1 Introduction

In the world of additive manufacturing (AM), electron beam powder bed fusion (PBF-EB) gains more and more importance [1]. The process is especially well-suited for high-performance parts in the aerospace or medical industry [1,2,3]. The process is composed of lowering the build platform by the layer thickness, applying the powder, and pre-heating the powder bed by a defocused electron beam. According to the sliced CAD geometry, the sintered powder is molten by a focused electron beam. Subsequently, an electron image of the molten layer can be taken if electron optics (ELO) are available [4]. For a comprehensive review of the process itself, the advantages and disadvantages, as well as the simulation of the process, the reader is referred to [1,2,3].

The influence of different process parameters on PBF-EB is studied for various materials [1, 5,6,7,8,9]. However, the influence of the acceleration voltage Ub has not been studied in depth due to the limitations of conventional machines to 60 kV. For the longest time, GE Additive, formerly known as Arcam AB, was the only manufacturer of PBF-EB machines. In recent years, companies like Freemelt AB or JEOL among others have brought new machines to market. However, these machines still are limited to a maximum acceleration voltage of 60 kV [1]. Therefore, a novel prototype machine from pro-beam GmbH & Co. KGaA is used for this study, which enables acceleration voltages of up to 150 kV. The advantages of a higher acceleration voltage for the PBF-EB process are described in the following sections.

First, pre-heating is conducted to set the process temperature and form a stable sinter cake, which increases the thermal and electrical conductivity of the powder bed. During pre-heating, the kinetic energy of the electrons E0 = 150 keV for 150 kV acceleration voltage, is absorbed by the powder particles. The kinetic energy is mostly transferred into heat and the temperature of the powder bed rises. To reach the desired process temperature, a certain beam power Pb has to be applied for a certain time. Simultaneously, the charge of the electrons is absorbed by the powder particles. Milberg and Sigl identified this electrostatic charge, introduced by the electron beam, as the main cause of smokes [10]. While the electric conductivity inside the metal powder particles is extremely high, the contact areas present bottlenecks for the flow of electrons, especially if oxides are present [11, 12]. In addition, surface oxides lead to charge accumulation enhancing the chance of smokes [11, 12]. When a smoke happens, powder particles are explosively spread in the building chamber, often leading to the failure of the process [10, 13].

There are different approaches to avoid smokes. The most common PBF-EB machines from GE Additive use helium as a rather expensive process gas to stabilize the process [14]. Other methods increase the conductivity of the powder by thermal [15] or thermo-mechanical treatment [11]. Further, there are several approaches to stabilize the process with additional technologies. For instance, “ProHeat” by Freemelt uses the thermal radiation of a tungsten plate, which is heated by the electron beam, to sinter the powder bed [16], or “NeuBeam” by Wayland provides a stream of positive ions toward the powder bed to neutralize the negative charge [17]. However, these methods either cost time and money or increase the complexity of the process.

While a higher acceleration voltage enables higher beam powers with the same beam current Ib, leading to shorter pre-heating times, it also allows the usage of smaller beam currents for the same beam power, which means fewer electrons are introduced into the powder bed. For instance, pre-heating with 3000 W needs 50 mA for 60 kV, while with 150 kV only 20 mA are necessary. Consequently, less charge is introduced into the powder bed and the risk of smokes is reduced.

Moreover, a higher acceleration voltage is expected to have an impact on the melting of the powder bed. The electron beam is focused on a smaller beam diameter than for pre-heating to locally provide sufficient energy for melting the powder. For 60 kV acceleration voltage and the TNM (Ti-43.5Al-4Nb-1Mo-0.1B) gamma titanium aluminide alloy (γ-TiAl), the maximum penetration depth, calculated by the formulation introduced by Kanaya and Okayama, is 16 µm [18]. Moreover, most energy and hence electrons are absorbed much nearer to the surface. For a detailed description of the absorption of the electron beam during PBF-EB, the reader is referred to [19]. Since the penetration depth is smaller than the typically used layer thickness for PBF-EB (50  to 150 µm), Galati et al. concluded that melting in the building direction is governed by heat conduction [20]. Consequently, the volume of initial energy absorption has to be heated above the melting temperature to ensure complete fusion between the molten layer and the previously built part. This effect is enhanced by the low thermal conductivity of the powder bed [20, 21]. A higher acceleration voltage leads to deeper penetration depths of the electrons (see Table 1), which enables a more homogeneous energy absorption within the powder particle [19]. As a result, smaller vertical volumes have to be molten via heat conduction, and less energy is needed to produce dense parts. Simulations by Klassen et al. support this conclusion since the number of defects is decreased if the same line energy is applied with higher acceleration voltages [19].

