Investigation of an additively manufactured modified aluminum 7068 alloy: Processing, microstructure, and mechanical properties

Many additively manufactured alloys exhibit higher strengths than compositionally identical alloys processed via conventional processing routes. However, this enhancement is not consistently observed in 7xxx series aluminum alloys. These alloys present two complications when printed via Laser Powder Bed Fusion (LPBF): significant evaporation of strengthening elements from the melt pool and hot cracking during solidification. To address these issues, we introduce two modifications to the feedstock powder: (i) we increase the concentration of alloying constituents to counteract evaporation during printing, and (ii) we disperse TiC nanoparticles within the feedstock powder to promote heterogeneous nucleation and limit grain growth, thus avoiding hot cracking and improving strength. Relationships between the evaporation of alloying elements and laser energy density are quantified experimentally using inductively-coupled-plasma mass-spectrometry and are well captured by simple analytical models. The microstructures in as-printed and heat-treated conditions are characterized using X-ray diffraction, scanning electron microscopy, and transmission electron microscopy. Printing parameters have been optimized to attain minimum porosity, resulting in tensile strengths up to 650 MPa, which are in good agreement with predictions from classic models of strengthening mechanisms.


Introduction
Additive Manufacturing (AM) is a developing technology that gained commercial relevance to rapidly prototype new part designs.Today, after many engineering and materials breakthroughs, AM is being utilized in the manufacturing of final products [1], ranging from custom printed titanium implants to rocket-nozzle-tip fuel injectors [2].
Metal AM techniques include binder jetting (BJ) [3], electron beam melting (EBM) [4], and laser powder bed fusion (LPBF) for small to medium part production, and wire arc additive manufacturing (WAAM) [5], direct energy deposition (DED) [6], and cold spray [7] for larger components and structural repairs.Throughout the development of these metal AM techniques, LPBF has emerged as the current market leader.LPBF is a process in which a thin layer of powder (20-100 μm) is deposited onto a build substrate before a laser scans and melts/solidifies according to a 2D slice of a 3D part.Another layer of powder is deposited over the previous layer, and this process is repeated until the entire 3D part is printed.Thanks to the combination of fine part resolution, high print density, high surface quality, and operational simplicity (e.g., no need for vacuum environment, significant substrate heating or extensive post-processing steps), LPBF is poised to maintain its lead in market penetration.Further innovations such as multiple lasers, larger build platforms, and most substantially improved LPBF process-specific materials will continue to provide LPBF with many advantages over comparable printing processes [1].
The two fundamental challenges affecting printability of highstrength aluminum alloys are the significant evaporation of the major strengthening elements from the melt pool and the prevalence of hot cracking during solidification.Overcoming these challenges is essential for the development of LPBF-processable high-strength Al alloys.The ubiquitous reporting of Al-Si alloys in the literature as well as the vast range of commercial applications of these alloys stem from their resistance to hot cracking, attributed to the narrow solidification range of the eutectic alloy and the low viscosity of the melt, which allows for interdendritic/cellular liquid flow [39][40][41][42][43].By contrast, hot cracking occurs during rapid solidification of 6xxx and 7xxx series aluminum due to the large solidification range, the high viscosity of the inter-dendritic fluid, and the high thermal contraction during the latter stages of solidification [16,44,45].While hot cracking of high-strength Al alloys is also observed during welding, it is greatly exacerbated in LPBF due to the significantly higher cooling rates and the directionality of thermal gradients from the build substrate towards the laser source, leading to cracks between large columnar grains [13].
Martin et al. first demonstrated that incorporating Zr nanoparticles (NPs) as inoculants in Al-7075 alloy powder feedstock eliminated hot cracking in LPBF-processed parts.The Zr nanoparticles promoted heterogeneous nucleation from in-situ formed Al 3 Zr inoculants within the melt pool and resulted in refined equiaxed grains [16].Molecular dynamics simulations have verified the amount of low-mismatch lattice planes between the inoculant and the matrix phase to be a good predictor of how well heterogeneous nucleation prevails [46,47].Additionally, Sistiega et al. have shown that increasing the silicon content up to 4% in Al-7075 reduces the tendency to hot crack, due to an increase in latent heat and decreased solidification range, enabling to successfully print nearly fully dense parts.However, Si stabilizes the β-MgSi 2 phase, which has a decreased strengthening effect compared to the desired η-MgZn 2 phase in 7xxx alloys, resulting in lower mechanical properties [48].Spierings et al. printed a novel aluminum-scandium alloy (Scall-malloy™) via LPBF; formation of L1 2 Al 3 Sc precipitates led to heterogeneous nucleation, resulting in an equiaxed grain structure, and mechanical properties competitive with those of wrought Al-7075 [49].Scandium inoculants have been consistently shown to eliminate hot cracking and provide improved high temperature strength from the ordered FCC L1 2 Al 3 Sc precipitates; however, scandium is not economically viable compared to other traditional inoculants such as Zr and Ti [50][51][52][53][54][55][56][57][58].TiC and TiB 2 inoculants have been previously shown to improve the processability of high-strength aluminum castings, weld filler, and LPBF powder [59][60][61], and additionally provide grain refinement, in a cost-effective manner [62][63][64][65].Recent work has shown that TiC's low lattice mismatch with the α-Al matrix makes it effective in increasing nucleation rate and restricting grain growth during processing of 2xxx, 6xxx, and 7xxx series Al [66].However, varying results in terms of the TiC phase stability and strengthening effect have been reported [32,60,61,67,68], warranting further studies.
Evaporation of volatile elements, such as Mg in 6xxx and Mg and Zn in 7xxx Al alloys, has been observed during LPBF processing [16,40].R. Mertens et al. enriched an Al-7050 + 1% Zr alloy with an additional 7% Zn to counteract evaporation and observed over 50% loss in Zn [69].Additionally, A. Martin et al. varied the Zn composition in the feedstock powders, and observed less than 15% loss in Zn [70].The literature continues to show varying results in terms of elemental vaporization in multiple aluminum alloys [17,54,[71][72][73][74][75], warranting further investigation on the role of processing parameters on post-print composition, and specifically the influence of the energy input into the melt pool on compositional changes.These insights will enable alloy design and control of the constituent elements with different volatilities during LPBF processing.
Al-7068 possesses excellent strength but is prone to hot cracking as well as compositional changes during LPBF processing [76][77][78][79].This study presents a modified Al-7068 aluminum powder with two key features: (i) the addition of TiC nanoparticles to prevent hot cracking by promoting heterogeneous nucleation within the melt pool and restricting grain growth during solidification; and (ii) a significant enrichment of highly volatile elements (Zn and Mg) to counteract evaporation during LPBF processing, to achieve the desired target composition in the LPBF printed parts.The influence and effect of LPBF processing parameters (laser power, scan speed, and energy density) on the resultant composition, microstructure and mechanical properties of the printed alloy are carefully investigated and discussed.

