Chern classes of reductive groups and an adjunction formula (Q877540): Difference between revisions

Let \(G\) be a connected complex reductive group of dimension \(n\) and rank \(k\) (= dimension of a maximal torus). \(G\times G\) acts on \(G\) by \((g_1,g_2)\cdot g:=g_1gg_2^{-1}\). For each representation \(\pi :G\rightarrow \text{GL}(V)\) with \(d:=\dim V<\infty \), one can endow the trivial vector bundle \(E=G\times V\) over \(G\) with the \(G\times G\)-action given by \((g_1,g_2)\cdot (g,v):=(g_1gg_2^{-1},\pi (g_1)(v))\) and in this way one gets all the \(G\times G\)-equivariant vector bundles over \(G\) (up to equivalence). Let \(\Gamma (E)\) denote the space of global sections of \(E\) which can be written as differences of left- and right-invariant sections. \(\Gamma (E)\) is the image of the linear map associating to \((v_1,v_2)\in V\oplus V\) the global section \(G\ni g\mapsto \pi (g)(v_1)-v_2\) of \(E\). Let \(d(\pi )\) be the dimension of a principal orbit of \(G\) in \(V\) (there exists an open dense \(G\)-invariant subset of \(V\) such that the stabilizers of any two elements of it are conjugate, hence their orbits are isomorphic; such an orbit is called principal). The author shows that the degeneracy locus \(S_i(E)\) of \(d-i+1\) generic elements of \(\Gamma (E)\) is empty for \(i>d(\pi )\) and has codimension \(i\) in \(G\) for \(i\leq d(\pi )\). She also shows that the class \([S_i(E)]\) of \(S_i(E)\) in the ring of conditions \(C^{\ast}(G)\) of \(G\) [introduced by \textit{C. de Concini} and \textit{C. Procesi}, in: Algebraic groups an related topics (Kyoto/Nagoya, 1983), Adv. Stud. Pure Math. 6, 481--513 (1985; Zbl 0596.14041)] does not depend on the choice of the generic elements of \(\Gamma (E)\). These classes are analogues of the classical Chern classes (with values in the cohomology ring) in this noncompact context. Assume, now, that \(E\) is the tangent bundle \(\text{T}G\) of \(G\), corresponding to the adjoint representation \(\text{Ad}:G\rightarrow \text{GL}(\mathfrak{g})\). One has \(d(\text{Ad})=n-k\). Let \(S_i:=S_i(\text{T}G)\). The author shows that \([S_1]\) can be represented in \(C^{\ast}(G)\) as the doubled sum of the closures of all codimension 1 Bruhat cells in \(G\) and that \([S_{n-k}]=| W| [T]\), where \(T\) is a maximal torus and \(W\) is the Weyl group. Then, using a result of \textit{M. Brion} [J. Algebra 134, No. 1, 115--143 (1990; Zbl 0729.14038)] about the (classical) Chern classes of the tangent bundle of a regular compactification of \(G\), she obtains the main result of the paper, a kind of adjunction formula, asserting that if \({\pi}_i:G \rightarrow \text{GL}(V_i)\), \(i=1,\dots ,m\), are faithful representations and if \(H_i\) is the inverse image by \({\pi}_i\) of an affine hyperplane of \(\text{End}(V_i)\) such that the collection \(H_1,\dots ,H_m\) is ``generic'' in a precise sense, then the (topological) Euler characteristic of the complete intersection \(H_1\cap \dots \cap H_m\) is equal to the term of degree \(n\) in the expansion in \(C^{\ast}(G)\) of the intersection product : \[ (1+S_1+\dots +S_{n-k}){\prod}_{1\leq i\leq m}H_i(1+H_i)^{-1}. \] When \(m=n-1\), i.e., when the complete intersection is a curve, this formula combined with a formula of \textit{M. Brion} [Duke Math. J. 58, No. 2, 397--424 (1989; Zbl 0701.14052)] and \textit{B. Ya. Kazarnovskij} [Funct. Anal. Appl. 21, No. 4, 319--321 (1987; Zbl 0662.22014)] gives an explicit answer for the Euler characteristic and the genus of the curve. In a final section, the author explains how one can extend her results about the Chern classes of the tangent bundle to the case of arbitrary spherical homogeneous spaces.