The Inﬂuence of the Process Parameters on the Densiﬁcation and Microstructure Development of Laser Powder Bed Fused Inconel 939

: This work aims to investigate the e ﬀ ect of the process parameters on the densiﬁcation and microstructure of Inconel 939 (IN939) alloy processed by laser powder bed fusion (LPBF). IN939 is a Ni-based superalloy with high creep and corrosion resistance that can be used up to around 850 ◦ C under load, resulting in higher operative temperatures than the ones commonly allowed for Inconel 718 and Inconel 625 alloys (around 650 ◦ C). However, this alloy can su ﬀ er from poor weldability involving possible crack formation. In order to minimize the residual porosity and the cracking density, speciﬁc process parameters were investigated. The parameters to generate IN939 samples almost pores-free (porosity ≤ 0.22%) with a cracking density ≤ 1.36 mm / mm 2 as well as samples almost crack-free ( ≤ 0.10 mm / mm 2 ) with limited residual porosity ( ≤ 0.89%) were determined. The microstructure revealed ﬁne dendritic / cellular structures with the formation of sub-micrometric phases. A high concentration of these phases was also found along the intergranular cracks, suggesting that their presence, coupled to the high thermal stresses, can be the primary reason for crack formation during the LPBF process. elongated grains along the building direction ( z -axis) by a strong along the 001 > Inside the grains, a network of dendritic / cellular architectures with sub-micrometric phases mainly along the grain boundaries and interdendritic areas were observed. This indicates that the high cooling rates of the did not inhibit the formation of phases during the solidiﬁcation.


Introduction
Laser powder bed fusion (LPBF) is an additive manufacturing process attracting considerable attention from both the research and industrial sectors. This technology provides an excellent way to fabricate complex geometries for materials difficult to machine by conventional manufacturing processes (i.e., casting and wrought) [1][2][3][4]. Among these materials, Ni-based superalloys are typically considered difficult to machine due to their high mechanical properties, which are also maintained at high temperatures [5].
The most commonly processed LPBF Ni-based superalloys are Inconel 718 and Inconel 625 fabricated with a relative density close to 100% due to their good weldability [6][7][8][9][10]. Nevertheless, Inconel 718 and Inconel 625 can be used up to around 650 • C for applications under load. In fact, it is well known that the operative temperature of these alloys is limited under high load due to the coarsening and transformation of the metastable γ" (Ni 3 Nb-body-centered tetragonal) phase into the δ (Ni 3 Nb-orthorhombic) phase, which can drastically reduce the mechanical performance [11][12][13][14].
Consequently, there has been growing interest in developing Ni-based superalloys for application at higher temperatures than 650 • C by the LPBF process. Such superalloys include Hastelloy X [15,16], level of LPBFed IN939 alloy. The current work mainly deals with the role of the process parameters on the production of IN939 by LPBF. Moreover, the microstructure and cracking mechanisms are also investigated.

Powder Characterization
The gas atomized powder was purchased by LPW (Carpenter Additive) with the main chemical composition of the used powder reported in Table 1. The declared chemical composition was checked by focused ion beam-scanning electron microscopy (FIB-SEM, TESCAN S9000G, Tescan Orsay Holding, a.s., Brno, Czech Republic) equipped with energy dispersive X-ray spectrometry (EDS) detector in order to have a semi-quantitative identification coupled to C identification by means of combustion infrared analysis, obtaining a good correlation. The declared chemical composition was inside the standard chemical composition of typical cast IN939 alloy, as indicated in Table 1. Table 1. IN939 main chemical composition of the IN939 in weight percent (wt. %): (1) chemical composition of the used powder provided from the supplier; (2) semi-quantitative analysis obtained by EDS analysis except for C determined by combustion infrared analysis of the used powder; (3) standard chemical composition range of cast IN939 alloy, data from [37]. The particles were mostly spherical with the presence of satellites and a few irregular particles, highlighted in Figure 1a by red arrows and red ovals, respectively. Figure 1b shows the particle size distribution and cumulative frequency of the powder, revealing a D10 of 16.8 µm, D50 of 26.1 µm and D90 of 40.7 µm, as determined by laser granulometry analysis. In order to remove possible large particles, the powder was sieved down to 50 µm.
The Hall flow test resulted in a flow time of 15.1 ± 0.2 s and an apparent density of 4.29 ± 0.01 g/cm 3 , following the ASTM B213 and ASTM B212 standards, respectively.
The cross-section of the particles (Figure 1c) polished up to 1 µm with diamond suspension revealed mainly the presence of particles with a low quantity of small spherical pores together with the limited existence of particles characterized by pores around 20 µm. These pores are typically formed due to the entrapped gas during the atomization process. Analyzing 3000 particles, the total residual porosity was assessed as 0.4 ± 0.2%. Finally, after chemical etching, the cross-section (Figure 1d) revealed dendrites with a width around 1-2 µm, enriched in Ti, W, and Nb and depleted of Ni, Cr, and Co within the interdendritic areas. The analysis was performed by means of the scanning electron microscope (SEM, Phenom XL, Phenom-World BV, Eindhoven, The Netherlands) equipped with EDS detector.

Process Parameter
IN939 cubic samples (10 × 10 × 10 mm 3 ) were fabricated by means of a CONCEPT Mlab Cusing R system (Concept Laser GmbH, Lichtenfels, Germany) with laser power up to 100 W with a laser spot size of approximately 50 µ m. For the process parameters optimization, it was chosen to modify the scanning speed and hatching distance keeping constant the laser power at 95 W and layer thickness at 20 µ m. All the chosen parameter conditions are provided in Table A1 in Appendix A.
In order to define the energy delivered to the powder, the volumetric energy density (VED) is one of the approaches used for the process optimization of LPBFed materials. The VED (J/mm 3 ) takes into account the laser power P (W), the scanning speed v (mm/s), the layer thickness t, and the hatching distance hd (both in mm) [44][45][46]: VED values from 30 to 320 J/mm 3 were selected to study the densification behavior of LPBFprocessed specimens. A fixed scanning strategy was used, consisting of stripes of 5 mm with a rotation of 67° between consecutive layers of powder. A schematic representation of the scanning strategy is illustrated in Figure 2.

Process Parameter
IN939 cubic samples (10 × 10 × 10 mm 3 ) were fabricated by means of a CONCEPT Mlab Cusing R system (Concept Laser GmbH, Lichtenfels, Germany) with laser power up to 100 W with a laser spot size of approximately 50 µm. For the process parameters optimization, it was chosen to modify the scanning speed and hatching distance keeping constant the laser power at 95 W and layer thickness at 20 µm. All the chosen parameter conditions are provided in Table A1 in Appendix A.
In order to define the energy delivered to the powder, the volumetric energy density (VED) is one of the approaches used for the process optimization of LPBFed materials. The VED (J/mm 3 ) takes into account the laser power P (W), the scanning speed v (mm/s), the layer thickness t, and the hatching distance h d (both in mm) [44][45][46]: VED values from 30 to 320 J/mm 3 were selected to study the densification behavior of LPBF-processed specimens. A fixed scanning strategy was used, consisting of stripes of 5 mm with a rotation of 67 • between consecutive layers of powder. A schematic representation of the scanning strategy is illustrated in Figure 2.

Microstructural Investigation
For the densification analysis (porosity and cracking density) and microstructure investigation, the zy plane was considered. Moreover, for the texture analysis, both the zy and xy planes were used, as schematically displayed in Figure 3. The samples were cut and then polished down to 1 µ m using diamond suspensions. The residual porosity was determined by a light optical microscope (LOM, Leica DMI 5000 M, Leica Microsystems, Wetzlar, Germany), taking 10 images at 50×, analyzing a total area of around 26 mm 2 . The optical micrographs were post-processed to determine the residual porosity in percentage through the Image J software. In order to determine the porosity, the cracks were removed from the LOM images.
The cracks resulted to be very tight, and it was difficult to obtain a reliable evaluation from the LOM images. Therefore, the cracking density was determined by SEM taking 10 images at 500×, investigating a total area around 3 mm 2 , and then post-processing the images by the Image J software. The cracking density reported as mm/mm 2 was determined using the Ferret length as typically considered for other LPBFed Ni-based superalloys subjected to crack formation [6,18,24]. Examples of the procedure employed to analyze the residual porosity by LOM images and cracking density by SEM images, including the obtained post-processed images, are provided in Figure 4. More in detail, the LOM images were processed by the Image J software (version: 1.49v, National Institutes of Health, Bethesda, MD, USA) in order to highlight the defects (pores and cracks) in black, while the austenitic matrix was pointed out in white. Afterward, for the residual porosity determination (Figure 4a), the cracks were removed (Figure 4b), giving the possibility to determine the black areas of the porosity. In the other case, for the cracking density evaluation (Figure 4c), the pores were removed from the image, allowing the determination of the cracks pointed out in black (Figure 4d).

Microstructural Investigation
For the densification analysis (porosity and cracking density) and microstructure investigation, the zy plane was considered. Moreover, for the texture analysis, both the zy and xy planes were used, as schematically displayed in Figure 3.

