5

I know that the shape that we see it below is homotopy equivalent of wedge of infinite circles,so the fundamental group of it is $\prod _{1}(\vee _{\alpha \in A}S^{1})=\ast _{\alpha \in A }\mathbb{Z}$,but I don't know how should I show that it is wedge of infinite circles,also I can't imagine what is happening for this shape.please help me with your knowledge,thanks.enter image description here

kpax
  • 2,911
  • 1
    I think it's easier to consider this space as the union of infinitely many pieces, where each piece is a torus minus two disjoint disks. From that, one applies Van Kampen. If you're interested, I can show you how to do it. – Ayman Hourieh Mar 01 '14 at 07:35
  • @kpax: How do you know that your surface is homotopy equivalent to the wedge sum of infinitely many circles? – Seirios Mar 01 '14 at 08:57
  • @Seirios,I just saw it http://math.stackexchange.com/questions/254318/fundamental-group-of-an-orientable-surface-of-infinite-genus,and I think its right. – kpax Mar 01 '14 at 09:16
  • @AymanHourieh,it will be great if you show me. – kpax Mar 01 '14 at 09:16
  • What is an infinite circle? You mean infinitely many circles. In practice, the usage of "infinite" is a particularly tricky issue for non-native speakers. – Carl Mummert Mar 02 '14 at 12:01

3 Answers3

7

enter image description here This is a difficult homotopy equivalence to visualize, but I've attempted to draw a picture of what's going on. First you poke a hole in your surface starting at $+\infty$ and pushing in from the right. Similarly poke a hole from the left. You can see this is a homotopy equivalence by analogy with an infinite cylinder, which can be visualized as having two dotted boundary components at $\pm\infty$. This transforms your surface into an infinite strip with a band connecting top and bottom and infinitely many tubes. Now look at the righ-hand side of my picture which shows a square with a tube in the middle. Once you draw in the indicated $1$-cells, the complement is a disk, which you can use to push the top boundary onto the remaining $1$-cells as indicated in the lower right. Do this for infinitely many squares in a row on your strip all at the same time, to get the bottom left picture. From here, it is easy to see this is a wedge of infinitely many circles. Note that you get one "extra" circle, besides the obvious infinitely many pairs.

  • Great pictures and great explanation. I would have been happy with a Whitehead argument and been done with it, but the visualisation makes it very clear. – Dan Rust Mar 10 '14 at 11:32
  • Thank you to whoever awarded me the bounty. I hadn't even noticed the question had a bounty. – Cheerful Parsnip Mar 15 '14 at 04:56
2

Let $X$ be the infinite-holed torus and $G$ its fundamental group. You can write $X$ as an increasing union $X= \bigcup\limits_{n \geq 2} X_n$, where $X_n$ is a $4$-punctured $n$-holed torus; let $G_n$ denote its fundamental group, which is a free group of rank $2n+2$. In the figure below, $X_3$ is represented with a free basis of $G_3$ in red.

enter image description here

Now, you may notice that $G_{n+1}=G_n \ast \mathbb{F}_2$ (in particular, $G_n$ is naturally a subgroup of $G_{n+1}$) and that the inclusion $X_n \hookrightarrow X$ induces an injective homomorphism $G_n \hookrightarrow G$. Therefore, $\bigcup\limits_{n \geq 2} G_n$ is isomorphic to $\mathbb{F}_{\infty}$ and can be viewed as a subgroup of $G$. Because $\bigcup\limits_{n \geq 2} X_n = X$, in fact $G= \bigcup\limits_{n \geq 2} G_n$, and $G$ is a free group of (countably) infinite rank.

Added: 1) To show that $G_n \to G$ is injective, it is sufficient to show that if $c : [0,1] \to X_n$ is a loop such that $c=1$ in $G$ then $c=1$ in $G_n$. If $H : [0,1]^2 \to X$ is a homotopy between $c$ and the trivial loop, by compactness $\mathrm{Im}(H) \subset X_m$ for some $m \geq n$, hence $c=1$ in $G_m$. But $G_n$ is a subgroup of $G_m$ hence $c=1$ in $G_n$.

