A single freeze steers ferrihydrite toward hematite a year after the ice is gone

Start with the arithmetic, which does not quite close. A micrometre divided by 30 is about 33 nanometres, comfortably above the few-nanometre size quoted for individual ferrihydrite particles ^[3], so the 30-fold growth figure is tracking aggregates rather than single crystallites ^[18]. Nor can it be read as a 30-fold loss of reactive surface. For a thin flake, specific surface area is set mostly by thickness and barely by lateral extent ^[19], and surface area is what governs how much nutrient, pollutant and organic carbon this mineral holds ^[14]. Surface area is the quantity that matters most downstream, and it is not what this account measures.

What it does report is persistence. The freeze lasts minutes ^[1]; twelve months later the mineral products of frozen and unfrozen material had still not converged ^[9]. A year is 525,600 minutes, so against a five-minute freeze that is a memory roughly a hundred thousand times longer than the event that wrote it ^[17]. The bonding result is what licenses extending this from a beaker to a melt stream: whatever joins those particles is not undone by returning them to liquid water ^[8].

Two distinctions are worth holding onto. The year in water shows suppression, an absence of goethite; the hematite, which is the mineral that inverts the climate reading, appeared under accelerated aging rather than in the ambient year ^[9]. And the freeze was at minus 20 C ^[5], with no other temperature reported. Ferrihydrite in a real active layer also arrives with organic carbon and nutrients bound to that same surface ^[14], and how those co-solutes behave once the shrinking liquid pockets concentrate them is not settled here ^[4].

The proxy consequence is the sharpest one. Goethite is the yellow-brown rust of cool, moist soils and hematite the red rust of warm, dry ones ^[10]; their ratio in ancient soils is used to reconstruct past climate, and oxygen isotopes in the same minerals to estimate past temperature, both on the assumption that the transformation happened in liquid water ^[11]. A cold event that yields the warm-dry mineral can push the inference in the wrong direction, not just add noise to it. Hematite also absorbs a broader band of sunlight than goethite, so a freeze-driven tilt toward it could speed light-driven iron cycling wherever polar soil thaws under sun ^[12]. Warming is intensifying freeze-thaw and pushing it into ground that did not have it ^[16], which means the share of samples carrying this history is growing rather than shrinking. Kevin M. Rosso of Pacific Northwest National Laboratory, in a Perspective accompanying the paper in the same issue of Science, treats ice itself as the reactor ^[15], which is also how Umea's Jean-Francois Boily puts it ^[2].