1 Introduction

The ATI C103™ Alloy, comprising of Nb–10Hf–1Ti (wt pct), was developed as a collaboration between ATI and Boeing in the late 1950s, primarily for use in rocket propulsion applications such as expansion skirts and combustion chambers.[1,2,3,4] The addition of reactive elements such as Hf, Ti, Zr, Y, and C is known to strengthen niobium alloys,[5] where during melt processing, ATI C103™ is conducive to the formation of grain boundary hafnium oxide. The key benefits of C103 stem from the additions of Hf and Ti. The addition of ~ 1 wt pct titanium significantly enhances the strength and stability of niobium alloys, striking a balance between increased strength without sacrificing high ductility.[6] This combination of properties allows for ease of fabrication through conventional welding, forming, and machining techniques. Additionally, unlike most refractories,[7] the low ductile-to-brittle transformation temperature has made C103 the preferred refractory alloy for propulsion applications over the last 60 years.[8]

Despite being fabricable through various means, the utilization of additive manufacturing (AM) for C103 is appealing due to the potential elimination of multi-step joining, increased geometric complexity, and decreased requirement for machining, especially for low-volume production runs. The advantages of material efficiency in AM design and construction are notably pronounced for refractory alloys, considering their high-value raw materials and the significance of weight reduction in rocketry and hypersonic flight requiring complex and highly optimized geometry. A recent study by Mireles et al., laser powder bed fusion (LPBF) AM of relevant C103 thruster components to be significantly more cost-effective than traditional machining, despite the higher cost of AM-specific powder feedstock, post-treatment, and finishing.[9] With the rapid commercialization of space access, numerous opportunities exist to evaluate and incorporate design flexibility through the development of lighter-weight designs. Overall, there are limited published data on AM of C103, and of these, only a few have investigated the effects of AM processing on the microstructure and properties of C103, with a majority of these studies focused specifically on LPBF AM.[8,9,10,11,12,13] The general success of identifying a stable AM process window with high material density without cracking in these studies can largely be attributed to the weldability of C103.[14,15,16] Awashti et al. and Mireles et al. each produced specimens without cracks and with relative densities exceeding 99 pct with LPBF. Similarly, Colon et al. demonstrated powder-based directed energy deposition of C103 with a relative density > 99 pct.[17] Philips et al. explored both LPBF and electron beam powder bed fusion (EB-PBF) AM methods. These results suggest that the strength properties of AM-produced ATI C103™ are comparable to and, in some cases, superior to wrought material. The LPBF achieved a higher yield and tensile strength compared to wrought at room temperature, with a slightly lower yield but higher tensile strength at elevated temperature (1755 K). The EB-PBF samples show similar yield strength but lower tensile strength at room temperature compared to wrought material, with comparable strengths at 1366 K.[8] The underlying phenomena driving the strengthening are not yet fully understood. Both methods share the common characteristic of depositing and then selectively melting thin metallic powder layers, yet they differ significantly in energy coupling, heat transfer, the processing environment, and process control. LPBF typically takes place in an inert gas environment (e.g., Ar) at room temperature or slightly elevated temperature. EB-PBF typically takes place under mid-level vacuum conditions (~ 2e−3 to  ~ 2e−7mBar) and at elevated ambient powder bed temperatures (typically near the sintering temperature).[18,19] Both EB-PBF and LPBF produced C103 show a strong [001] texture in the build direction, with the EB-PBF grains being large and columnar, and due to the high ambient build temperatures, typically exhibit low residual stress in the as-fabricated condition.[8] While not fully describing the observed effects, the increase in UTS and reduction in elongation associated with increased interstitial oxygen in niobium alloys are well documented.[20] Nevertheless, controlling, or at least quantifying, oxygen pickup within the AM parts is an important concern. In all reported cases, the initial LPBF powders (prior to AM processing), measured by inert gas fusion, were higher than the ASTM B655-10 oxygen limit of 250 ppm for C103. This is most likely due to the high specific surface area of the powders. Native surface oxides are directly incorporated into AM parts as interstitial contamination. Additional oxygen pickup can also occur in the imperfect inert LPBF chamber atmosphere. In Philips et al., the LPBF powder feedstock contained 283 ppm oxygen and the printed parts 370 ppm.[8] Mireles et al. reported 441 ppm oxygen in the powder and 1070 ppm in printed specimens after stress relieving and hot isostatic pressing (note the as-fabricated values were not reported).[10] In Awashti et al. the powder feedstock was 444 ppm oxygen, and the resulting parts showed 455 ppm oxygen.[11] The EB-PBF powders reported by Philips et al., on the other hand, have markedly lower initial oxygen levels, 180 ppm, and the solid parts, 240 ppm, both consistent with a typical wrought material. The lower initial oxygen content of the EB-PBF powder can be partially attributed to the larger powder typically used, nominally 45 to 150 μm, compared to 15 to 45 μm powder used in LPBF. For a given mass of spherical powder, the reduction of its diameter by half corresponds to an exponential, eightfold increase in the bulk powder surface area. It should be noted that this powder size distinction is an arbitrary convention rather than a necessity constrained by the physics of the process.

