Doctoral Thesis: Artificial Intelligence-Aided Synthesis and Characterization of 2D Materials

Tuesday, May 2
7:45 am - 9:00 am


Ang-Yu Lu


Semiconductor chips serve as the fundamental building blocks of modern electronics and form the core of artificial intelligence systems. However, as the technology node approaches 3 nm, silicon transistor scaling encounters its physical limits, leading to short-channel effects and performance degradation. To address this challenge, two-dimensional (2D) materials have emerged as promising candidates for next-generation transistors, to maintain the pace of Moore’s Law—doubling the number of transistors every 18 months. The integration of AI and automation in material science has recently drawn significant attention, offering the potential to expedite and enhance material development processes. This thesis aims to develop an autonomous platform to accelerate 2D material synthesis with four distinct projects. First, we employ named entity recognition (NER) and extractive question-answering (EQA) models to extract experimental recipes, including categorical and numerical data, illustrating how to trace the trajectories within a single material and between two different materials. Additionally, we use generative language models to summarize and generate synthesis recipes for knowledge connections and transfers in 2D materials. Second, we explore the correlation between growth parameters and provided the growth windows for high-quality hBN by the Gaussian process. Third, we demonstrate cost-effective automated synthesis and characterization systems for CVD-grown graphene by upgrading existing equipment and adopting open-source software and hardware solutions. Moreover, we propose an integrated autonomous platform that combines robotics, multiphysics simulations, machine learning, and automated synthesis and characterization systems for 2D material synthesis. Finally, we systematically investigate the connections between PL signatures and Raman modes employing statistical analysis, convolutional neural networks, interpretable models, and support vector machines, delivering comprehensive insights into the physical mechanisms linking PL and Raman features. This thesis may serve as a potential framework for developing and discovering novel materials for next-generation electronics.

Thesis Supervisor: Professor Jing Kong


  • Date: Tuesday, May 2
  • Time: 7:45 am - 9:00 am
  • Category:
  • Location: 34-401B
Additional Location Details:

Hybrid defense, please contact for the link