Performance of High-Layer-Thickness Ti6Al4V Fabricated by Electron Beam Powder Bed Fusion under Different Accelerating Voltage Values

The electron beam powder bed fusion (EB-PBF) process is typically carried out using a layer thickness between 50 and 100 μm with the accelerating voltage of 60 kV for the electron beam. This configuration ensures forming accuracy but limits building efficiency. The augmentation of the accelerating voltage enlarges the molten pool due to the rise in penetrability, suggesting that a higher layer thickness can be used. Therefore, the effects of layer thickness and accelerating voltage were investigated simultaneously in this study to explore the feasibility of efficiency improvement. Ti6Al4V was fabricated by EB-PBF using layer thicknesses of 200 and 300 μm. Two accelerating voltage values of 60 and 90 kV were used to study their effects under expanded layer thickness. The results reveal that dense parts with the ultimate tensile strength higher than 950 MPa and elongation higher than 9.5% could be fabricated even if the layer thickness reached 300 μm, resulting in a building rate of up to 30 mm3/s. The expansion of the layer thickness could decrease the minimum bulk energy density needed to fabricate dense parts and increase the α platelet thickness, which improved the energy efficiency. However, expanding layer thickness had a significant negative effect on surface roughness, but it could be improved by applying augmented accelerating voltage.


Introduction
Electron beam powder bed fusion (EB-PBF) is a kind of additive manufacturing technology which uses electron beams to selectively melt the powder layer by layer based on three-dimensional (3D) digital models and form 3D parts [1,2]. Compared with subtractive manufacturing technologies, parts with low thermal stress and arbitrary complex structures can be fabricated by EB-PBF due to the high powder-bed temperature and support effect provided by the sintered powder bed during the manufacturing process. The recycling of the sintered powder greatly increases the material utilization rate. In recent years, metal parts with high mechanical properties have been successfully fabricated by EB-PBF using different materials [3][4][5][6][7][8][9], which demonstrates promising prospects in aerospace, medicine and other fields.
As a powder bed fusion process, layer thickness plays a key role in EB-PBF. Generally, the typical layer thickness for the EB-PBF process varies between 50 and 100 µm [10,11]. This is larger than the typical layer thickness between 20 and 50 µm used in laser powder bed fusion (L-PBF) which uses a laser as the heat source to melt the metal powder. A smaller layer thickness results in higher dimensional accuracy and surface roughness, but a lower

Equipment and Manufacturing Process
The tests were carried out using EB-PBF-250 equipment developed by Tsinghua University [24,25]. The accelerating voltage of this equipment was variable between 60 and 90 kV. A Ti6Al4V alloy sheet with a thickness of 10 mm was used as the substrate. The substrate was preheated to 700 C before the powder spreading process and the powder bed was preheated to 700 C before the melting process. Building samples with a crosssection of 20 × 20 mm 2 and a height of 8 mm were fabricated under a constant pressure of 0.1 Pa with argon backfilling.
An "S-mode" strategy was adopted for the melting process, as depicted in Figure 2a. The scanning direction between adjacent layers was rotated by 90° and the scanning direction between adjacent scanning lines was rotated by 180°. The bulk energy density E (J/mm 3 ), referred to the energy obtained in unit volume by material during the electron beam melting, was used to characterize the heat input. It is calculated by where U denotes the accelerating voltage (kV), I is the electron beam current (mA), h denotes the hatching space between adjacent scanning lines (mm), v denotes the scan speed (mm/s) and L is the layer thickness (μm). In this experiment, the scanning speed was fixed at 500 mm/s and the hatching space was 0.2 mm. Two layer thicknesses, 200 and 300 μm, were used to investigate the effect of high layer thickness on the forming quality of the as-built samples. Besides the accelerating voltage of 60 kV, 90 kV was also used to evaluate the feasibility of forming specimens with high layer thickness using high accelerating voltage. In all experiments, the electron beam was focused to the smallest spot diameter that the equipment could achieve (approximately 400 μm at 60 kV and approximately 300 μm at 90 kV) during the melting process, which could suppress the keyhole and splashing effects because of the larger spot diameter than the spot diameter below 100 μm in L-PBF [19]. The processing parameters for all samples are listed in Table 2; three different bulk energy densities were used and the beam current was determined according to Equation (1).  An "S-mode" strategy was adopted for the melting process, as depicted in Figure 2a. The scanning direction between adjacent layers was rotated by 90 • and the scanning direction between adjacent scanning lines was rotated by 180 • . The bulk energy density E (J/mm 3 ), referred to the energy obtained in unit volume by material during the electron beam melting, was used to characterize the heat input. It is calculated by where U denotes the accelerating voltage (kV), I is the electron beam current (mA), h denotes the hatching space between adjacent scanning lines (mm), v denotes the scan speed (mm/s) and L is the layer thickness (µm). In this experiment, the scanning speed was fixed at 500 mm/s and the hatching space was 0.2 mm. Two layer thicknesses, 200 and 300 µm, were used to investigate the effect of high layer thickness on the forming quality of the as-built samples. Besides the accelerating voltage of 60 kV, 90 kV was also used to evaluate the feasibility of forming specimens with high layer thickness using high accelerating voltage. In all experiments, the electron beam was focused to the smallest spot diameter that the equipment could achieve (approximately 400 µm at 60 kV and approximately 300 µm at 90 kV) during the melting process, which could suppress the keyhole and splashing effects because of the larger spot diameter than the spot diameter below 100 µm in L-PBF [19]. The processing parameters for all samples are listed in Table 2; three different bulk energy densities were used and the beam current was determined according to Equation (1).

