In situ alloying of a CrMnNi steel via electron beam powder bed fusion process—microstructure and mechanical properties

In this study, a fine-grained and texture-free microstructure should be formed after the PBF-EB/M process by mixing steel powder X2CrMnNi16-7–3 with a low nickel content of 3 wt.% and steel powder X2CrMnNi16-7–9 with a high nickel content of 9 wt.%. The thermodynamic calculations for both precursor powders are presented in Supplementary Material (S1). The adjustment of the required primary-ferritic solidification as well as a high ferrite (\(\delta\)) → austenite (\(\upgamma\)) transition temperature of the alloy mixture should be achieved in analogy to [19]. Therefore, thermodynamic calculations for different mixing ratios of the precursor powders were performed with Thermo-Calc using the TCFE10 database.

As an initial step, equilibrium thermodynamic calculations were performed to determine the influence of the chemical composition on the \(\delta\)-ferrite content and, finally, the solidification behavior as well as the temperature at which the \(\delta\)→\(\gamma\) transformation is fully completed. The calculations revealed that mixing the powders in a weight proportion ratio of 4 parts X2CrMnNi 16–7-3: 1 part X2CrMnNi16-7–9 results in a \(\delta\)-ferrite content of 100% after completed solidification (see Fig. 1a, black arrow), which indicates a fully ferritic solidification. Additionally, at this mixing ratio, the transformation from ferrite to austenite is fully completed at 1152 °C (see Fig. 1a, blue dot) and thus far above the build temperature of 900 °C.

Bild vergrößern

However, the cooling rates reported in the literature for the PBF-EB/M process are around \({10}^{5}\) K/s, which is far from equilibrium solidification conditions. For this reason, additional Scheil–Gulliver calculations were performed to represent rapid cooling conditions (see Fig. 1b). Under these cooling conditions, the solidification starts as primary ferritic up to a solid fraction of approximately 85% at 1410 °C. Subsequently, the solidification continues either in the three-phase L + bcc + fcc region or the two-phase L + fcc region. Multiple changes in the solidification sequence occur, which can be attributed to the calculated local chemical composition at the solidification front and thus to segregation effects. Below a temperature of 1390 °C, the remaining solidification takes place in the three-phase L + bcc + fcc region. Therefore, it can be estimated that the fully ferritic solidification under equilibrium cooling conditions will be shifted to a primary ferritic solidification followed by a mixed solidification sequence under PBF-EB/M processing conditions.

It should be noted, that the phases M₂₃C₆, HCP and \(\sigma\), shown in Fig. 1a at temperatures below 750 °C, are considered as irrelevant since their formation starts below the build temperature (900 °C) of the PBF-EB/M process. The \(\alpha\)-ferrite phase transformation below 600 °C under equilibrium conditions indicates that the alloy exhibits a metastable austenitic microstructure when cooled at a sufficiently high rate to suppress further phase transformations and precipitations. This observation implies the possibility of a strain-induced martensitic transformation, i. e. the TRIP effect under tensile deformation.

The two feedstock powders X2CrMnNi16-7–3 and X2CrMnNi16-7–9 with a nominal particle size distribution of 50—150 µm have been produced by gas atomization using the Electrode Induction Melting Gas Atomization (EIGA) process under nitrogen gas atmosphere and were supplied by Eckart TLS GmbH (Bitterfeld, Germany). To achieve the desired chemical composition of the powder used in this study, powder X2CrMnNi16-7–3 and powder X2CrMnNi16-7–9 were mixed, according to the results from the thermodynamic calculations in a 4:1 ratio using a tumbler mixer for one hour at 60 rpm. The mixed powder, with a mean chemical composition of X2CrMnNi16-7–4.5, was subsequently sieved through a 150 µm mesh to remove larger residues and agglomerates from the manufacturing process that may have remained in the powder. In total, less than 20 g of material was removed from 80 kg powder mixture, therefore no significant effect on the chemical composition or the powder size distribution is expected.

The chemical composition of the feedstock powders were determined as described in [29]: C—combustion method (G4 Icarus, Bruker AXS); N – carrier gas method (G8 Galileo, Bruker AXS); Cr, Mn, Ni, Si, Al—optical emission spectrometry (Foundry Master UV, Hitachi). Afterwards, the powder particle size distribution was determined using the laser light scattering method, following the DIN 13320 standard, with a HORIBA LA-960 particle analyzer [29]. Furthermore, the chemical compositions as well as the \({d}_{10}\), \({d}_{50}\) and \({d}_{90}\) values of the particle size distributions from the feedstock powders and the mixed powder are shown in Table 1. It should be noted, that the chemical composition of the mixed powder was calculated from the measured chemical compositions of both precursor powders. The calculated mean composition was also used as nominal composition for the thermodynamic calculations given in Sect. 2.1.

Examination of the freshly mixed and sieved powder using SEM showed reasonably spherical particles containing a small number of satellites as shown in Fig. 2a. In Fig. 2b,c a representative metallographic cross-section of the mixed powder and the associated EDS area scan are presented, showing the nickel content for individual particles. It is clearly evident that the particles of the steel with 3 wt.% Ni predominate, while significantly fewer particles of the powder with 9 wt.% Ni are present. At higher magnifications, the particles revealed a relatively high amount of porosity in the individual powder particles of both powders (indicated by arrows in Fig. 2d).

Bild vergrößern

The chemical composition also affects the austenite stability and, therefore, the martensite start temperature \({M}_{\text{s}}\) as well as the \({M}_{\text{d}30}\) temperature, which describes the temperature at which 50% of strain-induced martensite was formed at an applied strain of 30%. Therefore, the \({M}_{\text{s}}\) temperature was determined according to Eq. 1 suggested by Eichelman [30], while the calculation of the \({M}_{\text{d}30}\) temperature was performed according to Eq. 2 of Nohara [31] (element concentrations in wt.%, respectively). This equation accounts for the influence of the grain size (according to ASTM E112) on the austenite stability.

Specimens for microstructural, chemical and mechanical investigations were manufactured on start plates made of stainless steel AISI 304L by PBF-EB/M using an Arcam A2X (Arcam EBM, Sweden) with machine software EBM control 4.2.201. Manufacturing of the parts was done at a process temperature of 900 °C. Preliminary tests on steel powder X2CrMnNi16-7–4.5 in several build jobs have shown that an energy input \({E}_{vol}\) between 24 and 30 J/mm3 leads to dense components. The energy input \({E}_{\text{vol}}\) was determined by means of well-known Eq. 3 with \(U\) the acceleration voltage, \(I\) the beam current, \(h\) the hatch distance, \(l\) the layer thickness and \(v\) the scan speed of the electron beam.

Melting of the whole specimen material was done using an acceleration voltage of 60 kV and a meandering scan strategy with a 90° rotation in scan direction for each layer. The focus offset (10 mA), the hatch distance (75 µm) and the layer thickness (50 µm) were constant for all build jobs.

