A Real-Space Genetic Algorithm for Crystal Structure Determination N.L. Abraham and M.I.J. Probert Department of Physics, University of York, Heslington, York, YO10 5DD http://www-users.york.ac.uk/~nla101/ Abstract There has been much interest in using genetic algorithms for determining the ground-state structure of clusters [Deaven and Ho, 1995] and more recently silicon surfaces [Chuang et al., 2004]. We present a real-space encoded genetic algorithm which is suitable not only for surface structure calculations, but also for bulk crystal structure determination. This algorithm makes use of a novel crossover technique in the generation of offspring. The method is also suitable as a polymorph search, and is flexible enough that population members can have different supercells. We will present results from a variety of empirical and ab initio systems, where all calculations have been performed using the CASTEP [Segall et al., 2002] code. 1 Introduction Genetic Algorithms (GAs) are becoming widely used to determine the minimum energy configurations of atomic clusters [Deaven and Ho, 1995, Johnston, 2003], nanowires [Wang et al., 2001], and surfaces [Chuang et al., 2004, 2005]. These previous studies made use of a planar cut in the crossover operation, slicing each parent in two, and then swapping halves to generate offspring. However, a planar cut does not take the system periodicity into account. The prediction of crystal structures from first prin- ciples has long been recognised as one of the out- standing challenges in solid state physics [Maddox, 1988, VanDeWalle, 2005]. The most recent methods of cluster expansion [Blum et al., 2005, Hart et al., 2005] assume the lattice structure of the crystal. In this poster we demonstrate a new method for unbi- ased ab initio crystal structure determination using a novel Genetic Algorithm which makes no assump- tions of atom number, unit cell or lattice structure. 2 Genetic Algorithms • Genetic Algorithms are a stochastic minimisation technique inspired by Darwin’s Theory of Evolu- tion [Holland, 1992] (see figure 1). • The “population” is made up of a number of mem- bers, each of which is a viable solution to the prob- lem being studied. Figure 1: A Genetic Algorithm Figure 2: Crossover