4.2.1. Formation and Effects of Ni4Ti3 Precipitates
Based on the XRD patterns, no traces of Ni4Ti3 precipitate can be found on the NiTi powder. However, the presence of this intermediate phase has been identified on the SLM NiTi samples. One possible explanation would be that during the fabrication procedure, the melting stage of the SLM process provided an additional opportunity for the atoms to complete diffusion. Thus, it resulted in the formation of Ni4Ti3 precipitates. No precipitation has been detected in the NiTi powder as the rapid solidification rate of the gas atomization process did not provide sufficient time for complete diffusion. Hence, only the transition region was observable in the DSC curve of the NiTi powder. Both the transition region and Ni4Ti3 precipitates found in the DSC curve and XRD pattern of the NT samples were due to the partial completion of the diffusion process.
Nevertheless, during the process of heat treatment, the occurrence of oxidation was inevitable. The manifestation of oxidation would lead to the depletion of Ti as it was more reactive to oxygen than Ni [
59]. Thus, the content of Ni in the NiTi matrix would increase, resulting in two phenomena; (1) the decrease of the transformation temperatures and the transformation of martensitic phase into austenitic phase and (2) the promotion of precipitate formation.
Precipitation will occur when the condition of supersaturation is met. The NiTi material would be in a supersaturated state when the content of Ni has exceeded its solubility limit but yet to form precipitation due to the quenching process. Hence, the supersaturation state of NiTi would provide a driving force for the initiation of precipitation during the heat treatment process [
35].
When the temperature of the heat treatment conducted was at 400 °C, the depletion of Ti would not be as much. This was because the main oxidation process did not start immediately at a lower temperature [
59]. Moreover, the oxidation rate tended to slow down and saturate after some time. Thus, the slight increase in the Ni content could be within the overlapped region between the metastable equilibrium phase (NiTi and Ni
4Ti
3) and the solubility limit of NiTi [
35,
41]. In this region, the increase in the Ni content would lead to the decrease in the transformation temperatures. This decrease was reflected in the M
f temperature of the H400 samples while the identification of the austenitic phase was shown in their XRD pattern. Likewise, Firstov et al. have also reported the detection of austenitic phase when their NiTi material (atomic percentage of 50% Ni and 50% Ti) was processed between a heat treatment temperature of 300 to 500 °C [
59]. Additionally, the small increase in the content of Ni would also result in a slight increase in the nucleation of the Ni
4Ti
3 precipitates.
However, when the heat treatment temperature increased to 500 and 600 °C, more Ti would be depleted. This is the result of oxidation proceeding earlier when at a higher temperature [
59]. Thus, the increase in the content of Ni could exceed the solubility limit of NiTi, extending into the metastable equilibrium state. Subsequently, the austenitic phase disappeared as presented in the XRD patterns of the H500 and H600 samples. Instead, the rate of precipitation increased as shown in
Table 4 and the existing Ni
4Ti
3 precipitates started to grow larger at the expense of the smaller particles [
60]. The results obtained were also supported by the similar observation of Ni
4Ti
3 precipitates found in samples that were heat-treated at 500 to 600 °C for 6 min [
35,
61].
Nonetheless, when the heat treatment temperature increased to 700 °C, the depletion in Ti was so strong that the Ni
4Ti
3 precipitates began to decompose into the equilibrium Ni
3Ti phase [
35,
56,
59,
62,
63]. Hence, both Ni
4Ti
3 and Ni
3Ti precipitates could be found in the XRD pattern of the H700 sample. The results obtained were confirmed by the observation of Ni
3Ti precipitates at a higher heat treatment temperature of above 600 °C [
59].
Conversely, a higher nucleation rate and the decomposition of Ni
4Ti
3 precipitates into Ni
3Ti phase may not be beneficial to the shape memory properties of repetitively scanned NiTi samples. Based on
Table 2, only H400 samples have demonstrated improvements in the transformation strain and shape recovery percentage as compared to the NT samples. The shape memory properties were found to deteriorate with increasing heat treatment temperature. The occurrence of these phenomena could be understood from the physical metallurgy principle.
The generation of a high density of fine precipitates is widely known to be the most effective method in preventing the movement of dislocations [
35,
56,
64]. Concurrently, it increases the critical stress magnitude for the slip to occur as well [
37]. The formation of dislocations is plastic deformation and an irreversible process. It is not possible to restore the strains produced by this defect via the mechanism of the reversible martensitic phase transformation. Therefore, heat treating the repetitively scanned NiTi samples at a temperature of 400 °C for a duration of 5 min would initiate the precipitations of fine Ni
4Ti
3 particles to deter the generation of plastic deformation. This result has also coincided with several past research where the conventionally-produced NiTi parts demonstrated the best SME and superelastic properties after heat treating at 400 °C [
37,
49,
63]. Other studies have attributed the improvement in the shape memory properties to the presence of fine Ni
4Ti
3 precipitates as well [
30,
63,
65,
66].
