Copper additive manufacturing using MIM feedstock: adjustment of printing, debinding, and sintering parameters for processing dense and defectless parts

Copper MIM feedstock (Cu999 from PolyMIM, Germany) was used as the raw material. This feedstock consists of ~ 3–5 mm granules containing 93.5 wt% of copper and two polymers, namely polyethylene glycol (PEG) and some wax (not precisely specified by the provider). PEG offers flexibility during printing due to its thermoplastic nature. Wax works as the backbone binder to hold the copper particles together after PEG removal. Figure 2 depicts an SEM image of the feedstock and suggests a particle size distribution in the range 2–20 μm for copper particles. Some voids with copper particles and binders in the feedstock granules were also observed. The spherical shape of the particles and their broad size distribution facilitate particle packing in the green body, making it possible to obtain high-density sintered materials [31].

3D printing of the green parts was performed with a screw-based extrusion printing machine supplied by AIM3D, Germany. Samples with dimensions up to 255 mm in the three directions can be obtained. As-received feedstock granules were directly fed into the hopper of the machine. A pneumatic controlled piston was used to push the feedstock in the screw extruder and maintain a homogenous flow. A hardened steel nozzle with 0.4 mm diameter was placed at the extruder end for the filament extrusion. A schematic diagram of the machine is given in Fig. 3a. Gcode was generated from the Simplify3D (USA) software for the control of the movement of the printing head. The CAD model for the printing was first generated with SolidWorks software. The CAD model was converted into the tessellation file with the minimum chordal error. Finally, the tessellated file was imported in Simplify3D software for orientation and slicing. Optimization of printing parameters was discussed in a previous paper [29]. The design of an experiment-based approach was used to obtain optimal parameters for maximum green density and minimum surface roughness. The optimal parameters were found to be a layer thickness of 0.05 mm, a nozzle speed of 20 mm/s, an extrusion flow rate of 120% of the standard value, an extrusion temperature of 196°C, with a bed temperature of 60°C, and an infill density of 100%. Furthermore, extrusion width and overlapping were adjusted in the present work to 0.51 mm and 65% to obtain a higher green density. A printing scan with a closed contour was chosen. The two perimeter contour shells were printed in each layer from outside to inside. A scan raster with a 45° angle offset on each layer was used for the infill pattern. The printing pattern design is displayed in Fig. 3b. For the first layer printing, the nozzle speed was set at 50% of the actual speed for better adhesion to the printing bed. For the present study, cylindrical samples with 8 mm diameter and 4 mm height were fabricated with optimized printing parameters for characterization and subsequent debinding and sintering. A sample with non-optimized, arbitrarily selected parameters (layer thickness 0.2 mm, nozzle speed 60 mm/s, extrusion multiplier 100%, extrusion temperature 210 °C) was also fabricated to highlight the benefit of the optimization process.

The debinding process was performed in two steps: solvent and thermal debinding, respectively, to remove all binders from the green body. Solvent debinding was performed to extract most of the PEG polymer. The samples were immersed in water with magnetic stirring for 12 h at room temperature, 40 or 60 °C. Then, the samples were dried in an oven at 100 °C for 2 h. The weight of the samples was measured before and afterwards, and the weight loss was calculated. Thermal and sintering steps were performed in two different tube furnaces to avoid any carbon deposition of binder residue on the sintered sample, which could decrease the density and purity of the sintered copper. The thermal cycles for debinding and sintering are shown in Fig. 4. First, the thermal debinding was performed to remove the backbone binder with a heating cycle of 1 °C/min up to 500°C and isothermal time of 1 h followed by cooling to room temperature at 4 °C/min rate. The thermally debinded samples were placed carefully in the sintering furnace. Sintering was performed by heating at 4 °C/min up to 950, 1000, 1030, or 1050 °C with a holding time of 3 h and then cooling at 4 °C/min to room temperature. The thermal debinding and sintering cycles were performed in He-4%H₂ atmosphere.

