Recent efforts to process 17-4 PH stainless steels via additive manufacturing (AM) have revealed pronounced variability in both microstructure and mechanical properties [1][2][3][4][5][6][7][8][9]. Such variability has limited the broader adoption of 17-4 PH in AM applications, despite the well-recognized advantages of AM for producing highly customized and geometrically complex components [10][11][12]. The cause of this variability lies in the nonequilibrium thermal conditions intrinsic to AM. Each deposited layer undergoes multiple cycles of melting, rapid solidification, and reheating [13,14], thereby producing solidification pathways far from those observed under conventional processing. Consequently, small changes in processing parameters or chemical compositions, even within specifications, produce dramatically different microstructures, ranging from fully martensitic (b.c.t.) [4] or ferritic (b. c.c.) [15] to fully austenitic (f.c.c.) [16], as well as any combination of these three phases [4,16,17].
This strong sensitivity to processing conditions translates into substantial scatter in mechanical properties. Reported tensile yield strengths for selective laser melted (SLM) 17-4 PH span ~600-1200 MPa, while elongation to failure varies from 2 % to 15 % [1,2,16,[18][19][20][21][22][23]. Moreover, the steep thermal gradients and rapid cooling rates inherent to AM promote significant residual stresses (hydrostatic or directional) [24], which influence strength, fatigue resistance, and fracture behavior [25,26]. Different post-processing heat treatments have been explored to mitigate the microstructural variability associated with AM processing. A common approach involves solution treatment (1000-1060 → C), followed by controlled aging, often combined with hot isostatic pressing (HIP) to reduce porosity and residual stresses [27][28][29]. Quenching from the solutionized state produces the desired martensitic matrix, while subsequent aging determines the balance between strength and toughness. Aging at lower temperatures, such as the H900 condition at 482 → C, yields maximum strength through the formation of nanoscale Cu-rich precipitates, but it reduces ductility. By contrast, higher-temperature aging treatments, such as H1150 at 621 → C, produce moderate strength combined with improved toughness [21,[30][31][32].
Because 17-4 PH commonly retains some austenite after AM, even after a proper post-processing treatment, a further design-critical concern is yield asymmetry between tension and compression arising from the transformation-induced plasticity (TRIP) effect [33][34][35][36][37]. Yield asymmetry, the difference between tensile and compressive yield strengths, directly impacts AM'ed component reliability, affecting fatigue resistance, failure modes, and overall service performance in multiaxial loading. Earlier studies on TRIP-assisted martensitic steels established the mechanistic basis for yield asymmetry under different loading modes [36]. In tension, retained austenite (f.c.c.) transforms readily to martensite (b.c.t.), producing additional transformation strains that couple with plastic flow and reduce the apparent tensile yield stress. Under compression, however, this transformation is largely suppressed due to crystallographic constraints and martensite variant selection, resulting in higher compressive yield stresses. Beyond the static strengthening from the newly formed martensite, this transformation acts as a concurrent deformation mode and contributes to "dynamic softening," when a stress-assisted transformation governs the plastic response [33,36,38]. The magnitude of the compressive-tensile yield asymmetry (CTYA) depends on the stability of retained austenite, precipitation state, crystallographic texture [36,39], and residual stress fields (hydrostatic or directional) [34,40,41], which in AM'ed steels can be substantial due to high cooling rates and steep thermal gradients. We note that this type of asymmetry is fundamentally different from those reported in non-cubic systems such as magnesium [42][43][44] or titanium [45][46][47] alloys, where the origin arises from limited slip systems, pronounced tension-compression twinning asymmetries [42,43], and anisotropic plasticity [46], rather than transformation-induced mechanisms.
The extent and origins of CTYA in additively manufactured 17-4 PH stainless steel, and its sensitivity to common post-processing treatments, remain unquantified. We examine the microstructure and mechanical response of an SLM'ed 17-4 PH stainless steel in the as-printed and different heat-treated states. We integrate CALPHAD-based equilibrium calculations to guide phase stability analyses, electron backscatter diffraction (EBSD) and magnetic measurements to quantify texture and retained austenite, and comprehensive tensile and compressive testing to measure yield asymmetry. Comparisons are made between the aged specimens with and without prior solution treatments, where we identify conditions that lead to aging-induced embrittlement.
