Characteristics and Processing of Hydrogen-Treated Copper Powders for EB-PBF Additive Manufacturing

: The fabrication of high purity copper using additive manufacturing has proven di ﬃ cult because of oxidation of the powder feedstock. Here, we present work on the hydrogen heat treatment of copper powders for electron beam powder bed fusion (EB-PBF), in order to enable the fabrication of high purity copper components for applications such as accelerator components and vacuum electronic devices. Copper powder with varying initial oxygen contents were hydrogen heat-treated and characterized for their chemistry, morphology, and microstructure. Higher initial oxygen content powders were found to not only reduce surface oxides, but also reduce oxides along the grain boundaries and form trapped H 2 O vapor inside the particles. The trapped H 2 O vapor was veriﬁed by thermogravimetric analysis (TGA) and residual gas analysis (RGA) while melting. The mechanism of the H 2 O vapor escaping the particles was determined by in-situ SEM heated stage experiments, where the particles were observed to crack along the grain boundaries. To determine the e ﬀ ect of the EB-PBF processing on the H 2 O vapor, the thermal simulation and the validation of single melt track width wafers were conducted along with melting single layer discs for chemistry analysis. A high speed video of the EB-PBF melting was performed in order to determine the e ﬀ ect of the trapped H 2 O vapor on the melt pool. Finally, solid samples were fabricated from hydrogen-treated copper powder, where the ﬁnal oxygen content measured ~50 wt. ppm, with a minimal residue hydrogen content, indicating the complete removal of trapped H 2 O vapor from the solid parts.


Introduction
Applications including particle accelerators and vacuum electronic devices (VEDs) require materials with the highest electrical and thermal conductivity, as well as ultra-high vacuum compatibility. The copper used in these applications approaches the theoretical maximum achievable quality in terms of the purity, density, and metallurgical properties, such as crystallographic texture and grain size. These applications also require intricate designs and extensive metallurgical processing routes, followed by the assembly and brazing or welding of multiple components into a final part. The consolidation of the component assemblies into complex monolithic parts, and the potential for novel designs achievable through additive manufacturing (AM), is desired in order to reduce the cost and

Materials and Methods
The experiments were performed in order to understand the influence of a hydrogen heat treatment on copper powder feedstocks of varying initial oxygen contents. We evaluated three levels of oxygen content, with and without hydrogen heat treatment. To generate these six conditions, the following methodology was used.
Two separate batches of nitrogen gas atomized copper powder were acquired. For the purposes of this discussion, the powders were categorized based on the initial measured oxygen content. The lowest oxygen content powder received was between 100-200 wt. ppm oxygen, and is referred to as low oxygen copper (LO-Cu). The second powder batch with 400-700 wt. ppm oxygen is referred to as medium oxygen content powder (MO-Cu). A selected amount of the received MO-Cu powder was intentionally oxidized to above 1000 wt. ppm oxygen, and is referred as high oxygen content copper (HO-Cu).

Feedstock Preparation
Spherical MO-Cu powder was received with a nominal 15-53 µm particle size distribution (d10-d90), smaller than what is typical for other materials used in EB-PBF (e.g., Ti6Al4V). Nevertheless, it is consistent with previous studies in copper AM by our group [4], and compared to early work by our group and others, also results in a significantly reduced surface roughness [1]. In order to maintain a comparatively low oxygen content in the LO-Cu powder, special consideration was given to the packaging, handling, and screening steps post atomization. This powder was received in the as-atomized (−250 µm) condition, directly from the manufacturer, and screened to a nominal 15-53 µm particle size distribution within an argon glove box. To generate the HO-Cu powder, a methodology similar to that of Nieh and Nix [28,30] was utilized. A portion of MO-Cu was treated at 150 • C for 1 h in one atmosphere of air, and cooled at a rate of 5 • C per minute. The temperature was kept lower in the current study, compared with the 800 • C used by Nieh and Nix, to prevent the sintering of the small powder size distribution. Given that the diffusion of oxygen in copper is 2e −11 m 2 /s at 400 • C [39], the oxygen penetration depth exceeded the particle diameters used in this study during the 2-h treatment, which is confirmed by subsequent focused ion beam (FIB) sectioning of individual powder particles.
Portions of the powder from each of the three batches, LO-Cu, MO-Cu, and HO-Cu, were then exposed to hydrogen treatment at 400 • C for 4 h under one atmosphere of flowing pure hydrogen, and cooled at a rate of 5 • C per minute in a dry hydrogen atmosphere heating cover retort furnace. A custom stainless-steel fixture comprised of stacked circular trays allowed the processing of 1 kg of powder per treatment cycle in 2 mm thick layers, as shown in Figure 1. To avoid oxygen contamination after hydrogen treatment, the powders were packed in vacuum-sealed foil bags after purging with dry nitrogen. The vacuum bags were then sealed in a second bag filled with dry nitrogen. All of the copper powders were subsequently stored, handled, and sampled in an argon or nitrogen atmosphere.

Feedstock Preparation
Spherical MO-Cu powder was received with a nominal 15-53 µ m particle size distribution (d10-d90), smaller than what is typical for other materials used in EB-PBF (e.g., Ti6Al4V). Nevertheless, it is consistent with previous studies in copper AM by our group [4], and compared to early work by our group and others, also results in a significantly reduced surface roughness [1]. In order to maintain a comparatively low oxygen content in the LO-Cu powder, special consideration was given to the packaging, handling, and screening steps post atomization. This powder was received in the as-atomized (−250 µ m) condition, directly from the manufacturer, and screened to a nominal 15-53 µ m particle size distribution within an argon glove box. To generate the HO-Cu powder, a methodology similar to that of Nieh and Nix [28,30] was utilized. A portion of MO-Cu was treated at 150 °C for 1 h in one atmosphere of air, and cooled at a rate of 5 °C per minute. The temperature was kept lower in the current study, compared with the 800 °C used by Nieh and Nix, to prevent the sintering of the small powder size distribution. Given that the diffusion of oxygen in copper is 2e −11 m 2 /s at 400 °C [39], the oxygen penetration depth exceeded the particle diameters used in this study during the 2-h treatment, which is confirmed by subsequent focused ion beam (FIB) sectioning of individual powder particles.
Portions of the powder from each of the three batches, LO-Cu, MO-Cu, and HO-Cu, were then exposed to hydrogen treatment at 400 °C for 4 h under one atmosphere of flowing pure hydrogen, and cooled at a rate of 5 °C per minute in a dry hydrogen atmosphere heating cover retort furnace. A custom stainless-steel fixture comprised of stacked circular trays allowed the processing of 1 kg of powder per treatment cycle in 2 mm thick layers, as shown in Figure 1. To avoid oxygen contamination after hydrogen treatment, the powders were packed in vacuum-sealed foil bags after purging with dry nitrogen. The vacuum bags were then sealed in a second bag filled with dry nitrogen. All of the copper powders were subsequently stored, handled, and sampled in an argon or nitrogen atmosphere.

Feedstock Characterization
Each of the three powder conditions were characterized before and after hydrogen treatment, for size distribution by laser diffraction (Microtrac S3500, Microtrac Inc., Montgomeryville, PA, USA), and density by helium pycnometry (ASTM B923-16, Quantachrome MicroUltrapyc 1200e, Quantachrome Instruments, Boynton Beach, Florida, USA). The oxygen and hydrogen content of all the powders were quantified by inert gas fusion analysis (LECO OHN836, LECO, St. Joseph, MI, USA). The reported values herein were acquired immediately prior to the experiments. The powder morphology was characterized by a JOEL 6010LA scanning electron microscope (JEOL USA, Inc., Peabody, MA, USA). An additional powder characterization was performed on individual powder particles using focused ion beam (FIB) milling for the cross-sectional analysis by secondary electron and gallium ion contrast on a FEI Quanta 3D Field Emission Gun (FEG) (Thermo Fisher Scientific,

Feedstock Characterization
Each of the three powder conditions were characterized before and after hydrogen treatment, for size distribution by laser diffraction (Microtrac S3500, Microtrac Inc., Montgomeryville, PA, USA), and density by helium pycnometry (ASTM B923-16, Quantachrome MicroUltrapyc 1200e, Quantachrome Instruments, Boynton Beach, Florida, USA). The oxygen and hydrogen content of all the powders were quantified by inert gas fusion analysis (LECO OHN836, LECO, St. Joseph, MI, USA). The reported values herein were acquired immediately prior to the experiments. The powder morphology was characterized by a JOEL 6010LA scanning electron microscope (JEOL USA, Inc., Peabody, MA, USA). An additional powder characterization was performed on individual powder particles using focused ion beam (FIB) milling for the cross-sectional analysis by secondary electron and gallium ion contrast on a FEI Quanta 3D Field Emission Gun (FEG) (Thermo Fisher Scientific, Hillsboro, Oregon, USA).
Scanning transmission electron microscopy (STEM) analysis and energy dispersive spectroscopy (EDS) were performed on a FEI Talos (Thermo Fisher Scientific, Hillsboro, Oregon, USA), to assess the chemical composition, size of surface, and internal oxide species.
A thermogravimetric analysis (TGA) was performed using a TA SDT 650 Simultaneous Thermal Analyzer (TA Instruments, New Castle, DE, USA) under flowing nitrogen, while the samples were heated at 5 • C/s to 600 • C in an alumina sample holder in order to identify the temperature at which the H 2 O can escape from the treated copper powders.

