List of Tables

• 2.1. Ignored derivatives during the first steps of the Newton iteration.
• 3.1. Admittance matrix calculated by MINIMOS-NT for a frequency $f = 100\$MHz and terminal voltages $\ensuremath{V_\mathrm{GS}}= 0.0\,V$ and $\ensuremath{V_\mathrm{DS}}= 0.0\,V$.
• 4.1. Eigenvalues of the compiled, pre-eliminated, sorted, and scaled system matrix.
• 4.2. Newton adjustment levels.
• 4.3. Comparison of terminal quantities with and without the correction. Note that the boundary charges are significantly different.
• 4.4. The transposed and untransposed transformation matrix before and after the correction. The column index stands for the source or the target of the transformation, respectively.
• 5.1. Six two-dimensional, five mixed-mode device/circuit (the number of devices is given in parenthesis), and five three-dimensional simulations were used for evaluating the solver performance. The dimension of the linear equation system, the number of non-zero entries, and the typical number of DC Newton iterations are given.
• 5.2. This table provides general quantitative information about the simulations using the in-house ILU-BICGSTAB on the IBM cluster. After the index of the simulation, the dimensions of the inner and complete equation system is given. The next column shows the share of the inner in respect to the complete system. In order to analyze how the simulation time is spent, the remaining columns contain the shares of the initialization, pre-processing, assembling, solving (includes all steps shown in Table 5.3), quantity update, and post-processing, respectively.
• 5.3. This table provides general quantitative information about the solving process which includes the MCSR conversion, compiling of the complete linear system, pre-elimination, sorting, scaling, preconditioning (ILU), and solving with the BICGSTAB. The simulations were started on the IBM cluster.
• 5.4. This table shows the same information as Table 5.3, but without the Newton adjustment.
• 5.5. Three two-dimensional device structures with the given gate lengths were simulated. The dimensions of the linear equation systems are given depending on the transport model used. For the drift-diffusion simulation and BICGSTAB, the CPU time as well as the share of the simulation time spent for compiling, pre-eliminating, sorting, scaling, and solving are specified.
• A.1. Keywords provided in the Contact section.
• A.2. Valid selections for external solver modules.
• B.1. Stepping functions provided by MINIMOS-NT.
• B.2. Information provided to delta stepping control algorithms.
• B.3. Information expected from delta stepping control algorithms.
• B.4. Information provided to conditional stepping control algorithms.
• B.5. Information expected from conditional stepping control algorithms.
• B.6. General stepping control keywords.
• C.1. List of all special-purpose arguments, which are automatically generated by the library.
• C.2. Definition of the test levels.