Scientists Unveil AI System for Graphene Transistor Assembly

Researchers unveil Qumus, an AI system for quantum material experiments.

Researchers have introduced Qumus, an autonomous AI system designed for experiments with quantum materials. The work is published on arXiv and is yet to be peer-reviewed.

The platform integrates robotics, computer vision, and a multi-agent architecture.

Qumus operates in a robotic mini-laboratory, executing the entire cycle from protocol planning to result analysis and report preparation.

Without human intervention, the system isolated graphene flakes using mechanical exfoliation and fabricated nanoscale devices from atomically thin materials, including a graphene field-effect transistor based on a van der Waals structure.

To find a method for obtaining a graphene flake larger than 200 μm², the AI examined four parameters: temperature, contact time, number of press cycles, and tape detachment speed. Over more than four hours and five optimization cycles, the platform produced a fragment measuring 245 μm².

Qumus is capable of correcting errors. In one test, the system detected the absence of a silicon chip, reassembled the plan, and repeated exfoliation on a new substrate. In another, it corrected the misclassification of hexagonal boron nitride as graphene.

During transistor assembly, the robot conducted a 90-minute dry transfer involving 30 physical operations and 18 decision-making stages.

A hierarchical network of agents based on large language models is used to manage the laboratory. Computer vision tracks objects and inventory, employing YOLOv8 for detection.

In May, CEO of Keeper Security Darren Guccione stated that AI and quantum technologies will threaten existing security systems.