Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.

This project will bring exciting advances to X-ray imaging by revealing the true nature of materials buried in 3-dimensional scans. The main limitation of conventional X-ray absorption imaging is that the image forming signals are a function of the attenuation coefficient, which tells us almost nothing about the chemical or crystallographic structure of the object under inspection. However, it is well understood that if diffracted flux, rather than the transmitted X-rays, is collected then slice images may be reconstructed using similar algorithms to conventional computed tomography (CT). The measurement of the energy or wavelength of the diffracted X-rays together with their associated diffraction angles enables the calculation of crystallographic parameters to identify, for example, the material phase of a sample. Scientists and engineers routinely measure diffracted flux from carefully prepared samples in instruments called diffractometers. Typically, this 'molecular fingerprinting' process uses relatively soft radiation and long inspection times of which both are impractical for security and in vivo diagnostic imaging. Despite significant efforts over the decades, there is little evidence of the 'gold standard' specificity and sensitivity achieved in laboratory settings being realised in time critical, commercially viable 3-dimensional imaging technologies. For example, the security screening industry has recognised the potential for X-ray diffraction as a 'gold standard' probe since the early 1990s. The challenge in this sector includes identifying powders, liquids, aerosols, and gels buried amongst the clutter of everyday objects in security scans of luggage. State-of-the-art CT spectroscopic scanners are limited fundamentally and are unable to deal adequately with homemade explosives. A main limitation of using diffracted radiation is that the signals are often orders of magnitude weaker in comparison with the primary incident beam. This fundamental limitation leads to long inspection times i.e. minutes or hours per point measurement, which in general is impractical for imaging. We have previously demonstrated a focal construct geometry (FCG) method where a hollow or conical shell beam produces high-intensity patterns or caustics in the diffracted flux from a sample. The bright caustics enable high-speed measurements that can be deconvoluted to form depth-resolved sectional images. Our novel method enables spatial features much smaller than the diameter of the interrogating beam to be resolved accurately in the reconstructed images. In keeping with standard computed tomography, FCG tomography in absorption and diffraction both use similar reconstruction principles. In this project, we propose reducing the total number of X-ray measurements and X-ray dose by more than 90% by applying sporadic sampling to FCG absorption/diffraction signals. We use a state-of-the-art flat panel X-ray source with multiple X-ray emission points optically coupled to energy resolving detectors. We treat the array of emitters as a virtual or spatially offset linear array (SOLA) to implement sporadic sampling independently of the minimum separation between emitter points (limited by the emitter physics) and to minimise crosstalk between measurements. We expect our method to enable the collection of diffraction and absorption signals at the same scan rate to realise depth-resolved material specific imaging. A successful demonstration of our method would establish a platform technology scalable in both X-ray energy and inspection space. This work will maintain the UK at the forefront of these unique and exciting scientific developments in security and diagnostic imaging.

The critical importance of capabilities for semiconductor research in the UK is recognised as part of a national strategy, as stressed in the recent BEIS Report 'The semiconductor industry in the UK'. Particular strength in research is centred around a number of cleanroom facilities located at academic institutuions. The University of Southampton hosts a range of cutting-edge nanofabrication tools which enable a range of research activities in electronic and photonic devices. Fabrication of semiconductor devices and circuits becomes cost effective when processed on a large wafer. However, process efficiency can only be achieved if an ultra-high-resolution scanning electron microscope (SEM) with material characterisation system is available to provide high throughput feedback results to improve fabrication and facilitate novel process development. Manually operated SEMs are a common imaging tool for characterisation used in academic research but automated in-line imaging of wafers throughout a process flow is required to achieve fast imaging and shorten inspection time from fabrication processes. The aim of the proposal is to acquire an ultra-high-resolution SEM (UHR-SEM) capable of material characterisation for wafers up to 200 mm in diameter at the University of Southampton. As device feature sizes are reduced, dimension and performance variations across the wafer become an issue which must be mitigated at the early stage of the fabrication. Therefore, the proposed UHR-SEM will be unique within the UK academic landscape since it will perform automated in-line imaging and analysis of entire wafers up to 200 mm in diameter at sub-nm resolution. The system will also have a low landing voltage on samples to reduce surface damage during imaging of delicate devices and patterned resists, as well as a good depth of focus for the inspection of thick multi-stack materials. The UHR-SEM will address the main challenges in large wafer imaging such as generating relevant surface metrology information at nanoscale dimensions and creating a detailed map showing various material parameters such as chemical composition and defect distribution.

Machine learning (ML), in particular Deep Learning (DL) is one of the fastest growing areas of modern science and technology, which has potentially enormous and transformative impact on all areas of our life. The applications of DL embrace many disciplines such as (bio-)medical sciences, computer vision, the physical sciences, the social sciences, speech recognition, gaming, music and finance. DL based algorithms are now used to play chess and GO at the highest level, diagnose illness, drive cars, recruit staff and even make legal judgements. The possible applications in the future are almost unlimited. Perhaps DL methods will be used in the future to predict the weather and climate, of even human behaviour. However, alongside this explosive growth has been a concern that there is a lack of explainability behind DL and the way that DL based algorithms make their decisions. This leads to a lack of trustworthiness in the use of the algorithms. A reason for this is that the huge successes of deep learning is not well understood, the results are mysterious, and there is a lack of a clear link between the data training DL algorithms (which is often vague and unstructured) and the decisions made by these algorithms. Part of the reason for this is that DL has advanced so fast, that there is a lack of understanding of its foundations. According to the leading computer scientist Ali Rahimi at NIPS 2017: 'We say things like "machine learning is the new electricity". I'd like to offer another analogy. Machine learning has become alchemy!' Indeed, despite the roots of ML lying in mathematics, statistics and computer science there currently is hardly any rigorous mathematical theory for the setup, training and application performance of deep neural networks. We urgently need the opportunity to change machine learning from alchemy into science. This programme grant aims to rise to this challenge, and, by doing so, to unlock the future potential of artificial intelligence. It aims to put deep learning onto a firm mathematical basis, and will combine theory, modelling, data, computation to unlock the next generation of deep learning. The grant will comprise an interlocked set of work packages aimed to address both the theoretical development of DL (so that it becomes explainable) and the algorithmic development (so that it becomes trustworthy). These will then be linked to the development of DL in a number of key application areas including image processing, partial differential equations and environmental problems. For example we will explore the question of whether it is possible to use DL based algorithms to forecast the weather and climate faster and more accurately than the existing physics based algorithms. The investigators on the grant will be doing both theoretical investigations and will work with end-users of DL in many application areas. Mindful that policy makers are trying to address the many issues raised by DL, the investigators will also reach out to them through a series of workshops and conferences. The results of the work will also be presented to the public at science festivals and other open events.