In-Situ Heated Stage
In-situ heated stage SEM degassing experiments were performed on a FEI Quanta 3D FEG so as to observe the mechanism of H 2 O gas release from a single copper powder particle. Both hydrogen heat treated copper particles and untreated copper particles were separated from the adjacent powder feedstock with a single fiber brush using a light microscope at 50× magnification. Individual powder particles were then placed onto a SiC heated stage (Protochips Fusion) for controlled heating under standard SEM operating conditions. The stage was ramped to a maximum temperature of 400 • C at 100 • C/s, and image sequences were made at 30 fps with secondary electron detectors.

Customized EB-PBF Platform
All of the electron beam melting (EBM) experiments were conducted using a customized Arcam A2 EB-PBF system (Arcam Build Control Software V3.2, SP2, Arcam AB, Sweden). While the EB-PBF operation and process parameters have been described in detail elsewhere [40], it is important to highlight that in this work, the internal components of the commercial system have been removed and replaced with a custom 304 stainless steel build tank, platform, and powder feeder system. The customized chamber was designed with viewports that are oriented radially about the vertical axis of the electron beam column at a 55 • angle to the surface of the experimental substrate/fixture, in order to view the entire build surface. The experiments were monitored with by an emissivity corrected single color pyrometer (Fluke Process Endurance Pyrometer, Fluke, Everett, WA, USA), a visible light camera (Nikon D7000, Nikon, Melville, NY, USA), a 200-amu residual gas analyzer (Stanford Research Systems, Sunnyvale, CA, USA), and an IR thermal imaging camera (FLIR A655sc, FLIR, Wilsonville, OR, USA). The viewport window materials varied as a function of the transmittance requirements for each instrument. Figure 2 shows a photograph of the customized setup. Hillsboro, Oregon, USA). Scanning transmission electron microscopy (STEM) analysis and energy dispersive spectroscopy (EDS) were performed on a FEI Talos (Thermo Fisher Scientific, Hillsboro, Oregon, USA), to assess the chemical composition, size of surface, and internal oxide species. A thermogravimetric analysis (TGA) was performed using a TA SDT 650 Simultaneous Thermal Analyzer (TA Instruments, New Castle, DE, USA) under flowing nitrogen, while the samples were heated at 5 °C /s to 600 °C in an alumina sample holder in order to identify the temperature at which the H2O can escape from the treated copper powders.

In-Situ Heated Stage
In-situ heated stage SEM degassing experiments were performed on a FEI Quanta 3D FEG so as to observe the mechanism of H2O gas release from a single copper powder particle. Both hydrogen heat treated copper particles and untreated copper particles were separated from the adjacent powder feedstock with a single fiber brush using a light microscope at 50× magnification. Individual powder particles were then placed onto a SiC heated stage (Protochips Fusion) for controlled heating under standard SEM operating conditions. The stage was ramped to a maximum temperature of 400 °C at 100 °C /s, and image sequences were made at 30 fps with secondary electron detectors.

Customized EB-PBF Platform
All of the electron beam melting (EBM) experiments were conducted using a customized Arcam A2 EB-PBF system (Arcam Build Control Software V3.2, SP2, Arcam AB, Sweden). While the EB-PBF operation and process parameters have been described in detail elsewhere [40], it is important to highlight that in this work, the internal components of the commercial system have been removed and replaced with a custom 304 stainless steel build tank, platform, and powder feeder system. The customized chamber was designed with viewports that are oriented radially about the vertical axis of the electron beam column at a 55° angle to the surface of the experimental substrate/fixture, in order to view the entire build surface. The experiments were monitored with by an emissivity corrected single color pyrometer (Fluke Process Endurance Pyrometer, Fluke, Everett, WA, USA), a visible light camera (Nikon D7000, Nikon, Melville, NY, USA), a 200-amu residual gas analyzer (Stanford Research Systems, Sunnyvale, CA, USA), and an IR thermal imaging camera (FLIR A655sc, FLIR, Wilsonville, OR, USA ). The viewport window materials varied as a function of the transmittance requirements for each instrument. Figure 2 shows a photograph of the customized setup.

Simulation and Validation of Variable Spacing Single Melt Track Width Wafers
We hypothesized that during melting, H 2 O vapor escapes from both the liquid melt pool and from powder in the heat affected zone. For the powder adjacent to the melt pool, the internal pressure of the H 2 O vapor overcomes the mechanical strength of the grain boundary, and the pore begins to expand in size. This leads to the fracturing of the particle along the grain boundaries and a release of H 2 O vapor into the vacuum system. To explore this, single melt track width scenarios were simulated at various hatch spacings, utilizing the 3D transient heat transfer model developed by Lee et al. [41] at the Oak Ridge National Laboratory. Calculations were made using FEniCS software (2019.1.0, FEniCS), utilizing python scripts, and the thermal model output was analyzed using ParaView visualization software (v5.7, Kitware Inc., Clifton Park, NY, USA). The simulation used variables including heat input, beam speed, thermal conductivity, and so on, to model the EB-PBF process and the temperature profile across a solid plate with the thermophysical properties of a comparable copper powder bed during different operating conditions. The simulation variables used in this study are listed in Appendix A.
Based on the model results, single melt track width wafers 13 mm long and 10 mm tall were fabricated using hydrogen-treated HO-Cu powder with 2.0 mm on center spacings, as shown in Figure 3. The build substrate was a machined OFE copper substrate that measured 90 mm in diameter and 30 mm in thickness. During the EB-PBF processing, the build temperature was maintained at 270 • C, measured by a thermocouple attached to the bottom of the plate, and the partial pressures of H 2 O vapor and other gas species were monitored using the RGA. Each wafer was melted with a line energy of 0.8 J/mm (600 mm/s, 8 mA, 0 mA focus offset) and layer thickness of 40 µm, with the power analyze function (Arcam specific thermal controls) disabled. The beam scan direction was constant (right to left) for each wafer and each layer.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 6 of 22 We hypothesized that during melting, H2O vapor escapes from both the liquid melt pool and from powder in the heat affected zone. For the powder adjacent to the melt pool, the internal pressure of the H2O vapor overcomes the mechanical strength of the grain boundary, and the pore begins to expand in size. This leads to the fracturing of the particle along the grain boundaries and a release of H2O vapor into the vacuum system. To explore this, single melt track width scenarios were simulated at various hatch spacings, utilizing the 3D transient heat transfer model developed by Lee et al. [41] at the Oak Ridge National Laboratory. Calculations were made using FEniCS software (2019.1.0, FEniCS), utilizing python scripts, and the thermal model output was analyzed using ParaView visualization software (v5.7, Kitware Inc., Clifton Park, NY, USA). The simulation used variables including heat input, beam speed, thermal conductivity, and so on, to model the EB-PBF process and the temperature profile across a solid plate with the thermophysical properties of a comparable copper powder bed during different operating conditions. The simulation variables used in this study are listed in Appendix A.
Based on the model results, single melt track width wafers 13 mm long and 10 mm tall were fabricated using hydrogen-treated HO-Cu powder with 2.0 mm on center spacings, as shown in Figure 3. The build substrate was a machined OFE copper substrate that measured 90 mm in diameter and 30 mm in thickness. During the EB-PBF processing, the build temperature was maintained at 270 °C , measured by a thermocouple attached to the bottom of the plate, and the partial pressures of H2O vapor and other gas species were monitored using the RGA. Each wafer was melted with a line energy of 0.8 J/mm (600 mm/s, 8 mA, 0 mA focus offset) and layer thickness of 40 µm, with the power analyze function (Arcam specific thermal controls) disabled. The beam scan direction was constant (right to left) for each wafer and each layer. Upon the build completion, each block of wafers was removed from the surrounding sintered powder and build surface using a razor blade. The sintered powder cake between the solid wafers was serially sectioned with a razor blade as a function of the distance from the melt pool, and placed on carbon tape for the SEM analysis. Additionally, the top surface of the intact sintered cake was analyzed on a JOEL 6010LA, in order to observe sintered powder characteristics at specific distances from the melt track.

