Bulk semiconductor potentials

    Interatomic potentials for bulk semiconductor materials, corresponding to the paper "Gaussian Approximation Potentials: The Accuracy of Quantum Mechanics, without the Electrons", PRL 104, 136403 (2010). As potential format has changed considerably since the paper was published, using the original potential files would be cumbersome. The fitting procedure has also become more efficient by using virial information. We have therefore refitted the potentials using the most recent SOAP-based GAP kernel. In some cases we used the original dataset, but where available we added new ab initio datapoints to provide an even better potential. Due to the changes in the fitting procedure and the dataset these potentials do not reproduce the figures that were published in the paper exactly, but they are extremely close and there is no degradation in the quality of the fit.

    NB: Do not use these potentials on configurations substantially different from those used in training. This scheme is not expected to produce any meaningful extrapolation for domains in the phase space outside of the training region. The type of databases we used are listed for each potential. We did not include defects, surfaces, disordered phases or crystalline phases other than those listed.

    Using any of the potentials below for configurations other than defect-free crystals at moderate temperatures will produce RUBBISH.

    Potential for carbon, fitted to diamond and graphite configurations. The cutoff is 3.7 . Initialise this potential with Potential xml_label=GAP_2015_6_1_60_16_37_12_201 in LAMMPS or QUIP.


    Potential for silicon, fitted to diamond configurations. The cutoff is 4.0 . Initialise this potential with Potential xml_label=GAP_2015_6_3_60_10_30_13_769 in LAMMPS or QUIP.


    Potential for germanium, fitted to diamond configurations. The cutoff is 5.5 . Initialise this potential with Potential xml_label=GAP_2015_6_2_60_10_45_24_208 in LAMMPS or QUIP.


    Potential for gallium nitride, fitted to zincblende configurations. This is a GAP + Coulomb potential, with fixed, opposite charge of 1.07 on the Ga and N atoms. The electrostatics are treated by the Ewald method within QUIP. The cutoff of GAP is 4.0 . Initialise this potential with Potential xml_label=GAP_2015_6_1_60_16_13_34_864 in LAMMPS or QUIP.


    Training Configurations C, Si, Ge and GaN training configurations with DFT data

    Topic revision: r9 - 04 Feb 2018, GaborCsanyi
    Molecular Modelling

    Engineering Laboratory


    Log In
    Copyright by the contributing authors. All material on this collaboration platform is the property of the contributing authors.