Metal-based additive manufacturing, or three-dimensional (3D) printing, is a potentially disruptive technology across multiple industries, including the aerospace, biomedical and automotive industries. Building up metal components layer by layer increases design freedom and manufacturing flexibility, thereby enabling complex geometries, increased product customization and shorter time to market, while eliminating traditional economy-of-scale constraints. However, currently only a few alloys, the most relevant being AlSi10Mg, TiAl6V4, CoCr and Inconel 718, can be reliably printed1,2; the vast majority of the more than 5,500 alloys in use today cannot be additively manufactured because the melting and solidification dynamics during the printing process lead to intolerable microstructures with large columnar grains and periodic cracks3,4,5. Here we demonstrate that these issues can be resolved by introducing nanoparticles of nucleants that control solidification during additive manufacturing. We selected the nucleants on the basis of crystallographic information and assembled them onto 7075 and 6061 series aluminium alloy powders. After functionalization with the nucleants, we found that these high-strength aluminium alloys, which were previously incompatible with additive manufacturing, could be processed successfully using selective laser melting. Crack-free, equiaxed (that is, with grains roughly equal in length, width and height), fine-grained microstructures were achieved, resulting in material strengths comparable to that of wrought material. Our approach to metal-based additive manufacturing is applicable to a wide range of alloys and can be implemented using a range of additive machines. It thus provides a foundation for broad industrial applicability, including where electron-beam melting or directed-energy-deposition techniques are used instead of selective laser melting, and will enable additive manufacturing of other alloy systems, such as non-weldable nickel superalloys and intermetallics. Furthermore, this technology could be used in conventional processing such as in joining, casting and injection moulding, in which solidification cracking and hot tearing are also common issues.
In metal-based additive manufacturing, application of a direct energy source, such as a laser or electron beam, to melt alloy powders locally results in solidification rates between 0.1 m s−1 and 5 m s−1, an order of magnitude increase over conventional casting processes. Given that rastering of this direct energy source (such that it follows a pattern of slightly overlapping lines in a back and forth pattern) to continuously fuse successive layers of powder is analogous to welding processes, it is not surprising that the suite of printable metal alloys are limited to those known to be easily weldable. Application of conventional 3D printing methods to ‘unweldable’ high-performance engineering alloys that cannot accommodate these solidification conditions, such as 6000 and 7000 series aluminium alloys and high-strengthening-phase (high-γ′) nickel superalloys, results in microstructures with columnar grains and cracks spanning dozens of successive print layers3,4,5. The limitations of the currently printable alloys, especially with respect to specific strength, fatigue life and fracture toughness, have hindered metal-based additive manufacturing from maturing to its full potential.
During solidification of these unweldable alloys, the primary equilibrium phase solidifies first at a different composition from the bulk liquid. This results in solute enrichment in the liquid near the solidifying interface, locally changing the equilibrium liquidus temperature and producing an unstable, undercooled condition6. As a result, there is a breakdown of the solid–liquid interface leading to cellular or dendritic grain growth with long channels of interdendritic liquid trapped between solidified regions. As temperature and liquid volume fraction decrease, volumetric solidification shrinkage and thermal contraction in these channels produces cavities and hot tearing cracks which may span the entire length of the columnar grain and can propagate through additional intergranular regions7,8 (Fig. 1e).
In contrast, fine equiaxed microstructures more easily accommodate strain in the semi-solid state by suppressing coherency that locks the orientation of these solid dendrites and promotes tearing9. Producing these ideal equiaxed structures requires large amounts of undercooling, which has thus far proven difficult in additive processes where high thermal gradients arise from rastering of a direct energy source in an arbitrary geometric pattern10. Here we present a general approach to control solidification microstructure by promoting nucleation of new grains with nanoparticle grain refiners (Fig. 1d). Alloy powder feedstock particles are decorated with lattice-matched nanoparticles (Fig. 1b) that heterogeneously nucleate the primary equilibrium phases during cooling of the melt pool. By providing a high density of low-energy-barrier heterogeneous nucleation sites ahead of the solidification front, the critical amount of undercooling needed to induce equiaxed growth is decreased11. This allows for a fine equiaxed grain structure that accommodates strain and prevents cracking under otherwise identical solidification conditions. Using this technology enables additive manufacturing of previously unattainable high-performance alloys, such as 7075 or 6061 aluminium, with improved properties over currently available systems.
