DynamO: A free O(N) general event-driven molecular-dynamics simulator

arXiv:1004.3501v2 [physics.comp-ph] 26 Jul 2011 DynamO: A free O(N) general event-driven molecular-dynamics simulator M. N. Bannerman,1, 2 R. Sargant,2 and L. Lue3, 4 1Institute for Multiscale Simulation, Universit¨at Erlangen-N¨urnberg, Erlangen, Germany 2School of Chemical Engineering and Analytical Science, The University of Manchester, Oxford Road, Manchester M13 9PL, United Kingdom 3School of Chemical Engineering and Analytical Science, The University of Manchester, Manchester, United Kingdom 4Department of Chemical and Process Engineering, University of Strathclyde, James Weir Building, 75 Montrose Street, Glasgow G1 1XJ, United Kingdom (Dated: Received: date / Accepted: date) Abstract Molecular-dynamics algorithms for systems of particles interacting through discrete or “hard” potentials are fundamentally different to the methods for continuous or “soft” potential systems. Although many software packages have been developed for continuous potential systems, software for discrete potential systems based on event-driven algorithms are relatively scarce and specialized. We present DynamO, a general event-driven simulation package which displays the optimal O(N) asymptotic scaling of the com- putational cost with the number of particles N, rather than the O(N log N) scaling found in most standard algorithms. DynamO provides reference implementations of the best available event-driven algorithms. These techniques allow the rapid simulation of both complex and large (> 106 particles) systems for long times. The performance of the program is benchmarked for elastic hard sphere systems, homogeneous cooling and sheared inelastic hard spheres, and equilibrium Lennard-Jones fluids. This software and its documentation are distributed under the GNU General Public license and can be freely downloaded from http://marcusbannerman.co.uk/dynamo. 1 I. INTRODUCTION Molecular-dynamics simulations have become an indispensable tool in the development of novel nanomaterials [1], drug discovery [2–4], and materials engineering in the estimation of ther- mophysical properties and phase behavior of complex solutions. Molecular dynamics (MD) not only allows the exploration of the link between inter-particle interactions and macroscopic struc- ture and dynamics, but it is also capable of providing quantitative predictions for real materials. Molecular-dynamics simulations have been dominated by time stepping methods for systems that interact with continuous potentials. This method was first used by Rahman [5] in 1964, and later popularized by Verlet [6]. Since then, many sophisticated software packages have been developed, such as LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) [7], GROMACS (Groningen Machine for Chemical Simulation) [8, 9], NAMD [10], Desmond [11] and ESPResSo (Extensible Simulation Package for Research on Soft Matter) [12], which are freely available and allow the simulation of complex systems. In addition, a wealth of force fields based on continu- ous potentials have been developed to describe real materials, such as small organic and inorganic molecules, polypeptides, proteins, and DNA (e.g., see AMBER [13, 14] or CHARMM [15]). MD has also been applied to granular systems, which was pioneered by Cundall and Strack [16]. Since then, Hertz’s law for elastic particles has been generalized for viscoelastic spheres [17], and a range of approximations for the tangential forces are now available [18–20]. An alternative approach to modeling many-body systems is through the use of discrete inter- action potentials, such as the hard-sphere or square-well potentials. These potentials contain only distinct energy level changes; however, they can be stepped to either approximate soft potentials, such as the Lennard-Jones potential [21], or directly reproduce thermodynamic data [22]. The “true” interactions between real molecules or atoms (or larger scale particles) are expected to be smooth and continuous, and so one may question the relevance of discrete potentials; they are an extreme approximation to the “true” interactions. However, in reality, all interaction potentials that are used in computer simulations are necessarily approximate, due to practical limitations in computing resources. In the case of continuous potentials, the main approximation is typically the assumption of pair-wise additivity, neglecting many-body interactions which are present in all “real” systems, or the use of restricted functional forms for the interaction potential (e.g., Lennard- Jones, Stockmayer, etc.). So most commonly used continuous potentials also only approximate the “true” interactions. 2 The main issue is, however, not whether a potential exactly reproduces the “true” interactions between real particles, but whether it captures the essential features of the interaction to be able to reproduce the physics/chemistry which is of interest in a particular study. This is the primary motivation of coarse grained sim

…(Full text truncated)…

This content is AI-processed based on ArXiv data.