Rectangular bars of 12 ↑ 12 ↑ 90 mm 3 (Table 1, Fig. S1) were SLM'ed horizontally on a 316 stainless steel build plate using argon gasatomized 17-4 PH stainless steel powder (Osprey®, Sandvik, Neath, UK) with a particle size distribution of 15-45 μm. The SLM process was performed in an argon environment with ω0.2 vol% oxygen employing a DMG MORI LASERTEC 12 SLM machine (DMG MORI, Bielefeld, DE) after determining the optimal processing parameters, Table 2.
Post-processing heat treatments were performed to assess microstructural and mechanical property changes. The experimental groups included: (i) as-printed condition; (ii) stress-relieved at 300 → C for 2 and 4 h; (iii) aged at 482 → C for 1 and 24 h, directly from the as-printed state; (iv) aged at 482 → C for 1 h following stress-relieving at 300 → C for 4 h; and (v) solution treated at 1000 → C for 1 h followed by an aging step at 482 → C for 1 h. Heat treatments were conducted in a box furnace in air, followed by air cooling. For group (v), samples were quenched into water at ambient temperature immediately after the solution treatment.
Tensile testing was performed according to ASTM E8 standard using subsize flat specimens with a gauge length of 25.4 mm, a width of 6 mm, and a thickness of 1.8 mm. Compression tests were conducted on cylindrical specimens with a diameter of 6 mm and a height of 12 mm according to ASTM E9 standard. To minimize friction, specimens were coated with boron nitride and placed between two mica sheets. Both tensile and compression tests were performed at a nominal strain rate of 10 ↓3 s ↓1 , using an MTS Criterion Model 45 machine. Strain was measured using an MTS extensometer for tensile tests and the digital image correlation [48] (DIC, Correlated Solutions, Irmo, SC, USA) method for compression tests. For each heat treatment condition, two specimens were tested to determine the stress-strain curves under tensile and compressive loading. All specimens were machined from the horizontally printed bars (Fig. S1) perpendicular to the build direction (Z), using wire electrical discharge machining (EDM).
Samples for optical microscopy (OM), electron backscattered diffraction (EBSD), and scanning electron microscopy (SEM) were prepared using standard metallographic techniques. Specimens were mechanically ground using a series of SiC-loaded papers, followed by W. Wang et al. Materials Science & Engineering A 950 149537
polishing with diamond suspensions (6, 3, and 1 μm). Final polishing was performed with a 0.05 μm colloidal silica suspension. For EBSD, specimens were additionally electropolished with a 5 % perchloric acid solution in methanol to remove all polishing-induced deformation layers. An FEI Quanta 650 field-emission-gun SEM equipped with an Oxford Symmetry S3 EBSD detector was used at 15 kV for microstructural investigations. EBSD was performed on two specimens without prior etching: an as-printed specimen and a specimen solution-treated at 1000 → C for 1 h, followed by aging at 482 → C for 1 h. Kikuchi patterns were acquired using AZtec software (Oxford Instruments, Abingdon, UK) over a 150 ↑ 200 μm 2 scan area with a step size of 100 nm on specimens tilted at 70 → relative to the electron beam. Indexing of Kikuchi patterns was performed in the real Kikuchi pattern space using the band refinement approach (refined accuracy [49]), achieving indexing rates of ε99 % for the as-printed and ε95 % for the heat-treated specimens.
Post-processing, including orientation and phase mapping as well as the generation of pole figures, was performed using the AZtecCrystal software package (Oxford Instruments, Abingdon, UK). Electron channeling contrast imaging (ECCI) was employed to examine the grain structure of the as-printed specimen at a relatively low magnification (500X), providing a larger view of the columnar grains. An accelerating voltage of 5 kV was used for the ECCI analyses. For OM, specimens were etched using a solution of 25 g iron chloride (FeCl 3 ), 25 mL hydrochloric acid (HCl), and 100 mL distilled water for ~10 s to reveal grain boundaries. A MSAT instrument (Metis Instruments, Leuven, Belgium) equipped with a Lakeshore 480 fluxmeter was utilized for determining the volume fraction of retained austenite by measuring the saturation magnetization of the steel samples and correcting for the chemical composition. Each specimen was measured five times; the mean and one standard deviation are reported.
In Fig. 1, we present equilibrium thermodynamic calculations for a 17-4 PH steel, Table 1, performed employing the CALPHAD methodology [50] with ThermoCalc [51] and the TCFE9 v9.3 database. The predicted phase volume fractions as a function of temperature indicate that δ-ferrite (b.c.c.) is the first phase to form from the liquid state, starting at ~1460 → C, reaching a maximum volume fraction of ~46 at.% ~1428 → C. Upon further cooling, δ-ferrite (b.c.c.) transforms completely to γ-austenite (f.c.c.) through a peritectic reaction. At ~804 → C, γ-austenite (f.c.c.) begins to transform back to α-ferrite (b.c.c.). At a slightly lower temperature, ~750 → C, the M 23 C 6 carbide phase forms as the γ-austenite (f.c.c.) continues to decompose with decreasing temperature.
