TELF AG analyzes the impact of AI in the discovery of new mineral compounds
New resources for the industry
The needs of the global energy transition seem to require an ever-increasing amount of resources and economic efforts to ensure a smooth conversion to renewable energy, often encountering practical and operational difficulties that could slow down the process unforeseeably. One of these has to do with raw materials deemed “critical” for the success of the transition, such as all those already used to power clean energy technologies or electric vehicles. The shortages of these resources and the uncertainties and possible interruptions that could characterize the supply chain have made it necessary to search for new deposits and new energy sources capable of supporting the ongoing change.
In this regard, only raw materials or resources already existing underground or all those materials that have yet to be identified and discovered in different points of the planet have been taken into consideration. No one seems to have ever considered the hypothesis that such materials could be assembled thanks to technology, with the support of artificial intelligence. The Google DeepMind laboratory is working on a project of this kind, which is said to have discovered more than two million new mineral compounds through an assembly process made possible by artificial intelligence. According to Google, the time equivalent of this discovery amounts to approximately 800 years of knowledge on the subject.
The potential of deep learning tools
From this point of view, one of the most interesting aspects is that the compounds created in this way represent entirely new materials with an infinite number of possible industrial applications. Among those already assembled, the compounds capable of reaching a certain degree of stability would be less than half, around 400,000. Still, some of these could contain the solution to today’s critical issues in the industry. The idea behind GNoME – as the deep learning tool capable of predicting the stability of new materials was called – was to combine the study of materials with the most advanced artificial intelligence techniques, putting together atoms and then testing the overall stability of the compound, thus obtained.
This technology is still in its infancy, but the premises undoubtedly appear promising. One of the main questions concerns the stability of atomic structures, which still needs to be verified. However, the aim seems to be represented by creating materials endowed with specific properties, which can prove helpful in the most modern industrial applications. Thanks to the support of very advanced supercomputers, technology now makes it possible to simulate the properties of materials, starting from the fundamental laws of physics, obtaining valuable information on their stability, conductive capabilities, biocompatibility, and many others.
The GNoME system has already identified 380,000 stable crystals that could be used in various industrial applications, such as superconductors or batteries for electric cars. These materials, the most stable among those discovered by GNoME, would already represent excellent candidates for experimental synthesis, thus demonstrating the full potential of artificial intelligence in finding and developing new resources. The results of this discovery were recently published in the journal Nature, with two different articles. In the second, the role of predictions triggered by artificial intelligence in the autonomous synthesis of materials is explained.
