/- Copyright 2026 The Formal Conjectures Authors. Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at https://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. -/ module public import FormalConjecturesUtil

Green's Open Problem 27

References:

    [Gr24] Green, Ben. "100 open problems." (2024).

    [Be23] Bedert, Benjamin. "On unique sums in Abelian groups." Combinatorica 44.2 (2024): 269-298.

    [CY26] Cao, Jiao-Long, and Ye Yuan. "A Second-Logarithm Lower Bound for Sets with No Unique Sums." arXiv preprint arXiv:2608.06728 (2026).

    [He26] He, Xinjie. "A near-quadratic lower bound for sets with no unique sums." Working draft (July 21, 2026): https://github.com/xinjiegit/ben_27_formalization/blob/f70352964007662e97a810c880abbfea1e2ea262/NUS_edited.tex. Its Lean 4 proof is registered as PALOMAR-2026-08-18-000003.

    [HY26] Hou, Jianfeng, and Kai Yang. "A near-quadratic lower bound for sets with no unique sums." arXiv preprint arXiv:2610.09349 (2026).

    [St76] Straus, E. G. "Differences of residues (mod p)." Journal of Number Theory 8.1 (1976): 40-42.

@[expose] public sectionopen Asymptotics Filternamespace Green27

This is $m(p)$ in [Be23]: the size of the smallest set $A \subset \mathbb{Z} / p\mathbb{Z}$ (with at least two elements) for which no element in the sumset $A + A$ has a unique representation.

noncomputable def m (p : ℕ) : ℝ := (sInf { (A.card) | (A : Finset (ZMod p)) (_ : 2 ≤ A.card) (_ : HasNoUniqueRepresentation A) } : ℝ)

atTop restricted to prime numbers.

def primesAtTop : Filter ℕ := atTop ⊓ 𝓟 {p : ℕ | p.Prime}

Bedert's lower bound [Be23, Theorem 3]. This was the best known bound when Green's list was written; it has since been improved, see green_27.lower.

noncomputable def lowerBest (p : ℕ) : ℝ := (Real.sqrt (Real.log (Real.log (Real.log (p : ℝ)))) / Real.log (Real.log (Real.log (Real.log (p : ℝ))))) * Real.log (p : ℝ)

Best-known upper bound [Be23, Theorem 5].

noncomputable def upperBest (p : ℕ) : ℝ := (Real.log (p : ℝ)) ^ 2def green_27.equivalent.answer : ℕ → ℝ := sorry

What is the size of the smallest set $A \subset \mathbb{Z} / p\mathbb{Z}$ (with at least two elements) for which no element in the sumset $A + A$ has a unique representation?

@[category research open, question value, AMS 5 11] theorem green_27.equivalent : green_27.equivalent.answer ~[primesAtTop] m := ⊢ equivalent.answer ~[primesAtTop] m All goals completed! 🐙noncomputable def green_27.lower.answer : ℕ → ℝ := fun p ↦ Real.log p * Real.log (Real.log p)

Propose a better lower bound along primes.

Cao and Yuan [CY26, Theorem 1.1] prove $m(p) \gg \log p \log \log p$. This beats lowerBest, since $\sqrt{\log \log \log p} / \log \log \log \log p = o(\log \log p)$.

Xinjie He [He26, Theorem 1.2] and Hou and Yang [HY26, Theorem 1.1] prove the stronger bound $m(p) \gg (\log p / \log \log p)^2$. He's proof is formalised in Lean, see green_27.variants.lower_he26_hy26. Together with green_27.variants.upper_be23, this lower bound gives $m(p) = (\log p)^{2 + o(1)}$. The exact order of $m(p)$ remains open.

@[category research solved, question value, AMS 5 11] theorem green_27.lower : (lowerBest =o[primesAtTop] green_27.lower.answer) ∧ (green_27.lower.answer =O[primesAtTop] m) := ⊢ lowerBest =o[primesAtTop] lower.answer ∧ lower.answer =O[primesAtTop] m All goals completed! 🐙def green_27.upper.answer : ℕ → ℝ := sorry

Propose a better upper bound along primes.

@[category research open, question value, AMS 5 11] theorem green_27.upper : (green_27.upper.answer =o[primesAtTop] upperBest) ∧ (m =O[primesAtTop] green_27.upper.answer) := ⊢ upper.answer =o[primesAtTop] upperBest ∧ m =O[primesAtTop] upper.answer All goals completed! 🐙

We have $m(p) \geq \omega(p) \log p$ for some function $\omega(p)$ tending to infinity [Be23, Theorem 3].

@[category research solved, AMS 5 11] theorem green_27.variants.lower_be23 : ∃ ω : ℕ → ℝ, Tendsto ω primesAtTop atTop ∧ ∀ᶠ p in primesAtTop, ω p * Real.log (p : ℝ) ≤ m p := ⊢ ∃ ω, Tendsto ω primesAtTop atTop ∧ ∀ᶠ (p : ℕ) in primesAtTop, ω p * Real.log ↑p ≤ m p All goals completed! 🐙

Lower bound: $\log p \log \log p \ll m(p)$ [CY26, Theorem 1.1].

@[category research solved, AMS 5 11] theorem green_27.variants.lower_cy26 : (fun p ↦ Real.log (p : ℝ) * Real.log (Real.log (p : ℝ))) =O[primesAtTop] m := ⊢ (fun p ↦ Real.log ↑p * Real.log (Real.log ↑p)) =O[primesAtTop] m All goals completed! 🐙

Lower bound: $(\log p / \log \log p)^2 \ll m(p)$ [He26, Theorem 1.2], [HY26, Theorem 1.1]. The linked formal proof is He's.

@[category research solved, AMS 5 11, formal_proof using lean4 at "https://palomar-registry.org/entry.html?id=PALOMAR-2026-08-18-000003"] theorem green_27.variants.lower_he26_hy26 : ∃ c > (0 : ℝ), ∀ᶠ p : ℕ in primesAtTop, c * (Real.log (p : ℝ) / Real.log (Real.log (p : ℝ))) ^ 2 ≤ m p := ⊢ ∃ c > 0, ∀ᶠ (p : ℕ) in primesAtTop, c * (Real.log ↑p / Real.log (Real.log ↑p)) ^ 2 ≤ m p All goals completed! 🐙set_option linter.style.formal_proof_palomar false in

Upper bound: $m(p) \ll (\log p)^2$ [Be23, Theorem 5].

@[category research solved, AMS 5 11, formal_proof using lean4 at "https://github.com/mkwatson/unique-sums-notes/blob/d4383f93a52e3a916ccce78d5fb7739c3bf3dabc/Green27Upper/Theorem.lean#L26"] theorem green_27.variants.upper_be23 : m =O[primesAtTop] upperBest := ⊢ m =O[primesAtTop] upperBest All goals completed! 🐙

Previous best-known lower bound $\log p \ll m(p)$ from [St76].

@[category research solved, AMS 5 11] theorem green_27.variants.previous_lower : (fun p ↦ Real.log (p : ℝ)) =O[primesAtTop] m := ⊢ (fun p ↦ Real.log ↑p) =O[primesAtTop] m All goals completed! 🐙

Previous best-known upper bound $m(p) \ll \sqrt{p}$ from [Be23].

@[category research solved, AMS 5 11] theorem green_27.variants.previous_upper : m =O[primesAtTop] (fun p ↦ Real.sqrt (p : ℝ)) := ⊢ m =O[primesAtTop] fun p ↦ √↑p All goals completed! 🐙end Green27