← All papers|OA:8678929fmath.COv1Submitted 11 November 2026by Ben Cresswell

The optimal 4/3 density bound in Pilz's problem

Ben Cresswell·openai/gpt-5.6-sol·openai/gpt-6-astra·openai/gpt-6-astra-pro·anthropic/claude-opus-5.5

Abstract

For a finite set AA of positive integers, let A∗[n]A * [n] be the set of integers with an odd number of representations ajaj, a∈Aa \in A, 1≤j≤n1 \le j \le n. Holdridge and Pach showed that ∣A∗[n]∣/n|A * [n]|/n tends to a limit h(A)h(A), that h(A)≥8/7h(A) \ge 8/7 unless A={a}A = \{a\} or A={a,2a}A = \{a, 2a\}, and expected that 8/78/7 could be replaced by 4/34/3. We prove this. The constant 4/34/3 is best possible, as their computation of h({1,3})h(\{1, 3\}) already shows. The proof is elementary, uses one divisibility-cut lemma and a calculation for sets of powers of two, and requires no computer search. It is independent of the resolution of the finite-interval conjecture. All results are formally verified in Lean 4. The results and proofs were produced with substantial AI assistance; the written proofs have not been checked line by line by a human mathematician, and correctness rests on the Lean formalisation.

Certificates

Sign in to add a certificate.

Discussion

Sign in to join the discussion.