Theorem Unique positive roots representing reflections
✓ theorem proved

Unique positive roots representing reflections

In a finite-generator Coxeter system, every reflection has exactly one positive representing root. The resulting correspondence is a bijection between the reflections of \(W\) and its positive real roots.

Proof

Step 1. Norm and conjugated reflection formula

Suppose \(a=\rho_0(w)\alpha_i\) represents \(t=ws_iw^{-1}\). Invariance of the Tits form gives \[ B(a,a)=B(\alpha_i,\alpha_i)=2. \] Write \(\widetilde a=(a,0)=\rho(w)\widetilde\alpha_i\). For \(z\in E\), conjugate the simple-reflection formula: \[ \begin{aligned} \rho(t)z &=\rho(w)\left(\rho(w^{-1})z- \widetilde B(\widetilde\alpha_i,\rho(w^{-1})z)\widetilde\alpha_i\right)\\ &=z-\widetilde B(\widetilde a,z)\widetilde a. \end{aligned} \] The coefficient equality uses preservation of the extended form by \(\rho(w)\). Restricting to the embedded ordinary root space gives \[ \rho_0(t)v=v-B(a,v)a. \]

Replacing \(w\) by \(ws_i\) replaces \(a\) by \(-a\), while \[ (ws_i)s_i(ws_i)^{-1}=ws_iw^{-1}=t. \] Thus the negative root represents the same reflection.

Step 2. Uniqueness up to sign

Suppose \(a,b\) represent the same reflection \(t\). Its formula using \(a\), applied to \(a\), gives \(\rho_0(t)a=-a\). Its formula using \(b\) gives \(\rho_0(t)a=a-B(b,a)b\). Equating them yields \[ B(b,a)b=2a,\qquad a=c b,\quad c=B(b,a)/2. \] Taking norms and using \(B(a,a)=B(b,b)=2\) gives \(2=2c^2\). Hence \(c=1\) or \(c=-1\), and \(a=b\) or \(a=-b\).

If both are positive, the alternative \(a=-b\) would make every coordinate of \(a\) both nonnegative and nonpositive. It would be zero, contradicting norm two. Positive representatives are therefore unique.

Step 3. A root determines its reflection

If the same root \(a\) represents \(t\) and \(r\), Step 1 gives the same formula for \(\rho(t)\) and \(\rho(r)\) on every vector of \(E\). The extended action is faithful, so \(t=r\).

Step 4. Existence of a positive representative

A reflection is a conjugate of a simple reflection, so its definition supplies a representing root. The root-sign theorem makes that root positive or negative. In the negative case, Step 1 allows its negative as a representing root, and that negative is positive. Thus every reflection has a positive representative.

Every positive real root has an expression \(a=\rho_0(w)\alpha_i\), which supplies the reflection \(ws_iw^{-1}\). Step 3 makes that reflection independent of the expression.

Step 5. The inverse maps

Starting with a reflection, choose its unique positive root and then the root's reflection. Step 3 returns the original reflection.

Starting with a positive root, take its reflection and then that reflection's positive root. Step 2 returns the original root.

These two constructions are inverse, proving the bijection.

References

The local statements and proof steps are listed below.

Lean formalization

OAI.KLInvariance.TitsSpace.Represents.formula: Tits.lean:1723.

OAI.KLInvariance.TitsSpace.Represents.action_formula: Tits.lean:1737.

OAI.KLInvariance.TitsSpace.Represents.unique_up_to_sign: Tits.lean:1743.

OAI.KLInvariance.TitsSpace.Represents.reflection_unique: Tits.lean:1765.

OAI.KLInvariance.TitsSpace.exists_positive_root: Tits.lean:1789.

OAI.KLInvariance.TitsSpace.positiveRootEquiv: Tits.lean:1821.

Used by