3.1 Density Changes and Porosity
The initial filament used for this study contains a maximum of 13 wt% polymer and binder materials that are thermally degraded and removed, resulting in a reduction in sample size. Table 2 and Fig. 6 show the results of the manufactured discs used for this analysis that were measured immediately following both printing and sintering steps to avoid potential differences as a result of atmosphere changes which can be particularly problematic in the green condition. Figure 6 also compares the green and sintered density results to pure annealed copper [41]. The green samples show a low density as compared to all sinter conditions as expected by the presence of the additives and the larger volume than present in the sintered states. This low variation in density indicates that the printing process is stable and repeatable across all samples. The high sinter samples were found to have the highest density and lowest variation across all sinter conditions while the control samples produced the lowest shrinkage and highest variation. This improved consistency is expected as the higher temperature would provide additional energy to assist with diffusion of the copper particles and results in a stronger connection.
Table 2
Average shrinkage disc results for each tested condition
Condition | N Samples | Volume (cm3) | Volumetric Shrinkage (%) | Density (g/cm3) |
Low | 8 | 0.219 ± 0.007 | 31.2 ± 1.4 | 6.31 ± 0.22 |
Control | 8 | 0.217 ± 0.005 | 29.5 ± 1.6 | 6.41 ± 0.15 |
High | 8 | 0.211 ± 0.002 | 31.1 ± 0.5 | 6.57 ± 0.11 |
Green | 24 | 0.306 ± 0.003 | | 5.03 ± 0.06 |
In comparison to the green samples, all sinter conditions show an increase in density as the PLA and binders are removed and mostly copper remains. Comparison of the density between sinter conditions shows that the high sinter condition exhibits the highest density with the smallest variation as expected based on volume loss. Following the same logic, it would be expected for the control condition to have the lowest density since it exhibited the least shrinkage, however, it is the low sinter condition which shows the lowest density and highest variation between the sintered condition. While the low sinter condition shows a decrease in density and high variation as expected, this is counter to the observed shrinkage where the control sample shrinks less. This would indicate that the reason for the lower density is the result of a lack of diffusion between copper particles resulting in a likely increase in porosity in both the control and low sinter conditions with no significant difference between them as tested using a one-way ANOVA. Figure 7 shows an example of the shrinkage observed in the dog bone samples as they are going from green, debound (brown), and finally sintered for the control condition, as well as the sintered dog bone sample in its polished state prior to uniaxial testing. The shrinkage results and this figure show that while the dog bone samples were printed to match ASTM-D638 Type V [39] specimens, the final tensile specimens were approximately 90% the scale of dimensions provided in Fig. 2.
The level of porosity in the gage section from each condition can be seen in Fig. 8 and Table 3 numerically shows the difference in pores between conditions. All detected pores that were less than three pixels were removed from the results. Views in Fig. 8 are from side views that are parallel to the FFF printing layer lines where there is a clear elongation of the pores in line with the printing layers. Figure 8a-c shows all pores identified while Fig. 8d-f shows only the larges 30 voids recorded.
Table 3
CT Scan Porosity Results of Copper Dog Bone Gage Sections
Sintered Sample | Pore Count | Overall Porosity (%) | Total Pore Volume (mm3) | Top 30 Pore Volume (mm3) |
Low | 9844 | 3.07 | 1.919 | 1.585 |
Control | 10225 | 2.31 | 1.738 | 1.300 |
High | 12107 | 3.09 | 2.322 | 1.784 |
When looking at the present voids in all samples, regardless of condition, it is clear they all contain large quantities of both large and small voids. There is no clear trend in the porosity of the samples in each condition from the single scan, but rather it shows the significant porosity present as a result of the manufacturing method that cannot be removed from an increase in sintering temperature alone. The use of pressure would likely help reduce the porosity further but was not a part of the scope.
3.2 Hardness
Rockwell Hardness results could not be obtained on any Rockwell scale. This can be explained by the low density and high porosity as compared to annealed copper discussed in the previous section. It is expected that during indentation, the pores below the top surface collapsed under the applied loads making it impossible to record a reliable reading. Micro-hardness results were able to be obtained as they interrogate a smaller region of the material, and the results reveal the similarities between the material produced in the sintering process and those obtained on annealed pure copper [41]. The sintered samples show a downward trend in hardness as the sintering temperature increased. Results of the HV hardness testing can be seen in Table 4. and show a difference between the sinter conditions, but no significant effect on where the indent was conducted in the sample. Figure 9 compares the results of green and all sinter condition samples hardness to annealed pure copper [41].
