Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Solving excited states for long-range interacting trapped ions with neural networks

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The NQES algorithm trains K restricted-Boltzmann neural states inside a determinant-shaped wavefunction so that K low-lying excited states of a long-range spin system are solved together, with demonstrated accuracy up to 300 ions.

desk verdict A genuine extension of Pfau's excited-state method to spin systems, with clean small-system benchmarks but unverified 100-300 ion claims. read the letter →

arxiv 2506.08594 v1 pith:WKKAPMUG submitted 2025-06-10 quant-ph cond-mat.dis-nncs.AIcs.LG

Yixuan Ma , Chang Liu , Weikang Li , Shun-Yao Zhang , L.-M. Duan , Yukai Wu , Dong-Ling Deng This is my paper
classification quant-phcond-mat.dis-nncs.AIcs.LG
keywords neuralquantumexcitedstatesrestrictedBoltzmannmachinelong-rangeIsingmodeltrapped-ionsimulatorHaldane-ShastrystochasticreconfigurationsignproblemvariationalMonteCarlo
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper introduces the neural quantum excited-state (NQES) algorithm, which combines K restricted Boltzmann machine wavefunctions into a Slater-determinant-like composite ansatz so that K low-lying excited states of a spin Hamiltonian are obtained in one variational run, with no explicit orthogonalization. Its central demonstration is that this works for long-range and all-to-all interactions where tensor-network methods struggle: a 20-site transverse-field Ising chain, a 128-site Haldane-Shastry chain with sub-$10^{-3}$ relative energy errors, a 100-ion 2D Wigner-crystal Ising model, and a power-law antiferromagnetic model with up to 300 ions. The paper also uses NQES to solve an experimental puzzle: a 300-ion simulator's quasi-adiabatic ramp is too short to be adiabatic, yet its measured correlations match ground-state predictions, and the computed low-lying excited states carry essentially the same correlation pattern. If correct, NQES offers a scalable classical route to excitation spectra, state-resolved correlations, and gap scaling in long-range interacting quantum systems, with direct use in benchmarking quantum simulators.

What carries the argument

The load-bearing object is the expanded wavefunction $\Psi(\mathcal{S}) = \det[\psi_k(S_{k'})]$, built from $K$ independently parameterized restricted Boltzmann machines—neural-network state ansätze in which visible spin units couple to hidden binary units. The determinant enforces mutual orthogonality automatically and prevents spectral collapse, while the cost function is the trace of a sampled local-energy matrix $E_{\mathrm{loc}} = \mathbb{E}[\,\Psi^{-1} H \Psi\,]$; diagonalizing it after training gives the eigenvalues and, through the same rotation, the correlation functions of each state. Three engineering pieces carry the scalability: stochastic reconfiguration with a matrix-free MINRES-QLP Krylov solver so the covariance matrix is never stored, a bit-packed spin encoding that reduces Pauli actions to shifts, XORs, and population counts, and a curriculum-learning schedule with a local unitary transformation that seeds the sign structure needed for the Haldane-Shastry model.

What would settle it

Measure the actual excited-state populations after the truncated ramp in the 100-ion alternating-sign system, or simulate the ramp time evolution classically for small $N$ and compare the resulting correlation matrix with NQES predictions; if the prepared state has substantial weight outside the low-lying manifold, or its correlations differ from the computed excited-state correlations, the explanation fails.

Watch

Extended reading notes

Core claim

The central claim, stated on the paper's own terms, is that multiple low-lying eigenstates can be learned simultaneously by minimizing the total energy of the expanded determinant ansatz, and that orthogonality comes for free because the determinant vanishes whenever the single states become linearly dependent. At convergence the sampled $K\times K$ local-energy matrix approaches the exact eigenvalues up to a basis rotation, and diagonalizing that matrix yields the individual excitation energies plus state-resolved expectation values via the same rotation. The numerical evidence spans the transverse-field Ising chain (all four lowest states with relative errors below $10^{-3}$ for $M\ge 4N$ hidden units), the Haldane-Shastry chain at $N=128$ after a curriculum-learning pass through an XXZ model, the 100-ion alternating-sign all-to-all Ising model in both ferromagnetic and paramagnetic regimes, and the power-law antiferromagnetic model up to 300 ions where the gap is size-independent at $h=2$ and closes near $h=0.5$.

