NATURE · ARTICLE · 30 APRIL 2025

U. Manzoor¹  ·  L. Mujica Roncery²  ·  D. Raabe³  ·  I. R. Souza Filho¹

¹ Max Planck Institute for Sustainable Materials, Düsseldorf, Germany

² Universidad Pedagógica y Tecnológica de Colombia, Tunja, Colombia

³ Institut Jean Lamour, CNRS (UMR 7198), Université de Lorraine, Nancy, France

Open access (CC BY 4.0) · Received 16 February 2024 · Accepted 14 March 2025 · DOI 10.1038/s41586-025-08901-7

Nickel is a critical element in the shift to sustainable energy systems, with the demand for nickel projected to exceed 6 million tons annually by 2040, largely driven by the electrification of the transport sector. Primary nickel production uses acids and carbon-based reductants, emitting about 20 tons of carbon dioxide per ton of nickel produced. Here we present a method using fossil-free hydrogen-plasma-based reduction to extract nickel from low-grade ore variants known as laterites. We bypass the traditional multistep process and combine calcination, smelting, reduction and refining into a single metallurgical step conducted in one furnace. This approach produces high-grade ferronickel alloys at fast reduction kinetics. Thermodynamic control of the atmosphere of the furnace enables selective nickel reduction, yielding an alloy with minimal impurities (<0.04 wt% silicon, approximately 0.01 wt% phosphorus and <0.09 wt% calcium), eliminating the need for further refining. The proposed method has the potential to be up to about 18% more energy efficient while cutting direct carbon dioxide emissions by up to 84% compared with current practice. Our work thus shows a sustainable approach to help resolve the contradiction between the beneficial use of nickel in sustainable energy technologies and the environmental harm caused by its production.

Key numbers

• 6 Mt/yr — projected Ni demand by 2040, doubled from 3 Mt today
• ~20 t CO₂ per ton of Ni in current practice — ten times the emission of a ton of steel
• −84% — potential cut in direct CO₂ emissions with hydrogen-plasma smelting reduction
• +18% — potential energy savings versus the rotary kiln–electric furnace route
• 1 step — calcination, smelting, reduction and refining merged into one furnace

Nickel is a strategic, hard-to-replace element: 1.97 million tons of stainless steel and 210 kilotons of non-ferrous alloys — especially superalloys — are made with it every year. But the batteries that will electrify transport need another 3 million tons by 2040, doubling global demand to 6 million tons per year.

The supply base is the wrong shape for this demand. Today 60% of annual nickel production comes from high-grade sulfide ores (1.5–4 wt% Ni), whose chemically simple minerals (NiS, Ni₃FeS₂, (Co,Ni)₃S₄) separate cleanly by froth flotation. The remaining 40% comes from low-grade laterites — saprolite and limonite, averaging 1.5 wt% Ni. In laterites there are no discrete Ni minerals: Ni²⁺/Fe²⁺ ions sit dissolved in complex magnesium silicates such as lizardite (Mg,Fe,Ni)₃Si₂O₅(OH)₄, or partially replace iron in goethite (Fe,Ni)OOH. Yet 60% of all land-based nickel reserves are locked in these laterites, and only 40% in sulfides (Fig. 1b).

Processing laterites is punishing. The energy demand runs from 230 to about 570 GJ per ton of Ni — far above the 22 GJ needed for steel. Emissions follow: the rotary kiln–electric furnace (RK-EF) and blast-furnace route emit about 45 t CO₂e per ton of Ni, high-pressure acid leaching (HPAL) about 14 t, pushing the industry average to 20–27 t CO₂e per ton — over ten times steel (2.3 t). Nickel is one of the most environmentally damaging metals to produce (Fig. 1c).

Instead of drying, calcining, smelting and refining in sequence, the whole dried ore charge is melted and reduced at once — hydrogen-plasma smelting reduction (HPSR) — in a single electric arc furnace (Fig. 1d).

Solid-state direct reduction with gaseous hydrogen has been tried, but the crystallographic complexity of laterites, the low Ni content (0.5–2 wt%) and up to 90% impurity oxides make the kinetics sluggish and the hydrogen utilization inefficient, and demand additional pre- and post-treatments. Direct reduction of saprolites is scarcely reported: their silicates are too thermodynamically stable at typical reduction temperatures (800–1,000 °C).

HPSR sidesteps that stability problem by melting first. In the molten state the crystal structures dissociate into ionic species — Fe²⁺, Ni²⁺, Mg²⁺, O²⁻, SiO₄⁴⁻ — without catalysts, and the melt is exposed to highly reactive hydrogen-plasma species (H₂, H and H⁺). Calcination, smelting, beneficiation, metal separation and refining collapse into one metallurgical step, operable entirely on renewable energy: renewable electricity replaces carbon-based fuels, and sustainably produced hydrogen replaces carbon- and sulfur-based reductants — eliminating direct CO₂ and SO₂ emissions and the acids used in HPAL.

