Algebraic magnetism, formally verified and extended: a survey
Arnaud Mayeux
math.AG
Sep 4, 2025 · v4
TL;DR
Formalizes three papers on algebraic magnetism and new results in Lean 4, in a 193k-line library built on Mathlib, AxiomMath Stacks and Tau Ceti.
Abstract
We formalize in the Lean proof assistant the three papers on algebraic magnetism published in Semigroup Forum: "Algebraic magnetism", "Algebraic magnetism invariants of self-actions of diagonalizable monoid schemes" and "Algebraic magnetism: $X$-products of attractors via $\mathbb{F}_1$-geometry". We also prove further results and comment on the relation with other work in the literature. The description of this formalization and of these results forms in itself a survey of the theory, from the perspective opened by the formalization. The first main mathematical contribution consists of a Bialynicki-Birula decomposition for faces and of proofs of the existence of strongly fixed-point-reflecting atlases, of the finiteness of pure magnets and of the vanishing of the higher monoid cohomology of $\mathrm{A}(M)$ with quasi-coherent coefficients, three conjectures of "Algebraic magnetism". The case of algebraic stacks, where the analogue of the existence of fixed-point-reflecting atlases fails, is treated in joint work with Sylvain Brochard, in which the finiteness of pure magnets was first proved privately, by a comparable method. The second main contribution, for split reductive groups, is a description of the pure magnets and of the join-irreducible ones: besides the classical roots, they are the join-irreducible pure magnets equal to $(1-w)M$ for some $w$ in the Weyl group, which encode the components of attractors; a prime $p$ is a torsion prime of the group, in the sense of Steinberg, if and only if some join-irreducible Weyl magnet $(1-w)M$ has $w$ of $p$-power order.
Problem
Algebraic magnetism studies actions of diagonalizable monoid schemes on algebraic spaces through their attractors. Its three foundational papers left several conjectures open, and their content had not been machine-checked.
Approach
Every definition and statement of the three Semigroup Forum papers, plus new results, is formalized in Lean 4. The work builds on Mathlib for algebra, on AxiomMath Stacks for the big fppf site and algebraic spaces, and on Tau Ceti. Each statement in the text is paired with the names of the Lean declarations that formalize it. The section on split reductive groups (Section 15) is only partially formalized, because the needed structure theory of reductive groups is not yet available in these libraries.
Results
The library has 656 modules, 193,193 lines and 7,449 theorems, depending only on the standard axioms. It proves three conjectures: the existence of strongly fixed-point-reflecting atlases, the finiteness of pure magnets, and the vanishing of higher monoid cohomology with quasi-coherent coefficients; it also gives a Bialynicki-Birula decomposition for faces. For split reductive groups it describes the pure and join-irreducible magnets and relates them to torsion primes; the Table 1 rows up to rank 3 are checked by the Lean kernel.