Microstructure Refinement of Bulk Inconel 718 Parts during Fabrication with EB-PBF Using Scanning Strategies: Transition from Bidirectional-Raster to Stochastic Point-Based Melting

1. Introduction

Metal additive manufacturing (AM) has transformed the modern aerospace and defense sectors. [1,2,3]. This manufacturing approach nowadays has been widely adopted and suited for manufacturing of aerospace alloys such as aluminum and nickel-based superalloys to produce operational components [4,5,6]. AM technologies like powder bed fusion (PBF), direct energy deposition (DED), and binder jetting have allowed the possibility of producing single-part components while substantially reducing weight without compromising structural integrity. [7,8,9].

Inconel 718 is a widely used nickel-based superalloy. It is a precipitation-hardened alloy and can retain strength at elevated temperatures up to 650°C. It compromises primarily 50-55% nickel and 17-21% chromium, alongside 4.8-5.5% niobium, 2.8-3% molybdenum, 0.65-1.15% titanium, and 1% cobalt, with iron and aluminum constituting the balance [10,11]. This alloy exhibits an excellent combination of resistance to creep, oxidation, and corrosion, making it suitable for aerospace applications and sectors, including aircraft, gas turbine engines, rocket engines, valves, flow control devices, turbine blades, etc. [12].

Traditional manufacturing processes like casting and powder metallurgy have been proven effective in producing functional components of Inconel 718[12,13,14]. However, due to the nature of the alloy, these methods face challenges during complex geometry fabrication, including issues like porosity from entrapped gas during casting and issues with sintering during powder metallurgy processes [13,15]. Moreover, these parts require extensive subtractive techniques after fabrication to achieve working parts with structural integrity [16]. Nowadays, advances in computer-aided design software, particularly through topology optimization and generative designs, allow for the transformation of multiple parts into a single, more efficient component with significant weight reduction[17]. Navigating the complexities of intricate geometries presents considerable challenges; however, additive manufacturing stands out as a solution to effectively address these difficulties[18].

One of the major issues with AM-fabricated parts is directional solidification in the build direction, leading to epitaxial grain growth along the build axis[19,20,21]. Such microstructural variations significantly influence the mechanical properties, deviating from those produced through conventional processes[22,23]. Additionally, these epitaxial boundaries can become preferential sites for crack propagation[24,25]. As a result, parts manufactured through AM require post-processing treatments, such as hot-isostatic pressing and heat treatment, before they are suitable for application use [26,27]. Both DED and PBF technologies are employed in the aerospace industry to fabricate IN718 parts, with PBF being particularly popular for producing complete components from this alloy.

Electron Beam Powder Bed Fusion (EB-PBF) is a type of PBF technology that uses an electron gun to melt layers of powder sequentially [28]. In an EB-PBF machine, the entire chamber is maintained under vacuum conditions. One distinctive feature of EB-PBF is that it operates as a hot working process; the parts are fabricated inside the vacuum chamber at elevated temperatures. This approach reduces thermal stresses in the melted parts, and stress relief occurs during the cooling phase of part fabrication.[29].

The electron beam's path during the melting of powders is termed a scanning strategy. The most common scanning strategy in EB-PBF is the bi-directional raster snake melt scan, where the melt scanning pattern moves in lines, and the key parameters are spot size, speed, and power[30,31]. There is another scanning strategy that utilizes beam spots instead of lines, with key parameters including spot size, spot dwell time, and power[32,33,34].

Previous studies done with raster scan on EB-PBF for IN718 fabrication showed directional solidification with columnar grains in the build direction[35,36,37,38]. To minimize this anisotropy, some research has explored rotating the raster scan layer by layer in builds, which has been observed to minimize epitaxial growth during the build process[39,40]. Additionally, research employing an offset method to control microstructure in IN718 fabrication has demonstrated that defects in the build direction can restrict directional solidification and promote the formation of equiaxed grains[41]. Moreover, some point-based spot melt techniques showed possibilities for columnar to the equiaxed transformation of grains [21,32,42,43]. The research established that – controlling temperature gradient (G), increasing solidification rate (R), and creating restrictions at the solidification front can allow the formation of equiaxed dendritic growth in microstructure[44,45,46].

Figure 1 provides a detailed solidification map explanation with temperature gradient (G) and solidification rate (R) adapted from the welding metallurgy process [47]. The diagram showcases the effect of G and R on the solidification structure of metals. It highlights that lowering the temperature gradient and increasing the growth rate R promotes the formation of equiaxed grains over columnar grains, leading to a more significant columnar to equiaxed grain transformation (CET).

