Microstructural and Microhardness Evolution from Homogenization and Hot Isostatic Pressing on Selective Laser Melted Inconel 718: Structure,Texture, and Phases

In this work, the microstructure, texture, phases, and microhardness of 45◦ printed (with respect to the build direction) homogenized, and hot isostatically pressed (HIP) cylindrical IN718 specimens are investigated. Phase morphology, grain size, microhardness, and crystallographic texture at the bottom of each specimen differ from those of the top due to changes in cooling rate. High cooling rates during the printing process generated a columnar grain structure parallel to the building direction in the as-printed condition with a texture transition from (001) orientation at the bottom of the specimen to (111) orientation towards the specimen top based on EBSD analysis. A mixed columnar and equiaxed grain structure associated with about a 15% reduction in texture is achieved after homogenization treatment. HIP treatment caused significant grain coarsening, and engendered equiaxed grains with an average diameter of 154.8 μm. These treatments promoted the growth of δ-phase (Ni3Nb) and MC-type brittle (Ti, Nb)C carbides at grain boundaries. Laves phase (Fe2Nb) was also observed in the as-printed and homogenized specimens. Ostwald ripening of (Ti, Nb)C carbides caused excessive grain growth at the bottom of the HIPed IN718 specimens, while smaller grains were observed at their top. Microhardness in the as-fabricated specimens was 236.9 HV and increased in the homogenized specimens by 19.3% to 282.6 HV due to more even distribution of secondary precipitates, and the nucleation of smaller grains. A 36.1% reduction in microhardness to 180.5 HV was found in the HIPed condition due to γ′′ phase dissolution and differences in grain morphology.


Introduction
Inconel 718 (IN718) is widely used primarily due to its excellent mechanical properties and corrosion resistance at high temperatures [1]. The alloying elements of Inconel 718 include iron (Fe), chromium (Cr), niobium (Nb), and molybdenum (Mo), as well as small amounts of titanium (Ti), regions [11]. Mostafa et al. [13] described the effects of homogenization and homogenization with HIP treatment on IN718 specimens SLM-printed perpendicular to the building direction. The study found that post-manufacturing treatments reorganized the {002} dominant columnar grain structure into equiaxed grains with a {111} dominant orientation [13]. (Nb 0.78 Ti 0.22 )C carbides and δ-phase precipitates were also reported to form due to the diffusion of constituent elements and the high temperature treatment which led to dissolution of γ [13].
Tucho et al. [20] reported differences in microhardness between the top and bottom of SLM IN718 printed vertically (with respect to the building direction). The cause of these discrepancies was not verified in the study [20]. The aim of this study is to analyze and quantify the local differences in microstructure and microhardness that occur due to the build orientation of 3D printed IN718 using SLM, as well as to determine the effects of building strategy on the microstructure and localized mechanical properties after heat treatment. Additional investigation into the effects of building strategy and heat treatments on the microstructure, texture, and phase evolution of laser melted IN718 is required to generate a full understanding of these processes. This added knowledge will be useful in optimizing the microstructural and mechanical properties of SLM manufactured parts.

Materials and Methods
The material used to manufacture SLM specimens was gas-atomized Inconel 718 powder provided by EOS-GmbH (Krailling, Germany) [21]. Composition of the powder was determined by Mostafa et al. [13].
Cylindrical specimens were printed using an EOSINT ® M280 400 W Yb:YAG fiber laser and parameter set Inconel 718_Performance 1.0 (285 W laser power, 0.11 mm hatching distance, 40 µm layer thickness, 100 µm laser beam diameter, and 960 mm/s scanning speed). The platform was preheated to 80 • C and held at this temperature to reduce the thermal gradient between fabricated parts and the platform [11,22]. The cylinders were 12 mm in diameter and 20 mm in length, and were printed at a 45 • with respect to the building direction as shown in Figure 1a. The representation of x, y, and z axes in the figure will be used as a reference in this paper. The laser scanning strategy consisted of bidirectional laser tracks and a hatch angle of 67 • in each consecutive layer as illustrated in Figure 1b [13]. The cylinders underwent homogenization (as-printed specimens heated to 1100 • C for an hour, then furnace cooled), and hot isostatic pressing (homogenized specimens were heated to 1160 • C under 100 MPa of pressure for 4 h, then furnace cooled) treatments.
found that post-manufacturing treatments reorganized the {002} dominant columnar grain structure into equiaxed grains with a {111} dominant orientation [13]. (Nb0.78Ti0.22)C carbides and δ-phase precipitates were also reported to form due to the diffusion of constituent elements and the high temperature treatment which led to dissolution of ′′ [13].
Tucho et al. [20] reported differences in microhardness between the top and bottom of SLM IN718 printed vertically (with respect to the building direction). The cause of these discrepancies was not verified in the study [20]. The aim of this study is to analyze and quantify the local differences in microstructure and microhardness that occur due to the build orientation of 3D printed IN718 using SLM, as well as to determine the effects of building strategy on the microstructure and localized mechanical properties after heat treatment. Additional investigation into the effects of building strategy and heat treatments on the microstructure, texture, and phase evolution of laser melted IN718 is required to generate a full understanding of these processes. This added knowledge will be useful in optimizing the microstructural and mechanical properties of SLM manufactured parts.

