Involutions of reductive Lie algebras in positive characteristic (Q875245)

Let \(k\) be an algebraically closed field\(.\) Let \(\theta\) be an involutive automorphism of \(G\) with corresponding linear involution \(d\theta :\mathfrak{g}\rightarrow\mathfrak{g}\) where \(\mathfrak{g}=\)Lie\(\left( G\right) .\) Then \(\mathfrak{g=k\oplus p}\) where \(\mathfrak{k}=\left\{ x\in\mathfrak{g\,}|\,d\theta\left( x\right) =x\right\} \) and \(\mathfrak{p} =\left\{ x\in\mathfrak{g}\,|\,d\theta\left( x\right) =-x\right\} .\) If we let \(G^{\theta}=\left\{ g\in G\,|\,\theta\left( g\right) =g\right\} ,\) then \(G^{\theta}\) acts on \(\mathfrak{p},\) and while this action is understood when \(k\) has characteristic zero, it is not as well understood in the positive characteristic case. The characteristic zero case, as developed in [\textit{B. Kostant} and \textit{S. Rallis}, Am. J. Math. 93, 753--809 (1971; Zbl 0224.22013)] which use compactness properties and \(\mathfrak{sl}\left( 2\right) \)-triples which do not work in positive characteristic. Here, the author considers this action in the case where char\(\;k=p\) for a good prime \(p\), that is, when \(G\) is a reductive algebraic group with a root system \(\Phi\) such that when the longest element of each irreducible component of \(\Phi\) is expressed as a linear combination of a basis each coefficient is less than \(p\). Much of the paper proceeds as in the work cited above, however some adjustments must be made to the characteristic zero theory. Suppose furthermore that the derived subgroup is simply connected and that there exists a symmetric \(G\)-invariant non-degenerate bilinear form \(\mathfrak{g\times g}\rightarrow k\). It is proved that this bilinear form can be chosen to be \(\theta\)-equivariant as well. The notion of \(\mathfrak{sl} \left( 2\right) \)-triples is replaced by associated cocharacters, which allows us to compute the number of irreducible components of the variety \(\mathcal{N}\) of nilpotent elements of \(\mathfrak{p}\). This variety has a dense open orbit, which is also true for each fibre of the map \(\mathfrak{p\rightarrow p}//G^{\theta}.\) The corresponding statement for \(G\), a conjecture of Richardson, is shown to be false.