Available Projects
The project list below helps self-funded and externally sponsored applicants (scholarship) connect with a prospective supervisor.
Applications are welcome from those who can support themselves or have funding already arranged. Such applications will go through the same level of academic assessment. Please see the fee arrangementsLink opens in a new window for PGR students.
From wildfire smoke to alien skies: replacing a 1910 formula with physics-informed machine learning
A formula fitted in 1910 to oil drops still decides how climate models, pollution monitors and engineering software predict the motion of tiny airborne particles. But real particles such as soot, ash, viruses and nanoplastics are not spheres. Building on new theory from our group, you will combine fast molecular-level simulations with physics-informed machine learning to create an open tool that predicts how any nanoparticle moves, in any gas, from Earth's air to the skies of other worlds.
Primary supervisor: Duncan Lockerby - Duncan.Lockerby@warwick.ac.uk
Co-supervisor:TBD
Project detail:
In 1910, Ebenezer Cunningham wrote down a correction to explain why tiny particles fall through air faster than the equations of fluid dynamics predict. More than a century later, a version of his formula, with coefficients fitted to experiments on oil drops and latex spheres, is everywhere: in international standards for air-quality monitors, in commercial engineering software, in climate models, in studies of airborne disease transmission, and even in models of the clouds of exoplanets.
The trouble is that it was fitted to spheres in air. Real airborne nanoparticles are rarely spheres: soot from engines and wildfires forms branching, chain-like clusters; volcanic ash and mineral dust are jagged; viruses, nanoplastics and engineered nanomaterials come in every shape. For these, the formula is a guess, and the guess feeds directly into predictions of how long smoke stays aloft, how much sunlight it absorbs, and where it lands in our lungs.
The formula survives because the true answer is expensive. Particles this small move through air that no longer behaves like a smooth fluid: beyond the Navier–Stokes equations, the gas must be described molecule by molecule using kinetic theory, and doing so for a complex shape can take hours of computing per particle. Climate models need answers for billions of particles, instantly.
Recent work in our group has shown that a simple formula built entirely from physics, with nothing fitted to experiment, matches the century-old standard for spheres in air and outperforms it for other gases and shapes. This project takes that idea much further by combining it with machine learning. You will use new, fast kinetic-theory solvers to generate high-quality data for particles of many shapes, from smooth spheroids to realistic soot clusters, and train machine-learning models that respect the known physics by construction. The goal is a fast, open tool that predicts how any nanoparticle moves, settles, diffuses and exchanges heat, in any gas: from Earth's atmosphere to the hazy skies of other worlds.
You will gain advanced skills in kinetic theory, high-performance computing and scientific machine learning. The project suits a graduate in engineering, physics, mathematics or computer science who is excited by both fundamental physics and modern data-driven methods.
How to apply for admission: www.warwick.ac.uk/pgrengineering
The University of Warwick provides an inclusive working and learning environment, recognising and respecting every individual’s differences. We welcome applications from individuals who identify with any of the protected characteristics defined by the Equality Act 2010.