Since the Nobel prize winning discovery of single layer graphene, the field of 2D materials has advanced significantly. Condensed matter physicists have used various 2D materials like graphene to create platforms to study strong correlated quantum phenomenon and the micro/nanoscale. Using 2D material flakes as fundamental building blocks, multilayer heterostructure can be assembled by stacking multiple flakes at controlled offset and orientation just like piecing together nanoscale LEGO blocks. Large-scale moiré patterns appear during stacking like interference patterns to form superlattice structure. A new field of twistronics arises from such studies where 2D material systems demonstrate significant property changes by stacking multiple layers at different angles. For example, stacking twisted bilayer graphene at the “magical angle” of 1.1 degree gives superconducting property at cryogenic temperature. With their superior properties, researchers are also developing electronic devices based on 2D materials for practical applications.
The development of 2D material heterostructure needs automation for improved scalability. To form a 2D material heterostructure, the typical process involves exfoliation, target flake selection, stamp-based transfer, stacking at various angles, various characterization/cleaning steps and lithography for defining electrodes as needed. The process is currently labor intensive with manual handling of transition between multiple instruments for different stages of the task. The overall goal of the project is to automate the process to improve efficiency, quality, positioning accuracy and assembly repeatability,
This Eureka post focuses on the initial stage of flake exfoliation and identification using data-driven machine vision algorithms. While 2D materials can be grown at large scale, exfoliated flakes are still needed for high-quality structures. The target flake needs to be identified using optical microscopy techniques. To automate this process, an automated exfoliations system, an automated optical microscope for imaging stitched large-area high-resolution images and a trained flake selection model are needed. Using manually exfoliated flakes or the automated system, a dataset of 2D materials will be created for training and validation of the model. This material detection system will be integrated into a fully automated system for 2D heterostructure production, manipulation and characterization.
Requirement: Junior or senior major in Computer Science or Electrical and Computer Engineering (highly motivated Sophomore students with relevant experiences also possible)
Undergraduate Research Assistants working on this project are expected to be interested in programming, machine learning, mechatronic systems, etc. Experience implementing neural networks in MATLAB or python is a must. In person work is required for access to the hardware and 2D material.
Students are expected to work closely with graduate students leading the project for 6 - 10 hours per week for pay or for credit with opportunities to continue in the future. Given the longer-term nature of the project, preference will be given to students who can commit to multiple terms.
Specific duties may include but not limited to the following:
- 2D material sample preparation
- Assistance to the development of automated exfoliation/imaging system
- Training data collection and curation
- Implementation and testing of a neural network using ResNet architecture
- Documentation and additional tasks assigned by the supervisor
