What is special about the homology of the compact spaces, what are the most elementary properties that are verified by homology of compact spaces that are not true for the homology of non compact spaces ? I'm asking because i read that Euler characteristic is not easily defined for non compact spaces while it is straightforward for compact spaces and i also read that we have a homology theory that is better suited for noncompact spaces rather than singular homology, what is making this difference between compact and noncompact spaces regarding homology and cohomology ? Thank you for your help !

  • 2
    $\begingroup$ When your spaces are compact, you have that various (co)homology groups are finite dimensional (necessary for the usual definitions of Euler characteristic), and also that you can integrate cochains over chains to make a pairing between homology and cohomology. Depending on what context you are working in (simplicial sets/complexes, CW-complexes, manifolds, topological spaces, etc.), you may be able to talk about (co)homology with compact support, or Borel-Moore (co)homology or some other theory that makes things work better. $\endgroup$ – Aaron Nov 9 '17 at 17:19
  • 2
    $\begingroup$ I thought the homology groups are finite dimensional when $X$ is compact manifold but when $X$ is a compact space i'm not able to find the result in Hatcher or other references, could you please give me a reference to read on this fact, thanks in advance $\endgroup$ – palio Nov 9 '17 at 17:48
  • $\begingroup$ It depends what you mean by space. Take a product of a cantor set and a circle. This space will have infinite rank $H_0$ and $H_1$ and it is clearly compact. Obviously, it is not homotopy equivalent to a CW complex however, so you may want to restrict to this class. (or if you don't like a Cantor set, take the one-point compactification of $\mathbb{N}$ as your first factor for the product). $\endgroup$ – Dan Rust Nov 9 '17 at 17:53
  • $\begingroup$ My statement might not have been correct in full generality. I do not have a reference, and so if I misspoke, I apologize. It has been a long time since I have thought about general topological space. $\endgroup$ – Aaron Nov 9 '17 at 17:53

There are essentially no special properties of (singular) homology and cohomology of compact spaces, in general. A concrete result along these lines is that every topological space which is not too large (where "too large" is so large you cannot even prove such a large set exists in ZFC) is weak homotopy equivalent to a compact Hausdorff space. Since a weak homotopy equivalence induces isomorphisms on singular (co)homology, this means anything you that can happen with them in general spaces can also happen in compact Hausdorff spaces (as long as your spaces aren't ridiculously huge). To be precise, "not too large" means "smaller than the least measurable cardinal". This is a theorem of Adam Przeździecki; see this MO post and the comments under it.

When you restrict to sufficiently nice spaces, though, (co)homology of compact spaces tends to be better behaved since it is guaranteed to be finitely generated. For instance, if your space is a CW-complex, then it is compact iff it has only finitely many cells, and if it has only finitely many cells its (co)homology will be finitely generated since the cellular (co)chain complex is finitely generated. This finiteness is required, for instance, in order for the usual definition of Euler characteristic to make sense.

Another difference is that sometimes you are interested not in ordinary cohomology but in cohomology with compact support (or other related variations), which arises naturally in studying things like Poincaré duality (in the context of de Rham cohomology, for instance, it arises naturally because you can integrate a form over a compact set but over a non-compact set the integral may diverge). For compact spaces, cohomology with compact support is the same thing as ordinary cohomology, so you don't need to make a distinction between them.


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.