We define a space $X$ by $$X=(S^2\times S^2)\cup_{\Delta} D^3$$ where $S^2$ is the $2$-sphere, $D^3$ is the $3$-disk, and $\Delta\colon S^2\to S^2\times S^2$ is the diagonal map, so we attach a $3$-cell to $S^2\times S^2$ using the diagonal.

I want to compute the (co)homology groups of this space, but I don't know how. I would say that this space is not a manifold since dimensions don't match, and it doesn't carry a CW structure, so I'm a bit lost. What do you think it would be the best strategy to solve this problem?

  • 1
    $\begingroup$ So the diagonal element is getting killed... What element the diagonal represents in cohomology of $S^2\times S^2$? $\endgroup$ – Anubhav Mukherjee May 22 '18 at 14:40
  • $\begingroup$ You can make this a CW complex. $\endgroup$ – Cheerful Parsnip May 22 '18 at 17:23
  • $\begingroup$ You might try visualizing what happens for $S^1\times S^1\cup_\Delta D^2$ as a warm-up. $\endgroup$ – Cheerful Parsnip May 22 '18 at 17:25
  • $\begingroup$ How can I make this space a CW complex? $\endgroup$ – Laszlo May 23 '18 at 0:46
  • $\begingroup$ @Laszlo: you have to subdivide the existing cells. That's why I suggested looking at one dimension down as a warm-up to see what needs to be done. $\endgroup$ – Cheerful Parsnip May 23 '18 at 19:28

I'll sketch an argument using the Mayer-Vietoris sequence, which is a little more low-tech than Tyrone's answer, and fleshes out a comment by Anubhav Mukherjee. Let $U$ be a small open neighborhood of $S^2\times S^2$ in your adjunction space $X$ and let $V$ be the interior of the attached $D^3$. Then we have that $H_i(D^3)$ is trivial (let's take reduced homology). Also $U\cap V$ deformation retracts onto the diagonal $S^2$ inside $S^2\times S^2$. Our sequence becomes $$\to H_i(U\cap V)\to H_i(S^2\times S^2)\to H_i(X)\to H_{i-1}(U\cap V)\to$$ Since $H_i(U\cap V)=0$ unless $i=2$, we get that $H_k(S^2\times S^2)\to H_k(X)$ is an isomorphism for $k=1,3,4$. In the middle of the sequence we have $$0\to H_2(U\cap V)\to H_2(S^2\times S^2)\to H_2(X)\to 0.$$ Moreover, we explicitly know the map is induced by the diagonal so we can set things up so that $\mathbb Z\to\mathbb Z\oplus \mathbb Z$ by the diagonal $n\mapsto (n,n)$. So $H_2(X)\cong(\mathbb Z\oplus \mathbb Z)/\langle (n,n)\rangle\cong \mathbb Z$.


We have homotopy cofibration sequences

$S^2\xrightarrow{c}S^2\vee S^2\xrightarrow{(1,-1)}S^2$

$S^3\xrightarrow{w}S^2\vee S^2\xrightarrow{j}S^2\times S^2$

where $c$ is the suspension comultiplication, $w$ is the attaching map of the top cell of the product, and $j$ is the inclusion.

Now $\Delta\simeq j\circ c$, so if we intertwine the previous homotopy cofibration sequences then we find that $X$ has the homotopy type

$X\simeq (S^2\times S^1)\cup_\Delta e^e\simeq S^2\cup_{[1,-1]} e^4$

where $[1,-1]=(1,1)\circ w$ is the Whitehead product. Now the Whitehead product is bilinear, so $[1,-1]=-[1,1]$. Moreover it is well known that $[1,1]=-2\eta$, where $\eta:S^3\rightarrow S^2$ is the Hopf map that generates $\pi_3S^2$. Hence $[1,-1]=2\eta$ and

$X\simeq S^2\cup_{2\eta}e^4.$

We consider cohomology. Thus

$H^nX=\begin{cases}\mathbb{Z}&n=0\\\mathbb{Z}\{x\}&n=2\\\mathbb{Z}\{y\}&n=4\\ 0&\text{otherwise}\end{cases}$

where $x$, $y$ denote generators. From the fact that $\eta$ is the map of Hopf invariant one we conclude that


(you may also want to recall that $\eta$ is the attaching map of the top cell of $\mathbb{C}P^2$).

This calculates the ring structure of $H^*X$. Since the groups are free you can dualise with the universal coefficient theorem to obtain the homology, which I will leave up to you.

  • $\begingroup$ This answer seems very complete but it is too advanced for me. I haven't seen the Whitehead product, nor homotopy cofibrations. $\endgroup$ – Laszlo May 22 '18 at 17:20
  • $\begingroup$ I was worried that might be the case, but wanted to show that the general problem of the space's homotopy type isn't really much more difficult that its homology. $\endgroup$ – Tyrone May 22 '18 at 17:28
  • $\begingroup$ I appreciate your effort -- I will try to understand it if I get a bit of time. $\endgroup$ – Laszlo May 22 '18 at 17:32
  • $\begingroup$ You can essentially take the above as the definition of the Whitehead product. If $f:S^n\rightarrow X$, $g:S^m\rightarrow X$ are maps, then let $w:S^{n+m-1}\rightarrow S^n\vee S^m$ be the attaching map for the top cell of $S^n\times S^m$. Then the composition $[f,g]=(f,g)\circ w:S^{n+m-1}\rightarrow X$ is the Whitehead product of $f$ and $g$. $\endgroup$ – Tyrone May 22 '18 at 17:33
  • $\begingroup$ As for "homotopy cofibration" it essentially means here little more than a cofibration sequence. Each map is not necessarily a cofibration, but if you turn it into one then all the spaces have the right homotopy type. $\endgroup$ – Tyrone May 22 '18 at 17:36

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.