Theorem Extracting the bottom R-tilde row from character comparison
✓ theorem proved

Extracting the bottom R-tilde row from character comparison

Suppose two intervals are isomorphic and, for every \(z\) strictly below the top, \[ \widetilde R_{u,z}=\widetilde R'_{u',\phi(z)}. \] If all upper kernels needed in the current schedule are free, then \[ \widetilde R_{u,b}(t^{-1}-t) \le \widetilde R'_{u',b'}(t^{-1}-t) \qquad(0<t<1). \]

Relation to the OpenAI notes

Different proof route.

OpenAI notes: The simultaneous induction compares normalized KL vectors, obtains P-invariance from the strict half-degree bound, and then recovers bottom terminal-kernel freeness from endpoint equality. sec:induction, eq:comparison, eq:reciprocal-vector.

Lean (primary): The simultaneous induction first proves normalized R-invariance together with suffix-kernel freeness in both transported directions. It cancels known bottom R-row entries against one actual source-character vector. Canonical KL equality is proved afterwards in a separate induction using the now common R-matrix and the canonical degree bound. Invariance:136, Invariance:190, Invariance:427.

Role in the proof: Proper upper intervals supply kernel freeness, and proper initial intervals supply known R-entries. The new R-equality gives exact endpoints and then current-rank suffix freeness. The canonical degree bound enters the subsequent P-induction.

Proof

Step 1. Define the vector and both matrices on the same index set

Fix \(0<t<1\) and \(T=t^{-1}-t\). Let \(X=[a,b]\). Using the actual source sheaf, set \[ v_z=c_z(t^{-1}),\qquad A_{x,z}=\widetilde R_{x,z}(T),\qquad A'_{x,z}=\widetilde R'_{\phi(x),\phi(z)}(T). \] Both matrices and the vector are indexed by \(X\). In particular, \(v\) is not changed to a target-system character vector when \(A'\) is used.

Step 2. Write the two identities to be compared

The source character reciprocity gives the exact equality \(c(t)=Av\). Upper-kernel freeness supplies the image inclusions required for The transported comparison for actual characters, giving \(c(t)\le A'v\).

Substitute the first equality into the second and take coordinate \(a\): \[ \sum_{z\in X}A_{a,z}v_z\le\sum_{z\in X}A'_{a,z}v_z. \] The interval is finite, so these are ordinary finite sums.

Step 3. Cancel the already known entries explicitly

The row hypothesis says \(A_{a,z}=A'_{a,z}\) for every \(z\ne b\). Put \(S=\sum_{z\in X,\ z\ne b}A_{a,z}v_z\). Splitting both sums into their top term and the remaining terms gives \[ S+A_{a,b}v_b\le S+A'_{a,b}v_b. \] Subtracting the same real number \(S\) from both sides gives \(A_{a,b}v_b\le A'_{a,b}v_b\). There is no sign assumption on the canceled summands.

Step 4. Compute the remaining scalar

The top stalk is \(A\) generated in degree zero, so \(C_{b,b}(q)=1\). Its normalization exponent is \(\ell(b)-\ell(b)=0\). Hence \[ v_b=c_b(t^{-1})=(t^{-1})^0C_{b,b}(t^{-2})=1. \] An interval order isomorphism takes the unique bottom to the unique bottom and the unique top to the unique top; thus \(\phi(a)=a'\), \(\phi(b)=b'\). The inequality from Step 3 is precisely \[ \widetilde R_{a,b}(T)\le\widetilde R'_{a',b'}(T). \]

The argument isolated one matrix entry using the same source vector on both sides. It did not require identifying source and target stalk characters.

References

6. Reciprocity and simultaneous induction.

Lean formalization

OAI.KLInvariance.BruhatGraph.rTilde_eval_le_of_upper_freeness: SimultaneousInvariance.lean:444.

Used by