Introduction

The synthesis and processing of metals is the largest source of greenhouse gas emissions1. Most metals occur in the Earth’s crust as mixed oxides or sulfides, traditionally extracted via energy-intensive pyrometallurgical reduction using fossil-based agents (e.g., coke, coal), releasing CO21. This approach is predominant for recovering transition metals such as iron, nickel, copper, chromium, cobalt, manganese, zinc, and tin from their oxidized forms2. After extraction, additional processing is needed to produce application-ready parts, including liquid alloying, thermomechanical treatment (e.g., forging, rolling), machining, or shaping through extrusion or atomization followed by powder metallurgy or additive manufacturing. Thus, starting from extraction all the way to the final part production has their associated energy footprint thereby negatively impact economic sustainability.

H2-based redox reactions have emerged as a sustainable alternative to fossil reductant-based extraction, oxidizing H2 to H2O rather than CO2, directly supporting decarbonization3. Vacancy formation during H2 redox promotes atomic interdiffusion, enabling solid-state alloying and mass transport for densification. Recent studies have demonstrated alloying from mixed oxides (e.g., Fe-, Ni-, Cu-, Co-, Mo-, Cr-oxides), but under idealized conditions: nano-sized, high-purity powders shaped into micro-lattices and free from naturally occurring gangue oxides such as silicates (SiO2) and alumina (Al2O3)4,5,6. In practice, gangue is inevitably present in industrial ores7; removing these gangues via separation or hydrometallurgy incurs substantial environmental costs. To realize the full potential of H2-based metallurgy at scale, two fundamental questions must be answered: (1) firstly, whether near-net-shape bulk components can be directly synthesized and (2) secondly, how trace gangue oxides influence alloying behavior and final microstructure. Resolving these questions is critical for advancing H₂-based metallurgy toward industrial-scale, sustainable implementation.

To this end, we investigate a sustainable manufacturing paradigm: net-shaping of bulk alloy parts directly from mixed transition metal oxides through H₂-based redox alloying, without extensive purification. This approach is demonstrated by manufacturing austenitic stainless steel one of the ubiquitous alloys with wide range of applications, while deliberately retaining realistic gangue levels. A suspension front bearing serves as a demonstrator to assess the feasibility of net-shaping. Furthermore, we rationalize the sequential oxide reduction and alloying within mixed system using thermodynamic calculations, offering a new pathway for design oxide precursors as alloying agents.

Figure 1 outlines the proposed paradigm. The case-study alloy, austenitic stainless steel, comprises iron as the base matrix, with chromium for passivation and nickel and molybdenum for corrosion resistance. Iron and chromium are commonly derived from oxide-bearing mineral sources (e.g., hematite, chromite) that often coexists with alumina/silicate gangue partially removable via beneficiation (e.g., leaching, physical separation)8; alkali roasting may be used later for the subsequent extraction9. Molybdenum is commonly produced from molybdenite concentrates via oxidative roasting to molybdenum trioxide (MoO3), which serves as the standard precursor to metallic Mo10. Nickel recovery is comparatively challenging because it only presented at low concentrations in lateritic or sulfidic ores11. Although H₂-based reduction of nickel sulfides is thermodynamically feasible, the presence of sulfur complicates processing12; therefore, nickel oxide (NiO) is used here as a simplified carrier to model nickel addition. Overall, the selected oxides serve as industrially relevant model feedstocks. These oxides are mixed and shaped into a preform with a geometry close to the final part. Upon exposure to H2 and high-temperature, the preform undergoes redox-driven reactions form metallic stainless steel part. During this process, shrinkage driven by oxygen removal and porosity closure is expected; however, this dimensional change can be compensated for through design.

Fig. 1
Fig. 1
Full size image

Conceptual schematic of ore-derived oxides to net-shape stainless steel via H2-based redox alloying.

