Three parallel algorithms for classical molecular dynamics are presented. The first assigns each processor a fixed subset of atoms; the second assigns each a fixed subset of inter-atomic forces to compute; the third assigns each a fixed spatial region. The algorithms are suitable for molecular dynamics models which can be difficult to parallelize efficiently—those with short-range forces where the neighbors of each atom change rapidly. They can be implemented on any distributed-memory parallel machine which allows for message-passing of data between independently executing processors. The algorithms are tested on a standard Lennard-Jones benchmark problem for system sizes ranging from 500 to 100,000,000 atoms on several parallel supercomputers--the nCUBE 2, Intel iPSC/860 and Paragon, and Cray T3D. Comparing the results to the fastest reported vectorized Cray Y-MP and C90 algorithm shows that the current generation of parallel machines is competitive with conventional vector supercomputers even for small problems. For large problems, the spatial algorithm achieves parallel efficiencies of 90% and a 1840-node Intel Paragon performs up to 165 faster than a single Cray C9O processor. Trade-offs between the three algorithms and guidelines for adapting them to more complex molecular dynamics simulations are also discussed.

Fast parallel algorithms for short-range molecular dynamics

: The unpolarized absorption and circular dichroism spectra of the fundamental vibrational transitions of the chiral molecule, 4-methyl-2-oxetanone, are calculated ab initio. Harmonic force fields are obtained using Density Functional Theory (DFT), MP2, and SCF methodologies and a 5S4P2D/3S2P (TZ2P) basis set. DFT calculations use the Local Spin Density Approximation (LSDA), BLYP, and Becke3LYP (B3LYP) density functionals. Mid-IR spectra predicted using LSDA, BLYP, and B3LYP force fields are of significantly different quality, the B3LYP force field yielding spectra in clearly superior, and overall excellent, agreement with experiment. The MP2 force field yields spectra in slightly worse agreement with experiment than the B3LYP force field. The SCF force field yields spectra in poor agreement with experiment.The basis set dependence of B3LYP force fields is also explored: the 6-31G* and TZ2P basis sets give very similar results while the 3-21G basis set yields spectra in substantially worse agreements with experiment. jg

Ab initio calculation of vibrational absorption and circular dichroism spectra using density functional force fields

The intrinsic reaction coordinate (IRC) approach has been used extensively in quantum chemical analysis and prediction of the mechanism of chemical reactions. The IRC gives a unique connection from a given transition structure to local minima of the reactant and product sides. This allows for easy understanding of complicated multistep mechanisms as a set of simple elementary reaction steps. In this article, three topics concerning the IRC approach are discussed. In the first topic, the first ab initio study of the IRC and a recent development of an IRC calculation algorithm for enzyme reactions are introduced. In the second topic, cases are presented in which dynamical trajectories bifurcate and corresponding IRC connections can be inaccurate. In the third topic, a recent development of an automated reaction path search method and its application to systematic construction of IRC networks are described. Finally, combining these three topics, future perspectives are discussed. © 2014 Wiley Periodicals, Inc.

https://onlinelibrary.wiley.com/doi/pdf/10.1002/qua.24757

Intrinsic reaction coordinate: Calculation, bifurcation, and automated search

Since their introduction about 25 years ago, machine learning (ML) potentials have become an important tool in the field of atomistic simulations After the initial decade, in which neural networks were successfully used to construct potentials for rather small molecular systems, the development of high-dimensional neural network potentials (HDNNPs) in 2007 opened the way for the application of ML potentials in simulations of large systems containing thousands of atoms To date, many other types of ML potentials have been proposed continuously increasing the range of problems that can be studied In this review, the methodology of the family of HDNNPs including new recent developments will be discussed using a classification scheme into four generations of potentials, which is also applicable to many other types of ML potentials The first generation is formed by early neural network potentials designed for low-dimensional systems High-dimensional neural network potentials established the second generation and are based on three key steps: first, the expression of the total energy as a sum of environment-dependent atomic energy contributions; second, the description of the atomic environments by atom-centered symmetry functions as descriptors fulfilling the requirements of rotational, translational, and permutation invariance; and third, the iterative construction of the reference electronic structure data sets by active learning In third-generation HDNNPs, in addition, long-range interactions are included employing environment-dependent partial charges expressed by atomic neural networks In fourth-generation HDNNPs, which are just emerging, in addition, nonlocal phenomena such as long-range charge transfer can be included The applicability and remaining limitations of HDNNPs are discussed along with an outlook at possible future developments

Four Generations of High-Dimensional Neural Network Potentials.

An increased synergy between experimental and theoretical investigations in heterogeneous catalysis has become apparent during the last decade. Experimental work has extended from ultra-high vacuum and low temperature towards operando conditions. These developments have motivated the computational community to move from standard descriptive computational models, based on inspection of the potential energy surface at 0 K and low reactant concentrations (0 K/UHV model), to more realistic conditions. The transition from 0 K/UHV to operando models has been backed by significant developments in computer hardware and software over the past few decades. New methodological developments, designed to overcome part of the gap between 0 K/UHV and operando conditions, include (i) global optimization techniques, (ii) ab initio constrained thermodynamics, (iii) biased molecular dynamics, (iv) microkinetic models of reaction networks and (v) machine learning approaches. The importance of the transition is highlighted by discussing how the molecular level picture of catalytic sites and the associated reaction mechanisms changes when the chemical environment, pressure and temperature effects are correctly accounted for in molecular simulations. It is the purpose of this review to discuss each method on an equal footing, and to draw connections between methods, particularly where they may be applied in combination.

