-
Boltzmann Generators and the New Frontier of Computational Sampling in Many-Body Systems
Authors:
Alessandro Coretti,
Sebastian Falkner,
Jan Weinreich,
Christoph Dellago,
O. Anatole von Lilienfeld
Abstract:
The paper by Noé et al. [F. Noé, S. Olsson, J. Köhler and H. Wu, Science, 365:6457 (2019)] introduced the concept of Boltzmann Generators (BGs), a deep generative model that can produce unbiased independent samples of many-body systems. They can generate equilibrium configurations from different metastable states, compute relative stabilities between different structures of proteins or other organ…
▽ More
The paper by Noé et al. [F. Noé, S. Olsson, J. Köhler and H. Wu, Science, 365:6457 (2019)] introduced the concept of Boltzmann Generators (BGs), a deep generative model that can produce unbiased independent samples of many-body systems. They can generate equilibrium configurations from different metastable states, compute relative stabilities between different structures of proteins or other organic molecules, and discover new states. In this commentary, we motivate the necessity for a new generation of sampling methods beyond molecular dynamics, explain the methodology, and give our perspective on the future role of BGs.
△ Less
Submitted 25 April, 2024;
originally announced April 2024.
-
Machine Learning for Mini-EUSO Telescope Data Analysis
Authors:
Mario Bertaina,
Mikhail Zotov,
Dmitry Anzhiganov,
Dario Barghini,
Carl Blaksley,
Antonio Giulio Coretti,
Aleksandr Kryazhenkov,
Antonio Montanaro,
Leonardo Olivi
Abstract:
Neural networks as well as other methods of machine learning (ML) are known to be highly efficient in different classification tasks, including classification of images and videos. Mini- EUSO is a wide-field-of-view imaging telescope that operates onboard the International Space Station since 2019 collecting data on miscellaneous processes that take place in the atmosphere of Earth in the UV range…
▽ More
Neural networks as well as other methods of machine learning (ML) are known to be highly efficient in different classification tasks, including classification of images and videos. Mini- EUSO is a wide-field-of-view imaging telescope that operates onboard the International Space Station since 2019 collecting data on miscellaneous processes that take place in the atmosphere of Earth in the UV range. Here we briefly present our results on the development of ML-based approaches for recognition and classification of track-like signals in the Mini-EUSO data, among them meteors, space debris and signals the light curves and kinematics of which are similar to those expected from extensive air showers generated by ultra-high-energy cosmic rays. We show that even simple neural networks demonstrate impressive performance in solving these tasks.
△ Less
Submitted 28 August, 2023;
originally announced August 2023.
-
Enhanced Sampling of Configuration and Path Space in a Generalized Ensemble by Shooting Point Exchange
Authors:
Sebastian Falkner,
Alessandro Coretti,
Christoph Dellago
Abstract:
The computer simulation of many molecular processes is complicated by long time scales caused by rare transitions between long-lived states. Here, we propose a new approach to simulate such rare events, which combines transition path sampling with enhanced exploration of configuration space. The method relies on exchange moves between configuration and trajectory space, carried out based on a gene…
▽ More
The computer simulation of many molecular processes is complicated by long time scales caused by rare transitions between long-lived states. Here, we propose a new approach to simulate such rare events, which combines transition path sampling with enhanced exploration of configuration space. The method relies on exchange moves between configuration and trajectory space, carried out based on a generalized ensemble. This scheme substantially enhances the efficiency of the transition path sampling simulations, particularly for systems with multiple transition channels, and yields information on thermodynamics, kinetics and reaction coordinates of molecular processes without distorting their dynamics. The method is illustrated using the isomerization of proline in the KPTP tetrapeptide.
△ Less
Submitted 22 March, 2023; v1 submitted 17 February, 2023;
originally announced February 2023.
-
Learning Map**s between Equilibrium States of Liquid Systems Using Normalizing Flows
Authors:
Alessandro Coretti,
Sebastian Falkner,
Phillip Geissler,
Christoph Dellago
Abstract:
Generative models are a promising tool to address the sampling problem in multi-body and condensed-matter systems in the framework of statistical mechanics. In this work, we show that normalizing flows can be used to learn a transformation to map different liquid systems into each other allowing at the same time to obtain an unbiased equilibrium distribution through a reweighting process. Two proo…
▽ More
Generative models are a promising tool to address the sampling problem in multi-body and condensed-matter systems in the framework of statistical mechanics. In this work, we show that normalizing flows can be used to learn a transformation to map different liquid systems into each other allowing at the same time to obtain an unbiased equilibrium distribution through a reweighting process. Two proof-of-principles calculations are presented for the transformation between Lennard-Jones systems of particles with different depths of the potential well and for the transformation between a Lennard-Jones and a system of repulsive particles. In both numerical experiments, systems are in the liquid state. In future applications, this approach could lead to efficient methods to simulate liquid systems at ab-initio accuracy with the computational cost of less accurate models, such as force field or coarse-grained simulations.
