using coarse grained Ni substrates. A similar experiment in solid Ni/liquid Al/solid Ni system is more difficult to analyze, as higher joining temperature is necessary to apply than the one for Sn/Ni diffusion pair and the nickel recovery and recrystallization phenomena occurs.

This paper shows the results concerning the relation between the reactivity (intermetallic phases growth kinetics) in Ni/Al system and the microstructure of nickel substrates. As the grain boundaries can act as fast diffusion paths, they can influence the growth kinetics of the intermetallic phases and therefore, different results can be reported for apparently the same experimental procedure. In this work Ni substrates of two crystallographic orientations was employed to study the growth of the intermetallic phases in solid/liquid and solid/solid state in Ni/Al/Ni system.

#### **2. Materials and Methods**

Substrates used in the experiment were prepared from high purity commercial Ni rod (99.999, (Goodfellow Cambridge Ltd., Huntingdon, UK) with a diameter of 5 mm. The rod was cut in two orientations: along and perpendicular to the rod elongation direction as it is showed in Figure 3. EBSD maps revealed the different crystallographic orientation of both types of substrates. Optical microstructures and EBSD maps show significant difference in the appearance of both substrates, for substrate NiA-type (Figure 3a) grains are elongated and narrow. On the other hand, for the NiB-type substrate (Figure 3b), grain possess irregular shape and on the map considerable refinement of the structure is visible.

**Figure 3.** The scheme of the preparation of nickel substrates with their optical microstructures and electron backscattered diffraction (EBSD) maps (**a**) NiA-type, (**b**) NiB-type.

In order to obtain the diffusion-soldered interconnections, two nickel slices with the same or different orientation were grinded, polished and cleaned in ultrasonic cleaner for 300 s. Then the thin (80 μm) slice of high purity aluminum (99.999, Goodfellow Cambridge Ltd., Huntingdon, UK) was clamped between two Ni substrates and held in specific temperature for different periods of time. As the aluminum melting point is of 660 ◦C, the 720 ◦C was applied as the joining temperature. Such temperature, higher than the one necessary to melt Al, was chosen for several reasons. First of all, it ensures that Al passes to the liquid state during the annealing in the vacuum. It also allows comparing obtained results with the data presented by Lopez et al. [13]. Only slight mechanical pressure was used to avoid of leakage of the solder and the samples were sealed in quartz ampules to prevent the samples' oxidation. Table 1 shows variety of applied experimental assemblies and conditions of their annealing.


**Table 1.** The experimental assemblies and conditions. NiA denotes the substrate cut to the parallel direction to the elongation rod, while the NiB in the perpendicular direction.

The cross-sections of the interconnections for scanning electron microscopy (SEM, Quanta 3D FEG, FEI, Hillsboro, OR, USA) examinations were prepared by standard metallographic procedure: embedding samples in epoxy, grinding and then polishing with the diamond paste (3 μm) and silica (0.04 μm). As a starting point the scanning electron microscopy observations with 20 kV accelerating voltage and energy dispersive X-ray spectroscopy (EDS, Trident (EDS-EBSD-WDS), EDAX Inc., Tilburg, The Netherlands) analysis were carried out for each sample, revealing its phase composition and thickness of particular intermetallics layers and also the chemical composition changes across them. The samples in SEM were inspected using the backscattered electrons mode (BSE). As a next step, the electron backscattered diffraction technique was used to expose the amount and character of the grain boundaries and to correlate it with the creating phases composition and thickness. The thin foils for the transmission electron microscopy (TEM, TECNAI G2 200 kV, FEI, Hillsboro, OR, USA) observations were prepared using the Focused Ion Beam (FIB, Quanta 2D, FEI, Hillsboro, OR, USA) technique. It is the only technique suitable for this type of thin foils (exact location of the place of interest). During the milling process, problems with uneven consumption of the sample material were encountered. Difficulties were associated with significant difference in hardness of the various intermetallics phases located at the reaction zone. The obtained thin foils with the thickness of about 100 nm were next examined by TEM.

#### **3. Results and Discussion**

#### *3.1. Sequence of Intermetallic Phases in Interconnections*

The diffusion-soldering at the temperature of 720 ◦C for different periods of time resulted in the growth of the several intermetallic phases in the joined area. It is important that the sequence of phase-creation in all cases was the same as it was predicted in [13,14,26]. They grew according to equilibrium phases diagram from the ones rich in low melting component to the ones with higher amount of nickel. Sequence of their appearance in the interconnection zone depended on the duration of reaction.

The initial stage of reaction in Ni/Al/Ni interconnection was observed after 15 min of annealing (Figure 4a). SEM observations using BSE mode showed the contrast differences at the Ni/solder interface, pointing the existence of two intermetallic phases. The measurements of the chemical composition within the interconnection zone confirmed that these phases were Al3Ni (76.0 at. % Al, 24.0 at. % of Ni) and Al3Ni2 (60.6 at. % of Al, 39.4 at. % of Ni). On the other hand, the middle of the joined zone was composed of Al3Ni-Al eutectics (97.1 at. % of Al, 2.9 at. % of Ni) instead of pure aluminum. Moreover, inside of the Al3Ni-Al eutectics, the primary precipitates of the Al3Ni intermetallic phase possessing the faced walls could be observed. Thanks to the channeling contrast, the dual-morphology of Al3Ni2 is visible, showing the larger grains to be located closer to the middle of the interconnection and finer grains being located close to the nickel substrates. As it is showed in Figure 4a, the Al3Ni phase, growing next to the nickel substrate, formed the areas of irregular shape at the interface with eutectics. Such a morphology is called scallops and it is typical for the growth of the intermetallics with assistance of the liquid. Additionally, it was noticed that the interface between Al3Ni and Al3Ni2 phases is wavy. At this point it can be summarized that the interconnection consisted of the following constituents:
