The \(p\)-torsion of the Farrell-Tate cohomology of the mapping class group \({\Gamma{}}_{p-1}\) (Q1196786)

Let \(S_ g\) be a closed, orientable surface of genus \(g\geq 2\). The mapping class group of \(S_ g\) is \(\Gamma_ g=\pi_ 0(\text{Diff}^ +(S_ g))\), where \(\text{Diff}^ +(S_ g)\) is the group of orientation preserving diffeomorphisms of \(S_ g\). The cohomology of these groups is a subject of great interest and many results are known, see the references of the author's Ph.D. thesis [Ohio State Univ. (1990)] for instance. The group \(\Gamma_ g\) has finite cohomological dimension \(4g- 5\) and so the Farrell-Tate cohomology groups \(\widehat {H}^ i(\Gamma_ g,\mathbb{Z})\) exist for all \(i\in \mathbb{Z}\) and equal the ordinary cohomology groups \(H^ i(\Gamma_ g,\mathbb{Z})\) for \(i>4g-5\). The author initiated a study of these groups in the his Ph.D. thesis [loc. cit.]. The Farrell- Tate groups are usually more easily computed than the ordinary groups, since the \(p\)-primary component of Farrell-Tate groups may be computed by calculating the \(\Gamma_ g\)- equivariant cohomology of the simplicial complex of elementary abelian \(p\)-subgroups of \(\Gamma_ g\). The simplest case occurs when all elementary abelian \(p\)-subgroups of \(\Gamma_ g\) have rank at most 1, for then the \(p\)-primary component of the Farrell-Tate cohomology is given by \[ \widehat{H}^*(\Gamma_ g,\mathbb{Z})_{(p)}\simeq\prod_{P\in{\mathcal P}}\widehat{H}^*(N(P),\mathbb{Z})_{(p)}, \] where \(\mathcal P\) is a set of representatives of conjugacy classes of elementary abelian \(p\)-subgroups of \(\Gamma_ g\). In some cases the normalizers have some simple form that allows calculation of the cohomology on the right hand side. Fortunately, the rank 1 condition on elementary abelian subgroups holds for ``almost all'' mapping class groups and ``almost all'' primes (\(p=2\) is the worst exception), in the following sense. There is a doubly indexed collection of families \({\mathcal F}_{t,\tau}\) of mapping class groups of the form \[ {\mathcal F}_{t,\tau}=\{\Gamma_ g: g=\alpha p+\beta,\quad p\text{ prime}\}, \] where \(\alpha=(2\tau+t-2)/2\), \(\beta=(2-t)/2\), such that in a given family \({\mathcal F}_{t,\tau}\), \(\Gamma_{\alpha p+\beta}\) has no elementary abelian \(p\)-subgroups of rank higher than 1, for all but finitely many primes \(p\). The parameters \(\tau,t\in\{0,1,2,\dots\}\) must satisfy \(t\neq 1\), \(t\geq 3\) if \(\tau=0\) and \(t\geq 2\) if \(\tau=1\), and, also, \(\Gamma_ g\) has \(p\)-torsion if and only if \(g=\alpha p+\beta\) for some \(\tau\) and \(t\) so restricted. So, for example, the cohomology decomposition above applies to \(\Gamma_ g\) in the following cases: \(g=(p-1)/2\), \(p\geq 5\), (\(\tau=0\), \(t=3\)); \(g=p-1\), \(p\geq 3\), (\(\tau=0\), \(t=4\)); \(g=(3p-3)/2\), \(p\geq 3\), \(p\neq 5\), (\(\tau=0\), \(t=5\)); \(g=p\), \(p\geq 3\), (\(\tau=1\), \(t=2\)); and \(g=p+1\), \(p\geq 7\), (\(\tau=2\), \(t=0\)). Furthermore, for a given family \({\mathcal F}_{t,\tau}\) each \(\Gamma_ g\) in \({\mathcal F}_{t,\tau}\) satisfies the following property, except for a finite number of primes. Each cyclic \(p\)-subgroup \(P\subset \Gamma_ g=\Gamma_{\alpha p+\beta}\) is realized by a \(\mathbb{Z}/p\) action on \(S_ g\) such that \(S_ g/P\) is a surface \(T\) of genus \(\tau\) and the quotient map \(S_ g\to T\) is branched over \(t\) points. This allows a combinatorial classification of the conjugacy classes of the groups in \(\mathcal P\) and the quotient \(N(P)/P\) can be identified with subgroup of finite index in the mapping class group of \(T\) with \(t\) punctures. The paper under review continues the author's previous work on the explicit determination of the \(p\)-torsion of the Farrell-Tate cohomology for the mapping class groups where \(\tau=0\). In his paper ``The \(p\)- torsion of the Farrell Tate cohomology of the mapping class group \(\Gamma_{(p-1)/2}\)'' [in: Topology '90, Proc. Res. Sem. Low-dim. Topol., Ohio-State Univ. (G. R. Baker et al. (eds.)) (to appear)], the cohomology is completely worked out for the case \(g=(p-1)/2\), \(p\geq 5\), i.e., \(\tau=0\), \(t=3\). This is the easiest case since \(N(P)/P\) is a finite group; but this is the only family for which this occurs. The paper examines the next simplest case \(g=p-1\), \(p\geq 3\), (\(\tau=0\), \(t=4\)) employing methods similar to those of the quoted paper, but with the new twist that \(N(P)/P\) is infinite. It is shown that there is an exact sequence \(1\to K_ 4\to N(P)/P\to M\to 1\) where \(K_ 4\) is the pure mapping class group of a quadruply punctured sphere, and \(M\) is one of six finite subgroups of \(\Sigma_ 4\), the symmetric group on four letters. Since \(K_ 4\) is the free group on 2 generators its cohomology is known, and by using the Leray-Hochschild-Serre spectral sequence on the above sequences as well as the sequence \(1\to P\to N(P)\to N(P)/P\to 1\), the cohomology of the normalizers are calculated. The main theorem gives explicit formulas for the rank of \(\widehat{H}^ i(\Gamma_ g,\mathbb{Z})_{(p)}\) as a product of \(\mathbb{Z}/p\)'s. The results and methods of proof are quite general and it appears that they could be extended to the families \({\mathcal F}_{t,0}\), at least for small \(t\). In particular, one gets exact sequences \(1\to K_ t\to N(P)/P\to M\to 1\), with \(M\subset \Sigma_ t\); the \(M\)'s are easy to determine. The only part that may get sticky is extending the spectral sequence computations to this sequence. \textit{F. Cohen's} results [``Artin's braid group and the homology of certain subgroups of the mapping class group'', Mem. Am. Math. Soc. (to appear)] (also used by the author) would be useful. The methods can also likely be used in cases where \(\tau>0\), for example, \(g=p\), \(p\geq 3\), (\(\tau=1\), \(t=2\)) and \(g=p+1\), \(p\geq 7\), (\(\tau=2\), \(t=0\)), however, the cohomology of possibly punctured surfaces of low genus must first be calculated.