The Future of AI & Frontier Materials: Materials Informatics

Since the last decade, there has been an explosion of data science applications, pushing the boundaries of science and technology helping to observe the non-observable from elemental particles to black holes, improving medical diagnosis, developing pharmaceuticals, and enhancing material properties among many more applications. In this sense, materials informatics is capable of taking frontier materials to the next level for achieving multiple sustainable development goals.

From a practical point of view, materials informatics can be applied in all the stages of materials development such as synthesis, characterization, applications, and scaling. From the synthesis side, it is possible to search for the best material property given a set of synthesis variables. At first, it is needed to define what are the lowest and maximum values of the variables given the knowledge of an expert and the literature. Then an optimization method is implemented to explore the possibilities and fine-tune towards the best synthesis conditions. This can take several iterations, however, the results can go much beyond prior achievements. It is particularly useful when the relationship between the synthesis conditions and properties is unknown and when there are multiple variables that increase the complexity for humans to track. Furthermore, materials synthesis and characterization can be automated and led by artificial intelligence (AI), capable of working 24 hours and 365 days a year, also known as a self-driving lab.

Once a frontier material is made it is necessary to characterize it to understand its properties and improve the material design for the desired application. Characterization analysis can be prone to misinterpretations, and bias, and can be noise sensitive. To overcome these issues a model can be developed to repeat characterizations with high accuracy and reproducibility. For example, microscopy requires careful observation to identify regions of interest, focus on the material and perform feature analysis. This process can be automated with the help of image classification to determine dimensions, the number of objects, object identification, real-time tracking, etc. Another example is spectroscopy, it is possible to create models that can instantly analyze spectra composition to identify the chemical nature of materials. These tools can be a great advantage when it is necessary to screen a large number of samples.?

"The convergence between AI and frontier materials will help shape the future of society."

When a potential application has been identified it is necessary to focus material development on a specific goal. In this regard, it is possible to use the synthesis variables, material properties, or both as input data to create optimization methods that enhance device and materials performance. Electrical signal processing has given further functionality for precise sensing classification. For example, the signals from an array of sensors can yield a unique fingerprint to distinguish organic solvents or even food products such as cheese and liquor.

Nowadays, multiple research groups and industries are leveraging AI to accelerate materials discovery to solve humanity's greatest challenges. Therefore, it can be expected that the time between lab-made materials to commercial production can be much shorter. The convergence between AI and frontier materials will help shape the future of society.

Author: Aaron Morelos