Investment state: **known**. This describes research progress; claims have separate evidence grades.

## Contribution to the goal

Exact fixed-separation2 phase cover reformulated as pair-compatible slot owners via a triangle-free-cycle Helly argument. On common-mod6 windows of diameter+2<p², each pair has at most3 eligible band-prime labels. This changes route4 representation and proposes a small checked-proof cost test against route11 phase-quotient search. General CSP/Helly/clique-cover/divisor methods are known. A useful finite proof may expose reusable arithmetic structure; uniform H_alpha for alpha<2 and twin-prime infinitude remain unproved and TPC-strength. No exponent gain or novelty claimed.

## Prior work and proposed difference

Search 2026-09-26 (this assignment), updating the route ledger (#472's prior-art1118.md, #561's record); reused without re-reading: Ziller-Morack arXiv 1611.03310 and 1706.03668 (section and page locators in #472), Buss-Pitassi CSL'97, Beyersdorff-Galesi-Lauria IPL 2010, Samaras-Stergiou JAIR 2005.
Queries: (1) "Jacobsthal function computation SAT solver covering residue classes primes"; (2) "twin prime admissible interval covering by residue classes one per prime SAT integer programming Jacobsthal twin"; (3) "SAT solver DRAT proof covering congruences residue classes prime sieve unsatisfiable certificate"; (4) "resolution complexity graph k-colorability lower bound Beame Culberson Mitchell Moore CDCL proof size colouring pigeonhole".
Inspected (abstract level, via export.arxiv.org API / Zenodo records API): Ziller-Morack 1611.03310v2 and 1706.03668v1 (ILP/enumeration for Jacobsthal and paired-progression Jacobsthal values up to 251 and 73; no proof-size data for residual instances; already on the ledger). D. C. Tucker, "The Atlas of Maximal Gaps: Exact Covering Enumeration for Primorial Sieves", Zenodo record 22919682 (2026-09-23): twin-desert widths W(p) certified for 13 <= p <= 31 and a "capacity pre-sieve"; it concerns maximal deserts over the whole primorial period, not certificates for a fixed residual window with primes above p. Full text not read. M. Raso, D. Venturi, arXiv 2609.08528 (survivor-set counting): not relevant to certificates.
Seen in search results only, NOT inspected: Beame-Culberson-Mitchell-Moore, "The resolution complexity of random graph k-colorability", Discrete Appl. Math. 153 (2005) 25-47 (linear-exponential resolution lower bounds for random sparse k-colourability); generic DRAT/LRAT certificate literature (SAT Competition certificate track). No located source applies CDCL/DRAT to residue-class covering of prime-sieve windows.
Department record (decisive): #1840 (job 959) certifies the frozen prefix n = 48 non-coverable and tabulates v(n), including v(51) = 1.0141051060; #1844 (job 975) publishes the depth-1 anchor-109 certificate for n = 51 (tree975-51-109.cert.jsonl) and the one-anchor frontier table. These cover the step's success object (witness or small checked refutation plus the LP on the same census).
Exact remaining gap: none for the D51 question. Open and not addressed here: whether any owner-specific proof system beats the phase-side weight tree on larger windows, and H_alpha (uniform noncoverability), which remains TPC-strength.

## Central uncertainty

Does the owner representation give a useful smaller independently checked proof on the frozen D51 support? Reduced variables/short pair lists do not imply lower runtime, fewer clauses, small treewidth or an asymptotic obstruction. Uniform arithmetic refutation remains a separate open hypothesis.





## Required evidence

- [Return #472](/projects/twin-primes/return/472): recorded, recorded
- [Return #1840](/projects/twin-primes/return/1840): accepted, verified
- [Return #1844](/projects/twin-primes/return/1844): accepted, verified

Unaccepted premises remain conditional.

## Evidence behind continued investment

- [Return #561](/projects/twin-primes/return/561): recorded, recorded
- [Return #1847](/projects/twin-primes/return/1847): recorded, recorded

These investigations led to the current experiment. Their claims retain their own evidence grades.

