A Highly Scalable NEGF Solver for Modeling Time-Dependent Quantum Transport in Nanomaterials

Advances in nanotechnology are enabling breakthroughs in miniaturized photon detectors, which will, in turn, have a tremendous impact on numerous scientific undertakings, such as the identification of remote galaxies and biomedical imaging. Such devices make use of highly conductive nanomaterials such as graphene or carbon nanotubes that are functionalized with photoactive molecules or quantum dots, which absorb photons of particular wavelength depending on their size and generate exciton pairs that dissociate at the interface of these nanomaterials. Either a hole or electron of the exciton pair is transferred to the transport channel, while the leftover charge modulates the electrostatic surface potential and affects the current. Quantum transport through these nanomaterials can be accurately characterized using the time-dependent non-equilibrium Green's function method that must be self-consistently coupled with Poisson's equations. In this work, we present an open-source, GPU-accelerated software framework to solve these equations and demonstrate its scaling on manycore/GPU architectures. The core of the solver is built on top of the AMReX library, which provides support for highly scalable applications employing structured grid implementations.