Various characterizations have been employed in the present study to explore the characteristics of the 3D-printed, debinded, and sintered samples. Scanning electron microscopy (SEM) images of cross-sections of the material at different processing steps were taken with a JEOL JSM-IT 500 HR apparatus to study the surface porosity of the samples at different stages of the process. The green density was calculated by measuring the mass and the volume of the printed samples. High green density ensures the lower number of voids in the sample. The mass was determined with an accuracy of 0.001 g, and the volume was deduced from the dimensions measured using a Vernier caliper with an accuracy of 0.01 mm. The surface roughness of the samples in the printing direction was measured using an Olympus DSX500 optical microscope as per the ISO 4287 standard. The surface roughness of the samples was studied to report the quality of the process. The weight density (ρ) of the sintered samples was calculated by the 3-mass Archimedes method in air and ethanol. The relative density of the samples was calculated by reporting to the solid copper density (8.96 g/cm³). For mass loss during thermal debinding, thermogravimetric analysis (TGA) was performed in He-4%H₂ atmosphere. The experiments and characterizations performed at different stages were carried out three times, and the average and standard deviation were calculated. X-ray diffraction (XRD) analysis was performed using X'Pert Pro MPD from PANalytical to check the purity of the copper sample. The samples were scanned from 25 to 100° angle at 0.97°/min to capture phases with small intensity. For the 3D observation of the samples, microtomography technique was used to capture the porosity of the 3D-printed and sintered samples. Many authors have explored the technique in additive manufacturing and sintering for the evaluation of porosity at different stages [24, 32, 33]. The microtomography machine with a laboratory X-ray source (Easytom XL) was used for the 3D scanning of green and sintered parts. The scan was operated at 150 kV with a voxel size of 5.2 μm (or 0.93 μm for a few high-resolution images). In the analysis of tomography results, ImageJ and Avizo Lite software were used. First, the raw μCT files were imported into the ImageJ software. A thresholding operation was carried out to capture the voids or pores in the samples. The term ‘void’ is given for the 3D-printed green samples, which contain copper particles and binder and ‘pore’ for the sintered samples, which only contain copper. After thresholding, the voids/pores were segmented by labeling with 3D connectivity as per the algorithm given by Boulos et al. [34]. The volume (V) and surface area (S_A) of each void/pore were calculated from the segmented data. The equivalent diameter (d_eq) and sphericity (S) (lying between 0 and 1, where 1 denotes a perfect sphere) were calculated as per Eqs, 1 and 2. Furthermore, the data were imported in the Avizo Lite for visualization of voids or pores in 3D.

The SEM image of the extruded material with the 0.4-mm nozzle is shown in Fig. 5a. Voids larger than the particle size (typically 10–20 μm), named as ‘extrusion voids’, are observed on the surface of the extruded filament. The melted feedstock is indeed strongly sheared inside the extruder, and the formation of voids may result from the resulting complex state of stress and also from the variation of viscosity as the filament spreads out of the nozzle and solidifies on the deposited layers. Two kinds of voids are observed on the printed green part (Fig. 5b), the ‘extrusion voids’ already mentioned in the printed filaments and ‘printing voids’ between the printed layers. Several printing parameters affect the green density during printing. The effect and optimization of significant parameters such as layer thickness, nozzle speed, extrusion multiplier (flow rate), and extrusion temperature were studied for the green density and surface roughness of the sample in a previous study [29]. The optimized results were obtained to be within 5.42 ± 0.11 g/cm³ for green density and 3.44 ± 1.59 μm for surface roughness. The optimal parameters with furthermore adjustments of overlapping and extrusion width resulted in the fabrication of a green body with a weight density of 5.7 g/cm³ and a surface roughness of 2.12 μm. The green density is thus improved, as compared to the previous study. However, the surface roughness can be varied for the overhang parts, in which the support structure could decrease the surface quality [35]. The SEM image of the sample printed with optimized parameters is shown in Fig. 4b. The sample printed with arbitrary parameters had a weight density of 4.16 g/cm³ and a surface roughness of 26.8 μm. From the weight density of the printed samples and the total weight loss during debinding (see next section), the relative density of copper particle packing with respect to solid copper can be estimated as 60 % and 43 % with optimized and non-optimized parameters, respectively.

