Development of a Multi-Objective Evolutionary Algorithm for Strain-Enhanced Quantum Cascade Lasers

Abstract

An automated design approach using an evolutionary algorithm for the development of quantum cascade lasers (QCLs) is presented. Our algorithmic approach merges computational intelligence techniques with the physics of device structures, representing a design methodology that reduces experimental effort and costs. The algorithm was developed to produce QCLs with a three-well, diagonal-transition active region and a five-well injector region. Specifically, we applied this technique to Al x Ga 1−x As/In y Ga 1−y As strained active region designs. The algorithmic approach is a non-dominated sorting method using four aggregate objectives: target wavelength, population inversion via longitudinal-optical (LO) phonon extraction, injector level coupling, and an optical gain metric. Analysis indicates that the most plausible device candidates are a result of the optical gain metric and a total aggregate of all objectives. However, design limitations exist in many of the resulting candidates, indicating need for additional objective criteria and parameter limits to improve the application of this and other evolutionary algorithm methods.

A2Cloud‐RF: A random forest based statistical framework to guide resource selection forhigh‐performancescientific computing on the cloud

“…The multi-objective evolutionary algorithm (MOEA) is used to search the massive design space of the diagonal transition three-well quantum cascade laser (QCL) device. 64 The evolutionary computation techniques used are based on the second generation Strength Pareto Evolutionary Algorithm (SPEA2). 65 To simulate the QCL candidate devices, the algorithm uses a third-party self-consistent Schrödinger-Poisson solver, nextnano 3.66 The evolutionary QCL performs frequent disk input/output and is computationally-intensive 64 with an AI value equal to 2.…”

“…The multi‐objective evolutionary algorithm (MOEA) is used to search the massive design space of the diagonal transition three‐well quantum cascade laser (QCL) device 64 . The evolutionary computation techniques used are based on the second generation Strength Pareto Evolutionary Algorithm (SPEA2) 65 .…”

“…To simulate the QCL candidate devices, the algorithm uses a third‐party self‐consistent Schrödinger‐Poisson solver, nextnano 3. 66 The evolutionary QCL performs frequent disk input/output and is computationally‐intensive 64 with an AI value equal to 2.…”