Published on

Geometric Topology 16

Authors

Geometric Topology 16

Paths in Topological Spaces

From now on XX will denote a topological space. Here are some examples:

  • Euclidean space Rn\mathbb{R}^n, including the real line, the plane and ordinary space (n=1,2,3n=1,2,3). The Argand plane C\mathbb{C}, which unlike R2\mathbb{R}^2 is 'pre-oriented'.
  • The nn-dimensional sphere Sn={vRn+1:v=1}S^n = \{\underline{v} \in \mathbb{R}^{n+1} : \|\underline{v}\| = 1\}, which is compact. Basic topology asserts that SnS^n minus a point is homeomorphic to Rn\mathbb{R}^n.
  • The nn-dimensional torus TnT^n, which is homeomorphic to the product S1××S1S^1 \times \cdots \times S^1 of nn circles, and to the quotient (group and top. space) Rn/Zn\mathbb{R}^n /\mathbb{Z}^n. The 33-torus T3T^3 is obtained from a solid cube by identifying opposite sides.
  • Any graph in R2\mathbb{R}^2 or R3\mathbb{R}^3, including the figure-eight ""q(P)"\infty" \approx q(\partial \mathscr{P}) arising from the boundary code aba1b1aba^{-1}b^{-1} of the square P\mathscr{P} representing TT.
  • Any surface, in particular an orientable MY,g,r\mathscr{M}_{Y,g,r} or a non-orientable MN,h,r\mathscr{M}_{N,h,r}.
  • More general quotients of polygons like xxx'xxx' that are not locally Euclidean.
  • The complement of a knot (this 33-manifold is an open subset of R3\mathbb{R}^3 or S3S^3).

Joining Paths

Let XX be a topological space, x0,x1Xx_0,x_1 \in X. Let I=[0,1]I = [0,1].

Definition: A path from x0x_0 to x1x_1 is a continuous map α:IX\alpha : I \rightarrow X such that α(0)=x0\alpha(0) = x_0 and α(1)=x1\alpha(1) = x_1. There is no requirement for α\alpha to be injective. If x0=x1x_0 = x_1 then it is called a loop based at x0x_0. Once x0x_0 is fixed, sx0 (s)s \mapsto x_0 \ (\forall s) defines the constant loop ε\varepsilon at x0x_0.

Assume that XX is path-connected, i.e. any two points of XX can be joined by a path. This implies that XX is connected; these two properties are equivalent for surfaces.

If α\alpha is a path from x0x_0 to x1x_1, and β\beta is a path from x1x_1 to x2x_2, then αβ\alpha \beta will denote the concatenated path, defined by traversing α\alpha then β\beta both at double speed:

s{α(2s)if s[0,12]β(2s1)if s[12,1].s \mapsto \begin{cases} \alpha(2s) & \text{if} \ s \in [0,\frac{1}{2}] \\ \beta(2s -1) & \text{if} \ s \in [\frac{1}{2}, 1]. \end{cases}

This is continuous: its image has no 'gap'.

We can define αβγ\alpha \beta \gamma as (αβ)γ(\alpha \beta) \gamma, or else divide into thirds (it won't really matter, we will see later). We also define a 'backwards path' α1\alpha^{-1} by α1(s)=α(1s)\alpha^{-1}(s) = \alpha(1-s).

Homotopy of Paths

Definition: Fix x0,x1Xx_0,x_1 \in X. Two paths α,β\alpha,\beta from x0x_0 to x1x_1 are path-homotopic if there exists a continuous mapping H:I2XH: I^2 \rightarrow X such that:

H(s,t)={α(s)if t=0β(s)if t=1, and  {x0if s=0x1if s=1.H(s,t) = \begin{cases} \alpha (s) & \text{if} \ t = 0 \\ \beta (s) & \text{if} \ t = 1, \end{cases} \ \text{and } \ \begin{cases} x_0 & \text{if} \ s=0 \\ x_1 & \text{if} \ s=1. \end{cases}

We will write αβ\alpha \approxeq \beta or αHβ\alpha \approxeq_{H} \beta.

The map HH represents the deformation of the paths fixing their endpoints, and it helps to write H(s,t)=Ht(s)H(s,t) = H_t(s) so H0=αH_0 = \alpha and H1=βH_1 = \beta.

Ordinary homotopy (αβ)(\alpha \simeq \beta) would only insist on the left-hand equations, and then any path starting from x0x_0 would be homotopic to ε\varepsilon by setting H(s,t)=α((1t)s)H(s,t) = \alpha((1-t)s).

Lemma 0: \approxeq is an equivalence relation on the set of all paths from x0x_0 to x1x_1. We will denote such an equivalence class by [α][\alpha].

Proof: To see αα\alpha \approxeq \alpha, take F(s,t)=α(s) t.F(s,t) = \alpha(s) \ \forall t. To see that αFβ\alpha \approxeq_F \beta implies βα\beta \approxeq \alpha, take:

G(s,t)=F(s,1t).G(s,t) = F(s,1-t).

To see that αFβ\alpha \approxeq_F \beta and βGγ\beta \approxeq_G \gamma implies that αγ\alpha \approxeq \gamma, take:

H(s,t)={F(s,2t) if t[0,12]G(s,2t1) if t[12,1],H(s,t) = \begin{cases} F(s,2t) & \ \text{if} \ t \in [0,\frac{1}{2}] \\ G(s,2t-1) & \ \text{if} \ t \in [\frac{1}{2},1], \end{cases}

so this time speed is doubled vertically. To see that HH is continuous, it suffices to show that if CC is a closed subset of XX then H1(C)H^{-1}(C) is a closed subset of II. Let A=I×[0,12]A = I \times [0,\frac{1}{2}] and B=I×[12,1]B = I \times [\frac{1}{2},1], noting that the squashed versions F~,G~\tilde{F},\tilde{G} agree on ABA \cap B (where t=12t = \frac{1}{2}). Then:

H1(C)=(F~1(C)A)(G~1(C)B)H^{-1}(C) = (\tilde{F}^{-1}(C) \cap A) \cup (\tilde{G}^{-1}(C) \cap B)

is the union of two closed subsets of II. \blacksquare

Loops in a Graph

Any graph can be regarded as a subset of R3\mathbb{R}^3, so that edges only meet at vertices. The diagram shows that K4K_4 (the complete tetrahedral graph with 44 vertices) is planar, since it can be embedded as a subset of R2\mathbb{R}^2. Each face defines a loop based at x0x_0 by concatenation:

α=bda1,β=aec1,γ=cf1b1,δ=bdef1b1.\alpha = bda^{-1}, \\ \beta = aec^{-1}, \\ \gamma = cf^{-1}b^{-1}, \\ \delta = bdef^{-1}b^{-1}.

Note that δ\delta resembles insertion of a cuff at the vertex x0x_0 of a polygon. Moreover αa=bda1abd\alpha a = bda^{-1}a \approxeq bd. This path-homotopy will be proved in the following section, but matches our treatment of boundary codes.

Using similar equivalences:

δ(αa)(a1βc)(c1γb)b1αβγ.\delta \approxeq (\alpha a)(a^{-1}\beta c)(c^{-1}\gamma b)b^{-1} \approxeq \alpha \beta \gamma.

We shall show that this path-homotopy of loops can be interpreted as a relation in the so-called fundamental group. The latter is generated by the path-homotopy classes of loops based at a fixed point of a topological space (like x0x_0 in K4K_4).