Thermal Bootstrap of Large- 𝑵 N Matrix Models via Conic Optimization
Sophia M. Adams · 2025
Read Thermal Bootstrap of Large- 𝑵 N Matrix Models via Conic Optimization by Sophia M. Adams (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-01209.txt.
Introduction
Sophia Adams improves thermal bootstrap methods for matrix quantum mechanics, a technique that bounds a system's energies using only symmetry and self consistency conditions. The targets are anharmonic oscillator models with one and two large matrices, studied in the large N limit.
Using a conic optimization solver, she bounds the thermal energies without the logarithmic relaxation earlier work relied on. Stricter bounds apply to the one matrix model. At low temperature it can be read as a theory of long strings. The bounds pin the first long string excitation energy to within a tiny fraction of a percent of its physical value. The same equations yield the first estimate of the first long string coupling coefficient obtained from symmetry and self consistency alone.
Catalog record
| CREATOR | Sophia M. Adams |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 27 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-01209
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Thermal Bootstrap of Large- 𝑵 N Matrix Models via Conic Optimization Sophia M. Adams Abstract This paper is aimed at improving thermal bootstrapping methods for matrix quantum mechanics and quantum field theory. The thermal energies of the large- N N one-matrix anharmonic oscillator and large- N N two-matrix anharmonic oscillator were bounded without logarithmic relaxation using the Quantum Information Conic Solver. For the one-matrix anharmonic oscillator, dual to a theory of “long strings”, stricter bounds yielded a value of the first long string excited energy within 0.001 % 0.001\% of the physical value and the first estimation from symmetry and self-consistency equations alone of the first long string coupling coefficient. 1 Background Bootstrapping is a method of solving a theory by specifying conditions that the theory must satisfy, such as self-consistency equations and symmetries. Solving a theory with bootstrapping can be advantageous because it can produce tight, rigorous bounds on observables without much computational cost. In particular, matrix quantum mechanics (MQM) is an important bootstrapping target because some MQM theories may not be analytically solvable or reliably solvable through numerical or perturbative methods but may contain many symmetries. Additionally, MQM provides simplified, nonperturbative models that capture essential features of quantum gravity and black hole dynamics, such as black hole entropy in the BFSS matrix model (Ref. PhysRevD.55.5112 ). In the thermal bootstrap, the main object of study is the thermal density matrix, constrained by symmetries and by the Kubo–Martin–Schwinger (KMS) condition that characterizes thermal equilibrium (Refs. ArakiSewell1977 , Sewell1977 ). Bootstrapping a thermal state involves determining which combinations of correlators and expectation values are consistent with these constraints. In this work, the thermal energy is bootstrapped for the one-matrix and two-matrix anharmonic oscillator. From the one-matrix model, we estimate the long string ground state energy, the first long string excited state energy, and the first long string coupling coefficient, which characterize the structure of the thermal spectrum and the interaction strength in the matrix model’s dual long string description. Previous thermal bootstrap studies have primarily used linear semidefinite programming (SDP) methods, which approximate nonlinear physical constraints by linear relaxations. While rigorous, such methods can fail to enforce nonlinear relations like the KMS condition exactly. To overcome this limitation, we