Materials and Methods
The material used to manufacture SLM specimens was gas-atomized Inconel 718 powder provided by EOS-GmbH (Krailling, Germany) [21]. Composition of the powder was determined by Mostafa et al. [13].
Cylindrical specimens were printed using an EOSINT ® M280 400 W Yb:YAG fiber laser and parameter set Inconel 718_Performance 1.0 (285 W laser power, 0.11 mm hatching distance, 40 µ m layer thickness, 100 µ m laser beam diameter, and 960 mm/s scanning speed). The platform was preheated to 80 °C and held at this temperature to reduce the thermal gradient between fabricated parts and the platform [11,22]. The cylinders were 12 mm in diameter and 20 mm in length, and were printed at a 45° with respect to the building direction as shown in Figure 1a. The representation of x, y, and z axes in the figure will be used as a reference in this paper. The laser scanning strategy consisted of bidirectional laser tracks and a hatch angle of 67° in each consecutive layer as illustrated in Figure 1b [13]. The cylinders underwent homogenization (as-printed specimens heated to 1100 °C for an hour, then furnace cooled), and hot isostatic pressing (homogenized specimens were heated to 1160 °C under 100 MPa of pressure for 4 h, then furnace cooled) treatments. Specimens were cut into horizontal and vertical cross-sections with respect to the cylinder axis, shown in Figure 2, using a diamond-bladed slow cutter in a mineral oil bath to prevent sample heating and for lubrication. Specimens were polished down to 0.25 µ m using glycol-based diamond suspension and etched for metallographic investigation using Kalling's Solution No. 2 (5 g CuCl2, 100 mL HCl, 100 mL ethanol). Specimens were cut into horizontal and vertical cross-sections with respect to the cylinder axis, shown in Figure 2, using a diamond-bladed slow cutter in a mineral oil bath to prevent sample heating and for lubrication. Specimens were polished down to 0.25 µm using glycol-based diamond suspension and etched for metallographic investigation using Kalling's Solution No. 2 (5 g CuCl 2 , 100 mL HCl, 100 mL ethanol). Optical microscopy (OM) was carried out using an Olympus BX51M (Tokyo, Japan) equipped with a digital camera. The optical micrographs were processed using Image J ® software (National Institutes of Health, Bethesda, MD, USA) [23] in order to determine the porosity percentage and melt pool dimensions. Scanning electron microscopy (SEM) was performed with an FEI Quanta 250 field emission gun (Hillsboro, OR, USA) equipped with an energy dispersive X-ray spectrometer (EDS, EDAX Inc., Mahwah, NJ, USA). SEM and EDS experiments were carried out at 20 KV and factory preset spot size number 5. An HKL electron backscattered diffraction detector (EBSD, EDAX Inc., Mahwah, NJ, USA) was used to obtain crystallographic orientations and grain size and shape distributions. EBSD images were retrieved with a step size of 0.8 µ m and a resolution of 1024 × 800 pixels. EBSD image post-processing consisted of grain confidence index standardization (angle: 5°, size: 2), neighbor confidence index correlation (≥0.1), neighbor orientation correlation, and grain dilation. Specimens were prepared for EBSD analysis by grinding and polishing down to 0.05 µ m without etching. X-ray diffraction (XRD) analysis was performed using a PANalytical Empyrean XRD (Spectris plc, Almelo, The Netherlands) to retrieve crystallographic information and phase quantities in bulk specimens. Experimental XRD patterns were refined using Rietveld analysis in the PANalytical X'pert Highscore software (Version 3.0.2, Almelo, The Netherlands) [24] and compared with standard patterns obtained from Pearson's crystal database [25].
Microindentation hardness tests were performed with a 50 g load using a Vickers HIGHWOOD hardness tester (TTS Unlimited Inc., Kita-Ku, Osaka, Japan). A minimum of seven (HV) tests were conducted along evenly-spaced 1 mm intervals at each of the vertical and horizontal cross-sections of as-printed, homogenized, and homogenized followed by HIP specimens with the exception of the bottom as-printed horizontal cross-section. Only two (HV) tests were possible on the bottom as-printed horizontal cross-section due to the deep surface inclination of the specimen. The data presented correspond to an average hardness observed at each location.

Microstructure of the As-Printed Inconel 718
Due to the unique printing strategy at 45° with respect to the building direction, microstructural properties of the cylindrical Inconel 718 specimen are highly location dependent. These locations can be divided into roughly three sections, the top (1), middle (2), and bottom (3) sections labeled in Figure 3a. The melt pool morphology at each location in the vertical section is displayed in Figure 3b. The images displayed in Figure 3b were taken from the top corner of the specimen (1), the center of the specimen (2), and the bottom corner of the specimen (3). In order to accurately compare the specimens, the micrographs of other conditions are taken at the same positions. The dotted lines illustrated in Figure 3a mark the transitions in melt pool morphology observed in optical micrographs of the cylinder. Morphological variations in these sections can be attributed to the large thermal gradient between older and newer deposited layers during the SLM process [10,13].

Horizontal Cross Section
Vertical Cross Section Optical microscopy (OM) was carried out using an Olympus BX51M (Tokyo, Japan) equipped with a digital camera. The optical micrographs were processed using Image J ® software (National Institutes of Health, Bethesda, MD, USA) [23] in order to determine the porosity percentage and melt pool dimensions. Scanning electron microscopy (SEM) was performed with an FEI Quanta 250 field emission gun (Hillsboro, OR, USA) equipped with an energy dispersive X-ray spectrometer (EDS, EDAX Inc., Mahwah, NJ, USA). SEM and EDS experiments were carried out at 20 KV and factory preset spot size number 5. An HKL electron backscattered diffraction detector (EBSD, EDAX Inc., Mahwah, NJ, USA) was used to obtain crystallographic orientations and grain size and shape distributions. EBSD images were retrieved with a step size of 0.8 µm and a resolution of 1024 × 800 pixels. EBSD image post-processing consisted of grain confidence index standardization (angle: 5 • , size: 2), neighbor confidence index correlation (≥0.1), neighbor orientation correlation, and grain dilation. Specimens were prepared for EBSD analysis by grinding and polishing down to 0.05 µm without etching. X-ray diffraction (XRD) analysis was performed using a PANalytical Empyrean XRD (Spectris plc, Almelo, The Netherlands) to retrieve crystallographic information and phase quantities in bulk specimens. Experimental XRD patterns were refined using Rietveld analysis in the PANalytical X'pert Highscore software (Version 3.0.2, Almelo, The Netherlands) [24] and compared with standard patterns obtained from Pearson's crystal database [25].
Microindentation hardness tests were performed with a 50 g load using a Vickers HIGHWOOD hardness tester (TTS Unlimited Inc., Kita-Ku, Osaka, Japan). A minimum of seven (HV) tests were conducted along evenly-spaced 1 mm intervals at each of the vertical and horizontal cross-sections of as-printed, homogenized, and homogenized followed by HIP specimens with the exception of the bottom as-printed horizontal cross-section. Only two (HV) tests were possible on the bottom as-printed horizontal cross-section due to the deep surface inclination of the specimen. The data presented correspond to an average hardness observed at each location.