Characterization
The sample densities were measured using the Archimedes principle with equipment ADVENTURER, AR423DCN. The nominal density of Ti6Al4V was calculated as 4.42 g/cm 3 [26]. The surface roughness measurements were carried out using a 3D confocal microscope (Phase Shift MicroXAM-3D). The arithmetic average value (Ra) was used for the surface roughness analyses. During the surface roughness measurements, the measured direction for the lateral surface was parallel to the build direction and the measured direction for the upper surface was perpendicular to the scanning direction.

Characterization
The sample densities were measured using the Archimedes principle with equipment ADVENTURER, AR423DCN. The nominal density of Ti6Al4V was calculated as 4.42 g/cm 3 [26]. The surface roughness measurements were carried out using a 3D confocal microscope (Phase Shift MicroXAM-3D). The arithmetic average value (Ra) was used for the surface roughness analyses. During the surface roughness measurements, the measured direction for the lateral surface was parallel to the build direction and the measured direction for the upper surface was perpendicular to the scanning direction.
The surface morphologies and microstructures were observed using a Sigma 300 field emission scan electron microscope (FESEM), which operated at 5 kV. The samples were polished and then etched for 20 s with Kroll's reagent (vol. 1% HNO3, vol. 2% HF, vol. 97% H2O) to better reveal the microstructure. Tensile samples, shown in Figure 2b, were tested using a Zwick Z2.5 TH tensile test machine at 25 C. The strain rate was 0.5 × 10 −3 /s and the test direction was perpendicular to the building direction. The dimensions of the tensile samples are depicted in Figure 2c   The surface morphologies and microstructures were observed using a Sigma 300 field emission scan electron microscope (FESEM), which operated at 5 kV. The samples were polished and then etched for 20 s with Kroll's reagent (vol. 1% HNO 3 , vol. 2% HF, vol. 97% H 2 O) to better reveal the microstructure. Tensile samples, shown in Figure 2b, were tested using a Zwick Z2.5 TH tensile test machine at 25 • C. The strain rate was 0.5 × 10 −3 /s and the test direction was perpendicular to the building direction. The dimensions of the tensile samples are depicted in Figure 2c.