To validate the optimized parameters and to study the internal voids of the printing parts, a micro-tomography analysis was performed. 3D renderings of the μCT scans are shown in Fig. 6a, b, for optimized and non-optimized conditions, respectively, with different magnifications. The various phases of the materials, like metal particles, polymers and voids, can be distinguished. Only voids are of interest in the present study. Therefore, the voids were extracted by thresholding and displayed in red color. The volume fraction of voids was calculated as 26% in the non-optimized sample with an average void equivalent diameter of ~ 32 μm. The frequency of the void diameters for the non-optimized sample is shown in Fig. 6c. For the optimized sample showing much smaller voids (typically < 20 μm) with a volume fraction less than 0.8%, the image contrast and resolution were considered not to be high enough for a relevant calculation of void size distribution. The sample with un-optimized parameters resulted in interconnected voids (they can be easily identified at higher magnification in Fig. 6a) between the layers and large voids in the extruded filaments. On the other side, the sample with optimized parameters resulted in a dense structure with tiny voids.

After printing, the samples were immersed in water for solvent debinding to remove PEG (water-soluble) polymer. Figure 6a depicts the effect of immersion time and temperature on the weight loss of the green body during this step. PEG is almost completely dissolved in water after a few hours of immersion. The dissolution rate increased with increasing temperature. The change in weight loss gets smaller after 10 h. Therefore, to reduce the processing time, a duration of 10 h at 60 °C was selected for experiments. After solvent debinding, the weight loss was ~ 3.8%, and the sample had enough strength to be handled.

Figure 7b, c depicts SEM images of green (3D printed) and brown (solvent debinded) materials. PEG and wax were observed in the green body. Interconnected voids were observed after solvent debinding The interconnected capillary voids could allow the flow of melted backbone polymer and gaseous decomposition products of degradation of the remaining binder harmlessly in a short time [36]. Also, the SEM image in Fig. 7b explains the strength of the sample after solvent debinding. Copper particles are still bonded to each other by the backbone polymer, ensuring interparticle bonds. Similar morphology changes and formation of interconnected voids have been reported in MIM literature [37, 38].

To remove the PEG residue and the remaining backbone polymer, a thermal debinding step was performed. Thanks to the open porosity created by the solvent debinding, the liquid and the gas fumes resulting from the heating of remaining polymers (PEG residues and wax) could escape [39]. A fast-heating rate of 4°C/min was first used and resulted in the formation of cracks on the surface, likely due to the rapid flow of melted backbone polymer, which damaged the structure. The heating rate of 1°C/min was observed to be an optimal value to perform defect-free thermal debinding. To study the weight loss during thermal debinding, TGA of the solvent debinded sample was performed. Figure 8a depicts the in-situ weight loss during a TGA test. The weight loss from 100 to 300°C is attributed to the decomposition of PEG residues remaining after solvent debinding. The sharp weight loss from 300 to 425°C is likely due to the removal of the backbone polymer. A total weight loss of 2.7% was observed during thermal debinding. Figure 8b depicts the SEM image of the sample after thermal debinding at 500 °C. The particles were free from any polymer. The total weight loss during solvent and thermal debinding was ~ 6.5%. After these steps, the material had low strength, and it must be handled carefully.

Direct thermal debinding and sintering of a sample without solvent debinding was also tried. However, the high amount of polymer to be extracted during thermal debinding resulted in the formation of cracks and in swelling, as shown in Fig. S1. Therefore, solvent debinding is a necessary step for the complete debinding of polymers without any defect formation.

After solvent and thermal debinding, sintering was performed for material consolidation and densification. It was carried out in a high flow of He-4%H₂ atmosphere to avoid any oxide formation at the particle surface. Sintering is driven by atomic diffusion mechanisms, which are thermally activated [31]. The relative density calculated as per Archimedes’ method after sintering at 950, 1000, 1030, and 1050°C during 3 h is shown in Fig. 9a. As expected, the density increases with increasing sintering temperature. The maximum relative density of 94.5% was achieved after sintering at 1050 °C for 3 h. Similar behavior of density with respect to sintering temperature was observed by Tang et al. [40] and Singh and Pandey [17].

A few feedstock granules were also sintered with similar debinding and sintering conditions. The resulting relative density was within the standard deviation of the printed samples (Fig. 9a). The residual porosity in sintered samples should then result from incomplete sintering rather than from large extrusion or printing voids which would have been challenging to eliminate. However, the small size of the feedstock granules could have made a bias in the calculation of the density by the Archimedes’ method.