Table 1 Maximum penetration depth for the Ti-43.5Al-4Nb-1Mo alloy calculated according to [18] in dependence of the acceleration voltage
Full size table

In addition, an increasing acceleration voltage enables smaller beam diameters and higher beam power densities [22, 23]. The higher the acceleration voltage, the shorter the time of flight of the electrons from the anode to the target, and hence the duration of the so-called “space charge effect” is limited [23, 24]. The space charge effect describes the repulsive Coulomb force, which is applied by the electrons of the beam on each other and broadens the beam [23, 24]. A recent study, conducted with an electron beam welding machine, confirmed that an increasing acceleration voltage and decreasing beam current reduce the minimal beam diameter [22].

It is to note that the lens current of the focus lens has to be adapted to the acceleration voltage. This is based on Eq. (1), which describes the focal length f of an electric lens [25]. m is the mass of an electron, Ub is the acceleration voltage, e− is the elemental charge of an electron, and B is the magnetic field generated by the electric lens. In the case of a PBF-EB machine, the focal length is equal to the working distance, and therefore has to remain constant. Consequently, it is necessary to increase the magnetic field by increasing the lens current for higher acceleration voltages.

$$f= \frac{8m{U}_{b}}{{e}^{-}{\int }_{-\infty }^{+\infty }{B(z)}^{2}dz}$$
(1)

Li et al. investigated the effect of an increased acceleration voltage on the PBF-EB process [26]. The acceleration voltage is increased from 60 to 90 kV for Ti-6Al-4V and argon is used as a process gas [26]. They can reduce the beam current for pre-heating while shortening the pre-heating time as well [26]. Further, it is shown that the energy needed for melting is reduced by 20–40% depending on the scanning velocity. However, they report that the full width at half maximum (FWHM) beam diameter for 450 W is lowered from 401 µm for 60 kV to 308 µm for 90 kV. Since the beam diameter has a strong influence on the power density and energy input during melting [27], this has to be considered apart from the influence of the acceleration voltage. Consequently, the beam diameter has to be the same to isolate the impact of the acceleration voltage on PBF-EB.

This also highlights one of the challenges when comparing PBF-EB machines. Each machine has its unique environment and especially beam quality. However, to investigate the impact of higher acceleration voltages, e.g., 150 kV, compared to conventionally used 60 kV, it is necessary to compare different PBF-EB machines, since no current commercial machine can operate on the whole range from 60 to 150 kV. For processing γ-TiAl alloys via PBF-EB, the Arcam A2X is the most commonly used machine [5, 28,29,30,31]. Consequently, the approach in the presented work is to take the beam diameter of the Arcam A2X 60 kV machine from literature [32] and imitate the electron beam on the HELIOS machine at higher acceleration voltages. This is possible due to a detailed beam diameter measurement. Moreover, all other process parameters as raking, pre-heating, or the atmosphere are kept as similar as possible. To the best of the authors’ knowledge, this is the first attempt of its kind to investigate the impact of the acceleration voltage over such a wide range and two different PBF-EB systems.

2 Experimental

The experiments are conducted with a γ-TiAl alloy called TNM. The gas-atomized Ti-44.5Al-4Nb-1Mo-0.1B (at%) powder from Allegheny Technology (Pittsburgh, US) was previously reused multiple times on another PBF-EB system over a span of 2 years. Therefore, degradation of the powder is expected. The powder possesses a particle size distribution of 45 µm to 150 µm, an oxygen content of 1250 ppm measured via EMGA 620W (HORIBA, Kyoto, Japan), and a gas porosity of 0.1%.

