OctoInformBeta

OpenAI preprint, September 2026 · §7. Flatness for every finite exponent

Corollary 7.1: One signing per length NN with |Pε(e(t))|/N→1|P_\varepsilon(e(t))|/\sqrt{N}\to 1 in every LpL^p, p>0p>0, and a family of least maximum modulus flat in L2L^2

There are Littlewood polynomials flat for every finite exponent.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Write e(x)=exp⁡(2πix)e(x)=\exp(2\pi i x) and 𝕋=ℝ/ℤ\mathbb{T}=\mathbb{R}/\mathbb{Z}, so that t∈𝕋t\in\mathbb{T} runs over the unit circle as z=e(t)z=e(t). A Littlewood polynomial of length NN has its NN coefficients in {−1,1}\{-1,1\}, and the flatness problem asks how closely its modulus on the unit circle, divided by N\sqrt{N}, can follow the constant 11.

Here closeness is taken in mean, for every finite exponent at once: is there a sequence of Littlewood polynomials, one of each length NN, whose normalized moduli tend to 11 in Lp(𝕋)L^p(\mathbb{T}) for every real p>0p>0? Beside this question stand the Littlewood polynomials of least maximum modulus on the circle, one chosen at each length, and the mean square of a Littlewood polynomial there.

Signings and Their Polynomials

A signing of length NN is a vector ε=(ε0,…,εN−1)\varepsilon=(\varepsilon_0,\dots,\varepsilon_{N-1}) of real numbers in which every entry is −1-1 or 11.✦

For a real coefficient vector ε\varepsilon of length NN, the polynomial PεP_\varepsilon is the function on ℂ\mathbb{C} given by Pε(z)=∑k<Nεkzk,P_\varepsilon(z)=\sum_{k<N}\varepsilon_k z^k,(1)✦ and when ε\varepsilon is a signing, PεP_\varepsilon is its Littlewood polynomial.

The trigonometric polynomial QεQ_\varepsilon of a real coefficient vector ε\varepsilon of length NN is the continuous function from 𝕋\mathbb{T} to ℂ\mathbb{C} given by Qε(t)=∑k<Nεke(kt),Q_\varepsilon(t)=\sum_{k<N}\varepsilon_k\,e(kt),(2)✦ so that Qε(t)=Pε(e(t))Q_\varepsilon(t)=P_\varepsilon(e(t)). The sup norm of the trigonometric polynomial QεQ_\varepsilon is ∥Qε∥∞=max⁡t∈𝕋|Qε(t)|\|Q_\varepsilon\|_\infty=\max_{t\in\mathbb{T}}|Q_\varepsilon(t)|, the maximum modulus of PεP_\varepsilon on the unit circle.

The circle 𝕋\mathbb{T} carries its Haar probability measure dtdt, and every integral over 𝕋\mathbb{T} below is taken against it.✦

The extremal signing εN*\varepsilon^\ast_N of length NN is a signing of length NN at which ∥Qε∥∞\|Q_\varepsilon\|_\infty attains its least value over the signings of length NN, one such signing being chosen for each NN.✦ The normalized modulus FNF_N of the extremal signing of length NN is the function on 𝕋\mathbb{T} given by FN(t)=|QεN*(t)|N.F_N(t)=\frac{|Q_{\varepsilon^\ast_N}(t)|}{\sqrt{N}}.(3)✦

The Claim

Theorem 1 (Corollary 7.1). (i) There is a family of signings ε(N)\varepsilon^{(N)}, one of each length NN, such that for every real p>0p>0 limN→∞∫𝕋||Pε(N)(e(t))|N−1|pdt=0.\lim_{N\to\infty}\int_{\mathbb{T}} \left|\frac{|P_{\varepsilon^{(N)}}(e(t))|}{\sqrt{N}}-1\right|^p dt=0.(4)✦ (ii) For every NN and every signing ε\varepsilon of length NN, ∥QεN*∥∞≤∥Qε∥∞.\|Q_{\varepsilon^\ast_N}\|_\infty\le\|Q_\varepsilon\|_\infty.(5)✦ (iii) Every signing ε\varepsilon of length NN satisfies ∫𝕋|Qε(t)|2dt=N.\int_{\mathbb{T}}|Q_\varepsilon(t)|^2\,dt=N.(6)✦ (iv) The normalized moduli of the extremal signings satisfy limN→∞∫𝕋(FN(t)−1)2dt=0.\lim_{N\to\infty}\int_{\mathbb{T}}\bigl(F_N(t)-1\bigr)^2\,dt=0.(7)✦

What Is Checked

The claim is stated over signings of length NN, the polynomial (1) at the points e(t)e(t) of the unit circle, the trigonometric polynomial (2) with its sup norm, the Haar probability measure on 𝕋\mathbb{T}, and the extremal signings εN*\varepsilon^\ast_N with their normalized moduli (3).