Theorem Linear membership and units at a root-plane prime
✓ theorem proved

Linear membership and units at a root-plane prime

Fix a finite-generator Coxeter system. Let \(V_0\) be its original root space, \(V\) its extended realization, and \(\iota_0:V_0\hookrightarrow V\) the embedding. Choose polynomial coordinates \[ \Phi:\operatorname{Sym}_{\mathbb R}(V)\xrightarrow{\sim}A. \] For a subspace \(P\subseteq V_0\), let \(\mathfrak p_P\subseteq A\) be the transported ideal generated by \(\iota_0(P)\). Write \[ \lambda(v)=\Phi(\iota(\iota_0(v))). \] Then \[ \lambda(v)\in\mathfrak p_P\quad\Longleftrightarrow\quad v\in P. \] In the localized ring \(A_{\mathfrak p_P}\), its image lies in the maximal ideal exactly when \(v\in P\), and is a unit exactly when \(v\notin P\).

Proof

Step 1. Apply the generic linear-ideal criterion

In the symmetric algebra, the linear-quotient theorem gives \[ \iota(\iota_0(v))\in I_{\iota_0(P)} \quad\Longleftrightarrow\quad \iota_0(v)\in\iota_0(P). \] The ideal is prime by that theorem.

Step 2. Use injectivity of the embedding

If \(\iota_0(v)\in\iota_0(P)\), choose \(w\in P\) with \(\iota_0(w)=\iota_0(v)\). The embedding is injective, so \(w=v\), and \(v\in P\).

Conversely, \(v\in P\) immediately puts its image in \(\iota_0(P)\). Thus Step 1 is equivalent to \(v\in P\).

Step 3. Transport the ideal through polynomial coordinates

By definition \(\mathfrak p_P=\Phi(I_{\iota_0(P)})\). Since \(\Phi\) is bijective, \[ \Phi(\iota(\iota_0(v)))\in\mathfrak p_P \quad\Longleftrightarrow\quad \iota(\iota_0(v))\in I_{\iota_0(P)}. \] Together with Step 2, this proves the membership equivalence. The algebra isomorphism also transports primality.

Step 4. Apply the prime-localization unit criterion

In localization at a prime, the maximal ideal consists of fractions with numerator in that prime, and an element from the original ring becomes a unit exactly when it lies outside the prime.

Apply these two facts to \(\lambda(v)\), then use Step 3. They give both localized assertions.

The construction is defined for any subspace \(P\). The geometric localization step takes \(P\) to be a root plane, so labels of roots outside that plane become invertible.

Step 5. Compute the generators for a two-root span

For positive roots \(a,b\), take \(P=P(a,b)=\operatorname{span}_{\mathbb R}\{a,b\}\). Any \(v\in P\) is \(v=sa+tb\), so \[ \lambda(v)=s\lambda(a)+t\lambda(b) \] lies in \((\lambda(a),\lambda(b))\). Hence the ideal generated by all \(\lambda(v)\), \(v\in P\), is contained in that two-generator ideal.

Conversely, \(a,b\in P\), so both generators belong to \(\mathfrak p_P\). Thus \[ \mathfrak p_{P(a,b)}=(\lambda(a),\lambda(b)). \] The source's planePrime_span obtains the same identity by transporting the linear span and its generated ideal through the embedding and polynomial coordinate map.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.TitsSpace.embedLinear: PlaneLocalization.lean:346.

OAI.KLInvariance.TitsSpace.embedLinear_injective: PlaneLocalization.lean:351.

OAI.KLInvariance.TitsSpace.planePrime: PlaneLocalization.lean:355.

OAI.KLInvariance.TitsSpace.planePrime_isPrime: PlaneLocalization.lean:359.

OAI.KLInvariance.TitsSpace.linear_mem_planePrime_iff: PlaneLocalization.lean:362.

OAI.KLInvariance.TitsSpace.linear_mem_maximal_iff: PlaneLocalization.lean:383.

OAI.KLInvariance.TitsSpace.linear_isUnit_iff: PlaneLocalization.lean:389.

OAI.KLInvariance.BruhatGraph.planePrime_span: TerminalKernels.lean:546.

Used by