Research Fellow in Machine Learning for Materials Design

at  University College London

ABOUT US

The Chemistry Department at UCL is one of the top-ranked departments in the UK, with 100% of its outputs ranked as being world-leading (4) or internationally excellent (3) in the recent REF2021. The Department is committed to supporting excellence in both research and teaching. The department offers undergraduate BSc and MSci programmes in Chemistry and currently teaches 700 undergraduates registered in Chemistry as well as students who select Chemistry on the Natural Sciences programme and first year Chemistry for life scientists. The Department also offers a number of Postgraduate Taught Masters courses with about 100 students per year. The Department has an overall PhD student school of around 200 students. The Chemistry Department has over 60 members of academic staff carrying out world-leading research. We specialise in areas of organic synthesis, chemical biology, computational chemistry, nanotechnology, inorganic and materials chemistry, physical chemistry and chemical physics. The department research income derived from many sources including UKRI (EPSRC, BBSRC, MRC, and NERC), European Commission and a wide range of charities and industrial partners in the UK, Europe and the USA. Details about our research can be found on the departmental website: http://www.ucl.ac.uk/chemistry

The post is funded through Prof. Butler’s grant: Designing and optimizing polar photovoltaics with physics informed machine learning. The aim is to design new polar materials, with light absorbing properties that can exploit the presence of spontaneous polarisation to enhance photovoltaic performance. The appointee will be developing new machine learning methods to predict polarisation and optical properties in crystalline materials. In this field (as in much of materials science) machine learning faces the challenge of relatively small datasets on which to train. To overcome this problem, we will use the latest developments in physics informed machine learning, where physical biases, for example symmetry of the system or known boundary conditions, are built into the ML model to greatly improve data efficiency. The project is closely linked to the research activity of Prof. Joe Briscoe, who’s group will attempt synthesis campaigns for materials predicted on this project. The appointee will also work closely with the UK’s Physical Sciences Data Infrastructure (https://www.psdi.ac.uk/) to develop data and model resources that will be used by the wider materials discovery community.