explore the use of the Quantum Information Conic Solver (QICS) (Ref. he2024qicsquantuminformationconic ), a nonlinear SDP solver designed to handle quantum conic constraints natively. The idea behind QICS is to extend the feasible space of SDPs to include nonlinear cone structures that naturally encode positivity and consistency conditions of quantum states. This allows us to impose nonlinear constraints directly, rather than approximating them through relaxations. In particular, QICS uses the Skajaa-Ye algorithm (Ref. SkajaaYe2015 ) to find the feasible space of a quantum relative entropy cone (which happens to correspond exactly to the form of the KMS constraint). In this work, we take steps toward improving thermal bootstrapping methods for MQM by applying QICS to bootstrap the one-matrix and two-matrix anharmonic oscillators, achieving larger system sizes and tighter bounds than previously obtained through linear methods (Ref. cho2025thermalbootstrapmatrixquantum ). 2 Models In this paper, we are interested in the one-matrix and two-matrix anharmonic oscillator in the ’t Hooft large- N limit. For each bootstrap problem, we consider operators built as products of elementary matrices X,P (and other matrices in the two-matrix model). The length L is the maximum number of operators appearing in a single product. Increasing L allows longer operator products in the bootstrap and generally improves the precision of the resulting bounds. 2.1 One-Matrix In one-matrix quantum mechanics, the quartic anharmonic oscillator Hamiltonian takes the form H=\operatorname{Tr}(\frac{1}{2}P^{2}+\frac{1}{2}X^{2}+\frac{g}{N}X^{4}) (1) where the canonical coordinate X and its conjugate momentum P are traceless N\times N Hermitian matrices. The gauged theory or “singlet sector” can be reformulated in terms of non-interacting fermions, while the ungauged theory contains “non-singlet sectors” dual to a theory of “long strings” (Refs. Maldacena2005LongStrings and Balthazar_2019 ) and is not known to be analytically solvable. One can derive bounds on the energy spectrum, which characterize the energy spectrum of long strings and long string interactions in the adjoint sector, by bootstrapping the ungauged theory at finite temperature where E(\beta)\equiv\langle H\rangle_{\beta}\equiv\operatorname{Tr}_{\mathcal{H}}(\rho_{\beta}H)=\operatorname{Tr}_{\mathcal{H}}(\frac{1}{Z(\beta)}e^{-\beta H}H)=\operatorname{Tr}_{\mathcal{H}}(\frac{1}{\operatorname{Tr}_{\mathcal{H}}\left(e^{-\beta H}\right)}e^{-\beta H}H). (2) Due to large- N factorization, \langle\operatorname{Tr}(\mathcal{O}_{1})...\operatorname{Tr}(\mathcal{O}_{1})\rangle_{\beta}=\langle\operatorname{Tr}(\mathcal{O}_{1})\rangle_{\beta}...\langle\operatorname{Tr}(\mathcal{O}_{1})\rangle_{\beta}, (3) so we only need to consider single-trace operators. Note that each trace is to be regarded as order N , and X and P are rescaled by \frac{1}{\sqrt{N}} , allowing us to omit the \frac{1}{N} multiplying X^{4} in the Hamiltonian. As in Ref. cho2025thermalbootstrapmatrixquantum , we express the components of X and P as X_{ab}=X^{A}T_{ab}^{A} and P_{ab}=P^{A}T_{ab}^{A} where T_{ab}^{A} are the traceless Hermitian SU(N) generators satisfying the completeless relation T_{ab}^{A}T_{cd}^{A}=\delta_{ad}\delta_{bc}-\frac{1}{N}\delta_{ab}\delta_{cd} so that the canonical commutation relation \frac{1}{N}[X^{A},P^{B}]=\frac{i}{N}\delta^{AB} is equivalent to \frac{1}{N}[X_{ab},P_{cd}]=\frac{i}{N}\delta_{ad}\delta_{bc}+\mathcal{O}(\frac{1}{N^{2}}). (4) We consider the same large- N simplifications as in Ref. cho2025thermalbootstrapmatrixquantum , including treating any remaining trace products as independent variables. We find bounds on the thermal energy by bootstrapping the system of operators \mathcal{O}_{i} in tuples of X and P up to maximum length \frac{L}{2} (e.g. for L=6 , O_{i}\in\{1,X,P,XP,PX,XXX,XXP,XPX,PXX,XPP,PXP,PPX, PPP\} ) subject to the following constraints: • Semidefinite positivity of the moment matrix \mathcal{M}_{ij}=\langle\operatorname{Tr}(\mathcal{O}_{i}^{\dagger}\mathcal{O}_{j})\rangle_{\beta} • Parity symmetry, \langle\operatorname{Tr}(\mathcal{O^{\prime}}_{1}\mathcal{O^{\prime}}_{2}\cdots\mathcal{O^{\prime}}_{n})\rangle_{\beta}=0 with n odd and \mathcal{O^{\prime}}_{i}\in\{X,P\} • Time-reversal symmetry, \operatorname{Re}(\langle\operatorname{Tr}(P_{1}P_{2}\cdots P_{n})\rangle_{\beta})=\operatorname{Im}(\langle\operatorname{Tr}(P_{1}P_{2}\cdots P_{m})\rangle_{\beta})=0 with n odd and m even • Reality conditions, \langle\operatorname{Tr}(\mathcal{O}_{i}^{\dagger})\rangle_{\beta}=\langle\operatorname{Tr}(\mathcal{O}_{i})\rangle_{\beta}^{*} • Stationary state conditions, \langle[H,\operatorname{Tr}(\mathcal{O}_{i})]\rangle_{\beta}=0 • Cyclicity of the trace with large- N factorization, \langle\operatorname{Tr}(\mathcal{O^{\prime}}_{1}\mathcal{O^{\prime}}_{2}\cdots\mathcal{O^{\prime}}_{n})\rangle_{\beta}=\langle\operatorname{Tr}(\mathcal{O^{\prime}}_{2}\mathcal{O^{\prime}}_{3}\cdots\mathcal{O^{\prime}}_{n}\mathcal{O^{\prime}}_{1})\rangle_{\beta} +\sum_{k=2}^{n}\langle\operatorname{Tr}(\mathcal{O^{\prime}}_{2}\cdots\mathcal{O^{\prime}}_{k-1})\rangle_{\beta}\langle\operatorname{Tr}([\mathcal{O^{\prime}}_{1},\mathcal{O^{\prime}}_{k}])\rangle_{\beta}\langle\operatorname{Tr}(\mathcal{O^{\prime}}_{k+1}\cdots\mathcal{O^{\prime}}_{n})\rangle_{\beta} with \mathcal{O^{\prime}}_{i}\in\{X,P\} • Normalization, \frac{1}{N}\langle\operatorname{Tr}(\mathds{1})\rangle_{\beta}=1 • Large- N commutation relations (Eq. ( 4 )) • KMS condition on a thermal state (Section 3 ) 2.2 Two-Matrix The two-matrix quartic anharmonic oscillator is described by the Hamiltonian H=\operatorname{Tr}(\frac{1}{2}(P_{1}^{2}+P_{2}^{2})+\frac{1}{2}(X_{1}^{2}+X_{2}^{2})-\frac{g}{N}[X_{1},X_{2}]^{2}) (5) where P_{1} is the conjugate momentum of X_{1} and P_{2} , X_{2} . Bootstrapping the two-matrix model is ultimately motivated by the BFSS matrix model, conjectured to be dual to M-theory in the large- N limit PhysRevD.55.5112 . Due to the computational cost associated with bootstrapping a large number of matrices and the fact that the thermal state is not supersymmetric, bootstrapping the full BFSS model, which has 9 bosonic and 16 fermionic matrices, is a much more nontrivial task using this method. Instead, we consider the bosonic BFSS model, dual to a theory of a black D0-brane in the large- N limit, truncated to two matrices. As an additional simplification, we include the mass term \frac{1}{2}(X_{1}^{2}+X_{2}^{2}) . The two-matrix system made from tuples of X_{1} , X_{2} , P_{1} , and P_{2} is subject to the same constraints as the one-matrix model with an additional constraint that comes from swapping the 1 and 2 operator labels: X_{1}\leftrightarrow X_{2} , P_{1}\leftrightarrow P_{2} . Reformulating the model in terms of complex operators P=\frac{1}{\sqrt{2}}(P_{1}+iP_{2}),\quad\bar{P}=\frac{1}{\sqrt{2}}(P_{1}-iP_{2}),\quad X=\frac{1}{\sqrt{2}}(X_{1}+iX_{2}),\quad\bar{X}=\frac{1}{\sqrt{2}}(X_{1}-iX_{2}) (6) as in Ref. lin2025highprecisionbootstrapmultimatrixquantum allows us to view the SO(2) symmetry as a U(1) symmetry and make manifest additional constraints from U(1) charge conservation. In terms of complex operators, the Hamiltonian can be written H=\operatorname{Tr}(P\bar{P}+X\bar{X}+\frac{g}{N}[X,\bar{X}]^{2}) (7) since the system is invariant under the relabeling of barred and unbarred operators. In practice, the U(1) charge of a term is the difference between the number of barred and unbarred operators, and charge conservation implies that any observable with a nonzero U(1) charge is zero. In this basis, we can consider the same constraints as in the real basis but can