Difference between revisions of "Fundamental group"

m (split off from algebraic topology article)
 
(remove nonexistent category)
 
(5 intermediate revisions by one other user not shown)
Line 1: Line 1:
Perhaps the simplest object of study in algebraic topology is the [[fundamental group]]. Let <math>X</math> be a [[path-connected]] topological space, and let <math>x\in X</math> be any point. Now consider all possible "loops" on <math>X</math> that start and end at <math>x</math>, i.e. all [[continuous function]]s <math>f:[0,1]\to X</math> with <math>f(0)=f(1)=x</math>. Call this collection <math>L</math>. Now define an [[equivalence relation]] <math>\sim</math> on <math>L</math> by saying that <math>p\sim q</math> if there is a continuous function <math>g:[0,1]\times[0,1]\to X</math> with <math>g(a,0)=p(a)</math>, <math>g(a,1)=q(a)</math>, and <math>g(0,b)=g(1,b)=x</math>. We call <math>g</math> a [[homotopy]]. Now define <math>\pi_1(X)=L/\sim</math>. That is, we equate any two elements of <math>L</math> which are equivalent under <math>\sim</math>.
+
Perhaps the simplest object of study in algebraic topology is the '''fundamental group'''.  
  
Unsurprisingly, the fundamental group is a group. The [[identity]] is the [[equivalence class]] containing the map <math>1:[0,1]\to X</math> given by <math>1(a)=x</math> for all <math>a\in[0,1]</math>. The [[Function/Introduction#The_Inverse_of_a_Function | inverse]] of a map <math>h</math> is the map <math>h^{-1}</math> given by <math>h^{-1}(a)=h(1-a)</math>. We can compose maps as follows: <math>g\cdot h(a)=\begin{cases} g(2a) & 0\le a\le 1/2, \\ h(2a-1) & 1/2\le a\le 1.\end{cases}</math> One can check that this is indeed [[well-defined]].
+
Let <math>(X,x_0)</math> be a [[based topological space|based]], [[topological space]] (that is, <math>X</math> is a topological space, and <math>x_0\in X</math> is some point in <math>X</math>). Note that some authors will require <math>X</math> to be [[path-connected]]. Now consider all possible "loops" on <math>X</math> that start and end at <math>x_0</math>, i.e. all [[continuous function]]s <math>f:[0,1]\to X</math> with <math>f(0)=f(1)=x_0</math>. Call this collection <math>\Omega(X,x_0)</math> (the '''loop space''' of <math>X</math>). Now define an [[equivalence relation]] <math>\sim</math> on <math>\Omega(X,x_0)</math> by saying that <math>f\sim g</math> if there is a (based) [[homotopy]] between <math>f</math> and <math>g</math> (that is, if there is a continuous function <math>F:[0,1]\times[0,1]\to X</math> with <math>F(a,0)=f(a)</math>, <math>F(a,1)=g(a)</math>, and <math>F(0,b)=F(1,b)=x_0</math>). Now let <math>\pi_1(X,x_0)=\Omega(X,x_0)/\sim</math> be the set of equivalence classes of <math>\Omega(X,x_0)</math> under <math>\sim</math>.
 +
 
 +
Now define a [[binary operation]] <math>\cdot</math> (called ''concatenation'') on <math>\Omega(X,x_0)</math> by
 +
<math>(g\cdot h)(a)=\begin{cases} g(2a) & 0\le a\le 1/2, \\ h(2a-1) & 1/2\le a\le 1.\end{cases}</math>
 +
One can check that if <math>f\sim f'</math> and <math>g\sim g'</math> then <math>f\cdot g\sim f'\cdot g'</math>, and so <math>\cdot</math> induces a well-defined binary operation on <math>\pi_1(X,x_0)</math>.
 +
 
 +
One can now check that the operation <math>\cdot</math> makes <math>\pi_1(X,x_0)</math> into a group. The identity element is just the constant loop <math>e(a) = x_0</math>, and the inverse of a loop <math>f</math> is just the loop <math>f</math> traversed in the opposite direction (i.e. the loop <math>\bar f(a) = f(1-a)</math>). We call <math>\pi_1(X,x_0)</math> the '''fundamental group''' of <math>X</math>.
  