The M 23 C 6 phase achieves a maximum volume fraction of ~0.3 at.% ω~650 → C. This carbide phase is, however, not anticipated to form in the as-printed or aged specimens due to its very high activation energy for nucleation [53] and the extremely small diffusivities of its constituents: Mo and Cr, at the aging temperature of ~482 → C (D ~ω10 ↓23 m 2 /s [54]). The NbC carbide phase becomes stable at below ~1126 → C, with a maximum volume fraction of ~0.25 %. This phase is anticipated to precipitate during multiple heating-cooling cycles of SLM or solution treatment/austenitization at ~1000 → C. Copper-rich precipitates are predicted to form below ~956 → C, achieving a relatively high volume fraction of ~3.7 at.% at 500 → C, which is close to the standard aging temperature for this steel, 482 → C (900 → F).
During laser melting (for example, SLM), extremely high cooling rates (~10 4 -10 6 → C/s [55]) are achieved, which means that the equilibrium predictions in Fig. 1 cannot describe the actual far-from-equilibrium microstructures formed under such conditions. At these rapid solidification rates, kinetic constraints dominate phase selection, suppressing many diffusion-controlled transformations that equilibrium calculations predict. Consequently, the solidification pathway and the retention of metastable phases depend strongly on both solidification dynamics and the concomitant cooling rates. Nevertheless, equilibrium calculations remain valuable for establishing the thermodynamically accessible phases and transformation sequences. They provide a critical framework for qualitatively assessing which solidification pathways are possible and which phases are most likely suppressed under rapid cooling. This thermodynamic baseline is used in Section 4.1 to interpret experimental observations and to compare the SLM-processed microstructures with the predicted equilibrium state.
Fan-shaped columnar grains (CG), revealed by ECCI, are evident in the as-printed microstructure, Fig. 2a, which reflect the steep thermal gradients and rapid solidification rates characteristic of SLM. At a higher magnification, Fig. 2b, between adjacent melt tracks and/or columnar grains, a fine network of equiaxed grains (FEG) is evident. Direct aging at 482 → C, Fig. S2, leads to negligible changes in grain structure, whereas solution treatment at 1000 → C followed by aging at 482 → C yields a distinctly different microstructure. The high-temperature step induces rapid, partition-less austenitization, concomitant with solution treatment and solute homogenization.
EBSD analyses of the as-printed specimen, Fig. 3a, reveal that the fan-shaped columnar grains observed in Fig. 2a , indicate the existence of a strong fiber texture along the build direction (BD) of the as-printed specimen, ω001ε//BD (maximum intensity of 4.9, (001) PF), which is consistent with the epitaxially grown elongated grains observed in Fig. S2a. A solution-treated and aged specimen, however, exhibits a significantly weaker, nearly random crystallographic texture (maximum intensity of 1.6, (001) PF).
The volume fractions of retained austenite measured by magnetometry V γ f , Table 3, demonstrate minimal variation for the different heat treatment conditions, with values ranging between 20.9 % and 24.0 %. Notably, the as-printed, stress-relieved (2 h at 300 → C), and aged specimens (1 h and 24 h at 482 → C) without prior solution treatments all exhibit nearly identical V γ f values (~21 %). This consistency in the measured V γ f values indicates that the retained austenite phase remains relatively stable within the 300-482 → C range. The specimen subjected to full solution treatment (austenitization) at 1000 → C followed by quenching and subsequent aging exhibits a modest increase in V γ f to 24.0 %.
The engineering stress-strain curves, Fig. 5, compare the mechanical behavior of the SLM'ed steel in the as-printed and different heat-treated conditions under uniaxial tensile (solid lines) and compressive (dashed lines) loading. The curves are offset by 1 % strain for clarity. The average yield strength, ultimate tensile strength, and elongation to fracture values for the tensile tests, and the yield strengths for the compression tests are displayed in Table 4. The as-printed specimens exhibit moderately good mechanical properties with a yield strength of ~843 MPa and elongation to failure of ~7 %. Under compression, the yield strength increases significantly to ~933.8 MPa, revealing pronounced compressive-tensile yield asymmetry (CTYA). Stress-relief treatments at 300 → C for 2 h and 4 h induce minimal changes in tensile and compressive responses, suggesting that this low-temperature treatment primarily relaxes internal residual stresses without significantly altering the microstructure. The essentially unchanged tensile yield strength demonstrates that precipitation hardening is negligible at this temperature, consistent with the small diffusivity of Cu in ferrite at 300 → C (D Cu α ↘ 4 ↑ 10 ↓28 m 2 /s [56]) and the nearly identical retained-austenite volume fractions, Table 3.