Results

Oxides-to-metal conversion

Among the transition metal oxides presented in the mixture (Supplementary Fig. S1), Fe2O3, NiO, or MoO3 are unstable phases under H2 below 1300 °C13,14, while reduction of Cr2O3 is only thermodynamically feasible under carbon-based reductant at 1300 °C, as widely utilized in ferrochromium production using CH4-H2 mixtures15. Experimentally, the reducibility of the mixed oxides under H2 was evaluated using a slow ramp of 1 K/min up to 1300 °C. The extent of oxides-to-metal conversion, defined as α = (oxygen removed)/(total oxygen initially present), was derived from TGA measurement and plotted in Fig. 2a. The α-curve climbs to α ≈ 0.81 by 700 °C, surpassing the theoratical maximum α if assuming Cr2O3 remains unreduced. The inset, which displays the instantaneous mass loss rate, highlights that most of the reduction occurred rapidly below 700 °C. Above this temperature, the reduction progresses asympototically as α approach unity where all transition metal oxides were fully metalized when approaching 1300 °C. Ex-situ XRD patterns in Fig. 2b corroborate the complete reduction, including that of Cr2O3, and reveal a transformation from mixed oxides phases to a single set of FCC reflections. This suggests that the reduced Fe, Ni, Cr, and Mo atoms have been successfully incorporated into a solid solution, forming an austenitic matrix. EDS analysis from four scan areas of 2500 µm2 across the TGA sample cross-section showed an average composition of Fe68.11±0.43 Cr16.74±0.34 Ni12.01±0.25 Mo2.34±0.61 (wt.%), confirming a near-complete reduction of all transition metal oxides to form the targetted Fe68Cr17Ni12Mo3.

Fig. 2: Characterization of oxide mass and phase change upon H2-based redox.
Fig. 2: Characterization of oxide mass and phase change upon H2-based redox.
Full size image

a Oxide-to-metal conversion (α) and instantaneous mass loss (∆m(t)/m0) derived from TGA. b XRD patterns before and after TGA.

Macro- and micro-structural evolution

The reducibility of the oxide mixture was further evaluated using gel-cast parts processed under a tube furnace with the same heating rate (1 K/min) as in the TGA experiment. X-ray diffractograms collected from the center of the sample cross-section reaffirmed a complete transformation to a single FCC phase following H2-redox (Fig. 3a), accompanied by a distinct color change from hematite red to metallic silver (Fig. 3b). Dimensional comparison against the original CAD model revealed a substantial volumetric shrinkage of 77.6% (Fig. 3c), attributable to both redox-driven volume changes and sintering-induced densification. The theoratical shrinkage due to metalization of oxides can be approximated from the molar volume ratio of the pure metal (\({V}_{{metal}}\)) and its oxide (\({V}_{{oxide}}\)): \(1-{{nV}}_{{metal}}{/V}_{{oxide}}\), where \(n\) is the number of metal atoms per mole volume of oxide. Applying a composite theory (i.e., mass-weighted rule of mixtures) to the present oxide composition yields a calculated shrinkage of 52.4 vol.%, due solely to molar volume change during reduction (Fig. 3c). The remaining shrinkage is ascribed to the densification via pore closure. Notably, despite the high shrinkage observed, the final geometry exhibits isotropic shrinkage and retains the intended design.

Fig. 3: Macro- and microstructural evolution during H2-based redox alloying.
Fig. 3: Macro- and microstructural evolution during H2-based redox alloying.
Full size image

a X-ray diffractograms of the oxide mixture before and after H2-based redox alloying in the tube furnace. b Photographs showing the geometrical changes of the gal-cast part. c Comparison between theoretical shrinkage calculated from oxide reduction and the experimentally measured value. d Optical micrograph (left) and SEM-EDS maps (right) showing the dense microstructure decorated by Al- and Si-based particulates.

Microstructural analysis (Fig. 3d) shows a fully-dense, crack-free matrix with grain-level uniform distribution of Fe, Cr, Ni, and Mo, as a result of H2-redox alloying. However, submicron inclusions, further resolved by SEM-EDS after Kalling etching, were identified as residual Si- and Al-based gangue oxides located at boundary areas and pore interfaces. These oxides remain chemically distinct, forming no detectable compound with transition metals, as further confirmed by EDS analyses on unetched microstructure (Supplementary Materials, Figs. S2 and S3). Prior studies on H2-reduced iron ores have linked the presence of Si/Al oxides to swelling and cracking due to their more than one order of magnitude thermal expansion mismatch with FCC-iron16,17,18. However, no such macroscopic defects are observed here. Nonetheless, the gangue oxides consistently coincide with residual pores or reside at interfaces, both indicative of mechanical degradtion due to stress concentration and interfacial debonding. It seems that purification of oxides to remove gangue elements are imperative if relying on ore-derived feedstock for direct net-shape manufacturing of metal components.

