Industry

Agromining leachate crystallizes to 99.99 wt% nickel sulfate after a single wash — and the paper's own table shows where the claim overshoots

Kalakech, Sanchez, Buch, Matar and Puel (CentraleSupélec, Université Paris-Saclay) have published in RSC Advances (2026, CC BY-NC) a process study that sits exactly at the agromining bottleneck: not growing the plants, but turning their ash into something a battery maker will sign for. Nickel sulfate hexahydrate (NiSO₄·6H₂⁠O, α-phase) is the standard precursor for Ni-rich NMC and NCA cathodes, and battery grade means very tight impurity ceilings — the specification the authors cite requires ≥220,000 ppm Ni with Co, Ca and Mg each below 20 ppm and Fe and Mn each below 10 ppm. Their question: does a leach solution derived from hyperaccumulator biomass behave better or worse than the laterite and electrolyte feeds the crystallization literature usually studies?

A feed that only a hyperaccumulator could produce

The feed was synthetic but formulated from compositional data of a real post-solvent-extraction (SX) bio-ore leach solution provided by the authors’ colleagues: Ni 122,151 ± 2,699 ppm against Co 20 ± 2, K 48 ± 2, Mg 50 ± 2, Mn 4 ± 2, Fe 11 ± 1 and Ca 45 ± 3 ppm (their Table 3). That ratio — a fifth of a percent or less of any impurity per unit Ni — is the signature of the biological step. Plants that hyperaccumulate nickel, from Mediterranean Odontarrhena lesbiaca (syn. Alyssum lesbiacum, with histidine-based Ni chelation shown by Krämer and colleagues in the 1990s) to South African Berkheya coddii — proposed as a phytomining crop by Robinson and co-workers in 1997 — do not take up the soil’s full metal menu proportionally. By the time their biomass is ashed, leached and solvent-extracted, the hardest part of refining has arguably already happened in the roots.

Crystallization sorts by ionic resemblance to nickel

Batch evaporative crystallization at 50 °C and 116 mbar produced phase-pure α-NiSO₄·6H₂⁠O (XRD peaks at 19.4° and 20.9° 2θ) and partitioned the impurities decisively. Rejection into the mother liquor was nearly total for Mn (100%, below detection in the crystals), Ca (89.3%) and K (78.8%), substantial for Mg (67.4%) and Fe (56.6%) — and weakest for Co at just 33.9%, with a Co/Ni distribution coefficient of 1.04 ± 0.14, meaning cobalt enters the crystal almost as readily as nickel itself (Table 5). The ordering Co > Fe > Mg > K ≈ Ca > Mn tracks ionic resemblance, not feed concentration: Co²⁺ (75 pm), Fe²⁺ (78 pm) and Mg²⁺ (72 pm) sit within ~13% of Ni²⁺ (69 pm) and can follow it into the lattice or be sealed into growth-induced liquid inclusions, while the much larger Ca²⁺ (102 pm) and K⁺ (138 pm) are excluded (we note the paper’s own “<8%” and “>50%” mismatch figures are arithmetically loose — computed honestly, Co is 8.7%, Fe 13%, Ca 48% — but the qualitative argument is sound). Controlled-dissolution impurity mapping added the mechanistic detail that justifies the washing step: Ca is mostly surface-adsorbed (released within the first 20% of dissolution), while Co, Fe and Mg hide in mother-liquor inclusions between 60–80% dissolution depth — reachable by wash solution, unlike true lattice substitution.

One wash, 99.99 wt% — and an asterisk the authors leave in their own table

Displacement-washing the filter cake once with saturated NiSO₄ solution raised purity from 99.9845 ± 0.0023 wt% (unwashed) to 99.9900 ± 0.0014 wt%, the authors’ headline “single wash meets battery grade” result; a fourth wash reached 99.9932 wt% (Table 6). Here the piece must be exact, because the paper is not: the text and abstract claim one wash brought every impurity below spec, but the paper’s own Table 6 shows Co at 24 ppm and Mg at 32 ppm after that single wash, against the <20 ppm limit the paper itself quotes — and Mg is still 23 ppm even after four washes. Fe sits at exactly 10 ppm, not below it. Co only drops under 20 ppm at wash ratio 3. So the honest statement is: one wash reaches 99.99 wt% overall purity and gets most impurities near or under spec, with cobalt and magnesium as the documented holdouts — in a synthetic feed at bench scale. (One table cell is also almost certainly misprinted: K = 216 ppm at wash ratio 2, which is irreconcilable with the row’s own purity of 99.9913 wt% and with the text’s “≈19% decrease” between ratios 2 and 4; the impurity sums reconcile exactly at every other wash ratio if the value is ~16.)

Why this matters anyway

The mechanistic story is genuinely good news for agromining economics, with the caveats attached. Crystallization from this feed rejected most impurities because the impurity-to-Ni ratio was already low — the authors themselves attribute the easy purification to “the HPA plant selectivity of Ni toward impurities,” which is a restatement of the core agromining bet: the plant is the first refinery. Laterite leachates, with their Mg and Mn loads, typically demand multistage repulping and re-crystallization to reach comparable purity; here a single displacement wash nearly suffices. The study was conducted with French agromining company Genomines acknowledged as industrial partner and Bpifrance as funder — which is where money and attention in this field are visibly flowing, and worth more than any award as evidence of sector momentum.

The real risks, stated plainly

Three cautions belong in the same paragraph as the opportunity. First, the feed was synthetic, built from a colleague’s BLS compositional data — no actual agromining leachate, with its variability and organic carryover from SX, has yet been through this crystallizer, and the authors say so, flagging Fe(III) hydrolysis artefacts, residual organics nucleating inclusions, fine-crystal filtration problems, cake channeling and vacuum-evaporation energy cost as open scale-up questions. Second, the one impurity the crystal cannot reject is cobalt — and cobalt is precisely the element many nickel hyperaccumulators co-accumulate from ultramafic soils, so feedstocks from Co-rich serpentines could behave worse than this feed. Third, Mn at 100% rejection is good news here but a reminder that the same process discards the companion metals that some agromining business models hope to co-recover. The pathway to battery-grade nickel from plant ash is narrower, and more plant-dependent, than the headline purity figure suggests.

Provenance: open access (CC BY-NC); the full text was read directly from the Europe PMC record (PMC13697633), independently re-fetched and confirmed verbatim by a separate research pass, and every load-bearing number was verified against the primary source by three independent reviewers on different vendors (reasoning/code/kimicode) using a shared 11-claim template. Paper-internal inconsistencies are reported above, not smoothed: the “single wash meets all battery-grade impurity limits” claim contradicts the paper’s own Table 6 (Co 24 ppm, Mg 32 ppm at wash ratio 1; Mg 23 ppm even at ratio 4, vs the paper’s own quoted <20 ppm spec); Table 6’s K = 216 ppm at wash ratio 2 is arithmetically irreconcilable with the same row’s purity and is likely a misprint for ~16; the paper’s “<8%”/“>50%” ionic-radius mismatch figures are arithmetically wrong (Co 8.7%, Fe 13%, Ca 48% — we quote the radii and computed values instead); and the bench feed was synthetic, not an actual agromining leachate. Database cross-references: the histidine result for Odontarrhena lesbiaca traces to Krämer et al., Nature 379, 635 (1996); the phytomining-crop proposal for Berkheya coddii traces to Robinson et al. 1997. Methods and limits at /methodology/.

Primary source: https://doi.org/10.1039/d6ra06711e

← All news & analysis