Influence of Electron Beam Powder Bed Fusion Process Parameters at Constant Volumetric Energy Density on Surface Topography and Microstructural Homogeneity of a Titanium Aluminide Alloy

In powder bed fusion additive manufacturing, the volumetric energy density E V is a commonly used parameter to quantify process energy input. However, recent results question the suitability of E V as a design parameter, as varying the contributing parameters may yield different part properties. Herein, beam current, scan velocity, and line offset in electron beam powder bed fusion (PBF‐EB) of the titanium aluminide alloy TNM–B1 are systematically varied while maintaining an overall constant E V. The samples are evaluated regarding surface morphology, relative density, microstructure, hardness, and aluminum loss due to evaporation. Moreover, the specimens are subjected to two different heat treatments to obtain fully lamellar (FL) and nearly lamellar (NLγ) microstructures, respectively. With a combination of low beam currents, low‐to‐intermediate scan velocities, and low line offsets, parts with even surfaces, relative densities above 99.9%, and homogeneous microstructures are achieved. On the other hand, especially high beam currents promote the formation of surface bulges and pronounced aluminum evaporation, resulting in inhomogeneous banded microstructures after heat treatment. The results demonstrate the importance of considering the individual parameters instead of E V in process optimization for PBF‐EB.

U represents the acceleration voltage of 60 kV, I beam is the beam current, P is the beam power, and v scan is the scan velocity. [12,13] Another commonly used option to describe the process energy input is the volumetric energy density E V (Equation (2)) which additionally considers the line offset l offset as the distance between adjacent hatch lines and the layer thickness t. [18] In many cases, a higher E V was found to be associated with a reduction in part porosity. [19][20][21] A correlation between process energy input and further component properties, such as surface morphology, [22][23][24][25] microstructure, [4,26,27] or evaporation of alloying elements, [28][29][30] was also observed by several authors. It is generally accepted that insufficient energy density results in small melt pool dimensions and low melt pool temperatures, [23,31] which in turn lead to unmelted powder particles [16] and lack-of-fusion defects [13] between the individual layers. Excessive energy input, on the other hand, is well known to create melt pool instabilities and swelling. [23,31] Nevertheless, numerous recent investigations point out that the parameters E L and E V in powder bed fusion processes have only limited validity with respect to the physical processes and thermal boundary conditions in the process zone. Using the same E L or E V , but with different combinations of input parameters, can lead to significantly different outcomes in terms of melt pool geometry and component properties. Bertoli et al. [32] compared the morphology and melting conditions of single tracks generated with different combinations of laser power and scan velocity at an overall constant E L in PBF-LB. Despite maintaining a constant energy input, the authors observed considerable differences in track morphology. Prashant et al. [33] analyzed the influence of varying laser power and scan velocity at constant energy density in PBF-LB of an aluminum alloy and reported deterioration of component density and mechanical properties when laser power and scan velocity were decreased simultaneously. In a comprehensive study using in situ X-ray analysis during PBF-LB, Guo et al. [34] examined the influence of laser power and scan velocity on melt pool evolution. Even at an overall constant E V , an increase in laser power was correlated to an enlargement of melt pool dimensions and an alteration of melt pool shape. As shown by Zhao et al. [35] and Larimian et al., [36] different thermal histories resulting from varying combinations of laser power and scan velocity can affect the microstructure and crystallographic texture as well as the mechanical properties.
Regarding PBF-EB, few studies have systematically investigated the effect of different parameter combinations at constant E V . Gui et al. [23] studied the correlation between top surface morphology and varying combinations of beam current, scan velocity, and line offset at different energy inputs. The authors observed that the same value of the corresponding E V could result in different surface morphologies and concluded that E V does not characterize the energy input that is used for melting. Silvestri et al. [18] varied beam current, scan velocity, and line offset at an overall constant E V and reported varying results in terms of top surface roughness, microhardness, and microstructure, which they ascribed to variations in energy absorption and thermal diffusion for the different parameter sets. In a study on PBF-EB of Ti-47Al-8Nb, Kan et al. [37] demonstrated that differences in melt pool geometry, thermal gradients, and the velocity of the solidification front occurred as a function of beam current and scan velocity. This in turn affected the resulting microstructure. The combination of high beam powers with medium-to-high scan velocities favored epitaxial grain growth during the PBF-EB process, resulting in the formation of columnar lamellar colonies with pronounced preferential orientation along the build direction. In contrast, other parameter combinations with a similar E V produced equiaxed microstructures. Yue et al. [21] outlined that the use of different scan velocities at the same energy density leads to different beam return times, which in turn affects the thermal boundary conditions. The beam return time t R is defined as the time required for the electron beam to return to the same point in the adjacent line and can be calculated according to Equation (3) as the ratio of the scan length L and the scan velocity v scan [22] Moreover, scan velocity and line offset influence the lateral velocity v lat , which is defined as the velocity of the melt pool perpendicular to the scan direction (Equation (4)) [38] v Hence, these results strongly suggest that the prediction or control of process conditions and microstructure evolution in PBF processes is not possible solely on the basis of E L or E V . Several authors proposed a modified volumetric energy density approach which also considers beam-material interactions and material properties [39] or the use of dimensionless parameters quantifying the normalized enthalpy and the thermal penetration depth in relation to the layer thickness [40] as alternatives to characterize the process energy input. Nevertheless, the contribution of the individual process parameters to the evolution of component properties during PBF additive manufacturing has not yet been fully understood.