Materials preparation
The powder feedstock used in this study was produced by MetaLi™ via Argon gas atomization of the alloy ingot with dispersed TiC nanoparticles [68].The metal charge of ~18.5 kg was held at 820 • C (180 • C superheat) in a vacuum furnace for 1 hr before atomization.The powder was sieved through 15-53 μm sieves prior to printing.The powder has an oxygen content of 349.2 ppm and a Carney Flow Rate of 13.4s/50g.The feedstock powder particle size distribution and chemical composition of the alloy are shown in Fig. 1 and Table 1, respectively.
An SLM 125HL printer with a Yb-fiber 400 W maximum output laser was used to print the specimens for this study.Small 5x5x10mm blocks were printed for density and compositional measurements, with laser power and laser scanning speed varying as described in Fig. 2 and Table 2. Samples are printed with laser powers of 200 W, 300 W, and 400 W. Laser scan speed was varied from 250 mm/s to 1250 mm/s.Hatch spacing and layer thickness are kept constant at 100 μm and 30 μm, respectively.A stripe scan pattern was used, with laser raster parallel to the recoating direction.All additional printing parameters and scan strategy were kept constant through all builds.
These parameters resulted in volumetric energy density (E D , defined as the laser power divided by the product of scan speed, layer thickness and hatching distance) varying by an order of magnitude, from 66 J/ mm 3 to 533 J/mm 3 .Recent studies on LPBF Al report that both Al-Si alloys and high-strength Al alloys (6xxx and 7xxx series) print with >99% relative density with E D in the 100-200 J/mm 3 range [21,40].The addition of ceramic particle inoculants at small volume fractions has shown to not substantially vary this optimal E D range [29,60,81,82].However, in this study we decided to investigate a much broader E D range, in order to more comprehensively capture the effect of processing parameters on material evaporation.The build chamber was purged with nitrogen, down to oxygen levels of 0.01%; the build plate was heated to 200 • C during the print.The powder was re-sieved before each print.All samples were printed directly on the build substrate, with electrical-discharge-machining (EDM) used to remove the parts from the substrate.
For tensile tests, larger rectangular blocks with cross-sectional dimensions of 100 mm × 12 mm were printed, as shown in Fig. 3. Two different processing settings were used: [200 W, 400 mm/s] and [400 W, 750 mm/s], henceforth referred to as 200W and 400W.These settings were determined to be optimal from a parametric sweep of the small builds, as discussed in Sec.3.2.The rectangular blocks were printed with heights ~30 mm.The printed tensile blocks were then machined into rectangular tensile coupons via EDM according to ASTM E8 [83].The tensile coupons were cut parallel to the build substrate, with their gauge length perpendicular to the build direction (Fig. 3a).Material was sectioned via EDM adjacent to the gauge section of the larger blocks and used for metallographic and X-ray diffraction (XRD) analyses.The middle section of the larger block was removed via EDM across the entire build direction and examined for structural integrity via CT imaging.
Post-processing of the samples consisted of homogenization and aging heat treatments.Additionally, select samples were treated with hot isostatic pressing (HIP) before further heat treatments.Samples were homogenized at 465 • C for 30 min in a MTI GSL-1100X tube furnace in ambient atmosphere with a heating rate of 10 • C/s, and then water quenched.Aging treatments were carried out under the same conditions at 130 • C, with peak aging being 24 hrs.HIP treatments were carried out in a Quintas MIH-9 Hot Isostatic Press at 400 • C for 4 hrs, with a 2 hr simultaneous pressure and temperature ramp-up from ambient conditions.

