郭硕鸿讲座（六）（6月21日 周四14:30-17:30）

欢迎广大师生前来参与！

主持人：姚道新 副院长、张宏浩 教授

【报告摘要】Topological physics has been studied for more than three decades, can we still find something new ? In this talk I will present three answers from our recent work. 1) Non-Equlibirum Dynamics:  Previous studies of topological effect are mostly focused on equilibrium or near equilibrium situation, we will show that the topological invariant can also manifest its physical effect in a quench dynamics far from equilibrium.  2) Interaction Effect: We utilize the recently proposed Sachdev-Ye-Kitaev model and construct an exactly solvable model to address the interaction effect in a topological band insulator. An interaction induced topological transition and its critical behaviors can be shown explicitly by this model. 3) Machine Learning: We show that we can train a neural network to accurately predict topological invariant from local input and without human knowledge as a prior. We also analyze the neural network to show that what is captured by the neural network is precisely the same mathematical formula for topological invariant. 

Reference: 1) Ce Wang, Pengfei Zhang, Xin Chen, Jinlong Yu and Hui Zhai, Phys. Rev. Lett. 118, 185701 (2017). 2) Pengfei Zhang, Huitao Shen and Hui Zhai, Phys. Rev. Lett. 120, 066401 (2018). 3) Pengfei Zhang and Hui Zhai, Phys. Rev. B 97, 201112(R) (2018).

【个人简介】翟荟，清华大学教授，教育部国家级人才项目学者和国家杰青获得者。1998年进入清华大学物理系首届基础科学班，2002年本科毕业，2005年1月在清华大学高等研究中心获物理学博士学位，导师杨振宁。2005-2009年期间先后在美国俄亥俄州立大学、加州大学伯克利分校做博士后。2009年起任清华大学高等研究院研究员，2012年获得长聘教授，2015年任高等研究院教授。他的主要研究方向包括冷原子和凝聚态等量子物质的理论研究，机器学习方法在物理学中的应用等。