Let $B$ be a cellular (simplicial, semi-simplicial etc) complex having $\mathbb{Z}^n$-symmetry (that is, there is a free action of $\mathbb{Z}^n$ on $B$, commuting with the boundary operators) and let $R$ be a commutative ring. In this case it is possible to establish an isomorphism (as $R$-modules) of the homology modules $H_*(B, R)$ and $H_*(B', R[\mathbb{Z}^n])$, where the "quotient" complex $B'$ is constructed in the following way:

1) In each dimension $d$, choose a representative cell in each orbit of $\mathbb{Z}^n$-action on $B$. Define the chain modules $C_d(B')$ as the free $R[\mathbb{Z}^n]$-modules on the basis of these representatives.

2) Let $e_l$ be a cell of dimension $d$ in $B$ and let $e$ be the corresponding representative in its $\mathbb{Z}^n$-orbit, with $l \in \mathbb{Z}^n$ bringing $e$ to $e_l$. Let also $f$ be the respective basis element in $C_d(B')$. Then the mapping of $e_l$ to $1l \cdot f$ extends to an $R$-module isomorphism $\alpha: C_d(B) \to C_d(B')$ (here $1l \in R[\mathbb{Z}^n]$ stands for the element of the group ring corresponding to $l \in \mathbb{Z}^n$).

3) Since the boundary operators $\partial$ in $B$ commute with the action of $\mathbb{Z}^n$, the operators $\alpha \circ \partial \circ \alpha^{-1}$ are $R[\mathbb{Z}^n]$-linear, and make $B'$ into a differential complex of $R[\mathbb{Z}^n]$-modules. Its homology modules are isomorphic (as $R$-modules) to those of $B$.

My question is: is there a more elegant and natural way (without choosing representatives) to construct this "quotient" complex $B'$ and the corresponding isomorphism of homology modules? Also, I have a feeling that this is a particular case of some more general theory, but do not see which one.

  • $\begingroup$ I am abit confused. Let's take $B=\mathbb{R}$ with the obvioux $\mathbb{Z}$ action. Then what does your $B'$ look like? $\endgroup$ – user43326 Oct 27 '19 at 14:29
  • $\begingroup$ Let us consider the case $R=\mathbb{Z}$. In this case $B'$ consists of 2 free modules of rank 1 (in dimensions 1 and 0) over $\mathbb{Z}[\mathbb{Z}]$. If we represent $\mathbb{Z}[\mathbb{Z}]$ as a ring of Laurent polynomials over a variable $X$, then the differential operator is (in the obvious basis) just the multiplication by $X-1$. Since this operation is injective, we have $H_1=0$, and since its image contains only Laurent polynomials with zero sum of coefficients, $H_0=\mathbb{Z}$, as it should be. $\endgroup$ – p_k Oct 27 '19 at 14:58
  • $\begingroup$ Yes $C_*(B)$ is naturally a free $R[G]$- module, but to explicitly choose a basis one has to pick one cell in each orbit. However it comes equipped with a natural equivalence class of bases given by the construction above. The base change matrices are always permutation matrices with where entries are allowed to be elements from $G\subset R[G]$. Note that such an equivalence class is really an additional structure; not all bases lie in that class. If you want to look further into those ideas, the term Whitehead-group might be helpful. $\endgroup$ – HenrikRüping Oct 27 '19 at 18:44

Suppose a group $G$ acts `properly discontinuously' on a friendly topological space $B$, so that $G$ is acting freely on $B$, and, if we let $B^{\prime} = B/G$, then $B \rightarrow B^{\prime}$ is a covering map. It is a nice exercise in the standard lifting theorem of covering space theory to show that cellular structures on $B^{\prime}$ correspond to $G$--equivariant cellular structures on $B$. (Similarly for simplicial structures, etc.)

If $M$ is a $R[G]$--module, $H_*(B^{\prime};M)$, the homology of $B^{\prime}$ with twisted coefficients $M$, is defined to be the homology of the complex $C_*(B) \otimes_{R[G]} M$. There are two extreme cases. Let $M = R$. As $C_*(B) \otimes_{R[G]} R = C_*(B^{\prime})$, we have $H_*(B^{\prime};R) = H_*(B^{\prime};R)$. (phew!). At the other extreme, let $M = R[G]$. As $C_*(B) \otimes_{R[G]} R[G] = C_*(B)$, we have - yes - $H_*(B^{\prime};R[G]) = H_*(B;R)$.

| cite | improve this answer | |
  • 1
    $\begingroup$ Oh, I just learned two things: the term "homology with twisted coefficients" and the fact that $C_*(B)$ is naturally a free $R[G]$-module. My problem was that I tried to reason in terms of a basis of this module, but this is an unnecessary compication. $\endgroup$ – p_k Oct 27 '19 at 17:50

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.