Chemical analysis
Inductively-coupled-plasma mass-spectrometry (ICP-MS) was conducted with a ThermoFisher Scientific iCAP RQ system using a single quadrupole detector.The feedstock powder and hand-ground sections from the small blocks were digested using 15 ml of trace metal grade 70% nitric acid (HNO 3 ), and further diluted 150x into 1% HNO 3 solution.Calibration standards for concentration analysis were carried out using the multi-element standard IV-ICPMS-71A from Inorganic Ventures (VA, USA), with a 10 mg/ml analyte concentration in a 3% volume HNO 3 solution.

Microstructural characterization
The Archimedes method was used for density measurements.Each surface of the small blocks was mechanically ground to 1200 grit, and then polished to 0.05 μm using colloidal silica.The weight of the small blocks (w) was measured in air, and the apparent weight submerged in pure ethanol (w a ) measured using a Mettler Toledo AG204 Archimedes scale.The temperature of the ethanol after submerging was allowed to equilibrate, allowing an accurate estimate of the density in ethanol (ρ f ).
The density of the alloy (ρ) was calculated as ρ = ρ f w /(w − w a ).The relative density of the samples was obtained as the measured density divided by theoretical density of the material; the theoretical density   was determined using CALculated PHAse Diagram (CALPHAD) equilibrium calculations with ThermoCalc Software's TCAL8 database from known post-print compositions measured by ICP-MS.We emphasize that theoretical densities depend on scan parameters, as rates of evaporation and hence part composition vary.Four samples were imaged using a VJ Technologies VedaPro X-ray Computed Tomography (CT) platform with a 200 μm pixel pitch digital detector array and a microfocus X-ray source with maximum voltage up to 225 kV: two small builds with optimal densities (200 W-400 mm/s, and 400 W-750 mm/s), and two middle sections EDM'd from the larger builds printed with the exact same parameters.20 μm and 45 μm voxel sizes were used for the small and large builds, respectively.Analysis of the pore distribution was conducted in the VGSTUDIO MAX software package.
For powder cross-sectional analysis, the feedstock powder was mounted in an epoxy mix with conductive graphite.The powder was sequentially polished down according to standard metallurgical polishing procedure, with a final polish using 0.05 μm colloidal silica.
Microstructural analysis was performed using a TESCAN GAIA-3 scanning electron microscope (SEM).Micrographs for particle size distribution analysis were captured using a secondary electron (SE) detector, whereas cross-section micrographs of phase distribution were captured using a back-scattered-electron (BSE) detector.For microstructural analysis of the as-printed builds, the small blocks were mounted in an epoxy mixture and polished down to a 0.05 μm colloidal silica finish.
Phase-distribution micrographs were obtained in BSE mode, and Energy-Dispersive-Spectroscopy (EDS) elemental maps were acquired with an Oxford Aztec Energy Advances EDS Microanalysis system with a X-MAX 150 mm 2 silicon drift detector.Phase fractions were calculated using ImageJ Software (NIH) [84].
Higher resolution microscopy and characterization of the lattice structures and precipitates were conducted with a JEOL JEM-2800 scanning/transmission electron microscope (S/TEM) equipped with a Gatan Oneview camera, operating at a beam voltage of 200 kV.The JEM-2800 is equipped with dual 100 mm 2 silicon drift detectors (SDD) for energy-dispersive X-ray spectrometry (EDS) for elemental characterization.The TEM lamellae samples were extracted using a focused ion beam (FIB) in a Quanta 3D FEG dual beam SEM/FIB (Thermo Fisher Scientific Inc.).
XRD patterns were collected using a Rigaku Ultima X-ray diffractometer equipped with a Cu K α (λ = 0.154 nm) radiation source configured in Bragg-Brentano geometry.40 kV and 30 mA emission was used, and scans were conducted varying 2θ in the range 15-90 • with 0.01 • step size and scan speed of 0.5 • /s.Each pattern was normalized for comparison.

Mechanical testing
Tensile tests were conducted on an Instron 5985 load frame equipped with a 250 kN load cell.An AVE2663-901 video extensometer with a Fujinon HF16HA-1S lens was used to track the strain of the gauge section.Tests were conducted according to ASTM E8 standards at quasistatic strain rate of 0.001s − 1 .The printed tensile blocks were machined via EDM into rectangular tensile coupons with gauge dimensions of 20 mm in length, 6 mm in width, and 3 mm in thickness.Vickers hardness measurements were taken with a 50 g load held for 10s averaged across 10 measurements, using a Buehler Wilson VH3300 instrument.Hardness was used as a metric to identify the peak aging condition.

