Let $E$ be a spectrum acted upon a finite group $G$. Is there a general way of computing the homology of the homotopy fixed point spectrum $E^{hG}$ in terms of that of $E$? (I'm aware that there is a spectral sequence for computing $\pi_* E^{hG}$ in terms of $\pi_* E$, but smashing with some other spectrum probably doesn't preserve homotopy fixed points.)

Here's a specific example I have in mind. Take connective $K$-theory $bu$. This has an action of $\mathbb{Z}/2$, which comes from the $\mathbb{Z}/2$-action on $K$-theory (given on the level of cohomology theories by complex conjugation of vector bundles). Then $bu^{\mathbb{Z}/2} = bo$. Actually, this is only true before taking connective covers.

Let's say I know how to compute the mod 2 homology of $bu$ (it's $\mathbb{Z}/2[\zeta_1^2, \zeta_2^2, \zeta_3, \zeta_4, \dots]$ as a comodule over the dual Steenrod algebra). Does that give any information about $H_*(bo; \mathbb{Z}/2)$? Is there a good reference for this material?

  • 7
    $\begingroup$ I think that, while $KU^{\mathbb Z/2}=KO$, it is NOT true that $ku^{\mathbb Z/2}=ko$. If you run the homotopy fixed point spectral sequence, it's ptretty obvious that you don't get $\pi_*(ko)$ from $H^*(\mathbb Z/2;\pi_*(ku))$. $\endgroup$ – André Henriques Jun 30 '12 at 20:17
  • $\begingroup$ I thought that taking homotopy fixed points would commute with truncation $\tau_{\geq 0}$ (which is a right adjoint, right?). Is there something that goes wrong here? $\endgroup$ – Akhil Mathew Jul 1 '12 at 2:36
  • 5
    $\begingroup$ @Akhil: The truncation is a right adjoint, but only if the target is a different category (connective spectra). In that category, the homotopy fixed point object of $ku$ is indeed $ko$. $\endgroup$ – Tyler Lawson Jul 1 '12 at 5:40

Consider $G$ of order $2$ acting trivially on the sphere spectrum $S^0$. In this key example, smashing with $H\mathbb Z/2$ drastically fails to commute with $^{hG}$: If it did commute, then the mod $2$ homology of the homotopy fixed point spectrum would be the homotopy of $(H\mathbb Z/2)^{hG}$, so $\mathbb Z/2$ in nonpositive dimensions and $0$ in positive dimensions. But the (solved) Segal conjecture shows that the homotopy fixed point spectrum is connective.


A good reference for the sort of spectral sequence that you're looking for is an article by Bruner-Rognes, "Differentials in the homological homotopy fixed point spectral sequence." It treats some of the convergence questions that you're worried about.

  • $\begingroup$ This paper looks pretty interesting. Thanks. $\endgroup$ – Akhil Mathew Jul 1 '12 at 12:55
  • 1
    $\begingroup$ I was just about to suggest this. Their construction is pretty explicit. While they are mostly interested in the case where $G=S^1$ you can get the general case by using your favorite model for $EG$ and following their construction. $\endgroup$ – Sean Tilson Jul 2 '12 at 4:47
  • $\begingroup$ Also, that spectral sequence works the same for any generalized cohomology theory. The only issue might be convergence if it is not connective (or bounded below rather). $\endgroup$ – Sean Tilson Jul 2 '12 at 4:49

There is a subject of equivariant stable homotopy theory. You are referring to $bu$ as what is called a naive $G$-spectrum (a spectrum acted on by a group), and your notation $E^G$ is confusing the categorical fixed point spectrum with the homotopical fixed point spectrum. The difference between the two is, for example, the subject of the Segal conjecture. There is a homotopy fixed point spectral sequence for homotopical fixed point spectra. There is a huge literature on equivariant stable homotopy theory (which was a key input to solution of the Kervaire invariant problem by Hill, Hopkins, and Ravenel). One expository introduction is:

Greenlees and May. Equivariant stable homotopy theory. Handbook of Algebraic Topology, edited by I.M. James, pp. 279-325. 1995.

But a lot has been done since then. Naive $G$-spectra are not so interesting. Genuine $G$-spectra have $RO(G)$-graded homology and cohomology.

  • $\begingroup$ I did mean the homotopy fixed point spectrum, though I realize my notation was off (I've changed it). I'm aware of the homotopy fixed point sequence for the homotopy groups; I guess my question is whether one can use it to compute homology, because smashing with a big spectrum like $H \mathbb{Z}/2$ probably doesn't preserve homotopy fixed points, right? $\endgroup$ – Akhil Mathew Jul 1 '12 at 2:36
  • 2
    $\begingroup$ In a sense "naive $G$-spectra" is the natural home for questions about homotopy fixed-point spectra and homotopy orbit spectra. But in order to answer a question you sometimes have to go to the larger world of "genuine $G$-spectra". $\endgroup$ – Tom Goodwillie Jul 1 '12 at 11:23
  • $\begingroup$ Just a note for people who find this thread in the future: after this answer the OP edited his question to replace $E^G$'s by $E^{hG}$'s in response to the first part of this answer. $\endgroup$ – David White Jul 4 '12 at 13:13

Maybe to expand on the relationship between naive and genuine G-spectra, Tom's answer is stated in terms of naive G-spectra, but the actual mathematics, the Segal conjecture, is all about genuine G-spectra. The genuine sphere $G$-spectrum $S_G$ has non-trivial $G$-action and nontrivial $G$-fixed point spectrum. Consider a finite $G$-CW complex. The $G$-map $EG\times X \to X$ induces a map of genuine function $G$-spectra \[ F(X_+, S_G) \to F((EG\times X)_+, S_G). \] If $G$ is a $p$-group, the Segal conjecture says that this map is an equivalence after $p$-adic completion. This remains true after passing to $G$-fixed point spectra. By definition, that passage means that you first take underlying naive $G$-spectra and then take actual fixed points. Tom is taking $X$ to be a point, and then the map is \[ (S_G)^G \to F(EG_+, S_G)^G, \] going from the fixed point spectrum to the homotopy fixed point spectrum of the genuine $G$-spectrum $S_G$. The target is equivalent to the naive homotopy fixed point spectrum $F(EG_+, S)^G$ of the naive sphere $G$-spectrum with trivial action by $G$, and that is equivalent to $F(BG_+,S)$, giving the cohomotopy groups of $BG$ on passage to homotopy groups. But all of the real mathematics in validating Tom's answer takes place on the level of genuine $G$-spectra.

  • $\begingroup$ Sure. I've been a bit loose about this in the question. I'll take a look at the equivariant homotopy book so that I can understand these distinctions. $\endgroup$ – Akhil Mathew Jul 6 '12 at 0:39
  • 1
    $\begingroup$ Your question was about spectra with $G$-action. There was no need to mention genuine $G$-spectra in the question. The remarkable thing is they come into the answer in an essential way. $\endgroup$ – Tom Goodwillie Jul 6 '12 at 12:24

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.