Russian scientists and industrialists aim to change the very principle of creating new materials. Instead of lengthy experiments and trial-and-error, artificial intelligence will be taught to calculate the necessary properties of a substance in advance for specific tasks – for example, for microelectronics, sensors, or electric vehicle electronics. The joint project is being launched by Nornickel and the N. S. Kurnakov Institute of General and Inorganic Chemistry of the Russian Academy of Sciences.
The main idea of the project is to transition from empirical search for new chemical substances to digital design of materials with pre-defined characteristics. The system will be based on a large database of experimental data that IONKh RAS has been collecting for decades.
This involves tens of thousands of measurements: chemical composition of materials, their crystal structure, physicochemical and functional properties. All this data is planned to be structured and used for machine learning.
In the first stage, researchers want to collect at least a thousand unique compositions with a full set of characteristics. The project combines two key competencies:
- IONKh RAS is responsible for systematizing and standardizing scientific data, forming the basis for digital materials science;
- Nornickel provides expertise in artificial intelligence and working with palladium systems. The company already has a Palladium Technologies Center, which focuses on predicting material properties based on their crystal structure.
The main task of the platform is to create materials for specific application conditions. This involves not only selecting existing alloys but also generating new compositions with pre-defined properties.
Special attention is paid to microelectronics. Currently, the global industry uses about 250 tons of gold annually – mainly in contact pads and connecting elements. Gold is valued for its corrosion resistance and good conductivity, but it remains a very expensive material. Palladium is considered a possible alternative for a number of tasks, as it is cheaper and lighter.
This is especially important in the context of the miniaturization of modern electronics, where material properties at the nanoscale become critically important. Developers expect that the AI platform will be able to design materials directly for specific use cases – from server boards and power electronics for electric vehicles to industrial sensors for aggressive environments.
Read more on the topic:
- AI to Take on Palladium: Nornickel Intends to Create Alloys Using Neural Networks
- Nornickel Presents MetalGPT-1 Industry AI Model with 32 Billion Parameters
- Nornickel Launches "Norda" Tunneling Complex for Record Ore Production