Proof of homotopy invariance of homology : any way to make it better?

271 Views Asked by At

Suppose you want to prove that homotopic maps induce the same morphisms in singular homology. One way to do that is the following : you have your homotopy $X\times I \to Y$, apply $Sing$ to it to get $Sing(X)\times Sing(I)\to Sing(Y)$, and then you have a natural map $\Delta^1\to Sing(I)$ which yields $Sing(X)\times \Delta^1\to Sing(Y)$, and since homology only depends on $Sing$, it's clear that it suffices to prove the following :

let $X\times\Delta^1\to Y$ be a simplicial homotopy, then the two maps $X\to Y$ induce the same map in simplicial homology.

To do that, it's again clear that it suffices to show that $X\to X\times \Delta^1$ is a homology isomorphism. One possible method of proof is to build the beginning of simplicial homotopy theory (but some purely formal bits) as follows :

1- Show it has the lifting property against Kan fibrations, by essentially proving the pushout product axiom; for that it suffices to check that for all $n$, $\partial \Delta^n \times \Delta^1\cup_{\partial \Delta^n \times \{0\}} \Delta^n\times \{0\}\to \Delta^n\times \Delta^1$ has the given lifting property. This can be done by hand, by expressing the given map as a composition of pushouts of horn inclusions.

2- Conclude by the small object argument that $X\to X\times \Delta^1$ is a retract of a cell-complex (built on $\Lambda^n_k\to \Delta^n$'s)

3- Prove that "induces an isomorphism on homology" is stable under retracts and pushouts and transfinite compositions.

4- Prove that $\Lambda^n_k$ and $\Delta^n$ both have zero homology.

Of these steps, only 1 and 4 have some "non conceptual" content, they contain some computations (for 1, you need to express explicitely the map as a pushout of horn inclusions, so you need to find those inclusions; for 4 you need to actually compute some homology, albeit quite simple)

But it seems to me that you can do this more quickly, by actually expressing directly $X\to X\times \Delta^1$ as some relative complex of horn inclusions, if that's possible; since this sort of computation is there anyways, if you can do it, you might as well bypass the whole argument.

So my first question is :

Is there an explicit construction of $X\to X\times \Delta^1$ as obtained by pushouts and transfinite compositions (and maybe a retraction) of horn inclusions $\Lambda^n_k \to \Delta^n$ ?

My second question is about the proof I sketched above, specifically about the computational points I noted :

Is there a way to bypass the computations ? More specifically, 1- is there a way to prove that $\partial \Delta^n \times \Delta^1\cup_{\partial \Delta^n \times \{0\}} \Delta^n\times \{0\}\to \Delta^n\times \Delta^1$ has the left lifting property against Kan fibrations without having to find out which horn inclusions do the right thing ? and 2- is there a way to prove that $\Lambda^n_k\to \Delta^n$ is a homology isomorphism without computing both homologies and discovering that they are $0$ ?

(note that for all these questions, the point is to avoid using powerful stuff, because the end goal is to prove some basic property : the homotopy invariance of homology; so of course I can't accept "well $|\Delta^n|$ is contractible and simplicial homology coincides with singular homology so $H_*(\Delta^n) = 0$" as that would help in no way what I'm trying to do)

1

There are 1 best solutions below

0
On BEST ANSWER

That's not an answer to my specific questions, but I just thought that if the end goal is proving that $X\to X\times \Delta^1$ is a homology isomorphism, there's a much quicker way, although perhaps less formal, through Dold-Kan.

Indeed, $\mathbb Z[X\times \Delta^1]\cong \mathbb Z[X]\otimes \mathbb Z[\Delta^1]$ (here $\otimes$ is dimensionwise), and the Dold-Kan correspondance is monoidal, so $C_*(X\times\Delta^1;\mathbb Z)\simeq C_*(X;\mathbb Z)\otimes C_*(\Delta^1;\mathbb Z)$, where now $\otimes$ is the usual chain complex tensor-product.

So we're reduced to computing $C_*(\Delta^1;\mathbb Z)$ up to homotopy (note that $C_*(X;\mathbb Z)$ is a complex of projectives, so $C_*(X;\mathbb Z)\otimes- $ preserves quasi-isomorphisms), but by Dold-Kan this is entirely explicit, and one easily checks $C_*(\Delta^1;\mathbb Z)\simeq \mathbb Z$ (in degree $0$), from which the result immediately follows.