In contrast, direct aging at 482 → C for 1 h or 24 h without prior solution treatment results in completely brittle behavior under tension, with fracture occurring with negligible plastic deformation. This embrittlement prevents accurate determination of tensile yield strength in these conditions. A similar brittle response is observed when aging at 482 → C for 1 h is performed after a stress relief at 300 → C for 4 h. Under compression, however, the specimen directly aged at 482 → C for 1 h exhibits a relatively high yield strength of ~1294 MPa, that is, an increase of ~320 MPa in yield strength when compared to the as-printed specimen. This substantial strengthening is attributed to precipitation hardening imparted by Cu-rich nanoprecipitates, as documented in prior studies [17,57], which is consistent with the much enhanced diffusivity of Cu in ferrite at 482 → C, D Cu α ↘ 7.5 ↑ 10 ↓22 m 2 /s [56]. The observed ~320 MPa increase in yield strength after direct aging is in quantitative agreement with prior reports attributing 250-350 MPa [58] of hardening in 17-4 PH stainless steels to the formation of coherent Cu-rich nanoprecipitates. These nanoprecipitates, typically 2-5 nm in diameter, contribute to the strengthening primarily through shearing, Orowan looping, and coherency strain mechanisms.
The most significant improvement in tensile behavior is observed for the specimen solution treated at 1000 → C for 1 h, followed by aging at 482 → C for 1 h. This condition leads to the highest tensile yield strength (~1210 MPa) and some ductility (~2 %), highlighting the importance of a full solution treatment and concurrent austenitization. The corresponding compressive yield strength reaches ~1337 MPa, again showing marked compression-tension asymmetry, Table 5. To better illustrate the observed CTYA, we replot the tensile and compressive stress-strain curves in Fig. 6 along with the corresponding true curves (corrected for specimen cross-sectional changes) for the as-printed and stress-relieved specimens, as well as a specimen aged after solution treatment. The ratios of compressive-to-tensile yield strengths (σ c /σ T ) and the yield strength differences (Δσ) are displayed in Table 5. For all conditions, the true curves display similar work-hardening slopes under tension and compression, indicating comparable hardening behavior in both loading modes.
The fracture surface morphologies observed in Fig. 7 offer critical insights into the failure mechanisms associated with different heat treatments, as further corroborated by the corresponding engineering stress-strain responses in Fig. 5. The as-printed specimen, Fig. 7a, exhibits a highly ductile fracture characterized by numerous equiaxed dimples with a deep conical shape and an average diameter of ωDε ~0.5 μm. This micro-void coalescence mechanism [59] is an indication of significant plastic deformation and energy absorption prior to failure, consistent with moderate elongation to fracture (~7 %) and a relatively moderate tensile strength (~1000 MPa), Fig. 5 and Table 4. In stark contrast, the specimen aged directly from the as-printed state, Fig. 7b, demonstrates a flat fracture surface with pronounced cleavage facets and river patterns, which are indicative of a brittle fracture mode. This agrees with the tensile curves that fail prematurely in the elastic regime, indicating negligible ductility and plasticity. Fig. 7c, corresponding to the specimen solution treated at 1000 → C for 1 h followed by aging at 482 → C for 1 h, reveals a mixed (brittle-ductile) mode fracture morphology comprising relatively shallow dimples (ωDε ~0.7 μm).
This fracture behavior is reflected in the mechanical response, which combines relatively high strength (UTS ~1310 MPa) with a small elongation (~2 %), Table 4.
