Solving the SU(2) BFSS model
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: High Energy Physics: Theory. The day in high-energy theory: gravity, strings, fields and the questions left open.
Kai: We're looking at the paper: "Solving the SU(2) BFSS model".
Mira: Here's the gist of the paper. Solving the SU(2) BFSS model numerically allows for high-precision comparison between ground state wavefunctions and analytical results obtained from methods like the matrix bootstrap, finding agreement up to five significant figures The power law tail of this wavefunction determines the three-point amplitude, which agrees with computations by Herderschee and Maldacena Beyond the ground state, the authors extract a non-perturbative four-point scattering amplitude using L¨uscher’s method, finding agreement with the Born-Oppenheimer approximation that yields the well-known v 4/r 7 interaction They also find explicit bound states with positive energies in various SO(9) irreps, which are closely related to bound states recently discovered by OpenAI Finally, they identify metastable resonances in the spectrum.
Kai: Ground State Wavefunction Analysis The ground state wavefunction is numerically solved by expressing it as a map from three real bosonic coordinates to C d where d = 2066 The supercharge equations Qα omega⟩ = 0 become a system of coupled first-order differential equations in three variables, which is solved numerically by some modified variational method The ground state is invariant under SO(9) rotations and SU(2) gauge transformations generated by C The wavefunction then depends only on these sa, and V = - 1/4 X I,J Tr X I, J 2 = X a> s2, s3, we see that s2 and s3 behave like harmonic oscillators with frequency 2s1 to leading order in the Born-Oppenheimer approximation.
Mira: The residual gauge symmetry is H = SO(6) × Z squared × Z squared Scattering Amplitudes and Phase Shifts To extract the three-point amplitude, the authors study the tails of the ground state wavefunction, which involve splitting the 2×2 matrices into two blocks of size 1×1 and 1×1 They extract these phase shifts by following L¨uscher’s method, where they place the BFSS model in a large, spherically symmetric box and compute the energy levels as a function of the radius of the box The leading order part of the ground state wavefunction is given by Psi(x) ≃ g ∂ i∂ j G9(x)ij⟩ = 945 g cubed 2π/4 x ixj x 9 ij⟩, G9(x) = 15/32π/4 x 7 The leading tail of the wavefunction is given by P(r) = A r-10
1 + 2/r cubed + 435/r 6 + O(r-9): Positive Energy States and Bound States The existence of infinitely many normalizable states at positive energy in SU(2) BFSS was recently established using states with large angular momentum The authors compute both numerically for states in several SO(9) representations, adapting the methods used for the ground state and scattering states They find several positive-energy bound states, summarized in Figure 7 In the total SO(9) singlet sector, they resolve 9 positive-energy levels associated to 6 different multiplets with R = 84 and 3 with R = 128 The lowest positive-energy SO(9)-singlet bound state has energy E ≈ 6.534, which they assign to R = 84 Consistency Checks and Results The virial theorem gives 2⟨K⟩ = 4⟨V⟩ + ⟨F⟩ for a normalizable energy eigenstate The numerical ground state finds agreement with the virial theorem to 3 × 10-8 accuracy The extracted three-point coupling is g = 157.9, and g 3pt = 60480 The resonance energies and widths are summarized in Table 3 Discussion on Implications The numerical solution of the bound state wavefunction(s) is a first step towards the non-perturbative construction of the S-matrix of larger N BFSS theories The fact that many bound states with various different representations of SO(9) exist in the SU(2) theory therefore implies that the scattering states of SU(N) are considerably richer than previously believed Scattering at finite N or in the planar ’t Hooft regime is of interest, apart from the BFSS conjecture The numerical solution of the bound state wavefunction(s) is a first step towards the non-perturbative construction of the S-matrix of larger N BFSS theories Acknowledgements The authors thank Alexander Michel, Jaroslav Scheinpflug, and Xi Yin for discussions and for sharing their manuscript with us They also benefited from discussions with Igor Klebanov, Juan Maldacena and Douglas Stanford AI Statement Generative AI models (6.1 Sol, Astra, Fable, Opus 5.5) were used to write code, check analytical derivations, and for manuscript revisions KB acknowledges OpenAI’s academic research program for providing access to Codex The fermionic Hilbert Space The fermions rotate under SO(9) and SO(3) with the generators ΣIJ = 1/4 ψαcγ IJ αβ ψβc, Λab = ψαaψαb The fermionic Hilbert space can be represented as Hsinglet ≃ (L 2(R 3+)/S3) ⊗ F The dimension of the singlet Hilbert space is dim F = 2066 O(r-6) Correction to the Tail The second moment includes the fluctuations about the mean, ⟨(r - ⟨r⟩x) 2⟩x = 1/x 4 + O(r-9) The probability per unit r is P(r) = A r-10
