Theorem Homogeneous evaluation and homogeneous lifts to word modules
✓ theorem proved

Homogeneous evaluation and homogeneous lifts to word modules

Fix a finite-generator Coxeter system and a basis of its real extended realization. Let \(A=\bigoplus_n A_n\) be its symmetric coefficient algebra with linear generators in degree one. Equip every tensor word module \(M_{\mathbf t}\) with the grading from its standard basis degrees \(d_\gamma\).

Each evaluation preserves degree: \[ E_{\mathbf t,\eta}((M_{\mathbf t})_n)\subseteq A_n. \] For \(J_x=\{\eta:w_{\mathbf t}(\eta)=x\}\), give \(A^{J_x}\) its coordinatewise grading. The image stalk \(M_{\mathbf t}^x\) is a homogeneous submodule, and every homogeneous stalk vector \(z\) of degree \(n\) has a homogeneous lift \(m\in(M_{\mathbf t})_n\) with \(\pi_x(m)=z\).

Proof

Step 1. Determine the degrees of matrix entries

We prove \[ (E_{\mathbf t})_{\eta\gamma}\in A_{d_\gamma} \] by word induction. For the empty word, the only matrix entry is \(1\in A_0\).

Suppose this is known for \(\mathbf t\). In the block matrix for \(s\mathbf t\), the first column block has entries \(E_{\mathbf t}\) or \(\tau(E_{\mathbf t})\). The coefficient action preserves degree, so each entry in column \((0,\gamma)\) has degree \(d_\gamma\).

The second column block has entries \(\alpha_s E_{\mathbf t}\) or \(-\alpha_s\tau(E_{\mathbf t})\). Since \(\alpha_s\in A_1\), these have degree \(1+d_\gamma\). Those are exactly the assigned degrees of columns \((1,\gamma)\). This completes the induction.

Step 2. Evaluate a homogeneous vector

If \(m=\sum_\gamma c_\gamma b_\gamma\) lies in degree \(n\), the basis grading means \[ c_\gamma\in A_{n-d_\gamma}. \] Evaluation expands as \[ E_{\mathbf t,\eta}(m) =\sum_\gamma c_\gamma(E_{\mathbf t})_{\eta\gamma}. \] Each product has degree \((n-d_\gamma)+d_\gamma=n\) by Step 1. The finite sum therefore lies in \(A_n\).

Step 3. Commute evaluation with homogeneous components

Write \(m=\sum_j m_j\) as its finite homogeneous decomposition. Step 2 gives \(E_{\mathbf t,\eta}(m_j)\in A_j\). Uniqueness of the homogeneous decomposition in \(A\) then yields \[ \operatorname{pr}_n(E_{\mathbf t,\eta}(m)) =E_{\mathbf t,\eta}(\operatorname{pr}_n m). \] Collecting the same identities over \(\eta\in J_x\) gives the component identity for the weight projection.

Step 4. Give the image stalk its grading

If a coordinate vector is in the image, write it as \(E_x(m)\). Its degree-\(n\) component is \(E_x(\operatorname{pr}_n m)\), by Step 3, and hence is again in the image.

Thus the image is closed under homogeneous components. Its grading is \[ (M_{\mathbf t}^x)_n =\{z\in M_{\mathbf t}^x:z_\eta\in A_n\text{ for every }\eta\in J_x\}. \] This is bottStalkPiece, obtained by restricting the coordinatewise grading bottCoordinatePiece. The finite decomposition and its uniqueness descend from the ambient coordinate module.

Step 5. Lift a homogeneous stalk vector

Let \(z\in(M_{\mathbf t}^x)_n\). Surjectivity of the stalk projection gives some \(v\) with \(\pi_x(v)=z\). Set \(m=\operatorname{pr}_n v\). Then \(m\) has degree \(n\), and Step 3 gives \[ \pi_x(m)=\operatorname{pr}_n(\pi_x(v)) =\operatorname{pr}_n(z)=z. \] This proves the required lift. It is the component projection of any initial lift, exactly as in bottStalk_homogeneous_lift.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.TitsSpace.bottEvaluation_basis_homogeneous: BottSamelson.lean:338.

OAI.KLInvariance.TitsSpace.bottEvaluation_homogeneous: BottSamelson.lean:367.

OAI.KLInvariance.TitsSpace.bottEvaluation_component: BottSamelson.lean:384.

OAI.KLInvariance.TitsSpace.bottCoordinatePiece: BottSamelson.lean:421.

OAI.KLInvariance.TitsSpace.bottWeightProjection_homogeneous: BottSamelson.lean:436.

OAI.KLInvariance.TitsSpace.bottStalk_homogeneous: BottSamelson.lean:442.

OAI.KLInvariance.TitsSpace.bottStalkPiece: BottSamelson.lean:449.

OAI.KLInvariance.TitsSpace.bottStalkProjection_homogeneous: BottSamelson.lean:462.

OAI.KLInvariance.TitsSpace.bottStalk_homogeneous_lift: BottSamelson.lean:467.

Used by