Problem: Let $M$ and $N$ be $n$-dimensional manifolds, where $n > 2$. Let $M \# N$ be their connected sum. Show that $\pi(M \# N) = \pi(M) \ast \pi(N)$.

RE-EDITED Attempt:

  1. Let $U_2$ and $V_2$ be two small open balls from $M$ and $N$ to be removed by the connected sum operation. Let $p_m \in U_2$ and $p_n \in V_2$, and then let $U_1 = M - \{p_m\}$ and $U_2 = N - \{p_n\}$.

  2. We can now view the connected sum $M\#N$ as the quotient of the disjoint union of $M$ and $N$ by an equivalence relation identifying $U_2 -\{p_m\}$ with $V_2-\{p_n\}$ defined by some homeomorphism between the two. Denote $W_1$ and $W_2$ as the respective images of $U_1$ and $V_1$ under the quotient map.

  3. We have that $U_2$ and $V_2$ are open by construction. Furthermore, since $M$ and $N$ are open, we have that $U_1 = M - \{p_m\}$ and $V_1 = N - \{p_n\}$ are also open (since open sets with a point removed are still open).

  4. Then we have the following open covers of $M$ and $N$ respectively:

    $$ M = U_1 \cup U_2 $$ $$ N = V_1 \cup V_2 $$

  5. We can then express

    $$ M \# N = \underbrace{W_1}_{\text{open}} \cup \underbrace{W_2}_{\text{open}} $$

    as well since $M \cup N = V_1 \cup U_1$ and $W_1$ and $W_2$ are just the images of $U_1$ and $V_1$ under the natural quotient map used to define $M \# N$.

  6. Consider that $W_1 \cap W_2$ is path connected since $W_1 \cap W_2$ is homeomorphic to $U_2 - \{p_m\} \cong V_2 - \{p_n\}$, both of which are punctured disks which deformation retract onto $S^{n-1}$. Since spheres of dimension greater than $2$ are simply connected (hence path connected), we have that $W_1 \cap W_2$ is path connected as well. This will allow us to later apply Van Kampen to $M \# N = W_1 \cup W_2$.

  7. Now we have that

    $$ U_1 \cap U_2 = U_2 - \{p_n\} \cong W_1 \cap W_2 \cong V_2 - \{p_m\} = V_1 \cap V_2 $$

    so that from above we can say

    $$ \pi_1(U_1 \cap U_2) \cong \pi(V_1 \cap V_2) \cong \pi(W_1 \cap W_2) \cong \{e\} $$

  8. Now since $n > 2$, we have that

    $$ \underbrace{\pi_1(U_1) = \pi_1(M - \{p_m\}) \cong \pi_1(M) \cong \pi_1(W_1)}_{\text{removing a point doesn't change fundamental group for $n > 2$}} $$

    and similarly

    $$ \pi_1(V_1) = \pi_1(N - \{p_n\})\cong\pi_1(N) \cong \pi_1(W_2) $$

  9. Then applying Van Kampen on $M \# N = W_1 \cup W_2$ yields that

    $$ \pi_1(W_1) \ast_{\pi_1(W_1 \cap W_2)} \pi_1(W_2) \cong \pi_1(M\#N) $$

  10. But since (8) yields that

    $$ \pi_1(W_1) \ast_{\pi_1(W_1 \cap W_2)} \pi_1(W_2) \cong \pi_1(W_1) \ast_{\{e\}} \pi_1(W_2) \cong \pi_1(W_1) \ast \pi_1(W_2) \cong \pi_1(M) \ast \pi_1(N) $$

    we then have from (9) that

    $$ \pi_1(M) \ast \pi_1(N) \cong \pi_1(M\#N) $$

    as desired.

  • 3
    $\begingroup$ Hint: use Seifert–van Kampen $\endgroup$ – Grigory M May 9 '14 at 10:55
  • 1
    $\begingroup$ $\pi_1(M)$ is a group not a number, so «$\pi_1(M)=\infty$» is not even wrong $\endgroup$ – Grigory M May 9 '14 at 10:56
  • $\begingroup$ @user1770201: What is $\pi$? $\endgroup$ – user99914 May 9 '14 at 11:07
  • $\begingroup$ @John: The fundamental group, or the set of all paths (up to homotopy) of $M$ or $N$. Since $M$ and $N$ are manifolds, I presume our choice of base points don't matter. $\endgroup$ – user1770201 May 9 '14 at 11:11
  • $\begingroup$ In light of the duplicate post flagged by Grigory M and other users, I edited my post above with (what I think is) a complete solution to the problem but with questions flagged on the steps where I'm specifically confused. I appreciate the comments. $\endgroup$ – user1770201 May 9 '14 at 15:28

Let's be a little more precise. The direct sum $M\#N$ is defined as quotient of the disjoint union of $M$ and $N$ by an equivalence relation identifying $U_2 -\{p_m\}$ with $V_2-\{p_n\}$ defined by some homeomorphism between the two. Then denote by $W_1$ the image of $U_1$ under the quotient map and by $W_2$ the image of $V_1$ under the quotient map.

5) $W_1\cap W_2$ is connected because it is homeomorphic to $U_2-\{p_m\}\cong V_2-\{p_n\}$. These are punctured disks, which deformation retract onto spheres. Spheres of dimension greater than $0$ are connected.

(Note that this still works for $n=2$.)

6) See above: everything here intersects in the punctured disk where we glued to create the connect sum. Punctured disks deformation retract to spheres.

7) The space $S^{n-1}$ is definitely not contractible for $n\geq 2$. It is simply connected.

8) Here is where you need the hypothesis $n>2$. For dimension greater than $2$, removing a point does not change the fundamental group. To see why, show that the homomorphism $\pi_1(U_1)\to \pi_1(M)$ induced by the inclusion is actually an isomorphism. The intuition is that if any of the homotopies between generators $\pi_1(M)$ hits the ball you've removed, it can be "moved" so it "goes around" the ball.

(Removing a point does change higher homotopy groups --- $M$ is not homotopy equivalent to $U_1$ --- but does not change $\pi_1$.)

  • $\begingroup$ In your first paragraph, why is the equivalence relation identifying $U_2 - \{p_m\}$ with $V_2 - \{p_n\}$ instead of just $U_2$ with $V_2$? Also in your first paragraph: how do we know what the image of $U_1$ and $V_1$ is under the quotient map for purposes of defining $W_1$ and $W_2$? $\endgroup$ – user1770201 May 11 '14 at 16:58
  • $\begingroup$ @user1770201 Because $U_2$ and $V_2$ are open disk; for a connect sum don't want to glue the two manifolds along a disk, we want to glue them along a collar of the removed disks. $\endgroup$ – Neal May 11 '14 at 17:24
  • $\begingroup$ @user1770201 I'm not sure I understand the question. The image of $U_1$ is the set of equivalence classes in $M\# N$ which have at least one representative in $U_1$. $\endgroup$ – Neal May 11 '14 at 17:26
  • $\begingroup$ Do you have a link which explains in more detail the definition of $M \# N$ you are using? Definitions seem to vary from site to site (wikipedia, proof wiki, etc) and I'm trying to find yours. $\endgroup$ – user1770201 May 12 '14 at 2:21
  • $\begingroup$ @user1770201 They're all the same: Delete a disk, glue along boundaries. :) Mine is equivalent to Wikipedia's in the case of a smooth manifold; the disk $U_i$ is a tubular neighborhood of $p_i$ diffeomorphic to the normal bundle of $p_i$. Wikipedia also specifies a choice of gluing map, which I did not do. $\endgroup$ – Neal May 12 '14 at 2:36

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.