L1 Regularization-Based Model Reduction of Complex Chemistry Molecular Dynamics for Statistical Learning of Kinetic Monte Carlo Models

MRS Advances ◽  
2016 ◽  
Vol 1 (24) ◽  
pp. 1767-1772 ◽  
Author(s):  
Qian Yang ◽  
Carlos A. Sing-Long ◽  
Evan J. Reed
Keyword(s):  
Molecular Dynamics ◽  
Monte Carlo ◽  
Kinetic Monte Carlo ◽  
A Priori ◽  
Reaction Network ◽  
Md Simulations ◽  
Rate Coefficients ◽  
Elementary Reactions ◽  
L1 Regularization ◽  
Complex Chemistry

ABSTRACTKinetic Monte Carlo (KMC) methods have been a successful technique for accelerating time scales and increasing system sizes beyond those achievable with fully atomistic simulations. However, a requirement for its success is a priori knowledge of all relevant reaction pathways and their rate coefficients. This can be difficult for systems with complex chemistry, such as shock-compressed materials at high temperatures and pressures or phenolic spacecraft heat shields undergoing pyrolysis, which can consist of hundreds of molecular species and thousands of distinct reactions. In this work, we develop a method for first estimating a KMC model composed of elementary reactions and rate coefficients by using large datasets derived from a few molecular dynamics (MD) simulations of shock compressed liquid methane, and then using L1 regularization to reduce the estimated chemical reaction network. We find that the full network of 2613 reactions can be reduced by 89% while incurring approximately 9% error in the dominant species (CH4) population. We find that the degree of sparsity achievable decreases when similar accuracy is required for additional populations of species.

scholarly journals Learning reduced kinetic Monte Carlo models of complex chemistry from molecular dynamics

2017 ◽  
Vol 8 (8) ◽  
pp. 5781-5796 ◽  
Author(s):  
Qian Yang ◽  
Carlos A. Sing-Long ◽  
Evan J. Reed
Keyword(s):  
Molecular Dynamics ◽  
Monte Carlo ◽  
Molecular Dynamics Simulations ◽  
Statistical Learning ◽  
Large Scale ◽  
Kinetic Monte Carlo ◽  
Reaction Networks ◽  
Learning Framework ◽  
Dynamics Simulations ◽  
Complex Chemistry

We propose a novel statistical learning framework for automatically and efficiently building reduced kinetic Monte Carlo (KMC) models of large-scale elementary reaction networks from data generated by a single or few molecular dynamics simulations (MD).

Coupled Kinetic Monte Carlo and Molecular Dynamics Simulations of Implant Damage Accumulation in Silicon

2003 ◽  
Vol 792 ◽  
Author(s):  
Gustav Otto ◽  
Gerhard Hobler
Keyword(s):  
Molecular Dynamics ◽  
Monte Carlo ◽  
Ion Implantation ◽  
Damage Accumulation ◽  
Kinetic Monte Carlo ◽  
Md Simulations ◽  
Implantation Damage ◽  
Complex Process ◽  
Light Ions ◽  
Dynamics Simulations

ABSTRACTDamage formation during ion implantation is a complex process that cannot accurately be modeled by binary collision simulations alone. Molecular dynamics (MD) simulations are suited to describe the quenching of collision cascades, while thermally activated processes may be treated with the kinetic Monte Carlo (kMC) method.MD and kMC simulations have widely but separately been used to investigate damage accumulation and annealing in silicon. Simulation of ion implantation at room temperature, however, requires both methods to be coupled. In this paper we describe for the first time a scheme of the coupling between MD and lattice kMC for damage accumulation. Using this scheme we study the dynamic annealing behavior of implantation damage for heavy and light ions.

Coarse kMC-based replica exchange algorithms for the accelerated simulation of protein folding in explicit solvent

2016 ◽  
Vol 18 (18) ◽  
pp. 13052-13065 ◽  
Author(s):  
Emanuel K. Peter ◽  
Joan-Emma Shea ◽  
Igor V. Pivkin
Keyword(s):  
Molecular Dynamics ◽  
Monte Carlo ◽  
Protein Folding ◽  
Kinetic Monte Carlo ◽  
Replica Exchange ◽  
Replica Exchange Molecular Dynamics ◽  
Accelerated Simulation ◽  
Explicit Solvent ◽  
Exchange Algorithms

In this paper, we present a coarse replica exchange molecular dynamics (REMD) approach, based on kinetic Monte Carlo (kMC).