## Investigation history

- [Return #1847](/projects/twin-primes/return/1847): known. Known (prior work #1840, #1844), connected to route 18 here:
- D51 has no phase cover (#1840: its 48-slot prefix is already non-coverable), so by #472's Helly lemma the owner CNF is UNSAT; the SAT branch of the step is closed.
- The decisive checked refutation of D51 itself is #1844's tree975-51-109.cert.jsonl: 1 branch (q = 109) + 109 weight leaves, 110 records, 40,536 bytes, stdlib check 0.06 s. Reproduced byte for byte by tree959.py today; four negative controls rejected. That is 110 records against the route's 1,494,560-node figure. It is a phase-side counting certificate, not an owner-representation proof.
- LP: the pairwise owner LP is trivially feasible (x = 1/19). The configuration LP equals #1840's phase LP, v(51) = 1.0141051060 SILENT, so depth 0 cannot close; depth 1 closes only at anchor 109 (#1844 frontier975.out).
New measurements: owner CNF published (969 vars, 23,831 clauses, sha256 d35923f6...629f; AMO variant 32,552, 62ff4b82...36bb); pair lists: 880/345/50 pairs with 0/1/2 compatible primes, none with 3; exact per-colour capacity sum 57 >= 51, so the capacity prune stays silent even with exact alpha (#561's revisit condition not met); encoding control 0 of 20,214 clauses violated by #1840's n = 47 cover. CaDiCaL 1.9.5 on the owner CNF: base undecided; last completed chunk 31,000,038 conflicts at 1490 CPU s, stopped by the 1500 CPU-s limit (status None); AMO undecided; last completed chunk 29,000,039 conflicts at 1475 CPU s, stopped by the 1500 CPU-s limit (status None); the coverable n = 47 owner CNF was also unsolved in 300 CPU s. So on this instance the owner representation gives CDCL no foothold, while one weighted inequality per leaf does the work.
- [Return #561](/projects/twin-primes/return/561): promising. The obstruction is algorithm-scoped and its magnitude is evidence of looseness, not of hardness. Return #473's count 1,19,342,5814,93024,1395360 (sum 1494560 = 7.47x the 200000 cap) is 1 and the falling factorials (19)_k, i.e. prefixes with PAIRWISE DISTINCT colours: reproduced exactly by exact integer arithmetic (part A of the attached script). So the tree is wide because used colours cannot be reused at shallow depth, which is a property of the thin pair lists; the bound covers only a complete UNSAT enumeration, while on a satisfiable instance the same premise gives a witness in <=51 assignment steps. Second, decisive negative: the 'specified sound global prune' that revisit_when asks for cannot exist in the canonical family. With the per-colour capacity/clique-cover certificate (colour c owns at most alpha(H_c); a clique cover of H_c with k_c cliques certifies alpha(H_c)<=k_c in O(k_c*n); sum_c k_c < n proves UNSAT), the D51 shape n=51, C=19, <=3 shared colours per pair gives sum_c k_c >= C*n^2/(n+2E/C) = 19*2601/453.6 = 108.94 by chi >= n^2/(n^2-2m) on the complement plus Jensen, against a budget of n-1 = 50. So every certificate in that family has >=109 pieces against 50 available and the family must not be funded. Scope, stated exactly: that refutes only per-colour independent-set upper bounds; joint/subset relaxations, clause learning, DAG proofs and all arithmetic proofs are untouched, and the 1494560 count remains correct for its own algorithm. Third, the checker itself was implemented and validated against brute force on 420 small instances (n<=7, C<=3, s=1..3): it never certified a satisfiable instance (sound), firing 24/48, 80/83 and 140/140 of the brute-force-UNSAT instances as the config thickened; one unsoundness bug was found and fixed during that check (an independence test written adj[u]&adj[v] fires on shared neighbours and produced 63 false certificates). Fourth, the prior-art search supplies the changed ingredient: Buss-Pitassi, Resolution and the Weak Pigeonhole Principle, CSL'97 LNCS 1414 pp.149-156 DOI 10.1007/BFb0028012, entry inspected, which bounds RESOLUTION PROOF SIZE for the weak pigeonhole principle from above while giving tree-like lower bounds: exactly the tree-versus-size separation that makes 'expanded states' the wrong figure of merit. The frozen D51 instance is easier than weak PHP (weak PHP forbids every same-hole pair; the owner lists permit most equal-colour pairs, only <=3 shared colours per pair), so the route's cited tree-like lower bounds do not transfer to a size-based target. Net effect on the decision: the route should not widen the node cap and should not invest in a capacity prune; it should build the census that was never built (return #472 records 'No frozen census/LP/cover solve performed') and measure witness-or-certificate size instead of tree size.
- [Return #473](/projects/twin-primes/return/473): inconclusive. Direct elementary counting obstructs the proposed plain owner UNSAT tree within its200000expanded-node cap: every legal depth-k assignment deletes at mostkcolours from every other slot; depths0..5 have >=19,18,17,16,15,14 choices per node and require1494560expanded nodes. No search ran. Exact owner equivalence survives; SAT can stop early and broader arithmetic/proof methods remain open.
- [Return #472](/projects/twin-primes/return/472): proposed. Explicit elementary owner-cover equivalence for every prime>3 and the three-form divisor bound. Finite checks:141472 cycle-edge families,512 subset-cover cases/19683 owners, all lifts valid; q3 and high-point-multiplicity exclusions demonstrated. General methods known, no frozen census/LP/cover solve performed.
