Cohomologie des modules de dimension finie sur la super algèbre de Lie \(\mathfrak{osp}(3,2)\) (Cohomology of finite dimensional modules on the Lie superalgebra \(\mathfrak{osp}(3,2)\)) (Q1867325)

It is well known that complex simple Lie superalgebras have in general non-semisimple finite dimensional representations (i.e. which do not split into a direct sum of simple modules). \textit{V. G. Kac} [Lect. Notes Math. 676, 597-626 (1978; Zbl 0388.17002)] classified modules which split in every finite dimensional representation (i.e. which are a direct summand in every finite dimensional representation in which they appear as a sub- or a factor module) as typical, otherwise as atypical. Let \({\mathfrak g}\) denote the basic classical Lie superalgebra \({\mathfrak o}{\mathfrak s}{\mathfrak p}(3,2)\) over \({\mathbb C}\). Then \({\mathfrak g}={\mathfrak g}_0\oplus{\mathfrak g}_1\) with \({\mathfrak g}_0={\mathfrak s}{\mathfrak l}_2\times{\mathfrak s}{\mathfrak l}_2\) and \({\mathfrak g}_1=V_0\otimes V_1\) where \(V_0\) (resp. \(V_1\)) denotes the standard representation of \({\mathfrak o}(3)\) (resp. of \({\mathfrak s}{\mathfrak p}(2)\)) (note the identification \({\mathfrak o}(3)\cong{\mathfrak s}{\mathfrak p}(2)\cong {\mathfrak s}{\mathfrak l}_2\)). Choosing a Cartan subalgebra \({\mathfrak h}\) of \({\mathfrak g}_0\), we have a root system involving two linearly independent roots \(\epsilon\) and \(\eta\). It follows from Kac's work [loc. cit.] that a \({\mathfrak g}\)-module of highest weight \(\lambda\) is atypical iff it is trivial or \(\lambda=(l+1)\epsilon+l\eta\) with \(l\in{\mathbb N}\). The first part of the article under review treats an explicit construction of the projective cover \(P_l\) of the atypical simple \({\mathfrak g}\)-module \(M_l\) (notation shifted with respect to \textit{J. Germoni} [Bol. Acad. Nac. Cienc., Córdoba 65, 147-163 (2000; Zbl 1032.17006)]) of heighest weight \((l+1)\epsilon+l\eta\) (and of the trivial one dimensional module) as the image of a coboundary operator in some Koszul complex associated to the standard representation \(V=V_0\oplus V_1\) of \({\mathfrak g}\). The author uses elementary methods from representation theory as opposed to the abstract methods (like Zuckerman functors, due to Santos in the super setting, and quivers) used by Germoni in (loc. cit.). Decisively, the former uses the decomposition of a typical irreducible highest weight \({\mathfrak g}\)-module into irreducible highest weight \({\mathfrak g}_0\)-modules due to Kac (loc. cit.). In the second part, the author computes the cohomology of \({\mathfrak g}\) with values in an atypical simple module and with values in an indecomposable module which is neither simple nor projective. This encompasses all cohomology with values in finite dimensional modules, as the one with values in a typical module is trivial (the reason is the semisimplicity, like in the case of a finite dimensional module over a semi-simple Lie algebra). Drawing on earlier work [\textit{C. Gruson}, Ann. Inst. Fourier 47, 531-553 (1997; Zbl 0974.17024)], the algebra \(H^*({\mathfrak g},{\mathbb C})=:H\) is known. \(H^*({\mathfrak g},M_l)\) for \(l>0\) turns out to be a free \(H\)-module with one generator in degree \(l\) and \(l+1\); \(H^*({\mathfrak g},M_0)\) as well, but with only one generator in degree 2. The proof uses the computation of \(H^*({\mathfrak g},P_l)\) and the fact that \(P_l\) is a non-trivial extension of the non-trivial extensions of \(M_l\) and \(M_{l+1}\), and \(M_{l+1}\) and \(M_{l+2}\); the corresponding spectral sequence leads to the conclusion. On the other hand, let \(M\) denote an indecomposable module which is neither simple nor projective. First, the author shows that \(M\) is an extension of its largest submodule \(S\), which is the direct sum of simple \({\mathfrak g}\)-modules by its largest quotient \(Q\), which is the direct sum of simple \({\mathfrak g}\)-modules: \[ 0\to S\to M\to Q\to 0. \] The cohomologies \(H^*({\mathfrak g},M)\), \(H^*({\mathfrak g},S)\) and \(H^*({\mathfrak g},Q)\) are free \(H\)-modules determined by exponents \(m_i\) (resp. \(s_i\) and \(q_i\)) defined by \(H^*({\mathfrak g},M)=\bigoplus_{i\geq 0}(H[-i])^{m_i}\) (where \(H[i]\) is the \(i\)th suspension of \(H\)), resp. \(H^*({\mathfrak g},S)=\bigoplus_{i\geq 0}(H[-i])^{s_i}\) and \(H^*({\mathfrak g},Q)=\bigoplus_{i\geq 0}(H[-i])^{q_i}\). Then she uses the above short exact sequence and the knowledge of the extension quiver of the \(M_i\)'s (due to Germoni (loc. cit.)) to describe the \(m_i\) in terms of \(s_i\) and \(q_i\): \[ m_i = (s_i-q_{i-1})_++(q_i-s_{i+1})_+, \] where \(()_+\) denotes the positive part.