Kinetic Monte Carlo simulations of cascades in Fe alloys

2000 ◽  
Vol 650 ◽  
Author(s):  
C. Domain ◽  
C.S. Becquart ◽  
J.C. van Duysen
Keyword(s):  
Monte Carlo ◽  
Kinetic Monte Carlo ◽  
Md Simulations ◽  
Atom Probe ◽  
Pressurized Water Reactor ◽  
Pressurized Water ◽  
Displacement Cascades ◽  
Cu Content ◽  
Primary Damage ◽  
Eam Potential

ABSTRACTThe Pressurized Water Reactor vessel steels are embrittled by neutron irradiation. Among the solute atoms, copper play an important role in the embrittlement and different Cu-rich defects have been experimentally observed to form. We have investigated by Kinetic Monte Carlo (KMC) on rigid lattices the evolution of the primary damage. Since the point defects created by the displacement cascades have very different kinetics, their evolution is tracked in two steps. In a first step, we have studied their recombination in the cascade region and the formation of interstitial clusters using “object diffusion”. The parameters of this model are based on MD simulations, or on first principles calculations. In a second part, we have investigated the subsequent evolution of the primary damage with a model based on a vacancy jump mechanism. These simulations which rely on an adapted EAM potential show the formation of copper rich defects. Some of the potential's predictions that played a key role in the model were checked by ab initio calculations. The defects obtained from these simulations, subsequent to the primary damage created by displacement cascades, exhibit similarities with the ones observed by atom probe. The influence of temperature and Cu content on the final damage was investigated.

scholarly journals Sorption, Structure and Dynamics of CO2 and Ethane in Silicalite at High Pressure: A Combined Monte Carlo and Molecular Dynamics Simulation Study

Molecules ◽  
2018 ◽  
Vol 24 (1) ◽  
pp. 99 ◽  
Author(s):  
Siddharth Gautam ◽  
Tingting Liu ◽  
David Cole
Keyword(s):  
Molecular Dynamics ◽  
Monte Carlo ◽  
Time Scales ◽  
Rotational Motion ◽  
High Pressures ◽  
Sorption Isotherms ◽  
Md Simulations ◽  
Dynamics Simulation ◽  
Structure And Dynamics ◽  
Low Pressures

Silicalite is an important nanoporous material that finds applications in several industries, including gas separation and catalysis. While the sorption, structure, and dynamics of several molecules confined in the pores of silicalite have been reported, most of these studies have been restricted to low pressures. Here we report a comparative study of sorption, structure, and dynamics of CO2 and ethane in silicalite at high pressures (up to 100 bar) using a combination of Monte Carlo (MC) and molecular dynamics (MD) simulations. The behavior of the two fluids is studied in terms of the simulated sorption isotherms, the positional and orientational distribution of sorbed molecules in silicalite, and their translational diffusion, vibrational spectra, and rotational motion. Both CO2 and ethane are found to exhibit orientational ordering in silicalite pores; however, at high pressures, while CO2 prefers to reside in the channel intersections, ethane molecules reside mostly in the sinusoidal channels. While CO2 exhibits a higher self-diffusion coefficient than ethane at low pressures, at high pressures, it becomes slower than ethane. Both CO2 and ethane exhibit rotational motion at two time scales. At both time scales, the rotational motion of ethane is faster. The differences observed here in the behavior of CO2 and ethane in silicalite pores can be seen as a consequence of an interplay of the kinetic diameter of the two molecules and the quadrupole moment of CO2.

Combining Molecular Dynamics and Monte Carlo Simulations to Model Chemical Vapor Deposition: Application to Diamond

1992 ◽  
Vol 278 ◽  
Author(s):  
D.W. Brenner ◽  
D.H. Robertson ◽  
R.J. Carty ◽  
D. Srivastava ◽  
B.J. Garrison
Keyword(s):  
Molecular Dynamics ◽  
Monte Carlo ◽  
Chemical Vapor Deposition ◽  
Vapor Deposition ◽  
Surface Reactions ◽  
Chemical Vapor ◽  
Md Simulations ◽  
Monte Carlo Algorithm ◽  
Surface Collision ◽  
Classical Trajectories

AbstractGas-surface reactions of the type that contribute to growth during the chemical vapor deposition (CVD) of diamond films are generally completed in picoseconds, well within timescales accessible by molecular dynamics (MD) simulations. For low-pressure deposition, however, the time between collisions for a surface site can be microseconds, which makes direct modeling of CVD crystal growth impossible using standard MD methods. To effectively bridge this discrepancy in timescales, the gas-surface reactions can be modeled using MD trajectories, and then this data can be used to define probabilities in a Monte Carlo algorithm where each step represents a gas-surface collision. We illustrate this approach using the reaction of atomic hydrogen with a diamond (111) surface as an example, where we use abstraction and sticking probabilities generated using classical trajectories in a simple Monte Carlo algorithm to determine the number of open sites as a function of temperature. We also include models for the thermal desorption of hydrogen that predict that growth temperatures are not restricted by the thermal loss of chemisorbed hydrogen.

Sign in / Sign up

Export Citation Format

Share Document