Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2611.16780

Learning to control and predict light-driven electron-phonon dynamics

Matteo Ballerini Puviani

Published Nov 12, 2026• 0 citations
arXivHTMLPDF

Abstract

Driving lattice vibrations provides a direct route to manipulating electronic correlations through nonlinear electron-phonon coupling. Finding suitable driving fields, however, requires control over the coupled dynamics beyond a simple choice of pulse amplitude and frequency. Here we develop a differentiable framework that combines field optimization with neural-network prediction in a two-site electron-phonon model. We compare piecewise-constant, B-spline, feed-forward, and recurrent control parameterizations, optimized to increase the accumulated electronic double occupation. For the dissipation-free model at zero Hubbard interaction, all the methods increase double occupation, with the B-spline raising the mean total double occupation of approximately 30%. Alongside control optimization, a separate recurrent network learns to predict the next-step double occupation and phonon observables from their preceding values and the applied field. These results establish a computational proof of concept for shaping local electronic correlations through a phonon drive and identify a promising route toward learned models for nonequilibrium materials control.

DOI
10.48550/arxiv.2611.16780

Related papers

Light-enhanced electron-phonon coupling from nonlinear electron-phonon coupling

Michael A. Sentef

We investigate an exact nonequilibrium solution of a two-site electron-phonon model, where an infrared-active phonon that is nonlinearly coupled to the electrons is driven by a laser field. The time-resolved electronic spectrum shows coherence-incoherence spectral weight transfer, a clear signature of light-enhanced electron-phonon coupling. The present study is motivated by recent evidence for enhanced electron-phonon coupling in pump-probe terahertz and angle-resolved photoemission spectroscopy in bilayer graphene when driven near resonance with an infrared-active phonon mode [E. Pomarico et al., Phys. Rev. B 95, 024304 (2017)], and by a theoretical study suggesting that transient electronic attraction arises from nonlinear electron-phonon coupling [D. M. Kennes et al., Nat. Phys. 13, 479 (2017)]. We show that a linear scaling of light-enhanced electron-phonon coupling with the pump field intensity emerges, in accordance with a time-nonlocal self-energy based on a mean-field decoupling using quasiclassical phonon coherent states. Finally, we demonstrate that this leads to enhanced double occupancies in accordance with an effective electron-electron attraction. Our results suggest that materials with strong phonon nonlinearities provide an ideal playground to achieve light-enhanced electron-phonon coupling and possibly light-induced superconductivity.

arXiv:1702.00952•Physical review. B./Physical review. B•May 8, 2017•105 citations

Machine learning nonequilibrium phase transitions in charge-density wave insulators

Yunhao Fan, Sheng Zhang, Gia-Wei Chern

Abstract Nonequilibrium electronic forces play a central role in voltage-driven phase transitions but are notoriously expensive to evaluate in dynamical simulations. Here we develop a machine-learning (ML) framework for adiabatic lattice dynamics coupled to nonequilibrium electrons, and demonstrate it for a gating-induced insulator-to-metal transition out of a charge-density-wave state in the Holstein model. Although exact electronic forces can be obtained from nonequilibrium Green’s-function (NEGF) calculations, their high computational cost renders long-time dynamical simulations prohibitively expensive. By exploiting the locality of the electronic response, we train a neural network to directly predict instantaneous local electronic forces from the lattice configuration, thereby bypassing repeated NEGF calculations during time evolution. When combined with Brownian dynamics, the resulting ML force field quantitatively reproduces domain-wall motion and nonequilibrium phase-transition dynamics obtained from full NEGF simulations, while achieving orders-of-magnitude gains in computational efficiency. Our results establish direct force learning as an efficient and accurate approach for simulating nonequilibrium lattice dynamics in driven quantum materials.

arXiv:2601.07583•Machine Learning Science and Technology•May 5, 2026•1 citation

Learning Memory-Dependent Neural Network Correlation Potentials for Accurate Electron Dynamics