For the microstructural investigations, including chemical analysis and measurements of ferromagnetic phase content as well as the preparation of tensile specimens, four cubes with an edge length of 23 mm were built in a minimized build unit on a 50 × 50 × 10 mm3 start plate, as shown in Fig. 3a. Two cubes with energy inputs of \({E}_{\text{vol}}\) = 24 J/mm³ (C24) and \({E}_{\text{vol}}\) = 30 J/mm³ (C30) were produced, respectively. According to the build parameters given in Table 2, the different energy inputs of the cubes were achieved by selecting scan speeds of \(\nu\) = 4000 and \(\nu\) = 5000 mm/s. Furthermore, two batches of cylindrical rods were built for fatigue testing (F24 and F30) in a 4 × 4 pattern on a 100 × 100 × 10 mm³ start plate in the standard build unit of the machine, as shown in Fig. 3b. The specimens with a diameter of 16 mm and a height of 68 mm were manufactured with the same melting parameters as cubes C24 and C30 (cf. Table 2), respectively. Additionally net-shaped fatigue specimens (FN30) with a gauge length of 14 mm and a gauge diameter of 6 mm were built (see Fig. 3c) with the same melting parameters as C30 (cf. Table 2). However, the contour of FN30 specimens was additionally melted with a lower scan speed of 2000 mm/s. After the build process, each component was powder blasted in a powder recovery system with powder of its own kind until the sintered powder was completely removed.

Bild vergrößern

It is known from previous studies that evaporation of manganese occurs during the PBF-EB/M process [9, 12, 19]. This effect is more pronounced, the higher the energy input is. For the determination of the manganese content after the PBF-EB/M process, one test specimen of the as-built material was taken from the center of cubes C24 and C30 (size: 23 × 23 × 5 mm³), respectively, as shown in Fig. 4a. The measurements were done by using optical emission spectrometry (OES) (Foundry Master UV, Hitachi). In addition, EDS measurements were conducted on tensile specimens C24 and C30 as well as fatigue specimens F24, F30 and FN30, respectively.

Bild vergrößern

Microstructural characterization of the steel after PBF-EB/M was conducted only in the as-built state. For cubes C24 and C30, one tensile specimen per cube was selected. The microstructural investigations were carried out on the central region of the respective tensile specimen, as illustrated by the red rectangle in Fig. 4c. In the case of fatigue specimens F24, F30 and FN30, the gauge lengths were removed and subsequently cut in half along the BD. The microstructural investigations were then carried out in the central region of the respective sections, as indicated by the red rectangle in Fig. 5.

Bild vergrößern

All specimens were ground under hot water (approximately 90 °C) to suppress deformation-induced α'-martensite formation during preparation. This process was followed by polishing and subsequent vibration polishing for 24 h using 0.02 µm colloidal silica to achieve a mirror-like finish. Microstructural investigations were performed using a high-resolution field-emission scanning electron microscope (MIRA 3 XMU, Tescan, Czech Republic), operated at an acceleration voltage of 15 kV. The SEM was equipped with detectors for secondary electron (SE) and backscattered electron (BSE) contrast. Energy-dispersive X-ray spectroscopy (EDS) was employed as an additional method to assess the local chemical composition and differences in manganese and nickel contents in the polished specimens. Electron backscattered diffraction (EBSD) was also conducted to determine the texture, phase compositions, and average grain size. The average grain size \(\overline{d }\) was calculated using the area-weighted diameter, applying a misorientation angle threshold of 12° for identifying high-angle grain boundaries. In addition, the aspect ratio of austenitic grains was determined as a quotient of the two axes \({a}_{\text{minor}}\) and \({a}_{\text{major}}\) (both in µm). Furthermore, SEM investigations were carried out to examine the fracture surfaces of both tensile and fatigue specimens in SE contrast, where non-metallic inclusions (NMIs) were analyzed using EDS.

The density of specimens C24 and C30 in the as-built state was measured using the Archimedes method, following the DIN EN 623–2 standard. For this, one miniature tensile specimen for each condition was utilized, respectively. Each specimen was measured five times in its dry condition (\({m}_{1}\)), its immersed state it water (\({m}_{2}\)) and its water-saturated state (\({m}_{3}\)), respectively. The bulk density \({\rho }_{\text{bulk}}\) was then calculated according to Eq. 4, where \({\rho }_{\text{water}}\) represents the density of water.

To study the mechanical properties of the as-built material under uniaxial, monotonic quasi-static tensile loading, flat miniature specimens with a rectangular cross section in the gauge section of 1.5 × 1.8 mm2 were manufactured via electro discharge machining (EDM) from the bulk material of the cubes C24 and C30. Figure 4 shows the specimen geometry and its orientation within the built cubes. Two specimens each with loading axis (LA) parallel and perpendicular to the BD were manufactured from the center of each cube. Afterwards, the surface layer affected by the EDM was removed by grinding the specimens under hot water (approximately 90 °C) to a thickness of 1.3 mm (removal of 150 µm per side).

The tensile tests were performed on a miniature load frame (Kammrath & Weiss GmbH, Germany) at room temperature with a nominal strain rate of \(4 x {10}^{-3}{s}^{-1}\). Strain measurements were conducted by a video extensometer using digital image correlation (DIC) (VEDDAC strain, Chemnitzer Werkstoffmechanik GmbH, Germany). The 0.2% offset yield strength (\({R}_{\text{p}0.2}\)), the ultimate tensile strength (\(UTS\)) and the elongation at fracture (\(A\)) were determined from the engineering stress–strain curves. To study the fatigue behavior under uniaxial loading conditions, fatigue specimens with a gauge length of 14 mm and a gauge diameter of 6 mm were used, see Fig. 5. Cylindrical rods of batches F24 and F30 were machined to obtain the desired geometry. In addition, the surface was mechanically polished to a finish of 1 µm.

The net-shaped specimens FN30 (see Fig. 3c), were built directly in the geometry shown in Fig. 5. Only the clamping areas were designed with a larger diameter of 16 mm to allow for subsequent machining to ensure a good grip of the clamping jaws. Strain-controlled low-cycle fatigue tests were carried out on a servohydraulic testing machine (MTS Landmark 250). The tests were performed in symmetrical tension–compression loading at \({R}_{\upvarepsilon }\) = -1 at total strain amplitudes of \({0.2 \%\le \Delta \varepsilon }_{\text{t}}/2 \le 0.8 \%\) with a constant strain rate of \(4 x {10}^{-3}{s}^{-1}\). The estimation of fatigue life as a function of total strain amplitude was evaluated according to the Basquin-Manson-Coffin relationship (see Eq. 5):where \({\sigma {\prime}}_{\text{f}}\) represents the fatigue strength coefficient, \(E\) the Young’s modulus, \(b\) the fatigue strength exponent, \({\varepsilon {\prime}}_{\text{f}}\) the fatigue ductility coefficient, and \(c\) the fatigue ductility exponent. The Young’s modulus was determined from the initial loading curves of the fatigue tests. Furthermore, the fatigue failure was determined either by a decrease in the stress amplitude to 90% of the maximum stress or final fracture of the specimen.