Composition analysis and selective elemental evaporation
As the boiling points of Zn and Mg (907 • C and 1091 • C, respectively) are much lower than the boiling points of Al and Cu (2470 • C and 2562 • C, respectively) and not much higher than the solidus of Al-7068 (~465 • C), selective evaporation of Zn and Mg during LPBF processing is expected.As MgZn 2 is the strengthening precipitate in 7xxx alloys, it is critical to accurately determine the loss of Zn and Mg during processing.At low Zn:Mg ratios, other intermetallic phases, e.g.S-phase and Tphase, become stable and the strengthening contribution from precipitation hardening is reduced [85,86].Chemical analysis of the small builds was conducted across the parameter sweep to systematically evaluate the role of the volumetric energy density, E D .Fig. 4 shows the compositional changes in Zn and Mg as a function of E D and laser power.The composition for every sample is shown in Table 9 in the supplementary information.As energy density increases, the concentrations of both Zn and Mg decrease, as shown in Fig. 4a and b, respectively.The Zn:Mg ratio also decreases as much as 33% from the initial starting ratio in the feedstock powder (Fig. 4c), since Zn evaporates at a faster rate than Mg.Importantly, notice that these trends are not directly affected by the laser power alone, confirming that volumetric energy density is a viable metric for predicting evaporation of constituent elements.
It has been frequently observed that LPBF-printed 7xxx Al alloys experience significant compositional change relative to the powder feedstock chemistry, with the literature reporting wide ranges of selective evaporation, depending on the process parameters and measurement method used [69,70].The effect of printing parameters on the temperature of the melt pool, and ultimately on the selective evaporation of low-boiling point elements, can be estimated analytically through calculations of vapor pressures and vaporization fluxes via the Langmuir equation [72].The melt pool geometry can be estimated using the Rosenthal equation [87,88], under the simplifying assumptions that the melt pool shape is semi-elliptical, the heat flow is two-dimensional, the heat transfer through the melt pool is purely conductive (with vaporization from the top surface), the laser beam is a point source, and all properties of each phase are temperature independent.Details of the calculations are provided in Appendix 1.The experimental measurements of Zn and Mg evaporation are compared with the analytical predictions in Fig. 5.The predicted trend of a large initial increase in evaporation with a tailoring off towards higher energy densities matches the experimental measurements from this work.It should be noted that our experiments and analytical predictions best agree in the energy density range of 150-200 J/mm 3 , which is the region in which we observe maximum density and avoid both lack-of-fusion and keyholing porosity (see Sec. 3.2 and Fig. 6).In summary, significant compositional changes occur during LPBF processing of Al-7068, with 30-55% of volatile elements being removed, depending on the energy density (Fig. 4d).While Mg concentration and Zn:Mg ratios fall within the prescribed ranges for Al-7068 at nearly any energy density, energy densities higher than 300 J/mm 3 are needed to bring the Zn amount in the prescribed range.This suggests that the Zn concentration in the initial powder could be slightly decreased, as further discussed in the next sections.Even with this caveat, we demonstrated that enriching the initial powder of volatile elements is a viable strategy for achieving compositional accuracy in LPBF-processed 7xxx series aluminum alloys.

