Counterexamples to a conjecture of Lemmermeyer (Q1291142)

Let \(K\) be a quadratic number field, and let \(L\) be an unramified \(2\)-extension of \(K\) which is normal over \(\mathbb Q\) (the Hilbert \(2\)-class field of \(K\) has this property). By Chebotarev's monodromy theorem, \(\Gamma = \text{Gal}(L/\mathbb Q)\) is generated by involutions (elements of order \(2\)); in particular, \(G = \text{Gal}(L/K)\) is a subgroup of index \(2\) in \(\Gamma\). Let us say that a \(2\)-group \(G\) has property P if \(G\) is a subgroup of index \(2\) in a group generated by involutions. In an unpublished manuscript, the reviewer conjectured that every finite \(2\)-group has property P; clearly finite abelian \(2\)-groups do. In this paper, the authors show by a computation using the MAGMA databases that exactly two out of the \(267\) groups of order \(64\) do not have property P, and then go on to construct infinitely many counterexamples by hand.