For measurements of the ferromagnetic phase content on cubes C24 and C30, a magnetic saturation device (MSAT) of the type Lake Shore 480 Fluxmeter was used. The specimens were magnetized by an external magnetic field, allowing the determination of the ferromagnetic phase fraction. According to [32], this method typically achieves an accuracy within ± 1 vol%. For this purpose, one section with a size of 23 × 23 × 3 mm3 of both C24 and C30 (cf. Figure 4a) was tested in the as-built state, respectively.

Measurements of the ferromagnetic phase content were performed on all tensile specimens before and after tensile testing using a Ferritscope® (Fischerscope® FMP30, Germany). Since the main deformation takes place in the gauge length, three measurement points were taken in this area before the tensile tests. After the tensile tests, measurements were conducted as close as possible to the fractured site on both specimen pieces. It should be noted, that the Ferritscope is calibrated for measurements of the \(\delta\)-ferrite phase of austenitic steels. Since the magnetic permeabilites of \(\delta\)-ferrite and \(\alpha {\prime}\)-martensite are different, the measured values were multiplied by a correction factor of 1.7 to obtain the true \(\alpha {\prime}\)-martensite content [33]. However, this correction factor is only valid up to a ferrogmagnetic phase content of 55 wt.%, since higher values can lead to a measurement deviation of up to ± 16%. During the fatigue tests, the evolution of martensite content of the batches F24 and F30 was also measured continuously by using a Ferritscope® (Fischerscope® MMS® PC, Germany).

The initial surface roughness values of specimens F30 and FN30 were determined by confocal microscopy (MarSurf CM Explorer, Mahr GmbH, Germany). The line roughness value \({R}_{\text{z}}\) (average maximum profile height) was measured along the BD for a minimum length of 9 mm. In addition, the area roughness parameters \({S}_{\text{z}}\) (single maximum peak-to-valley distance within the area) and \({S}_{\text{v}}\) (maximum depth of the deepest surface valley within the area) were measured in an area of at least 1.5 mm × 9 mm. All surface roughness values were determined according to ISO 25178.

A summary of the manganese contents in the as-built material, which were determined using OES and EDS, is presented in Table 3. The OES measurements revealed a reduction in manganese content in both specimens C24 (5.67 wt.%) and C30 (5.32 wt.%). Based on an initial Mn content of 7.03 wt.% in the powder, this corresponds to a Mn loss of 1.36 wt.% and 1.71 wt.%, respectively. The same trend was observed in the EDS analysis, wherby the measured values for C24 and C30 were even slightly lower (cf. Table 3). These variations can be attributed to the different sampling locations used for the respective analyses. The EDS measurements were conducted on tensile specimens extracted from the center of the cubes, where manganese evaporation is known to be more pronounced due to heat accumulation [19]. In contrast, the OES analysis was performed across the entire cross-section, leading to a more averaged chemical composition. Additionally, the manganese content in one fatigue specimen of each batch F24, F30 and FN30 was determined by EDS. The results show a similar trend, with Mn contents of 5.36 wt.% for F24 and 5.16 wt.% for F30, respectively. For specimen FN30, a relatively low Mn content of 4.76 wt.% was determined, which is the result of a higher Mn evaporation due to the smaller cross section compared to F30.

The \({M}_{\text{s}}\) and \({M}_{\text{d}30}\) temperatures were calculated according to Eq. 1 and Eq. 2 based on the initial powder composition listed in Table 1. However, due to the evaporation of manganese the values determined by OES and EDS analysis from the as-built material were set into Eq. 1 and Eq. 2 for Mn. In addition, the ASTM grain size number \(G\) was determined from EBSD data for each specimen in order to calculate the \({M}_{\text{d}30}\) temperature according to Eq. 2. The results are given in Table 3. To illustrate the approach, the calculation of \({M}_{\text{s}}\) and \({M}_{\text{d}30}\) temperatures is explained using the OES-derived Mn values for both specimens C24 and C30. It is evident, that the initially calculated \({M}_{\text{s}}\) temperature of − 65.0 °C for the mixed powder increases with increasing Mn loss, which is generally observable for all specimens. The \({M}_{\text{s}}\) temperature rises to –21.5 °C for specimen C24 and further increases to –8.0 °C in the case of specimen C30. For the calculation of \({M}_{\text{d}30}\), the grain size of 15 µm for C24 was obtained from the EBSD data (cf. Figure 7d). The grain size of C30, however, had to be estimated based on the EBSD results, since the grains exceed the image section (cf. Figure 8c-e). Therefore, the largest suitable ASTM grain size number \(G\) of 0 (corresponding to an average grain diameter \(\overline{d }\) > 359.2 µm) was selected. In comparison to the \({M}_{\text{d}30}\) temperature of 84.1 °C for powder particles, the values increase to 96.5 °C for specimen C24 and 113.0 °C for specimen C30, respectively. The trend of increasing \({M}_{\text{d}30}\) temperatures with decreasing Mn content and increasing grain size is consistently observed across all specimens analyzed by EDS, as shown in Table 3.

General results of SEM investigations on the initial microstructure of as-built cubes C24 and C30 are summarized in Fig. 6. It should be noted, that the presented microstructure results are representative for the tensile specimens given in Sect. 3.3. In Fig. 6a, the microstructure of C24 is presented in BSE contrast. The as-built state exhibits a predominantly uniform microstructure with small and nearly globular grains. However, numerous lack-of-fusion defects (indicated by yellow arrows) and gas porosities were distributed over the cross section. In addition, Fig. 6b shows the largest defect in the gauge length of the examined tensile specimen, identified as a NMI consisting primarily of Al and O by means of EDS. The determined density for the specimen was 7.6016 g/cm³. Since the specimens were fabricated from a mixture of two powders, the chemical homogeneity of the components had to be verified. For this purpose, high resolution EDS area scans (step size = 0.35 µm) were performed on both tensile specimens. Figure 6c shows localized enrichment in nickel content for specimen C24.

Bild vergrößern

Specimen C30 shows a significantly distorted microstructure at the surface (Fig. 6d). The grain morphology is not clearly visible which might be caused by the preparation-induced martensitic transformation due to evaporation of manganese, leading to a decrease in the austenite stability, as already described in [9]. The density of specimen C30 was 7.6050 g/cm3 and it also reveals smaller LOF defects in the gauge length (see Fig. 6e). In addition, a non-metallic inclusion is visible in Fig. 6e, as indicated by the arrow. Furthermore, the specimen predominantly contained only small gas pores. The nickel content was very homogeneous, with only minor nickel enrichments (indicated by the white arrow in Fig. 6f), thus verifying a homogeneous alloy composition.