Microstructure of the As-Printed Inconel 718
Due to the unique printing strategy at 45 • with respect to the building direction, microstructural properties of the cylindrical Inconel 718 specimen are highly location dependent. These locations can be divided into roughly three sections, the top (1), middle (2), and bottom (3) sections labeled in Figure 3a. The melt pool morphology at each location in the vertical section is displayed in Figure 3b. The images displayed in Figure 3b were taken from the top corner of the specimen (1), the center of the specimen (2), and the bottom corner of the specimen (3). In order to accurately compare the specimens, the micrographs of other conditions are taken at the same positions. The dotted lines illustrated in Figure 3a mark the transitions in melt pool morphology observed in optical micrographs of the cylinder. Morphological variations in these sections can be attributed to the large thermal gradient between older and newer deposited layers during the SLM process [10,13]. The contour layer of the as-printed specimen can be seen at location 1 of Figure 3b. A higher average concentration of porosity is observed to have developed in the contour layer (~0.7%) compared to the interior of the specimen (~0.04%) due to a higher energy density used for Due to the unique printing strategy of the cylinder at a 45 degree angle with the building direction, the microstructural properties are highly location dependent. These locations can be divided up roughly into three sections, the top (1), middle (2), and bottom sections (3) shown in Figure 4. The microstructural variation in these sections can be attributed to the large thermal gradient between older and newer deposited layers during melting. Cooling rates between layers were estimated to be between 200-5000 °C /s for stainless steel during laser melting deposition (LMD) [24]. Section 1 is subjected to lower rates of cooling since successive layers add heat to the matrix and the distance that the laser must travel dwindles near the top causing faster layer deposition and lowering the time for cooling between layers. The first layers deposited in section 3 undergo the highest cooling rates due to the steep thermal gradient near the base plate. Section 2 is located between 1 and 3, and is characterized by cooling rates between these two sections.   The contour layer of the as-printed specimen can be seen at location 1 of Figure 3b. A higher average concentration of porosity is observed to have developed in the contour layer (~0.7%) compared to the interior of the specimen (~0.04%) due to a higher energy density used for contouring. Lower scanning speeds and smaller hatching distances and, therefore, higher energy densities, are used for contouring with the objective of improving surface roughness [26]. During exposure to laser power with a relatively high energy density, a higher concentration of pores with high sphericity form during printing due to gas entrapped in the powder particles [27]. Differences in the morphology at each location in Figure 3 can be explained by a process of vertical and horizontal melt pool interactions [13]. Melt pools in location 3 appear to be relatively shallow and have an average depth of 51.1 ± 18.3 µm. The significant difference in temperature between the building platform and the melted powder causes a large temperature gradient and high cooling rates at the beginning of the 3D printing process [28,29]. Rapid cooling towards the bottom of the specimen causes melted powder to solidify and cool quickly. This cooled material requires more energy to remelt, which reduces the penetration depth of the laser.
A more typical semicircular melt pool morphology is observed at location 2. The average melt pool depth at this location (56.5 ± 19.1 µm) is larger than the value found at location 3, yet smaller than the depth at location 1 (72.8 ± 25.0 µm). Every layer added to the specimen gives additional heat to the matrix. New layers at location 2 are deposited on previous layers which are at relatively higher temperatures than those at location 3. It, therefore, takes less energy to remelt the previously-deposited layers at location 2 than it does at location 3, allowing the laser to penetrate further into the previous layer. Relatively low cooling rates and a reduction in the path that the laser travels at the top corner of the specimen cause this region to have the deepest melt pools. The decreased laser path accelerates the deposition of new layers, reducing the amount of time for previous layers to cool. Figure 4 displays SEM images of melt pool boundaries and inter-dendritic grain structures in both etched and non-etched specimens. These images were taken at locations 1, 2, and 3 of the vertical cross-section of the specimen displayed in Figure 3. Melt pool boundaries are clearly visible in both the etched and non-etched SEM images. The average diameter of melt pools at location 3 is 175.1 ± 83.1 µm. These relatively wide and shallow melt pools appear to be flat layered boundaries in the SEM images at location 3 in Figure 4. Average melt pool diameters at locations 2 and 1 are 108.0 ± 26.6 µm and 130.44 ± 38.1 µm, respectively. Visible melt pool lines were reported to disappear in the upper layers of horizontally-printed specimens (with respect to the building direction) [13]. In comparison, this study finds that the melt pool boundaries at location 1 are still visible.
Long and thin columnar grains demonstrating epitaxial growth over multiple layers of melt pool boundaries in the building direction and faint remnants of micro-dendritic structures can be observed at location 3 in Figure 4. Grain growth over melt pool boundaries is due to the remelting of previous layers at the bottom of the specimen [30]. Columnar grains crossing over melt pool boundaries are also observed in locations 1 and 2 of Figure 4. However, the grains at these locations appear to have less elongated and relatively more equiaxed structures. Regions (a) and (b) indicated in the non-etched image at location 2 in Figure 4 exhibit differences in primary dendrite arm spacing (PDAS) within the melt pool structures, corresponding to the changing cooling rates. Region (a) contains dendrites with an average PDAS of~0.37 µm, whereas the dendrites in region (b) have undergone partial dissolution and have an average arm spacing of~0.2 µm. Region (a) is located within an area of the previous layer that has been re-melted. Due to this vertical melt pool overlap, region (a) is expected to have had a larger heat content than region (b). These overlapping melt pool boundaries have been reported to act as nucleation sites for new grains in the successive printed layers [13]. Dendritic structures form due to rapid cooling rates in the laser melting process [10], which were estimated to be between 10 5 -10 8 • C/s [31,32]. Micro-dendrites appear to have almost vanished in the bottom layers (location 3) of the specimen. This is due to the increased interaction between the horizontal and vertical melt pools in this region [13].