Aluminium alloys are a good demonstration platform for our approach, because the only printable aluminium alloys are based on the binary Al–Si system and have a wide range of reported properties, but tend to converge around a yield strength of approximately 200 MPa with a low ductility of 4% (refs 1, 12). The exception is Scalmalloy13,14, which relies on alloying additions of scandium, a rare high-cost metal. In contrast, most aluminium alloys used in automotive, aerospace and consumer applications are wrought alloys of the 2000, 5000, 6000 or 7000 series, which can exhibit strengths exceeding 400 MPa and ductility of more than 10% but cannot currently be additively manufactured15,16,17. These systems have low-cost alloying elements (Cu, Mg, Zn, Si) carefully selected to produce complex strengthening phases during subsequent ageing. These same elements promote large solidification ranges, leading to hot tearing during solidification, an issue that has been difficult to surmount for the more than 100 years since the first age-hardenable alloy, duralumin, was developed18,19. The most complete study of elemental effects dates back to the late 1940s; however, the mechanistic effect was not fully described until 1999 by Rappaz, Drezet and Gremaud (RDG)8,20. The RDG model incorporated both deformation of the semi-solid network and fluid backfill to capture the composition and microstructure effects on cavitation-assisted tearing. Additionally, Gourlay and Dahle21 demonstrated experimentally that strain can be accommodated more readily in a fine equiaxed material owing to an increase in the solid fraction at which dendrite coherency occurs and the suppression of large dilatant shear bands which require additional backfilling. Combining the mechanistic effects addressed by Gourlay and Dahle and the predictions of the RDG model to minimize crack susceptibility has not been effective for highly crack-susceptible alloys such as Al7075 and Al6061, owing to a lack of processing paths to produce fine equiaxed grains. We have developed a scalable and alloy-agnostic approach to incorporate grain-refining particles into conventional hot-tear-susceptible alloy powders directly to additively manufacture high-strength crack-free alloys with a fine equiaxed microstructure (Fig. 2). Conventional alloy powders and nanoparticles are electrostatically assembled, producing a powder feedstock with uniformly distributed nanoparticles. Nanoparticle compositions targeted to each alloy were selected using a software tool that identifies matching crystallographic lattice spacing and density to provide a low-energy nucleation barrier on the basis of classical nucleation theory (Fig. 2f). The software analysed more than 4,500 different powder and nanoparticle combinations corresponding to more than 11.5 million matching pairs. Potential matches were sorted by a combined set of constraints: minimized lattice misfit, similar atomic packing along matched crystallographic planes, thermodynamic stability in the desired alloy, and availability. For the aluminium alloys tested, hydrogen-stabilized zirconium particles were selected for their stability in air and ability to decompose at the melting temperature, resulting in formation of the favourable Al3Zr nucleant phase22. This phase has previously been described as a ‘mild’ grain refiner, but can be difficult to incorporate in many aluminium alloys owing to rapid coarsening and a high liquidus temperature, making gas atomization of additive feedstock difficult22,23. In our approach, incorporation of this particulate at the instant of melting provides a high level of mixing and a high density of nucleation sites.
Pre-alloyed gas-atomized 7075 and 6061 spherical powders with an average particle size of 45 μm were coated with 1 vol% hydrogen-stabilized zirconium nucleants using an electrostatic assembly technique to ensure uniform distribution in the powder bed and avoid settling. Assembled powders were additively manufactured via selective laser melting using a Concept Laser M2 400W system with an 80 mm × 80 mm build volume. Standard machine parameters provided by the manufacturer for a conventional AlSi10Mg alloy were used for all nanoparticle-functionalized 7075 and 6061 powders. After completion of the build, components were homogenized on the build plate and aged to a T6 condition in accordance with conventional wrought materials. For direct property comparison, parts were also manufactured from stock 7075, 6061 and AlSi10Mg powders under the same conditions.
