Presentation Information
[8a-N302-11]Machine learning-based dispersion optimizer for carbon nanotubes
〇Hirokuni Jintoku1 (1.AIST)
Keywords:
Machine learning,Carbon nanotube,Dispersion
Designing high-quality carbon nanotube (CNT) dispersions requires the simultaneous optimization of dispersant chemistry, solvent environment, and processing conditions. However, experimental exploration of this combinatorial space is slow and often nonsystematic. In this study, we developed a machine-learning-based CNT dispersion optimizer using a dataset CNT dispersions prepared by several conditions. Molecular descriptors, experimental variables, and solvent–dispersant similarity metrics were used as input features to predict two key outputs: an optical-microscopy-derived dispersion score and a Raman-based crystallinity index. The XGBoost model achieved screening-grade accuracy and was used for virtual screening of formulation and process conditions. Experimental validation confirmed that the model could identify promising combinations within the learned domain. These results demonstrate that machine learning can reduce empirical trial-and-error and guide CNT dispersion design for downstream applications such as films, composites, and printed electronics.