Zbigniew A. Rycerz

The Force-Biased Algorithm for the Irregular Close Packing of Equal Hard Spheres

Abstract We present a “force-biased” algorithm for generating the irregular close packing of hard spheres. The algorithm is partly based on Jodrey and Torys ideas [9] and incorporates methods from Molecular Dynamics. Packings generated by means of the two algorithms are consistent up to final packing fraction of 0.65, which seems to be the limit density of Jodrey and Torys method. Significantly higher densities (up to 0.71) can be achieved for small numbers of spheres by the force-biased algorithm. However the shape of the radial and angle distribution functions implies that a partial short-range ordering occurs in packings of those densities.

Ewald Summation in the Molecular Dynamics Simulation of Large Ionic Systems: The Cohesive Energy

Abstract The accuracy and efficiency of the direct Ewald summation are discussed in terms of the size of a Molecular Dynamics (MD) ionic system and the ranges of the r-space and q-space summations. The dependence of the convergence parameter α on the size of the system and on the choice of cut-off radius for the short-range potential is given. The possibility of neglecting the q-space term for large ionic systems is discussed in terms of the accuracy and efficiency of the simulation.

A vectorized algorithm on the ETA 10-P for molecular dynamics simulation of large number of particles confined in a long cylinder

Abstract In this paper we describe an algorithm suitable for molecular dynamics (MD) computer simulation of particles confined in a cylinder and interacting by short-ranged forces. A procedure for determining neighbours is based on sorting the particles according to values of one coordinate along the axis of a cylinder and gathering the values of the other two coordinates due to indices of the sorted list. In addition to the classical cut-off distance an integer cut-off number is introduced. This and the organization of data structures make that the calculating forces loop is completely vectorized. The algorithm has been devised for the CYBER 205 and ETA 10 computers and is specially efficient for MD simulation in cylindrical micropores.

Vectorized program of order N for molecular dynamics simulation of condensed matter: II. MDSLAB1: Slab, short-range interactions

Abstract A vectorized program of order N for molecular dynamics simulation of condensed matter systems is presented. The program handles short-range interactions and it enables one to simulate efficiently very large systems. The program can be used on any vector computer and on the ETA 10-P supercomputer it calculates the interactions between the particles with a speed of about 50 000 particles/cpu s.

Molecular dynamics simulation program of order N for condensed matter: I. MDPYRS1: scalar pyramid, short-range interactions

Abstract A program for molecular dynamics simulation in which the cpu time proportionally to the number of particles N is presented. The program enables one to stimulate efficiently very large condensed-matter systems involving short-range interactions (e.g. simple liquids with pairwise-additive forces). The program is optimized for serial computers and it does not involve any neighbour lists, so it requires relatively small computer memory.

Molecular Dynamics Simulation of the Fast Ion Conductor δ-Bi2o3

Abstract Molecular Dynamics simulations of the superionic δ-phase of Bi2O3 are reported. Initial configurations were based on 100,000 time-step (Δt = 5 × 10−15s) equilibrations. Making use of recently determined thermal expansivity data on δ-Bi2O3, a series of simulations were performed at increasing temperatures, with appropriate lattice constants, until oxygen sub-lattice melting was observed. Calculations were restricted to the rigid-ion approximation and accordingly short-range potentials, adjusted to reproduce the static dielectric constant were employed, in addition to modified electron-gas potentials. The dielectrically adjusted potentials proved superior in most respects although they sometimes prompted phase changes. In agreement with static-lattice calculation predictions the primary charge migration mechanism is a cube-edge or > vacancy migration. Cube-edge interstitials, half way along a cube-edge, predicted in static-lattice simulations and observed in neutron studies on yttria stabiliz...

Acceleration of molecular dynamics simulation of order N with neighbour list

Abstract A method for removing data dependency in an indirectly indexed DO loop is presented. Vectorization of such DO loops may, in some applications and on those computers which cannot do it themselves, result in a considerable speed-up of the calculations. Timings for scalar/vector calculations have been performed on an ETP 10P supercomputer and the application of the algorithm for a speed-up of molecular dynamics (MD) simulations of order N, handling short-range interactions, is discussed. The algorithm presented does not use any special calls and, therefore, it can be applied on most vector computers. The method makes it possible to execute an MD program of order N with practically the same efficiency on any supercomputer no matter if such a loop can be automatically vectorized or not.

On the Efficiency of Vectorized Molecular Dynamics Algorithms of Order N

Abstract We discuss in this paper several factors which decide the efficiency of a vectorized Molecular Dynamics algorithm of order N, handling short-range interactions. General rules have been formulated and a fulfilment of these ought to ensure a very high speed of vector processing. The principles described make it possible to perform on presently available supercomputers the simulation of a 3-D system as large as 105 (or even more) particles in rather modest cpu time.

Vectorized program of order N for molecular dynamics simulation of condensed matter: I. MDPYRV1: Vector pyramid, short-range interactions

Abstract A vectorized program of order N for molecular dynamics simulation of condensed matter systems is presented. The program handles short-range interactions and it enables one to simulate efficiently very large systems. This program consists of three — to a certain degree independent — versions which differ in their central memory requirements and their speed. The first version does not involve any neighbour lists, so it requires relatively small computer memory, but it is the slowest version. The two others are much faster, but they use some neighbour lists. The last (fastest) version calculates the interactions between the particles with a speed of about 50 000 particles/cpu s on the ETA 10-P supercomputer.

Monte Carlo computations of the order—disorder transition temperature on the lattice model of the binary BCC system

Abstract An order—disorder transition in a lattice model of the binary bcc system is simulated for four compositions by means of the Monte Carlo technique. The transition temperature of the model as a function of concentration is presented.