The mere chemical makeup of a material doesn’t always provide clear insights into its characteristics. Often, it’s the way molecules are arranged within the atomic structure or on the surface that matters most. Materials science leverages this by positioning individual atoms and molecules on surfaces using advanced microscopes. Now, an innovative research group is aiming to enhance the assembly of nanostructures through artificial intelligence.
The mere chemical makeup of a material doesn’t always give us a clear understanding of its characteristics. Often, the key aspect is how the molecules are arranged in the atomic lattice or on the material’s surface. Materials science takes advantage of this by strategically placing individual atoms and molecules on surfaces with high-performance microscopes. This process is currently very labor-intensive, and the resultant nanostructures tend to be relatively simple.
With the help of artificial intelligence, a new research team at TU Graz aims to revolutionize the creation of nanostructures: “Our goal is to create a self-learning AI system that can quickly and accurately position individual molecules autonomously,” states Oliver Hofmann from the Institute of Solid State Physics, who leads the group. This technology could enable the construction of intricate molecular structures, including nanoscale logic circuits. The research group, known as “Molecule arrangement through artificial intelligence,” has secured €1.19 million in funding from the Austrian Science Fund.
Positioning with a scanning tunneling microscope
The placement of individual molecules on a surface is performed using a scanning tunneling microscope. The probe’s tip sends an electrical impulse to apply a molecule it carries onto the surface. “For a simple molecule, it takes a person several minutes to complete this step,” explains Oliver Hofmann. “However, creating more elaborate structures that hold intriguing properties requires the precise positioning of thousands of complex molecules, and testing the results can take a considerable amount of time.”
Fortunately, a scanning tunneling microscope can be operated by a computer. Hofmann’s team plans to employ various machine learning techniques to allow the computer system to autonomously position the molecules accurately. Initially, AI is used to formulate the best strategy for constructing the desired structure. Then, self-learning AI algorithms direct the probe tip to place molecules exactly as planned. “Achieving high precision in positioning complex molecules is challenging, as their alignment always involves some degree of randomness, despite optimal control efforts,” says Hofmann. The researchers will incorporate this element of uncertainty into the AI system to ensure consistent performance.
Nanostructures resembling gates
By using an AI-controlled scanning tunneling microscope that operates continuously, the team aims to create so-called quantum corrals. These are gate-shaped nanostructures capable of trapping electrons from the underlying material. The wave-like behavior of these electrons can result in quantum-mechanical interferences with useful applications. Previously, quantum corrals have primarily been constructed from single atoms. Hofmann’s team intends to develop them using more complex molecules: “We believe this will enable us to create a wider variety of quantum corrals and enhance their functional properties.” The team plans to leverage these sophisticated quantum corrals to build logic circuits, facilitating a fundamental exploration of their molecular-level functionality. In theory, such quantum corrals could eventually be employed in chip manufacture.
Collaborative expertise from two universities
Over its five-year project, the research group is combining knowledge from artificial intelligence, mathematics, physics, and chemistry. Bettina Könighofer from the Institute of Information Security leads the development of the machine learning model. Her team needs to ensure that the self-learning system doesn’t accidentally damage the nanostructures it creates. Jussi Behrndt from the Institute of Applied Mathematics will establish the core theoretical properties of the structures in development, while Markus Aichhorn from the Institute of Theoretical Physics will convert these theoretical findings into real-world applications. Leonhard Grill from the Institute of Chemistry at the University of Graz will conduct the actual experiments using the scanning tunneling microscope.
