VibeMathedMath problems solved with AI

The number of planar central configurations of four bodies: is it always between 32 and 50?

Four bodies with masses m1,…,m4>0m_1,\dots,m_4>0 in the plane form a central configuration if the gravitational acceleration of each body is proportional to its displacement from the centre of mass; these are the shapes of the relative equilibria of the Newtonian four-body problem. Let N(m)N(m) be the number of central configurations up to rotation, translation and scaling, mirror images counted separately. Hampton and Moeckel (2006) proved that N(m)N(m) is finite for all positive masses, which settled Smale's sixth problem for four bodies, with 32≤N(m)≤847232\le N(m)\le 8472. Simó's numerical experiments (1978) suggest 32≤N(m)≤5032\le N(m)\le 50, and Figueras, Tucker and Zgliczyński (2024) named the unrestricted four-body count as their next challenge, expecting between 32 and 50. What is N(m)N(m) for each mass vector, and is it always between 32 and 50?

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI co-developed
Method
Computation
Field
Celestial mechanics; central configurations
Posed by
Simó (1978), numerical study of four-body relative equilibria; proved bounds 32 to 8472 by Hampton and Moeckel (2006)
Year posed
1978
Years open
48y
Solved
2026-10-10
Model
Generative AI tools (unspecified)
Vendor
—
Collaborators
Tejasvi Singh Tomar
Verification
Lean-checked, statement unaudited
Publication
Preprint
Significance
28 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Theorems 1 and 2: for all positive masses, N(m)=12+6+2(c1+c2+c3+c4)N(m)=12+6+2(c_1+c_2+c_3+c_4), where 12 counts the collinear configurations (Moulton), 6 the convex ones, and ci(m)∈{1,2,3,4}c_i(m)\in\{1,2,3,4\} the concave ones with body ii inside, read off from the position of the mass ratios relative to one surface of critical values. Hence N(m)N(m) is even with 32≤N(m)≤5032\le N(m)\le 50, every even value occurs, N=32N=32 exactly at Palmore's masses (1,1,1,m∗)(1,1,1,m^*) with m∗=(81+643)/249m^*=(81+64\sqrt3)/249, and N(1,1,1,1)=50N(1,1,1,1)=50. Caveats: the surface has no closed form, so NN is decided by certified enclosures only for masses off a closed null set; and the convex count 6, hence the upper bound 50, rests on the author's own unrefereed preprint on convex uniqueness (arXiv:2609.35632), whose formalization is part of the Lean development. This is not a new case of Smale's sixth problem: finiteness for four bodies was already known.

What the AI did

The paper's section Use of AI tools, in full: "Generative AI tools were used to find and prove the results, to write the programs and the Lean formalization, and to draft the text." No model or vendor is named. Hexagon labels the text a mix of human-written and AI-generated, with human understanding of all parts.

Verification

No independent mathematician has checked this yet. Lean-checked on this site's ladder, not Lean-verified. Checked here on 19 November 2026: the Hexagon record and 77-page PDF (hexagon:2611.00378v1) and the archive cc50-lean1.zip of the Zenodo supplement (Lean 4.35.0-rc3 with Mathlib; 1,588 Lean files, about 404,000 lines). A grep of those files for sorry, axiom, native_decide, implemented_by, extern and ofReduceBool finds only a message string and two tooling scripts. The headline CC50.Final.main, the conjunction of Theorem1 and Theorem2, was read against the paper: Ncc is the extended-natural cardinality of central configurations modulo orientation-preserving similarity, cConc counts concave ones with body ii inside modulo similarity, and Theorem2a states that Ncc lies in {32,34,…,50}\{32,34,\dots,50\} for all positive masses, so finiteness is part of the statement. The author's stored axiom report lists only propext, Classical.choice and Quot.sound. Not rebuilt here (the author reports about 240 CPU-hours of kernel checking), axioms not printed here, and the second archive of certificates was not opened. Nothing independent anchors the formal statement. The paper's own proof and its C and Python certificates were not checked. That the count was expected to lie between 32 and 50 was confirmed in Figueras, Tucker and Zgliczynski (arXiv:2204.08812).

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion