Identification of Important Sugar Binary Mixtures Found in Biorefineries Using Terahertz Time-Domain Spectroscopy.

Area of Science:

• Spectroscopy
• Biochemistry
• Machine Learning

Background:

• Terahertz (THz) spectroscopy is effective for analyzing biomolecules.
• Accurate sugar composition analysis is crucial for chemical synthesis and product development.

Purpose of the Study:

• To develop a framework for determining sugar compositions using THz spectroscopy.
• To evaluate machine learning models for predicting sugar content in binary mixtures.

Main Methods:

• Prepared solid binary mixtures of sugars (glucose-sorbitol, xylose-xylitol, glucose-fructose).
• Acquired THz spectra using a THz time-domain spectroscopy (THz-TDS) system.
• Applied four linear and nonlinear machine learning models, including support vector regression (SVR).

Main Results:

• Support vector regression (SVR) demonstrated superior performance in predicting sugar compositions.
• Achieved average and best root-mean-square errors (RMSE) of 5.42% and 2.83% w/w, respectively.
• THz spectroscopy showed potential for identifying functional isomers (glucose/fructose) but struggled with enantiomers (d-xylose/l-xylose).

Conclusions:

• THz spectroscopy coupled with SVR provides a promising, non-destructive method for sugar composition analysis.
• This approach enables rapid monitoring of chemical reactions in sugar synthesis.
• Further research may enhance THz spectroscopy's capability in differentiating molecular isomers.