Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userA coarse-grained deep neural network model for liquid water
80
0
0.0
(
0
)
Ask ChatGPT about the research
No Arabic abstract
We introduce a coarse-grained deep neural network model (CG-DNN) for liquid water that utilizes 50 rotational and translational invariant coordinates, and is trained exclusively against energies of ~30,000 bulk water configurations. Our CG-DNN potential accurately predicts both the energies and molecular forces of water; within 0.9 meV/molecule and 54 meV/angstrom of a reference (coarse-grained bond-order potential) model. The CG-DNN water model also provides good prediction of several structural, thermodynamic, and temperature dependent properties of liquid water, with values close to that obtained from the reference model. More importantly, CG-DNN captures the well-known density anomaly of liquid water observed in experiments. Our work lays the groundwork for a scheme where existing empirical water models can be utilized to develop fully flexible neural network framework that can subsequently be trained against sparse data from high-fidelity albeit expensive beyond-DFT calculations.
rate research
Read More
Classical density-functional theory provides an efficient alternative to molecular dynamics simulations for understanding the equilibrium properties of inhomogeneous fluids. However, application of density-functional theory to multi-site molecular fluids has so far been limited by complications due to the implicit molecular geometry constraints on the site densities, whose resolution typically requires expensive Monte Carlo methods. Here, we present a general scheme of circumventing this so-called inversion problem: compressed representations of the orientation density. This approach allows us to combine the superior iterative convergence properties of multipole representations of the fluid configuration with the improved accuracy of site-density functionals. Next, from a computational perspective, we show how to extend the DFT++ algebraic formulation of electronic density-functional theory to the classical fluid case and present a basis-independent discretization of our formulation for molecular classical density-functional theory. Finally, armed with the above general framework, we construct a simplified free-energy functional for water which captures the radial distributions, cavitation energies, and the linear and non-linear dielectric response of liquid water. The resulting approach will enable efficient and reliable first-principles studies of atomic-scale processes in contact with solution or other liquid environments.
Integral equation theory is applied to a coarse-grained model of water to study potential of mean force between hydrophobic solutes. Theory is shown to be in good agreement with the available simulation data for methane-methane and fullerene-fullerene potential of mean force in water; the potential of mean force is also decomposed into its entropic and enthalpic contributions. Mode coupling theory is employed to compute self-diffusion coefficient of water, as well as diffusion coefficient of a dilute hydrophobic solute; good agreement with molecular dynamics simulation results is found.
By means of Metropolis Monte Carlo simulations of a coarse-grained model for flexible polymers, we investigate how the integrated autocorrelation times of different energetic and structural quantities depend on the temperature. We show that, due to critical slowing down, an extremal autocorrelation time can also be considered as an indicator for the collapse transition that helps to locate the transition point. This is particularly useful for finite systems, where response quantities such as the specific heat do not necessarily exhibit clear indications for pronounced thermal activity.
Water modeling is a challenging problem. Its anomalies are difficult to reproduce, promoting the proliferation of a large number of computational models, among which researchers select the most appropriate for the property they study. In this chapter, we introduce a coarse-grained model introduced by Franzese and Stanley (FS) that accounts for the many-body interactions of water. We review mean-field calculations and Monte Carlo simulations on water monolayers for a wide range of pressures and temperatures, including extreme conditions. The results show the presence of two dynamic crossovers and explain the origin of diffusion anomalies. Moreover, the model shows that all the different scenarios, proposed in the last decades as alternative explanations of the experimental anomalies of water, can be related by the fine-tuning of the many-body (cooperative) interaction. Once this parameter is set from the experiments, the FS model predicts a phase transition between two liquids with different densities and energies in the supercooled water region, ending in a liquid-liquid critical point. From this critical point stems a liquid-liquid Widom line, i.e., the locus of maxima of the water correlation length, that in the FS model can be directly calculated. The results are consistent with the extrapolations from experiments. Furthermore, they agree with those from atomistic models but make predictions over a much wider thermodynamic region, allowing for a better interpretation of the available experimental data. All these findings provide a coherent picture of the properties of water and confirm the validity of the FS model that has proved to be useful for large-scale simulations of biological systems.
Small metal clusters are of fundamental scientific interest and of tremendous significance in catalysis. These nanoscale clusters display diverse geometries and structural motifs depending on the cluster size; a knowledge of this size-dependent structural motifs and their dynamical evolution has been of longstanding interest. Classical MD typically employ predefined functional forms which limits their ability to capture such complex size-dependent structural and dynamical transformation. Neural Network (NN) based potentials represent flexible alternatives and in principle, well-trained NN potentials can provide high level of flexibility, transferability and accuracy on-par with the reference model used for training. A major challenge, however, is that NN models are interpolative and requires large quantities of training data to ensure that the model adequately samples the energy landscape both near and far-from-equilibrium. Here, we introduce an active learning (AL) scheme that trains a NN model on-the-fly with minimal amount of first-principles based training data. Our AL workflow is initiated with a sparse training dataset (1 to 5 data points) and is updated on-the-fly via a Nested Ensemble Monte Carlo scheme that iteratively queries the energy landscape in regions of failure and updates the training pool to improve the network performance. Using a representative system of gold clusters, we demonstrate that our AL workflow can train a NN with ~500 total reference calculations. Our NN predictions are within 30 meV/atom and 40 meV/AA of the reference DFT calculations. Moreover, our AL-NN model also adequately captures the various size-dependent structural and dynamical properties of gold clusters in excellent agreement with DFT calculations and available experiments.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