/pdf/towards-operando-computational-modeling-in-heterogeneous-48sn74uzhy.pdf

Towards operando computational modeling in heterogeneous catalysis.

The determination of crystal structures and the solid-to-solid phase transition mechanisms are two important and related subjects in material science. Here we develop an unbiased general-purpose potential energy surface (PES) searching method, namely, SSW-crystal method, for prediction of both the crystal structure and the crystal phase transition pathway. The SSW-crystal method features with stochastic surface walking (SSW) via repeated small structural perturbation by taking into account the second derivative information on both the lattice and the atom degrees of freedom. The SSW-crystal method is capable of overcoming the high barrier of phase transition and identifying the desirable phase transition reaction coordinates. By applying the SSW-crystal method to a set of examples, including SiO2 crystal up to 162 atoms per cell, Lennard-Jones model crystals up to 256 atoms, ternary SrTiO3 crystal of 50 atoms and the rutile-to-anatase TiO2 phase transition, we show that the SSW-crystal method can efficiently locate the global minimum (GM) from random initial structures without a priori knowledge of the system, and also allows for exhaustive sampling of the phase transition pathways, from which the lowest energy pathway can be obtained.

Stochastic surface walking method for crystal structure and phase transition pathway prediction

Here we introduce the LASP code, which is designed for large‐scale atomistic simulation of complex materials with neural network (NN) potential. The software architecture and functionalities of LASP will be overviewed. LASP features with the global neural network (G‐NN) potential that is generated by learning the first principles dataset of global PES from stochastic surface walking (SSW) global optimization. The combination of the SSW method with global NN potential facilitates greatly the PES exploration for a wide range of complex materials. Not limited to SSW‐NN global optimization, the software implements standard interfaces to dock with other energy/force evaluation packages and can also perform common tasks for computing PES properties, such as single‐ended and double‐ended transition state search, the molecular dynamics simulation with and without restraints. A few examples are given to illustrate the efficiency and capabilities of LASP code. Our ongoing efforts for code developing and G‐NN potential library building are also presented.

LASP: Fast global potential energy surface exploration

Toward the activity prediction with large-scale computations, here a double-ended surface walking (DESW) method is developed for connecting two minima on a potential energy surface (PES) and locating the associated transition state (TS) using only the first derivatives. The method operates two images starting from the initial and the final states, respectively, to walk in a stepwise manner toward each other. The surface walking involves repeated bias potential addition and local relaxation with the constrained Broyden dimer method to correct the walking direction. We apply the method to a model PES, a large set of gas phase Baker reactions, and complex surface catalytic reactions, which demonstrates that the DESW method can establish a low energy pathway linking two minima even without iterative optimization of the pathway, from which the TS can be located readily. By comparing the efficiency of the new method with the existing methods, we show that the DESW method is much less computationally demanding a...

Double-Ended Surface Walking Method for Pathway Building and Transition State Location of Complex Reactions

The solid-phase transitions of zirconia are important phenomena for many industrial applications. Because of the lack of tools for resolving the atom displacement pattern, the transition kinetics has been disputed for over 60 years. Here, first-principles-based stochastic surface walking (SSW) pathway sampling is utilized for resolving the mechanism of ZrO2 tetragonal-to-monoclinic solid-phase transition. Two types of lattice and atom correspondence allowed in phase transition are determined for the first time from energy criterion, which are originated from two nearly energy-degenerate lowest-transition pathways and one stress-induced ferroelastic transition channel of tetragonal phase. An orthorhombic crystal phase (Pbc2/1) is discovered to be a trapping state at low temperatures in phase transition, the presence of which does not create new orientation relation but deters transformation toughening significantly. This new finding may facilitate the design of new functional oxide materials in ceramic ind...

Energy Landscape of Zirconia Phase Transitions.

It is of general concern whether the automated structure prediction of unknown material without recourse to any knowledge from experiment is ever possible considering the daunting complexity of potential energy surface (PES) of material. Here we demonstrate that the stochastic surface walking (SSW) method can be a general and promising solution to this ultimate goal, which is applied to assemble carbon fullerenes containing up to 100 atoms (including 60, 70, 76, 78, 80, 84, 90, 96, and 100 atoms) from randomly distributed atoms, a long-standing challenge in global optimization. Combining the SSW method with a parallel replica exchange algorithm, we can locate the global minima (GM) of these large fullerenes efficiently without being trapped in numerous energy-nearly degenerate isomers. Detailed analyses on the SSW trajectories allow us to rationalize how and why the SSW method is able to explore the highly complex PES, which highlights the abilities of SSW method for surmounting the high barrier and the p...

Xiao-Jie Zhang

Papers

Stochastic surface walking method for crystal structure and phase transition pathway prediction

LASP: Fast global potential energy surface exploration

Double-Ended Surface Walking Method for Pathway Building and Transition State Location of Complex Reactions

Energy Landscape of Zirconia Phase Transitions.

From Atoms to Fullerene: Stochastic Surface Walking Solution for Automated Structure Prediction of Complex Material