Current research explores raster and stochastic spot melt scanning strategies to build bulk IN718 parts. The primary goal is to promote grain refinement by affecting meltpool dynamics using stochastic spot melt and comparing it with raster builds. The objective is to lower the mean aspect ratio—denoted as microstructure refinement—thus minimizing anisotropy. The research begins with process optimization for both techniques for the dense build, followed by electron backscattered diffraction (EBSD) evaluation and comparing hardness values to provide an overall impact of scanning strategies in microstructure and performance.

2. Materials and Methods

2.1. Powder Characteristics

This study utilized plasma-atomized spherical powder from AP&C IN718. The powder size distribution ranged between 45 and 106µm, with corresponding values for D10, D50, and D90 being 52µm, 73µm, and 105µm, respectively. The flowability measure was 11s/50g. The powder was reused multiple times after proper sieving for the fabrication.

2.2. Fabrication Process

An open-source EB-PBF machine, FreemeltOne (Freemelt AB, Sweden), was used for solid part fabrication. The print process in FreemeltOne EB-PBF is divided into four stages: start heat, preheat, melt, and post-heat, which are executed sequentially in each cycle during printing. During its run time, the machine operates in a high vacuum in the range of 10-5 and 10-6 torr.

In each build- the build plate was heated up to 1050°C as a startheat function. Following this – two custom scan preheat strategies were employed – rectangular preheat scan for raster melt and circular preheat scan with bezier curves for spot melt. These customs preheat strategies were created using the OBPLIB function in a Python library. Before the beginning of the build – the cathode was calibrated with a beam calibration function where spot size at 1% corresponded to 0.265mm beam diameter and 150% corresponded to 1.2mm beam diameter. For preheating, a maximum spot size (150%) or defocused beam was used for part fabrication, and a 1% spot size was used for the melt to ensure proper fusion of the powders. Figure 2 represents the adopted preheat strategies for both builds. The distance from each beam path was increased to 1.2mm (termed line offset) to ensure enough distance so that the powers could retain the heat without getting overly sintered and minimize smoke events. Also, the lines or curves were shifted in a way- so that there is no overlap between each.

The melt process was explored employing two different scanning strategies—one with a bidirectional raster scan method and another with a randomized spot melt method. Figure 3 represents the scanning melt strategy implemented in this research work. The distance between two consecutive raster lines and points was kept at 0.1mm for melt. Based on powder distribution, Layer height was maintained at 75 microns.

A heuristic design approach was followed for all the DOE parameter development. This was used to develop the parameter to achieve dense builds. The density of the builds were assessed by measuring optical relative density using ImageJ software.

n the raster scanning method – the pattern implemented was bidirectional raster, as shown in Figure 3a - the build geometries evaluated were 20mm x20 mm x10 mm cuboids, as shown in the left of Figure 4a. Once a solid dense build (>99% relative optical density) was achieved with optimized parameters as shown in Figure 4a, center - 15mm x15mm x30 mm cuboids (Figure 4a, Right) were built for the verification build.

An open-source Python library, OBPLIB, created by (Freemelt AB, Sweden) was used to make the bidirectional beam pattern for raster builds. In this scanning method- the beams changed their direction like snake in each subsequent scan line.

For the spot melt method- randomized controlled point melt (stated as stochastic spot melt) was implemented using PixelMelt Cloud Software (Freemelt AB, Sweden). The build geometry for spot melt was 15mm diameter cylinders with 10mm height. The choice of geometry kept cylinders to ensure easier removal of sintered powder- and since both cuboid and cylinder were bulk parts – primary analysis showed no variation in results.

Four builds with different parameters were evaluated in one go. This randomized point melt converts the 2D print area into a small mesh (0.1mm x 0.1mm) and takes each center of the mesh as a point of melt. For each subsequent layer, the distribution of which point to choose for melt inside the mesh is randomized, as demonstrated in Figure 3b. Neighborhood cooling—an option in the software—has been turned on to minimize the remelting of the points from previous layers in the ongoing layer.

