A. Characterization of the starting powders

Table II shows the properties of the powders used in the experiments, specifically particle sizes, measured using light scattering particle size analysis, and apparent densities and powder flow rates, measured as described by ASTM standard B212.

Figure 2 shows the powders used for the experiments. Figure 2(a) is a secondary electron image (SEI) showing the irregular shape of the uncoated powder; Fig. 2(b) shows the blend of the same copper powder with smaller and rounder tin particles. Figure 2(c) is a backscattered electron image (BSEI) of the cross-section of the composite powder where the tin-rich areas are brighter. The coating appears to have a relatively uniform thickness of approximately 0.5–1 μm, as well as essentially complete coverage of the surface. The somewhat porous internal structure of the powder is also evident, and is presumed to be the porosity inherent in a powder manufactured by hydrogen reduction from copper oxide. ^{Reference German11} Figure 2(d) is an SEI of the surface of the tin-coated composite powder. The tin covering the surface has a nodular morphology, which is probably the cause of the low apparent density of the composite powder. ^{Reference German12}

FIG. 2.

SEM micrographs of (a) uncoated Cu powder, (b) blended Cu–Sn powder, (c) cross-section of composite Cu–Sn powder, and (d) surface of composite Cu–Sn powder.

The XRD results obtained from both powders are shown in Fig. 3. The blended powder shows only peaks corresponding to pure tin (JCPDS card 00-004-0673) and pure copper (JCPDS card 00-004-0836). The composite powder contains no apparent elemental tin; instead, the tin-rich outer layer detected is Cu₆Sn₅ (η phase, JCPDS card 00-045-1488) while the interior of the powder is elemental copper. Intermediate layers of other intermetallics were not detected by XRD, despite the expectation that at least a thin layer of Cu₃Sn (ε phase) exists as depicted on the phase diagram. The absence of metallic tin and ε has been seen in other types of chemical tin coatings on copper as well. ^{Reference Fujiwara and Enomoto13} A calculation of the tin content for this powder, based on the assumption of a uniform 0.75 μm layer of Cu₆Sn₅ (having a density of 8.28 g/cm³) on a 27 μm diameter copper particle, yields 9.95 wt% Sn, thus showing that this type of coating represents the nominal composition.

FIG. 3.

XRD patterns of composite and blended powders.

B. Evolution of the liquid phase formation during heating for both types of powders

Figure 4 shows DSC data for the blended and composite powders, where both powders were analyzed under apparent density, to mimic as closely as possible the behavior during fabrication. Each sample was heated to 830 °C at a rate of 25 °C/min in nitrogen. The heating rate used was much lower than that expected in SLS, because of the difficulty in obtaining high-quality measurements for very fast heating by this technique. It may be expected, therefore, that the magnitude and temperature of each of the events observed would be different during heating in SLS, particularly those that require substantial diffusion time (for instance, the formation of intermetallic compounds). The endothermic peaks noted for the blended and composite powders are listed in Tables III and IV, respectively, along with the calculated apparent volume of liquid formed based on the integral peak area. This analysis was done by comparing the integrated area (in J/g) to values of enthalpy of fusion from the literature. ^{Reference Zhai, Wang, Geng and Wei14} The results thus obtained are intended as a qualitative comparison of liquid content, rather than a strictly quantitative determination of liquid present for each event.

FIG. 4.

DSC traces for the blended and composite powders. Labels: 1 = Sn ⇒ L, 2 = η → L + ε, 3 = Cu + δ ⇒ γ, 4 = δ ⇒ γ + ζ, 5 = ε ⇒ γ + L, 6 = β ⇒ L + Cu.

TABLE III.

Reactions and apparent liquid volume detected in DSC for blended powder.

TABLE IV.

Reactions and apparent liquid volume detected in DSC for composite powder.