Microstructure analysis reveals a substantial difference between components additively manufactured from stock powders and those produced with nanoparticle-functionalized powder (Figs 1 and 3). Stock 7075 and 6061 (Extended Data Figs 6 and 7) exhibit a series of large columnar grains oriented parallel to the build direction, with cracks present in the intercolumnar region and extending through multiple build layers. This is consistent with previously documented attempts at printing wrought aluminium alloys and is driven by the high, directional heat flux in the additive process, which provides high thermal gradients and minimal undercooling during solidification4. Previous additive routes to producing equiaxed grains have focused on manipulating the thermal gradient and solidification velocity to induce substantial undercooling for nucleation of equiaxed microstructures. This requires extensive manipulation of parameters including scan strategy and build temperature and is not extensible to multiple alloy systems, additive hardware or build geometries10,24. Although the solidification velocity is relatively high, it alone is not sufficient to induce equiaxed growth per the conventional Hunt criterion for a columnar-to-equiaxed transition. In particular, the high thermal conductivity of aluminium and the large liquid diffusivities of alloying elements make substantial undercooling extremely difficult to achieve with the accessible ranges of solidification velocities and thermal gradients25. In addition, the Hunt criterion assumes a steady-state solidification front, whereas the additive process deviates substantially from steady state owing to the raster pattern and accumulation of residual heat26. As such, solidification preferentially occurs through nucleation on existing grains, leading to the observed grain growth vertical to the build direction with grains extending across multiple build layers, as in the inverse pole figure map in Fig. 1e.
In contrast, the 7075 and 6061 alloys manufactured with grain-refining nanoparticles show uniform equiaxed growth with no cracking. Upon melting, zirconium particulates are pulled into the melt pool and react to form Al3Zr. Al3Zr has more than 20 matching interfaces with the primary face-centred-cubic aluminium phase, exhibiting less than 0.52% lattice mismatch and 1% variation in atomic density, providing an ideal low-energy heterogeneous nucleation site. Nucleation of new grains ahead of the solidification front requires both an energetically favourable condition and a large number of nucleation sites to ensure new grains can form before the main solidification front overtakes new grains. The columnar growth demonstrated in the unmodified material indicates that undercooling is present, providing an energetically favourable condition; however, without additional nucleation sites, homogenous nucleation requires a substantially higher energy barrier. The large number of low-energy-barrier heterogeneous nucleation sites ahead of the solidification front induces a fine equiaxed structure under the same processing conditions as for the unmodified powder. This results in crack-free microstructure with grain sizes of about 5 μm, 100 times smaller than the grains in the unmodified material (Fig. 1e, f). The nucleant particles are uniformly incorporated into the microstructure, which can provide additional strengthening and resistance to grain growth owing to pinning effects.
