On the normality of commuting scheme for general linear Lie algebra
Authors
Artan Sheshmani, Xiaopeng Xia, Beihui Yuan
Abstract
The commuting scheme $\mathfrak{C}^{d}_{\mathfrak{g}}$ for reductive Lie algebra $\mathfrak{g}$ over an algebraically closed field $\mathbb{K}$ is the subscheme of $\mathfrak{g}^{d}$ defined by quadratic equations, whose $\mathbb{K}$-valued points are $d$-tuples of commuting elements in $\mathfrak{g}$ over $\mathbb{K}$. There is a long-standing conjecture that the commuting scheme $\mathfrak{C}^{d}_{\mathfrak{g}}$ is reduced. Moreover, a higher dimensional analog of Chevalley restriction conjecture was conjectured by Chen-Ngô. We show that the commuting scheme of $\mathfrak{C}^{2}_{\mathfrak{g}l_{n}}$ is Cohen-Macaulay and normal. As a corollary, we prove a 2-dimensional Chevalley restriction theorem for general linear group in positive characteristic.
Concepts
The Big Picture
Imagine studying all the ways two people in a room can simultaneously agree on a direction to walk. Now replace “people” with matrices (grids of numbers encoding rotations, reflections, and other geometric operations). Unlike ordinary multiplication (where 3 × 4 = 4 × 3), matrices usually don’t commute: rotating a shape then flipping it gives a different result than flipping then rotating. When two matrices do commute (when order makes no difference), the set of all such pairs forms a geometric object called the commuting scheme.
The commuting scheme sits at the crossroads of representation theory, algebraic geometry, and mathematical physics. It appears naturally when studying Higgs bundles, packages of geometric data encoding how symmetry transformations vary across a curved surface, with deep ties to the Langlands program and string theory. Yet a fundamental question has lingered since the 1970s: is this space as well-behaved as its construction suggests? Does it harbor hidden redundancies or pathological singularities?
A new paper by Artan Sheshmani, Xiaopeng Xia, and Beihui Yuan delivers a decisive answer. They prove that the commuting scheme for pairs of n×n matrices is Cohen-Macaulay and normal, two stamps of geometric respectability meaning the space has no hidden extra structure, no unexpected pathologies at its edges, and that its defining equations fully capture its geometry without redundancy. Mathematicians have long conjectured this; now it’s proved in full generality.
Key Insight: The commuting scheme of pairs of matrices, for any n and any field including positive characteristic, is geometrically well-behaved, with no hidden structure, unexpected singularities, or redundant components. This resolves a cornerstone case of a decades-old open conjecture.
How It Works
The commuting scheme C²_{gl_n} is concrete: take all pairs of n×n matrices (X, Y) such that XY − YX = 0. This forms an algebraic variety (the solution set to a system of polynomial equations, just as a circle is the solution set to x² + y² = 1). The question is whether this variety is reduced (no phantom solutions counted multiple times), irreducible (connected, not a disjoint union of pieces), and normal (well-behaved near almost-smooth points).
The strategy is elegant: induction on n. Assume the result holds for (n−1)×(n−1) matrices and extend it to n×n. Bridging the inductive step requires a chain of intermediate algebraic objects, rings R̃(n), R₁(n), R′(n), and R₂(n), each encoding slightly different geometric information about how the commuting condition constrains matrix entries.
The key technical lemmas establish that Cohen-Macaulay properties propagate: well-behavedness in a smaller setting passes intact through algebraic transformations to the full problem. The argument runs roughly as follows:
- Show R(n−1) being Cohen-Macaulay and reduced forces certain localizations of R(n) to inherit those properties.
- Use a local-to-global principle: patch together local Cohen-Macaulay results to obtain the global statement.
- Establish normality, the hardest piece, by analyzing the singular locus and showing it has high codimension.
A crucial ingredient comes from an earlier result of Vaccarino: once the commuting scheme is known to be reduced, an isomorphism between two quotient spaces, the Chevalley restriction map Φ: t²//W → C²_g//G, follows automatically. This isomorphism says you can read off all G-invariant information about commuting matrices just from their eigenvalues, compressing n²+n² variables down to just 2n eigenvalue parameters.
The paper also handles positive characteristic fields (fields where arithmetic wraps around at a fixed prime, like clock arithmetic modulo 5, where 3 + 3 = 1). This is significantly harder than the characteristic-zero case, where classical tools apply freely. Earlier work by Vaccarino, Chen, and Ngô established the Chevalley restriction isomorphism in characteristic zero. This result extends it to all characteristics, closing a conspicuous gap.
Why It Matters
The commuting scheme is a foundational object in the geometric Langlands program (one of mathematics’ most ambitious unification projects, seeking deep correspondences between number theory, representation theory, and geometry, with connections to quantum field theory and string theory).
The Hitchin morphism extracts compact spectral summaries from spaces of Higgs bundles, much as eigenvalues summarize a matrix. It governs the structure of moduli spaces (parameter spaces where each point represents a distinct geometric object) of Higgs bundles on higher-dimensional varieties. This morphism depends fundamentally on whether the commuting scheme has the geometric properties proved here.
Concretely, the result guarantees a well-defined universal spectral data morphism sd: C²_g → t²//W for GL_n. This allows mathematicians and physicists to extract eigenvalue-like invariants from pairs of commuting fields in a gauge theory. Without normality and Cohen-Macaulayness, this map could fail to exist or behave pathologically. With it established, several downstream conjectures about Higgs bundles on surfaces become tractable.
Open questions remain. The Chen-Ngô conjecture in full generality, for d ≥ 3 commuting matrices and for groups beyond GL_n, is still unresolved. Reducedness for d ≥ 3 is unproved. But each result like this one maps the terrain and narrows the frontier.
Bottom Line: Sheshmani, Xia, and Yuan prove the commuting scheme for pairs of general linear matrices is Cohen-Macaulay and normal over any algebraically closed field, resolving a foundational case of a 50-year-old conjecture and unlocking the Chevalley restriction theorem in positive characteristic for GL_n.
IAIFI Research Highlights
This work bridges abstract algebraic geometry (commuting schemes and Cohen-Macaulay rings) to the Hitchin morphism and Higgs bundle theory central to the geometric Langlands program and mathematical physics.
Results like this inform the algebraic structures used in equivariant neural networks and symmetry-aware machine learning architectures, where the geometry of matrix-valued fields governs representational capacity.
The proof of the 2-dimensional Chevalley restriction theorem in positive characteristic resolves a key open problem in the mathematical framework underlying gauge theories and string-theoretic dualities.
Future work will seek to extend these normality results to d ≥ 3 commuting matrices and to other reductive groups; the full paper is available at [arXiv:2505.13013](https://arxiv.org/abs/2505.13013).