Definition Linear reflections from a root and a normal
✓ definition formalized

Linear reflections from a root and a normal

Let \(V\) be a real vector space, \(\alpha\in V\), and \(f:V\to\mathbb R\) a linear functional. The associated pre-reflection is \[ r_{\alpha,f}(v)=v-f(v)\alpha. \] When \(f(\alpha)=2\), this map is a linear reflection. Assume this normalization in the following calculation; it implies that both \(\alpha\) and \(f\) are nonzero.

Fixed and moving directions

For \(v\in\ker f\), the formula gives \(r_{\alpha,f}(v)=v\). It also gives \(r_{\alpha,f}(\alpha)=-\alpha\). Every vector decomposes as \[ v=\left(v-\frac{f(v)}2\alpha\right)+\frac{f(v)}2\alpha. \] The first term lies in \(\ker f\). If \(c\alpha\in\ker f\), then \(0=f(c\alpha)=2c\), so the intersection of the two summands is zero. Thus \[ V=\ker f\oplus\mathbb R\alpha. \] The reflection acts by \(+1\) on the first summand and by \(-1\) on the second.

For a direct check, linearity gives \(f(r_{\alpha,f}(v))=-f(v)\). Therefore \[ r_{\alpha,f}(r_{\alpha,f}(v)) =v-f(v)\alpha-(-f(v))\alpha=v. \] The reflection is an invertible linear map and is its own inverse.

The image of \(r_{\alpha,f}-1\) is \(\mathbb R\alpha\): every value is a multiple of \(\alpha\), and its value at \(\alpha\) is \(-2\alpha\). Its fixed space is \(\ker f\), of codimension one.

Reflection defined by a bilinear form

Let \(B\) be symmetric and satisfy \(B(\alpha,\alpha)=2\). Taking \(f=B(\alpha,-)\) gives \[ r_\alpha(v)=v-B(\alpha,v)\alpha. \] Put \(a=B(\alpha,v)\) and \(b=B(\alpha,w)\). Bilinearity gives \[ \begin{aligned} B(r_\alpha(v),r_\alpha(w)) &=B(v,w)-bB(v,\alpha)-aB(\alpha,w)+abB(\alpha,\alpha)\\ &=B(v,w)-ab-ab+2ab\\ &=B(v,w). \end{aligned} \] Hence \(r_\alpha\) preserves \(B\). This uses symmetry and the normalization \(B(\alpha,\alpha)=2\).

For the Tits form, take \(\alpha=\alpha_i\) and \(f=\nu_i\). These are the simple operators used to construct the geometric action.

References

Realizations in the OpenAI notes.

Lean formalization

OAI.KLInvariance.Realization.geometricSimple: Tits.lean:311.

OAI.KLInvariance.Realization.simple_normal_value: Tits.lean:307.

OAI.KLInvariance.Realization.geometricSimple_preserves: Tits.lean:339.

Used by