An official website of the United States government
The development of reliable, yet computationally efficient interatomic forcefields is key to facilitate the modeling of glasses. However, the parameterization of novel forcefields is challenging as the high number of parameters renders traditional optimization methods inefficient or subject to bias. Here, we present a new parameterization method based on machine learning, which combines ab initio molecular dynamics simulations and Bayesian optimization. By taking the example of glassy silica, we show that our method yields a new interatomic forcefield that offers an unprecedented agreement with ab initio simulations. This method offers a new route to efficiently parameterize new interatomic forcefields for disordered solids in a non-biased fashion.
more »
« less
INDEEDopt: a deep learning-based ReaxFF parameterization framework
Abstract Empirical interatomic potentials require optimization of force field parameters to tune interatomic interactions to mimic ones obtained by quantum chemistry-based methods. The optimization of the parameters is complex and requires the development of new techniques. Here, we propose an INitial-DEsign Enhanced Deep learning-based OPTimization (INDEEDopt) framework to accelerate and improve the quality of the ReaxFF parameterization. The procedure starts with a Latin Hypercube Design (LHD) algorithm that is used to explore the parameter landscape extensively. The LHD passes the information about explored regions to a deep learning model, which finds the minimum discrepancy regions and eliminates unfeasible regions, and constructs a more comprehensive understanding of physically meaningful parameter space. We demonstrate the procedure here for the parameterization of a nickel–chromium binary force field and a tungsten–sulfide–carbon–oxygen–hydrogen quinary force field. We show that INDEEDopt produces improved accuracies in shorter development time compared to the conventional optimization method.
more »
« less
- Award ID(s):
- 1660477
- PAR ID:
- 10229579
- Publisher / Repository:
- Nature Publishing Group
- Date Published:
- Journal Name:
- npj Computational Materials
- Volume:
- 7
- Issue:
- 1
- ISSN:
- 2057-3960
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Yamashita, Y.; Kano, M. (Ed.)In this work surrogate assisted optimization is utilized to calibrate predictive molecular models, called force fields, used in molecular simulations to reproduce the liquid density of a hydrofluorocarbon refrigerant molecule. A previous calibration workflow which relied on Gaussian process regression models and large Latin hypercube samples to screen force field parameter space is extended to include Bayesian optimization methods to efficiently guide the search for force field parameters. In comparison to the previous work, the Bayesian-based calibration workflow finds a parameter set which results in a lower objective function value than the original workflow after evaluating approximately 50% fewer parameter sets. It is envisioned that this updated workflow will facilitate rapid force field optimization enabling screening of vast molecular design space.more » « less
-
Molecular dynamics at the atomistic scale is increasingly being used to predict material properties and speed up the materials design and development process. However, the accuracy of molecular dynamics predictions is sensitively dependent on the force fields. In the traditional force field calibration process, a specific property, predicted by the model, is compared with the experimental observation and the force field parameters are adjusted to minimize the difference. This leads to the issue that the calibrated force fields are not generic and robust enough to predict different properties. Here, a new calibration method based on multi-objective Bayesian optimization is developed to speed up the development of molecular dynamics force fields that are capable of predicting multiple properties accurately. This is achieved by reducing the number of simulation runs to generate the Pareto front with an efficient sequential sampling strategy. The methodology is demonstrated by generating a new coarse-grained force field for polycaprolactone, where the force field can predict mechanical properties and water diffusion in the polymer.more » « less
-
Lian, T (Ed.)The fast and accurate simulation of chemical reactions is a major goal of computational chemistry. Recently, the pursuit of this goal has been aided by machine learning interatomic potentials (MLIPs), which provide energies and forces at quantum mechanical accuracy but at a fraction of the cost of the reference quantum mechanical calculations. Assembling the training set of relevant configurations is key to building the MLIP. Here, we demonstrate two approaches to training reactive MLIPs based on reaction pathway information. One approach exploits reaction datasets containing reactant, product, and transition state structures. Using an SN2 reaction dataset, we accurately locate reaction pathways and transition state geometries of up to 170 unseen reactions. In another approach, which does not depend on data availability, we present an efficient active learning procedure that yields an accurate MLIP and converged minimum energy path given only the reaction end point structures, avoiding quantum mechanics driven reaction pathway search at any stage of training set construction. We demonstrate this procedure on an SN2 reaction in the gas phase and with a small number of solvating water molecules, predicting reaction barriers within 20 meV of the reference quantum chemistry method. We then apply the active learning procedure on a more complex reaction involving a nucleophilic aromatic substitution and proton transfer, comparing the results against the reactive ReaxFF force field. Our active learning procedure, in addition to rapidly finding reaction paths for individual reactions, provides an approach to building large reaction path databases for training transferable reactive machine learning potentials.more » « less
-
Abstract Morphogenetic programs coordinate cell signaling and mechanical interactions to shape organs. In systems and synthetic biology, a key challenge is determining optimal cellular interactions for predicting organ shape, size, and function. Physics-based models defining the subcellular force distribution facilitate this, but it is challenging to calibrate parameters in these models from data. To solve this inverse problem, we created a Bayesian optimization framework to determine the optimal cellular force distribution such that the predicted organ shapes match the experimentally observed organ shapes. This integrative framework employs Gaussian Process Regression, a non-parametric kernel-based probabilistic machine learning modeling paradigm, to learn the mapping functions relating to the morphogenetic programs that maintain the final organ shape. We calibrated and tested the method onDrosophilawing imaginal discs to study mechanisms that regulate epithelial processes ranging from development to cancer. The parameter estimation framework successfully infers the underlying changes in core parameters needed to match simulation data with imaging data of wing discs perturbed with collagenase. The computational pipeline identifies distinct parameter sets mimicking wild-type shapes. It enables a global sensitivity analysis to support the regulation of actomyosin contractility and basal ECM stiffness to generate and maintain the curved shape of the wing imaginal disc. The optimization framework, combined with experimental imaging, identified that Piezo, a mechanosensitive ion channel, impacts fold formation by regulating the apical-basal balance of actomyosin contractility and elasticity of ECM. This workflow is extensible toward reverse-engineering morphogenesis across organ systems and for real-time control of complex multicellular systems.more » « less
