The essential ideal in group cohomology does not square to zero. (Q1877759)

Let \(G\) be a finite group and \(k\) a field of characteristic \(p\). A class \(x\) in the cohomology ring \(H^*(G,k)\) is called essential if its restriction to every proper subgroup is zero. The essential classes in \(H^*(G,k)\) form an ideal, called the essential ideal \(\text{Ess}(G)\). If the group \(G\) is elementary Abelian, then the product of the Bocksteins of all nonzero elements of \(H^1\) is a non-nilpotent essential class. But in all remaining cases that could be calculated, the pattern that emerged seemed to be that every product of essential classes was zero. This led to the Essential Conjecture: Let \(G\) be a finite \(p\)-group which is not elementary Abelian, then \(\text{Ess}(G)^2=0\). \textit{P. A. Minh} showed [in Bull. Lond. Math. Soc. 32, No. 3, 285-291 (2000; Zbl 1021.20038)] that if \(x\) is essential then \(x^p=0\) (assuming \(G\) is not elementary Abelian). In this paper the author disproves the Essential Conjecture. He proves the following: Let \(G_0\) be a Sylow 2-subgroup of the finite unitary group \(\text{SU}_3(4)\). Then there are essential classes in \(H^4(G_0,\mathbb{F}_2)\) and \(H^{10}(G_0,\mathbb{F}_2)\) whose product is nonzero. This nonzero element of \(H^{14}(G_0,\mathbb{F}_2)\) is the last survivor in the sense of \textit{D. J. Benson} and \textit{J. F. Carlson} [Trans. Am. Math. Soc. 342, No. 2, 447-488 (1994; Zbl 0816.20044)]. A very interesting corollary concerning \(G_0\) comes from the work of \textit{J. Pakianathan} and \textit{E. Yalçin} [Topology 42, No. 5, 1155-1183 (2003; Zbl 1047.20041)] where they showed that if the essential cohomology conjecture is not true for a \(p\)-group \(G\), then for any connected \(G\)-graph \(X\) the following is true: \(X\) has a fixed point if and only if the map \(H^*_G(pt)\to H^*_G(X)\) is injective. The fact that \(\text{Ess}(G_0 )^2\) in nonzero for this group \(G_0\) was first observed using the author's computer program [Gröbner bases and the computation of group cohomology, (Lect. Notes Math. 1828) (2003; Zbl 1050.20036)]. The argument presented here uses \textit{J. Clark}'s spectral sequence calculation [Commun. Algebra 22, No. 4, 1419-1434 (1994; Zbl 0804.20036)].