Load-bearing premise

The trapped-ion explanation assumes the fast experimental ramp mainly excites only the first few low-lying eigenstates that NQES computes; the paper never simulates the ramp dynamics or verifies those populations.

Editorial extensions

If this is right

  • System sizes of 100–300 sites with all-to-all or power-law long-range couplings become accessible for excited-state spectra, correlations, and gaps, beyond exact diagonalization and area-law tensor-network methods.
  • The fast-ramp experiment on the 100-ion alternating-sign model is explained: the first and second excited states show correlation maps nearly indistinguishable from the ground state, in both ferromagnetic and paramagnetic phases.
  • For the power-law antiferromagnetic model the first excited state has a different, longer-range correlation pattern, so ground-state correlations are not robust to nonadiabatic excitations; NQES-computed gaps indicate how slow adiabatic ramps must be.
  • Because the determinant construction is agnostic to the underlying ansatz, the same simultaneous-excited-state scheme can be combined with other network architectures.
  • The method provides a quantitative tool for benchmarking quantum simulators: compute spectra and correlations classically, then compare against device measurements.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: the paper does not simulate the ramp dynamics, so a natural next test is to evolve the actual time-dependent Hamiltonian for $N=20$–$100$ and check whether the population-weighted correlation matrix matches NQES's static low-lying eigenstates.
  • Editorial extension: the correlation robustness in the alternating-sign model is likely tied to the single-phonon-mode structure of that Hamiltonian; in broader classes of long-range models, excited states should be expected to differ from the ground state, as the paper's antiferromagnetic example itself shows.
  • Editorial extension: a direct experimental probe would be to selectively drive or measure the first excited state of the power-law model at $N=300$; NQES predicts long-range correlations in that state, which would distinguish it sharply from the short-range-ordered ground state.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This paper introduces NQES, a variational neural-network method for simultaneously computing several low-lying excited states of quantum spin Hamiltonians. K restricted Boltzmann machine ansätze are combined into a determinant Ψ(S)=det[ψ_i(S_j)]; the total energy of the expanded Hamiltonian is minimized, which yields the sum of the K lowest eigenvalues, and the local energy matrix is diagonalized to extract individual energies and state-resolved observables. The method is benchmarked on a 20-site transverse-field Ising chain against exact diagonalization and on 20/128-site Haldane-Shastry chains against analytic spectra; the latter requires a curriculum-learning initialization from the XXZ model. The paper then applies NQES to two trapped-ion Ising models: a 100-ion single-phonon-mode model with alternating-sign all-to-all couplings, and a power-law antiferromagnetic model up to 300 ions, reporting gap scaling and correlation maps. The authors argue that the similar correlation patterns of low-lying excited states and the ground state explain the surprising robustness of the experimental ground-state correlations to nonadiabatic ramps.

Significance. If correct, the paper establishes a scalable classical tool for excited-state spectra and correlations in long-range spin systems, a regime where tensor-network methods are limited and where existing neural excited-state methods often require explicit orthogonalization penalties. The determinant construction is a clean way to enforce linear independence without penalty terms, and the benchmarks at N=20 (ED) and N=20/128 (analytic Haldane-Shastry) are meaningful and appear correctly implemented. The memory-efficient Krylov stochastic-reconfiguration solver and the bit-encoding Monte Carlo updates are useful engineering contributions. The main caveat is that the headline trapped-ion results at N=100 and N=300 are not certified: the only convergence evidence is a decrease of the variational energy with hidden-neuron density, which is necessary but not sufficient for the claim that individual excited states are accurately represented. The physical explanation of the experiment also relies on an unverified assumption about which excited states are populated during the ramp.

