Essential dimensions of \(A_7\) and \(S_7\) (Q2275691)

Let \(k\) be a field and \(G\) be a finite group. A compression is a dominant rational \(G\)-map of faithful \(G\)-varieties over \(k\). Let \(V\) be a faithful linear representation of \(G\) over \(k\); thus \(V\) can be regarded as a faithful \(G\)-variety over \(k\). The essential dimension of \(G\) over \(k\), denoted by \(\text{ed}_{k}(G)\), is the minimal value of \(\dim(X)\) where \(X\) is taken from the set of all faithful \(G\)-varieties sitting under a compression \(V \rightarrow X\). It is not an easy job to determine \(\text{ed}_{k}(G)\) even for a group \(G\) which we meet almost everyday (see page 13 of the book [\textit{S. Garibaldi, A. Merkurjev} and \textit{J.-P. Serre}, Cohomological invariants in Galois cohomology. University Lecture Series 28. Providence, RI: American Mathematical Society (2003; Zbl 1159.12311)]). For example, let \(S_n\) and \(A_n\) be the symmetric group and the alternating group of degree \(n\). It was known long before that, if \(k\) is a field with \(\text{char}\,k = 0\), then \(\text{ed}_k(A_5)=2\), \(\text{ed}_k(A_6)=2\) or \(3\), \(\text{ed}_k(S_6)=3\), \(\text{ed}_{k}(S_7)=3\) or \(4\). But it was an open question for many years to determine the exact values of \(\text{ed}_{k}(A_6)\) and \(\text{ed}_{k}(S_7)\). The essential dimension of \(A_6\) over a characteristic zero field \(k\) was found by \textit{J.-P. Serre} [Sémin. Bourbaki Vol. 2008/2009, Astérisque 332, 75--100, Exp. No. 1000 (2010; Zbl 1257.14012)] by using Manin-Iskovskikh's classification of minimal rational \(G\)-surfaces. A sketch of Serre's proof was included in the second paragraph of Section 2 of the paper under review. The main result of this paper is to show that \(\text{ed}_k(S_7)=\text{ed}_k(A_7)=4\). By standard results for essential dimensions, it suffices to show that \(\text{ed}_{C}(A_7)=3\) is impossible. The proof is rather similar to Serre's Theorem, but this time Prokhorov's classification of rationally connected faithful \(A_7\)-threefolds will be used [\textit{Yu. Prokhorov}, J. Algebr. Geom. 21, No. 3, 563--600 (2012; Zbl 1257.14011), \url{http://arxiv.org/abs/0908.0678}], instead of Manin-Iskovskikh's classification of \(G\)-surfaces. Here is an outline of the proof. Suppose \(\text{ed}_{C}(A_7)=3\). Then there is a compression \(V\to X\) where \(V\) and \(X\) are faithful \(A_7\)-varieties over \(C\), \(V\) is rational and \(\dim X=3\). It follows that \(X\) is a rationally connected faithful \(A_7\)-threefold. By Prokhorov's classification, \(X\) is \(A_7\)-birationally equivalent to one of the two threefolds: (i) the subvariety of \(P^6\) (with the ``obvious'' \(A_7\)-action) cut out by symmetric polynomials of degrees \(1,2\) and \(3\), and (ii) \(P^3\) with a linear action of \(A_7\). On the other hand, the compression \(V\to X\) implies that \(X\) has an \(A\)-fixed point for any abelian subgroup \(A\) of \(A_7\) by an old result of \textit{Z. Reichstein} and \textit{B. Youssin} [Can. J. Math. 52, No. 5, 1018--1056 (2000; Zbl 1044.14023)]; Proposition 5.3. Thus, for any abelian subgroup \(A\) of \(A_7\), the above two threefolds have \(A\)-fixed points. It is easy to see this is not the case. Hence the proof is finished.