Microstructural Development of Heat Treated Inconel 718
After the initial homogenization treatment, melt pool boundaries are no longer visible. Inter-dendritic regions seen in as-printed specimens have mostly dissolved into the matrix, however, faint remnants are still discernable. Precipitate phases along grain boundaries have replaced Nb-rich white inter-dendritic regions within the matrix after homogenization of the as-printed specimens. Observable precipitated phases distributed throughout the specimen include the needle-like δ phase, globular MC-type carbide phase, and a white plate-like phase that is most likely a Laves phase. The ′′ phase that existed in the interdendritic regions has disseminated and precipitated more evenly into the matrix, similarly to what was reported in the literature [4]. Phases , ′′ , δ, MC-carbide, and Laves phase are shown in Figure 6 and labeled 1 to 5, respectively. A semi-quantitative EDS analysis of these phases is covered in Section 3.2. Grain size and aspect ratio comparisons obtained by EBSD analysis are discussed in Section 3.3.2.

Microstructural Development of Heat Treated Inconel 718
After the initial homogenization treatment, melt pool boundaries are no longer visible. Inter-dendritic regions seen in as-printed specimens have mostly dissolved into the γ matrix, however, faint remnants are still discernable. Precipitate phases along grain boundaries have replaced Nb-rich white inter-dendritic regions within the matrix after homogenization of the as-printed specimens. Observable precipitated phases distributed throughout the specimen include the needle-like δ phase, globular MC-type carbide phase, and a white plate-like phase that is most likely a Laves phase. The γ phase that existed in the interdendritic regions has disseminated and precipitated more evenly into the matrix, similarly to what was reported in the literature [4]. Phases γ, γ , δ, MC-carbide, and Laves phase are shown in Figure 6 and labeled 1 to 5, respectively. A semi-quantitative EDS analysis of these phases is covered in Section 3.2. Grain size and aspect ratio comparisons obtained by EBSD analysis are discussed in Section 3.3.2. Specimens that undergo hot isostatic pressing after homogenization contain the matrix phase, as well as larger MC-carbide and δ phase precipitates labeled 1, 2, and 3 in Figure 7a, respectively. The Laves phase has dissolved completely and ′′ is not visible in the SEM images, as shown in Figure 7d. It is possible that ′′ particles are not visible in HIP-treated specimens due to their small size, which can be as small as 20 nm [4,33]. Unlike homogenized specimens, MC-carbides are not evenly distributed across the HIPed specimens. Dense regions of these globular precipitates can be seen in a ring about ~175 µ m thick around the edges of horizontal cross-sections of the sample and also around the outer edges of vertical cross-sections, displayed in Figure 8b,c. The area percentage of these carbides is calculated from SEM images of vertical and horizontal cross-sections using image analysis, as summarized in Figure 8.
The ′′ and δ phases dissolve at temperatures above 1032 °C [34]. Dissolution of ′′ stimulates the growth of MC-carbides during HIP treatment [13]. Carbides with the composition NbC were reported in homogenized and HIPed SLM IN718 printed horizontally with respect to the building direction [13]. NbC carbides have a solvus temperature between 1040-1093 °C [34]. Rao et al. [35] observed Nb-and Ti-rich MC-carbides along grain boundaries in powder compaction-sintering specimens of IN718 fabricated by HIP (1200 °C at 120 MPa for 3 h), a subsequent heat treatment (955 °C for one hour, then quenched), and double aging (720 °C for 8 h and furnace cooled followed by 620 °C for 8 h, then air cooled). They reported that the dissolution of these MC-carbides occurred during solution treatment of the specimens at 1150 °C for 1 h [35]. Lower percentages of carbides in the middle of the specimen can be attributed to differences in cooling rates between the surface and interior of the specimen, stimulating carbide dissolution during HIP treatment. Although furnace cooling occurs at a rate of around 2 × 10 −3 °C/s [13], specimen outer layers cool at a higher rate than the interior. Due to comparatively slower cooling, carbide dissolution occurs for a longer period of time in the interior. Area fractions calculated in Figure 8 indicate larger percentages of MC-carbides residing at the top horizontal cross-section of the HIPed specimen as compared to the bottom. This is verified by Rietveld analysis of XRD data in Section 3.3.1. Segregation of C and Nb during non-equilibrium solidification is the driving force for carbide formation [36]. This occurs during homogenization and HIP treatments due to Nb-rich ′′ dissolution, as mentioned earlier. However, as relative amounts of ′′ decrease during HIP treatment, carbide formation slows and the phase begins to dissolve through Ostwald ripening [34]. Specimens that undergo hot isostatic pressing after homogenization contain the γ matrix phase, as well as larger MC-carbide and δ phase precipitates labeled 1, 2, and 3 in Figure 7a, respectively. The Laves phase has dissolved completely and γ is not visible in the SEM images, as shown in Figure 7d. It is possible that γ particles are not visible in HIP-treated specimens due to their small size, which can be as small as 20 nm [4,33]. Unlike homogenized specimens, MC-carbides are not evenly distributed across the HIPed specimens. Dense regions of these globular precipitates can be seen in a ring about~175 µm thick around the edges of horizontal cross-sections of the sample and also around the outer edges of vertical cross-sections, displayed in Figure 8b,c. The area percentage of these carbides is calculated from SEM images of vertical and horizontal cross-sections using image analysis, as summarized in Figure 8.
The γ and δ phases dissolve at temperatures above 1032 • C [34]. Dissolution of γ stimulates the growth of MC-carbides during HIP treatment [13]. Carbides with the composition NbC were reported in homogenized and HIPed SLM IN718 printed horizontally with respect to the building direction [13]. NbC carbides have a solvus temperature between 1040-1093 • C [34]. Rao et al. [35] observed Nband Ti-rich MC-carbides along grain boundaries in powder compaction-sintering specimens of IN718 fabricated by HIP (1200 • C at 120 MPa for 3 h), a subsequent heat treatment (955 • C for one hour, then quenched), and double aging (720 • C for 8 h and furnace cooled followed by 620 • C for 8 h, then air cooled). They reported that the dissolution of these MC-carbides occurred during solution treatment of the specimens at 1150 • C for 1 h [35]. Lower percentages of carbides in the middle of the specimen can be attributed to differences in cooling rates between the surface and interior of the specimen, stimulating carbide dissolution during HIP treatment. Although furnace cooling occurs at a rate of around 2 × 10 −3 • C/s [13], specimen outer layers cool at a higher rate than the interior. Due to comparatively slower cooling, carbide dissolution occurs for a longer period of time in the interior. Area fractions calculated in Figure 8 indicate larger percentages of MC-carbides residing at the top horizontal cross-section of the HIPed specimen as compared to the bottom. This is verified by Rietveld analysis of XRD data in Section 3.3.1. Segregation of C and Nb during non-equilibrium solidification is the driving force for carbide formation [36]. This occurs during homogenization and HIP treatments due to Nb-rich γ dissolution, as mentioned earlier. However, as relative amounts of γ decrease during HIP treatment, carbide formation slows and the phase begins to dissolve through Ostwald ripening [34]. Due to the higher percentage of ′′ segregated within interdendritic regions at the top of SLM specimen, discussed further in Section 3.3.1, the region undergoes more carbide growth. Dissolution of the carbide phase through Ostwald ripening is linked with rapid grain growth in wrought IN718 [34]. This excessive grain growth did not occur alongside NbC dissolution in heat-treated compacted powder specimens due to the presence of prior particle boundaries (PPB) restricting the growth of grains [35]. However, SLM-manufactured IN718 specimens are free of PPBs [37]. Without PPBs inhibiting grain growth, Ostwald ripening of NbC precipitates is expected to lead to excessive grain coarsening in heat treated SLM specimens. This is evident through the presence of larger grains in the bottom horizontal cross-section of the HIPed cylinder compared to the top, which is discussed in Section 3.3.2. Due to the higher percentage of γ segregated within interdendritic regions at the top of SLM specimen, discussed further in Section 3.3.1, the region undergoes more carbide growth. Dissolution of the carbide phase through Ostwald ripening is linked with rapid grain growth in wrought IN718 [34]. This excessive grain growth did not occur alongside NbC dissolution in heat-treated compacted powder specimens due to the presence of prior particle boundaries (PPB) restricting the growth of grains [35]. However, SLM-manufactured IN718 specimens are free of PPBs [37]. Without PPBs inhibiting grain growth, Ostwald ripening of NbC precipitates is expected to lead to excessive grain coarsening in heat treated SLM specimens. This is evident through the presence of larger grains in the bottom horizontal cross-section of the HIPed cylinder compared to the top, which is discussed in Section 3.3.2.

