High-speed manufacturing-driven strength-ductility improvement of H13 tool steel fabricated by selective laser melting

ABSTRACT H13 tool steel was additively manufactured by selective laser melting (SLM). The sample printed at a higher laser scan speed exhibited higher strength and ductility than those of the sample printed at a lower speed. The samples were repeatedly exposed to a massive heat input during the SLM. The in-situ tempering effect was applied to the sample; the phase fraction is changed by varying the heat input by controlling the laser scan speed. The microstructure analysis showed that the sample printed at a higher scan speed had a higher fraction of retained austenite than at a lower speed. The former was affected by deformation-induced martensitic transformation with enhanced strain-hardening ability. This study entailed the control of process parameters to improve the mechanical properties and the productivity of SLM-printed H13 tool steel. It investigated the relationship between the laser scan speed and the phase fraction, whose effect on the mechanical properties was confirmed.


Introduction
Hot-work tool industries have grown rapidly owing to the increasing demand for high-performance products (e.g.speed drills, moulds and dies) with high cooling efficiencies [1,2].H13 steel is suitable for hot-work tools because of its excellent wear and high thermal shock resistance [3][4][5].H13 steel is manufactured by casting with machining for fabricating moulds and dies; however, recent industries demand novel manufacturing techniques to produce complex-shaped products to overcome the limited design freedom of traditional manufacturing techniques [6][7][8][9][10].Moreover, producing engineering parts with geometric complexity using traditional processes such as cutting or tooling is difficult because H13 steel exhibits high hardness and low formability [11].
Recently, the emergence of metal additive manufacturing (MAM), known as metal 3D printing, has offered unprecedented merits for producing engineering parts with a high degree of design freedom for user-customised applications.Among the various MAM techniques, selective laser melting (SLM) is a promising process with high precision and good spatial resolution for producing geometrically optimised products [15][16][17][18].Over the past decade, SLM of H13 steel has been extensively investigated to overcome the challenge of traditional manufacturing methods for producing high-performance hot-work tools, moulds and dies [19][20][21][22][23].
However, the SLM process window for obtaining a defect-free sample is not wider than that of stainless steel because H13 steel is vulnerable to cracking owing to its brittleness and the large thermal residual stress induced by rapid heating or cooling (10 6 -10 8 K/s) [23][24][25][26][27].A low scan speed is often adopted for sample fabrication because a high heat input is required to build fully dense H13 steel to effectively avoid lack-of-fusion (LOF) defects during SLM, leading to low SLM productivity [11,23].Furthermore, fully dense H13 steel can be successfully fabricated by controlling the SLM process parameters; however, the ductility of SLM-processed H13 steel is not as favourable as its wrought counterpart [28].
Therefore, numerous material scientists have recently established effective guidelines to optimise SLM process parameters for obtaining a fully dense sample and a promising strength-ductility combination of the processed H13 steel with high productivity [23,29,30].This study successfully fabricated H13 steel with a sound microstructure by SLM.Furthermore, the relationship between the microstructure and the resulting mechanical properties of SLMprocessed H13 steel was investigated to find a strategy for improving the strength-ductility combination with high productivity of H13 steel via SLM.

Raw materials
This study used a commercial gas-atomized powder of H13 steel (Sandvik Inc., Sweden) for sample building by SLM. Figure 1(a) shows an image of the spherical-shaped powder used in this study This study used a commercial gas-atomised powder of H13 steel (Sandvik Inc., Sweden) for sample building by SLM. Figure 1(a) shows an image of the sphericalshaped powder used in this study.The chemical compositions of the powder were presented in Table S1.The chemical composition was analysed by using energy dispersive spectrometry (EDS, Oxford Instruments, UK) and a carbon/sulfur determinator (True-Spec Micro CHNS, LECO Co., Netherlands).The particle size distribution of the powder was determined using a particle size analyser (PSA; FPAR-1000, Formulaction, France).The measured values of D10, D50 and D90 were 21, 33 and 47 µm, respectively (Figure 1(b)).Table 1 lists the properties of the powder.The Hall flow rate of the powder was measured using a Hall flowmeter funnel (ACuPowder International LLC) according to ASTM B213-17.The apparent density of the powder was estimated using a Hall flowmeter funnel based on ASTM B212-17.The mass per unit volume of the tapped powders was measured following the standard test of ASTM B527-15 to measure the tap density of the powders.Furthermore, the Hausner ratio (i.e.tap density/ apparent density) was calculated to estimate the flowability of the powder.

