I am trying to figure out the number of orbits of a two-point stabilizer of the action of $Sp(2m,2)$ on its two orbits $\Omega^+$ and $\Omega^-$ as detailed in Dixon and Mortimer's "Permutation Groups", the paper https://doi.org/10.1016/S0021-8693(02)00551-3, as well as this question https://mathoverflow.net/a/85094/172098. In the last one, Derek Holt says that the two-point stabilizers will have four orbits, with the two nontrivial ones given by the vectors orthogonal and not orthogonal to the fixed isotropic point. However, I cannot seem to figure out how to show this.

I am new to the area, but here is my (possibly mistaken) understanding so far. Let us fix a symplectic form $b$ on $V:=(GF(2))^{2m}$. The group $Sp(2m,2)\leq GL(2m,2)$ which preserves $b$ acts on the set $\Omega$ of quadratic forms on $V$ and has two orbits: $\Omega^+$ and $\Omega^-$. These are respectively the orbits of the forms $q^+=\sum_i^m x_i y_i$ and $q^-=x_1^2 + y_1 ^2 + \sum_i^m x_i y_i$. Take any quadratic form $q$ and for now assume that $q\in \Omega^+$. Then the subgroup $O(V,q)$ of $Sp(2m,2)$ which fixes $q$ can be shown (by Witt's Lemma) to act transitively on $\Omega^+\setminus\lbrace{q\rbrace}$. This can be done because there is a natural bijection from $V$ to $\Omega$ such that the actions of $O(V,q)$ on each will be isomorphic. In particular, $\Omega^+$ will correspond to the isotropic vectors in $V$ with respect to $q$ and $\Omega^-$ will correspond to the anisotropic vectors.

Thus, to show the desired goal that a two-point stabilizer of the action of $Sp(2m,2)$ on $\Omega^+$ will have four orbits, it suffices to pick a nonzero isotropic vector $v\in V$ with respect to a bilinear form $q\in \Omega^+$ and show that its stabilizer $O(V,q)_v$ on $O(V,q)$ will have two orbits on $V\setminus \lbrace 0,v\rbrace$. Derek Holt's claim was that these two orbits are precisely those that are orthogonal and not orthogonal to $v$ (with respect to the bilinear form $b$), and this is what I cannot show. Of course, it is clear that if $w$ is orthogonal to $v$, then so is any member of its orbit in $O(V,q)_v$; an analogous statement holds for when $w$ is not orthogonal to $v$. However, I am unable to show that all isotropic vectors (distinct from 0 and $v$) that are orthogonal (resp. not orthogonal) to $v$ will necessarily be in the same orbit of $O(V,q)_v$. A completion of this idea, an alternative approach, a reference confirming this result, and any corrections to the above will be greatly appreciated.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy