RUI: Structural and compositional modification of memristive niobium oxide films for neuromorphic computing applications

在过去的七十年里，电子电路的快速和看似无情的改进在很大程度上是由电子元件的小型化驱动的。 然而，在极小的长度尺度上的不利量子效应对这些电路的缩小提出了一个即将到来的限制，许多研究人员已经从生物系统中寻找进一步改进的灵感。 神经形态或大脑启发的计算具有提高性能和计算速度的潜力，同时通过模仿神经元的生物功能来降低功耗。 来自纽约州立大学布罗克波特和伊萨卡学院物理系的研究小组，沿着来自美国海军研究实验室的合作者，正在研究用于神经形态电路的氧化铌薄膜。 氧化铌薄膜是神经形态电路的理想候选材料，因为它丰富、廉价、无毒，并且可以模拟大脑的神经元和突触行为。 该项目的重点是薄膜的生长，在薄膜中加入其他元素（如锌和铝），生长后的热退火，以及制造成电子电路元件。 研究小组正集中精力将各种材料变化（例如，氧化物组成、厚度、生长参数）与器件的最终电子行为。 最终，该项目的目标是开发基于氧化铌的电子元件，可以与当前最先进的硅基电子产品无缝集成。 本科生是研究团队不可或缺的成员，参与这项研究往往是有吸引力的物理学代表性不足的群体的成员。 研究团队的本科生成员在夏季和学年期间参与研究项目的各个方面，并在区域和国家会议上展示他们的工作。 PI定期在当地学校与学生代表性不足的群体，包括成功的信息，挑战，并在材料科学和计算机科学的机会，点燃对科学和技术的兴趣。技术提示铌氧化物是一种多晶型材料，根据化学计量，具有许多有趣的和潜在有用的电气和光学性能。晶体二氧化铌（NbO2），特别是，显示挥发性忆阻行为，是一个领先的候选人架构，合并传统的金属氧化物半导体组件与大脑启发的神经形态电路元件，这通常需要有突触样和神经元样的组件。该项目的重点是更好地了解NbO2，它在800摄氏度左右经历了从高电阻到低电阻的挥发性相变。这种转变通过一旦达到温度阈值就突然改变电阻来模仿神经元的尖峰电行为。该研究团队由两名材料开发和电力运输专业的主要研究人员、他们的本科研究生以及美国海军研究实验室的合作者组成，正在研究材料沉积和沉积后处理过程，以及所得薄膜的光学和电学行为。在材料方面，研究团队使用原子层沉积（ALD）、掺杂和生长后结晶技术，以与现有半导体制造工艺完全兼容的方式制造高质量的NbO2。具体来说，该项目研究了在ALD过程中添加掺杂剂，以鼓励结晶，并具有与低功率器件操作兼容的热预算。 在结晶之后，超高真空系统用于使材料热循环通过其相变，同时观察反射和透射的光信号以快速建立生长条件或掺杂剂对相变温度的影响。在器件方面，进行电测量以确定材料制备和性质对关键器件操作参数的影响，例如在失效之前可以进行的转变次数、器件成品率，该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查进行评估，被认为值得支持的搜索.

Non-technical abstractThe rapid and seemingly relentless improvement in electronic circuitry over the last seven decades has been driven in large part by miniaturization of the electronic components. However, adverse quantum effects at extremely small length scales present an impending limit to shrinking of these circuits, and many researchers have looked to biological systems for inspiration for further improvement. Neuromorphic, or brain-inspired, computing has the potential to enhance performance and computational speed while reducing power consumption by mimicking the biological function of neurons. The research team from the undergraduate-only physics departments at SUNY Brockport and Ithaca College, along with collaborators from the U.S. Naval Research Laboratory, are studying thin films of niobium oxide for use in neuromorphic circuits. Thin-film niobium oxide is an ideal candidate for neuromorphic circuits, as it is plentiful, inexpensive, non-toxic, and can mimic both the brain’s neuronal and synaptic behaviors. This project focuses on the growth of the thin films, incorporation of other elements (such as zinc and aluminum) in the films, post-growth thermal annealing, and fabrication into electronic circuit components. The research team is focusing their effort on correlating the various material changes (e.g., oxide composition, thickness, growth parameters) with the device’s resulting electronic behavior. Ultimately, the project’s goal is to develop niobium oxide based electronic components that can seamlessly integrate with the current state-of-the-art silicon-based electronics. Undergraduate students are integral members of the research team, and participation in this research is often attractive to members of groups underrepresented in physics. Undergraduate student members of the research team participate in all aspects of the research project during both the summer and during the academic year, and present their work at regional and national conferences. The PIs regularly present at local schools in areas with students from underrepresented groups and include information on successes, challenges, and opportunities in materials science and computer science to ignite interest in science and technology.Technical descriptionNiobium oxide is a polymorphic material that, depending on stoichiometry, has a number of interesting and potentially useful electrical and optical properties. Crystalline niobium dioxide (NbO2), in particular, displays volatile memristive behavior, and is a leading candidate for architectures that merge traditional metal-oxide-semiconductor components with brain-inspired neuromorphic circuit elements, which are generally required to have both synapse-like and neuron-like components. This project focuses on developing a better understanding of NbO2, which undergoes a volatile phase transition from high to low resistance around 800 degrees Celsius. This transition mimics the spiking electrical behavior of neurons by abruptly changing resistance once a temperature threshold is achieved. The research team – which consists of two principal investigators, with specialties of materials development and electric transport, their undergraduate research students, and collaborators from the U.S. Naval Research Laboratory – is studying both the material deposition and post-deposition treatment processes, as well as the optical and electrical behavior of the resulting films. On the materials side, the research team uses atomic layer deposition (ALD), doping, and post-growth crystallization techniques to fabricate high-quality NbO2 in a way that is fully compatible with existing semiconductor manufacturing processes. Specifically, the project examines the addition of dopants during ALD to encourage crystallization with a thermal budget compatible with low-power device operation. Following crystallization, an ultra-high vacuum system is used to thermally cycle the material through its phase transition while observing reflected and transmitted optical signals to quickly establish the effect of growth conditions or dopants on the phase transition temperature. On the device side, electrical measurements are performed to establish the effect that material preparation and properties have on key device operation parameters, such as the number of transitions that can be performed before failure, the device yield, and the switching power requirements.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.