The recent explosion of "big data" has positioned data analytics and machine learning techniques at the centre of social, marketing, medicine and manufacturing research. In nanomaterials discovery, data analytics tools are also playing a fundamental role to successfully tackle the exponential increase in size and complexity of functional nanomaterials. In this work, the intrinsic relationships between nanoscale features and functional properties of nanomaterials will be discussed for nanodiamonds, graphene nanoflakes[1-3] and nanoporous metal-organic frameworks (MOFs)[4-6]. Multivariate statistics are applied to identify the most relevant individual features in conjunction with the most representative nanostructures in nanoparticle ensembles and structural libraries. Meanwhile, machine learning techniques, from simple decision tree predictors to more complex suppert vector machines and neural networks, are utilized to predict the electronic properties of graphene nanoflakes and to identify “high-performing” nanoporous MOFs for carbon capture and storage applications. It will be demonstrated that simple predictors provide useful insights into the structure-property relations paradigm at the nanoscale, whilst more complex machine learning models are efficient tools to rapidly discriminate among potential candidate materials at a fraction of the traditional computational cost.
1. Fernandez, M. & Barnard, A. S. ACS Nano 9, 11980–11992 (2015).
2. Fernandez, M., Shi, H. & Barnard, A. S. J. Chem. Inf. Model. 55, 2500–2506 (2015).
3. Fernandez, M., Shi, H. & Barnard, A. S. Carbon 103, 142–150 (2016).
4. Fernandez, M., Boyd, P. G., Daff, T. D., Aghaji, M. Z. & Woo, T. K. J. Phys. Chem. Lett. 5, 3056–3060 (2014).
5. Fernandez, M. & Barnard, A. S. ACS Comb. Sci. (2016). doi:10.1021/acscombsci.5b00188
6. Aghaji, M. Z., Fernandez, M., Boyd, P. G., Daff, T. D. & Tom, K. Eur. J. Inorg. Chem. (2016) in press.