Wiki
Wiki

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

Updated


Statement

For a measurable 11-periodic set A⊆RA\subseteq\mathbb R let ρ(A)=m(A∩[0,1])\rho(A)=m(A\cap[0,1]), the measure of AA over one period; the manuscript at times treats such a set as a subset of the circle T=R/Z\mathbb T=\mathbb R/\mathbb Z. Let D={2−n:n≥1}D=\{2^{-n}:n\ge1\}.

Lemma 2.1 (Periodic hitting sets). For every p∈(0,1)p\in(0,1) there is an open 11-periodic set H⊆RH\subseteq\mathbb R with ρ(H)≤6p\rho(H)\le6p such that

∀x∈R  ∀t∈[1,2]  ∃n∈Z≥1:x+t2−n∈H.\forall x\in\mathbb R\ \ \forall t\in[1,2]\ \ \exists n\in\mathbb Z_{\ge1}: \qquad x+t2^{-n}\in H.

A single HH is uniform in both the center xx and the normalized scale tt; only the hitting index nn is allowed to vary with them.

Source. OpenAI, The dyadic case of the Erdős similarity conjecture, release folder preprints/The-dyadic-case-of-the-Erdos-similarity-conjecture-September-25-2026; TeX sections/periodic.tex, environment lem:periodic (lines 13--20), PDF p. 4; proof spread over sections/windows.tex, sections/routing.tex, sections/tests.tex, sections/scales.tex and sections/repair.tex (Sections 3--5, PDF pp. 4--14), completed in sections/repair.tex lines 13--130 (PDF pp. 13--14); read. The card records the provenance and attestations.

Read depth. Claims checked: the statement and the definition of ρ\rho were read clause by clause in the TeX source and located in the PDF, together with the statements of the intermediate results Lemma 3.1, Lemmas 4.1--4.3, Lemma 5.1 and Proposition 5.2. The proofs were read for their structure only (below); no step was checked. Nothing here is independently reviewed.

Proof pointer

The proof fixes pp and runs through five deterministic choices, a random construction, and a repair.

