Speaker
Description
Molecular Dynamics simulations are the main techniques used to study a vast variety of atomic systems, spanning biological matter and inorganic materials. Such simulations are computationally demanding if they are carried at the quantum mechanical level; they might become unaffordable if large geometric models are required in order to produce data comparable with experiments. To overcome these difficulties, we propose the Normal Dynamics (ND) technique [1], which provides a way to integrate the Newton’s equations of motion by sampling the reciprocal space. We implemented the method in the FORTRAN code named “PINDOL”, freely available from the github repository [2]. In the ND method, the atomic interactions can be parameterized at the quantum mechanical level by including either ground or excited electronic configurations, magnetic or not, external electric or magnetic fields, and the hyperfine Hamiltonian. The sampling strategy is capable to produce dynamical trajectories at the ab initio level on an ordinary desktop computer. The reciprocal space sampling allows to: i) obtain a systematic improvement of the results accuracy and a fine control of the computational load, ii) account for distortions realized across large atomic distances without the use of large unit cells. The use of High Performance Computing facilities is mandatory for the preliminary calculation of the interatomic forces; once they are obtained, they can be collected in web-accessible databases for future uses. An optimal parallelization of the PINDOL code will enable its execution on HPC clusters for the study of very large systems. We consider three case studies: the first two are aimed to show the capabilities of the method while in the latter we investigate on the characterization of Raman spectra at different temperatures in MoS2/MX2 transition metal dichalcogenide heterostructures. Finally, we discuss how the sampling procedure is general and can be used to simulate periodic, semiperiodic and finite systems such as crystals, slabs, nanoclusters or molecules.
References
[1] A. Cammarata, M. Dašić, P. Nicolini, “Normal Dynamics: solving Newton’s equations of motion in the reciprocal space”, submitted, available at SSRN: https://ssrn.com/abstract=4550608
[2] PINDOL: Phonon-Inspired Normal Dynamics of Lattices, https://github.com/acammarat/pindol