major comments (3)
  1. [Trapped ions, Fig. 3c and Fig. 4] For the N=100 and N=300 systems, the only evidence that NQES has found the true low-lying eigenstates is the decrease of the variational total energy as M/N increases from 1 to 3 (Fig. 3c) and the qualitative similarity of the correlation maps. A variational energy decrease is necessary but not sufficient: the optimization could be trapped in a local minimum, or the Monte Carlo estimate of the local-energy matrix could be biased by the determinant sampling, while the total energy still declines with M. The small-N=20 exact-diagonalization checks in Supplementary Figs. S5-S6 do not control for expressivity or optimization failures at N=100/300. The nonmonotonic behavior at M=5N further shows that the M-dependence is not a simple convergence certificate. Please report per-state local-energy variances (or equivalently the variance of Tr[Eloc] and of the diagonalized eigenvalues) and, where possible, cross-check at intermediate sizes (e.g., N=32/64) against ED or high-quality tensor-network results. Without this, the claims in the abstract and the statement that NQES 'successfully uncovers gap scaling and correlation features' for up to 300 ions are not established.
  2. [Trapped ions, first scenario] The explanation of the experimental robustness rests on the premise that the quasi-adiabatically prepared state is dominated by the first few low-lying eigenstates computed by NQES. The paper states that truncated ramps inevitably populate excited states, but it never simulates the ramp dynamics or quantifies the overlap of the time-evolved state with the computed eigenstates. For N=20 (or another size accessible to ED), please compute the actual time-dependent Schrödinger evolution under the experimental ramp and compare the instantaneous state's projection onto the NQES eigenstates. Only then can one conclude that these specific eigenstates, rather than a broad superposition of higher states, are responsible for the measured correlation patterns.
  3. [Haldane-Shastry results, Fig. 2e] At N=128, the paper reports that both ground and first excited state energies agree with the analytic spectrum to relative error 10^-3, but the only state-resolved correlation function shown (Fig. 2e) is for the ground state. Since the abstract claims faithful reproduction of long-range spin correlations for multiple excited states, please also show excited-state correlation functions at N=128 (or at least at N=20 where ED is available) to verify that the state-resolved observables, not only the energies, are accurate at scale.
minor comments (5)
  1. [Abstract] The text contains a typo: 'Y et' should be 'Yet' in the opening sentence of the abstract/full text.
  2. [Fig. 3c] The x-axis is labeled 'Hidden neuron density' with tick values extending to 6, but the text describes M/N = 1,...,5; please clarify the axis range and tick labels to match the described protocol.
  3. [Trapped ions, second scenario] The sentence 'For models which in general does not possess the ground-state correlation robustness' has a grammatical error; please rephrase for clarity.
  4. [Supplementary Section IV C] The statement that the ground states are configured with σ_i^z = ±sign(b_ik) should specify that the sign choice is global up to the Z2 symmetry; as written, it could be misread as allowing independent per-site signs.
  5. [Code and data availability] For a computational methods paper, consider making the code and data available at submission time or providing a reviewer access link, rather than promising public release only upon publication.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the variational objective is the bare Hamiltonian and benchmarks are external (exact diagonalization, analytic Haldane-Shastry); the only self-citation, Ref. 13, is an input source, not a load-bearing proof.

full rationale

The NQES derivation chain is self-contained. The cost function minimized is the trace of the local energy matrix, Tr[E_loc], which by the standard trace-minimization (Ky Fan) principle equals the sum of the K lowest eigenvalues of the input Hamiltonian; the determinant construction is explicitly credited to Pfau et al. (Ref. 32), an external source, and the implementation is then validated against exact diagonalization for the transverse-field Ising chain (Fig. 2a-b) and against the analytic Haldane-Shastry spectrum for N=128 (Fig. 2d-e), with relative energy errors below 10^-3. No parameter is fitted to the target energies or correlations: the errors are computed after optimization against these independent references, and the analytic HS energies follow from Eq. (S27). The only self-citation is to the authors' own trapped-ion experiment (Ref. 13), which supplies the ion-crystal geometry, phonon-mode parameters, and J_ij values used as inputs to the model; it is not used to establish that NQES finds eigenstates. The large-N (100-300 ion) results are not checked against an independent reference, and the evidence there is variational energy decrease plus qualitative correlation structure; this is a verification/robustness limitation rather than circularity, because the objective remains the physical Hamiltonian and no target excited-state data are encoded in the ansatz or loss. Accordingly, no circular steps are present.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The main burden is carried by the trace-minimization theorem from Ref. 32, by the expressivity and trainability of RBMs for sign-problematic Hamiltonians, and by the assumption that the calculated low-lying eigenstates correspond to states populated in the quasi-adiabatic ion experiment. No physical constants are fitted to the target data; the only hand-chosen numbers are network capacity and SR hyperparameters.