  • Section 3 fixes the combinatorics. Integers M≥2M\ge2 and d≥1d\ge1 with p(M−1)/2≥10p(M-1)/2\ge10 and (1−21−M)d<p(1-2^{1-M})^d<p give a complete ordered MM-ary tree of height dd with KK edges; a gap gg with K22−g<pK2^{2-g}<p; a threshold r∗r_* with 20Mr(1+22r+2)exp⁡(−p(M−1)r/2)<p20Mr(1+2^{2r+2})\exp(-p(M-1)r/2)<p for all r≥r∗r\ge r_*; and window lengths rhr_h by a bottom-up recursion so that a child block's whole index span is at most 2rh2r_h (display (5)). Each edge ee receives a block WeW_e of consecutive dyadic indices in edge preorder, starting at 33, with gg unused indices between blocks; N\mathcal N is their union. Periodic grid keys Jb(z)=⌊2b+2{z}⌋J_b(z)=\lfloor2^{b+2}\{z\}\rfloor are nested. A center xx is stable when no translate indexed by a later window crosses a boundary of the preceding window's grid; Lemma 3.1 shows the stable set GG has ρ(Gc)<p\rho(G^c)<p and that for x∈Gx\in G every translate x+t2−nx+t2^{-n}, n∈Wen\in W_e, t∈[1,2]t\in[1,2], keeps every earlier window's key.
  • Section 4 builds the random set. Each nondefault edge e=(P,i)e=(P,i), i<Mi<M, carries a fair random table SeS_e indexed by the keys at resolution beb_e (the window's last index); each leaf carries a Bernoulli-pp table TLT_L at its incoming window's resolution. A point is routed from the root by the first child whose selector reads 11, the last child by default, and BB is the set of points whose terminal entry reads 11; Eρ(B)=p\mathbb E\rho(B)=p. Lemma 4.1: a fixed center's route never takes a default child with probability (1−21−M)d<p(1-2^{1-M})^d<p. For a stable center whose route first defaults at node UU of height hh, Lemma 4.2 shows that the translates indexed by the window of child ii of UU repeat the center's route to UU and its rejections of children 1,…,i−11,\dots,i-1, so a local predicate QiQ_i (selector on (U,i)(U,i) times the terminal entry reached from UiU_i) equal to 11 forces the translate into BB. Lemma 4.3: at a fixed t∈[1,2]t\in[1,2], conditional on the center's exposed selector entries, the (M−1)rh(M-1)r_h tested selector and terminal addresses are pairwise distinct, so all local tests fail with probability exactly (1−p/2)(M−1)rh(1-p/2)^{(M-1)r_h}.
  • Section 5 passes to all scales and all centers. Lemma 5.1: the local predicates depend on the key at resolution bi∗b_i^* (the block's largest endpoint), so as tt runs over [1,2][1,2] each tested translate crosses at most 1+22rh+21+2^{2r_h+2} grid boundaries, and a set of at most 20Mrh(1+22rh+2)20Mr_h(1+2^{2r_h+2}) representative scales reproduces every pattern of predicate values. Proposition 5.2 combines Lemmas 4.1--4.3 and 5.1 by a union bound over representatives: P(∃t∈[1,2] ∀n∈N: x+t2−n∉B)≤2p\mathbb P(\exists t\in[1,2]\ \forall n\in\mathcal N:\ x+t2^{-n}\notin B)\le2p for every stable xx. Subsection 5.1 completes the lemma: enlarge each outcome's BωB_\omega to an open periodic Bω+B_\omega^+ adding at most pp to its density; the exceptional-center set RωR_\omega (centers with some t∈[1,2]t\in[1,2] missing Bω+B_\omega^+ at every n∈Nn\in\mathcal N) is closed and periodic, by compactness of T×[1,2]\mathbb T\times[1,2], with Eρ(Rω)≤2p+ρ(Gc)≤3p\mathbb E\rho(R_\omega)\le2p+\rho(G^c)\le3p; one outcome has ρ(B+)+ρ(R)≤5p\rho(B^+)+\rho(R)\le5p; an open periodic ε\varepsilon-neighborhood VV of RR has ρ(V)≤ρ(R)+p\rho(V)\le\rho(R)+p, and H=B+∪VH=B^+\cup V. A center outside RR is hit inside N\mathcal N; a center in RR is hit by every x+t2−nx+t2^{-n} with 21−n<ε2^{1-n}<\varepsilon, which lies in VV because x+t2−n→xx+t2^{-n}\to x.

The hypothesis p>0p>0 makes MM and dd exist; p<1p<1 keeps pp a probability for the terminal tables. The restriction t∈[1,2]t\in[1,2] is what makes the grid crossings per translate finite and is removed in Section 6 by dyadic rescaling.

Dependencies

None at statement level. The manuscript names Kolountzakis 1997 (random cells, scale discretization, open-cover repair), Chlebík 2015, Kolountzakis and Papageorgiou 2025 and the periodic blocking sets of Iosevich, Kulkarni, Mora Cuéllar, Rojas Aravena and Yavicoli 2026 as precedents for the method and says that all estimates needed are supplied in the text. None was checked here.

Bears on

  • Problem 120: the lemma is the input from which the manuscript deduces the claimed dyadic case of the question (Theorem 1.1); on its own it says nothing about the problem, and the claim is unverified here. The page's status rests on its acceptance evidence.
  • [[analysis/openai_2026_geometric_case_erdos_similarity_conjecture/proposition_2_1|The companion's Proposition 2.1]]: the geometric-case manuscript's periodic hitting statement for {qn}\{q^n\} with general q∈(0,1)q\in(0,1) plays the same role there as this lemma does here; comparison only, neither verified here.