 
Note that the fundamental group is not in general [[abelian group|abelian]]. For example, the fundamental group of a figure eight is the [[free group]] on two [[generator]]s, which is not abelian. However, the fundamental group of a circle is <math>{\mathbb{Z}}</math>, which is abelian.
 
Note that the fundamental group is not in general [[abelian group|abelian]]. For example, the fundamental group of a figure eight is the [[free group]] on two [[generator]]s, which is not abelian. However, the fundamental group of a circle is <math>{\mathbb{Z}}</math>, which is abelian.
Line 13: Line 19:
  
 
If <math>\circ,\cdot</math> share the same unit <math>e</math> (such that <math>a \cdot e = e \cdot a = a \circ e = e \circ a = a</math>) then <math>\cdot = \circ</math> and both are abelian.
 
If <math>\circ,\cdot</math> share the same unit <math>e</math> (such that <math>a \cdot e = e \cdot a = a \circ e = e \circ a = a</math>) then <math>\cdot = \circ</math> and both are abelian.
 +
 +
== Independence from base point ==
 +
 +
At this point, one might wonder how significant the choice of base point, <math>x_0</math>, was. As it turns out, as long as <math>X</math> is path-connected, the choice of base point is irrelevant to the final group <math>\pi_1(X,x_0)</math>.
 +
 +
Indeed, pick consider any other base point <math>x_1</math>. As <math>X</math> is path connected, we can find a path <math>\alpha</math> from <math>x_0</math> to <math>x_1</math>. Let <math>\bar\alpha(t) = \alpha(1-t)</math> be the reverse path from <math>x_1</math> to <math>x_0</math>. For any <math>f\in\Omega(X,x_0)</math>, define <math>\varphi_\alpha = \bar\alpha\cdot f\cdot\alpha \in\Omega(X,x_1)</math> by
 +
<cmath>\varphi_\alpha(f)(t) = (\bar\alpha\cdot f \cdot \alpha)(t) = \begin{cases} \bar\alpha(t) & 0\le t\le 1/3, \\
 +
f(3a-1) & 1/3\le t\le 2/3,\\
 +
\alpha(3t-2) & 2/3\le t\le 1.
 +
\end{cases}</cmath>
 +
One can now easily check that <math>\varphi_\alpha</math> is in fact a well-defined map <math>\pi_1(X,x_0)\to\pi_1(X,x_1)</math>, and furthermore, that it is a [[group homomorphism|homomorphism]]. Now we may similarly define the map <math>\varphi_{\bar\alpha}:\pi_1(X,x_1)\to\pi_1(X,x_0)</math> by <math>\varphi(g) = \alpha\cdot g\cdot\bar\alpha</math>. One can now easily verify that <math>\varphi_{\bar\alpha}</math> is the inverse of <math>\varphi_\alpha</math>. Thus <math>\varphi_\alpha</math> is an [[isomorphism]], so <math>\pi_1(X,x_0)\cong \pi_1(X,x_1)</math>.
 +
 +
Therefore (up to isomorphism), the group <math>\pi_1(X,x_0)</math> is independent of the choice of <math>x_0</math>. For this reason, we often just write <math>\pi_1(X)</math> for the fundamental of <math>X</math>.
 +
 +
== Functoriality ==
 +
 +
Given a (based) continuous map <math>f:(X,x_0)\to(Y,y_0)</math> (that is, <math>f</math> is continuous and <math>f(x_0) = y_0</math>), one may define a group homomorphism <math>f_*:\pi_1(X,x_0)\to\pi_1(Y,y_0)</math> by sending each loop <math>a\in\pi_1(X,x_0)</math> to <math>f\circ a\in \pi_1(Y,y_0)</math>. It is easy to see that <math>f_*</math> sends homotopic loops to homotopic loops (indeed  if <math>F</math> is a homotopy from <math>a</math> to <math>a'</math>, then <math>f\circ F</math> is a homotopy from <math>f\circ a</math> to <math>f\circ a'</math>), and thus <math>f_*</math> is a well-defined map. Also <math>f_*</math> clearly preserves concatenations, so <math>f_*</math> is indeed a homomorphism.