Nevertheless, it can be inferred that the vacuum environment of the EB-PBF process likely mitigates oxygen pickup in C103, as has been observed in other materials.[18,21] Previous studies have shown that water vapor adsorbs onto the powder surfaces during handling (e.g., machine changeovers, recycling). Subsequently, when the powder bed is heated during processing, this adsorbed water vapor dissociates, releasing oxygen, which oxidizes the powder.[21,22] Careful powder handling significantly reduces this effect.[23]

On the other hand, fabricating in a high-temperature, vacuum environment of EB-PBF leads to the vaporization of alloying elements with high vapor pressure. This phenomenon occurs in both laser and electron beam processing but is more pronounced in the latter due to the vacuum conditions.[24,25,26] This is especially problematic when there is a large disparity in the vapor pressures of elements in a given alloy. For instance, the loss of aluminum has been widely observed in EB-PBF of Ti6Al4V and TiAl alloys,[18,22,27,28] as well as the observed loss of zinc or magnesium from aluminum alloys.[29] In the previous study on EB-PBF of C103 by Philips et al., evaporative loss of titanium was observed, which was discussed but not investigated in detail.[8] The loss of titanium, which acts as a low-temperature solid solution strengthener, correlated well with observed property changes in the material. This suggests a need for further development of the process to elucidate and potentially counteract this loss.

In the context of the present work, Block-Bolten and Eagar[26] demonstrated in their study on gas tungsten arc welding, by utilizing the Langmuir equation, that the vaporization flux is dependent on the equilibrium vapor pressure, a variable that has not been widely explored for EB-PBF.

Given the documented loss of titanium and oxygen uptake during AM of C103, this study aims to further investigate its fabrication through EB-PBF. This research is motivated by the need to address the following objectives: (1) To better understand the evolution of ATI C103™ powder and solid specimens during the EB-PBF process over multiple uses, including changes in particle size distribution, the pickup of oxygen and the evaporation of alloying elements as a function of variations in the EB-PBF chamber atmosphere. (2) Building upon previous EB-PBF research, which indicated the feasibility of producing dense C103 parts with mechanical properties comparable to wrought products, this study also aims to characterize the bulk microstructural and mechanical characteristics of ATI C103™ produced via EB-PBF as a function of Ti loss. (3) to explore how typical additive manufacturing post-processing treatments, such as hot isostatic pressing (HIP) and annealing, influence the mechanical properties of EB-PBF produced C103.

2 Materials and Methods

2.1 Powder Characterization

Alloy ATI C103™ powder was produced by electrode induction gas atomization (EIGA) from wrought bar and mechanically classified through − 100/ + 325 mesh screen to a nominal 45 to 150 µm prior to the EB-PBF process. The composition of the powder and AM-produced specimens was determined by a combination of inductively coupled plasma optical emission spectroscopy and mass spectroscopy (ICP-OES/MS), X-ray fluorescence (XRF), inert gas fusion (IGF), and combustion analysis. The powder feedstock composition listed in Table I is representative of typical ATI C103™ and is within typical wrought specification limits.[30] Qualitative assessment of powder morphology was performed using a JOEL 6010LA SEM in backscatter composition mode (BEC) with energy dispersive x-ray spectroscopy (EDS) and the powder size distribution (PSD) was determined by laser diffraction on a Microtrac S3500 using a wet, ultrasonic technique.

Table I Chemical Composition of ATI C103 Powder Feedstock Used for EB-EBF Fabrication