To compare the high acceleration voltage of the novel prototype to a standard 60 kV Arcam A2X (GE Additive, US) machine, the process conditions are kept as similar as possible. After raking, both machines perform a standard pre-heating pattern on a 130 mm x 130 mm square. The velocity (16 900 mm/s), line offset (1 mm), and scan strategy (line order 10) during pre-heating are identical. The target of the pre-heating is a process temperature of 1000 °C, which is achieved in roughly 60 s on both machines. It has to be emphasized that the 150 kV machine only operated at around 25% of its maximum power. Consequently, a shortening of the pre-heating time would be possible by increasing the beam power. For improved comparability, the authors refrained from increasing the beam power. After pre-heating, the cuboids are molten with the same standard cross-snake scanning strategy, which is rotated after each layer by 90°. The melting parameters are described in detail below. The only difference in the PBF-EB processes is the electron optic (ELO) image, which is taken at the end of each layer on the prototype machine.

The 150 kV and 125 kV experiments are carried out on a prototype PBF-EB machine called HELIOS developed and built by pro-beam GmbH & Co. KGaA (Gilching, Germany), see Fig. 1 [33]. The thermionic electron gun, for further information see [25], consists of a tungsten filament, a Wehnelt cup, and an anode, which is capable of adjusting the acceleration voltage between 100 and 150 kV. It is to note, that after changing the acceleration voltage, an automated calibration of the electromagnetic lens system is necessary. Therefore, the acceleration voltage can only be changed between build jobs. With the electron gun, beam powers of up to 45 kW are possible. Additionally, the prototype, with a build volume of 400 × 400 × 500 mm3 (xyz), is equipped with an ELO system for process monitoring. Excess powder is moved into an overflow tank. Lastly, to accelerate the cool down after the build job, a gas inlet is added after the machine acceptance by Neue Materialien Fürth GmbH (Fürth, Germany).

Fig. 1
figure 1

Schematic of the PBF-EB prototype, developed by pro-beam GmbH & Co. KGaA (Gilching, Germany). The so-called HELIOS is capable of 100 to 150 kV acceleration voltage and 45 kW beam power. It has an electron optic (ELO) detector for process monitoring and a build tank of 400 × 400 x 500 mm3 (xyz). The schematic is adapted from [33]

Full size image

The free programmable PBF-EB prototype runs on a script developed by the Chair of Material Science and Engineering for metals at the Friedrich-Alexander-Universität Erlangen-Nürnberg (Erlangen, Germany). The flexibility of the in-house developed script on a freely programmable machine allows us to mimic the PBF-EB process of an Arcam A2X 60 kV machine as presented in this work, while also providing the opportunity to develop new process strategies in future research.

All experiments are conducted at a pressure of 2 × 10–5 mbar without process gas and with constant layer thickness and line offset lo of 100 µm. The acceleration voltage is set to 125 kV and 150 kV with beam powers of 300 W, 600 W, 900 W, 1200 W, and 1500 W. Cuboids with 125 kV are built with a 2720 mA focus current, and cuboids with 150 kV are built with 3000 mA. Both focus currents represent a slightly defocused electron beam. Layered cubes with changing beam powers after 3.5 mm build height are built to realize a large number of parameter combinations. The total dimensions of the cubes are 15 × 15 × 17.5 mm3 (xyz) with 3 mm rod support connected to the start plate. Additionally, cubes with a dimension of 15 × 15 × 10 mm3 (xyz) and selected constant parameters are built for arc optical emission spectroscopy analysis.

The melting during the PBF-EB process is determined by the energy input over time. For a constant scan length ls, this is well described by the area energy Ea. The area energy is calculated by the acceleration voltage Ub, the beam current Ib, the beam velocity vb, and the line offset lo, according to Eq. (2). The acceleration voltage multiplied by the beam current is the beam power Pb. The beam velocity is varied for each beam power and acceleration voltage to achieve area energies between 0.90 J/mm2 and 2.50 J/mm2. All velocities are listed in Table 2 in the appendix.

$${E}_{a}= \frac{{U}_{b} \times {I}_{b}}{{v}_{b }\times {l}_{o}}= \frac{{P}_{b}}{{v}_{b} \times {l}_{o}}$$
(2)

