Nonmaximal sums of maximally monotone operators under Rockafellar's constraint qualification
cs.LG · math.FA
Quick verdict
Claims first disproof of Rockafellar's sum conjecture via explicit c0 and ℓ1 counterexamples; internally consistent, needs independent verification.
Summary
This preprint claims the first complete disproof of Rockafellar's sum conjecture — the 1970 question of whether the sum of two maximally monotone operators on a general Banach space must be maximally monotone whenever dom A ∩ int(dom B) ≠ ∅. The author establishes a general construction theorem (Theorem 1): for graphs G = {(La, a)} defined by a bounded linear map M: X* → H and a Lipschitz curve constraint Ma = F(r(a)), the entire monotone polar is computed explicitly (Eq. 8), maximality reduces to unsolvability of r(p) = τ, Mp = (τ, η) on [−1,1] (Eq. 9), and adding the everywhere-defined positive rank-one operator P = ⟨·,g⟩g produces a nonmaximal sum, witnessed by (0,0) lying in the polar but off the graph. On c0(I), block-sum and triangular maps (Definitions 6–8) verify every hypothesis; the explicit curve ωs(n) = max{1 − sn^{1/4}, 0}/4n yields S = π²/96 < 1/8 and bounds V(gra A) ≤ 2σ < 1/4 and ⟨x,b⟩ > 7/8 on gra(A+P). An explicitly constructed bounded surjection Q: ℓ1 → c0 with ‖Q‖ = 1 and lifting bound 2‖x‖∞ (Lemma 5) transfers the counterexample to standard ℓ1 (Theorem 3) via a pullback lemma (Lemma 1). A Lean formalization of the c0 instance and pullback lemma is linked. Key limitations: the detailed proofs are disclosed as AI-generated (GPT-5.6), the Lean artifact's coverage of Theorem 1's generality is unstated, and a claim of this magnitude requires independent expert verification.
Review complete
Assessment at a glance
4
risk flags identified
● high 1
● medium 2
● low 1
5 strengths · 4 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- Two maximally monotone operators on c0(I) satisfy dom A ∩ int(dom P) ≠ ∅ (dom P = c0) yet A + P is not maximally monotone, disproving Rockafellar's sum conjecture.
- Theorem 1 computes the entire monotone polar of the constructed graph (Eq. 8) and reduces maximal monotonicity to unsolvability of r(p) = tau, Mp = (tau, eta) on [-1,1] (Eq. 9).
- The same counterexample transfers to standard l1 through an explicitly constructed bounded surjection Q: l1 → c0 and a pullback lemma for maximal monotonicity (Lemma 1).
- Both counterexamples carry quantitative certificates: V(gra A) ≤ 2σ < 1/4 with σ = π²/96, and <x,b> > 7/8 for all (x,b) in gra(A+P).
- A Lean formalization of the c0 counterexample and the pullback lemma is publicly available.
Strengths
Exact monotone-polar computation
Theorem 1(i) computes the entire polar G^mu in closed form (Eq. 8) and derives a unique maximally monotone extension; I independently re-verified Eqs. (10)-(13) and the two-case analysis (|tau|>1 vs |tau|<=1) without finding gaps.
Fully concrete c0 instance
Definitions 6-8 use elementary block-sum and triangular maps; Lemma 4 gives explicit constants sigma = pi^2/96 ≈ 0.103 < 1/8 and Lipschitz constant c < 3/16; the witness a2 = -2e_{0,1} + 3e_{0,3} yields La2 = -6e_{0,1} - 8e_{0,2} - 3e_{0,3} exactly as claimed in the proof of Theorem 2.
Quantitative failure certificates
Transfer machinery without a bounded right inverse
Machine-checkable artifact and transparent AI disclosure
Weaknesses
Extraordinary claim, externally unvalidated
Core proofs AI-generated
Lean coverage ambiguity
Positioning and exposition gaps
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
The preprint (arXiv:2609.10487v3, filed primarily under cs.LG with math.FA secondary) claims the first complete disproof of Rockafellar's 1970 sum conjecture: on a general Banach space, must A+B be maximally monotone when A, B are maximally monotone and dom A ∩ int(dom B) ≠ ∅? Three results carry the claim. Theorem 1 is a general construction: for the graph G = {(La, a): a ∈ D} of Definition 5 — built from a bounded linear M: X* → H, a scalar functional r(a) = ⟨h,a⟩, and the 1-Lipschitz curve constraint Ma = F(r(a)) — it computes the entire monotone polar (Eq. 8), characterizes maximality by unsolvability of r(p) = τ, Mp = (τ,η) on [−1,1] (Eq. 9), and shows the sum with the everywhere-defined positive rank-one operator P = ⟨·,g⟩g is nonmaximal, since (0,0) ∈ gra(A+P)^μ but (0,0) ∉ gra(A+P). Theorem 2 instantiates this on c0(I), I = N0×N; Theorem 3 transfers it to standard ℓ1 via an explicit bounded surjection. A Lean formalization of the c0 counterexample and the pullback lemma is linked on GitHub.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 5 strengths & 4 weaknesses, analyzed in detail
- 4 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What is Rockafellar's sum conjecture?
A 1970 question in monotone operator theory: on a general (possibly nonreflexive) Banach space, is the pointwise sum of two maximally monotone operators A and B necessarily maximally monotone whenever the domain of A intersects the interior of the domain of B? It is settled (yes) in reflexive spaces and remained open for over five decades; this paper claims the first counterexample, on c0 and l1.
Has Rockafellar's sum conjecture been disproved?
This arXiv preprint (2609.10487) claims a complete disproof with explicit counterexamples on c0 and l1, partially supported by a Lean formalization. As of this review the arguments check out on internal recomputation, but the claim has not yet been independently validated by the research community.
Did AI write this paper?
Partly, by the author's own disclosure: the Use-of-AI statement says the author supplied prior results, constructions and direction, while the model GPT-5.6 Sol worked out the detailed constructions and proofs, including the transfer from c0 to l1.
Where is the Lean code for the counterexample?
The paper links github.com/Weifeng-Yang/RockafellarSumConjecture, said to formalize the c0 counterexample and the general pullback lemma; the paper itself does not specify whether Theorem 1's full generality is covered.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- PDF extraction flattens display mathematics (e.g., F-bar and G-hat render as 'bF' and 'eG'; subscripts/superscripts are linearized). All cited formulas were reconstructed from context and re-verified, but a single mis-extracted symbol in Definitions 6-8 could in principle evade this check.
- The provided text runs complete from abstract through reference [14]; no truncation is claimed. The linked GitHub Lean repository was not inspectable within this review, so artifact-scope statements rest on the paper's own description.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.