Table 4
HV Hardness results for sintered and green discs
Condition | N Samples | Measurements Per Zone | Zone 1 | HV Zone 2 | 50 Zone 3 | Overall |
Low | 3 | 30 | 65.5 ± 15.6 | 72.7 ± 22.8 | 69.1 ± 28.4 | 69.1 ± 23.1 |
Control | 3 | 30 | 54.5 ± 16.0 | 55.6 ± 10.5 | 55.1 ± 12.0 | 55.1 ± 13.0 |
High | 3 | 30 | 50.9 ± 13.4 | 51.4 ± 11.6 | 55.1 ± 15.6 | 52.4 ± 13.8 |
Green | 3 | 9 | 102.6 ± 26.3 | 111.4 ± 29.6 | 102.2 ± 33.3 | 105.4 ± 30.1 |
As all heating and hold times remained the same and only the hold temperature changed, the reduction in hardness is likely a result of the furnace cooling process that would be slower for a sample that is cooling from a higher temperature. This would allow for more annealing to occur as the furnace cools naturally resulting in slightly longer times between 200 and 400°C. Moreover, once the control temperature is reached, there is little improvement from the control in hardness which indicates that sintering at temperatures above the control yields no significant change in quality of sinter. The results indicate that this process of obtaining additively manufactured copper can result in hardness values on par with traditionally manufactured copper.
3.3 Tensile Behavior
The higher density coupled with a lower hardness indicate that the high sinter condition should be more ductile and have higher yield and ultimate strengths compared to the other two conditions. The monotonic tensile behavior revealed such generally expected differences between the behavior of different sinter conditions. The stress-strain plots in Fig. 10 show each sample of the three sintered dog bone conditions. From the tensile testing, Sy, Sult, E, and εfail were extracted to be compared to both green samples and literature values of traditional annealed copper [41] in Table 5 showing a clear mechanical advantage for the high sinter condition in comparison to the other sinter conditions. Given the variability in E for these materials, mean values of E and Sy are provided with the variability as inner quartile range (IQR) as opposed to Sult and εfail which are provided with standard deviation.
Table 5
This study’s tensile parameter results based on sample condition compared to literature values of annealed pure copper [41]
Sinter Condition | N Samples | Sy (MPa) | E (GPa) | Sult (MPa) | ε at Fail (%) |
Mean | IQR | Mean | IQR |
Low | 7 | 19.4 | 2.7 | 28.9 | 16.0 | 55.7 ± 7.9 | 7.7 ± 1.0 |
Control | 7 | 18.8 | 4.4 | 26.3 | 15.1 | 43.3 ± 11.5 | 6.1 ± 2.1 |
High | 7 | 21.0 | 2.5 | 56.4 | 42.4 | 70.7 ± 6.5 | 11.2 ± 1.8 |
Green | 3 | 4.4 | 0.3 | 11.1 | 10.6 | 6.2 ± 0.9 | 0.4 ± 0.2 |
Annealed Pure Copper [41] | 33.3 | - | 110 | - | 210 | 60 |
All conditions exhibit substantial drops in stress during testing in the plastic deformation region. These drops occur just after the yield point in all conditions while they also appear near final fracture in the low sinter condition. Additionally, this behavior is less prevalent in the high sinter condition and does not occur in annealed copper as demonstrated by [42]. While these drops are not expected in traditionally manufactured copper material, they can be explained when compared to other material systems that are additively manufactured or in lattice type structures [43, 44]. These load drops are consistent with materials that experience incremental failures or serrations in their stress-strain curves as the area carrying the load fails and the stress gets redistributed in the material [45–47]. In this case, the porous structure resulted in many voids that grow and coalesce such as seen in [47, 48]. When these coalesced pores reach a critical size a significant fracture event of the primary load pathway can occur resulting in a sudden release of energy that is observed as a sudden drop in the load carrying capacity of the material in a displacement-controlled test. Once the stresses are redistributed to new load paths in the remaining material the load will increase as before [47]. The high sinter condition shows less of this effect because of the higher density that results in more load paths existing at the start of the test making any load drops less severe. In addition to the load behavior over the load curve, the difference in the ultimate tensile strength and the stress at failure can be seen in Fig. 11a,b which shows the high sinter condition exhibiting improved strength and ductility as compared to the other conditions. Moreover, each condition exhibited differences in mean strength as quantified in Table 5, however, as a result of variation, there is overlap observed between the behavior of all conditions particularly in the range of 8–9% strain in Fig. 11a,b where all three conditions are present.