One of AM's main advantages is the possibility of producing complex parts. However, these geometries consist of varying scan lengths, which change the energy input over time. While the energy input is still well described by the area energy, the time variable changes. Therefore, the lateral velocity vlat is introduced as a scan length-independent description of the time variable during melting [34]. The lateral velocity is defined by the scan length ls, the beam velocity vb, and the line offset lo, according to Eq. (3). A numerical study concluded, that the melt pool and hence the resulting defects and microstructure are constant over varying scan lengths if the area energy and lateral velocity are constant [34]. This conclusion is supported by experimental work [28, 35]. More recent studies revealed that the approach is feasible if the power-dependent beam diameter, which has a strong influence on the energy input, is considered [27]. Therefore, in this study, the process map is evaluated for the area energy over the lateral velocity.

$${v}_{lat}= \frac{{v}_{b }\times {l}_{o}}{{l}_{s}}$$
(3)

The 60 kV experiments are performed on an Arcam A2X PBF-EB system. A maximum focused beam with a 10 mA focus offset is used for melting without helium as process gas. The process pressure achieved by the Arcam A2X is around 1×10–4 mbar. Twenty-four cuboids with constant parameters and a dimension of 15 × 15 x 15 mm3 (xyz) are built with 5 mm of rod support. The final dimension of the cuboids is 15 × 15 x 12 mm3 (xyz) due to process failure after 17 mm build height because of the lack of Helium. The beam power is set to 300 W, 600 W, 900 W, and 1200 W, while the line offset and layer thickness are kept constant at 100 µm. The area energies range from 0.06 J/mm2 to 3.00 J/mm2. The velocities of all cuboids can be found in the appendix in Table 3.

The beam diameter of the HELIOS is measured by the machine manufacturer pro-beam with their standardized measurement system, which is well established in the electron beam welding industry. The sensor is based on a Faraday cup with a pinhole and the electron signal is measured by the EB Vision software provided by pro-beam. The sensor is placed in the center of the build tank in the xy-direction and the aperture of the sensor is positioned at the height of the raking plane (z-direction), where the melting during the PBF-EB process takes place. The beam diameter is defined as the second-order moment of the power distribution, which equates to the 4σ definition of a perfect Gaussian beam. The calculation of the beam diameter is adapted from ISO 11146. The measurement has to be carried out carefully, to avoid damaging the pinhole by overheating. To constantly monitor the beam quality, the optimum focus value for 5 mA beam current is checked before every build job. The focus value is determined by an automated script provided by pro-beam, which uses a perforated plate and electron optics. The optimum focus value is rather constant for each acceleration voltage, respectively, indicating a constantly good beam quality.

The produced cuboids are cut, ground, and finally polished with a suspension of 50 ml OPS, 50 ml distilled water, 10 mg KOH, and 10 ml 30% H2O2. Images for defect analysis are taken with a Zeiss AxioImager.M1m (Carl Zeiss AG, Jena, Germany). At least 20 images, taken with 100 × magnification, are analyzed per parameter by an automated script to determine the amount of misconnection and gas porosity. The swelling of the surface and the geometrical inaccuracy is detected via the electron optics in the HELIOS machine and optical examination for the A2X 60 kV samples. For selected parameters, the aluminum content is measured via arc optical emission spectroscopy (OES) with a SPECTROMAXx (SPECTRO Analytical Instruments GmbH, Kleve, Germany).

3 Results

3.1 Beam diameter measurement

Figure 2 shows the beam diameter, which is used for melting, for each beam power and acceleration voltage, respectively. For the HELIOS machine, the beam diameter is measured from 1 to 10 mA in 1 mA steps for each acceleration voltage. All measurements are done without process gas. The 150 kV measurements are conducted first. During the measurement of the beam diameters for 125 kV, the pinhole was damaged during the 125 kV 11 mA measurement. Since the measurement device is no longer available, beam diameters for 125 kV are only available up to 1250 W. All beam diameters are according to the second moment of area, which is equal to the 4σ-diameter for a perfect Gaussian beam.

