A picture worth a thousand atoms: 2D image to 3D nanostructure mapping by merging simulated and measured data

Understanding how local atomic structure and long-range emergent magnetic and electronic properties in defective 2D materials are connected is critical for the development of next generation functional materials. However, modern atomically-resolved imaging techniques only give an integrated snapshot of the structure, without revealing the details of the three-dimensional morphology or the stability: There are many ways to arrange atoms that give essentially the same 2D image. The project will employ electronic structure calculations of well-characterized 2D materials, simulations of electron microscopy images, and machine learning methods to reconstruct the 3D atomic positions of materials from a 2D microscopy image. The student will work closely with experts at national spectroscopy and imaging facilities to deliver scientific software applicable to experimental imaging data.

Quantum, Atomistic, Physics, Chemistry, materials characterisation, catalysis.

Aluminium-steel fusion welding: What happens at the interface?

Aluminium and steel are widely employed metallic materials for automotive applications, such as in vehicle frames. Joining of these two dissimilar metals by fusion welding results in formation of brittle aluminium-iron compounds at the interface, which degrade the performance of the weld. Looking at the motion of individual atoms, we will use modelling techniques such as Molecular Dynamics (MD) and Kinetic Monte Carlo (KMC) combined with transmission electron microscopy analysis to study how iron and aluminium atoms react at the weld interface to find weld conditions where favourable intermetallic compounds form at interface.

Atomistic, Engineering, WMG, materials characterisation, alloys, molecular dynamics, kinetic, Monte Carlo.

AMPLIFY: Advancing Membrane Protein Learning for Improved Factorial Yield

Membrane proteins are vital for cellular function and are targets for over 50% of modern pharmaceuticals. Despite their importance, producing these proteins at scale remains a major challenge. Here we aim to AMPLIFY the production of membrane proteins through cutting-edge metabolic and molecular modelling tools. This project combines state-of-the-art protein engineering with AI-driven analysis to optimise yields and study protein stability. By mastering these techniques, you’ll contribute to breakthroughs in biomedical research and drug development, gaining invaluable skills in membrane protein science and computational (bio)chemistry. Join us to tackle this global challenge and AMPLIFY your scientific expertise.

Engineering, Chemistry, Life Sciences, Biological, Biomedical, Pharmaceuticals, Biomolecular modelling

This project aims to enhance the understanding of 3D atomic structures and defects in materials using advanced imaging and computational techniques. We will integrate super-resolution computational imaging (ptychography), chemical spectroscopy (EELS), and density functional theory (DFT) to achieve high-resolution reconstructions of atomic configurations. By fusing data from these methods, we will improve both lateral and depth resolution while reducing electron induced sample damage. Machine learning will be employed to refine our models iteratively, leading to precise predictions of material properties, particularly around defects. This research will provide valuable insights into the physical characteristics that impact material performance.

Boosting Battery Life with Hybrid Machine Learning of Degradation Mechanisms

From here to where? Mapping out crystal polymorph formation mechanisms using directed walks

Organic molecular crystals can form in different polymorphs that may exhibit vastly different properties, such as melting point, light absorption, solubility and pharmaceutical activity. However, predicting the mechanisms by which different polymorphs crystallize remains a frontier challenge for the chemical sciences.

In this project, we will develop a new simulation strategy to predict the key mechanistic pathways that form different polymorphs. Here, we will develop and test new strategies based on directed random walks in contact-map space to enable generation of intermediate ensembles associated with nucleation and crystallization to different polymorphs.

Advancements in materials synthesis have allowed the realization of many novel materials and their alloys, which are gradually finding their ways into numerous applications including energy, sustainability, medicine, novel computation, etc. A major direction of interest is their electronic properties. However, the accurate assessment and prediction of electronic transport is a highly challenging task. The project uses Machine Learning (ML), in combination with DFT and state-of-the-art Boltzmann transport methods, to predict, accelerate, and scale the computation of electronic properties of complex materials and their alloys. The richness of experimental data from literature and project partners will aid towards model validation.

materials characterisation, statistics, novel UQ, SciML, electronic devices, alloys.

Machine learning of energy barriers for reaction network discovery of drug-like molecules

In drug discovery, millions of molecules need to be screened for their viability as drug candidate, including their synthetic viability. Yields of chemical reactions are often limited by the formation of unforeseen by-products, which are not accounted for in synthesis planning. The exploration of kinetically accessible by-products requires the accurate prediction of reaction enthalpies and activation free energies for all relevant intermediates. In this project, a deep learning and generative design toolchain will be developed resulting in an ML model of reaction barriers. This will enable the development of more accurate and advanced high-throughput reaction network discovery and by-product prediction.

