11

I know the fundamental group of the torus is $\mathbb Z^2$. I am wondering is it possible to find a subset of the plane such that the fundamental group of this subspace is $\mathbb Z^2$? I can't seem to come up with an example.

Parcly Taxel
  • 105,904
Dan
  • 521
  • What about two circles with a unique common point? – Stefan Mesken Feb 09 '18 at 13:23
  • 1
    @StefanMesken That would be a free group with two generators. – Dan Rust Feb 09 '18 at 13:23
  • @DanRust Oh, right... – Stefan Mesken Feb 09 '18 at 13:25
  • I'm fairly sure this is impossible; if you set includes a wedge of two circles and not some point within each, then the inclusion of your set into the twice punctured plane gives a surjective homomorphism to the free group on two elements. (I'm not totally sure how one shows this to be the case though - I would bet if you just study the generators of $\mathbb Z^2$, you would find that this situation would have to happen) – Milo Brandt Feb 09 '18 at 13:30
  • What is the intuitive reason why it can't happen in $R^2$? I'm not sure if I see it or not. – Dan Feb 09 '18 at 14:06
  • 2
    This will follow from the paper by Fischer and Zastrow mentioned in https://math.stackexchange.com/questions/36279/is-the-fundamental-group-of-every-subset-of-mathbbr2-torsion-free. You prove that if $A$ is a subset of $R^2$ and $\alpha, \beta\in \pi_1(A,a)$ are commuting elements then they generate a cyclic subgroup of $\pi_1(A,a)$. (It suffices to treat the case of compact subsets of $R^2$.) – Moishe Kohan Feb 09 '18 at 14:59
  • 2
    The "right" claim is that fundamental groups $\pi_1(A,a)$ for subsets $A\subset R^2$ are locally free. This was proven by Fischer and Zastrow for compact subsets (according to Lowther's answer to the linked question). From that, it is easy to prove the claim in general. I will write a proof when/if I have time. – Moishe Kohan Feb 09 '18 at 15:33

1 Answers1

7

The following is likely to be in the literature, but it is easier to prove than to find a reference.

Proposition. If $A\subset R^2$ is a nonempty subset then for every $a\in A$ the fundamental group $\pi_1(A,a)$ is locally free.

(Recall that a group $G$ is called locally free if every finitely generated subgroup of $G$ is free. For instance, the additive group of rational numbers is locally free, but not free.)

I first give an intuitive explanation and then a formal proof. Our intuition comes from connected polygonal subsets $A\subset R^2$, i.e. finite planar simplicial subcomplexes. Every such subcomplex is homotopy-equivalent to a finite graph: just keep collapsing 2-dimensional simplices $s$ in $A$ which have one "free edge" i.e. an edge $e$ such that $e$ is contained in a unique 2-dimensional simplex in $A$, namely, $s$. After finitely many such collapses we obtain a finite connected subgraph $B\subset A$. It is a standard exercise in algebraic topology (using van Kampen's theorem) to show that the fundamental group of a finite connected graph is free. The trouble with this intuition is that there are (even compact) subsets in $R^2$ which are quite different from simplicial complexes, e.g. the topologists's sine curve or the Lakes of Wada continuum, which tend to defy our intuition of planar sets. Below is a formal proof of the proposition.

Proof. First, I need a bit of group theory.

Definition. The rank, $rank(G)$, of a (finitely generated) group $G$ is the minimal number of generators needed to generate $G$. (Note that $G$ need not be free!)

It is clear that if $Q$ is a quotient group of $G$ then $rank(Q)\le rank(G)$. What's less obvious is the following

Lemma. 1 $rank(F_n)=n$ for every $n$.

(This lemma follows by considering the abelianization of $F_n$, which is ${\mathbb Z}^n$; alternatively, it follows from Lemma 2 below.)

Lemma 2. If $\phi: F_n\to F_m$ is a non-injective epimorphism of free groups then $rank(F_m)< rank(F_n)$.

Lemma 2 follows from the Hopfian property of free groups of finite rank.

Given this, let's prove the proposition. The proof is by induction on rank of finitely generated subgroups of $\pi_1(A,a)$. There is nothing to prove for subgroups of rank $0$ (they are trivial), so assume that the claim holds for all subgroups of rank $<n$.

The key fact that we need is the following theorem due to Fischer and Zastrow mentioned by George Lowther in his answer to this question:

Theorem. If $K\subset R^2$ is a compact subset, then the fundamental group $\pi_1(K,x)$ is locally free for every $x\in K$.

Now, suppose that $G$ is a rank $n$ subgroup of $\pi_1(A,a)$. I claim that $G$ is free. Let $g_1,...,g_n$ be generators of $G$, they are represented by loops $L_1,...,L_n$ in $A$ based at $a$. If the elements $g_1,...,g_n$ do not satisfy any nontrivial relation in $\pi_1(A,a)$, then $G$ is free and we are done. Suppose, hence, that there exists $R$, a reduced word in the alphabet $g_i^{\pm 1}$ such that $R$ represents the identity element in $\pi_1(A,a)$. The word $R$ corresponds to a continuous map $f: D^2\to A$ (such that the restriction of $f$ to the boundary of $D$ is a certain concatenation of the loops $L_i$ and their inverses which is given by the word $R$). The union of the images of the loops $L_i$ and the map $f$ is a compact path-connected subset $C\subset A\subset R^2$. Let $h_1,...,h_n\in \pi_1(C,a)$ denote the elements represented by the loops $L_1,...,L_n$.

By the above theorem, $\pi_1(C,a)$ is locally free, hence, the subgroup $H$ of $\pi_1(C,a)$ generated by $h_1,...,h_n$, is free. We then obtain a non-injective epimorphism $$ F_n\to H $$ sending the free generators to $h_1,...,h_n$. By Lemma 2 above, $rank(H)< n$. Hence, by the induction assumption, the image of $H$ in $\pi_1(A,a)$ (under the homomorphism induced by the inclusion map $(C,a)\to (A,a)$), which is our subgroup $G$, is free. qed

Now, this proposition immediately implies that for every planar subset $A\subset R^2$, $\pi_1(A,a)$ cannot contain ${\mathbb Z}^2$, since ${\mathbb Z}^2$ is finitely-generated (actually, 2-generated) and is not free.

Edit. An example of a path-connected (even compact and locally path-connected)) planar subset with locally free but not free fundamental group is given by Hawaiian Earring. I do not know, however, if, ${\mathbb Q}$ is isomorphic to the fundamental group of a planar set.

Moishe Kohan
  • 111,854