Although much research has been done on zirconium hydrides, one of the major challenges has been the lack of an efficient and unbiased method to analyze these hydrides in both 2D and 3D. This is where the use of deep learning (DL) algorithms comes in. DL methods have shown great potential in tackling various materials science problems, especially in recognizing and classifying microstructural features with a high degree of accuracy and reliability.

The aim of this research is to apply DL techniques to detect and extract hydride features from datasets, allowing for the development of functions that can quantify hydride characteristics such as their length, orientation, and connectivity. By achieving this, the project hopes to provide, for the first time, a reliable quantitative analysis of hydride microstructures and, ultimately, gain a deeper understanding of their precipitation behaviour and how it impacts the overall performance of the cladding material.

The polymer materials field has yet to fully embrace the potential of modern digital technologies, relying instead on time-consuming traditional methods for polymer discovery and development. To meet the growing demand for high-performance, sustainable materials, we must transform how polymers are developed.

My PhD project is focused on addressing this challenge by advancing artificially intelligent polymer synthesis platforms. I am developing an automated flow reactor that leverages machine learning to self-optimize within a defined parameter space, streamlining the polymer development process.

By the end of my research, I aim to contribute to a new era of innovation in polymer science while gaining expertise in polymer synthesis, machine learning, programming, and flow chemistry.

X-ray imaging techniques, using both synchrotron and laboratory sources, have emerged as a powerful tool to study dynamic phenomena in materials science, from metal solidification to the functioning of lithium batteries. However, the vast and complex data sets generated during time-resolved experiments present profound technical and practical problems for quantification, especially for multi-modal experiments and fast time resolve tomography where 10s of TB can be generated in a single experiment. Applying Artificial Intelligence (AI) to X-ray imaging has so far mainly focused on speeding up cumbersome human operations on uni-modal tomographic data, such as volume reconstruction and segmentation, and radiograph post-acquisition analysis. Little work has been carried out on multi-modal deep learning, which therefore remains a difficult challenge as well as an enormous opportunity. One reason deep learning has not been applied extensively to multi-modal data is that training deep models requires large-scale annotated datasets, e.g. millions of images with human supplied labels. This project will capitalise on recent developments in self-supervised training methods to overcome the need of large-scale datasets and develop AI models for multi-modal X-ray imaging. Deep learning models will be trained directly from the data without human-supplied annotations, and then adapted to new tasks with a relatively small number of human-supplied labels for training. The newly created models will be applied to the study of metal solidification and the extraction of information in real-time during in-situ experiments 

This PhD project aims to develop a machine learning (ML)-based method for rapid and efficient characterisation of dislocations in semiconductors using cathodoluminescence (CL) data. Dislocations degrade semiconductor performance, reducing yield, performance, and reliability. Current methods like atomic force microscopy (AFM) and transmission electron microscopy (TEM) can provide precise defect analysis but are slow and unsuitable for in-line inspection. CL microscopy, which maps optical and electronic properties through light emission under an electron beam, offers potential for faster defect analysis, but its industrial application is currently limited to signal intensity-based techniques, which only measure dislocation density without identifying specific types or properties.

This research harnesses CL’s hyperspectral imaging capabilities to capture detailed spectral and polarisation information, enabling a multidimensional view of dislocations. By integrating data from multiple high-resolution techniques, such as AFM, the project will train ML algorithms to classify dislocation types (edge or screw) and properties like the Burgers vector. These ML models will automate the analysis of hyperspectral CL data, offering a fast, scalable alternative to existing methods. The outcome will be a novel tool for real-time defect monitoring in semiconductor manufacturing, enhancing yield and device performance through improved defect characterisation.

Materials in jet engines undergo extreme conditions in terms of both temperature and loading. Failure of critical parts can be catastrophic, and therefore we must have reliable techniques at hand to both prevent failures and understand failures when they occur. In recent years, there have been significant advances in microscopy capabilities that can be used to assess crack path damage, and the plastic wake beneath a crack. In this project, we use electron backscatter diffraction to analyse the plastic wake in different loading regimes in titanium alloys, and under elevated temperature. This will be combined with transmission electron microscopy to investigate the fundamental failure mechanisms. This will be used to develop robust methods for analysing material failures to predict the failure mode. The development of hydrogen powered gas turbine engines presents a further question, in whether we can use these tools to assess hydrogen related failures. It is well known that hydrogen embrittles engineering alloys, so we must ensure the tools that are developed can be applied to a new chemical environment. This includes the use of cryogenic atom probe tomography to connect the chemical signature of the failure to the plastic wake field, to fully understand the failure mechanisms occurring. 

In response to significant global warming over the past decade, global efforts have been focused on developing innovative methods for energy generation to decarbonize the atmosphere, with the goal of achieving net zero in the coming decades. Many of these methods, such as renewable energy and electric vehicles, rely heavily on energy storage systems. As a result, batteries have become a critical technology for achieving net zero. However, current batteries must become lighter, last longer, and possess higher energy densities to accommodate the demands of various applications.

