Research Interests

Topological Materials

Topological materials are systems that hold nontrivial topological properties, usually characterized by the "Chern number". The most commonly seen topological material is topological insulator -- an insulator in the bulk and a conductor at the edge.

We study mainly the three-dimensional topological insulators, which has a conducting surface. Such material including Bi2Se3, Bi2Te3, etc. Further, we study the layered structure of such material with magnetic materials such as MnBi2Te4. That interplay of these two types, although breaks time-reversal symmetry, can give rise to Chern insulator that exhibits anomalous quantum Hall effect.

不同種拓樸材料，被不同種類對稱性保護。而各自具有獨特的拓樸特性。最常見的拓樸材料是拓樸絕緣體。這種絕緣體在塊材中是絕緣體，卻在表面或邊緣是導體。我們研究的是一種三維的拓樸絕緣體化合物Bi2Te3，他在表面是類似石墨烯能帶的狄拉克錐 (Dirac cone)。我們更進一步研究將Bi2Te3 與 MnBi2Te4 做成層狀結構。雖然會破壞時間反轉對稱，但可能成為另一種拓樸材料: 陳絕緣體。這種絕緣體也具有拓樸特性，並可以在沒有外加磁場的情況下展現出量子霍爾效應。

這部分研究與德州大學奧斯丁分校的Alllan MacDonald 教授(沃爾夫獎得主)，施志剛教授(實驗)，以及清華大學的徐斌睿教授(實驗)合作。

Deep Learning in Physics

We apply different artificial neural networks, mainly the convolutional neural network (CNN) to identify or even to predict the property of materials. We train the network to recognize crystal structures with subtle differences so that it can quickly identify the different grains in a sample. We also teach the network to learn topology so that by given Hamiltonian it can give out the system's Berry curvature and Chern number.

我們主要使用圖形辨識原理的卷積神經網路來辨識，甚至是預測材料性質。我們可以訓練網路來快速辨識同一個養品中不同的晶體結構。另外，我們也可以教導神經網路來學習預測拓樸性質。之後輸入哈密頓函數給神經網路，他就可以判斷牠的貝里曲率與陳數。

這部分研究與周苡嘉教授(實驗)合作。