The cracking that we observed in the stock material appears consistent with the mechanisms of the RDG model8. Columnar grains grow in the direction of the heat flux, leaving a thin layer of interdendritic fluid and leading to cavity formation (Fig. 3b). Further thermal shrinkage allows this initial cavity to ‘unzip’ and to propagate through interdendritic colonies, resulting in large cracks oriented parallel to the columnar grains8. Although the RDG model does not explicitly describe the effects of equiaxed microstructure on crack susceptibility, the shift to equiaxed growth drastically reduces the effect of entrapped liquid as the grains begin to behave as a low-resistance granular solid21. Fine equiaxed semi-solid structures allow easier grain rotation and deformation, providing a means to accommodate strain in the semisolid state and thus preventing crack initiation and growth. The complete elimination of cracks that we observed is attributed to the change in microstructure. Hot-tearing models, including the RDG model, are dominated by the final stage of solidification when the fraction of solid is greater than 0.8 (ref. 27). Many hot-tear-susceptible materials can be identified from their solidification curves (Fig. 3a). The shapes of these curves are dictated by the compositions of the constituent alloys and can be described using a Scheil–Gulliver solidification model based on the equilibrium phase diagram28. Thermo-Calc software was used to simulate sequential steps from the liquidus temperature to an approximate solidus temperature, calculating the fraction of solid and composition of the new liquid at each point. Susceptible alloys have large solidification ranges between the liquidus and solidus temperatures and sharp turnover in the solidification curves at high fractions of solid. The sharp turnover is typically associated with the increased levels of strengthening solute that partitions in the liquid to a high degree during solidification. Associated thermal shrinkage leads to tearing and cavitation in that thin films of interdendritic liquid that are present at the high solid fraction. Decreasing the solid fraction at which the turnover occurs or reducing the difference between the solidus and liquidus temperatures will improve resistance to tearing. A conventional additive aluminium alloy such as AlSi10Mg has both an early turnover and a small difference in liquidus and solidus temperature, leading to a low tendency for cracking during solidification. This is in stark contrast to 7075 and other hot-crack-susceptible alloys (Extended Data Fig. 5 with Al6061 as well).
The shapes of the solidification curves can be shifted with increasing solidification velocity owing to the non-equilibrium partition coefficients; however, we found no evidence of substantial departure from equilibrium29. The addition of zirconium might be expected to shift the solidification curve into a more favourable shape; however, as shown in Fig. 3c, this is not the case. The Al–Zr binary phase diagram indicates a peritectic reaction at high mass fractions of aluminium22. As such, any Al–Zr reactions occur at the beginning of solidification where tear resistance is not critical owing to the low volume fraction of solid. More importantly, the addition of zirconium does not substantially alter the shape of the solidification curve at high fractions of solid, where hot tearing is initiated. As discussed above, the early inclusion of zirconium induces equiaxed growth, which can more easily accommodate the thermal contraction strains associated with solidification, ultimately resulting in an alloy system that is highly tear resistant, despite conventional wisdom.
We performed tensile testing and compared the results against equivalent specimens produced from unmodified powder to verify the crack-free nature of the additively manufactured material. Figure 4 displays typical stress–strain curves for each material; the associated yield strength, modulus, ultimate tensile strength and elongation to failure are summarized in Table 1. As shown, stock 7075 retains almost no strength owing to the large volume of cracks caused by hot tearing. Conventional AlSi10Mg shows about 7% ductility but less than half the strength of the wrought 7075 system, consistent with data provided by multiple selective laser melting companies. In comparison, additively manufactured 7075 with the incorporation of Al3Zr nucleant particles shows an 80% increase in strength over AlSi10Mg, and is within the expected bounds for its wrought counterpart. The modified Al7075 demonstrates Luders banding during deformation, which is indicative of an aluminium alloy with grain sizes of less than 10 μm (Fig. 4)30. The yield strength and elongation of functionalized Al7075 was produced within reported ranges of wrought Al7075 (Table 1); however, the ultimate strength difference and lower limit of yield strength can be explained by strain softening from the reduced grain size and evaporation of zinc, a major strengthening element, during the laser melting process. Differences between additively manufactured 7075 and conventional wrought material can be remedied by increasing the zinc concentration in the feedstock powder to improve strength and by optimizing the heat treatment to target an optimum final grain size to eliminate strain softening. Likewise, ductility and elastic modulus can be increased by improving processing parameters to reduce the porosity caused by excessive laser energy density and trapped gas, a feature that was not optimized in this study. Additively manufactured metal parts are often hot-isostatic-pressed to reduce porosity and to improve properties, but this was not carried out for this study to preserve equivalent processing conditions between functionalized and stock aluminium powders.
