Since the invention of the integrated circuit in the late 1950-ies the feature size of semiconductor devices has been decreased continuously. For the development of the first device generations experiments and analytical methods were sufficient. The dramatically reduced cost of computer hardware and the simultaneous increase in computational power made the development and use of process and device simulation programs feasible. During the last decade device simulation has become a standard tool for the development and improvement of semiconductor devices.
The increasing complexity of semiconductor devices and the thereby raised manufacturing cost and time increase the demand for sophisticated simulation tools. Additionally, many effects can be investigated more detailed and more efficiently by simulation than would be possible by experiments. Many properties of semiconductor devices depend on the specific processing and manufacturing methods. Therefore simulations always have to be calibrated to appropriate measurements.
The continuous down scaling of the feature size makes it necessary to extend existing physical models and to take additional effects into account. Therefore it is very important for a generic device simulator to be able to integrate new models which is achieved by the use of an object oriented model server.
The progress in the field of software engineering and simulator development allows to implement generic device simulators which are capable of simulating a wide range of different semiconductor devices. In comparison to simulators specialized on simulation of a certain class of devices generic device simulators provide the development engineer with more flexibility and the ability to investigate completely new device concepts. These advantages of generic device simulators come at the cost of higher system complexity and higher demand on computational resources. But the continuous increase of available computational resources and the simultaneous decrease of the cost of these resources reduces the effect of these disadvantages.
Generic device simulators can be extended to be capable of so-called mixed-mode simulations in which a circuit of distributed and lumped devices is simulated. This allows the simulation of the device characteristic within a test circuit without the need of the extraction of compact model parameters and a subsequent circuit simulation.
Device simulation is not only used for simulation of the electric characteristic of semiconductor devices. In combination with optimizers and process simulators device simulation is used for inverse modeling and doping characterization.
The simulations presented in this thesis were performed by the generic device simulator MINIMOS-NT [1][2]. MINIMOS-NT is capable of simulating devices made from Si, SiGe, and III-V compound semiconductors and including heterojunctions using the drift-diffusion transport model or a hydrodynamic transport model. An especially useful feature is the simulation of small circuits of distributed and lumped devices, so-called mixed-mode simulation [3].
Chapter 2 presents the application of MINIMOS-NT for the investigation and simulation of the newly developed AVC method for doping characterization. New models had to be implemented to account for the effects of high energy electron beams used by this method to probe the doping in a semiconductor.
In Chapter 3 the treatment of material interfaces and the thereby arising problems are discussed. By an appropriate transformation of the equation system the spectral condition number can be reduced and convergence is improved. This method is applied to the hydrodynamic simulation of a HEMT.
The two-dimensional transient simulation of a surface-channel charge-coupled device and the time-step control used for transient simulations is described in Chapter 4.
The optimization of the on-resistance of a VDMOS using process and device simulation is presented in Chapter 5 and Chapter 6 demonstrates the extraction of the compact model parameters for a BJT. The results of a MINIMOS-NT mixed-mode simulation of a Colpitts oscillator are compared to the results obtained by a circuit simulator using the extracted compact model parameters.