About 55 years ago, Gordon Moore speculated that transistors will become smaller and more energy efficient every year. Since then, we have enjoyed exponentially increasing computer performance owing to what has been called the Moore's law. However, Moore's law is coming to an end and has already begun to disrupt the semiconductor industry. Absent the exponential performance and energy gains due to device scaling, industry has pivoted to hardware specialisation: targeting hardware to a specific computation class generally leads to orders of magnitude improvement in energy and performance. We are well and truly in the age of heterogeneous computing. A modern smartphone today has dozens of devices within a single chip, including CPUs, GPUs, and other accelerators. But efficiency hinges on reducing data movement between these devices; otherwise, it can seriously jeopardise the benefits of heterogeneous computing. Sadly, an analysis of Google workloads on a mobile device reveals that, on average, more than 60% of the overall energy is spent on moving around data. One promising approach to reducing data movement is called cache coherence. The cache coherence protocol, which automatically replicates data consistently, enables data to be accessed locally when it is safe to do so. Thus, it not only minimises data movement but it also does so in a programmer-transparent fashion. However, cache coherence protocols are notoriously hard to design and verify even for homogeneous multicores, where they have been deployed today. To make matters worse, we do not know how to keep the devices of a heterogeneous computer coherent correctly, in part because we do not yet understand what it means to be correct. In this project, we propose an entirely new way of designing coherence protocols. Instead of manually designing them and verifying them later, we propose an automatic method to generate them correctly. Our method is based on a new foundation of heterogeneous coherence called compound consistency models, which formally answers the question of how distinct coherence protocols should compose. If successful, the project will not only lift the major roadblock to efficient heterogeneous computing (data movements costs), it will also catalyse the burgeoning open hardware movement by democratising one of its trickiest components: cache coherence protocols.

Internet and telecoms are facing an explosive growth in data traffic, increasing at 50% per year. This requires the development of monolithic on-chip integration of electronics and photonics, which offers a massive reduction in both footprint and processing costs. Such a compact system will require a high power density and excellent high temperature tolerance. Monolithically integrating III-nitride based electronics and photonics on silicon on a single chip will represent the most promising approach to meeting the requirements in the telecoms regime. The photonic parts include active (laser diodes) and passive (photodetectors) components linked by waveguides, where the laser diodes are controlled by high electron mobility transistors. The electronic and photonic parts both need to meet the requirements for high power, high frequency and high temperature operation, as well as excellent temperature stability and robust mechanical properties. Conventional III-V semiconductors (GaAs or InP) suffer a number of fundamental limitations such as intolerance to high-temperatures, temperature sensitivity, limited power density capacity and fragility. They also exhibit high losses due to scattering (high refractive index) and multiphoton absorption. III-nitride semiconductors all have direct bandgaps and cover a vast spectral region from deep ultraviolet to infrared. Compared with conventional III-V materials, the III-nitrides exhibit major advantages in the fabrication of high power, high frequency and high temperature devices due to their intrinsically high breakdown voltage, high saturation electron velocity and excellent mechanical hardness. III-nitrides exhibit low free carrier absorption, negligible multiphoton absorption, low refractive index (2.3 for GaN compared with 3.5 for GaAs) and superior temperature stability of the refractive index (one order of magnitude higher than that of InP). Therefore, III-nitrides offer great potential to revolutionise current internet and telecoms and enable ultra-fast speed and ultra-broad bandwidths, going far beyond that so-far achieved in the telecoms regime (1.3-1.55 um). Up to now research on III-nitrides has mainly been confined to the visible spectral range but this is not a limit. III-nitrides based devices exhibit superior properties in terms of delivering the power/efficiency required for next-generation telecoms. This is important to the communications industry, which is expected to use 20% of the global electricity by 2025, where a large proportion (>30%) is consumed by the data centre cooling systems. Monolithically integrating III-nitride electronics and photonics on silicon on a single chip by direct epitaxy in the telecoms regime would therefore offer transformative performance. Our ambitious vision is to employ the two major leading epitaxial growth techniques (MOVPE and MBE) for III-nitrides, combining the leading-expertise established at Sheffield, Cardiff and Strathclyde along with a world-leading research team at Michigan in USA in order to demonstrate the first monolithic on-chip integration of III-nitride based electronics and photonics on silicon with operation in the telecoms regime. This is expected to revolutionise current internet and telecoms.