Relative Densities and Surface Morphologies
The density results for the as-built samples are listed in Table 3 and plotted in Figure 3. The dense parts were defined with a density greater than 99.5%. It can be observed that dense parts could be achieved at the bulk energy density of 25 J/mm 3 for both layer thicknesses and accelerating voltages. Furthermore, when the layer thickness expanded from 200 µm to 300 µm, the minimum bulk energy density needed to fabricate dense parts reduced from 25 J/mm 3 to 20 J/mm 3 at the accelerating voltage of 60 kV and from 20 J/mm 3 to 15 J/mm 3 at the accelerating voltage of 90 kV. 3. The dense parts were defined with a density greater than 99.5 %. It can be observed that dense parts could be achieved at the bulk energy density of 25 J/mm 3 for both layer thicknesses and accelerating voltages. Furthermore, when the layer thickness expanded from 200 μm to 300 μm, the minimum bulk energy density needed to fabricate dense parts reduced from 25 J/mm 3 to 20 J/mm 3 at the accelerating voltage of 60 kV and from 20 J/mm 3 to 15 J/mm 3 at the accelerating voltage of 90 kV.    Figures 6 and 7), respectively. The surface roughness values of those samples are plotted in Figure 8.
As shown in Figures 4 and 6, the single-line scanning tracks could be observed, which caused the surface fluctuations and resulted in roughness (Ra) values between 5 and 10 µm of the upper surface. No significant differences could be observed for the four upper surfaces fabricated using different parameters. However, the surface roughness of the upper surface at higher layer thickness was slightly increased and decreased by augmenting the accelerating voltage from 60 to 90 kV, as shown in Figure 8a. Compared to the upper surface, the lateral surface was more uneven and had a greater surface roughness. The semi-elliptical cross-section of the molten pool, combining the step effect caused by the layer thickness, resulted in a structure with alternating peaks and valleys perpendicular to the building direction, as shown in Figures 5 and 7. Furthermore, the fluctuation of the molten pool and the incompletely melted powder adhered to the lateral surface could also exaggerate the lateral surface roughness. These effects resulted in lateral surface roughness values between 39 and 60 µm. Moreover, the surface roughness of the lateral surface at higher layer thickness was also increased and decreased by augmenting the accelerating voltage from 60 to 90 kV, as shown in Figure 8b, which shows the same trend as the upper surface.
With the expansion of layer thickness, the stochastic effect of the powder bed and the step effect caused by the melting tracks intensify, which exaggerates the surface roughness. The augmentation of accelerating voltage enlarges the melting depth, which improves the stability of the melt pool and reduces the track fluctuation [27]. Therefore, augmenting the accelerating voltage improves the surface roughness.           As shown in Figures 4 and 6, the single-line scanning tracks could be observed, which caused the surface fluctuations and resulted in roughness (Ra) values between 5 and 10 μm of the upper surface. No significant differences could be observed for the four upper surfaces fabricated using different parameters. However, the surface roughness of the upper surface at higher layer thickness was slightly increased and decreased by augmenting the accelerating voltage from 60 to 90 kV, as shown in Figure 8a. Compared to the upper surface, the lateral surface was more uneven and had a greater surface roughness. The semi-elliptical cross-section of the molten pool, combining the step effect caused by the layer thickness, resulted in a structure with alternating peaks and valleys perpendicular to the building direction, as shown in Figures 5 and 7. Furthermore, the fluctuation of the molten pool and the incompletely melted powder adhered to the lateral surface could also exaggerate the lateral surface roughness. These effects resulted in lateral surface roughness values between 39 and 60 μm. Moreover, the surface roughness of the lateral surface at higher layer thickness was also increased and decreased by augmenting the accelerating voltage from 60 to 90 kV, as shown in Figure 8b, which shows the same trend as the upper surface.
With the expansion of layer thickness, the stochastic effect of the powder bed and the step effect caused by the melting tracks intensify, which exaggerates the surface roughness. The augmentation of accelerating voltage enlarges the melting depth, which improves the stability of the melt pool and reduces the track fluctuation [27]. Therefore, augmenting the accelerating voltage improves the surface roughness.
The reduction in the minimum bulk energy density required for dense part fabrication due to the escalation of layer thickness can be reflected by the upper surface morphologies. Figure 9 shows the schematic diagram of the melting tracks during the PBF process. The melting depth needed to achieve a dense part was much higher than the nominal layer thickness, which was caused by the stochastic effect that the effect layer thickness varies between zero and several times the nominal layer thickness [28]. Therefore, the section of the melting tracks could be divided into the remelting area and the new melting area and the remelting area should be significantly larger than the new melting area because the molten width and depth are significantly larger than the hatching spacing and layer thickness [22], as shown in Figure 9. When the layer thickness expanded from 200 μm to 300 μm, while the hatching spacing was maintained at 0.2 mm, the new melting area should have been extended by 50%. However, as shown in Figure 8a, the upper surface roughness (Ra) of the dense samples under different layer thicknesses showed a small difference, which indicates that dense parts could be obtained with only a small increase in the height of the overlapping area when the layer thickness expands and the increase in the remelting area should be less than 50%. These facts mean that, when the layer thickness is expanded by 50%, the increase in minimum line energy density (P/v) needed to achieve a dense part should be less than 50%, which results in a decrease in bulk energy The reduction in the minimum bulk energy density required for dense part fabrication due to the escalation of layer thickness can be reflected by the upper surface morphologies. Figure 9 shows the schematic diagram of the melting tracks during the PBF process. The melting depth needed to achieve a dense part was much higher than the nominal layer thickness, which was caused by the stochastic effect that the effect layer thickness varies between zero and several times the nominal layer thickness [28]. Therefore, the section of the melting tracks could be divided into the remelting area and the new melting area and the remelting area should be significantly larger than the new melting area because the molten width and depth are significantly larger than the hatching spacing and layer thickness [22], as shown in Figure 9. When the layer thickness expanded from 200 µm to 300 µm, while the hatching spacing was maintained at 0.2 mm, the new melting area should have been extended by 50%. However, as shown in Figure 8a, the upper surface roughness (Ra) of the dense samples under different layer thicknesses showed a small difference, which indicates that dense parts could be obtained with only a small increase in the height of the overlapping area when the layer thickness expands and the increase in the remelting area should be less than 50%. These facts mean that, when the layer thickness is expanded by 50%, the increase in minimum line energy density (P/v) needed to achieve a dense part should be less than 50%, which results in a decrease in bulk energy density, as shown in Figure 3. In addition, the augmentation of the accelerating voltage enlarges the melting depth and results in a decrease in the bulk energy density to achieve dense parts [22]. density, as shown in Figure 3. In addition, the augmentation of the accelerating voltage enlarges the melting depth and results in a decrease in the bulk energy density to achieve dense parts [22].

