TY - JOUR

T1 - Smart Monte Carlo for accurate simulation of rare-event dynamics

T2 - Diffusion of adsorbed species on solid surfaces

AU - Kumar, P. V.

AU - Raut, Janhavi S.

AU - Warakomski, Steven J.

AU - Fichthorn, Kristen A.

PY - 1996/1/1

Y1 - 1996/1/1

N2 - We introduce a dynamical Smart Monte Carlo algorithm and assess its applicability for simulating the rare-event dynamics of adsorbate diffusion. Using the dynamical Smart Monte Carlo method, we simulate the self-diffusion of an adatom in the Cu/Cu(001) and Rh/Rh(111) systems and we compare the simulated diffusion coefficients to values arising from molecular dynamics and transition-state theory. We find that the accuracy of Smart Monte Carlo is sensitive to details of the potential-energy surface. For Cu/Cu(001), the agreement between dynamical Smart Monte Carlo, molecular dynamics, and transition-state theory is excellent. A similar comparison for the Rh/Rh(111) systems shows discrepancies between these three techniques. We find that the origins of the discrepancies in the Rh/Rh(111) system are transition-state recrossings, for small simulation time steps, and low escape rates of the adatom from the binding sites, at large time steps. We examine the sampling and dynamics in trajectories using a smaller time step for motion perpendicular to the surface than that for parallel motion. These studies show that low Smart Monte Carlo escape rates in the Rh/Rh(111) system can be correlated to excessive sampling, beyond the configurational space of the potential-energy minimum, at large time steps. Recrossings can be understood to arise from the absence of velocity correlations in the low-friction, transition-state region and can be minimized through the use of a large time step for parallel motion. With the appropriate choice of simulation time steps it is possible to improve the agreement between dynamical Smart Monte Carlo and more rigorous dynamical techniques.

AB - We introduce a dynamical Smart Monte Carlo algorithm and assess its applicability for simulating the rare-event dynamics of adsorbate diffusion. Using the dynamical Smart Monte Carlo method, we simulate the self-diffusion of an adatom in the Cu/Cu(001) and Rh/Rh(111) systems and we compare the simulated diffusion coefficients to values arising from molecular dynamics and transition-state theory. We find that the accuracy of Smart Monte Carlo is sensitive to details of the potential-energy surface. For Cu/Cu(001), the agreement between dynamical Smart Monte Carlo, molecular dynamics, and transition-state theory is excellent. A similar comparison for the Rh/Rh(111) systems shows discrepancies between these three techniques. We find that the origins of the discrepancies in the Rh/Rh(111) system are transition-state recrossings, for small simulation time steps, and low escape rates of the adatom from the binding sites, at large time steps. We examine the sampling and dynamics in trajectories using a smaller time step for motion perpendicular to the surface than that for parallel motion. These studies show that low Smart Monte Carlo escape rates in the Rh/Rh(111) system can be correlated to excessive sampling, beyond the configurational space of the potential-energy minimum, at large time steps. Recrossings can be understood to arise from the absence of velocity correlations in the low-friction, transition-state region and can be minimized through the use of a large time step for parallel motion. With the appropriate choice of simulation time steps it is possible to improve the agreement between dynamical Smart Monte Carlo and more rigorous dynamical techniques.

UR - http://www.scopus.com/inward/record.url?scp=0041827878&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0041827878&partnerID=8YFLogxK

U2 - 10.1063/1.471895

DO - 10.1063/1.471895

M3 - Article

AN - SCOPUS:0041827878

VL - 105

SP - 686

EP - 695

JO - Journal of Chemical Physics

JF - Journal of Chemical Physics

SN - 0021-9606

IS - 2

ER -