Abstract
We discuss the practicalities of applying the metadynamics method to sampling crystallisation events in molecular systems. Suitable choices for collective coordinates are presented along with criteria for their parameterisation. Issues arising from finite-size effects are discussed with particular reference to the generation of multiple clusters when biasing global order parameters. We also consider the applicability of two methods for enhancing the accuracy of the reconstructed free-energy landscape. The discussion is illustrated with example data from freezing in the Lennard-Jones and ice–water systems.
Acknowledgements
Computing facilities were provided by the Centre for Scientific Computing of the University of Warwick and by the HECToR service. Early access to this service is gratefully acknowledged. Funding was provided by the EPSRC under grants GR/S80127 and EP/F054785 (HECToR Capability Challenge). The authors would also like to thank Dr Martyn Foster and Dr Ilian Todorov for their assistance with the DL_POLY 3 code on HECToR. We also thank Prof Mike Allen for useful discussions.
Notes
1. In this context we mean a purely geometrical connection vector rather than chemical bond.