Microstructural Investigation
For the densification analysis (porosity and cracking density) and microstructure investigation, the zy plane was considered. Moreover, for the texture analysis, both the zy and xy planes were used, as schematically displayed in Figure 3. The samples were cut and then polished down to 1 µ m using diamond suspensions. The residual porosity was determined by a light optical microscope (LOM, Leica DMI 5000 M, Leica Microsystems, Wetzlar, Germany), taking 10 images at 50×, analyzing a total area of around 26 mm 2 . The optical micrographs were post-processed to determine the residual porosity in percentage through the Image J software. In order to determine the porosity, the cracks were removed from the LOM images.
The cracks resulted to be very tight, and it was difficult to obtain a reliable evaluation from the LOM images. Therefore, the cracking density was determined by SEM taking 10 images at 500×, investigating a total area around 3 mm 2 , and then post-processing the images by the Image J software. The cracking density reported as mm/mm 2 was determined using the Ferret length as typically considered for other LPBFed Ni-based superalloys subjected to crack formation [6,18,24]. Examples of the procedure employed to analyze the residual porosity by LOM images and cracking density by SEM images, including the obtained post-processed images, are provided in Figure 4. More in detail, the LOM images were processed by the Image J software (version: 1.49v, National Institutes of Health, Bethesda, MD, USA) in order to highlight the defects (pores and cracks) in black, while the austenitic matrix was pointed out in white. Afterward, for the residual porosity determination (Figure 4a), the cracks were removed (Figure 4b), giving the possibility to determine the black areas of the porosity. In the other case, for the cracking density evaluation (Figure 4c), the pores were removed from the image, allowing the determination of the cracks pointed out in black (Figure 4d). The samples were cut and then polished down to 1 µm using diamond suspensions. The residual porosity was determined by a light optical microscope (LOM, Leica DMI 5000 M, Leica Microsystems, Wetzlar, Germany), taking 10 images at 50×, analyzing a total area of around 26 mm 2 . The optical micrographs were post-processed to determine the residual porosity in percentage through the Image J software. In order to determine the porosity, the cracks were removed from the LOM images.
The cracks resulted to be very tight, and it was difficult to obtain a reliable evaluation from the LOM images. Therefore, the cracking density was determined by SEM taking 10 images at 500×, investigating a total area around 3 mm 2 , and then post-processing the images by the Image J software. The cracking density reported as mm/mm 2 was determined using the Ferret length as typically considered for other LPBFed Ni-based superalloys subjected to crack formation [6,18,24]. Examples of the procedure employed to analyze the residual porosity by LOM images and cracking density by SEM images, including the obtained post-processed images, are provided in Figure 4. More in detail, the LOM images were processed by the Image J software (version: 1.49v, National Institutes of Health, Bethesda, MD, USA) in order to highlight the defects (pores and cracks) in black, while the austenitic matrix was pointed out in white. Afterward, for the residual porosity determination (Figure 4a), the cracks were removed (Figure 4b), giving the possibility to determine the black areas of the porosity.
In the other case, for the cracking density evaluation (Figure 4c), the pores were removed from the image, allowing the determination of the cracks pointed out in black ( Figure 4d). After etching with Kalling's No.2 solution, the microstructure was evaluated by LOM, SEM, and FIB-SEM using EDS detectors. The crystallographic texture of the as-built samples was characterized using FIB-SEM equipped with an electron backscatter diffraction (EBSD) analyzer. The specimens were tilted by 70° and scanned at 20 kV, with a 1-2 µ m step, analyzing them along the building direction (z-axis) and perpendicular to the building direction (xy plane).

Influence of the Process Parameters on the Densification
Porosity and cracking density of the different samples, correlated to the VED values, are reported in Figure 5a,e, respectively. Moreover, the assessed values of residual porosity and cracking density for the various combinations of process parameters are provided in Table A1 in Appendix A.
Considering the porosity (Figure 5a), it was possible to define three distinct areas based on the VED:  VED < 50 J/mm 3 generated large pores and irregular lack of fusion, with sizes also larger than 200 µ m, due to insufficient energy delivered to the material ( Figure 5b);  A wide range of VED values between 50 and 160 J/mm 3 provided an adequate amount of energy to melt the powder, and the predominant defects resulted to be the spherical pores ( Figure 5c), with sizes typically up to around 20 µ m, while some rare irregular pores (lack of fusion) could be still detected. However, it should be noted that some conditions with a VED ranging from 100 to 160 J/mm 3 triggered a high residual porosity, and consequently, the effect of the single parameters should also be considered. An in-depth analysis will be provided in the following part of this paper; After etching with Kalling's No.2 solution, the microstructure was evaluated by LOM, SEM, and FIB-SEM using EDS detectors. The crystallographic texture of the as-built samples was characterized using FIB-SEM equipped with an electron backscatter diffraction (EBSD) analyzer. The specimens were tilted by 70 • and scanned at 20 kV, with a 1-2 µm step, analyzing them along the building direction (z-axis) and perpendicular to the building direction (xy plane).

