Calculating group cohomology: tests for completion (Q5928892)

This article deals with the computation of the mod \(p\) cohomology of a finite \(p\)-group \(G\). Of particular interest are stopping conditions: Having computed the cohomology up to some degree~\(N\), how can we know that we have a generating set for the cohomology ring? Carlson first explains how group cohomology can be computed degree by degree, using a minimal projective resolution of \(k\), a field of characteristic~\(p\). Then he puts up two conditions on the cohomology ring. The conditions are in fact formulated for the subring \(R_N\) generated by all cohomology up to some degree~\(N\), and they are conjectured to be true for the complete cohomology ring. The first condition is an assumption on the Koszul homology given by a parameter system for \(R_N\). The second condition is a statement on the essential cohomology, i.e., that part of the cohomology which vanishes when restricted to any maximal subgroup. Both conditions are somewhat technical to state, but they can be checked algorithmically after \(R_N\) has been calculated. The main theorem of the paper states that if the conditions are satisfied for \(R_N\) with \(N\) large enough (as specified in the theorem), then \(R_N\) coincides with the complete cohomology ring. The proof of the main theorem appears in the last section and uses the hypercohomology spectral sequence. Thus the following strategy emerges: Calculate the cohomology degree by degree until reaching a degree~\(N\) which is large enough (as specified in the main theorem) and has the property that both conditions are satisfied for \(R_N\). Then the main theorem tells us that \(H^*(G,k)=R_N\). As a desired side effect, a successful run of this strategy verifies the conjecture that the conditions hold for \(H^*(G,k)\). Carlson has implemented the algorithms in the MAGMA computer algebra system. The conjectures hold for all \(p\)-groups tested so far.