An Explicit Removal Estimate for Reciprocal-Sum-Free Sets

Primarily AI-generated textHuman understanding: no partsmath.NT — Number Theory

Contributed by Will Howe, Grant Huffman, Jensen Kohlmeyer, Liam Kruer ↗

Version 1 / Nov 21, 2026 / CC BY 4.0

Abstract

Let f(N)f(N) be the largest cardinality of a subset of {1,…,N}\{1,\ldots,N\} containing no three distinct elements with 1/a=1/b+1/c1/a=1/b+1/c. We give an explicit uniform removal estimate for finite-prime relaxations of this problem. Together with the smooth-number decomposition, the estimate yields f(N)=CN+o(N)f(N)=CN+o(N) and an effective description of CC by finite optimization problems. If UVU_V is the limiting density obtained by retaining triples with smooth primitive coordinates over the primes at most 2V2V, then 0≤UV−C≤(1023+561000)(log⁡V)−1/1400\le U_V-C\le (10^{23}+561000)(\log V)^{-1/140} for V≥4V\ge4. The proof uses elementary prime estimates, a finite quadratic sieve, and a dyadic decomposition indexed by the larger primitive parameter. We also give rational enclosures of arbitrary prescribed width and an explicit modulus of convergence for f(N)/Nf(N)/N. The finite-prime framework and the underlying sieve are due to the earlier density-limit argument. The emphasis here is the uniform quantitative estimate.

Author comments: Manuscript accompanying Justin Sun Prize submission JSP-000254. Pinned source and Lean proof materials: https://github.com/Ghuffy1/erdos302-jsp-submission/tree/1af724ea53a44310c5386062baa22849e6eddafd

Provenance statement

OpenAI ChatGPT assisted with argument development, automated checking and exposition. The Lean audit covers the main density, removal and enclosure results. Two supplementary convergence estimates are proved in the manuscript.

References

  1. T. F. Bloom, Erdős problem 302 , https://www.erdosproblems.com/302, including the construction attributed to Stijn Cambie.link
  2. D. Della Pietra, A positive-density improvement for all-length unit-fraction-free sets , manuscript, July 30, 2026, version of July 30, 2026 (https://github.com/donalddellapietra/erdos-301-proof/blob/789c6f045dbc81da3811031247d186a7128dafce/paper/main.tex).link
  3. P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory , Monographies de L'Enseignement Mathématique, vol. 28, Université de Genève, Geneva, 1980, p. 37.
  4. E. H. Jiang, J. Zhang, J. Yu, J. Lu, Z. Zhang, M. Simon, et al., Continual Graph Memory and Community Infrastructure for AI-Assisted Mathematical Discovery , arXiv:2610.02945 (https://arxiv.org/abs/2610.02945), 2026. See also the companion manuscript Erdős Problem #302: A Consolidated Candidate Proof of the Density Limit , version of September 26, 2026 (https://github.com/uclanlp/continual-graph-memory/blob/23367c6568c344d9c4dbf59be3f39b39f5d54324/results/erdos_302.md), Sections 2--6 and Appendix B.arXiv

Version history

  1. v1 · 12 pagesSubmitted by Liam KruerManuscript accompanying Justin Sun Prize submission JSP-000254. Pinned source and Lean proof materials: https://github.com/Ghuffy1/erdos302-jsp-submission/tree/1af724ea53a44310c5386062baa22849e6eddafdCurrentNov 20, 2026