The linear shrinkage was also calculated for different sintering temperatures (Fig. 9a). The shrinkage increased with increasing sintering temperature, in accordance with the density increase. The shrinkage was observed to be approximately isotropic. Therefore, compensation of ~ 13.1% in each axis could be added to the CAD model during slicing to get the desired dimensions of the sintered body with high density. Also, the surface roughness of the sintered sample was measured as ~ 2.9 μm. XRD diffraction on the sample sintered at 1050°C highlighted pure copper face-centered cubic phase (Fig. 9b). The sintered body was found to be free from any contamination or oxide formation. Figure 9c shows several complex shapes fabricated by the 3D printing method. This shows the efficacy of the process to fabricate pure copper customized parts.

For an in-depth study of porosity, pore size, and pore shape of sintered samples, a μCT analysis was also performed with samples sintered after 3D printing and with sintered feedstock granules. The μCT reconstruction and porosity extraction are shown in Fig. 10. The pores in feedstock granules are observed to be more randomly distributed (see high-magnification image in Fig. 10b) than in the 3D printed part, where the formation of small chains of pores can be seen at certain positions (enlarged image in Fig. 10a), which could be due to the 3D printing defects. The images with 5.2 μm voxel size, i.e., the same order of magnitude as the initial copper particle (5.87 μm), are unable to catch the tiny pores: the porosity calculated with these images is ~ 1.8 and ~ 1.0% for the sintered 3D print body and feedstock granules, respectively, which is much smaller than the one deduced from overall relative density measurement (5.5%). The porosity appears to be mainly constituted by tiny voids, which are below 10 μm. The 3D print sintered sample was machined to 1 mm diameter to enable higher resolution of the μCT scan. The reconstructed body and extracted porosity with low voxel size (0.9 μm) are shown in Fig. 10c. The porosity was calculated as ~ 4.1%, which is close to the value calculated from relative density (5.5%).

Analysis of the size, shape, and orientation of the pores for the sintered body, the sintered feedstock granules, and the sintered body at high resolution is shown in Fig. 11. The average equivalent diameter of large pores, calculated from 5.2 μm resolution images, is ~ 15 μm for the sintered body. A similar value of ~ 16 μm is obtained for the sintered granules. Figure 11a represents the frequency distribution of the pore diameter in the sintered body and in the sintered granules. The size distributions at low resolution are similar for both materials. A higher fraction of pores more prominent than 25 μm can be observed in the sintered granules. This means that the remaining large pores (> 10 μm) after sintering are not due to the 3D printing, as already assumed from the density values. The average equivalent diameter of pores in the sintered sample at high resolution was calculated to be ~ 7.5 μm. The higher-resolution image (Fig. 11b) clearly shows that most of the porosities are in the range 3–10 μm, which was not visible in the low-resolution images.

The pore morphology in the sintered body and the feedstock and granules were observed to significantly depart from spherical shape, with a significant contribution in a 0.6–0.8 range of sphericity (Fig. 11c). The small pores captured at high resolution in the sintered body appear to be more spherical, with a significant amount in the 0.75–1 sphericity range. The small pores are isolated or spherical, and characteristics of the final stage of sintering [31].

Figure 11d compares the pore orientation distribution in the sintered body, the sintered feedstock, and the sintered body at high resolution. Each pore orientation is defined as the direction along the ellipsoid’s longest axis [41]. The 3D orientation is represented by the angle of altitude (the angle of the orientation vector relative to the vertical axis or direction of the printing). A slight angle of altitude near 0° means that the pores are vertically oriented, whereas a pore that is smooth on the horizontal plane is around 90°. The low-resolution data, which captured the large pores, led to an almost homogeneous distribution of pore orientation in the sintered granules but a significantly larger amount of orientation between 45 and 90° in the sintered printed body. This difference in pore orientation suggests that sintering did not entirely delete the initial local anisotropy of pores introduced by 3D printing. It could also be due to the rearrangement of the particles in the 3D printed sample during the sintering process. In the high-resolution image, where the number of tiny pores is much larger, a homogeneous distribution of orientations is observed, which indicates that the 3D printing process does not influence the final minor porosity.