Conference Logo ProgramMaster Logo
Conference Tools for MS&T25: Materials Science & Technology
Login
Register as a New User
Help
Submit An Abstract
Propose A Symposium
Presenter/Author Tools
Organizer/Editor Tools

About this Abstract

Meeting MS&T25: Materials Science & Technology
Symposium Enhancing the Accessibility of Machine Learning-Enabled Experiments
Presentation Title ATOMIC: Autonomous Characterization of 2D Materials Through Foundation Models
Author(s) Haozhe "Harry" Wang
On-Site Speaker (Planned) Haozhe "Harry" Wang
Abstract Scope Nanomaterial characterization traditionally requires extensively trained experts and remains challenging even for specialists examining novel 2D structures. We present ATOMIC (Autonomous Technology for Optical Microscopy & Intelligent Characterization), a zero-shot framework integrating foundation models for fully autonomous characterization of 2D materials. Our system combines the Segment Anything Model, large language models, unsupervised clustering, and topological analysis to automate microscope control, sample scanning, image segmentation, and intelligent analysis through prompt engineering—without additional training. When analyzing MoS2 samples, our approach achieves 99.7% segmentation accuracy for single-layer identification, equivalent to human experts, while detecting grain boundary slits invisible to the human eye. The system maintains robust performance despite variable conditions including defocus, color temperature, and exposure variations. ATOMIC works across common 2D materials (graphene, MoS2, WSe2, SnSe) regardless of fabrication method, establishing a scalable, data-efficient paradigm that transforms nanoscale materials research.

OTHER PAPERS PLANNED FOR THIS SYMPOSIUM

Accelerating Scientific Discovery with Machine Learning: Data Analysis for Computational Beamlines
Adaptive Workflows for Lab of the Future
ATOMIC: Autonomous Characterization of 2D Materials Through Foundation Models
Autonomous Atomic Force Microscopy using Large Language Model Agents
DiffractGPT: Atomic Structure Determination from X-ray Diffraction Patterns Using a Generative Pretrained Transformer
Foundational Workflows for Processing Legacy Data and Realizing Domain-Specific Multi-Modal AI Models
From automated to autonomous – creating a general active learning service for self-driving laboratories
High-throughput, Ultra-fast Laser Sintering of Ceramics and Machine-learning Based Prediction on Processing-Microstructure-Property Relationships
Hypothesis Formation and Predictive Modeling of 2D Perovskite Spacer Cations Using Retrieval Augmented LLMs and Deep Kernel Learning
Ptychography Data Pipelines at the Advanced Photon Source
Pycroscopy, AEcroscopy, and data workflows: integrating customized control, data analysis and workflows in an autonomous microscopy facility

Questions about ProgramMaster? Contact programming@programmaster.org | TMS Privacy Policy | Accessibility Statement