Berlin 2024 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
O: Fachverband Oberflächenphysik
O 9: Semiconductor Substrates I: Adsorption of Small Molecules, Metallic Nanowires, Overlayers
O 9.1: Talk
Monday, March 18, 2024, 10:30–10:45, MA 144
Bayesian and Multi-Objective Optimization for Sensor Technologies — •Ransell D’Souza1, Shuja Malik2, Fatima Annanouch2, Eduard Valero2, and Milica Todorović1 — 1University of Turku, Turku, Finland — 2Universitat Rovira i Virgili
Sensors, a device that can detect toxic and combustible gasses, play a crucial role in identifying potentially hazardous gas leaks. Sensors typically function by employing inorganic substrates that respond to the adsorption of target molecules. To design selective and sensitive sensors, density functional theory studies should explore the interplay between adsorption energies, charge transfer, recovery time, and sensor response during interactions between toxic targets and the substrate.
While conventional computational approaches have focused on individual material properties, we need to balance different considerations of these properties for a truly predictive sensor design. This presents a multi-objective optimization problem, which can be effectively addressed using Bayesian Optimization (BO)-based machine learning techniques.
In this study, we employ multi-objective BO techniques to investigate the stable adsorbate structures of NH3 on WS2. We identify optimal tradeoffs between energetically stable structures and optimal response functions (Pareto optimal solutions) for the rational design of new sensors. Our resistivity and sensor response functions agree with experimental data, validating active learning’s success in gas sensor material optimization.
Keywords: Sensors; Multi-objective Bayesian Optimization; Density functional theory; 2D materials; Machine learning