EB-PBF Processing of Hydrogen-Treated Copper Powder
To directly observe and quantify the H2O release during EB-PBF, single layer discs were melted using the various hydrogen-treated and untreated powders. Figure 4 illustrates the experiment. The fixture upon which the experiments were carried out was an annealed OFE copper cylinder, 100 mm in diameter and 50 mm thick. Six equally spaced cylindrical pockets, measuring 25 mm in diameter by 3 mm deep were milled into the surface at a radial distance of 30 mm from the plate center, such that each pocket is effectively isolated from the others, as seen in Figure 4. After the machining, the Upon the build completion, each block of wafers was removed from the surrounding sintered powder and build surface using a razor blade. The sintered powder cake between the solid wafers was serially sectioned with a razor blade as a function of the distance from the melt pool, and placed on carbon tape for the SEM analysis. Additionally, the top surface of the intact sintered cake was analyzed on a JOEL 6010LA, in order to observe sintered powder characteristics at specific distances from the melt track.

EB-PBF Processing of Hydrogen-Treated Copper Powder
To directly observe and quantify the H 2 O release during EB-PBF, single layer discs were melted using the various hydrogen-treated and untreated powders. Figure 4 illustrates the experiment. The fixture upon which the experiments were carried out was an annealed OFE copper cylinder, 100 mm in diameter and 50 mm thick. Six equally spaced cylindrical pockets, measuring 25 mm in diameter by 3 mm deep were milled into the surface at a radial distance of 30 mm from the plate center, such that each pocket is effectively isolated from the others, as seen in Figure 4. After the machining, the copper substrate was cleaned using a standard ultra-high vacuum (UHV) protocol for machined copper.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 7 of 22 copper substrate was cleaned using a standard ultra-high vacuum (UHV) protocol for machined copper. Figure 4. Illustration of the in-situ H2O vapor out-gassing experiment. The electron beam melting (EBM) vacuum chamber is outfitted with a residual gas analysis (RGA) and IR camera, and the surface temperature is monitored by a two-color pyrometer and thermocouple (A). The base fixture contains six 3 mm deep pockets (B), which are filled with hydrogen-treated and untreated powders with varying initial oxygen content. A single layer disc is melted on the top surface of each powder type. During melting, H2O is released from the powder or melt pool, and detected by the RGA (C). Upon completion, the powder is manually removed from the solid disc of copper, which is analyzed for oxygen and hydrogen content by inert gas fusion (D).
Each of the pockets in the substrate were filled and leveled to the top surface of the plate with ~5 grams of LO-Cu, MO-Cu, or HO-Cu powder before or after hydrogen treatment, each having a replicate. The oxygen and hydrogen content of the powder, as noted previously, was measured by inert gas fusion prior to this experiment. A type K thermocouple was affixed to the bottom of the substrate, and the substrate was placed on top of the build piston, which allows for the thermal isolation from the system.
The vacuum chamber was evacuated to a pressure of 3.75 × e −5 Torr, and the copper substrate was heated by repeatedly scanning the defocused electron beam in a circular area, 25 mm in diameter at the center of the substrate, with a beam current of ~2-4 mA. The substrate temperature was measured by the thermocouple at the bottom, and a two-color pyrometer through a sapphire window positioned in one of the upper viewports.
The substrate heating was manually modulated until both of the temperature readings agreed within 5% of the target value (300 °C ), and then this temperature was maintained for a 2 h outgassing cycle. A 200 amu residual gas analyzer (SRS-RGA200) was used to measure the partial pressure of the gas species of interest for this study (O, H, N, NO, NO2, and H2O). During the outgassing cycle, the background H20 partial pressure was observed to decrease and stabilize at ~8 × e −7 Torr. This outgassing period at 300 °C also allowed sufficient time for the initial stages of sintering to take place within the powder. This was observed with the IR camera by a uniform emissivity shift across all six powder pockets.
The focused electron beam was then raster scanned over an area measuring 20 mm in diameter on the surface of each one of the powder pockets. A line energy of 0.8 J/mm (600 mm/s, 8 mA, 0 mA focus offset, 100 µ m hatch offset) was sufficient to melt and consolidate a single layer of copper supported underneath by the sintered powder. The melting process was recorded by the thermal imaging camera through a germanium port at 25 Hz, and by in-situ monitoring of the back scattered electron signal. A thermal video of the melting process is available in the supplemental data. During the melting, the outgassing of the key gas species was monitored by the RGA at 1 Hz. This process The electron beam melting (EBM) vacuum chamber is outfitted with a residual gas analysis (RGA) and IR camera, and the surface temperature is monitored by a two-color pyrometer and thermocouple (A). The base fixture contains six 3 mm deep pockets (B), which are filled with hydrogen-treated and untreated powders with varying initial oxygen content. A single layer disc is melted on the top surface of each powder type. During melting, H 2 O is released from the powder or melt pool, and detected by the RGA (C). Upon completion, the powder is manually removed from the solid disc of copper, which is analyzed for oxygen and hydrogen content by inert gas fusion (D).
Each of the pockets in the substrate were filled and leveled to the top surface of the plate with 5 grams of LO-Cu, MO-Cu, or HO-Cu powder before or after hydrogen treatment, each having a replicate. The oxygen and hydrogen content of the powder, as noted previously, was measured by inert gas fusion prior to this experiment. A type K thermocouple was affixed to the bottom of the substrate, and the substrate was placed on top of the build piston, which allows for the thermal isolation from the system.
The vacuum chamber was evacuated to a pressure of 3.75 × e −5 Torr, and the copper substrate was heated by repeatedly scanning the defocused electron beam in a circular area, 25 mm in diameter at the center of the substrate, with a beam current of~2-4 mA. The substrate temperature was measured by the thermocouple at the bottom, and a two-color pyrometer through a sapphire window positioned in one of the upper viewports.
The substrate heating was manually modulated until both of the temperature readings agreed within 5% of the target value (300 • C), and then this temperature was maintained for a 2 h outgassing cycle. A 200 amu residual gas analyzer (SRS-RGA200) was used to measure the partial pressure of the gas species of interest for this study (O, H, N, NO, NO 2 , and H 2 O). During the outgassing cycle, the background H 2 0 partial pressure was observed to decrease and stabilize at~8 × e −7 Torr. This outgassing period at 300 • C also allowed sufficient time for the initial stages of sintering to take place within the powder. This was observed with the IR camera by a uniform emissivity shift across all six powder pockets.
The focused electron beam was then raster scanned over an area measuring 20 mm in diameter on the surface of each one of the powder pockets. A line energy of 0.8 J/mm (600 mm/s, 8 mA, 0 mA focus offset, 100 µm hatch offset) was sufficient to melt and consolidate a single layer of copper supported underneath by the sintered powder. The melting process was recorded by the thermal imaging camera through a germanium port at 25 Hz, and by in-situ monitoring of the back scattered electron signal. A thermal video of the melting process is available in the Supplementary Materials. During the melting, the outgassing of the key gas species was monitored by the RGA at 1 Hz. This process was repeated on each powder pocket with approximately 10 min to pass between each raster melt in order to allow for the RGA partial pressures and substrate temperature to stabilize. The solids samples were then cooled under a vacuum to room temperature.
The material in each pocket was removed from the substrate and the sintered powder was mechanically removed from the solid wafer with a wire brush in an inert gas glove box. The oxygen and hydrogen content of both the sintered powder and solid wafer from each pocket were measured by inert gas fusion (LECO Form 203-821-455, LECO OHN836, LECO, St. Joseph, MI, USA). The experiment was repeated several times, varying the position of each powder type and the order in which powders were melted.

High Speed Footage of Melt Pool
In order to directly observe the H 2 O vapor escaping from the melt pool, a Photron SA-X2 high-speed camera was utilized to capture the footage at 30,000 frames per second, at a focal length of 30 cm. The footage was captured for both a hydrogen-treated LO-Cu and HO-Cu melt pool using the same procedure for the EB-PBF processing as described in Section 2.6.