free parameters (4)
  • Hidden neuron density M/N = 3-4 (3N or 4N)
    Network capacity chosen by hand. Fig. 3c shows errors improve up to M=3N/4N and degrade at M=5N. Large-N results use M=3N for 300 ions and M=4N in benchmarks, so this choice directly affects reported accuracy.
  • Stochastic reconfiguration diagonal regularization
    Methods state 'adding a small positive constant to its diagonal entries'; the value is not reported and affects numerical stability.
  • Learning-rate schedule gamma(p)
    SR updates require a learning rate whose schedule is not given in the main text or supplementary; replication depends on unspecified tuning.
  • Curriculum phase-switching criterion
    The switch from XXZ (Phase I) to Haldane-Shastry (Phase II) is shown in Fig. 2d at a vertical dashed line, but the stopping criterion or schedule is not stated.
assumptions (5)
  • standard math Minimizing Tr[S^-1 H] over the span of K variational wavefunctions yields an upper bound to the sum of the K lowest eigenvalues, tight when the span contains the exact eigenstates.
    Invoked in the Methods and around Eq. (1); proof is cited to Ref. 32 rather than derived here.
  • domain assumption Restricted Boltzmann machines can represent the relevant low-lying states of the studied Hamiltonians with sufficient accuracy and are trainable at the required size.
    The paper argues RBM is a universal approximator and imposes no area-law constraint, but does not prove convergence for these specific long-range, sign-problematic Hamiltonians. Fig. 3c shows training can degrade at M=5N.
  • domain assumption The phonon-mediated Ising Hamiltonian with Jij from Eq. (S33), using parameters from Ref. 13, accurately describes the trapped-ion experiment.
    Used in the trapped-ion sections; the authors compute Jij from adiabatic elimination and fix Rabi rate and Lamb-Dicke parameter from the experiment, without giving experimental error bars.
  • domain assumption The experimentally prepared quasi-adiabatic state is dominated by the low-lying eigenstates computed by NQES.
    This is the unstated premise connecting static eigenstate correlations to the observed experimental correlations; no ramp dynamics are simulated.
  • standard math A local unitary U maps the XXZ solution to the antiferromagnetic Heisenberg solution and provides the dominant sign structure for the Haldane-Shastry model.
    Used in the curriculum learning pipeline; the unitary transformation is explicit, but the claim that it transfers the dominant phase structure is empirical.
invented entities (1)
  • K-copy expanded Hilbert space and grand determinant ansatz Ψ(S)=det[ψ_i(S_j)] independent evidence
    purpose: To enforce automatic orthogonality among the K variational states and to define a single variational cost function for the K lowest energies.
    Mathematical device borrowed from Ref. 32; it has external support through the trace-minimization theorem and through the paper's benchmarks against exact diagonalization of Ising and Haldane-Shastry models.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Solving excited states for long-range interacting trapped ions with neural networks." pith.science (2026). https://pith.science/paper/WKKAPMUG

@misc{pith2026250608594,
  author       = {Pith},
  title        = {Pith review of: Solving excited states for long-range interacting trapped ions with neural networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WKKAPMUG}},
  note         = {Machine review of arXiv:2506.08594}
}
read the original abstract

