Four-dimensional spacetime atomistic artificial intelligence models

Abstract

We demonstrate that artificial intelligence (AI) can learn four-dimensional (4D) atomistic systems in the spacetime continuum. Given the initial conditions – nuclear positions and velocities at time zero – the proposed 4D-atomistic AI (4D-A2I) models can predict nuclear positions at any time in the future or past for the simplest systems as we show for H2. For larger polyatomic molecules, AI is capable of learning distant but finite future as we demonstrate for an ethanol molecule. 4D-A2I models provide direct access to a multitude of properties at a given time such as geometries, velocities, forces, and energies which can be used in simulating physicochemical transformations and spectra. Our approach can be used as a cost-efficient alternative to traditional molecular dynamics. We show an example of a 4D-A2I model describing the dynamical behavior of ethanol at the coupled-cluster level with the speed of one nanosecond simulation time per one hour wall-clock time on a single GPU card – a previously unachievable feat with traditional Born–Oppenheimer molecular dynamics. 4D-A2I model is also demonstrated to provide direct access to atomistic time-resolved details of physicochemical transformations.

Automatic Evolution of Machine-Learning-Based Quantum Dynamics with Uncertainty Analysis

“…Several previous works have already demonstrated the possibility to use various ML models (FCNN, KRR, CNN, and so on) to propagate quantum dynamics. − For instance, our own short work proposed using the RNN to realize such an idea and give the initial discussion on the importance of the model uncertainty. However, it is necessary to know when to stop the initial dynamics propagation and start to build the ML model.…”

“…With the rapid development of computer artificial intelligence, machine learning (ML) plays more and more important roles in theoretical chemistry, including the construction of the molecular Hamiltonian, − the analysis of trajectory-based molecular dynamics evolution, − the prediction of molecular properties and chemical reactions, − and the design of novel functional materials. , In addition to these, considerable efforts were made to employ the ML approach to simulate dynamics evolutions recently. − …”

Trajectory Propagation of Symmetrical Quasi-classical Dynamics with Meyer-Miller Mapping Hamiltonian Using Machine Learning

“…These works show the great potential of the ML application in nonadiabatic quantum dynamics. Besides, the ML approach was also used to propagate the classical dynamics of molecular systems. ,, …”

“…Besides, the ML approach was also used to propagate the classical dynamics of molecular systems. 22,49,50 In the simulation of the nonadiabatic dynamics, trajectorybased dynamics methods received the greatest attention, 29,36−39,51−59 because they can be used to treat highdimensional systems with complicated molecular motions at all atomic levels. Among them, the symmetrical quasi-classical dynamics method based on the Meyer-Miller mapping Hamiltonian (MM-SQC) 39,60 was proven to be a promising method that gives the proper description of the nonadiabatic dynamics in both the model and the realistic molecular systems.…”

“…With the rapid development of computer facilities and artificial intelligence sciences in recent years, machine learning (ML) methods are extensively used as the important auxiliary tools in theoretical simulations of chemical dynamics, such as the construction of the potential energy surfaces, − the automatic analysis of molecular dynamics evolution, − and the numerical solution of the dynamical equation. − Among these works, considerable efforts were made to employ the various ML approaches as the supplement but powerful computational tools to propagate the nonadiabatic dynamics evolutions, ,− ,,, in which the strong electronic-nuclear couplings exist.…”