Theorem Normalized KL reciprocity in the square parameter
✓ theorem proved

Normalized KL reciprocity in the square parameter

Let \(t\ne0\) lie in a field and let \(x\le b\). Write \[ p_x(t)=t^{-d(x,b)}P_{x,b}(t^2). \] Then \[ p_x(t)=\sum_{z\in[x,b]} \widetilde R_{x,z}(t^{-1}-t)\,p_z(t^{-1}). \] For real \(0<t<1\), the parameter \(t^{-1}-t\) is positive.

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. Evaluate the canonical reciprocity equation

Take a nonzero field element \(v\). The canonical normalization gives \[ \operatorname{reflect}_{d(x,b)}P_{x,b} =\sum_{z\in[x,b]}R_{x,z}P_{z,b}. \] The canonical P-value equals the recursive P-value, whose weak degree bound is at most \(d(x,b)\). The bounded reflection evaluation theorem therefore applies at the unit \(v^2\), giving \[ v^{2d(x,b)}P_{x,b}(v^{-2}) =\sum_{z\in[x,b]}R_{x,z}(v^2)P_{z,b}(v^2). \] Evaluation preserves the finite sum and each product.

Step 2. Multiply by the scale

Multiply both sides by \(v^{-d(x,b)}\). On the left the exponents add: \[ v^{-d(x,b)}v^{2d(x,b)}P_{x,b}(v^{-2}) =v^{d(x,b)}P_{x,b}(v^{-2}) =p_x(v^{-1}). \] The final equality uses \((v^{-1})^{-d}=v^d\) and \((v^{-1})^2=v^{-2}\).

Step 3. Convert each R-factor and combine the powers

For each \(x\le z\le b\), the square-parameter normalization gives \[ R_{x,z}(v^2)=v^{d(x,z)} \widetilde R_{x,z}(v-v^{-1}). \] Length differences add: \(d(x,b)=d(x,z)+d(z,b)\). Thus the corresponding scaled summand becomes \[ \begin{aligned} v^{-d(x,b)}R_{x,z}(v^2)P_{z,b}(v^2) &=\widetilde R_{x,z}(v-v^{-1}) v^{-d(x,b)+d(x,z)}P_{z,b}(v^2)\\ &=\widetilde R_{x,z}(v-v^{-1})\,p_z(v). \end{aligned} \] This proves the inverse-direction formula \[ p_x(v^{-1})=\sum_z\widetilde R_{x,z}(v-v^{-1})p_z(v). \]

Step 4. Substitute the inverse parameter

Set \(v=t^{-1}\), which is nonzero. Then \(v^{-1}=t\) and \(v-v^{-1}=t^{-1}-t\). Substitution gives the stated formula.

For real \(0<t<1\), division of \(t<1\) by \(t>0\) gives \(1<t^{-1}\). Hence \(t^{-1}>1>t\), proving \(t^{-1}-t>0\).

Step 5. Evaluate the rank-one example

For \([1,s]\), the scaled values are \(p_1(t)=t^{-1}\), \(p_s(t)=1\), and \(\widetilde R_{1,s}(T)=T\). The equation reads \[ t^{-1}=t+(t^{-1}-t)\cdot1. \] Its two terms correspond to the bottom and top vertices in the reciprocal interval sum.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.scaledKL_reciprocity_inverse: Dihedral.lean:737.

OAI.KLInvariance.scaledKL_reciprocity: Dihedral.lean:790.

Used by