△ Less
Submitted 22 August, 2022;
originally announced August 2022.
-
Conditioning Normalizing Flows for Rare Event Sampling
Authors:
Sebastian Falkner,
Alessandro Coretti,
Salvatore Romano,
Phillip Geissler,
Christoph Dellago
Abstract:
Understanding the dynamics of complex molecular processes is often linked to the study of infrequent transitions between long-lived stable states. The standard approach to the sampling of such rare events is to generate an ensemble of transition paths using a random walk in trajectory space. This, however, comes with the drawback of strong correlations between subsequently sampled paths and with a…
▽ More
Understanding the dynamics of complex molecular processes is often linked to the study of infrequent transitions between long-lived stable states. The standard approach to the sampling of such rare events is to generate an ensemble of transition paths using a random walk in trajectory space. This, however, comes with the drawback of strong correlations between subsequently sampled paths and with an intrinsic difficulty in parallelizing the sampling process. We propose a transition path sampling scheme based on neural-network generated configurations. These are obtained employing normalizing flows, a neural network class able to generate statistically independent samples from a given distribution. With this approach, not only are correlations between visited paths removed, but the sampling process becomes easily parallelizable. Moreover, by conditioning the normalizing flow, the sampling of configurations can be steered towards regions of interest. We show that this approach enables the resolution of both the thermodynamics and kinetics of the transition region.
△ Less
Submitted 19 May, 2023; v1 submitted 29 July, 2022;
originally announced July 2022.
-
Mass-Zero constrained dynamics for simulations based on orbital-free density functional theory
Authors:
Alessandro Coretti,
Taylor J. Baird,
Rodolphe Vuilleumier,
Sara Bonella
Abstract:
A new algorithm for efficient and fully time-reversible integration of first-principles molecular dynamics based on orbital-free density functional theory (OFDFT) is presented. The algorithm adapts to this nontrivial case the recently introduced Mass-Zero (MaZe) constrained dynamics. The formalism ensures that full adiabatic separation is enforced between nuclear and electronic degrees of freedom…
▽ More
A new algorithm for efficient and fully time-reversible integration of first-principles molecular dynamics based on orbital-free density functional theory (OFDFT) is presented. The algorithm adapts to this nontrivial case the recently introduced Mass-Zero (MaZe) constrained dynamics. The formalism ensures that full adiabatic separation is enforced between nuclear and electronic degrees of freedom and, consequently, that the exact Born-Oppenheimer probability for the nuclei is sampled. Numerical integration of the MaZe dynamics combines standard molecular dynamics algorithms, e.g. Verlet or velocity Verlet, with the SHAKE method to impose the minimum conditions on the electronic degrees of freedom as a set of constraints. The developments presented in this work, that include a bespoke adaptation of the standard SHAKE algorithm, ensure that the quasilinear scaling of OFDFT is preserved by the new method for a broad range of kinetic and exchange-correlation functionals, including nonlocal ones. The efficiency and accuracy of the approach is demonstrated via calculations of static and dynamic properties of liquid sodium in the constant energy and constant temperature ensembles.
△ Less
Submitted 9 October, 2022; v1 submitted 17 June, 2022;
originally announced June 2022.
