Theorem Exact kernels for a reflection distinct from the first letter
✓ theorem proved

Exact kernels for a reflection distinct from the first letter

Fix a finite-generator Coxeter system with symmetric coefficient algebra \(A\). Assume the tail word \(\mathbf t\) has exact reflection kernels and free stalks at every weight. Prepend a simple reflection \(s\), with coefficient action \(\tau=\chi_s\) and root polynomial \(\alpha_s\).

Let \(t\ne s\) be a reflection represented by a root \(a\), put \[ y=tx,\qquad \beta=\iota(\operatorname{embed}(a)), \] and assume \(\ell(x)<\ell(y)\). If \(m\in M_{s\mathbf t}\) satisfies \(\pi_x^{s\mathbf t}(m)=0\), then \[ \pi_y^{s\mathbf t}(m)=\beta z \quad\text{for some }z\in M_{s\mathbf t}^y. \]

Relation to the OpenAI notes

Word-image construction before the boundary sheaf.

OpenAI notes: The moment-graph sheaf theorem states the quotient description of an edge module and freeness for the normalized indecomposable boundary sheaf, through the cited moment-graph and Soergel constructions. bg:bmp.

Lean (primary): The tensor word first supplies explicit subexpression evaluations and image stalks. Right coefficient characters give label annihilation; a first-factor reflection proves the exact kernel across the initial simple reflection. Reflected pair lattices then describe stalks after prepending a letter. Frobenius:1558, Frobenius:2018, BottSamelson:214, BottSamelson:964, BottSamelson:1130, BottSamelson:1510, Support:426.

Role in the proof: The word-length induction proves stalk freeness and exact kernels together. Its two tail maps, quotient regularity and lattice saturation prove the non-head reflection step; the ascent/descent split sequences prove new stalk freeness. Coefficient transport then gives the polynomial word-sheaf input to the boundary-summand construction.

Proof

Step 1. Transport the upward edge by the first reflection

Because \(t\ne s\), the equality \(tx=sx\) is impossible by cancellation on the right. The simple edge-transport theorem therefore takes the upward edge \(x\to y\) to an upward edge \[ sx\longrightarrow sy. \] Its reflection is \(sts^{-1}=sts\), since \[ (sts^{-1})(sx)=s(tx)=sy. \] The transformed root \(s(a)\) represents this conjugate reflection. Indeed, if \(a=z(r_i)\) and \(t=z s_i z^{-1}\), then \(s(a)=(sz)(r_i)\) and \(sts^{-1}=(sz)s_i(sz)^{-1}\). These are the two defining equations for the transformed root to represent the conjugated reflection.

Step 2. Extract two tail vectors with zero lower projections

Let \(m_0=\mu(m)\) be the ordinary tail vector and let \(m_1\) be the underlying vector of the reflected tail \(\nu(m)\).

The local-pair zero criterion applied at \(x\) turns the hypothesis into \[ \pi_x^{\mathbf t}(m_0)=0,\qquad \pi_{sx}^{\mathbf t}(m_1)=0. \]

Step 3. Apply old-word exactness to each upward edge

Apply exact reflection kernels of the shorter word to \(x\to tx=y\), root \(a\), and vector \(m_0\). It supplies \(p\in M_{\mathbf t}^y\) with \[ \pi_y^{\mathbf t}(m_0)=\beta p. \]

Apply the same induction hypothesis to the transported edge \(sx\to sy\), root \(s(a)\), and vector \(m_1\). It supplies \(q\in M_{\mathbf t}^{sy}\) with \[ \pi_{sy}^{\mathbf t}(m_1) =\iota(\operatorname{embed}(s(a)))q. \] The extended action restricts to the geometric action on embedded roots, so \[ \iota(\operatorname{embed}(s(a)))=\chi_s(\beta)=\tau(\beta). \] Thus the two old projections are \(\beta p\) and \(\tau(\beta)q\).

Step 4. Assemble the divided pair in the twisted module

The local pair of \(m\) at the new upper weight \(y\) is \[ P_y(m)=\bigl(\pi_y(m_0),\pi_{sy}(m_1)\bigr) =(\beta p,\tau(\beta)q). \] The second factor has scalar action \(r\star q=\tau(r)q\). Therefore this is exactly \[ P_y(m)=\beta\cdot(p,q) \] in \(M_{\mathbf t}^y\oplus{}^\tau M_{\mathbf t}^{sy}\).

Step 5. Exclude divisibility by the first root

We claim \(\alpha_s\nmid\beta\). If divisibility held, the linear-generator divisibility criterion would give \[ \operatorname{embed}(a)=c\,\operatorname{embed}(r_s) \] for a real scalar \(c\), where \(r_s\) is the simple root. Projection to the original realization gives \(a=c r_s\).

Let \(B\) be its geometric bilinear form. Both roots have squared norm two, so \[ 2=B(a,a)=c^2B(r_s,r_s)=2c^2. \] Thus \(c^2=1\), and the real scalar \(c\) is \(1\) or \(-1\). Hence \(a=r_s\) or \(a=-r_s\). Both represent \(s\), and uniqueness of the reflection represented by a root forces \(t=s\), a contradiction.

This is the norm calculation behind the source's independence lemma and represents_polynomial_not_dvd.

Step 6. Remove the scalar inside the head pair lattice

The local-pair theorem gives \(P_y(m)\in L_y\). By Step 4, the multiple \(\beta(p,q)\) lies in \(L_y\). Head-pair saturation applies because the tail has exact reflection kernels and free stalks, and Step 5 gives \(\alpha_s\nmid\beta\). Hence \((p,q)\in L_y\).

The head-pair isomorphism now supplies \(z=C_y(p,q)\in M_{s\mathbf t}^y\), and its linearity gives \[ \pi_y^{s\mathbf t}(m)=C_y(P_y(m)) =C_y(\beta(p,q))=\beta z. \] This is the required exact-kernel divisibility.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.TitsSpace.represents_polynomial_not_dvd: BottSamelson.lean:982.

OAI.KLInvariance.TitsSpace.bottKernelExact_cons_nonhead: BottSamelson.lean:1130.

OAI.KLInvariance.TitsSpace.Represents.act: Tits.lean:1714.

OAI.KLInvariance.TitsSpace.Represents.independent_of_ne: Tits.lean:1983.

OAI.KLInvariance.TitsSpace.Represents.extended_independent_of_ne: Tits.lean:1998.

Used by