Supervisors: Prof. Duncan Lockerby (Eng.), Prof. Julian Gardner (EE), Dr. Radu Cimpeanu (Maths)

Summary:

The World Health Organisation recently classified air pollution as “the single biggest environmental threat to human health”. The airborne particulate matter thought to be most harmful is at the nanoscale – particles so small that they can evade our respiratory defence systems. Evidence indicates that the shape of nanoparticles has a big influence on health outcomes, but currently there are no ways to detect this property in isolation. This project will combine continuum fluid-mechanics models with probabilistic particle simulators to train a predictive tool capable of inferring shape, and other properties, from measurable quantities and limited experimental data.

Background:

This project will, for the first time, incorporate nanoscale flow physics within a Brownian simulator of particle motion in channels. In this largely unexplored space we can expect that a complex interplay of multiscale physics will determine nanoparticle dynamics and deposition: gas-kinetic effects (e.g. velocity slip and kinetic boundary layers); translational and rotational diffusion (6-axis Brownian motion); and Brownian drift (an average migration towards particle positions and orientations having greater mobility).

The project will integrate a diverse range of simulation approaches and tools: stochastic simulation of Brownian motion (stochastic ODEs); continuum solutions to advection-diffusion equations; and boundary-element-type methods for solving linear PDEs. These models will feed into Machine Learning libraries to perform the inference task of assessing aerosol properties, with expressible certainty, from practically measurable quantities.