-
Mass-Zero constrained dynamics and statistics for the shell model in magnetic field
Authors:
David D. Girardier,
Alessandro Coretti,
Giovanni Ciccotti,
Sara Bonella
Abstract:
In several domains of physics, including first principle simulations and classical models for polarizable systems, the minimization of an energy function with respect to a set of auxiliary variables must be performed to define the dynamics of physical degrees of freedom. In this paper, we discuss a recent algorithm proposed to efficiently and rigorously simulate this type of systems: the Mass-Zero…
▽ More
In several domains of physics, including first principle simulations and classical models for polarizable systems, the minimization of an energy function with respect to a set of auxiliary variables must be performed to define the dynamics of physical degrees of freedom. In this paper, we discuss a recent algorithm proposed to efficiently and rigorously simulate this type of systems: the Mass-Zero (MaZe) Constrained Dynamics. In MaZe the minimum condition is imposed as a constraint on the auxiliary variables treated as degrees of freedom of zero inertia driven by the physical system. The method is formulated in the Lagrangian framework, enabling the properties of the approach to emerge naturally from a fully consistent dynamical and statistical viewpoint. We begin by presenting MaZe for typical minimization problems where the imposed constraints are holonomic and summarizing its key formal properties, notably the exact Born-Oppenheimer dynamics followed by the physical variables and the exact sampling of the corresponding physical probability density. We then generalize the approach to the case of conditions on the auxiliary variables that linearly involve their velocities. Such conditions occur, for example, when describing systems in external magnetic field and they require to adapt MaZe to integrate semiholonomic constraints. The new development is presented in the second part of this paper and illustrated via a proof-of-principle calculation of the charge transport properties of a simple classical polarizable model of NaCl.
△ Less
Submitted 23 April, 2021;
originally announced April 2021.
-
Fluctuation relations for dissipative systems in constant external magnetic field: theory and molecular dynamics simulations
Authors:
Alessandro Coretti,
Lamberto Rondoni,
Sara Bonella
Abstract:
It has recently been pointed out that Hamiltonian particle systems in constant magnetic fields satisfy generalized time-reversal symmetries that enable to prove useful statistical relationships based on equilibrium phase-space probability distributions without the need to invert, as commonly considered necessary, the magnetic field. Among these relations, that hold without need of Casimir modifica…
▽ More
It has recently been pointed out that Hamiltonian particle systems in constant magnetic fields satisfy generalized time-reversal symmetries that enable to prove useful statistical relationships based on equilibrium phase-space probability distributions without the need to invert, as commonly considered necessary, the magnetic field. Among these relations, that hold without need of Casimir modifications, one finds the standard linear response Green-Kubo relations, and consequently the Onsager reciprocal relations. Going beyond linear response is also possible, for instance in terms of transient and steady state Fluctuation Relations (FRs). Here we highlight how the generalized time-reversal symmetries ensure that the (transient) FRs theory directly applies also for systems in external magnetic fields. Furthermore we show that transient FR can indeed be verified in nonequilibrium molecular dynamics simulations, for systems subjected to magnetic and electric fields, which are thermostatted à la Nosé-Hoover. The result is nontrivial because, since it is not immediate within which sizes and time scales the effects can actually be observable, it is not obvious what one may obtain by real molecular dynamics simulations.
△ Less
Submitted 17 October, 2020;
originally announced October 2020.
-
Fluctuation relations for systems in constant magnetic field
Authors:
Alessandro Coretti,
Lamberto Rondoni,
Sara Bonella
Abstract:
The validity of the Fluctuation Relations (FR) for systems in a constant magnetic field is investigated. Recently introduced time-reversal symmetries that hold in presence of static electric and magnetic fields and of deterministic thermostats are used to prove the transient FR without invoking, as commonly done, inversion of the magnetic field. Steady-state FR are also derived, under the t-mixing…
▽ More
The validity of the Fluctuation Relations (FR) for systems in a constant magnetic field is investigated. Recently introduced time-reversal symmetries that hold in presence of static electric and magnetic fields and of deterministic thermostats are used to prove the transient FR without invoking, as commonly done, inversion of the magnetic field. Steady-state FR are also derived, under the t-mixing condition. These results extend the predictive power of important statistical mechanics relations. We illustrate this via the non-linear response for the cumulants of the dissipation, showing how the new FR enable to determine analytically null cumulants also for systems in a single magnetic field.
△ Less
Submitted 11 May, 2020; v1 submitted 8 May, 2020;
originally announced May 2020.
