TheoremComplete

Alexander's Theorem

Theorem5.1Alexander's Theorem (Braids and Links)

Every oriented link LL in S3S^3 is the closure β^\hat{\beta} of some braid βBn\beta \in B_n for some nn. That is, the map ^:n=1Bn{oriented links}\widehat{\cdot}: \bigsqcup_{n=1}^\infty B_n \to \{\text{oriented links}\} sending a braid to its closure is surjective.


Proof

Proof

Step 1: Setup. Given an oriented link LL, we may assume LL is in general position with respect to a chosen axis AA (a line in R3\mathbb{R}^3). We will isotope LL so that it winds monotonically around AA, making it the closure of a braid.

Step 2: Choose an axis and projection. Consider R3\mathbb{R}^3 with cylindrical coordinates (r,θ,z)(r,\theta,z) and let AA be the zz-axis. In the projection to the (r,θ)(r,\theta)-plane, the link LL appears as a closed curve. The angular coordinate θ\theta may not be monotonically increasing along LL -- there may be arcs where θ\theta decreases (the link "goes backwards" around the axis).

Step 3: Eliminate bad arcs. A bad arc is a subarc of LL along which θ\theta decreases. For each bad arc α\alpha:

  • Choose a point PP far from LL on the axis AA (far above or below the projection).
  • Replace α\alpha by a detour: isotope α\alpha by swinging it around the axis AA in the positive direction, passing near PP, so that θ\theta is now increasing along the modified arc.

More precisely, if α\alpha goes from angle θ1\theta_1 to θ0<θ1\theta_0 < \theta_1, we replace it by an arc that increases angle by 2π(θ1θ0)2\pi - (\theta_1 - \theta_0), achieving the same net change while always moving in the positive θ\theta direction. The point PP is chosen so that the detour avoids all other strands of LL.

Step 4: Result is a braid closure. After eliminating all bad arcs, the modified link LL' (isotopic to LL) has θ\theta monotonically increasing along every component. Cutting LL' along a half-plane θ=θ0\theta = \theta_0 yields a braid βBn\beta \in B_n (where nn is the number of intersection points with the half-plane), and L=β^L' = \hat{\beta}.

Step 5: Finiteness. There are only finitely many bad arcs (since LL is a finite union of smooth arcs in general position), so the procedure terminates. \square


ExampleTrefoil as a Braid Closure

The trefoil knot 313_1 is the closure of σ13B2\sigma_1^3 \in B_2 (the braid on 2 strands with 3 positive crossings). Equivalently, it is the closure of (σ1σ2)2B3(\sigma_1\sigma_2)^2 \in B_3 (viewing the trefoil as a (2,3)(2,3)-torus knot). The figure-eight knot 414_1 is the closure of σ1σ21σ1σ21B3\sigma_1\sigma_2^{-1}\sigma_1\sigma_2^{-1} \in B_3. Every torus knot T(p,q)T(p,q) is the closure of (σ1σ2σp1)qBp(\sigma_1\sigma_2\cdots\sigma_{p-1})^q \in B_p.

RemarkNon-Uniqueness and Markov's Theorem

The braid representative of a link is far from unique. Markov's theorem characterizes when two braids have isotopic closures: β^1β^2\hat{\beta}_1 \cong \hat{\beta}_2 if and only if β1\beta_1 and β2\beta_2 are related by a sequence of Markov moves: (I) conjugation βαβα1\beta \mapsto \alpha\beta\alpha^{-1} in BnB_n, and (II) stabilization ββσn±1\beta \mapsto \beta\sigma_n^{\pm 1} (adding a strand, BnBn+1B_n \hookrightarrow B_{n+1}). Alexander's theorem combined with Markov's theorem reduces link theory to braid group algebra, enabling the construction of link invariants (such as the Jones polynomial) via braid group representations.