4

I need help verifying and completing my solution to problem 2.1.19 of Hatcher's book Algebraic Topology.

Calculate the homology groups of the subspace of $I \times I$ consisting of its 4 boundary edges and all the points in its interior with rational first coordinate.

Here is my partial solution: Let $X$ be the given space, let $Y$ be the top and bottom edge, and let $Z = I \cap \mathbb{Q}$. We have $H_k(Y) = 0$ for $k > 0$, so $H_k(X) \approx H_k(X, Y)$ for $k>1$ using the long exact sequence for the pair $(X, Y)$. Note that $(X, Y)$ is a good pair (i.e. $Y$ is a deformation retract of a neighbourhood in $X$), and $X/Y$ is the suspension $SZ$. Therefore, $H_{k+1}(X, Y) \approx \widetilde H_{k+1}(X/Y) \approx \widetilde{H}_k(Z)$, using the relationship between the homology of a space and the homology of its suspension. This gives $H_{k+1}(X) \approx \widetilde{H}_k(Z)$ for $k>0$. The latter is $0$ since $Z$ is totally disconnected, so $X$ has trivial homology in dimensions 2 and above. I don't see an easy way to continue for dimension $1$. Any hints would be appreciated.

abhi01nat
  • 1,611
  • Why is $X/Y$ a wedge of circles topologically? The open sets seem to be different. – Chris H Jul 28 '19 at 09:31
  • @user277182 ah yes of course, thanks for pointing that out. – abhi01nat Jul 28 '19 at 09:39
  • $X/Y$ is not the suspension of $Z$. It is obtained from $\Sigma Z$ by identifying the two suspension points. – Paul Frost Jul 30 '19 at 21:26
  • @PaulFrost I can't believe I did not notice that, thanks for pointing it out. I think in my head I was hastily thinking of the quotient as collapsing Y to two points... Either ways, as you pointed out in your answer I think the method still works. – abhi01nat Jul 31 '19 at 08:19

2 Answers2

5

This space can be described as $$ X = \{(x,y) \in I \times I : x \in \mathbb{Q}\}. $$

Take $A = \{(x,y) \in X : y < 3/4 \}$ and $B = \{(x,y) \in X : y > 1/2\}$ open sets. The lower and upper edges of $I \times I$ are deformation retracts of $A$ and $B$ respectively, so both $A$ and $B$ are contractible. Their intersection is

$$ A \cap B = \{(x,y) \in X : x \in \mathbb{Q} , y \in (1/2,3/4) \} \simeq I \cap \mathbb{Q}, $$

the homotopy given by collapsing each interval to its midpoint. In particular we see that $$H_k(A \cap B) \simeq \bigoplus_{q \in I \cap \mathbb{Q}} H_k(\{q\})$$ for each $k$. Since a point is trivially contractible, the intersection has trivial homology for positive degrees and $H_0(A \cap B) \simeq \mathbb{Z}^{(\mathbb{N})}$.

Thus, using Mayer-Vietoris we have the following exact sequence of reduced homology:

$$ 0 \to \widetilde{H}_1(X) \to \widetilde{H}_0(A \cap B) \to \widetilde{H}_0(A) \oplus \widetilde{H}_0(B) \to \widetilde{H}_0(X)\ \to 0. $$

Since both $A$, $B$ and $X$ are path connected their (reduced) homology at degree zero vanishes, so

$$ \widetilde{H}_1(X) \simeq \widetilde{H}_0(A \cap B) \simeq \mathbb{Z}^{(\mathbb{N})} $$

and thus $H_1(X) \simeq \mathbb{Z}^{(\mathbb{N})}$.

qualcuno
  • 17,121
  • It is not true that $A \cap B$ is the disjoint union of open intervals (the topology is different). But this is irrelevant, you have$A \cap B \simeq Z = I \cap \mathbb Q$. Thus $H_k(A \cap B) \approx \bigoplus_{q \in Z}H_k({q})$ since the ${q}$ are the path components of $Z$. – Paul Frost Jul 30 '19 at 21:35
  • @PaulFrost fixed it, thanks! I'd appreciate a sanity check though, just in case. – qualcuno Jul 30 '19 at 21:43
  • Two remarks: 1. $A$ and $C$ are not deformation retracts of the lower and upper edges respectively - the converse is true. 2. The direct sum should be taken over $I \cap \mathbb Q$. – Paul Frost Jul 30 '19 at 21:47
  • Final comment: You should write "... hence $A$ and $B$ are contractible". – Paul Frost Jul 30 '19 at 21:52
  • It's not my day today when it comes to wording :/ I think I've fixed it now. As for the sum, even though it's clearer to take the sum over $I \cap \mathbb{Q}$, at the end of the day the former was equivalent, no? As we only have to care about the index being countable. – qualcuno Jul 30 '19 at 21:55
  • Don't worry, your idea was absolutely correct! And for the sum you are right. But I think it is clearer to refer to the path components as the index set. – Paul Frost Jul 30 '19 at 22:00
  • Agreed, it's clearer that way. Thanks again for your help. – qualcuno Jul 30 '19 at 22:03
1

I think Guido A.'s answer is most elegant, but also your approach can be made working.

Let us first observe that if $(X,A)$ is a good pair with a contractible $A$, then the projection $q : X \to X/A$ induces isomorpisms $q_* : H_k(X) \to H_k(X/A)$ for all $k$. This follows from Hatcher's Proposition 2.22. Just consider the long exact sequences of the pairs $(X,A)$ and $(X/A,A/A)$ which are connected "levelwise" by the induced maps $q_* : H_k(A) \to H_k(A/A)$ (which are isomorphisms), $q_* : H_k(X) \to H_k(X/A)$ and $q_* : H_k(X,A) \to H_k(X/A,A/A)$ (which are isomorphisms). Now the Five Lemma applies.

Let $B$ and $T$ the bottom and top edges of $X$, respectively. They are contractible. Consider the quotient maps $p : X \to X/B$ and $q : X/B \to (X/B)/T = S Z$. Both $(X,B)$ and $(X/B,T)$ are good pairs. Now the above result applies to show that $q \circ p : X \to S Z$ induces isomorphisms in homology.

But then we can use $\tilde{H}_{k+1}(S Z) \approx \tilde{H}_k(Z)$.

Paul Frost
  • 76,394
  • 12
  • 43
  • 125