In PBF-EB of titanium aluminides, the choice of process parameters has a fundamental impact on the resulting microstructure. A particular challenge that occurs during processing is the evaporation of aluminum due to its comparatively high vapor pressure. The extent of aluminum evaporation largely depends on the size and temperature of the melt pool, which in turn are closely related to the selected process parameters. A correlation between the applied E L and the resulting aluminum loss was suggested by several authors. [8,22,28,30] The proposed approaches for mitigating evaporation include the reduction of the energy input at the cost of relative component density, [13] increasing the scan velocity while simultaneously decreasing the line offset to maintain a constant E V , [30] or adjusting the vacuum pressure in the build chamber. [5] Since the phase transition temperatures in titanium aluminide alloys-particularly the γ-solvus temperature-are strongly affected by the aluminum content, [1] changes in the aluminum concentration can considerably alter the microstructure formation during subsequent heat treatments. While this effect can be used on purpose to locally adapt the aluminum content by applying different process parameters in different areas of the same component and thus create functionally graded microstructures, [29,[41][42][43] pronounced evaporation may also complicate the adjustment of desired microstructures. Since evaporation occurs preferentially at the top of the melt pool, fluctuations in aluminum content can occur within the individual layers, which may lead to the formation of a banded microstructure. [7,44] During subsequent heat treatments, these local differences in aluminum content can result in inhomogeneous microstructures, [26,28,45] which could adversely affect the mechanical properties and therefore might require further homogenization treatments. [26] The aim of this study was to investigate whether certain combinations of the parameters beam current, scan velocity and line offset in PBF-EB at an overall constant E V lead to a different result regarding aluminum evaporation and microstructural evolution during heat treatment compared to other parameter combinations. Thus, it should be examined whether microstructural inhomogeneity can be avoided during the process through an adequate parameter selection.

Experimental Section
Gas-atomized, spherical titanium aluminide powder (GfE Metalle und Materialien GmbH) with a particle size distribution between 60 and 145 μm was used as powder raw material. To compensate for the expected aluminum loss during PBF-EB, the aluminum content had been increased by %3 at% in comparison to the conventional TNM-B1 alloy with the nominal composition Ti-43.5Al-4Nb-1Mo-0.1B. PBF-EB was performed using an Arcam EBM A2X machine (GE Additive, Mölnlycke, Sweden) equipped with the software EBM Control 3.2. In previous work focusing on the determination of a suitable process window for PBF-EB of TNM-B1, an E V of 32.1 J mm À3 (I beam = 15 mA, v scan = 4000 mm s À1 , l offset = 0.1 mm, t = 70 μm) was found to yield relative densities above 99.9% and even top surface morphologies. [19] The same set of parameters was chosen as the central point for the experimental design in this study. The experimental design was conceptualized to maintain a constant E V by keeping one of the three parameters beam current, scan velocity, and line offset at a fixed value while varying the other two in nine different combinations for each of the three sample series BC (constant beam current), SV (constant scan velocity), and LO (constant line offset). The corresponding parameter combinations are specified in Table 1. Accordingly, 25 cuboids with an edge length of %23 Â 23 Â 15 mm 3 were manufactured within one build job (since the parameters for the specimens BC-5, SV-5, and LO-5 are identical, only one cuboid was built with this parameter set).
A schematic illustration of the snake-like scanning strategy applied during the melting step is presented in Figure 1.
