The project of our team "Investigation of phase transitions in carbon materials at the atomic level using modern modelling techniques" was supported by the Russian Science Foundation!»
A variety of first order phase transitions in their development pass through the same stages, the first of which is the nucleation, the most interesting and the most complicative for investigation. In the theory of this stage the questions of thermodynamics of small systems and the description of the process of overcoming the energy barrier by nucleating particles are closely intertwined. Computer simulation techniques are required to achieve a detailed understanding of nucleation. The small size of the new phase nucleus require taking into account the contributions of interface energy, surface energy, the relaxation of mechanical stresses in the energy of curvature and other features of low-dimensional materials. This demands high-precision simulation that takes all these parameters into account, which, however, is an extremely challenging task for the current toolset of computational material science. Indeed, traditional methods of the density functional theory, although they allow one to accurately enough calculate the properties of atomic systems from first principles, are nevertheless limited by the computing power available. This limits their applicability to periodic structures consisting of hundreds of atoms. At the same time the problem of describing nucleation of new phases requires the description of systems with a number of atoms up to 10^4-10^6. On the other hand, empirical potentials, which are not demanding for computational resources, allow to describe large systems containing millions of atoms. But until recently, parametrization of these potentials has been limited to their (often rather narrow) model systems, not intended for the simulation of transition states and new phases, which is a necessary condition for the study of phase transformations. However, the situation has changed dramatically recently with the developing of empirical machine learning potentials that can be trained on a large data set derived from ab initio calculations. Thus, one of the challenges of the project is to develop such potentials describing interactions with the accuracy of first principles methods to model the required number of atoms in the structures. Parametrized potentials will be applied to describe phase transformations in carbon systems, graphite-diamond transition and multilayer graphene-diamond ultrathin film (diamane) and may be further used to describe phase transitions in other systems as well.
The project PI is Dr. Sergey Erohin.