7
$\begingroup$

Consider an extension\begin{equation}1\rightarrow N\rightarrow G\xrightarrow{\rho} K\rightarrow 1\end{equation}Let $K$ act on a $K$-module $A$ by $\phi_k: a\mapsto k\cdot a$. Define a $G$-action $a\mapsto g\cdot a:=\rho(g)\cdot a$; under this action, $A$ is a $G$-module. Now consider the Hochschild-Serre spectral sequence\begin{equation}E^2_{p,q}=H_p(K,H_q(N,A))\Rightarrow H_{p+q}(G,A)\end{equation}It would appear that the LHS, the page $E^2$, does not depend on the $K$-action $\phi_k$, while the RHS, the limit of the sequence, does depend on it. How does one resolve this apparent contradiction?

To be clear, I understand that, in general, $``H_p"$ is homology with local coefficients: $K$ acts on $H_q(N,A)$ according to $\rho$. I am comfortable with this.

I am concerned with the action $\phi_k$ of $K$ on $A$. Since $N\simeq\ker\rho$, $N$ always acts trivially on $A$, and so the effect of $\phi_k$ is erased on the second page. Is the data of $\phi_k$ encoded in the differentials?

I am primarily concerned with the case where $K=\mathbb{Z}_2=\{1,T\}$. I want to detect the difference between the trivial action $\phi^{triv}_T:a\mapsto a$ and the inversion action $\phi^{inv}_T:a\mapsto -a$. Presumably, these should yield different spectral sequences since they yield different limits $H_n(G,A)$.

$\endgroup$
1
  • 1
    $\begingroup$ I migrated this over at the request of the OP. $\endgroup$ Jun 4, 2014 at 12:28

1 Answer 1

3
$\begingroup$

The action of $K$ on $H_q(N;A)$ may be non-trivial, even though $A$ is a trivial $N$-module. This may even happen if the extension is central, so that the conjugation action of $K$ on $N$ is trivial. You have to take into account the coefficient homomorphisms $$ (\phi_k)_\ast : H_q(N;A)\to H_q(N;A),\quad k\in K. $$ Since $A$ is a non-trivial $K$-module, these may be non-trivial.

This is treated in section III.8 of Ken Brown's "Cohomology of groups" textbook, where he explains how to view group cohomology as a functor of two variables.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.