Skip to content

Documentary state: DRAFT_UNACCEPTED. This draft has NOT been editorially admitted. Mathematical claims require independent checking.

OAI-26 family 076: ultraflat real Littlewood polynomials

Editorial status: DRAFT_UNACCEPTED. NO_CERTIFICATION. This article explains three pinned publisher-reported manuscripts; it does not independently verify the constructions.

The question in plain language

A Littlewood polynomial has coefficients that are all either plus one or minus one. Despite this restriction, could one arrange the signs so that the polynomial's magnitude stays almost constant as its input moves around the unit circle?

The pinned OAI-26 overview claims the answer is yes, even for real Littlewood coefficients and for every sufficiently large polynomial length. It also reports binary sequences whose merit factors grow without bound, contradicting a conjectured uniform bound.

A small example

A real Littlewood polynomial of length \(N\) has the form

\[ P_N(z)=\sum_{j=0}^{N-1}a_jz^j,\qquad a_j\in\{-1,1\}. \]

For \(|z|=1\), orthogonality implies that the average of \(|P_N(z)|^2\) around the circle equals \(N\). So the natural magnitude scale is \(\sqrt N\).

Being ultraflat means much more than attaining that scale on average: the magnitude is close to \(\sqrt N\) uniformly at every point of the circle. Real coefficients make the values at \(z=1\) and \(z=-1\) sums of signs; ensuring the bound at these endpoints is an especially important qualification.

What is known

The protected catalogue identifies three manuscripts: Ultraflat real Littlewood polynomials and Nearly minimal maxima and positive minima of Littlewood polynomials (both October 5, 2026), and Asymptotically minimal maxima of real Littlewood polynomials (September 23, 2026).

The pinned publisher overview claims constructions for every sufficiently large integer length \(N\) satisfying

\[ \sup_{|z|=1}\left|\frac{|P_N(z)|}{\sqrt N}-1\right|\longrightarrow0. \]

The stated uniformity covers the entire unit circle, including the real endpoints. The indexed overview also claims that the binary merit factors associated with the constructions tend to infinity, contradicting Turyn's bounded-merit-factor conjecture.

Uniform approximation of the modulus is stronger than a near-minimal maximum alone; a lower envelope that approaches the same scale is required. The exact definition and normalization of the merit factor, and whether all lengths or a selected sequence are used, demand source-level checking.

GCL verification state

Source lock: openai/math@1581d4cbdd69a0faf34b2089614fe2a814316a6a. The protected catalogue identifies three PDFs, and the publisher overview states the ultraflatness and merit-factor claims.

This editorial pass has not independently reproduced the sign constructions, verified their uniform estimates, or proved the claimed merit-factor divergence. No Family 076-specific Lean formalization, Comparator replay, semantic-equivalence audit, or MATHCERT certificate is established.

What remains

Read all three pinned manuscripts, identify the main theorems, and check how the upper and lower modulus estimates combine. Verify the quantification over all sufficiently large lengths, not merely an infinite subsequence, and explicitly inspect the endpoint behavior at \(z=1\) and \(z=-1\).

Recompute the connections between polynomial modulus and aperiodic binary autocorrelation, establish the exact merit-factor convention, and verify that the constructions indeed produce an unbounded sequence of merit factors. Distinguish the claim that factors diverge from any claimed rate of divergence.

Complete mathematical Markdown and rendered-HTML validation, source-specific editorial criticism and protected merge/readback. Documentary admission is not mathematical certification.

Sources and claim boundaries

Boundary: NO_CERTIFICATION. These are publisher-reported claims, not independently verified or certified results.


Indexed primary manuscript paths: