**Ni/Nisolidsolution/AlNi3/AlNiNi-rich/AlNi/AlNiNi-rich/AlNi3/Nisolidsolution/Ni.**

Average chemical composition of all intermetallic phases after annealing for different periods of time is collected in Table 3.

**Figure 7.** SEM microstructures of the Ni/Al/Ni couples obtained at 720 ◦C after annealing for: (**a**) 20 and (**b**) 72 h using A- and B- type of substrates of Ni. Numbers 1–5 denote particular intermetallic phases: 1-AlNiNi-deficient, 2-AlNi, 3-AlNiNi-rich, 4-AlNi3, 5-Ni solid solution.

**Table 3.** Chemical composition of the intermetallic phases in Ni/Al/Ni joints after annealing for different periods of time.


The interfaces, where the solid/solid diffusion occurs were also examined by EBSD technique. Due to sufficient width of particular phases, which grow with time, the sample annealed for 20 h at 720 ◦C was selected. The EBSD map in Figure 8a indicates the existence of three main areas, which are indexed starting from the nickel as: Ni, AlNi3 and AlNi. The map shows that the zone of interest consists of grains with a random crystallographic orientation and the one large grain of nickel. The differences in size of grains for particular areas are visible. Much finer grains in comparison to other areas are observed for the intermetallic phase identified as AlNi3. On the other hand, in the case of AlNi phases large grains are observed. In both cases range of these phases grain sizes is variable. Complementary to the EBSD map, for the same area the EDS maps were collected (Figure 8b). As can be noticed, these EDS maps show more individual phases compared to EBSD measurement. Area between Al3Ni and AlNi intermetallic phases is rich in nickel, in comparison to AlNi phase, while at the second side of AlNi phase, the area deficient in nickel is present. This dependency is compatible with SEM micrograph registered in BSE mode for the same area and it is shown in Figure 9. For full understanding of the observed relation, the EBSD indexing confidence map was imposed in the EDS maps for Al and Ni elements (Figure 8c). The result was surprising, as it was mentioned above, the EBSD map does not indicate the existence of more than three phases. However, the combination of EBSD and EDS maps reveals the concentration gradient throughout the AlNi phase grains. Some grains of AlNi phase are enriched in nickel, confirming previous suspicion of existence of AlNi rich phase. The most interesting observation is that the AlNi Ni-rich phase does not create new grains but rather changes the composition of the grains of already existing phases. In contrary to this, AlNi deficient in nickel possess own grains, being separated from the AlNi stoichiometric phase ones.

**Figure 8.** (**a**) EBSD map with (**b**) EDS maps of Al and Ni elements distribution for the sample annealed for 20 h at 720 ◦C. (**c**) The imposition of the EBSD indexing confidence map and EDS map analysis for Al and Ni elements.

**Figure 9.** SEM micrograph in BSE mode of the intermetallic phases formed in the solid state after 20 h at 720 ◦C.

The TEM investigation for sample annealed for 3 h at 720 ◦C indicates different results in comparison to the EBSD and EDS overlapping for sample after 20 h of annealing. The diffraction pattern from the area taken by the AlNiNi-rich phase (Figure 10), determined based on the chemical composition, did not unambigously confirmed its presence. Two possible phases were taken into consideration, namely AlNi and Al3Ni5. Important is that AlNi is always identified based on the same crystallographic data [31], however, the content of elements is different, so this phase is considered as AlNi without division into rich and deficient in nickel types. The degree of mismatch is very high for AlNi phase reaching 38% (Figure 10c). For the second considered phase Al3Ni5, the degree of mismatch is of only 9% (Figure 10d). This orthorombic phase is metastable below 700 ◦C, as the samples were cooled with furnace after the annealing process, it could be formed. However, due to the fact that growth of this phase takes place at 720 ◦C, in the manuscript it is noted as Ni-rich AlNi phase. Further TEM investigations would be of grea<sup>t</sup> benefit for the description and understanding of the phase evolution, especially for the early stage of the Ni-rich AlNi phase growth. As the phase thickness was below the analytical resolution in SEM (see Figure 11) such examination in TEM is of essential need. The results of TEM-EDS and SEM-EDS measurements are similar and were collected in Table 4, however, in case of the doubtful phase substantial difference is visible. From SEM-EDS it follows that the phase present in the joint is the Al3Ni5 but the TEM-EDS results point at the AlNi rich in nickel one. As mentioned above, this phase was extremely narrow, the thickness is on the border of the resolving power of the method. Examined areas for both methods are shown in Figure 11.

**Figure 10.** (**a**) TEM bright field image showing the microstructure of the AlNi intermetallic phase for sample annealled for 3 h at 720 ◦C together with the corresponding (**b**) selected area diffraction pattern taken from the grain marked with circle. Simulation of the solve for (**c**) AlNi and (**d**) Al3Ni5.

**Figure 11.** SEM (**a**) and TEM (**b**) microstructures of a Ni/Al/Ni interconnections (Ni substrate of B-type) after 3 h of reaction time at 720 ◦C with indicated EDS point analysis presented in Table 4. Numbers 1–5 denote particular intermetallic phases: 1-Ni solid solution, 2-AlNi3, 3-AlNiNi-rich, 4-AlNi, 5-AlNiNi-deficient.


**Table 4.** Comparison of the chemical compositions obtained by SEM and TEM.

#### *3.2. Growth Rate of the Intermetallic Phases*

Determination of the thickness of the intermetallics (Table 5) allowed revealing their growth kinetics. For the short time of the reaction, the interconnection zones were broadening as it is assumed in the diffusion soldering process. Between 1 and 3 h of annealing only the subtle difference of the thickness of the whole joint was observed, nevertheless, different intermetallic phases in reaction zones appeared. Changes of phases composition points that the reaction after 1 h probably takes place in the solid state. After 5 h of annealing, due to isothermal solidification stage of DS, for both types of substrates, the interconnection zones shrink. As it was mentioned earlier, phases composition between 3 and 5 h stays the same, while the main difference is associated with the thickness of the interconnection zones for both types of substrates. For samples of A-type, the interconnection zones after 5 h of annealing is four times narrower and for B-type two times thinner than after 3 h of annealing. Comparing the samples after the same annealing conditions, for A-type to B-type Ni substrates, the following results are observed: after 1 h of annealing the thickness of the entire joint is comparable and widths of particular phases are similar. When time of annealing is extended to 3 h, the differences in diffusion process and in the overall appearance of the interconnection zones are not observed. After 5 h of annealing more visible differences appeared. First of all, the thickness of the joint, where substrate B-type was used, is twice broader than in Ni/Al/Ni reaction zone with A-type substrates. Additionally, in case of NiA/Al/NiA, the phase AlNi deficient in nickel is not observed. Thicknesses of the individual phases are similar, beside of the total width of AlNi phases (of every type), which in case of B-type nickel is broader, however, stoichiometric type of AlNi phases are comparable. This difference between thickness of whole joint after 5 h could be caused by the leakage of liquid solder during the experiment due to too high pressure applied. The comparison of the thickness of the particular layers of the intermetallic phases formed in Ni/Al/Ni interconnection in different time of reaction is collected in Table 5. Authors conducted experiment which allowed to eliminate the necessity of application of two separated systems: NiA/Al/NiA and NiB/Al/NiB. Numerous attempts prove that the localization of the Ni substrates (above or below the Al solder) does not affect the width and sequence of created phases. This approach resulted in simplifying the experimental procedure and allowed for producing of NiA/Al/NiB system (and conversely), which shortened experiment time. This procedure was used for shorter and longer annealing times. In early stages of diffusion soldering processes phase Al3Ni disappears fast—only after 30 min it is completely consumed and replaced by Al3Ni2 phase, which, in turn, after 3 h of annealing no longer exists. The longest times of annealing cause further phases broadening. Finally, after 72 h of annealing the interconnection zone contains only high nickel phases (50 at.% of Ni and higher). Three phases: AlNi3, AlNi, AlNiNi-rich for which the growth kinetic is calculated expand gradually with the annealing time.


**Table 5.** The thickness of the particular layers of the intermetallic phases formed in Ni/Al/Ni interconnection in different time of reaction at 720 ◦C.

