Rapid Screening of Zeolites Using Computational and Machine Learning Approaches

The aim of this project is to develop software to computationally screen known zeolites (a family of > 200 microporous aluminosilicates) and assess their suitability as storage materials and molecular seiving of fine chemicals. The student will carry out his/her doctoral research at UCL supervised by Prof Ben Slater and will have the opportunity to undertake secondments in Johnson Matthey at the Sonning Common Technology Centre and Chilton.

Zeolites are established and essential catalysts in the chemical and (in particular) petrochemical industry. For new applications, such as scavenging of ethylene, or separation of isomers, one significant challenge is to identify the most suitable zeolite to perform this function, including assessing the stoichiometry of the structures. In principle, modelling approaches based on relatively simple classical forcefields models can provide high throughput assessment of all known zeolite structures through Monte Carlo and Molecular Dynamics approaches (such as RASPA (https://github.com/numat/RASPA2)). A next level sift would be to examine leading candidates materials using more sophisticated periodic density functional theory calculations (such as CP2K (www.cp2k.org)). In addition to devloping a piece of software to screen zeolites, it is anticipated that the student will explore machine learning approaches to, for example, provide an initial sift of the structures to priortise which structures should be assessed by classical models.