Cellulose I crystallinity estimation using a combination of infrared spectroscopy and machine learning approaches

Area of Science:

• Materials Science
• Spectroscopy
• Machine Learning

Background:

• Cellulose crystallinity is a key factor influencing its properties and applications.
• Accurate and rapid methods for determining cellulose crystallinity are needed.
• Existing methods like X-ray diffraction can be time-consuming.

Purpose of the Study:

• To develop a novel, rapid, and reliable method for estimating the crystallinity index (CrI) of cellulose I materials.
• To integrate infrared (IR) spectroscopy with machine learning for CrI prediction.
• To establish a set of crystalline and amorphous standards for cellulose calibration.

Main Methods:

• Preparation of cellulose mixture samples with known CrI using microcrystalline cellulose, ball-milled MCC, xylan, and lignin.
• IR spectral data acquisition and preprocessing (second-derivative transformation, vector normalization).
• Application of machine learning models (Random Forest for feature selection, Multilayer Perceptron for prediction).

Main Results:

• Identification of the 1400-900 cm⁻¹ spectral region as highly informative for CrI prediction using Random Forest.
• Development of a Multilayer Perceptron model achieving high accuracy (R² = 0.959, RMSE = 0.052).
• Validation of the model on diverse cellulose materials, showing strong agreement with X-ray diffraction.

Conclusions:

• The integrated IR spectroscopy and machine learning approach offers a robust and accurate method for cellulose CrI estimation.
• This technique provides a faster alternative to traditional methods, benefiting cellulose science and material characterization.
• The study provides valuable calibration standards for future research in cellulose analysis.