The computation of excited states in strongly interacting quantum many-body systems is of fundamental importance. Yet, it is notoriously challenging due to the exponential scaling of the Hilbert space dimension with the system size. Here, we introduce a neural network-based algorithm that can simultaneously output multiple low-lying excited states of a quantum many-body spin system in an accurate and efficient fashion. This algorithm, dubbed the neural quantum excited-state (NQES) algorithm, requires no explicit orthogonalization of the states and is generally applicable to higher dimensions. We demonstrate, through concrete examples including the Haldane-Shastry model with all-to-all interactions, that the NQES algorithm is capable of efficiently computing multiple excited states and their related observable expectations. In addition, we apply the NQES algorithm to two classes of long-range interacting trapped-ion systems in a two-dimensional Wigner crystal. For non-decaying all-to-all interactions with alternating signs, our computed low-lying excited states bear spatial correlation patterns similar to those of the ground states, which closely match recent experimental observations that the quasi-adiabatically prepared state accurately reproduces analytical ground-state correlations. For a system of up to 300 ions with power-law decaying antiferromagnetic interactions, we successfully uncover its gap scaling and correlation features. Our results establish a scalable and efficient algorithm for computing excited states of interacting quantum many-body systems, which holds potential applications ranging from benchmarking quantum devices to photoisomerization.

Figures

Figures reproduced from arXiv: 2506.08594 by the authors.

Figure 1
Figure 1. FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Grassmann Variational Monte Carlo with neural wave functions

    quant-ph 2025-07 conditional novelty 6.0 of 10

    Grassmann Variational Monte Carlo generalizes neural-network excited-state optimization to subspaces and accurately reproduces low-lying spectra of the 2D Heisenberg model.

Reference graph

Works this paper leans on

56 extracted references · 51 canonical work pages · cited by 1 Pith paper

  1. [1]

    Defenu, T

    N. Defenu, T. Donner, T. Macr `ı, G. Pagano, S. Ruffo, and A. Trombettoni, Long-range interacting quantum systems, Rev. Mod. Phys. 95, 035002 (2023)

    work page 2023

  2. [2]

    S. R. White, Early days of DMRG, Nat. Rev. Phys. 5, 264 (2023)

    work page 2023

  3. [3]

    Schollw ¨ock, The density-matrix renormalization group, Rev

    U. Schollw ¨ock, The density-matrix renormalization group, Rev. Mod. Phys. 77, 259 (2005)

    work page 2005

  4. [4]

    M. B. Hastings, An area law for one-dimensional quantum sys- tems, J. Stat. Mech. 2007, P08024 (2007)

    work page 2007

  5. [5]

    Eisert, Colloquium: Area laws for the entanglement entropy, Rev

    J. Eisert, Colloquium: Area laws for the entanglement entropy, Rev. Mod. Phys. 82, 277 (2010)

    work page 2010

  6. [6]

    R. P . Feynman, Simulating physics with computers, in Feynman and computation (CRC Press, 2018) pp. 133–153

    work page 2018

  7. [7]

    I. M. Georgescu, S. Ashhab, and F. Nori, Quantum simulation, Rev. Mod. Phys. 86, 153 (2014)

    2014

  8. [8]

    J. I. Cirac and P . Zoller, Goals and opportunities in quantum simulation, Nat. Phys. 8, 264 (2012)

    work page 2012

Show all 56 references
  1. [9]

    Y . Kim, A. Eddins, S. Anand, K. X. Wei, E. van den Berg, S. Rosenblatt, H. Nayfeh, Y . Wu, M. Zaletel, K. Temme, et al., Evidence for the utility of quantum computing before fault tol- erance, Nature 618, 500 (2023)

    2023

  2. [10]

    A. D. King, A. Nocera, M. M. Rams, J. Dziarmaga, R. Wiersema, W. Bernoudy, J. Raymond, N. Kaushal, N. Heins- dorf, R. Harris, et al., Beyond-classical computation in quantum simulation, Science 388, 199 (2025)

    2025

  3. [11]

    Ebadi, T

    S. Ebadi, T. T. Wang, H. Levine, A. Keesling, G. Semeghini, A. Omran, D. Bluvstein, R. Samajdar, H. Pichler, W. W. Ho, et al., Quantum phases of matter on a 256-atom programmable quantum simulator, Nature 595, 227 (2021)

