Microstructural Control of a Multi‐Phase PH Steel Printed with Laser Powder Bed Fusion

The established approach to materials design for additive manufacturing (AM) consists of attempting to reproduce the uniform structures and properties of conventionally processed materials. While this certainly helped facilitate material certification and the rapid introduction of AM technologies in several industries, the opportunity to exploit unique features of specific AM processes to generate spatially varying microstructures–and hence novel materials, remains largely untapped. This work presents a method for manufacturing materials through laser powder bed fusion (LPBF), in which control over the spatial variation in phase composition and mechanical properties is achieved. This technique is demonstrated using 17‐4 precipitation‐hardened stainless steel (17‐4PH), by controlling spatial modulation of energy densities during printing. This results in local control of ferrite/martensite volume fractions, allowing the fabrication of metal/metal architected composites with hard/brittle regions interspersed with soft/tough regions. Local variations of ~20% in tensile strength and ~150% in elongation are achieved, with a spatial resolution of ~100 microns. The approach is general and robust, fully compatible with commercially available LPBF equipment, and applicable to virtually any multi‐phase alloy system. This work shifts the paradigm from attempting to print components with uniform properties to manufacturing alloys with controlled spatial property gradients.


Introduction
Additive Manufacturing (AM), also known as 3D printing, is a disruptive technology that allows for the rapid fabrication of parts of nearly arbitrary geometrical complexity without the need for expensive tooling, vastly simplifying the production of elaborate DOI: 10.1002/admt.202301037assemblies for a wide range of industries.The ability to print metallic components is particularly critical for structural applications.Existing AM technologies for metals include Powder Bed Fusion (either via laser (LPBF) [1] or electron beam (EBM) [2] ), Direct Energy Deposition (DED), [3] Metal Binder Jetting (MBJ), [4] Fused Filament Fabrication (FFF), [5] and more recently, Wire Arc Additive Manufacturing (WAAM), [6] Friction Stir Additive Manufacturing (FSAM) [7] and Cold Spray Additive Manufacturing (CSAM). [8]In spite of limitations related to residual stress evolution and difficulties in achieving uniform microstructures in a controllable way, at present LPBF is the most technologically mature process for the production of small and medium-sized parts, enabling high resolution at acceptable throughput. [9]he LPBF process consists of depositing a fine layer of powder (20-100 μm in thickness) over a build platform, where a laser scans and locally melts the powder according to a 2D slice of a 3D geometry file, thus solidifying a pattern on that layer.A subsequent layer of powder is coated on top of the previous layer, and this process is repeated until the entire 3D part is printed from the powder bed.
The intrinsic point-by-point nature of AM, combined with the enormous and location-dependent temperature gradients that affect laser and electron beam melting processes, present significant challenges in controlling residual stresses and achieving uniform microstructures across geometrically complex parts.Historically, much of the research in LPBF, EBM, and DED has focused on tackling these challenges, while depositing materials with properties approaching (and occasionally exceeding) those of commercially available wrought or cast alloys. [10,11]While this approach has certainly facilitated the adoption and certification of AM parts for several industries, a huge opportunity remains largely unexplored: the development of strategies to print heterogeneous "designer microstructures", where composition and/or microstructure are locally controlled at the microscopic scale and tailored to optimize component performance.The ability to fabricate large-scale components with local microstructural control has the potential to revolutionize manufacturing.
[14][15] Beese et al. reviewed recent advances in additive manufacturing of functionally graded metallic materials, highlighting the potential for multiple AM techniques to produce microstructures impossible to manufacture just a few years before. [16][24] The formation of undesirable brittle phases is a challenge that is exacerbated when multiple principal element alloys are used, for which the complex multidimensional thermodynamic space is not fully explored.