Phase Analysis of Heat-Treated Inconel 718
Energy dispersive spectroscopy (EDS) was used in combination with XRD to identify and compare observed phases with results reported in the literature. Table 1 summarizes the results of EDS spot analysis on the matrix and precipitates in homogenized specimens. δ and Laves phase precipitates are associated with larger experimental error due to their small size and the volumetric interaction of the electron beam with neighboring phases. Table 1. EDS spot analysis (in at %) of phases 1-5 illustrated in Figure 6.  Inter-dendritic regions of as-printed specimens have been observed to contain the ′′ strengthening phase [11,13,38,39]. MC-type carbides and Laves phase have also been reported to exist in these regions [12,[18][19][20]. The Laves phase ((Fe, Ni, Cr)2(Nb, Ti)) precipitates as a plate-like phase and forms in inter-dendritic regions due to the segregation of Nb and Ti [40,41]. The Laves phase with the composition (Fe2Nb) was detected in fusion zones of Inconel 718 welds [41,42]. EDS analysis showed that the atomic ratio of Fe:Nb in the white plate-like phase observed in homogenized specimens (Spot 5 in Figure 6) is 2.5:1. Although the plate-like phase could not be detected by X-ray diffraction due to a relatively small amount of precipitates, its morphology and composition are comparable with the Laves phase reported in the literature [3,19]. Qi et al. [3] reported the partial dissolution of the Laves phase during solution treatment at 980 °C of laser net shape manufactured specimens. Zhang et al. [19] similarly observed incomplete dissolution of Laves phase in SLM specimens after solution treatment at 980 °C and double aging at 720 °C. The instability of the Laves phase at homogenization temperatures leads to the assumption that it precipitated during SLM, concurrently with ′′ , within the Nb-rich inter-dendritic spaces, and began to dissolve during heat treatment. The