 +
 +
Furthermore, it is easy to see that if <math>f</math> and <math>g</math> are maps <math>f:(X,x_0)\to(Y,y_0)</math> and <math>g:(Y,y_0)\to (Z,z_0)</math>, then:
 +
<cmath>(g\circ f)_* = g_*\circ f_*,</cmath>
 +
and if <math>1_X</math> is the identity map on <math>(X,x_0)</math> and <math>1_{\pi(X,x_0)}</math> is the identity map on <math>\pi_1(X,x_0)</math>, then
 +
<cmath>(1_X)_* = 1_{\pi_1(X,x_0)}.</cmath>
 +
Thus we may in fact regard <math>\pi_1</math> as a (covariant) [[functor]] from the [[category of based topological spaces]] to the [[category of groups]].
 +
 +
One can also show that the induced map <math>f_*</math> depends only on the homotopy type of <math>f</math>, that is if <math>f,g:(X,x_0)\to (Y,y_0)</math> are (based) homotopic maps that <math>f_* = g_*</math>. Indeed, for any loop <math>a\in\pi_1(X,x_0)</math>, if <math>F</math> is a based homotopy from <math>f</math> to <math>g</math>, then <math>F\circ a</math> is a based homotopy from <math>f\circ a</math> to <math>g\circ a</math>, and thus <math>f\circ a = g\circ a</math> in <math>\pi_1(Y,y_0)</math>.
 +
 +
== Homotopy invariance ==
 +
 +
In order for the fundamental group to be a useful topological concept, any two spaces that are topologically "the same" must have the same fundamental group. Specifically, if <math>X</math> and <math>Y</math> are [[homeomorphic spaces|homeomorphic]] then <math>\pi_1(X)</math> and <math>\pi_1(Y)</math> are isomorphic.
 +
 +
We will in fact show that <math>\pi_1(X)</math> and <math>\pi_1(Y)</math> are isomorphic  if <math>X</math> and <math>Y</math> satisfy the weaker notion of equivalence: [[homotopy equivalence]].
 +
 +
Say that <math>(X,x_0)</math> and <math>(Y,y_0)</math> are based homotopy equivalent (<math>(X,x_0)\simeq (Y,y_0)</math>) with homotopy equivalences <math>f:(X,x_0)\to (Y,y_0)</math> and <math>g:(Y,y_0)\to (X,x_0)</math>. (By definition, this means that <math>g\circ f \simeq 1_X</math> and <math>f\circ g\simeq 1_Y</math>.) Now consider the induced maps <math>f_*:\pi_1(X,x_0)\to\pi_1(Y,y_0)</math> and <math>g_*:\pi_1(Y,y_0)\to\pi_1(X,x_0)</math>. From the previous section we get that:
 +
<cmath>f_*\circ g_* = (f\circ g)_* = (1_Y)_* = 1_{\pi_1(Y,y_0)}</cmath>
 +
and
 +
<cmath>g_*\circ f_* = (g\circ f)_* = (1_X)_* = 1_{\pi_1(X,x_0)}.</cmath>
 +
Therefore <math>g_*</math> is the inverse of <math>f_*</math>, so in particular <math>f_*</math> must be an isomorphism. Hence <math>\pi_1(X,x_0)\cong \pi_1(Y,y_0)</math>.
 +
 +
This gives us a very useful method for distinguishing topological spaces: if <math>X</math> and <math>Y</math> are topological spaces whose fundamental groups are not not isomorphic then <math>X</math> and <math>Y</math> cannot be homeomorphic (and in fact, they cannot be homotopy equivalent). For instance, one can show that <math>\pi_1(S^1)\cong\mathbb Z</math> and <math>\pi_1(S^2) \cong 0</math> (where <math>S^n</math> is the [[n-sphere]]), and hence a circle is not homeomorphic to a sphere.
 +
[[Category:Topology]]

Latest revision as of 20:11, 23 January 2017

Perhaps the simplest object of study in algebraic topology is the fundamental group.