    2021

  4. [12]

    J. G. Bohnet, B. C. Sawyer, J. W. Britton, M. L. Wall, A. M. Rey, M. Foss-Feig, and J. J. Bollinger, Quantum spin dynam- ics and entanglement generation with hundreds of trapped ions, Science 352, 1297 (2016)

    2016

  5. [13]

    Guo, Y .-K

    S.-A. Guo, Y .-K. Wu, J. Y e, L. Zhang, W.-Q. Lian, R. Y ao, Y . Wang, R.-Y . Y an, Y .-J. Yi, Y .-L. Xu,et al. , A site-resolved two-dimensional quantum simulator with hundreds of trapped ions, Nature 630, 613 (2024)

    2024

  6. [14]

    Tindall, A

    J. Tindall, A. Mello, M. Fishman, M. Stoudenmire, and D. Sels, Dynamics of disordered quantum systems with two- and three- dimensional tensor networks, arXiv:2503.05693 (2025)

    2025 arXiv

  7. [15]

    Sharir, Y

    O. Sharir, Y . Levine, N. Wies, G. Carleo, and A. Shashua, Deep Autoregressive Models for the Efficient V ariational Simulation of Many-Body Quantum Systems, Phys. Rev. Lett. 124, 020503 (2020)

    2020

  8. [16]

    T. D. Barrett, A. Malyshev, and A. I. Lvovsky, Autoregressive neural-network wavefunctions for ab initio quantum chemistry, Nat. Mach. Intell. 4, 351 (2022)

    2022

  9. [17]

    Carleo and M

    G. Carleo and M. Troyer, Solving the quantum many-body problem with artificial neural networks, Science 355, 602 (2017)

    2017

  10. [18]

    D.-L. Deng, X. Li, and S. Das Sarma, Quantum Entanglement in Neural Network States, Phys. Rev. X 7, 021021 (2017)

    2017

  11. [19]

    Carrasquilla and G

    J. Carrasquilla and G. Torlai, How To Use Neural Networks To Investigate Quantum Many-Body Physics, PRX Quantum 2, 040201 (2021)

    2021

  12. [20]

    Nomura, Helping restricted Boltzmann machines with quantum-state representation by restoring symmetry, J

    Y . Nomura, Helping restricted Boltzmann machines with quantum-state representation by restoring symmetry, J. Con- dens. Matter Phys. 33, 174003 (2021)

    2021

  13. [21]

    Hermann, J

    J. Hermann, J. Spencer, K. Choo, A. Mezzacapo, W. M. C. Foulkes, D. Pfau, G. Carleo, and F. No ´e, Ab initio quantum chemistry with neural-network wavefunctions, Nat. Rev. Chem. 7, 692 (2023)

    2023

  14. [22]

    Z. Li, Z. Lu, R. Li, X. Wen, X. Li, L. Wang, J. Chen, and W. Ren, Spin-symmetry-enforced solution of the many-body Schr¨odinger equation with a deep neural network, Nat. Com- put. Sci. 4, 910 (2024)

    2024

  15. [23]

    M. T. Entwistle, Z. Sch ¨atzle, P . A. Erdman, J. Hermann, and F. No ´e, Electronic excited states in deep variational Monte Carlo, Nat. Commun. 14, 274 (2023)

    2023

  16. [24]

    W. A. Wheeler, K. G. Kleiner, and L. K. Wagner, Ensemble variational Monte Carlo for optimization of correlated excited state wave functions, Electron. Struct. 6, 025001 (2024)

    2024

  17. [25]

    Schautz and C

    F. Schautz and C. Filippi, Optimized Jastrow–Slater wave func- tions for ground and excited states: Application to the lowest states of ethene, J. Chem. Phys. 120, 10931 (2004)

    2004

  18. [26]

    Cordova, L

    F. Cordova, L. J. Doriol, A. Ipatov, M. E. Casida, C. Fil- ippi, and A. V ela, Troubleshooting time-dependent density- functional theory for photochemical applications: Oxirane, J. Chem. Phys. 127, 164111 (2007)

