Inverse-basis transforms and the Hecke bar involution
Let \(U_w\) be the standard invertible action on the Hecke module. The inverse-basis vector attached to \(w\) is \[ b_w=U_{w^{-1}}^{-1}e_1. \] The linear bar transform is the \(C\)-linear operator \(B\) defined by \(B(e_w)=b_w\). Thus for any finitely supported vector, \[ B\left(\sum_w c_we_w\right)=\sum_w c_wb_w. \] The standard action is independent of the chosen reduced word, so this definition is independent of that choice as well.
Write \(b_w(a),B_a\) in the normalized convention, and \(b_w(q),B_q\) in the unnormalized convention. The latter uses a unit \(q\).
A reduced-word formula
If \(w=s_1\cdots s_m\) is reduced, then the reverse word is reduced for \(w^{-1}\). Reversing the product when taking its inverse gives \[ b_w=L_{s_1}^{-1}\cdots L_{s_m}^{-1}e_1. \] In particular \(b_1=e_1\). For a simple reflection, \[ b_s(a)=H_s-aH_1,\qquad b_s(q)=q^{-1}T_s+(q^{-1}-1)T_1. \]
These expressions are the vectors denoted conventionally by \(H_{w^{-1}}^{-1}\) and \(T_{w^{-1}}^{-1}\). The source defines them by inverse operators.
Parameter interchange
The parameter-interchange theorem proves \[ B_aL_s(-a)=L_s(a)^{-1}B_a,\qquad B_aB_{-a}=I, \] and \[ B_qL_s(q^{-1})=L_s(q)^{-1}B_q,\qquad B_qB_{q^{-1}}=I. \] The maps \(B_a,B_q\) themselves are coefficient-linear.
Semilinearity over Laurent coefficients
In the normalized Laurent convention, \(C=\mathbb Z[v,v^{-1}]\) and \(a=v-v^{-1}\). Let \(\sigma\) replace \(v\) by \(v^{-1}\) in every coefficient and leave the basis labels fixed. Then \(\sigma(a)=-a\).
The semilinear bar is \[ \overline f=B_a(\sigma f). \] Consequently \(\overline{cf}=\sigma(c)\overline f\), and \(\overline{H_w}=b_w(a)\). Applying coefficient inversion to the defining operator formulas changes \(B_a\) to \(B_{-a}\). Hence \[ \overline{\overline f}=B_aB_{-a}f=f. \]
For \(H_s\), this can be read directly: \[ \overline{H_s}=H_s-aH_1,\qquad \overline{H_s-aH_1}=(H_s-aH_1)+aH_1=H_s. \] The unnormalized Laurent convention similarly uses coefficient inversion \(q\mapsto q^{-1}\) before \(B_q\).
A bar-invariant vector satisfies \(\overline f=f\). The linear transforms are the objects used by the coefficient proofs in Lean; the semilinear formulas express them in the usual Laurent notation.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.Hecke.barBasis: Hecke.lean:303.
OAI.KLInvariance.Hecke.barLinear: Hecke.lean:339.
OAI.KLInvariance.HeckeQ.barBasis: Hecke.lean:790.
OAI.KLInvariance.HeckeQ.barLinear: Hecke.lean:833.
Used by
- The reciprocal convolution identity for R-polynomials
- The normalized R-family as bar coefficients
- The descent recursion for normalized R-polynomials
- Parameter interchange and inverse bar transforms
- Bruhat support and diagonal coefficients of the bar basis
- R-values extracted from the unnormalized bar basis
- Hecke operators, the bar map, and normalized R-polynomials