Influence of the Process Parameters on the Densification
Porosity and cracking density of the different samples, correlated to the VED values, are reported in Figure 5a,e, respectively. Moreover, the assessed values of residual porosity and cracking density for the various combinations of process parameters are provided in Table A1 in Appendix A.
Considering the porosity (Figure 5a), it was possible to define three distinct areas based on the VED: • VED < 50 J/mm 3 generated large pores and irregular lack of fusion, with sizes also larger than 200 µm, due to insufficient energy delivered to the material ( Figure 5b); • A wide range of VED values between 50 and 160 J/mm 3 provided an adequate amount of energy to melt the powder, and the predominant defects resulted to be the spherical pores (Figure 5c), with sizes typically up to around 20 µm, while some rare irregular pores (lack of fusion) could be still detected. However, it should be noted that some conditions with a VED ranging from 100 to 160 J/mm 3 triggered a high residual porosity, and consequently, the effect of the single parameters should also be considered. An in-depth analysis will be provided in the following part of this paper; • For VED > 160 J/mm 3 , the energy gradually started inducing melt pool instability causing the formation of keyholes, with a large number of mostly spherical pores with sizes up to around 100 µm (Figure 5d).
In order to minimize the defects, the presence of an optimized range of VED values is typically determined for LPBFed materials (e.g., Inconel 625, Inconel 718, and Ti6Al4V) [2,6].
On the contrary, the effect of VED values on the cracking density ( Figure 5e) did not show a significant trend. More in detail, different samples built with similar VED exhibited different cracking densities, as also indicated by the LOM images (Figure 5f-h) that show three specimens built with a VED from 85 to 88 J/mm 3 . Additionally, the reduced quantity of cracks for VED < 50 J/mm 3 or VED > 300 J/mm 3 can be associated with the dominant presence of large defects, with a release of the internal stresses that do not promote the formation of cracks, as displayed in the LOM images in Figure 5b,d.  For VED > 160 J/mm 3 , the energy gradually started inducing melt pool instability causing the formation of keyholes, with a large number of mostly spherical pores with sizes up to around 100 µ m (Figure 5d).
In order to minimize the defects, the presence of an optimized range of VED values is typically determined for LPBFed materials (e.g., Inconel 625, Inconel 718, and Ti6Al4V) [2,6].
On the contrary, the effect of VED values on the cracking density ( Figure 5e) did not show a significant trend. More in detail, different samples built with similar VED exhibited different cracking densities, as also indicated by the LOM images (Figure 5f-h) that show three specimens built with a VED from 85 to 88 J/mm 3 . Additionally, the reduced quantity of cracks for VED < 50 J/mm 3 or VED > 300 J/mm 3 can be associated with the dominant presence of large defects, with a release of the internal stresses that do not promote the formation of cracks, as displayed in the LOM images in Figure 5b,d.
The same VED values obtained with different parameters have a different effect on the melt pool solidification and its final shape [47]. The VED approach can be employed only for limiting the tested conditions, but successively, the effect of the singular parameters must be carefully examined to obtain a more accurate parameters optimization.  The hatching distance from 0.11 to 0.15 mm did not allow the fabrication of dense samples, even applying a slow scanning speed (100-400 mm/s). For these conditions, it is evident that the energy provided to the material was not sufficient to achieve a high densification level, also using a slow scanning speed. Hatching distances of 0.07 and 0.09 mm revealed a minimum of residual porosity (<0.5%) for a scanning speed of 1000 and 800 mm/s, respectively. Moreover, a hatching distance of 0.03 and 0.05 mm generated a low residual porosity (<0.5%) using a scanning speed in the range 1000-1400 mm/s and 1000-1200 mm/s, respectively.
Therefore, each value of hatching distance is related to an optimized range of scanning speed. Exceeding the optimized values of scanning speed results in providing less energy to the material, thus producing residual porosities such as lack of fusion. On the contrary, a scanning speed inferior The same VED values obtained with different parameters have a different effect on the melt pool solidification and its final shape [47]. The VED approach can be employed only for limiting the tested conditions, but successively, the effect of the singular parameters must be carefully examined to obtain a more accurate parameters optimization. Figure 6a,b shows the average values of residual porosity for samples built with a different scanning speed and hatching distance, keeping constant the laser power and layer thickness (95 W and 20 µm).
The hatching distance from 0.11 to 0.15 mm did not allow the fabrication of dense samples, even applying a slow scanning speed (100-400 mm/s). For these conditions, it is evident that the energy provided to the material was not sufficient to achieve a high densification level, also using a slow scanning speed. Hatching distances of 0.07 and 0.09 mm revealed a minimum of residual porosity (<0.5%) for a scanning speed of 1000 and 800 mm/s, respectively. Moreover, a hatching distance of 0.03 and 0.05 mm generated a low residual porosity (<0.5%) using a scanning speed in the range 1000-1400 mm/s and 1000-1200 mm/s, respectively.
Therefore, each value of hatching distance is related to an optimized range of scanning speed. Exceeding the optimized values of scanning speed results in providing less energy to the material, thus producing residual porosities such as lack of fusion. On the contrary, a scanning speed inferior to the optimized range tends to transmit too much energy to the system, generating defects due to the melt pool instability.
Metals 2020, 10, x FOR PEER REVIEW 8 of 19 to the optimized range tends to transmit too much energy to the system, generating defects due to the melt pool instability. It is worth noting that a hatching distance of 0.02 mm always generates high residual porosity within the range of the scanning speed employed. This could be associated with too extensive overlapping between two adjacent laser scannings, involving melt pools instability.
In conclusion, low porosities (<0.5%) were determined for the following combinations of scanning speed and hatching distance: (i) 800 mm/s with 0.09, 0.07 and 0.05 mm; (ii) 1000 mm/s with 0.07, 0.05 and 0.03 mm; (iii) 1200 mm/s with 0.05 and 0.03 mm; (iv) 1400 mm/s with 0.03 mm. Similarly, the effect of different scanning speeds and hatching distances on the cracking density are provided in Figure 7a,b, respectively.
For all the operating ranges of hatching distance (0.02-0.15 mm), scanning speed plays a fundamental role, and values inferior to 600 mm/s are typically related to a high crack density. Differently, higher scanning speeds provoked a minor cracking formation, reaching an average cracking density ≤1 mm/mm 2 mainly for scanning speeds from 1000 mm/s to 2200 mm/s, combined with hatching distances in the range of 0.09 to 0.02 mm, as pointed out in Figure 7b. It is worth noting that a hatching distance of 0.02 mm always generates high residual porosity within the range of the scanning speed employed. This could be associated with too extensive overlapping between two adjacent laser scannings, involving melt pools instability.
Similarly, the effect of different scanning speeds and hatching distances on the cracking density are provided in Figure 7a,b, respectively. It is possible to speculate that higher scanning speeds reduce the energy delivered to melt the powder, thus contributing to mitigate the thermal stresses within the material. The cracks are typically formed when the thermal stress overcomes the tensile strength of the material. A more detailed discussion on cracking formation will be reported in Section 3.3.

Selection of Optimal Parameters
The current work investigated the parameters optimization performed keeping constant both laser power (at 95 W) and layer thickness (at 20 µ m), modifying scanning speed and hatching distance. From the results, there was no set of parameters to minimize residual porosity and cracking density simultaneously. This mechanism can be discussed by the dependence of defects (pores and cracks) using some representative combined parameters reported in Figure 8.
Increasing the VED (low scan speed and reduced hatching distance) contributes to providing the full melting of the powder, reducing the risk of lack of fusion, but at the same time it triggers high For all the operating ranges of hatching distance (0.02-0.15 mm), scanning speed plays a fundamental role, and values inferior to 600 mm/s are typically related to a high crack density. Differently, higher scanning speeds provoked a minor cracking formation, reaching an average cracking density ≤1 mm/mm 2 mainly for scanning speeds from 1000 mm/s to 2200 mm/s, combined with hatching distances in the range of 0.09 to 0.02 mm, as pointed out in Figure 7b.
It is possible to speculate that higher scanning speeds reduce the energy delivered to melt the powder, thus contributing to mitigate the thermal stresses within the material. The cracks are typically formed when the thermal stress overcomes the tensile strength of the material. A more detailed discussion on cracking formation will be reported in Section 3.3.

Selection of Optimal Parameters
The current work investigated the parameters optimization performed keeping constant both laser power (at 95 W) and layer thickness (at 20 µm), modifying scanning speed and hatching distance. From the results, there was no set of parameters to minimize residual porosity and cracking density simultaneously. This mechanism can be discussed by the dependence of defects (pores and cracks) using some representative combined parameters reported in Figure 8.

