SWIFT: Electric Field Controlled Integrated Multiferroic Radio Frequency Devices for Interference Immune Broadband Wireless Systems

无线网络是现代社会的支柱，通过第5代网络(5G)和物联网(IoT)实现通信、同步和组织。网络参与者数量的增长以及机器对机器通信量的增加加剧了无线电频谱的稀缺，使其越来越有价值：有效利用无线电频谱至关重要。智能手机、收音机、电视、路由器、车辆和计算机等设备正在接近现有频谱共享方法的极限。智能手机开发商已经需要100多个独特的过滤器来满足不同通信方案的需求，如蓝牙、LTE、AM和FM无线电、5G和Zigbee。该项目将研究使用新的自偏置、多铁性材料以及新的算法，以了解其他网络参与者如何使用时间和频谱，并合作适应，以在有限的频谱内增加可能的通信。新的多铁性材料、设备和电路将实现非常快速的自适应过滤器，这反过来将使监控、预测、适应和与其他网络参与者共存。该项目没有要求所有网络参与者遵守特定的频谱共享协议，而是提出了一个分散的框架，在该框架中，“智能”网络参与者学习其他潜在的遗留或被动参与者的预测模型，并使用这些模型来最大化频谱共享效率。最终结果是更多的网络参与者使用相同的无线频谱进行更多的通信，具有更强的抗干扰性，以及更紧凑的无线设备，如智能手机和无线传感器。该项目使用新的材料和设备技术，在接收器内实现灵活、快速可调的过滤器，进而实现隐式无线环境监控和预测。收集的信息将用于干扰预防技术和传输调度预测算法，从而能够调度未来的传输，以最大限度地提高频谱共享效率。传统的基于铁磁的射频组件需要直流或可变的外部磁场偏置来操作和调谐，这使得它们耗电量大、体积大，而且集成起来不切实际。该项目利用自偏置多铁异质结构，通过分层薄的、高磁致伸缩的金属铁磁层和铁电材料层之间的界面交换耦合来增强磁绝缘体的自偏置，从而实现总电场控制，而不需要任何固定或可变的磁偏置。铁磁谐振器的总电场控制实现了可调、极低延迟的集成紧凑型滤波器，不需要直流偏磁。算法将与这些高度可重新配置的过滤器一起开发，以实现不受各种干扰影响的接收器，并能够更好地描述其无线环境，这些信息将用于学习传输模式的预测模型，从而在网络参与者之间实现高效的频谱共享，而不需要他们遵守特定的共享协议。该项目将评估传输时间序列预测的传统方法，如长期短期记忆，但也将考虑更好地支持快速搜索有效传输时间表的替代方法，包括因果学习和分布式学习。这些预测将被用于通知传输调度和信道使用的优化，以避免干扰并提高吞吐量。该项目将在多铁材料、模拟射频电路设计、网络频谱共享、分散网络协议和用于预测建模的机器学习领域做出科学和工程贡献。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Wireless networks are a mainstay of modern society, enabling communication, synchronization, and organization though 5th generation networks (5G) and the Internet-of-Things (IoT). Growth in the number of network participants and therefore the amount of machine-to-machine communication has exacerbated radio spectrum scarcity, making it increasingly valuable: its efficient utilization is vital. Devices like smartphones, radios, televisions, routers, vehicles, and computers are approaching the limits of existing approaches to spectrum sharing. Smartphone developers already require upwards of 100 unique filters to serve the needs of different communication schemes such as Bluetooth, LTE, AM and FM radio, 5G, and Zigbee. This project will investigate the use of novel self-biased, multiferroic materials in conjunction with novel algorithms to learn how other network participants are using time and spectrum and cooperatively adapt to increase the communication possible within a limited spectrum. The new multiferroic materials, devices, and circuits will enable very rapidly adaptable filters, which in turn will enable monitoring, predicting, adapting to, and coexisting with other network participants. Instead of requiring that all network participants conform to a particular spectrum sharing protocol, the project advances a decentralized framework within which "smart" network participants learn predictive models of other, potentially legacy or passive, participants and use these models to maximize spectrum sharing efficiency. The end result is more communication among more network participants using the same amount of wireless spectrum, with greater resistance to jamming and more compact wireless devices such as smartphones and wireless sensors.The project uses novel material and device technologies to enable flexible, rapidly tunable filters within receivers that in turn enable implicit wireless environment monitoring and prediction. The gathered information will be used by jamming prevention techniques and transmission schedule prediction algorithms that enable scheduling of future transmissions to maximize spectrum sharing efficiency. Conventional ferromagnetic based RF components require a DC or variable external magnetic field bias for operation and tuning, making them power hungry, bulky, and impractical to integrate. This project leverages self-biased multiferroic hetero-structures enabled by enhancing the self-biasing of a magnetic insulator through interface exchange coupling between a stratified thin, highly magnetostrictive, metallic ferromagnetic layer and a layer of ferroelectric material to achieve total electric field control without needing any fixed or variable magnetic bias. The total electric field control of ferromagnetic resonators enables tunable, very low latency, integrated compact filters that require no DC magnetic bias. Algorithms will be developed alongside these highly reconfigurable filters to realize receivers that are impervious to various kinds of interference and more capable of characterizing their wireless environments, and this information will be used to learn predictive models of transmission patterns, enabling efficient spectrum sharing among network participants, without requiring them to adhere to a particular sharing protocol. The project will evaluate conventional approaches to transmission timeseries prediction such as long shortterm memories, but will also consider alternatives better supporting rapid search for efficient transmission schedules, including causal and distributed learning. These predictions will be used to inform the optimization of transmission schedules and channel use to avoid interference and improve goodput. The project will make scientific and engineering contributions in the areas of multiferroic materials, analog RF circuit design, network spectrum sharing, decentralized network protocols, and machine learning for predictive modeling.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.