Take a hacksaw to a bar of stainless steel. For a moment the cut face is bright bare metal, every atom exposed. Leave it on the bench; nothing visible happens. Come back tomorrow and the face is as stainless as the rest of the bar.
The usual explanation is that stainless steel is unreactive. It is closer to the opposite. That bright face began corroding before the saw finished its stroke, and it has not stopped since.
What stainless steel has is chromium — at least 10.5 per cent, by the definition written into the European standard — and chromium’s oxide behaves unlike iron’s. Rust is porous and poorly bonded: it flakes, admits air and water to the metal beneath, and lets an ordinary steel bar corrode to its centre. Chromium’s oxide grows dense, tightly adherent, nearly closed. Once there, it obstructs the very traffic of ions that built it. So the corrosion does not halt; it throttles itself, settling at a thickness of one to three nanometres at room temperature.
Sit with that. Three nanometres is on the order of ten atoms stacked — a layer no optical microscope could resolve, and the whole difference between a kitchen knife and a brown stain. It exists only because the metal under it is still being consumed at a rate that never quite reaches zero.
Which is why “self-healing” is both the right word and a misleading one. Scratch the film off and the bared metal does what the sawn face did: corrodes at once, and the products of that corrosion seal it. Scratch-electrode experiments — drag a stylus across a submerged sample and watch the current — show the spike collapsing within a few tenths of a second. But the film that closes a wound that fast is not yet the film that was there before. Followed with atomic force microscopy, the same scratches on 304 and 316 steel took one to three hours to rebuild to a steady state. Two clocks, not one: the bleeding stops immediately, and the scar takes the afternoon to become skin.
Now the part that changes the picture. It is natural to imagine the film as a uniform glaze. It is not. Grown on even a single crystal it comes out polycrystalline — a mosaic of oxide grains five to thirty nanometres across, and the chromium is not evenly shared among them. Map its electrical resistance grain by grain with a conducting probe and the values spread over an order of magnitude, with local excursions of two or three. Some grains are close to pure chromium oxide. Others are iron-rich and comparatively leaky. Between them run the seams.
So the surface does not have a level of protection. It has a distribution, and its fate is decided at the bottom of that distribution. Corrosion does not begin somewhere random; it begins at the poorest grains and the joins between them. Which is why stainless steel, when it fails, so rarely fails the way plain steel does. It is not thinned evenly. It is punctured.
Chloride is what finds those places. Chloride ions adsorb at the film’s step edges and pull metal away as soluble complexes, thinning it locally; where the edges are saturated, they penetrate instead. Corrosion scientists say chloride poisons the self-healing, and the word is exact: the repair is not overwhelmed, it is disabled in one spot while working perfectly a micrometre away.
What follows is a running score rather than a breach. Tiny pits nucleate constantly on stainless steel in salt water and almost all repassivate within moments; the current traces are full of small failures healing. A pit turns permanent only when its own chemistry becomes self-sustaining — when the solution trapped inside goes acidic and chloride-rich enough that no film can form there at all. Then the hole that was one of thousands becomes the only one that matters. Healing also draws on a stock: the alloy immediately beneath the breach must still be chromium-rich enough to rebuild from. Strip that often enough in one place and the surface keeps its shine everywhere except where it can no longer recover.
Anodised aluminium is a different case, worth not confusing with this one. Aluminium grows its own native film, two or three nanometres, by the same spontaneous route; anodising abandons spontaneity. The part goes into an acid bath under current and the oxide is driven to microns — a thousand times thicker, hard, dyeable, manufactured rather than found. It is a coating, and breaching it is permanent.
The archive has healing that runs down and healing that runs on. No. 44’s concrete carries dormant bacteria that wake once, seal one crack, and are spent. No. 45’s cell digests its own components on a schedule it sets itself. This is neither. Nothing here is stored, scheduled, or decided. The film is simply the ash of an ongoing fire, and it lasts exactly as long as the fire does.
≈
One loop I’m watching
Next: the carbon-14 in the air you just breathed. Cosmic rays manufacture it continuously in the upper atmosphere; radioactive decay destroys it continuously everywhere; and the ratio every living thing carries is just the level where those two rates happen to meet. Radiocarbon dating works because that level is standing rather than fixed — which also means it drifts with the sun, and that we shoved it twice in one century, once by burning coal and once by testing bombs.