Multiscale simulation and optimization of copper electrodeposition

This thesis lays the groundwork for performing predictive simulations that can be used to guide the design of new copper electrodeposition additive baths for the filling of on-chip interconnects. Several continuum and stochastic algorithms were developed for accurate and efficient simulations.;A multiscale simulation code was created to simulate shape evolution during copper electrodeposition in the presence of additives. The model dynamically couples a kinetic Monte Carlo (KMC) model (for surface chemistry and roughness evolution) with a finite-volume model (for transport and chemical reactions in the electrolyte) and a level-set code (for tracking the macroscopic movement of the metal/electrolyte interface). The approach simulates the dynamic behavior during macroscopic shape evolution over extended periods of time while simultaneously tracking microscopic roughness evolution associated with nearly-molecular-scale events at the surface. The multiscale simulation code is designed to run on parallel computers with large numbers of processors.;A more efficient method was developed for copper electrodeposition when only macroscopic shape evolution is wanted and surface reactions can be modeled with mean-field reaction rate equations. In this method the KMC model is replaced with reaction rate equations. The new approach results in a simulation code that can run on a PC and achieve about 10 times speedup compared to the multiscale simulation code. The simulation code showed good agreement with experimental data which enabled its incorporation into simulation model-based optimal design framework. In the framework, the simulation code was treated as a "black box" and incorporated into derivative-free optimization package. This is the first time that such tools were developed for copper electrodeposition and shows the promise of computed-aided design (CAD) tools for such complex multiscale processes.;Two parallel KMC algorithms were also developed for the simulation of electrochemical nucleation and growth. Compared to the regular serial KMC algorithm, the developed parallel algorithms are able to speedup simulations and study large systems which may not be addressed by a single processor due to memory limitations.