Long and double hop kinetic Monte Carlo: Techniques to speed up atomistic modeling without losing accuracy

According to the ITRS, the development of predictive TCAD models and simulators is one of the keystone for future CMOS technologies. The kinetic Monte Carlo (kMC) technique is particularly well placed between the poles of physical ab initio but very slow simulations, and more empirical, fast simulations by partial differential equations. Nevertheless, in other to fulfill the necessities of the semiconductor industry, faster kMC simulations would be desirable.This work shows techniques to increase the simulation speed of kMC simulations an average of 2 ×× while still maintaining accuracy. These techniques modify the original kMC method by using different jump distances and performing two jumps at once during the same simulation cycle. We also show the problems and limitations of these strategies, how to detect them, and how to overcome them, if possible.Finally, a comprehensive set of simulations – including amorphization, recrystallization, extended defect ripening, diffusion and activation/deactivation of several dopants used in CMOS technologies – with and without speed up techniques are compared to experimental SIMS to elucidate how reliable these techniques are.