Fig. 2
figure 2

Beam diameters used for melting over the beam power. The used and minimal beam diameters of the HELIOS machine are measured, while the 60 kV Arcam A2X values are only calculated from literature [32]. The electron beam of the HELIOS machine is slightly defocused to fit the values of the Arcam A2X machine (60 kV). The minimum measured beam diameter of the HELIOS machine mainly depends on the acceleration voltage

Full size image

Additionally, the smallest possible beam diameter, which is achieved during the measurements on the HELIOS, is shown. It cannot be emphasized enough, that the beam on the HELIOS is intentionally defocused during melting to imitate the beam diameter of the Arcam A2X 60 kV machine from literature [32]. The reason for this approach is to isolate the impact of acceleration voltage, which is only possible if similar beam diameters are used.

With the available setup, a measurement of the beam diameter on the Arcam A2X 60 kV machine is not possible. Hence, literature beam diameters according to [32] are displayed as blue triangles, which can be treated as a guideline for comparison. Figure 2 shows that the minimal beam diameter of the HELIOS is smaller compared to 60 kV. This highlights the future potential for further improvements of the PBF-EB process by high acceleration voltages. Since the beam diameter has a strong influence on the melt pool and defect formation [27], similar beam diameters are required to compare the resulting defects and microstructures for different acceleration voltages. Therefore, the electron beam with 150 kV (black cubes) and 125 kV (red circles) is slightly defocused to fit the beam diameters for 60 kV from literature. These defocused beam diameters are given in Fig. 2. It is noted that defocusing the electron beam can change the shape and the power distribution within the beam. Since the beam is only slightly defocused, it is assumed that the beam shape remains round and the power distribution is similar to Gaussian.

3.2 Processing without process gas possible

All samples analyzed in this study are built in a vacuum atmosphere without process gas. On the Arcam A2X machine (60 kV), the process pressure is 1×10–4 mbar and the unstable process was shut down after 17 mm due to several smokes and delamination. Although the 12 mm high cuboids are sufficient for the evaluation of the process parameters, no larger samples for mechanical characterization or any parts could be built this way.

On the HELIOS machine, the process pressure is 2×10–5 mbar and all build jobs are finished successfully. The higher acceleration voltage reduces the risk of smokes by decreasing the required beam current to maintain the process temperature. The high vacuum makes the PBF-EB process cleaner and reduces the heat loss of the pre-heating area due to less interaction of the atmosphere with the hot process area. The build jobs with 125 kV and 150 kV are stable and have no issues with smokes. With 150 kV acceleration voltage, it is possible to build turbine blade demonstrators with a height of 230 mm at 1000 °C process temperature in 43 h. Ultimately, the high acceleration voltage enables the processing in a true vacuum atmosphere without any process gas.

3.3 Process windows

The process maps of the γ-TiAl TNM alloy for all three tested acceleration voltages are shown in Fig. 3. It has to be pointed out that the process windows, comprising all dense samples, are not optimized for 150 and 125 kV, since the electron beam with high acceleration voltage is defocused to fit the literature values for 60 kV. A widening of the process window to lower area energies can be expected when using the minimal beam diameter. However, to compare the different acceleration voltages, a similar beam diameter is essential.

Fig. 3
figure 3

Processing window of the TNM γ-TiAl alloy for three different acceleration voltages. The area energy is shown over the lateral velocity. The beam power used and beam diameter is given within the diagrams. Samples with misconnections > 0.05% are classified as misconnected (empty circles), while bulged samples are marked as swelling (triangles). If the molten surface strongly exceeds the CAD geometry, the sample is named geometrical inaccurate (diamonds). The process window is given by the dense samples (cubes)

Full size image

For all process maps, the area energy is plotted over the lateral velocity. Additionally, the beam powers used and the corresponding beam diameters are given in the diagrams. A sample is classified as misconnected (empty circles) if the cross-section shows more than 0.05% misconnections. A misconnection is formed at the bottom of the melt pool if the melt pool depth is not sufficient to melt the whole powder layer. Due to its shape, misconnections are detrimental to mechanical properties, especially for brittle materials like γ-TiAl. Therefore, the amount of misconnection is the criteria for dense samples in this work. The gas porosity is not considered for the process window but is discussed separately in the following section. Samples with surface swelling, resulting from a partial permanent persistent melt pool [36], are shown as triangles. For high energies and high lateral velocities, the melt pool gets bigger and more dynamic and exceeds the boundaries of the CAD geometry. These samples are shown by diamonds and are named geometrical inaccurate. The process window is displayed by the dense samples (cubes). Additionally, the expected transition from a trailing to a persistent melt pool is given by a dashed line. It is derived from a combination of experiment and previous simulation. The trailing melt pool is already solidified when the beam returns in the subsequent scan line because the melt pool lifetime is shorter than the beam return time [36]. Contrarily, a persistent melt pool is still liquid when the electron beam returns [36].

