Quasi-Static Tensile Properties of Unalloyed Copper Produced by Electron Beam Powder Bed Fusion Additive Manufacturing.

Mechanical properties of powder bed fusion processed unalloyed copper are reported majorly in the as-fabricated condition, and the effect of post-processes, common to additive manufacturing, is not well documented. In this study, mechanical properties of unalloyed copper processed by electron beam powder bed fusion are characterized via room temperature quasi-static uniaxial tensile test and Vickers microhardness. Tensile samples were extracted both perpendicular and parallel to the build direction and assigned to three different conditions: as-fabricated, hot isostatic pressing (HIP), and vacuum annealing. In the as-fabricated condition, the highest UTS and lowest elongation were obtained in the samples oriented perpendicular to the build direction. These were observed to have clear trends between sample orientation caused primarily by the interdependencies between the epitaxial columnar grain morphology and dislocation movement during the tensile test. Texture was insignificant in the as-fabricated condition, and its effect on the mechanical properties was outweighed by the orientation anisotropy. The fractographs revealed a ductile mode of failure with varying dimple sizes where more shallow and finely spaced dimples were observed in the samples oriented perpendicular to the build direction. EDS maps reveal that grain boundary oxides coalesce and grow in HIP and vacuum-annealed specimens which are seen inside the ductile dimples and contribute to their increased ductility. Overall, for the post-process parameters chosen in this study, HIP was observed to slightly increase the sample's density while vacuum annealing reduced the oxygen content in the specimens.

1. Introduction

The benefits of complex geometries, reduced assemblies, and rapid build time attributed to powder bed fusion (PBF) additive manufacturing (AM) have recently been extended to unalloyed copper for a variety of applications requiring very high thermal and electrical conductivity, complicated geometries, and extensive processing routes. Examples include, but are not limited to, electric motors [1], vacuum electronics [2,3,4,5,6], and heat exchangers [7,8]. However, compared to other common AM metals, the feasibility of PBF-AM of copper has only recently been demonstrated, and for all intents and purposes is at a relatively early stage of developmental maturity. Within the current literature, studies have typically focused on process parameter development with the predominant metrics investigated being material density and, to a lesser degree, material purity. The documented AM processing space is primarily driven by relatively small, prismatic geometries. As a result, the very basic property characterizations are found only sporadically in the scientific knowledge base, and then only a limited number of these studies have reported mechanical property data for AM fabricated copper; the highlights of which are summarized in Table 1 [9,10,11,12,13,14].

Table 1

Summary of the mechanical properties of BJ, L-PBF, and EB-PBF-processed copper (Best values are taken in each case; horizontal = loading transverse to build direction; vertical = loading parallel to build direction).

Process	UTS (MPa)	Yield Strength (MPa)	Elongation to Failure (%)	Relative Density (%)	Oxygen Content (wt. ppm)	Treatment Condition	Ref.
L-PBF	149	-	-	88.1	-	As-fabricated	[9]
L-PBF	248 ± 8.5	187 ± 5.3	9.2 ± 2.12	99.1 ± 0.5	446	As-fabricated	[10]
L-PBF	215 ± 7.2	51 ± 8.2	30 ± 3.04	-	446	Vacuum annealed	[10]
L-PBF	211 ± 4	122 ± 1	43 ± 3	99.3 ± 0.2	54	As-fabricated	[14]
EB-PBF	-	76	-	-	-	As-fabricated	[2]
EB-PBF	231.6 ± 5.4	149.8	56.2	>99.5	188	As-fabricated (horizontal)	[11]
EB-PBF	177 ± 3.3	78.1 ± 0.9	59.3 ± 7.5	>99.5	188	As-fabricated (vertical)	[12]
BJ	176.4 ± 6.5	N/A	67.2 ± 2.2	97.3 ± 0.1	N/A	Sintered, HIP	[13]

In part, the lack of component-level data can be attributed to the significant processing challenges associated with PBF-AM of copper. Recent data from the literature suggest that the feasible AM process window for both laser PBF (L-PBF) and electron beam PBF (EB-PBF) is effectively narrow and highly sensitive to thermal boundary conditions and geometric influences within a layer. These effects are exacerbated by the high thermal conductivity of the material [5,14]. Therefore, while AM-produced copper components exhibiting complex geometries have been reported, the process parameter and properties data are typically associated with small prismatic specimens.