    2007

  19. [27]

    Filippi, M

    C. Filippi, M. Zaccheddu, and F. Buda, Absorption Spectrum of the Green Fluorescent Protein Chromophore: A Difficult Case for ab Initio Methods?, J. Chem. Theory Comput. 5, 2074 (2009)

    2009

  20. [28]

    Cuzzocrea, A

    A. Cuzzocrea, A. Scemama, W. J. Briels, S. Moroni, and C. Fil- ippi, V ariational Principles in Quantum Monte Carlo: The Trou- bled Story of V ariance Minimization, J. Chem. Theory Comput. 16, 4203 (2020)

    2020

  21. [29]

    M. Dash, S. Moroni, C. Filippi, and A. Scemama, Tailoring CIPSI Expansions for QMC Calculations of Electronic Excita- tions: The Case Study of Thiophene, J. Chem. Theory Comput. 17, 3426 (2021)

    2021

  22. [30]

    D. M. Ceperley and B. Bernu, The calculation of excited state properties with quantum Monte Carlo, J. Chem. Phys. 89, 6316 (1988)

    1988

  23. [31]

    M. P . Nightingale and V . Melik-Alaverdian, Optimization of Ground- and Excited-State Wave Functions and van der Waals Clusters, Phys. Rev. Lett. 87, 043401 (2001)

    2001

  24. [32]

    D. Pfau, S. Axelrod, H. Sutterud, I. V on Glehn, and J. S. Spencer, Accurate computation of quantum excited states with neural networks, Science 385, eadn0137 (2024)

    2024

  25. [33]

    V erstraete, V

    F. V erstraete, V . Murg, and J. I. Cirac, Matrix product states, projected entangled pair states, and variational renormalization group methods for quantum spin systems, Adv. Phys. 57, 143 (2008)

    2008

  26. [34]

    Cincio, J

    L. Cincio, J. Dziarmaga, and M. M. Rams, Multiscale Entan- glement Renormalization Ansatz in Two Dimensions: Quantum Ising Model, Phys. Rev. Lett. 100, 240603 (2008)

    2008

  27. [35]

    F. D. M. Haldane, Exact Jastrow-Gutzwiller resonating- valence-bond ground state of the spin- 1 2 antiferromagnetic Heisenberg chain with 1/ r2 exchange, Phys. Rev. Lett. 60, 635 (1988)

    1988

  28. [36]

    B. S. Shastry, Exact solution of an S =1/2 Heisenberg antiferro- magnetic chain with long-ranged interactions, Phys. Rev. Lett. 60, 639 (1988)

    1988

  29. [37]

    Bravyi, G

    S. Bravyi, G. Carleo, D. Gosset, and Y . Liu, A rapidly mixing Markov chain from any gapped quantum many-body system, Quantum 7, 1173 (2023)

    2023

  30. [38]

    Bengio, J

    Y . Bengio, J. Louradour, R. Collobert, and J. Weston, Curricu- lum learning, in Proceedings of the 26th Annual International Conference on Machine Learning , ICML ’09 (Association for 10 Computing Machinery, New Y ork, NY , USA, 2009) pp. 41–48

    2009

  31. [39]

    Monroe, W

    C. Monroe, W. C. Campbell, L.-M. Duan, Z.-X. Gong, A. V . Gorshkov, P . W. Hess, R. Islam, K. Kim, N. M. Linke, G. Pagano, et al. , Programmable quantum simulations of spin systems with trapped ions, Rev. Mod. Phys. 93, 025001 (2021)

    2021

  32. [40]

    M. K. Joshi, C. Kokail, R. van Bijnen, F. Kranzl, T. V . Zache, R. Blatt, C. F. Roos, and P . Zoller, Exploring large-scale entan- glement in quantum simulation, Nature 624, 539 (2023)

    2023

  33. [41]

