It is known that group cohomology class $H^d[U(1),Z]$ is Z for even d and 0 for odd d. Do we know $H^d[G,Z]$ for $G=SO(3)$, $SU(2)$ and other compact Lie group?

Also is the Borel-group-cohomology class $H^d[G,R]$ alway trivial for compact Lie group $G$?

  • $\begingroup$ Your first example suggests you are talking about the cohomology of the classifying space $BG$. If so, why not ask this? Also, your first example gives a counter-example for your second question (unless I'm misinterpreting what the "Borel-group-cohomology class" is). Please clarify. $\endgroup$
    – Mark Grant
    Sep 14 '11 at 13:55
  • $\begingroup$ Dear Mark, Thanks for the question. I just know the algebraic definition of group cohomology and I am not familiar with classifying space (I am a physicist). By Borel-group-cohomology of $H^d[G,R]$ I mean that we take cochains as measurable function over $G$. The issue of "continuity" comes up since both the group $G$ and the module $R$ are continuous. If $H^d[G,R]$ is trivial, I am hoping to get $H^{d+1}[G,Z] = H^d[G,U(1)]$, again $H^d[G,U(1)]$ is the Borel-group-cohomology described above. I learned those from some math papers that I only half understand. I hope they are right. $\endgroup$ Sep 14 '11 at 15:07

For the group $SU(2)=S^3$ we just have $H^*(BSU(2);\mathbb{Z})=\mathbb{Z}[c_2]$ (where $c_2\in H^4$). More generally, for all $n$ we have \begin{align*} H^*(BU(n);\mathbb{Z}) &= \mathbb{Z}[c_1,\dotsc,c_n] \\\\ H^*(BSU(n);\mathbb{Z}) &= \mathbb{Z}[c_2,\dotsc,c_n] \\\\ H^*(BSp(n);\mathbb{Z}) &= \mathbb{Z}[p_1,\dotsc,p_n] \end{align*} with $c_i\in H^{2i}$ and $p_i\in H^{4i}$.

Now let $V$ be the tautological $3$-plane bundle over the space $X=BSO(3)$. This has Stiefel-Whitney classes $w_2\in H^2(X;\mathbb{Z}/2)$ and $w_3\in H^3(X;\mathbb{Z}/2)$. There is also a Bockstein element $v=\beta(w_2)\in H^3(X;\mathbb{Z})$ (which satisfies $2v=0$) and a Chern class $c=c_2(\mathbb{C}\otimes V)\in H^4(X;\mathbb{Z})$. The mod two reduction map $\rho$ satisfies $\rho(v)=Sq^1(w_2)=w_3$ and $\rho(c)=w_2^2$. If I've got everything straight, one can check using the Bockstein spectral sequence that $$ H^*(BSO(3);\mathbb{Z}) = \mathbb{Z}[v,c]/(2v). $$

It is not possible to be similarly explicit about $H^*(BSO(n);\mathbb{Z})$ for general $n$ (although $H^*(BSO(n);\mathbb{Z}/2)$ and $H^*(BSO(n);\mathbb{Q})$ are fairly straightforward).

  • $\begingroup$ Excuse me what is your precise definition of $H$ in your $c_i \in H^{2i}$? $\endgroup$
    – wonderich
    Mar 30 '17 at 20:39

A collection of relevant references (general, standard as well as specific ones) is here:



Check out


  • 1
    $\begingroup$ Thank you very much, Igor. But I am a physicist. May I ask how to obtain $H^d[G,Z]$ from its torsion free part and $H^d[G,Z_2]$. In particular, what are $H^4[SO(3),Z]$ and $H^5[SO(3),Z]$? I guess $H^3[SO(3),Z]=Z_2$ corresponding to the projective representations of $SO(3)$ and $H^2[SO(3),Z]=Z_1$ corresponding the absence of non-trivial 1D representation of $SO(3)$. ( I am using $H^{d+1}[SO(3),Z] = H^d[SO(3),U(1)]$ here, if I use $H^d[SO(3),U(1)]$ to mean Borel group cohomology where the conchains are measurable functions over $SO(3)$.) $\endgroup$ Sep 14 '11 at 14:54
  • $\begingroup$ In planetmath.org/encyclopedia/GroupCohomology3.html , it is stated that $H^d[SU(2),Z]=Z$ if $d=0$ mod 4 and 0 other wise. This seems contradict with the result in your note: The torsion free part of $H^d[SO(3),Z]$ is Z for $d=0$ mod 3 is zero other wise. (I am guessing that the torsion free part of $H^d[SU(2),Z]$ and $H^d[SO(3),Z]$ are the same, and I thought $H^4[G,Z]$ contains a $Z$ for any Lie group.) Please correct me if I misunderstand your note $\endgroup$ Sep 14 '11 at 23:48
  • 2
    $\begingroup$ After thinking more, now I feel that math.cornell.edu/~hatcher/SO/comments.pdf may not be the answer. It calculates cohomology of the topological space of $SO(n)$. What I am seeking is the group cohomology of $SO(3)$ which is the cohomology of the topological space $BSO(3)$ -- the classifying space of $SO(3)$. $\endgroup$ Sep 15 '11 at 0:14

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.