Porosity analysis
Metallic AM components are known to exhibit porosity, which results in limited ductility and reduced fatigue life [89].Hence the relative density (defined as 100% -porosity%) is a key metric for process optimization.The ICP-MS composition measurements shown in Fig. 4 indicate that samples processed at different energy densities have different compositions and therefore different theoretical densities, complicating determination of the relative density.To address this challenge, the compositional information determined by ICP-MS was used as the input for CALPHAD equilibrium calculations to determine the theoretical density of each sample.The experimentally measured density of each sample (Fig. 6a) was then divided by its calculated theoretical density to determine the true relative density of each sample (Fig. 6b).
For all sample compositions evaluated, CALPHAD predictions showed that the desired FCC-Al and MgZn 2 precipitate phases are stable across solid state temperatures, with no additional deleterious phases.Thus, process parameters for the larger prints were optimized simply to maximize relative density (i.e., minimize porosity), based on the data collected on the smaller prints (Fig. 6b).The optimal energy density along the 200W curve is obtained for E D = 166.6J/mm 3 , resulting from a scan speed of 400 mm/s; the optimal energy density along the 400W curve is obtained for E D = 177.7 J/mm 3 , resulting from a scan speed of 750 mm/s.These optimal parameters are henceforth adopted for large scale prints, to be used for microstructural analysis, CT analysis, and mechanical testing.Both optima are practically identical in chemistry (10.5% Zn and 2.7% Mg for the 400W condition, and 10.3% Zn and 2.7% Mg for the 200W condition).The full composition for every sample is shown in Table 9 in the supplementary information.
The amount of porosity observed in each sample (Fig. 7) confirms the trends in density discussed above and quantified in Fig. 6.Fig. 7a and d exemplify the lowest E D value, where lack of fusion porosity between consecutive build layers and between adjacent scan lines are present.Fig. 7b and e show prints produced with optimal E D , resulting in 99.6% relative density, verified by CT imaging.Finally, Fig. 7c and f show parts produced with the maximum E D , leading to uncontrolled turbulence in the melt pool causing keyholing and large circular pores.The increased evaporation rate present in the alloy also led to gas porosity across the entire processing range, but most notably at maximum E D .The trend of lack of fusion at low E D values and keyholing at high E D values is consistent with literature on LPBF processing [21,23].
The optimal process parameters for 200W and 400W conditions were used to print the larger sized tensile builds.The central portion of these larger builds were sectioned using EDM for CT imaging (shown in Fig. 3a).CT data was used to quantify both the size and distribution of the pores.CT imaging shows similar relative densities when scaling up from the small 5x5x10mm builds to the larger 12x100x30mm tensile builds.Fig. 8a and e are digitally reconstructed 3D volume renderings of the 400W parameter set prints of the small build and large build, respectively.Fig. 8b and f show the pore distribution along the build height, in the small and large builds, respectively.No trend in porosity distribution along the build height is observed.Fig. 8c and g show the distribution across the scanning and recoating direction (referred to as the X-Direction), in the small and large builds, respectively.Both builds have a bimodal pore distribution along the X-Direction.Fig. 8d and h show porosity radius histograms, in the small and large builds, respectively.The median pore radius of the large build is twice that of the smaller build, 101 μm vs 43 μm, respectively.One possible explanation for the increased porosity size in the large build is an increase in powder spatter, as the larger build layers have nearly a 50X increase in cross-sectional area in comparison to the small build layers.Powder spatter disrupts the recoating process and creates pores as large as several powder particles, spanning multiple build layers.The powder spatter on the surface of each layer interferes with the bi-directional recoating process and gets swept towards the middle of the build, from each side (see Fig. S2 and Video S1).This could additionally explain the bimodal pore distribution in the X-direction, observed in Fig. 8c and g.This bimodal porosity distribution across the recoating direction is not observed in CT imaging of previously investigated LPBF alloys [90,91], strongly suggesting the connection for the increase in spatter from increased evaporation.
The measured porosity from CT imaging is within 0.3% of Archimedes density measurements for both the large and small builds.Even at the optimal energy input yielding a 99.7% dense sample, porosity is still present due to spattered powder (Fig. S2 and Video S1).High speed imaging of the printing process has previously shown the vapor flux plume created during evaporation of constituent elements to be the main driver of ejected powder spatter [92].Indeed, a large increase in spatter for the presently printed alloy processed via LPBF can be expected, as the print evaporated 8% of its total mass during printing.The effect of spatter-induced defects on mechanical properties, and their evolution throughout processing, is discussed in section 3.4.

Microstructural evolution and phase analysis
The XRD patterns of the optimal 400W sample presented in Fig. 9 allow clear identification of the phases at each processing step.shows a slight diffraction peak shift towards higher angles of the MgZn 2 phase after printing compared to the powder feedstock, attributed to the loss of Mg and Zn during processing: as more Al substitutional atoms occupy lattice sites in the intermetallic MgZn 2 phase (predicted by CALPHAD calculations), the smaller Al atoms decrease the lattice plane spacing, thus increasing the diffraction angle.The TiC peak angles remain stable throughout processing, as shown in Fig. 9c.
Microstructural analysis was conducted on the initial feedstock  powder, and on samples printed with different processing conditions, to understand the effect of processing on microstructure.Fig. 10 shows representative microstructures for the powder feedstock (Fig. 10a), asprinted (Fig. 10b) and homogenized (Fig. 10c) samples, printed with 400W optimal parameters.All SEM-BSE micrographs clearly show three different phases, with the grey-scale intensity correlating with atomic mass.The phases from brightest to darkest are η-MgZn 2 , TiC, and α-Al, with CALPHAD-predicted densities of 5.11 g/cm 3 , 4.93 g/cm 3 , and 2.71 g/cm 3 , respectively.EBSD analysis (Fig. S5) confirmed that the equiaxed α-Al grains observed in Fig. 10 are individual grains separated by  distinct MgZn 2 intermetallic regions, without the presence of a larger columnar grain texture.The α-Al grain sizes are similar in both the feedstock powder and in the as-printed conditions, at 1.57 ± 0.55 μm and 1.61 ± 0.79 μm, respectively.This microstructure is consistent with the similar cooling rates achieved in gas atomization and LPBF [13].A full homogenization window was not observed, with 1.6% volume MgZn 2 intermetallic remaining after homogenization (Fig. 10c), which agrees with CALPHAD predictions.The comparison between the experimentally measured post-homogenization MgZn 2 volume fraction and the CALPHAD predicted remaining intermetallic fraction immediately below the solidus temperature is quantified and shown in Table 3.
In addition to appearing at α-Al grain boundaries, spherical MgZn 2 intermetallic particles are present within the α-Al grains in the as-printed condition.These additional particles are attributed to the build continuing to equilibrate on the 200 • C heated substrate for several hours while the entire print is completed.During the printing process, the solidified material is exposed to sufficiently high temperatures to enable diffusion of solute elements saturated in the α-Al grains due to the high solidification rates achieved.This diffusion leads to the uniform precipitation of spherical MgZn 2 intermetallic particles in the matrix.Fig. 10a-c confirm that the TiC particles remain stable throughout the LPBF process.TiC in the as-printed condition is mostly observed along the grain boundaries, bonded to the intermetallic phase due to low lattice-plane mismatch, with some TiC also present within the α-Al grains (Fig. 10b).TiC agglomerates are 530 ± 270 nm, ± 180 nm, 630 ± 400 nm in the powder, as-printed, and homogenized samples respectively; however, TEM observations (Fig. 11) show that the TiC nanoparticles form agglomerates made up of smaller particles, with an average diameter of 180 ± 83 nm.The as-printed microstructure shown in Fig. 10b clearly illustrates an equiaxed grain structure, confirming that columnar grain growth (a precursor to hot cracking) is avoided by the addition of TiC nanoparticles in the alloy.TiC nanoparticles act as heterogeneous nucleation sites within the melt pool, as the TiC particles are located within the α-Al grain.A high-resolution TEM (HRTEM) image of the TiC/Al interface is given in Appendix 2 (Fig. 15).Previous work has also verified coherent bonding between the planes of the intermetallic phase and TiC through HR-STEM in a nano-treated Al-7075 weld filler [66].Additionally, Li et al. have observed an increased nucleation rate of cast Al-7075 with TiC compared to Al-7075 without TiC [44].Restriction of the solidification front is also observed as TiC is located at the grain boundaries along with the MgZn 2 phase.This is due to TiC's high melting point and its stability in the melt pool throughout LPBF processing.
Selected samples for mechanical testing were additionally treated with HIP prior to homogenization and aging treatments.TEM micrographs of the 400W optimal parameter samples in different heat-treated conditions are shown in Fig. 11.Excessively large, elongated precipitates of η-MgZn 2 are clearly shown along the grain boundaries in the as-printed condition in Fig. 11a and d.The HIP treatment was subsequently performed at 400 • C, which is lower than the maximum solubility homogenization temperature (465 • C).As the alloy is supersaturated, much of the Zn and Mg coalesces into the larger MgZn 2 precipitates during HIP treatment, leading to coarse regions of MgZn 2 at the grain boundaries.The post-HIP homogenization solutionizes most of the Mg and Zn from the overaged intergranular spherical MgZn 2 precipitates, back into the matrix, as shown in Fig. 11b and e.The final strengthening heat treatment (24-hr ageing at 130 • C) reforms the nanoscale MgZn 2 precipitates, as shown in Fig. 11c and f.The mean radius and spacing of the MgZn 2 precipitates and TiC nanoparticles are reported in Table 4.The strengthening effects of both the MgZn 2 precipitates and TiC nanoparticles in the different heat-treated conditions are discussed in section 3.4.

