Predictions at the Nanoscale

Jan 3, 2022   •   Luis M. Antunes

The fields of Machine Learning (ML) and Artificial Intelligence (AI) continue to advance at a rapid pace. As their capabilities grow, their reach grows as well. Historically, the problems addressed by AI were generally academic in nature, and involved, for example, solving a puzzle, or navigating through a maze. Early examples of practical applications of ML include the recognition of hand-written characters. But as of today, in 2022, it is difficult to find an activity or problem of human interest that has not been touched by these technologies.

One area very worthy of attention is materials science and chemistry. The proliferation of human civilization is intimately tied to the development of technology, and technology is enabled by our mastery of materials. The world currently faces environmental challenges, and civilization demands technological solutions. At the core of any such solution will surely be a material. Our ability to persist in the universe will increasingly depend on our ability to be masters of matter.

What gives a material its properties? There are perhaps less than 80 types of atoms that we generally encounter, and far fewer are encountered regularly. We could take a collection of Carbon, Nitrogen, Oxygen, Hydrogen, and Phosphorus atoms and arrange them into an inert solid slab, or we could arrange them into DNA. Clearly, a material's properties are determined by the arrangement of the atoms of which it is comprised. However, it isn't currently obvious how to assemble a collection of atoms into any shape one desires. Thus, the development of fabrication techniques at the nanoscale is required. Today, manufacturing is largely conducted top-down. One starts with a large aggregation of matter, and chips away until the desired shape is obtained. From an atomic perspective, such an approach is coarse and crude. With the exception of certain classes of substances, such as small organic molecules, our materials are generally not built to atomic specifications.

Mastering nanoscale manufacturing will be key in enabling the development of future technologies. We believe that ML and AI techniques will aid in achieving this goal. But even after we have advanced our ability to build with atomic precision, we are still faced with the question of what to build. What arrangement of atoms is required for a given task? There are countless ways that different atoms can be arranged in 3D space, and, as we stated above, it is the arrangement of atoms that determines a material's properties. Can ML and AI help us to choose which arrangements are best? We believe so.

AI and ML technologies can enhance our ability to work at the nanoscale. The application of these technologies to the nano world is only in its infancy, but it is certainly an exciting and worthy endeavor.