Quantum, Atomistic, Chemistry, maths, materials characterisation, statistical methods, novel UQ, SciML

Modelling interfacial flows in porous media: a hybrid asymptotic-computational approach

Interfacial fluid flows around solid obstacles and through porous materials are important in numerous applications, including carbon sequestration, materials science, filtration, and manufacturing. For instance, resin must be injected into a porous mesh, without trapping air bubbles, to manufacture composite materials. Interfacial flows are difficult to model and simulate accurately, and in porous media the multiple disparate lengthscales further complicate matters. In this project we will develop and use hybrid modelling approaches for moving fluid-fluid interfaces around obstacles, incorporating mathematical modelling, state-of-the-art asymptotic methods and high-fidelity numerical simulations, to investigate questions like how to minimise air trapping during composite material manufacturing.

Continuum, Maths, materials characterisation, statistics, novel UQ, SciML, composites.

Multimodal sampling with ballistic-style Markov processes: from atomistic water to rapid simulation of polymer folding

Many major problems in predictive modelling involve multimodal energy landscapes. For example, proteins in a biological cell stabilise in a variety of folded configurations – or modes. Cells function correctly in the low-energy mode, but rare misfolds at higher energy lead to cell malfunction. Capturing the misfolds is, however, a challenge because simulations jam in certain modes on long timescales. Biasing simulations away from visited configurations should resolve this problem, but convergence is poor due to numerical instabilities of state-of-the-art simulations. This project leverages the fast yet stable dynamics of ballistic-style Markov processes to produce rapid multimodal sampling of polymer models.

Atomistic, Engineering, Physics, biomolecular modelling, biomedical systems, pharmaceuticals, maths, statistics, novel UQ, SciML.

Physics-informed machine learning-based swelling models for future battery cells

Efficient batteries for automotive industry are critical for achieving net zero goals and the future of our planet. During their lifetime, those energy storage systems can experience complex electrochemical-thermomechanical phenomena that can result in their volumetric changes (so called swelling). Swollen batteries are at risk of rupturing which may significantly shorten their lifetime. Development of advanced computer models is critical for understanding and optimization of batteries against the swelling phenomenon. This project aims at developing scientific machine learning approaches based on the Bayesian paradigm and electrochemical-thermomechanical models in close collaboration with our project partner Jaguar Land Rover.

Probabilistic approaches to fusion reactor analysis: A fusion-specific T-effective model with applications to neutron and tritium transport

Achieving controlled fusion for power generation requires precise mathematical models to handle the inherent complexities and uncertainties of reactor systems. In nuclear fission, k-effective (effective multiplication factor) provides critical information about the reactor's state of criticality. For fusion reactors new stochastic models are needed to account for the distinct challenges posed by particle transport and fuel cycling, particularly tritium management.

Quantum, Atomistic, Engineering, Maths, Statistics, novel UQ, SciML, materials characterisation

Programmable fluids: Continuum models for designing multi-stable meta-fluids

Adding deformable capsules to fluids allows for the realisation of “meta-fluids” with programmable mechanical, thermal and optical properties [1,2]. This newfound tunability unlocks materials with multi-physical properties not otherwise possible with single-phase fluids. This has far-reaching implications for applications to smart robotics, mechanical computation and energy harvesting [1,2]. This is a transformational technology, in the same way that metamaterials have transformed wave physics in the last two decades by exploiting multi-scale structure to yield novel, previously impossible functionality. This project will use a combination of asymptotic approximation [3] and direct numerical simulation [4] to develop novel reduced-order models for meta-fluids.

Reconstructing extreme space weather events to quantify risks for modern society

When Things Won't Let Go: A Sticky Problem in Industrial Drying

Many industrial companies dry their chemical products into powders to aid wide transportation and distribution. This process often takes place in large scale driers where the material is heated and broken up mechanically with mixing blades. However, under certain conditions the process can break down as the material sticks to the edges of the drier, causing wastage of the product, and in extreme cases, deterioration of the drier itself. This research aims to develop a new multiphysics model to describe the material properties of these drying powders, with the aim of identifying the culprits of wall adhesion. The project is suitable for an applied mathematician, or physicist with a strong mathematical background.