2.3. Microstructure and Hardness Characterization

The microstructure of the builds was evaluated using Electron Backscatter Diffraction (EBSD) on a Tescan Mira 3 Scanning Electron Microscope (SEM). Four different parameter sets of stochastic spot melt prints were assessed through EBSD characterization and compared with the EBSD data obtained parallel to the build direction for the raster builds. Microhardness measurements were conducted on the samples using a QATM CHD Master Microhardness tester on the HV1 scale following ASTM E384-17. The indentations are made diagonally along the plane parallel to the build direction.

3. Results & Discussion

3.1. Parameter Development

The parameter development process was split into two parts: one for the raster scan-build and one for the stochastic spot melt build.

For the raster parameter development, the energy density, Ea, was kept between 2.5 and 3.75J/mm². Variations of power and speed were experimented with during this process. Table 1 represents the successful builds from the experimentation process with respect to their optical relative density. Builds that appeared dense upon visual analysis were classified as "Solid" in Table 1, while those showing a lack of fusion (LOF) in optical image analysis were classified as "LOF." Lower speed and lower power facilitated a stable melt pool, resulting in a higher-density build.

For stochastic spot scan- a large range of E_as -Energy per spot area was explored between 5.8 to 16.9J/mm² as shown in Figure 5. It was explored to evaluate the effect of different power and dwell time in the optical relative density of parts The samples that had lack of fusion was labeled LOF, looked solid was labeled Solid, and had swelling after the build due to excessive energy was labeled as Swell. E_as in the range of 9 J/mm² to 10.5J/mm² had overall solid dense builds compared to others as shown in Figure 5. It has been observed that – the parts with identical E_as had higher density when they were built with lower power and higher spot dwell time.

3.2. Microstructure Characterization

3.2.1. Raster Scan Build

The 30mm height sample was used for the EBSD in the build direction (BD) and performed at the top and bottom sides, as demonstrated in Figure 6. Figure 6a and bottom Figure 6b represent the top and bottom in terms of the build height of the fabricated sample.

Both raster top and bottom- exhibit columnar grain formation in build direction Z due to directional solidification. The top section in Figure 6a exhibits a more uniform grain orientation, predominantly [001] observed from the IPF coloring map. The bottom section in Figure 6b also exhibits [001] orientation with greater variability of grain orientation. This variability can be caused by the initial thermal gradient and solidification dynamics at the beginning of the print.

In addition, the pole figures at the bottom of Figure 6 further illustrate the crystallographic texture in both sections. It is observed that both sections exhibit a strong {001} texture component but with differences in orientation clustering. The pole figures corresponding to the top section show concentrated intensity peaks, indicating a stronger and more uniform texture, whereas the pole figures for the bottom display slightly more dispersed peaks, reflecting the greater orientation variability.

3.2.2. Stochastic Spot Scan Build

Four individual samples from the stochastic spot builds were chosen to observe for EBSD to evaluate the impact of power and spot dwell in the microstructure of the build direction. The Eas varied in these samples between 8.25 and 10.25 J/mm2. Table 2 shows the samples that were evaluated. Samples were done EBSD—parallel to the build direction (parallel to rake movement) and perpendicular to the build direction as shown in Figure 7.

Figure 8 and Figure 9 consecutively show EBSD results perpendicular and parallel to the build direction. Upon observation- EBSD results perpendicular to the build direction (Figure 8) showed no significant microstructure variations. As the observation was from the top (XY Plane) – they exhibited equiaxed grains. In addition – it was also seen that SP1 and SP2 had larger grain sizes than SP3 and SP4 upon naked observation.

On the other hand, the EBSD result parallel to the build direction in the XZ plane (Figure 9) shows more significant variation in microstructures denoting columnar to equiaxed transformation (CET) from the SP1 to SP4 samples. In the SP1 sample, the EBSD map showed dominant elongation of grains with fewer equiaxed grains. In SP2, similar characteristics of elongated grains were observed, as some grains appeared smaller than SP1. This indicated microstructural refinement as we lowered the power from SP1 to SP2 while maintaining the same dwell time.

Furthermore, The SP3- EBSD map showed a transition to a more refined grain by being less elongated and more equiaxed. Lower power and Higher dwell time showed further grain refinement and CET introduction. Finally, In SP4 samples, further lowering the power and keeping a higher dwell time resulted in a more refined grain structure with prominent uniform grains and fewer elongated grains, resulting in a successful CET transformation.

Figure 10 and Figure 11 show pole configuration for SP1 to SP4 samples in perpendicular and parallel directions. The pole configurations perpendicular to the build direction (Figure 9) show typical pole configurations similar to equiaxed grain and were consistent across all the samples.