2.2 Sample Fabrication and Characterization

All ATI C103™ AM specimens were fabricated using a customized Arcam A2 EB-PBF machine (software version 3.2, SP2). The build chamber, shown in Figures 1(a) and (b), was modified from the OEM hopper-style powder feeder to a build piston, feeder piston, and rake powder delivery system. Further details regarding this setup can be found elsewhere.[18,21,22,23,29,31] The AM process temperature was monitored using a type K thermocouple placed underneath the AM substrate and a dual-wavelength pyrometer (Fluke Endurance E1RL) through the front viewport to measure the surface temperature of the powder bed, shown in Figure 1(c). Prior to AM fabrication, the system vacuum chamber was evacuated to 8.4e−6 mBar, then backfilled with helium to 2.1e−3 mBar. Specimens were fabricated directly upon a titanium alloy (Ti6Al4V) substrate without supporting structures. The substrate measured 88 mm in diameter and 35 mm thick. Initially, a preheat pattern with a 9 mA beam current, a scan speed of 25,000 mm/s, and a diameter of 50 mm was scanned over the build substrate to raise the temperature to a range of 1350 to 1400 °C as measured by the dual-wavelength pyrometer. The beam current was decreased to 3 mA and maintained at that level for 30 minutes, while the temperature was allowed to reach stability, which was indicated by the dual-wavelength pyrometer, and the type K thermocouple readings, both stabilizing at 1100 °C ± 20 °C. The Arcam EB-PBF process consists of 2 process steps during each layer: preheating and melting. At the beginning of each layer, the table was lowered by 70 µm, and powder was deposited onto the build substrate from the feeder tank. Next, the powder bed was preheated using a defocused raster pattern (example shown in Figure 1(d)). The preheating was employed to maintain the elevated build temperature and lightly sinter the powder to establish an electrical pathway to ground. Each sample was then individually melted using a focused beam, and the beam raster direction was rotated 90 deg for each subsequent layer. In all experiments, the Arcam EB-PBF automatic melt mode was used. The specific parameters for the preheat and melt steps required to duplicate these experiments are displayed in Table II, and only the altered parameters from the standard, commercially available settings for Ti6Al4V are shown.

Fig. 1
figure 1

Photographs showing the custom build chamber with the powder feed cylinder on the left and the build cylinder on the right (a), the build surface showing the powder platform and Ti6Al4V substrate (b), the configuration of the pyrometer through the front viewport (c) and an example showing the powder bed preheating (d)

Table II List of EB-PBF Control Parameters Used to Fabricate Solid Articles

Two sets of experimental geometries were produced for the purpose of this study. In the initial trials, metallurgical specimens in the form of 16 mm cubes were produced to validate the repeatability of the processing conditions and specimen density of ATI C103™, which have been previously reported.[8] After each run, the chamber was backfilled with helium to accelerate cooling. When the temperature of the substrate, measured with a thermocouple, reached 22 °C, the chamber was ventilated to atmospheric pressure with dry argon. The powder, including the lightly sintered powder from the preheat region, was manually fragmented and passed through a 210 µm sieve within an argon glovebox. This coarse screening step aimed to remove any large agglomerated particles generated during the process and to blend this recovered material with unused powder from the powder feeder for subsequent runs. However, no new powder was introduced into the system after the initial build. Powder samples were collected for chemical and PSD analysis from this blended powder before each new run. Samples for metallurgical examination were sectioned with a low-speed SiC saw blade and hot mounted in conductive resin (Pace technologies “conducto 5p”). Using a Struers Pendimen S, the samples were ground with 320, 600, and 1000 grit SiC at 10N and 125 rpm for 2 minutes each. The sample surface was swabbed between each grit with an etching solution (10 ml each of HF, HNO3, H2O2, and H2O) for 15 seconds. Polishing was carried out at 40N load at 125 rpm for 4 minutes with DiaPro AllegroLargo 9, 9-micron diamond abrasive slurry, then with DiaPro MolR3, 3-micron diamond abrasive slurry using Struers MD-Mol surface plate, 25N and 125 rpm for 2 minutes. Final polishing with Struers OP-U non-dry 0.04-micron diamond slurry using Struers MD Chem surface plate, 10N, and 125 rpm for 2 minutes. The samples were cleaned in an ethanol ultrasonic bath between each polishing step for 5 minutes. For samples observed with optical microscopy, grain structure was revealed by using the same etchant as above by swabbing for 10 seconds, followed by DI water rinse, ethanol rinse, and hot air blow drying. Samples for SEM and EBSD were not etched. The microstructure of the samples was analyzed using a Hirox KH-7700 optical microscope to observe macro-scale features such as grain size, porosity, and process-generated layer defects. Optical relative density was measured with a contrast threshold optical technique (ImageJ) applied to unetched metallography samples. For selected specimens, absolute density was measured using a Quantachrome pycnometer with nitrogen at ambient test conditions; samples were obtained from the ends of each tensile bar prior to machining. Three samples of each post-treatment were tested (As-fabricated, annealed, and HIP). Finer microstructural features were observed using a JOEL 6010LA SEM with EDS, a FEI Verios 460L field-emission scanning electron microscope (FESEM), a ThermoFisher Quanta 3D FEG with electron backscatter diffraction (EBSD), and a ThermoFisher Helios Hydra Dual-beam FEG with EDS capabilities.

