Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Statement

Conventions as on the Theorem 7 page.

Proposition 16 (pp. 10--11). Let p>0p>0 and let f(x,y)=∑k=1Ke(nkx+mky)f(x,y)=\sum_{k=1}^Ke(n_kx+m_ky) as in (16), with K∈NK\in\mathbb N, the nkn_k and mkm_k nonnegative integers and the mkm_k strictly increasing. Suppose its marginal pp-integral F(x)=∫01∣f(x,y)∣p dyF(x)=\int_0^1|f(x,y)|^p\,dy of (17) has a strict maximum at aa, where a=0a=0 or a=1/2a=1/2. Then there is full p-concentration with gap at aa.

The paper notes (Remark 18, p. 12) that ∣f(−x,−y)∣=∣f(x,y)∣|f(-x,-y)|=|f(x,y)|, so FF is even, and a unique maximum on T\mathbb T can occur only at 0 or 1/21/2. It credits the use of bivariate idempotents and Riesz products to a suggestion of Terence Tao (p. 8).

Source. Aline Bonami and Szilárd Gy. Révész, Integral concentration of idempotent trigonometric polynomials with gaps, arXiv:0707.3023v2 (16 October 2008): Proposition 16 on pp. 10--11, proof pp. 11--12; Lemma 17 on p. 11. The edition is the one identified on the source card.

Read depth. Claims checked: the statement and Lemma 17 were read clause by clause on the printed pages. The proof was read but not checked step by step.

Proof pointer

Take MM above all nk,mkn_k,m_k and the Riesz product gR,J(x)=∏j=1Jf(x,Rjx)g_{R,J}(x)=\prod_{j=1}^Jf(x,R^jx) of (18). For R>M(J+1)R>M(J+1) it is an idempotent, and its gaps exceed any given NN once RR is large in terms of JJ, MM and NN. First fix JJ so large that FJF^J puts all but a fraction ε\varepsilon of its integral on [a−δ,a+δ][a-\delta,a+\delta], which the strict maximum allows. Then Lemma 17 (p. 11), an equidistribution statement proved from the Riemann--Lebesgue lemma, shows that the integral of ∣gR,J∣p|g_{R,J}|^p over any fixed interval tends to that of FJF^J as R→∞R\to\infty.

Dependencies

Lemma 17 of the paper.

Bears on

No Erdős problem page directly.