Microstructural Evaluation of LPBFed IN939
The microstructure of samples characterized by the lowest cracking density (sample S38scanning speed of 1800 mm/s and hatching distance of 0.03 mm) is provided in Figures 9.
The EBSD maps (Figure 9a,b) reveal a pronounced texture along the <001> orientation (represented by the red color). The high angle grain boundaries (higher 10°) are marked in black, revealing large elongated grains that can reach a length of around 400 µ m along the z-axis, while the cross-section of the columnar grains with a size mainly around 35 ± 15 µ m are visible along the xy plane. The presence of a strong <001> texture has been often observed for LPBFed Ni-based superalloys [17,19,42,48,49].
The EBSD map in Figure 9c shows an area with smaller grains (highlighted by an oval) with a configuration similar to the melt pool architectures marked in red in the LOM image in Figure 9d. Increasing the VED (low scan speed and reduced hatching distance) contributes to providing the full melting of the powder, reducing the risk of lack of fusion, but at the same time it triggers high thermal residual stresses leading to cracks formation (see samples S19, S28, S21 and S14 in Figure 8). Moreover, too high energy can generate keyholes (see samples S7 and S2 in Figure 8). On the other hand, reducing the VED (high scan speed and large hatching distance) limits the crack formation but involves a higher concentration of residual pores (pores and lack of fusion) throughout the material (see samples S38, S35, S22, S46, S48 and S31 in Figure 8).
Considering the average value with the addition of standard deviation (Appendix A-Table A1), the lowest porosity (≤0.4%) with a moderate cracking density (≤1.5 mm/mm 2 ) was determined for a scanning speed of 1000 and 1200 mm/s combined with a hatching distance of 0.05 and 0.03 mm. Differently, the samples with the lowest cracking density (≤0.4 mm/mm 2 ) and moderate porosity (≤1%) were determined for scanning speed of 1600 and 1800 mm/s combined with a hatching distance of 0.03 mm. Therefore, a hatching distance of 0.03 mm can be selected as the best compromise in order to maximize the warranty to minimize porosity or cracks. Using a larger hatching distance, the too low overlapping of the laser scans promotes the formation of a lack of fusion. On the other hand, inferior hatching distances create melt pool instability due to larger overlapping between scans triggering defects.
Among these conditions, the sample S38 (scanning speed of 1800 mm/s and hatching distance of 0.03 mm) revealed the lowest cracking density of 0.08 ± 0.02 mm/mm 2 with a maximum residual porosity of 0.67 ± 0.22%. On the other hand, the sample S19 (scanning speed of 1000 mm/s and hatching distance of 0.03 mm) resulted in the lowest residual porosity of 0.14 ± 0.08% coupled to a cracking density of 0.80 ± 0.56 mm/mm 2 .
It should be noted that the current level of densification is not suitable for industrial applications due to the presence of large pores or cracks that could reduce the lifetime of the components. However, one of the two selected conditions with the lowest porosity (S19) or with the lowest cracking density (S38) could be subjected to hot isostatic pressing in order to obtain crack-free samples with an extremely low porosity level, making it suitable for industrial applications.