We have used secondary particulates in additive manufacturing to induce grain refinement of high-strength aluminium alloys of wrought compositions, producing crack-free materials with strengths double that of the most common additively manufactured aluminium alloy. This metallurgical approach is applicable to other industrially relevant crack-susceptible alloys and can be extended to new families of additive manufacturing materials such as non-weldable nickel alloys, superalloys and intermetallics. Furthermore, our approach provides a metallurgical tool for metals processing, affording a diverse range of alloys for additive manufacturing, accelerating broad adoption of additive processes and enabling the design of new alloy systems specifically for additive processing.
Aluminium alloy 7075 micropowder. Aluminium alloy 7075 micropowder was purchased from Valimet Inc. The powder consisted of Al (balance), Zn (5.40%), Mg (2.25%), Cu (1.54%), Cr (0.19%), Fe (0.17%), Si (0.13%), Mn (0.02%) and Ti (<0.01%), in weight per cent. The particle size distribution was bimodal with peak values at 45 μm and 15 μm (Extended Data Fig. 1).
Aluminium alloy 6061 micropowder. Aluminium alloy 6061 micropowder was purchased from Valimet Inc. The powder consisted of Al (balance), Mg (0.83%), Si (0.62%), Fe (0.25%), Cu (0.23%), Cr (0.08%), Mn (0.04%), Zn (0.04%) and Ti (0.02%), in weight per cent. The average particle size was 45 μm.
CL31 aluminium–silicon–magnesium alloy micropowder. Aluminium–silicon–magnesium alloy micropowder was purchased from Concept Laser Inc. The powder consisted of Al (balance), Si (9.0%–10.0%), Mg (0.2%–0.45%), Fe(<0.55%, trace), Mn(<0.45%, trace) and Ti(<0.15%, trace), in weight per cent. Particle size optimized for selective laser melting and proprietary to the manufacturer.
Hydrogen-stabilized zirconium. ZrH2 powder was purchased from US Research Nanomaterials Inc.
To determine the particulate compositions and crystallographic faces with the highest probability of inducing epitaxial, heterogeneous nucleation in the given alloy system, we developed software with Citrine Informatics that uses lattice matching algorithms to search through crystallographic databases for the highest matching crystal structures. By minimizing the lattice strain between an inoculating nanoparticulate and the nucleating phase of a molten material, the free-energy barrier to nucleation is reduced, enabling higher nucleation rates for any given undercooling. Lattice matching is calculated through an area strain measurement that accounts for the change in surface area of one indexed lattice plane that is needed to match to the area of another indexed lattice plane of another material.
Selective laser melting
Additive manufacturing of the stock aluminium alloy and functionalized aluminium alloy powders were performed on a Concept Laser M2 selective laser melting machine (specifications are listed in Extended Data Table 1). Samples consisted of 60 mm × 20 mm × 40 mm tensile block specimens and 10 mm × 10 mm × 40 mm blocks for examining microstructure. Images of the as-printed samples on the build plates can be seen in Extended Data Fig. 2. Samples were processed with the Concept Laser ‘islanding’ scan strategy, which was specifically developed for the CL31 AlSi10Mg alloy material to minimize thermal and residual stress build-up in the part. Islands that compose the core of the build geometry were 2 mm × 2 mm in size. Standard machine parameters provided by the Concept Laser for conventional AlSi10Mg alloy were used for all builds. The parameter values are considered proprietary by Concept Laser and cannot be accessed by the user. The 70 mm × 70 mm build plates were machined out of aluminium alloy 6061 and sandblasted on the surface. Layers of the build were incremented by a range from 25 μm to 80 μm depending on part geometry and location in the build. Processing was done under a flowing, inert argon atmosphere with oxygen monitoring. All processing was completed at room temperature with no applied heat to the build plate. Samples were removed from the machine and cleaned of extra powder by sonicating in water. Parts were then dried with clean compressed dry air.
Samples were then heat treated to a T6 condition. Samples were solutionized at 480 °C in air with a ramp rate of 5 °C min−1 for 2 h and then quenched with water at 25 °C. Samples were subsequently aged at 120 °C with a ramp rate of 4 °C min−1 in air for 18 h and allowed to cool to room temperature.
