Foreword

The Semiconductor Industry^†

Semiconductor devices such as diodes, transistors and integrated circuits can be found everywhere in our daily lives, in Walkman, televisions, automobiles, washing machines and computers. We have come to rely on them and increasingly have come to expect higher performance at lower cost.

Personal computers clearly illustrate this trend. Anyone who wants to replace a three to five year old computer finds that the trade-in value of his or her computer is surprising low. On the bright side, one finds that the complexity and performance of the today’s personal computers vastly exceeds that of their old computer and that for about the same purchase price, adjusted for inflation.

While this economic reality reflects the massive growth of the industry, it is hard to even imagine a similar growth in any other industry. For instance, in the automobile industry, no one would even expect a five times faster car with a five times larger capacity at the same price when comparing to what was offered five years ago. Nevertheless, when it comes to personal computers, such expectations are very realistic.

The essential fact which has driven the successful growth of the computer industry is that through industrial skill and technological advances one manages to make smaller and smaller transistors. These devices deliver year after year better performance while consuming less power and because of their smaller size they can also be manufactured at a lower cost per device.

The purpose of this text is to explore the internal behavior of semiconductor devices, so that we can understand the relation between the device geometry and material parameters on one hand and the resulting electrical characteristics on the other hand.

This text provides the link between the physics of semiconductors and the design of electronic circuits. The material covered in this text is therefore required to successfully design CMOS-based integrated circuits.

The Metal-Oxide-Silicon Field-Effect-Transistor (MOSFET) is the main subject of this text, since it is already the prevailing device in microprocessors and memory circuits. In addition, the MOSFET is increasingly used in areas as diverse as mainframe computers and power electronics. The MOSFET’s advantages over other types of devices are its mature fabrication technology, its successful scaling characteristics and the combination of complementary MOSFETs yielding CMOS circuits.

The fabrication process of silicon devices has evolved over the last 25 years into a mature, reproducible and reliable integrated circuit manufacturing technology. While the focus in this text is on individual devices, one must realize that the manufacturability of millions of such devices on a single substrate is a minimum requirement in today’s industry. Silicon has evolved as the material of choice for such devices, for a large part because of its stable oxide, silicon dioxide (SiO₂), which is used as an insulator, as a surface passivation layer and as a superior gate dielectric.

The scaling of MOSFETs started in the seventies. Since then, the initial 10 micron gatelength of the devices was gradually reduced by about a factor two every five years, while in 2000 MOSFETs with a 0.18 micron gatelength were manufactured on a large scale. This scaling is expected to continue well into the 21^st century, as devices with a gatelength smaller than 30 nm have already been demonstrated. While the size reduction is a minimum condition when scaling MOSFETs, successful scaling also requires the reduction of all the other dimensions of the device so that the device indeed delivers superior performance. Devices with record gate lengths are typically not fully scaled, so that several years go by until the large-scale production of such device takes place.

The combination of complementary MOSFETs in logic circuits also called CMOS circuits has the unique advantage that carriers only flow through the devices when the logic circuit changes its logic state. Therefore, there is no associated power dissipation if the logic state must not be changed. The use of CMOS circuits immediately reduces the overall power dissipation by a factor ten, since less that one out of ten gates of a large logic circuit switch at any given time.

Bart Van Zeghbroeck, Boulder, August 2000