Entropy-driven densification of medium-entropy alloys with laser powder bed fusion Authors: Zairan Luo, Qian Liu, Jiang Yi, Boyang Wu, Zhiqian Rao, Mujin Yang, Shi Shi, Shuai Wang Published: 2026 ABSTRACT AbstractAnomalies inherent to laser-based additive manufacturing (AM), including porosity and structural defects, present significant barriers to fabricating high-integrity metallic structural components. These unavoidable anomalies can significantly degrade the long-term properties of metals, including fatigue, creep, and corrosion, thereby limiting their use in critical environments. In this study, we overcome this challenge with a single-step entropy engineering strategy. By integrating CoCrNi and 316 L powders and enabling temporal changes in entropy evolution during printing, we obtained a nearly void-free (density > 99.99%) medium-entropy alloy with exceptional fatigue resistance and corrosion performance, surpassing conventional PBF-LB/Med medium- and high-entropy alloys. By combining macro-scale densification analysis, micro-scale SEM/EBSD observations, and nano-scale TEM investigations, our results revealed that the increase in configurational entropy during the AM process, together with changes in microstructure, can modify the thermal history, thereby optimizing melt pool spreading. This modification promoted the formation of a uniformly distributed, subgrain-predominated microstructure with a relatively low geometrically necessary dislocation density, which is relatively stable under external loading. This work presents a single-step approach to overcoming the long-term property limitations of AM-fabricated components.Subject terms: Metals and alloys, Laser material processing Introduction Powder bed fusion of metal with a laser beam (PBF-LB/M) is a state-of-the-art additive manufacturing (AM) methodology for metallic materials, characterized by its precise, layer-wise fabrication process that achieves micron-scale dimensional tolerances1,2. This laser-based technique has shown remarkable potential in aerospace, biomedical, and high-end equipment applications due to its unique capabilities for fabricating complex geometries, lightweight structures, and customized components3,4. Anomalies ranging from tens to thousands of microns, such as porosity, lack of fusion (LOF), and balling, are commonly observed in PBF-LB/M-fabricated metallic materials5,6. Interestingly, these defects do not necessarily impact the immediate quasi-static mechanical properties, such as strength and ductility. Numerous studies have demonstrated that many as-printed metals can overcome traditional strength-ductility trade-offs7,8. This has fueled optimism around the potential of PBF-LB/M technology to meet practical mechanical requirements. However, these benefits are less evident in the long-term application. During prolonged service, these anomalies may trigger localized stress concentrations, crack initiation, and degradation of mechanical properties9,10, progressively compromising fatigue resistance, corrosion resistance, and creep characteristics. These factors pose a critical limitation to the scalability and reliability of PBF-LB/M technology for industrial applications that require high durability11,12. Significant efforts have been dedicated to mitigating these defects and enhancing the structural integrity of PBF-LB/M-fabricated components. External field-assisted densification is probably the most straightforward way to produce dense parts, such as hot isostatic pressing13, layer-by-layer rolling14, magnetic field-assisted densification15, and ultrasonic vibration-assisted densification16. However, many of these approaches suffer from limited sample sizes and add complexity to the 3D printing process. It is worth noting that some studies have shown that micro-sized pores are difficult to remove after their formation and can progressively reappear and grow during isostatic heat treatment, due to the internal argon gas pressure within the as-printed parts17. Therefore, it is crucial to eliminate the source of defects during the printing process. Many researchers have attempted to address this problem by adjusting processing parameters to search for a high-density-oriented processing window18,19. In addition to the melting and solidification, the gravitational effects of powder spreading are also believed to play a significant role in void formation20. Many studies also used the laser rescanning method to increase the relative density21. This method can increase the density of 316 L stainless steel by 0.21–0.42% (with the highest density reaching 99.96%), and also modify the characteristic size of solidification microstructures. Some researchers densify the as-printed part by modifying the powders. A typical case is that plasma processing of tungsten powders can enhance their laser absorption rate and thermal diffusion, resulting in a high density of 98.7% at a relatively low preheating temperature22. Adding grain refiners, such as LaB6 particles23, can also result in a crack-free alloy, and it is suggested that the refined grains have a high capacity to reduce thermal-stress-induced cracking. Although these efforts improve the immediate tensile and compressive properties, residual voids and inclusions still prevent as-printed metals from reaching densities comparable to those of forged or rolled counterparts, and they act as stress concentrators that accelerate fatigue crack initiation and propagation24,25. Removing these anomalies and achieving the highest density in as-printed metal remains a significant challenge. Recently, it has been proposed that the voids formed in PBF-LB/M can be eliminated by controlling the asynchronism between phase transformation and grain growth26, and a near-void-free AM technique for Ti-alloy27 has been developed, significantly enhancing its fatigue performance. The optimized alloy reached a fatigue limit of approximately 1 GPa, outperforming all previously reported AM and forged Ti-alloys. However, for alloys that lack phase transformation variations during the PBF-LB/M process, such as high- and medium-entropy alloys (MEAs), there remains a lack of long-term, property-oriented strategies to eliminate anomalies (i.e., densification). Given the compositional complexity and sluggish diffusion characteristics of MEAs, modulating elemental ratios by mixing them with other alloy powders may offer a route to densification. It has been reported that blending metal powders can induce in situ partial homogenization of two dissimilar alloy melts, thereby controlling the local concentration and phases28. Other studies have also shown that optimizing powder blends in PBF-LB/M in-situ alloying of HEAs improves powder bed density, homogeneity, and microstructure refinement29. Moreover, introducing composition-gradient layers between dissimilar metals can suppress interfacial defects and enhance bonding strength, as demonstrated by an AlSi40 alloy made from mixed Al and Si powders, which showed a 5% lower coefficient of thermal expansion than samples from pre-alloyed AlSi40 powder30. Modeling of mixed Ni–steel powder systems further reveals that the melt pool size plays a crucial role in achieving chemical uniformity and high density31. Although previous studies have demonstrated that powder mixing and proper choice of the printing processing window can effectively enhance properties across different material systems, the mixing method remains empirical, and the principles underlying mixing and printing remain unclear. In this study, we aimed to establish a mixing-printing principle based on measurable physical quantities to effectively reduce the number of anomalies in as-printed metal. We occasionally observed that PBF-LB/M with a mixture of CoCrNi and 316 L powders produced a compact bulk with an unexpectedly smooth surface finish. To investigate the key factors that control densification and properties, we designed a series of CoCrNi-to-316L ratios, optimized the laser processing window, and fabricated MEA bulks with high density and strong resistance to fatigue crack growth. By carefully examining the characteristics of as-printed microstructure and melt pools, our results suggested a strong correlation between configurational entropy evolution and densification behavior. Specifically, the configurational entropy can reach a peak of 11.53 J/mol·K during printing, thereby modifying the thermal history and facilitating a transition from heterogeneous dislocation networks to a uniform