PhD opportunity
Deep diffusion models for finding order in chaos
Unfunded
31 January 2027
- Funding – China Scholarship Council PhD Programme
- Application deadline 31st January 2027
- Standard Entry date - September 2027
- Applicants are expected to have a degree (equivalent of Honours or Masters) in a relevant discipline.
Many physical systems governed by partial differential equations (PDEs) exhibit chaotic behaviour, whereby even small uncertainties in their initial conditions can lead to dramatically different outcomes. This makes long-term prediction challenging, even when the underlying governing equations are known precisely. A particularly important example is turbulence in fluid dynamics, governed by the Navier–Stokes equations. Despite the apparent complexity of turbulent flows, their dynamics are often organised around a hidden structure of relatively simple, dynamically unstable solutions, including equilibria, travelling waves, periodic orbits and invariant tori. Identifying these solutions is important for understanding the fundamental mechanisms underlying turbulence, developing reduced descriptions of fluid flows, and improving our ability to analyse and predict complex dynamical systems.
Numerical methods for computing these invariant solutions are now well established. However, they typically rely on an initial guess that is sufficiently close to the desired solution for the numerical algorithm to converge. Finding such initial guesses, particularly in high-dimensional and strongly chaotic systems, remains a significant challenge. This project will investigate whether modern generative artificial intelligence, and in particular deep diffusion models, can provide an effective way of exploring the state space of chaotic systems and identifying previously unknown invariant solutions.
Building on the supervisor's recent work using diffusion models to identify periodic orbits in two-dimensional Navier–Stokes flow, the project will develop and extend these methods to more complex settings. This will include investigating three-dimensional turbulent flows and broadening the range of solutions that can be identified, with a particular focus on equilibria, travelling waves and invariant two-tori. The research will address several key questions: Can diffusion models learn the underlying structure of turbulent dynamics sufficiently well to generate useful initial guesses for invariant solutions? How effectively can these methods scale to higher-dimensional systems? Can generative models uncover dynamically important solutions that are difficult to find using conventional numerical approaches? As the field of generative AI develops, the project will also explore opportunities to incorporate new architectures and machine-learning techniques.
The project offers an interdisciplinary training opportunity at the interface of applied mathematics, computational fluid dynamics and artificial intelligence. The candidate will develop expertise in nonlinear dynamical systems, numerical methods for PDEs, deep learning and generative modelling. They will also gain practical experience in scientific programming, GPU computing and high-performance computing (HPC), making use of the computational resources available at the University of Dundee.
The research has potential applications beyond fluid dynamics, as the underlying methodology could be adapted to identify coherent structures and dynamically important solutions in other complex systems governed by PDEs. There may also be opportunities to engage with researchers and stakeholders working in computational science, engineering and AI, where understanding and modelling complex dynamical behaviour is of growing importance.
Candidates should have a working understanding of PDEs and neural networks. An interest in, or previous experience of, fluid dynamics and nonlinear dynamical systems would be advantageous, but the project provides scope to develop expertise across these areas.
How to apply
Please apply through the link below, you may contact the principal project supervisor to discuss your interest further, see supervisor details below.
Related PhD programme
- Type
- PhD funding
China Scholarship Council (CSC) Programme
An opportunity for Chinese Students to undertake a PhD programme in any research field in Medicine, Life Sciences, Science and Engineering