EBSD analyses of the as-printed 17-4 PH specimen, Fig. 3a, reveal a predominantly ferritic microstructure (V f ~80 %, Table 3) consisting of columnar grains aligned along the build direction. These grains exhibit strong epitaxial growth across multiple melt pools, Fig. S2a, resulting in a pronounced ω001ε//BD fiber texture, Fig. 4a. A similar ω001ε//BD texture was reported for an SLM'ed 17-4 PH steel solidifying in the ferritic mode [15,22]. Such microstructural features strongly indicate direct solidification of ferritic grains from the liquid phase, in strong Fig. 4. The (100), ( 110) and (111) pole figures (PFs) computed from the EBSD datasets displayed in Fig. 3 (a and e) along the build direction (BD) for the matrix of the (a) as-printed ferritic specimen; and (b) solution treated (1000 → C, 1 h) and aged (482 → C, 1 h) martensitic specimen. A strong (100) texture (max intensity 4.9, right-hand side legend) in the as-printed specimen is due to the epitaxial growth of columnar ferritic grains. Fig. 5. Engineering stress-strain curves under tensile (solid lines) and compressive (dashed lines) loadings for the SLM'ed specimens in the as-printed and different heat-treatment conditions. For clarity, curves are offset by 1 % strain. The heat treatments include: (i) stress-relieved at 300 → C for 2 h (blue) and 4 h (Magenta); (ii) aged at 482 → C for 1 h (red) and 24 h (green); (iii) stressrelieved at 300 → C followed by aging at 482 → C 1 h (brown); and (iv) solution treated at 1000 → C for 1 h followed by aging at 482 → C for 1 h (gold). All tests were conducted at ambient temperature at a nominal strain rate of 10 ↓3 s ↓1 . The average yield strength, ultimate tensile strength, and elongation to fracture values for the tensile tests, and the yield strengths for the compression tests, are displayed in Table 4. The specimens aged at 482 → C without prior solution treatment exhibit premature brittle fracture (minimal plasticity) under tensile loading, which does not permit the evaluation of their yield strengths under tensile stresses. Repeat tests are displayed using different shades of the corresponding colors. Compression tests were stopped at ~4 % strain. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Table 3 Volume fractions of austenite, V γ f , in a SLM'ed 17-4 PH stainless steel for different heat treatment states, determined by the magnetic saturation methodology.
As-printed 2 h @ 300 → C contrast to equilibrium thermodynamic predictions, Fig. 1, which predict an initial δ-ferrite formation followed by a δ → γ transformation upon cooling, and the termination of solidification by the liquid → γ reaction. The absence of a significant volume fraction of martensite in the as-printed condition further corroborates the ferritic solidification mode.
Our results suggest that the diffusion-controlled δ → γ transformation is completely suppressed, despite the wide temperature interval over which this massive partition-less [60] reaction is expected to occur (~1428-804 → C, Fig. 1). These results are consistent with in-situ XRD analyses of Guo et al. [19], which demonstrated that during laser melting (cooling rates ~10 4 → C/s) of a commercial 17-4 PH stainless steel, δ-ferrite solidified first, followed by γ-austenite, but the δ → γ transformation did not occur, leading to a final microstructure dominated by coarse δ-ferrite grains. Furthermore, the columnar morphology and epitaxial grain growth suggest that pre-existing δ-ferrite acts as a favorable nucleation template for subsequent ferrite growth during rapid melt-solidification cycles, reducing the nucleation barrier for δ-ferrite relative to γ-austenite and thereby further promoting ferritic growth across layers.
Fine equiaxed γ-austenite grains (V γ f ~20 %, Table 3) are observed between the columnar grains, suggesting their formation at the end of solidification. This is attributed to constitutional undercooling, caused by the rejection of solute elements such as Cu and Ni, which are strong austenite stabilizers. This mechanism is consistent with a columnar-toequiaxed transition (CET) [61,62], which is restricted to the final stages of solidification. Importantly, the γ-austenite grains form directly from the liquid phase, rather than by a solid-state δ → γ transformation, as demonstrated by their equiaxed morphology.
The solidification sequence in the present study can be summarized as: liquid → columnar δ-ferrite (b.c.c.) followed by liquid → limited equiaxed γ-austenite (f.c.c.) at the final stages of solidification. This nonequilibrium solidification pathway results in the retention of a predominantly ferritic microstructure in the as-printed specimens. Heat treatment at 1000 → C for 1 h is shown to austenitize the microstructure completely, and subsequent quenching produces a predominantly martensitic structure, Fig. 3c, with a much weaker crystallographic texture, Fig. 4b.
Tensile tests, Fig. 5, reveal that aging the as-printed alloy directly, that is, without a prior solution treatment/austenitization, leads to a marked loss of ductility and a brittle cleavage-type fracture response, Fig. 7b. Specimens fail under tensile stresses in the elastic regime at ~650-860 MPa, Table 4, well below their yield strength, ~1200 MPa, estimated from the compression tests. A prolonged aging treatment at 482 → C for 24 h yields a similar fracture response with minimal plasticity.