Sample fabrication
Cuboidal blocks (160 × 3 × 14 mm 3 ), designed to resemble dog bone-shaped tensile specimens, were built using a commercial SLM machine (Concept Laser M2 Series 5, GE additive, U.S.A.) in a meander scanning pattern with 90°layer-by-layer rotation under a 99.7% Ar atmosphere.A schematic of this process is shown in Figure 2. The red arrows depict the laser-scanning tracks in the N-th and (N + 1)-th layers during the SLM.Note that the longitudinal direction of the cuboidal sample is labelled as the X direction, and the direction perpendicular to the X direction and building direction (Z direction) is labelled as the Y direction.The extraction of the plate-type tensile sample is shown schematically in Figure 2. The heat input is a major factor in building a defect-free sample by SLM [23].Sharp pores between the layers can be produced as LOF defects in the sample when the heat input is too low for the sample during SLM.Gas pores or keyhole defects are generated during SLM when the heat input is significantly high.The preliminary process optimisation was performed on the H13 powder under a hatch space and layer thickness of 77 and 30 μm, respectively, by controlling the laser power (280-360 W) and laser scan speed (400-1200 mm/s) without build plate preheating during the SLM process to obtain dense samples (see Figure 3 and Supplementary Figure S1).Based on our preliminary study, when the volumetric energy density (VED) was below 100 J/mm 3 , lack of fusion (LOF) defects having sharp shape were detected in the samples due to low heat input [11][12][13][14].On the other hand, when the  VED is over 245 J/mm 3 , hot-crack defects could be produced due to excessive heat input to the samples during SLM process.We investigated the samples without cracks or keyhole defects to explore the relationship between the microstructure and mechanical properties; The samples were prepared at a laser power of 310 W with scan speeds of 600 and 700 mm/s, denoted as the V600 and V700 samples, respectively.

Microstructural characterisation and mechanical testing
The cuboidal block samples were machined into platetype tensile samples (gauge length: 25 mm; gauge width: 6 mm; thickness: 3 mm) as the ASTM E8 subsize standard to investigate the mechanical behaviours of the samples (Figure 2).The tensile direction of the sample was parallel to the longitudinal direction of the  block sample (X-direction).Uniaxial tensile tests were performed using a universal testing system (5982, Instron Co., U.S.A.) equipped with an extensometer at a strain rate of 10 −3 s −1 at room temperature.All tensile tests were repeated at least thrice for each sample condition to ensure the reproducibility of the data.All samples were mechanically polished using SiC abrasive paper and diamond suspension up to a surface roughness of 1 μm for microstructural analysis.X-ray diffraction (XRD) was conducted using an Xray diffractometer (D/MAX-2500, RIGAKU Co. Japan) with Cu-Kα radiation (wavelength: 0.15418 nm) to identify the constituent phases of the as-built samples.The XRD scans were performed in the scattering (diffraction) 2θ range of 30°-100°with a scan speed of 10°/min.Electron backscatter diffraction (EBSD) analysis was performed to investigate the microstructure of the samples using a field-emission scanning electron microscope (FE-SEM, JSM-7900F, JEOL Ltd., Japan) system.The EBSD data were interpreted using phase imaging microscopy software (TSL OIM Analysis 7).X-ray diffraction (XRD) and electron backscatter diffraction (EBSD) analyses were conducted in an area located 0.5 cm away from the fracture zone, as illustrated in Supplementary Figure S2.

Initial microstructure
Figure 4 shows the XRD patterns of V600 and V700 samples.The presence of retained austenite in the SLMed H13 steel has been widely reported [11,28].The segregation of austenite-stabilising elements to the cell boundaries in SLMed H13 steel promotes forming of austenite at the cell boundaries [31].Holzweissig et al. [28] investigated the microstructural evolution of hot-work tool steel during SLM.They reported that austenite could be stabilised by carbon diffusion, like the quenching and partitioning (Q&P) process, during the complex thermal cycle of the SLM process.The samples underwent multiple reheating thermal cycles during the SLM process, which led to the partitioning and diffusion of carbon from the supersaturated martensite to the cellular boundaries with austenite reversion and the growth of retained austenite [28,31].
Figure 5 presents SEM-back scattered electron (BSE) micrographs, indicating the cell structures of the V600 and V700 samples.The cell boundaries can be identified by different contrasts between the cell interiors and boundaries, implying a different composition between the cell interiors and boundaries in the SEM-BSE micrographs [32].C, Co, Mo and V are usually enriched at the cell boundaries in the SLMed H13 steel shown in Figure 6 and Supplementary Figure S3 [19,20].Kong et al. [33] reported that solute atoms (i.e.C, Co, Mo and V) are segregated at the cell boundary due to compositional supercooling as a thermodynamic factor and the surface instability of Bernard Marangoni in terms of kinetics during the formation of the cell structure.
Figure 7 shows the EBSD inverse pole figure (IPF) map and phase map of the V600 and V700 samples.Figure 7(a) and (b) presents the IPF maps, showing diversified grain orientations of both retained austenite and the lath martensite.The microstructure was dominated by the lath martensite (a ′ -phase).The fine block martensite formed by a group of parallel lath martensite occupied the inner space of the retained austenite.The average block martensite size of the V600 and V700 samples are 2.45 and 2.03 µm, respectively.The high cooling rate (10 6 -10 8 K/s) formed martensite during melt pool solidification [26,34,35].The columnar grain and the grain boundary consisted of martensite and retained austenite, respectively (see Figures 7(c) and (d)) [3,23,29,36].A fully austenitic microstructure was formed after surpassing the temperature of austenite formation.During cooling, martensite transformed.Austenitic grains gradually transformed into the lath martensite, except for the cell walls of the solidification structure, which remained austenitic due to segregation.Retained austenite grains were trapped between the lath martensite, forming a discontinuous network of small austenitic grains because the full transformation was not attained [28,38,39].
Localised retained austenite observed at cell-to-cell boundaries.Austenite appeared as a long needle in the EBSD phase map (when the dendrite axes were in the image plane) or as small circles (when the dendrite axes were perpendicular to the image plane) depending on the orientation of the dendrites.The interdendritic areas were enriched with Ti, Mo and C owing to microsegregation during solidification, i.e. partitioning of solute elements into the remaining liquid [41].The solute enrichment causes retained austenite to be found only in specific areas.The lower the martensitic transformation temperature, the higher the elemental content and the more active the austenite stabilisation effect owing to C, Co, Mo and V being ferrite stabiliser elements [37].

