Atomically Thin Oxides for Ultralow Power Non-volatile Memory Technology

新的高性能，超低功耗非易失性存储器（NVM）技术对于各种各样的和巨大增长的以数据为中心的技术至关重要，这些技术涵盖物联网，运输，医疗，安全，娱乐，神经形态计算和人工智能，所有这些都将在未来十年内显着发展，并及时彻底改变我们的生活方式。NVM对于大幅提高高能耗数据中心的效率也至关重要，在这些数据中心中，内存占总功耗的很大一部分。忆阻器可以模仿能够计算和存储数据的人工神经元，它们有可能大大减少传统计算机中的能量和时间损失。在忆阻器结构本身内，某些氧化物（inc. TiO 2和HfO 2）是优良的活性层，部分原因是在施加电压时不稳定的氧空位影响层的电阻。本项目将探索并专注于精确设计的二维（2D）氧化物材料，以开发忆阻器中的新型有源“开关”层。这提供了令人兴奋的潜力，原因如下：首先，我们可以缩小整体设计，使其更紧凑，并避免“短沟道”效应，通过限制在原子级薄的有源区的电子。其次，我们可以测试新的货车范德华束缚的活性层，其中可能有新的现象引起增强开关。最后，大范围的可用分层材料提供了探索作为多层活性层的大量潜在组合的范围。首先在材料向下选择过程的指导下，我们将使用最先进的设施制造设备并对其进行重新设计，同时考虑未来的潜在用例，特别强调神经网络/机器学习应用。

New high-performance, ultralow power non-volatile memory (NVM) technology is essential to a wide range of diverse and hugely growing data centric technologies spanning IoT, transport, medicine, security, entertainment, neuromorphic computing, and AI, all which will evolve significantly over the next decade, and in time will radically change the way we live. NVM is also essential for strongly improving the efficiency of energy-hungry data centres where memory accounts for a large fraction of the overall power usage. Memristors can mimic artificial neurons capable of both computing and storing data, and they have the potential to dramatically reduce the energy and time lost in conventional computers. Within the memristor structure itself, certain oxides (inc. TiO2 and HfO2) are excellent active layers, owing in part to labile oxygen vacancies affecting the resistance of the layer, upon an applied voltage. This project shall explore and focus on precisely engineered two-dimensional (2D) oxide materials to develop novel active 'switching' layers in memristors. This offers exciting potential for the following reasons: Firstly, we can downscale the overall design, making it more compact and avoid the 'short channel' effect, by confining electrons within the atomically-thin active region. Secondly, we can test novel van der Waals bound active layers where there could be new phenomena giving rise to enhanced switching. Lastly, the large range of available layered materials provides scope to explore a wealth of potential combinations as a multi-layered active layer. Guided first by a material down-selection process, we shall fabricate devices and characterise them using state-of-the-art facilities, all the while considering future potential use cases, with a special emphasis on neural network/machine learning applications.