Semiconductor Power Devices

In a competitive market, technical systems rely on automation and process control to improve their productivity. Initially, these productivity gains were focused on attaining higher production volumes or less (human) labor-intensive processes to save costs. Today, attention is paid toward energy efficiency because of a global awareness of climate change and, above all, questions related to increasing energy prices, as well as security of energy and increasing urbanization. Consequently, it is expected that the trend toward more electrical systems will continue and accelerate over the next decades. As a result, the need to efficiently process electrical energy will dramatically increase.

Research on semiconductors has a long history [Smi59, Lar54]. Phenomenologically, they are defined as substances whose electrical resistivity covers a wide range, about 10⁻⁴−10⁹ Ω cm, between that of metals and insulators and which at high temperatures decreases with increasing temperature. Other characteristics are light sensitivity, rectifying effects, and an extreme dependency of the properties on impurities. After reaching a basic understanding of their physical nature in the 1930s and 1940s, semiconductors are defined now often by the band model and impurity levels leading to the observed phenomena: Semiconductors are solids whose conduction band is separated from the valence band by an energy gap E _g and at sufficiently low temperatures is completely empty, whereas all states of the valence band are occupied. Most important for application in devices, however, is that the conductivity can be controlled over a wide temperature range by impurities and that there are two types of impurities, donors which release electrons causing n-type conductivity and acceptors which provide positive carriers, the holes, leading to p-type conductivity. This allows the fabrication of pn-junctions.

pn-Junctions are the basic element of nearly all power devices. They are formed when the type of conductivity changes from p-type to n-type within the same crystal. pn-junctions are rectifying, they conduct current only in one direction of the applied voltage, called forward direction, whereas in the opposite direction, the blocking direction, the current is extremely small. Although the function of pn-junctions has been described theoretically already in 1938 [Dav38], their full technological significance became obvious only after the invention of the transistor and essential further advances in theory and technology [Sho49, Sho50]. This was a starting point of the enormous development of the semiconductor sector till now. Today, even the former poly-crystal rectifiers are thought for the most part to function by a pn-junction.

In the following some basic aspects of power device production technology will be described. The selection was done with the aim to describe the process steps which are important for the understanding of the power device operation and limitations.

Most power diodes are pin-diodes, i.e. they possess a middle region with a much lower doping concentration than the outer p- and n-layers enclosing it. Compared with unipolar devices (see Chap. 6), pin-diodes have the advantage that the on-resistance is strongly reduced by high-level injection in the base region, which is known as conductivity modulation. Hence pin-diodes can be used up to very high blocking voltages. The base region is not intrinsic, as suggested by the name. The intrinsic case – doping in the range of < 10¹⁰ cm⁻³ – would not only be difficult to attain by technology, extremely low doping would cause essential disadvantages in the turn-off behavior and other properties. Power diodes usually have a p⁺n⁻n⁺-structure, hence the so-called i-layer is actually an n⁻-layer. Since it is several orders of magnitude lower than the doping of the outer layers, the name pin-diode has become the usual denotation in almost every case

Schottky diodes are unipolar devices, which means that only one type of carrier is available for the current transport. If they are designed for large blocking voltages, the resistance of the base will increase strongly due to the lack of charge carrier modulation, as will be shown in the following. Schottky power diodes have been used for a long time, but in the last years they have gained an increased importance in the medium power range:

The transistor was invented in 1947 first in the form of a point contact transistor, whose emitter and collector were formed by sharp metal wires on a germanium block as base [Bar49, Sho49]. Soon it was clear that the metal semiconductor junctions at the point contacts can be replaced by two closely coupled pn-junctions. The first paper on a bipolar transistor of silicon with diffused emitter and base was published in 1956 [Tan56]. For a bipolar transistor as power switch, the emitter and base are fine interdigitated structures whose distance in the range of 30 μm have to be mastered; the technology for this was available in the 1970s. For a certain time, the bipolar transistor was the most important switching device in power electronics. But already at the end of the 1980s, the IGBT was introduced (see Chap. 10), and the IGBT began to replace the bipolar power transistor. Nowadays power converters are no longer equipped with bipolar transistors. In niche markets, e.g., as line deflection transistors in TVs, they have survived. However, recently, activities to develop bipolar SiC transistors were started.

The thyristor was the dominating switching device in power electronics for a long time. It was described already in 1956 [Mol56] and introduced to the market in the early 1960s [Gen64]. The acronym SCR (Silicon Controlled Rectifier) was primarily used for a thyristor in early publications and is still occasionally in use today. In its basic structure, a thyristor can be fabricated without very fine structures and with low-cost photolithography equipment. Thethyristor is still widely used in applications with low switching frequencies, such as controlled input rectifiers which are applied at the grid frequency of 50 or 60 Hz. A further actual application field of the thyristor is the power range that cannot be reached with other power devices – the range of very high blocking voltages and very high currents. For high-voltage DC power transmission, thyristors with 8 kV blocking voltage and more than 5.6 kA rated current have been introduced in 2008 as a single device in the size of a 6-in. wafer [Prz09].

The MOSFET basic structure was investigated early [Hof63]. For the comprehension of the function of a MOSFET (metal oxide semiconductor field effect transistor), the surface of the semiconductor may be examined at first. The surface of a semiconductor is always a disturbance of the ideal lattice due to the lack of the neighboring atom. Therefore, a thin oxide will always be built up on the surface or other atoms and molecules are adsorbed. Thus, these surface layers are normally electrically charged.

A lot of work was spent to combine bipolar devices with their superior current density with the possibility of voltage control as given in MOSFETs. Early works tried to combine thyristor-related structures with MOS gate control. However, a transistor-based device won the race. The insulated gate bipolar transistor (IGBT) was invented in the United States by Wheatley and Becke [Bec80]. The advantage compared to the bipolar transistor and MOSFET was described by [Bal82]. About 10 years later, IGBTs were introduced in the market by manufacturers from Japan and Europe. In a short time, IGBTs won an increasing share of applications and they replaced the formerly used bipolar power transistors, and nowadays even GTO-thyristors in the high-power range.

The operation of a power semiconductor device produces dissipation losses. The order of magnitude of these losses shall be estimated in the following example:

This chapter will deal with some destructive mechanisms in power devices, and typical failure pictures for them will be shown. Failure analysis requires a lot of experience, especially regarding the conditions in the power circuit at failure, which must be carefully considered. Although some of the failure pictures appear to be similar, it is difficult to draw conclusions only from pictures. However in practice the engineer often has the problem to find the reason for failures, and the following sections might be helpful.

Every power electronic switching action results in a deviation from the ideal sinusoidal AC current or the ideal homogeneous DC current. Switching events are usually done periodically in time. Every periodic event can be separated into a row of sinus and cosinus terms by means of Fourier transformation. With this tool, the generated frequencies, the harmonics, and their intensity can be calculated.

The expression “power electronic system” is used in different contexts with different meanings. A monograph on fundamental power electronic circuittopologies might well be found under the search criterion ‘systems.’ It is therefore appropriate to start with a precise definition of the object of discussion.