Solid Sample Fabrication
To validate the process for fabricating the components of high purity copper with hydrogen heat-treated powers, solid cylindrical specimens measuring 15 mm in diameter were fabricated for the microstructural and chemical analysis, using both untreated and hydrogen-treated MO-Cu powder. The untreated powder had an oxygen content (~600 wt. ppm) similar to what was used in the studies by Frigola et al. [1], Lodes et al. [8], and Guschlbauer et al. [9]. The processing parameters used were 600 mm/s, 8 mA, 18 mA focus offset, and 120 µm hatch offset with a 40 µm layer thickness. This falls within the processing space (~166 J/mm 3 ) used by both Frigola et al. [1] and Guschlbauer et al. [9]. The build temperature was maintained at 300 ± 10 • C for the duration of the fabrication. The gas species in the chamber were monitored during the fabrication process using the RGA, as described earlier.

Powder Characterization
The oxygen and hydrogen content were carefully tracked after the atomization, screening, and hydrogen treatment. Table 1 shows the oxygen and hydrogen content for the untreated and hydrogen treated LO-Cu, MO-Cu, and HO-Cu screened powders. The LO-Cu powder was procured as-atomized and was sieved in-house under argon in order to maintain a low oxygen content of~225 wt. ppm and~1 wt. ppm hydrogen. After the hydrogen treatment of the LO-Cu powder, the oxygen content was reduced to~50 wt. ppm, with minimal change in the hydrogen content. The untreated MO-Cu powder contained~450 wt. ppm oxygen and~2 wt. ppm hydrogen, and after hydrogen treatment, the oxygen content was reduced to~280 wt. ppm, but the hydrogen content increased to~30 wt. ppm, indicating that the hydrogen was retained within the powder after treatment, likely as trapped H 2 O, which is consistent with the reported literature by Nieh and Nix [28,30]. As a third condition, the HO-Cu was intentionally oxidized in order to exaggerate the effects of oxygen, and produced an extreme condition with oxygen measured at~1500 wt. ppm and~2 wt. ppm hydrogen. After the hydrogen treatment of the HO-Cu powder, the oxygen content was reduced to~580 wt. ppm, but the hydrogen content increased to~75 wt. ppm, again most likely trapping H 2 O inside the particles. No significant sintering is observed after the powders undergo the hydrogen treatment. Figure 5 shows a plot of the atomic percent of oxygen versus the atomic percent of hydrogen for all of the hydrogen heat-treated copper powders used in this study. The 2:1 stoichiometric relationship is a key indicator that hydrogen is retained in the powder in the form of H 2 O vapor. It is also notable that the linear relationship is offset by~0.06% at. % oxygen, which can be accounted for volumetrically on the surface of the powder particles. produced an extreme condition with oxygen measured at ~1500 wt. ppm and ~2 wt. ppm hydrogen.
After the hydrogen treatment of the HO-Cu powder, the oxygen content was reduced to ~580 wt. ppm, but the hydrogen content increased to ~75 wt. ppm, again most likely trapping H2O inside the particles. No significant sintering is observed after the powders undergo the hydrogen treatment. Figure 5 shows a plot of the atomic percent of oxygen versus the atomic percent of hydrogen for all of the hydrogen heat-treated copper powders used in this study. The 2:1 stoichiometric relationship is a key indicator that hydrogen is retained in the powder in the form of H2O vapor. It is also notable that the linear relationship is offset by ~0.06% at. % oxygen, which can be accounted for volumetrically on the surface of the powder particles. Further evidence for the formation of H2O vapor induced cavities in the powder is supported by the helium pycnometry results for each powder condition (shown in Figure 6). No significant change in density is seen after the hydrogen treatment of the LO-Cu particles from the untreated condition. Without internal, grain boundary Cu2O, the hydrogen does not react to form H2O vapor. However, both the MO-Cu and HO-Cu particles show a decrease in density after hydrogen treatment, which is likely caused by the formation of H2O in the particles due to the significant difference in densities between H2O (1 g/cm 3 ) and copper (8.96 g/cm 3 ). The decrease in density among the as received powders can be attributed to the increasing contribution of Cu2O with the increasing oxygen content, which has a density of only ~6 g/cc.  Further evidence for the formation of H 2 O vapor induced cavities in the powder is supported by the helium pycnometry results for each powder condition (shown in Figure 6). No significant change in density is seen after the hydrogen treatment of the LO-Cu particles from the untreated condition. Without internal, grain boundary Cu 2 O, the hydrogen does not react to form H 2 O vapor. However, both the MO-Cu and HO-Cu particles show a decrease in density after hydrogen treatment, which is likely caused by the formation of H 2 O in the particles due to the significant difference in densities between H 2 O (1 g/cm 3 ) and copper (8.96 g/cm 3 ). The decrease in density among the as received powders can be attributed to the increasing contribution of Cu 2 O with the increasing oxygen content, which has a density of only~6 g/cc.
Because of its relatively high oxygen content, the MO-Cu powder was selected to observe the effect of hydrogen heat treatment on the surface oxides. Figure 7 shows the representative low and high magnification scanning electron micrographs of ( Figure 7A,C) and the hydrogen-treated ( Figure 7B,D) MO-Cu powder. Figure 7C shows the surface of the untreated MO-Cu powder, where bright spots were identified as oxygen-rich copper particles. Figure 7D shows the surface at the same magnification after the hydrogen treatment of the MO-Cu powder, where the surface oxides have been largely removed. These results were typical for untreated and hydrogen-treated MO-Cu and HO-Cu particles.
Without internal, grain boundary Cu2O, the hydrogen does not react to form H2O vapor. However, both the MO-Cu and HO-Cu particles show a decrease in density after hydrogen treatment, which is likely caused by the formation of H2O in the particles due to the significant difference in densities between H2O (1 g/cm 3 ) and copper (8.96 g/cm 3 ). The decrease in density among the as received powders can be attributed to the increasing contribution of Cu2O with the increasing oxygen content, which has a density of only ~6 g/cc.    The representative particles were serial sectioned using a FIB to reveal the internal structure of the untreated and hydrogen-treated powder of each type using ion channel imaging, as shown in Figure 8. Figure 8A is an ion image of a sectioned untreated LO-Cu particle showing no obvious porosity or oxide-rich particles. Figure 8B,C are cross-sectioned MO-Cu and HO-Cu particles, respectively, which have fine equiaxed microstructures and submicron oxygen-rich particles along the grain boundaries. The distribution of oxide particles along the grain boundaries vary by size, where larger oxides tend to be located at triple points, and smaller oxides along the grain boundaries. Figure 8D shows a hydrogen-treated LO-Cu particle with minimal change after hydrogen treatment because of its initially low oxygen content. The LO-Cu particles, produced with a special handling consideration, consistently showed few to no internal oxides (within the resolution of the instrument). A grain boundary is shown in Figure 8A,D. However, the lack of contrast between the two grains can be attributed to the similarity of the grain orientations. Figure 8E shows a hydrogen-treated MO-Cu particle with both a micron and submicron scale porosity throughout the grain boundaries. The high oxygen content of the HO-Cu particle leads to the formation of a significant interconnected porosity at the grain boundaries after hydrogen heat treatment. Several large pores are evident after hydrogen treatment in Figure 8E,F for MO-Cu and HO-Cu, respectively; however, in both cases, small dark features are evident along the grain boundaries. It was not immediately clear from this SEM analysis whether these were residual, unreacted oxides or small water vapor induced pores. Videos of the serial sectioning process using FIB are provided in the Supplementary Materials for each of the cases in Figure 8. The HAADF STEM imaging and EDS ( Figure 9) of a grain boundary triple point of hydrogen heat-treated MO-Cu confirms the presence of the H 2 O vapor induced porosity, consistent with the observations of Nieh and Nix [28,30].
Appl. Sci. 2019, 9, x FOR PEER REVIEW 11 of 22 contrast between the two grains can be attributed to the similarity of the grain orientations. Figure  8E shows a hydrogen-treated MO-Cu particle with both a micron and submicron scale porosity throughout the grain boundaries. The high oxygen content of the HO-Cu particle leads to the formation of a significant interconnected porosity at the grain boundaries after hydrogen heat treatment. Several large pores are evident after hydrogen treatment in Figure 8E and Figure 8F for MO-Cu and HO-Cu, respectively; however, in both cases, small dark features are evident along the grain boundaries. It was not immediately clear from this SEM analysis whether these were residual, unreacted oxides or small water vapor induced pores. Videos of the serial sectioning process using FIB are provided in the supplemental data for each of the cases in Figure 8. The HAADF STEM imaging and EDS ( Figure 9) of a grain boundary triple point of hydrogen heat-treated MO-Cu confirms the presence of the H2O vapor induced porosity, consistent with the observations of Nieh and Nix [28,30].      Figure 10 shows the results of the thermogravimetric analysis (TGA) of untreated and hydrogentreated powders for the three initial oxygen contents. A TGA was carried out to ascertain the temperature at which the internal pressure of the H2O vapor pores exceeds the grain boundary strength of the copper powder particles. All of the untreated powders demonstrate no significant change in weight loss after heating to 600 °C at 5 °C /s. However, both hydrogen-treated MO-Cu and   To directly observe the mechanism for the H2O release during the heating of the treated powders, treated and untreated particles were heated to 400 °C , and then to melting at 1200 °C , using an in-situ SEM heating stage (Protochips Fusion). It is important to note that the reported temperature is that of the surface of the heating elements, not necessarily the equilibrium temperature of the particles. The temperature of the stage is calibrated with a pyrometer and correlated to the input current. Here, it is assumed that the mass of the particle is small compared with the input power, and that the equilibrium temperature is reached quickly. This assumption is corroborated by the data acquired from multiple sources in this study (TGA, in-situ EB-PBF, etc.). A temperature of 400 °C was chosen for imaging, based on the onset temperature of weight loss observed by TGA. Figure 11A shows a typical hydrogen-treated HO-Cu particle at room temperature. Additional treated and untreated particles are included as supplemental video files. This particle was placed on the SiC chip in the SEM, and heated in-situ to 400 °C at 100 °C /s. Figure 11B is an SEM image of the particle surface at 400 °C immediately after fracture lines are observed, as H2O vapor escapes from the particle, illustrated by the arrows. Subsequently, as shown in the supplemental videos, the fracture lines close and heal through diffusion mechanisms. To directly observe the mechanism for the H 2 O release during the heating of the treated powders, treated and untreated particles were heated to 400 • C, and then to melting at 1200 • C, using an in-situ SEM heating stage (Protochips Fusion). It is important to note that the reported temperature is that of the surface of the heating elements, not necessarily the equilibrium temperature of the particles. The temperature of the stage is calibrated with a pyrometer and correlated to the input current. Here, it is assumed that the mass of the particle is small compared with the input power, and that the equilibrium temperature is reached quickly. This assumption is corroborated by the data acquired from multiple sources in this study (TGA, in-situ EB-PBF, etc.). A temperature of 400 • C was chosen for imaging, based on the onset temperature of weight loss observed by TGA. Figure 11A shows a typical hydrogen-treated HO-Cu particle at room temperature. Additional treated and untreated particles are included as Supplemental video files. This particle was placed on the SiC chip in the SEM, and heated in-situ to 400 • C at 100 • C/s. Figure 11B is an SEM image of the particle surface at 400 • C immediately after fracture lines are observed, as H 2 O vapor escapes from the particle, illustrated by the arrows. Subsequently, as shown in the Supplemental videos, the fracture lines close and heal through diffusion mechanisms.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 13 of 22 Figure 11. Scanning electron micrographs of a HO-Cu particle after hydrogen treatment, before (A) and after (B-D) in-situ SEM heating on a SiC chip to 400 °C at 100 °C /s under a vacuum.

