Polarizable Water Model with Ab Initio Neural Network Dynamic Charges and Spontaneous Charge Transfer

Journal of Chemical Theory and Computation, Год журнала: 2025, Номер unknown

Опубликована: Март 29, 2025

Simulating water accurately has been a challenge due to the complexity of describing polarization and intermolecular charge transfer. Quantum mechanical (QM) electronic structures provide an accurate description in response local environments, which is nevertheless too expensive for large systems. In this study, we have developed polarizable model integrating Charge Model 5 atomic charges at level second-order Mo̷ller–Plesset perturbation theory, predicted by transferable neural network (ChargeNN) model. The spontaneous transfer explicitly accounted for, enabling precise treatment hydrogen bonds out-of-plane polarization. Our ChargeNN successfully reproduces various properties gas, liquid, solid phases. For example, correctly captures hydrogen-bond stretching peak bending-libration combination band, are absent spectra using fixed charges, highlighting significance Finally, molecular dynamical simulations liquid droplet with ∼4.5 nm radius reveal that strong interfacial electric fields concurrently induced partial collapse surface-to-interior study paves way QM-polarizable force fields, aiming large-scale high accuracy.

Язык: Английский

---

Transferability of Buckingham Parameters for Short-Range Repulsion between Topological Atoms

The Journal of Physical Chemistry A, Год журнала: 2024, Номер 128(22), С. 4561 - 4572

Опубликована: Май 28, 2024

The repulsive part of the Buckingham potential, with parameters

Язык: Английский

---

Incorporating Noncovalent Interactions in Transfer Learning Gaussian Process Regression Models for Molecular Simulations

Journal of Chemical Theory and Computation, Год журнала: 2024, Номер 20(14), С. 5994 - 6008

Опубликована: Июль 9, 2024

FFLUX is a quantum chemical topology-based multipolar force field that uses Gaussian process regression machine learning models to predict atomic energies and multipole moments on the fly for fast accurate molecular dynamics simulations. These have previously been trained monomers, meaning many-body effects, example, intermolecular charge transfer, are missed in Moreover, dispersion repulsion modeled using Lennard-Jones potentials, necessitating careful parametrization. In this work, we take an important step toward addressing these shortcomings show clusters, case, dimer, can be used simulations by preparing benchmarking formamide dimer model. To mitigate computational costs associated with training higher-dimensional models, rely transfer of hyperparameters from smaller source model larger target model, enabling order magnitude faster than direct approach. The allows account two-body including polarization penetration, do not require nonbonded potentials. We limitations closer mechanics possible monomeric models.

Язык: Английский

---

Modeling Many-Body Interactions in Water with Gaussian Process Regression

The Journal of Physical Chemistry A, Год журнала: 2024, Номер 128(42), С. 9345 - 9351

Опубликована: Окт. 11, 2024

We report a first-principles water dimer potential that captures many-body interactions through Gaussian process regression (GPR). Modeling is upgraded from previous work by using custom kernel function implemented the KeOps library, allowing for much larger GPR models to be constructed and interfaced with next-generation machine learning force field FFLUX. A new synthetic data set, called WD24, used model training. The resulting can predict 90% of geometries within chemical accuracy test set in simulation. curvature energy surface captured models, successful geometry optimization completed total error just 2.6 kJ mol–1, starting structure where molecules are separated nearly 4.3 Å. Dimeric modeling flexible, noncrystalline system FFLUX shown first time.

Язык: Английский

---

Going beyond the Ordered Bulk: A Perspective on the Use of the Cambridge Structural Database for Predictive Materials Design

Crystal Growth & Design, Год журнала: 2024, Номер 24(17), С. 6911 - 6930

Опубликована: Авг. 19, 2024

When Olga Kennard founded the Cambridge Crystallographic Data Centre in 1965, Structural Database was a pioneering attempt to collect scientific data standard format. Since then, it has evolved into an indispensable resource contemporary molecular materials science, with over 1.25 million structures and comprehensive software tools for searching, visualizing analyzing data. In this perspective, we discuss use of CSD CCDC address multiscale challenge predictive design. We provide overview core capabilities demonstrate their application range design problems recent case studies drawn from topical research areas, focusing particular on mining machine learning techniques. also identify several challenges that can be addressed existing or through new varying levels development effort.

Язык: Английский

---

Transfer learning of hyperparameters for fast construction of anisotropic GPR models: design and application to the machine-learned force field FFLUX

Physical Chemistry Chemical Physics, Год журнала: 2024, Номер 26(36), С. 23677 - 23691

Опубликована: Янв. 1, 2024

The polarisable machine-learned force field FFLUX requires pre-trained anisotropic Gaussian process regression (GPR) models of atomic energies and multipole moments to propagate unbiased molecular dynamics simulations. outcome simulations is highly dependent on the predictive accuracy underlying whose training entails determining optimal set model hyperparameters. Unfortunately, traditional direct learning (DL) procedures do not scale well this task, especially when hyperparameter search initiated from a (set of) random guess solution(s). Additionally, complexity space (HS) increases with number geometrical input features, at least for kernels, making optimization hyperparameters even more challenging. In study, we propose transfer (TL) protocol that accelerates GPR by facilitating access promising regions HS. based seeding-relaxation mechanism in which an excellent solution identified rapidly building one or several small source over subset target before readjusting previous entire set. We demonstrate performance assessing DL TL charges various conformations benzene, ethanol, formic acid dimer drug fomepizole. Our experiments suggest can be built order magnitude faster while preserving quality their analogs. Most importantly, deployed simulations, compete outperform analogs it comes performing geometry computing harmonic vibrational modes.

Язык: Английский

---

Toward Gaussian Process Regression Modeling of a Urea Force Field

The Journal of Physical Chemistry A, Год журнала: 2024, Номер unknown

Опубликована: Сен. 20, 2024

FFLUX is a next-generation, machine-learnt force field built on three cornerstones: quantum chemical topology, Gaussian process regression, and (high-rank) multipolar electrostatics. It capable of performing molecular dynamics with near-quantum accuracy at lower computational cost than standard

Язык: Английский

---