Autonomous robotic nanofabrication with reinforcement learning.

Area of Science:

• Nanotechnology and Materials Science
• Robotics and Artificial Intelligence

Background:

• Handling single molecules for construction faces challenges in controlling atomic-scale conformations and poor observability.
• Current methods for building complex supramolecular structures are limited by self-assembly and manual manipulation.

Purpose of the Study:

• To demonstrate autonomous robotic nanofabrication by manipulating single molecules.
• To overcome variability and observability issues in atomic-scale manipulation.
• To enable the construction of complex supramolecular structures beyond self-assembly capabilities.

Main Methods:

• Utilized a reinforcement learning (RL) agent for autonomous control.
• Employed a scanning probe microscope for single-molecule manipulation.
• Trained the RL agent to perform molecule removal from a supramolecular structure with sparse feedback.

Main Results:

• Achieved autonomous removal of single molecules from a supramolecular assembly.
• Demonstrated excellent performance of the RL agent in a complex nanofabrication task.
• Successfully automated a task previously requiring human intervention.

Conclusions:

• The developed RL strategy effectively addresses challenges in single-molecule handling and nanofabrication.
• Autonomous robotic nanofabrication using RL opens new avenues for constructing functional supramolecular structures.
• This approach promises enhanced speed, precision, and perseverance in molecular construction.