Shown in Fig. 12, the mean values of yield stress were found to be quite similar, with a notable decrease in spread in the high sinter samples. This indicates that the yield stress for all conditions is consistent as is expected given all samples are the same material with identical print conditions. The reduced variation observed in the high sinter condition is a result of the higher material density and suggests the variation in measured yield strength is a function of the pore size and distribution present after sintering. It is also likely an indication of the quality of bonding between copper particles after sintering as a function of sinter soak temperature.
Another potential cause of variation within the tensile specimens is the presence of contaminants from our ambient thermal process as mentioned by [21, 22, 26]. To quantify the potential influence of contaminants within our samples, EDS was conducted on the fracture surface of the control sintered dog bone condition. The elemental composition of the dog bone’s gage section was measured at the outer edge and core, while this measurement quantity is not enough to provide a full representation of the sample population, it is sufficient to acknowledge the presence of contaminants that contribute to the final wt%. The results near the surface (edge) and in the center (core) of the sample are shown in Table 6 for the three primary elements present. The quantity of oxygen present is from a combination of the ambient air exposure and from the degradation of the PLA binder, which is primarily composed of oxygen and carbon. It is clear that the concentration of the contaminants are higher near the surface than the core and that these will result in material properties that are not the same as pure copper, however, the presence of the contaminants that result from this sintering process in ambient conditions demonstrates mechanical properties that are in the range of pure copper and could allow for the method to be used in similar applications.
Table 6
EDS Elements wt% Results of Control Sinter Dogbone Fracture Surface
Location | Elemental wt% |
Cu | C | O |
Edge | 61.44 | 30.72 | 7.32 |
Core | 69.82 | 25.30 | 4.88 |
3.4 Nondestructive Evaluation Results
Significant variation in AE activity was observed not only between conditions on average, but also within each condition. This is expected based on the variability in porosity and subsequently the stress strain behavior that is common in additively manufactured components [15, 49, 50]. Figure 13 shows the variation in the form of AE absolute energy accumulation across and within sinter conditions where each row is a different sinter condition. The columns represent the test results with the least, median, and maximum AE observed AE activity as defined by the number or recorder hits.
In general the samples that produced the least amount of activity Fig. 13a,d,and g show intermittent instantaneous jumps in energy throughout the loading. In the case of the low sinter condition, this type of activity was observed regardless of the number of AE events recorded suggesting periodic large fracture events occur as expected from the observed stress drops in the stress-strain curve. The control sinter samples show that there is a more progressive failure process as the amount of AE activity increases. The increase in activity and lack of energy jumps observed particularly in Fig. 13f combined with the lack of any observed stress drop suggests that this material fails more gradually like a traditionally manufactured metals typically do and does not have large fracture events resulting from the porosity present. The large jump in energy seen at final fracture is expected in any ductile metal. This is further observed in the high sinter conditions where in the median activity case we are seeing a more progressive failure than observed in the other conditions for the same level of defined activity. The difference between conditions can be seen more clearly in Fig. 14 where AE amplitude, peak frequency, and absolute energy for the median AE activity cases.
Figure 14 shows similar frequency bands and a clear reduction in the AE activity in the median case between sinter conditions, however, general trends remain the same in amplitude and Peak frequency. As sinter temperature increases the density also increases allowing for an increase in load pathways and properties that trend to annealed copper [41]. As annealed copper is a highly ductile metal that exhibits minimal AE amplitude as compared to brittle materials as demonstrated by [51, 52], it is expected that as ductility increases AE activity will reduce due to the detectability limit of the method. In this work, an increase in ductility is observed to correlate with the sintering temperature. In all conditions, most of the activity occurs near the yield point of the samples as is consistent with AE observed in metal fracture [37, 53, 54]. This trend is strongest in the high sinter condition that shows activity reduces drastically just after yield which is consistent with AE observations in metallic materials [37, 55] where crack nucleation activity is concentrated at this point. As with previous observations this trend indicates an improved bond between the sintered copper particles with an increase in sinter temperature [56, 57]. While there is significant reduction in activity in the high sinter case, the energy continues to grow gradually over the life of the sample suggesting that there is progressive growth occurring after the nucleation at yield. Both the control and low sinter conditions show relatively consistent activity in the amplitude and peak frequency up to the ultimate strength before dying off indicating crack nucleation and brittle type failures occur to a much later stress than observed in the high sinter condition. The most notable AE trends were seen in the energy outputs of the samples, which are quantified in Table 7 and are based on the mean across all samples in each condition. The most notable differences in this data include the total energy and absolute energy per hit of the low samples being on a different order of magnitude in comparison to other conditions. However, the median energy per hit has the lowest value on the low sinter, which suggests that this condition resulted in a lot of low energy hits with a few high energy hits likely from significant brittle like failure resulting from a weaker particle bond during sinter that dominates the total energy trend. With the results shown in Table 8, it should also be noted that the results of each sample, within the same condition, were highly variable. Table 8 shows the range of hit quantities that were seen across samples of the same condition. For both Tables 7and 8, that the data show is representative of the full tensile test and not only on hits prior to the εfail cut-off.
