5
$\begingroup$

Let $G$ be a group and $F$ a field. I am particularly interested in the case where $G$ is a uniform lattice in a Lie group and $F=\mathbb{F}_2$, or in finite groups $G$ where $\operatorname{char} F$ divides $|G|$, but the discussion applies to general $G$ and $F$.

Let $\rho:G\to \mathrm{GL}(V)$ be a representation of $G$ on a finite-dimensional $F$-vector space $V$. Letting $G$ act trivially on $F$, we may form the cohomology ring $H^*(G,F)$, which acts via the cup product on the graded module $H^*(G,V)$. In particular, the cup product induces a linear map $$ p: H^1(G,F)\otimes_F H^1(G,V)\to H^2(G,V). $$ My question is whether there are $G$ and $V$ for which:

  • $p$ is injective,
  • $\dim H^1(G,V) \geq 1$,
  • $\dim H^1(G,F)\geq 2$.

More generally, is there a recipe for constructing such examples? I am not aware of even a single example of this kind.

The injectivity of $p$ is the real issue. For example, if we take $V=F$, then the cup product $\cup :H^1(G,F)\times H^1(G,F)\to H^2(G,F)$ is well-known to be anti-symmetric, which means that $\dim \ker p\geq \frac{1}{2}\dim H^1(G,F)(\dim H^1(G,F)-1)$ if the characteristic of $F$ is $2$, or $\dim \ker p\geq \frac{1}{2}\dim H^1(G,F)(\dim H^1(G,F)+1)$ if $F$ is not of characteristic $2$, so $\ker p$ would not be trivial when $\dim H^1(G,F)\geq 2$. (This special example is studied in a paper of Hillman.)

$\endgroup$

1 Answer 1

3
+100
$\begingroup$

$\newcommand{\bZ}{\mathbb{Z}}$Let $G$ be a finite group of order divisible by $p:=\mathrm{char}\, F$ such that $\dim_F \mathrm{Hom}(G,F)\geq 2$ (e.g. $G=\bZ/p\times\bZ/p$). Take $V$ to be the kernel of the augmentation map $e(\sum a_g\cdot g)=\sum a_g$ from $F[G]$ to $F$. Since $F[G]$ is an injective $G$-module, the short exact sequence $0\to V\to F[G]\xrightarrow{e} F\to 0$ yields isomorphisms $H^i(G, F)\simeq H^{i+1}(G,V)$ for $i\geq 0$ (this is automatic for $i>0$ and for $i=0$ follows from the fact that $e:F[G]^G\to F$ is the zero map because the order of $G$ is divisible by $p$).

These isomorphisms are compatible with the cup-product in the sense that the following diagram commutes for every $i$ (this follows from the associativity of cup-product using the observation that the isomorphism $H^i(G,F)\to H^{i+1}(G,V)$ is obtained by cupping with a class in $H^1(G,V)$)

$\require{AMScd}$ \begin{CD} H^1(G,F)\otimes H^{i+1}(G,V) @>\cup >> H^{i+2}(G,V)\\ @| @|\\ H^1(G,F)\otimes H^{i}(G,F) @>>\cup > H^{i+1}(G,F) \end{CD}

For $i=0$ the bottom map is obviously an isomorphism, hence the top map is an isomorphism, as desired.

$\endgroup$
2
  • $\begingroup$ Thank you! I think that the connecting map always compatible with the cup product (up to sign if done on the left), is it not? $\endgroup$ Commented Apr 26, 2022 at 6:23
  • $\begingroup$ @UriyaFirst Right, and one way to prove this is to observe that any connecting homomorphism is given by cupping with the cohomology class of the extension that gave rise to the connecting homomorphism. $\endgroup$
    – SashaP
    Commented Apr 26, 2022 at 14:50

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .