Liquid electrolytes and melts
Calculation of self-diffusion coefficients, viscosity, and thermal conductivity

• Molecular dynamics simulations using MTP machine-learned potentials
• Reduction of computational cost
• Achievement of quantum-mechanical accuracy

Thermal conductivity calculation

Thermal conductivity is one of the most important properties in the design of devices and systems. For thermoelectric modules, for example, a material with the lowest possible thermal conductivity is required. The same applies to thermal barrier coatings. For microelectronics, however, a material that enables efficient heat dissipation from components is preferable. The calculation of thermal conductivity from first principles is an extremely resource-intensive procedure. The use of machine learning allows for the rapid execution of these calculations without loss of accuracy.

Crystal structure prediction

Development of a software package for the first-principles evolutionary search of inorganic compounds with variable composition.

• Performance of property calculations during evolutionary search
• Construction of Pareto fronts
• Use of graph neural networks for the rapid screening of existing material databases to identify materials with desired parameters

Molecular crystals

Investigation of the most energetically stable polymorphs of molecular crystals under given thermodynamic conditions.

• Screening of conformational space using MTP machine-learned potentials
• Performance of density functional theory calculations for the most promising polymorphs
• Use of machine learning for the inclusion of temperature effects and quantum phenomena