Parallel eigensolvers in plane-wave Density Functional Theory
It provides a scalable solution for large-scale electronic structure computations, which is critical for materials science and chemistry researchers using high-performance computing.
The paper addresses parallelization of eigensolvers for plane-wave DFT, showing that a Chebyshev polynomial-based algorithm scales to tens of thousands of processors and outperforms block conjugate gradient methods for large computations.
We consider the problem of parallelizing electronic structure computations in plane-wave Density Functional Theory. Because of the limited scalability of Fourier transforms, parallelism has to be found at the eigensolver level. We show how a recently proposed algorithm based on Chebyshev polynomials can scale into the tens of thousands of processors, outperforming block conjugate gradient algorithms for large computations.