Figure 7a shows a phase map of the predominantly austenitic microstructure of specimen C24 in the as-built state, with an area fraction of the fcc/bcc indexed pixels of 0.94/0.06. This finding roughly agrees with MSAT measurements, which revealed a ferromagnetic phase fraction of only 1.4 vol.%. The crystallographic orientation map of the austenitic phase (fcc) is shown in Fig. 7b. It should be noted that black areas in Fig. 7b are points, which were not indexed or belong to the bcc phase. As already evident from the inverse pole figure color-coded map, the material exhibited a weak texture which was confirmed by the texture index (TI) of 3.0 obtained from the texture calculations (see Fig. 7c). Thus, the microstructure exhibited a weak < 011 > texture parallel to the BD. Furthermore, an average grain size \(\overline{d }\) of about 15 µm ± 7 µm was evaluated using a misorientation angle of 12° for high angle grain boundaries. Thus, the microstructure can be described as fine-grained (see Fig. 7d). The grains were slightly elongated in BD causing an average aspect ratio of 0.49.

Bild vergrößern

Figure 8 shows the results of EBSD measurements obtained on cube C30. The phase map in Fig. 8a confirms that the microstructure is predominantly of bcc crystal lattice structure, which is in agreement with the assumption made from BSE imaging in Fig. 6d. Only a few regions exhibit a fcc crystal lattice structure, indicating a lower overall austenite stability in this material condition (see Sect. 3.1). While the area fraction of fcc/bcc indexed pixels was 0.02/0.98, the MSAT measurements revealed a ferromagnetic phase content of only approximately 2.9 vol.% in the bulk material. This, together with the evaporation of Mn and the increase in both \({M}_{\text{s}}\) and \({M}_{\text{d}30}\) (see Sect. 3.1), confirms that the observed high amount of bcc phase in the EBSD analysis is the consequence of a preparation-induced martensitic phase transformation which was even not suppressed by the hot-water preparation procedure at approximately T = 90 °C.

Bild vergrößern

In the band contrast map (Fig. 8b), small local areas with brighter contrast are visible (marked by red arrows) which correspond to regions with higher contrast values of Kikuchi bands compared to the lower contrast values (darker grey values) throughout the majority of the image. The contrast of the Kikuchi bands is related also to the defect density (dislocations and distortions) in the microstructure. Therefore, the non-transformed austenite with low dislocation density, in particular, appears bright. These fcc areas might be stabilized by a locally increased nickel content, as indicated in Fig. 6f. Additionally, smaller regions with bcc lattice structure also exhibit a bright contrast, which is likely attributed to primary solidified \(\delta\)-ferrite (exemplarily marked by the white arrows). All the other grains exhibiting a bcc lattice structure revealed lower band contrast (dark grey values), which is probably attributed to the higher dislocation density in the preparation-induced martensite. The band-like structure of bcc regions with different crystallographic orientations (e.g. see marked area in Fig. 8e) may also be taken as an indicator for the formation of preparation-induced martensite. The overall microstructure exhibited a columnar grain morphology with elongated grains oriented along the BD, as shown in Fig. 8c-e. This columnar structure is clearly distinguishable by comparison of all three specimen directions BD, ND and TD, highlighting the columnar grain growth during PBF-EB/M processing. Compared to the specimen manufactured with 24 J/mm3 (cf. Figure 7), cube C30 displayed a significant shift in crystallographic texture (see Fig. 8f). It should be noted that the texture was calculated for the bcc phase for this material condition. The weak < 112 > texture parallel to the BD is combined with a stronger < 111 > texture of the ND. In addition, the grain size in cube C30 was significantly larger than in cube C24. Due to the columnar growth, individual grains extended beyond the field of the EBSD scan area of several hundred µm edge length, preventing an accurate grain size determination within the examined image section.

To further evaluate the differences in the transformation behavior of specimens C24 and C30, a comparison of the Kernel Average Misorientation (KAM) from the EBSD data presented in Fig. 7 and Fig. 8 was conducted, as shown in Fig. 9. The fcc regions in Fig. 9a,c exhibited an average misorientation of 0.8 ± 0.3° (C24) and 0.9 ± 0.4° (C30), respectively. A comparison of the arrow-marked area in Fig. 9c with Fig. 6f and Fig. 8b (red arrows) reveals that this region of low misorientation also corresponds to an area with increased nickel content. This indicates a locally enhanced austenite stability, which prevented the martensitic phase transformation in this region, leading to a low KAM in the marked area. In contrast, the KAM values for the bcc phase in Fig. 9b,d are significantly higher, with values of 1.4 ± 0.5° (C24) and 1.5 ± 0.4° (C30), respectively. These elevated values are attributed to the martensitic phase transformation, which causes substantial lattice distortion.

Bild vergrößern

Figure 10 summarizes the results of combined EBSD and EDS measurements for fatigue specimen batches F24 (Fig. 10a-e), F30 (Fig. 10f-j) and FN30 (Fig. 10k-o). It should be noted that microstructure investigations were conducted on one specimen per batch in the as-built condition. The presented results are representative of the microstructure characteristic of each respective batch.

Bild vergrößern

The phase maps in Fig. 10a,f show a certain degree of austenite banding perpendicular to the BD. Based on the EDS maps in Fig. 10b,g, this banding can be attributed to locally increased nickel concentration due to insufficient mixing during the melting step, which led to a local stabilization of the austenitic phase. Therefore, no martensitic phase transformation occurred in these areas. Furthermore, a similar trend is observed as in the cubic specimens (cf. Figure 6): The specimen processed with a lower \({E}_{\text{vol}}\) of 24 J/mm3 displayed a more inhomogeneous distribution of nickel, whereas the specimen manufactured at 30 J/mm3 showed improved elemental mixing. For specimen FN30, however, the phase map in Fig. 10k shows no such banding, which is the result of the very homogeneous nickel distribution (see Fig. 10l) as a consequence of the small cross section and, therefore, the decreased beam return time. The white arrows in Fig. 10b,g indicate areas with a very low Ni content. These correspond to Al- and O-rich inclusions, with the respective EDS scans provided in the Supplementary Material (S2).

Overall, only a few regions exhibit a fcc structure, indicating a lower austenite stability. The crystallographic orientation map for the fcc phase in Fig. 10c and the corresponding texture calculations in Fig. 10e for specimen F24 show a fine-grained microstructure (\(\overline{d }\) = 8 µm ± 6 µm) with a weak < 011 > texture (TI = 3.4) in the BD, similar to specimen C24 (TI = 3.0, cf. Figure 7). For specimen F30, the microstructure is also fine-grained with \(\overline{d }\) = 12 µm ± 9 µm (see Fig. 10h), but the < 011 > TI in the BD increases to 5.2 (see Fig. 10j). The crystallographic orientation maps in Fig. 10d,i show a shift towards < 111 > for the bcc phase, which is confirmed by the bcc texture indices in Fig. 10e,j. Furthermore, the area fractions for fcc/bcc indexed pixels obtained from the phase maps in Fig. 10a,f are 0.44/0.56 for F24 and 0.52/0.48 for F30. For specimen FN30, the crystallographic orientation maps show a pronounced < 011 > TI of 10.4 for the fcc phase (see Fig. 10m,o) and a shift of the TI towards < 111 > in the bcc indexed regions (see Fig. 10n,o). In addition, the crystallographic orientation map in Fig. 10n reveals columnar grain growth with elongated grains along the BD, which is similar to the grain morphology of specimen C30 (cf. Figure 8). The area fraction for fcc/bcc indexed pixels is 0.24/0.76. As observed for specimen C30, the grains extended beyond the image section, preventing a reliable determination of their size, which was around several hundreds of micrometers and, therefore, definitely larger than for F24 and F30.