The DSC trace for each powder was interpreted using the well-characterized copper–tin phase diagram. ^{Reference Saunders and Miodownik9,Reference Fürtauer, Li, Cupid and Flandorfer15} The melting of tin in the blended alloy is indicated by the large endothermic peak (peak #1 in Fig. 4) with the onset at 226 °C; this peak is completely absent from the composite powder. The absence of melting in the composite alloy at low temperature is due to rapid interdiffusion of elements ^{Reference Mei, Sunwoo and Morris16} in the coated layer that leads to the formation of the η phase during the coating process. The broad endothermic peak (#2) beginning at 415 °C is present in both powders, and corresponds well to the peritectic melting of the η phase, supporting the idea that the composite powder has formed the η phase through a solid-state transformation. The onset of this peak is also convoluted with the decline of the preceding exotherm (which denotes the formation of η), giving rise to the apparently very large amount of liquid formed (51 vol%); this is likely a much smaller liquid content in reality. The next peak (#3), which is also present in both powders, indicates the eutectoid boundary at 520 °C between δ + Cu and γ. It is followed by a peak (#4) at 589 °C, only present in the blended powder. This peak could result from one of the three events: the eutectoid reaction between the copper solid solution and γ phase, forming the β phase (at 586 °C on the phase diagram – although newer data ^{Reference Fürtauer, Li, Cupid and Flandorfer15} place this reaction at a lower temperature); the eutectoid reaction of ε and δ to make ζ at 582 °C; or the decomposition of δ to make γ and ζ at 603 °C. At 640 °C (peak #5), both powders show the transition from ε to γ with simultaneous melting of γ, and at 680 °C, a small exotherm shows that any remaining ε is converted to γ. The next event (peak #6) occurs at 792 °C, where the only possible reaction, which is much larger in the blended powder, is from β to copper solid solution and liquid. Following this is an exotherm showing dissolution of the liquid phase into copper, which makes estimation of the true liquid content at this point difficult.

In terms of factors affecting liquefaction (and therefore the sintering behavior), the composite powder is most notable for the prolonged absence of liquid during heating. The initial and most tin-rich liquid (92.4 wt%) is formed at 420 °C, a difference of more than 7 wt% in composition and nearly 200 °C compared to the melting of pure tin. At least four of the intermetallic phases (η, ε, δ, and γ) are apparently present at some point during heating. The large speed of diffusion and formation of intermetallics at higher temperatures, in combination with the small length scales inherent in the composite structure, led to a continuously changing composition of both liquid and solids; it must be noted, however, that the much faster heating in SLS probably does not allow enough time for all of these phases to form on heating, even if the small diffusional distances are taken into account.

C. SLS of both types of powders

Figures 5(a) and 5(b) show the powder blend after SLS as a baseline result, in which the bonding was so slight that no solid piece could be removed from the powder bed; because of this, no analysis beyond SEM was performed with the blended alloy. In these images, the only change when compared with the starting powder blend is the attachment of small tin droplets to the copper particles and to one another. This indicates tin melting, but without much tin–copper interaction since times on the order of 10³ s are typically required for good wetting and subsequent homogenization of this system. ^{Reference Hasouna, Nogi and Ogino3}

FIG. 5.

SEM micrographs of blended powder after SLS: (a) low magnification and (b) high magnification.

In the coated powder, shown in Figs. 6(a) and 6(b), there appears to be extensive bonding between particles, largely rounded particles and significant necking apparent. There is obvious connectivity between particles, though the extent of melting is not immediately apparent and there is not a complete elimination of pore space between particles.

FIG. 6.

SEM micrographs of composite powder after SLS: (a) low magnification and (b) high magnification.

Figures 7(a) and 7(b) show backscattered electron image cross sections of the composite powder after SLS at different magnifications. The laser-sintered powder has regions of clear compositional contrast, with tin-rich bands forming necks and connections between copper-rich regions. This indicates that, while melting and bonding of the tin coating did occur, there was no sufficient time for full diffusional homogenization to take place. In addition, when compared with the unsintered powder an increase in thickness of the tin-rich layer from <1 μm to approximately 3 μm can be seen, which is evidence of redistribution of the liquid toward necks and of dissolution of copper into the liquid, ^{Reference Tuah-Poku, Dollar and Massalski17,Reference Zhou18} which in turn shifts the composition of the tin-rich liquid toward the bulk composition (as indicated by DSC measurements). The non-rounded particle shapes and remnant internal porosity visible in the lower magnification image show that the original copper particles were not melted during sintering and that the bonding between particles is occurring only by liquid-phase sintering through the tin-rich layer. The sintered density of the composite material, measured from x-ray tomography, was 36 ± 2% of the theoretical value. The increase in density (of 17%, compared with the loose powder) is further evidence that sintering processes are effective in creating interparticle bonding and densification.

FIG. 7.

SEM micrographs of cross-section of composite powder after SLS at (a) low magnification and (b) high magnification.