After completion of the build process, the top surfaces of all specimens were photographed with a Canon EOS 5D camera (Canon, Tokyo, Japan) for visual inspection of the surface morphology. Since pronounced differences between the individual samples were observed, top surface morphology and waviness of selected samples were analyzed with a Keyence VK-X250 laser scanning confocal microscope (Keyence Corporation, Osaka, Japan). Per sample, 61 vertically aligned equidistant measuring lines were drawn across the sample surface and evaluated using the software VK-X Series MultiFileAnalyzer (Keyence Corporation, Osaka, Japan). Based on these data, the arithmetical mean waviness Wa of the surfaces was determined. In order to evaluate only the waviness, the roughness was filtered out by applying a cutoff value of λ c = 250 μm, which was slightly above the largest line offset value of l offset = 0.200 mm.
Following the analysis of the surface morphologies, the samples were cut into several parts as shown in Figure 2. One part remained in the as-built (AS) condition, while the other two were subjected to two different heat treatments according to Table 2. Subsequently, metallographic cross sections of the xz planes of all three conditions were prepared. Cross-section images of the AS condition were taken with an Olympus GX51 inverted optical microscope (Olympus K.K., Tokyo, Japan). Based on these images, pore size and relative sample density were evaluated on three measuring fields per sample using the software ImageJ (National Institutes of Health, Bethesda, USA).
Macrohardness HV10 was measured on the polished cross sections of the AS condition in accordance with DIN EN ISO 6507-1:2018-07 [46] using a hardness testing device (KB Prüftechnik GmbH, Hochdorf-Assenheim, Germany). All hardness values were calculated as the average of ten measurements per sample. HV0.05 microhardness mappings were generated with a LECO AMH55 automatic hardness tester (LECO Corporation, St. Joseph (MI), USA). Indentations were placed at a distance of 50 μm to cover an area of 200 μm in x-direction and 600 μm along the z-direction (build-up direction).
The sample parts for heat treatment were first subjected to hot isostatic pressing (HIP) (see Table 2) as a common step to eliminate residual porosity in additively manufactured parts. [47] HIP was carried out under an argon atmosphere in a HIP200-300*450G hot isostatic press (EPSI International, Temse, Belgium) at the Fraunhofer Institute for Ceramic Technologies and Systems (Fraunhofer IKTS, Dresden, Germany). Subsequently, the specimens underwent solution heat treatment in a tube furnace under an argon atmosphere at either 1290°C or 1245°C. The solution heat treatment temperatures were chosen to adjust a fully lamellar (FL) microstructure or a nearly lamellar microstructure with globular γ-grains (NLγ), respectively. [48] In a second step, a precipitation heat treatment was carried out on the solution annealed samples in a Xerion XVAC furnace (Xerion Berlin Laboratories GmbH, Berlin, Germany) under vacuum atmosphere. Similar two-step heat treatments are well established for titanium aluminides and were described in detail by Schwaighofer et al. [49] and investigated in a previous own publication. [45] The software Thermo-Calc with the thermodynamic database TCTi4 (Thermo-Calc Software, Stockholm, Sweden) was used to calculate the phase fractions present at the selected solution heat treatment temperatures. [50] For microstructural analysis, scanning electron microscopy (SEM) was carried out on the polished metallographic cross sections of the AS, HT-1290 and HT-1245 conditions of the different samples, respectively. For image acquisition, a JEOL JSM-6610LV scanning electron microscope (JEOL Ltd., Tokyo, Japan) was utilized. Aluminum contents were determined via energy-dispersive X-Ray spectroscopy (EDX) on 25 measuring fields per sample with an approximate edge length of 250 Â 250 μm 2 using a Zeiss DSM 950 SEM (Carl Zeiss AG, Oberkochen, Germany).    www.advancedsciencenews.com www.aem-journal.com 3. Results

Surface Morphology
It is well established in the literature that different parameter combinations in PBF-EB can result in different surface morphologies. [8,17,22,23] This effect is often attributed to variances in process energy input. While too low energy densities are characterized by porous surface morphologies, excessive energy input can cause wavy surfaces or even swelling. [8,19,22] Even though all samples were manufactured with a constant E V of 32.1 J mm À3 in this study, the manufactured parts exhibited significantly different surface morphologies, as shown in Figure 3.