-
Mass-zero constrained molecular dynamics for electrode charges in simulations of electrochemical systems
Authors:
Alessandro Coretti,
Laura Scalfi,
Camille Bacon,
Benjamin Rotenberg,
Rodolphe Vuilleumier,
Giovanni Ciccotti,
Mathieu Salanne,
Sara Bonella
Abstract:
Classical molecular dynamics simulations have recently become a standard tool for the study of electrochemical systems. State-of-the-art approaches represent the electrodes as perfect conductors, modelling their responses to the charge distribution of electrolytes via the so-called fluctuating charge model. These fluctuating charges are additional degrees of freedom that, in a Born-Oppenheimer spi…
▽ More
Classical molecular dynamics simulations have recently become a standard tool for the study of electrochemical systems. State-of-the-art approaches represent the electrodes as perfect conductors, modelling their responses to the charge distribution of electrolytes via the so-called fluctuating charge model. These fluctuating charges are additional degrees of freedom that, in a Born-Oppenheimer spirit, adapt instantaneously to changes in the environment to keep each electrode at a constant potential. Here we show that this model can be treated in the framework of constrained molecular dynamics, leading to a symplectic and time-reversible algorithm for the evolution of all the degrees of freedom of the system. The computational cost and the accuracy of the new method are similar to current alternative implementations of the model. The advantage lies in the accuracy and long term stability guaranteed by the formal properties of the algorithm and in the possibility to systematically introduce additional kinematic conditions of arbitrary number and form. We illustrate the performance of the constrained dynamics approach by enforcing the electroneutrality of the electrodes in a simple capacitor consisting of two graphite electrodes separated by a slab of liquid water.
△ Less
Submitted 10 March, 2020;
originally announced March 2020.
-
Adiabatic motion and statistical mechanics via mass zero constrained dynamics
Authors:
Sara Bonella,
Alessandro Coretti,
Rodolphe Vuilleumier,
Giovanni Ciccotti
Abstract:
In recent work [Coretti et al., The Journal of Chemical Physics, 2018, 149, 191102], a new algorithm to solve numerically the dynamics of the shell model for polarization was presented. The approach, broadly applicable to systems involving adiabatically separated dynamical variables, employs constrained molecular dynamics to strictly enforce the condition that the force on the fast degrees of free…
▽ More
In recent work [Coretti et al., The Journal of Chemical Physics, 2018, 149, 191102], a new algorithm to solve numerically the dynamics of the shell model for polarization was presented. The approach, broadly applicable to systems involving adiabatically separated dynamical variables, employs constrained molecular dynamics to strictly enforce the condition that the force on the fast degrees of freedom, modeled as having zero mass, is null at each time step. The algorithm is symplectic and fully time reversible, and results in stable and efficient propagation. In this paper we complete the discussion of the mechanics of mass zero constrained dynamics by showing how to adapt it to problems where the fast degrees of freedom must satisfy additional conditions. This extension includes, in particular, the important case of first principles molecular dynamics. We then consider the statistical mechanics of the mass zero constrained dynamical system demonstrating that the marginal probability sampled by the dynamics in the physical phase space recovers the form of the Born-Oppenheimer probability density.
△ Less
Submitted 10 January, 2020;
originally announced January 2020.
-
Charge fluctuations from molecular simulations in the constant-potential ensemble
Authors:
Laura Scalfi,
David T. Limmer,
Alessandro Coretti,
Sara Bonella,
Paul A. Madden,
Mathieu Salanne,
Benjamin Rotenberg
Abstract:
We revisit the statistical mechanics of charge fluctuations in capacitors. In constant-potential classical molecular simulations, the atomic charge of electrode atoms are treated as additional degrees of freedom which evolve in time so as to satisfy the constraint of fixed electrostatic potential for each configuration of the electrolyte. The present work clarifies the role of the overall electron…
▽ More
We revisit the statistical mechanics of charge fluctuations in capacitors. In constant-potential classical molecular simulations, the atomic charge of electrode atoms are treated as additional degrees of freedom which evolve in time so as to satisfy the constraint of fixed electrostatic potential for each configuration of the electrolyte. The present work clarifies the role of the overall electroneutrality constraint, as well as the link between the averages computed within the Born-Oppenheimer approximation and that of the full constant-potential ensemble. This allows us in particular to derive a complete fluctuation-dissipation relation for the differential capacitance, that includes a contribution from the charge fluctuations (around the charges satisfying the constant-potential and electroneutrality constraints) also present in the absence of an electrolyte. We provide a simple expression for this contribution from the elements of the inverse of the matrix defining the quadratic form of the fluctuating charges in the energy. We then illustrate numerically the validity of our results, and recover the expected result for an empty capacitor with structureless electrodes at large inter-electrode distances. By considering a variety of liquids between graphite electrodes, we confirm that this contribution to the total differential capacitance is small compared to that induced by the thermal fluctuations of the electrolyte.
△ Less
Submitted 21 November, 2019;
originally announced November 2019.