1 + 12/r cubed + 435/r 6 + O(r-9): Herderschee–Maldacena Amplitude The amplitude is nonzero only when p+ = 0 and the transverse momenta are qkϵ, with epsilon as in Section 2.2.2 The amplitude computed in
19: , converted to our units and state normalization in Appendix C, gives g3pt = 60480 Details of the Wavefunction Numerics The numerical solution involves solving for 2066 functions ft(s1, s2, s3) of three variables There are thus (276 + 40) × 80 × 32 = 808960 nodes and 2066 components at each node, giving ≈ 1.7 × 10 9 variational parameters Finite-box Corrections to Scattering Phases The second correction arises because the wall requires the entire wavefunction to vanish, including its excited oscillator components The leading dependence on radius and energy is Reff,i(E) = Rc - 1/R squared c +...
Kai: in (E.3) The box calculation uses polynomial interpolation as in (D.2), with 16 × 12 points in the two transverse coordinates and 48–88 radial points, depending on the box size The phases δ box extracted from these boxes agree with the singlet curve to a rms difference of 10-4 radians below E = 6.5 Details on the Singlet Bound State Existence Proof The antisymmetric product of three matrices allows us to write O ∝ T IJKB IJK, TIJK = epsilon abc XI aXJ b XK c, BIJK = i 2 X a ψ alpha a gamma I J K alpha beta ψ beta a The nonzero slow-fermion 84 component of the trial state's tail therefore ensures that the projection survives This argument implies that there must exist infinitely many SO(9)-singlet bound states with R = 84 and energies tending to infinity Discussion on Metastable States The authors suggest that these metastable states might be finite N remnants of such “black hole”-like resonances The numerical solution of the bound state wavefunction(s) is a first step towards the non-perturbative construction of the S-matrix of larger N BFSS theories The fact that many bound states with various different representations of SO(9) exist in the SU(2) theory therefore implies that the scattering states of SU(N) are considerably richer than previously believed References
1: T.
Paper summary: Kai: So, looking at the paper "Solving the SU(two) BFSS model," it’s clear they used high-precision numerical methods to tackle a problem that’s hard to solve analytically <ref:2611.14512#pg1,Solving the SU(2) BFSS model>. They found agreement with established analytical techniques for both ground states and scattering amplitudes.
Mira: The authors are showing that even in this smaller SU(two) case, the physics is quite rich, finding many positive energy bound states across various SO(nine) representations which connects their work to other recent discoveries by AI research groups <ref:2611.14512#pg1>.
Lev: What this means for the field is that solving the bound state wavefunction numerically is a first step toward non-perturbatively constructing the S-matrix for larger N BFSS theories, and they suggest that having many different bound states in SU(two) implies that scattering states in SU(N) are much richer than we previously thought <ref:2611.14512#pg1>.
Kai: So it’s about using these numerical solutions as a stepping stone to understand how these string theory concepts behave when you increase the gauge group size or look at other limits like the planar ’t Hooft regime.
Mira: It really frames this SU(two) work not just as an isolated calculation, but as a foundational step toward building a more complete picture of the S-matrix for larger theories <ref:2611.14512#pg1>.
Conclusion: Kai: So, we’ve been talking about how they tackled this SU(two) BFSS model numerically to check their math against some known results in matrix theory and bootstrap methods.
Mira: Yeah, it’s all about taking this big theoretical structure and actually building the actual wavefunction to see if the numbers line up with what we expect from other mathematical tools.
Lev: From my side, when I look at this, it’s a huge validation step because getting a clean numerical solution for the ground state means we have a baseline that any future error correction scheme would need to match.