Harish S. Bhat, Christine M. Isborn

Time-dependent density functional theory, though exact in theory, is in practice applied in an adiabatic approximation using exchange-correlation functionals with only local temporal dependence. Simultaneously, the exact correlation potential formally depends on, among other quantities, the time-history of electron densities. Here we develop a framework to learn neural network models of correlation functionals that feature explicit memory-dependence. Our framework includes two different approaches, both linked through their use of adjoint-based optimization. One approach decouples the learning of the functional from inversion to find ground truth values of the correlation potential. The other approach learns the functional directly without requiring inversion. We apply these methods to modeling the electron dynamics of two-electron systems in two spatial dimensions. In both cases, our methods yield correlation functionals with low test set propagation error, outperforming standard local density and generalized gradient approximation functionals by one to two orders of magnitude. Overall, our framework points at one strategy to move beyond the adiabatic approximation and develop memory-dependent correlation functionals that yield accurate propagation for excited state and/or non-equilibrium dynamics.

arXiv:2609.04527•Sep 3, 2026•2 citations

Thermalization Dynamics in the Two-Dimensional Hubbard Model with Neural-Network Quantum States

Alessandro Sinibaldi, Luciano Loris Viteritti, Riccardo Rende +2 more

Thermalization in strongly correlated fermionic systems remains a central open problem in quantum many-body physics. In this work, we investigate the real-time dynamics and the approach to thermalization in the two-dimensional Hubbard model, a paradigmatic framework for correlated electrons, relevant to high-temperature superconductivity and ultracold quantum simulation. Focusing on the half-filled square lattice, we monitor the time evolution of the double occupancy following a quench in the on-site interaction $U$, and assess whether its long-time value is captured by a canonical thermal ensemble. We employ time-dependent variational Monte Carlo methods combined with transformer-based Neural-Network Quantum States to accurately describe the nonequilibrium dynamics of fermions, especially for the behavior at long times, thereby accessing regimes that were previously inaccessible to numerical simulations. Our results reveal two dynamical behaviors: for weak to intermediate interactions, the double occupancy rapidly approaches the thermal prediction, consistent with ergodic evolution; beyond a critical interaction $U_{C}$, the dynamics remains distinct from the thermal expectation on the timescales investigated, revealing signatures of a prethermal plateau delaying fast relaxation. These results establish numerical simulation as a powerful tool to probe nonequilibrium quantum phenomena in correlated fermionic matter.

cond-mat.str-elCross-listedarXiv:2606.05293•Jun 5, 2026•2 citations

Predicting interacting Green's functions with neural networks

Agapov, Egor, Oriol Bertomeu, Carballo, Andrés +2 more

Strongly correlated materials exhibit complex electronic phenomena that are challenging to capture with traditional theoretical methods, yet understanding these systems is crucial for discovering new quantum materials. Addressing the computational bottlenecks in studying such systems, we present a proof-of-concept machine learning-based approach to accelerate Dynamical Mean Field Theory (DMFT) calculations. Our method predicts interacting Green's functions on arbitrary two-dimensional lattices using a two-step ML framework. First, an autoencoder-based network learns and generates physically plausible band structures of materials, providing diverse training data. Next, a dense neural network predicts interacting Green's functions of these physically-possible band structures, expressed in the basis of Legendre polynomials. We demonstrate that this architecture can serve as a substitute for the computationally demanding quantum impurity solver in DMFT, significantly reducing computational cost while maintaining accuracy. This approach offers a scalable pathway to accelerate simulations of strongly correlated systems and lays the groundwork for future extensions to multi-band systems.

