The report in 1958 of the invention of the tunnel diode by Sony Corp.'s Leo Esaki created almost as much excitement as the announcement of the transistor by Bell Labs a decade earlier. This may have been because tunneling was a fundamental idea from quantum physics, and here it was in a simple p-n junction. Here was a super-fast device with switching speeds that left most transistors in the dust. As the fame of the tunnel diode spread through the electronics world, the device often took on the name of its inventor. It was widely known as the Esaki Diode. And the development earned Esaki a Nobel Prize in 1973. Before the award of that Nobel, the world was shaken when it learned that Esaki was leaving Sony in Japan to join IBM in New York State, an almost unprecedented happening.

Japanese companies, especially large and reputable ones like Sony, were known for lifetime employment. Once you earned employment at an outfit like Sony, you knew that you would stay there forever or, at least, till you retired or died. So it was almost unheard of for somebody to leave, especially a high-ranking individual like Leo Esaki.

So remarkable was Esaki's departure that both Sony and IBM issued press releases indicating that Esaki's move had the blessings of both companies and he was leaving under the most honorable circumstances. (Esaki is now president of the University of Tsukuba in Japan.)

Switching times for the Esaki Diode were in the picosecond range while most transistors at the time were really stepping along if they managed milliseconds. This was clearly the device for microwave circuitry and new computer circuits, especially as a fast switch.

Just two problems

There were, however, two small problems. First, nobody was manufacturing the tunnel diode just yet. So you couldn't design those super-fast circuits that many customers wanted. But surely, any day now, somebody would announce that he had tunnel diodes in production.

The second problem was that the tunnel diode, like any other diode, was a two-terminal device. And engineers had no experience designing amplifiers with two-terminal devices. In fact, they really didn't know how. Vacuum tubes, which were still the preferred device for high-speed and microwave circuits, included a control grid where you could apply a signal to be amplified. A voltage applied to the grid would modify a current flowing from cathode to anode to a load resistor, across which the voltage would follow the voltage applied to the grid.

Corresponding to the vacuum tube's grid, cathode and anode, transistors had base, emitter and collector. But the tunnel diode had only an anode and cathode — nothing comparable to a grid or a base.

With the optimism that preceded many other great breakthroughs, the technical community assumed that the problem of amplifying with a two-terminal device would quickly be overcome as soon as engineers could get their hands on some tunnel diodes. Then they would apply their ingenuity and design dazzling circuitry.

The technical world waited for production. In 1959, a year after General Electric announced the silicon-controlled rectifier, a device that was to enjoy stunning success, the company also announced that it planned to offer experimental samples of tunnel diodes.

In a press release dated July 23, 1959, Guy Suits, GE's vice president and director of research, said that the tunnel diode, a little over a year old, would find its way into high-speed computers, TVs, communications equipment, nuclear controls and space vehicles. He hailed the work of Bob Hall and J. Tiemann in Schenectady, N.Y., and Nick Holonyak and Arnold Lesk at GE's Advanced Semiconductor Lab in Syracuse, N.Y.

Suits stressed the speed of the device. He cited oscillation frequencies higher than 2,000 megacycles (this was before they invented megahertz), and told of tunnel diodes switching in a fraction of a milli-microsecond (before they had nanoseconds). He hailed the tunnel diode's resistance to nuclear radiation and its operation at temperatures up to 650°F.

Going further, Suits predicted that the tunnel diode's unique properties, including negative resistance over part of its operating-voltage range, would allow its use as an amplifier, an RF power generator and a switching device. He added that its simplicity made possible development of integrated circuits, in which entire circuits for some applications could be formed on a single semiconductor structure. (This was just four months after Texas Instruments announced integrated circuits, which TI called solid circuits, invented by Jack Kilby. The announcement was made in New York at the annual show and convention of the Institute of Radio Engineers, ancestor of the IEEE.)

Negative resistance is what intrigued engineers. For part of the tunnel diode's operating range, an increase of bias voltage resulted in a decrease in current, a very rapid decrease, then an increase again.

Nick Holonyak, Arnold Lesk, Uri Davidsohn and Mel Aarons published a major technical paper, "Germanium and Silicon Tunnel Diodes — Design, Operation and Application." This was the authoritative article. Anybody who was anybody in the field read it. Articles appeared in the 1959 Wescon Convention Record, in Electronics magazine, in Electronic Design, in EDN. Leading electronics publications begged for articles on tunnel diodes and rushed them into print.

And then, with little fanfare, the anticipated market disappeared, as did the passionate interest.

