ITMO Scientists Propose New Digitization Technique for Nanomaterials
Researchers from ITMO University have developed an algorithm that creates digital versions of nanomaterials by automatically detecting their morphological parameters such as size, shape, and surface structure. The technology will make it possible to create advanced artificial intelligence (AI)-based predictive models in materials science, as well as design structures using an inverse design approach. As a result, this will pave the way for the development of novel materials with predetermined properties for biomedicine, optics, and biotechnology.
Although the advances of machine learning (ML) and AI are widely used in organic and medicinal chemistry, scientists still do not have enough nanomaterial-related data and informative digital representations to make their application possible in materials science.
The existing methods cannot fully represent surface morphology, that is the distribution of surfaces depending on their size, shape, and structure properties. Scientists often use rather trivial names to describe the shape of particles (e.g., spherical or cubic), which not only fail to quantify objects but also make it impossible to compare different shapes. Researchers use average values or even ranges of values to describe sizes, which are insufficient to characterize most systems—not to mention the absolute exclusion of surface structures. The lack of a systemic approach in the parametrization and simplified description of systems hamper the development of more efficient tools that could help scientists to discover novel nanomaterials with unique properties, as well as research in the field, including AI-driven projects.
The novel method proposed by ITMO researchers solves this problem. The scientists have designed the algorithm that provides a comprehensive description of nanomaterial morphology, finds similar materials based on their images, and generates synthesis parameters to develop new structures with desired properties.
The algorithm analyzes a nanomaterial image taken with an electron microscope, extracts the morphological characteristics, and then draws up lists of parameters using numerical values. Usually, such preliminary work is performed manually and scientists have to go through multiple papers to find the right parameters and synthesis methods related to such factors as the concentration of substances or synthesis temperatures.
"Our algorithm extracts morphological data removing any noises. Such compression results in a rather small set of accurate properties, though not so easy to interpret in this compressed form. We can use that not only to obtain parameters we're interested in but also to restore or create new structures following the principles of inverse design," explains Nikita Serov, a Ph.D. student at ITMO and an author of the paper. "Apart from that, the system can also find images of similar nanomaterials just like Google Lens or Yandex Image Search. You make an inquiry, add an image, and the system looks for similar materials in the database, as well as provides their production methods. The entire process takes less than a minute but the developers plan to optimize their method further, thus offering new opportunities for future research."
As noted by the author of the method, the developed algorithm recognizes not only microscope images but also hand drawings. All you need to do is to upload an image of the object under study. What's more, the technology makes it possible to mark the presence of spikes, bumps, or holes on surfaces, all of which will be considered by the algorithm.
The researchers ran their algorithm on their own database of calcium carbonate materials synthesized from scratch for testing. Calcium carbonate is one of the abundant substances on Earth so it is no wonder that many living beings create their shells out of this mineral. This is a rather simple compound when it comes to chemical synthesis as it can be easily modified and used to produce structures with highly varied and complex morphology. Having tested their method on the database, the scientists demonstrated the efficacy of their system using data from the synthesis of gold nanoparticles extracted from various scientific articles.
The control over the morphological parameters of nanoparticles is necessary to ensure their safety and shape their behavior in living beings, as well as facilitate the development of cutting-edge materials in biomedicine, therapy, and biotechnology.
"The morphology of nanomaterials greatly affects their behavior in cells and how the immune system of the organisms responds to them. For instance, by knowing the size of a nanoparticle, we can say in which human organs it will accumulate, whether it will stay in kidneys or a liver, or circulate in a bloodstream for a long time. Their shapes could have a direct impact on metabolism. The changes in surface areas of nanomaterials can increase their antibacterial or toxic properties," emphasizes Nikita Serov.
The team plans to carry on their research, test their model by synthesizing the systems not from the database following the system recommendations, as well as expand their database and turn their technology into an accessible and user-friendly product for users from all over the world. The research is conducted with the support of the ITMO'ы 2030 Development Strategy project.
Provided by ITMO University
