Molecular dynamics (MD) simulations of chemical vapor deposition (CVD) of carbon dimer on a diamond (100) surface and application of neural networks (NN) for event probability predictions

Carbon dimers are found to be an important growth species in the growth of nanocrystalline diamond (NCD) through CVD process. Events, such as chemisorption, reflection, and desorption occur during the deposition of carbon dimers on to the substrate on which the diamond films are to be grown. The probabilities of each of these events have a significant effect on diamond growth. Molecular Dynamics (MD) simulations are widely used to predict the probabilities of such events. Though, MD simulations give agreeable results with experimental values, the calculation of the effect of different input parameters on various events involve time consuming numerical methods and hence the process is cumbersome. In this study, initially MD simulations of carbon dimer deposition on diamond (100) surface were performed using a many body empirical potential and the probabilities of the aforesaid events were calculated by varying the input conditions. This information was used to implement Neural Networks (NN) to predict the probabilities of the events. The neural network was also used to predict the underlying relationship between various input parameters and event probabilities. The computational time for the prediction of the events using molecular dynamics is generally several days while implementation of neural networks reduces it to mere minutes. The functional relationship between various input parameters and event probabilities predicted by NN is found to agree well with the MD simulation results.