时间：2019年4月1日（星期一），下午13:30
地点：浙江大学玉泉校区硅材料国家重点实验室1号楼会议室
报告人：夏钶教授
邀请人：杨德仁院士
报告人简介：Ke Xia received his PhD in Physics at Nanjing University in 1997. He then joined the University of Twente and the Technical University of Delft as a postdoctoral researcher focusing on the ab initio calculation of magnetoelectronics. In 2002, he joined the Institute of Physics, Chinese Academy of Sciences under the 100 Talents Program granted by the CAS. He received the NSFC for Distinguished Young Scholars granted by the NSFC in 2008. He became a Professor of physics at Beijing Normal University. Since 2018, he has been a Professor of Physics at SUSTech at Sehzhen.
报告摘要：In the past decade, significant progress has been made in artificial Intelligence, where advanced algorithms using artificial neural networks (ANNs) have been successfully applied in many areas. There have been many attempts to design and fabricate neuromorphic hardware devices, which are not limited by the von Neumann bottleneck and intrinsically possess all the aforementioned advantages. In this talk, we will report a theoretical realization of RNNs with magnetic tunnel junctions, which are used as the basic units of spin-transfer-torque magnetic random access memory. By performing a micromagnetic simulation, we demonstrate that an RNN consisting of as few as 40 MTJs can generate and recognize sequential signals after an efficient training process. The capability of the network can be significantly improved by increasing the number of MTJs.
Magnon-polaritons are hybrid bosonic quasiparticles consisting of strongly interacting magnons and photons in microwave cavities. In contrast to Hermitian systems, we proposed a system where level attraction can be observed by introducing an additional phase-controlled field to drive the magnetization. Thermal induced generation of monochromatic microwave radiation in magnon-polariton is also proposed.
