coxeter groups

References

Summary

Coxeter groups are those groups that have presentations of the form

⟨x0, …, xn ∣ (xixj)mij = 1⟩, where M is a symmetric matrix with 1s on the diagonal (so xi2 = 1).

This determines a graph, called the Coxeter-Dynkin graph

Note that for i ≠ j we must have mij ≥ 2, since if mij = 1, that implies xixj = 1 so xi = xj−1 = xj.

When drawing the Coxeter-Dynkin graph you

If we have (xixj)3 = 1 that looks like a transposition / braiding

(xixj)2 = xjxi

So xixjxi = xjxixj

If (xixj)4 = 1 that says

xixjxixj = xjxixjxi

Somehow xi corresponds to taking a loop around a fixed pole, and so does xj

you think of xj as wrapping around a fixed pole, and xi as just an ordinary transposition

Then intuitively, maybe you think of the number as the braiding + 2

xyxyxy = yxyx

In xyxyxy = 1 we have the first vertex goes to location 2 then 3 then stays, then 2 then 1 then stays

In xyxyxyxy = 1 it goes 1 to 2 to 2 to make it work

xyxyxy = yxyx 1 to 2 to 3 to 3 to 3 to 2 to 1 to 1 compared to 1 to 1 to 2 to 3 to 3

so that doesn't work

(xy)m = 1 means that

crystallographic restriction in tiling

somehow you want m/2 strings involved in the reflection

but m=3 is braiding so m=5 should be something like braiding and a wall?

maybe m=4 is like a ribbon; like we don't think of wrapping a string around a wall, but as a ribbon with a twist? no it's not exactly that. oh yeah maybe it is; but it's like a double twist; you cross over twice

Lie Groups

Let G be a compact connected Lie group. Then it contains a maximal torus (a maximal abelian subgroup isomorphic to Tk). The Weyl group is

W = NG(T)/T, where T is the maximal torus. The Weyl group turns out to be a finite Coxeter group.

The maximal torus corresponds to the Cartan subalgebra of the Lie algebra, 𝔥. We look at [h,x] = α(h)x with h ∈ 𝔥.

The linear functionals α ∈ 𝔥* are the roots of the system.

Type An

We have generators xn, which behave like adjacent transpositions.

This corresponds to braid groups and Yang-Baxter equations

Has to do with simplices

Root System

For the root system our matrix has entries Mij = cos (π/mij)

We take ⟨xi, xj⟩ = Mij, giving us a bilinear form.

The idea is that we work in euclidean space and take ⟨xi, xj⟩ = ∥xi∥∥xj∥cos (ϕ)

We view (sisj)m = 1 as saying that sisj is a rotation by 2π/m. If you have two lines over which you reflect, reflecting over them in sequence is the same as rotating by 2θ, where θ is the angle between the mirrors. So θ = π/m

Now you view the xi as the normal vectors of the mirrors. You find that ϕ = π − θ, and so

⟨xixj⟩ =  − ∥xi∥∥xj∥cos (ϕ/m)

Now let's say we take our vectors to be of the

Type Bn

Correspond to signed permutations

Has to do with hypercubes

Type Dn

Correspond to even-signed permutations

Type I2(m)

Two-dimensional dihedral groups

Type E6

Type E7

Type E8

Type F4

Type H3

Type H4

Affine Types