trureturingGitHub MATHEMATICAL ATLAS

RESEARCH / Focused target

CKZ Conjecture 2: the four actual record distributions

Literature status: not recheckedRepository record: proved

RELEASED FOUNDATIONS / PROPOSED CONNECTION

Research connections

0 released anchors
All source anchors (1)

Problem

The question

Chen, Kitaev and Zhang, Distributions of statistics on separable permutations, arXiv:2404.18517v1, Section 3, Conjecture 2, page 17, asks:

The distribution of rmax or lmin (resp., lmax or rmin) on irreducible (resp., reducible) separable permutations is unimodal with the peak being at k=3 for n>=5 (see Table 4).

The journal reference is Discrete Applied Mathematics 355 (2024), 169–179. The carrier is the actual classical $2413/3142$-avoiding permutations, rather than a recursive replacement class. A proper direct cut separates a nonempty prefix whose values are all below those of the nonempty suffix. Irreducibility means positive length and no proper direct cut; reducibility means a proper direct cut. The empty permutation belongs to neither positive class. The singleton is irreducible and has one record of each type; the reducible singleton class is empty.

The four strict record statistics are right maxima, left minima, left maxima and right minima. A maximum exceeds every value on its indicated side, and a minimum is below every value on its indicated side. These counts are unshifted. For any of the pairs $(\mathrm{irreducible},\mathrm{rmax})$, $(\mathrm{irreducible},\mathrm{lmin})$, $(\mathrm{reducible},\mathrm{lmax})$ and $(\mathrm{reducible},\mathrm{rmin})$, let $a(n,k)$ be the corresponding actual record-fiber cardinality. The complete target is, for every $n\ge5$ and every natural $k$,

$$ \begin{aligned} k<3&\implies a(n,k)\le a(n,k+1),\ k\ge3&\implies a(n,k+1)\le a(n,k),\ a(n,k)&\le a(n,3). \end{aligned} $$

A maximum at three is interpreted weakly. Strict inequalities on zero tails and uniqueness are not required. Issue 12664 registers this full target. A partial rising or transport result does not settle the conjecture.

Motivation

Our foothold

The actual rising side holds for every $n\ge4$ and every $k<3$ in all four distributions. RecordPeak.actual_record_rising supplies the actual right-maximum comparison from two to three. The RecordTransport.actual_four_record_transports_and_rising theorem proves the all-length exact actual record transports, their singleton corrections, the low-record zero fibers and the complete four-distribution rising side. Writing $i_t,d_t$ for the positive irreducible and reducible record counts, these exact transports are

$$ \begin{aligned} i_{\mathrm{lmin}}(n,k)&=i_{\mathrm{rmax}}(n,k),\ d_{\mathrm{lmax}}(n,k)+\delta(n,k)&=i_{\mathrm{rmax}}(n,k),\ d_{\mathrm{rmin}}(n,k)+\delta(n,k)&=i_{\mathrm{rmax}}(n,k), \end{aligned} $$

where $\delta(n,k)=1$ exactly when $n=k=1$. They reduce the remaining four declining laws at $n\ge5$ to the actual irreducible right-maximum law. RecordFirstDecline.actual_four_record_first_decline proves the first declining comparison $a(n,4)\le a(n,3)$ at every natural length, and its strict form for every $n\ge3$, on all four actual class/statistic pairs. The proof establishes its exact positive quotient and infinite tail inside the actual weighted inference.

Gap

Missing bridges

RecordAllIndex.actual_ckz_conjecture_two proves the complete displayed target on literal Equiv.Perm (Fin n) avoiders, with the stated positive irreducible and proper-direct-cut classes and strict, unshifted statistics. It has no assumed generating-function bridge, finite length cutoff or uniqueness claim.

RecordNewtonPositivity.actual_q_record_newton_nonnegative proves the unbounded coefficient sign law $[t^n]G_r(q(t))\ge0$ for every natural $r,n$. Its proof constructs the Motzkin series from Mathlib's Catalan supplier, derives its differential recurrence at every index, and propagates two pairs of consecutive positive-composition certificates to the universal tail. This removes the universal kernel-sign obligation.

RecordAllIndex.actual_weighted_quadratic derives the actual weighted equation from exact least-cut record fibers. Its coefficient formula J_coeff and Newton transform actual_adjacent_difference identify, for every $k\ge3$,

$$ J_k-J_{k+1}=t^k\sum_{r=0}^{k-3}\binom{k-3}{r}G_r(q(t)). $$

These are rational-series differences. Taking every length coefficient and applying the sign law proves actual_rmax_decline for all natural lengths and every $k\ge3$, including zero tails. Exact record transports, the rising side and induction yield all four distributions and their global weak maximum at three for every $n\ge5$.

Route

Proposed approach

The source linkage is the existing SEPARABLE_SIGNED_RECORD_WEIGHTS theory, especially its actual weighted quadratic and Section 8 coefficient and kernel identities. The Lean proof constructs the actual quadratic, normalizes it using the Catalan/Motzkin series, proves the exact binomial coefficient identities, substitutes the actual scalar $q$, and extracts coefficients before applying the actual record transports.

Falsifier

What would falsify this route

An actual source-class permutation count at some $n\ge5$ and natural $k$ violating a required adjacent inequality, or exceeding the count at three, refutes the corresponding conjectured law. A counterexample to an auxiliary coefficient estimate refutes that estimate; it does not automatically refute the actual permutation conjecture.

Evidence

Evidence to collect

The source v1 and the author and institutional manuscript bodies retain Conjecture 2. The bounded literature screen, detailed in the Library entry, reports no equivalent proof or refutation in the source manuscripts, five known citing bodies and the related papers inspected. The journal version of record body remains unverified. This is a non-discovery in the searched scope, not exhaustive global novelty, priority or exclusive ownership. Fu–Lin–Zeng's descent real-rootedness problem and arXiv:2608.27583v1 Conjecture 15 are separate open problems.

Triage

Scope assessment

Theorem: the literal weak-peak formulation of the published conjecture is proved in Lean. This mathematical conclusion does not establish worldwide priority or a newly solved-problem KPI. The Scribe's typed resolution links the theorem to this source-v1 target; it supplies no literature priority evidence or completed registration audit.

Proved: the controlling structure is the exact binomial expansion of each adjacent difference into nonnegative Newton kernels. The right-maximum decline holds at every natural length; the full peak theorem uses $n\ge5$. Open in the digestion ledger: separate coverage of the stronger all-length four-statistic decline and the multi-clause coefficient exposition. No neighbouring descent or Conjecture 15 claim is settled by this result.

ASSUMED-UNVERIFIED

Unverified assumptions

The journal version of record body, exhaustive citation coverage and global novelty are unverified. Reg/enrollment remains owner-paused under issue 12664. The public theorem audit is unfinished under the linked-issue exception; no declared_validated status is claimed. Final independent review and normal integration remain separate from the mathematical result.

Dossier & anchor history

Loading release observations...