With the exception of some recently introduced multi-material powder bed systems, [25][26][27] most powder bed processes have uniform feedstock and do not allow compositional grading.Thus, the only possible route in LPBF to control microstructural evolution is by local tailoring of the processing parameters.[30][31][32][33] Mukherjee et al. have thoroughly reviewed and outlined the different approaches enabling the control of grain structures, phases, and defects in metal AM parts. [34]Most notably, Sofinowski et al. have introduced highly controllable layer-wise engineering of grain orientation (LEGO) microstructures, where the crystallographic grain orientation in LPBF 316L steel [35] and Ti-Nb [36] can be locally manipulated with great accuracy, by using the laser scan speed and printing strategy to tune thermal gradients.This capability has been recently further refined by Gao et al., who employed careful choices of the hatch spacing to control the grain orientation texture of LPBF 316L at a resolution of 125 μm. [37]Similarly, Plotkowski et al. have shown the ability to control grain texture and morphology in EBM Inconel 718 by locally tuning the amount of time the material is above the melting point, generating equiaxed and columnar microstructures on a point-by-point basis. [38]These approaches indicate the feasibility of manipulating the processing parameters to affect the resultant microstructures.However, the work to date does not result in significant spatial gradation in mechanical properties nor the ability to generate local microstructures with competing properties (e.g., strong/brittle VS soft/ductile). [28,36,39]n this manuscript, we propose an alternative approach for local microstructural control in LBPF, based on tailoring of volume fractions in multi-phase alloys.We demonstrate the potential of this approach on 17-4 precipitation hardening (17-4PH) stainless steel (Type 630, or UNS S17400).[46][47][48][49][50][51][52][53][54][55][56][57][58] Most recently, Haines et al. observed austenite, ferrite, and martensite all present in 17-4PH, [59] whereas An et al. found mostly ferrite with little martensite. [60] "concentric borders" scan strategy and a "rotating stripe" scan strategy were used, respectively.Both studies employed the same printer (3D systems ProX 300 LPBF) and powder, with similar energy densities (53 and 56 J mm −3 ), further highlighting the variability present in the as-printed microstructures between similar feedstocks printed under similar conditions.While the as-printed steel can always be solutionized and quenched back into a fully martensitic homogeneous microstructure, the highproperty contrast between these phases in the as-printed condition may be exploited for local microstructural control.Here we show that we can locally control the volume fraction of martensite and ferrite by spatially varying the energy density during LPBF printing.We characterize the limit of the microstructures we can achieve while preserving nearly 100% density and demonstrate the ability to locally and systematically vary strength by ˜20% and ductility by ˜150% at a spatial resolution ˜150 μm.This enables the fabrication of metal-matrix-composites (MMC) with a hard/brittle BCT martensite phase and a soft/ductile BCC ferrite phase.[63][64][65][66][67][68][69][70][71][72] This approach opens a vista of opportunities to manufacture metallic systems with locally tailored "designer microstructures".

Phase Stability and Microstructural Control
It is well known that the quality of metal parts printed by LPBF is dramatically influenced by multiple processing parameters, most notably the laser power (P), the laser scan speed (v), the hatch spacing (h), the thickness of each powder layer (t) and the scan strategy (laser path).An intuitive combination of these parameters is the volumetric energy density (E), calculated as E = P v⋅h⋅t , expressing the amount of energy deposited in a given volume of powder.For a given material, the ratio of E to the minimum amount of energy necessary to melt the powder strongly correlates with sample density: [73,74] low values of E cause (e-h) Optical micrographs of wedge builds printed with an upwards build direction, increasing the energy density from 50 to 400 J mm −3 .Higher magnification of wedge builds printed with an out-of-page build direction, showing dual phase versus martensitic microstructures corresponding to (i) E = 100 J mm −3 and (j) E = 400 J mm −3 respectively.(See Figure S1, Supporting Information for geometric details).
excessive lack-of-fusion porosity, whereas excessively high E induces significant turbulence in the melt pool, resulting in keyhole porosity.Both scenarios negatively affect mechanical properties, in particular ductility.Intermediate values of E generally yield maximum density and hence optimal mechanical behavior.To determine the processing window for 17-4PH steel, we print a number of wedge-shaped specimens along two different printing directions, over a range of energy densities between 50 and 400 J mm −3 (see Figure S1, Supporting Information for geometric details).The "borders inside-out" scan strategy is used in all our prints, whereby an outline of the shape is printed from the inside of the sample, followed by progressively larger adjacent outline scans until the final border is scanned.The results are shown in Figure 1, with porosity data reported in Table 1. Figure 1a-d,e-h shows optical images of the samples printed horizontally and vertically, respectively.A significant lack of fusion porosity is clearly visible in the samples processed at the lowest energy density (E = 50 J mm −3 ), regardless of the printing orientation.The vertically printed sample (Figure 1e) shows more porosity near the base, consistent with the higher heat flow induced by the vicinity of a large heat sink.All other samples are almost fully dense (porosity ≲ 0.1%), except for the vertically printed specimen at the highest energy density (E = 400 J mm −3 ), in which pores are visible near the tip.Indeed, this region is the furthest away from the heat sink, consistent with keyhole porosity.The conclusion is that 17-4PH steel presents a wide processing window, where near fully dense specimens can consistently be printed with energy densities in the range of 100-400 J mm −3 .This is in striking contrast with many conventional LPBF alloys, which enter the keyholing porosity regime at much lower energy densities. [43,75]he optical micrographs in Figure 1 have been etched to reveal microstructural details.A clear difference emerges between the samples produced at low (50-100 J mm −3 ) and high (250-400 J mm −3 ) energy densities, with the former showing a two-phase microstructure with finely spaced martensitic laths interspersed with larger grains between scan lines (Figure 1i), and the latter exhibiting a very uniform martensitic microstructure throughout the specimen (Figure 1j).All samples clearly show a uniform martensitic outer border, regardless of the printing conditions.As the outer border is printed last, this suggests that the initial solidification of any single laser scan has a sufficiently high cooling rate to induce a martensitic transformation.This is consistent with well-established estimates of cooling rates in Table 1.Porosity of wedge samples as a function of energy density.][78] The larger grains present can be linked to retained ferrite [40] ; in the present work, this is proposed to have formed through re-transformation upon reheating from an adjacent scan line.Since dual-phase microstructures are only observed in our lower energy density prints, and fully martensitic microstructures are observed in our higher energy density prints, the reheating effect necessary to retransform martensite during adjacent scans is dependent on the processing parameters.The important conclusion is that two processing conditions exist (corresponding to E = 100 J mm −3 and E = 400 J mm −3 ) that result in fully dense specimens with remarkably different microstructures, clearly indicating the possibility of microstructural control.These two processing conditions were selected for further analysis.
To further investigate the difference in microstructure between samples printed at different energy densities, 1 mm-thick dog bone specimens are printed with E = 100 J mm −3 and E = 400 J mm −3 , respectively.A larger specimen is fabricated with alternating 1mm-thick layers printed with E = 100 and 400 J mm −3 conditions, allowing visualization of the interface between different processing conditions.(See Figure S1, Supporting Information for geometric details).Scanning Electron Microscopy (SEM) images of the dog bone gauge sections are depicted in Figure 2a,d.While the sample printed at lower energy density displays a martensitic border (Figure 2c) encom-passing a two-phase core consisting of fine martensite intermixed with larger grains (Figure 2b), the sample printed at higher energy density shows a uniform martensitic microstructure (Figure 2e,f).This finding is consistent with the "borderinside-out" printing strategy, as discussed above.The interfacial region between volumes printed with different energy densities is analyzed by Electron BackScatter Diffraction (EBSD), with Figure 2h-j showing Inverse Pole Figure (IPF-X), phase map, and EBSD band contrast, respectively.The phase map (Figure 2i) confirms that the region processed with E = 100 J mm −3 (top half) has a dual phase microstructure, composed mainly of columnar ferrite grains (shown in red) and regions of martensite laths (shown in blue); by contrast, the region processed with E = 400 J mm −3 (bottom half) shows a fully martensitic microstructure.We note that the several red pixels indexed as ferrite in this region are due to the challenge in distinguishing between body-centered cubic (BCC) ferrite and the low aspect ratio centered tetragonal (BCT) martensite, resulting in an EBSD detection limit of ˜1 μm, which is comparable with the laths dimension. [43,44,79]From the large area EBSD phase map (Figure 2i), ferrite fractions are measured as 79% and 28% in the regions processed with E = 100 J mm −3 and E = 400 J mm −3 respectively.While the phase fractions from the phase map in Figure 2i are not precise values, they nonetheless offer good estimates and exclude the presence of austenite.Most importantly, Figure 2h-j clearly shows that regions printed with different energy densities consistently exhibit different (and repeatable) microstructures.To fully validate the hypothesis that powder always initially solidifies in a fully martensitic microstructure (regardless of the energy density), with dual-phase ferritic/martensitic regions developing into inter-hatch regions upon subsequent heating/cooling cycles in samples printed at lower energy density, we print simple vertical walls, either one or two laser tracks wide, with the two different energy densities, 100 and 400.The programmed hatching distance for the two-track walls is 100 μm.Optical images of the etched specimens are shown in Figure 3. Notice that the width of single-track walls increases significantly with energy density, almost doubling from 158.8 ± 20.0 μm at 100 J mm −3 to 327.1 ± 22.4 μm at 400 J mm −3 , implying that the remelted region between adjacent melt pools is significantly larger in the 400 J mm −3 two-track sample than in the 100 J mm −3 two-track sample.It should be noted that both single and dual-track walls printed at high energy density display a uniform martensitic microstructure, consistent with the results on all parts displayed in Figures 1 and 2. Single-track walls printed at low energy density also display a uniform martensitic microstructure, indistinguishable from that of the walls printed at higher energy density.Conversely, dual-track walls printed at low energy density show a clear martensitic border encapsulating a dual-phase microstructure in the core, with large ferritic grains extending through multiple print layers.Both the ferritic and martensitic phases are identical to those seen in the large-scale specimens (Figure S3, Supporting Information).This clearly confirms that powder always initially solidifies in martensitic microstructure at all energy densities (consistent with the very fast cooling rates experienced in LPBF), and that the ferritic microstructure is formed in the hatch between adjacent scan tracks upon reheating-and only at low energy densities.
It is well known that the vaporization of lower boiling point elements in LPBF melt pools is a function of energy density, potentially resulting in variations in alloy composition at different E values.We measure these differences via OES and perform Scheil solidification calculations with ThermoCalc (Figure S5, Supporting Information).With the caveat that martensitic transformations are not accounted for in the model, we demonstrate that these compositional differences are not responsible for the observed differences in microstructure between samples printed with different energy densities reported herein.Fully elucidating the precise pathways that induce the martensite to ferrite solid phase transformation in the hatch upon the heating/cooling cycle requires a more complex combination of numerical modeling .STEM and STEM-EDS images of a sample processed with E = 100 J mm −3 in the as-printed (a-e) and heat treated conditions (f-j), and a sample processed with E = 400 J mm −3 in the as-printed (k-o) and heat treated conditions (q-t).Nb-rich precipitates are visible in the heat treated samples.and experimental work, which is beyond the scope of this article.Our working hypothesis is that the temperature distribution and history are such that some volumes of material near the hatch get reheated to a temperature where the -ferrite phase is stable, inducing fast athermal transformation to the  phase.Once nucleated, these grains no longer transform to martensite upon fast cooling, "locking in" the ferritic microstructure.Extensive simulations of thermal history coupled with thermodynamic modeling and selected experiments are currently underway to verify this hypothesis.