subgrain-dominated architecture. Based on the results of this study, we propose a strategy termed temporal entropy engineering that leverages the dynamic melting–mixing–solidification process inherent to PBF-LB/M to modulate configurational entropy. This approach enables an in-situ increase in configurational entropy during printing, enhancing melt pool spreading and modulating the solidification microstructure. The results of this study provide new insights into the densification mechanisms in AM processing and propose a powerful strategy to enhance the performance of AM-fabricated alloys for long-term load-bearing. Results Entropy increasing and densification behavior of as-printed alloys Before mixing, the 316 L and CoCrNi raw powders exhibited a similar size distribution, ranging from 20 to 59 μm (see Supplementary Note 1). The morphologies of the raw powders were near-spherical with adherent satellite particles, a typical feature of gas-atomized metallic powders32. After mixing, most of these satellite particles were removed due to friction between the different types of powders (Supplementary Fig. 2), and the 75T-MEA size range was reduced to 17–50 μm. After simulating the processing window with small cubes, porosity was used as the sole criterion to determine the optimal printing parameters for dense samples with varying compositions. Figure 1a shows the effect of adding CoCrNi on the densification of T-MEA (blue boxes), which is presented along with the configurational entropy (red spheres). The configurational entropy, Smix, of T-MEA was calculated based on the concentration shown in Supplementary Table 1, using the equation Smix=−R∑i=1nxilnxi33, where R is the gas constant (8.314 J/mol·K), xi is the mole fraction of each constituent element, and n is the number of elements present in the alloy. As shown in Supplementary Table 1, with increasing CoCrNi content, the entropy of T-MEA first increases and then decreases, reaching a maximum value of 11.53 J/mol·K at 75T-MEA. This value is close to the theoretical maximum entropy of 11.82 J/mol·K that is expected for a mixture of these two materials. Densification behavior and defect morphology of as-printed 316 L, CoCrNi, and T-MEA alloys. a The variation of as-printed average density (blue cubic symbols) as a function of the mixing ratio of CoCrNi. Error bars represent the standard error of the mean. The density change exhibits a non-linear densification trend following a similar pattern to that of configurational entropy (red circular symbols). In a, error bars are standard error (SE), n = 5. b–e CT reconstructions of 75T-MEA reveal the distribution of micro-voids, represented by color mapping with regular, circular contours. f Typical morphology of voids of PBF-LB/M 316 L, highlighting larger, more complex defects with sharper contours. g–i Cross sectional morphologies of 316 L, 75T-MEA, and CoCrNi, indicating that 75T-MEA contains fewer void defects. The average porosity for 75T-MEA is 7 and 10 times lower than that in CoCrNi and 316 L. j Side-view of the surface region of as-printed 75T-MEA bulk. The inset presents the measured surface roughness. In (j), error bars are the standard deviation (SD), n = 3. k Internal defect density of T-MEA samples as a function of mixing ratio of CoCrNi, in which 75T-MEA has the lowest defect density (1.17 defects/mm3). The dashed lines and the dotted lines marked the defect density in near-void-free AM alloy reported by Qu et al.27. Source data are provided as a Source Data file. At the same PBF-LB/M parameters, by using micro-computed tomography (micro-CT), the density of as-printed 316 L and CoCrNi was measured as ~ 99.74% and 99.84%, respectively. If the thermal history in manufacturing each T-MEA is always identical, theoretically, the density of mixed 316 L and CoCrNi should increase linearly as the portion of CoCrNi increases. However, the density of the PBF-LB/M nT-MEA does not exhibit a linear trend, as revealed by the CT-characterized density in Fig. 1a, in which the lengths of the error bars are determined by using the standard error of the mean. Overall, the density plots indicate that as the CoCrNi content in T-MEA increases, the sample density first increases, and then decreases, reaching a maximum value at 75 wt.% CoCrNi (average density 99.9991%, for details of the density data, see Supplementary Note 2). Notably, the change in the density exhibits the same tendency as that of configurational entropy. As depicted in Fig. 1b–e, the CT reconstruction of a 2 × 2 × 2 mm3 75T-MEA cube reveals the lowest porosity, with defects that are smooth and nearly spherical. In contrast, in the as-printed 316 L, the defects exhibit larger sizes and more irregular shapes with sharp outlines (Fig. 1f). These sharp and irregular voids can be attributed to the uneven energy absorption and higher residual stresses that occur during the laser melting of the irregular distribution of satellite particles in the raw powder32. Furthermore, we confirmed the variation trend in density using random cross-sectional observations in SEM, and the typical cross-section morphologies are shown in Fig. 1g–i. The average porosity shown in the inset of Fig. 1h for 75T-MEA is 7 and 10 times lower than that in CoCrNi (Fig. 1i) and 316 L (Fig. 1g), respectively (see Supplementary Note 3). The similar trend observed in both CT and SEM measurements demonstrates that the densification behavior is not due to random error but follows the pattern of entropy evolution. As shown in Fig. 1j, the side-polished surface of the 75T-MEA, observed under SEM using electron-channeling contrast imaging (ECCI) conditions, exhibits a dense, uniform microstructure with well-fused layers and shallow melt pools. The inset in Fig. 1j shows the arithmetic mean surface roughness (Sa) of the printed bulk, showing that the 75T-MEA has the lowest Sa of 0.94 μm. The trend is consistent with configurational entropy and density. Using defect volumes obtained from CT scanning, we calculated the average internal defect density for all as-printed bulk (details are provided in Supplementary Note 3). As depicted in Fig. 1k, mixing printing reduces the defect density to a level lower than that of the single varieties on both ends. The 75T-MEA exhibits the lowest defect volume, 1.17 defects per mm3. Recently, by rebuilding the microstructure of a PBF-LB/M fabricated Ti-6Al-4V using post-processing, Qu et al. obtained a void-free AM (Net-AM) alloy and enhanced its fatigue resistance to a level higher than most metallic materials27. For comparison, we add a dotted line and a dashed line in Fig. 1k, indicating the level of defect volumes (measured using CT) in Net-AM and near-Net-AM alloys, respectively. Both alloys exhibit better fatigue resistance than their conventional counterparts. As highlighted by the shallow blue region, the defect volume of our as-printed 50T-MEA and 75T-MEA are between the Net-AM and near-Net-AM conditions. We must note that the defect volume of our T-MEA was obtained in an as-printed condition, which differs from the post-processing condition (hot isostatic pressing, solution heat treatment, and aging) reported in the reference27. Fatigue crack growth and electrochemical properties Fatigue properties, particularly fatigue crack initiation and growth, are highly sensitive to metal density. For instance, it has been reported that increasing the density of the alloy from 97.19% to 97.89% can significantly enhance the fatigue strength to a level higher than that of the as-cast ones34; surface densification (e.g., laser shock peening and shot peening) has been envisaged as an effective method for anti-fatigue of PBF-LB/M fabricated metal35. The fatigue crack growth resistance can be quantified by using the Paris exponent, which is an exponent of a power-law fatigue equation derived by Paris to accurately describe the steady-state crack growth regime36: in which da/dN represents the crack growth rate, measured as the length of crack extension per loading cycle (mm/cycle), ΔK denotes the stress intensity factor range, m is the Paris exponent (slope in a log ΔK -log da/dN plot), and C is a constant coefficient. Under the same environment, cyclic loading frequency, and stress ratio, materials with large m values are generally considered sensitive to crack growth under fluctuations in ΔK, indicating a lower resistance to fatigue crack growth. Contrastingly, materials with lower m values are generally preferred for load-bearing structures that require high fatigue