Mechanical behavior
The mechanical properties of LPBF-processed samples were measured via tensile testing, to reveal the effects of processing parameters, heat treatments and HIP on strength and ductility.Fig. 12 shows representative stress-strain curves for all processing conditions tested.The results are provided in Tables 5 and 6, for the 200W and 400W samples, respectively.For Tables 5 and 6, for both print conditions, homogenized samples show yield strengths σ y ∼ 370 MPa and strains to failure ε f ∼ 4%.HIP has a relatively small effect on the yield strength but increases strain to failure; i.e., ε f ∼ 6% for the 400W print condition.
While this increase is consistent with porosity closure (Fig. S4), HIP is not sufficient to provide the 11% elongation typical of wrought Al-7068 [93,94].
Aluminum 7xxx series alloys are precipitation hardened, with optimally sized and distributed MgZn 2 precipitates increasing both yield strength (σ y ) and ultimate tensile strength (σ ult ), albeit at the cost of reduced ductility.In this work, all aging treatments were performed at 130 • C, with a peak aging time of 24 hrs determined by hardness microindentation (Fig. S1).The length of the peak aging treatment is in good agreement with reported data for wrought 7xxx alloys [93,95].The effects of peak 24 hr aging and HIP on strength and ductility are reported in Fig. 12.As expected, for both 200W and 400W printing conditions, peak aging increases the strength significantly (up to a maximum σ y ∼ 650 MPa when combined with HIP treatment), but substantially embrittles the alloy.
There are no clear differences in properties when comparing the optimal settings printed with the 200W and 400W conditions.As both conditions had nearly identical energy densities (E D = 166.6J/mm 3 for the 200W sample and E D = 177.7 J/mm 3 for the 400W sample), E D emerges as a solid metric to predict part density and mechanical properties achievable during LPBF processing of 7xxx Al alloys.
The limited effect of HIP treatment on mechanical properties was unexpected, as porosity is generally responsible for the low ductility of LPBF-processed materials.Fig. S4 confirmed that HIP treatment resulted in near-total elimination of internal porosity in all samples.The contributing strengthening mechanisms in different conditions are discussed below and in detail in Appendix 2. Fig. 13 is representative of the observed defect evolution throughout LPBF processing.EDS of oxide defects are shown in Fig. 13d-f.Fig. 13a  and d show that within the as-printed pores, there is an increase in Zn, Mg, and O content throughout the pore.Fig. 13b and e show that HIP successfully reduces the size of the pores as expected; however, oxide inclusions persist throughout the processing steps (Fig. S4 additionally shows HIP fully reduced porosity).Fig. 13c and f show that the areas enriched with oxide inclusions are the locations where fracture initiates and propagates.Fig. S3 additionally compares the brittle fracture surface shown in Fig. 13c with that of a more ductile sample printed and aged with the same conditions.It should be noted that the difference in scale bar size between Fig. 13a and b is due to HIP-induced pore closure.
There is a wide variance in ductility in each condition, quantified in Tables 5 and 6, attributed to the presence of defects introduced during spattering.Since this variance similarly occurs in the post-HIP condition, the defects causing ductility variance and premature failure are not the pores themselves, but rather oxides that may be deposited during spatter.Both our fracture surfaces and large pores are observed to have similar discolorations.Confocal micrographs of the fracture surface in Fig. 13c are shown in Fig. S3, highlighting the discoloration and defect accumulation in the most brittle specimens.While most of the spatter      Remarkably, we find that the calculated yield strengths are within 15% of the experimentally determined yield strength values for the HIP'ed and homogenized alloy, and within 2% for the 24-hr aged alloy.Dislocation strengthening may be anticipated to make a large contribution, as fast solidification rates in LPBF result in high dislocation density values.Additionally, both grain boundary and solid solution strengthening are expected, due to the significant grain refinement induced by TiC and the large additions of Zn and Mg, which are well above the ranges for conventional 7xxx Al alloys.The calculated effect of Orowan strengthening in the HIP + Homogenized condition is small, as the majority of Mg and Zn is solutionized and has not been aged into nanoscale η-MgZn 2 precipitates.Accurate precipitate radii and spacing for the larger elongated particles along the grain boundaries are difficult to quantify, thus leading to variances between the theoretical calculations and experimental measurements.Precipitation strengthening due to TiC particles is only responsible for Δσ orowan ~4 MPa, which can be linked to the large spacing of 6 μm between TiC agglomerates, with the remaining Orowan strengthening due to the larger MgZn 2 phases at grain boundaries and smaller MgZn 2 precipitates that did not fully solutionize during the homogenization treatment.However, the TiC nanoparticles do provide a significant grain boundary strengthening effect overall, by restricting grain growth during homogenization at 465 • C, leading to relatively high strengths even without aging.When aging at 130 • C for 24 hrs after HIP + Homogenization, much smaller η-MgZn 2 phase precipitates form, with a much closer 79 nm interprecipitate spacing, leading to substantial Orowan strengthening in comparison to the homogenized sample.Overall, the ability to maintain high precipitation strengthening and solid solution strengthening associated with 7xxx series Al alloys by compensating for evaporation of Mg and Zn, along with TiC nanoparticles refining grain size and thus increasing grain boundary strengthening, result in a printable alloy with remarkably high strength.

