When device physicists, process engineers and mathematical model developers assemble this year at the 50^th anniversary of International Electron Devices Meeting (IEDM) in December, they will pause to look back and see how far they have come from the days when electron tubes reigned, and what the road ahead looks like. Even scientists who first played with germanium junction transistors in the early days of semiconductor technology will not believe the form and shape it has taken today. The first integrated circuit (IC), a phase-shift oscillator, developed in 1958 by Texas Instruments' Jack Kilby boasted one transistor and few passives. With Moore's Law still holding true today, we are talking about packing several hundred million transistors on a single state-of-the-art silicon microprocessor chip and moving toward placing nearly a billion by next year. By the way, the popular Moore's Law is an observation that says the number of transistors per IC will double roughly every couple of years. This prophecy by Gordon Moore has been true since the mid-1960s and is expected to take us through the nano era.

Only a few years ago, nanometer CMOS process was seen as a big challenge. While few have taken 90 nm CMOS process to production, several researchers are ready to discuss 65 nm transistor structures this year at the IEDM and will unravel the roadmap to scaling down to 45 nm and below. While silicon and CMOS have been at the center of the digital revolution, these advances have also trickled into other areas. Thus, it is not surprising to see the gap between the latest digital CMOS process and RF/analog CMOS to be narrowing. Lately, as an example, we have seen demonstration of 0.5 mm CMOS RF antenna switches at 5 GHz going head-on against GaAs counterparts, with conversion to 0.25 µm in the works.

And, with advances in process technology and material science, so many other new transistor structures and materials have emerged on the scene in the last five decades to serve a variety of RF and microwave applications, including wireless and mobile communications. While gallium arsenide (GaAs)-based compound semiconductor devices have become popular in the microwave realm over the years, we have seen Indium phosphide (InP) and gallium nitride (GaN) come to the rescue of developers in several wireless and wireline applications. Consequently, GaAs and InGaP-based heterojunction bipolar transistors (HBTs) with ultrafast speed are making strong inroads into the wireless handset niche, along with pseudomorphic high electron mobility transistors or pHEMTs. GaN HEMTs are competing for sockets in the base station applications. For high-power RF applications, silicon carbide (SiC) FETs have also emerged to give designers another option. Concurrently, silicon germanium (SiGe) transistors are driving biCMOS to new frontiers in the RF and microwave space. With the merger of SiGe HBTs with mainstream CMOS, biCMOS process promises to accomplish unprecedented levels of integration of RF and analog/mixed signal on the same die.

The pace of development continues unabated. Thus, at the upcoming IEDM, we will hear InP HBTs flaunting a maximum frequency of 430 GHz with the potential of achieving 500 GHz fmax performance. With the interest to squeeze more RF performance from silicon, efforts are in progress to make SiGe HBTs much faster. Results indicate that SiGe HBTs have improved significantly in speed to deliver fmax in the range of 300 GHz.

Despite all the progress that IEDM has reported in the last 50 years, the fundamental transistor has remained the same — an on/off switch. The focus has been on scaling, density, speed, power and cost. With feature sizes of silicon transistors approaching physical limits, it is time to consider new technologies and devices. Don't you think it's time to replace semiconductors, just as semiconductors displaced tubes to spark a revolution?