Multi-Fidelity simulation in the exascale era
DateThursday, October 4, 2018 - 12:30pm
AbstractWith every generational shift in computing capability, we need to consider both what we can do with the new capabilities and how to do it. For the combustion community exascale computing offers the opportunity for a transformation increase in realism of near-first principles calculations (‘DNS’) in terms of complexity of the geometry and physics that can be included. However, even with this change the impact on real-world design considerations will still be dependent on the efficiency of transferring knowledge developed from the high fidelity simulations to engineering calculations that can be used directly for parameter space exploration and design optimization. Machine learning has the potential for accelerating this linkage. This presentation will explore some specific examples of how these three avenues: engineering simulation, exascale computing code development, and machine learning are being pursued at NREL to address combustion research challenges.
A parallel grid-free method for the radiative transfer equation
DateThursday, October 11, 2018 - 12:30pm
AbstractIn high-energy photon-based imaging systems, the steady-state energy propagation model takes the form of a radiative transfer equation, which is an integro-differential equation in 6 dimensions. In order to simulate such imaging systems for the purposes of design and optimization, fast and adequately accurate numerical methods are required. We discuss a method based on fast parallel grid-free X-ray transforms and iterative Neumann series, then show its application to computing spectral approximations to system singular functions.
Finite-Element Simulation of Optical Phenomena on 2D Materials
DateThursday, October 18, 2018 - 12:30pm
AbstractIn the terahertz frequency range, the effective (complex-valued) surface conductivity of atomically thick 2D materials such as graphene has a positive imaginary part that is considerably larger than the real part. This feature allows for the propagation of slowly decaying electromagnetic waves, called surface plasmon-polaritons (SPPs), that are confined near the material interface with wavelengths much shorter than the wavelength of the free-space radiation. SPPs are promising ingredients in the design of novel optical applications promising "subwavelength optics" beyond the diffraction limit. There is a compelling need for controllable numerical schemes which, placed on firm mathematical grounds, can reliably describe SPPs in a variety of geometries. In this talk we present an adaptive, higher-order finite element approach for the simulation of SPPs on 2D materials and layered structures. Aspects of the numerical treatment such as absorbing perfectly matched layers, local refinement and a-posteriori error control are discussed. We will present a number of applications of the framework to optical device simulations. Corresponding analytical results elucidate the solution structure. We conclude by introducing a homogenization theory of layered heterostructures to design novel devices.