Conclusions
A modified Al-7068 alloy enriched with higher Zn and Mg content and TiC nanoparticles was designed and printed to eliminate hot cracking, improve mechanical behavior, and counter-act evaporation of volatile elements.The key conclusions can be summarized as follows: (1) The TiC particles in the feedstock powder promoted equiaxed grains in the as-printed condition, avoiding hot cracking.TiC particles are present both within the grain (indicating heterogeneous nucleation sites) and at grain boundaries (indicating the grain restriction during solidification).This results in small grains and high grain boundary strengthening effect after conventional homogenization and aging treatments.(2) A post-print target composition was achieved by counteracting the evaporation of constituent elements.This provided a nearoptimal distribution of MgZn 2 precipitates upon homogenization and aging, resulting in a very significant Orowan strengthening effect.(3) Simple analytical models of selective evaporation agree well with experimental measurements of compositions in samples printed at intermediate energy densities, when porosity is minimal and   value, routinely reached under LPBF conditions [96]), R is the universal gas constant, T is the temperature of the melt pool, and M i and a i are the molar mass and the activity of element i, respectively.a i has been shown to be modeled accurately by the mole fraction (ideal solution model) [97], taken as 0.048 and 0.079 for Mg and Zn, respectively.The saturated vapor pressure can be calculated from the Clausius-Clapeyron equation [71]: where p 0 is the ambient or chamber pressure, ΔH v,i is the latent heat of vaporization of element i, and T b,i is the boiling point of element i.The thermophysical properties of our bulk material and constituent elements are reported in Table 7.A chamber pressure p 0 = 4 mbar is measured during printing.The melt pool temperature can be estimated from the Rosenthal equation as [88]: with α the absorptivity of the powder bed, P the laser power, ρ, C p and D t the density, specific heat and thermal diffusivity of the alloy, respectively, v the scan speed, and r the laser spot radius.Temperature calculations for our process parameters result in melt pool temperatures much higher than the boiling point of the highest boiling point element, in our case aluminum.As this model does not account for the latent heat of vaporization, these predictions merely indicate that the temperature reaches the boiling point of aluminum.As the melt pool temperature cannot exceed this value [13,72], we take T = T b,Al = 2470 • C. With the known vaporization flux, the mass of evaporated element i, Δm i , and the relative evaporation from its initial mass m 0,i can be calculated as [98]: where L mp , A s and V mp are the length, surface area and volume of the melt pool, respectively, J i is the vaporization flux (Eq.( 1)), v is the scan speed, ρ is the density, and f i is the initial mass fraction of element i.The melt pool cross section was estimated using the Rosenthal model, which provides analytical expressions for the shape and size of a semielliptical melt pool.These have been shown to agree with LPBF melt pool width and depth measurements [87,88].While the Rosenthal solutions only provide two-dimensional section predictions (width, W mp and depth, D mp ), the pool length, L mp can be estimated by empirical proportional scaling laws.Hence, we have [87,88]:  with G, b, M defined as above, and ν as the Poisson's ratio of the alloy (Table 8).Using the measured mean radius (r) and mean inter-precipitate distance (λ p ) from TEM image analysis (Table 4), the total precipitate/dispersoid strengthening for the homogenized and aged condition are Δσ Orowan = 75 MPa and 232 MPa, respectively.It is noted that HRTEM imaging (Fig. 15) confirms the MgZn 2 precipitates to be incoherent, implying that the operative mechanism is indeed Orowan dislocation bypassing as opposed to dislocation shearing [104], the former being described by Eq. ( 16).The quantitative contributions of these four strengthening mechanisms are reported in Fig. 14, in good agreement with experimental measurements.