In conventionally processed 17-4PH steel, the martensitic microstructure is introduced via solutionizing and quenching, and further enriched with Nb-and Cu-rich precipitates upon 1hr aging at 482 °C (900 °F), a thermal process denoted as H900. [40,41,43]n LPBF-printed and H900-treated 17-4PH, the nature and distribution of precipitates appear to depend on the printing conditions, in ways that have not been fully elucidated. [48,40,80]As solutionizing would effectively erase any gradient in microstructure intentionally introduced in our samples, in this study we eliminate this process and directly age the as-printed specimens.To verify the efficacy of this approach in generating precipitationstrengthened microstructures and identify potential differences as a function of the applied energy density, Scanning Transmission Electron Microscopy (STEM) and Energy Dispersive Spectroscopy (STEM-EDS) are performed on samples printed at E = 100 and E = 400 J mm −3 , both in as-printed and heat treated conditions (Figure 4).The TEM lamellae samples processed at E = 100 J mm −3 are taken from a large-grained ferritic region, and those processed at E = 400 J mm −3 are taken from the martensitic region.Several important results emerge: i.Samples printed at lower energy density (Figure 4a,f) exhibit a much lower dislocation density than samples printed at higher energy density (Figure 4k,p), both before and after the aging process.This observation, combined with visual observation of the lath structure in the sample printed at a high energy density, confirms that samples printed at high energy density (400) have a martensitic microstructure, whereas the large grains observed in samples printed at lower energy density (100) are non-martensitic.TEM diffraction patterns from all samples (Figure S4, Supporting Information) confirm the large grains to have a BCC ferrite structure (as opposed to FCC austenite).ii.STEM micrographs of both samples in the as-printed condition show large spherical particles (˜50-150 nm in diameter) dispersed throughout the samples (Figures 4a,k).STEM-EDS identified these particles as alumino-silicates in the low-energy density specimen (Figure 4b-e) and silicates in the high energy density specimen (Figure 4l-o).Similar silicate particles have been observed in LPBF-processed 17-4PH samples before, and are reported as inclusions, as opposed to strengthening precipitates. [50,55]Upon aging, these particles appear to nucleate Nb-rich precipitates in samples printed at both energy densities (Figure 4j,t).Interestingly, no Cu-rich precipitates are visible in either sample.iii.Both the ferritic and martensitic microstructures remain unaltered after the aging process (except for the formation of precipitates), confirming that this treatment maintains microstructural gradients intentionally introduced during printing.
In conclusion, the STEM investigation reveals that the aging process is successful in precipitating Nb-rich particles in microstructures printed with any energy density, without altering the martensite and ferrite distribution.

Microstructure-Mechanical Properties Relations
The ability to consistently program heterogeneous microstructures in samples of different shapes, dimensions, and printing orientations is demonstrated in Figure 5.(See Figure S1, Supporting Information for geometric details).The low-energy density ferritic/martensitic microstructure and the high-energy density fully martensitic microstructures are clearly visible in the etched specimens via optical microscopy (Figure 5a-c).Note that the sample displayed in Figure 5a exhibits a microstructural gradient parallel to the printing direction, while the two are perpendicular in the samples shown in Figure 5b,c.Figure 5d-f displays hardness maps obtained by Vickers micro-indentation on the same samples.Moreover, the fully martensitic microstructure is consistently harder than the two-phase microstructure, with a difference that is more pronounced along the printing direction: while the fully martensitic region consistently displays a hardness of ˜500 HV, the dual-phase microstructure ranges in hardness from ˜400 HV along the print path (Figure 5d) to ˜450 HV in the perpendicular direction (Figure 5e,f).The dependence of the hardness on the printing orientation for the dual-phase microstructure is consistent with the highly  (a) Engineering stress-strain curves from tension tests on dog bone specimens printed with differing processing strategies: (i) uniformly low energy density of 100 J mm −3 (black); (ii) uniformly high energy density of 400 J mm −3 (blue), and (iii) a "brick-and-mortar"-inspired architecture, embedding prismatic domains (bricks) printed at 400 J mm −3 in a matrix (mortar) printed at 100 J mm −3 (red).(b) Stress-strain curve with DIC strain maps for a "brick-and-mortar"-inspired sample, clearly showing localized plastic deformation in the softer regions printed at E = 100 J mm −3 .(See Figure S1, Supporting Information for geometric details).elongated shape of the ferritic grains along the printing direction (Figure 2h).