First, the transition of a trailing melt pool to a persistent melt pool is examined, since these melt pool characteristics have a strong influence on the formation of defects. As shown in Fig. 3, dense samples with a persistent melt pool need less energy than samples with a trailing melt pool for all acceleration voltages. For high area energies, the transition from trailing to persistent is characterized by the formation of surface swelling (triangles), which is detected by the ELO system. The swelling originates from a partial persistent melt pool as explained in [36]. For lower area energies, the transition from trailing to persistent has to be determined by analysis of the melt surface of the last layer. In this study, layered cuboids are built to investigate a large range of parameters with 125 kV and 150 kV. Therefore, the melt surface is only available for the last parameter set built. In the case of 125 kV, these are the 300 W samples and for 150 kV, these are the 600 W samples. All the melt surfaces show a trailing melt pool. As a result, the course of the transition at lower area energies is adapted from literature [36]. Nevertheless, a shift of the transition to higher lateral velocities for 150 kV acceleration voltage is observed.

High acceleration voltages lead to a larger process window of the γ-TiAl alloy (see Fig. 3). On the one hand, less energy is required to produce dense samples. This is observed for a trailing melt pool as well as for a persistent melt pool. On the other hand, the process window is expanded to higher lateral velocities. For 60 kV, no samples with 1500 W are built because a massive rise in beam diameter (891 µm) is expected according to literature [32].

The process windows for 125 kV and 150 kV are similar since the difference in acceleration voltage is rather small. It has to be emphasized that these process windows are not optimized. Using the minimal beam diameter instead of the value adapted to the 60 kV system, the process window can be further enlarged, especially toward lower area energies for dense samples.

3.4 Gas porosity

In Fig. 4, the gas porosity of the dense parameters (black cuboids in Fig. 3) is shown. Additionally, the trailing to persistent melt pool transition is indicated by a dotted line.

Fig. 4
figure 4

Gas porosity as a bubble plot with the lateral velocity on the x-axis and the area energy on the y-axis. Every dense parameter from Fig. 3 is represented by a bubble. The area of the bubble indicates the gas porosity. Additionally, the transition from a trailing to a persistent melt pool is indicated by a dotted line

Full size image

For higher area energy and lateral velocity, the gas porosity is reduced. To decrease the gas porosity, the gas pores need more time to ascend in the melt pool and leave the sample. Therefore, the gas porosity is a good indicator of melt pool lifetime for dense samples. When comparing samples with similar parameters in the persistent melt pool regime, the acceleration voltage seems to have a minor effect on the gas porosity. For a trailing melt pool, slight differences can be observed. Due to the small number of dense samples in this region, these differences have to be analyzed very carefully to avoid over-interpretation of small fluctuations in the process.

3.5 Aluminum evaporation

To analyze the aluminum evaporation via OES, additional samples with a dimension of 15 × 15 × 10 mm3 (xyz) are produced. Since only a limited number of samples can be examined, the beam powers 600 W, 900 W, and 1200 W are chosen for this investigation.

In Fig. 5, the aluminum content is displayed over the area energy a) and the lateral velocity b). For all acceleration voltages, the aluminum content decreases with increasing area energy, which agrees well with the literature [5]. Although the OES contains some measurement uncertainty indicated by the error bars, a linear fit, which is not based on a physical model, including all samples can be calculated and is given within the diagram. Area energies below 0.75 J/mm2 will not increase the aluminum content further, since the values for 0.75 J/mm2 are already close to the composition of the powder. In addition, increasing the area energy above 2.25 J/mm2 will lead to strong swelling and eventually to an unstable melt pool. Therefore, the linear decrease will not continue for very high area energy. Nevertheless, the linear fit describes the aluminum content dependence on the area energy within the range of parameters presented in this study well. The grey area displays the linear fit ± 0.2 wt.%, which is the average measurement uncertainty of the OES. When considering the measurement uncertainty of the OES, neither the acceleration voltage nor the beam power seems to influence the aluminum evaporation strongly.