The samples manufactured with low beam current and low line offset at constant scan velocity (up to SV-3) or with low beam current and low scan velocity at constant line offset (up to LO-3) were found to have flat and even surfaces. With increasing beam current, surface unevenness becomes noticeable at both constant scan velocity or constant line offset, which culminates in pronounced bulges on SV-9 and LO-9. It is conspicuous that these bulges are arranged in a distinct pattern on the surfaces. At constant beam current, all samples show such features, which are most prominent on BC-1. Comparing BC-1 and BC-9, the increasing overlap of the melt lines due to the reduced line offset is evident. The melt track boundaries of sample BC-1 appear more erratic, indicating a more pronounced horizontal movement of the melt driven by surface tension and wetting phenomena on the surrounding powder particles and the already solidified material. [17] The curved melt tracks on SV-9 and LO-9 mark the presence of surface bulges. At the top of these bulges, the surface appears blistered and the individual melt tracks can no longer be easily distinguished.
To quantify the evenness of the surfaces, their arithmetical mean waviness Wa was determined based on the data obtained with laser scanning confocal microscopy. The results are presented in Figure 5 along with 3D images of the surface morphologies of the corresponding maximum and minimum parameter settings.
In agreement with the results of the visual inspection (Figure 3), the measurements show a clear correlation between the waviness and the selected parameter combination. In particular, a higher beam current seems to be associated with an increase in waviness. At a constant beam current, a slower scan velocity appears to foster a higher waviness. The distinctive localization of the material accumulation in the bulges results in the observed high standard deviation at high values for Wa.

Microstructure and Aluminum Evaporation
The relative densities and pore sizes derived from the metallographic cross sections are displayed in Figure 6.
With a line offset below or equal to 0.100 mm, relative densities above 99.9% and maximum pore diameters of less than  www.advancedsciencenews.com www.aem-journal.com 0.15 mm could be achieved. Larger line offsets resulted in a noticeable increase in porosity, indicating that in these cases there may have been an insufficient overlap between adjacent melt lines. The microstructural images obtained by SEM analysis are displayed in Figure 7. In all cases, the microstructure consists of lamellar α 2 /γ-colonies, globular γ-grains (dark grey), and β O -phase (white). However, differences between the samples in terms of size and distribution of microstructural constituents can be observed.
Samples SV-1 and LO-1 show a comparatively uniform distribution of microstructural features, whereas in all other samples more or less pronounced inhomogeneities, for example, in the form of an irregular distribution of β O -phase or large lamellar colonies can be detected.
These local variations in phase distribution across several layers are also indicated in the microhardness mappings in Figure 8. Similar micro-and nanohardness mappings have been used by several researchers to investigate microstructural homogeneity. [26,28,51] BC-9 and SV-9 show occasional softer areas, which could be caused, for example, by the prevalence of soft globular γ-grains or an overall coarser microstructure, as shown in the SEM image of SV-9 (Figure 7). In contrast, areas of higher hardness can be distinguished in LO-9, which could indicate the enhanced occurrence of the hard and brittle β O -phase. The mappings of samples BC-1, SV-1, and LO-1 exhibit comparatively uniform microhardness values within the measured area. The corresponding macrohardness values, which also can be taken from Figure 8, are all within a similar range. The lowest hardness was measured for sample SV-9, which is again in line with the coarser microstructure (Figure 7). Sample LO-9 exhibited the highest average macrohardness, which is in good agreement with the microhardness mappings. Figure 9 plots the measured aluminum contents of all samples along with the corresponding line energies E L . In addition, representative low-magnification SEM images of selected samples are presented to illustrate the effect of aluminum distribution on the microstructure.
Although no statistically significant conclusions about differences in total aluminum content can be derived from the EDX measurements, the data suggest a correlation between the applied process parameters and the variation of the aluminum content as a consequence of evaporation, especially with increasing beam current (Figure 9b,c). Accordingly, the microstructural images of SV-9 and LO-9 reveal a pronounced banded microstructure. As aluminum tends to evaporate close to the top of the melt pool, the upper part of the melt track becomes depleted in aluminum. In the metallographic cross section, the Al-lean regions, therefore, appear brighter than the Al-rich bottom of the melt tracks, which has a higher proportion of γ-phase that appears in a darker shade. The most pronounced evaporation effects were found in the areas beneath the surface bulges of SV-9 and LO-9, which is the reason why the layers in the corresponding SEM images are tilted with respect to the build-up direction. The specimens manufactured with the lowest beam current (SV-1 and LO-1) show significantly less pronounced fluctuations in the aluminum content. Judging from the SEM images of the samples manufactured with constant beam current (Figure 9a), the combination of slow scan velocities and high line offsets (BC-1) in particular seems to provoke a nonuniform aluminum distribution, even though no clear trend can be identified from the EDX data of this sample series. Interestingly, the results reveal no correlation between E L and aluminum loss. This is particularly evident from the significant www.advancedsciencenews.com www.aem-journal.com differences in aluminum evaporation at constant line offset despite the constant E L (Figure 9c).