Microstructural Evaluation of LPBFed IN939
The microstructure of samples characterized by the lowest cracking density (sample S38-scanning speed of 1800 mm/s and hatching distance of 0.03 mm) is provided in Figure 9.
The EBSD maps (Figure 9a,b) reveal a pronounced texture along the <001> orientation (represented by the red color). The high angle grain boundaries (higher 10 • ) are marked in black, revealing large elongated grains that can reach a length of around 400 µm along the z-axis, while the cross-section of the columnar grains with a size mainly around 35 ± 15 µm are visible along the xy plane. The presence of a strong <001> texture has been often observed for LPBFed Ni-based superalloys [17,19,42,48,49].
The EBSD map in Figure 9c shows an area with smaller grains (highlighted by an oval) with a configuration similar to the melt pool architectures marked in red in the LOM image in Figure 9d. When the laser beam creates the melt pools, the columnar grains tend to be formed along the thermal flux orientation (towards the building platform). The majority of the grains grow, extending across several melt pools, along the z-axis, even though the overlapping between consecutive melt pools can give birth to smaller grains with different orientations.
The melt pools were composed of fine dendritic/cellular architectures (Figure 10a,b). The dendrites were characterized by sub-micrometric dimensions. From the primary dendritic arm spacing (PDAS), it is possible to roughly estimate the cooling rates by the following equation [49]: where a and b are constants with values 50 µm and 1/3, respectively, for nickel-based alloys, and ε is the cooling rate. The melt pools were composed of fine dendritic/cellular architectures (Figure 10a,b). The dendrites were characterized by sub-micrometric dimensions. From the primary dendritic arm spacing (PDAS), it is possible to roughly estimate the cooling rates by the following equation [49]: = · − (2) where a and b are constants with values 50 μm and 1/3, respectively, for nickel-based alloys, and ε is the cooling rate.
The values of PDAS were determined for the samples built using the parameters to minimize the cracking density (sample S38) and residual porosity (sample S19). For the samples S38 and S19, the PDAS was 0.35 ± 0.10 µ m and 0.28 ± 0.11 µ m, respectively. The PDAS of the two conditions resulted to be very similar, evaluating a cooling rate of around 10 6 °C/s, which is in agreement with the typical range of cooling rates associated with Ni-based superalloys processed by LPBF [25,49].
High-magnification views (Figure 10c,d) revealed the presence of sub-micrometric precipitates, with the largest ones mainly located along the interdendritic areas. These precipitates are most likely to be MC carbides and γ' precipitates [35,41,50]. The formation of larger precipitates along the interdendritic areas is associated with the solidification process. When the dendrites are formed, the remaining liquid surrounding the dendritic core (interdendritic areas) starts to be enriched in segregated elements such as Ti, Nb and Ta (for IN939), leading to the formation of larger precipitates or possible segregations in these areas. For Ni-based superalloys, the presence of interdendritic areas The values of PDAS were determined for the samples built using the parameters to minimize the cracking density (sample S38) and residual porosity (sample S19). For the samples S38 and S19, the PDAS was 0.35 ± 0.10 µm and 0.28 ± 0.11 µm, respectively. The PDAS of the two conditions resulted to be very similar, evaluating a cooling rate of around 10 6 • C/s, which is in agreement with the typical range of cooling rates associated with Ni-based superalloys processed by LPBF [25,49].
High-magnification views (Figure 10c,d) revealed the presence of sub-micrometric precipitates, with the largest ones mainly located along the interdendritic areas. These precipitates are most likely to be MC carbides and γ' precipitates [35,41,50]. The formation of larger precipitates along the interdendritic areas is associated with the solidification process. When the dendrites are formed, the remaining liquid surrounding the dendritic core (interdendritic areas) starts to be enriched in segregated elements such as Ti, Nb and Ta (for IN939), leading to the formation of larger precipitates or possible segregations in these areas. For Ni-based superalloys, the presence of interdendritic areas characterized by chemical segregations can influence the formation of the phases under heat treatments, promoting a different microstructure evolution with respect to the traditional conditions, i.e., casting or wrought states [51][52][53]. In addition, for the LPBF process, the continuous temperature cycling caused by the remelting could provoke the formation of precipitates. characterized by chemical segregations can influence the formation of the phases under heat treatments, promoting a different microstructure evolution with respect to the traditional conditions, i.e., casting or wrought states [51][52][53]. In addition, for the LPBF process, the continuous temperature cycling caused by the remelting could provoke the formation of precipitates. In order to better study the cracking phenomena, it was chosen to observe the sample S19 (scanning speed of 1000 mm/s and hatching distance of 0.03 mm), which is characterized by the lowest residual porosity and moderate cracking density. The micrographs (Figure 11a,b) show that the cracks were mainly located along the grain boundaries, as noted for other LPBFed Ni-based superalloys [16,18,21,24].
An EDS scan line (Figure 11b,c), performed across one of the typical "bridges" of a crack (an example is visible in the yellow circle in Figure 11b), revealed a slight enrichment of Nb, Ti and C which can suggest the formation of carbides, while the increment of Ti may also indicate γ′ formation. Based on the literature on traditional cast IN939, MC carbides, such as TiC and NbC, and γ′ phases could be formed [35,50].
The high-magnification view (Figure 11d) underlines that the cracks are covered by submicrometric intergranular precipitates, as indicated by the yellow arrow. It is possible to assume that the high concentration of precipitates along the grain boundaries can embrittle these areas. Consequently, the thermal stresses can overcome the tensile strength triggering the cracks. Based on the microstructural observations, the cracks may be ascribed to two different mechanisms: (i) liquation cracking due to the melting of eutectic phases such as γ′ precipitates and MC carbides; and (ii) strain-age cracking due to the carbides and γ′ precipitates formation during the heating of the material in the solid-state. In this scenario, a reduction in the C content could limit the formation of the carbides reducing the cracking formation.
Similarly, the critical role of the precipitates in the cracking formation was also reported to explain the limited weldability of the alloy for welding processes [35]. In order to better study the cracking phenomena, it was chosen to observe the sample S19 (scanning speed of 1000 mm/s and hatching distance of 0.03 mm), which is characterized by the lowest residual porosity and moderate cracking density. The micrographs (Figure 11a,b) show that the cracks were mainly located along the grain boundaries, as noted for other LPBFed Ni-based superalloys [16,18,21,24].
An EDS scan line (Figure 11b,c), performed across one of the typical "bridges" of a crack (an example is visible in the yellow circle in Figure 11b), revealed a slight enrichment of Nb, Ti and C which can suggest the formation of carbides, while the increment of Ti may also indicate γ formation. Based on the literature on traditional cast IN939, MC carbides, such as TiC and NbC, and γ phases could be formed [35,50].
The high-magnification view (Figure 11d) underlines that the cracks are covered by sub-micrometric intergranular precipitates, as indicated by the yellow arrow. It is possible to assume that the high concentration of precipitates along the grain boundaries can embrittle these areas. Consequently, the thermal stresses can overcome the tensile strength triggering the cracks. Based on the microstructural observations, the cracks may be ascribed to two different mechanisms: (i) liquation cracking due to the melting of eutectic phases such as γ precipitates and MC carbides; and (ii) strain-age cracking due to the carbides and γ precipitates formation during the heating of the material in the solid-state. In this scenario, a reduction in the C content could limit the formation of the carbides reducing the cracking formation.
Similarly, the critical role of the precipitates in the cracking formation was also reported to explain the limited weldability of the alloy for welding processes [35].