Tensile properties
Figure 8 shows the results of the tensile tests for the V600 and V700 samples (for the detailed results fabricated at different power included in Supplementary Figure S4).The yield strength (YS) and ultimate tensile strength (UTS) of the V700 sample were higher than those of the V600 sample; however, the elongation (EL) of the V700 sample was higher than that of the V600 sample.When the strength of metallic materials is increased, the ductility is compromised, known as the strength-ductility trade-off.However, we found that increasing the laser scan speed from 600 mm/s to 700 mm/s for the sample building increased the strength and ductility of the SLMed H13 steel.Specifically, YS, UTS and EL increased from 937 ± 13 MPa to 1101 ± 108 MPa, 1513 ± 63 MPa to 1811 ± 104 MPa and 3.19 ± 019% to 4.73 ± 0.70%, respectively.The YS and UTS values increased slightly at different laser scan speeds; however, the difference was insignificant.Meanwhile, EL increased by ∼1.5%, ∼1.5 times.It is noted that the strain hardening ability of the V700 sample is much better than that of the V600 sample.Moreover, the region of the positive strain hardening rate can be observed in the tensile curve of the V700 sample, while the strain hardening rate is gradually decreased with increasing strain for the V600 sample.

Effect of retained austenite on tensile properties
Figure 9 observed the EBSD kernel average misorientation (KAM) maps of the V600 and V700 samples (for the XRD patterns of the deformed the samples, see Supplementary Figure S5).For the average KAM value, the V600 and V700 samples are 0.66 and 0.81, respectively.The average KAM value of the V700 sample is higher than the 600 sample because the  V700 sample has a higher cooling rate.The higher scan speed leads to a higher cooling rate as its exposure a longer time to the laser beam [17].
Generally, some researchers have reported that the YS value is reduced when the excessive retained austenite volume fraction is generated [42][43][44].Meanwhile, in this study, the faster the scan speed, the higher the average KAM value of the V700 sample than the V600 sample due to the relatively higher cooling rate.The average martensitic block size of the V700 sample is also smaller than the V600 sample due to the difference in cooling rate, so it is confirmed that the YS increases despite the high retained austenite volume fraction [29,[45][46][47][48][49].
Figure 10(a-d) represents the EBSD IPF map and phase maps of the V600 and V700 samples after deformation.Most specimens had a reduced volume fraction of retained austenite due to the deformation.The specimen with a volume fraction of retained austenite of 12.2% decreased the most to 4.5% after deformation at the V700 sample.The phenomenon of decreasing fraction was similar to that explained in Figure 9.The deformation-induced martensite transformation (DIMT) effect occurs in which the retained austenite transforms into the martensite phase.DIMT occurred locally in the retained austenite and was dispersed entirely in a dot form, unlike before deformation, in which the retained austenite was formed in a lined form.
The DIMT was suppressed by the low diffusion distance of carbon in austenite, which also suppressed dislocation annihilation and movement during  deformation.Therefore, austenite deformation allowed martensite formation; the carbon atoms could not diffuse out of the prior metastable austenite phase, producing lattice strain [50].Therefore, the lattice strain of austenite to ε and a ′ -martensite increased strain hardenability and ductility [51,52].In martensitic steel, a large volume fraction of metastable austenite is undesirable because it lowers the overall strength; however, metastable austenite simultaneously increases the EL and UTS because of the DIMT [11,23,31,33,51].Therefore, the DIMT phenomenon becomes more active, and the mechanical properties are improved with increasing volume fraction of retained austenite [11,23].

