Wdass’s CMOS circuit

Dec. 5, 1967 F. M. WANLASS 3,356,858 LOW STAND-BY POWER COYPLEYENTARY FIELD EFFECT CIRCUITRY Filed June 18. 1963 5 Sheets-Sheet 5

Va Vi - IUL._- Vi00 TGATE ,8 la

Most people remember the ’60s as the era of the Beatles, the Vietnam Wx, Woodstock, and 35-cent-a-gallon gasoline. But electron- ics engineers of a certain age remember them as the days of free-wheeling ex- perimentation when-at some companies, at least-bright young Ph.D.s were given ‘83 latitude to see what they could create with- so7 out interference from corporate managers. One bright young Ph.D. of that time was Ianlass S patent portrayed an integrated CMOS inverteer. Frank Wanlass, who this February received Me investigating that possibility, he found MOSFET was not. It would be years before the 1991 IEEE Solid-state Circuits Award for ) his surprise that aluminum electrodes MOS surface physics was well enough un- his invention of complementary-MOS eposited by an electron beam evaporation derstood to permit the fabrication of such (CMOS) logic circuitry. Lachine yielded quite stable devices. The devices. Consequently, Wanlass made Wanlass’s interest in MOS technology roblem, he began to suspect, was not alu- depletion-mode n-channel and dates to 1962, when he was still studying for hum diffusion after all, but contamination. back-biased their bodies negative with re- his doctorate in solid-state physics at the The usual way to deposit aluminum in spect to their sources to turn them into University of Utah in Salt Lake City. Upon lose days was to evaporate it by placing an enhancement-mode units. reading about the Radio Corporation of uminum wire in contact with a heated tung- The concept worked. The first demon- America’s work with thin-film cadmium sul- .en filament. Wanlass reasoned that the stration circuit, a two- inverter, fide field-effect (FETs), he be- rocess was introducing positive ions into consumed just a few nanowatts of standby came intrigued by the simple structure of the le aluminum and thence the . power. Equivalent bipolar and PMOS gates devices, which he thought would make it Further investigation inculpated sodium consumed milliwatts of power even in stand- easy to design fairly complex integrated cir- intamination from both the tungsten and by. CMOS shrank the standby power con- cuits. But the devices were unstable. When ie aluminum. Electron beam evaporation sumption by six orders of magnitude! left on a shelf for just a few hours, their elec- dved the problem because the electron- The speed was impressive enough, too. trical parameters changed dramatically. Sam apparatus had a shutter mechanism Propagation delay times were on the order Wanlass believed that making the FETs in iat protected the silicon wafers from a car- of 100 nanoseconds-about half the speed silicon would solve the problem. After all, m crucible of molten aluminum until the of bipolar, but almost an order of magnitude Fairchild Research and De- uminum was at evaporation temperature. faster than PMOS. velopment had made some very stable and he sodium, having a much lower boiling On June 18, 1963, Frank Wanlass applied reliable bipolar transistors in silicon using its iint, boded away before the shutter opened. for patent protection for his CMOS concept, , so why not use the Same ma- Wanlass next had the idea for CMOS. “It and in due course, was granted US. patent terial to make stable MOSFETs? :curred to me,” he told IEEE Spectrum no. 3 356 858, the rights to which became So he went to work for the Fairchild Semi- I a recent interview, “that a complemen- part of Fairchild’s patent portfolio. That pat- conductor subsidiary of Fairchild Camera iry circuit of NMOS and PMOS devices, ent described the overall concept and three and Instrument Co. in Pa10 Alto, Calif., in it could be made, would use very little specific circuits-an inverter, a NOR gate, August 1962, unaware that other research- 3wer. In standby, it would draw practically and a NAND gate-from which any digital ers were already working on silicon MOS- 3thing-just the leakage current. ” function can be built. In addition to the dis- FETs at both RCA and Bell Laboratories. A His boss Gordon Moore, now the chairman crete implementations that were actually few months later, he made his first p-channel Intel Corp., gave him a free hand to pur- built, the patent includes the representation silicon MOSFETs, which, like all early ie his idea. At first, Wanlass tried to build of an integmted CMOS inverter shown here. MOSFETs, were a great disappointment. At CMOS circuit monolithically, but that was (Wanlass’s laboratory notebooks, with his 10-20 volts, their threshold voltages were difficult he decided to prove the concept on& drawings, were lost, probably when very high and very unstable. They were no ith discrete p-channel and n-chan- was taken over by better than RCA’s CdS devices. MOSFETs instead. Because only p-chan- National Semiconductor Corp. in 1987.) Wanlass speculated that the aluminum :I devices were available, he had to start Neither Wanlass nor Fairchild Semicon- gate electrode was diffusing into the gate VT building an n-channel silicon MOSFET. ductor grew rich from the invention. In those oxide. If that were the case, the use of a The CMOS concept requires that both of days companies traded the rights to their more inert metal would solve the problem. s transistors be enhancement-mode patent portfolios. Still, the integmted CMOS ?vices. But whereas PMOS transistors inverter shown, although never built, is the Michael J. Riezenman Contributing Editor ere inherently of that type, the n-channel progenitor of all CMOS ICs today. +

44 0018-9235/91/0005-0044$1.0001991 IEEE IEEE SPECTRUM MAY 1991