EAGER: Nano-Patterned Coupled Spin Torque Oscillator (STO) Arrays - a Potentially Disruptive Multipurpose Nanotechnology

由于传统的CMOS技术遇到了多重障碍，可能会减缓甚至阻止摩尔？在未来十年的某个时候，需要新的材料发现和新的纳米器件和电路范例，以允许使用传统方法不切实际甚至不可能的新应用。 这项工作的重点是自旋力矩振荡器（STO）的建模，制造和应用，重点是纳米阵列。使用多铁性材料和电场，该团队将调整自旋扭矩振荡器的频率以及相邻STO之间的耦合。该电场将改变自由层的矫顽场，这反过来将调节STO的旋进频率和STO之间的耦合。当呈现可能来自不同类型的环境和/或生物传感器的实时不同数据记录的连续流时，STO阵列可以通过分析流并提供关于流中的各种数据实例及其类成员资格的匹配程度信息来提取知识。STO阵列的关联存储器能力对于数据源是不可知的，因此它们可以用于计算和组合来自以不同操作模式工作的混合源的数据流。这种特殊形式的机器学习可以在真实的时间内完成，而不需要过多的本地数据缓冲。它还可以适应概念漂移，因为用于数据挖掘的基础统计数据可以通过控制对各个STO的输入而随时间轻松地改变。所提议的技术能够实现的实时数据流挖掘等应用对国家的安全和安保至关重要。多学科活动利用基本的纳米电子学和自旋电子学的概念，通过在材料科学，电路设计和新的纳米计算范例的贡献。PI在跨学科合作以及让本科生和研究生参与这些合作活动方面有着良好的记录，该项目将使材料科学家，物理学家以及电气和计算机工程师之间的进一步合作成为可能。

As conventional CMOS technologies are running into multiple barriers that are likely to slow down or even stop Moore?s Law scaling sometime during the next decade, there is a need for new material discoveries and new nanodevice and circuit paradigms that allow new applications that would be impractical or even impossible using traditional methods. This work focuses on modeling, fabrication and applications for spin torque oscillators (STOs), with an emphasis on nano-arrays. Using multiferroic materials and an electric field, the team will tune both the frequency of a spin torque oscillator and also the coupling between adjacent STOs. This electric field will change the coercive field of the free layer which, in turn, will adjust the precessional frequency of the STO and the coupling between STOs. When presented with a continuous stream of real-time disparate data records that could come from different types of environmental and/or biological sensors an STO array can extract knowledge by analyzing the stream and providing a degree of match information about various data instances in the stream and their class membership. The associative memory capabilities of STO arrays are agnostic to the source of data, thus they can be used for computation and composition of data streams from hybrid sources working in different modes of operation. This special form of machine learning can be done in real time with no need for excessive local buffering of data. It can also accommodate concept drift as the underlying statistics used for data mining can easily be changed over time by controlling the inputs to the individual STOs. Such applications as real-time data stream mining as enabled by the proposed technology are critical to the safety and security of the country. The multidisciplinary activities exploit fundamental nanoelectronics and spintronics concepts through contributions in materials science, circuit design and novel nano-computing paradigms. The PIs have a strong track record of collaborating across disciplines and of involving their undergraduate and graduate students in these collaborative activities, and this project will enable further collaboration between material scientists, physicists, and electrical and computer engineers.