Microstructure
Figures 10 and 11 demonstrate the microstructures of the as-built samples for the accelerating voltage values of 60 and 90 kV at different layer thicknesses and bulk energy densities, respectively. It can be seen that the microstructure of all the as-built samples consisted of the α and β phases with a Widmanstӓtten-like structure; the gray α phase was surrounded by the white β phase, which showed no differences in phase compared with the microstructure obtained by low layer thickness [10].  consisted of the α and β phases with a Widmanstätten-like structure; the gray α phase was surrounded by the white β phase, which showed no differences in phase compared with the microstructure obtained by low layer thickness [10]. Figure 9. Schematic diagram of the melting tracks during the PBF process. Figures 10 and 11 demonstrate the microstructures of the as-built samples for the accelerating voltage values of 60 and 90 kV at different layer thicknesses and bulk energy densities, respectively. It can be seen that the microstructure of all the as-built samples consisted of the α and β phases with a Widmanstӓtten-like structure; the gray α phase was surrounded by the white β phase, which showed no differences in phase compared with the microstructure obtained by low layer thickness [10].  The difference in the microstructure of the as-built samples could be characterized by the change in the α platelet thickness, which is plotted in Figure 12. The α platelet thickness is determined by the cooling rate. When the cooling rate decreases, the α platelet has more time to grow up and coarsen [29]. Therefore, with the increase in the bulk energy density or the expansion of the layer thickness under the same bulk energy density, the line energy density (UI/v, defined as the beam power divided by scanning speed) increases and a larger molten pool can be obtained, which results in a lower cooling rate and a thicker α platelet, as shown in Figure 12. In addition, the α platelet thickness also increases with the augmentation of the accelerating voltage, which is caused by the higher energy efficiency and penetrability of the higher accelerating voltage [22]. The difference in the microstructure of the as-built samples could be characterized by the change in the α platelet thickness, which is plotted in Figure 12. The α platelet thickness is determined by the cooling rate. When the cooling rate decreases, the α platelet has more time to grow up and coarsen [29]. Therefore, with the increase in the bulk energy density or the expansion of the layer thickness under the same bulk energy density, the line energy density (UI/v, defined as the beam power divided by scanning speed) increases and a larger molten pool can be obtained, which results in a lower cooling rate and a thicker α platelet, as shown in Figure 12. In addition, the α platelet thickness also increases with the augmentation of the accelerating voltage, which is caused by the higher energy efficiency and penetrability of the higher accelerating voltage [22]. density or the expansion of the layer thickness under the same bulk energy density, the line energy density (UI/v, defined as the beam power divided by scanning speed) increases and a larger molten pool can be obtained, which results in a lower cooling rate and a thicker α platelet, as shown in Figure 12. In addition, the α platelet thickness also increases with the augmentation of the accelerating voltage, which is caused by the higher energy efficiency and penetrability of the higher accelerating voltage [22].