The influence of the energy input \({E}_{\text{vol}}\) during the PBF-EB/M process on the mechanical properties of the material was investigated by tensile tests of specimens taken parallel and perpendicular to the BD of cubes C24 and C30. Each material batch and load direction with respect to the BD was tested twice. The results are summarized in Fig. 11a, whereby only the engineering stress–strain curves with the highest elongation for each batch and loading direction are presented.

Bild vergrößern

In general, the two material batches C24 and C30 exhibit significantly different mechanical behavior under tensile loading, summarized in Table 4. The engineering stress vs. strain curves of cube C24 exhibited higher values for \({R}_{\text{p}0.2}\) (250 MPa) compared to specimens of cube C30 (200 MPa) for both loading orientations, respectively. This is caused by the different grain sizes: \(\overline{d }\) = 15 µm for batch C24 and several hundred of micrometers for batch C30. Comparing the UTS, specimens of C24 with BD ⊥ LA exhibited the highest UTS of 950 MPa. However, specimens manufactured with the same energy input and a LA ‖ BD showed significantly lower UTS of 850 MPa. Specimens of cube C30, displayed comparable average UTS values (around 890 MPa) both for LA ‖ BD and LA ⊥ BD, respectively.

Regarding the achieved strain values—uniform elongation and strain to failure—significant differences were observed both for the two batches C24 and C30 as well as for the different alignment of the LA with respect to the BD. Thus, the uniform elongation of batch C30 is significantly lower compared to C24 and no significant difference was observed regarding different alignments of the LA with respect to the BD. In the case of C24, a difference in uniform elongation and strain to failure was observed for BD ⊥ LA and BD ‖ LA. Whereas the specimen with BD ‖ LA failed immediately after reaching the UTS, the specimen with BD ⊥ LA exhibits a pronounced necking after passing UTS. Furthermore, a pronounced strain hardening is observed in both batches regardless the alignment of the LA with respect to the BD, resulting in a sigmoidal shape of all engineering stress vs. strain curves, visible in Fig. 11a. Figure 11b shows the strain hardening rate curves obtained from calculated true stress vs. true strain values. It is evident, that the strain hardening rate of C30 specimens is significantly higher and peaks at lower true strain values \({\epsilon }_{\text{true}}\) = 0.125 for C30 compared to \({\epsilon }_{\text{true}}\) = 0.2 for C24. This behavior is related to the strain-induced martensitic phase transformation from austenite into martensite causing the TRIP effect, which has been previously reported for similar steels [4, 8, 9].

The occurrence of the strain-induced martensitic phase transformation during the quasi-static tensile tests was further confirmed by measurements of the ferromagnetic phase fraction conducted on the specimens prior and after tensile testing, as presented in Table 4. It shows the significant increase in ferromagnetic phase fraction after tensile testing of batches C24 and C30, although batch C30 revealed slightly higher values prior to mechanical testing.

The fracture surfaces of all tested tensile specimens were analyzed by SEM. Figure 12 shows a comparison of selected fracture surfaces of the specimens from batches C24 and C30. First, it is evident that the C24 specimen with the LA parallel to the BD (Fig. 12a) maintained an almost square cross-section after tensile testing, whereas the C30 specimen (Fig. 12d) displayed significant necking (cf. Figure 11), indicating a higher ductility. The fracture surfaces of both specimens primarily revealed ductile fracture features characterized by dimples (Fig. 12a-e) caused by the coalescence of microvoids during mechanical loading. Moreover, the fracture surface of specimen C24 contained a high number of LOF defects. This is also reflected in the slightly lower density for C24 in comparison to C30. In the areas of LOFs (Fig. 12a-c), but also in the center of some dimples, non-metallic inclusions were found identified via EDS analysis as Al-O-containing non-metallic inclusions [34]. In the C24 specimen, several irregularly shaped non-metallic inclusions were visible, the largest located in the top left corner of Fig. 12a with a diameter of approximately 80 µm. Apart from occasional inclusions, the largest remaining inhomogeneities are gas pores (marked by red arrows in Fig. 12e), which, as previously mentioned in Sect. 2.2, most likely originate from the powder feedstock. In addition, Fig. 12f,g shows the fracture surfaces of specimens from batches C24 and C30 with the LA perpendicular to the BD. Again, the C24 specimen reveals numerous LOF defects, as indicated by white arrows in Fig. 12f, which is in agreement with the observations in Fig. 12a. For the C30 condition, the largest voids were gas pores which were distributed across the whole fracture surface (marked by red arrows).

Bild vergrößern

The results of the cyclic loading experiments are summarized in Fig. 13, whereby the batch with 24 J/mm3 (F24) was only tested in machined condition (Fig. 13a) and the batch with 30 J/mm³ in machined (F30, Fig. 13c) and net-shaped (FN30, Fig. 13e) conditions. All batches were tested at various total strain amplitudes. The evolution of the fatigue-induced martensitic volume fraction as a function of the number of cycles is given for batches F24 and F30 in machined condition in Fig. 13b,d. Due to the high surface roughness, the martensitic phase evolution of specimens FN30 in net-shape condition was not determined. In general, the cyclic deformation curves of all tested conditions reveal similar trends in the evolution of the stress amplitude as a function of the number of cycles. A cyclic softening throughout the entire fatigue live was observed for the lowest total strain amplitude of \({\Delta \varepsilon }_{\text{t}}/2\) = 0.2%.

Bild vergrößern

An increase in the total strain amplitude causes, in general, both an increase in the initial stress amplitude as well as a secondary hardening. The onset of the secondary hardening is shifted to lower number of cycles and becomes more pronounced as higher the total strain amplitude is (see Fig. 13a,c,e). The secondary hardening is the consequence of a fatigue-induced martensitic phase transformation. With an increase in total strain amplitude also the fraction of plastic strain increases, causing a faster accumulation of the plastic strain needed for triggering the phase transformation, and, therefore, the incubation time is shortened. For medium strain amplitudes of \({0.3 \% \le \Delta \varepsilon }_{\text{t}}/2\) \(\le 0.5\text{ \%}\), a primary increase in stress amplitude is observed, which tends to saturate in some cases after a short period. This phenomenon has already been described in the literature and is attributed to an increase in dislocation density during the first cycles, followed by rearrangement and annihilation processes [4, 35, 36].

