Noise-Aware Machine Learning for Multimetal Surface Plasmon Resonance Parameter Estimation from Angular Reflectance Curves

Abstract Surface plasmon resonance (SPR) sensors are widely used for refractive index sensing due to their high sensitivity to changes in the surrounding medium. However, extracting multiple physical parameters from an SPR reflectance curve remains an inverse problem. This work investigates machine learning-assisted inverse modelling of SPR sensors in the Kretschmann configuration using reflectance spectra generated through the Transfer Matrix Method (TMM). Machine learning models are employed to estimate the analyte refractive index and metal thickness, while a metal classification model identifies the material. A preliminary benchmark was performed to compare the machine learning approach with physics-based inverse modelling using TMM. On a noise-free held-out set of 60 samples, the machine learning model achieved a mean absolute error (MAE) of 0.003884 in refractive index and 0.21065 nm in metal thickness. The physics-based inverse model achieved errors of approximately 2.02 × 10⁻¹¹ in refractive index and 2.88 × 10⁻⁸ nm in thickness. The machine learning approach provided a prediction time of 4.574 ms, compared with 44.038 ms for the physics-based inverse method, corresponding to an approximately 9.63-fold speed advantage in this benchmark. These preliminary results demonstrate the trade-off between prediction accuracy and computational speed. Further evaluation under noisy conditions and across a broader parameter range is required to establish the robustness and generalisation capability of the proposed approach.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-13
DOI
https://doi.org/10.21203/rs.3.rs-11711593/v1
Primary Topic
Plasmonic and Surface Plasmon Research
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Noise-Aware Machine Learning for Multimetal Surface Plasmon Resonance Parameter Estimation from Angular Reflectance Curves

Mayank Goswami, Honey Thapa
4 links
Research Square
Plasmonic and Surface Plasmon Research
preprint

Noise-Aware Machine Learning for Multimetal Surface Plasmon Resonance Parameter Estimation from Angular Reflectance Curves

Mayank Goswami, Honey Thapa
preprint en

Abstract

Abstract Surface plasmon resonance (SPR) sensors are widely used for refractive index sensing due to their high sensitivity to changes in the surrounding medium. However, extracting multiple physical parameters from an SPR reflectance curve remains an inverse problem. This work investigates machine learning-assisted inverse modelling of SPR sensors in the Kretschmann configuration using reflectance spectra generated through the Transfer Matrix Method (TMM). Machine learning models are employed to estimate the analyte refractive index and metal thickness, while a metal classification model identifies the material. A preliminary benchmark was performed to compare the machine learning approach with physics-based inverse modelling using TMM. On a noise-free held-out set of 60 samples, the machine learning model achieved a mean absolute error (MAE) of 0.003884 in refractive index and 0.21065 nm in metal thickness. The physics-based inverse model achieved errors of approximately 2.02 × 10⁻¹¹ in refractive index and 2.88 × 10⁻⁸ nm in thickness. The machine learning approach provided a prediction time of 4.574 ms, compared with 44.038 ms for the physics-based inverse method, corresponding to an approximately 9.63-fold speed advantage in this benchmark. These preliminary results demonstrate the trade-off between prediction accuracy and computational speed. Further evaluation under noisy conditions and across a broader parameter range is required to establish the robustness and generalisation capability of the proposed approach.

Research Square
Indian Institute of Technology Roorkee (IN)
Plasmonic and Surface Plasmon Research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon