The relative Hochschild-Serre spectral sequence and the Belkale-Kumar product (Q2849032)

For a complex semisimple algebraic group with parabolic subgroup \(P\) of \(G\), let \(m=\dim(H^2(G/P))\). For each \(t\in\mathbb C^m\), Belkale and Kumar defined a product \(\odot_t\). This product degenerates the usual cup product on \(H^\ast(G/P)\), and it gives applications to the eigenvalue problem and to the problem of finding \(G\)-invariants of the tensor products of representations. Previously, the authors have given a new construction of this product, and they have proved that \((H^\ast(G/P,\mathbb C),\odot_t)\) is isomorphic to a relative Lie algebra cohomology ring \(H^\ast(\mathfrak g_t,\mathfrak l_\Delta)\), where \(\mathfrak l_\Delta\) are subalgebras of \(\mathfrak g\times\mathfrak g\). The present article focuses on \((H^\ast(G/P),\odot_t)\). Let \(B\subset P\) be a Borel subgroup, let \(H\subset B\) be a Cartan subgroup, and let \(\alpha_1,\dots,\alpha_m\) be the simple roots with respect to \(B\). Let \(L\) be the Levi factor of \(P\) containing \(H\), let \(\mathfrak l\) be the Lie algebra of \(L\), and let the simple roots of \(\mathfrak l\) be \(I=\{\alpha_{m+1},\dots,\alpha_n\}\) where \(\alpha_1,\dots,\alpha_m\) are roots of the nilradical \(\mathfrak u\) of \(P\). For \(t=(t_1,\dots,t_m)\in\mathbb C^m\), \(J(t)=\{1\leq q\leq m:t_q\neq 0\}\), \(K=J(t)\cup I\). \(\mathfrak l_K\) denotes the Levi subalgebra generated by the Lie algebra \(\mathfrak h\) of \(H\) and the root spaces \(\mathfrak g_{\pm\alpha_i}\), \(i\in K\), \(L_K\) denotes the corresponding subgroup, and \(P_K=BL_K\) denotes the corresponding standard parabolic.NEWLINENEWLINEThe main result is perfectly and precisely given as follows (direct quote): {Theorem 1.1.} For parabolic subgroups \(P\subset P_K\) of \(G\), with \(P_K\) determined by \(t\in\mathbb C^m\) as above,NEWLINENEWLINE(1) \(H^\ast(P_k/P)\) is isomorphic to a graded subalgebra \(A\) of \((H^\ast(G/P),\odot_t)\).NEWLINENEWLINE(2) The ring \((H^\ast(G/P_K),\odot_0)\cong (H^\ast(G/P),\odot_t)/I_+\), where \(I_+\) is the ideal of \((H^\ast(G/P),\odot_t)\) generated by positive degree elements of \(A\). \vskip0,2cm So \((H^\ast(G/P),\odot_t)\) has a classical part which is the usual cohomology ring, with the associated quotient given by the degenerate Belkale-Kumar product. The theorem is proved by applying the Hochschild-Serre spectral sequence in relative Lie algebra cohomology, that is, by using the relative Lie algebra cohomology description of the product. There is a need to show that the spectral sequence degenerates at the \(E_2\)-term, to compute the edge morphisms, and to determine the product structure on the \(E_2\)-term. Thus the main content of this article is to fulfill this need, that is carrying out these computations using the original approach of Hochschild and Serre. This approach is generalized to the the relative setting, with a main new point: A Lie algebra \(\mathfrak g\) is given, together with an ideal \(I\) and a subalgebra \(\mathfrak k\) which is reductive in \(\mathfrak g\). To construct the spectral sequence, there is a need for an action of \(\mathfrak g/I\) on the relative cohomology group \(H^\ast((I,I\cap\mathfrak k,M)\), \(M\) a \(\mathfrak g\)-module. If \(I\cap\mathfrak k\) is nonzero, the Lie algebra \(\mathfrak g/I\) acts in a special way on the space of cochains \(C^\ast(I,I\cap\mathfrak k)\). The authors construct a an action on the cohomology group by a formula involving cochains. Then they verify that this yields the \(d_1\) differential in the spectral sequence . The confirmation of the last fact is the main technical complication of the article.NEWLINENEWLINEThe authors give a nice introduction to Lie algebra cohomology, and they generalize this to the relative setting. This involves explicit formulas for the differentials, needed in the relative computations. Then they give a rather complete module on spectral sequences, including the computation of general edge morphisms. Finally they apply the spectral sequences on the relative Hochscild Serre cohomology. This means that they give explicit conditions of convergence and descriptions of the edge morphisms. Finally, the article apply the Hochschild-Serre spectral sequence to prove the main theorem, and this proves that the cohomology of a generalized flag variety, equipped with the Belkale-Kumar product, has a structure analogous to the cohomology of a fibre bundle.NEWLINENEWLINEThe article also includes some interesting theorems on the structure of cohomology rings in the relative case. The article is detailed and precise, and in addition to giving nice results, it is also a nice introduction to the techniques of Lie algebra cohomology rings.