Lower-interval Hecke sums and their correction vectors
Fix a rank-two Coxeter system and a commutative coefficient ring \(C\). In the unnormalized standard Hecke module, define \[ S_w=\sum_{z\le w}T_z. \] The lower ideal is finite. Its coefficient function is the indicator of that ideal: \[ S_w(x)= \begin{cases} 1,&x\le w,\\ 0,&x\not\le w. \end{cases} \] Thus every included basis vector occurs with coefficient one.
The first sums are \[ S_1=T_1,\qquad S_s=T_1+T_s. \] The vector \(S_w\) depends on the basis and order, while the parameter \(q\) enters when a Hecke operator acts on it.
Simple-reflection strings
For a fixed simple reflection \(s\), a simple-reflection string is the pair \(\{r,sr\}\). Multiplication by \(s\) exchanges the two members, since \(s^2=1\). The members are distinct, and their lengths differ by one.
Choose \(r\) to be the shorter member. Then \(r<sr\) in Bruhat order. Distinct strings are disjoint: sharing either member determines the other by multiplying by \(s\). Thus these pairs partition the group.
The lower ideal of \(v\) can contain neither member, just the shorter member, or both members. If it contains the longer member, downward closure also includes the shorter one.
Prefix and tail data
Suppose \(\omega_i(n+2)\) is reduced, and write \[ w=\operatorname{prod}\omega_i(n+2),\quad v=\operatorname{prod}\omega_{\bar i}(n+1),\quad p=\operatorname{prod}\omega_i(n). \] With \(s=s_i\), we have \(w=sv\), and their lengths are \(n+2,n+1,n\). The tail and prefix are reduced because the original word is reduced.
Define the correction vector \[ C_{i,n}= \begin{cases} 0,&n=0,\\ S_p,&n>0. \end{cases} \] Its coefficient is one exactly when \(n>0\) and \(x\le p\), and zero otherwise.
The first two corrections
For \(w=st\), the tail is \(v=t\), and \(n=0\). The correction is the zero vector.
For a reduced \(w=sts\), the tail is \(v=ts\), the prefix two letters shorter is \(p=s\), and \[ C_{s,1}=S_s=T_1+T_s. \] The Hecke-sum product theorem explains these vectors by identifying the simple-reflection pairs whose two members already lie below \(v\).
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.Dihedral.lowerSum: Dihedral.lean:197.
OAI.KLInvariance.Dihedral.lowerSum_coeff: Dihedral.lean:200.
OAI.KLInvariance.Dihedral.lowerSum_one: Dihedral.lean:205.
OAI.KLInvariance.Dihedral.lowerSum_simple: Dihedral.lean:490.
OAI.KLInvariance.Dihedral.correction: Dihedral.lean:322.
OAI.KLInvariance.Dihedral.correction_coeff: Dihedral.lean:325.