An MPI-based Parallel Multiphysics Discontinuous Galerkin Framework for Photoconductive Devices
Permanent link to this recordhttp://hdl.handle.net/10754/662548
MetadataShow full item record
AbstractThe proper functioning of photoconductive devices (e.g., photoconductive antennas and photomixers) relies on the nonlinear interactions between the electromagnetic (EM) fields/waves induced on the whole device and the carriers inside the semiconductor region [1-2]. Therefore, simulator developed for the numerical characterization of these devices should be capable of accurately modeling EM field interactions and carrier dynamics individually and also the (nonlinear) coupling between them [3-4]. The characteristic scales of EM fields and carriers differ by several orders of magnitude in space and time, which makes the simulation computationally very challenging. Additionally, the use of geometrically intricate nanostructures, which are used to increase the optical-to-terahertz (THz) conversion efficiency by focusing optical EM fields, has further increased the computational requirements of the simulator being developed. In this talk, we present an MPI-based parallel multiphysics discontinuous Galerkin (DG) framework for the analysis of photoconductive devices. This framework couples Maxwell and drift-diffusion equations and discretizes them using DG schemes. The resulting (coupled) semi-discrete system is integrated in time using an explicit time marching scheme to yield the sampled values of the EM field and the carrier density. The two solvers for two sets of equations are MPI-parallelized within the same object-oriented framework. The device is discretized with a single mesh, while different (geometric) partitions of the mesh are used in individual solvers. Several concerns on improving the computational efficiency of the parallel DG framework are discussed next. First, carriers response slower than the variation of EM fields, consequently different time integration schemes and different time step sizes are used by the two solvers. Second, because the two solvers have different computation domains, different workload weights are assigned to different partitions of the mesh to obtain higher parallel efficiency and better scalability. Third, the photoconductive antenna simulation requires modeling of both optical- and THz-frequency EM waves. Direct simulation of the THz-frequency EM radiation of the antenna using the mesh size and time step size required by the optical-frequency EM wave is prohibitively expensive. A two-step simulation procedure is used to alleviate this problem. First, the transient photocurrent is obtained from the interaction between the optical pump and the semiconductor device using the multiphysics DG framework. Then this current is fed into the THz antenna, and the THz radiation is modeled using an EM-only DG scheme as a post-processing step.
SponsorsKing Abdullah University of Science and Technology (KAUST) Office of Sponsored Research (OSR) under Award No 2016-CRG5-2953. The authors would like to thank the King Abdullah University of Science and Technology Supercomputing Laboratory (KSL) for providing the required computational resources.
Conference/Event namePhotonIcs & Electromagnetics Research Symposium 2019
The following license files are associated with this item:
Except where otherwise noted, this item's license is described as Attribution-NonCommercial-NoDerivs 3.0 United States