Conclusions
The current work regards the effect of the process parameter on the densification level of LPBFed IN939 alloy, including the investigation of the microstructure.
The process parameters optimization permitted to identify specific conditions to limit the porosity and cracking density.
From the results, the most appropriate hatching distance is 0.03 mm, which provides a good overlapping between consecutive laser scans for the reduction in cracking density and residual porosity. Hatching distances inferior to 0.03 mm create melt pool instability resulting in keyhole formation, while too large hatching distances create lack of fusion due to incomplete overlapping between consecutive laser scans. For the scanning speed, 1000 mm/s is effective in almost suppressing all the porosity, but the high energy delivered to the material results in crack formation. On the other hand, by increasing the scanning speed up to 1800 mm/s, it is possible to reduce the cracking formation, but at the same time, the residual porosity increases due to the lower energy delivered to the material.
From the analysis, the sample S38 (scanning speed of 1800 mm/s and hatching distance of 0.03 mm) exhibited the lowest cracking density of 0.08 ± 0.02 mm/mm 2 with a residual porosity of 0.67 ± 0.22%. On the other hand, the sample S19 (scanning speed of 1000 mm/s and hatching distance of 0.03 mm) presented the lowest residual porosity of 0.14 ± 0.08% together with a cracking density of 0.80 ± 0.56 mm/mm 2 .
The microstructure revealed elongated grains along the building direction (z-axis) characterized by a strong texture along the <001> orientation. Inside the grains, a network of dendritic/cellular architectures with sub-micrometric phases mainly located along the grain boundaries and interdendritic areas were observed. This indicates that the high cooling rates of the process did not inhibit the formation of phases during the solidification.

Conclusions
The current work regards the effect of the process parameter on the densification level of LPBFed IN939 alloy, including the investigation of the microstructure.
The process parameters optimization permitted to identify specific conditions to limit the porosity and cracking density.
From the results, the most appropriate hatching distance is 0.03 mm, which provides a good overlapping between consecutive laser scans for the reduction in cracking density and residual porosity. Hatching distances inferior to 0.03 mm create melt pool instability resulting in keyhole formation, while too large hatching distances create lack of fusion due to incomplete overlapping between consecutive laser scans. For the scanning speed, 1000 mm/s is effective in almost suppressing all the porosity, but the high energy delivered to the material results in crack formation. On the other hand, by increasing the scanning speed up to 1800 mm/s, it is possible to reduce the cracking formation, but at the same time, the residual porosity increases due to the lower energy delivered to the material.
From the analysis, the sample S38 (scanning speed of 1800 mm/s and hatching distance of 0.03 mm) exhibited the lowest cracking density of 0.08 ± 0.02 mm/mm 2 with a residual porosity of 0.67 ± 0.22%. On the other hand, the sample S19 (scanning speed of 1000 mm/s and hatching distance of 0.03 mm) presented the lowest residual porosity of 0.14 ± 0.08% together with a cracking density of 0.80 ± 0.56 mm/mm 2 .
The microstructure revealed elongated grains along the building direction (z-axis) characterized by a strong texture along the <001> orientation. Inside the grains, a network of dendritic/cellular architectures with sub-micrometric phases mainly located along the grain boundaries and interdendritic areas were observed. This indicates that the high cooling rates of the process did not inhibit the formation of phases during the solidification.
The presence of high thermal residual stresses, due to the rapid cooling rate and subsequent heating/cooling cycles, coupled to a high fraction of intergranular precipitates, seems to provoke the formation of cracks along the grain boundaries.
The main results of the current study are the indication of specific parameters to process Inconel 939 by LPBF, revealing parameters for the fabrication of samples with a low cracking density and low residual porosity. Additionally, understanding the cracking mechanisms can help to identify strategies to mitigate or eliminate crack formation. For example, the reduction of C within the starting powder may limit the intergranular carbides, reducing the number of cracks.
Future studies will be performed to evaluate the mechanical properties of the as-built materials, as well as to investigate the effect of hot isostatic pressing on the densification, microstructure and mechanical properties. Hot isostatic pressing, in fact, can be effective in completely consolidating the material, making it suitable for industrial applications. Finally, future investigations will be carried out on the optimization of specific heat treatments to improve the mechanical properties. Table A1. List of the process parameters employed to fabricate the different conditions coupled to the resulting total void content obtained using porosity content and crack density. P = 95 W and t = 20 µm are fixed parameters.