    Iqbal, N

    M. Iqbal, N. Tantivasadakarn, R. V erresen, S. L. Campbell, J. M. Dreiling, C. Figgatt, J. P . Gaebler, J. Johansen, M. Mills, S. A. Moses, et al., Non-Abelian topological order and anyons on a trapped-ion processor, Nature 626, 505 (2024)

    2024

  34. [42]

    Lloyd, Quantum-mechanical computers and uncomputabil- ity, Phys

    S. Lloyd, Quantum-mechanical computers and uncomputabil- ity, Phys. Rev. Lett. 71, 943 (1993)

    1993

  35. [43]

    T. S. Cubitt, D. Perez-Garcia, and M. M. Wolf, Undecidability of the spectral gap, Nature 528, 207 (2015)

    2015

  36. [44]

    Bausch, T

    J. Bausch, T. S. Cubitt, A. Lucia, and D. Perez-Garcia, Unde- cidability of the Spectral Gap in One Dimension, Phys. Rev. X 10, 031038 (2020)

    2020

  37. [45]

    Carrasquilla and R

    J. Carrasquilla and R. G. Melko, Machine learning phases of matter, Nat. Phys. 13, 431 (2017)

    2017

  38. [46]

    E. P . L. van Nieuwenburg, Y .-H. Liu, and S. D. Huber, Learning phase transitions by confusion, Nat. Phys. 13, 435 (2017)

    2017

  39. [47]

    Sorella, M

    S. Sorella, M. Casula, and D. Rocca, Weak binding between two aromatic rings: Feeling the van der waals attraction by quantum Monte Carlo methods, J. Chem. Phys. 127 (2007)

    2007

  40. [48]

    S.-C. T. Choi, C. C. Paige, and M. A. Saunders, MINRES-QLP: A Krylov Subspace Method for Indefinite or Singular Symmet- ric Systems, SIAM J. Sci. Comput. 33, 1810 (2011). Supplementary: Solving excited states for long-range interacting trapped ions with neural networks CONTENTS...

    2011

  41. [49]

    Optimization scheme for the ground state 2

  42. [50]

    Learning excited states 3

    Efficient sampling of configurations 3 C. Learning excited states 3

  43. [51]

    Optimization scheme for excited states 3

  44. [52]

    Curriculum learning pipeline for a long-range antiferromagnetic model 5 A

    Efficient sampling of configurations 5 III. Curriculum learning pipeline for a long-range antiferromagnetic model 5 A. Theoretical solution 5 B. Sign problem of the Haldane-Shastry model 6 C. Curriculum learning scheme 6 IV . Solving the many-body Hamiltonian of strongly interac...

  45. [53]

    In this framework, the quantum state ψW is evolved under an imaginary time iτ to approximatee−τHψW , a process that drives the state toward the ground state as τ→∞

    Optimization scheme for the ground state To optimize the variational wavefunction, we employ the stochastic reconfiguration (SR) method, which can be understood as a variational approximation to an imaginary-time evolution. In this framework, the quantum state ψW is evolved und...

  46. [54]

    First, we apply the Metropolis-Hastings method to generate a Markov chain of configurations S1,...,S P that obeys the probability distribution |ψW (S)|2

    Efficient sampling of configurations During the training process, the average of quantities X = ES∼ψ2 [X(S)] can be computed through the standard Monte Carlo paradigm. First, we apply the Metropolis-Hastings method to generate a Markov chain of configurations S1,...,S P that obey...

  47. [55]

    We define the matrix for the composite function Ψ(S) =      ψ1(S1) ··· ψK ( S1)

    Optimization scheme for excited states In this section, we present the scheme to efficiently evaluate the cost function for multiple excited states. We define the matrix for the composite function Ψ(S) =      ψ1(S1) ··· ψK ( S1) . . . . . . ψ1 ( SK) ··· ψK ( SK)     . ...

  48. [56]

    !" # ($$!$$

    Efficient sampling of configurations In comparison to the single-state scenario, the number of network parameters has increased by a factor of K. Additionally, there is a shift in the cost function to Tr[Eloc]. These changes are reflected in the SR method when updating the networ...

    2017

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.