Collectively, these findings establish the feasibility of direct ore-to-part fabrication using H2-based redox alloying, producing bulk components without geometrical deformation. However, the complete recovery of each element, specifically, Cr and Mo from their oxides under the solid-state H2 sintering condition reported here, is still not well understood. Previous studies have reported significant Cr recovery (up to ~50 wt.%) from chromite ores via H2 plasma smelting reduction to produce ferrochrome as the stainless steel precusor19,20. Such plasma-assisted processes, however, operate under non-equilibrium conditions that differ fundamentally from the thermally driven gas–solid reduction reactions relevant to the present process. To gain insight into the reduction sequence leading to the formation of the final austenitic stainless steel, thermodynamic calculations were therefore performed.

Discussion

Firstly, equilibrium phase evolution for the ore-derived oxide mixture in H2 is presented in Fig. 4a. Under equilibrium at low temperature, Fe3O4, FeCr2O4, NiO, and MoO2 are stable. It is interesting to note that, none of the starting oxides, i.e., Fe2O3, MoO3, Cr2O3, are stable at low temperature in the presence of H2, indicating reaction between oxides to form stable products. As temperature increases, the reduction proceeds sequentially, beginning with MoO2 → Mo, followed by NiO → Ni, and finally Fe3O4→FeO→Fe and Cr2O3→Cr. Upon formation of elemental Fe in BCC phase, subsequent reduction products like Ni, Mo, and Cr dissolve progessively into the BCC of Fe matrix, ultimatly forming austenitic stainless steel (FCC phase in Fig. 4a). To quantify this reduction sequence under expreimental conditions as shown in the TGA data in Fig. 2, thermodynamic calculations were performed based on the reduction of 10 g of ore derived oxide mixture using 100 g of H2 (100 g was assumed to simulate the continuous flow of H2).

Fig. 4: Thermodynamic and process metallurgy calculations of oxide reduction and metal formation during H2-based redox alloying.
Fig. 4: Thermodynamic and process metallurgy calculations of oxide reduction and metal formation during H2-based redox alloying.
Full size image

a Equilibrium phase evolution for a mixture of oxide powders and excess H2. b Process metallurgy calculation indicating amount of oxide and metal formed when considering 10 g of mixed oxides in H2. c Relative mass change and metal conversion rate corresponding to (b). d Evolution in constitents of oxides with reduction temperature. e Evolution of metal composition with reduction temperature.

Figure 4b shows the evolution of oxide and metal masses as a function of temperature. Initially, the system contains 10 g of oxide. As the temperature increases, the formation of metallic species begins, accompanied by a progressive reduction in oxide mass. Approaching 1250 °C, the metal mass reaches ~ 7 g, while the oxide mass is fully depleted. Beyond 1250 °C, the masses of both metal and oxide pleateau, indicating that the completion of the reduction process. To validate the thermodynmaic simulation against the TGA data presented in Fig. 2, the relative mass change and metals conversion were calculated and are plotted in Fig. 4c. The close agreement bewteen the nature of the curve in Fig. 4c and the experimental TGA in Fig. 2b supports confidence in the thermodynamic calculations. To further resolve the reduction sequence and the composition of the metal obtained, the constituents of oxides and metal composition are shown in Fig. 4d, e, respectively. The results indicate that NiO readily reduces first, followed by MoO2. The absence of MoO3 in Fig. 4d suggests an immediate conversion to MoO2. Subsequently, Fe2O3 reduces further into FeO, while both FeO and Cr2O3 gradually reduce as temperarure increases. This stepwise reduction is directly reflected in the evolving metallic composition shown in Fig. 4e.

From the results presented in Fig. 4, it is clear that the reduction proceeds sequentailly as NiO→Ni, MoO3→Mo, Fe2O3→Fe, and finally Cr2O3 →Cr. Going purely by the Ellingham diagram21, only the reduction of NiO→Ni by H2 can be directly explained. The reduction mechanisms for the remaining oxides, necessitate further discussion.

Considering the sequential reduction of the four compounds, Fig. 5 demonstrates the thermodynamic spontaneity of their respective reduction steps. Figure 5a shows the Gibb’s free enegy curves for oxidation reactions of Ni and H2. The diagram clearly indicates that NiO reduction by H2 is thermodynamically favorable, elucidating why Ni is the first metallic species formed during the reduction (Ref. Fig. 4e).

Fig. 5: Gibbs free energy curves for reduction reactions.
Fig. 5: Gibbs free energy curves for reduction reactions.
Full size image

Curves showing reduction behaviour of: a NiO, b MoO2, c Fe2O3, and d Cr2O3.