Subsequent builds produced three batches of EB-PBF cylinders, each consisting of 9 cylinders with a diameter of 14 mm and a height of 60 mm. Each build was carried out under different chamber vacuum levels. The Arcam EB-PBF control system features a function called ‘Controlled Vacuum’ (CV), which introduces a regulated amount of helium into the chamber to maintain a consistent vacuum level during the process. One batch was produced with a vacuum level of 8 × 10−2 mBar using the CV feature; another batch was produced with a lower vacuum level of 2 × 10−2 mBar using the CV feature, and the third batch with the CV feature disabled, resulting in a pressure level of 1 × 10−5 mBar. Each of these batches was carried out with new (fresh) powder. Samples from the 3 batches were then randomly assigned one of 3 post-processing conditions; as-fabricated, annealed, and HIP. Annealing was performed in a vacuum furnace at 1500 °C for 4 hours, and HIP was performed at 1593 °C at 207 MPa pressure for 4 hours. ASTM E8 sub-size tensile test specimens were harvested from the cylinders by CNC turning the cylinders with a 6 mm diameter and 12 mm length (Mazak Integrex 100i). The specimens were tested under uniaxial load conditions using an ATS model 1620C test machine equipped with a 20 kN load cell, controlled by displacement at a crosshead speed of 0.008 mm/s, resulting in a strain rate of 7e−4/s. The strain was measured using digital image correlation (DIC), with the gage surface painted white and marked with black speckles. A Point Grey Grasshopper with 12 MP resolution was recorded images at 2 frames per second, and the strain was analyzed using GOM Correlate software.

3 Results

3.1 Powder Feedstock and AM Sample Characterization

Figure 2(a) shows the volumetric powder size distribution of the starting ATI C103™ feedstock. The powder size ranges from d10 to d95 (10th to 95th percentiles) of 58 to 158 μm with a median size (d50) of 87 μm. The particles shown in Figures 2(b) and (c) suggest that the powder is primarily spherical, while Figure 2(d) shows an example of occasionally observed features like satellites and trapped gas porosity which are commonly observed in powders produced by gas atomization.[32,33]

Fig. 2
figure 2

Powder size distributions of ATI C103™ by volume (a) SEM image of ATI C103TM powder (b) and typical (c), and occasionally observed particle morphologies (d)

The images shown in Figure 3 present the typical geometries and microstructural characteristics of the ATI C103™ alloy at progressively smaller feature size scales to investigate the characteristics of EB-PBF melting of the alloy. Figure 3(a) shows an example of the 14 mm diameter cylindrical tensile specimens on the Ti6Al4V build substrate. Figure 3(b) shows an optical micrograph of an etched metallographic specimen sectioned in the XZ plane, while Figure 3(c) shows a BEC SEM image of the same sample at higher magnification. Optical analysis of these images shows a density of over 99 pct.

Fig. 3
figure 3

Photograph showing the 14 mm dia. x 60 mm cylindrical samples (a) Tiled optical microscopy (b) and backscatter FEG-SEM (c) of a dense as-fabricated microstructure, and EBSD results (d, e) showing tiles of nearly 16 mm across the sample at the bottom and middle regions represented by the yellow boxes in (b)

The results of the examination revealed a lack of prominent defects associated with the processing, as shown in Figure 2(c). The utilization of gas pycnometry revealed the existence of some micro-porosity in the as-manufactured and heat-treated specimens. The density measurements obtained from these analyses exhibited little variability, with values exceeding 99 pct relative density as a function of the applied post-processing conditions, with the HIP samples exhibiting the highest measured density.

The microstructure observed is typical for components manufactured using EB-PBF. The crystallographic texture shown in the EBSD analysis, presented in Figures 3(d) and (e), suggests that the primary growth occurred along the < 100 > directions, which is attributed to the solidification conditions and the heat extraction along the build direction.[19] The crystallographic texture is dominated by the presence of {001} grains in the reconstructed Inverse Pole Figure (IPF) maps and their associated pole figures in most of the sections analyzed in the sample. However, the presence of a mixed texture is observed along the edges of the sample. This is attributed to both the different solidification conditions due to roughness and thermal boundary conditions as well as fluctuations in the electron beam melting parameters in these regions. These results are similar to results from a recent C103 LPBF and EBM study where the fabricated articles reported approaching full density and demonstrated epitaxial grains in the Z-direction.[8]

3.2 Observations on the Effects of Powder Reuse

Maintaining uniform processing conditions and composition in C103 throughout the EB-PBF process is important to produce repeatable material performance, particularly when reusing the powder over multiple runs. In practice, likely sources of unintentional compositional changes during EB-PBF include improper powder handling and storage, oxygen and moisture pickup within the AM process, or elemental loss through evaporation.