For L-PBF, the high thermal conductivity, high reflectance (for typical ~1µm wavelengths used in AM), and oxidation are well documented and often cited as the most significant process barriers for fabrication [8,15]. Recent work with high-power lasers (400–1000 W) [10,14,16,17,18], shorter laser wavelengths in the visible spectrum [19,20,21,22], and micro-alloying [23,24] have incrementally improved processing, material density, and have shown promising properties. EB-PBF of unalloyed copper generally circumvents many of the barriers faced by L-PBF techniques. The energy coupling between the beam and the powder is significantly higher [5], the mid-level vacuum environment (~2 × 10−3–2 × 10−6 mBar) contributes toward minimizing oxygen pickup, and the high beam power and scan speeds facilitate efficient bed preheating to reduce global thermal gradients [8,25], although the latter results in powder sintering which can complicate powder removal.

As expected, the reported ultimate tensile strength (UTS) and yield data are skewed by sample density. This is notably evident in the L-PBF studies, where the reported density has a wide range. Lykov et al. [9] reported a UTS of 149 MPa with a relative density of 88.1%, whereas Yan et al. [10] and Jadhav et al. [14] fabricated copper with a relative density up to 99.1% and 99.3% that resulted in an UTS of 248 and 211 MPa, respectively. For EB-PBF, however, most of the reported relative densities exceeded 99.5% [11,12,26,27,28]. Guschlbauer et al. reported a maximum UTS of 231.6 MPa tested transverse to the build direction; the lowest UTS reported in the same study was 181.3 MPa oriented parallel to the build direction [11]. It should be noted that these data, for both EB-PBF and L-PBF, are typically extracted from sub-size tensile specimens where the influence of grain size, compositional contamination, or a particularly large defect will have a significant effect.

Apart from powder bed fusion AM processes, Kumar et al. have reported UTS values of 176.4 ± 6.5 MPa with 67.2 ± 2.2% elongation for copper produced using a binder jet (BJ) AM that was sintered in a reducing atmosphere and subsequently subjected to hot isostatic pressing (HIP) at 1075 °C [13]. While of lower purity and subject to contamination introduced by the binder material, the microstructure reported by the authors is refined and equiaxed as a result of the solid-state processing. This is in contrast to the directional microstructure typical of the powder bed fusion process of unalloyed copper reported in the literature.

Alongside the solid density, the effect of oxygen content on ductility, particularly outside the range of what would typically be considered pure copper, is well documented [5]. Myers and Blythe reported that in cast copper, ductility was maintained up to an oxygen content of 280 wt. ppm, beyond which it led to a significant loss of ductility [29], in accord with the findings of Guschlbauer et al. [11] that attributed copper oxides acting as stress concentration sites during the tensile test of EB-PBF processed copper. Further, with the exception of the data presented by Ledford et al. [1,26], which do not include any mechanical properties, the EB-PBF automatic control parameters were explicitly disabled. While open-loop on most EB-PBF systems, such parameters serve to regulate beam speed and power to maintain stable melt pool size and powder bed temperature as part geometries evolve. It is therefore unclear how the reported properties data translate to more complicated geometries and beam control strategies. Therefore, exploration of mechanical properties of AM copper subjected to post-processing steps common to AM such as HIP and annealing could have the potential to widen the effective processing space and enable broader geometric variation with AM copper components. In the present study, results of room temperature quasi-static uniaxial tensile test and microhardness measurements were recorded for copper specimens produced via EB-PBF with automatic control settings enabled across multiple runs with X-oriented (perpendicular to build direction) and Z-oriented (parallel to build direction) specimens. A subset of the samples was subjected to HIP to explore the influence of densification on residual micro porosity while another group of samples from the same builds were subjected to vacuum annealing which is known to reduce oxygen content through diffusion and evaporation [30]. The remaining specimens were tested in the as-fabricated condition.

2. Materials and Methods

2.1. Powder Feedstock

Nitrogen gas-atomized copper powder screened to −270/+635 mesh was obtained from Sandvik Osprey (Sandvik Osprey, Neath, UK). The initial oxygen content of the powder was 457 wt. ppm, measured using inert gas fusion (LECO OH836, copper protocol, LECO, St. Jospeh, MI, USA). Powder particle size measurement was carried out on a Microtrac S3500 (Microtrac Inc., Montgomeryville, PA, USA) using a laser diffraction technique. The unalloyed copper powder chemistry was obtained using the inductively coupled plasma mass spectroscopy (ICP-MS, Laboratory Testing Inc., Hartfield, PA, USA) method. Powder morphology was observed using a JEOL 6010LA scanning electron microscope (JEOL, Peabody, MA, USA).

