A finite coordinate presentation over an arbitrary weight set
Fix a finite-generator Coxeter system with symmetric coefficient algebra \(A\), and a word \(\mathbf t\). Let \(J_{\mathbf t}\) be its finite set of subexpressions, with weight map \(w:J_{\mathbf t}\to W\).
For any set \(U\subseteq W\), define \[ J_U=\{\varepsilon\in J_{\mathbf t}:w(\varepsilon)\in U\}. \] This finite set is BottUpperIndex. It indexes subexpressions whose weights lie in \(U\).
The source uses the finite coordinate map \[ \widehat p_U:M_{\mathbf t}\longrightarrow \prod_{\varepsilon\in J_U}M_{\mathbf t}^{w(\varepsilon)}, \qquad \widehat p_U(m)_\varepsilon=\pi_{w(\varepsilon)}(m). \] This is bottIndexSetEval. Each coordinate is an entire weight-stalk vector. Several indices may carry the same stalk.
Repeated weights
For the word \((s,s)\), the four choices and their weights are \[ (0,0)\mapsto1,\qquad (0,1)\mapsto s,\qquad (1,0)\mapsto s,\qquad (1,1)\mapsto1. \] With \(U=\{1,s\}\), the finite target is \[ M_{ss}^{1}\times M_{ss}^{s}\times M_{ss}^{s}\times M_{ss}^{1}. \] The map repeats each of the two stalk projections twice. If \(U=\{1\}\), its target repeats the \(1\)-stalk at the two corresponding subexpressions.
A tuple \(z=(z_x)_{x\in U}\) induces the finite tuple \[ \widehat z_\varepsilon=z_{w(\varepsilon)}. \]
Recovering the original tuple
Suppose \(\widehat p_U(m)=d\widehat z\). To prove \(\pi_x(m)=dz_x\) for \(x\in U\), compare any subexpression coordinate \(\varepsilon\) in that stalk. Its weight is \(x\), so \(\varepsilon\in J_U\). The finite equality at this outer index is an equality of \(x\)-stalk vectors; evaluating its \(\varepsilon\)-coordinate gives the required scalar equality.
Every coordinate of the \(x\)-stalk therefore agrees. This proves \(p_U(m)=dz\) on the original set \(U\), including stalks with an empty coordinate set.
Stable coefficient weights on the finite presentation
For \(a\in A\), define \[ \widehat\chi_a(\varepsilon)=\chi_{w(\varepsilon)}(a). \] The image of \(\widehat p_U\) is preserved by these diagonal weights, because \[ \widehat p_U(m\cdot a)_\varepsilon =\chi_{w(\varepsilon)}(a)\pi_{w(\varepsilon)}(m). \] This supplies the finite stable-character module needed by the class-lift assembly theorem. Its residue classes may repeat a weight many times; the geometric classification concerns the distinct group weights represented in each class.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.TitsSpace.BottUpperIndex: IntegralPairings.lean:2013.
OAI.KLInvariance.TitsSpace.bottIndexSetEval: IntegralPairings.lean:2015.
OAI.KLInvariance.TitsSpace.bottIndexSetEval_stable: IntegralPairings.lean:2020.