Figure 4a, d indicate clearly that MoO3 is not a stable oxide under these conditions; instead, MoO2 is thermodynamically favored. As illustrated by the Gibbs free energy curves for Mo-based oxides in Fig. 5b, that the initial reduction of MoO3 to MoO2 is highly favorable at low temperatures (Reaction a). This explains why MoO2 is the dominant oxide phase at low temperatures, as observed in Fig. 4a, d. However, direct reduction of MoO2 into Mo is thermodynamically unfavorable at low temperatures, suggesting an intermediate reduction step involving Mo4O11 Reaction a’ in Fig. 5b. Lalik22, using in situ neutron diffraction, have confirmed Mo4O11 as an intermediate oxide formed during MoO3 reduction to MoO2. Therefore, we propose that MoO3 initially reduces to MoO2 and Mo4O11 (Reaction a’, Fig. 5b), and subsequently, the combined MoO2 and Mo4O11 phases undergo further reduction to metallic Mo via Reaction b. The final reduction becomes thermodaynamically feasible above 280–300 °C, corresponding precisely to the observed temperature at which Mo partitions into the metallic phase in Fig. 4e. Thus, the reduction of MoO2 to metallic Mo occurs fully between 280 and 400 °C.

Fe2O3 readily reduces into magnetite (Fe3O4) under H2, as illustrated by Reaction c in Fig. 5c. Near approximately 400 °C, the formed Fe3O4 is further reduced to metallic Fe (Reaction e, Fig. 5c), explaining the observed increase in Fe content in the metal phase. Furthermore, above 650–700 °C, any remaining Fe3O4 can further further reduce to FeO (Reaction d, Fig. 5c), accounting for the presence of FeO observed in Fig. 4d.

Similar to MoO2, Cr2O3 is challenging to reduce solely under H2. However, several studies have have reported the reduction of Cr2O3 under H2 at relatively lower temperatures. For example, Radomysel’sii et al.23 investigated the co-reduction of Fe2O3 and Cr2O3 under H2 targetting Fe-20Cr (wt.%) composition. They observed only trace amount of Cr2O3 after 7 h at 1100 °C and spectualted that the thermodynamic driving force originating from reduced Fe2O3 facilitated the reduction of Cr2O3. This observation underscores a critical limitation of the classical Ellingham diagram, which assumes that the activity of the metal αMe as unity, thus it models oxidation of pure metals exclusively. In multicomponent alloy systems, where αMe < 1, the Gibbs free energy barrier for oxide reduction is effectively lowered. The equilibrium conditions for oxidation in such multicomponent systems are more accurately described by the modified relation24:

$$RT\,ln{p}_{{O}_{2}}={{\rm{G}}}^{{\rm{o}}}-\frac{2a}{b}RT\,ln\,{\alpha }_{{Me}}$$
(1)

where \({G}^{o}\) is the standard oxidation reaction of the pure metal, and \({\rm{a}}\) and \({b}\) refer to the stoichiometry of the metal oxidation \(\frac{2{\rm{a}}}{{\rm{b}}}{\rm{Me}}+{{\rm{O}}}_{2}\to \frac{2}{{\rm{b}}}{{{\rm{M}}}_{{\rm{e}}}}_{{\rm{a}}}{{\rm{O}}}_{{\rm{b}}}\). For instance, assuming an ideal Fe68Cr17Ni12Mo3 (wt.%) solution with αMe ≈ 0.18, the driving force for Cr2O3 reduction (quantified by \({\mathrm{\varDelta G}}_{\mathrm{oxide}}^{{\rm{o}}}-{\mathrm{\varDelta G}}_{{{\rm{H}}}_{2}{\rm{O}}}^{{\rm{o}}}\) value) shifts by ~ 39 kJ/mol toward spontaneity. Similarly, more recent study conducted by Shankar et al.25 reported that in mixed oxide systems, the reduced activity of Cr in the metallic phase could aid in lowering the reduction temperature of Cr2O3.

From the perspective of lowering thermodynamic barrier for such reaction, we propose an additional mechanism to explain the observed low temperature reduction of Cr2O3. As previously discussed for Fe2O3, residual Fe3O4 can undergo reduction into FeO beyond 600 °C (Reaction d, Fig. 5c), with the presence of FeO further supported by thermodynamic calculations in Fig. 4d. FeO and Cr2O3 can react to form spinel based FeCr2O4 via Reaction f in Fig. 5d, which is thermodynamically favorable, as indicated by the equlibrium calculations in Fig. 4a. The FeCr2O4 spinel can undergo reduction to metallic Fe and Cr via Reaction g in Fig. 5d. It should be noted that this reaction is thermodynamically feasible above ~ 1000 °C, corresponding to the temperature range where Cr partition into the metallic phase is observed. Furthermore, the inset in Fig. 5d confirms that the reduction of FeCr2O4 under H2 becomes thermodynaimcally favorable above 1000 °C, with the complete reduction achieved close to 1280 °C.

