Abstract:We present new results on the underlying guest-host interactions and spectral characterization of a CO2 molecule confined in the cages of the sI clathrate hydrate. Such types of porous solids raise computational challenges, as they are of practical interest as gas storage/capture materials. Accordingly, we have directed our efforts toward addressing their modeling in a proper manner, ensuring the quality of the input data and the efficiency of the computational approaches. The computational procedure for spectral simulations, within the multi-configurational time-dependent Hartree framework, involves the development of a fully coupled Hamiltonian, including an exact kinetic energy operator and a many-body representation of the potential, along with dipole moment surfaces, both obtained through neural network machine learning techniques. The resulting models were automatically trained and tested on extensive datasets generated by PW86PBE-XDM calculations, following the outcome of previous benchmark studies. Our simulations enable us to explore various aspects of the quantized dynamics upon confinement of CO2@D/T, such as constrained rotational-translational quantum motions and the averaged position/orientation of the CO2 guest in comparison to the experimental data available. Particularly notable are the distinct energy patterns observed in the computed spectra for the confined CO2 in the D and T cages, with a considerably high rotational-translational coupling in the CO2@T case. Leveraging reliable computations has proved instrumental, highlighting the sensitivity of the spectral features to the shape and strength of the potential interactions, with the explicit description of many-body contributions being significant.

Multi-dimensional neural network interatomic potentials for CO on NaCl(100)

Ab initio Accuracy Neural Network Potential for Drug-like Molecules

Temperature-transferable coarse-graining of ionic liquids with dual graph convolutional neural networks

The Potential of Neural Network Potentials

Quantum-Enhanced Neural Exchange-Correlation Functionals

Extending the atomic decomposition and many-body representation, a chemistry-motivated monomer-centered approach for machine learning potentials

Neural Network Potential with Multi-Resolution Approach Enables Accurate Prediction of Reaction Free Energies in Solution

Modeling Intermolecular Interactions With XDM Dispersion Corrections to Neural Network Potentials

AIMNet2: A Neural Network Potential to Meet your Neutral, Charged, Organic, and Elemental-Organic Needs

Neural Network Interatomic Potentials For Open Surface Nano-mechanics Applications

Ab Initio Molecular Dynamics Simulations of Atmospheric Molecular Clusters Boosted by Neural Networks

Analyzing Atomic Interactions in Molecules as Learned by Neural Networks

A deep learning interatomic potential developed for atomistic simulation of carbon materials

Global Neural Network Potential with Explicit Many-Body Functions for Improved Descriptions of Complex Potential Energy Surface.

Peering inside the black box: Learning the relevance of many-body functions in Neural Network potentials

Transferable Machine Learning Interatomic Potential for Carbon Hydrogen Systems

Spline-based neural network interatomic potentials: blending classical and machine learning models

A high-dimensional neural network potential for molecular dynamics simulations of condensed phase nickel and phase transitions

CO2 inside sI clathrate-like cages: Automated construction of neural network/machine learned guest-host potential and quantum spectra computations

Ab initio quantum chemistry with neural-network wavefunctions

MolNet‐3D: Deep Learning of Molecular Representations and Properties from 3D Topography