2.2. Sample Fabrication

Five builds, each containing three rectangular blocks measuring 58.5 mm × 12.7 mm × 58.5 mm, were fabricated upon an 86 mm diameter and 40 mm thick oxygen-free copper base plate using a customized Arcam A2 EB-PBF system (Software version 3.2, SP2, Arcam AB), the details of which are described by Ledford et al. [31]. The nominal layer thickness was 40 µm. The powder bed temperature was measured by a thermocouple affixed to the bottom of the build substrate and was observed to be stable throughout the build duration within a range of 290–320 °C. The vacuum level inside the chamber was maintained at 5 × 10−4 mbar. Table 2 and Table 3 show the process parameters used to fabricate the samples. As illustrated in Figure 1, the rectangular blocks were oriented 45° about the z-axis. The beam raster patterns were oriented parallel and perpendicular to the x-axis and rotated 90° each layer. The blocks were melted “model-wise,” one by one so that each experienced nearly uniform scan line length (shown schematically on the front block within the build layout) and thermal history. Upon completion, the build volume was allowed to cool to below 30 °C prior to removal from the build chamber. The powder cake surrounding the parts was removed manually in an argon-filled glove box maintained with an atmospheric oxygen content below 10 wt. ppm. The powder was passed through a 70-mesh screen to remove large agglomerations and blended with the primary batch prior to each reuse. Powder size and oxygen content of both the reused powder and fabricated solids were measured after each build.

Table 2

EB-PBF melt parameters used to fabricate the specimens used in this study (the parameters shown are specifically those that have been modified from the standard commercial parameter sets for Ti6Al4V, V3.2 SP2).

Melting Parameter	Value
Surface Temp (°C)	600
Power Analyze Max Current (mA)	25
Power Analyze Min Current (mA)	8
Beam Speed (mm/s)	1000
Beam Current (mA)	8
Max Current (mA)	8
Focus Offset (mA)	18
Speed Function	20
Line Offset (mm)	0.13
Change for Each Depth (mm)	0.02
Heating Enable	TRUE
Max Heat Time (s)	5
Melt Heating Use Process Power	TRUE
Heating Between Models	TRUE

Table 3

EB-PBF preheat parameters used to fabricate the specimens used in this study (the parameters shown are specifically those that have been modified from the standard commercial parameter sets for Ti6Al4V, V3.2 SP2).

Parameter	Value
Focus Value (mA)	70
Focus Value Heater (mA)	150
Box Enable	TRUE
Box Size (mm)	85
Offset to Part (mm)	0.1
Max Current for Box (mA)	22.5
Jump Safe Sweep Max Current (mA)	13.5
Jump Safe Sweep Min Current (mA)	0.1
Jump Safe Sweep Speed (mm/s)	14600
Jump Safe Sweep Total Repetitions of Sweep	40
Jump Safe Sweep Max Number of Sweeps	40
Heating Enable	TRUE
Max Heat Time (s)	5
Heating Between Models	TRUE

Figure 1

Overall experimental framework highlighting the build layout and sample orientations, treatment conditions, and the ASTM Standard design of the test coupons used in the study.

Figure 1 illustrates the experimental design. The orientation of the tensile bars within each block was randomly assigned subject to the constraint that only specimens with a single orientation could be extracted from any one block. As illustrated in Figure 1, each as-fabricated block was cut into four tensile blanks using a bandsaw assigned as Z when the blank was cut with the long sample axis parallel to the build direction and X when the long sample axis was perpendicular to the build direction using the ASTM nomenclature [32]. Each of the tensile blanks was then randomly assigned to one of three treatment groups. Samples in the as-fabricated group were tested without treatment, samples from the HIP group were subjected to HIP using a QIH15LMURC high pressure furnace (Quintus Technologies, Lewis Center, OH, USA) with a heating rate of 10 °C/min to 950 °C, a dwell of 2 h, and cooling at a rate of 10 °C/min. The pressure is proportional to the temperature, and at 950 °C was 165 MPa. Several preliminary experiments were conducted at Quintus Technologies to identify this particular parameter that yielded high densities. Samples assigned to the vacuum-annealing group were subjected to the same temperature profile in a tube furnace (Thermo-Scientific Lindberg/Blue M HTF55322C, Hillsboro, OR, USA), but under vacuum at 5 × 10−5 mbar (5 × 10−9 MPa). After treatment, all samples were machined into ASTM subsize specimens (specimen type 4) [33]. The surface finish of the machined specimens in the gauge region measured an average of 16 µin R using a stylus profilometer (Mitutoyo Surftest SJ210, Aurora, IL, USA).