Heat Treatment
Using the software Thermo-Calc, the expected equilibrium phase fractions at the solution heat treatment temperatures of 1290 and 1245°C were calculated as a function of the aluminum content, respectively. The results are plotted in Figure 10.
To obtain an FL microstructure with particularly favorable creep properties, solution annealing has to be carried out in the single α-phase field. [48,52,53] Figure 10a demonstrates that at 1290°C, this single α-phase field only exists within a very narrow range of the aluminum content. At higher aluminum contents, γ-phase occurs in addition to α-phase, whereas α-phase and β-phase are in equilibrium at aluminum contents below %44.5 at%. At 1245°C, the relevant aluminum contents yield a heat treatment within the (α þ β/β O þ γ)-region, but the ratio of the phase fractions strongly depends on the aluminum content. Consequently, local fluctuations in the aluminum content caused by pronounced evaporation can lead to undesirable microstructures as a result of heat treatment. This is also evident from the microstructural images of the different samples after HT-1290 and HT-1245 presented in Figure 11.
In the case of SV-1 and LO-1, homogeneous FL microstructures which solely consist of lamellar α 2 /γ-colonies could be obtained with HT-1290. In all other samples, β O -and/or γ-grains can be found in addition to the lamellar colonies. This implies that the variance of the aluminum content in these samples was too large to facilitate heat treatment in the single α-phase field. The SEM images after HT-1245 reveal a similar result. A mostly homogeneous NLγ microstructure composed of lamellar α 2 /γ-colonies and globular γ-grains, which is known for improved ductility, [53] could be adjusted in the samples SV-1 and LO-1. Again, all other investigated samples show inhomogeneous microstructures due to an irregular distribution of the γ-phase. Especially in the case of SV-9 and LO-9, a periodic pattern of the microstructure is noticeable, which is in good agreement with the layerwise manufacturing and the preferential aluminum evaporation at the top of each layer.

Discussion
Several studies have demonstrated that different combinations of beam power, scan velocity, and distance between adjacent hatch  www.advancedsciencenews.com www.aem-journal.com lines in powder bed fusion processes at an overall constant energy input level can lead to significantly different outcomes in terms of surface morphology, [18,21,23,54] microstructure, [35,54] and mechanical properties. [18,35,36,54] Most of these investigations focussed on PBF-LB, [32][33][34][35][36]54,55] while only few addressed the influence of different parameter combinations at constant energy density in PBF-EB [18,21] . In this study, the effect of varying beam current, scan velocity, and line offset at constant E V on the resulting surface topography, microstructure, and aluminum evaporation in PBF-EB of TNM-B1 was systematically investigated. The results demonstrate that flat surfaces and homogeneous microstructures can be achieved with certain parameter combinations, while others lead to surface bulges and pronounced aluminum evaporation. Generally, an increase in beam current at constant scan velocity or constant line offset led to the development of uneven surfaces, more pronounced aluminum evaporation, and the occurrence of inhomogeneous microstructures after heat treatment. At a constant beam current, inhomogeneities were found both for high line offsets and slow scan velocities and for low line offsets and high scan velocities, although the different combinations led to different microstructures and surface qualities. In all cases, line offsets below 0.100 mm seemed to be essential for attaining high relative densities and small pore sizes. In order to understand the observed phenomena, it is essential to take a closer look at the influence of the process parameters on the melting conditions in PBF additive manufacturing. The interaction of beam power, scan velocity, and line offset can affect the melt pool dimensions, [32,34,55,56] melt pool dynamics, [23] and thermal boundary conditions [34,35,38,57] of the process.
