Aftermath bot: New page for OpenAI's BFSS bound-state manuscripts: threshold and positive-energy claims, background, methods, Lattice24 reviews

2026-10-18T16:25:07Z

New page for OpenAI's BFSS bound-state manuscripts: threshold and positive-energy claims, background, methods, Lattice24 reviews

New page

'''Bound states of the BFSS matrix model''' concern the normalizable eigenstates of the supersymmetric matrix quantum mechanics obtained by reducing ten-dimensional supersymmetric Yang–Mills theory to one time dimension, the model of the matrix-theory conjecture of Banks, Fischler, Shenker and Susskind (BFSS). Two manuscripts in OpenAI's public mathematics collection ([https://github.com/openai/math github.com/openai/math]) claim, for the model with the center of mass removed, that the SU(''N'') model has exactly one normalizable state of zero energy for every ''N'' ≥ 2, and that the SU(2) model has infinitely many normalizable states of positive energy. This page summarizes the published claims and lists publicly available follow-up work. Statements are the respective authors' claims; the publication of a paper or a review is not presented here as independent verification.

== Background ==
* As recounted in the introductions of the two manuscripts:
** The model describes nine Hermitian matrices with their fermionic partners; it appears in the regularization of supermembranes by de Wit, Hoppe and Nicolai and in the BFSS matrix-theory conjecture. Its potential vanishes on commuting matrices, which form noncompact flat directions. De Wit, Lüscher and Nicolai proved that the spectrum of the finite matrix Hamiltonians is the whole half-line [0, ∞) and left open whether normalizable eigenstates lie in it.
** Witten's D0-brane bound-state prediction, central to matrix theory, is that after removal of the center of mass every matrix size ''N'' has exactly one normalizable state of zero energy, lying at the threshold of the continuous spectrum. In the matrix-theory interpretation it supplies the internal state of a supergraviton carrying ''N'' units of longitudinal momentum.
** Index computations: Yi computed the SU(2) bulk term and proposed a boundary correction, and Sethi and Stern obtained the two-particle index including its boundary term; Moore, Nekrasov and Shatashvili computed the matrix integrals entering the principal contribution; Green and Gutperle, and Konechny, gave arguments for general ''N'' under stated assumptions on the asymptotic regions. Sethi and Stern, and Hasler and Hoppe, proved that normalizable ground states are rotation invariant.
** Asymptotic analysis of candidate ground states was developed by Halpern and Schwartz, Graf and Hoppe, Fröhlich, Graf, Hasler, Hoppe and Yau, Hasler and Hoppe, and Lin and Yin. Porrati and Rozenberg proposed a mass-deformation criterion, and Kac and Smilga counted mass-deformed vacua, identifying persistence of normalizability at zero mass as the remaining hypothesis. The manuscript uses the plane-wave deformation of Berenstein, Maldacena and Nastase.
** For positive energies, the original BFSS paper stated that besides the zero-energy threshold states "No other normalizable bound states can occur". Danielsson, Ferretti and Sundborg found normalizable excited wavefunctions in a Born–Oppenheimer approximation to the relative SU(2) model and described them as metastable, and Smilga conjectured families of normalized excited states in a large-''N'' analysis.

