DFT-FE — a massively parallel real-space density functional theory code using higher-order adaptive spectral finite-element discretization, and its large-scale application to study dislocation core energetics in crystalline materials

Kohn-Sham density functional theory (DFT) calculations have been instrumental in providing many crucial insights into materials behavior and occupy a sizable fraction of world's computational resources today. However, the stringent accuracy requirements in DFT needed to compute meaningful material properties, in conjunction with the asymptotic cubic-scaling computational complexity with number of electrons, demand huge computational resources. Thus, these calculations are routinely limited to material systems with at most few thousands of electrons. In this talk, we present a massively parallel real-space DFT framework (DFT-FE), which is based on a local real-space variational formulation of the Kohn-Sham DFT energy functional discretized with higher-order adaptive spectral finite-element, and handles pseudopotential and all-electron calculations in the same framework. We will present the efficient and scalable numerical algorithms in conjunction with mixed precision strategies for the solution of Kohn-Sham equations, that has enabled computationally efficient, fast and accurate DFT calculations on generic material systems reaching ~100,000 electrons, on both many-core and hybrid CPU-GPU architectures. DFT-FE achieves an order of magnitude performance advantage over widely used plane-wave codes both in CPU-times and wall-times. Further, we will present our recent work on developing a robust and efficient preconditioner for self-consistent field iterations in Kohn-Sham DFT. Finally, we demonstrate the successful application of DFT-FE to accurately study the core energtics of pyramidal dislocations in magnesium, which have so far been out of reach as large system sizes containing thousands of atoms are required to accurately resolve the relevant physics in the dislocation core.