To investigate the scale of microstructural control, a gradient block sample is printed by alternating high energy density (E = 400 J mm −3 ) and low energy density (E = 100 J mm −3 ) regions of decreasing thickness, from 450 μm down to 30 μm (the thickness of a single print layer), along the print direction.An optical image of the etched sample is shown in Figure 6a,b.As detecting the boundaries between different regions becomes challenging as the regions get progressively thinner, a hardness map is produced in Figure 6c, with a scan across the gradient shown in Figure 6d.Note that differences in hardness are appreciable down to a thickness region of ˜150 μm (equivalent to ˜5 print layers).As this value is partly determined by the resolution of our micro-indentation maps, it can be taken as a very conservative estimate of the scale of our microstructural control along the print direction.As the hatch spacing perpendicular to the print direction is 100 μm, and microstructural differences are evident at this scale (Figure 1i), we expect the resolution perpendicular to the build direction to be of the order of 50-100 μm.
While non-instrumented indentation provides a highthroughput mechanism to ascertain the impact of microstructural features on mechanical properties as well as allowing rapid 2D mapping of large heterogeneous regions (Figures 5  and 6), it cannot capture all the information embedded in a full stress-strain curve extracted from a tensile test.To fully appreciate the impact of our microstructural control on the mechanical response of LPBF-printed 17-4PH components (and in particular on yield strength, ultimate strength, hardening behavior, and strain to failure), we perform tensile tests on dog bone-shaped specimens printed with three different conditions: (i) samples printed with uniformly low energy density of 100 J mm −3 ; (ii) samples printed with uniformly high energy density of 400 J mm −3 , and (iii) hybrid samples with a "brick-and-mortar"-inspired architecture, embedding prismatic domains (bricks) printed with high energy density in a matrix (mortar) printed with low energy density (Figures S1 and S2, Supporting Information).Ten nominally identical samples were tested for each processing condition.Stress-strain curves are displayed in Figure 7a, with average mechanical properties Table 2. Mechanical properties of dog bone specimens printed at E = 100 J mm −3 , hybrid "brick-and-mortar" strategy, and E = 400 J mm −3 .(See Figure S1, Supporting Information for geometric details).E = 100 J mm −3 Hybrid Sample E = 400 J mm reported in Table 2. Several important results clearly emerge.The samples printed with low energy density exhibit substantial ductility (strain to failure ˜18%), and yield and tensile strength of 1,151 and 1,378 MPa, respectively.By contrast, the samples printed with high energy density are remarkably more brittle (strain to failure ˜7%) but significantly stronger, with yield and tensile strength of 1,394 and 1,561 MPa, respectively.Remarkably, these results are consistent across all samples tested, with most properties exhibiting variations of ˂2% (Table 2).The only exception is the strain to failure in the sample printed at higher energy density, which shows variations in excess of 30%; this is consistent with a brittle mechanical response and is likely exacerbated by the presence of keyhole porosity (Figure 1h).The hybrid sample printed with a "brick-and-mortar" pattern displays intermediate characteristics, with strain to failure ˜13% and yield and tensile strength of 1,213 and 1,444 MPa, respectively.A full-field strain map obtained during the uniaxial test of a hybrid specimen by Digital Image Correlation (DIC) is shown in Figure 7b, clearly showing that plastic strain can be effectively localized in the softer domains.Collectively, two key results emerge: i.By simply tuning the energy density upon printing, we can generate two different microstructures, both yielding desirable but different mechanical properties.Both phases are easily printable with high quality and exhibit differences in yield strength and ductility of the order of 20% and 150%, respectively, which are much larger than variations achievable with any other previously demonstrated microstructural control strategy in LPBF.ii.Functionally graded hybrid samples can be readily fabricated by interpenetrating domains printed with different conditions at a resolution of the order of 100μm.These hybrid printing strategies can be used to fully tune the mechanical response within the bounds presented by the two end microstructures, as well as a mechanism to localize plastic deformation in specific regions of the sample.

Conclusion
In summary, we demonstrated local microstructural control in a PH stainless steel printed via LPBF.When printed at high energy density (400 J mm −3 ), the material develops a fully martensitic microstructure; by contrast, when printed at lower energy density (100 J mm −3 ), a dual-phase microstructure ensues, with martensitic laths interspersed with large ferritic grains, which largely populate the inter-hatch regions of the samples.The two microstructures exhibit significantly different mechanical properties, with ˜20% differences in strength and ˜150% differences in ductility.We demonstrate local microstructural control with a resolution of ˜100 microns, in samples of different shapes and sizes.