Phase Analysis of Heat-Treated Inconel 718
Energy dispersive spectroscopy (EDS) was used in combination with XRD to identify and compare observed phases with results reported in the literature. Table 1 summarizes the results of EDS spot analysis on the matrix and precipitates in homogenized specimens. δ and Laves phase precipitates are associated with larger experimental error due to their small size and the volumetric interaction of the electron beam with neighboring phases. Table 1. EDS spot analysis (in at %) of phases 1-5 illustrated in Figure 6.  Inter-dendritic regions of as-printed specimens have been observed to contain the γ strengthening phase [11,13,38,39]. MC-type carbides and Laves phase have also been reported to exist in these regions [12,[18][19][20]. The Laves phase ((Fe, Ni, Cr) 2 (Nb, Ti)) precipitates as a plate-like phase and forms in inter-dendritic regions due to the segregation of Nb and Ti [40,41]. The Laves phase with the composition (Fe 2 Nb) was detected in fusion zones of Inconel 718 welds [41,42]. EDS analysis showed that the atomic ratio of Fe:Nb in the white plate-like phase observed in homogenized specimens (Spot 5 in Figure 6) is 2.5:1. Although the plate-like phase could not be detected by X-ray diffraction due to a relatively small amount of precipitates, its morphology and composition are comparable with the Laves phase reported in the literature [3,19]. Qi et al. [3] reported the partial dissolution of the Laves phase during solution treatment at 980 • C of laser net shape manufactured specimens. Zhang et al. [19] similarly observed incomplete dissolution of Laves phase in SLM specimens after solution treatment at 980 • C and double aging at 720 • C. The instability of the Laves phase at homogenization temperatures leads to the assumption that it precipitated during SLM, concurrently with γ , within the Nb-rich inter-dendritic spaces, and began to dissolve during heat treatment.
The γ-phase matrix composition is consistent with the literature [4,13,30]. Wang [13]. This study found a ratio of 1:2.84 (11.22 at % Nb and 31.82 at % Ni) which is close to the theoretical stoichiometric composition (Ni 3 Nb) of δ phase and in agreement with the findings of other investigations [4,13,43].
Formation of δ phase in heat-treated SLM IN718 is due to segregation of Nb throughout inter-dendritic regions in the as-printed condition. These inter-dendritic regions contain γ identified using XRD and EDS analysis, and the Laves phase identified by EDS analysis. Idell et al. [44] found that the concentration of Nb in interdendritic spaces of direct metal laser sintered (DMLS) ATI 718 Plus ® resulted in the formation of a δ phase during solution treatment. Body-centered tetragonal γ undergoes a phase transformation to orthorhombic δ during long-term exposure to high temperatures (100 h at 700 • C and 50 h at 750 • C) [45]. Zhang et al. [19] and Qi et al. [3] reported the growth of the δ phase after partial Laves phase dissolution during solution treatment. Observations of homogenized specimens in this study show the dissemination and precipitation of γ more evenly into the matrix, the partial dissolution of the Laves phase, and the growth of~1-0 µm long and~0.20-µm thick δ phase precipitates. HIP treatment resulted in the dissolution of Laves and γ phases, allowing δ precipitates to grow~1.9 µm long and~0.25 µm thick.
Carbide formation due to the dissolution of Nb-rich γ was discussed for HIP treatment in Section 3.1.2 and is applicable to the formation of carbides after homogenization of laser melted specimens. Reports of the observed MC-carbide stoichiometry in the literature include (Nb 0.78 Ti 0.22 )C [13] and (Nb 0.9 Ti 0.1 )C [46]. This study reports a (Nb 0.76 Ti 0.24 )C carbide stoichiometry which is comparable to the aforementioned observations. Figure 9 displays EDS spectra of the γ, MC-type carbide, δ, and Laves phases.  [13]. This study found a ratio of 1:2.84 (11.22 at % Nb and 31.82 at % Ni) which is close to the theoretical stoichiometric composition (Ni3Nb) of δ phase and in agreement with the findings of other investigations [4,13,43]. Formation of δ phase in heat-treated SLM IN718 is due to segregation of Nb throughout inter-dendritic regions in the as-printed condition. These inter-dendritic regions contain ′′ identified using XRD and EDS analysis, and the Laves phase identified by EDS analysis. Idell et al. [44] found that the concentration of Nb in interdendritic spaces of direct metal laser sintered (DMLS) ATI 718 Plus ® resulted in the formation of a δ phase during solution treatment. Body-centered tetragonal ′′ undergoes a phase transformation to orthorhombic δ during long-term exposure to high temperatures (100 h at 700 °C and 50 h at 750 °C) [45]. Zhang et al. [19] and Qi et al. [3] reported the growth of the δ phase after partial Laves phase dissolution during solution treatment. Observations of homogenized specimens in this study show the dissemination and precipitation of ′′ more evenly into the matrix, the partial dissolution of the Laves phase, and the growth of ~1-0µ m long and ~0.20-µ m thick δ phase precipitates. HIP treatment resulted in the dissolution of Laves and ′′ phases, allowing δ precipitates to grow ~1.9 µ m long and ~0.25 µ m thick.
Carbide formation due to the dissolution of Nb-rich ′′ was discussed for HIP treatment in Section 3.1.2 and is applicable to the formation of carbides after homogenization of laser melted specimens. Reports of the observed MC-carbide stoichiometry in the literature include (Nb0.78Ti0.22)C [13] and (Nb0.9Ti0.1)C [46]. This study reports a (Nb0.76Ti0.24)C carbide stoichiometry which is comparable to the aforementioned observations. Figure 9 displays EDS spectra of the , MC-type carbide, δ, and Laves phases.  Figure 10 displays the XRD spectra of the top and bottom horizontal cross-sections and the vertical cross-section of the as-printed, homogenized, and HIP-treated specimens. Table 2 lists the phase quantification of the horizontal cross-sections from Rietveld analysis of the XRD data. The -matrix and ′′ phase were identified in all specimens, however, the volume fraction of ′′ in heat-treated specimens is listed to be 0%. This could mean that ′′ is too small to be quantified by Rietveld analysis but exists in enough relative amounts to be detected by the XRD. The (002) and (022) ′′ XRD peaks overlap with those of , demonstrating that the precipitates formed in a  Table 2 lists the phase quantification of the horizontal cross-sections from Rietveld analysis of the XRD data. The γ-matrix and γ phase were identified in all specimens, however, the volume fraction of γ in heat-treated specimens is listed to be 0%. This could mean that γ is too small to be quantified by Rietveld analysis but exists in enough relative amounts to be detected by the XRD. The (002) and (022) γ XRD peaks overlap with those of γ, demonstrating that the precipitates formed in a columnar microstructure parallel to the matrix. This phenomenon has been reported in the literature [11,13]. Small γ precipitates were observed in SEM images of homogenized specimens as discussed in Section 3.1.2. However, γ has been reported to dissolve into the γ-matrix after HIP treatment at 1120 • C [38] for 4 h. A larger volume fraction (3.7%) of γ was detected in the top horizontal cross-section of as-printed specimens compared to the bottom cross-section (2.3%). columnar microstructure parallel to the matrix. This phenomenon has been reported in the literature [11,13]. Small ′′ precipitates were observed in SEM images of homogenized specimens as discussed in Section 3.1.2. However, ′′ has been reported to dissolve into the -matrix after HIP treatment at 1120 °C [38] for 4 h. A larger volume fraction (3.7%) of ′′ was detected in the top horizontal cross-section of as-printed specimens compared to the bottom cross-section (2.3%).   Melt pool overlap promotes the dissolution of the white interdendritic regions containing the γ precipitate, as mentioned in Section 3.1.1. If γ dissolution occurred in the bottom of the as-printed specimens at a greater rate than in the top, microhardness would be expected to decrease in the lower layers of the specimens. However, a greater microhardness is observed in the bottom of the as-printed specimens and is discussed in Section 3.4. Increased melt pool overlap in the bottom section of as-printed specimens creates local reheating cycles that could act similarly to heat treatments observed in the literature [4,11,19], causing the dispersion and precipitation of small γ precipitates more evenly into the lower layers of the specimens. This would be consistent with the observations of an increased hardness in the region. Since the segregated γ is disseminated more evenly as smaller particles in the matrix, lower quantities are detected by XRD in the lower layers.