Electron Beam Fabrication
EBM involves the selective melting of an elevated temperature powder bed for part fabrication. In this study, the powder bed temperature for part fabrication in copper is below the critical temperature for H2O release in hydrogen-treated powder. The reported powder bed temperatures for the EB-PBF of copper range from 380 °C (Lodes et al. [8]) to 530 °C (Guschlbauer et al. [9]), with oxygen contents comparable to the MO-Cu used in this study (although with a larger powder size distribution). It was observed early in the development of this study that the removal of surface oxides with hydrogen heat treatment resulted in difficulty in spreading the powder at elevated powder-bed temperatures (>350 °C ). Without the presence of surface oxides, the activation energy for sintering is greatly reduced. In some cases, sintering was observed during spreading, resulting in homogenous and uneven layering. A lower powder bed temperature alleviates this problem to some extent, and, from a practical standpoint, facilitates the removal of powder from closed cavities. Unlike other common materials utilized in EB-PBF (e.g., Ti6Al4V, TiAl, and 718), copper powder is less susceptible to charge-induced scattering, because of its high conductivity, which also enables EB-PBF processing at lower temperatures, with minimal sintering.
The single melt track width experiments, described in Section 2.5, were carried out to determine whether the powder adjacent to the melt-pool front reaches a sufficiently high temperature for a sufficiently long duration, in order to induce the powder cracking H2O vapor release mechanism observed in the heated stage experiments. The thermal input conditions were modeled using the 3D transient heat transfer model developed by Lee et al. [41] at the Oak Ridge National Laboratory, to guide the experimental design. A 2D contour plot of the temperature profile during melting with a 2000 µ m hatch spacing is shown in Figure 12. The heat affected zone in the powder bed surrounding the melt pool extends several hundred microns below and to the sides of the melt pool, ranging from 500 °C at 150 µ m from the center of the melt pool, to 350 °C at a distance of 250 µ m.
The powder samples were harvested from between each of the EB-PBF wafers from the edge/wall wafers and in ~0.1 mm increments so as to determine the influence of the heat affected zone of the melt pool on the surrounding powder bed particles. Figures 12A-C are the SEM images of the sintered powder adhered to the fabricated part ( Figure 12A), ~0.1 mm distance ( Figure 12B) and ~0.3 mm distance ( Figure 12C) from the part where the surface characteristics of the particles correlate to their respective associated temperature. The sintered powder at the edge of the melt pool  Figure 11C shows images of another particle surface at 400 • C immediately after the onset of the grain boundary fracture. In this case, the particle was quenched from this condition, and subsequently cross sectioned with a FIB, which reveals interconnected porosity along the grain boundaries inside the particle ( Figure 11D). Here, the release of H 2 O vapor at 400 • C causes the interconnected porosity along the grain boundaries underneath the particle surface, with an additional porosity towards the center of the particle.