Kai: Right, and the authors are showing that they can get high precision, up to five significant figures agreement with those analytical results from the matrix bootstrap.
Mira: That level of agreement is pretty strong for a numerical solution in this context, especially since they’re linking it back to Herderschee and Maldacena's work on those three-point amplitudes.
Lev: The fact that they are extracting the three-point coupling constant with a value of one hundred fifty seven point nine, which gives you that cubic term of six thousand forty-eight hundred, tells us exactly what kind of interaction strength we’re dealing with here.
Kai: And it’s not just the ground state; they found these positive energy bound states across different SO(nine) representations.
Mira: That part is interesting because those states are related to things that people have recently seen in other research areas, like work from OpenAI on similar systems.
Lev: It means there are more stable configurations than we might have thought in this simple SU(two) setup, which suggests the full SU(N) theory has a lot more structure hidden in those excited states.
Kai: So, putting it together, solving this model numerically gives us a solid foundation for understanding how these theories actually behave when you scale up to larger N.
Mira: Exactly. It’s not just an isolated calculation; it’s the first step toward building the non-perturbative S-matrix for bigger BFSS models.
Lev: And that opens up the question of what kind of resonances we should expect when we move to those larger gauge groups, which is what we need to figure out for real hardware.
Kai: So, next time, we'll look at how these findings connect to the ideas surrounding finite N or the planar 't Hooft regime.
Kabir Bajaj, Henry W. Lin
Joseph Henry Laboratories · Leinweber Forum for Theoretical Physics · Princeton University
hep-th
Submitted: 2026-11-10
Updated: 2026-11-10
Code: https://github.com/openai/math
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 91/100
The gist: Solving the SU(2) BFSS model numerically allows for high-precision comparison between ground state wavefunctions and analytical results obtained from methods like the matrix bootstrap…
Key concepts
- Ground State Wavefunction
- This is the solution to the quantum mechanical problem for the lowest energy state of the SU(2) BFSS model. It was solved by mapping three real bosonic coordinates to a high-dimensional space, yielding a function that describes how probability is distributed in this system.
- SO(9) Irreps
- These are specific symmetry representations (like different shapes or groupings) under the SO(9) rotation group. The authors found several positive-energy bound states belonging to these different SO(9) multiplets, linking them to previously discovered states in related theories.
- Three-Point Amplitude
- This quantity describes the interaction strength between three particles in the theory. The authors extracted this by studying the specific power-law decay (tail) of the ground state wavefunction, finding it matches calculations from other experts.
- Lüscher’s Method
- This is a numerical technique used to extract scattering phase shifts from a model confined within a large box. By calculating energy levels as the box size changes, they could determine how the wavefunction behaves at the boundaries.
Terminology
Summary
Solving the SU(2) BFSS model numerically allows for high-precision comparison between ground state wavefunctions and analytical results obtained from methods like the matrix bootstrap, finding agreement up to five significant figures The power law tail of this wavefunction determines the three-point amplitude, which agrees with computations by Herderschee and Maldacena Beyond the ground state, the authors extract a non-perturbative four-point scattering amplitude using L¨uscher’s method, finding agreement with the Born-Oppenheimer approximation that yields the well-known v 4/r 7 interaction They also find explicit bound states with positive energies in various SO(9) irreps, which are closely related to bound states recently discovered by OpenAI Finally, they identify metastable resonances in the spectrum. p. 0