Nevertheless, as the temperature of the heat treatment increases, particle agglomeration of the Ni
4Ti
3 precipitates occurs [
30,
48,
62,
65,
66]. Experimental observation of Ni
4Ti
3 enlargement by 30 times as the heat treatment temperature raised from 400 to 500 °C has been reported by Yan et al. [
63]. Thus, the increment of the heat treatment temperature would decrease the density distribution of the precipitates. It would then lead to a reduction of their effects on the shape memory properties of the heat treated NiTi samples. In addition, it has been reported that coarse Ni
4Ti
3 would lose its coherency with the NiTi matrix [
48,
62,
63]. Hence, dislocations would be introduced to relieve the stress fields generated around the precipitates [
58,
62,
63]. Consequently, it would lead to an overall decrease in the fatigue strength of the NiTi samples.
Meanwhile,
Table 4 provides another explanation for the improvement in the shape memory properties exhibited by the H400 samples. It has identified that the highest volume fraction of martensitic phase could be found in the H400 samples. Thus, a larger transformation volume is available for the reversible martensitic phase transformation during the process of heating [
40]. Correspondingly, the H400 samples would demonstrate a higher transformation strain than the other samples. In summary, the heat treatments of repetitively scanned NiTi samples above 400 °C have resulted in overaging of the material [
64]. Eventually, the samples demonstrated poorer shape memory responses than the NT samples.
4.2.2. Formation and Effects of Grain Boundary Migration
Other than introducing the formation of Ni
4Ti
3 metastable precipitates, the implementation of heat treatment was also found to alter the microstructures of the repetitively scanned NiTi samples. According to
Figure 4 and
Table 5, the grain size of the NT and heat treated samples increases with the application of heat treatment and with rising heat treatment temperature. Specifically, the grains were observed to elongate lengthwise.
At first glance, it might appear that the heat treated samples have undergone through the recovery and recrystallisation processes. The migration of the grain boundary was the result of ordinary grain growth after recrystallisation as the grains grew larger at the expense of the other grains. However, on closer inspection, it is inferred that the heat treated samples did not experience recrystallisation. This deduction came about based on the following factors.
For the recrystallisation process to occur, there are a few requirements that need to be fulfilled. Firstly, the samples have to be subjected to a certain magnitude of deformation [
67]. It has been reported that the recrystallisation phenomenon did not happen for samples deformed below the strain of 20% [
68]. Moreover, the recrystallisation temperature is a function of the degree of deformation [
67]. A less deformed sample would have a higher recrystallisation temperature than a severely deformed sample. As reported in the past research, NiTi has a recrystallisation temperature of between 550 to 600 °C when they experienced cold-working of about 30% [
48,
49]. During the sample preparation, precautions were taken to ensure that the samples did not suffer any unnecessary deformation prior to the heat treatment and the determination of grain size. Thus, it is expected that the heat treated samples would not experience recrystallization within the implemented temperature range.
The second factor is the lack of time for the formation and growth of the recrystallised nuclei to microscopic size [
69,
70]. During a typical recrystallisation process, an initial incubation period is required for sufficient energy to develop such that the first strain-free nucleus could grow to a visible size [
67]. In addition, the recrystallisation temperature for a particular material denotes the approximated temperature at which its highly cold-worked form would completely recrystallise in 1 h [
67,
70]. A heat treatment of 5 min is unlikely able to accumulate an adequate amount of energy to form defect-free nuclei and grow to appreciable size.
The third condition is the additional energy required to initiate the formation of the strain-free lattice [
67]. During the process of cold-working, plastic deformation was introduced into the samples. The internal energy evolved from the deformation would increase the energy state of the atoms. However, it is not possible for the atoms or dislocations to revert back to a defect-free lattice from the distorted lattice at room temperature due to the nature of strain hardening. Hence, additional energy is required to bring the atoms to the next energy state level to overcome the rigidity of the distorted lattice. The additional amount of energy would be supplied in terms of heat energy. Nonetheless, as the repetitively scanned samples were not severely deformed prior to the heat treatment, the internal energy state of the samples would be much lower. Thus, a larger amount of additional energy needs to be provided to the samples to initiate recrystallisation. Since the highest heat treatment temperature tested is only about 100 °C higher than the recrystallisation temperature of a heavily deformed NiTi material, the energy supplied may not be sufficient to start up the recrystallisation process.