As shown in Figure 4(a), the initial oxygen content in the powder was 180 ppm, which was within the ASTM specification limit of 250 ppm (indicated by the dotted line in Figure 4(a)).[34] The oxygen content increased over the repeated cycles of powder reuse until it reached 250 ppm after six cycles and ultimately peaked at 310 ppm (without refreshing with fresh powder). Nandwanda, et al. conducted a focused study on powder reuse in EB-PBF using Ni-718 and Ti6Al4V powders and found similar results to previous studies,[22,35] with a linear increase in oxygen content over five cycles of reuse without powder refresh.[22] Recall that the powder feedstock comprises the powder cake from the previous build cycle, unused powder from the feeder, and any material recovered from the raking overflow. A similar effect is observed in the data in Figure 4, which also shows the oxygen content of the preheat cake harvested after each run, which is consistently higher than the pre-build powder and consistently above the ASTM specification.[34] Such oxygen pickup in EB-PBF processes under typical operating conditions is challenging to avoid without refreshing or changing the powder[22,35] where the oxygen pickup is attributed to the handling of the powder and the release of adsorbed atmospheric water vapor from the vacuum chamber walls during build turnaround, which evaporates at elevated preheat temperatures and is captured by the reactive metal powder cake.

Fig. 4
figure 4

Oxygen content in the powder feedstock after consecutive uses (a), 10th, 50th, and 95th percentiles of the powder feedstock size distribution in count and volumetric modes over consecutive runs (b). ASTM B655/B655M (2018) specification limits[34]

The particle size data presented in Figure 4(b) were obtained by evaluating both the count-based and volume-based particle size distributions of each powder sample collected, starting from a new powder and then from samples collected from the powder bed before each EB-PBF build iteration up to a total of 10 builds without refreshing the powder, as described in the methodology. The particle size measurements (d10, d50, and d95) were plotted against the number of times the powders were reused. It is worth noting that the size distributions obtained through count-based measurements consistently exhibit a shift toward smaller particle sizes when compared to the distributions obtained through volume-based measurements, indicating the presence of finer particles. The shift in the d10 and d50 count distributions, which stabilized after the first three builds, can be attributed to the finest particles agglomerating or adhering to larger particles in the sinter cake during the initial use of the powder. The volume distribution is largely unaffected by a relatively small volume of fine particles (< 20 μm), where none of the size percentiles shift with the reuse maintaining a nominal size of 60 μm, 90 μm, and 150 μm, respectively. It is important to highlight that the PSD measurements did not show the accumulation of larger particles over reuse events due to the screening of agglomerates or spatter after each build completion with the 210 µm (70 mesh) screen, as described in the methodology.

3.3 Compositional Changes in EB-PBF Samples

Another factor affecting the composition of materials produced through EB-PBF is the evaporation that occurs in the vacuum environment during fabrication, which in this study resulted in decreased titanium content in the fabricated specimens. Table III presents the effect of operating chamber vacuum on the composition of ATI C103™ cylinders, using fresh powder before each run, to control for any influence that could result from powder reuse. Table III shows no measurable change in Hf concentration across different vacuum conditions. However, titanium decreases from 1.03 pct in the powder to 0.78 pct content in the specimens when fabricated under the low vacuum environment (Hi-CV) and stabilizes to 0.74 to 0.76 pct as the vacuum levels increase.

Table III Hafnium and Titanium Content (Measured by ICP) from Powder and Samples Fabricated Under Varying Chamber Vacuum Levels

During EB-PBF processing, evaporated species are deposited on the walls and viewport shutters of the vacuum chamber. These deposited films were carefully collected after a standard run under controlled vacuum (CV) conditions and analyzed by SEM/EDS as shown in Figure 5. These data corroborate the observed compositional shift in the bulk sample chemistry measurements in Table III. Two distinct types of metallic flakes were identified based on the titanium (Ti) content as determined by EDS mapping. One type was found to be highly enriched in Ti, while the other type had Nb and Hf contents that were more similar to those of the base alloy.

Fig. 5
figure 5

Metallic condensate collected from the EB-PBF heat shield after specimen fabrication. Green arrows indicate Ti-rich flakes, and the white arrows are flakes rich in Nb and Hf. The respective elements are identified in the bottom left of each EDS image

The darker flakes, which are enriched in Ti, are indicated by green arrows in Figs. 5(a) and (b) and are differentiated from the bright feedstock particles, which contain higher levels of Nb and Hf. EDS quantitative data show these low Z flakes contain a nominal 70 to 75 pct Ti, along with residual Nb and Hf. The white arrows identify a second type of condensation with higher Z in Figure 5, which contained a considerable amount of Hf and Nb. Quantitative EDS map data showed this population contained a nominal 50 pct Ti, 40 pct Nb, and 10 pct Hf. These data suggest that in both populations, Ti preferentially evaporates from the melt pool during EB-PBF fabrication as anticipated based on elemental vapor pressure.[36,37] It should be noted that flakes relatively rich in Hf and Nb are muted compared with adjacent feedstock particles due to the presence of elevated titanium present (and lower-Z) in all flakes.

