Definition Chebyshev coefficients for rank-two root calculations
✓ definition formalized

Chebyshev coefficients for rank-two root calculations

Define polynomials \(S_n(z)\), indexed by integers, by \[ S_{-1}(z)=0,\qquad S_0(z)=1,\qquad S_{n+1}(z)=zS_n(z)-S_{n-1}(z). \] The recurrence determines all indices: it can also be solved for \(S_{n-1}\). For example, \[ S_1=z,\quad S_2=z^2-1,\quad S_3=z^3-2z,\quad S_{-2}=-1. \] This fixes the normalization of the Chebyshev polynomials used in the root calculations.

Trigonometric evaluation

For a real angle \(\theta\) with \(\sin\theta\ne0\), \[ S_n(2\cos\theta)=\frac{\sin((n+1)\theta)}{\sin\theta}. \] The right side has values zero and one at indices \(-1\) and zero. The addition formulas for sine give \[ \sin((n+2)\theta) =2\cos\theta\sin((n+1)\theta)-\sin(n\theta). \] Thus these values satisfy the same recurrence, in both directions, which proves the formula.

At \(z=2\), the values are \(S_n(2)=n+1\). The sequence \(n+1\) has the required initial values and satisfies \(n+2=2(n+1)-n\), so the recurrence again proves the assertion.

Coefficients for a Coxeter pair

For simple generators \(s_i,s_j\), define \[ c_n(i,j)=S_n\!\left(2\cos\frac{\pi}{m_{ij}}\right). \] For an infinite exponent the argument is two, so \(c_n(i,j)=n+1\).

For distinct generators with finite exponent \(m=m_{ij}\ge2\), put \(\theta=\pi/m\). Then \[ c_n(i,j)=\frac{\sin((n+1)\theta)}{\sin\theta}. \] If \(-1\le n\) and \(n+1\le m\), the numerator has angle in \([0,\pi]\) and is nonnegative; the denominator is positive. Hence \(c_n(i,j)\ge0\) in this range.

For \(m=3\), the argument is one and the first values are \[ c_{-1}=0,\qquad c_0=1,\qquad c_1=1,\qquad c_2=0. \] For an infinite exponent they are \(0,1,2,3,\ldots\).

The alternating-root formula uses these coefficients for the two simple-root coordinates. The finite reflection-relation proof uses the same polynomials at the angle \(2\pi/m\), where their vanishing gives the required powers of the two-reflection operator.

References

The local source statements and calculations are listed below.

Lean formalization

OAI.KLInvariance.TitsSpace.rootCoeff: Tits.lean:813.

OAI.KLInvariance.TitsSpace.rootCoeff_zero: Tits.lean:817.

OAI.KLInvariance.TitsSpace.rootCoeff_neg_one: Tits.lean:821.

OAI.KLInvariance.TitsSpace.rootCoeff_succ: Tits.lean:826.

OAI.KLInvariance.TitsSpace.rootCoeff_nonneg: Tits.lean:867.

Used by