Besides needing to fulfil the three requirements, the observed grain shape also provided evidence of the absence of recrystallisation. In
Figure 4, the grains were noticed to elongate upon heat treatment and with increasing heat treatment temperature. Moreover, the elongation seems to start promptly at the start of the heat treatment despite the short duration of 5 min. However, in a typical recrystallisation process, no observable differences can be seen in the microstructures of the material during the recovery phase [
67]. Even when the recrystallisation temperature has been reached, it requires a certain period of incubation for the formation and growth of new nuclei. Furthermore, the recrystallised grains were observed to be equiaxed [
67], which contradicted what was captured on the microscopic images. Thus, it is concluded that the heat treated samples did not experience recrystallisation. The elongation of the grains is postulated to be a result of strain-induced boundary migration [
69].
The difference between strain-induced boundary migration and the recrystallisation process is that in the former, the annealed material left behind due to the movement of the grain boundary has the same orientation as the strain hardened parent grain [
69]. However, in the recrystallisation process, nuclei with different orientations are produced. The movement of the grain boundaries in the strain-induced boundary migration is towards the distorted regions due to the strain gradients generated [
71]. Correspondingly, this would result in the increase of the size of one grain and the disappearance of the other grain. One characteristic of strain-induced boundary migration is that the produced annealed material would be constricted by the ordinary grain boundary on one side, while bounded by the parent grain on the other side without the intervention of grain boundary [
69]. Nevertheless, in a typical recrystallisation process, the produced recrystallised grains would be located in between the strain hardened grains. These grains were separated by grain boundaries.
However, the reason for the occurrence of the strain-induced boundary migration on repetitively scanned samples is not known yet at this moment. Nonetheless, the grains were observed to elongate together in one direction. The direction of the grain elongation was generally aligned perpendicularly to the direction of the laser scanning path as illustrated in
Figure 5. One possible explanation for both phenomena would be due to the directional solidification nature of the SLM process. During the fabrication of a single layer, the laser would begin scanning in a line-by-line manner, starting from the top region. As the laser proceeds down to the adjacent line, the previous scanned section could have solidified and cooled down. Hence, a temperature gradient is generated between the molten and solidified regions, resulting in the grains to grow towards the direction of the temperature gradient. In addition, the non-uniform thermal conductivity and the thermal expansion coefficient of the liquid and solid phase of NiTi could also lead to the development of residual stress and/or strain. Thus, during the heat treatment process, the grains further elongate lengthwise according to the produced strain gradients. Nevertheless, more studies are needed to confirm this hypothesis.
In general, the occurrence of the strain-induced boundary migration did not contribute to the improvement in the shape memory properties of repetitively scanned NiTi samples. As presented in
Figure 4,
Table 2 and
Table 5 the transformation strain and shape recovery percentage of the heat treated samples decreased with increasing grain size. The only exception is the H400 samples, where they demonstrated an improvement in their shape memory responses as compared to the NT samples. However, this enhancement could be attributed to the formation of fine Ni
4Ti
3 precipitates and high content of martensitic phase, where they outweigh the negative impact of the strain-induced boundary migration.
As reported by various authors, the increase in grain size could be detrimental to the mechanical properties of repetitively scanned NiTi samples [
36,
48,
72]. Based on the results obtained by Delville et al., they observed that samples with a larger grain size tend to be significantly prone to the formation and build-up of dislocations [
72]. The increase in the dislocation density has led to the rapid accumulation of permanent strain build-up during the cyclic test performed. Moreover, they determined that their NiTi samples had the highest resistance to slip deformation when the samples just entered into the recrystallisation phase. In this phase, the new defect-free recrystallised grains would be coexisting with the polygonised microstructure. Furthermore, the exhibition of small grain size contributed to its high strength and stability for the cyclic test as well. However, when the heat treatment process proceeded into the phase of grain growth, the yield stress for plasticity and strength dropped significantly.
The improvement in the mechanical properties of NiTi SMA after a decrease in the grain size has also been presented by the other researchers [
36,
48]. It was found that a reduction in grain size could slow down the propagation of cracks formed during the cyclic test of NiTi material. Hence, samples with small grains would have a higher fatigue life than samples with coarse grains. The reason for the improvement in the NiTi shape memory properties is due to the increase in the volume fraction of the grain boundary [
36]. The grain boundary acts as a non-transformable barrier that separates the crystallites with different orientations. When the grain size decreases, the mechanical constraint of the grain boundary on the deformation of the crystallites becomes more significant as compared to samples with coarse grains. Thus, the movement of the dislocation was impeded, resulting in a decrease in the accumulation of permanent strain. Sequentially, samples with finer grain size would have a higher transformation strain and shape recovery percentage. Therefore, combining the effects of grain size and Ni
4Ti
3 precipitates, H400 samples have the best shape memory properties among the other heat treated samples.