While the kinetic pathways that lead to different microstructural evolutions different processing conditions are still under investigation, we experimentally confirm that ferritic grains consistently form exclusively in the inter-hatch region during subsequent thermal cycles, when the material is reheated within the temperature range of -ferrite stability.Computational studies are currently underway to develop a mechanistic understanding of these complex phase evolutions.We expect that the findings presented in this study can be extended to other multi-phase alloy systems (e.g., Titanium alloys), significantly expanding the design space for LPBF-processed gradient structures.

Experimental Section
Nitrogen atomized 17-4PH powder from Carpenter Additive was used as feedstock for this study.The particle size distribution was between 20 μm and 40 μm.All samples printed using an SLM Solutions 125HL printer, with the build chamber in a 99.99% Nitrogen atmosphere and the build substrate pre-heated to 200 °C.A laser power of 200 W, a layer thickness of 30 μm, a hatch spacing of 100 μm, and a scan strategy of "borders inside-out" were kept constant throughout the printing."Borders insideout" is a scan strategy in which concentric outlines (borders) of the parts are printed, from the smallest inside the core of the part toward the outer border of the part.The scan speed was varied from 166 to 1333 mm −1 s −1 , in order to vary the volumetric energy density from 50 to 400 J mm −3 .Energy density was calculated by E = P v⋅h⋅t where P is the laser power, v is the laser scan speed, h is the hatch spacing, and t is the layer thickness.
Wedge specimens were printed at four different energy densities (50, 100, 250, and 400 J mm −3 ), and with two different printing directions (in the plane of the triangle and perpendicular to it).Thin-wall specimens (with single track and double track width) were printed vertically, with two different energy densities (100 and 400 J mm −3 ).Dog-bone-shaped specimens were printed in the vertical direction, with three different printing conditions: E = 100 J mm −3 throughout, E = 400 J mm −3 throughout, and a hybrid "brick-and-mortar" pattern, with harder (E = 400 J mm −3 ) prismatic bricks embedded in a softer (E = 100 J mm −3 ) matrix.Gradient blocks were printed with alternating layers at low energy density (E = 100 J mm −3 ) with layers at high energy density (E = 400 J mm −3 ), resulting in microstructural gradients parallel to the printing direction.Finally, hybrid specimens (the UCI logo and the soft core/hard shell cylinder) were printed with microstructural gradients parallel to the build platform: multiple energy densities in a single layer were achieved by interpenetrating two geometrical files, each printed with uniform energy density, in a single build.
Samples were removed from the building substrate via wire electrodischarge-machining (EDM).Post-processing heat treatments of the samples consist only of aging, without any solutionizing step.Samples were aged at 482 °C for 60 min in a Nabertherm B400 furnace in an ambient atmosphere with a heating rate of 10°C min −1 , and then air quenched.
Microstructural analysis for lower magnifications was performed using an Olympus DSX10-UZH Digital Optical Microscope.Higher magnification microscopy was conducted using a TESCAN GAIA-3 Scanning Electron Microscope (SEM).Electron-Backscatter-Diffraction (EBSD) phasedistribution maps were acquired with an Oxford NordlysMax3 detector, and the Kikuchi patterns were processed via Aztec software.Particle size distribution of the feedstock was imaged in secondary electron (SE) mode.Phase fractions were calculated using ImageJ Software.Samples were mounted in epoxy and polished using standard metallurgical procedures down to 0.05 μm.To differentiate between the phases, samples were etched using Waterless Kalling's, aka Kalling's No.2 Reagent.Samples were submerged in etchant for 40 s before immediately being rinsed, sonicated for 1 min, and air dried.
Higher resolution microstructural characterization of the lattice structure and precipitates is conducted in a JEOL JEM-2800 scanning/transmission electron microscope (S/TEM) with a Gatan Oneview camera, operating at a beam voltage of 200 kV.JEM-2800 was equipped with dual 100 mm 2 silicon drift detectors (SDD) for energy-dispersive Xray spectrometry (EDS).TEM lamellae were extracted using a focused ion beam (FIB) in a Quanta 3D FEG dual beam SEM/FIB (Thermo Fisher Scientific Inc.).
