Relative density and chemical composition

Table 4.3 shows the ANOVA obtained in publication IV for relative density and Mn content, including homogenization temperature (Th) and time (th) as continuous predictors. Overall, the results demonstrate the high repeatability of L-PBF for manufacturing high-density polycrystalline Ni-Mn-Ga with a consistent and predictable composition. All samples in publication IV were highly dense (average ~98.4%) with only minor between-sample variation and standard deviations in the range of ±0.15–

0.41% for all samples. The ANOVA does not show any dependency between applied homogenization treatment and observed density. Considerable density changes were not expected during the heat treatment because the density of the as-built samples was relatively high.

The optimized sample in publication III exhibited a ~1.1 at.% decrease in the fraction of Mn and a corresponding increase in the fraction of Ni compared to the initial powder, with no clear change in the fraction of Ga. This was expected based on the parameter optimization conducted earlier in the same publication. The as-built samples in publication IV exhibited a ~1.6 at.% decrease on average in the fraction of Mn compared to the initial powder. Consequently, the fractions of Ni and Ga each increased by ~0.8 at.%. Using a lower VED value (44.4 J/mm³) in publication IV than in publication III likely led to the reduced evaporation of Ga during L-PBF, which explains this observation. Additionally, the studies were conducted using different Ni-Mn-Ga powders and substrates, further contributing to the observed differences. The ANOVA in Table 4.3 shows that the heat treatment did not influence the samples’ chemical composition. The

4.2 Characterization of as-built and heat-treated samples 49

observed compositional differences between samples (see Table 4.1) were not large enough to impact their magneto-structural properties.

The XRF measurements did not reveal any distinct change of composition within each sample in each publication. The EDS analysis conducted on a section along the build direction of a sample (III-OPT) in publication III did not show any segregation or scattering of the chemical elements. The composition of the sample was moderately homogenous, with only a minor variation that could not be reliably distinguished from the noise and uncertainty of the measurement itself. Importantly, the samples in both publications exhibited standard deviations of the measured compositions approximately within the measurement accuracy. In publication IV, the standard deviations of the measured compositions were smaller in samples with longer homogenization times or higher homogenization temperatures, implying an increase in chemical homogeneity.

Figure 4.5: SEM images of the microstructure obtained from: (a) a section of an as-built sample (III-OPT) along the build direction; sections perpendicular to the build direction: (b) as-built sample (IV-1), (c) a sample ordered at 800 ℃ for 4 hours (IV-2), (d) a sample homogenized at 1040 ℃ for 6 h (IV-6), and (e) a sample homogenized at 1080 ℃ for 24 h (IV-11). The dotted red lines indicate observed boundaries between adjacent laser-scanning tracks. (Modified from publications III and IV.)

4.2.2 Microstructure

Figure 4.5 shows the SEM images obtained from a section of the as-built sample III-OPT along the build direction and sections of the as-built sample IV-1 and selected heat-treated samples (IV-2, IV-6, and IV-11) perpendicular to the build direction. Some of the observed boundaries between adjacent laser scanning tracks are marked with red dotted lines. The reader is advised to note the composition difference between the sample produced in publication III and the samples produced in publication IV.

The microstructures of the as-built samples, shown in Figure 4.5a-b, are highly anisotropic, with the boundaries and profiles of each deposited track clearly distinguishable in the SEM images. Both samples exhibit a columnar grain structure without the presence of dendritic cooling structures that are sometimes observed with L-PBF-built materials. Some finer grains are located between the relatively larger columnar grains at the boundaries between adjacent laser-scanning tracks. This grain texture develops when columnar grains grow along the normal direction to the edges of melt pools induced by the large directional thermal gradients during L-PBF manufacture.

Additionally, martensitic twin variants, observable as a parallel stripe-like surface relief with contrasting areas, can be observed throughout the SEM image of sample III-OPT in Figure 4.5a. Some of the observed twins cross the boundaries between adjacent tracks and consecutive deposition layers. The orientation and width of the twins vary from grain to grain. The twins are not as visible in the SEM images of the samples produced in publication IV, mostly because the used SEM parameters were selected to enable high grain contrast for grain size measurements; see the section about SEM in Chapter 3.

Figure 4.5e shows that the sample homogenized at 1080 ℃ for 24 h displayed considerable recrystallization, exhibiting a microstructure that predominantly consists of large equiaxed grains. The boundaries between adjacent laser-scanning tracks are no longer observable. The SEM images of intermediate samples, shown in Figure 4.5c-d, demonstrate the gradual change from the columnar grain structure of the as-built samples to the equiaxed grain structure of the sample homogenized at 1080 ℃ for 24 h.

Some intergranular cracking was observed within the built samples in publication IV, which was expected based on the results obtained earlier in publication II. The location, size, number and distribution of the observed cracks appeared to be random. Overall, Ni-Mn-Ga based MSM alloys possess low deformability (Wei et al., 2018), which makes them brittle and highly susceptible to cracking in response to internal stresses accumulated in L-PBF or during cutting and grinding in sample preparation. Cracking was not observed with the as-built sample (III-OPT) in publication III, possibly due to its different chemical composition and the different L-PBF processing conditions in comparison to the samples in publication IV.

In publication III, the volume-weighted average grain size of the as-built sample was determined as 16.6 µm. A similar value, 13.5 µm, was obtained for the as-built sample in publication IV despite the obvious compositional difference between the two samples.

Overall, the average grain sizes of the as-built samples can be considered small but are

4.2 Characterization of as-built and heat-treated samples 51

within the typical range for L-PFF-manufactured materials (DebRoy et al., 2018). Figure 4.6 shows the effect of the applied heat-treatment parameters on the volume-weighted average grain sizes of the samples built in publication IV. All heat-treated samples exhibited grain growth in comparison to the as-built material. The ANOVA in Table 4.4 shows a statistically significant dependency between the observed grain size and the applied homogenization temperature and time. The most significant average grain size increase was observed for samples homogenized at 1080 ℃ for 12 and 24 h. The observed grain growth was overall equal throughout the volume of each sample, although some singular larger grains formed near the edges of these samples. For example, the sample homogenized at 1080 ℃ for 24 h exhibited a large quantity of grains exceeding 300 µm in diameter; see the SEM image in Figure 4.5e and the polarized light optical image in Figure 4.7e.

Figure 4.6: Effect of the applied heat-treatment on the average grain size of the L-PBF-built samples, as measured and averaged from multiple sections of each sample perpendicular to the build direction. (Modified from publication IV.)

Table 4.4: ANOVA table for the volume-weighted average grain size, including homogenization temperature (Th) and time (th) as continuous predictors.

(Modified from publication IV.) Average grain size (µm)

Source DF Adj SS Adj MS F-value P-value Th 1 803.610 803.610 12.38 0.039 th 1 776.410 776.410 11.96 0.041 Th2 1 180.620 180.620 2.78 0.194 th2 1 24.510 24.510 0.38 0.582 Th*th 1 150.690 150.690 2.32 0.225 Error 3 194.770 64.920 - -

Total 8 2029.410 - - -