The separating variety for matrix invariants
Jonathan Elmer · 2025
Read The separating variety for matrix invariants by Jonathan Elmer (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-13865.txt.
Introduction
Jonathan Elmer, writing alone in 2025, studied a geometric object that records when matrix data cannot be told apart by invariant measures. A group acts on sets of matrices by a shared change of basis. Pairs of sets that every invariant scores the same form the separating variety, whose shape limits how small a telling set of invariants can be.
Elmer links the parts of this variety to a purely ordered list built from counts. He gives its size for all cases and counts its parts by depth in key ranges. In several ranges the result shows that no single polynomial, and no simple surface set, can separate all pairs. The paper also settles a past conjecture for two quiver settings.
Catalog record
| CREATOR | Jonathan Elmer |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 97 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-13865
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The separating variety for matrix invariants Jonathan Elmer Abstract . Let G G be a linear algebraic group defined over an algebraically closed field 𝕜 \Bbbk , and let 𝒱 \mathcal{V} be a vector space on which G G acts linearly. Introduced in [ 29 ] , the separating variety 𝒮 G , 𝒱 \mathcal{S}_{G,\mathcal{V}} is the subvariety of 𝒱 2 \mathcal{V}^{2} consisting of pairs of points indistinguishable by invariant polynomials in 𝕜 [ 𝒱 ] G {\Bbbk[\mathcal{V}]^{G}} . Its geometry places restrictions on the existence of small separating sets, i.e. sets of invariants which distinguish the same points as the full algebra of invariants. It is well known that the cardinality of a separating set for 𝕜 [ 𝒱 ] G {\Bbbk[\mathcal{V}]^{G}} is bounded below by dim ( 𝕜 [ 𝒱 ] G ) \dim({\Bbbk[\mathcal{V}]^{G}}) . A separating set of cardinality dim ( 𝕜 [ 𝒱 ] G ) \dim({\Bbbk[\mathcal{V}]^{G}}) is called a polynomial separating set, and a separating set of cardinality dim ( 𝕜 [ 𝒱 ] G ) + 1 \dim({\Bbbk[\mathcal{V}]^{G}})+1 is called a hypersurface separating set. The purpose of this article is to study the separating variety in the important special case where G = GL p ( ℂ ) G=\operatorname{GL}_{p}(\mathbb{C}) acts on the set 𝒱 = ℳ p n \mathcal{V}={\mathcal{M}}_{p}^{n} of n n -tuples of p × p p\times p matrices by simultaneous conjugation. We define a purely combinatorial poset, 𝒫 p , n \mathcal{P}_{p,n} , whose maximal elements are in 1-1 correspondence with the irreducible components of 𝒮 G , 𝒱 \mathcal{S}_{G,\mathcal{V}} . We show that 𝒮 G , 𝒱 \mathcal{S}_{G,\mathcal{V}} is a variety of dimension ( n + 1 ) p 2 − 1 (n+1)p^{2}-1 , and determine its subdimension (i.e. the dimension of its smallest component) for all n n and p p . In particular we show the subdimension is ( n + 1 ) p 2 − p (n+1)p^{2}-p if n ≥ 3 n\geq 3 , or n ≥ 2 n\geq 2 and p ≥ 4 p\geq 4 . In the case n ≥ 3 n\geq 3 , we give a formula for the number of components of given codimension in 𝒮 G , 𝒱 \mathcal{S}_{G,\mathcal{V}} . We give explicit decompositions of 𝒮 G , 𝒱 \mathcal{S}_{G,\mathcal{V}} for all n n where p = 2 , 3 p=2,3 or 4 4 . Our results in particular show that when n ≥ 2 n\geq 2 and p ≥ 4 p\geq 4 , or n ≥ 3 n\geq 3 and p = 3 p=3 , ℂ [ 𝒱 ] G \mathbb{C}[\mathcal{V}]^{G} does not contain a polynomial or hypersurface separating set. It was proven in [ 20 ] that the same is true if n ≥ 4 n\geq 4 and p = 2 p=2 . The author made a conjecture in [ 19 ] generalising the Skronowski-Weyman theorem for representations of quivers. The results of this paper prove that conjecture in two important special cases: for the quiver with one vertex and an arbitrary number, n n , of loops, and for the quiver with two vertices and n n arrows between them. 2020 Mathematics Subject Classification 16G20, 13A50, 05E40, 14R20 1. Background 1.1. Representations of quivers A quiver is a quadruple Q=(Q_{v},Q_{a},t,h) , consisting of two ordered sets Q_{v} (vertices) and Q_{a} (arrows), along with two functions t,h:Q_{a}\rightarrow Q_{v} (tail and head respectively). It is usually visualised as a directed graph with a node for each element of Q_{v} , and for each a\in Q_{a} a directed edge leading from t(a) to h(a) . We will deal only with finite quivers, so we may choose an order on the vertices and on the arrows, setting Q_{v}=\{x_{1},\ldots,x_{k}\} and Q_{a}=\{a_{1},\ldots,a_{n}\} . Let \Bbbk be any field. A representation V of the quiver Q over \Bbbk is an assignment to each vertex x\in Q_{v} of a vector space V(x) , and to each arrow a\in Q_{a} of a linear map V(a):V(t(a))\rightarrow V(h(a)) . We write \alpha=(\alpha(x):x\in Q_{v}) where \alpha(x)=\dim(V(x)) for all x\in Q_{v} ; this is called the dimension vector of V . A homomorphism \phi:V\rightarrow W between representations V,W of Q is a collection of linear maps (\phi(x):x\in Q_{v}) such that \phi(x):V(x)\rightarrow W(x) for all x\in Q_{v} , and for all a\in Q_{a} the following diagram commutes: V(t(a)) V(h(a)) W(t(a)) W(h(a)) V(a) W(a) \phi(t(a)) \phi(h(a)) We write \operatorname{Hom}_{Q}(V,W) for the set of homomorphisms from V to W , which is naturally a \Bbbk -vector space. It is clear that the composition of two homomorphisms is again a homomorphism, and there is an obvious identity homomorphism \operatorname{id}:V\rightarrow V . Thus, we have a category \operatorname{Rep}_{\Bbbk}(Q) of representations of Q over \Bbbk . Denote by \operatorname{Rep}_{\Bbbk}(Q,\alpha) the set of representations of Q with dimension vector \alpha . To ease notation we will drop the \Bbbk where the field is known. A homomorphism \phi:V\rightarrow W is an isomorphism if \phi(x) is an isomorphism for each x\in Q_{v} . We write V\cong W if V and W are isomorphic. We say that W is a subrepresentation of V if W(x)\leq V(x) for all x\in Q_{v} and the collection of inclusion maps i(x):W(x)\hookrightarrow V(x) is a homomorphism W\rightarrow V . We write W\leq V if W is a subrepresentation of V . It is clear that V\leq V and 0\leq V , where 0 denotes the representation of Q in which 0(x) is the zero space for all x\in Q_{v} and 0(a) is the zero map for all a\in Q_{a} . We say V is simple if it has no subrepresentations other than these. Further, if W\leq V then we may define a quotient representation V/W as follows: (V/W)(x)=V(x)/W(x) for all x\in Q_{v} and (V/W)(a) is the map V(t(a))/W(t(a))\rightarrow(V(h(a))/W(h(a)) induced by V(a) (the commuting of the diagram above ensures this is well-defined). There is a natural notion of direct sum for representations of a given quiver: if V and W are representations of Q over \Bbbk with dimension vectors \alpha and \beta respectively, then V\oplus W is the representation of Q over \Bbbk with dimension vector \alpha+\beta\in\mathbb{Z}^{k} defined by (V\oplus W)(x)=V(x)\oplus W(x) for all x\in Q_{v} and (V\oplus W)(a)=V(a)\oplus W(a) for all a\in Q_{a} . One can show that this notion of direct sum makes \operatorname{Rep}_{\Bbbk}(Q) into a \Bbbk -linear abelian category. If V is isomorphic to a direct sum of simple representations we say V is semisimple . A representation is said to be indecomposable if it is not isomorphic to the direct sum of two non-trivial representations. The Krull-Remak-Schmidt Theorem [ 6 , Theorem 1.7.4] states that every representation of Q may be written as a direct sum of indecomposable representations and this decomposition is unique up to reordering the summands. Thus, describing the representations of Q up to isomorphism is reduced to the problem of describing the isomorphism classes of indecomposable representations. The quiver Q is said to have: (1) finite representation type if Q has only finitely many isomorphism classes of indecomposable representations; (2) tame representation type if the indecomposable representations of Q in each dimension vector up to isomorphism occur in finitely many one-parameter families; (3) wild representation type otherwise. Let V\in\operatorname{Rep}(Q,\alpha) . If V