\documentclass[11pt]{amsart} \usepackage[letterpaper,margin=1in,headheight=14pt]{geometry} \usepackage{amsmath,amssymb} \usepackage[hidelinks]{hyperref} \newtheorem{theorem}{Theorem}[section] \numberwithin{equation}{section} \newcommand{\N}{\mathbb N} \newcommand{\Z}{\mathbb Z} \newcommand{\paperroot}{https://github.com/Ghuffy1/erdos-335/blob/29e64ce67b4999f76657bb4b51e306076f59e689/paper} \title[A periodic-summand characterization for Erd\H{o}s problem 335]{A periodic-summand characterization for Erd\H{o}s problem 335} \author{Will Howe} \author{Grant Huffman} \author{Jensen Kohlmeyer} \author{Liam Kruer} \date{November 2026} \keywords{sumsets, asymptotic density, periodicity, inverse theorems} \begin{document} \begin{abstract} Erd\H{o}s and Graham asked which pairs of positive-density sets of positive integers have a sumset whose ordinary density equals the sum of their densities. The formalization gives an unrestricted characterization. A numerical boundary case is controlled by literal modular supports. In the strict-gap case, equality forces parallel Bohr intervals with finite backgrounds, two periodic templates with a literal fringe test, or one sparse exceptional section with an exact terminal condition. This overview presents the statement and proof structure. The accompanying mathematical notes give the detailed arguments, and the Lean project contains the complete formal proof. \end{abstract} \maketitle \section{Introduction} Throughout, \(A\) and \(B\) are subsets of the positive integers, and both ordinary densities exist and are positive. All sumsets use the actual elements of the sets. For a positive integer \(q\), let \(R_q(A)\) be the set of residues represented by \(A\). Put \begin{align} d(A)&=\lim_{N\to\infty}\frac{|A\cap[1,N]|}{N},& s&=d(A)+d(B),\\ R_q(A)&=\{a\bmod q:a\in A\},& \gamma&=\inf_{q\ge1}\frac{|R_q(A)+R_q(B)|}{q}. \end{align} \begin{theorem}[Classification] The equality under study is \begin{equation} d(A+B)=d(A)+d(B). \end{equation} Equality always holds at the modular boundary: \begin{equation} \gamma=s\quad\Longrightarrow\quad d(A+B)=s. \end{equation} If \(\gamma0\) and \(V+g\sim_d V\). For finite \(U\) this is the remaining terminal condition. For infinite \(U\), the literal sums of \(D/g\) with each rescaled \(g\)-residue section of \(V\) must satisfy the escape-packing test outside that section. \section{The literal interval-packing test}\label{sec:packing} The terminal tests operate on runs of consecutive block indices. They keep initial elements, singleton runs and sparse sections. A residue section and the carry in a sum are given by \begin{equation} E_{q,r}=\{k\ge0:qk+r\in E\},\qquad \kappa(r,t)=\left\lfloor\frac{r+t}{q}\right\rfloor, \qquad 0\le r,t0,\ b_t(y)>0. \end{equation} Two times with the same label have that label at their sum, almost everywhere. A bounded positive-measure piece of a label has a self-convolution positive on an interval. Repeated addition gives a tail. Disjoint labels therefore cannot both have positive measure, so a single label holds almost everywhere on the positive axis. The normalized prefix-profile sequence has vanishing successive increments. Its cluster points lie in finitely many closed slices, and the label rules separate those slices. The sequence therefore selects one slice. One input coordinate converges to a fixed periodic mask, giving an ordinary periodic-in-density summand. \paragraph{Exact reduction and sufficiency.} Once one summand is periodic in density, use its mask and the other input's actual residue support to form the fixed-translate baseline. If that baseline fills the output density, both inputs have periodic templates. If the baseline has smaller density, the reduction gives the finite background frame with one null section. Finite sparse sets use gcd-shift invariance and finite unions. Infinite sparse sets use the escape test to give \(d((D+V)\setminus V)=0\). The Bohr alternative uses literal arc containment, equidistribution and the Kneser lower bound. Recombining the background and active sections proves the converse in every alternative. The \href{\paperroot/README.md}{mathematical source package} provides the reading order and the detailed supporting arguments. The \href{\paperroot/notes/FULL_CANDIDATE_PERIODIC_FORM_20261111T0010Z.md}{classification proof} links its local inverse arguments and exact reductions. The \href{\paperroot/gluing-details.md}{expanded gluing proof} gives the common scale selection, comparison of windows and passage to one ordinary periodic summand. \section{Formal verification and provenance} The complete formal result is the closed theorem \begin{quote} \texttt{Erdos335.Proposed.full\_periodic\_classification}. \end{quote} Its type is \texttt{PeriodicRefinedSolution}. The theorem \begin{quote} \small\texttt{Erdos335.Proposed.full\_periodic\_classification\_original\_positive\_counts} \end{quote} writes the same classification using the original counts on \(\{1,\ldots,N\}\). \begin{center} \begin{tabular}{@{}ll@{}} \hline Check&Recorded result\\ \hline New supporting modules&45 compiled\\ Exact declaration and axiom audit&2,602 targets passed\\ Allowed proof axioms&\texttt{propext}, \texttt{Classical.choice}, \texttt{Quot.sound}\\ Fresh full-closure kernel replay&Passed in 1,460.406 seconds\\ Project object files bound&890\\ Source, object and runtime hashes&Unchanged across replay\\ \hline \end{tabular} \end{center} The runtime is Lean 4.34.0, with Mathlib revision \texttt{5ed2965256430c3649e86755f9576b54eca72435}. The checked proof commit is \texttt{63e4377a3f62cdf4a989b2559d8f9c7341c78789}. The checker ran \texttt{leanchecker --fresh -v Challenge}. The \href{https://github.com/Ghuffy1/erdos-335/tree/29e64ce67b4999f76657bb4b51e306076f59e689/verification}{verification record} includes the proof revisions, logs and reproduction instructions. The \href{\paperroot/erdos335-overview.pdf}{original overview} accompanies \href{https://github.com/TheJustinSunPrize/awards/pull/5574}{Justin Sun Prize submission JSP-000278}. The prize submission credits OpenAI Codex assistance in the Lean formalization. The original overview reports that its literature search found no earlier matching full characterization in the public record it checked. Publication priority and Justin Sun Prize acceptance are submitted for independent review. \begin{thebibliography}{4} \bibitem{erdos} P.~Erd\H{o}s and R.~L.~Graham, \emph{Old and New Problems and Results in Combinatorial Number Theory}, 1980, pp.~50--51. \bibitem{kneser} M.~Kneser, Absch\"atzung der asymptotischen Dichte von Summenmengen, \emph{Mathematische Zeitschrift} 58 (1953), 459--484. \bibitem{ar} E.~Ackelsberg and F.~K.~Richter, An inverse theorem for sumsets of sets of positive density in the integers, \href{https://arxiv.org/abs/2604.12864}{arXiv:2604.12864}, 2026. \bibitem{griesmer} J.~T.~Griesmer, Compact and Loeb inverse-theorem inputs, \href{https://arxiv.org/abs/1112.6403}{arXiv:1112.6403} and \href{https://arxiv.org/abs/1807.01694}{arXiv:1807.01694}. The mathematical source notes identify the exact propositions. \end{thebibliography} \end{document}