Theorem The ordinary descent recurrence at every unit parameter
✓ theorem proved

The ordinary descent recurrence at every unit parameter

Let \(q\) be a unit in a commutative ring. If \(s\) is a simple reflection with \(sy<y\), then for every \(x\), \[ r_q(x,y)= \begin{cases} r_q(sx,sy),&sx<x,\\ (q-1)r_q(x,sy)+q\,r_q(sx,sy),&sx>x. \end{cases} \] Every left descent of \(y\) satisfies this identity.

Proof

Step 1. Express the longer bar vector through the shorter one

Put \(v=sy\), so \(sv=y\) is an ascent and \(\ell(v)=\ell(y)-1\). The bar ascent formula gives \(b_y(q)=L_s(q)^{-1}b_v(q)\).

Write \(c_x=b_v(q)(x)\) and \(c_{sx}=b_v(q)(sx)\). The inverse-coefficient formula gives \[ b_y(q)(x)= \begin{cases} q^{-1}c_{sx},&sx<x,\\ c_{sx}+(q^{-1}-1)c_x,&sx>x. \end{cases} \]

Step 2. Normalize the simultaneous descent case

Suppose \(sx<x\). Put \(m=\ell(sx)\) and \(n=\ell(v)\). Then \(\ell(x)=m+1\) and \(\ell(y)=n+1\). Therefore \[ \begin{aligned} r_q(x,y) &=(-1)^{m+1}(-q)^{n+1}q^{-1}c_{sx}\\ &=(-1)^m(-q)^n\bigl((-1)(-q)q^{-1}\bigr)c_{sx}\\ &=(-1)^m(-q)^nc_{sx}=r_q(sx,v). \end{aligned} \] The scalar in parentheses is one.

Step 3. Normalize the ascent case at the lower endpoint

Suppose \(sx>x\). Put \(m=\ell(x)\) and \(n=\ell(v)\), so \(\ell(sx)=m+1\) and \(\ell(y)=n+1\). Let \(A=(-1)^m(-q)^n\). Then \[ r_q(x,y)=(-q)A\bigl(c_{sx}+(q^{-1}-1)c_x\bigr). \] The \(c_x\)-term is \[ (-q)(q^{-1}-1)A c_x=(q-1)r_q(x,v). \] For the other term, \(r_q(sx,v)=-A c_{sx}\), hence \[ (-q)A c_{sx}=q\,r_q(sx,v). \] Adding gives the second branch.

Step 4. Identify the smaller instances

Every entry on the right has upper endpoint \(sy\), whose length is one less than that of \(y\). The calculation uses any chosen left descent and the already defined bar vector. It consequently proves the recurrence simultaneously for every such descent.

For \(x=1,y=s\), the second branch reads \[ r_q(1,s)=(q-1)r_q(1,1)+q\,r_q(s,1)=q-1, \] using the diagonal and support values.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.HeckeQ.rValue_recursion: Hecke.lean:1000.

Used by