Significant differences between batches F24 and F30 in contrast to batch FN30, comparing identical total strain amplitudes, are (i) a higher initial stress amplitude, (ii) lower maximum stress amplitude, and (iii) lower number of cycles to failure for F24 and F30. The higher initial stress amplitudes of F24 and F30 are the consequence of the lower grain size, as illustrated in Fig. 10, whereas the lower number of cycles to failure in batch F24 was most probably caused by the high amount of LOFs. As a consequence of the lower fatigue life, the amount of fatigue-induced martensite and, therefore, also the strain hardening and the final maximum stress amplitude were lower. The evolution of the volume fraction of fatigue-induced \(\alpha {\prime}\)-martensite as a function of the number of cycles and in dependence on the total strain amplitude is shown for batches F24 and F30 in machined condition in Fig. 13b,d, respectively. It is obvious that the incubation time for the fatigue-induced martensitic phase transformation, which causes the cyclic strengthening [37, 38], is shortened and the volume fraction of \(\alpha {\prime}\)-martensite is increased significantly with an increase in the total strain amplitude. It should be noted, that the minor irregularities for total strain amplitudes of 0.3% and 0.6% in Fig. 13d are most likely caused by a slip of the Ferritscope®. Furthermore, the lower the strain amplitudes are, the more pronounced is the saturation behavior in martensite formation after a certain point towards failure. This is because the driving force for martensite formation is no longer sufficient after a certain cumulative plastic deformation at lower strain amplitudes due to decreasing plastic strain amplitudes. As the deformation progresses, the materials ability to transform further into martensite diminishes, leading to the observed plateau in the \(\alpha {\prime}\)-martensite content [37]. It is also evident that the F30 specimens exhibit higher martensite contents than the F24 batch, as they withstand a higher number of cycles.

The calculated fatigue lives according to Basquin-Manson-Coffin relationship for the three tested batches are shown in Fig. 13f. The fatigue lives of F30 and FN30 are very close to each other showing, however, some scatter. In contrast, the fatigue lives of batch 24 are significantly lower, which is caused by the higher amount of LOFs. The similar fatigue lives of batches F30 and FN30 with different surface conditions are particularly surprising. Since failure during LCF testing initiates in most cases from the surface, especially with decreasing strain amplitudes, mechanical machining and polishing of the surface were performed for batch F30 in an attempt to enhance fatigue life. Nevertheless, no significant difference in fatigue life was observed despite the significantly different surface roughness conditions.

To illustrate the differences between the surface conditions, the respective gauge lengths of specimens F30 and FN30 are shown in Fig. 14a,b prior to fatigue testing. The values given in Table 5 were calculated as the arithmetic mean of three single measurements. For the net-shaped specimens, the roughness results are in agreement with values reported in the literature for \({R}_{\text{z}}\) [39, 40] and for area roughness parameters \({S}_{\text{z}}\) and \({S}_{\text{v}}\) [41, 42]. It is evident that the roughness values for the FN30 condition are significantly higher than those for the F30 condition, which usually results in decreased fatigue life. It should be noted, that the relatively high standard deviation for F30 can be explained by several pores (valleys) and inclusions (peaks) located on the surface.

Bild vergrößern

At higher magnification, the F30 specimen revealed several clusters of NMIs (yellow arrows) at the surface, as illustrated in Fig. 14c. Following the fatigue tests, the specimen’s surface was re-examined and crack initiation was observed at the NMIs (see Fig. 14d), which supports the finding from the fracture surfaces. In addition, Fig. 14e shows the metallographic cross section of a net-shaped FN30 specimen in the as-built condition. Here, notches with a depth of up to 300 µm were detected, indicated by yellow arrows. This observation can be directly correlated with Fig. 15g, where a smooth structure without sufficient material fusion was also identified directly at the surface. It should be noted that subsurface defects cannot be detected by roughness measurement methods, resulting in underestimated effective surface valley values [43].

Bild vergrößern

The unusual fatigue life behavior was further investigated by analysis of the fracture surfaces of all three batches, which are displayed in Fig. 15. The machined specimens F24 and F30 in Fig. 15a,c, respectively, reveal a high number of NMIs distributed over the whole cross section, similar to the fracture surfaces of C24 tensile specimens with BD || LA (see Fig. 12). For specimen F24, multiple crack initiation sites were identified along the surface area, indicated by black arrows. Figure 15b shows one selected crack initiation site containing NMIs area in the near surface area of a machined F24 specimen with higher magnification. It is well known, that NMIs are detrimental for fatigue properties, since they act as stress concentration sites [44, 45].

Upon higher magnification, the fracture surface of a machined F30 specimen, tested with a total strain amplitude of 0.8%, in Fig. 15c reveals a relatively smooth structure covered with NMIs. It becomes evident, that this region was not fully fused, which is a common feature in many powder bed-based additive manufacturing processes. Such incompletely melted regions and inclusions serve as preferred sites for crack initiation due to stress concentration. In addition, Fig. 15d,e show EDS maps of the NMI cluster located near the surface of the specimen. It was found that these inclusions mainly consist of aluminum and oxygen, most probably Al₂O₃. These findings indicate that mechanical processing of the specimen surface cannot be effective, as the defects are distributed throughout the entire volume of the material. During the specimen preparation, the initially fabricated cylinders were turned down from a diameter of 16 mm to 6 mm in the gauge length area. This means that there is a high probability that a larger defect will always be located near the surface, ultimately leading to crack initiation. Therefore, machined specimens have not improved lifetimes compared to specimens with as-built surface, as already demonstrated in Fig. 13e.

An overview of the fracture surface of a net-shaped FN30 specimen, however, does not show as many NMIs as for the machined batches. Once again, multiple crack initiation sites can be identified in the (sub)surface area, indicated by white arrows (see Fig. 15f). Upon higher magnification, Fig. 15g displays the differences between machined and net-shaped condition: First, multiple unmelted powder particles are revealed in the outermost region of the surface. In addition, this area is very smooth, which indicates an insufficient melting as previously mentioned for Fig. 15c. These surface areas can be attributed to the surface roughness of the specimens in net-shape condition and are just as detrimental for the fatigue life as NMIs and LOF defects, as shown in Fig. 13f.

The approach of developing an alloy by mixing two precursor powders with different nickel content to obtain a fine-grained microstructure after the PBF-EB/M process was investigated.

While a certain inhomogeneity in the distribution of nickel was observed for cubic specimen C24, increasing the energy input to \({E}_{\text{vol}}\) = 30 J/mm³ for C30 resulted in a very homogeneous Ni distribution. This can be attributed to the larger and more sustained melt pool, which enhances mixing [46‐49]. In addition, studies in this field confirm that Marangoni convection is stronger for such melt pools, further increasing the chemical homogeneity [50]. The same effect was observed for the fatigue specimens F24, F30 and FN30, where an increase of \({E}_{\text{vol}}\) led to a more homogeneous Ni distribution.

Furthermore, the specimen C24 is characterized by an austenitic (ferromagnetic phase content of 1.4 vol.%) and fine-grained microstructure (average grain size of \(\overline{d }\) = 15 µm), supporting the effectiveness of the selected alloy design approach. The chemical composition of the powder mixture met all the criteria outlined by Burkhardt et al. [19]: (1) a primary ferritic solidification of the melt, (2) the completion of the bcc → fcc solid phase transformation upon cooling to build temperature, (3) during reheating of the material upon the deposition of subsequent layers, the solid-state transformation from fcc to bcc is suppressed kinetically. The thermodynamic calculations under equilibrium conditions (cf. Figure 1a) served as a reliable indicator for comparing the solidification and transformation behavior of the steel variants with each other in previous studies [19]. This was further confirmed by the Scheil solidification simulation. While the latter does not predict a fully ferritic solidification path, the simulation still indicates primary ferritic solidification of the melt up to a high solid fraction of approximately 85%.

In contrast to C24, specimen C30 was manufactured with an increased energy input of \({E}_{\text{vol}}\) = 30 J/mm³ by using a lower scan speed of the beam. This causes a significant evaporation of manganese as shown by Günther et al. on steel X2CrMnNi16-6–6 [8, 9] and Burkhardt et al. in [19], as well as higher temperatures in the component. Burkhardt et al. [19] demonstrated that both factors have a significant impact on the resulting microstructure. Due to the lower scan speed, the electron beam remains a longer time in a certain area, resulting in reduced heat dissipation. Consequently, higher temperatures are achieved in the process. As described by Burkhardt et al. [19], the higher temperature within the component can inhibit the complete bcc → fcc solid phase transformation up to the building temperature, causing a part of the material to retain its ferritic structure. In addition, it is suggested in [19] that manganese evaporation might lead to an enlargement of the two-phase region bcc + fcc, thereby hindering the complete transformation from bcc to fcc during cooling down to build temperature. This also results presumably to a premature transformation from fcc to bcc upon reheating of the material during building of subsequent layers. Consequently, ferritic grains from previous build layers can grow epitaxially into the melt with primary ferritic solidification leading to columnar grain growth over multiple layers. According to the phase diagram shown in Fig. 16, it can be observed globally that a lower manganese content shifts the complete transformation to austenite to lower temperatures. However, manganese evaporation does not significantly alter the size of the two-phase field for the determined Mn content in the present case of specimens C24 and C30 (see Fig. 16). Therefore, it can be assumed that the increased energy input and finally the higher temperatures are the main reasons for columnar grain growth in cube C30.

Bild vergrößern

Upon cooling to room temperature, the ferrite fully transforms into austenite, but the columnar grain structure remains, cf. Figure 8. This obviously led to a change from the fine-grained microstructure observed in C24 to a columnar microstructure in C30 with grains elongated in the BD with a length of more than 400 µm. Nevertheless, the bulk of C30 was still overwhelmingly austenitic, which was confirmed by a ferromagnetic phase content of only 2.9 vol.%.

This transformation behavior and the microstructure observed in the EBSD analysis can be directly attributed to several interrelated factors. While the increased energy input led to a more homogeneous nickel distribution, it also caused enhanced manganese evaporation. Consequently, specimens manufactured with an \({E}_{\text{vol}}\) of 30 J/mm³ experienced a more pronounced Mn loss compared to specimens of identical geometry produced with a lower \({E}_{\text{vol}}\) of 24 J/mm³. As previously discussed, the higher energy input was achieved by reducing the electron beam scan speed, which in turn decreased heat dissipation and promoted columnar grain growth. As a result, the grains became coarser but still retained the austenitic phase at room temperature. However, due to their larger size, these columnar grains exhibited lower austenite stability, as described by Nohara et al. [31]. In summary, specimens with elongated grains and reduced Mn content (C30 and FN30) were more susceptible to martensitic phase transformation during metallographic preparation. In the case of specimen F30, the material showed a lower Mn evaporation and a small grain size, partially preventing the phase transformation during the preparation process.

In addition, KAM analyses were performed to investigate potential differences in the average misorientation between fine-grained microstructures and columnar grains. No significant differences were observed within the \(fcc\) phase between the two specimens C24 (0.8°) and C30 (0.9°). Similarly, the \(bcc\) phases of both specimens showed comparable misorientation values (1.4° and 1.5°, respectively). However, the overall misorientation within the \(bcc\) phase was notably higher than in the \(fcc\) phase, which can be attributed to lattice distortions caused by the martensitic phase transformation occurring during specimen preparation.

Besides the grain size and grain morphology, specimen C24 displayed a slight < 011 > texture with a TI of 3.0 in the BD. In contrast, alloys processed in powder bed AM processes often show a pronounced < 001 > texture in BD [51], which was also observed for the austenitic steels AISI 316L [52‐55] and AISI 304L [19]. Furthermore, the steel X2CrMnNi16-7–9 also displayed a very pronounced < 001 > texture and columnar grain growth in the BD with a TI of up to 14.5 [12]. The similar steel X5CrMnNi16-6–6, however, showed very weak texture indices with values of less than 2.0 [2‐4, 8, 9]. Nonetheless, the presence of a < 011 > texture in the BD, as observed for X2CrMnNi16-7–4.5, is not uncommon in powder bed AM processes and has been reported in literature for various alloys [56‐59]. However, the precise mechanism leading to this specific texture is currently under investigation, since the X2CrMnNi16-7–3 has not been processed in PBF-EB/M so far. Furthermore, a shift in crystallographic texture for specimen C30 from fcc < 011 > towards the bcc < 111 > direction can be observed. As previously explained, this is only a martensitic phase transformation at the surface of the specimens, which was caused by the metallographic preparation, although grinding was done with hot water.

The transition from a fine-grained to a columnar microstructure as a result of changes in energy input is less pronounced in the cylindrical specimens, as both F24 and F30 exhibit a fine-grained microstructure. For the fatigue specimens of batch F24, a grain size of 8 µm and a TI of 3.6 for < 011 > in the BD were observed. An increase of the \({E}_{\text{vol}}\) to 30 J/mm³ for specimens F30 resulted only in a small increase in average grain size to 12 µm, while the TI significantly increased to 5.2. Obviously, the heat dissipation in the cylindrical specimens with a scanned diameter of 16 mm is significantly more efficient than in the cubic specimens C24 and C30, which had a square cross-section with an edge length of 23 mm. The improved heat dissipation is also evident from the EDS measurements for Mn presented in Table 1. The difference between F24 (5.38 wt.% Mn) and F30 (5.16 wt.% Mn) is very small, whereas significantly larger local variations in Mn content were observed in the cubes C24 (5.36 wt.%) and C30 (4.26 wt.%). As a result, the transformation from ferrite to austenite upon cooling to build temperature can be completed sufficiently fast in both F24 and F30 despite the increase of the volumetric energy density from \({E}_{\text{vol}}\) = 24 J/mm³ to \({E}_{\text{vol}}\) = 30 J/mm³. Due to their smaller grain size and the limited Mn loss, both material conditions partially remained their austenitic microstructure after metallographic preparation.

The net-shaped specimens FN30 with a significantly smaller diameter in the gauge length of 6 mm exhibited columnar grain growth, preventing an accurate determination of the grain size, which is, however, significantly larger than for F24 and F30. This is probably caused by the return time of the beam, which describes the time it takes for the electron beam to travel from the middle of one fusion line to the middle of an adjacent one [60, 61]. Because of the small cross section of the gauge length, the return time of the beam was very low, resulting in a pronounced heat accumulation in the component. Therefore, the transformation from \(\delta\) to \(\gamma\) was not completed at build temperature, since the transformation temperature was shifted to lower values. In addition, the reduced heat dissipation resulted in a stronger Mn evaporation for FN30 (4.76 wt.% Mn) in comparison to F24 (5.38 wt.% Mn) and F30 (5.16 wt.% Mn). In combination with the large grain size, this led to the pronounced formation of preparation-induced martensite.