Fig. 5
figure 5

a The measured aluminum content over the area energy. With increasing area energy, the aluminum content of the samples decreases. A linear fit including all samples is given. The grey area represents the linear fit ± 0.2 wt.%, which is the measurement uncertainty of the spark measurement. b The same aluminum measurements are plotted over the lateral velocity

Full size image

Analyzing the aluminum content dependence of the lateral velocity reveals that the area energy remains the crucial element for aluminum evaporation (see Fig. 5b). However, an influence of the lateral velocity cannot be excluded, due to the measurement uncertainty of the OES. In theory, a higher velocity should reduce the peak temperature, and hence less aluminum should evaporate [32].

4 Discussion

For future research, the minimum beam diameter of the high acceleration voltage is of great interest. For higher acceleration voltages, smaller beam diameters are possible due to the limitation of the space charge effect by reducing the time of flight (see Fig. 2). A smaller beam diameter should be beneficial to melting and enables higher resolution for electron optical process observation. It should lower the necessary area energies for dense samples and widen the processing window for higher lateral velocities if the beam diameter is kept constant. In this study, the effect of increasing the acceleration voltage from 60 to 125 kV and 150 kV is discussed. Therefore, a beam diameter similar to 60 kV is chosen.

The feasibility of running a PBF-EB process without any process gas has to be highlighted for high acceleration voltages. Less beam current is needed to reach the beam power necessary for maintaining the process temperature of 1000 °C. Hence, less charge is introduced into the powder bed, enabling the processing of titanium aluminide powder in a 2×10–5 mbar vacuum atmosphere without helium. The process in a true vacuum atmosphere is cleaner than using process gas, which can contain residual gasses like oxygen. Moreover, the lower the process pressure, the better isolated the hot process surface, which leads to less heat loss due to interactions with the atmosphere. The powder is processed for over 40 h in a single build job at 150 kV. Despite the many previous reuse cycles on an Arcam A2X and its high oxygen content of 1250 ppm, no smoke issues occurred for 125 kV and 150 kV. This outlines the robustness of the process without process gas. Moreover, the high oxygen content shows no negative influence on the mechanical properties of the alloy [37].

Examining the melting with a higher acceleration voltage, first, the melt pool formation is discussed, since it is crucial for sample quality, such as misconnections, gas porosity, chemical composition, microstructure, surface topography, or surface roughness [5, 7, 27, 34, 36, 38]. Therefore, the influence of the acceleration voltage on the melt pool formation is touched upon briefly. Nevertheless, a lot of future research is required to develop a sophisticated model of the effect of acceleration voltage on melt pool formation.

To achieve a melt pool depth sufficient to produce dense samples, the persistent melt pool requires less area energy than the trailing melt pool, due to the enhanced heat transfer from previous layers [34]. Furthermore, the persistent melt pool stays liquid longer than the trailing melt pool [34]. This allows the gas entrapped in the powder particles to rise and escape the melt pool. As a result, less gas porosity is expected for longer-living melt pools [27]. Both effects are observed for all three tested acceleration voltages in a similar fashion (see Fig. 3). During the transition from a trailing to a persistent melt pool, a partial permanent persistent melt pool is formed, which leads to a characteristic surface swelling [36]. This is observed for high area energies independent of the acceleration voltage. However, the transition for 150 kV is shifted to higher lateral velocities compared to 125 kV and 60 kV. A possible explanation is the absorption of the energy over a larger depth. Therefore, the surface of the melt pool should be cooler and solidify faster. Consequently, higher lateral velocities are necessary to achieve a persistent melt pool.