This pronounced embrittlement seems to be correlated with the highly textured ferritic microstructure (ω001ε//BD, Fig. 4a) developed during SLM. Previous studies have demonstrated that brittle fracture in high-purity iron (b.c.c.) frequently occurs along {001} cleavage planes [63]. Similar observations have been made for polycrystalline steels tested at low temperatures (77 K) [64,65]. In these specimens, it was Table 4 Tensile and compressive properties of a selective laser melted 17-4 PH steel subjected to different heat treatments (HT), determined from the stress-strain curves in Fig. 5. Values represent the average ↗ one standard deviation, except where otherwise noted. Test HT YS (MPa) UTS (MPa) El. (%) Uni. EL a (%) Tensile As-printed 859.6 ↗ 11.1 978.4 ↗ 7.2 6.5 ↗ 0.3 5.7 ↗ 0.2 2 h @ 300 → C 855.1 ↗ 15.9 955.2 ↗ 25.5 9.0 ↗ 4.2 6.6 ↗ 1.7 4 h @ 300 → C 865.1 ↗ 15.1 941.2 ↗ 18.2 7.0 ↗ 0.4 7.0 ↗ 0.4 4 h @ 300 → C ≃ 1 h @ 482 → C Brittle fracture at ~650 MPa ---1 h @ 482 → C Brittle fracture at ~820 MPa ---24 h @ 482 → C Brittle fracture at ~860 MPa ---1 h @ 1000 → C ≃ 1 h @ 482 → C 1210.4 ↗ 8.4 1310.3 ↗ 19.2 1.7 ↗ 0.2 1.7 ↗ 0.2 Compression As-printed 972.8 ↗ 18.6 ---
a Uniform elongation measured up to the ultimate tensile strength (UTS).
b Data obtained from a single test specimen.
Compressive-to-tensile yield strengths ratios, σ c /σ T , and Δσ, defined as the difference between the compressive and tensile yield strengths for an SLM'ed 17-4 PH steel in the as-printed and two different heat-treated states. Values were determined from the true stress-strain curves in Fig. 6.
Heat treatment as-printed 2 h @ 300 → C 1 h @ 1000 → C ≃ 1 h @ 482 → C σc σT
1.13 ↗ 0.03 1.12 a 1.11 ↗ 0.02 Δσ (MPa) 113.2 ↗ 21.7 109 ↗ 15.9 127.0 ↗ 19.1 a One compression test was performed.
Fig. 6. Engineering (solid lines, replotted from Fig. 5) and true (dashed lines) stress-strain curves under tensile (red) and compressive (black) loading for specimens in the as-printed, stress-relieved at 2 h at 300 → C, and solutiontreated at 1000 → C for 1 h followed by aging at 482 → C for 1 h conditions. For clarity, the curves are offset by 3 % strain. All true tensile curves are truncated at about the UTS value, beyond which necking occurs. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
W. Wang shown that a strong {001} texture component lowers the local critical stress for cleavage and promotes delamination along planes parallel to the rolling plane, as documented for textured ferritic steels [65]. In different steels, toughness values are shown to decrease significantly in the presence of {100} texture components [66,67].
Although the as-printed specimen exhibits ductile fracture, Fig. 7a, the strong crystallographic texture of ferrite likely promotes cleavagetype fracture and contributes to the observed loss in ductility after direct aging. Direct aging leads to the formation of Cu-rich nanoprecipitates within the ferritic matrix, which increases the yield strength significantly by ~320 MPa. This precipitation hardening simultaneously elevates the ductile-to-brittle transition temperature (DBTT) [68] into the ambient-temperature regime, particularly in a matrix already favorably aligned for {001} cleavage. We note that the volume fraction of the austenite phase remains essentially unchanged after direct aging (Table 3), suggesting that comparable TRIP-assisted strain accommodation [69] should, in principle, remain available.
It is worth noting that carbide precipitation at grain boundaries, which is known to cause embrittlement in certain steels, is not operative under the present aging conditions. Given the low carbon content of the alloy, the sluggish diffusivity of substitutional solutes such as Mo and Cr at 482 → C, and the high activation energy required for the nucleation of M 23 C 6 carbides, the formation of such carbides at grain boundaries during short-time aging is highly unlikely. This conclusion is further supported by the absence of any type of carbide precipitates in the fracture surface analyses. Furthermore, since the solution-treated specimens, aged at the same temperature (482 → C) as the directly aged ones, do not exhibit similar embrittlement, the possibility of M 23 C 6 or other carbide precipitates contributing to the embrittlement of the directly aged specimens can be ruled out conclusively.