Figure 8 shows the XRD pattern of the composite alloy after SLS. There is no indication of tin or η, while the overlapping peaks near 42° showing some combinations of metastable δ, metastable ζ, and metastable β- and γ-related phases (JCPDS card 04-007-1295 for δ, card 04-010-7872 for ζ, and cards 04-004-6308, 01-071-7875, and 04-010-7872 for the β- and γ-type phases, as well as other Refs. Reference Kennon and Bowles19,Reference Saunders and Miodownik20) are very pronounced. There is also a small peak at 38° indicating that the ε phase (JCPDS card 04-001-2885) is present. The absence of both η and tin shows that the degree of homogenization in this sintered product is closer to equilibrium than in the starting powder. This is likely due to the short distances required for diffusion and the large contact area between the tin-rich and copper phases. The copper peak is still present for the sintered powder, and is not significantly shifted by a change in lattice parameter, showing that there is relatively little tin in solid solution after sintering. This is reasonable, because the extensive formation of a copper solid solution would require the solid-state interdiffusion of tin and copper across distances on the order of tens of micrometers, while the formation of the intermetallic phases can occur directly from the solidification of the tin-rich liquid. This is also verified by the two-layered structure as can be seen in Fig. 7. The XRD results also confirm what is shown by DSC: that the composite SLS specimens, on reheating, contain ε, ζ, γ, and β phases as well as a copper solid solution, and that there is no apparent η or tin phase remaining.

FIG. 8.

XRD pattern of composite powder after SLS.

Figure 9 shows the DSC curve for the composite powder after SLS. After reheating the laser-sintered compact of composite powder to 830 °C at 25 °C/min, only two events are noted. The endothermic peak (#1 in Fig. 9) has an onset at 640 °C, where the simultaneous formation (from ε) and melting of the γ phase begin. This is followed closely by an exotherm indicating any remnant ε phase transforming to γ, which then almost immediately begins melting along with any pre-existing γ phase, as indicated by the downward slope beginning at 708 °C. The lack of a distinct melting event near 800 °C is evidence that homogenization has occurred by this point to yield a solid solution of tin in copper.

FIG. 9.

DSC trace of the composite powder after SLS. Label: 1 = ε ⇒ γ + L.

The apparent absence of the η phase is strong evidence that temperatures higher than 420 °C were reached throughout the sample during laser sintering. Furthermore, the prominence of the ε decomposition peak is an indication that there is a relatively large amount of that phase present after laser sintering (corresponding to 8.2 vol% liquid), which also implies that η has been eliminated.

To evaluate the mechanical properties of the sintered product, the composite powder specimens (having a density of 36 ± 2% of the theoretical value for this alloy, 8.77 g/cm³) were tested in 3-point bending. The flexural strength (based on 3 tests) of these was found to be 6.2 ± 3.7 MPa, with no plastic deformation noted. The relatively low strength might be expected, given the low sintered density of the product as well as the presence of brittle intermetallic phases at particle necks.

After a homogenization treatment (600 °C for 2 h), to diffuse tin out of the intermetallics and into solid solution and to promote further solid-state sintering, the lattice parameter of the specimen was measured by XRD and found to be 0.3671 ± 0.0003 nm, corresponding to a solid solution of 9.1 ± 0.4 wt% of Sn in Cu, ^{Reference Saunders and Miodownik9} while no phases aside from FCC Cu were found. This indicates a fully homogeneous solid solution, which was verified by SEM as shown in Fig. 10. The 3-point bending strength, calculated with 0.2% offset strain, increased after the homogenization treatment to 14.7 ± 2.2 MPa (with 3 specimens tested), while the density remained essentially unchanged at 39 ± 2% of the theoretical value. More significantly, the samples that had been homogenized were ductile and did not fracture during testing. The bending angle in each case was limited by the geometry of the testing jig to 100°. The substantial ductility of these samples is shown in their flexural stress–strain curves, an example of which is shown in Fig. 11. The inflection point at which the sample reaches maximum bending angle and contacts the jig is also noted.

FIG. 10.

Backscattered SEM image of cross-section of homogenized SLS specimen.

FIG. 11.

Flexural stress–strain curve for 3-point bending of homogenized SLS sample.

Fabrication of highly porous copper-based materials for applications in filtration, electrical, and heat-transfer applications has been carried out by a variety of methods ^{Reference Castrodeza and Mapelli21–Reference Parvanian and Panjepour24} ; recently, this has also been attempted by additive manufacturing using electron beam melting. ^{Reference Ramirez, Murr, Li, Tian, Martinez, Martinez, Machado, Gaytan, Medina and Wicker22} The specific yield strengths seen in these materials vary from as low as 0.56 to 15 MPa cm³ g⁻¹, depending on manufacturing method, pore size, and final density. ^{Reference Castrodeza and Mapelli21,Reference Shehata Aly, Almajid, Nakano and Ochiai23} The specific strength achieved with the homogenized samples in the current work (4.3 MPa cm³ g⁻¹) falls in the same range, and as such, shows that the additional versatility and flexibility in design gained by using an additive manufacturing method do not adversely affect the attained properties.