Corrosion is usually imagined as surface erosion — material dissolving layer by layer, measurable by weight loss. But molten-salt corrosion in metals can thread through grain boundaries in a single dimension, forming a connected path through the metal's depth while consuming very little material.
This “wormhole” corrosion travels along grain boundaries — the interfaces between crystal grains in a metal — creating a connected path from one surface to the other. The penetration rate per unit of corroded mass is extraordinarily high because the attack is localized to boundaries just nanometers wide. If it persists, a container could be breached by corrosion that a standard mass-loss measurement would classify as small.
The failure mode matters for next-generation molten salt nuclear reactors, which depend on metal containment resisting hot, chemically aggressive salt for decades. Traditional corrosion monitoring — weighing coupons, measuring thickness — would not detect a wormhole penetration until after the breach. The damage is structurally catastrophic but volumetrically trivial.
The engineering insight: the most dangerous corrosion is not the kind that consumes the most material but the kind that finds the thinnest continuous path. A crack needs only to connect, not to widen. The topology of the damage matters more than its volume — a lesson that extends to any system where failure is about connectivity, not magnitude.
Correction (2026-11-04): I checked this essay against the paper it describes, “One dimensional wormhole corrosion in metals” (Nature Communications, February 2023; I read the arXiv version, 2203.16312), which studies a nickel-20% chromium alloy in molten fluoride salt. It supports the core idea: tunnels along grain boundaries, a connected percolating network, and a very high ratio of penetration depth to mass lost. It does not show a wall breached through its full thickness. The paper says the implication for reactors follows “if the 1D form of corrosion occurs and persists”, and that the voids can weaken a part by grain-boundary embrittlement even before complete penetration. My lines that such corrosion bridges both sides of a containment wall, and that standard monitoring “would not detect” it before a breach, were stated more firmly than the paper supports; I have softened the first two and the last is my inference. The essay also gave no source or date: this is a 2023 result in one alloy.