Orbits in degenerate compactifications of symmetric varieties (Q1024625)

The author proves that the normalization map of some completions of a symmetric space is an homeomorphism (in characteristic zero). Let \(G\) be a semisimple and simply connected group over an algebraically closed field and let \(\sigma\) be an involution of \(G\). Let \(H^{0}\) be the subgroup of \(\sigma\)-fixed points and let \(H\) be its normalizer. The variety \(G/H\) is called a symmetric space (of adjoint type). For example, any semisimple (adjoint) group \(\dot{G}\) is a \(\dot{G}\times\dot{G}\)-symmetric space (and the involution is such that \(\sigma(x,y)=(y,x)\)). A wonderful completion of any fixed symmetric space (of adjoint type) has been defined by \textit{C. De Concini} and \textit{C. Procesi} [in: Invariant theory, Proc. 1st 1982 Sess. C.I.M.E., Montecatini/Italy, Lect. Notes Math. 996, 1--44 (1983; Zbl 0581.14041)]. More generally, normal completions of a symmetric space (i.e., the symmetric varieties) are been studied by many authors, remarkably by Brion, Luna and Vust [see \textit{T. Vust}, Ann. Sc. Norm. Super. Pisa, Cl. Sci., IV. Ser. 17, No. 2, 165--195 (1990; Zbl 0728.14041)] for an application of the general theory of spherical varieties to the symmetric varieties). But there are only a few works on the completions which are non-normal (see for example [\textit{W. Bruns} and \textit{A. Conca}, J. Algebra 322, No. 9, 2927--2949 (2009; Zbl 1201.14036)] for the case of \(PGL(n)\) and [\textit{D. A. Timashev}, Sb. Math. 194, No. 4, 589--616 (2003); translation from Mat. Sb. 194, No. 4, 119--146 (2003; Zbl 1074.14043)] for the case where \(G/H\) is isomorphic to a group). The author studies the following completions of a fixed \(G/H\). Let \(V\) be an irreducible \(G\)-representation which contains a non-zero vector \(h\) fixed by \(H^{0}\); then \(V\) is said to be spherical and the stabilizer of \([h]\in \mathbb{P}(V)\) is \(H\). The author studies the closure \(Z_{V}\) of \(G\cdot [h]\) in \(\mathbb{P}(V)\) and its normalization \(Y_{V}\). Remark that there are some rare cases of irreducible representations \(W\) which contain a non-zero vector \(v\) such that the stabilizer of \([v]\) is \(H\) but \(v\) is not fixed by \(H^{0}\). First, the author describes the orbits of \(Y_{V}\) and \(Z_{V}\), by using the orbit structure of the wonderful completion of \(G/H\). Indeed \(Y_{V}\) and \(Z_{V}\) are both dominated by \(X\) through morphisms, say \(q_{V}:X\rightarrow Z_{V}\) and \(p_{V}:X\rightarrow Y_{V}\), extending the identity automorphism of \(G/H\) (\(q_{V}\) is the composition of \(p_{V}\) with the normalization \(r_{V}:Y_{V}\rightarrow Z_{V}\)). In Theorem 9 are described conditions so that the images in \(Y_{V}\) (resp. in \(Z_{V}\)) of two \(G\)-orbits of \(X\) coincides. The author uses also a description of \(Y_{V}\) given by \textit{R. Chirivì, C. De Concini} and \textit{A. Maffei} [Tohoku Math. J. (2) 58, No. 4, 599--616 (2006; Zbl 1141.14033)] utilizing the wonderful completion \(X\). The orbits of \(X\) are indexed by the subsets of the set of simple restricted roots of \(G/H\). The author gives a similar description for \(Y_{V}\) and \(Z_{V}\). In particular, there is an one-to-one correspondence between the orbits of \(Y_{V}\) and the orbits of \(Z_{V}\). This last fact is due, in a much greater generality, to Timashev [loc. cit.]. Because of the cited result of Timashev, the Theorem 9 holds also in positive characteristic. Each \(G\)-orbit \(\mathcal{O}\) of \(X\) contains a distinguished point \(x_{\mathcal{O}}\) which is obtained as \(\lim_{t\rightarrow\infty }\gamma(t)[h]\), where \(\gamma\) is an appropriate one-parameter subgroup of \(G\). Moreover two \(G\) orbits of \(X\) have the same image in \(Y_{V}\) (resp. in \(Z_{V}\)) if and only if the associated points have the same images. In the second part of this work are described explicitly the stabilizers of \(p_{V}(x_{\mathcal{O}})\) and \(q_{V}(x_{\mathcal{O}})\) for each \(G\)-orbit \(\mathcal{O}\) of \(X\) (see Theorem 14). In particular, it is proved that these two subgroups of \(G\) coincide. Thus the normalization map is an homeomorphism. The description of such stabilizers is similar to the description of the \(Stab_{G}(x_{\mathcal{O}})\) given in [De Concini and Procesi, loc. cit.].