Electron Beam Fabrication
EBM involves the selective melting of an elevated temperature powder bed for part fabrication. In this study, the powder bed temperature for part fabrication in copper is below the critical temperature for H 2 O release in hydrogen-treated powder. The reported powder bed temperatures for the EB-PBF of copper range from 380 • C (Lodes et al. [8]) to 530 • C (Guschlbauer et al. [9]), with oxygen contents comparable to the MO-Cu used in this study (although with a larger powder size distribution). It was observed early in the development of this study that the removal of surface oxides with hydrogen heat treatment resulted in difficulty in spreading the powder at elevated powder-bed temperatures (>350 • C). Without the presence of surface oxides, the activation energy for sintering is greatly reduced. In some cases, sintering was observed during spreading, resulting in homogenous and uneven layering. A lower powder bed temperature alleviates this problem to some extent, and, from a practical standpoint, facilitates the removal of powder from closed cavities. Unlike other common materials utilized in EB-PBF (e.g., Ti6Al4V, TiAl, and 718), copper powder is less susceptible to charge-induced scattering, because of its high conductivity, which also enables EB-PBF processing at lower temperatures, with minimal sintering.
The single melt track width experiments, described in Section 2.5, were carried out to determine whether the powder adjacent to the melt-pool front reaches a sufficiently high temperature for a sufficiently long duration, in order to induce the powder cracking H 2 O vapor release mechanism observed in the heated stage experiments. The thermal input conditions were modeled using the 3D transient heat transfer model developed by Lee et al. [41] at the Oak Ridge National Laboratory, to guide the experimental design. A 2D contour plot of the temperature profile during melting with a 2000 µm hatch spacing is shown in Figure 12. The heat affected zone in the powder bed surrounding the melt pool extends several hundred microns below and to the sides of the melt pool, ranging from 500 • C at 150 µm from the center of the melt pool, to 350 • C at a distance of 250 µm.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 14 of 22 from the grain boundaries, and the particle surface is smoothed due to diffusion at these relatively high temperatures. Powder images at 0.1 mm from the edge of the wall experienced temperatures of approximately 400-500 °C , according to the thermal model, and had features similar to the in-situ SEM heating (Figure 11), where obvious grain boundary cracking occurred, but no surface smoothing was visible. The collected powder imaged at 0.3-0.5 mm from the melt pool showed no obvious surface changes because of the relatively low temperature (<350 °C ) estimated by the thermal model, and still contained the characteristic H2O vapor pores. These data support the hypothesis that H2O is released in the heat affected zone ahead of the melt pool in EB-PBF. To determine the extent at which H2O is released from hydrogen-treated copper powders during melting as a function of the initial oxygen content, single layer discs were fabricated from each of the treated and untreated powder conditions. Figure 13A shows an image of the melted wafers on the OFE copper fixture after removal from the EB-PBF system. Figure 13C shows an optical image of the surface of the hydrogen-treated LO-Cu melted disc, which exhibits a uniform surface texture with evenly spaced melt pools. However, a non-uniform surface texture is observed on the hydrogentreated HO-Cu melted disc, where exposed powder can be seen in circular pores on the surface ( Figure 13B). The oxygen and hydrogen content of the starting powder, and the resulting solid discs are tabulated in Figure 13D for the treated powders. The untreated powders showed no change in oxygen content. A minimal change in the oxygen content was detected in the treated LO-Cu solid discs from the starting powder. However, the MO-Cu and HO-Cu solid discs show an approximately 50% reduction in the oxygen content from the starting powder. In both cases, a reduction in the hydrogen content is also observed as being approximately proportional to the amount of reduction in the oxygen content. The powder samples were harvested from between each of the EB-PBF wafers from the edge/wall wafers and in~0.1 mm increments so as to determine the influence of the heat affected zone of the melt pool on the surrounding powder bed particles. Figure 12A-C are the SEM images of the sintered powder adhered to the fabricated part ( Figure 12A),~0.1 mm distance ( Figure 12B) and~0.3 mm distance ( Figure 12C) from the part where the surface characteristics of the particles correlate to their respective associated temperature. The sintered powder at the edge of the melt pool was estimated to reach temperatures of 700 • C, where the subsurface H 2 O is expected to be expelled from the grain boundaries, and the particle surface is smoothed due to diffusion at these relatively high temperatures. Powder images at 0.1 mm from the edge of the wall experienced temperatures of approximately 400-500 • C, according to the thermal model, and had features similar to the in-situ SEM heating (Figure 11), where obvious grain boundary cracking occurred, but no surface smoothing was visible. The collected powder imaged at 0.3-0.5 mm from the melt pool showed no obvious surface changes because of the relatively low temperature (<350 • C) estimated by the thermal model, and still contained the characteristic H 2 O vapor pores. These data support the hypothesis that H 2 O is released in the heat affected zone ahead of the melt pool in EB-PBF.
To determine the extent at which H 2 O is released from hydrogen-treated copper powders during melting as a function of the initial oxygen content, single layer discs were fabricated from each of the treated and untreated powder conditions. Figure 13A shows an image of the melted wafers on the OFE copper fixture after removal from the EB-PBF system. Figure 13C shows an optical image of the surface of the hydrogen-treated LO-Cu melted disc, which exhibits a uniform surface texture with evenly spaced melt pools. However, a non-uniform surface texture is observed on the hydrogen-treated HO-Cu melted disc, where exposed powder can be seen in circular pores on the surface ( Figure 13B). The oxygen and hydrogen content of the starting powder, and the resulting solid discs are tabulated in Figure 13D for the treated powders. The untreated powders showed no change in oxygen content. A minimal change in the oxygen content was detected in the treated LO-Cu solid discs from the starting powder. However, the MO-Cu and HO-Cu solid discs show an approximately 50% reduction in the oxygen content from the starting powder. In both cases, a reduction in the hydrogen content is also observed as being approximately proportional to the amount of reduction in the oxygen content. The residual gas analysis was measured as a function of time during the EB-PBF of HO-Cu, MO-Cu, and LO-Cu single layer discs, and is shown in Figure 14. Each disc was melted as a single step and the H2O partial pressure was allowed to stabilize before moving to the next step. The HO-Cu and MO-Cu powders show an increase in partial pressure of the H2O vapor during the melting of the disc with the intensity of the peak being indicative of the relative amount of hydrogen in the powder, as seen in Table 1. The peaks gradually diminish as the H2O vapor is removed from the system via the vacuum pumps. A minimal peak is witnessed during melting for the LO-Cu powder because of the absence of appreciable H2O vapor trapped in the particles. The presence of porosity on the surface of the discs produced from hydrogen-treated HO-Cu and MO-Cu powder is indicative of excessive spatter, and suggests that a portion of the H2O vapor The residual gas analysis was measured as a function of time during the EB-PBF of HO-Cu, MO-Cu, and LO-Cu single layer discs, and is shown in Figure 14. Each disc was melted as a single step and the H 2 O partial pressure was allowed to stabilize before moving to the next step. The HO-Cu and MO-Cu powders show an increase in partial pressure of the H 2 O vapor during the melting of the disc with the intensity of the peak being indicative of the relative amount of hydrogen in the powder, as seen in Table 1. The peaks gradually diminish as the H 2 O vapor is removed from the system via the vacuum pumps. A minimal peak is witnessed during melting for the LO-Cu powder because of the absence of appreciable H 2 O vapor trapped in the particles.
The presence of porosity on the surface of the discs produced from hydrogen-treated HO-Cu and MO-Cu powder is indicative of excessive spatter, and suggests that a portion of the H 2 O vapor escapes from the liquid copper, in addition to the cracking mechanism observed in both the EB-PBF wafer tests and the heated stage experiments. To examine this possibility, high speed imaging (30 kfps) was conducted during the melting process. Full videos are included as part of the Supplementary Materials in this report. Selected frames of high-speed footage of the EB-PBF of hydrogen-treated LO-Cu powder ( Figure 15) show a relatively stable melt pool with minimal spatter. However, Figure 16 shows selected frames from the high-speed footage of hydrogen treated HO-Cu particles during melting, and in this case, significant spatter is observed, caused by the escape of H 2 O vapor and the expansion in the molten copper. Large bubbles, on the order of 1-2 mm in diameter, are observed expanding and collapsing from liquid copper several mm behind the EB spot. Figure 16B shows a bubble forming in the molten pool, which then collapses and ejects molten metal from the pool shown, in Figure 16C (indicated by arrows). and the H2O partial pressure was allowed to stabilize before moving to the next step. The HO-Cu and MO-Cu powders show an increase in partial pressure of the H2O vapor during the melting of the disc with the intensity of the peak being indicative of the relative amount of hydrogen in the powder, as seen in Table 1. The peaks gradually diminish as the H2O vapor is removed from the system via the vacuum pumps. A minimal peak is witnessed during melting for the LO-Cu powder because of the absence of appreciable H2O vapor trapped in the particles. The presence of porosity on the surface of the discs produced from hydrogen-treated HO-Cu and MO-Cu powder is indicative of excessive spatter, and suggests that a portion of the H2O vapor escapes from the liquid copper, in addition to the cracking mechanism observed in both the EB-PBF wafer tests and the heated stage experiments. To examine this possibility, high speed imaging (30 kfps) was conducted during the melting process. Full videos are included as part of the supplemental data in this report. Selected frames of high-speed footage of the EB-PBF of hydrogen-treated LO-Cu powder ( Figure 15) show a relatively stable melt pool with minimal spatter. However, Figure 16  shows selected frames from the high-speed footage of hydrogen treated HO-Cu particles during melting, and in this case, significant spatter is observed, caused by the escape of H2O vapor and the expansion in the molten copper. Large bubbles, on the order of 1-2 mm in diameter, are observed expanding and collapsing from liquid copper several mm behind the EB spot. Figure 16B shows a bubble forming in the molten pool, which then collapses and ejects molten metal from the pool shown, in Figure 16C (indicated by arrows).  Despite the presence of porosity in the single layer melt scenario caused by the escape of H2O vapor from the liquid zone, it was speculated that the layerwise remelting of a material typical in additive manufacturing would serve to reduce or eliminate this porosity in the fabrication of components. In order to determine the effect of successive melts on the removal of H2O, as well as the solid sample density, cylinders were fabricated using hydrogen-treated MO-Cu powder.
In the EB-PBF validation runs, the precursor untreated powders had an oxygen content of ~450 ppm and a hydrogen content of ~2 ppm, and the solid copper components fabricated with untreated powder show no change in the oxygen content. Hydrogen heat treatment of the MO-Cu powder reduces the oxygen content to ~280 ppm because of the volumetric elimination of surface oxides, and additionally, the hydrogen content of the powder is increased to ~33 ppm. Solid samples fabricated with EB-PBF using hydrogen-treated MO-Cu powder results in an oxygen content of ~50 ppm, and a negligible hydrogen content. The in-situ RGA during both trials shows the partial pressure of the layerwise outgassing of H2O vapor, which while present on all AM powders, is two orders of magnitude higher for the treated powders compared with the untreated powders. These data are reported in Figure 17. Figure 18 shows the representative microstructures of the solid samples produced with both treated and untreated powders. In both cases, a columnar grain structure is observed, aligned with the build direction, and a high density is achieved. shows selected frames from the high-speed footage of hydrogen treated HO-Cu particles during melting, and in this case, significant spatter is observed, caused by the escape of H2O vapor and the expansion in the molten copper. Large bubbles, on the order of 1-2 mm in diameter, are observed expanding and collapsing from liquid copper several mm behind the EB spot. Figure 16B shows a bubble forming in the molten pool, which then collapses and ejects molten metal from the pool shown, in Figure 16C (indicated by arrows).  Despite the presence of porosity in the single layer melt scenario caused by the escape of H2O vapor from the liquid zone, it was speculated that the layerwise remelting of a material typical in additive manufacturing would serve to reduce or eliminate this porosity in the fabrication of components. In order to determine the effect of successive melts on the removal of H2O, as well as the solid sample density, cylinders were fabricated using hydrogen-treated MO-Cu powder.
In the EB-PBF validation runs, the precursor untreated powders had an oxygen content of ~450 ppm and a hydrogen content of ~2 ppm, and the solid copper components fabricated with untreated powder show no change in the oxygen content. Hydrogen heat treatment of the MO-Cu powder reduces the oxygen content to ~280 ppm because of the volumetric elimination of surface oxides, and additionally, the hydrogen content of the powder is increased to ~33 ppm. Solid samples fabricated with EB-PBF using hydrogen-treated MO-Cu powder results in an oxygen content of ~50 ppm, and a negligible hydrogen content. The in-situ RGA during both trials shows the partial pressure of the layerwise outgassing of H2O vapor, which while present on all AM powders, is two orders of magnitude higher for the treated powders compared with the untreated powders. These data are reported in Figure 17. Figure 18 shows the representative microstructures of the solid samples produced with both treated and untreated powders. In both cases, a columnar grain structure is observed, aligned with the build direction, and a high density is achieved. Despite the presence of porosity in the single layer melt scenario caused by the escape of H 2 O vapor from the liquid zone, it was speculated that the layerwise remelting of a material typical in additive manufacturing would serve to reduce or eliminate this porosity in the fabrication of components. In order to determine the effect of successive melts on the removal of H 2 O, as well as the solid sample density, cylinders were fabricated using hydrogen-treated MO-Cu powder.
In the EB-PBF validation runs, the precursor untreated powders had an oxygen content of 450 ppm and a hydrogen content of~2 ppm, and the solid copper components fabricated with untreated powder show no change in the oxygen content. Hydrogen heat treatment of the MO-Cu powder reduces the oxygen content to~280 ppm because of the volumetric elimination of surface oxides, and additionally, the hydrogen content of the powder is increased to~33 ppm. Solid samples fabricated with EB-PBF using hydrogen-treated MO-Cu powder results in an oxygen content of 50 ppm, and a negligible hydrogen content. The in-situ RGA during both trials shows the partial pressure of the layerwise outgassing of H 2 O vapor, which while present on all AM powders, is two orders of magnitude higher for the treated powders compared with the untreated powders. These data are reported in Figure 17. Figure 18 shows the representative microstructures of the solid samples produced with both treated and untreated powders. In both cases, a columnar grain structure is observed, aligned with the build direction, and a high density is achieved.

