*3.1. Densification Analysis*

Being porosity-free is of importance to the mechanical properties of the final products. Thus, it is necessary to investigate the e ffects of the process parameters on the formation of defects in order to obtain high relative density parts. In this section, the relationships between the relative density and process parameters are analyzed. Figure 6 shows the variation of the relative density with di fferent volumetric energy densities. With an increase of VED, the relative density first became higher and tended to be stable. The relative density of the final samples fabricated in this work can were nearly 100% free from cracks. Figure 7 depicts the actual transverse morphology of the samples obtained under di fferent VEDs. With an increase in the value of the VED, the number of defects in the samples gradually decreased. When the input of the VED was insu fficient, the low-melting-point metals (Ni/Fe) were melted, and the liquid phase was formed. The un-melted tungsten particle gaps were filled with the liquid phase. However, a low VED indicated a combination of a low laser power (200 W) and a high scan speed (300 mm/s). This result implies that the residence time of the formed liquid phase was too short to fully fill the gaps, so irregular defects were formed, as shown in Figure 8a. Poor densification was caused by inadequate liquid phase content or a short residence time under the processing conditions. Figure 8b presents the high-magnification transverse morphology of the samples fabricated with a high VED. Nearly full densification could be obtained under the process parameters.

**Figure 6.** Variation of the relative density with the volumetric energy density. The inserted figure shows the different transverse morphologies corresponding to the VEDs.

**Figure 7.** The transverse morphologies of different VEDs. (**a**) 185.19 J/mm3; (**b**) 277.78 J/mm3; (**c**) 396.83 J/mm3; (**d**) 518.52 J/mm3; (**e**) 592.60 J/mm3; (**f**) 617.28 J/mm3; (**g**) 634.92 J/mm3; (**h**) 648.15 J/mm3; (**i**) 777.78 J/mm3.

**Figure 8.** Transverse morphology with a high magnification. (**a**) 253.97 J/mm3; (**b**) 648.15 J/mm3.

The pores were nearly fully filled with the formed liquid phase. A combination of high laser power (350 W) and low scan speed (150 mm/s) produced a longer residence time for the liquid phase. Therefore, the best rearrangemen<sup>t</sup> characteristics for the tungsten powder particles were observed under the capillary force of liquid, and the densification of samples was improved. When the laser power was 350 W, the viscosity of the formed molten pool decreased, and partial tungsten powder particles could be melted. In this way, an adequate liquid phase and better fluidity for the molten pool were obtained, and the densification process of the samples could be completed.

It can also be seen that under the same VED, there are di fferent relative densities corresponding to distinct transverse morphologies, as shown in Figure 9. This phenomenon could be ascribed to the varying process parameter combinations and similar results of other materials that have been reported [28,29]. Here, the signal-to-noise (S/N) ratio of Taguchi's method was used to compare the effect of every process parameter on the relative density, which means the sensitivity of the relative density to the selected process parameters. In general, the signal-to-noise ratio in Taguchi's method can be classified into three categories: "lower is better", "nominal is best", and "higher is better". In this study, the highest relative density for the sample was expected. Thus, the "higher is better" category was adopted, which can be calculated as follows [30]:

$$\frac{S}{N} = -10 \log \left( \frac{1}{n} \sum\_{1}^{n} \frac{1}{y\_i^2} \right) \tag{5}$$

where *n* is the number of experiments and *yi* is the tested data of relative density for the *i*th sample.

**Figure 9.** Transverse morphology of the same VED (370.37 J/mm3).

With the tested data for the relative density under different combinations of process parameters, the calculated results for the S/N ratios are presented in Figure 10 and Table 3. It can be seen that the laser power was identified as the most important factor influencing the final relative density of the samples. The scan speed had a relatively insignificant effect on relative density, and the hatching distance showed the least significant effects among all the selected process parameters. Therefore, the sensitivity of the relative density to the three process parameters decreased according to the following order: laser power > scan speed > hatching distance. This result is consistent with previous results and could be ascribed to the special thermophysical properties of tungsten and tungsten heavy alloys [31].

**Figure 10.** Main effect plots of signal-to-noise (S/N).



In this section, the nearly full densification of the 90W-7Ni-3Fe sample fabricated by SLM was realized. The densification behaviors under different process parameters were analyzed and discussed. Laser power is a dominant factor in the SLM process of the 90W-7Ni-3Fe samples. For the 90W-7Ni-3Fe sample with a high relative density in SLM, a higher laser power (≥250 W) was preferred, and the value of volumetric energy density had to be no less than 300 J/mm3. The morphology of the 90W-7Ni-3Fe sample presented similar sintering, characteristic in traditional powder metallurgy [32]. The microstructure is mainly composed of a refractory tungsten particle skeleton and a liquid phase formed by low-melting-point metals (Ni/Fe).

#### *3.2. Phase Identification and Microstructure*

The XRD patterns of the 90W-7Ni-3Fe samples processed using different volumetric energy densities are shown in Figure 11. The main phases of the 90W-7Ni-3Fe samples fabricated by SLM were W and Ni-Fe solid solution phases, although the volumetric energy density varied from 185.19 J/mm<sup>3</sup> to 777.78 J/mm3. There seems to have been no significant difference in the phase composition among the samples obtained under different volumetric energy densities. This means that low-melting-point metals (Ni/Fe) still melted, even under low volumetric energy densities.

**Figure 11.** The XRD pattern of different VED.

Figure 12a shows the typical SEM morphology of the 90W-7Ni-3Fe sample fabricated by SLM. Many tungsten particles were distributed in the matrix. The gaps between the tungsten particles were filled with liquid phases formed by low-melting-point metals (Ni/Fe). In addition, partial tungsten particles were contacted under the driving force of the liquid phase during the processes of melting and solidification. Fine tungsten grains can also be found in Figure 12a. The microstructural region can be divided into three main regions: the W particle phase, the fine W dendrite region, and the Ni–Fe matrix region with dissolved W. Scanning electron microscope (SEM) and energy-dispersive spectrometer (EDS) analyses were performed in order to further observe the microstructural constitution and confirm whether the tungsten element had been dissolved into the matrix. EDS analysis results (Figures 12b and 13a) confirmed that the particles were nearly composed of 100% W with little Ni and Fe. EDS analysis results of the Ni–Fe matrix indicated that partial tungsten had been dissolved into the matrix (Figure 13b). This phenomenon could be ascribed to the different solubilities of W, Ni, and Fe in W-Ni-Fe systems. Tungsten has a high solubility in the Ni/Fe matrix, but the solubility of Ni and Fe in tungsten is practically negligible. Moreover, fine W dendrites formed in the matrix when W particles were partially melted and W particles acted as heterogenous nucleation sites. Similar tungsten dendrites were also observed in the WHA samples fabricated by SLM [33,34].

**Figure 12.** (**a**) The microstructure of 90W-7Ni-3Fe under VED = 518.52 J/mm3; (**b**) element distribution.

**Figure 13.** Energy-dispersive spectrometer (EDS) element analysis. (**a**) Point A; (**b**) point B.
