# Machine Learning-Enhanced Dual-Band Plasmonic Sensing for Simultaneous Qualitative and Quantitative Detection of Biomolecules in the Mid-Infrared Region

## TL;DR

This paper introduces a machine learning-enhanced plasmonic sensor that can detect multiple biomolecules simultaneously in the mid-infrared region.

## Contribution

The novel use of PCA machine learning enables versatile quantitative analysis of mixed biomolecules in plasmonic sensing.

## Key findings

- Strong coupling between plasmonic resonance and molecular vibration enhances absorption signals.
- PCA machine learning effectively determines molecular proportions in complex mixtures.
- Rabi splitting serves as both a biomolecule presence indicator and concentration measure.

## Abstract

Recently, sensing for biomolecules has become increasingly popular in the fields of environmental monitoring, personal health, and food safety. Plasmonic biosensors have been a powerful tool due to their high sensitivity and label-free operation. However, when it comes to molecules with different kinds and concentrations, detection technology and data processing remain a challenging task. In this study, we investigate the qualitative and quantitative detection of two kinds of biomolecules in the mid-infrared region simultaneously by the utilization of a plasmonic sensor. The strong coupling between each plasmonic resonance and the corresponding molecular vibration is found to significantly enhance the absorption signal of molecules, and the obtained Rabi splitting is not only a proof of molecular existence but also an indicator of molecular concentration. However, the amount of the molecular solution with a background refractive index in turn affects the plasmonic resonance position. In more general situations, it is not easy to achieve the match between plasmonic resonance and molecular resonance, and thus the quantitative detection by the Rabi splitting depth is not always feasible. Hence, we propose a machine learning algorithm called principal component analysis (PCA), providing a versatile approach for analyzing the proportion of each molecule in the mixture. Our work opens up new routes in noninvasive optical sensing and the integration of AI-driven data analysis further strengthens its potential for real-world applications.