Although additional insight is needed to determine the appropriate heat treatment methods for additively manufactured high-strength aluminium alloys, ageing time for the samples was determined by performing four ageing treatments with hold times of 6 h, 12 h, 18 h and 24 h at 120 °C. Vicker’s hardness measurements were used to determine a range of appropriate ageing time around 18 h (see Extended Data Fig. 3).
Sectioning and sample preparation
All samples were removed from the build plates via wire electro discharge machining (EDM). Tensile specimens were sectioned with wire EDM to a thickness of 2 mm. Tensile specimens were prepared for mechanical testing by polishing the surfaces of the gauge section with 240, 360, 400, 800 and 1,200 grit sand paper by hand. One side of the mechanical test samples was painted with white and spackled with black paint with an airbrush for digital image correlation using a GOM ARAMIS-3D Motion and Deformation Sensor.
Microstructure blocks were sectioned with a water-cooled saw and mounted in epoxy resin for polishing. Grinding was done with 240, 360, 400, 800 and 1,200 grit sand paper. Final polishing of the samples was accomplished with 1-μm diamond and 50-nm Al2O3 polishing compounds from PACE Technologies. Some polished samples were etched with Keller’s Etch for 10 s to reveal microstructure. Additional imaging was conducted using SEM and electron backscatter diffraction (EBSD).
To observe microstructural differences, mounted samples were observed with an optical microscope under polarized light and with SEM.
Vicker’s micro hardness was performed on mounted samples. The load applied was 200 g and indentation sizes were measured and compared (see Extended Data Fig. 3). These hardness measurements indicated that the appropriate ageing time is about 18 h for additively manufactured 7075.
Inductively coupled plasma optical emission spectroscopy (ICP-OES) was completed on raw Al7075 + Zr powder and on a Al7075 + Zr component after being additively manufactured. Analysis showed about 25% loss of zinc and about 32% loss of magnesium, both of which are strengthening elements for the 7075 alloy.
Tensile tests were performed on a servo-electric INSTRON 5960 frame equipped with a 50-kN load cell (INSTRON). Samples were clamped by the ends of the dog-bone-shaped samples. The extension rate was 0.2 mm min−1 and samples were loaded until fracture. Testing was conducted following ASTM E8. A U-joint was used to account for any misalignment in the sample. Extended Data Fig. 4 displays the stress–strain curves for the Al7075 material tested in this study and typical stress–strain curves for the CL31 stock powder produced.
Because cracking tended to orient parallel to the build direction, all tensile testing was conducted perpendicular to the expected crack orientation. This ensured that any residual cracks would have the maximum effect on the tensile properties. Observed ductility in the nanoparticle-functionalized material indicates a complete elimination of deleterious cracking.
Scheil simulations were conducted as described in the main text. Extended Data Fig. 5 shows the solidification curve for Al6061 in comparison to the solidification curves of Al7075 and AlSi10Mg (which are also depicted in Fig. 3a). The shape of the Al6061 Scheil curve indicates a high susceptibility to hot cracking.
Extended Data Fig. 6 shows micrographs of additively manufactured 6061 with and without Zr additions, indicating identical behaviour to 7075. Additional micrographs of 7075 with and without Zr can also be seen in Extended Data Fig. 7. Cracking is seen in the unmodified powder, whereas the addition of the Zr nucleant induces fine equiaxed grains and eliminates hot cracking. Further evidence of cracking between columnar grains in additive manufacturing of 7075 without Zr is shown in Extended Data Fig. 8. The EBSD map is used to clearly show that the cracking is restricted to intergranular regions.
The data that support the findings of this study are available from the corresponding author on reasonable request.
We acknowledge financial support by HRL Laboratories, LLC, and thank D. Martin for her artistic contribution to the figures, as well as B. Carter of HRL Laboratories, LLC, X. Li of the University of California, Los Angeles, and K. Hemker of John Hopkins University for discussions.