2) Thanks to the relation $G_{n+1}= G_n \ast \mathbb{F}_2$, there exist loops $a_1, a_2, \dots$ such that $\{a_1,\dots,a_{2+2n}\}$ is a free basis of $G_n$ for all $n \geq 2$. Therefore, $\{a_1,a_2, \dots\}$ is a free basis of $\bigcup\limits_{n \geq 2} G_n$.

Seirios
  • 33,157
  • thank you very much for your explanation. – kpax Mar 01 '14 at 09:19
  • How do you continuously attach $3$-cells to fill up $X$? – Ayman Hourieh Mar 01 '14 at 16:27
  • What is bounded cohomology with coefficients in $\mathbb Z_2$? Aren't all such cochains bounded? – Cheerful Parsnip Mar 01 '14 at 21:23
  • 2
    I think this does have the homotopy type of a wedge of circles. The inclusion of the 1 skeleton is an isomorphism on $\pi_1$ and all higher homotopy groups vanish (since maps of spheres land in finite genus pieces of the space). So by Whitehead's theorem, the inclusion should be a homotopy equivalence. – Cheerful Parsnip Mar 01 '14 at 21:30
  • 1
    Also your filling-in argument is not correct. You need to add 2-cells and 3-cells to fill in, and the 2-cells will affect the fundamental group. – Cheerful Parsnip Mar 01 '14 at 21:32
  • @GrumpyParsnip Yeah, this can't be right. One cannot fill $X$ with $3$-cells alone. Otherwise one would similarly fill a torus minus two disjoint disks and get an incorrect answer for the fundamental group of this space. I added another answer using van Kampen. – Ayman Hourieh Mar 01 '14 at 22:35
  • @Seirios: I'm guessing you are mixing up bounded cohomology which is homotopy invariant with cohomology with compact support which is not. – Cheerful Parsnip Mar 02 '14 at 00:42
  • @GrumpyParsnip: You are right, my answer was definitely wrong... I completely changed it. – Seirios Mar 02 '14 at 08:43
  • @AymanHourieh: I agree and I completely changed my answer. – Seirios Mar 02 '14 at 08:43
  • @Seirios Thank you. You need to justify why the inclusion is injective in this case (it's not always so, but it is here since there is a corresponding retraction). Also this shows that $\pi_1(X)$ contains free subgroups of arbitrarily large finite ranks, but I don't think it shows that it must be free of countably infinite rank. For example, the free group of rank two also contains free subgroups of arbitrary finite ranks. – Ayman Hourieh Mar 02 '14 at 10:50
  • @AymanHourieh: I added clarifications. – Seirios Mar 02 '14 at 11:57
2

Break up the space $X$ into infinitely countably many open pieces $\{X_i\}$, where each piece is homotopy equivalent to a torus minus two disjoint disks, and the intersection of each two consecutive pieces is homotopy equivalent to a circle. Pick a straight line $L$ that extends through all pieces, and pick a base point $y_0$ on this line. Let $Y_i$ be the union of $X_i$ and a small neighborhood of $L$ that deformation retracts onto $L$.

The open cover $\{Y_i\}$ of $X$ satisfies the conditions of the van Kampen theorem. The inclusion map $Y_i \cap Y_{i+1} \hookrightarrow Y_{i}$ maps the generator of $Y_i \cap Y_{i+1}$ to a generator of $Y_i$. The same is true for $Y_i \cap Y_{i+1} \hookrightarrow Y_{i+1}$. All other intersections have trivial fundamental groups. Since each $\pi_1(Y_i)$ is the free group on three generators, it follows that $\pi_1(X)$ is the free group on countably infinitely many generators.

Ayman Hourieh
  • 39,603