In contrast, solution treatment at 1000 → C for 1 h austenitizes the microstructure, and subsequent quenching results in a predominantly martensitic structure, Fig. 3c, with a much weaker crystallographic texture, Fig. 4b. Specimens aged after this solutionizing step exhibit significantly enhanced tensile ductility, when compared to those aged directly from the as-printed state. The susceptibility for cleavage fracture is substantially reduced, and the fracture mode transitions toward a mixed brittle-ductile response, Fig. 7c. The improved combination of strength and ductility is consistent with precipitation hardening of tempered fine martensite and its significantly weaker crystallographic texture, as observed in wrought 17-4 PH steels. Further enhancements of the tensile ductility are anticipated at the expense of strength by aging the solution-treated specimens at higher temperatures, as demonstrated for wrought [31] and AM'ed [21] 17-4 PH steels.
The true stress-strain curves, Fig. 6, reveal distinct differences in their mechanical behavior under tensile and compressive loading. A pronounced compressive-tensile yield asymmetry (CTYA) is observed for all tested conditions, Table 5, which is characterized by consistently higher yield strengths in compression compared to tension. In the asprinted condition, the compressive yield strength exceeds the tensile value by Δσ ~113 MPa. Stress-relieving at 300 → C results in a slightly reduced asymmetry, Δσ ~109 MPa, but does not eliminate it, which indicates that the contribution of the residual stresses developed during SLM processing to the measured CTYA is relatively insignificant. The near-constant volume fraction of the retained austenite after stress relieving suggests that TRIP may play a dominant role in governing the CTYAs. This interpretation is further supported by the increased asymmetry, Δσ ~127 MPa, observed in solution-treated and aged martensitic specimens, which have a slightly higher retained austenite volume fraction (24 % vs. 21 %, Table 3).
Since all the tensile and compression tests are performed at ambient temperatures, well below the calculated M s temperature (~75-100 → C, Fig. S3), the stress-assisted transformation [36] is anticipated to dominate the TRIP effect as observed in wrought steels rather than strain-induced transformation [70]. Olson and Azrin [33] demonstrated that below M s , initial yielding in TRIP steels is controlled by transformation plasticity. An analogous study by Breedis and Robertson [35] on the temperature dependence of the critical resolved shear stress (CRSS) in an Fe-16Cr-12Ni alloy (C plus N ↘ 0.007 wt%) demonstrated that, below 200 → C, the austenite → martensite transformation initiates under tensile stresses, and the CRSS first increases and then decreases with decreasing temperature. At ↓32 → C, the dilation associated with the spontaneous transformation becomes so significant that it leads to an almost zero CRSS [35]. Similarly, Karaman et al. [34] observed that in carburized 4320 steels with ~15 % retained austenite, stress-assisted martensitic transformation occurs predominantly in tension, contributing up to an ~4 % volumetric expansion, which facilitates plasticity before significant slip activation. Their results indicate that higher retained austenite volume fractions reduce the flow stress and trigger earlier transformation, which is consistent with the higher CTYA of our solution-treated and aged martensitic specimens exhibiting a slightly higher retained austenite volume fraction. Both ferritic and martensitic matrices can accommodate stress-induced transformation of retained austenite, although the local stress state may differ between the two. Therefore, the observed differences in CTYA primarily reflect variations in the retained-austenite volume fraction rather than fundamental differences between the matrix phases. Secondary effects of the matrix phase and its texture may, however, influence CTYA and warrant further investigation.
In agreement with these findings, increased mechanical stability of retained austenite, achieved through solute atom partitioning (for example, in high-Ni martensitic steels [71]), was demonstrated to reduce yield asymmetry in TRIP-assisted steels by delaying the onset of stress-assisted transformation [34,39]. In the stress-relieved specimens, the annealing temperature of 300 → C is not sufficiently high, however, to enable significant partitioning of the substitutional solute atoms, such as Ni, to the retained austenite and thereby increasing its mechanical stability. Although interstitial carbon is highly mobile within the matrix at this temperature (D C α ↘ 1 ↑ 10 ↓12 m 2 /s [54]), the austenite stabilization effect by carbon partitioning appears to be small. We note that the multiple heating and cooling cycles during SLM and the gradual elevation of the overall build plate temperature during printing (up to 120 → C after ~19 h, Table 2) provide sufficient time and temperature for partitioning of fast diffusing carbon in the as-printed specimens. Consequently, given the small bulk carbon concentration (0.04 wt%), no significant differences in austenite carbon content are anticipated between as-printed and stress-relieved specimens. The same argument is relevant to the specimens aged at 482 → C.
