Is the homology of the chain complex of singular simplices in a topological group with point wise addition the usual homology?

288 Views Asked by At

An, at first sight, surprising exercise in the higher homotopy groups is that if your space is a topological group then addition in the higher homotopy groups is the same as point-wise multiplication of representative maps (where we are thinking of elements of $\pi_n(X)$ as maps $(I^n, \partial I^n) \to (X,x_0)$ where $x_0$ is the identity element of $X$, and addition is done in the last component of $I^n$).

Now I'm wondering: is there any analogous result for singular homology?

If $X$ is a topological group then we have a function $ev: C_n(X) \to C_n(X)$ given by $ev(\sum_{i=1}^m \alpha_i \sigma_i)(t) =\sum_{i=1}^n \alpha_i \sigma_i(t)$ for $t \in \Delta^n$. Here $\sigma_i:\Delta^n \to X$ and $\alpha_i = \pm 1$ so $\sum_{i=1}^m \alpha_i \sigma_i$ is a generic element of $C_n(X)$. The only ambiguity in how we chose to express an element of $C_n(X)$ in this way is the order of the sum and canceling pairs, and this considered, $ev$ is still well defined.

The best I can hope for here is that $ev$ sends homologous cycles to homologous cycles. This seems to work for the circle at least with reduced homology.

I believe an equivalent result would be that in a topological group we can redefine the chain complex so that $C_n(X)=\{\sigma: \Delta^n \to X\}$, with addition of simplices being point-wise addition of maps and boundary maps now $ev\circ \partial$, and the homology would still be the same. Note that the zero element of $C_n(X)$ would be the constant map at the identity element. This is a legitimate chain complex because $ev \circ \partial \circ ev \circ \partial=ev \circ \partial \circ \partial$ using the relation $ev \circ \partial \circ ev =ev \circ \partial$, although a more convincing argument may be to just run through the argument that $\partial^2 =0$ applying $ev$ at every step and observing that the same cancellations happen.

If there isn't any analogous result then I would like to know why not.

1

There are 1 best solutions below

0
On BEST ANSWER

You can do what you suggest: if $X$ is a topological abelian group, then you can form a chain complex $A_\bullet$ where $A_n$ is the set of all maps $\Delta^n\to X$, which form a group under pointwise addition of maps. The boundary map is given by taking the usual boundary of an $n$-simplex as an alternating sum of $(n-1)$-simplices and then adding those $(n-1)$-simplices together pointwise, just as you suggest with your $ev$ map.

However, the homology of this chain complex $A_\bullet$ is not the homology of the space $X$! Instead, $H_n(A_\bullet)$ is naturally isomorphic to the $n$th homotopy group $\pi_n(X)$. The proof is a bit complicated but essentially amounts to showing that in computing the homology of $A_\bullet$, you can restrict to just those simplices which map all but one of their boundary faces to the identity element of $X$. When you restrict to such simplices, a cycle is then just a simplex $\Delta^n\to X$ which maps the entire boundary $\partial\Delta^n$ to the identity element of $X$, which induces a map $\Delta^n/\partial\Delta^n\cong S^n\to X$, and thus an element of $\pi_n(X)$. You can find the details in Section III.2 of Goerss and Jardine's Simplicial Homotopy Theory (the specific result is Corollary 2.7, taking $A$ to be the singular simplicial set of $X$).