Full Rank-Order Closure Reduces To Pattern Upset Lattice¶
Claim/Theorem¶
For the reconstructed f_3 certificate from [[f3-fan-certificate-passes-necessary-multimorphism-tests-but-not-operation-lift.md]], the meet/join closure generated by the seven observed positive-support columns is not the full 2^50 Boolean lattice. It is exactly the upset lattice of the realized 7-bit support-pattern poset.
More generally, let P be the positive support of a signed certificate and let
be the observed domain columns. Each support point p has a pattern
The closure generated from the A_j by finite unions and intersections is canonically isomorphic to the family of upsets of the realized pattern set R_P=\{\alpha(p):p\in P\}, ordered by coordinatewise comparison.
For the current f_3 certificate:
|P|=50;- all
50positive support patterns are distinct; - the exact domain closure size is
2,849,631; - the old
3001cutoff was only an artifact of naive subset enumeration.
The same representation gives:
- six-qubit control domain closure size
35,664; - toy submodularity control domain closure size
4.
Proof¶
Every expression in the generators A_j using only union and intersection is a positive monotone Boolean formula in the m pattern bits. Therefore its truth set on R_P is an upset.
Conversely, if U\subseteq R_P is an upset and M is the antichain of its minimal elements, then
For a fixed minimal pattern \alpha, the principal upset \{\beta:\alpha\le\beta\} is represented by
Taking the union over M realizes U. Hence the closure is exactly the upset lattice of the realized pattern poset.
This is the exact closure object computed by docs/project_QEM-QEC/tmp/scripts/route_d_full_rank_order_extension.py.
Dependencies¶
- [[conservative-hamming-operation-extension-is-rank-order-lattice-extension.md]]
- [[f3-certificate-passes-pairwise-rank-order-extension-but-not-full-closure.md]]
- [[certificate-lift-frontier-is-operation-extension-not-cone-duality.md]]
Conflicts/Gaps¶
- This node computes the exact domain closure object; it does not decide whether a rank-preserving order-preserving extension exists.
- The codomain image of a closure element is not required to lie in the analogous codomain pattern closure, so the output-pattern closure size is diagnostic only.
- The result compresses the closure from
2^50to a finite2,849,631-element lattice, but the full extension problem still has one monotone label function per negative support point.
Sources¶
10.1016/j.dam.2009.07.001- local computation:
docs/project_QEM-QEC/tmp/certificates/full_rank_order_extension_report.json - local script:
docs/project_QEM-QEC/tmp/scripts/route_d_full_rank_order_extension.py