Recovering a group from few orbits (Q6858023)

Let \(V\) be a finite-dimensional Hilbert space over \(\mathbb{F}\in\{\mathbb{R},\mathbb{C}\}\) and let \(G\le \operatorname{Aut}(V)\) be an unknown finite group of automorphisms. The paper studies how many generic \(G\)-orbits in \(V\), observed as unstructured subsets, suffice to recover \(G\) up to isomorphism (abstract recovery) and as a concrete subgroup of \(\operatorname{Aut}(V)\) (concrete recovery).\N\NThe Gram graph of a finite set \(S\subset V\) is the complete directed graph on vertex set \(S\) in which the edge \(s\to t\) is labeled by the inner product \(\langle s,t\rangle\). For abstract recovery, the main complex result (Theorem 2) shows that one generic orbit suffices: for generic \(v\in V\) the isomorphism class of the Gram graph of \(Gv\) determines \(G\) up to isomorphism and the canonical map \(G\to \operatorname{Aut}(Gv)\) is an isomorphism. In the real case, two generic orbits suffice (Theorem 3), while the analogous one-orbit statement remains open as a conjecture.\N\NFor concrete recovery, the authors first show that, for generic \((v_1,\dots,v_k)\in V^k\), the canonical map \(G\to \operatorname{Aut}(Gv_1\cup\cdots\cup Gv_k)\) is an isomorphism provided \(k\ge 1\) if \(\mathbb{F}=\mathbb{C}\) and \(k\ge 2\) if \(\mathbb{F}=\mathbb{R}\) (Theorem~4). They then give sharp representation-theoretic criteria ensuring that \(k\) generic orbits span a subspace of codimension \(<r\), where \(r\) is the dimension of the smallest nontrivial representation of \(G\) (Theorem 5), using the regular representation and multiplicity bounds. Corollary 1 combines these ingredients: if \(k\) satisfies both the orbit-number condition from Theorem 4 and the spanning criterion from Theorem 5, then \(k\) generic orbits determine \(G\) as a concrete subgroup of \(\operatorname{Aut}(V)\); if the spanning criterion fails, then no choice of \(k\) orbits determines \(G\).