resistance. To obtain the Paris exponent, the fatigue crack growth behaviors of as-printed 316 L, 50T-MEA, 75T-MEA, and CoCrNi were compared, and their steady-state crack growth diagrams are shown in Fig. 2. According to Fig.2a, the m values are 2.95 for 75T-MEA, 3.01 for 50T-MEA, 3.96 for 316 L, and 3.17 for CoCrNi, respectively. The lowest m value of 75T-MEA suggested it has a high resilience to load fluctuations. The remarkable enhancement in fatigue property is linked to the defect produced during PBF-LB/M. In addition, the inset in Fig. 2a shows the J-integral for fatigue crack growth, which represents the energy release rate at the crack tip (details are described in Supplementary Note 4). For a crack length a=10 mm, the J-integral in 75T-MEA is 3.549 × 105 J/mm2, which is an order of magnitude lower than that of 316 L (2.415 × 106 J/mm2) and CoCrNi (1.885 × 106 J/mm2). This substantial reduction in the energy release rate highlights the enhancement in 75T-MEA fatigue resistance. In Fig. 2b, the fatigue crack growth of the three T-MEA is compared with literature-reported PBF-LB/M or traditionally fabricated alloys37–46 tested under similar fatigue conditions to this study (a stress ratio of R = 0.1 and ΔK > 10 MPa·m1/2). It is noteworthy that the m value of 50T-MEA and 75T-MEA are significantly lower than most PBF-LB/M Ti alloy and Ni alloy, and its porosity is also much lower than that of MEA with a similar m value. The measured value of 2.95 for 75T-MEA is ~ 25% lower than the typical m value of 3.9–4.0 reported for PBF-LB/Med conventional MEAs, indicating improved resistance to fatigue crack growth. Fatigue crack growth behavior of as-printed T-MEA alloys, 316 L, and CoCrNi. a Fatigue crack growth curves of 316 L (red), 50T-MEA (green), 75T-MEA (blue), and CoCrNi (yellow) at a stress ratio of R = 0.1. The inset depicts the corresponding J-integral. b Comparison of the Paris exponent (m) of PBF-LB/M 75T-MEA obtained in this study (hexagons) with data reported for other PBF-LB/M alloys in the literature37–46. Source data are provided as a Source Data file. Figure 3a compares the potentiodynamic polarization curves of 316 L, 75T-MEA, and CoCrNi, with key electrochemical parameters summarized (statistical analysis of at least 3 curves for each material) in Fig. 3b using the Tafel slope method. The dashed lines in Fig. 3a represent the polarization curves for each sample, all of which show a tendency for self-passivation in the anodic region and similar self-corrosion current values. The pitting potential (Epit) of 75T-MEA is 0.816 V (vs. Ag/AgCl), which is higher than that of 316 L (0.743 V) and CoCrNi (0.658 V). This increase in Epit reflects improved resistance of the passivation film to localized breakdown and chloride-ion attack, indicating superior pitting corrosion resistance for 75T-MEA compared to the other material47. Electrochemical corrosion behavior of as-printed alloys in a 3.5 wt.% NaCl aqueous solution at room temperature. a Potentiodynamic polarization curves of 316 L, 75T-MEA, and CoCrNi tested in a 3.5 % NaCl solution. All alloys exhibit spontaneous passivation in the anodic region, yet 75T-MEA displays the most positive breakdown potential and a wider passive range, indicating better resistance to localized corrosion initiation than 316 L and CoCrNi. b Comparison of corrosion potential Epit (blue bars) and corrosion current density jcorr (red bars) from the polarization curves via Tafel extrapolation. 75T-MEA possesses the highest Epit of 0.816 V (vs. Ag/AgCl), exceeding 316 L (0.743 V) and CoCrNi (0.658 V), while maintaining a comparably low jcorr. This directly demonstrates that 75T-MEA forms a more robust passive film and delivers markedly enhanced pitting corrosion resistance in the chloride-containing environment relative to the other two alloys. In (b), error bars are the standard deviation (SD), n = 3. Source data are provided as a Source Data file. Phase and lattice distortion analysis To investigate how entropy variations influence the thermal history and densification of T-MEA, we analyzed microstructural features sensitive to entropy changes. Increased configurational entropy alters melt-pool cooling and solidification behavior, leading to changes in phase transitions, lattice distortion, and defect formation, which are first reflected in the evolution of XRD peak positions and widths. As shown in Fig. 4a, all samples exhibit a consistent austenitic (γ) structure. However, zooming in on the γ(111) peaks shows a rightward shift as the CoCrNi concentration increases. This shift corresponds to an increase in the scattering angle in XRD, indicating a reduction in the average lattice constant according to Bragg’s law. This phenomenon is likely due to the smaller atomic radii of Co and Ni compared to Fe, resulting in lattice contraction as Fe is progressively replaced by Co and Ni atoms48. XRD analysis of as-printed bulks. a XRD patterns listed in the order of CoCrNi ratio (from 0% in 316 L to 100% in CoCrNi). All samples show a single austenitic (γ) phase. The enlarged view in (b) highlights the γ(111) diffraction peak, which shifts to higher diffraction angles (to the right) with increasing CoCrNi content. c Relationship between CoCrNi ratio, grain size, and the half-peak width of the γ(111) reflection. The half-peak width increases from 316 L to 75T-MEA and then slightly decreases in CoCrNi, while grain size shows a marginal increase with higher CoCrNi content, indicating that the half-peak width variation primarily reflects changes in dislocation density and lattice distortion rather than grain size. Error bars in (c) are the standard deviation (SD). Source data are provided as a Source Data file. Moreover, as shown in Fig. 4c (left axis), the half-peak width (red boxes) of the γ(111) diffraction peak in the T-MEA samples gradually increases from 0.3394° (in 316 L) to 0.3689° (in 75T-MEA), before slightly decreasing to 0.3498° in CoCrNi. The increase in the half-peak width is generally taken as a sign of a finer grain size49. However, as plotted in Fig. 4c (right axis), the grain size (measured using the crystallographic orientation maps in Supplementary Fig. 5) changed from 23.04 µm in 316 L to 25.10 µm in 75T-MEA, and then 25.50 µm in CoCrNi, i.e., it proportionally increases as the mixing ratio of CoCrNi increases. Also, it is worth noting that the grain size variation is marginal. While XRD is sensitive to the crystallographic nature of the material, including microstrain, it does not provide a direct assessment of grain size; however, it offers a straightforward indication of internal disorder or lattice distortion within the crystalline structure. Therefore, the change in half-peak width as a function of CoCrNi concentration should be more directly connected with the change in dislocation density or substructure inside grains. Crystallographic orientation mapping and analysis of local misorientation The impact of CoCrNi addition on crystallographic orientation, geometrically necessary dislocations (GNDs; details shown in Supplementary Note 5), and grain morphology in T-MEA was further investigated using electron backscattering diffraction (EBSD) mapping in the building plane (see Supplementary Note 6). The local misorientation within each grain was analyzed by comparing the orientation of each point with the mean orientation of its host grain, as shown in the grain reference orientation deviation (GROD) maps (Fig. 5a, d, and g). In the GROD map, higher-value points indicate greater local strain accumulation and potential dislocation density50. In Figs. 5a and 5g, the GROD values in 316 L and CoCrNi appear heterogeneous, with distinct higher-misorientation regions (> 25°) along the melt pool. These higher values suggest greater residual strain/stress and increased dislocation density, particularly near melt pool boundaries. In contrast, the GROD map of T-MEA in Fig. 5d exhibits the lowest overall misorientation, as indicated by the lighter color. This suggests a microstructure with reduced lattice distortion and residual stress. The lower GROD values in T-MEA indicate that blending 316 L with CoCrNi modified the thermal conditions during PBF-LB/M, resulting in a more uniform temperature gradient. Crystallographic orientation analysis of PBF-LB/M 316 L (a-c), 75T-MEA (d-f), and CoCrNi (g-i) along the building plane. a, d, and g, grain reference orientation deviation (GROD) maps that characterize the misorientation of all scanning points with respect to the mean