Scientists from Germany have developed an artificial intelligence system that autonomously learns to capture and move individual molecules using a scanning tunneling microscope. This method, published in Science Advances, is relevant not only for research but also for new manufacturing technologies such as molecular 3D printing.
The rapid and economical production of prototypes or models, better known as 3D printing, has long established itself as an important tool in the industry. If this concept could be translated to the nanoscale so that individual molecules could be joined or separated again like LEGO bricks, the possibilities would be almost limitless, given that there are about 1,060 possible types of molecules, explains Dr. Christian Wagner, ERC Working Group Leader on molecular manipulation at Forschungszentrum Jülich.
While the scanning tunneling microscope is a useful tool for moving individual molecules back and forth, it always takes a special “recipe” to guide the microscope tip to position the molecules in a spatially targeted manner. This method cannot be calculated or deduced intuitively since mechanics at the nanoscale is too volatile and complex. After all, a microscope tip is not a flexible grip, but a rigid cone. The molecules simply adhere slightly to the tip of the microscope and can only be placed in the desired location using complex motion patterns.
Artificial intelligence was tasked with removing individual molecules from a closed molecular layer. To do this, a bond is first established between the microscope tip (top) and the molecule (middle). The AI then tries to remove the molecule by moving the tip without breaking contact. Initially, the movements are random. After each pass, the AI learns from experience and gets better and better.
In our case, the agent was tasked with removing individual molecules from the layer in which they are held by a complex network of chemical bonds. To be precise, these were perylene molecules, such as those used in dyes and organic light bonds that emit diodes. The particular problem here is that the force required to move them must never exceed the bond strength with which the tip of the scanning tunneling microscope attracts the molecule, otherwise the bond would break. Therefore, the tip of the microscope must follow a special pattern of movement that we previously had to literally detect manually.
Christian Wagner, Doctor and Head of the ERC Working Group on Molecular Manipulation at Forschungszentrum Jülich
While the software agent over time develops rules that determine which movement is most promising for success in a given situation and, therefore, gets better with each cycle.