Physics-Informed Neural Network Modeling of Adsorption sensors for Multi-species Classification

Liu Hua-Hsing^1*, Yi-Ting Pan¹, Min-Syun Li¹, Kuang-Yao Lo¹, Yun-Chorng Chang²

¹Department of Physics, National Cheng Kung University, No. 1, University Road, East District, Tainan City, Taiwan, Taiwan
²Research Center for Applied Sciences, Academia Sinica, No. 128, Sec. 2, Academia Road, Nangang District, Taipei City, Taiwan

* Presenter:Liu Hua-Hsing, email:joan99884@gmail.com

Adsorption-based sensors, including metal-oxide gas sensors, ion-sensitive field-effect transistors (ISFETs), chemical-sensitive field-effect transistors (ChemFETs), and Biologically-sensitive Field-Effect Transistors (BioFETs), share a common limitation: when multiple target species coexist in the same environment, their signals often overlap, resulting in reduced selectivity. To address this challenge, specific receptors or selective filtering membranes are typically employed to enhance sensing performance. In this study, we leverage a Physics-Informed Neural Network (PINN) to distinguish among multiple gaseous and ionic species, thereby overcoming the inherent selectivity limitations of adsorption-based sensors. This approach not only unlocks the potential of low-cost adsorption-based sensors fabricated using advanced semiconductor technologies, but also enables miniaturization by replacing conventional sensor arrays previously required for multi-species discrimination. In the future, this technology could be applied to robotic olfactory, gustatory, and tactile systems, such as electronic noses, electronic tongues, and electronic skins.

Keywords: Physics-Informed Neural Network, Surface physics, Multi-ion Classification, Multi-Gas Classification, Adsorption sensors