Definition Kernel Hilbert values and normalized states
✓ definition formalized

Kernel Hilbert values and normalized states

Let \(A=\mathbb R[X_i:i\in\Sigma]\), where \(\Sigma\) is finite and the variables have degree one. Let \(\mathcal B\) be a graded sheaf with graded module data at its vertices and edges, including finite generation and zero negative-degree pieces.

The selected-kernel Hilbert value

For a vertex \(x\) and outgoing selection \(F\), let \(K_x(F)\) be the selected kernel. Its degree-\(n\) piece is the actual intersection \[ K_x(F)_n=K_x(F)\cap(\mathcal B^x)_n. \] The kernel Hilbert value is \[ H_x(F;q)= \sum_{n\in\mathbb Z}\dim_{\mathbb R}K_x(F)_n\,q^n, \qquad 0<q<1. \] This is the convergent polynomial-module specialization of GradedSheaf.kernelHilbert, which evaluates the source's Hilbert.value on this actual submodule.

The Hilbert-convergence theorem gives finite dimensions and convergence. The field ModuleData.nonneg sets every negative stalk piece to zero. The selected kernel inherits this support, so its Hilbert sum over \(\mathbb Z\) equals the sum over \(n\ge0\).

The state scale on a Coxeter interval

Now fix a finite-generator real Coxeter system \(W\), \(u\le_Wb\), a finite real basis of its extended realization, and a boundary sheaf \(\mathcal B\) on \([u,b]_W\) over the corresponding polynomial algebra \(A\). Put \[ N=|\Sigma|, \] the number of basis coordinates, equivalently the number of polynomial variables.

For \(x\in[u,b]_W\) and \(0<t<1\), the source's state scale and normalized kernel are \[ s_x(t)=(1-t^2)^N\,t^{\,\ell_W(x)-\ell_W(b)}, \qquad D_x(t,F)=s_x(t)\,H_x(F;t^2). \] These are stateScale and normalizedKernel. The exponent \(\ell_W(x)-\ell_W(b)\) is an integer. The comparison uses \(0<t<1\), so \(0<t^2<1\) and the Hilbert value converges.

The factor \((1-t^2)^N\) removes the coefficient-ring denominator. If the kernel has a finite homogeneous free basis with generator degrees \(d_1,\ldots,d_r\ge0\), the formula becomes \[ D_x(t,F)=t^{\,\ell_W(x)-\ell_W(b)} \sum_{i=1}^r t^{2d_i}. \] The definition by intersection dimensions also applies to partial kernels before their freeness is established.

The coefficient used when a condition is removed

Define \[ T(t)=t^{-1}-t. \] The edge-normalization theorem proves that the correction degree makes this same coefficient occur at every edge. It also proves \(s_x(t)>0\) and \(T(t)>0\) for \(0<t<1\).

A single-edge comparison uses the normalized values at the lower endpoint with and without its current condition, and the unchanged selected-kernel value at the upper endpoint. A full schedule collects these numbers into a vector indexed by the interval vertices.

Relation to the OpenAI notes

Equivalent parameter conventions.

OpenAI notes: The comparison uses \(0<q<1\), scale \(q^{-d/2}(1-q)^N\), and transfer coefficient \(q^{-1/2}-q^{1/2}\). eq:state, eq:step.

Lean (primary): The comparison uses \(0<t<1\) and substitutes \(q=t^2\). Its scale is \(t^{-d}(1-t^2)^N\) and its transfer coefficient is \(t^{-1}-t\). Dihedral:730, PositiveTransfer:591, PositiveTransfer:621, EdgeLocalization:377, EdgeLocalization:423, PositiveTransfer:535, PositiveTransfer:603.

Role in the proof: The substitution \(q=t^2\) identifies the two normalizations. Linear root labels have degree one throughout the calculation.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.MomentGraph.GradedSheaf.kernelHilbert: PositiveTransfer.lean:359.

OAI.KLInvariance.BruhatGraph.stateScale: PositiveTransfer.lean:591.

OAI.KLInvariance.BruhatGraph.normalizedKernel: PositiveTransfer.lean:599.

Used by