ReflectoRNN: AI-Enabled In-Operando Optical Reflectometry for Evolving Materials Using a Recurrent Neural Network

Area of Science:

• Materials Science
• Optoelectronics
• Artificial Intelligence

Background:

• Complex refractive indices are crucial for light-matter interactions and designing photonic devices.
• Real-time tracking of refractive index changes in materials under external stimuli is challenging.
• Conventional methods like ellipsometry are often impractical for dynamic or multilayered systems.

Purpose of the Study:

• To develop an AI-powered framework for real-time extraction of complex refractive indices from reflectance spectra.
• To enable accurate in-operando optical characterization of materials undergoing external stimuli.
• To overcome limitations of traditional methods in dynamic and complex material systems.

Main Methods:

• Development of ReflectoRNN, an AI framework utilizing recurrent neural networks (RNN).
• Application of ReflectoRNN to analyze reflectance spectra under thermal, electrical, magnetic, or mechanical stimuli.
• Validation using generated datasets and experimental measurements on MoS2 and WS2 across various substrates and structures.

Main Results:

• ReflectoRNN achieved high accuracy with a median Pearson's correlation coefficient (PCC) of 0.998 and a relative accuracy score (RAS) of 0.968 on generated data.
• Experimental validation on MoS2 and WS2 demonstrated high accuracy and physical consistency across diverse substrates and multilayer stacks.
• Temperature-dependent measurements showed resonance energy matching Bose-Einstein predictions, confirming physical validity.

Conclusions:

• ReflectoRNN enables accurate, real-time in-operando extraction of complex refractive indices.
• The AI framework significantly advances optical characterization for dynamic materials and complex photonic structures.
• This work paves the way for automated material monitoring and accelerated materials discovery.