Discussion
The oxygen content of each of the three starting powder conditions results from surface oxide films and oxides along the grain boundaries. The copper oxide will continue to grow when exposed to oxygen, as it does not passivate [42]. The surface oxide thickness as a function of time at room temperature was empirically derived by White and Germer [43]. A typical oxide thickness would fall between 3-5 nm, but could be much thicker. The overall contribution of this oxide layer becomes a significant portion of the chemistry for a powder with a high specific surface area [44]. Figure 19 shows the calculated wt. ppm of oxygen as a function of the particle diameter for cuprous oxide surface films of various thicknesses. In the case of a 5 nm coating, and a typical Gaussian powder size distribution centered at a 35 µ m diameter, the oxygen content is estimated to be roughly 75 wt. ppm oxygen, assuming all oxygen neglecting the 2 wt. ppm in solid solution at room temperature [39]. The experimental data gathered in this study suggest that the oxide thickness on the LO-Cu particles used in this study is ~15-20 nm, consistent with our STEM observations. Volumetrically, this accounts for the measured 200 wt. ppm oxygen, which was mainly surface oxides in the LO-Cu material for particles in the range of 15-53 µ m.

Discussion
The oxygen content of each of the three starting powder conditions results from surface oxide films and oxides along the grain boundaries. The copper oxide will continue to grow when exposed to oxygen, as it does not passivate [42]. The surface oxide thickness as a function of time at room temperature was empirically derived by White and Germer [43]. A typical oxide thickness would fall between 3-5 nm, but could be much thicker. The overall contribution of this oxide layer becomes a significant portion of the chemistry for a powder with a high specific surface area [44]. Figure 19 shows the calculated wt. ppm of oxygen as a function of the particle diameter for cuprous oxide surface films of various thicknesses. In the case of a 5 nm coating, and a typical Gaussian powder size distribution centered at a 35 µ m diameter, the oxygen content is estimated to be roughly 75 wt. ppm oxygen, assuming all oxygen neglecting the 2 wt. ppm in solid solution at room temperature [39]. The experimental data gathered in this study suggest that the oxide thickness on the LO-Cu particles used in this study is ~15-20 nm, consistent with our STEM observations. Volumetrically, this accounts for the measured 200 wt. ppm oxygen, which was mainly surface oxides in the LO-Cu material for particles in the range of 15-53 µ m.

