Optimization of parameters for process simulation

One of the major tasks in process simulation is the estimation of dopant distributions in the device region. From experience it is known that different production lines may give different results under the same process conditions due to influences which are not yet understood. Recurrent tasks in process simulation are therefore calibrations of process parameters for particular sites. Likewise, advanced diffusion models require the indirect determination of physical parameters which cannot be measured dirctly. To minimize the requirement of user interaction for both purposes, a general parameter optimizer based on the method of Rosenbrock has been implemented in the one-dimensional process simulator ICECREM. This paper describes why the authors have this particular algorithm and outlines the experiences with the method used.