TROCADERO ROCADERO : a multiple-algorithm multiple-model atomistic simulation program R. Rurali a , E. Hern andez b, * a Centre Nacional de Microelectr  onica (CNM––CSIC), Campus de Bellaterra, 08193 Barcelona, Spain b Institut de Ci  encia de Materials de Barcelona (ICMAB––CSIC), Campus de Bellaterra, 08193 Bellaterra, Barcelona, Spain Abstract In this paper a technical description of the simulation package TROCADERO ROCADERO is provided. This program implements large-scale atomistic simulations of materials, using a variety of simulation techniques which include several forms of molecular dynamics, structural relaxation and phonon calculations. Also a growing number of different models are available in the program, ranging from empirical potentials to semi-empirical tight-binding models. A detailed dis- cussion of the coding strategy is given, followed by a summary of the capabilities of the program, both in terms of simulation algorithms and models available, which are further illustrated by a review of recent applications of the code in several branches of computational materials science. Ó 2003 Elsevier B.V. All rights reserved. PACS: 71.15.Pd; 07.05.Tp Keywords: Atomistic simulation; Molecular dynamics; Modular programming 1. Introduction In the last three decades, computer simulation of materials [1–3] has evolved from a tool for statis- tical mechanics of simple model systems (such as Lennard-Jones fluids or spin systems) to a tool for the study of realistic models, thanks to ever grow- ing computer power on the one hand, and to im- portant developments in simulation techniques and fundamental advances in the theoretical descrip- tion of interactions at the atomic scale, on the other. These advances have turned materials sim- ulation into a tool of increasing usefulness. Traditionally, simulation packages are built around a single atomistic model. If a user wants to compare the relative merits of two different models (e.g. the Tersoff potential vs the Stillinger–Weber model), then typically two different programs would be used, and this is likely to complicate the task of making accurate comparisons, since it is frequently the case that different programs do not always have the same capabilities. Yet, as ex- plained in Section 2, the structure of a simulation package is independent of the details of the model. Why not then construct simulation packages that can work with different models? Then all the simulation capabilities of the program will be Computational Materials Science 28 (2003) 85–106 www.elsevier.com/locate/commatsci * Corresponding author. E-mail address: ehe@icmab.es (E. Hern andez). 0927-0256/$ - see front matter Ó 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0927-0256(03)00100-9