Theorem The normalized single-edge inequality and its equality criterion
✓ theorem proved

The normalized single-edge inequality and its equality criterion

Fix a finite-generator real Coxeter system \(W\), \(u\le_Wb\), a finite real basis of the extended realization and its polynomial coefficient algebra \(A\). Let \(\mathcal B\) be a boundary sheaf on \([u,b]_W\).

For an edge \(e:x\to y\) with positive source root \(a_e\), choose arbitrary outgoing selections \(F,H\) at \(x,y\). Let \[ I=\rho_e^-(K_x(F)),\qquad J=\rho_e^+(K_y(H)), \] and let \(\mu_e\) be multiplication on the actual edge module by the correction lift \(F_{x,a_e}\). Assume the embedded image inclusion \[ I\subseteq\mu_e(J). \] Write \(H_x(F;q)\) for the kernel Hilbert value. With \(N\) the cardinality of the chosen coefficient basis, the state scale and normalized kernel value are \[ s_x(t)=(1-t^2)^N t^{\ell_W(x)-\ell_W(b)}, \qquad D_x(t,F)=s_x(t)H_x(F;t^2). \] For every \(0<t<1\), they satisfy \[ D_x(t,F)\le D_x(t,F\cup\{e\})+T(t)D_y(t,H), \qquad T(t)=t^{-1}-t. \] Equality at the chosen \(t\) holds if and only if \[ I=\mu_e(J). \]

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.

Proof

Step 1. Apply the raw Hilbert increment at q=t squared

Set \(q=t^2\). The assumptions \(0<t<1\) give \(0<q<1\). The actual correction map is homogeneous and injective, with degree \[ k=\deg(F_{x,a_e}) =\frac{\ell_W(y)-\ell_W(x)-1}{2}. \] The raw edge-increment theorem applies to the actual boundary sheaf and the supplied inclusion: \[ H_x(F;q)-H_x(F\cup\{e\};q) \le q^k(1-q)H_y(H;q). \] It also states that equality in this raw formula is equivalent to \(I=\mu_e(J)\).

Step 2. Multiply by the positive state scale

The edge-normalization theorem gives \(s_x(t)>0\). Multiplying the raw inequality by that positive number preserves its direction: \[ s_x(t)H_x(F;t^2)-s_x(t)H_x(F\cup\{e\};t^2) \le s_x(t)(t^2)^k(1-t^2)H_y(H;t^2). \] The same theorem identifies the factor on the right as \(T(t)s_y(t)\). By the definitions of the normalized kernel values, this becomes \[ D_x(t,F)-D_x(t,F\cup\{e\}) \le T(t)D_y(t,H). \] Moving the subtracted term to the right proves the stated inequality.

Step 3. Recover the raw equality by cancelling a nonzero scale

Suppose the normalized inequality is an equality. Substitute the definitions and the state-scale identity to obtain \[ s_x(t)H_x(F;t^2) =s_x(t)H_x(F\cup\{e\};t^2) +s_x(t)\bigl((t^2)^k(1-t^2)\bigr)H_y(H;t^2). \] The factor \(s_x(t)\) is nonzero. Divide by it and move the second term to the left. This gives exactly the raw equality in Step 1.

Conversely, multiplying that raw equality by \(s_x(t)\) and applying the same factor identity gives the normalized equality. This is the source's scaled_increment_eq_iff cancellation.

Step 4. Apply the actual image equality criterion

Step 1 identifies raw equality with \(I=\mu_e(J)\). Step 3 identifies normalized equality with raw equality. Combining the two equivalences proves the asserted criterion for the two actual embedded images.

Application to a chronological step

Take the canonical boundary sheaf restricted from \([1,b]_W\), an interval order isomorphism to a second finite-generator system, and a chronological suffix consisting of \(e\) followed by \(p\). Set \[ F=\mathcal R_p(x),\qquad H=\mathcal R_p(y). \] If the upper selected kernel \(K_y(H)\) is free over \(A\), the schedule edge-inclusion theorem supplies the premise \(I\subseteq\mu_e(J)\).

The condition set before processing \(e\) is \(F\cup\{e\}\) at \(x\), and remains \(H\) at \(y\). At every other vertex it is unchanged as well. Thus the theorem gives the one-coordinate update bound used in the full schedule comparison, with the same positive coefficient \(T(t)\) at each edge.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.BruhatGraph.normalized_edge_increment_le: PositiveTransfer.lean:621.

OAI.KLInvariance.BruhatGraph.normalized_edge_increment_eq_iff: PositiveTransfer.lean:637.

OAI.KLInvariance.Comparison.scaled_increment_eq_iff: PositiveTransfer.lean:562.

Used by