Table 7
Average AE Energy Results
Condition | N Samples | Mean Abs Energy (aJ) | Median Abs Energy (aJ) | Total Energy (aJ) |
Control | 7 | 1.82 | 0.37 | 914.63 |
Low | 7 | 29.79 | 0.33 | 17417.18 |
High | 7 | 2.77 | 0.52 | 443.71 |
Table 8
Condition | Min Hits | Max Hits | Median Hits | Mean Hits |
Control | 100 | 2250 | 433 | 847.7 |
Low | 258 | 1490 | 527 | 663.9 |
High | 88 | 445 | 184 | 247.0 |
3.5 Fractography
Each sample’s fracture surface was observed using optical microscopy with the only difference in surfaces being the distribution of voids. No difference in fracture initiation sites was observable. The voids on both fracture surfaces of each sample were extracted using an overlayed polygonal map after tensile testing for further analysis. A void is defined here as any area inside the materials bounding box that has no material. Figure 15 shows an example of the porosity map where the regions highlighted in red in Fig. 15b are the areas considered to be pores. These are the final voids after mechanical loading and thus can be used to compare the presence of the combination of initial and evolved voids in the material. They cannot, however, be directly correlated to initial porosity.
Figure 16 shows the difference in bottom fracture surfaces across all three sinter conditions. Values from the top and bottom surfaces were averaged together to create an overall value for each sample, the mean of the averaged values for each condition were calculated and results of which can be seen in Table 9.
Table 9
Fracture surface porosity results, shows mean values of condition using averaged top and bottom porosities to represent each sample
Condition | N Samples | Porosity (%) | Quantity Pores | Pore Size (µm2) | Total Map Area (µm2) | Total Pore Area (µm2) |
Low | 7 | 15.572 ± 3.407 | 14231 | 73.574 | 6.7E + 06 | 1.0E + 06 |
Control | 7 | 13.372 ± 2.045 | 11494 | 78.543 | 6.7E + 06 | 8.9E + 05 |
High | 7 | 12.825 ± 1.479 | 9293 | 90.365 | 6.4E + 06 | 8.3E + 05 |
When comparing the fracture surface porosity measurements with the total AE absolute energy and hits during testing, no correlation can be observed. Supporting the hypothesis that AE activity is primarily based upon the strength of bonds created between copper particles due to changes in sintering temperature, and not due to internal pore growth with the exception of the large load drops that result from such failure.
The porosity results follow the same trend as hardness where the higher sintering temperature sees the lowest porosity as well as the fewest number of pores per sample, which also agrees with the higher density and improved mechanical properties over the other conditions. The fracture surfaces and values show that the control condition generally has less pores compared to the low condition, but a larger pore size that can explain the control condition’s generally lower mechanical properties than observed in the low condition. These trends presented in Table 9 are also not consistent with the results of CT scan of the dog bone cross sections. From the CT scans, the control sinter sample had a higher pore quantity but lower total pore volume than the low condition. The high sinter condition is also on the opposite end of the porosity and pore count extremes in CT scans in comparison to fracture porosity. The discrepancies of the porosity before and after loading can be used to examine the evolution of said pores as a result of the loading where the pores grow more in the control and high sinter conditions than they do in the low condition over the course of the loading which would explain the progressive failure observed in AE. Moreover, this could be an indication of the superior bonds between individual copper particles and subsequent better quality sintering achieved by the high sinter temperature but would need more CT samples to ensure the trends there are consistent. Based on the density values alone it was expected that the mechanical properties of the control would be better than the low, however, it is shown here that the presence of porosity is not sufficient to explain the behavior. While initial pore position and geometry contribute significantly to the materials response, the evolution of these pores as a response to the loading appears to have the final say in the resulting material behavior.