6. Conclusion and Outlook

During the last 15 years several a flexible, efficient, physically based ion implantation simulator has been developed by Hobler [30], Stippel [78], Bohmayr [7]. The history of the development is summarized in Chapter 4. The goal of this work was to improve the quality and to optimize the performance of the simulator and to develop new models and to implement new features required for modern semiconductor process simulation for ULSI technology. All desirable new features have been added to improve as well the flexibility as the efficiency of the simulator.

Originally the simulator existed in two versions, one for two-dimensional applications and one for three-dimensional applications. These version have been unified to the multidimensional simulator MCIMPL. Now the simulator is applicable to arbitrarily shaped one-, two- and three-dimensional input structures consisting of various materials.

As an additional feature several succeeding ion implantations without an annealing process in between can be simulated. The damage generated during one ion implantation simulation step can be reused as an additional input information by the next ion implantation simulation step.

Since the implantation of molecular ions has gained importance especially for the implantation of boron, two models have been developed to deal with this problem. On the one hand side a rigorous model for the simulation of the implantation of molecular ions has been implemented, which allows to calculate the distribution of all atom species contained in the molecule. On the other hand side a simplified method has been added which requires less computational resources, but calculates just the distribution of one significant atom species.

To improve the accuracy of the simulation results, especially of the point defect distributions the Follow-Each-Recoil method has been developed. This method allows a rigorous treatment of the implantation induced damage generation process and to provide a very accurate input for the simulation of rapid thermal annealing processes.

A point response interface method has been developed to couple the Monte-Carlo simulator with an analytical ion implantation simulator. This gives shorter simulation times of an analytical simulator without loosing the flexibility of a Monte-Carlo ion implantation simulator. The accuracy of a coupled simulation is as high as the accuracy of a full Monte-Carlo simulation for many applications.

Besides the development and the implementation of new models this work focused on a further reduction of the simulation time. Therefore several numerical algorithms have been optimized, which could be identified as having a strong impact on the total simulation time.

Additionally the simulation time could be significantly reduced by implementing the Trajectory-Reuse-Method. By using this method certain trajectories can be copied instead of being calculated, which results in an enormous reduction in the simulation time. Especially if the simulation domain consists of large areas made of amorphous materials, because the Trajectory-Reuse-Method can only be applied to trajectories passing through amorphous materials.

Finally a Parallelization method has been implemented to distribute the simulation among several workstations, which is the typical environment for TCAD simulations. The Parallelization method has been designed to minimize the communication overhead. Therefore the simulation can be performed on a workstation clusters connected by fairly slow networks.

Further improvements of the simulator are possible by calibrating the empirical parameters of the electronic stopping and the damage generation models for a wider range of ion species and target materials, especially for compound semiconductors.

Since the calibration of the existing empirical model requires a lot of measurements, it would be useful to extend the empirical electronic stopping model to a more physically based model. Thereby the number of empirical parameters could be reduced which would ease the calibration.

Finally a development of a temperature dependent self annealing model would be very important. Especially if the implantation temperature is raised or if the implantation dose rate becomes very high, the self annealing during ion implantation can no longer be neglected. Therefore coupling with a molecular dynamic simulator may be useful.