The evolution of power device technology

Power semiconductor devices and their associated technology have come a long way from their beginnings with the invention of the bipolar transistor in the late 1940's. Presently, the spectrum of what are referred to as "power devices" span a very wide range of devices and technology from the massive 4 in, 3000-A thyristor to the high-voltage integrated circuit and the power MOSFET, a device of VLSI complexity containing up to 150 000 separate transistors. In this paper, the past, present, and future of power devices will be reviewed. The first section will be a historical perspective indicating the key events and developments of the past that brought the power devices of today to their present state. The second section of the paper will review the technology and characteristics of bipolar power devices with separate subsections on thyristors, the gate turnoff thyristor (GTO), and the bipolar transistor. Within the thyristor subsection there will be discussions of the phase control thyristor, the inverter thyristor, the asymmetric thyristor (ASCR) the reverse conducting thyristor (RCT), the gate-assisted turn-off thyristor (GATT), and finally the light-triggered thyristor (LTT). The third section of the paper is devoted to the new field of integrated power devices and will review the evolution to the present power MOS devices including the power MOSFET, the insulated gate transistor (IGT), and the high voltage IC (HVIC). The last section of the paper reviews the future of power devices with projections as to future ratings of power devices for both the traditional bipolar devices, such as the thyristor, GTO, and bipolar transistor, as well as the integrated devices such as the MOSFET and the IGT. In case of the former, in particular the thyristor, the maximum device ratings will be tied to the availability of large area float zone material, currently difficult to obtain in the high resistivities needed for power devices. In the case of integrated devices, the maximum ratings will be limited by the maximum die area for which acceptable device yields can be obtained. This is identical to the situation for conventional IC's since much of the unit processing is the same.