The stress–strain curves of the tensile tests (Fig. 11) show that miniature tensile specimens produced from cube C24 exhibit a significantly higher \(\text{YS}\) compared to those from cube C30. This can be explained by the columnar microstructure in C30 after the PBF-EB/M process, resulting in a larger average grain size (compare Fig. 6 and 7, respectively). According to the Hall–Petch effect, this increased grain size resulted in a reduced \(\text{YS}\) for the C30 specimens compared to C24. Although the grain size for C30 could not be directly determined, it is obvious that individual grains exceeded a length of 400 µm in BD, which was significantly higher than for C24 specimens. Another reason for the lower \(\text{YS}\) might be the presence of a larger volume fraction of \(\delta\)-ferrite in the C30 specimens (between 5 to 6 vol.%), as confirmed by EBSD (cf. Figure 8) and MSAT (cf. Table 4) measurements. In the softer ferrite, initial plastic slip processes can occur at relatively low stress levels before the macroscopic yield point is reached, which is known as micro-yielding [62].

After passing the \(YS\), the material exhibited a certain amount of strain hardening followed by the start of martensitic phase transformation at the inflection point of the engineering stress–strain curve, correlated to the TRIP effect. However, the strain hardening rate observed in the 30 J/mm³ tensile specimens was higher than in the 24 J/mm³ specimens (see Fig. 11b). As previously stated, this phenomenon can be explained by the more pronounced manganese evaporation in combination with columnar grain growth. These factors led to a decrease in austenite stability and, therefore, a higher tendency for the martensitic phase transformation under loading. The lower austenite stability is also displayed by the increasing values for \({M}_{\text{s}}\) and \({M}_{\text{d}30}\), resulting in a significant increase in the hardening rate of C30 with lower Mn content [14, 15].

Due to the lower austenite stability of C30, a higher \(\alpha {\prime}\)-martensite content was achieved at the same strain values compared to C24. In the case BD ⊥ LA, the elongation to fracture of C30 was reduced in comparison to C24, since a significant portion of the martensite forming capability in C30 has already been depleted at lower strain levels (see Fig. 11). In contrast, C24 still exhibited the potential for \(\gamma\) → \(\alpha {\prime}\) phase transformation under loading at higher strains, due to the higher austenite stability. Generally, for specimens with BD ⊥ LA, higher \(UTS\) and elongation to fracture were observed. Although the strain hardening behavior in the case BD || LA of both batches C24 and C30 remained similar to the BD ⊥ LA condition, they showed a decrease in elongation to fracture. Especially C24 with BD || LA exhibited premature failure without reaching the maximum strength potential, leading to fracture without any necking, as illustrated in Fig. 12a. This particular fracture surface showed a high number of NMIs and LOF defects in their vicinity. However, such a defect density was not observed for the C30 specimen in Fig. 12d. It is, therefore, assumed that the higher energy input in specimen C30 also melted areas in the vicinity of NMIs more sufficiently. It is known that such NMIs can lead to defects during AM [63] since they generally act as a heat shield in powder bed additive manufacturing, as reported in [64]. In electron beam processes, NMIs can additionally accumulate charge due to their low electrical conductivity, potentially leading to electron beam deflection and improper melting of the surrounding material [64‐66]. The positive effect of an increased energy input seems reasonable, since it is well established that the width and depth of a melt pool increase in proportion to the energy input [67], resulting in enhanced density and mechanical properties to a certain extent. As a consequence of proper melting, the amount of NMI-induced LOF areas, which act as an easy crack nucleation and a fast path for crack propagation, diminish. Consequently, the mechanical properties for both BD ⊥ LA and BD ‖ LA specimens converge for C30 (see Fig. 11a). In conclusion, the high number of NMIs can only be attributed to the initial powders. Since the X2CrMnNi16-7–9 powder was already thoroughly investigated [2, 12] and NMIs of such quantity were never observed, the impurities must originate from the X2CrMnNi16-7–3 powder. In addition, the powder was not further processed after atomization. Therefore, the impurities must originate from the raw material. Most likely these NMIs were introduced during the casting of the electrodes for the EIGA process.

In the following, the tensile properties of the new material will be compared to other CrMnNi steels that were also produced using PBF-EB/M. The steel X2CrMnNi16-7–4.5 achieved a maximum \(UTS\) of up to 950 MPa, which was significantly higher than for PBF-EB/M processed steel X2CrMnNi16-7–9 with an \(UTS\) of 550 MPa [12] and steel X5CrMnNi16-6–6 with an \(UTS\) of 840 MPa [4, 8]. However, the elongation to fracture of steel X2CrMnNi 16–7-4.5 is with a maximum of \(A\) = 50% significantly lower than for steel X2CrMnNi16-7–9 [12], which in some cases reached up to \(A\) = 110% as a result of the occurring TWIP effect under loading. Furthermore, steel X5CrMnNi16-6–6 showed a higher elongation to fracture of up to \(A\) = 70% as well [5, 8]. In comparison to conventionally cast material, the C24 tensile specimens in the BD ⊥ LA loading case almost reached the \(UTS\) values for as-cast X2CrMnNi16-6–3 (984 MPa) while also achieving a higher elongation to fracture (26% for as-cast X2CrMnNi16-6–3) [7]. Simultaneously, the steel X2CrMnNi 16–7-4.5 processed by PBF-EB/M exceeded the \(UTS\) of an as-cast X5CrMnNi16-6–6 (792 MPa) and achieved slightly higher elongation to fracture values (up to 45%) [7].

For fatigue specimens manufactured with a lower energy input of 24 J/mm³ (F24), the fine-grained microstructure, combined with a significantly higher yield strength, had no positive effect on the fatigue life due to the extensive presence of NMIs and LOF defects. Due to the small average grain size, specimen batches F24 and F30 exhibited higher initial stress amplitudes than the FN30 condition with larger, columnar grains. An increase in energy input to 30 J/mm³ (F30 and FN30) led to an improved melting for fatigue specimens, resulting in a longer fatigue life. This agrees with the improved \(UTS\) and elongation to fracture for tensile specimens with a loading direction parallel to the BD and built with an increased \({E}_{\text{vol}}\) of 30 J/mm³. In analogy to the tensile tests, it was observed that the F30 batch showed a higher tendency for martensite formation in comparison to the F24 specimens. This can be explained due to the lower austenite stability in F30 as a result of higher manganese evaporation during PBF-EB/M. Surprisingly, surface machining did not show any improvement in the lifetime under cyclic loading for the 30 J/mm³ specimens. This can, however, be explained by the high number of large NMIs distributed throughout the entire specimen volume (see Fig. 15). As a result, even after machining, impurities remain at or near the surface, which acted as stress concentration sites, as shown in Fig. 15c,d [44, 45].