An increase in melt pool depths and lengths at higher laser powers was documented by Bertoli et al. [32] in single-track experiments at constant energy density in PBF-LB. Guo et al. [34] used in situ X-ray characterization to examine the evolution of melt pool dimensions at different parameter sets in PBF-LB. When beam power and scan velocity were increased at the same time while maintaining a constant energy input, the length, depth, and width of the melt pool were increased. Hence, the total melt pool volume also increased. This effect was confirmed in a study by Yuze et al. [55] At the same time, the increase in the length of the melt pool along the scan direction was found to be more pronounced than the increase in width. [34] Simulation results by Zäh and Lutzmann suggest a similar preferential dimension development with increasing beam current and scan velocity for the PBF-EB process. At constant beam currents, they reported a significant increase in the length-to-width ratio with increasing scan velocity, while a variation of the beam current at constant scan velocities only had a minor effect on that ratio. [58] Melt pool depth in PBF-EB was reported to be enlarged with increasing E L . [59] Regarding the melt pool dimensions in this study, it can be assumed that the largest melt pool volumes of the individual sample series were generated during the fabrication of BC-1 due to the combination of intermediate beam currents and slow scan velocities (i.e., high E L ) as well as of SV-9 and LO-9 due to the high beam currents. It also seems likely that the melt pools of BC-9 and LO-9 exhibited the highest length-to-width ratios within their sample series as a result of the high scan velocities, while the length-to-width ratio might have been largely unaffected by the variation of beam current and line offset at constant scan velocity.
Both the laser beam in PBF-LB and the electron beam in PBF-EB are commonly assumed as Gaussian-shaped moving heat sources. [60] The peak temperature T peak of a material that is heated with a Gaussian distributed heat source is correlated to the beam power P and the scanning velocity v scan according to Equation (5) T peak % P ffiffiffiffiffiffiffiffiffi v scan p (5) www.advancedsciencenews.com www.aem-journal.com It can therefore be assumed that the beam power has a stronger impact on the peak temperature in the melt pool than the scanning velocity. [32] Nevertheless, the local interaction time between electron beam and powder bed and the beam return time are reduced at high scan velocities. Hence, thermal diffusion time decreases, and more heat accumulates in the material, which is available during melting of the subsequent scan lines. [18,22,55,56] Consequently, the simultaneous increase of www.advancedsciencenews.com www.aem-journal.com beam power and scan velocity at constant energy input may lead to an increase in the occurring melt pool temperatures [18,32,55] and cooling rates. [35] If the beam current is kept constant, melt pool temperature considerably increases with decreasing scan velocity (i.e., increasing E L ). [30] Thus, it could be concluded that the highest peak temperatures within the sample series at constant scan velocity and constant line offset in this study occurred during manufacturing of SV-9 and LO-9, respectively. At a constant beam current (BC series), the high local heat input at slow scan velocities and the heat accumulation due to reduced thermal losses at high scan velocities might be competing factors in terms of melt pool temperature, although the results obtained by Klassen et al. [30] provide evidence that the highest peak temperatures might have occurred in BC-1. High melt pool temperatures and thus high thermal gradients are known to amplify the Marangoni convection, which is caused by the temperature dependence of surface tension. Moreover, the vapor recoil pressure, which is induced by evaporation effects, is related to the melt pool temperature. Marangoni convection and vapor recoil pressure are the main driving forces for melt pool flow in PBF-EB, leading to a vigorous and high-velocity liquid flow behavior in the case of high peak temperatures and large thermal gradients. [23,31] As a result of the associated material transport, bumps can be created at the surface, which can accumulate over the course of several layers and finally lead to swelling. [23] This supports the results obtained in this study, where the samples with presumably higher melt pool temperatures were found to have more irregular surface morphologies (see Figure 3 and 5). The increased waviness of BC-1, SV-9, and LO-9 could therefore have been provoked by enhanced Marangoni convection and higher-vapor recoil pressures. Depending on the process parameters and the scanning strategy, different degrees of remelting and varying thermal boundary conditions may occur locally within each layer, which could be partly responsible for the emergence of bulges in distinct patterns. In a recent research paper, Breuning et al. [57] proposed an additional explanation for the occurrence of pronounced bulges in certain areas of the surface: At high beam currents and scan velocities, the melt pool regime can change from a lens-shaped trailing melt pool to a persistent melt pool, that is, the melt pool is still active when the beam returns to the same position in the adjacent scan line. [56,57,61] If, for example, a rectangular area is melted with a snake-like scan strategy, a persistent melt pool regime may occur near the turning points, whereas the  www.advancedsciencenews.com www.aem-journal.com melt in the center of that area is already completely solidified within the beam return time. Breuning et al. [57] called this melt pool regime "partially persistent." The occurrence and the extent of such persistent domains largely depend on the applied parameters, which determine the beam return time and the lateral velocity perpendicular to the scan direction (see Figure 1). Since the material transport governed by melt pool dynamics can significantly differ for trailing and persistent melt pools, material accumulations can form at the transition between the two melt pool regimes. As the scan strategy is commonly rotated by 90°with each layer, these accumulations can be amplified by superposition and create a distinct pattern of bumps and bulges on the top surface. Some of the top surfaces in Figure 3 show such rotationally symmetric patterns, which might at least partly be explained by this effect.
