Characterization of the Activation-Relaxation Technique: Recent Results on Models of Amorphous Silicon
- PDF / 295,486 Bytes
- 6 Pages / 612 x 792 pts (letter) Page_size
- 21 Downloads / 141 Views
Characterization of the Activation-Relaxation Technique: Recent Results on Models of Amorphous Silicon Rachid Malek1 , Normand Mousseau1 y and G. T. Barkema2 (1) Department of Physics and Astronomy and CMSS, Ohio University, Athens, OH 45701, USA. (2) Theoretical Physics, Universiteit Utrecht, Utrecht, The Netherlands ABSTRACT The activation-relaxation technique (ART) is a method for finding saddle points in high-dimensional energy landscapes. ART has already been applied to a wide range of materials including amorphous semiconductors, Lennard-Jones glasses, and proteins. In spite of its successes, a number of fundamental questions remain to be answered regarding the biases associated with its sampling of the saddle points. We present here results of a detailed analysis of the biases in the simulation of amorphous silicon. We focus in particular on the biases of the method in sampling saddle points, the completeness of the sampling and the sensitivity of these quantities to variations of the different parameters.
INTRODUCTION One of the major obstacles for scientific progress is the gap between the time scales relevant for experiment and those reachable in atomistic computer simulations. To bridge this time-gap, one approach is to speed up molecular dynamics either by deforming the potential energy landscape or by reweighting high-temperature fast dynamics in a manner appropriate for low temperature simulations. Such algorithms, developed mostly by Voter and collaborators, have already been applied with success to study a number of growth problems on metallic surfaces [1, 2, 3]. Although this represents a major step forward, these methods are extremely computer intensive and scientific results usually require access to massively parallel machines. Another approach, developed by us [4] and others [5, 6], is to concentrate solely on the activated mechanisms, defined by the crossing of first-order saddle points in the configurational energy landscape. With sufficient sampling, Henkelman and J´onsson showed, for a simple metal-on-metal diffusion problem, that it is possible to build up an instantaneous table of events and pursue directly the dynamics in a kinetic Monte-Carlo fashion, using only a workstation. While the original algorithms were formally different, they have converged over the last few years, now only differing in the details of the implementation. The latest version of the activationrelaxation technique, ‘ART nouveau’, can converge directly onto first-order saddle points and provide a unique way to sample high-dimensional (d>100) energy landscapes. Although the proof of concept was established by Henkelman and J´onsson, we are not yet able to perform dynamical simulations of complex materials such as amorphous semiconductors, glasses or polymers using ‘ART nouveau’. At this point, there are still too many unknowns regarding the structure of the energy landscape of these materials for kinetic Monte Carlo simulations. y Present
address: D´epartement de physique, Universit´e de Montr´eal, C.P. 6128 succ.
Data Loading...
