Nanotechnology

Development of an autonomous AI assistant to build nanostructures

Positioning of individual molecules on the material surface is performed using scanning tunneling microscopy. The tip of the probe emits electrical impulses that deposit molecules carried by the probe. Credit: Bernhard Ramsauer—TU Graz

A material’s chemical composition alone may reveal little about its properties. Often the determining factor is the arrangement of molecules within the atomic lattice structure or on the surface of the material. Materials science uses this element to create specific properties by applying individual atoms and molecules to surfaces using high-powered microscopes. This is still very time consuming and the nanostructures constructed are relatively simple.

A research group at Graz University of Technology now wants to use artificial intelligence to take the construction of nanostructures to a new level. Their paper is published in the journal Computer Physics Communications.

“A self-learning AI that quickly and specifically positions individual molecules in the right orientation, and does all of this completely autonomously,” said Oliver Hoffmann of the Solid State Physics Institute, head of the research group. I would like to develop a system.” This should enable the construction of highly complex molecular structures containing logic circuits in the nanometer range.

Positioning using a scanning tunneling microscope

Positioning of individual molecules on the material surface is performed using scanning tunneling microscopy. The tip of the probe emits electrical impulses that deposit molecules carried by the probe.

“For simple molecules, this step takes several minutes to complete,” Hoffman says. “But building complex structures with potentially exciting effects requires placing thousands of complex molecules individually and testing the results. Of course, this requires a relatively long It takes time.”

Autonomous AI assistant builds nanostructures

A visual representation of the state-action space. a) Each state is given by the angle 𝜑 and the distance 𝑑 between the goal and the moving part. The state space encompasses all possible distances and angles. b) The working space is defined as the combination of possible parameters given by the bias voltage 𝑉 , the approach distance of the tip 𝑍 , and the relative position of the STM tip from the molecule position 𝑋 , 𝑌 . . Credit: Computer Physics Communications (2024). DOI: 10.1016/j.cpc.2024.109264

However, scanning tunneling microscopes can also be controlled by a computer. Hoffman’s team now hopes to use a variety of machine learning techniques to force such computer systems to independently place molecules in the correct position.

First, we use AI techniques to calculate an optimal plan that describes the most efficient and reliable approach to building the structure. A self-learning AI algorithm then controls the probe tip and precisely positions the molecules according to the plan.

“Positioning complex molecules with the highest precision is a difficult process because, despite the best possible control, it is always subject to some degree of chance,” Hoffman explains. . Researchers integrate this conditional probability element into AI systems to ensure that they work reliably.

Nanostructure in the shape of a gate

The researchers eventually hope to use AI-controlled scanning tunneling microscopes that can operate 24 hours a day to build so-called quantum enclosures. These are gate-shaped nanostructures that can be used to trap electrons from the material being deposited. The wave-like nature of electrons gives rise to quantum mechanical interferences that can be used in practical applications.

Until now, quantum enclosures have been constructed primarily from single atoms. Hoffman’s team now hopes to manufacture them from molecules with complex shapes. “Our hypothesis is that this will allow us to build more diverse quantum enclosures and specifically scale up their effects.”

Researchers hope to use these more complex quantum enclosures to build logic circuits and fundamentally study how they work at the molecular level. In theory, such quantum enclosures could one day be used to build computer chips.

Further information: Bernhard Ramsauer et al, MAM-STM: Software for autonomous control of single parts directed to specific surface locations, Computer Physics Communications (2024). DOI: 10.1016/j.cpc.2024.109264

Provided by Graz University of Technology

Citation: Developing an autonomous AI assistant to build nanostructures (January 16, 2025), from https://phys.org/news/2025-01-autonomous-ai-nano Structures.html 2025.1 Retrieved on March 19th

This document is subject to copyright. No part may be reproduced without written permission, except in fair dealing for personal study or research purposes. Content is provided for informational purposes only.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button