Intrinsic residual stresses during thin-film growth typically result from the complex interplay between the atomic deposition processes (e.g., adsorption, diffusion, and desorption). Therefore, unravelling the underlying atomic mechanisms is key to optimize the deposition process, currently done by trial-and-error, and to improve the efficiency, performance, and reliability of MEMS/NEMS devices. To this end, an atomistic simulation method is required that can simulate thin film growth up to a thickness of at least 100 nm while mimicking the experimentally observed equilibrium growth rates, to enable in-depth studies of relevant atomistic growth mechanisms (e.g. island growth), the evolution of crystal defects, and the build-up of intrinsic residual stresses. Yet, all existing atomistic simulation methods have limitations that severely hamper the above investigations; specifically, the methods that are accurate enough to simulate realistic crystal defects are (far) too slow to achieve relevant film thicknesses.

To overcome the above-mentioned limitations, we propose a simple, yet powerful and computationally efficient atomistic method to directly deposit atoms at the minimum energy positions (MEPs) of the potential energy landscape, neglecting the system evolution through the intermediate states, hence the name Minimum Energy Atomic Deposition (MEAD). The MEAD method is motivated by the fact that high-density films with low crystal defect density and high Young’s modulus and yield strength are preferred for almost all applications, which is achieved in the cleanroom by slow growth rates at high deposition temperature to maximize the surface diffusion length of adatoms in order to let them find and deposit at the MEPs.

The MEAD method has been implemented in the LAMMPS Molecular Dynamics code and demonstrated using three examples of large scale depositions, i.e. deposition of Al on a ~25×25nm2 Al substrate, Al on ~25×25nm2 Si, and Si on ~25×25nm2 Si, achieving ≥100nm film thickness at near-equilibrium deposition rates while mimicking experimental growth rates and high densities (thus overcoming key limitations of existing kinetic Monte Carlo and Molecular Dynamics methods). Observed are realistic atomic growth mechanisms, realistic crystal defects such as stacking faults, twin boundaries, and grain boundaries, and the build-up of residual stress as a function of film thickness, all of which can be studied in detail with the MEAD method.