The UKRI Centre for Doctoral Training in Machine Intelligence for Nano-electronic Devices and Systems (MINDS-CDT) will operate as a centre of training excellence in the next generation of systems that employ Artificial Intelligence (AI) algorithms in low-cost/low-power device technologies: hardware-enabled AI. The use of AI in real-world applications through systems of interconnected devices (so-called Internet of Things) is increasingly important across the global economy. Various market surveys estimate the sector to be valued in the hundreds of billions, and project levels of compound annual growth of 25-30%. Applications of these technologies include smart cities, industrial IoT and robotics, connected health and smart homes. It is widely agreed that new advances in artificial intelligence and machine learning are key to unlocking the potential of these systems. Significant challenges remain, however, in the development of robust algorithms and coordinated systems that are efficient, secure, and work in concert with modern devices. Advances in electronics will soon hit atomic scales, requiring new approaches if we are to continue to improve hardware speed and power consumption. Novel nanotechnologies such as memristors have the potential to play a key role in addressing these challenges, but critical to their employment in real-world applications is how algorithms work in the context of device physics. Further, there are significant challenges around how resources available to devices (energy, memory, etc.) can more effectively adapt to the computational tasks at hand, again requiring us to think about how hardware and software work together. The MINDS CDT is unique in its cross-disciplinary research programme crossing emerging AI algorithms and models with advances in device technologies that underpin and enable their potential. To quote from one of our industry partners, "innovation is to come from software and hardware co-development" and that "this joined-up thinking as a potential game changer". The MINDS-CDT will train a substantial number of experts with the knowledge and skills to lead the development of this next generation of intelligent, embedded systems. The training programme will draw from both computer science and electronics expertise at the University of Southampton, and a substantial network of stakeholders from across industry, government and the broader economy. Core to our training ethos is the up-front investigation of the potential impacts of technological innovation on society, security and safety, and in the engagement of interest groups and the public in understanding the benefits as well as the risks of the use of these new developments in AI and technology for our society and economy. The processes we will use here include that all projects and research activities will be informed by in-depth impact assessment, and we will instigate an ambassadors programme for public engagement and, in particular, the engagement of underrepresented groups in AI and engineering.

The term microwave is used in reference to electromagnetic radiation with wavelengths ranging from about one meter to one millimetre. In the electromagnetic spectrum microwave wavelengths are shorter than those of radio waves but longer than those of infrared waves. Microwaves are used extensively in modern communication systems, including: mobile networks, WiFi, GPS, satellite TV, etc.. Other applications, include: heating, radar, imaging, etc.. The number of applications for microwaves is increasing due to the increasing use of electronic devices and the convenience of communication without wires. In the future microwaves will be used in 5G mobile networks, which will see the introduction of a multitude of new devices, all relying on communication via wireless signals. Those new devices and applications include: driverless cars, remote surgery, virtual reality, internet of things, etc.. Today most of the components within a system, operating at microwave frequencies, are designed specifically for that particular application. This increases the cost, and time required to bring a new product to market. In turn, this impacts the price which consumers pay for goods and services e.g. mobile handsets. In this research we ask the question; what if a communication system could be assembled from a collection of standardised bricks in just the same way that anything can be constructed from standard Lego(TM) bricks? Then the design task would reduce to that of devising and designing a suitable set of bricks with which to create a range of different systems. To some extent this already happens; for example, companies produce a range of frequency selective filters having different specifications, and one can select a filter for a particular application. However, the enormous variety of different systems means that a large number of different variations are required. So a huge amount of design effort is still required. In this research we consider what would happen if, we could devise a generic Lego(TM) brick that would assume different sizes and forms. This would enable us to construct any system from a collection of this single almost magical Lego(TM) brick. If this could be achieved the task of designing a complex microwave system, such as