Discussion
The oxygen content of each of the three starting powder conditions results from surface oxide films and oxides along the grain boundaries. The copper oxide will continue to grow when exposed to oxygen, as it does not passivate [42]. The surface oxide thickness as a function of time at room temperature was empirically derived by White and Germer [43]. A typical oxide thickness would fall between 3-5 nm, but could be much thicker. The overall contribution of this oxide layer becomes a significant portion of the chemistry for a powder with a high specific surface area [44]. Figure 19 shows the calculated wt. ppm of oxygen as a function of the particle diameter for cuprous oxide surface films of various thicknesses. In the case of a 5 nm coating, and a typical Gaussian powder size distribution centered at a 35 µm diameter, the oxygen content is estimated to be roughly 75 wt. ppm oxygen, assuming all oxygen neglecting the 2 wt. ppm in solid solution at room temperature [39]. The experimental data gathered in this study suggest that the oxide thickness on the LO-Cu particles used in this study is~15-20 nm, consistent with our STEM observations. Volumetrically, this accounts for the measured 200 wt. ppm oxygen, which was mainly surface oxides in the LO-Cu material for particles in the range of 15-53 µm. Figure 19. Wt. ppm oxygen as a function of the particle size, assuming a 5 nm cuprous oxide film.
The MO-Cu and HO-Cu particles greatly exceed this oxygen content and show a large number of submicron oxides along the grain boundaries, as seen in Figure 7B and Figure 7C. Naturally, the HO-Cu particles show the largest number of submicron oxides along the grain boundaries, with several larger oxides at triple points of the microstructure because of its exposure to air at elevated temperatures.
To remove the oxide particles, a hydrogen atmosphere heat treatment is employed to reduce the copper oxide to copper and H2O. This was first observed in the MO-Cu powder in Figure 7C and Figure 7D, where the surface oxides were removed by hydrogen treatment. As hydrogen diffuses rapidly in copper (900 µ m 2 /s at 400 °C [34]), it allows for the reduction of oxides present inside the particles in a relatively short amount of time. The hydrogen treatment temperature and time was chosen to allow for sufficient hydrogen diffusion through the particles, while not reaching sintering temperatures for the fine powder size distribution. Although not directly reported herein, it is worth noting that, not since the completion of this study, we have subsequently demonstrated the hydrogen reduction of copper powder in several furnace types (including batch and continuous) in both pure hydrogen and in forming gas (5% h, 95% Ar), and we have observed consistent results in all scenarios for the equivalent treatment times. After hydrogen treatment, the MO-Cu and HO-Cu powders exhibit a micron scale porosity at the triple points of the grain boundaries, and a smaller submicron porosity along the grain boundaries. These pores have previously been shown in the literature to be trapped H2O vapor [28,30,35].
According to Ito and Hayashi [35], the formation of H2O vapor in closed pores is as a result of the reduction of copper oxide during hydrogen heat treatment. The H2O vapor pressure inside a closed pore is higher than that of the surface stress of the pore, which inhibits the pore from collapsing. However, the H2O vapor pressure is not great enough to overcome the mechanical strength of the grain boundaries of the particle at room temperature.
At around 350 °C , the strength of the copper is significantly reduced, and the increasing internal pressure of the H2O vapor overcomes the mechanical strength of the grain boundary and the pore begins to expand in size. This leads to the fracturing of the particle, seen as an extensive network of interconnected pores along the grain boundaries, and a release of H2O vapor into the vacuum system, as seen in the SEM images in Figure 11C and Figure 11D, and the RGA data from Figure 14. Notably, our work is presented for a given size distribution of gas atomized powder. While it is apparent that the reduction of internal Cu2O and the formation of steam pores is, for all intents and purposes, independent of particle size (owing to the rapid diffusion of hydrogen in copper at 400 °C ), the practical upper and lower limits on particle size, for which the observed phenomena hold true, is not clear from this research. It is likely that the combined influences of grain boundary strength, grain size, radius of curvature, and oxygen content will contribute to the steam embrittlement/fracture mechanism in different ways.
The observations shown here suggest that H2O is released by two mechanisms during EB-PBF. A chemical analysis of the sintered powder bed shows a similar oxygen and hydrogen content as the Figure 19. Wt. ppm oxygen as a function of the particle size, assuming a 5 nm cuprous oxide film.
The MO-Cu and HO-Cu particles greatly exceed this oxygen content and show a large number of submicron oxides along the grain boundaries, as seen in Figure 7B,C. Naturally, the HO-Cu particles show the largest number of submicron oxides along the grain boundaries, with several larger oxides at triple points of the microstructure because of its exposure to air at elevated temperatures.
To remove the oxide particles, a hydrogen atmosphere heat treatment is employed to reduce the copper oxide to copper and H 2 O. This was first observed in the MO-Cu powder in Figure 7C,D, where the surface oxides were removed by hydrogen treatment. As hydrogen diffuses rapidly in copper (900 µm 2 /s at 400 • C [34]), it allows for the reduction of oxides present inside the particles in a relatively short amount of time. The hydrogen treatment temperature and time was chosen to allow for sufficient hydrogen diffusion through the particles, while not reaching sintering temperatures for the fine powder size distribution. Although not directly reported herein, it is worth noting that, not since the completion of this study, we have subsequently demonstrated the hydrogen reduction of copper powder in several furnace types (including batch and continuous) in both pure hydrogen and in forming gas (5% h, 95% Ar), and we have observed consistent results in all scenarios for the equivalent treatment times. After hydrogen treatment, the MO-Cu and HO-Cu powders exhibit a micron scale porosity at the triple points of the grain boundaries, and a smaller submicron porosity along the grain boundaries. These pores have previously been shown in the literature to be trapped H 2 O vapor [28,30,35].
According to Ito and Hayashi [35], the formation of H 2 O vapor in closed pores is as a result of the reduction of copper oxide during hydrogen heat treatment. The H 2 O vapor pressure inside a closed pore is higher than that of the surface stress of the pore, which inhibits the pore from collapsing. However, the H 2 O vapor pressure is not great enough to overcome the mechanical strength of the grain boundaries of the particle at room temperature.
At around 350 • C, the strength of the copper is significantly reduced, and the increasing internal pressure of the H 2 O vapor overcomes the mechanical strength of the grain boundary and the pore begins to expand in size. This leads to the fracturing of the particle, seen as an extensive network of interconnected pores along the grain boundaries, and a release of H 2 O vapor into the vacuum system, as seen in the SEM images in Figure 11C,D, and the RGA data from Figure 14. Notably, our work is presented for a given size distribution of gas atomized powder. While it is apparent that the reduction of internal Cu 2 O and the formation of steam pores is, for all intents and purposes, independent of particle size (owing to the rapid diffusion of hydrogen in copper at 400 • C), the practical upper and lower limits on particle size, for which the observed phenomena hold true, is not clear from this research. It is likely that the combined influences of grain boundary strength, grain size, radius of curvature, and oxygen content will contribute to the steam embrittlement/fracture mechanism in different ways.
The observations shown here suggest that H 2 O is released by two mechanisms during EB-PBF. A chemical analysis of the sintered powder bed shows a similar oxygen and hydrogen content as the starting powder, as seen in Table 1, and suggests that the H 2 O vapor is not released during the preheating step, which is measured in the range of 250-270 • C. While not explored in this study, it is anticipated that higher preheat temperatures, similar to those used by Frigola [1] and Guschlbauer [9], would be sufficiently high to elicit the effect observed by the TGA and heated stage SEM experiments. In the present case, the simulation results and experiments of a single electron beam melt line in a copper powder bed show that as the electron beam scans across the sample during the melting step, the highly localized heat in front and to the sides of the beam causes the particles in the heat affected zone to fracture because of the H 2 O vapor expansion at 400 • C. The scanning electron microscopy of the particles surrounding the single wafer tracks show fracture surfaces similar to that of Figure 11C.
The chemical analyses of the single layer melt solid discs of MO-Cu and HO-Cu show that some amount of H 2 O vapor is trapped during the solidification of the melt pool, as seen in Figure 14. We theorize that as the particles become molten, the volume of the H 2 O vapor rapidly expands because of the significant pressure drop due to the vacuum environment, and become bubbles of H 2 O vapor in the melt pool. According to Boyle's law, a pressure of 5.7 MPa confined to a sphere with a diameter of 1 um will expand to a sphere with a diameter of 1 mm at a pressure of 5 × e −9 MPa. This is approximately a 10 9 increase in the volume of the trapped H 2 O vapor from within a closed pore. The H 2 O vapor bubbles are able to escape the melt pool, which is evident in the decrease in oxygen and hydrogen content of the single melt layer discs. However, because of the rapid solidification of the melt pool (10 −4 to 10 −6 • C/s) and the high viscosity of molten copper, and the large amount of spatter associated with bubble collapse, a portion of these H 2 O vapor bubbles may become trapped in the solid single layer discs as pores. An observation of the multiple layer build validation runs suggest that these pores are largely eliminated by the subsequent remelting of the previous layers, which allows repeated opportunities for the H 2 O vapor to escape to the surface. This is in stark contrast to the traditional copper processing routes (powder metal, sintering, annealing etc.) for which oxygen containing copper powder exposed to a high temperature hydrogen atmosphere results in deleterious embrittlement, cracking, swelling, and porosity. This is clearly shown in the density and microstructures of Figure 18. Also evident is an apparent difference in grain size. This was not investigated in this study, but it is reasonable that the differences in oxygen content and the turbulent liquid pool/spatter brought about by the escape of water vapor during solidification could lead to the disruption of grain growth observed. Further analyses would be required in order to verify this hypothesis.
While the present study has focused on the EB-PBF processing of copper, it is entirely plausible that a comparable methodology could be repeated for the LPBF processes. However, the poor absorptivity, lower ambient temperatures, and lack of a vacuum atmosphere may contribute to a significantly different outcome. In the work by El-Wardany et al. [6], a relatively high oxygen copper powder was heat treated in a forming gas environment in order to remove the surface oxides and subsequently be coated with a polydimethylsiloxane (PDMS) polymer. No further data on the chemistry, properties, or microstructure of this material were reported, other than the observation of the formation of unexpectedly large pores in single track welds that were not present in the untreated powders. These results, and the results of the current study suggest that further research is warranted, which also accounts for the potentially important effects of chamber atmosphere and the partial pressures of the relevant gas species during the AM processing of copper.

Conclusions
By utilizing a hydrogen furnace treatment to reduce the oxygen content of atomized copper powder feedstock for use in EBM processing, we have shown the ability to produce pure copper parts with a lower oxygen content than that of the initial feedstock. Depending on the initial oxygen content, the mechanism for reduction shifts from the removal of surface oxides of a low starting oxygen content, to a combination of surface oxide removal and internal H 2 O vapor formation as the starting oxygen content increases. The release of internally trapped H 2 O vapor is shown to occur in the heat affected zone to the front and side of the melt pool, in addition to release from the melt pool. Successive remelting allows sufficient time for H 2 O removal from hydrogen-treated powder, to produce solids resulting in an oxygen content of 50 wt. ppm.