In the experiments, ore fines were compacted into ~10 g green pellets, placed on the water-cooled copper hearth of an arc-melting furnace, and reduced for 2 min by a 200 A plasma arc in Ar–10% H₂ at 900 mbar. Even this brief exposure precipitated pure metal droplets.

Melted in argon alone, the ore solidifies as olivine (forsterite, Mg₂SiO₄, ~86 wt%) and pyroxene (MgSiO₃, ~10 wt%) — no Ni-rich phase at all. Nickel is dissolved in the silicates, confirmed by SEM–EDX mapping. The as-received ore loses ~23 wt% above 1,600 °C: moisture, dehydroxylation of serpentine minerals and decomposition of goethite to haematite.

The reduction mechanism is oxygen extraction, not direct attack on silicate. Free oxide ions (O²⁻) diffuse to the arc–melt interface and are taken by hydrogen-plasma species as water vapour or hydroxyl radicals. Each oxygen leaves two electrons behind, and those electrons are consumed preferentially by the cation with the lowest oxygen affinity — Ni²⁺, precipitating first, followed by Fe²⁺. The two metals form a homogeneous molten phase that solidifies into a single Fe–Ni alloy (Fig. 2c).

After just 2 min the metal has fully separated from the silicate melt — driven by a factor of two to three in density — and can simply be hammered out: ~5 mm nodules embedded in the unreduced slag, or 1–3 mm particles at the sample bottom (Fig. 2d).

A narrow thermodynamic window

Equilibrium calculations (ThermoCalc) for molten ore at 1,600 °C exposed to Ar–10% H₂ show that metal precipitation begins only after roughly 20% oxygen removal — and that first 20% is lost thermally, without any hydrogen. Beyond that point, grade and recovery trade off sharply: a narrow interval of 20.7–21.3% oxygen removal yields 70–90 wt% Ni ferronickel at 60–90 wt% recovery (Fig. 3a). Pushing past it feeds the alloy with iron and starves it of nickel.

The microstructure confirms the picture. Small droplets that solidified before coalescing act as micro-labs of the earliest reduction: one analysed at 97.3 wt% Ni, 0.44 wt% Fe and 2.22 wt% O — right in the calculated selective-precipitation range (Fig. 3d). After 2 min, 55 wt% of the ore's Ni was recovered as 52 wt% Ni alloy; after 4 min the alloy fell to 42 wt% Ni (74 wt% recovery) — the signature of Fe enrichment. Re-reducing the leftover silicate gave an alloy of only 9 wt% Ni and 90 wt% Fe.

Atmosphere control beats exposure time

A second ore (Ni ore 2, 1.57 wt% Ni, 16 wt% Fe) was reduced in two 2-min steps to test whether the leaner atmosphere in step 2 can hold grade without sacrificing recovery (Fig. 4):

Same 4 min of processing, same recovery — but nearly double the grade when the step-2 atmosphere is kept H₂-lean. The authors note the sequence need not be stop–start: in practice the furnace can run continuously, as in conventional EAF operation, with gas composition adjusted on the fly and the remaining slag cycled through to recover residual Ni (a further 4 min in Ar–2.5% H₂ tapped 10 wt% Ni, lifting total recovery to 78 wt%).

Emissions and energy: where the gains come from

In the RK-EF route, 87.7% of all CO₂ falls on primary extraction — rotary dryer, rotary kiln and electric furnace — while mining and ore preparation contribute only 4.4% and 6.6% (Fig. 5c). The pre-treatments also eat 57% of the process energy (rotary dryer ~15%, rotary kiln ~42%), of which only ~27% is actually needed to run the chemistry: the rest leaves as flue gas, heat lost in material transfer (calcined ore cools from 880 to 400 °C before the furnace) and dust (Fig. 5d).

HPSR deletes that loss chain. The dried ore goes straight into the arc furnace on renewable power; the H₂/H₂O off-gas can heat the dryer, with unreacted H₂ combusted for extra heat or recirculated after dust filtration and water condensation. The result: up to ~18% lower energy consumption and an 84% cut in CO₂ emissions, assuming renewable electricity and green hydrogen.

And the by-products have homes. The ~30% Ni ferronickel feeds stainless steel, Invar and magnet alloys without energy-intensive refining; the Mg-silicate slag can go into bricks, roads and cement — or be refined for cosmetics. For scale-up, the paper points to established EAF practice: short high-current arcs, more electrodes, electromagnetic stirring and bottom gas injection to keep free oxygen reaching the arc–melt interface, plus the natural drop in melt viscosity as complex silicates simplify. A detailed cost analysis in the supplementary information reports a positive financial outcome.