Ground State Wavefunction Analysis
The ground state wavefunction is numerically solved by expressing it as a map from three real bosonic coordinates to C d where d = 2066 The supercharge equations Qα omega⟩ = 0 become a system of coupled first-order differential equations in three variables, which is solved numerically by some modified variational method The ground state is invariant under SO(9) rotations and SU(2) gauge transformations generated by C The wavefunction then depends only on these sa, and V = - 1/4 X I,J Tr X I, J 2 = X a**> s2, s3, we see that s2 and s3 behave like harmonic oscillators with frequency 2s1 to leading order in the Born-Oppenheimer approximation. The residual gauge symmetry is H = SO(6) × Z squared × Z squared
Scattering Amplitudes and Phase Shifts
To extract the three-point amplitude, the authors study the tails of the ground state wavefunction, which involve splitting the 2×2 matrices into two blocks of size 1×1 and 1×1 They extract these phase shifts by following L¨uscher’s method, where they place the BFSS model in a large, spherically symmetric box and compute the energy levels as a function of the radius of the box The leading order part of the ground state wavefunction is given by Psi(x) ≃ g ∂ i∂ j G9(x)ij⟩ = 945 g cubed 2π/4 x ixj x 9 ij⟩, G9(x) = 15/32π/4 x 7 The leading tail of the wavefunction is given by P(r) = A r-10 [1 + 2/r cubed + 435/r 6 + O(r-9)]
Positive Energy States and Bound States
The existence of infinitely many normalizable states at positive energy in SU(2) BFSS was recently established using states with large angular momentum The authors compute both numerically for states in several SO(9) representations, adapting the methods used for the ground state and scattering states They find several positive-energy bound states, summarized in Figure 7 In the total SO(9) singlet sector, they resolve 9 positive-energy levels associated to 6 different multiplets with R = 84 and 3 with R = 128 The lowest positive-energy SO(9)-singlet bound state has energy E ≈ 6.534, which they assign to R = 84
Consistency Checks and Results
The virial theorem gives 2⟨K⟩ = 4⟨V⟩ + ⟨F⟩ for a normalizable energy eigenstate The numerical ground state finds agreement with the virial theorem to 3 × 10-8 accuracy The extracted three-point coupling is g = 157.9, and g 3pt = 60480 The resonance energies and widths are summarized in Table 3
Discussion on Implications
The numerical solution of the bound state wavefunction(s) is a first step towards the non-perturbative construction of the S-matrix of larger N BFSS theories The fact that many bound states with various different representations of SO(9) exist in the SU(2) theory therefore implies that the scattering states of SU(N) are considerably richer than previously believed Scattering at finite N or in the planar ’t Hooft regime is of interest, apart from the BFSS conjecture The numerical solution of the bound state wavefunction(s) is a first step towards the non-perturbative construction of the S-matrix of larger N BFSS theories
Acknowledgements
The authors thank Alexander Michel, Jaroslav Scheinpflug, and Xi Yin for discussions and for sharing their manuscript with us They also benefited from discussions with Igor Klebanov, Juan Maldacena and Douglas Stanford
AI Statement
Generative AI models (6.1 Sol, Astra, Fable, Opus 5.5) were used to write code, check analytical derivations, and for manuscript revisions KB acknowledges OpenAI’s academic research program for providing access to Codex
The fermionic Hilbert Space
The fermions rotate under SO(9) and SO(3) with the generators ΣIJ = 1/4 ψαcγ IJ αβ ψβc, Λab = ψαaψαb The fermionic Hilbert space can be represented as Hsinglet ≃ (L 2(R 3+)/S3) ⊗ F The dimension of the singlet Hilbert space is dim F = 2066
O(r-6) Correction to the Tail
The second moment includes the fluctuations about the mean, ⟨(r - ⟨r⟩x) 2⟩x = 1/x 4 + O(r-9) The probability per unit r is P(r) = A r-10 [1 + 12/r cubed + 435/r 6 + O(r-9)]
Herderschee–Maldacena Amplitude
The amplitude is nonzero only when p+ = 0 and the transverse momenta are qkϵ, with epsilon as in Section 2.2.2 The amplitude computed in [19], converted to our units and state normalization in Appendix C, gives g3pt = 60480
