Linear-scaling density-functional theory with the ONETEP code
University of Cambridge, UK
First-principles simulations based on density-functional theory (DFT), in particular the plane-wave pseudopotential (PWP) method, have become established as a powerful tool for gaining insight into complex atomistic processes and predicting the properties of new materials. Methods for performing such calculations are being developed and applied by a growing number of scientists including not just physicists, chemists and materials scientists but also biochemists and geologists.
However the system-sizes accessible to first-principles simulations is limited by the computational scaling of traditional implementations, which grows with the cube of the number of atoms and restricts them to the study of several hundreds of atoms even with modern supercomputers.
There has therefore been much interest in the development of so-called linear-scaling methods for insulators, which promise to revolutionise the scope and scale of simulations based upon DFT and facilitate calculations involving thousands of atoms. These new methods all abandon the conventional description of the fictitious Kohn-Sham system in terms of extended Bloch states in order to exploit the localisation of the density-matrix and/or Wannier functions. This also means that linear-scaling calculations are more amenable to embedding within other calculations and hence incorporation within multiscale simulations. This is reflected by the incorporation within Working Group 2 (Multiscale Methods) of the Psi-k Network.
However only a few general purpose linear-scaling codes have emerged over the last decade. The ONETEP code has been applied to systems consisting of up to thirty thousand atoms and ranging from proteins to nanostructures. In ONETEP, local orbitals associated with each atom are described in terms of a systematic basis set equivalent to a set of plane-waves and individually optimised in situ to obtain high accuracy and transferability.
While ONETEP inherits a number of desirable features from its relationship with the PWP method, it is nonetheless based on a reformulation of DFT in terms of the density-matrix whose truncation requires a considerably more complex (and sometimes conflicting) convergence procedure. Hence this tutorial is required to introduce the new principles and practices associated with ONETEP both to experienced practitioners and novices alike.
Although ONETEP is marketed commercially by Accelrys, it is available to academic users worldwide direct from the University of Cambridge via an inexpensive license to cover administrative costs. These users are encouraged to participate in the self-supporting ONETEP user community through the Wiki: www.onetep.org.
The first ONETEP summer school was held in Cambridge in July 2008 and was intended mainly for prospective developers. The attendees were almost exclusively from the UK. The aim of this tutorial is rather different: to provide training for new users from across Europe and beyond and to help them to exploit the new opportunities that ONETEP provides for their research.
Peter Haynes (Imperial College London) - Organiser & speaker
Nicholas Hine (Imperial College London) - Organiser & speaker
Arash Mostofi (Imperial College London) - Organiser & speaker
Mike Payne (University of Cambridge) - Organiser & speaker
Chris-Kriton Skylaris (University of Southampton) - Organiser & speaker