3.4 Mechanical Properties of EB Fabricated Specimens

Room-temperature tensile properties for the ATI C103™ alloy fabricated by EB-PBF are shown in Table IV. Representative engineering stress-strain curves for the tensile tests are shown in Figure 6. The yield stress is calculated from a 0.2 pct strain offset. The ASTM E8 sub-size bars used for testing were harvested in the vertical build direction (z-dir) from cylinders represented in Figure 2(a) and fabricated with three build chamber conditions: control vacuum (CV), high control vacuum (Hi-CV), and no control vacuum (No-CV) to explore the influence of vacuum level on mechanical properties. Additionally, each of the samples printed at varying vacuum levels was tested in one of the three conditions: as-fabricated or post-process by HIP or vacuum annealed. Therefore, nine (9) tested conditions were used to determine the influence of varying EB process and post-processing conditions on mechanical properties.

Table IV Room-Temperature Tensile Properties ATI C103™ Fabricated by EB-PBF
Fig. 6
figure 6

Representative engineering stress-strain curve for tensile tests at room temperature for EB-PBF ATI C103™ alloy

The impact of post-processing conditions on the room-temperature mechanical properties is highlighted in Table IV, which shows that the as-fabricated and annealed specimens had the lowest yield stress and UTS values consistently below 270 MPa and 355 MPa, respectively. However, the hot isostatic pressing (HIP) specimens exhibited higher yield stress values of 281 MPa and UTS of 372 MPa. The standard deviations of the post-processed specimens were also very low, suggesting a less significant impact of chamber vacuum levels compared with HIP. Chamber pressure did not have a significant effect on mechanical properties. Using the reference density value of 8.85 g/cm3 for wrought material, all samples showed a high relative density of > 99 pct. The average relative density of the as-fabricated samples is 99.77 pct +/− 0.14, while the samples subjected to HIP are higher at 99.97 pct +/− 0.20 but still within error.

3.5 Microstructural Evaluation of Tensile Specimens

To understand the effects of EB-PBF process vacuum and post-processing conditions, fracture surfaces and microstructure in the X–Z plane were evaluated. The annealed and the as-fabricated conditions demonstrate a fibrous cup and cone fracture morphology with evident porosity at the 100 µm and 10 µm length scales (Figure 7 Rows 1 and 2). In comparison, the HIP condition fracture surfaces (Figure 7, row 3) reveal that the HIP specimen contained a blend of cup and cone fracture and areas with smooth regions.[39] While the as-fabricated specimens met the typical criterion for “fully dense” due to lack of apparent porosity in the etched macro-structures (Figure 2(b)), pycnometry indicated the presence of residual micro-porosity in the As-fabricated/CV and annealed/CV specimens compared to the HIP specimens which may have reduced the strength and failure strain. The application of HIP transitions the failure mode from void-driven failure to a fracture morphology significantly influenced by the columnar grains in the direction of loading and showing grain boundary delamination.

Fig. 7
figure 7

SEM images showing the fracture surfaces of tensile specimens at different magnifications. Specimens were fabricated in the standard vacuum configuration with different post-processing; as-fabricated (row A), vacuum annealed (row B), and HIP (row C)

The as-fabricated-CV specimens contain submicron grain boundary oxides, which typically range from 0.1 to 1 µm in size as shown in Figures 7(a) and 8(a). These oxides appear to have coarsened after the HIP and anneal due to the high post-processing temperatures. Even after coarsening, the oxide morphologies were typically angular, as demonstrated in Figure 7(c), and cylindrical, as observed from the high magnification fractography in Figure 7(b). The annealed-CV condition in Figures 7(b) and 8(b) demonstrated a bimodal size distribution of grain boundary oxides where significant coarsening of high-density oxide colonies was frequently observed in the fractography dimples.

Fig. 8
figure 8

SEM images showing grain boundary HfO2 for each post-processing condition: As-Fabricated (a), vacuum annealed, 4 h @ 1500 °C (b), and HIP, 4 h @ 1593 °C, 207 MPa (c). EDS maps for Nb and Hf are shown inset in (c)

