Introduction to the FitLJ package
Interatomic potentials based on analytical, physically-motivated functions are fundamental ingredients in molecular simulations. Using them as surrogate models of more elaborate ab-initio methods can speed-up simulations by a factor up to \(10^4\).
The accuracy of interatomic potential models heavily relies on their parameters. Well parametrized models can match the accuracy of ab-initio methods, whereas poorly parametrized models can result in highly unphysical results. A key aspect in the application of any interatomic potential model lies therefore an accurate parametrization strategy.
The FitLJ package allows one to fit interatomic potential models based on the well-established Lennard-Jones potential, using a reference data set obtained from calculations performed with more accurate and computationally expensive ab-initio methods. The reference FitLJ program leverage on the database functionalities of the Atomic Simulation Environment (ASE) package. Any ab-initio software that can be coupled to ASE can therefore be used to produce the reference dataset to be used within FitLJ.
The functional form of the Lennard-Jones potential is
where the parameters :math:espilon and :math:sigma determine the depth of the potential well and the equilibrium distance, as shown Figure Fig. 1:
Fig. 1 Lennard-Jones potential for \(\epsilon = 3.3\) and \(\sigma = 1.0\). The vertical and horizontal lines indicate the potential defined by the parameters \(\epsilon\) and \(\sigma\), respectively.
FitLJ takes as input the energies of a discrete set of atomic configuration stored in an ASE database, and fit the functional form of Eq. (1). The program then produces as output the values of \(\epsilon\) and \(\sigma\) that provide the best match to the data.