Definition The symmetric coefficient algebra and its action
✓ definition formalized

The symmetric coefficient algebra and its action

Fix a finite-generator Coxeter system and let \(V\) be its extended real root space. The coefficient algebra is \[ A=\operatorname{Sym}_{\mathbb R}(V). \] It is the commutative real algebra generated by \(V\), subject to the linear relations already holding in \(V\). Write \(\iota:V\to A\) for the linear map taking a vector to its generator.

Its defining property is that every linear map \(f:V\to D\), where \(D\) is a commutative real algebra, extends uniquely to an algebra homomorphism \(\widehat f:A\to D\). The extension satisfies \(\widehat f(\iota(v))=f(v)\). Thus a map on the vector generators determines a map on every sum and product of generators.

Polynomial coordinates

Choose a basis \(e_1,\ldots,e_N\) of \(V\). Sending \(\iota(e_i)\) to \(X_i\) gives \[ A\simeq\mathbb R[X_1,\ldots,X_N]. \] The reverse homomorphism sends \(X_i\) to \(\iota(e_i)\). The composites fix all the generators, so both composites are the identity.

For \(v=\sum_i c_i e_i\), its generator becomes the linear polynomial \(\sum_i c_iX_i\). For example, \(2e_1-e_2\) becomes \(2X_1-X_2\). A basis of the extended space includes the added dual directions, so those directions also supply polynomial variables.

A nonzero vector gives a nonzero degree-one polynomial, since some basis coefficient is nonzero. In particular, a root vector gives the nonzero linear label used on an edge.

The transported symmetric grading

Let \(\Phi:A\to\mathbb R[X_1,\ldots,X_N]\) be the chosen polynomial-coordinate isomorphism. The source defines \[ A_n=\Phi^{-1}\bigl(\mathbb R[X_1,\ldots,X_N]_n\bigr). \] Equivalently, it takes the image of the polynomial degree-\(n\) subspace under \(\Phi^{-1}\).

To decompose \(a\in A\), decompose \(\Phi(a)\) into finitely many homogeneous polynomial components and apply \(\Phi^{-1}\) to each. Uniqueness follows by applying \(\Phi\) again. Since \(\Phi\) is an algebra isomorphism, products of pieces have the sum of their degrees.

Thus each \(\iota(v)\) is degree one, constants are degree zero, and negative pieces vanish. This is the symmetric grading used by the word-module constructions.

Extending the group action to coefficients

The extended action \(\rho(w):V\to V\) induces the algebra automorphism \(w_A\) determined by \[ w_A(\iota(v))=\iota(\rho(w)v). \] On a product it acts factor by factor: \[ w_A(\iota(v_1)\cdots\iota(v_m)) =\iota(\rho(w)v_1)\cdots\iota(\rho(w)v_m). \] The operators \((wz)_A\) and \(w_Az_A\) agree on each generator by the group-action law on \(V\); hence they agree on all of \(A\).

For a simple root \(\alpha_i\), the reflection formula gives \[ (s_i)_A(\iota(v)) =\iota(v)-\widetilde B(\alpha_i,v)\iota(\alpha_i). \] In particular, \((s_i)_A(\iota(\alpha_i))=-\iota(\alpha_i)\). The fixed subalgebra consists of those polynomials \(a\) for which \((s_i)_A(a)=a\). It is the coefficient ring over which the next tensor-induction step is taken.

Every generator has degree one. Linear actions preserve that degree, so their algebra extensions preserve the total-degree grading.

References

The local source declarations below define the constructions and prove the stated properties.

Lean formalization

OAI.KLInvariance.TitsSpace.SymmetricCoefficient: Frobenius.lean:18.

OAI.KLInvariance.TitsSpace.coefficientAction: Frobenius.lean:20.

OAI.KLInvariance.TitsSpace.coefficientAction_ι: Frobenius.lean:23.

OAI.KLInvariance.TitsSpace.simpleRootPolynomial: Frobenius.lean:48.

OAI.KLInvariance.TitsSpace.symmetricGrading: Frobenius.lean:777.

OAI.KLInvariance.TitsSpace.mem_symmetricGrading: Frobenius.lean:782.

OAI.KLInvariance.TitsSpace.coefficientAction_mul: Frobenius.lean:33.

OAI.KLInvariance.TitsSpace.coefficientAction_simple: Frobenius.lean:28.

Used by