After reading Cohen and Voronov's notes on string topology, one can find the following construction: Suppose we have a topological space $X$ with continuous action of $S^1$. This means we have a map $\rho: S^1 \times X \to X$. If we choose a fundamental class $[S^1]$ of the circle, we can form the operator $\Delta: H_\ast(X) \to H_{\ast+1}(X)$ by setting $\Delta(a) = \rho_\ast([S^1] \times a)$. For dimensional reasons, $\Delta$ squares to zero and therefore turns $H_\ast(X)$ into a cochain complex. Let's denote the cohomology of this complex by $H_\ast^\Delta(X)$. Because in this cohomology we have as representatives homology classes in $X$ which are annihilated by the action of $S^1$ "in a homological sense", one would think this is relevant to the $S^1$-equivariant homology of $X$.

However, a few simple examples show that $\Delta$-cohomology and equivariant homology are certainly not equal. For example, take the point with trivial action, then $H^\Delta_\ast(pt)$ is $\mathbb{Z}$ in degree 0 and zero in all other degrees, but $H^{S^1}_\ast(pt) = \mathbb{Z}[a_2]$ where $|a_2| = 2$. Another example is $LS^1$.

So my question is: is there a different (hopefully geometric) description of $\Delta$-cohomology? Is it related to $S^1$-equivariant homology in any way?

If this is not possible in the general case, is it at least possible for $X$ of the form $LM$, with $M$ a manifold? My main motivation for considering $\Delta$-cohomology is string topology, where $\Delta$ is also known as the BV-operator.


1 Answer 1


By $G$-equivariant homology of $X$ you mean the homology of the Borel construction or homotopy orbit space $EG\times_GX$. (I only mention this because the same phrase can refer to other kinds of homology for $G$-spaces, satisfying a weaker kind of homotopy axiom. This "Borel homology" gives you an isomorphism for every $G$-equivariant map $X\to Y$ that is nonequivariantly a homotopy equivalence.)

Yes, the operator that you call $\Delta$ is related to $S^1$-homology. There is a first-quadrant spectral sequence for the $G$-homology of $X$, coming from the fact that $EG\times _GX$ is a bundle over $BG$ with fiber $X$. The $E^2$ is $H_i(BG;H_j(X)$. It is a spectral sequence of modules for the ring $H^{-*}BG$. When $G=S^1$, the groups are $E^2_{2k,j}=H_j(X)$ if $k\ge 0$ and otherwise $0$, and the differential $d^2_{2k,j}\to E^2_{2k-2,j+1}$ is independent of $k$. This first differential $d^2$ is the operator $\Delta$. The $2$-periodicity continues in $E^r$ until the differentials start crossing the vertical axis. So basically your $\Delta$-homology is $E^3=E^4$ and then there is an operator $d^4$ going from $ker(\Delta)/im(\Delta)$ to $ker(\Delta)/im(\Delta)$ and raising $j$ by $3$; and so on.

This is related to cyclic homology.

  • 2
    $\begingroup$ Can one get skupers's $\Delta$ composing the two maps (the only ones with the correct domain and codomain) in the homology Gysin sequence for the fibration $S^1\to X\times EG\to X\times_{S^1}EG$? $\endgroup$ Jul 28, 2010 at 4:19
  • $\begingroup$ Yes, that's true, too. $\endgroup$ Jul 28, 2010 at 10:34
  • $\begingroup$ Yes, I did mean the Borel construction. $\endgroup$
    – skupers
    Jul 28, 2010 at 10:36
  • 1
    $\begingroup$ An example of why $\Delta$ really is naive is to consider the case of $S^3$ acted on by $S^1$ the Hopf fibration or acted upon trivially in both cases $\Delta$ is 0. But equivariant cohomology distinguishes these two cases, the difference is reflected in higher differentials in a spectral sequence. From the point of view of cyclic homology, $\Delta$ should be basically the Connes operator, but you need to remember it's action at the chain level to compute cyclic homology, again because of the possibility of higher operations. $\endgroup$ Jul 29, 2010 at 12:45

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.