Formal Proof of an Anisotropic Calderón Uniqueness Theorem
OpenAI announced uniqueness for the smooth anisotropic Calderón inverse problem but did not release a Lean proof of it. The authors seek a machine-checked proof of a restricted case: smooth metrics on the 3D unit ball that are Euclidean outside the ball, with full boundary measurements.
Boundary measurements are encoded through the minimized Dirichlet energy (the DN-map quadratic form), so no separately defined DN map is needed. The theorem is stated as a comparator challenge (OAI.AnisotropicCalderon.main), and an AI-assisted autoformalization (OpenAI Codex, Claude Code) produced a solution that uses Mathlib and 96 Tau Ceti modules. The proof is not a line-by-line translation of the manuscript, and its prerequisites are formalized as well.
The Lean proof shows that equal Dirichlet energies imply the metrics agree up to a smooth diffeomorphism fixing the exterior. The comparator replays the proof in Lean's kernel and confirms that only the standard axioms (propext, Classical.choice, Quot.sound) are used. The repository is public.
