Phase transitions of the first kind, despite their diversity, pass through common stages of development, and the most important and complex of them is the nucleation stage. The study of this stage requires careful analysis, since it involves the intersection of the thermodynamics of small systems and the processes associated with the overcoming of the energy barrier by the nascent particles of a new phase. For a detailed understanding of the nucleation processes, the use of computer modelling methods is extremely important. Given the small size of the nucleating nucleus of the new phase, a number of features such as interface contributions, surface energy or mechanical stress relaxation must be taken into account. A correct description of this process requires highly accurate modelling of the nucleate structure and the surrounding material, which is a challenging task for current computational materials science methods.
First-principles methods of density functional theory (DFT) provide the possibility to calculate the properties of atomic systems with high accuracy. However, their applicability is limited by computational power, which allows modelling only systems consisting of hundreds of atoms with periodic structure. The description of phase nucleation processes requires the modelling of much larger systems including up to 10⁶ atoms, which is beyond the capabilities of traditional TFP methods.
On the other hand, empirical potentials are less demanding on computational resources and can describe systems up to a million atoms. Nevertheless, earlier they were practically not used to describe nucleation because traditional empirical potentials had significant limitations: their parameterisation was usually based on a narrow set of model systems, which made them unsuitable for modelling transition states and phase transformations. This situation has changed in recent years with the advent of potentials based on machine learning (MO potentials). These potentials can be trained on a large amount of data obtained from calculations by first-principles methods. Thus, they combine the accuracy of TFP with the ability to model systems containing a large number of atoms.
The aim of the project is to model at the atomic level the nucleation of diamond and HPAI in the HPHT process in the presence of catalysts. In particular, the first steps of nucleation of a new phase, the possibility of intermediate phases of wurtzite or tetragonal BN, which was previously inaccessible for both experiment and theory, will be determined. To achieve the goal, MO potentials will be developed to describe interactions with TFP accuracy and to perform modelling of phase transitions in carbon and BN systems.
The project is headed by Dr. P.B. Sorokin, Ph.D.-M.Sc.
First-principles methods of density functional theory (DFT) provide the possibility to calculate the properties of atomic systems with high accuracy. However, their applicability is limited by computational power, which allows modelling only systems consisting of hundreds of atoms with periodic structure. The description of phase nucleation processes requires the modelling of much larger systems including up to 10⁶ atoms, which is beyond the capabilities of traditional TFP methods.
On the other hand, empirical potentials are less demanding on computational resources and can describe systems up to a million atoms. Nevertheless, earlier they were practically not used to describe nucleation because traditional empirical potentials had significant limitations: their parameterisation was usually based on a narrow set of model systems, which made them unsuitable for modelling transition states and phase transformations. This situation has changed in recent years with the advent of potentials based on machine learning (MO potentials). These potentials can be trained on a large amount of data obtained from calculations by first-principles methods. Thus, they combine the accuracy of TFP with the ability to model systems containing a large number of atoms.
The aim of the project is to model at the atomic level the nucleation of diamond and HPAI in the HPHT process in the presence of catalysts. In particular, the first steps of nucleation of a new phase, the possibility of intermediate phases of wurtzite or tetragonal BN, which was previously inaccessible for both experiment and theory, will be determined. To achieve the goal, MO potentials will be developed to describe interactions with TFP accuracy and to perform modelling of phase transitions in carbon and BN systems.
The project is headed by Dr. P.B. Sorokin, Ph.D.-M.Sc.