Isolation of Cellulose Nanocrystals: Data-Driven Identification of Key Factors Affecting Yield and Morphology

Cellulose nanocrystals (CNCs) are rod-like colloidal nanoparticles derived from natural cellulose sources. Among the various methods developed for their isolation, the treatment of cellulose fibres with sulfuric acid remains the most widely used. Although conceptually simple, this process involves a complex interplay of parameters that can significantly influence the yield and morphology of the resulting CNCs. However, many of these critical parameters are either inconsistently reported or entirely overlooked in the literature, making it difficult to compare results across studies and leaving important aspects of the CNC isolation process uncertain. This thesis investigates the key factors involved in the sulfuric acid isolation of CNCs, influencing not only their yield but also their surface charge and morphology. A database was compiled from literature data and revealed to contain a bias in reported yields due to the frequent absence of a phase separation method in optimisation studies. Despite this, the dataset enabled the training of a neural network model incorporating reaction kinetics to quantify the relative influence of the isolation parameters on the reaction yields. This highlighted a significant dependence of the yield on the experimental context, likely due to differences in cellulose sources and to the neglect of certain experimental parameters. Among the overlooked parameters, the post-hydrolysis centrifugation step was selected for further experimental investigation. This study revealed that centrifugation plays a crucial role in inducing morphological changes in the CNCs, strongly affecting their liquid crystalline behavior. To facilitate further examination of the process while minimising experimental bias, a continuous laboratory-scale setup was also engineered. By combining data-driven modelling, experimental investigation, and process development, this work improves general understanding of CNC isolation, facilitating the optimisation of both industrial and laboratory processes for targeted applications.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2025

Author dc:creator

• Ballu, Kevin

Advisors dc:contributor.advisor

• Vignolini, Silvia
• Frka-Petesic, Bruno
• Lapkin, Alexei

• Automation
• Cellulose nanocrystals
• CNC
• Colloids
• Data-driven
• Engineering
• Kinetics
• Machine learning
• Morphology
• Neural network
• NN
• Physiclly informed neural network
• PINN
• Process development
• Self-assembly
• Soft-matter
• Sulfuric acid hydrolysis
• Yield

Licence

• https://www.repository.cam.ac.uk/bitstreams/c64e8325-6417-49fc-8331-8f97b428824e/download
• https://creativecommons.org/licenses/by/4.0/

Language dc:language

eng

Author Identifier

0000-0001-8641-5175

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/391707