trureturingGitHub MATHEMATICAL ATLAS

RESEARCH / Focused target

Werner's all-column formula for the affine polynomial pair

Literature status: not recheckedRepository record: proved

RELEASED FOUNDATIONS / PROPOSED CONNECTION

Research connections

0 released anchors
All source anchors (1)

Problem

The question

OEIS A208509 revision 28, FORMULA, attributes this conjecture to Knud Werner:

Conjecture: T(n,k) = binomial(n-1,2k+1) + binomial(n,2k+1). - Knud Werner, Jan 11 2022

Clark Kimberling authored the sequence entry. Its companion A208508 revision 10 supplies the jointly generated first polynomial. The complete defining laws are u(1,X)=v(1,X)=1, u(n,X)=u(n-1,X)+X*v(n-1,X), and v(n,X)=u(n-1,X)+v(n-1,X)+1 for every n>=2, over Z[X]. For every natural n>=1 and every natural k>=0, T(n,k) is the coefficient of X^k in this actual v(n,X), including all zero coefficients beyond its degree. The formula's two natural binomial coefficients are cast to integers. No assertion at n=0 is included.

Motivation

Our foothold

The identity gives every coefficient of an unbounded family specified by a coupled affine polynomial recursion. It connects that actual recursion to a closed binomial expression without restricting rows, columns or degree.

Gap

Missing bridges

The bounded source and literature screen recorded in Library/RecurrenceComposition/werner2022a208509.md found no complete proof or refutation of this exact conjecture. The private-inclusive repository, pinned Mathlib and third-party Lean screen found no full supplier. The frozen A208342 pair has v_next=Xu+Xv and a different endpoint; no identification with this affine second family was established. Elementary coefficient and Pascal identities supply normalization, while the coupled coefficient invariant supplies the missing whole-family argument. These findings do not establish global openness or priority; unavailable searches are not zero-hit searches.

Route

Proposed approach

Use the actual pair recursion with stage m corresponding to source index m+1. Simultaneously maintain, for every natural m,k, the first coefficient choose(m,2k)+choose(m+1,2k)-[k=0] and the second coefficient choose(m,2k+1)+choose(m+1,2k+1), interpreted in the integers. The odd successor uses both preceding components. The even successor uses the shifted odd component, with k=0 handled separately. Pascal identities close the coupled steps; the degree-zero correction cancels the affine constant. Project the second component at stage n-1 and use n-1+1=n for n>=1.

Falsifier

What would falsify this route

A natural n>=1 and k>=0 for which the coefficient of X^k in the polynomials from precisely the two initial values and both recurrences differs from the stated integer binomial sum would refute the conjecture. A failure of either component of the simultaneous invariant, the affine cancellation, or the source-stage identification would invalidate this proof route. Agreement on a finite prefix or a single column cannot prove the universal statement.

Evidence

Evidence to collect

The single current public proof endpoint is D5/S1/Recurrence/Invariants/AffinePolynomialPairBinomialCoefficients.werner_a208509, with statement_id sha256:2611e9479e93d268bcc052fadc93f7932faa6757d6162378face9f7003475552. Its owning module is D5/S1/Recurrence/Invariants/AffinePolynomialPairBinomialCoefficients.lean; its current frozen member is Golden/Frozen/state/D5/S1/Recurrence/Invariants/AffinePolynomialPairBinomialCoefficients.lean.json. The actual definition pair and the proof-local simultaneous invariant are consumed by this endpoint; no second theorem or finite certificate is claimed. The owning Scribe records the typed proved-resolution edge for this dossier.

The source statement and full indices were preregistered as Tier 1 before the probe in #26010. The admission basis is open-problem-resolution; utility: none describes this unbounded symbolic proof. The public theorem has proof_shape: content through its live coupled invariant. Local arithmetic and normalization helpers are bind-only and consumed by the invariant or final index transport; induction syntax alone is not the content criterion. The theorem's accepted axiom closure is propext, Classical.choice, and Quot.sound.

The original source snapshots are:

  • A208509 revision 28: https://raw.githubusercontent.com/oeis/oeisdata/main/seq/A208/A208509.seq, SHA-256 aad0c0fd287110c8b9f5ca1f7c4d4af77bbe3e71dcf284faf7e6baba1efb6c5a.
  • A208508 revision 10: https://raw.githubusercontent.com/oeis/oeisdata/main/seq/A208/A208508.seq, SHA-256 c88c749749f5e4944ef974eab6217576a5195f85fd65a34f671aa5f4e53d3175.

Triage

Scope assessment

theorem: the cited frozen endpoint proves the complete all-column conjecture for the actual second polynomial, every n>=1 and every k>=0, including zero tails. The first-component formula remains proof-local; it is not a separate public result.

What the settlement shows

  • Proved by the cited endpoint: the coupled affine recursion has the stated odd-binomial coefficient formula throughout the full declared index range. The decisive mechanism is the simultaneous even/odd coefficient invariant and its degree-zero correction, which absorbs the added constant.
  • Proved internally on the endpoint's live path: the first-component invariant is needed to propagate the second component, and the shifted second component propagates the first. This explains why an isolated second-component Pascal rewrite does not supply the whole induction.
  • Open: changing the initial polynomials, the coefficient of X, the affine constant, or the coefficient ring requires a new statement and corresponding proof. This result supplies no such generalization. There is no finite range or numerical bound whose sharpness is being claimed.
  • Supporting consequence: the exact A208509 conjecture can be used in its stated source range. No further source theorem depending on it was identified by the bounded screen. Consequences for other entries or conjectures, including an independent public A208508 endpoint, remain unasserted here.

ASSUMED-UNVERIFIED

Unverified assumptions

The bounded literature and supplier screen does not prove worldwide openness, originality, exhaustive absence of a stronger supplier, or publication priority. Semantic Scholar returned 429 and other restricted searches did not provide usable coverage. The source-to-Lean interpretation is supported by the explicit recursion, index correspondence and independent fidelity review; natural-language source equivalence is not itself a kernel theorem. The typed resolution edge records this repository's claim about the cited frozen result and does not certify the conjecture's status elsewhere. Generalizations and unexamined source-dependent consequences remain open.

Dossier & anchor history

Loading release observations...