It turned out, despite the optimism, that using a two-terminal device as an amplifier was no trivial matter. You really needed to separate the input from the output. And there was something else. It was a low-impedance device.

An old-timer's thought

Nick Holonyak, one of the tunnel-diode pioneers at GE in 1958 and 1959, offers some insight. Holonyak is now John Bardeen professor of electrical and computer engineering and physics at the University of Illinois at Urbana-Champaign. This chair, by the way, was sponsored by the Sony Corp. (Esaki's outfit).

Holonyak points to the problem with a low-impedance device. You can have currents — even monstrous currents — with very little voltage.

In 1984, GE sold its inventory of tunnel diodes to Germanium Power Devices, a company that specializes in carrying obsolete or obsolescent components.

At one time, several companies fabricated tunnel diodes for their own use. In the late 1960s, Tektronix used its own tunnel diodes in a sampling-scope step generator with a 25-ps rise time. The company no longer manufactures tunnel diodes.

Now, says Tek's John Rettig, "we have better technology to get flatter steps for roughly the same speed. We use a switch made of two Schottky diodes back to back. You drive current into a base line, then, as quickly as you can, you switch currents out of a front diode and into a back diode. And since you cut off to zero volts all the time, there is no thermal associated with the switch." The tunnel diode generated heat, Rettig points out.

In Boulder, Colo., Picosecond Pulse Labs, a spinoff from what used to be called the National Bureau of Standards, still sells a 20-year-old pulse generator that uses a tunnel diode as the switching element. The generator, the Model TD 1107, is a direct replacement for the discontinued Hewlett-Packard Model 1106. It can deliver 230-mV pulses with rise times in the range of 20 to 40 ps, depending on the quality of the tunnel diode. HP now buys these generators from Picosecond for use as calibration standards.

In the 1960s, says Jim Andrews, Picosecond's founder and chief technology officer, the tunnel diode was the only semiconductor you could use to get amplification at microwave frequencies. It was really tricky. But in a field of slow transistors, it was useful.

Where does the company buy its tunnel diodes today? Says Andrews, "I consider that extremely proprietary information." Andrews concedes that, though the tunnel diode isn't dead, it's pretty feeble. (It's been estimated that a few thousand germanium and silicon tunnel diodes are used per year in microwave detectors and picosecond pulse generators.)

Well, then, is the tunnel diode dead? Not quite. It is still breathing — just barely. But wait.

Several universities are conducting research on tunnel diodes. At the University of Delaware's department of electrical and computer engineering, associate professor Paul Berger leads a group that includes researchers from the Naval Research Laboratory, led by Phillip Thompson. Berger states that the tunnel diode is very useful in mixed-signal RF applications and digital CMOS to boost speed and reduce component count and, as a byproduct, slash power consumption. Tunnel diode/transistor circuits are a candidate for extending CMOS technology as the industry marches down Moore's Law. He cites as an example a comparator circuit at the heart of an A/D converter that might conventionally require 12 transistors and six Schottky diodes. A circuit with just two tunnel diodes and two transistors with one-sixth the area can provide the same arithmetic operation at faster speeds.

Newer silicon tunnel diodes, Berger points out, can be used in a wide array of circuits as switching elements, because they switch very rapidly from peak to valley currents in their negative-resistance region, and as storage devices. Tunnel-diode SRAM memory, using the bistability of a tunnel-diode pair, offers the refresh-free, fast storage of an SRAM cell, but with the footprint of a DRAM cell. Raytheon has already demonstrated a reduction in standby power over other SRAM cells by a factor of 3600.

Resonant tunneling diodes have been used at Raytheon Systems and elsewhere in very fast A/D converters with clock frequencies exceeding 10 GHz and switching time constants as low as 1.5 ps. And the diodes have proven themselves in a wide variety of other super-speed circuits. Says Berger, who has produced an array of tunnel diodes across a wafer, "This may prove to be the first viable technology for integrating silicon-based tunnel diodes with conventional CMOS semiconductors to boost circuit performance."

At the University of Illinois, Holonyak has graduate students experimenting with tunnel diodes. But he feels that tunnel-diode oscillators are not nearly as important as some people suggest. Holonyak just earned a patent using a tunneling contact in a laser because that arrangement enabled him to use sideways currents, allowing full light output from the top of the laser.

Says Holonyak, "Tunneling has its uses, but it isn't mainstream and it doesn't threaten anything like a good transistor."

But maybe, just maybe, it's coming back.

The Century of the Engineer: Misunderstood Milestones