I asked a question on m.se about generalised cohomology and spectra. Not having received any specific answer I attempted to draw more attention by offering a bounty. But I still could not get any help. If it does not violate this site's policies, I would like to reproduce my question below. I would really appreciate any help. Thanks


I am reading the paper Quadratic functions in geometry, topology and M theory by M.J.Hopkins and I.M.Singer, and in section 4.8 they say :

'Recall that for any compact $S$, and for any cohomology theory $E$ there is a canonical isomorphism $E^{\ast}(S)\otimes\mathbb{R} = H^{\ast}(S;\pi_{\ast}E\otimes \mathbb{R})$.When $E$ is $K$-theory, this isomorphism is given by the Chern character. The isomorphism arises from a universal cohomology class $ i_{E}\in H^{0}(E;\pi_{\ast}E\otimes\mathbb{R}) = \varprojlim H^{n}(E_{n};\pi_{\ast+n}E\otimes\mathbb{R})$ and associates to a map $f:S\to E_{n}$ representing an element of $E^{n}(S)$ the cohomology class $f^{\ast}i_{n}$, where $i_{n}$ is the projection of $i_{E}$ to $H^{n}(E_{n};\pi_{\ast+n}E)$.

I do not understand most of the above paragraph. My specific questions are :

1.) Where can I find a proof of the isomorphism mentioned ? I am familiar with the basic definitions of spectra and generalized cohomology but can not quite understand the isomorphism. Is this some kind of Brown representability ?

2.) I do not understand the system of abelian groups over which the limit is taken.

Of course, it is unreasonable to ask for detailed explanations for my doubts. I would really appreciate some hints, and pointers to good texts where I can find a discussion of the above issues and educate myself. Thanks so much !

  • $\begingroup$ This comes from the fact that over rationals (so a fortiori over reals) all spectra are wedge of spheres, thus the Atiyah-Hirzebruch spectral sequence is trivial. Strictly speaking, only the first part of the previous sentence suffices, but the second makes things clearer, if you are familiar with Atiyah-Hirzebruch spectral sequence. I don't know where you can find a complete proof of this, but this follows from the equivalence of Eilenberg-Maclane and sphere spectra, and this is proved, for example, in Adams' Chicago lecture note. The limit is simply over $n$. $\endgroup$
    – user43326
    Jun 17, 2018 at 7:47
  • $\begingroup$ @user43326 Thanks. I have studied Serre sprectral sequence but am not familiar with AHSS. Can you suggest some modern source for learning AHSS and the statements in your comment ? I have heard a lot of praise about Adams' book, but I usually prefer modern expositions. Thanks a lot $\endgroup$
    – user90041
    Jun 17, 2018 at 16:39
  • $\begingroup$ I guess Switzer's book isn't modern for you? I think any textbook covering generalized (co)homology covers AHSS. $\endgroup$
    – user43326
    Jun 18, 2018 at 6:37

1 Answer 1


For any spectra $X$ and $Y$ put $$ M(X,Y)_* = \text{Hom}(H_*(X;\mathbb{Q}),H_*(Y;\mathbb{Q})) = \prod_{i\in\mathbb{Z}} H^i(X;H_{*-i}(Y;\mathbb{Q})) $$ There is an evident natural map $$ h_{XY}\colon [X,Y]_* \to M(X,Y)_*, $$ which induces a map $h'_{XY}\colon [X,Y]_*\otimes\mathbb{Q}\to M(X,Y)_*$. The main point is that $h'_{XY}$ is an isomorphism whenever $X$ is a finite spectrum. To prove this, we first need to know that $h'_{S^p,S^q}$ is an isomorphism. This is essentially the fact that $\pi_0(S^0)=\mathbb{Z}$, with $\pi_i(S^0)$ being finite for $i\neq 0$. Now put $$ \mathcal{A} = \{Y : h'_{S^0,Y} \text{ is an isomorphism }\}. $$ This is easily seen to be closed under suspension and desuspension and under arbitrary coproducts. If we have a cofibre sequence $Y_0\to Y_1\to Y_2\to\Sigma Y_0$, then we get long exact sequences of homotopy and homology groups. Thus, if two of the spectra $Y_i$ lie in $\mathcal{A}$, then we can use the five lemma to show that the third does also. It follows by cellular induction that all spectra lie in $\mathcal{A}$, so $h'_{S^0,Y}$ is always an isomorphism. In particular, this means that $\pi_*(Y)\otimes\mathbb{Q}=H_*(Y;\mathbb{Q})$ so we can rewrite $M(X,Y)$ as $$ M(X,Y) = \prod_iH^i(X;\pi_{*-i}(Y)\otimes\mathbb{Q}) $$ if desired.

Now put $$ \mathcal{B} = \{X:h'_{XY} \text{ is an isomorphism for all } Y\}. $$ We have seen that $S^0\in\mathcal{B}$. It is also easy to see that $\mathcal{B}$ is closed under suspension and desuspension and under finite coproducts. Moreover, if we have a cofibre sequence $X_0\to X_1\to X_2\to\Sigma X_0$ then we again get long exact sequences for $[X_i,Y]_*\otimes\mathbb{Q}$ and $M(X_i,Y)_*$. Thus, we can again use the five-lemma to see that if two of the spectra $X_i$ are in $\mathcal{B}$, then the third is also. It follows by cellular induction that all finite spectra lie in $\mathcal{B}$, so $h'_{XY}$ is an isomorphism whenever $X$ is finite. The same kind of argument also shows that when $X$ is finite we have $H^*(X;\pi_*(Y)\otimes\mathbb{Q})=H^*(X;\pi_*(Y))\otimes\mathbb{Q}$.

To reconcile this with your notation, take $Y=E$. The case $X=S^0$ gives $H_*(E;\mathbb{Q})=\pi_*(E)\otimes\mathbb{Q}$, so $$ E^*(X)\otimes\mathbb{Q} = H^*(X;\pi_*(E)\otimes\mathbb{Q}) = H^*(X;\pi_*(E))\otimes\mathbb{Q}. $$ You can now tensor everything with $\mathbb{R}$ if you like.


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.