orientation of each grain, (b, e, and h) the geometrically necessary dislocation density maps. c, f, and i the distribution of crystallographic orientations in the inverse pole figure of the building direction (BD), colored by using the multiple of random distribution (MRD, the normalized probability density with respect to random distribution). The corresponding grain boundary misorientation angle distribution histograms are depicted below each inverse pole figure. Source data are provided as a Source Data file. The GND maps in Fig. 5b, e, and h reveal that the overall distribution of GND in as-printed 316 L, CoCrNi, and 75T-MEA shares a similar pattern, with dislocations primarily concentrated along the melt pool boundaries, forming aggregation zones, as marked by arrowheads. However, differences in the overall GND density are evident. Specifically, 316 L has a GND density of 2.48 × 1014 m−2, while CoCrNi exhibits a slightly lower value of 2.44 × 1014 m−2. In contrast, 75T-MEA exhibits the lowest GND density of 1.98 × 1014 m−2, indicating a relatively lower level of dislocation multiplication. This lower geometrically necessary dislocation density suggests that 75T-MEA experiences the least cyclic stress induced by laser melting and fast solidification. This relatively uniform internal stress distribution should correspond to a more stable dislocation structure than the other materials51. To evaluate the arrangement of crystallites in as-printed meltas, the distribution of crystallographic orientations in 316 L, CoCrNi, and 75T-MEA was colored using the multiple of random distribution (MRD) in the inverse pole figures, and the results are shown in Fig. 5c, f, and i. The 75T-MEA has an MRD of around 3.7, which is significantly higher than that of 316 L and CoCrNi, indicating its building direction is well aligned with the crystallographic <101> direction. A high MRD value in as-printed metal usually indicates facilitated grain reorientation during laser-metal interaction, which can be achieved through recrystallization (e.g., laser rescanning) and an extended heating time. Histograms of grain boundary misorientation angles for all samples are shown below each MRD figure. The 316 L exhibits a pronounced peak near 60˚ (Blue dashed line), corresponding to the typical {111} < 110>twinning in FCC metals52, whereas this peak is much less visible in CoCrNi. In contrast, almost no twinning misorientation was observed in 75T-MEA. As the concentration of T-MEA falls between 316 L and CoCrNi, its twinning energy is unlikely to be outside the range of these two alloys; therefore, the absence of twinning misorientation could be attributed to recovery or recrystallization during manufacturing. The overall misorientation variance values for 316 L, 75T-MEA, and CoCrNi are 210.7, 161.6, and 198.7, respectively. The results of MRD and misorientation distributions suggest that the microstructure in 75T-MEA underwent a different evolution than that in samples printed from single-variety powders. Microstructure analysis Figure 6 presents the micrographs obtained by using bright-field scanning transmission microscopy (BF-STEM). As shown in Fig. 6a, the predominant microstructure in the as-printed T-MEA sample was outer subgrains delineated with fine precipitates and embracing an inner cellular structure (mainly composed of dislocations). As shown in Fig. 6b, the as-printed 316 L exhibits only cellular structures without discernible subgrains, whereas Fig. 6c reveals that CoCrNi contains a few subgrains located primarily along grain boundaries. Microstructure in the as-printed and its evolution during deformation. a Bright-field STEM image of 75T-MEA showing subgrain boundaries (arrows) surrounding an inner cellular structure (dashed circle, ~ 350 nm). b As-printed 316 L exhibits only cellular structures (dashed circle) without subgrains. c As-printed CoCrNi contains a few subgrains (arrows) and a predominant cellular structure (dashed circle). d EDS mapping of 75T-MEA reveals Mn-, Si-, and O-rich precipitates (indicated by down-pointing triangles) distributed along subgrain boundaries with minimal segregation in the matrix. e–g Micrographs and corresponding selected area diffraction patterns (SADP) of 75T-MEA, 316 L, and CoCrNi after 20% tensile deformation. The deformed 316 L exhibits extensive deformation twins (dashed box in f) across grains. CoCrNi shows prevalent stacking faults (dashed box in g) while its cellular morphology becomes indistinct. In contrast, 75T-MEA maintains stable and well-defined subgrain and cellular structures. Source data are provided as a Source Data file. The cellular structure exists in all three alloys. It has a shape similar to a deformation-induced dislocation cellular structure53, and is commonly observed in PBF-L/M, laser cladding, and directional solidification of metals54. Its formation is attributed to the rapid cooling and steep thermal gradients. The size of the cellular structure, measured by the intercept method, is similar in 316 L, CoCrNi, and 75T-MEA samples, with an average value of ~ 350 nm. By contrast, the outer subgrains are uncommonly seen in PBF-L/M samples. Numerous studies have demonstrated that the formation of subgrain boundaries can occur due to the coarsening or recovery of cellular structures during PBF-LB/M or subsequent heat treatment. For example, in PBF-LB/M Ti-35Nb alloys, lower cooling rates and reduced laser power promote the growth of cellular structures into subgrains, thereby enhancing structural uniformity55. Similarly, in PBF-LB/Med Cu–Sn alloys, post-heat treatment transformed cellular structures into recovered and recrystallized subgrain morphologies, confirming that subgrain boundaries represent a recovered state of the cellular structure56. By comparing those micrographs in Fig. 6a–c, we noticed that the portion of subgrain-cellular structure is also consistent with variations in density, indicating there is a connection between the densification and microstructure. In addition, in Fig. 6d, the 75T-MEA EDS mapping reveals that Mn-, Si-, and O-rich precipitates were primarily located at the subgrain boundary of T-MEA. Despite these precipitates, no significant solute segregation was detected within the matrix of 75T-MEA, indicating that the mixing printing altered the solute distribution during the PBF-LB/M process. To further evaluate the microstructural stability under deformation, we also investigated the deformation structure in these alloys. Figure 6e–g presents micrographs of 75T-MEA, 316 L, and CoCrNi subjected to ~20% engineering strain, with the insets showing selected area diffraction patterns (SADP) observed along the [1¯10] zone axis. After deformation, 316 L exhibits dense deformation twins traversing the grains, as evidenced by the pronounced twin reflections in the SADP. In contrast, the deformed CoCrNi sample is characterized by a predominance of stacking faults, exhibiting distinct streak features in the SADP, while its cellular structure becomes increasingly indistinct. Remarkably, 75T-MEA retains a well-defined cellular and subgrain structure, with neither twins nor stacking-fault streaks observable in its SADP. We only observed a few of the twins in the fractured 75T-MEA (see Supplementary Note 6). Similar to 316 L and CoCrNi, dislocation slip in 75T-MEA traverses the cellular walls; however, the presence of subgrain boundaries terminates the motion of dislocations, indicating that the subgrain network effectively impedes dislocation motion and contributes to structural stability during plastic deformation. Overall, the unique microstructure of 75T-MEA originates from its distinct thermal history during PBF-LB/M. The higher MRD value, lower GND density, and the coexistence of subgrains and cellular structures reflect a thermally stabilized state, while the weak twinning peak in misorientation distribution and the absence of deformation twins after straining confirm that this subgrain–cellular configuration enhances structural stability. Melt pool morphology analysis The above microstructural analysis demonstrated the influence of entropy variations on the thermal history of T-MEA. To characterize the variations of the thermal process, it is necessary to investigate the detailed morphology of the melt pool. The typical melt pool morphologies at 800 mm/s and different laser powers are shown in Fig. 7a–c. In Fig. 7a, the 316 L sample exhibits fluctuations in track width at lower energy inputs, accompanied by occasional discontinuities