AI meets MD: Machine Learning Potentials for Nanomaterial-Liquid Systems

Whether it is the substantial cooling requirements of future data centres or energy-dense batteries for next-generation electric vehicles, the need for energy-efficient electronics cooling systems is ubiquitous. This is because while recent developments have produced ever-smaller and ever-denser devices, heat fluxes comparable to the surface of the Sun can be generated at hot spots, producing high temperatures that adversely impact their performance and raise risk of catastrophic failure. In the last decade and a half, novel 2D nanomaterials have been developed with unique thermal properties (e.g. ultrahigh thermal conductivity). These nanomaterials can be used to form surface coatings to enhance heat transfer from the extremely hot surfaces of electronic devices into the adjacent coolant liquid. 

However, our understanding of thermal transport at this nanomaterial/liquid interface is currently limited. For 2D nanocoatings, the nanomaterial can be either carbon-based (graphene nanoparticles or nanoflakes, nanopores, graphene oxide nanosheets etc), boron-based (boron nitride nanosheets, nanotubes, etc) or hybrid (e.g. boron carbon nitride). Similarly, while water is the most studied coolant liquid, realistic applications involve dielectric fluids (e.g. benzene, pentane). Molecular dynamics (MD) simulations represent a powerful tool to study such interfaces, but MD of nanomaterial/liquid interfaces require well-calibrated intermolecular potentials, which don’t currently exist. This project will rely on recent advances in neural networks to develop machine learning potentials (MLPs) for MD simulations of realistic nanomaterial/coolant-liquids and use these to gain fundamental insights into interfacial thermal transport. The goals are to:

1) run ab-initio molecular simulations to sample relevant nanomaterial/liquid interfaces.

2) construct new MLPs by using generated data from 1) and validate them.

3) use MLPs to run classical MD simulations and characterise thermal transport.

This PhD project will be based within the School of Engineering, University of Edinburgh. This PhD project will be supervised by Dr Rohit Pillai and Dr Eleonora Ricci, and the successful applicant will join an active, friendly, and collaborative research group (see https://multiscaleflowx.github.io/). Our group makes extensive use of ARCHER2 – the UK’s national supercomputer, which is based in Edinburgh. This PhD will give the successful applicant the skills and experience to become a future leader in either academia or industry. The supervisors will provide the successful applicant with exceptional research and training opportunities, including:

• regular weekly meetings to discuss the research progress.

• opportunities for travel to participate in workshops/summer schools dedicated to advanced computational methods, as well as present results in international conferences.

• training and experience in state-of-the-art engineering research.

• mentoring from other investigators and experienced postdoctoral researchers.

• exceptional career development opportunities with strong institutional support of early career researchers.

Minimum entry qualification - an Honours degree at 2:1 or above (or International equivalent) in a relevant science or engineering discipline, possibly supported by an MSc Degree. Further information on English language requirements for EU/Overseas applicants.

Tuition fees + stipend are available for Home/EU and International students

Further information and other funding options.