VibeMathedMath problems solved with AI

The coercivity conjecture for nonlocal energies (Imbert-Silvestre)

For a measurable kernel K≥0K\ge0 on Rn×Rn\mathbb{R}^n\times\mathbb{R}^n and 0<s<10<s<1, let EK(u)=∬∣u(x)−u(z)∣2K(x,z) dz dx\mathcal{E}_K(u)=\iint|u(x)-u(z)|^2K(x,z)\,dz\,dx. Assume the symmetry K(x,z)=K(z,x)K(x,z)=K(z,x), the upper bound r2s∫B2r(x)∖Br(x)K(x,z) dz≤Λr^{2s}\int_{B_{2r}(x)\setminus B_r(x)}K(x,z)\,dz\le\Lambda and the directional lower bound r2s−2inf⁡∣e∣=1∫Br(x)(e⋅(z−x))+2K(x,z) dz≥λr^{2s-2}\inf_{|e|=1}\int_{B_r(x)}\big(e\cdot(z-x)\big)_+^2K(x,z)\,dz\ge\lambda, for all r>0r>0 and almost every xx. Does it follow that [u]Hs(Rn)2≤C EK(u)[u]_{H^s(\mathbb{R}^n)}^2\le C\,\mathcal{E}_K(u) for every u∈Hs(Rn)u\in H^s(\mathbb{R}^n), with C=C(n,s,λ,Λ)C=C(n,s,\lambda,\Lambda)? Here [u]Hs2=∬∣u(x)−u(z)∣2∣x−z∣−n−2s dz dx[u]_{H^s}^2=\iint|u(x)-u(z)|^2|x-z|^{-n-2s}\,dz\,dx. Imbert and Silvestre, working on the Boltzmann equation without cut-off, wrote that they knew no kernel satisfying the directional lower bound without coercivity and that it is natural to conjecture the implication; Chaker and Silvestre restated it as a conjecture in 2020 and proved coercivity under a stronger density assumption.

Result
Disproved(see note)
Status
Candidate (review pending)
AI contribution
AI co-developed
Method
Construction
Field
Nonlocal elliptic equations; fractional Sobolev spaces
Posed by
Cyril Imbert and Luis Silvestre (weak Harnack inequality for the Boltzmann equation without cut-off, arXiv 2016, JEMS 2020); restated by Chaker and Silvestre (2020)
Year posed
2016
Years open
10y
Solved
2026-11-12
Model
ChatGPT versions 5.5, 5.6 and 6
Vendor
OpenAI
Collaborators
Linus Behn, Roberto Colombo, Xavier Fernández-Real, Florian Grube, Harsh Prasad, Xavier Ros-Oton
Verification
Unreviewed
Publication
Preprint
Significance
18 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Theorem 1.1 of arXiv:2611.16682 (44 pages): for every n≥1n\ge1 and s∈(0,1)s\in(0,1) with n>2sn>2s there are λ,Λ>0\lambda,\Lambda>0 and a kernel KK satisfying all three hypotheses with inf⁡0≠u∈Cc∞EK(u)/[u]Hs2=0\inf_{0\ne u\in C_c^\infty}\mathcal{E}_K(u)/[u]_{H^s}^2=0. Kernels smooth away from the diagonal admit no uniform constant either, and a function w∈Hlocsw\in H^s_{\mathrm{loc}} with 0≤w≤10\le w\le1 and EK(w)=0\mathcal{E}_K(w)=0 has oscillation 1 on every ball, so the associated equation has bounded discontinuous weak solutions. This covers every ss when n≥2n\ge2, and s<1/2s<1/2 when n=1n=1. Theorem 1.15: for n=1n=1 and 1/2<s<11/2<s<1 the conjecture is true. The case n=1n=1, s=1/2s=1/2 is covered by neither theorem. Positive results: coercivity holds under an annular cancellation condition (Theorems 1.3 and 1.4), for kernels comparable to variable stable kernels (Theorem 1.8) and for non-cutoff Boltzmann kernels under a nonconcentration assumption (Theorem 1.12).

What the AI did

From Section 1.9 (Use of AI): the project began with a counterexample the authors constructed for small ss in dimension one, with its extension to higher dimensions following Stein's construction, and with their own heuristic for coercivity under double symmetry. It then developed over several months through discussions among the authors and sustained interaction with ChatGPT versions 5.5, 5.6 and 6. The authors formulated the questions, proposed approaches and assessed the suggestions; the models served as sounding boards and contributed mathematical ideas, technical extensions, refinements of proofs and help with exposition. All arguments kept in the paper were independently checked and rewritten by the authors.

Verification

No independent mathematician has checked this yet. Read by this site on 19 November 2026: arXiv 2611.16682v1 (Sections 1.1, 1.2 and 1.7 to 1.9, Theorems 1.1 and 1.15, the AI statement). Checked against the originals: Imbert and Silvestre (arXiv 1608.07571, introduction) write that they know no kernel satisfying their nondegeneracy condition (1.4) but not the coercivity condition (1.3), and that it is natural to conjecture the implication; Chaker and Silvestre (arXiv 1904.13014) restate it as a conjecture. Their setting is local and kinetic, with a lower-order term; the global symmetric statement is the paper's own formulation. Biswas (arXiv 2610.06408) was checked for priority: by its abstract it proves the conjecture under an extra LpL^p bound and does not disprove it. The proof, including the counterexample constructions of Sections 2 and 3, was not refereed here, and there is no formal proof.

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion