Definition Patching the two tail assignments on a reflected union
✓ definition formalized

Patching the two tail assignments on a reflected union

Fix a finite-generator Coxeter system, a simple reflection \(s\), and a tail word \(\mathbf t\). Let \(\tau=\chi_s\), and let \(\alpha=\alpha_s\) be the simple-root polynomial.

For \(m\in M_{s\mathbf t}\), write \[ p=\mu(m),\qquad q=\nu(m) \] for its two underlying tail vectors in \(M_{\mathbf t}\). They satisfy \[ p-q=\alpha n \] for some tail vector \(n\). Indeed, for a pure tensor \(a\otimes v\), the difference is \[ av-\tau(a)v=(a-\tau(a))v=\alpha(bv) \] with \(a-\tau(a)=\alpha b\). Summing this calculation over a finite pure-tensor expression proves the assertion for every \(m\).

The patched assignment

For \(U\subseteq W\), put \(J=U\cup sU\) and \(K=U\cap sU\). Define the head patch on \(J\) by \[ z_x= \begin{cases} \pi_x(p),&x\in U,\\ \pi_x(q),&x\in J\setminus U. \end{cases} \] This is the source's bottHeadPatch.

When the two tail projections agree on \(K\), the patch agrees with \(p\) throughout \(U\) and with \(q\) throughout \(sU\): on the overlap, the two prescriptions give the same value.

The patched-tail lifting theorem proves compatibility on \(J\) when \(U\) is upper. The root multiple \(p-q=\alpha n\) handles a crossing edge between the two pieces. If the tail word has compatible-tuple lifting, the whole patch then has a single tail-word preimage.

Vanishing on the overlap

If the new word vector \(m\) vanishes on \(K\), set-tail detection and \(sK=K\) show that both \(p\) and \(q\) vanish there. Their values on \(K\) therefore agree. This is the overlap condition used in the two-stage kernel argument.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.ReflectionPairTranslation.tensor_difference: IntegralPairings.lean:761.

OAI.KLInvariance.TitsSpace.bottHead_difference: IntegralPairings.lean:821.

OAI.KLInvariance.TitsSpace.bottHeadPatch: IntegralPairings.lean:827.

Used by