2.3. Quasi-Static Uniaxial Tensile Test Procedure

Figure 2A–G show a photograph of the tensile testing setup. Tensile specimens harvested from the fabricated blocks (Figure 2A) were tested to failure after respective treatment conditions at room temperature using an ATS 1620C universal testing machine (Applied Test Systems, Inc., Butler, PA, USA) with wedge grips and a constant crosshead speed of 0.254 mm/min using a 5 kN load cell. Strains were calculated using digital image correlation (DIC). One face of the gauge section of each sample was painted with white background with black speckles. As the specified ductility of the dried paint material is lower than that of pure annealed copper, tests were conducted while the speckle paint was still wet. Images were acquired at 30 frames per minute using a 12.3 MP FLIR Grasshopper camera (FLIR, Wilsonville, OR, USA) under appropriate illumination conditions obtained using two F&V K4000S bi-color studio lighting kits (F&V, Mundelein, IL, USA), illustrated in Figure 2C. Strain field mapping and elongation measurement were carried out using GOM Correlate software (2018 Hotfix 6, Rev. 117418), where selected speckles were tracked for making strain measurements. Figure 2D–G show an example of such strain field mapping of a Z- and X-oriented vacuum-annealed specimen at the beginning and end of the test. The yield stress was calculated from a 0.2% offset from the linear portion of the stress-strain curve. The mechanical properties results were compared using ANOVA (JMP Pro 15. Ink) with pooled variance. Prior to analysis, the data were tested for normality using the Shapiro-Wilk test with the α-value of 0.05.

Figure 2

Fabricated blocks on a copper substrate (A), Machined tensile sample according to [33] (B), Overall tensile test setup showing the arrangement of DIC camera and lighting kits (C), Strain mapping of a Z- and X-oriented vacuum-annealed tensile sample in GOM Correlate software in the beginning (D,F) and the end (E,G) of the test, respectively.

2.4. Metallographic Analyses and Microhardness Measurements

The grip section of each tensile bar was cut perpendicular to the longitudinal axis approximately 10 mm from each end. These samples, in turn, were sectioned longitudinally with a low-speed diamond saw (Buehler Isomet, Lake Bluff, IL, USA). The cut face was potted in phenolic resin and progressively ground using 400, 600, 1000, 2000 and 4000 grit SiC paper with water lubricant. Samples were polished with 1-micron, 0.3-micron, and 0.05-micron alumina slurry, and etched with Ammonium Persulphate solution by submersion/swabbing for 10 s and rinsing to reveal the grain structure. Samples were indented for Vickers microhardness measurements in a LECO M400 hardness tester (LECO, St. Joseph, MI, USA) under 100 gf load for a dwell time of 10 s. A total of ten indentations, 500 µm apart, were made along the direction of tension for both types of sample orientation, and the average values are reported. The sectioned grip end that was not used for metallographic analysis was used to measure the density of the sample using gas pycnometry (Quantachrome Ultrapyc 1200E according to [34]). Approximately 1 +/− 0.0002 g of the same was sectioned longitudinally for oxygen content measurement with inert gas fusion (as mentioned in Section 2.1). After tensile testing, a thin slice of the grip end along the build direction was cut and polished for as-fabricated, HIP, and vacuum-annealed X-oriented samples for electron backscatter diffraction (EBSD) which was performed on the FEI Quanta 3D Field Emission Gun (Thermo Fisher Scientific, Hillsboro, OR, USA) equipped with an Oxford Instruments EBSD detector on a 3 mm × 3 mm area on samples extracted with a X-sample orientation. Inverse pole figure (IPF) maps and pole figures were generated with Channel 5 Tango and Mambo software, respectively. The fracture surfaces of the tensile specimens were observed with the JEOL 6010LA SEM. Characterization of grain boundary oxides was carried out with STEM-EDS using a ThermoFisher Talos F200X TEM (Thermo Fisher Scientific, Hillsboro, OR, USA). The TEM samples were prepared with a Hitachi Ar-blade 5000 ion mill for 15 min prior to sample lift out, thinning, and mounting with the FEI Quanta 3D FEG.