Furthermore, two main trends are observed when increasing the acceleration voltage from 60 to 125 kV or 150 kV. First, less area energy is needed to produce dense samples (see Fig. 3). The larger penetration depth, provided by the higher acceleration voltage [18], reduces the amount of powder, which has to be molten via heat conduction (effective layer thickness of the powder minus the penetration depth of the electrons). Therefore, the overheating of the melt pool can be decreased and less energy is required to produce dense samples. Second, higher lateral velocities can be used with higher acceleration voltage. Contrarily to the 60 kV machine, beam widening for beam powers larger than 1200 W is avoided, which enables the usage of higher beam powers, and hence higher lateral velocities. Further improvement of the process window is expected for the minimal beam diameter, which is not investigated in this study.

The process windows of 125 kV and 150 kV are similar and no clear influence of the acceleration voltage is observed. However, the change in acceleration voltage is much larger from 60 to 125 kV than from 125 to 150 kV, so a bigger impact is expected. This is also the reason for conducting the study on two different machines. Otherwise, the influence of the acceleration voltage can only be studied over a limited range.

No influence of the acceleration voltage on the gas porosity is found in this work (Fig. 4). Nevertheless, the effects of area energy and lateral velocity on the gas porosity, known for 60 kV, can be confirmed for 125 kV and 150 kV in similar magnitude.

Analyzing the aluminum content, the acceleration voltage shows no substantial influence on the aluminum evaporation, if the same beam diameter is used (see Fig. 5a). However, the measurement uncertainty of the OES, which was chosen due to its short measurement time to analyze a multitude of parameters, is rather high. Therefore, a small influence of the acceleration voltage on the aluminum cannot be excluded. The same conclusion can be drawn for the lateral velocity (see Fig. 5b). Regarding the data presented in this work, the area energy seems to be the key factor for aluminum evaporation. The absence of process gas does not lead to enhanced aluminum evaporation. The amount of aluminum evaporation is in the same range as similar samples from literature built with 60 kV and helium [5, 27, 28]. If a higher acceleration voltage leads to more homogenous aluminum evaporation, it is the subject of future investigations.

When combining the lower required area energy for 125 kV and 150 kV (Fig. 3) with the notion that the area energy is crucial for aluminum evaporation (Fig. 5a), higher acceleration voltages provide the opportunity to adjust the aluminum content over a larger range. This enables or improves innovative process strategies, for example locally adjusted aluminum contents for the dual microstructure concept of γ-TiAl turbine blades in jet engines [28].

5 Conclusion

To compare the high acceleration voltage of 125 kV and 150 kV to the conventionally used 60 kV, the beam diameter of the HELIOS machine is adapted to literature values of the 60 kV GE Additive A2X machine. This is of paramount importance when comparing different PBF-EB machines and acceleration voltages since the beam diameter has a strong influence on the final part quality. It is to note that electron beam powder bed fusion with high acceleration voltage and minimal beam diameter should yield even better results.

With the data presented in this study, several advantages of a higher acceleration voltage for PBF-EB can be shown:

  • With a higher acceleration voltage, PBF-EB is feasible without any process gas.

  • Higher available beam power can shorten the pre-heating duration or enables larger pre-heating areas. Both will boost the productivity of the PBF-EB process on its way to broader industrial application.

  • Increasing the acceleration voltage from conventional 60 kV to 125 kV or 150 kV improves the process window for dense samples toward lower area energies and higher lateral velocities.

  • An increase from 125 to 150 kV acceleration voltage did not significantly improve the process window in this work.

  • The absence of process gas did not lead to increased aluminum evaporation compared to literature values.

  • There is no effect of the acceleration voltage on the aluminum evaporation found in the scope of this work. However, due to the limitations of the OES used to determine the aluminum content of the samples, a small effect cannot be excluded.

  • The same holds for the lateral velocities investigated here.

Although machines with higher acceleration voltage have to be equipped with stronger shielding due to enhanced stray radiation, the increased acceleration voltage benefits the PBF-EB process greatly. Similar advantages as demonstrated in this study can be expected for other materials and will boost the development of novel scan strategies. Furthermore, optimizing the process for higher acceleration voltages using the smallest possible beam diameter should enhance the discussed benefits further. Therefore, the true potential of 150 kV acceleration voltage is even higher than presented in this study. Consequently, 150 kV acceleration voltage has the potential to enable the next generation of PBF-EB machines on the way to broader industrial applications.