arXiv:2411.13644•Nov 20, 2024•7 citations

Deep learning parameter estimation and quantum control of single molecule

Juan M. Scarpetta, Omar Calderón-Losada, Morten Hjorth-Jensen +1 more

Coherent control, a central concept in physics and chemistry, has sparked significant interest due to its ability to fine-tune interference effects in atoms and individual molecules for applications ranging from light-harvesting complexes to molecular qubits. However, precise characterization of the system's dissipative dynamics is required for its implementation, especially at high temperature. In a quantum control experiment, this means learning system-bath parameters and driving coupling strengths. Here, we demonstrate how to infer key physical parameters of a single molecule driven by spectrally modulated pulses at room temperature. We develop and compare two computational approaches based on two-photon absorption photoluminescence signals: an optimization-based minimization scheme and a feed-forward neural network. The robustness of our approach highlights the importance of reliable parameter estimation in designing effective coherent control protocols. Our results have direct applications in ultrafast spectroscopy, quantum materials and technology.

quant-pharXiv:2601.02517•ACS Physical Chemistry Au•Jan 7, 2026•0 citations
Driving lattice vibrations provides a direct route to manipulating electronic correlations through nonlinear electron-phonon coupling. Finding suitable driving fields, however, requires control over the coupled dynamics beyond a simple choice of pulse amplitude and frequency. Here we develop a differentiable framework that combines field optimization with neural-network prediction in a two-site electron-phonon model. We compare piecewise-constant, B-spline, feed-forward, and recurrent control parameterizations, optimized to increase the accumulated electronic double occupation. For the dissipation-free model at zero Hubbard interaction, all the methods increase double occupation, with the B-spline raising the mean total double occupation of approximately 30%. Alongside control optimization, a separate recurrent network learns to predict the next-step double occupation and phonon observables from their preceding values and the applied field. These results establish a computational proof of concept for shaping local electronic correlations through a phonon drive and identify a promising route toward learned models for nonequilibrium materials control.
We investigate an exact nonequilibrium solution of a two-site electron-phonon model, where an infrared-active phonon that is nonlinearly coupled to the electrons is driven by a laser field. The time-resolved electronic spectrum shows coherence-incoherence spectral weight transfer, a clear signature of light-enhanced electron-phonon coupling. The present study is motivated by recent evidence for enhanced electron-phonon coupling in pump-probe terahertz and angle-resolved photoemission spectroscopy in bilayer graphene when driven near resonance with an infrared-active phonon mode [E. Pomarico et al., Phys. Rev. B 95, 024304 (2017)], and by a theoretical study suggesting that transient electronic attraction arises from nonlinear electron-phonon coupling [D. M. Kennes et al., Nat. Phys. 13, 479 (2017)]. We show that a linear scaling of light-enhanced electron-phonon coupling with the pump field intensity emerges, in accordance with a time-nonlocal self-energy based on a mean-field decoupling using quasiclassical phonon coherent states. Finally, we demonstrate that this leads to enhanced double occupancies in accordance with an effective electron-electron attraction. Our results suggest that materials with strong phonon nonlinearities provide an ideal playground to achieve light-enhanced electron-phonon coupling and possibly light-induced superconductivity.
Abstract Nonequilibrium electronic forces play a central role in voltage-driven phase transitions but are notoriously expensive to evaluate in dynamical simulations. Here we develop a machine-learning (ML) framework for adiabatic lattice dynamics coupled to nonequilibrium electrons, and demonstrate it for a gating-induced insulator-to-metal transition out of a charge-density-wave state in the Holstein model. Although exact electronic forces can be obtained from nonequilibrium Green’s-function (NEGF) calculations, their high computational cost renders long-time dynamical simulations prohibitively expensive. By exploiting the locality of the electronic response, we train a neural network to directly predict instantaneous local electronic forces from the lattice configuration, thereby bypassing repeated NEGF calculations during time evolution. When combined with Brownian dynamics, the resulting ML force field quantitatively reproduces domain-wall motion and nonequilibrium phase-transition dynamics obtained from full NEGF simulations, while achieving orders-of-magnitude gains in computational efficiency. Our results establish direct force learning as an efficient and accurate approach for simulating nonequilibrium lattice dynamics in driven quantum materials.