Determination of the main mechanisms, which control the growth of the intermetallic phase is based on simple and useful formula (Equation 1), in details discussed in [32], giving the relation between the thickness of the intermetallics with the time of annealing.

$$
\Delta d = k t^n \tag{1}
$$

where: Δ*d* is a thickness of the intermetallic phase layer, *k*—the growth rate constant and *t*—time of annealing. To define the mechanism of the intermetallic phase growth, it is necessary to determine the value of *n* exponent. Depending on this value, the growth can be controlled either by the volume diffusion (*n* = 0.5), or by the chemical reaction at the interfaces (*n* = 1), grain boundary diffusion (*n* < 0.5) and finally by mixed mechanism of growth (0.5 < *n* < 1). The growth rate constant *k* can be determined from two types of plot, namely Δ*d* vs. *t* ( 1 2 ) or Δ*d*<sup>2</sup> vs. *t*. First type of plot is better for the growth of the intermetallic phases, where the layer is formed at the initial period of growth and does not influence the course of further stabilized growth. Second type should be used, when the initial growth of the obtained layer affects the period of the parabolic growth [32]. Diffusion process may be controlled by the dislocation mechanism but only at lower temperature, while at the temperature close to the melting point of metals the amplitude of thermal vibrations of atoms is too high and dislocations as a structural effects disappear. Therefore, in such a case, the dislocations are not taken into account and the mechanisms of diffusion are either volume diffusion or diffusion by grain boundaries or reactive diffusion. The growth kinetics results are collected in Table 6 and shown in Figure 12.


**Table 6.** The growth kinetics of AlNi, AlNiNi-rich and AlNi3.

**Figure 12.** Time dependence of the layer width for stoichiometric AlNi (**a**), AlNiNi-rich (**b**) and AlNi3 (**c**) in Ni/Al/Ni joints.

The growth kinetics data for three phases formed in the solid state in Ni/Al/Ni interconnections was determined. Plots (Figure 12) Δ*d* vs. *t* for AlNi stoichiometric (Figure 12a), AlNiNi-rich (Figure 12b) and AlNi3 phases (Figure 12c), showed that the growth of AlNi phases was controlled by different mechanism in comparison to AlNiNi-rich and AlNi3. The growth of AlNi phase involved two mechanisms: at first, the reaction at the interface took place and then it was replaced by the volume diffusion. The time exponent *n* for AlNi phase equals 0.67 for substrates of A-type and 0.65 for the B-type ones. The calculation showed that growth mechanism for AlNi rich in nickel and AlNi3 are similar and governed by the volume diffusion (*n* is 0.5 ± 0.1). In the case of AlNi rich in nickel

phase growth with the short incubation time occurred only for substrate of A-type. The AlNi3 phase grew due to the volume diffusion mechanism and no incubation time was observed in its growth. Finally, it was verified, that neither of these phases grew due to the grain boundary diffusion, therefore, the differences between the samples of A and B Ni substrates were not observed. In the study made by Lopez et al. [13] phases grew only by volume diffusion, while in present work the growth mechanism for the AlNi stoichiometric is found to be mixed (chemical reaction and volume diffusion). Lopez et al. [13] determined that the fastest growth occurred for stoichiometric AlNi and the slowest growth for Ni-rich AlNi one. They also calculated that the values of n factor equaled 0.5 ± 0.1 for all three phases: AlNi, AlNiNi-rich and AlNi3. Therefore, the authors assumed that the growth of all these layers obeys a parabolic law—it is governed by the volume diffusion. What is interesting, they also noticed a transition period of AlNi3 growth for short time of annealing. In this study presented graphs in Figure 12 and Table 6 revealed new insight for the Al/Ni interaction. At present work the differences in behavior of phase-growth at short time of annealing and after longer time was noticed. Growth of AlNi rich in Ni phase is governed by volume diffusion but only after longer time of annealing. Focusing on the shorter time of process (2–5 h), the calculated *n* exponent shows that for substrates of A-type, the grain boundary mechanism dominates (*n* = 0.37). On the other hand, considering of B-type substrate, data obtained at the beginning of DS suggesting the significant contribution of the reaction at interface (*n* = 0.91). Similar behavior for AlNi3 phase for both types of substrates was observed, however in the range of 0.5–5 h time of annealing, only reaction at interface was evidenced as the governing mechanism of growth (for A-type *n* = 0.73; for B-type *n* = 0.76). In the case of AlNi stoichiometric, independently on time of annealing, two mechanisms of growth, as it was mentioned above - reaction at interface and volume diffusion occurred.