the radio within a mobile handset, would merely involve deciding how to assemble a collection of these "magic" Lego(TM) bricks to create the required system. The idea, although attractive, sounds like a fantasy because from our everyday experience we "know" that no object cannot mutate to assume any form and then hold that form, at will. Surely, such a concept is pure science fiction and the stuff of movies like the terminator... Well, no in fact it is not, since 2014 researcher have been working intensively on a new and exciting material which behaves in a way very much like the metal seen in the terminator movies. This material is a metal and yet it is also a liquid at room temperature. Excitingly it can be caused to move under direct electrical control and to hold its shape, at will. In this research we plan to use that material to a create this "magic" Lego(TM) brick which behaves as a universal microwave component. Being made from liquid the component can be flowed into different sizes and forms and thus we obtain 'liquid wires'. To create larger systems, we will simply need to decide how to join the bricks together so that they can operate in unison to perform more complex functions. Our research is highly interdisciplinary in nature and will benefit the U.K. economy across a wide range of different areas, including: chemistry, materials science, and engineering. The technology could revolutionise the way that communications systems are designed and built, resulting in entire new industries.

Semiconductors are able to efficiently convert electrical energy into light; this is the basis of light emitting diodes (LEDs) and semiconductor lasers. Such devices produce classical light, consisting of many trillions of photons every second. However there are applications in quantum computing and cryptography which require non-classical light, for example a regular stream of single photons or entangled photon pairs; two spatially separated photons which form a single quantum system. Such non-classical light can be created by semiconductor quantum dots; semiconductor nanostructures in which the size of the semiconductor in any dimension is no greater than a few 10's nanometres. Electrons trapped within a quantum dot are unable to move; resulting in dramatically different properties compared to conventional bulk semiconductors in which free electron motion is possible. In addition to the production of non-classical light quantum dots can be used to improve the performance of both lasers and solar cells. There are a number of approaches for the formation of quantum dots. The most studied is self-assembly where the dots form spontaneously on a semiconductor surface; this process is driven by the strain that results when the deposited semiconductor has a different atomic spacing to that of the underlying semiconductor. However the spontaneous nature of this process results in the quantum dots having a distribution in their shape and size; no two dots are identical. In addition controlling the position at which the dots form is very difficult. Recently the formation of quantum wires which grow vertically upwards from a semiconductor surface has been demonstrated. Growth of these wires is initiated either by initially depositing tiny metal droplets on the surface or by forming nanoscale holes in an oxide mask. The quantum wires can have lengths in excess of 1um and diameters below 100nm. During the growth of the quantum wire it is possible to change the semiconductor type and hence insert a small disk of a different semiconductor within the quantum wire. This disk forms a quantum dot and it is this new type of quantum dot that forms the subject of our research. These so-called nanowire quantum dots have a number of significant advantages in comparison to self-assembled ones. For example their position can be accurately controlled by placing the hole in the oxide mask at the desired position. There is also much greater control of the quantum dot shape and size; one consequence of this is the possibility to form many closely spaced identical dots within the wire. Such vertical stacking of quantum dots is not possible in the self-assembled system but is advantageous in lasers where a large number of quantum dots are required to achieve sufficient amplification of the light. In addition the nanowire acts as a cavity to confine photons, allowing the fabrication of nanoscale lasers. Nanowire quantum dots is a very immature field and significant growth development complemented by extensive optical and structural characterisation is required to optimise their properties for a range of applications. We will develop the system based on GaAs quantum dots in GaAsP nanowires grown by molecular beam epitaxy on silicon substrates. Growth on silicon is important as it provides the potential for integration with conventional electronics. Structures will be characterised by transmission electron microscopy and optical spectroscopy of single nanostructures. Following optimisation we will develop structures for a number of applications, including sources of single photons and entangled photon pairs, and nanoscale lasers. We will initially develop devices which are excited by light from a laser but a major later aim is to achieve all electrical devices.