Thesis title: Frontiers in Magnetism

人们经常预言摩尔定律的终结：早在2004年，为了防止集成电路过热，每年递增的计算机时钟速度被冻结在4GHz左右，这就预示着摩尔定律的终结。这个僵局的一个潜在解决方案是引入新的计算硬件，其操作基于对准粒子的操作，准粒子可用于以最小的能源成本存储、传输和处理信息。最有用途和前途的准粒子是磁振子。它已经在一系列模拟和数字计算设备中证明了它的能力。尽管如此，目前还没有为磁计算设定硬性和快速的总体范式。这就是我们提出的研究的吸引力和动机。选择最终的磁振子计算架构是可能的。最直接的结果是，它将改变我们目前使用的基于阿贝尔运算和布尔代数的计算范式。在不太可能的情况下，磁计算的潜力会走到这一步，但它仍然是计算技术的一个变革步骤，它将使布尔计算更快，使用类似或更少的空间，并且在相同的计算能力下消耗的热量减少三个数量级。一个更有野心的问题——这篇论文的动机——是磁振计算的潜在性能是否能超越硅所提供的。为了解决这个问题，我们借鉴了量子计算中的方法，旨在对磁振子器件的行为进行全面的代数理解，并利用这种理解来对磁振子系统架构进行严格的研究。我们的策略是在三个现有或预计的磁振学领域（波计算设备，量子式逻辑门和磁振模拟设备）中对可用的代数工具进行数学检查和形式化，并使用这种方法开发这些基本代数工具能够支持的更高级别的新型计算机体系结构。我们旨在解决的关键问题是：1。该技术的可用扇出是什么：即将一个逻辑设备的输出耦合到多个顺序设备的输入的能力？这是决定这项技术最终能力极限的关键考虑因素。在实践中，我们如何利用磁振子器件的能力，在不同频率通道上并行运行的多个数据流上同时应用不同的操作？我们能否在同一块硬件上依次运行这些数据流，并对每个通道上的数据执行不同的逻辑操作？具体来说，我们能否将一个硬件的输出作为输入注入同一硬件，但在不同的通道上？我们如何安排这些数据流的物理指示，使它们耦合到正确的处理通道而不产生串扰？在这种情况下，我们可以使用我们在最近的磁振子相位共轭工作中制造和测试的非简并四波混频构建块吗？它有两个有用的特性：输出相对于输入是频移的，并且它沿着完全相同的物理路径跟踪。我们怎样才能将这样的磁器件彼此连接起来，又与传统的电子器件连接起来，既节能又能可靠地将数据从波格式传输到脉冲格式？我们是否有能力制造出与量子计算风格逻辑元件具有类似代数性质的磁振子器件，这些器件是否可以用于改进或简化现有的磁振子逻辑配置？模拟磁振子处理器的作用是什么？它们能否成功地与数字波技术集成？合作者：生态效益与EPSRC目标的一致性：新的计算范例

The end of Moore's Law is frequently predicted: the harbingers have been with us since 2004, when the progressive annual increase in computer clock speeds was frozen at around 4GHz to prevent the integrated circuits from overheating. A potential solution to this impasse is the introduction of new computing hardware whose operation is predicated on manipulating quasiparticles that may be used to store, transfer and process information with minimal energy cost. The quasiparticle with the most versatility and promise is the magnon. It's already proved its capabilities in a range of analogue and digital computing devices. Still, no hard and fast overall paradigm has yet been set for magnonic computing. Therein lies the attraction and motivation for our proposed research. A selection of eventual magnon computing architectures are possible. The most straightforward outcome is that it will transform the computing paradigms that we currently use and that are based on Abelian operation and Boolean algebra. In the unlikely event that magnonic computing potential goes this far and no further, it will still be a transformative step in computing technology which will enable Boolean computing to happen faster, use similar or smaller amounts of real estate and dissipate three orders of magnitude less heat for the same computing power. A much more ambitious question - the motivation of this thesis - is if the potential performance of magnonic computing could go beyond what silicon offers. To address this we are borrowing from approaches taken in Quantum Computing, aiming to develop a thorough algebraic understanding of the behaviour of magnonic devices and exploit this understanding to have a rigorous approach to magnonic systems architecture. Our strategy is to examine and formalise mathematically the algebraic tools available to us in three existing or projected areas of magnonics - wave computing devices, quantum-style logic gates and magnonic analogue devices - and to use this approach to develop higher levels of novel computer architectures that these basic algebraic tools are capable of supporting.Key questions we aim to address are: 1.What is the available fanout of this technology: that is the ability to couple outputs of one logic device into the inputs of more than one sequential device? This is a pivotal consideration in deciding the limits of what this technology is ultimately capable of.2.How in practice can we exploit the ability of magnon devices to apply different operations simultaneously to multiple datastreams running in parallel on different frequency channels?3. Can we run those datastreams sequentially through the same piece of hardware and perform different logical operations to the data on each passage. Specifically, can we take an output of a piece of hardware and inject it as an input to the same piece of hardware, but on a different channel?4.How can we arrange the physical signposting of such datastreams such that they get coupled to the correct processing channels without crosstalk? In this capacity, can we use the non-degenerate four wave mixing building block that we made and tested in our recent work on magnon phase conjugation and that has two useful characteristics: the output is frequency shifted relative to the input and it tracks back along exactly the same physical path?5. How can we interface such magnonic devices with one another and with conventional electronics in a way that is both energy efficient and is capable of transferring data reliably from wave to pulse format?6. Are we capable of making magnonic devices with similar algebraic properties to quantum computing style logic elements and can these be used to improve or streamline existing magnonic logic configurations?7. What role is there for analogue magnonic processors and can they be integrated successfully with digital wave technology?Collaborators: EcotricityAlignment with EPSRC aims: Novel Computing Parad