B
. Fields et al.

Fig. 2 .
Fig. 2. Parametric sweep of laser power and laser scan speed, covering a wide range of energy density (E D ), 66 J/mm 3 to 533 J/mm 3 .Each set of four 5x5x10mm blocks has nominally identical processing conditions.(a) CAD design file.(b) Printed samples on the build plate, with top surfaces ground.

Fig. 3 .
Fig. 3. Large build geometry used to assess mechanical behavior and porosity distribution: (a) Schematic of the build, highlighting the orientation of tensile coupons and the portion used for CT porosity analysis, and (b) image of the as-printed block to demonstrate build surface finish.

Fig. 4 .Fig. 5 .
Fig. 4. Compositional analysis of the resulting builds as a function of energy density: (a) Zn loss, (b) Mg loss, (c) Zn:Mg ratio, and (d) total weighted loss of Zn and Mg. Fig.9b

Fig. 6 .
Fig. 6.(a) Archimedes density and (b) relative density as a function of energy density.

Fig. 8 .
Fig. 8. CT porosity measurements for 400W specimens: 3D reconstructions of the (a) small and (e) large builds.Porosity distributions across the build height of the (b) small and (f) large builds.Porosity distributions along the recoat/scan direction of the (c) small and (g) large builds.Pore radius histogram across the (d) small and (h) large builds.

Fig. 9 .
Fig. 9. (a) Full range of XRD peaks observed in the feedstock powder, as-printed and homogenized 400W sample.(b) Zoomed view of the 19-23 • range, showing nearly complete homogenization.(c) Zoomed view of the 36-42 • range, showing near full homogenization and the TiC stability.

Fig. 10 .
Fig. 10.Representative SEM-BSE micrographs of the (a) feedstock powder, (b) as-printed and (c) homogenized 400W specimen.Representative EDS showing phase constituents in the (d) powder, (e) as-printed and (f) homogenized samples.The build direction is upwards.
and vapor are flown into an outlet filter via a laminar gas flow inlet pressure, much of the spatter inevitably lands on the print, enriching the pores with Zn and Mg oxides.Images of these spattered particles after the recoating process are shown in Fig.S2.The exceptionally high strengths achieved in this modified alloy warrant discussion into the specific roles of multiple strengthening mechanisms, including grain boundary or Hall-Petch strengthening (Δσ gb ), solid-solution strengthening (Δσ ss ), dislocation strengthening (Δσ dis ), and/or precipitate/dispersoid strengthening (Δσ Orowan ).The contributions of each strengthening mechanism to the yield strength of the alloy, before and after the aging process, can be analytically estimated (see Appendix 2 for details) and are illustrated in Fig.14, in

Fig. 12 .
Fig. 12. Representative tensile curves for optimal (a) 200W (b) 400W samples in the homogenized and 24hr aged conditions, with and without HIP.

B
. Fields et al. comparison with experimentally measured values.

Fig. 13 .
Fig. 13.Representative SEM and EDS of defects in the 400W sample: (a,d) a pore in the as-printed condition, with oxide and Zn, Mg accumulation; (b,e) an oxide inclusion in the post-HIP condition; (c,f) the fracture surface of a post-HIP sample showing beach marks propagating from a region with overall increased amount of oxygen, along with oxide inclusions.

Fig. 15 .
Fig. 15.HRTEM of interfaces between (a) MgZn 2 and FCC-Al, and (b) TiC and FCC-Al, in the sample printed with 400W conditions.

Table 2
Processing parameters for all 5x5x10mm blocks.

Table 3
Volume percent of the intermetallic phase: experimental measurements versus CALPHAD predictions.

Table 4
TEM measurements of average precipitate diameter and spacing.

Table 5
Tensile properties of samples printed with the 200W conditions (* only one test exceeded yielding).

Table 6
Tensile properties of samples printed with the 400W conditions (* only one test exceeded yielding).

Table 7
Thermophysical properties used in vaporization calculations.