4 Discussion

4.1 Modeling the EBM Process for Elemental Evaporation and Chemistry Effects During EB-PBF

The variation in Ti concentration in the fabricated solids is attributed to the preferential evaporation of Ti at the varying operating vacuum pressures. To explore this, a thermal model[38,40] was used to estimate the melt pool temperature during EB-PBF processing using C103 thermophysical data (ρ: 8580 kg/m3, Cp: 340 J/kg-K, k: 62 w/m2-K and e = 0.6) and the EB melt parameters of 18 mA beam current, 300 mm/s beam speed ,and hatch of 0.00025 m with a bed temperature at 1623 K.[30,41] Figure 9(a) shows the predicted melt pool temperatures for C103 where the nominal bed temperature of 1723 K (1350 °C) begins to increase as the beam raster approaches a region of interest (ROI) where the maximum temperature is predicted to exceed 3200 K in the center of the melt pool with model temperatures vary from 3250 K to 3677 K, depending on reported thermophysical properties. Partial vapor pressure of the alloy elements is shown in Figure 6(b) over a range of temperatures and reveals that while evaporation of all elemental components occurs during the process, as the melt pool is exposed to temperatures above 3273 K for full-density part fabrication, the evaporation of Nb and Hf is not expected to significantly impact the component composition, as observed in Table III. However, the data in Table III and prediction in Figures 9(c) and (b) show that while Hf remains largely unaffected by the EB AM process, Ti loss through evaporation is expected. In order to mature AM applications for alloys like C103, which consist of elements with a nearly 1500 K difference in boiling temperature, compositional change during AM must be addressed.

Fig. 9
figure 9

Thermal model of EB melting of ATI C103™ using parameters listed in Table II with the melt pool temperature calculated by an FEA thermal model of the EB-PBF process the arrow in (b) indicates the point of measurement for the temperature profile in (a), calculated elemental vapor pressure (c), and evaporation rate (d) during vaporization[42,43,45]

Evaporation rate (J) was also predicted using the Langmuir expression after Nawanda et al.,[22,42] and is shown in Figure 6(c) as a function of temperature, T:

$$ J = p_{i} \left( {2\pi MRT} \right)^{{\frac{ - 1}{2}}}, $$

where M is the molecular mass of the evaporating atomic monomer for each species, and R is the universal gas constant.[43,43,44,45,46,47,48] The CALPHAD model predicts activity coefficients between 1 and 1.5 for the constituents in the liquid phase C103 and suggests that the Ti vapor pressure reaches 2 orders of magnitude higher than Hf and Nb at temperatures predicted by the thermal model. While the predicted rate of evaporation for Hf and Nb at 3200 K is relatively similar at just above 0.005 kg/m2sec, the Ti evaporation rate was calculated over 0.5 kg/m2s. This model suggests that the melt pool approaches the atmospheric boiling temperature of Ti (3637 K)[45] and that preferential loss of Ti is consistent with the much higher boiling points of Hf and or Nb (Tb = 4893 K, 5200 K, respectively).

Similar recent results indicated that Ti loss did not noticeably occur in LPBF. However, oxygen increased to 370 ppm, while EB-PBF fabricated articles lost 0.14 wt pct Ti.[8] The key differentiator in the Ti loss during EB-PBF vs LPBF is that the vacuum environment dramatically increases Ti transport away from the pool. Evaporation also occurs in LPBF, but the much lower transport rates, due to the slow-moving argon atmosphere (relative to the beam raster velocity), allow the establishment of a quasi-equilibrium vapor composition over the melt pool, which limits evaporation. This understanding of the process is in agreement with the observed loss of 0.2 pct to 0.25 pct Ti in the printed component and the presence of a metallic condensate enriched with titanium, as depicted in Figure 5. The process requirements to avoid layer defects and macro-scale porosity during fabrication of ATI C103™ drive melt pool temperature and size to achieve overlap during rastering should be expected to cause some titanium loss in the vacuum environment. Balancing composition and operating conditions are critical to maintain reasonable levels of titanium yet counteract the potentially detrimental effects of oxygen pickup. The elevated oxygen content will most likely appear on the grain boundaries as HfO2.

Additive manufacturing of alloy C103 has been reported using LPBF and EB-PBF with varying mechanical property results. Table V summarizes a number of AM-related studies reporting mechanical properties data for alloy C103. Recent studies using laser PBF reported higher strength than observed herein but with elongation consistently at or below 20 pct[10,11] in one case and 35 to 37 pct, depending on specimen orientation and post-processing conditions in another.[8]

Table V Mechanical Test Results Reported in the Literature Compared with AVG and HIP Mechanical Results from the Present EB-PBF Study