Laser scan speed effect on phase fraction
The fast cooling rate (10 6 -10 8 °C/s) during the SLM process results in pronounced undercooling during solidification, eventually hindering or delaying phase transformation.The non-equilibrium phases in the H13 steel fabricated by SLM were predicted from their continuous cooling transformation diagrams (CCTs), as shown in Figure 11(a).The cooling rates of the deposited samples were greater than 100°C /s, as depicted by the simulated results, which implies that the samples fabricated by SLM were mainly composed of martensite.When the in-situ tempering temperature over the martensitic transformation temperature, bainite is formed, while it does not form when the in-situ tempering temperature remains below this threshold.The volume fractions of retained austenite and martensite differed depending on the laser scan speed.The as-solidified layers or tracks can be tempered under the thermal influence of the next adjacent layers or tracks.Hence, the final microstructure depends on the temperature that triggers solidification and in situ tempering shown in Supplementary Figure S6 [23,29,[53][54][55][56][57].In V600, bainite was formed (see CCT diagram, Figure 11(a)) while remaining above the martensitic transformation temperature owing to the high heat input.However, V700 maintained a temperature below the martensite transformation with a lower heat input than V600.Hence, bainite did not form.Therefore, the V700 sample has a higher retained austenite volume fraction than the V600 sample [27,40,53].Similarly, Fonseca et al. [23] reported that the volume fraction of retained austenite was higher when the heat input was lower than the higher heat input.The volume fraction of retained austenite depends on the temperature above or below the martensite transformation temperature because the insitu tempering temperature depends on the heat input.
Another in-situ tempering effect is the main cause for the generation of carbide precipitates in the H13 steel fabricated by SLM (see Figure 10(b)).The insitu tempering effect leading to the precipitation of carbide precipitates has been verified during the conventional SLM of H13 steel [11,29,58,59].The lower the heat input, the higher the fraction of carbidebased precipitates owing to the development of the microstructure at a relatively low temperature, probably because the fraction of carbide increased as the temperature decreased (Figure 11(b)).

Conclusions
This study investigated the effect of phase transformation on the mechanical properties of H13 tool steel manufactured using a high-speed SLM process.The following conclusions were drawn: (1) The H13 tool steel fabricated by SLM using a variable laser scan speed exhibited excellent mechanical properties.The YS, UTS and EL increased from 937 to 1101 MPa, 1513 MPa to 1811 MPa and 3.19% to 4.73%, respectively.(2) Segregation of solute atoms such as C, Co, Mo and V was formed at cell boundaries due to compositional supercooling and Bernard-Marangoni surface instability during cell structure formation.Segregation is a stabilising element that lowers the martensitic transformation temperature, forming locally long needles and small circular retained austenite.
(3) The in-situ tempering effect is varied by different scan speed between V600 and V700 samples during SLM process.Bainite formed when the in-situ tempering temperature exceeded the martensitic transformation temperature.However, it did not form when the temperature was lower than the martensitic transformation temperature.Therefore, a fast scan speed (i.e.lower heat input) resulted in a high retained austenite volume fraction.Additionally, the lower the heat input, the higher the fraction of carbide-based precipitates because the microstructure develops at a relatively low in-situ temperature.(4) The excellent mechanical properties of the H13 steel samples fabricated by high-speed SLM were mainly attributed to the retained austenite (FCC, γ-Fe) volume fraction in the as-built sample after the SLM process.The influence on the mechanical properties increased with decreasing block size of martensite a non-equilibrium microstructure and increasing volume fraction of the retained austenite owing to the active occurrence of DIMT.

Figure 1 .
Figure 1.(a) SEM image showing the morphology of the gas-atomized H13 powder.(b) Particle size distribution of the powder used in this study.

Figure 2 .
Figure 2. Schematics of the scanning strategy for sample printing using SLM.The red arrows schematically represent the laserscanning paths.

Figure 3 .
Figure 3. Optical micrographs for the fine polished surface of the samples fabricated by different laser powers (280-360 W) and scan speed (400-1200 mm/s).

Figure 6 .
Figure 6.SEM-EDS maps of the solidification structure of SLM-fabricated H13 tool steel.

Figure 8 .
Figure 8. Tensile engineering strain-stress curves of the V600 and V700 samples.Inset: the measured tensile properties of the samples.

Figure 11 .
Figure 11.Equilibrium phase diagram of H13 steel elaborated by thermodynamic calculation: (a) CCT diagram at the different cooling rate and (b) equilibrium phase diagram.