INTEGRATED CIRCUIT MANUFACTURING is nowadays a multi-billion dollar business and one of the most complex industry branches in the world.
Historically the development costs for introducing a new technology generation (technology node) have been steadily increasing. This trend is even accelerated by entering the nanotechnology regime with its highly complicated new lithography processes. To keep this trend under control, Technology Computer Aided Design (TCAD) has gained more and more importance.
This ``front loaded'' approach in development has improved the speed and quality of the semiconductor process technology development significantly. It was also very successful in reducing the development cost significantly (by 35% in 2005[1]).
However TCAD is still mainly a very specialized tool for only a small group of engineers inside semiconductor companies. Despite the big savings potential, the consistent application of this methodology on the control of the manufacturing process is still in the fledgling stages.
Technology simulation has not done the leap, from a tool for a small group from specialists to a tool for a larger number of users in the semiconductor production area, yet.
Main problems of the broader use result from the complex, often little intuitive use of the tools, the complex underlying physical models and particularly by the missing integration into the work flow of a modern semiconductor production line.
This work aims to bridge the distance between the semiconductor manufacturing line and the TCAD group, which exists in nearly each semiconductor production company. Special attention has to be paid to the transfer of the operational work flow sequences to the TCAD system in a similar form.
This new system enables non-specialized engineers to profit from the advantages of a theoretical evaluation through the closely integrated TCAD framework.
In this work the overall situation of both worlds was analyzed and categorized in a structured, hierarchical way. A consistent and effective TCAD work flow was set up. The necessary information for this work flow was identified. Additionally the existing data interfaces between manufacturing and simulation were analyzed and their structure and coupling nodes were represented.
In the following the interfaces between TCAD and manufacturing, identified thereby, were subject to a more throughout analysis and finally an integrated interface system between the two "worlds" was implemented. A strong emphasis was put on the border condition that all available TCAD software packages could be used together with this new integrated interface system.
In order to be able to generate the necessary input information for simulations as automatically and resistant to errors as possible, based on this analysis a compact set of converters and data transfer procedures was defined. The interaction with the user of the TCAD system was limited to the absolutely necessary minimum. This led to a strong improvement of the quality, reliability and also predictability of the results of the TCAD simulations.
The converters were integrated into the entire TCAD work flow. Several examples of nearly each aspect of the typical TCAD work flow show the most important effects of this new approach. In addition, during the development of new process technologies, particularly for the support of typical production problems (like yield problems in the manufacturing) this approach has been tested successfully. Furthermore, optimizations of furnace programs (a task which may frequently occur during manufacturing of qualified processes too) have been performed. The use of diffraction correction calculation for a better representation of interconnect structures could be shown. Thus a substantial improvement in the computation of parasitic elements, the optimization of an EEPROM cell and the three-dimensional simulation of a lateral pin diode could be obtained. As a further application area the inverse modeling of a polycrystalline fuse was shown. The difficult to measure thermoelectric material parameters of the materials used in the fuse (e.g. tungsten, titanium, titanium nitride) were determined through inverse modeling. Finally the use of technology simulation within the area of statistic process control was demonstrated.
The work closes with a short outlook and open problems.