Exact reflection kernels for a tensor word
Fix a finite-generator Coxeter system and its symmetric coefficient algebra \(A\). A word \(\mathbf t\) satisfies exact reflection kernels if the following condition holds for every group element \(x\), every reflection \(t\), and every root \(a\) representing \(t\).
Put \(\beta=\iota(\operatorname{embed}(a))\). Whenever
\[
\ell(x)<\ell(tx),
\]
every vector \(m\in M_{\mathbf t}\) with \(\pi_x(m)=0\) has
\[
\pi_{tx}(m)=\beta z
\quad\text{for some }z\in M_{\mathbf t}^{tx}.
\]
This is the source's proposition BottKernelExact. The length inequality orients the reflection edge upward. The divisibility takes place in the actual image stalk \(M_{\mathbf t}^{tx}\).
Equivalent kernel formulations
For an upward edge, let \(f=\pi_x\), \(g=\pi_{tx}\), and \(u:M_{\mathbf t}^{tx}\to Q\) be the upper map to their pair-projection quotient. The condition gives \[ g(\ker f)\subseteq\beta M_{\mathbf t}^{tx}. \] Reflection-label annihilation supplies the reverse inclusion: \(u(\beta z)=0\), and the pair-projection criterion lifts \((0,\beta z)\). Therefore exact reflection kernels give \[ g(\ker f)=\ker u=\beta M_{\mathbf t}^{tx}. \] Equivalently, \[ \bigl(\exists m,\ f(m)=0,\ g(m)=y\bigr) \quad\Longleftrightarrow\quad \bigl(\exists z,\ y=\beta z\bigr). \] These are the source's upper_kernel and kernel_iff consequences.
Reindexing an edge
If the upper endpoint is written as \(y\) with \(tx=y\), the same condition reads \[ \ell(x)<\ell(y),\quad \pi_x(m)=0 \quad\Longrightarrow\quad \pi_y(m)=\beta z. \] The source's kernel_iff_of_mul and projection_of_mul provide this substitution explicitly. They let the word induction compare the original edge with its conjugated edge after prepending a simple reflection.
The free-word-stalk theorem proves this property for every word together with stalk freeness. In that simultaneous induction, the first-reflection case uses the reflection flip; the other reflections use the tail's exact kernels and saturation of the head pair lattice.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.TitsSpace.BottKernelExact: BottSamelson.lean:964.
OAI.KLInvariance.TitsSpace.BottKernelExact.upper_kernel: BottSamelson.lean:1012.
OAI.KLInvariance.TitsSpace.BottKernelExact.kernel_iff: BottSamelson.lean:1020.
OAI.KLInvariance.TitsSpace.BottKernelExact.kernel_iff_of_mul: BottSamelson.lean:1028.
OAI.KLInvariance.TitsSpace.BottKernelExact.projection_of_mul: BottSamelson.lean:1036.