3.2 Overview

This section concentrates on the overall structure of the TCAD system as implemented during this work. The simulation work flow is outlined and certain specialities of this approach are shown. This work flow refers to the TCAD software suite of former ISE AG (now acquired by Synopsys) including the main simulators DIOS [82] (process simulator) and DESSIS [83] (device simulator). Since this work flow is only one of multiple possible implementations, some parts of this section cannot be generalized or applied to other TCAD installations (like TSUPREM4 and MEDICI installations) but care has been taken to show the TCAD-work flow as general as possible.
The different levels of functionality of the TCAD system are shown in Figure 3.2.

Figure 3.2: TCAD work flow scheme showing possible iteration loops

The purpose of the process simulation is to provide the structural information of the device under scope, consisting of the boundary including the composition of the different materials involved (e.g. polycrystalline silicon, single crystalline silicon, silicon dioxide, metals etc.). In addition, the doping concentration inside the silicon has to be available. The process simulation takes the photo mask information and the process flow to model the evolution of above mentioned information (boundary and dopant) over the multiple steps of the process.
The mesh for solving the partial-differential equations typical for the physical and chemical processes occurring during processing is normally of unstructured type, to model the steep gradients of the doping distributions with good accuracy, but with a low number of mesh points where the physical fields (doping concentration, point defect concentrations etc.) are not varying much. A detailed description of a process simulator can be found in, e.g., [113],[114],[115]
The boundary and dopant information is then used as an input to describe the electrical behaviour of the device under scope by calculating the potential distribution and the carrier transport phenomena (current concentrations etc.) via solving the PDEs describing their physics. A detailed description of the underlying principles of a device simulator can be found in, e.g., [116],[117],[106]
Since the requirements on meshes for process and device simulations, respectively, are very different, a re-meshing step is necessary to minimize the numerical error, and the number of mesh points necessary for a certain accuracy of the solution. This re-meshing is normally based on the gradient or difference refinement criteria. In some cases this approach is not sufficient to get a good mesh. The inversion region of a MOSFET channel is a good example for the problems gradient refinement criteria are facing. However in recent investigation approaches are outlined to overcome or, at least, to tackle these limitations [118],[119].