On infinite dimensional algebraic transformation groups (Q2249229): Difference between revisions

Let \(X\) be an irreducible algebraic variety over an algebraically closed field \(k\). The automorphism group \(\mathrm{Aut}(X)\) is not a finite-dimensional algebraic group in general. Nevertheless, one may introduce the families of elements in \(\mathrm{Aut}(X)\) parametrized by some irreducible \(k\)-variety \(T\), leading to the notions to injective, unital or algebraic families. If \(T\) is a connected algebraic \(k\)-group and the family \((\varphi_t)_{t \in T}\) gives rise to an algebraic action \(T \times X \to X\), we say \(\varphi_T\) is a connected algebraic subgroup of \(\mathrm{Aut}(X)\). The author proves the following fundamental result: Theorem. Let \(G\) be a subgroup of \(\mathrm{Aut}(X)\). Then \(G\) is a connected subgroup if and only if it is generated by a collection of unital algebraic families in \(\mathrm{Aut}(X)\). Using this, one deduces many applications on rational invariants and orbits under \(G\), including a generalization of Rosenlicht's theorem, whose finite-dimensional counterparts are well-known in invariant theory. Another application is a sufficient condition for unirationality. As corollaries, one shows that every Calogero-Moser space \[ \mathcal{C}_n := \left\{ (A,B) \in \mathrm{Mat}_n(\mathbb{C})^2 : \mathrm{rk}([A,B]+1) = 1 \right\} /\!/ \mathrm{PGL}_n(\mathbb{C}) \] is irreducible and unirational. So is the variety \[ Q_{m,n}(\tau) := \left\{ x \in \mathrm{Mat}_n(k)^m /\!/\mathrm{PGL}_n(k) : \;\mathrm{representation} \; \mathrm{type} = \tau \right\}, \] whenever \(m \geq 3\) and \(k\) is of characteristic zero.