On a property of special groups (Q5959060): Difference between revisions

The authors prove a converse of certain conjectures of Serre, and thus show those conjectures are sharp. Their converse reads as follows. Let \(G\) be an algebraic group defined over an algebraically closed field \(k\). Assume \(H^1(K,G)=\{1\}\) for some finitely generated field extension \(K\) of \(k\) of transcendence degree \(d\). Then (a) If \(d\geq 1\), then \(G\) is connected. (b) If \(d\geq 2\), then the universal covering map \(\overline G\to G/\text{Rad}(G)\) is purely inseparable. In particular, in characteristic zero \(G/\text{Rad}(G)\) is simply connected. (c) If \(d\geq 3\) and the set of torsion primes for \(G\) is not \(\{\text{char}(k)\}\), then \(G/\text{Rad}(G)\) is `special', that is it is a product of connected simply connected groups of type SL or Sp. The proof is by construction of a nontrivial class in \(H^1(K,G)\) for each \(G\) that does not satisfy the conclusion.