In summary, we demonstrate that using ore derived oxides as feedstock for additive manufacturing, followed by sintering under H2 atmosphere is a viable manufacturing route to tranform extracted ore into near-net-shape parts. We demonstrate this concept using mixed oxide system comprising Fe2O3, Cr2O3, NiO, and MoO3 to fabricate bulk near-net-shaped austenitic stainless steel part. Based on experimental and thermodynamic results, the following conclusions can be drawn:

  1. (i)

    H2-based redox directly transforms sustainable oxide preforms into net-shape bulk metal parts. Alloying and densification can proceed even in the presence of minor Al and Si-based gangue impurities; however, these impurities tend to reside within residual pores and at interfaces. Shrinkage during oxide metalization is significant but uniform, in constrast to classical powder sintering where differential shrinkage often leads to slumping and distortion.

  2. (ii)

    Phase transformations between oxides within the mixed system can alter the reduction pathway and lower the reduction temperature relative to pure single oxides, thererby assisting in the reduction of otherwise difficult-to-reduce under H2. We propose this mechanism as the main pathway for the reduction of MoO3 and Cr2O3 during the formation of austenitic stainless steel.

  3. (iii)

    Complete reduction of difficult-to-reduce oxides by exploiting oxide-to-oxide phase transformation can open up new avenues in CO2-free steel production and on-the-fly alloying and can potentially provides a pathway to reduce refractory oxides (Nb, W, and V) with comparable stability under H2.

Together, these findings lay the foundation for utilization metallic ores as of sustainable, low-cost feedstocks for direct net-shape metal manufacturing.

Methods

Near-net-shape manufacturing of oxide mixture

For our investigation, micronized Fe2O3 (KREMER Pigmente, D90 = 5.91 µm), Cr2O3 (KREMER Pigmente, D90 = 3.42 µm), NiO (Alfa Aesar, 400 mesh), and MoO3 (Sigma Aldrich, ~100 nm) powders were blended to target a nominal composition of 316 grade austenitic stainless steel (17Cr–12Ni–3Mo–68Fe (wt.%)) after complete oxide reduction (Fig. 1). The oxide mixture was homogenized via low-energy ball milling (250 rpm, 4 h, ball-to-powder ratio 10:1), then shaped into a suspension bearing using gel-casting following established protocols described in ref. 26. Gangue impurities were introduced through the Fe2O3 pigment, contributing 0.34 wt.% Si and 0.36 wt.% Al.

Oxide reduction experiments

Redox test of the oxide mixture (without binder) was performed using thermogravimetric analysis (TGA; STA 449F3, Netzsch). A sample mass of 107.41 mg was subjected to a flowing H2 atmosphere (99.999% purity, Linde AG) at 100 mL/min. The thermal profile consists of heating at 1 °C/min to 1300 °C, followed by a 2 h of isothermal hold. The measurement was repeated under identical conditions. Redox test of gel-casted demonstrators was conducted in a tube furnace (Elektromat Dresden) under flowing H2 at 2.5 L/min. A debinding step was first performed at 500 °C for 1 h, followed by redox processing using the same thermal profile as employed in the TGA experiments.

Macro- and micro-structural characterizations and analysis

Post-reduction geometries were measured using a contact-free optical scanner (ALCON, Hexagon) and compared with the original computer-aided design (CAD) model to quantify volumetric shrinkage (vol%) and evaluate geometrical distortion. Phase analysis of the ball-milled mixture and redox specimens were conducted via X-ray diffraction (XRD; D8 DISCOVER, Bruker) using a Cobal source (Co-Kα, λ = 1.79026 Å), operated at 40 kV and 45 mA. Microstructural analaysis was carried out using a field emission scanning electron microscopy (FESEM; Leo 1530, Zeiss) equipped with an energy dispersive spectroscopy (EDS) detector. Thermodynamic calculations presented were performed using TCOX10 oxides database of ThermoCalc in conjunction with the Process Metallurgy module of ThermoCalc27,28.