XRD Analysis of Structure and Phases
MC-type carbide, δ, and Laves phases were not detected by XRD analysis of homogenized specimens due to their low quantities. However, (Ti, Nb)C carbides were identified in HIP-treated specimen, indicating an overall increase in carbide volume. XRD analysis confirms that greater volumes of (Ti, Nb)C carbides reside at the top (0.2%) of HIP specimens compared to the bottom (0.1%). The top to bottom ratio of γ in the as-printed specimen quantified by Rietveld analysis is 1.6:1, which is very close to the 2:1 top-to-bottom ratio of MC-carbides in the HIPed specimen. Since γ is more segregated within interdendritic regions in the upper layers of the specimens, more MC-carbide formation occurs.
A strong (002) texture with weaker (111) and (022) peaks was reported in cross-sections parallel to the printing direction of SLM specimens fabricated horizontally with respect to the building direction [13]. Amato et al. [11] observed a more dominant (002) and (111) texture with a smaller (022) peak in a vertical cross-section of as-fabricated SLM specimens built in the vertical direction (with respect to the building direction). Horizontal cross-sections retained the prominent (002) and (111) texture, but not (022) [11].
This work identifies a strong (111) texture in all three cross-sections of the as-printed specimen with a strong (002) texture observed in the vertical cross-section. The peaks indicating the (002) and (022) planes in the top and bottom horizontal cross-sections of the as-printed specimens are small. Significant growth in the (111) direction during homogenization and homogenization followed by HIP treatments was reported for SLM specimens printed horizontally with respect to the building direction [13]. Similarly, XRD analysis of the vertical cross-sections revealed that the (111) orientation remains prominent throughout both the heat treatments in this work. A reduction in the peak of the (002) orientation is observed after homogenization treatment. In contrast, grains oriented in the (002) direction grow significantly during HIP treatment and it becomes the dominant orientation in the specimen. A reduction in the (002) orientation after HIP treatment is evident in the bottom horizontal cross-section. Peak splitting is observed in the (111), (002), and (022) diffraction peaks of HIPed specimens at every cross-section, most notably in the (022) plane. This splitting could be due to the phase transformation of γ" to a lower symmetry phase: from (bct-Ni 3 Nb) to orthorhombic δ-(Ni 3 Nb). Evidence of this is indicated by the higher relative amount of δ phase in the HIPed specimens, as compared to other conditions determined by phase analysis.

Evolution of Structure and Texture
EBSD mapping was used to determine the evolution in structure and texture of as-printed and heat treated Inconel 718. 1450 µm × 1100 µm maps were taken from both the top and bottom horizontal cross-sections and from the vertical cross-section at the locations illustrated in Figure 3. These maps reveal grain size, grain shape, and crystallographic texture at each of the listed locations. Figure 11 displays EBSD maps of the as-printed specimen. These maps reveal grain size, grain shape, and crystallographic texture at each of the listed locations. Figure 11 displays EBSD maps of the as-printed specimen. The evolution of orientation in vertical cross-sections of the as-printed specimen from the mainly (001) and (101) directions at the bottom to a (001) dominant orientation with some (111) and (101) oriented grains at the top, as can be seen in Figure 11. Average grain sizes and morphologies also differ dramatically between the top and bottom, as can be seen in Table 3. Long, thin grains are visible at the bottom of the vertical cross-section. This is quantified by the aspect ratios at the locations 3 (0.253), 2 (0.275), and 1 (0.277) with the bottom having the smallest aspect ratio and the top having the largest. Deviation in grain size and shape is also larger at location 3 in the vertical section, indicating a mixture of large and small columnar and equiaxed grains. Due to the high relative cooling rates at the bottom of the vertical cross-section, grains are stretched towards the cooling direction. Average equivalent grain diameter is larger at the bottom of the vertical section (65 µ m) than at the middle (56 µ m) which is close to the top (55 µ m). Both the bottom and top horizontal cross-sections show that the tops of the columnar grains have a more equiaxed shape with aspect ratios of (0.392) at location 3 and (0.371) at location 1 in Figure 11. The top horizontal section displays larger average grain sizes and a larger aspect ratio in the corresponding vertical section than the bottom due to lower cooling rates. The evolution of orientation in vertical cross-sections of the as-printed specimen from the mainly (001) and (101) directions at the bottom to a (001) dominant orientation with some (111) and (101) oriented grains at the top, as can be seen in Figure 11. Average grain sizes and morphologies also differ dramatically between the top and bottom, as can be seen in Table 3. Long, thin grains are visible at the bottom of the vertical cross-section. This is quantified by the aspect ratios at the locations 3 (0.253), 2 (0.275), and 1 (0.277) with the bottom having the smallest aspect ratio and the top having the largest. Deviation in grain size and shape is also larger at location 3 in the vertical section, indicating a mixture of large and small columnar and equiaxed grains. Due to the high relative cooling rates at the bottom of the vertical cross-section, grains are stretched towards the cooling direction. Average equivalent grain diameter is larger at the bottom of the vertical section (65 µm) than at the middle (56 µm) which is close to the top (55 µm). Both the bottom and top horizontal cross-sections show that the tops of the columnar grains have a more equiaxed shape with aspect ratios of (0.392) at location 3 and (0.371) at location 1 in Figure 11. The top horizontal section displays larger average grain sizes and a larger aspect ratio in the corresponding vertical section than the bottom due to lower cooling rates. Table 3. Average grain shape and diameter obtained from EBSD maps calculated by EDAX's OIM EBSD analysis software. Horizontal cross-section data was averaged from three locations on each section.