Tensile tests were conducted on an Instron 5985 frame 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 0.001 s −1 .Strain mapping via Digital Image correlation (DIC) was conducted using a Correlated Solutions system, and the captured speckle patterns were processed via Vic3D software.Vickers hardness measurements were taken with a 500 g load held for 10 s for each indent.Hardness maps were created with indents spaced 150 μm apart, across the entire sample.A Wilson VH3300 automated indenter was used for mappings.

Figure 1 .
Figure 1.(a-d) Optical micrographs of wedge builds printed with an out-of-page build direction, increasing the energy density from 50 to 400 J mm −3 .(e-h)Optical micrographs of wedge builds printed with an upwards build direction, increasing the energy density from 50 to 400 J mm −3 .Higher magnification of wedge builds printed with an out-of-page build direction, showing dual phase versus martensitic microstructures corresponding to (i) E = 100 J mm −3 and (j) E = 400 J mm −3 respectively.(See FigureS1, Supporting Information for geometric details).

Figure 2 .
Figure 2. (a-d) SEM images of dual phase versus fully martensitic microstructures from gauge sections of dog bone specimens processed with E = 100 J mm −3 and E = 400 J mm −3 , respectively.Microstructures of the inner core versus the outer border of the samples are shown in (b,c) for E = 100 J mm −3 and (e,f) for E = 400 J mm −3 .(g) Optical image of an etched hybrid sample printed by alternating 1 mm layers at E = 100 and 400 J mm −3 , with IPF-X maps, phase maps, and band contrast maps shown in (h-j) for the inset in (g), respectively.The build direction is upward throughout the figure.(See Figure S1, Supporting Information for geometric details).

Figure 3 .
Figure 3. Single-track walls printed with (a) E = 100 J mm −3 and (b) E = 400 J mm −3 .Dual-track walls printed with (c) E = 100 J mm −3 and (d) E = 400 J mm −3 .Notice that ferritic grains are only present in the inter-hatch spacing in dual-track walls printed with the lower energy density.

Figure 4
Figure 4. STEM and STEM-EDS images of a sample processed with E = 100 J mm −3 in the as-printed (a-e) and heat treated conditions (f-j), and a sample processed with E = 400 J mm −3 in the as-printed (k-o) and heat treated conditions (q-t).Nb-rich precipitates are visible in the heat treated samples.

Figure 5 .
Figure 5. Optical images (a-c) and Vickers HV0.5 hardness maps (d-f) of etched hybrid samples: (a,d) a block printed by alternating 1 mm-thick layers processed with E = 100 and 400 J mm −3 ; (b,e) a cylinder printed with an E = 100 J mm −3 core and a 400 J mm −3 outer shell; and (c,f) a block printed at 100 J mm −3 encompassing the UCI logo printed at E = 400 J mm −3 .(See Figure S1, Supporting Information for geometric details).

Figure 6 .
Figure 6.(a) Optical image of an etched hybrid sample, printed by alternating layers processed at E = 100 and 400 J mm −3 , with gradually lower thickness.The microstructural gradient is aligned with the printing direction.(b) Magnification displaying the microstructural differences between the layers.(c,d) Hardness map and hardness profile (averaged across the width) of region in (b), demonstrating a resolution for microstructural control of the order of 150 μm.(See Figure S1, Supporting Information for geometric details).

Figure 7 .
Figure 7. (a) Engineering stress-strain curves from tension tests on dog bone specimens printed with differing processing strategies: (i) uniformly low energy density of 100 J mm −3 (black); (ii) uniformly high energy density of 400 J mm −3 (blue), and (iii) a "brick-and-mortar"-inspired architecture, embedding prismatic domains (bricks) printed at 400 J mm −3 in a matrix (mortar) printed at 100 J mm −3 (red).(b) Stress-strain curve with DIC strain maps for a "brick-and-mortar"-inspired sample, clearly showing localized plastic deformation in the softer regions printed at E = 100 J mm −3 .(See Figure S1, Supporting Information for geometric details).