# a CW-complex homotopic to a manifold

I'm reading a paper and here the authors say that a connected 4-manifold with zero rational top homology has a homotopy type of 3-dimensional CW-structure. I can't figure out how it can be done.

-
For this to be true, you need to assume that your 4-manifold is orientable. Assuming this, your conditions imply that your 4-manifold is not compact (otherwise the 4th homology group would be Q). It is a general fact that smooth noncompact $n$-manifolds are homotopy equivalent to $(n-1)$-dimensional CW complexes. For details, see Mohan Ramachandran's answer to my question here : mathoverflow.net/questions/18454 – Andy Putman Oct 11 '10 at 19:39
Consider $M=RP^4$ (or any other non-orientable closed 4-manifold). It has $H_4(M;\mathbb Q)=0$, but $H_4(M,\mathbb Z/2)\ne 0$, hence it is not homotopy equivalent to a 3-dimensional CW-complex. – Sergei Ivanov Oct 11 '10 at 19:44
Thank you everybody :) – hjjang Oct 11 '10 at 20:01
@Andy @Sergei Could someone post an answer so that the question gets closed? – Greg Kuperberg Oct 12 '10 at 19:59
Sure Greg, I'll make it into an answer. – Andy Putman Oct 13 '10 at 2:39

For this to be true, you need to assume that your $4$-manifold $M$ is not a compact nonorientable manifold. Otherwise, you would have $H_4(M;\mathbb{Q}) = 0$ but $H_4(M;\mathbb{Z}/2) \neq 0$, so there is no hope that your manifold is homotopy equivalent to a $3$-dimensional CW-complex.
Assuming this, your conditions imply that your $4$-manifold is not compact (otherwise the 4th homology group would be $\mathbb{Q}$). It is a general fact that smooth noncompact $n$-manifolds are homotopy equivalent to $(n−1)$-dimensional CW complexes. For details, see Mohan Ramachandran's answer to my question here.