3. Results and Discussion

3.1. Powder Characterization

Figure 3A depicts the particle size distribution of the powder used in this study with a d10–d90 size range of 22–63 µm and d50 value of 39 µm and demonstrates no significant change in particle size with reuse over the course of this study. Table 4 shows the chemistry of the initial feedstock as measured by ICP-MS where no major contamination was detected. The oxygen content of the powder does show an increase during the study from an initial 457 ppm to a final value of 630 ppm after the fifth run. While the raw material used to produce the powder was oxygen free (class 1 oxygen free electrolytic grade (OFE) copper [35], C10100) with <5 ppm oxygen, the measured level of oxygen in the powder is similar to that of electrolytic tough pitch (ETP) C11000 with approximately 400–600 ppm oxygen [5], and can be attributed to the thin layer of non-passivating cuprous oxide and the high specific surface area of AM powders. This oxygen pickup can be difficult to avoid in practice and originates in the screening of powder to size fractions suitable for AM as well as from build-to-build powder handling. It is likely that special handling and monitoring or techniques to reduce surface oxides may be required. While the oxygen is elevated in the feedstock powder, it is within the range of recent reports for 3D-printed copper powder [14,31,36], where no AM-related study has reported oxygen values for copper powder that approach the values of C10100.

Figure 3

Particle size distribution of feedstock unalloyed copper powder used across the builds (A), and SEM image of the powder prior to the first build (B); scale bar shows 100 µm.

Table 4

ICP-MS analysis of feedstock unalloyed copper powder prior to the first build. Values are reported in wt. ppm.

Ag	Cd	N	P	Pb	S	Sb	Sn	Zn
10	<1	2	5	2	8	<1	<1	<1

The SEM image shown in Figure 3B illustrates chiefly spherical morphology with scattered satellite particles attached to the powder surfaces, a typical feature of gas atomized metal powder. While the powder distribution is lower than that typically used by EB-PBF (45–106 µm), no detrimental effects on powder bed formation or electrostatic charging due to the size distribution were observed. It should also be noted that the higher electrical conductivity of copper mitigates the charge-induced scattering common in other AM materials [37]. Horn et al. [6] have also previously demonstrated improvements in the surface finish of EB-PBF-produced copper associated with finer powder size distributions.

3.2. Processing Space of Copper in EB-PBF

Figure 4 illustrates the processing space of unalloyed copper via the EB-PBF method reported in the literature and used in this study. A closer look at Figure 4 and properties reported in Table 1 suggests that the reported EB-PBF processing space is indeed narrow and is sensitive to powder characteristics like size distribution and oxygen content that collectively lead to significant variation in properties. As seen in Figure 4, Ramirez et al. [36] studied EB-PBF of an oxidized copper powder with 99.8% purity that was processed at a high substrate temperature and a low beam speed, ultimately leading to an increased hardness of 88 HV in the solids. Lodes et al. and Raab et al. reported a stable process window at substrate temperatures close to 400 °C for fabricating >99.9% pure copper articles at various scan speeds and melt currents resulting in average relative densities greater than 99.95% [27,28]. Guschlbauer et al. [11,12] expanded this processing space and reported mechanical properties data for two different powder types (pure and oxidized), three build orientations (vertical, horizontal, and 45°), and variations in process parameters. The work by Guschlbauer et al. also highlights, in a more systematic way than previous studies, the influence of copper oxides on the strength and ductility in as-fabricated conditions. However, the processing space reported by Ledford et al. [1,26] is in contrast with others as the substrate temperature is relatively low because of the low initial oxygen content of the powder that tends to sinter more easily at high temperature preheat settings. Moreover, in literature it is generally seen that oxygen content in the powder feedstock ranges from 40 wt. ppm [31] to ~2000 wt. ppm [36], and warrants for different processing conditions to be adopted. Small powder size distribution and a powder purity of ~99.95% used in this study led to the use of a low substrate temperature, and as a result, low preheat and melt currents to avoid any flowability issues during the raking process. In terms of volumetric energy density (E), defined as: where P is the beam power in J/s, v is the beam velocity in mm/s, h is the scan line offset in mm and t is the layer thickness in mm, the used parameter set resulted in an average volumetric energy density of 92.3 J/mm3, as shown in Figure 4.

Figure 4

Processing space of EB-PBF fabrication of unalloyed copper by taking the process parameters used in the literature. Major observations of such studies are highlighted along with reported solid densities in Table 1.

3.3. Material Characterization

Figure 5A reports the density of samples from each treatment condition. The error bars shown represent a 95% confidence interval based on pooled variance of all samples. Within each treatment group no difference in specimen density was detected based on sample orientation, build number, or location within the build. Mean density of the specimens ranges from 8.88 to 8.92 g/cm3 (99.4% to 99.9% of theoretical). Specimens subjected to both vacuum-anneal heat treatment and HIP exhibited a slightly higher density (8.91 ± 0.01 and 8.92 ± 0.01 g/cm3, respectively) compared to as-fabricated samples (