Definition Subexpression evaluations of tensor word modules
✓ definition formalized

Subexpression evaluations of tensor word modules

Fix a Coxeter system with finite simple-generator set \(I\), its real extended realization \(V\), and \(A=\operatorname{Sym}_{\mathbb R}(V)\). Write \(\chi_w:A\to A\) for the coefficient action of \(w\in W\). For a simple reflection \(s\), put \(\tau=\chi_s\) and \(R=A^s\).

A word \(\mathbf t=(s_1,\ldots,s_m)\) has a tensor word module \(M_{\mathbf t}\), with \[ M_\varnothing=A,\qquad M_{s\mathbf t}=A\otimes_R M_{\mathbf t}. \] Its set \(J_{\mathbf t}\) of subexpressions consists of choices of \(0\) or \(1\) at each position. The empty word has one empty choice. Recursively, \[ J_{s\mathbf t}=\{0,1\}\times J_{\mathbf t}. \]

Group weights

The weight of a subexpression is the product of the retained letters in their original order: \[ w_\varnothing(())=1,\qquad w_{s\mathbf t}(\varepsilon,\eta)= \begin{cases} w_{\mathbf t}(\eta),&\varepsilon=0,\\ s\,w_{\mathbf t}(\eta),&\varepsilon=1. \end{cases} \] This is bottWeight. For example, the choices \((1,0,1)\) in \((s,t,u)\) have weight \(su\).

The coefficient character at a group weight \(x\) is the algebra automorphism \(\chi_x:A\to A\). The evaluation right-action theorem proves that right multiplication by \(a\) transforms the weight-\(x\) coordinate by the scalar \(\chi_x(a)\).

Evaluation maps

For each \(\eta\in J_{\mathbf t}\), evaluation is an \(A\)-linear map \[ E_{\mathbf t,\eta}:M_{\mathbf t}\to A. \] The empty-word evaluation is the identity of \(A\). Prepending \(s\) gives \[ \begin{aligned} E_{s\mathbf t,(0,\eta)}(a\otimes m)&=a\,E_{\mathbf t,\eta}(m),\\ E_{s\mathbf t,(1,\eta)}(a\otimes m)&=a\,\tau(E_{\mathbf t,\eta}(m)). \end{aligned} \] The outer coefficient \(a\) stays outside the reflection.

These formulas respect the tensor relations over \(R\). If \(r\in R\), then \(\tau(r)=r\); hence \[ \tau(E_{\mathbf t,\eta}(rm)) =\tau(rE_{\mathbf t,\eta}(m)) =r\,\tau(E_{\mathbf t,\eta}(m)). \] Thus both \(m\mapsto E_{\mathbf t,\eta}(m)\) and \(m\mapsto\tau(E_{\mathbf t,\eta}(m))\) are \(R\)-linear. Multiplication by the outer factor extends each to \(A\otimes_R M_{\mathbf t}\). The source implements these two steps as weightFunctional and inducedFunctional.

Full evaluation collects all coordinates: \[ E_{\mathbf t}:M_{\mathbf t}\to A^{J_{\mathbf t}}, \qquad m\mapsto(E_{\mathbf t,\eta}(m))_{\eta\in J_{\mathbf t}}. \] Its injectivity is proved by the reflection block matrix.

A vector with every evaluation equal to one

Define the unit tensor recursively by \[ u_\varnothing=1,\qquad u_{s\mathbf t}=1\otimes u_{\mathbf t}. \] Both evaluation formulas send this tensor to \(1\), since \(\tau(1)=1\). Induction gives \(E_{\mathbf t,\eta}(u_{\mathbf t})=1\) for every \(\eta\).

For any \(a\in A\), the vector \(a u_{\mathbf t}\) therefore evaluates to \(a\) at each coordinate. In particular, every individual evaluation is surjective. The source calls these constructions and properties bottUnit, bottEvaluation_unit and bottEvaluation_surjective.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.TitsSpace.bottWeight: Frobenius.lean:1056.

OAI.KLInvariance.TitsSpace.weightFunctional: Frobenius.lean:1062.

OAI.KLInvariance.TitsSpace.inducedFunctional: Frobenius.lean:1078.

OAI.KLInvariance.TitsSpace.bottEvaluation: Frobenius.lean:1085.

OAI.KLInvariance.TitsSpace.bottEvaluation_tmul: Frobenius.lean:1091.

OAI.KLInvariance.TitsSpace.bottUnit: BottSamelson.lean:276.

OAI.KLInvariance.TitsSpace.bottEvaluation_unit: BottSamelson.lean:280.

OAI.KLInvariance.TitsSpace.bottEvaluation_surjective: BottSamelson.lean:289.

Used by