Documentation

TauCeti.Algebra.Lie.HighestWeight.Weyl.Character

The Weyl character formula #

Let L be a finite-dimensional Lie algebra with non-degenerate Killing form over an algebraically closed field of characteristic zero, H a Cartan subalgebra, b a base of its root system, and M a finite-dimensional module generated by a highest weight vector of weight lam. With c(χ) = ⟨χ + ρ, χ + ρ⟩ - ⟨ρ, ρ⟩ the Casimir scalar (TauCeti.casimirScalar), this file proves that c takes the single value c(lam) on the support of the product ch M · Δ of the formal character with the Weyl denominator Δ = ∏_{α>0}(1 - e^{-α}).

Since c separates the dominant integral weights below lam, it follows that ch M · Δ vanishes at every dominant integral weight other than lam, where its coefficient is 1. These are the dominant integral coefficients through which TauCeti.formalCharacter_mul_weylDenominator_eq_of_forall_coeff_isDominantIntegral_eq determines ch M · Δ. The Weyl numerator N(lam) = ∑_{w ∈ W} sgn(w) e^{w ⬝ lam} has the very same dominant integral coefficients, by TauCeti.coeff_weylNumerator_self_of_dominantIntegral and TauCeti.coeff_weylNumerator_eq_zero_of_dominantIntegral_of_ne, and is supported in lam - Q⁺ as well, so the two are equal: that is the Weyl character formula, proved at the end of this file.

The argument #

Write m(χ) = dim M_χ and σ_T = ∑_{α ∈ T} α for a set T of positive roots, so that the coefficient of ch M · Δ at χ is ∑_T (-1)^{|T|} m(χ + σ_T). Multiply the T-th term by c(lam) - c(χ) and split this as (c(lam) - c(χ + σ_T)) + (c(χ + σ_T) - c(χ)).

For each positive root α, the string sum at μ is the string sum at μ + α plus m(μ + α) ⟨μ + α, α⟩; pairing the sets T ∌ α with T ∪ {α} therefore cancels the string sums against the root pairings. What survives is -∑_T (-1)^{|T|} c(-σ_T) m(χ + σ_T), which groups into -∑_x c(x) Δ_x m(χ - x). This vanishes because c vanishes on the support of Δ (TauCeti.casimirScalar_eq_zero_of_coeff_weylDenominator_ne_zero): restricted to the weights where c ≠ 0, the denominator is still alternating for the dot action (c is dot invariant), lies in the negative root cone and has no dominant integral weight in its support, so it is zero by TauCeti.IsDotAlternating.eq_zero_of_forall_coeff_dominantIntegral_eq_zero.

Main results #

References #

The Casimir scalar and the dot action #

@[simp]

The Casimir scalar is invariant under the dot action: c(w ⬝ x) = c(x), since w ⬝ x + ρ = w (x + ρ) and the invariant form is Weyl invariant.

The Casimir scalar vanishes on the support of the Weyl denominator.

The Weyl character formula #

The Casimir scalar is constant on the support of ch M · Δ: for a finite-dimensional module generated by a highest weight vector of weight lam, every weight carrying a nonzero coefficient of ch M · Δ has the Casimir scalar of lam.

ch M · Δ vanishes at every dominant integral weight other than the highest weight. A dominant integral weight in the support lies below lam and has the Casimir scalar of lam, and the Casimir scalar separates dominant integral weights along the root cone order.

The Weyl character formula #

The Weyl character formula. For a finite-dimensional module M generated by a highest weight vector of weight lam, the formal character of M times the Weyl denominator is the Weyl numerator of lam: ch M · ∏_{α>0} (1 - e^{-α}) = ∑_{w ∈ W} sgn(w) e^{w(lam + ρ) - ρ}.

Both sides are alternating for the dot action and supported in lam - Q⁺, so TauCeti.formalCharacter_mul_weylDenominator_eq_of_forall_coeff_isDominantIntegral_eq reduces the identity to the coefficients at the dominant integral weights. There both sides are 1 at lam and 0 elsewhere: on the left by the two coefficient computations above and coeff_formalCharacter_mul_weylDenominator_eq_one_of_isHighestWeightVector_of_lieSpan_eq_top, on the right because the dot orbit of a dominant integral weight is free and meets the dominant integral weights only at lam.

Such an M is irreducible, by TauCeti.isIrreducible_of_isHighestWeightVector_of_lieSpan_eq_top, so this is the character of L(lam); it is stated for M itself so that no identification with L(lam) is needed to use it.