Presentation
20 March 2023 A link between micro-LED simulation and Bayesian reasoning
Nicolas Michit
Author Affiliations +
Abstract
We present an overview of kriging benefits when dealing with optical simulations of non-coherent light emitting µLED. Such device is composed of quantum wells (QW) forming a surface embedded in high index material and connected by metallic electrodes. The traditional way to simulate such object is using FDTD with a discretized QW surface into multiple dipoles. It leads to tremendous computational effort when conducting a parameter space exploration in order to optimize the µLED optical behaviour. To diminish the required number of one-dipole simulations, we propose to build a metamodel using Bayesian inference with Gaussian process iteratively building a non-uniform dipole grid more adapted to the QW surface. Gaussian process inference also enables calculating an estimate of the final result’s error margin to provide a stopping criteria for the simulation : it allows to considerably reduce the number of simulations one theoretically needs to perform.
Conference Presentation
© (2023) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Nicolas Michit "A link between micro-LED simulation and Bayesian reasoning", Proc. SPIE PC12415, Physics and Simulation of Optoelectronic Devices XXXI, PC1241501 (20 March 2023); https://doi.org/10.1117/12.2649912
Advertisement
Advertisement
KEYWORDS
Error analysis

Quantum wells

Computer simulations

Bayesian inference

Electrodes

Finite-difference time-domain method

Optical simulations

Back to Top