Let $(X,x_0)$ be a based, topological space (that is, $X$ is a topological space, and $x_0\in X$ is some point in $X$). Note that some authors will require $X$ to be path-connected. Now consider all possible "loops" on $X$ that start and end at $x_0$, i.e. all continuous functions $f:[0,1]\to X$ with $f(0)=f(1)=x_0$. Call this collection $\Omega(X,x_0)$ (the loop space of $X$). Now define an equivalence relation $\sim$ on $\Omega(X,x_0)$ by saying that $f\sim g$ if there is a (based) homotopy between $f$ and $g$ (that is, if there is a continuous function $F:[0,1]\times[0,1]\to X$ with $F(a,0)=f(a)$, $F(a,1)=g(a)$, and $F(0,b)=F(1,b)=x_0$). Now let $\pi_1(X,x_0)=\Omega(X,x_0)/\sim$ be the set of equivalence classes of $\Omega(X,x_0)$ under $\sim$.

Now define a binary operation $\cdot$ (called concatenation) on $\Omega(X,x_0)$ by $(g\cdot h)(a)=\begin{cases} g(2a) & 0\le a\le 1/2, \\ h(2a-1) & 1/2\le a\le 1.\end{cases}$ One can check that if $f\sim f'$ and $g\sim g'$ then $f\cdot g\sim f'\cdot g'$, and so $\cdot$ induces a well-defined binary operation on $\pi_1(X,x_0)$.

One can now check that the operation $\cdot$ makes $\pi_1(X,x_0)$ into a group. The identity element is just the constant loop $e(a) = x_0$, and the inverse of a loop $f$ is just the loop $f$ traversed in the opposite direction (i.e. the loop $\bar f(a) = f(1-a)$). We call $\pi_1(X,x_0)$ the fundamental group of $X$.

Note that the fundamental group is not in general abelian. For example, the fundamental group of a figure eight is the free group on two generators, which is not abelian. However, the fundamental group of a circle is ${\mathbb{Z}}$, which is abelian.

More generally, if $X$ is an h-space, then $\pi_1(X)$ is abelian, for there is a second multiplication on $\pi_1(X)$ given by $(\alpha\beta)(t) = \alpha(t)\beta(t)$, which is "compatible" with the concatenation in the following respect:

We say that two binary operations $\circ, \cdot$ on a set $S$ are compatible if, for every $a,b,c,d \in S$, \[(a \circ b) \cdot (c \circ d) = (a \cdot c) \circ (b \cdot d).\]

If $\circ,\cdot$ share the same unit $e$ (such that $a \cdot e = e \cdot a = a \circ e = e \circ a = a$) then $\cdot = \circ$ and both are abelian.

Independence from base point

At this point, one might wonder how significant the choice of base point, $x_0$, was. As it turns out, as long as $X$ is path-connected, the choice of base point is irrelevant to the final group $\pi_1(X,x_0)$.

Indeed, pick consider any other base point $x_1$. As $X$ is path connected, we can find a path $\alpha$ from $x_0$ to $x_1$. Let $\bar\alpha(t) = \alpha(1-t)$ be the reverse path from $x_1$ to $x_0$. For any $f\in\Omega(X,x_0)$, define $\varphi_\alpha = \bar\alpha\cdot f\cdot\alpha \in\Omega(X,x_1)$ by \[\varphi_\alpha(f)(t) = (\bar\alpha\cdot f \cdot \alpha)(t) = \begin{cases} \bar\alpha(t) & 0\le t\le 1/3, \\  f(3a-1) & 1/3\le t\le 2/3,\\ \alpha(3t-2) & 2/3\le t\le 1. \end{cases}\] One can now easily check that $\varphi_\alpha$ is in fact a well-defined map $\pi_1(X,x_0)\to\pi_1(X,x_1)$, and furthermore, that it is a homomorphism. Now we may similarly define the map $\varphi_{\bar\alpha}:\pi_1(X,x_1)\to\pi_1(X,x_0)$ by $\varphi(g) = \alpha\cdot g\cdot\bar\alpha$. One can now easily verify that $\varphi_{\bar\alpha}$ is the inverse of $\varphi_\alpha$. Thus $\varphi_\alpha$ is an isomorphism, so $\pi_1(X,x_0)\cong \pi_1(X,x_1)$.