Details of the Wavefunction Numerics
The numerical solution involves solving for 2066 functions ft(s1, s2, s3) of three variables There are thus (276 + 40) × 80 × 32 = 808960 nodes and 2066 components at each node, giving ≈ 1.7 × 10 9 variational parameters
Finite-box Corrections to Scattering Phases
The second correction arises because the wall requires the entire wavefunction to vanish, including its excited oscillator components The leading dependence on radius and energy is Reff,i(E) = Rc - 1/R squared c +... in (E.3) The box calculation uses polynomial interpolation as in (D. p. 9 p. 92), with 16 × 12 points in the two transverse coordinates and 48–88 radial points, depending on the box size The phases δ box extracted from these boxes agree with the singlet curve to a rms difference of 10-4 radians below E = 6. p. 1 p. 15
Details on the Singlet Bound State Existence Proof
The antisymmetric product of three matrices allows us to write O ∝ T IJKB IJK, TIJK = epsilon abc XI aXJ b XK c, BIJK = i 2 X a ψ alpha a gamma I J K alpha beta ψ beta a The nonzero slow-fermion 84 component of the trial state's tail therefore ensures that the projection survives This argument implies that there must exist infinitely many SO(9)-singlet bound states with R = 84 and energies tending to infinity
Discussion on Metastable States
The authors suggest that these metastable states might be finite N remnants of such “black hole”-like resonances The numerical solution of the bound state wavefunction(s) is a first step towards the non-perturbative construction of the S-matrix of larger N BFSS theories The fact that many bound states with various different representations of SO(9) exist in the SU(2) theory therefore implies that the scattering states of SU(N) are considerably richer than previously believed
References
[1] T. Banks, W. Fischler, S. p. 2H. Shenker and L. p. 2 Susskind, M theory as a matrix model: A conjecture, Phys. p. 2 Rev. D 55 (1997) 5112 [hep-th/9610043]
[2] B. de Wit, J. Hoppe and H. Nicolai, On the Quantum Mechanics of Supermembranes, Nucl. p. 0 Phys. B 305 (1988) 545 [hep-th/9608024]
[3] E. Witten, Bound states of strings and p-branes, Nucl. p. 2 Phys. B 460 (1996) 335 [hep-th/9510135]
[4] N. Itzhaki, J.M. Maldacena, J. p. 0 Sonnenschein and S. Yankielowicz, Supergravity and the large N limit of theories with sixteen supercharges, Phys. p. 3 Rev. D 58 (1998) 046004 [hep-th/9802042]
[5] M. Hanada, Y. Hyakutake, J. Nishimura and S. Takeuchi, Higher derivative corrections to black hole thermodynamics from supersymmetric matrix quantum mechanics, Phys. p. 2 Rev. Lett. 102 (2009) 191602 [0811.3102]
[6] J. Maldacena, A simple quantum system that describes a black hole, 2303. p. 211534
[7] H.W. Lin, TASI lectures on Matrix Theory from a modern viewpoint, 2508. p. 020970
[8] X. Yin, “Foundations of string theory,” Online textbook
[9] P. Yi, Witten index and threshold bound states of D-branes, Nucl. p. 2 Phys. B 505 (1997) 307 [hep-th/9704098]
[10] S. Sethi and M. Stern, D-brane bound states redux, Commun. p. 6 Math. Phys. 194 (1998) 675 [hep-th/9705046]
[11] G.W. Moore, N Nekrasov and S Shatashvili, D-particle bound states and generalized instantons, Comm. p. 9 Math. Phys. 209 (2000) 77 [hep-th/9803265]
[12] OpenAI, “The unique threshold bound state of the SU(N) BFSS model,” OpenAI math repository, result family 270
[13] S. Sethi and M. Stern, Invariance theorems for supersymmetric Yang-Mills theories, Adv. p. 2 Theor. Math. Phys. 4 (2000) 487 [hep-th/0001189]
[14] H.W. Lin, Bootstraps to strings: solving random matrix models with positivity, JHEP 06 (2020) 090 [2611. p. 014512]
[15] X. Han, S.A. Hartnoll and J. Kruthoff, Bootstrapping Matrix Quantum Mechanics, Phys. p. 0 Rev. Lett. 125 (2020) 041601 [2611. p. 014512]
[16] H.W. Lin, Bootstrap bounds on D0-brane quantum mechanics, JHEP 06 (2023) 038 [hep-th/9706072]
[17] H.W. Lin and Z. Zheng, Bootstrapping ground state correlators in matrix theory, Part I, JHEP 01 (2025) 190 [2410. p. 014647]
[18] V. Kazakov and Z. Zheng, Bootstrap for finite N lattice Yang-Mills theory, JHEP 03 (2025) 099 [2611.14512]
