TimeSOAP: Tracking high-dimensional fluctuations in complex molecular systems via time-variations of SOAP spectra

Abstract

Many molecular systems and physical phenomena are controlled by local fluctuations and microscopic dynamical rearrangements of the constitutive interacting units that are often difficult to detect. This is the case, for example, of phase transitions, phase equilibria, nucleation events, and defect propagation, to mention a few. A detailed comprehension of local atomic environments and of their dynamic rearrangements is essential to understand such phenomena, and also to draw structure-property relationships useful to unveil how to control complex molecular systems. Considerable progresses in the development of advanced high-dimensional structural descriptors (e.g., Smooth Overlap of Atomic Position (SOAP), etc.) have certainly enhanced the representation of atomic-scale simulations data. However, despite such efforts, local dynamic environment rearrangements remain still difficult to elucidate. Here, exploiting the structural-rich description of atomic environments of SOAP and building on the concept of time-dependent local variations, we developed a time-dependent SOAP-based descriptor, TimeSOAP (tSOAP), which essentially tracks the time variations in the local SOAP environments surrounding each molecule (i.e., each SOAP center) in complex molecular systems along ensemble trajectories. We demonstrate how analysis of the time-series {tSOAP data and of their time-derivatives allows detecting dynamics domains and tracking instantaneous changes of local atomic arrangements (i.e., local fluctuations) in a variety of molecular systems. The approach is simple and general, and we expect will help to shed light on a variety of complex dynamical phenomena.

arXiv comment20 pages, 9 figures

References in corpus (23)

• SciPy 1.0--Fundamental Algorithms for Scientific Computing in Python
• Array Programming with NumPy
• Gaussian Approximation Potentials: the accuracy of quantum mechanics, without the electrons
• On representing chemical environments
• DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics
• Structure identification methods for atomistic simulations of crystalline materials
• Identification of slow molecular order parameters for Markov model construction
• Pressure control using stochastic cell rescaling
• Comparing molecules and solids across structural and alchemical space
• Enhanced Sampling in Molecular Dynamics Using Metadynamics, Replica-Exchange, and Temperature-Acceleration
• Estimating the intrinsic dimension of datasets by a minimal neighborhood information
• Efficient and Accurate Machine-Learning Interpolation of Atomic Energies in Compositions with Many Species
• Atom-Density Representations for Machine Learning
• Unsupervised machine learning in atomistic simulations, between predictions and understanding
• Graph Dynamical Networks for Unsupervised Learning of Atomic Scale Dynamics in Materials
• Discovering the Building Blocks of Atomic Systems using Machine Learning
• Unravelling the Origins of Ice Nucleation on Organic Crystals
• Exploring the robust extrapolation of high-dimensional machine learning potentials
• Predicting the phase diagram of titanium dioxide with random search and pattern recognition
• Detecting dynamic domains and local fluctuations in complex molecular systems via timelapse neighbors shuffling
• A Physically-informed Graph-based Order Parameter for the Universal Characterization of Atomic Structures
• Innate Dynamics and Identity Crisis of a Metal Surface Unveiled by Machine Learning of Atomic Environments
• The microscopic mechanism of bulk melting of ice

Cited by in corpus (7)

• "Layer-by-layer" Unsupervised Clustering of Statistically Relevant Fluctuations in Noisy Time-series Data of Complex Dynamical Systems
• Reconstructions and Dynamics of $\beta$-Lithium Thiophosphate Surfaces
• Data-driven assessment of optimal spatiotemporal resolutions for information extraction in noisy time series data
• Relevant, hidden, and frustrated information in high-dimensional analyses of complex dynamical systems with internal noise
• Change Point Detection of Events in Molecular Simulations using dupin
• A data driven approach to classify descriptors based on their efficiency in translating noisy trajectories into physically-relevant information
• Maximum Information Extraction Via Clustering and Minimization of Shannon Entropy