Transition Path Sampling of Exchange Processes in Solids

The controlled manufacturing of nanostructures with tailored mechanical, electrical and optical properties is key to the development of future nanotechnologies. The aim of the project is the computer simulation of the microscopic processes leading the formation of nanostructures in PbTe/CdTe systems. These two compunds are immiscible, which means that they become separated as the system reaches thermodynamic equilibrium, a feature which is often employed in self- organisation of nanostructures. PbTe quantum wells and quantum dots embedded in CdTe are characterised by their intense emission in the mid-infrared spectrum. Our studies are motivated by experiments in which morphological transformations of PbTe/CdTe structures during crystal growth or annealing were observed. The atomistic mechanism by which those transformations occur is unknown. In the project the microscopic processes that control the growth will be analysed as well as the diffusive processes that occur in the bulk of the sample. The study will be performed by means of computational simulation techniques such as density functionary theory, molecular dynamics and transition path sampling. It is expected that the results of the project will help in understanding the morphological transformations occurring in PbTe/CdTe and other immiscible systems. That will lead to better control over design and manufacturing of nanostructures.

In the project, diffusion of cation interstitials in lead telluride (PbTe) and cadmium telluride (CdTe) was studied by means of neural network based molecular dynamics simulations. For both materials three neural network potentials were independently trained using energies and forces obtained from ab initio calculations. The molecular dynamics simulations of PbTe and CdTe supercells containing a single cation interstitials were carried out with these potentials in the temperature range 700-1200 K with the interval of 50 K. The obtained trajectories were subsequently analysed. The interstitial diffusion coefficient for each trajectory was measured by calculating the mean square displacement. Two different mechanisms of diffusion were identified, referred to as hops and exchanges. A hop is a transition of the interstitial between two neighbouring energetic minima in the lattice, while an exchange involves pushing out a lattice atom by the interstitial from its position into a new interstitial site. PbTe trajectories contained more exchanges than hops, while in CdTe it was the other way around. However, in both systems exchanges are effectively longer than hops. The results for the three independently trained neural networks were compared. The activation energies for the total diffusion and for the separate mechanisms were extracted from the temperature dependence of the diffusion coefficient and of the number of hops and exchanges. For both systems the values of the activation energies were comparable for different neural networks. Taking into account separate mechanisms of diffusion allowed to fit the analytical temperature dependence of the diffusion coefficient to the simulation data more precisely than by using only one single activation energy.