Automated Workflows for Reconfigurable Nanomaterials Published in Advanced Functional Materials

Cross-disciplinary approaches, including lab automation, machine learning, and computer vision methods, open new possibilities in analysing complex colloidal systems with reversible features based on microstructural periodicities.

Researchers from Manufacturing Futures Lab: Mr. Yueyang Gao (First author), Dr. Maximilian O. Besenhard, collaborated with System Identification Group (Prof. Federico Galvanin) and AdReNa Group (Prof. Stefan Guldin, Corresponding author) in the Department of Chemical Engineering at UCL developed automated data-driven workflows for predicting and analyzing aggregation behavior of nanoparticles in liquid crystals, advancing the discovery of nanomaterial systems with programmable and reconfigurable features. Their latest breakthrough was published in Advanced Functional Materials, which was also selected for themed collections in hot topics of Artificial Intelligence and Machine Learning, and Gold. 

Gold nanoparticles (AuNPs) have gained prominence in chemical and biomedical fields as versatile nanoscale building blocks, Liquid crystals (LCs) offer a promising matrix for fundamental research and AuNPs applications, but this potent system exhibits complex and dynamic behavior due to colloidal instability. Thus, optimizing the solubility of AuNPs here remains difficult due to the interplay of multiple experimental variables, necessitating extensive combinatorial trials. 

In this study, an automated platform combined with a Design of Experiment (DoE) framework was employed to streamline the optimization, using robotics to accelerate traditional experimental trials. A random forest model, trained on a relatively small dataset, successfully predicted aggregate classifications with high accuracy. Aggregate behavior was further analyzed by UV-vis spectroscopy with automated data processing for feature reduction. These steps enhanced the closed-loop optimization by iteratively constructing a generalized additive model for predicting spectral features. AuNPs with optimized solubility were validated for hierarchical assembly during thermocycling tests. Computer vision methods were used to quantify the reversibility of LC-AuNPs systems, utilizing information entropy derived from pattern recognition algorithms to realize the high throughput analysis on pixel-level.

Yueyang Gao, the first author and PhD student at UCL Manufacturing Futures Lab (MFL), commented: “The automated data-driven workflows not only facilitate the design and analysis for comprehensive materials systems but also provide more chances for researchers to aggregate with families just like nanoparticles.”

Useful Links:

• Predicting Aggregation Behavior of Nanoparticles in Liquid Crystals via Automated Data-Driven Workflows - Published in Advanced Functional Materials
• Mr Yueyang Gao's UCL Profile
• Dr Max Besenhard's UCL Profile
• Prof Federico Galvanin's UCL Profile
• Prof Stefan Guldin's UCL Profile

Image credit

Mr Yueyang Gao
CC BY 4.0

---