Stoicheia, Inc. Trains AI to Predict Unknown Complex Novel Materials

(US and Canada) Northwestern University and the Toyota Research Institute have successfully used technology from Stoicheia, Inc. to accurately predict novel — and previously unknown — materials for the clean energy, chemical and automotive industries.

In a groundbreaking new study, the researchers combined Stoicheia's Megalibrary technology with artificial intelligence (AI) to come up with 19 new materials that have selected characteristics. Eighteen proved correct, representing 95% accuracy.

Stoicheia's Megalibrary technology creates over 200 million positionally encoded and different by design nanomaterials on a two-by-two-centimeter chip, becoming the foundation for unparalleled datasets of a size and quality not previously achieved in materials science.

"This outcome is profound," said Chad A. Mirkin, chairman and co-founder of Stoicheia and world-renowned nanotechnology pioneer. "The predictions were outside the bounds of human scientific postulation, proving the power of Stoicheia's Megalibrary technology and machine learning and illuminating Stoicheia's path to breakthrough future discoveries critical to the clean energy, automotive and chemical industries."

The research is detailed in the Dec. 23, 2021, issue of the journal Science Advances.

Mirkin, who is the George B. Rathmann Professor of Chemistry at Northwestern University, is one of the study's senior authors.

To conduct the study, researchers used Stoicheia's Megalibrary technology to generate structural data for nanoparticles with complex compositions, structures, sizes and morphologies. These data were then used to train machine learning (ML) algorithms.

With limited physics and chemistry first-principles, the ML models based upon the pure training data were able to predict complex structures of materials never before synthesized on earth.

"This study illustrates the ability of Stoicheia to create novel complex materials faster than anybody on earth, generate vast amounts of first-party, high-quality materials data and train algorithms that will rapidly accelerate our understanding of the materials genome," said Dr. Andrey Ivankin, Stoicheia's co-founder and chief scientific officer. "As our models accelerate understanding, the gap between traditional discovery methodologies and Stoicheia's platform will compound. Stoicheia's virtuous cycle will finally harness the power of artificial intelligence in materials science to drive incredibly valuable solutions to the climate crisis."

The design space for new nanomaterials, which are often compared to the human genome, is inherently more complex. Rather than dealing with four-letter alphabet, the design space comprises 118 elements on the periodic table. And, when reduced to the nanoscale, changes in shape, size, phase morphology, crystal structure and more, compound the enormity of discerning structure and function of new nanomaterials. Interrogating this design space through traditional serial experimentation is a prohibitively expensive, slow and inefficient process that leads to local winners, versus the global, or best, winner for the problem at hand.

Stoicheia's technology is fueling processes in clean energy, automotive and chemical industries. Identifying new green catalysts will enable the conversion of waste products and plentiful feedstocks to useful matter, green hydrogen generation, carbon dioxide utilization and the development of fuel cells among other applications. Novel catalysts also could be used to replace expensive and rare materials like iridium, the metal used to generate green hydrogen and CO2 reduction products. As rare earth metals become more important globally, Stoicheia's technology is critical to a green and geopolitically secure future.