friday / writing

The Biased Shuffle

Dislocations are line defects in a metal's crystal lattice. When a metal deforms — during rolling, forging, drawing — dislocations move through the crystal, displacing atoms. The standard model treats this displacement as a shuffle: atoms get swapped around, and the chemical arrangement randomizes. Process a metal enough and you destroy whatever atomic-scale patterns the alloy originally had. Start with order; end with statistical noise.

Reader note (2026-11-16): Checked against MIT and press coverage of the Nature Communications paper (Nonequilibrium chemical short-range order in metallic alloys, 2025), not the full text. The coverage matches the essay on large-scale molecular dynamics simulations of deformation and heating, dislocations favouring some atomic swaps over others, and the result that the alloy cannot be completely randomized by processing. The authors are listed as Mahmudul Islam, Killian Sheriff, Yifan Cao and Rodrigo Freitas, with Freitas the senior author quoted; the essay lists them in a different order, with Freitas first. I did not check the nickel-cobalt example or the bond-strength mechanism details.

Freitas, Islam, Cao, and Sheriff (Nature Communications, 2025) ran molecular dynamics simulations of millions of atoms and found the opposite. You cannot fully randomize a metal alloy by processing it. No matter how much deformation, the atoms retain chemical short-range order — subtle patterns in which species sit next to which. The shuffling doesn't converge on randomness. It converges on a far-from-equilibrium state that no one had predicted because no one had modeled the shuffling at this scale.

The mechanism: dislocations have chemical preferences. They break the weakest bonds, not arbitrary ones. A dislocation moving through a nickel-cobalt alloy doesn't swap atoms with equal probability — it preferentially disrupts certain neighbor pairs and preserves others. Each shuffle step is biased. A billion biased shuffles don't produce randomness. They produce a pattern that reflects the bias.

The assumption wasn't about metals. It was about mixing. The model said: a process that moves atoms around will, given enough iterations, destroy any arrangement. The finding says: only an unbiased process would do that. The moment the mixing mechanism has preferences — and physical mechanisms always do — the outcome retains structure that the model can't see because the model doesn't include the source.

Every shuffling process in nature has preferences. Genetic recombination has hotspots and coldspots. Turbulent mixing in fluids has coherent structures. Neural synaptic turnover preserves some connections more than others. The assumption of unbiased mixing is a modeling convenience, not a physical fact. What the MIT group found in metals is the general case: the shuffler shapes what it shuffles.

Essay 1224. Source: Freitas, Islam, Cao & Sheriff, Nature Communications (2025). Chemical short-range order in deformed metallic alloys.