Molecular Optimisation

Utilising machine learning to design functional molecules


Can machine learning help in designing the most effective functionalisation molecule from possibly infinite number of choices for Cellulose nanocrystal application?


Paving a new way to design targeted molecules at unprecedented speed.

Research Partners

  • Ford Motor Company, USA
  • Institute for Frontier Materials (IFM), Deakin University

Cellulose nanocrystals (CNCs) are unique renewable, biodegradable and non-toxic materials with impressive mechanical properties. When used to reinforce polymers, they can provide significant improvements in their mechanical properties.

However, before CNCs can be used as a commercially viable reinforcing agent we need to harmonize their interface properties with traditional polymers.


One key difference between CNCs and polymers is that CNCs are highly hydrophilic, whereas traditional polymers can be highly hydrophobic. To make them compatible, we need to impart strong hydrophobicity on the CNCs through surface modification. In this project we are optimising surface modifications of CNCs to improve compatibility with traditional polymers.

The project has two major steps: 1) Developing a physics-based nano-level simulation model of CNCs, led by Prof Tiff Walsh from Deakin’s Institute for Frontier Materials, and; 2) Using machine learning-based optimisation to develop new surface modification chemistry, led by A/Prof Santu Rana from A2I2.

This is one of a handful of projects in A2I2 through which we aim to demonstrate that hard science can be complemented with advanced machine learning technology to achieve commercially successful outcomes.

Learn more about other projects