The fluctuations in aluminum content due to evaporation were found to be particularly enhanced in the vicinity of surface bulges, which might be the consequence of longer melt pool lifetimes and larger lateral melt pool dimensions in the persistent areas. [56] This coincidence of distinct surface features and increased evaporation implies that surface morphology may be used as one indicator to assess the homogeneity of the component. Renner et al. [24] presented a system for in situ melt surface monitoring during PBF-EB based on electron optical imaging, which is capable of capturing the evolution of surface bulges and could prospectively assist in the early detection of swelling and pronounced evaporation. Gui et al. [23] introduced different criteria for surface flatness to identify internal defects and accelerate process parameter optimizations using different machine learning technologies, which might also offer the potential to account for evaporation.
Another potentially important factor in PBF-EB, which has rarely been considered in the literature, is the increase in beam diameter with increasing beam currents. In a recent study, Reith et al. [38] demonstrated the influence of the beam diameter on melt pool dimensions, microstructure, and aluminum evaporation in PBF-EB of a titanium aluminide alloy. At constant E L , the authors reported an increase in aluminum evaporation with increasing beam power up to a certain value, while evaporation decreased again when the beam power was increased further. They attributed this observation to the power-related increase in beam diameter and the associated reduction in beam power density and melt pool temperature. Klassen et al. [30] found the following relationship between beam power P and beam diameter d beam (Equation (6)) The corresponding beam power density q A can be calculated according to Equation (7) q A ¼ ηÃP Reith et al. [38] assumed a value of 0.85 for the absorption coefficient η. Using Equation (3), (4), (6), and (7), the beam return time t R , the lateral velocity v lat , the beam diameter d beam and the beam power density q A for the parameter combinations investigated in this study were calculated (see Figure 12). It should be noted that the beam diameter may strongly depend on the PBF-EB machine and the applied focus offset. Hence, the values for d beam and q A presented here can only provide a rough estimation of the actual conditions and should rather be regarded as a qualitative comparison between the individual parameter sets. It is evident from Figure 12 that the thermal boundary conditions during manufacturing were unique for each sample as a result of the different parameter combinations. While all samples were subjected to the same energy input by keeping E V constant, the differences in beam power density and the interaction time between the electron beam and the material may have altered energy dissipation. [34] Consequently, the specimens underwent different solidification and cooling rates, which are reflected in the different microstructures in Figure 7.
The calculations were performed under the assumption that the absorption coefficient η remains constant throughout the entire process parameter field. However, several studies outlined that energy absorption can vary significantly depending on the process parameters. In PBF-LB, absorption is known to increase for higher laser powers, particularly when the melting regime transitions from conduction mode to keyhole mode. [34,62] Variations in energy absorption might also occur in PBF-EB as a function of the beam power because the backscattering coefficient is related to beam intensity. [18] An influencing factor on the thermal boundary conditions which was not considered in this study is the scan length L (see Figure 1), which was identical for all samples investigated. Especially in complex geometries commonly encountered in additive manufacturing, scan lengths can vary both between layers and within a cross-sectional area to be melted. With otherwise constant beam parameters, this results in local differences with regard to beam return time, lateral velocity, heat accumulation, melt pool regime, etc. [56] To ensure homogeneous part properties throughout complex components, advanced scanning strategies which optimally take all of these factors into account might therefore be required.
Considerable differences regarding aluminum evaporation were identified for the different specimens (see Figure 9). In the literature, it is often claimed that the aluminum loss is directly correlated to the line energy E L . [8,22,28,30] However, this correlation could not be confirmed in this study, since the samples in the LO series (constant line offset) show different aluminum distributions despite the constant E L , suggesting a more complex relation between process parameters and evaporation. Studies have shown that evaporation is increased with higher melt pool temperatures, longer melt pool lifetimes, and larger melt pool volumes, [22,30] although the surface-to-volume ratio of the melt pool could possibly be more decisive than the actual melt pool size. [16,31] As explained earlier, it is reasonable to assume that samples manufactured with low beam currents, such as SV-1 and LO-1, exhibited lower melt pool sizes and temperatures during manufacturing than the ones produced at intermediate-to-high beam currents (e.g., BC-1, SV-9 and LO-9). Moreover, it could be presumed that evaporation might be enhanced in areas with partially persistent melt pools, which might occur at high scan velocities and high lateral velocities, due to the prolonged melt pool lifetime and the enlarged melt pool surface-to-volume ratio. [56] These assumptions are in good agreement with the observed aluminum evaporation.