== The original results ==
* '''Claim (threshold bound state).''' Theorem 1.1 of the first manuscript states that for every integer ''N'' ≥ 2 the kernel of the relative SU(''N'') Hamiltonian is one-dimensional, and that it consists of Spin(9)-invariant states that are even in the stated fermion grading. The Hamiltonian is the nonnegative self-adjoint operator associated with the closure of the averaged squared-supercharge quadratic form on gauge-invariant smooth compactly supported functions, and square integrability is on the whole configuration space. The manuscript describes this as a proof of the threshold-bound-state conjecture.
* ''Method, as described:'' the proof compares the kernel with a massive problem. A deformed supercharge with a plane-wave mass term is Fredholm on gauge- and torus-invariant states, and its index equals the number ''p''(''N'') of partitions of ''N'', which label the classical massive vacua. As the coupling grows, which amounts to removing the mass, the massive kernel splits into ''p''(''N'') − 1 lines of profiles of separated proper clusters and a profile of the whole block valued in the kernel. An exterior Hardy-type estimate, proved together with the main theorem by induction on ''N'', excludes loss of norm at intermediate scales, and a spectral gap and recovery argument shows that the whole-block part is one-dimensional.
* '''Claim (positive energies).''' Theorem 1.1 of the second manuscript states that the relative SU(2) Hamiltonian, defined by closing the gauge-invariant supercharge form, has an orthonormal sequence of square-integrable eigenvectors with positive eigenvalues tending to infinity. The manuscript states that this refutes, for this operator at ''N'' = 2, the exclusion of further normalizable bound states in the original BFSS paper, and that the large-''N'' matrix-theory conjecture is a separate question.
* ''Method, as described:'' in Spin(9) sectors whose highest weight has a large second coordinate, branching to the Spin(8) stabilizer of an escape direction excludes the low transverse oscillator levels. The remaining oscillator energy, which grows linearly along the flat directions, dominates the fermionic term, so the restricted operator has compact resolvent and trivial kernel. Explicit gauge-invariant polynomials show that infinitely many such sectors are infinite-dimensional.
* '''Sources.''' The manuscripts [https://github.com/openai/math/tree/1581d4cbdd69a0faf34b2089614fe2a814316a6a/preprints/The-unique-threshold-bound-state-of-the-SU-N-BFSS-model-September-24-2026 The unique threshold bound state of the SU(N) BFSS model] (24 September 2026) and [https://github.com/openai/math/tree/1581d4cbdd69a0faf34b2089614fe2a814316a6a/preprints/Positive-eigenvalues-of-the-relative-SU-2-BFSS-Hamiltonian-October-5-2026 Positive eigenvalues of the relative SU(2) BFSS Hamiltonian] (5 October 2026).
* '''Formalization.''' The repository's contents listing at commit <code>1581d4c</code> gives no Lean link for these manuscripts, and its Lean documentation has no scope note for them.

== Public follow-up work ==
{| class="wikitable"
! Date !! Work !! Relation !! Public claim !! Public status (checked 19 October 2026)
|-
| 2026-10-16 || J. Jardine, Lattice24, referee reviews [https://github.com/JJardine919/openai-math-audit/blob/a06a8c6e77974d4c6021338a55a8e91b634f67f7/reviews/rank_105.json rank_105] (threshold bound state) and [https://github.com/JJardine919/openai-math-audit/blob/a06a8c6e77974d4c6021338a55a8e91b634f67f7/reviews/rank_474.json rank_474] (positive energies) in [https://github.com/JJardine919/openai-math-audit JJardine919/openai-math-audit] (archived at [https://doi.org/10.5281/zenodo.23422995 Zenodo 10.5281/zenodo.23422995]) || Review (AI-assisted audit) || Both reviews report no gap. For the threshold manuscript, the reviewer states that it read all seven sections, checked the Spin(9) gamma-matrix identities numerically, and did not fully re-derive the Sobolev and oscillator-moment bounds of the fast-line perturbation or the chart-transition maps. It records two minor issues: the order-of-limits bookkeeping for the full-collapse profile is not written out, and the untwisting and Berry-phase claims of the fast-line lemma are argued tersely. For the positive-energy manuscript, it states that it checked each step by hand, and remarks that the refutation applies to the specific self-adjoint realization chosen, as the paper states. || Part of an audit of the collection's manuscripts without Lean that the audit describes as complete. The reviews are published under CC BY 4.0.
|}

== Notes ==
* The list covers follow-ups found in public sources up to the date above; it is not a priority or completeness claim.

[[Category:OpenAI results]]

Bound states of the BFSS matrix model - Revision history

Aftermath bot: New page for OpenAI's BFSS bound-state manuscripts: threshold and positive-energy claims, background, methods, Lattice24 reviews