The perfect trace pairing of a reflection extension
Let \(A\) be a commutative domain over a characteristic-zero field \(k\), and let \(\tau:A\to A\) be a \(k\)-algebra involution. Suppose \[ \alpha\ne0,\qquad \tau(\alpha)=-\alpha,\qquad \alpha\mid a-\tau(a)\quad(a\in A). \] With \(R=A^\tau\) and the reflection trace \(t\), the map \[ D:A\longrightarrow\operatorname{Hom}_R(A,R), \qquad D(a)(b)=t(ab) \] is an \(R\)-linear isomorphism. Thus \(R\to A\) is a finite free Frobenius extension with basis \(1,\alpha\), and \[ D(ca)(b)=D(a)(cb). \]
Relation to the OpenAI notes
Explicit tensor construction underlying the duality.
OpenAI notes: The grading-normalization proof starts from a normalized self-dual projective, uses the cited moment-graph/Soergel equivalence and character theorem, and translates doubled degrees back to the degree-one convention. bg:normalization.
Lean (primary): The local construction begins with the simple reflection fixed algebra and the trace extracting the root coefficient, normalized by t(alpha)=1. It recovers linear functionals through the trace, extends the resulting duality by tensor product, and proves a one-degree pairing shift for each letter. Decomposition:1588, Frobenius:168, Frobenius:270, Frobenius:602.
Role in the proof: These nodes supply the actual word-module pairing and its homogeneous basis calculation. The boundary-pairing construction subsequently transports this pairing to a boundary summand. The source uses roots of degree one throughout this tensor calculation.
Proof
Step 1. Express the pairing in invariant coordinates
The reflection coordinate theorem gives \[ C:A\xrightarrow{\sim}R^2,\qquad a\mapsto(a_0,a_1). \] The trace-product formula states \[ t(ab)=a_0b_1+a_1b_0. \] Consequently, in these coordinates the duality is the map \[ K:R^2\longrightarrow\operatorname{Hom}_R(R^2,R),\qquad K(c_0,c_1)(b_0,b_1)=c_0b_1+c_1b_0. \] Each formula is linear over \(R\), including linearity in the representing vector \((c_0,c_1)\).
Step 2. Invert the coordinate pairing
For an \(R\)-linear functional \(\lambda:R^2\to R\), define \[ J(\lambda)=(\lambda(0,1),\lambda(1,0)). \] Every vector decomposes as \[ (b_0,b_1)=b_0(1,0)+b_1(0,1). \] Hence \[ K(J(\lambda))(b_0,b_1) =\lambda(0,1)b_1+\lambda(1,0)b_0 =\lambda(b_0,b_1). \] Conversely, evaluating \(K(c_0,c_1)\) at \((0,1)\) gives \(c_0\), and evaluating at \((1,0)\) gives \(c_1\). Thus \(J K\) and \(K J\) are identities. This is the source's crossPairEquiv.
Step 3. Transport the isomorphism back to the algebra
Precomposition with \(C\) identifies functionals on \(R^2\) with functionals on \(A\): \[ \lambda\longmapsto \lambda\circ C. \] It is an isomorphism with inverse precomposition by \(C^{-1}\). Compose it with \(K\) and \(C\). For \(a,b\in A\), the composite has value \(a_0b_1+a_1b_0=t(ab)\), so it is precisely \(D\).
An explicit inverse sends a functional \(f:A\to R\) to \[ D^{-1}(f)=f(\alpha)+\alpha f(1). \] Testing on the basis \(1,\alpha\) verifies the formula. Equivalently, the trace-pairing matrix is \[ \begin{pmatrix}0&1\\1&0\end{pmatrix}, \] whose square is the identity over \(R\).
Step 4. Verify the multiplication compatibility
For \(a,b,c\in A\), associativity and commutativity give \[ D(ca)(b)=t(cab)=t(a(cb))=D(a)(cb). \] The basis \(1,\alpha\) already proves finite freeness over \(R\). Together with \(D\) and its trace formula, this supplies every part of the Frobenius extension data.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.ReflectionFrobenius.crossPair: Decomposition.lean:1552.
OAI.KLInvariance.ReflectionFrobenius.crossPair_bijective: Decomposition.lean:1560.
OAI.KLInvariance.ReflectionFrobenius.crossPairEquiv: Decomposition.lean:1577.
OAI.KLInvariance.ReflectionFrobenius.duality: Decomposition.lean:1583.
OAI.KLInvariance.ReflectionFrobenius.duality_apply: Decomposition.lean:1588.
OAI.KLInvariance.ReflectionFrobenius.duality_balanced: Decomposition.lean:1593.