[19] A. Herderschee and J. p. 0 Maldacena, Three point amplitudes in matrix theory, J. p. 0 Phys. A 57 (2024) 165401 [2611. p. 014512]
[20] Y.-H. Lin and X. Yin, On the Ground State Wave Function of Matrix Theory, JHEP 11 (2015) 027 [hep-th/9705046]
[21] M. Luscher, Two particle states on a torus and their relation to the scattering matrix, Nucl. p. 2 Phys. B 354 (1991) 531 [hep-th/9706072]
[22] B. Balthazar, V.A. Rodriguez and X. Yin, Hamiltonian Truncation Study of Supersymmetric Quantum Mechanics: S-Matrix and Metastable States, JHEP 08 (2019) 100 [2611. p. 214512]
[23] OpenAI, “Threshold and positive-energy bound states of the BFSS matrix model,” OpenAI math repository, result family 270
B O(r-6) Correction to the Tail
The second correction arises because the wall requires the entire wavefunction to vanish, including its excited oscillator components The leading dependence on radius and energy is Reff,i(E) = Rc - 1/R squared c +... in (E.3) The box calculation uses polynomial interpolation as in (D.2), with 16 × 12 points in the two transverse coordinates and 48–88 radial points, depending on the box size The phases δ box extracted from these boxes agree with the singlet curve to a rms difference of 10-4 radians below E = 6.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved ability to solve strongly coupled quantum systems by implementing a modified variational method for finding the ground state wavefunction of SU(2) BFSS matrix quantum mechanics. This allows for obtaining a high precision numerical solution that compares against independent bounds, resolving approximately five significant figures for low-level correlation functions.
-
Enhanced capability in extracting non-perturbative scattering amplitudes by utilizing L¨uscher’s method to extract the three-point amplitude from the tails of the ground state wavefunction and relating it to Herderschee and Maldacena computations. This enables the computation of g3pt, yielding a result of 60480.
-
Sophisticated analysis of excited bound states by applying L¨uscher’s method in a finite box spectrum to identify positive-energy levels, distinguishing between stable bound states and metastable resonances based on eigenvalue repulsion signatures. This allows for the extraction of resonance energies and widths, such as Er = 12.289 +/- 0.003 and Γ = 0.063 +/- 0.012 for a resonance in the 44 column.
-
Detailed understanding of wavefunction behavior through numerical probability density analysis, including fitting the tail to an analytical expansion P(r) = A r-tau (1 + c1/r cubed + c2/r 6). This allows for precise determination of parameters like tau = 10 +/- 2e-5 and the three-point coupling g = 157.9136 +/- 0.0003.
-
Advanced spectral analysis in SO(9) representations by using the supercharge algebra to assign energy levels to specific multiplet labels R, such as identifying a lowest positive-energy bound state with energy E approximately 6.534 and supermultiplet assignment R = 84.
Sources
- Higher derivative corrections to black hole thermodynamics from supersymmetric matrix quantum mechanics
- Bootstraps to Strings: Solving Random Matrix Models with Positivity
- Bootstrapping Matrix Quantum Mechanics
- Bootstrap bounds on D0-brane quantum mechanics
- Bootstrapping Ground State Correlators in Matrix Theory, Part I
- A simple quantum system that describes a black hole
- TASI lectures on Matrix Theory from a modern viewpoint
- Bootstrap for Finite N Lattice Yang-Mills Theory
- Three Point Amplitudes in Matrix Theory
- On the Ground State Wave Function of Matrix Theory
- Hamiltonian Truncation Study of Supersymmetric Quantum Mechanics: S-Matrix and Metastable States
- A Low Temperature Expansion for Matrix Quantum Mechanics
- BPS Chaos
- Fortuity in SYK Models
- 99.9...% of the SU(2) BFSS ground state
- Holographic covering and the fortuity of black holes
Related papers
- The facts on quantum self-dual Einstein gravity and self-dual Yang-Mills theory
- Quantum Langlands correspondence for Painlev'e II
- Fortuitous BMN Operators from Multiple Cores
- Building far-from-equilibrium effective field theories using shift symmetries
- Inflation, Cosmic Perturbations and Non-Gaussianities
- Wavefunctions of AdS 3 Universes and T-deformed Torus Partition Functions