In the modern world, electronic components based on semiconductor devices have become a very important part of an uncountable number of applications. The performance gain and the miniaturization of semiconductor devices continuously open up new possibilities. Of comparable importance for the entire development is the decreasing costs per device. The main factor for cost reduction is the dramatic increase of the device density on a silicon waver [1,2]. Therefore, a considerable amount of industrial and academical research is being performed to enable the continued shrinking of the devices.
Another important method for the cost reduction of electronic components is to combine different functional groups in a single integrated semiconductor die. This technique is applied, for example, in smart power devices which incorporate power and high-voltage devices with additional functions, such as power control, sensing and protection, and interfacing. To accomplish this, different technologies have to be combined. A majority of the smart power applications integrates low-voltage CMOS logic with high-voltage and/or power device technologies [3,4,5,6].
When talking about high-voltage devices, one has to clarify more specifically the voltage range, because the terminus ``high-voltage'' strongly depends on the application field. In power engineering and in the area of transmission and distribution, for example, voltages above 1,000V are considered as high-voltages [7]. In long-distance electric power transmission lines this lower limit is as high as 100kV. At the other end of the voltage range, especially in smart power devices in which CMOS logic is integrated, one considers voltages starting from 5V as high voltages [5]. In the automotive industry, for example, the high-voltage part of smart power devices is dominated by the 14V vehicle power supply. Contributing to the disturbances on the supply and signal lines encountered in automotive environments, these devices are often rated in the range of 50-100V [8,9,10]. In this thesis the term high-voltage refers to the range from 5 to 60V. In Chapter 2 various device architectures, which are used in this voltage range, are presented. Due to the wide usage in industry, the focus is put on the double-diffused metal-oxide-semiconductor field-effect transistor (DMOSFET) and its variants in lateral and vertical orientation. Also several device design approaches such as field shaping and isolation techniques are shortly described.
Modern semiconductor devices have to fulfill many requirements in terms of performance, reliability, and costs. Certain reliability goals must be met, which depend on the field of application and other considerations, for example, safety, security, and liability issues. An overview on reliability in general followed by some specific reliability concerns found in semiconductor devices is given in Chapter 3. However, the down-scaling and the increasing complexity of devices and integrated circuits make it very challenging to reach all specified design goals. Therefore, more and more often device simulation tools are employed in development, research, and optimization. These tools, commonly referred to as technology computer aided design (TCAD) tools, aim to reproduce the physical mechanisms and hope to predict the device behavior [11,12,13,14]. The most important formulas and physical models which are needed for device simulation within the drift-diffusion model are described in Chapter 4. This chapter also discusses the possibilities and limitations of this model to describe hot-carrier phenomena, which are of crucial importance for high-voltage device reliability. Strictly speaking, a physics-based modeling approach of these effects requires an exact solution of the Boltzmann transport equation. In this context the Monte Carlo method [15,16] proved to be one of the most popular approaches. In fact, this method gives accurate results and allows to easily incorporate various physical models. Unfortunately, the computational cost required for this method are very high, which make them not too appealing for industrial use. On the other hand, the drift-diffusion model is numerically stable and can be solved efficiently [11,14]. The modeling approaches presented in this thesis aim to deliver good results in reasonable simulation times and, therefore, are based on the drift-diffusion model.
The two hot-carrier related reliability effects discussed in this thesis are the impact-ionization generation and hot-carrier degradation. Chapter 5 is devoted to the physical phenomenon of impact-ionization and presents different modeling strategies and aspects. The importance of impact-ionization for the reliability of smart power devices is demonstrated in a case study. In Chapter 6 hot-carrier degradation in MOS devices is discussed. After a review of currently used modeling techniques, a distribution function based model is presented, which is currently under development [17,18]. This model is based on results obtained within the Monte Carlo simulations and is therefore computationally very demanding. To overcome this disadvantage, possible approximations using the drift-diffusion scheme are presented and discussed.
The simulation of semiconductor devices in TCAD tools requires the solution of a system of non-linear differential equations. To solve this system, a spatial discretization scheme has to be used to transform the equations into a system of difference equations. In the context of device simulation, the box integration method proved to be very reliable. Iterative solution techniques are required to obtain a solution for this numerical problem. Simulations in high-voltage devices turned out to be numerically challenging, especially in combination with impact-ionization. Hence, investigations on various vector discretization schemes were done, which are presented in Chapter 7.