In SP1, pole density is concentrated and shows the early stages of CET transformation. Followed by that- in SP2 and SP3 – increased and diffused densities in the {100}, {110}, and {111} pole configurations are observed. In SP4 – the pole configuration exhibited the highest diffusion, indicating further refinement and progression to CET transformation.

In Table 2 –both SP1 and SP2 had identical dwell times of 2.21 seconds. These similar melt times allowed consistent thermal exposures and resulted in identical microstructural features. Lowering melt power- in such situation, stabled further heat distribution within the melt pool and showed increased promotion of CET in SP2. In SP3 and SP4- the dwell time was doubled to 4.42 seconds. In these cases, higher melt time has also been attributed to superior heat distribution within the melt pool, and lowering the power allowed more CET transformation in SP4 samples.

3.2.3. Comparison between Raster and Stochastic Spot Melt EBSD Results

Table 3 shows the comparative EBSD data of all the build samples parallel to the build direction (XZ). Upon observation, the raster build shows a greater grain count at the bottom (813) because of a lower thermal gradient at the beginning, which lowered at the top (425) due to an increased thermal gradient. On the other hand, from SP1 to SP4, a consistent increase in grain count was observed from 192 to 6433.1. Subsection

Observing the mean aspect ratio, the raster build had a mean aspect ratio of 3.6 at the bottom, which increased to 5.9 at the top. The area-weighted mean aspect ratio was also significantly higher for the raster at the top (10.9) compared to the bottom (5.6). For both the top and bottom- a minimum aspect ratio of almost 1.1 and a maximum aspect ratio of 36.3 indicates a larger columnar grain presence throughout the sample. A standard deviation of 5.4 at the top and 2.9 at the bottom states that a greater grain size variation was observed at the top due to thermal turbulence.

In contrast, while observing stochastic spot melt samples from SP1 to SP4, a gradual decrease in mean aspect ratio was observed from 2.9 in SP1 to 2.5 in SP4. This lowering trend suggests further refinement for CET. The minimum and maximum aspect ratio ranged from 1 to 12.7 in all the samples, suggesting fewer columnar grains than raster builds. The lower standard deviation around 1.5 for all the samples indicates a higher degree of grain refinement in spot melt samples.

3.3. Microhardness Value in HV1 Scale

Figure 12 and Table 4 show ickers microhardness values evaluated on the parallel plane in the build direction of the HV1 scale. The raster samples exhibited the highest mean hardness of 434 HV1, with a standard deviation of 11.9. This higher hardness can be attributed to the predominance of columnar grain presence in the XZ plane, as seen in Figure 6.

On the other hand, stochastic spot melt samples from SP1 to SP4 showed lower hardness values overall than raster samples. In SP1, a more significant standard deviation (72.9 HV1) was observed with a mean hardness of 298. In SP2 to SP4, we saw a gradual increase in hardness value to 408 in SP4 and a gradual decrease in standard deviation to 4.1 in SP4. This increased hardness can be attributed to a gradual increase in grain count in Table 3 and refined microstructure.

4. Conclusions

This study evaluates two scanning strategies—raster and stochastic spot melt—and their effect on the microstructure and mechanical properties of EB-PBF-fabricated IN718 superalloys. The parameter development phase identified the optimal condition for each scanning strategy for dense builds. Raster scan was observed to achieve dense build with lower power and speed at identical energy per area, whereas, for stochastic spot melt, higher dwell time and lower power were observed to produce dense parts at an energy range of 9 to 10.25J/mm2. EBSD analysis revealed a columnar grain structure- where the [100] orientation was prominent in raster samples aligned in the build direction. On the other hand- stochastic spot melt samples -showed higher distributions of {100}, {110}, and {111} oriented grains, consistent with an equiaxed grain structure. Vickers microhardness measurement further highlights these differences. Raster build demonstrated the highest hardness due to their increased columnar grains and enhanced directional strength. All stochastic samples had lower hardness values than raster build -denoting CET. From SP1 to SP3, a gradually higher hardness value and lower standard deviation were observed. SP4 showed the highest hardness value for stochastic spot melt and lower standard deviation, denoting a more isotropic grain formation with refined gains. These findings underscore the importance of scanning strategies during the fabrication process for microstructure formation and highlight the potential for microstructure refinement through controlled scanning strategies in the EB-PBF process.