and spatter. This is due to insufficient energy, leading to rapid cooling and uneven solidification, which causes inconsistent track width and incomplete melting at the edges. At higher energy inputs, the larger melt pool size enhances fluidity but also intensifies surface tension effects, leading to humping and track instability. In contrast, the 75T-MEA samples demonstrated a more stable melt pool, as shown in Fig. 7b, with a uniform and continuous track along the entire scan direction. This improved melt track stability can be attributed to the addition of CoCrNi. As the portion of CoCrNi increases, the entropy of the alloy system increases, and consequently, it induces a more disordered atomic arrangement, which can mitigate the formation of undesirable inhomogeneities and improve the uniformity of solidification. At lower energy levels, discontinuities were also evident in the melt pool of CoCrNi (Fig. 7c), similar to those observed in the 316 L melt pool. However, as the input laser energy increases, the melt pool of CoCrNi has better overall continuity and stability than that of 316 L. The melt pool behavior of these three alloys exhibited a similar tendency to that of the change in entropy and density values shown in Supplementary Table 1. Comparison of single-track stability and melt pool behavior of 316 L, 75T-MEA, and CoCrNi at a fixed scanning speed of 800 mm s−1. a–c Representative optical images of single tracks for (a, 316 L, b, 75T-MEA and (c) CoCrNi under varying laser power. 316 L exhibits irregular track width, discontinuities and spatter at low energy input owing to insufficient energy and rapid uneven solidification and humping at high energy input. 75T-MEA maintains a continuous and uniform track across the entire power range. CoCrNi displays discontinuities at low energy density similar to 316 L, but achieves better continuity and stability than 316 L at higher energy inputs. d Relationship between melt pool width and energy density, showing that 75T-MEA exhibits the widest and most uniform melt pool under identical processing conditions. e Surface roughness parameters Ra and Rz of single tracks as a function of the CoCrNi fraction. The Ra decreases from 2.83 µm for 316 L to a minimum of 2.39 µm for 75T-MEA, and then slightly increases to 2.49 µm for CoCrNi; Rz follows the same trend. This roughness minimum aligns well with the highest entropy and densification observed in 75T-MEA, indicating a more stable melt pool. In (e), error bars are the standard deviation (SD), n = 3. Source data are provided as a Source Data file. Figure 7d depicts the melt pool width variation under different energy densities for 75T-MEA, 316 L, and CoCrNi, with energy density as the horizontal axis and melt pool width as the vertical axis. The linear fit of pool width vs. energy density shows that, for the same energy input, 75T-MEA exhibits the widest melt pool, indicating better spreading ability than 316 L and CoCrNi. This enhanced spreading would improve fusion quality and facilitate efficient material deposition in the 75T-MEA. Figure 7e shows the variation in surface roughness of single-track melt pools as the CoCrNi ratio increases. Specifically, the Ra (average surface roughness) decreases from 2.83 µm in 316 L to a much lower value of 2.39 µm in 75T-MEA, before slightly increasing to 2.49 µm in CoCrNi. A similar trend is also evident in the Rz (mean roughness depth) curve. These patterns align well with the entropy values and densification tendency in these as-printed metals (Fig. 1). The smoother surface indicates a more stable melt pool and improved wetting behavior of the liquid phase, which can contribute to higher densification57. A smoother surface also suggests fewer irregularities and defects at the track level, promoting better layer bonding in the subsequent solidification process (see Supplementary Note 7). Figure 8 presents the optical micrograph, SEM morphology, and EBSD mapping of the 75T-MEA melt pools. The optical appearance in Fig. 8a provides a top-down view of the melt pool along the SD-TD plane, where explicit ripple-like features are evident, which are generally considered a sign of balanced heat input and melt flow58. As shown in Fig. 8b, the cross-sectional view along the SD-BD direction reveals that the melt track of 75T-MEA has a dense interior, free of visible defects, suggesting a well-developed solidification process. Figure 8c presents the corresponding EBSD mapping of Fig. 8b, in which the shape and the fusion boundary in the melt pool can be distinguished according to the grain distribution, as indicated by dashed lines. This crystallographic orientation map indicates that the single-track depth of 75T-MEA is approximately 57 μm, with a depth-to-width ratio of 2.42. This optimal depth-to-width ratio can enhance fluidity of molten metal59, facilitating uniform heat dissipation and preventing keyhole formation or excessive vaporization. This, in turn, contributes to improved melt pool stability and promotes material densification. Morphology and crystallographic characterization of single/multiple tracks in 75T-MEA. a Optical micrograph of a single melt pool showing ripple-like morphology along the SD-TD plane. b Cross-sectional SEM image along the SD-BD direction, showing a dense interior. c EBSD mapping of the cross-section, showing the single-track depth (~ 57 μm) and depth-to-width ratio (2.42). d, e Optical micrograph and SEM image of three overlapping tracks showing good connectivity and continuity. f crystallographic mapping of the cross-section for overlapping tracks showing crescent-shaped grains beneath the remelted surface, with a grain length-to-width ratio of ~ 1. Note that the EBSD noise (non-indexable points) in (c, f) is due to the sample mounting process. Source data are provided as a Source Data file. To connect these observations with the entropy-induced thermal history variations mentioned earlier, we also investigated the morphology evaluation during multi-pass laser melting. As shown in Figs. 8d and 8e, the multi-pass overlapping of melts can remelt a portion of the solidified single tracks, thereby providing much wider space for gas bubble escape, further reducing the number of defects. Beneath the multi-pass melt surface, in Fig. 8f, the formation of distinct crescent-shaped grains is evident, where the ratio of grain length to grain width is approximately 1, in contrast to the ~ 1.2 reported in CoCrNi60. This result indicates that the enhanced melt-pool fluidity in 75T-MEA facilitated uniform grain growth. We also confirmed that the surface grain morphology of the as-printed 75T-MEA cube is similar to that of the single track, indicating that single-track quality directly impacts the densification of the bulk sample (see Supplementary Note 8). Discussion Thermal conductivity and microstructural stability The PBF-LB/M process is characterized by rapid, localized solidification. The high density of the as-printed T-MEA is attributed to several factors that influence its microstructure and thermal history. One key factor is in-situ entropy evolution during laser processing, which leads to on-the-fly changes in atomic arrangement, limiting the motion of electrons and phonons61,62. To better understand the relation between thermal properties and material performance, the thermal conductivity (K), electronic thermal conductivity (Ke), and lattice thermal conductivity (Kl) of T-MEA were quantified. The thermal conductivity K is calculated using the equation K=α/(ρCp), where the thermal diffusivity α and the specific heat capacity cp can be measured using the laser flash apparatus (LFA) method, and the sample density, ρ, has already been determined in the previous section. The electron contribution Ke can be calculated using the Wiedemann-Franz law63,64, which relates the electronic thermal conductivity to the electrical resistivity and temperature via Ke=LσT, where σ represents electrical conductivity, T denotes temperature, and L signifies the Lorentz number with a value of 2.44 × 10−8 W⋅Ω/K2. The Kl is then derived by subtracting Ke from the total thermal conductivity K. Figure 9a shows the values of measured thermal diffusivity (α) for different as-printed samples, from which the thermal conductivity of each sample can be calculated, and the corresponding values of thermal conductivity K at room temperature are presented in Supplementary Table 4. As shown in Supplementary Table 4, Kl decreases gradually in T-MEA, which correlates with a decrease