Horizontal Cross-Section
Vertical Cross-Section  Figure 12 displays EBSD maps of homogenized and HIP-treated IN718 at the locations indicated in Figure 11. The breakdown of columnar grains and formation of smaller equiaxed grains after heat treatment of SLM as-printed specimens has been termed recrystallization in the literature [11,19]. Recrystallization during homogenization causes average grain sizes to decrease due to the disintegration of columnar structures [13]. Aspect ratios in vertical sections of homogenized specimens have increased, but remnant grains from the printing process are still visible. Location 3 in the vertical cross-section of the homogenized specimen contains larger (54 µm) and more columnar grains (0.333 aspect ratio) with greater deviation in comparison to the other locations. Homogenized horizontal cross-sections are more equiaxed than those of as-printed specimens. Significant grain coarsening occurs during HIP treatment of the homogenized specimens. Grains in the HIPed specimens are also more equiaxed, columnar grains are no longer observable. HIPed specimens have significantly larger grains at the bottom (153 µm) horizontal cross-section compared to the top (70 µm). This can be attributed to the lesser amounts of segregated γ prior to HIP treatment to feed carbide formation, leading to Ostwald ripening of MC-type carbides mentioned in Section 3.1.2. The larger carbide content at the top of the HIPed cylinders has the effect of pinning down moving grain boundaries, contributing to the grain size disparity [12]. HIP-treated specimens contain an equiaxed (0.399-0.419 aspect ratio) grain structure. Location 3 in the vertical cross-sections of homogenized and HIP specimens are observed to retain a more (001) orientation than the other locations.  Figure 11) of the homogenized and HIP-treated specimens.

Vickers Microhardness Measurements
Popovich et al. [30] reported that SLM IN718 specimens fabricated at lower laser power (250 W) had a higher average hardness of 320 HV compared to 287 HV for those fabricated at higher laser power (900 W) while volumetric energy density was held constant. The observed difference in mechanical properties was attributed to a strong (001) texture and the coarser grains in the specimens fabricated by higher laser power [30]. The average Vickers hardness measurements are displayed in Table 4. As-printed hardness values are lower than those printed at lower laser power and scanning speeds in the literature: 365 HV printed with a 170 W laser and 417 mm/s scanning speed [4], 305 HV printed (horizontally with respect to the building direction) with a 175 W laser and 619 mm/s scanning speed [20], and 319 HV (x-z plane) printed with a 100 W laser and 85.7 mm/s scanning speed [12].  Figure 11) of the homogenized and HIP-treated specimens.

Vickers Microhardness Measurements
Popovich et al. [30] reported that SLM IN718 specimens fabricated at lower laser power (250 W) had a higher average hardness of 320 HV compared to 287 HV for those fabricated at higher laser power (900 W) while volumetric energy density was held constant. The observed difference in mechanical properties was attributed to a strong (001) texture and the coarser grains in the specimens fabricated by higher laser power [30]. The average Vickers hardness measurements are displayed in Table 4. As-printed hardness values are lower than those printed at lower laser power and scanning speeds in the literature: 365 HV printed with a 170 W laser and 417 mm/s scanning speed [4], 305 HV printed (horizontally with respect to the building direction) with a 175 W laser and 619 mm/s scanning speed [20], and 319 HV (x-z plane) printed with a 100 W laser and 85.7 mm/s scanning speed [12]. Table 4. Average values of Vickers hardness of the as-printed and heat treated Inconel 718.

Horizontal Cross-Section
Vertical Cross-Section cylinders. However, grains located towards the upper layers of the specimen had greater aspect ratios due to lower cooling rates. Homogenization treatment partially dissolved the interdendritic regions containing the γ and Laves phases, and caused partial recrystallization of the columnar grains. Smaller and more equiaxed grains were observed in the homogenized condition compared to the as-fabricated cylinders. Long and thin columnar grains left over after the printing process still remained in the homogenized condition, particularly at the bottom of the cylinder. Homogenization promoted the nucleation and growth of MC-type carbide and δ phase precipitates, as well as the dissemination of small γ precipitates into the matrix.
A~175-µm thick ring with higher carbide content was observed around the edges of the HIP condition. Lower carbide content was observed towards the interior of the HIPed cylinders due to slower cooling rates. The relative amount of MC-type carbides increased during HIP treatment. However, Ostwald ripening of carbides also occurred. Excessive grain growth towards the bottom of the cylinder during HIP treatment occurred as a result of Ostwald ripening of (Ti, Nb)C carbides, and a lower relative amount of grain-pinning carbide precipitates in the region.
The texture of all cross-sections in the as-printed condition is characterized by a strong (111) orientation. The (111) orientation remains prominent in the vertical cross-sections throughout both homogenization and HIP treatments. A reduction in the (002) peak was observed after homogenization treatment. The (002) peak grew significantly during HIP treatment and became the dominant orientation.
Vickers microhardness in the vertical cross-section was 236.9 HV for the as-fabricated material, 282.6 HV for the homogenized material, and 180.5 HV for the HIPed material. The bottom cross-section of as-printed and homogenized conditions had greater hardness values than their respective top cross-sections due to increased γ dispersion and differences in grain morphology that occurred as a result of the printing direction.