I've been stuck for a while on Exercise 1.9 of Bjorn Poonen's "Rational Points on Varieties". We start with $L/K$ a finite Galois extension with Galois group $G$, some $r \in \mathbb{Z}_{\geq 0}$, and for each 1-cochain $\xi: G \rightarrow GL_r\left(L\right)$, we define $W_{\xi}$ to be $L^r$ with equipped with the function $G \times L^r \rightarrow L^r$, $\sigma \mapsto \xi_{\sigma}\left(\sigma w\right)$.

In the first part of the question, we are asked to show that this function describes a semilinear $G$-action on $W_{\xi}$ iff $\xi$ is a 1-cocycle, i.e., $\xi_{\sigma \tau} = \xi_\sigma {}^{\sigma}\xi_{\tau}$. I have managed this.

The second part is to show that, for two $1$-cocycles $\xi$ and $\xi'$, the spaces $W_\xi$ and $W_{\xi'}$ are isomorphic as $L$-vector spaces with semilinear $G$-action iff $\xi$ and $\xi'$ are cohomologous. This is where I am stuck.

I am new to group cohomology, but I have studied other types of cohomology before and have read a few things on group cohomology in order to try and solve this problem. As I understand it, $\xi$ and $\xi'$ are cohomologous iff their difference is a $1$-coboundary, i.e., a principal crossed homomorphism. Translating the standard definitions into multiplicative language, I believe this amounts to showing that there is some $f \in GL_r\left(L\right)$ such that, for all $w \in W = L^r$ and all $\sigma \in G$, we have $$\xi'_{\sigma} \xi_{\sigma}^{-1}\left(w\right) = \left({}^{\sigma}f\right) f^{-1}\left(w\right),$$ i.e.,

\begin{equation} \xi'_{\sigma} \xi_{\sigma}^{-1} = \left({}^{\sigma}f\right) f^{-1}. \end{equation}

Now, from what I can gather, $W_{\xi}$ and $W_{\xi'}$ are isomorphic as $L$-vector spaces with semilinear $G$-action iff there is $g \in GL_r\left(L\right)$ such that for all $w \in W = L^r$ and all $\sigma \in G$, we have $$g\left(\xi_\sigma \left(\sigma w\right)\right) = \xi'_{\sigma}\left(\sigma g\left(w\right)\right),$$ i.e., $$g\left(\xi_\sigma \left(\sigma w\right)\right) = \xi'_{\sigma}\left( {}^{\sigma}g\left(\sigma w\right)\right),$$ meaning precisely that

\begin{equation} g \xi{_\sigma} = \xi'_{\sigma}{}^{\sigma}g. \end{equation}

My idea for tackling this is to show that we can manipulate the second expression until it is in the form of the first one.

From $\xi$ being a 1-cocycle, we can deduce that $\xi_{id_G} = id_{GL_r\left(L\right)}$ and then that $\xi_{\sigma}^{-1} = {}^{\sigma}\xi_{\sigma^{-1}}$, so, assuming the latter highlighted equation (involving $g$), we get $$\xi'_{\sigma} \xi_{\sigma}^{-1} = \xi'_{\sigma} {}^{\sigma}\xi_{\sigma^{-1}} = g\xi_\sigma {}^{\sigma}\left(g^{-1}\right) {}^{\sigma}\xi_{\sigma^{-1}} = g\xi_\sigma {}^{\sigma}\left(g^{-1}\xi_{\sigma^{-1}}\right),$$ but I can't work that into the form I want by a suitable choice of $f$. I have tried other manipulations, but with no success.

My conclusion is that I have probably misunderstood or mistranslated some concept. In particular, I fear I might have the wrong notion of isomorphism for $L$-vector spaces with semilinear $G$-action. The definition I have here is one I found for isomorphism of $G$-sets on Wikipedia. Initially, I had wondered whether an isomorphism should be a pair $\left(\phi,h\right)$ where $\phi: G \rightarrow G$ is an automorphism, $h \in Gl_r\left(L\right)$ and, for all $w \in W$ and $\sigma \in G$, $$h\left(\xi_\sigma \left(\sigma w\right)\right) = \xi'_{\phi\left(\sigma\right)}\left(\phi\left(\sigma\right)h\left(w\right)\right),$$

so in particular I get the first definition back if $\phi$ is the identity map on $G$, but I'm not sure.

If anybody can point out the erroneous definitions/where I am going wrong, then I would be very grateful.


Your mistake is in the meaning of $\xi$ and $\xi'$ being cohomologous. The condition should be $\xi_\sigma =f^\sigma \xi'_\sigma f^{-1}$, for some $f\in GL_r(L)$. This is the correct definition of cohomologous cocycles in the non-abelian cituation, and then $f$ just give you the isomorphism.

In general one should be careful about order and directions when dealing with non-abelian cohomologies. The correct formulas are usually not the first one you would guess.

  • 4
    $\begingroup$ Let me add that an excellent reference is Serre Local fields, Part III. $\endgroup$ – abx Oct 12 '18 at 13:11
  • $\begingroup$ Ah, brilliant! So the question is in fact quite straightforward with the correct definition. Thank you very much for your help! $\endgroup$ – Sam Streeter Oct 12 '18 at 13:13
  • 1
    $\begingroup$ I would also recommend reading Serre, Galois cohomology, Ch. I, Section 5. $\endgroup$ – Mikhail Borovoi Oct 13 '18 at 11:20

Your Answer

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

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