Therefore (up to isomorphism), the group $\pi_1(X,x_0)$ is independent of the choice of $x_0$. For this reason, we often just write $\pi_1(X)$ for the fundamental of $X$.

Functoriality

Given a (based) continuous map $f:(X,x_0)\to(Y,y_0)$ (that is, $f$ is continuous and $f(x_0) = y_0$), one may define a group homomorphism $f_*:\pi_1(X,x_0)\to\pi_1(Y,y_0)$ by sending each loop $a\in\pi_1(X,x_0)$ to $f\circ a\in \pi_1(Y,y_0)$. It is easy to see that $f_*$ sends homotopic loops to homotopic loops (indeed if $F$ is a homotopy from $a$ to $a'$, then $f\circ F$ is a homotopy from $f\circ a$ to $f\circ a'$), and thus $f_*$ is a well-defined map. Also $f_*$ clearly preserves concatenations, so $f_*$ is indeed a homomorphism.

Furthermore, it is easy to see that if $f$ and $g$ are maps $f:(X,x_0)\to(Y,y_0)$ and $g:(Y,y_0)\to (Z,z_0)$, then: \[(g\circ f)_* = g_*\circ f_*,\] and if $1_X$ is the identity map on $(X,x_0)$ and $1_{\pi(X,x_0)}$ is the identity map on $\pi_1(X,x_0)$, then \[(1_X)_* = 1_{\pi_1(X,x_0)}.\] Thus we may in fact regard $\pi_1$ as a (covariant) functor from the category of based topological spaces to the category of groups.

One can also show that the induced map $f_*$ depends only on the homotopy type of $f$, that is if $f,g:(X,x_0)\to (Y,y_0)$ are (based) homotopic maps that $f_* = g_*$. Indeed, for any loop $a\in\pi_1(X,x_0)$, if $F$ is a based homotopy from $f$ to $g$, then $F\circ a$ is a based homotopy from $f\circ a$ to $g\circ a$, and thus $f\circ a = g\circ a$ in $\pi_1(Y,y_0)$.

Homotopy invariance

In order for the fundamental group to be a useful topological concept, any two spaces that are topologically "the same" must have the same fundamental group. Specifically, if $X$ and $Y$ are homeomorphic then $\pi_1(X)$ and $\pi_1(Y)$ are isomorphic.

We will in fact show that $\pi_1(X)$ and $\pi_1(Y)$ are isomorphic if $X$ and $Y$ satisfy the weaker notion of equivalence: homotopy equivalence.

Say that $(X,x_0)$ and $(Y,y_0)$ are based homotopy equivalent ($(X,x_0)\simeq (Y,y_0)$) with homotopy equivalences $f:(X,x_0)\to (Y,y_0)$ and $g:(Y,y_0)\to (X,x_0)$. (By definition, this means that $g\circ f \simeq 1_X$ and $f\circ g\simeq 1_Y$.) Now consider the induced maps $f_*:\pi_1(X,x_0)\to\pi_1(Y,y_0)$ and $g_*:\pi_1(Y,y_0)\to\pi_1(X,x_0)$. From the previous section we get that: \[f_*\circ g_* = (f\circ g)_* = (1_Y)_* = 1_{\pi_1(Y,y_0)}\] and \[g_*\circ f_* = (g\circ f)_* = (1_X)_* = 1_{\pi_1(X,x_0)}.\] Therefore $g_*$ is the inverse of $f_*$, so in particular $f_*$ must be an isomorphism. Hence $\pi_1(X,x_0)\cong \pi_1(Y,y_0)$.

This gives us a very useful method for distinguishing topological spaces: if $X$ and $Y$ are topological spaces whose fundamental groups are not not isomorphic then $X$ and $Y$ cannot be homeomorphic (and in fact, they cannot be homotopy equivalent). For instance, one can show that $\pi_1(S^1)\cong\mathbb Z$ and $\pi_1(S^2) \cong 0$ (where $S^n$ is the n-sphere), and hence a circle is not homeomorphic to a sphere.