While research into enhancing battery performance is already underway, progress is hindered by human limitations in processing and analysing complex data. To accelerate this process, my PhD aims to leverage artificial intelligence to optimize battery performance by predicting ideal microstructural parameters for electrodes that can be practically manufactured.

This project focuses on developing sustainable electrocatalysts for ammonia synthesis using artificial intelligence (AI). It will use multi-scale simulations to generate high-quality catalysis datasets and train generative AI models to design novel catalysts, improving electrochemical ammonia production from air, water, and renewable energy. This collaboration between Imperial College and Entalpic aims to accelerate the design and optimisation of clean energy materials.

Solar energy is a cornerstone of Net-Zero. Metal halide perovskites (MHPs) are promising candidates for next-generation solar panels. However, their long-term stability is low and poorly understood, complicating their commercialisation. This project will develop an open-source AI system to assess the ‘health’ of MHPs across device-relevant areas, accessed through luminescence imaging. This will enable the community to predict, visualise, understand and overcome degradation.

The project aim is to support the industrial translation of emerging ultra-thin 2D semiconductor materials by developing smart metrology and best-practice approaches to fingerprint and monitor material quality characteristics across entire process pathways. The project brings together expert academic expertise at the University of Cambridge and the National Physical Laboratory (NPL) to form new international metrology strategies.

Nitride semiconductors in energy efficient LED light bulbs use nitride material grown on an foreign substrates such as sapphire. New applications of nitrides, in power conversion and control for electric vehicles and the future grid, require devices to be grown not on foreign substrates but on bulk nitride materials, particularly bulk AlN. Growth of bulk AlN crystals is challenging, and efficient characterisation of the resulting material is imperative to allow widespread commercial application.

Controlling particle properties is key to achieving desired material properties. The stability, downstream processability and bioavailability of a drug is dependent on the particle morphology, and surface characteristics, while electronic and photonic properties of crystals are anisotropic and are dependent on both particle size and the dominant crystal facets. Process control can help achieve property control to some extent, however, it is heavily dependent on the chemical environment and the crystallisability of the material. Hence, being able to predict such properties from big data can minimize experimental needs and can prove to be a more sustainable means to materials discovery.

This project seeks to exploit modelling and simulation methods to establish a holistic understanding of the effect of hydrogen environments on fatigue crack initiation, from nano- to micro-scales. As well as fundamental initiation mechanisms, key influential factors, such as surface condition, gaseous impurities and the presence of surface oxides, will be addressed, considering the design of new, and the assessment of existing materials. This will provide exploitable benefits to industries, and contribute to informing best practice, as we embark on a green transition.

Ferroelastic materials feature dynamic topological structures, like domain walls, that can host hidden phases with unique properties. This project combines 4D-STEM and EELS with machine learning to uncover and map these phases. You will develop multimodal deep learning tools to correlate structural and spectroscopic data, enabling automated detection of emergent behaviours. The project sits at the intersection of materials physics, AI, and advanced microscopy.

The goal is to discover new corrosion inhibitors that can be added to coatings for protection of metallic infrastructure. Initially, you will quantify two key performance indicators for candidate molecules, i.e., corrosion reduction and diffusion through the coating matrix. Subsequently, you will use these data, combined with interfacial analysis results, as input for computational methods, including artificial intelligence (AI), that will allow identification of new high-performance species.

Industry has expressed a clear need for better understanding of polymer quenching, particularly the physics and chemistry of polymer interactions with hot metal surfaces. From this will come predictive tools for modelling cooling rates in industrial scale parts, set-ups and conditions, to achieve desired microstructures and mechanical properties, whilst controlling residual stress. Experimental and modelling techniques will generate a wealth of data and provide robust predictive models.

This project will develop sophisticated data analysis workflows to effectively gain insights from large multidimensional data. It will use novel time lapse X-ray diffraction contrast tomography which non-destructively captures the 3D shape, crystallography and grain structure over time. This will be applied to the phenomenon of Hydrogen Embrittlement of dual phase stainless steel and high strength aluminium both of which have important applications in sustainable futures and attainement of net zero targets using H as an energy source and developing new high performance materials which are resilient to extreme environments.

In fusion reactors, materials experience extreme temperatures, stresses, and radiation damage. Safe operation requires identification of deformation patterns that are early warning signs of materials failure. These characteristic patterns result from the interaction of deformation mechanisms across multiple scales making detection via traditional analytical methods extremely challenging. This project will apply pattern recognition and machine learning techniques to a large database of experimental data to reveal early-stage fingerprints for damage hidden in the data.

The project will develop a microfluidic manufacturing process to create low-density metal and metal-oxide aerogels with bespoke micro-structural features. Aerogel fabrication will be accompanied by advanced 3D microstructure imaging (FIB-SIMS, nanoXRT) and the development of machine-learning-based data processing for quantitative image analysis. The most promising materials will be taken forward for advanced plasma-physics experimentation by the project partner AWE.