in the number of precipitates as the CoCrNi content increases. Recent studies by Körmann et al.65 have demonstrated that in high-entropy alloy systems, the increase of secondary phases induces significant chemical disorder and local lattice distortion, both of which cause pronounced phonon broadening. This broadening reduces the phonon mean free path, thereby fundamentally lowering the lattice thermal conductivity. Furthermore, as elucidated by Qian et al.66, the introduction of configurational entropy alters the phonon spectral characteristics through mass fluctuation, force-constant disorder, and lattice distortion. On one hand, these factors intensify phonon–phonon interactions by enhancing anharmonic scattering processes (e.g., Umklapp and four-phonon scattering) and may even lead to phonon localization. On the other hand, electronic carriers are also scattered by the chemically disordered lattice, which enhances electron–phonon coupling and suppresses electronic thermal conductivity (Ke). Zhang et al.67 further confirmed that such chemically complex structures effectively modulate thermal transport in the electron subsystem, thereby delaying heat dissipation during solidification and reducing defect stabilization. Thermal diffusivity analysis of T-MEA, 316 L, and CoCrNi alloys. a Room temperature thermal diffusivity of as printed T-MEA, 316 L stainless steel and CoCrNi measured by the laser flash analysis (LFA) method. b Thermal diffusivity of T-MEA alloys as a function of the CoCrNi mixing ratio recorded over the temperature range 300–700 K. Curves follow a parabolic trajectory, decreasing initially and then increasing with the CoCrNi mixing ratio. The 75T-MEA consistently exhibits the lowest thermal diffusivity across the entire temperature interval. Source data are provided as a Source Data file. As shown in Fig. 9b, over a wide temperature range of 300–700 K, the thermal diffusivity α as a function of the CoCrNi mixing fraction exhibits a parabolic trajectory, decreasing first and then increasing. According to the tendency, we applied a quadratic polynomial fitting to the curves of thermal conductivity and CoCrNi mixing fraction and aimed to find the alloy composition that corresponds to the lowest K. For example, the fitted equation at 300 K is Kx=0.00047x2−0.0717x+14.86, where x is the CoCrNi fraction. Based on this fitted equation, we can calculate the theoretical minima and maxima of the thermal conductivity curve. The results indicated that the maximum thermal conductivity occurs at 0% CoCrNi, with a value of 14.86 W/m·K, while the minimum thermal conductivity is achieved at approximately 77% CoCrNi, with a value of 12.10 W/m·K (see Supplementary Note 9). As suggested by Wang et al.68, a lower thermal conductivity can significantly prolong the cooling time of the melt pool, which allows for more thorough diffusion of the liquid metal and promotes more uniform in situ alloying. At the same time, the slower heat dissipation rate reduces the temperature gradient across the melt pool, as noted by Zhou et al.55. The enhanced diffusion and slower heat dissipation are directly correlated with the ultra-high densification of 99.99% and the unique microstructures observed in this study. This is because the extended cooling time ensures that the alloying elements are dispersed and also induces recovery of dislocations during the layer-by-layer laser process, leading to less internal stress and a more homogenous microstructure. Our TEM observations revealed dislocation annihilation and the formation of subgrain boundaries, which are likely related to dynamic recrystallization occurring under low thermal conductivity conditions. As reported by Zhang et al.55, at low laser power or high scanning speed, slow cooling rates allow the cell structure to transition towards low-energy subgrain boundary structures. The formation of subgrain boundaries redistributes dislocations generated by cyclic thermal stress, thereby reducing the likelihood of localized deformation or failure. Moreover, the low thermal conductivity can also reduce the residual stress generated in PBF-LB/M69, which lowers the internal driving force for crack initiation. This is closely related to the observed decrease in the Paris law m-value for 75T-MEA, indicating improved stability and resistance to crack propagation. Solidification time and melt pool spreading behavior The spreading behavior of the melt pool during the PBF-LB/M process can be understood as a competitive interaction between the diffusion and solidification of the molten liquid, with heat loss in the capillary playing a critical role in governing this process. As the laser beam irradiates the molten droplets, these droplets adhere to the solidified surface, creating a solidification angle (θsolid) at the interface with the solid. Concurrently, the unsolidified liquid portion diffuses at a diffusion angle (θdiff) at the same interface. As wetting continues, θdiff decreases as the droplets spread across the solid surface. In contrast, θsolid increases as the melt solidifies, leading to a larger volume of solidified material. The spreading behavior of the entire melt pool ceases when θsolid equals θdiff. It can be reasonably inferred that for effective spreading on the bulk substrate before solidification, the solidification time must exceed the spreading time. For 316 L and CoCrNi, which have similar densities and melting points, their spreading times are also comparable. Thus, variations in melt pool microstructure and density between 316 L, CoCrNi, and T-MEA are largely influenced by differences in their solidification times. The specific solidification times t for these materials can be calculated using the formula proposed by Gao et al. 70: where a denotes the radius of the melt pool, Tf is the melting temperature (also referred to as the liquid-phase line temperature), Tt is the ambient temperature, P is the laser power, and R is a parameter related to the position of the laser light source. Although this formula does not account for the increased heat loss associated with droplet diffusion, it provides a useful approximation for understanding the behavior during solidification. For 316 L, the parameters are as follows: ρ = 7.75 g/cm³, cp ≈ 500 J/(kg·K), and Tf ≈ 1400 °C71. For CoCrNi, the values are: ρ = 8.11 g/cm³, cp ≈ 425 J/(kg·K), and Tf ≈ 1400 °C72. The product of density and specific heat capacity is denoted as ρcpa2, measured as 2.29 × 10−2 J/(m·K) for 316 L and 2.31 × 10−2 J/(m·K) for CoCrNi. The melting temperature Tf of T-MEA is measured as ~ 1400 °C. Based on Supplementary Table 4, the ρcpa2 value for T-MEA is around 2.30 × 10−2 J/(m·K), akin to the values for 316 L and CoCrNi. According to those parameters, the solidification times for T-MEA, 316 L, and CoCrNi are calculated as 1.265 × 10−3 × (R-6.611) seconds, 1.028 × 10−3 × (R-6.817) seconds, and 1.261 × 10−3 × (R-6.622) seconds, respectively. Therefore, the 75T-MEA has the lowest thermal conductivity among all samples and exhibits the longest solidification time, despite having a spreading time comparable to that of the other materials. This prolonged solidification time results in a continuous increase in θsolid, as the molten droplets adhere to the solidified surface while spreading. Given that the spreading behavior of the melt pool ceases when θsolid equals θdiff, the extended solidification time of 75T-MEA allows θsolid to grow more significantly before reaching equilibrium with θdiff. This calculated result is consistent with the experimental observations in Fig. 7, where 75T-MEA exhibits the widest melt pool and the lowest melt pool surface roughness. Differences between printing with mixed-powders and pre-alloyed powders In PBF-LB/M technology, pre-alloyed powders are generally preferred for their ability to produce components with high chemical homogeneity. However, it has also been reported that mixed powder has more advantages in terms of densification and improved properties. In the PBF-LB/M processing of AlSi alloys, it has been found that some pre-alloyed powders are prone to cracking during laser processing, leading to high macroscopic residual stress73. In contrast, Roberts et al.74. found that AlSi produced from a mixture of pure Al and Si powders had fewer cracks than those printed using pre-alloyed powders. Similarly, Kang et al.75 have shown that the energy density required to process AlSi12 from mixed powders is lower than that required for pre-alloyed powders. During melting