Time-dependent density functional theory, though exact in theory, is in practice applied in an adiabatic approximation using exchange-correlation functionals with only local temporal dependence. Simultaneously, the exact correlation potential formally depends on, among other quantities, the time-history of electron densities. Here we develop a framework to learn neural network models of correlation functionals that feature explicit memory-dependence. Our framework includes two different approaches, both linked through their use of adjoint-based optimization. One approach decouples the learning of the functional from inversion to find ground truth values of the correlation potential. The other approach learns the functional directly without requiring inversion. We apply these methods to modeling the electron dynamics of two-electron systems in two spatial dimensions. In both cases, our methods yield correlation functionals with low test set propagation error, outperforming standard local density and generalized gradient approximation functionals by one to two orders of magnitude. Overall, our framework points at one strategy to move beyond the adiabatic approximation and develop memory-dependent correlation functionals that yield accurate propagation for excited state and/or non-equilibrium dynamics.
Thermalization in strongly correlated fermionic systems remains a central open problem in quantum many-body physics. In this work, we investigate the real-time dynamics and the approach to thermalization in the two-dimensional Hubbard model, a paradigmatic framework for correlated electrons, relevant to high-temperature superconductivity and ultracold quantum simulation. Focusing on the half-filled square lattice, we monitor the time evolution of the double occupancy following a quench in the on-site interaction UUU, and assess whether its long-time value is captured by a canonical thermal ensemble. We employ time-dependent variational Monte Carlo methods combined with transformer-based Neural-Network Quantum States to accurately describe the nonequilibrium dynamics of fermions, especially for the behavior at long times, thereby accessing regimes that were previously inaccessible to numerical simulations. Our results reveal two dynamical behaviors: for weak to intermediate interactions, the double occupancy rapidly approaches the thermal prediction, consistent with ergodic evolution; beyond a critical interaction UCU_{C}UC​, the dynamics remains distinct from the thermal expectation on the timescales investigated, revealing signatures of a prethermal plateau delaying fast relaxation. These results establish numerical simulation as a powerful tool to probe nonequilibrium quantum phenomena in correlated fermionic matter.
Strongly correlated materials exhibit complex electronic phenomena that are challenging to capture with traditional theoretical methods, yet understanding these systems is crucial for discovering new quantum materials. Addressing the computational bottlenecks in studying such systems, we present a proof-of-concept machine learning-based approach to accelerate Dynamical Mean Field Theory (DMFT) calculations. Our method predicts interacting Green's functions on arbitrary two-dimensional lattices using a two-step ML framework. First, an autoencoder-based network learns and generates physically plausible band structures of materials, providing diverse training data. Next, a dense neural network predicts interacting Green's functions of these physically-possible band structures, expressed in the basis of Legendre polynomials. We demonstrate that this architecture can serve as a substitute for the computationally demanding quantum impurity solver in DMFT, significantly reducing computational cost while maintaining accuracy. This approach offers a scalable pathway to accelerate simulations of strongly correlated systems and lays the groundwork for future extensions to multi-band systems.
Coherent control, a central concept in physics and chemistry, has sparked significant interest due to its ability to fine-tune interference effects in atoms and individual molecules for applications ranging from light-harvesting complexes to molecular qubits. However, precise characterization of the system's dissipative dynamics is required for its implementation, especially at high temperature. In a quantum control experiment, this means learning system-bath parameters and driving coupling strengths. Here, we demonstrate how to infer key physical parameters of a single molecule driven by spectrally modulated pulses at room temperature. We develop and compare two computational approaches based on two-photon absorption photoluminescence signals: an optimization-based minimization scheme and a feed-forward neural network. The robustness of our approach highlights the importance of reliable parameter estimation in designing effective coherent control protocols. Our results have direct applications in ultrafast spectroscopy, quantum materials and technology.