In comparison to the present study, a recent report describing EB-PBF to fabricate articles for room-temperature tensile testing with reported average values of 281 MPa yield, 354 MPa UTS, and 34 pct elongation,[8] with some differences in data depending on vertical or horizontal harvesting orientations. The data in Table V show that the average of all specimens reported in the present study exceeded the B655 specification limits for yield strength and strain (pct); however, EBM specimens in the current study without HIP post-processing tended to have below-average yield strengths. The specimens in the present study post-processed by HIP well exceeded the YS and strain, where both data sets, with and without HIP, struggled to exceed the specification limit for UTS. Similar test results for EBM fabricated specimens were reported by Philips et al.,[8] for both X–Y and Z-dir properties, as shown in Table V as averages of the data set. The EB- fabricated data sets reported very similar YS & UTS; however, elongation in the current study is consistently lower than those reported in the literature, especially in the Z-direction ambient tensile results.[8] It should be noted that, a sub-size gage cross section in Reference 8 may affect these results. The reduced ultimate strength in both EBM data sets is most likely due to the presence of epitaxial grains in the direction of tensile deformation in comparison with equiaxed structures. In the context of EB-PBF in general, the UTS is not consistently lower relative to wrought materials. The most common example for EB-PBF, due to its widespread use, is Ti-6Al4V which exhibits higher YS and UTS compared to its wrought counterpart, albeit with reduced elongation along the build direction.[49] An important distinction is that while both EB-PBF Ti6Al4V and C103 exhibit columnar microstructures, Ti–6Al–4V processed through EB-PBF often demonstrates superior strength to wrought Ti–6Al–4V, likely due to the PBF-induced microstructures resembling a quenched state, characterized by finely interspersed α + β laths or complex Widmanstätten structures within the columnar grains.[50] For pure metals such as EB-PBF of Cu, the UTS and Yield have been reported to be similar to annealed copper but are influenced by porosity, oxygen content, and residual stress, in addition to significant differences in ductility between transverse and epitaxial loading conditions.[50] In this study, the ASTM B655-10 YS and UTS properties of wrought materials were similar to those of EB-PBF C103, particularly when subjected to HIP. As previously noted, the As-fabricated/CV and annealed/CV specimens ranged from 99.25 to 99.77 pct relative density as compared to the HIP specimens at 99.97 pct. Therefore, it appears that residual micro-porosity in the as-fabricated and annealed specimens significantly reduced the strength and failure strain. A second influence in lower ultimate strength may be the loss of Ti as a solid strength solution element in the EB fabricated specimens, which was reported from 0.899 wt pct[8] to a minimum measured value of 0.74 wt pct in the present study, where small amounts of titanium can significantly increase the UTS in Nb.[51] In the present study, oxides, primarily located along grain boundaries, as shown in Figures 7 and 8, are unlikely to notably impact strength. These oxides, while often submicron in size, are situated on the grain boundaries rather than being intergranular as is seen in powder metal HIP and wrought preparations wherein the particles are typically moved off grain boundaries during thermo-mechanical processing and recrystallization.[41,52,53] Consequently, in EB-PBF prepared articles, their presence is expected to have minimal effect on dislocation pinning and the resultant strengthening of the material.[5,8,52] The elimination of porosity after HIP, rather than the dispersed oxides, appears to be the dominant influence on mechanical properties. Nevertheless, the additional factors of lower Ti in solution due to evaporation, HfO2 at grain boundaries reducing Hf solid solution strengthening, and columnar semi-epitaxial grains contributed to the reduction of ultimate tensile stress measured in the present study.

5 Conclusions

These results highlight the potential for process optimization to further enhance the material performance and pave the way for the wider use of electron beam powder bed fusion in industrial applications.

  • Control EB-PBF environment to minimize the superheat and dwell to retain Ti in the alloy. Control of vacuum process pressure showing minimal effect on evaporative loss of titanium.

  • Process optimization to further increase density beyond 99.5 pct and/or apply Hot Isostatic Pressing to mitigate the reduction of strength due to porosity defects.

  • As with most powder bed materials, anisotropic microstructure can reduce performance, and process control to reduce elongated columnar grains is warranted.

The results obtained in this study are significant as they demonstrate the feasibility and limitations of using EB-PBF to fabricate high-density articles in ATI C103™. The microstructure and texture are consistent with other reports from the EB-PBF literature, including recent reports regarding refractory metals.[8,9,10,11,13] The results also show that the oxygen content of the powder remained below the specification limit, even after several recycling uses. However, it is important to note a limitation of this study, which is primarily related to the limited quantity of powder available for experimentation. Specifically, mechanical testing was not performed on these samples due to the small sample sizes used to evaluate reuse and oxygen pickup. Consequently, it remains unclear whether the oxygen specification limit, which is applicable to wrought material, is also appropriate for AM-processed material. Moreover, the impact of exceeding this limit on the mechanical properties of the AM-processed material has yet to be determined. Nevertheless, the results suggest that powder can be reused multiple times without exceeding the existing specification limits.

  • Oxygen content remained below the 250 ppm specification limit up to 6 recycling uses, reaching a maximum of 310 ppm after 10 cycles without refresh.

  • The mechanical properties of the ATI C103™ EB-PBF articles under normal CV EB-PBF operating conditions had an average yield stress of 261 MPa, ultimate tensile strength of 358 MPa, and 28 pct elongation.

  • The highest mechanical strengths were achieved after hot isostatic pressing (HIP) removal of porosity, with a yield stress of 281 MPa, an ultimate tensile strength of 372 MPa, and 27 pct elongation, all meeting or exceeding the specification limits.