and solidification, compositional differences within mixed powders can generate additional chemical gradients and enhance melt-pool convection, resulting in fewer defects and longer solidification duration76. The thermal/chemical exchange and associated melt convection resulting from the melting of different powder species should also play a significant role in controlling the melt morphology of T-MEA alloy. This powder-mixing effect and the temporal entropy-increasing effect should explain our observation that the T-MEA melt pool is wider and more continuous than those of 316 L and CoCrNi. Localized compositional differences can also alter thermal expansion and contraction, which can help alleviate stress concentrations that are typically more pronounced in pre-alloyed systems. The effect of powder mixing may help redistribute residual stress and improve defect tolerance. As a result, the mixed-powder approach provides a convenient method for controlling melt behavior and mitigating stress accumulation, thereby enhancing densification and performance of the as-printed material. Beyond T-MEA manufacturing, the proposed design and printing approach also holds promise for other materials in AM. For instance, in high-strength alloys like titanium or metal matrix composites, which are susceptible to cracking and high residual stresses during AM, this strategy could modify the solidification process, minimize defects, and improve mechanical properties, making it suitable for demanding applications in industries such as aerospace and automotive. In summary, the time-dependent increase in configurational entropy during PBF-LB/M of mixed powders drives densification, producing a near-fully dense T-MEA with enhanced anti-fatigue crack growth properties and corrosion resistance. Entropy-driven modulation, together with precipitate formation during laser–material interaction, increases lattice disorder and phonon scattering, lowering thermal conductivity by 18.4% relative to single-variety alloys. The extended solidification process modifies melt pool thermal history, reduces residual stress, and lowers the internal defect density of the PBF-LB/Med MEA; it also promotes the formation of a stable microstructure dominated by subgrain boundaries and dislocation cellular structures, which remains stable under external loading. The results from the study suggest that time-dependent entropy evolution is a viable mixing powder principle for density-oriented additive manufacturing, providing a promising avenue for the single-step printing of structural materials with enhanced long-term properties. Methods Raw Material The raw materials consisted of 316 L stainless steel and CoCrNi alloy powders (AVIMETAL), both prepared by vacuum induction melting and gas atomization, with particle sizes ranging from 15 to 53 μm. The 316 L powder had the following composition, determined by ICP-MS: Fe (bal) – 16.83 at.% Cr – 10.64 at.% Ni – 2.44 at.% Mo – 0.76 at.% Mn – 0.42 at.% Si. The CoCrNi powder composition was Co (bal) – 33.14 at.% Cr – 33.61 at.% Ni. Powders were weighed with a high-precision analytical balance and mixed for 5 h in a 3D powder mixer (SC-110HF). The mixing process didn’t alter the spherical morphology or flowability of the powders. Mixed powders with CoCrNi contents of 0, 25, 50, 75, and 100 wt.% were prepared, corresponding to 316 L, 25T-MEA, 50T-MEA, 75T-MEA, and CoCrNi samples, respectively (see Supplementary Table 1 for chemical concentrations). PBF-LB/M Process Additive manufacturing was performed using a PBF-LB/M system (Hans-M-100, Hans Laser) with a continuous Ytterbium fiber laser (1070 nm, 50 μm spot diameter). The process parameters (detailed in Supplementary Table 5) included a hatch spacing of 70 μm, a layer thickness of 30 μm, and a scanning speed of 800 mm/s. Each layer was rotated 67° relative to the previous one to reduce anisotropy, enhance interlayer bonding, and minimize periodic defects by avoiding repetitive melt-track overlaps77. The PBF-LB/M energy density is defined by E=P/(v×t×h)(E, energy density; P, laser power; v, scanning speed; t, layer thickness; h, hatch spacing). The calculated energy densities for each sample are summarized in Supplementary Table 5. Single-track experiments were conducted to assess the stability and track continuity of the melt pool directly. These experiments varied the energy input by adjusting the laser power and scanning speed. Scanning speeds ranged from 600 mm/s to 1000 mm/s (in increments of 200 mm/s), and laser power was varied from 100 W to 200 W (in increments of 20 W) to establish the relationship between energy input and melt pool characteristics. Density and porosity measurement Process instabilities during localized laser–powder interactions can lead to the stochastic formation of large spatters and pores, which serve as primary sources of defects and hinder final densification of as-printed metals. Therefore, to filter out other influences and highlight the effects of entropy and microstructure, we carefully controlled the laser processing window. Specifically, identical laser processing parameters were applied to all samples, ensuring that factors such as laser power, scanning speed, layer thickness, and hatch spacing remained consistent. In addition, the position from which each sample was taken during printing was standardized across all runs, minimizing variation due to location-dependent effects. By keeping these parameters constant, we minimized random errors arising from process fluctuations or sample heterogeneity. We employed computed tomography (CT) scanning and measurement of cross-sectional defects in scanning electron microscopy (SEM, Apreo2 S Lovac, FEI) to systematically evaluate the densities of the samples. To compare the density of samples with different percentage-based porosity, we also measured the relative density by using Archimedes’ method. A micro- and nano-focused CT inspection system (Diondo d2) operating at 120 kV and 100 μA with a maximum resolution of 3 μm was utilized to visualize three-dimensional defects. To ensure accuracy and reproducibility of the results, multiple regions at different locations (upper, middle, and lower) in each sample were scanned. Each sample underwent at least three repeated measurements under identical conditions to verify consistency. For SEM measurement, random cross-sections of as-printed samples were manually polished using SiC sandpapers with grit sizes ranging from 180 to 2000 grit, followed by further polishing using a VibroMet polishing machine (BUEHLER) with a 0.5 μm particle-sized polishing solution to achieve a higher-quality surface. The surface of the polished samples was then examined under an SEM (Apreo2 S Lovac, FEI), and their density/porosity was statistically characterized using ImageJ software. Mechanical tests Before mechanical tests, all specimens were polished with SiC papers, progressing through grit sizes from 180 to 2000 to ensure a smooth surface. To obtain the hardness of the PBF-LB/M fabricated sample, we performed Vickers microhardness tests on the scanning direction-transverse direction (SD-TD) plane using a Vickers diamond pyramid indenter (HXD-1000TMC/LCD, Shanghai Taiming Optical Instrument Co., Ltd.) with a 10 N load and a holding time of 10 s. For details on the definition of the SD-TD side of the printing process, see Supplementary Note 10. Tensile tests were performed on a universal testing machine (Wance X10A) equipped with an optical extensometer (TestStar-UTM) at a strain rate of approximately 2.5 × 10−4 s−1. Dogbone-shaped tensile specimens were cut to dimensions of 42 mm in length, 13 mm in width, and 2 mm in thickness using electrical discharge machining (SodickAG400l). The fatigue crack propagation experiment was conducted in accordance with the ASTM E647 standard. Compression and tension specimens were used for the fatigue test. The specimen dimensions are shown in Supplementary Note 11. Constant-load fatigue tests were conducted with a stress ratio of R = 0.1, a frequency of 25 Hz, and a constant maximum load of 3500 N78. The non-contact measurement of specific heat capacity and thermal diffusivity of the sample is conducted using the laser scattering method. This process is carried out with the LFA467 instrument (NETZSCH), which is equipped with an infrared detector. The infrared detector captures thermal radiation emitted from the sample surface, allowing for the accurate determination of temperature changes during the experiment. This method ensures that temperature measurements are obtained without direct physical contact with the sample, minimizing potential interference with the thermal response of the material. In addition, both tensile and compression performance tests were also conducted to further evaluate the mechanical response of the materials. Detailed experimental procedures and results are provided in Supplementary Note 12. Electrochemical tests Electrochemical corrosion studies were conducted at a controlled temperature of 25 °C within a 3.5 wt.