Theorem Tail pairs recover word-stalk projections and their divisibility
✓ theorem proved

Tail pairs recover word-stalk projections and their divisibility

Fix a finite-generator Coxeter system with symmetric coefficient algebra \(A\), a simple reflection \(s\), a tail word \(\mathbf t\), and a weight \(x\). Let \[ P_x:M_{s\mathbf t}\to M_{\mathbf t}^x\oplus{}^\tau M_{\mathbf t}^{sx} \] be the local pair of the ordinary and reflected tail maps, where \(\tau=\chi_s\). Denote the ordinary tail map by \(\mu\) and the reflected tail map by \(\nu\). Let \(L_x\) be the reflected pair lattice and \(C_x:L_x\xrightarrow{\sim}M_{s\mathbf t}^x\) its coordinate isomorphism.

Every \(P_x(m)\) lies in \(L_x\), and \[ C_x(P_x(m))=\pi_x^{s\mathbf t}(m). \] Moreover, \[ \pi_x^{s\mathbf t}(m)=0 \quad\Longleftrightarrow\quad \pi_x^{\mathbf t}(\mu(m))=0 \quad\text{and}\quad \pi_{sx}^{\mathbf t}(\nu(m))=0. \] If \(P_x(m)=\beta p\) with \(p\in L_x\), then \(\pi_x^{s\mathbf t}(m)=\beta C_x(p)\).

Proof

Step 1. Check the coordinate identity on tensor generators

First regard the coordinate map \(C_x\) as defined on the whole pair module, with values in the ambient new weight-coordinate module. Both \(C_xP_x\) and the new weight projection \(E_x\) are \(A\)-linear.

On a tensor \(1\otimes n\), the tail formulas give \(P_x(1\otimes n)=(\pi_x(n),\pi_{sx}(n))\). The head-coordinate formula then gives \[ C_xP_x(1\otimes n)=E_x(1\otimes n). \] For \(a\otimes n=a(1\otimes n)\), linearity gives the same equality after multiplication by \(a\). Every tensor is a finite sum of pure tensors, so \(C_xP_x=E_x\) on the entire word module.

Step 2. Recover lattice membership from coordinates

The head-pair equivalence says that every vector in the image stalk has coordinates \(C_x(p)\) for some \(p\in L_x\). Apply this to \(E_x(m)\), choosing \(p\in L_x\) with \(C_x(p)=E_x(m)\).

Step 1 gives \(C_x(p)=C_x(P_x(m))\). The ambient coordinate map is injective, so \(p=P_x(m)\). Thus \(P_x(m)\in L_x\). Restricting the coordinate identity to the image stalk gives \(C_x(P_x(m))=\pi_x^{s\mathbf t}(m)\).

Step 3. Recover the two vanishing conditions

Since \(C_x\) is injective and linear, \[ C_x(P_x(m))=0\quad\Longleftrightarrow\quad P_x(m)=0. \] A vector in a product module is zero exactly when both its components are zero. Substituting the definition of \(P_x\) gives the stated equivalence. The second zero is the zero of the underlying tail stalk; the twist preserves zero.

Step 4. Descend a divisible lattice pair

Suppose \(P_x(m)=\beta p\) with \(p\in L_x\). Both sides are now elements of the lattice, and its coordinate isomorphism is \(A\)-linear. Therefore \[ \pi_x^{s\mathbf t}(m) =C_x(P_x(m)) =C_x(\beta p) =\beta C_x(p). \] This supplies an actual stalk vector representing the quotient by \(\beta\).

The remaining input in the non-head reflection argument is to prove that the divided pair \(p\) lies in \(L_x\). Saturation of the lattice provides exactly that assertion.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.TitsSpace.bottTensor_maps_eq: BottSamelson.lean:70.

OAI.KLInvariance.TitsSpace.bottHeadLocalPair_coordinates: BottSamelson.lean:820.

OAI.KLInvariance.TitsSpace.bottHeadLocalPair_mem: BottSamelson.lean:828.

OAI.KLInvariance.TitsSpace.bottHeadLocalPair_zero_iff: BottSamelson.lean:836.

OAI.KLInvariance.TitsSpace.bottHeadLocalPair_value: BottSamelson.lean:853.

OAI.KLInvariance.TitsSpace.bottHeadLocalPair_toStalk: BottSamelson.lean:859.

OAI.KLInvariance.TitsSpace.bottHeadStalk_divisible_of_pair: BottSamelson.lean:866.

Used by