Both the tensile and compression tests were performed perpendicular to the build direction (X-direction, Fig. S1) in this study. Directional anisotropy of the mechanical properties, that is, different yield strengths and ductilities along different directions, is anticipated for the highly textured as-printed and directly aged specimens, as demonstrated previously [72,73]. However, we do not anticipate substantial differences between the CTYA values measured along the build (Z) direction and those perpendicular to the build direction (X-or Y-directions). This conclusion is supported by a negligible difference in CTYA measured for a Cu-free variant of 17-4 PH stainless steel [74].
Despite the widespread use of 17-4 PH stainless steels in structural components at both ambient and elevated temperatures, the reliable performance of additively manufactured variants remains a significant challenge. These steels are extensively employed in aerospace, chemical processing, and nuclear power systems because of their desirable combination of mechanical strength, corrosion resistance, and oxidation resistance [31,32,[75][76][77]. The findings of this study demonstrate, however, that the as-printed ferritic specimens, while exhibiting a reasonable balance of ductility and strength at ambient temperature, may undergo severe embrittlement during prolonged service at elevated temperatures, as demonstrated by our direct aging experiments. For components exposed to hydrogen-containing environments, the embrittlement risk becomes even more pronounced [78,79]. This highlights the importance of full solution treatments and concurrent austenitization for dissolving pre-existing precipitates, homogenizing solute distributions, and, most importantly, creating a martensitic microstructure. This heat treatment, however, presents significant technological challenges. High-temperature exposure may induce distortion in components with complex geometries, as well as decarburization and surface oxidation, necessitating additional machining operations. Consequently, the need for costly post-processing undermines the advantage of AM, which aims to deliver near-net-shape components with minimal finishing requirements.
Reducing the sensitivity of phase selection to solidification conditions and minor compositional variations can substantially improve the predictability of microstructures, and thereby mechanical properties. Prior studies focused on powder chemistry and solidification conditions have shown that the as-printed microstructure is highly dependent on the chromium-to-nickel equivalent ratio (Cr eq /Ni eq ) [4]. By adjusting the composition, martensitic structures can be obtained directly in the as-built condition [20]. This composition-driven approach enables better control of phase fractions during rapid solidification. A recently developed variant, QT17-4≃ [6,7,80], aims to stabilize desirable phase fractions over a broader range of AM process parameters, reduce sensitivity to thermal history, and improve as-printed mechanical performance. Modifications in scan strategy, or grain refinement, which randomize or weaken texture in the as-printed microstructures, are anticipated to enhance ductility by enhancing plasticity.
The present study highlights another critical step towards designing a reliable AM stainless steel. The yield asymmetry as identified herein can be eliminated or reduced by stabilization of the retained austenite. Without stabilization, this asymmetry is anticipated to increase at lower temperatures, where diminished austenite stability accelerates a stressassisted transformation [36], posing significant risks for components intended for service at cryogenic or sub-ambient temperatures. Achieving robust microstructures directly during solidification remains a key pathway toward minimizing costly downstream treatments and enabling the reliable use of AM 17-4 PH components.
We examined the microstructures and mechanical properties of a selective laser melted (SLM'ed) 17-4 PH stainless steel for different heat treatment conditions. For all tested conditions, specimens exhibit a pronounced compressive-tensile yield asymmetry (CTYA), leading to significantly lower yield strengths under tensile loading. The following conclusions are drawn.
⇐ EBSD analyses, Fig. 3, indicate that SLM processing yields a predominantly ferritic microstructure (~80 vol%, Table 3), with columnar grains exhibiting a strong texture, ω001ε//build direction, Fig. 4a, and ~20 % fine equiaxed retained austenite. The following non-equilibrium solidification sequence is deduced: liquid → columnar δ-ferrite (b.c.c.) terminated by liquid → equiaxed γ-austenite (f.c.c.). A limited fraction of the γ-austenite phase transforms to martensite upon cooling to ambient temperature. ⇐ Direct aging at 482 → C without a prior solution treatment embrittles the specimens, Fig. 5, leading to cleavage-type fracture under tension with negligible plasticity, Fig. 7b. This is linked to Cu-rich nanoprecipitation hardening in a strongly textured ferritic matrix, elevating the ductile-to-brittle transition temperature and promoting {001} cleavage. ⇐ Solution treatment at 1000 → C followed by aging at 482 → C austenitizes the microstructure, producing lath martensite with a weaker texture upon quenching, resulting in better tensile properties (yield strength ~1210 MPa, elongation ~2 %) and a mixed ductile-brittle fracture mode, Fig. 7c.