Tensile Properties
The mechanical properties of the as-built samples are listed in Table 4. When the relative density was higher than 99.5%, an ultimate tensile strength higher than 950 MPa and elongation higher than 9.5% could be achieved, which agrees with the results obtained by

Tensile Properties
The mechanical properties of the as-built samples are listed in Table 4. When the relative density was higher than 99.5%, an ultimate tensile strength higher than 950 MPa and elongation higher than 9.5% could be achieved, which agrees with the results obtained by EB-PBF process using layer thickness between 50 and 100 µm [11,22]. Porosity is the dominant factor in tensile properties [11]. The variation in tensile properties can be explained by unfused defects and entrapped gas in powder particles, as shown in Figures 13 and 14. There were no significant differences between the layer thicknesses of 200 and 300 µm when the relative density was higher than 99.5%, which indicates that Ti6Al4V parts with desirable mechanical properties could be obtained with high layer thickness up to 300 µm via the EB-PBF process.

Building Rates
For the PBF process, the building rate is a key parameter that determines the forming efficiency and cost. The building rate can be defined as the volume melted per unit time, which can be calculated by where B denotes the building rate (mm 3 /s).

Building Rates
For the PBF process, the building rate is a key parameter that determines the forming efficiency and cost. The building rate can be defined as the volume melted per unit time, which can be calculated by where B denotes the building rate (mm 3 /s). The typical building rates of Ti6Al4V for different PBF processes are listed in Table  5. For the L-PBF process, the building rate is lower than 5 mm 3 /s and can be raised up to 7.2 mm 3 /s by expanding the layer thickness from 30 μm to 200 μm [14]. As for the EB-PBF process, the typical building rate achieved by the ARCAM machine ranges from 2.5 to 30 mm 3 /s due to different scanning speeds [3]. In this research study, a building rate up to 30 mm 3 /s was achieved using the layer thickness of 300 μm and the scanning speed of 500 mm/s, which is 12 times the typical EB-PBF building rate at the scanning speed of 500 mm/s and equivalent to the building rate at the scanning speed of 6000 mm/s. It should be noted that, when considering the time spent on powder spreading and powder preheating process to build parts of the same height with a high layer thickness, it is significantly lower than that obtained using a low layer thickness; the forming efficiency using The typical building rates of Ti6Al4V for different PBF processes are listed in Table 5. For the L-PBF process, the building rate is lower than 5 mm 3 /s and can be raised up to 7.2 mm 3 /s by expanding the layer thickness from 30 µm to 200 µm [14]. As for the EB-PBF process, the typical building rate achieved by the ARCAM machine ranges from 2.5 to 30 mm 3 /s due to different scanning speeds [3]. In this research study, a building rate up to 30 mm 3 /s was achieved using the layer thickness of 300 µm and the scanning speed of 500 mm/s, which is 12 times the typical EB-PBF building rate at the scanning speed of 500 mm/s and equivalent to the building rate at the scanning speed of 6000 mm/s. It should be noted that, when considering the time spent on powder spreading and powder preheating process to build parts of the same height with a high layer thickness, it is significantly lower than that obtained using a low layer thickness; the forming efficiency using layer thickness of 300 µm and scanning speed of 500 mm/s should be greater than that obtained using layer thickness of 50 µm and scanning speed of 6000 mm/s.

Conclusions
In this research study, layer thicknesses up to 300 µm and accelerating voltage values up to 90 kV were used to fabricated Ti6Al4V samples by EB-PBF with different bulk energy densities. The relative densities, surface morphologies, microstructure and mechanical properties of the as-built samples were analyzed and compared using the layer thicknesses of 200 and 300 µm and the accelerating voltage values of 60 and 90 kV. The conclusions are as follows: (1) Parts with relative density greater than 99.5%, ultimate tensile strength higher than 950 MPa and elongation higher than 9.5% were fabricated when the layer thickness expanded to 300 µm under both accelerating voltage values. The variation in tensile properties was mainly caused by porosities.
(2) With the increase in the layer thickness or accelerating voltage, the minimum bulk energy density needed to fabricate dense parts decreased due to the decrease in the proportion of the remelted area in the molten pool and the α platelet thickness increased due to the increase in the molten pool.
(3) A higher accelerating voltage enlarged the melting depth, which improved the surface finish even if it was impaired due to the expansion of layer thickness.
(4) A building rate up to 30 mm 3 /s without sacrificing the mechanical properties could be achieved when the layer thickness was 300 µm, the hatch space was 200 µm and the scanning speed was 500 mm/s.