During PBF-EB, evaporation primarily is located at the surface of the melt pool, which becomes depleted in aluminum. [22,30,63] Even though mixing with aluminum-rich material from the bottom of the melt pool takes place, [63] this effect can lead to banded microstructures in the final component, [26,28,44] as shown in the cross-section images in Figure 9. This can severely impact www.advancedsciencenews.com www.aem-journal.com subsequent heat treatments, especially if they are carried out close to the γ-solvus temperature. [28] The results of the thermodynamic calculations with Thermo-Calc shown in Figure 10 provide evidence that even small variations in the aluminum content can have major effects on the phase transformation temperatures and the phase fractions during heat treatments. Accordingly, the samples with pronounced local fluctuations in the aluminum content (i.e., BC-1, BC-9, LO-9, and SV-9) were particularly susceptible to developing inhomogeneous microstructures during heat treatment (Figure 11), which might deteriorate the mechanical properties and should therefore be avoided. [8] In agreement with results presented by Seifi et al., [64] HIP preceding solution annealing did not homogenize the aluminum distribution. To counteract microstructural inhomogeneity in PBF-EB-manufactured Ti-48Al-2Cr-2Nb, Wartbichler et al. [26] applied a complex homogenization treatment consisting of a 1 h dwell time above the γ-solvus temperature followed by oil quenching and a subsequent annealing step at 1100°C for 24 h. Thus, the aluminum distribution could be sufficiently homogenized to obtain the desired microstructures after the actual heat treatment. However, this homogenization annealing is energy intensive and costly, so avoiding inhomogeneities already during the PBF-EB process is highly relevant. The results shown in Figure 9 and 11 demonstrate the possibility of reducing fluctuations in the aluminum content at the process level by a careful selection of the process parameters, thus facilitating homogeneous microstructures and eliminating the necessity of additional annealing. At the applied energy density level of E V = 32.1 J mm À3 , a low beam current was particularly favorable for achieving even top surfaces and uniform aluminum distributions. Presumably, this is linked to lower melt pool temperatures and smaller melt pool dimensions occurring at low beam powers. In combination with low line offsets, TNM-B1 parts with relative densities above 99.9% could be obtained. Overall, the results of this study support the fact that although E L and E V might be useful tools for narrowing down a parameter field for PBF processing of new materials, the contributing parameters such as beam current, scan velocity, or line offset should be optimized individually for attaining the best possible part quality. Future research should further enhance the understanding of the complex interrelation between these parameters and the phenomena in the process zone, for example, melt pool dynamics and evaporation effects.

Conclusion
This study investigated the influence of different combinations of beam current, scan velocity, and line offset in PBF-EB of TNM-B1 on surface morphology and microstructural heterogeneity in different heat treatment conditions. Despite a constant volumetric energy density for all parameter sets, the results show considerable differences. The following conclusions can be drawn from the experiments: 1) The overall best results were obtained with low beam currents, line offset values below 0.1 mm, and low-to-intermediate scan velocities. Specimens fabricated under these conditions exhibited even surfaces, low porosity, fine and isotropic microstructures, and uniform aluminum distribution, which allowed the adjustment of homogeneous FL and NLγ microstructures through appropriate heat treatments; 2) The beam current seems to be the most dominant factor in terms of swelling and aluminum evaporation, since high beam currents invariably led to surface bulges and pronounced fluctuations in the aluminum content, which in turn resulted in anisotropic banded microstructures after heat treatment. This was also the case when beam current and scan velocity were increased simultaneously at constant line energy E L ; 3) The presence of surface bulges and an increase in waviness coincided with increased aluminum evaporation, suggesting that top surface quality could be used as an indicator for microstructural homogeneity; 4) The observed variations between the samples indicate that the individual parameter combinations differ in terms of energy absorption and energy dissipation, which may be linked to differences in melt pool size, melt pool dynamics, and peak temperature; 5) While line energy E L and volumetric energy density E V can be useful tools for reducing the degrees of freedom in process parameter optimization for new materials, they fail to account for the complex interactions in the process zone encountered in powder bed-based additive manufacturing. Therefore, careful adjustment of the individual parameters such as beam power or scanning velocity is recommended, especially for materials that are susceptible to evaporation or microstructural variations.