% sodium chloride (NaCl) solution. Polarization curves were measured using the VersaSTAT 3 F electrochemical workstation (Princeton Applied Research), equipped with a standard three-electrode configuration. The reference electrode was a silver-chloride (Ag-AgCl) electrode, the working electrode was the specimen, and a platinum foil acted as the counter electrode. To ensure the attainment of a stable electrochemical equilibrium, open circuit potential (OCP) measurements were conducted over a 30 min duration before the commencement of potentiodynamic polarization measurements. Dynamic potentiodynamic polarization curves were recorded at a steady scan rate of 0.167 mV/s. The potential was swept from an initial − 0.2 V vs. Ag/AgCl to a final 0.8 V vs. Ag/AgCl. The electrical conductivity (σ) of the as-printed sample was measured using a digital eddy current tester (FD-102) following ASTM E1004-09. This method relies on electromagnetic induction, where an alternating magnetic field induces eddy currents in the material, and the resulting impedance changes are analyzed to determine conductivity. To ensure accuracy, a dual-coil probe was calibrated using 100% IACS copper and 61% IACS aluminum standards. Measurements were conducted at 23 ± 1 °C, with polished sample surfaces (Ra < 1.6 μm) to minimize variability. Five readings were taken per sample, and the average value was reported. Microstructure characterization and analysis The PBF-LB/M fabricated T-MEA specimens underwent various characterizations, including grain size, elemental composition, phase composition, melt pool boundary and segregation. Before these analyses, the sample surfaces were subjected to pre-treatments, including abrasion and polishing. A scanning electron microscope (SEM, Apreo2 S LoVac, FEI) equipped with electron backscatter diffraction (EBSD, MTEX toolbox and TSL-OIM analysis software) and electron channeling contrast imaging (ECCI) capabilities was used to analyze the surface morphology. EBSD measurements were conducted with a step size of 0.1 μm, an accelerating voltage of 20 kV, a probe current of 13 nA, a detector tilt of 70°, and a working distance of 15 mm. SEM imaging was performed at 5 kV with a beam current of 1.3 nA, using an ETD signal detector and a working distance of 10 mm. By applying the methodology proposed by Pantleon79, we estimated the density of the geometrically necessary dislocations (GNDs) from EBSD results using the MTEX toolbox80. By comparing melt pool cross-sectional EBSD maps with morphologies obtained by scanning electron microscopy, we found that the melt pool in our sample consistently forms a semi-elliptical zone, where grains have different sizes, morphologies, and orientations from those in the surrounding substrate. These differences are a direct consequence of the rapid melting and solidification experienced within the melt pool compared to the solid-state substrate. Therefore, we identified the melt pool boundary as the outlines of the semi-elliptical zones (delineated by black dashed lines in Supplementary Fig. 5). Within these boundaries, the melt pool depth (D) and width (W) were measured using ImageJ software81,82: depth was taken as the maximum vertical distance from the top surface of the track to the deepest point of the boundary, and width as the maximum horizontal extent at the original substrate surface level. The depth-to-width ratio (D/W) was then calculated from the cross-sectional geometries obtained at a scanning speed of 800 mm/s and laser powers ranging from 100 W to 180 W. For TEM observations, 3-mm disks were thinned using electrochemical twin-jet polishing in a solution containing 8% perchloric acid and 92% ethanol until electron-transparent regions appeared in the center of the disks. In addition, a 200 keV transmission electron microscope (TEM, Talos F200X G2, FEI) equipped with scanning transmission electron microscopy (STEM) functionality. A spot size of 6, a C1 magnetic condenser lens with an aperture size of 50 µm, and a camera length of 330 mm were used for BF-STEM imaging, providing optimal diffraction-contrast conditions for high-resolution dislocation observation while reducing unwanted phase contrast. An energy-dispersive X-ray spectroscope (EDS) was employed for even more detailed microstructural analysis. In addition, the surface roughness of the printed samples was measured using a surface profiler (SNEX031DX-12, Japan). The morphology of the single-track melt pool was examined through confocal microscopy (VK-X1000, Japan). X-ray diffraction (XRD) analysis was conducted with a SmartLab-9 kW instrument (Rigaku) using Cu-Kα radiation (λ = 1.54 Å, 45 kV, 200 mA) to assess changes in phase composition. Furthermore, the chemical composition of the bulk sample was also double-checked by using inductively coupled plasma mass spectrometry (ICP-MS, Agilent 7700). Thermal diffusivity measurements The thermal diffusivity of the T-MEA samples was measured using a laser flash apparatus (LFA 467, NETZSCH) over a temperature range of − 100 to 500 °C. Measurements above 100 °C were conducted under high-purity nitrogen to prevent oxidation. A short laser pulse was applied to the specimen surface, and the temperature rise on the opposite side was recorded using a non-contact infrared detector. Disc-shaped samples (12.7 mm in diameter, 2 mm thick) were prepared according to the instrument specifications, and both surfaces were coated with graphite to enhance laser absorption and reduce radiative losses. The instrument was calibrated using Pyroceram 9606 standards, with measurement uncertainty controlled within ± 3%. Each specimen was tested three times to ensure reproducibility. Thermal conductivity (k) was calculated using k = αρCp, where α is thermal diffusivity, ρ is density measured by Archimedes’ method, and Cp is specific heat capacity obtained using a sapphire reference (the measured values are provided in Supplementary Table 4). Supplementary information Supplementary Information Transparent Peer Review file Source data Source Data Acknowledgements All authors gratefully acknowledge the assistance from Dr. Yang Qiu and Dr. Dongsheng He at SUStech Core Research Facilities. We thank the Electron Microscope Center of KAIPLE Co., Ltd. (Changsha) for the support of microstructural characterizations. Author contributions Z.L. Conceptualization, Methodology, Experimental Design, Formal Analysis, Investigation, Writing - original draft. Q.L. Investigation, Data curation, Visualization, Investigation, Experimental Support, Validation. J.Y. Visualization, Investigation, Experimental Support, Validation. B.W. Experimental Support, Validation. Z.R. Experimental Support, Validation. M.Y. Investigation, Validation, Funding acquisition, Supervision, Writing - review & editing. S.S. Visualization, Writing - review & editing. S.W. Conceptualization, Methodology, Funding acquisition, Supervision, Writing - review & editing. Peer review Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Funding This work is supported by the National Key R&D Program of China (No. 2022YFB4600700). We thank the financial support from the National Natural Science Foundation of China (No. 52475364), Shenzhen Science and Technology Innovation Commission (KJZD20240903101400001), and the Development and Reform Commission of Shenzhen Municipality (XMHT20240115003). SW appreciates the support from the Fang Keng Faculty Award Fund. M.Y. acknowledges the support from the Guangdong Basic and Applied Basic Research Foundation (No. 2023A1515240048) and the Shenzhen Science and Technology Program (No. JCYJ20240813094806009). Data availability Source data are provided in this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Mujin Yang, Email: yangmj@sustech.edu.cn. Shuai Wang, Email: swangi@zju.edu.cn. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76700-3. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review file Source Data Data Availability Statement Source data are provided in this paper.