Quantitative analysis of bisphenol analogue mixtures by terahertz spectroscopy using machine learning method.

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning

Background:

• Quantitative analysis of complex mixtures presents significant challenges in spectral analysis.
• Bisphenol A (BPA) is a widely used chemical facing replacement due to toxicity concerns, necessitating reliable analytical methods.
• Accurate identification and quantification of BPA in mixtures are crucial for safety and regulatory compliance.

Purpose of the Study:

• To develop and validate a novel strategy for determining the composition of BPA within its analogue mixtures.
• To explore the efficacy of terahertz (THz) spectroscopy coupled with machine learning for quantitative analysis of bisphenol mixtures.
• To establish a robust method for predicting BPA concentrations in unknown samples.

Main Methods:

• Terahertz (THz) spectra of four bisphenol components were acquired.
• A Support Vector Regression (SVR) model was employed to learn the relationship between spectral frequency and BPA concentration.
• A hold-out validation scheme was utilized to reconstruct absorption spectra and assess model performance.

Main Results:

• The SVR model accurately predicted BPA concentrations in mixtures, achieving a decision coefficient (R²) of 0.98.
• Absorption spectra of bisphenol mixtures were successfully reconstructed, demonstrating the model's predictive power.
• The combined THz spectroscopy and SVR approach proved robust and accurate for quantitative mixture analysis.

Conclusions:

• Terahertz spectroscopy in conjunction with Support Vector Regression (SVR) offers a powerful and accurate tool for quantitative analysis of complex bisphenol mixtures.
• This methodology shows significant potential for widespread industrial applications requiring precise chemical composition determination.
• The study highlights the robustness and accuracy of THz spectroscopy and SVR for reliable BPA quantification.