From high-speed transistors to solid-state lasers, IEEE Medal of Honor recipient Herbert Kroemer's theories have led to a wealth of semiconductor applications
An unusual condition was imposed on Herbert Kroemer at the start of his research career 50 years ago. He was not allowed to touch anything in his workplace, the Telecommunications Laboratory of the German Postal Service. The fear was that this recent graduate in theoretical physics would break something. Far from constraining him, the restriction expanded his horizons.
With just pencil and paper, he began sketching out theories that would resonate across the entire world of semiconductor science. And that work would culminate in a Nobel Prize in Physics in 2000 and this year's IEEE Medal of Honor, the latter for "contributions to high-frequency transistors and hot-electron devices, especially heterostructure devices from heterostructure bipolar transistors to lasers, and their molecular beam epitaxy technology."
While his theories led to products that earned their manufacturers billions of dollars, none of the profits came to Kroemer. "That really doesn't bug me," he says, sitting in his small and sparsely decorated office on the Santa Barbara campus of the University of California, where he is now professor of electrical and computer engineering and materials.
IEEE Fellow Kroemer never tried to develop applications of his work--or even predict them. "I like lemmas," he told IEEE Spectrum, "and this one about applications is perhaps my most important message. It's called 'The futility of predicting applications,' and states: 'The principal applications of any sufficiently new and innovative technology always have been and will continue to be applications created by that new technology.' " So he doesn't begrudge others the fruits of his ideas.
"I've always called myself an opportunist," he says. "This is supposed to be a derogatory term, but I'm not one bit ashamed of accepting opportunities. In the scientific sense, I was an opportunist who was looking for challenging problems."
Too many lists
In high school in Germany, Kroemer played around with chemistry experiments but soon turned to physics. "I liked the beautiful logic of a structure with a relatively small number of fundamental principles from which you could draw far-reaching conclusions," he says. A university chemistry course that required rote memorization of lists and lists of chemical reactions destroyed any remaining interest in that science.
College was a breeze. He entered the University of Jena in East Germany in 1947, then left for West Germany the next year during the Berlin airlift and was accepted at the University of Göttingen. Four years later he received his Ph.D. for a theoretical dissertation on germanium transistors that discussed electron transport in high electrical fields. It broke little new ground, and he takes no particular pride in it. He explained some experiments, he says, but the explanation later proved completely wrong.
As he told Spectrum, his actual knowledge of the subject matter was rather limited. But what his research advisor really cared about was methodology. Does a student know how to tackle a problem with no background in the subject? And does he or she know how to acquire the knowledge needed? And that Kroemer knew how to do. To this day, his view of education is that accumulating methodology matters more than accumulating knowledge of subject matter.
"It was not until a number of years after working with him that I realized how unique this is," says William Frensley, a one-time graduate student of Kroemer's and now professor of electrical engineering at the University of Texas, Dallas. "Other students worked for professors who were specialists and became specialists in the same thing, whereas we said we have a problem, and we are going to master whatever techniques it takes to solve it."
In 1952, when Kroemer received his Ph.D., an academic career was out of the question. The lines of succession at existing German universities were long, and no new ones were being established. So he joined the Telecommunications Research Laboratory of the German Postal Service in Darmstadt.
This is less of a stretch than it seems. The postal service ran the telephone system and had a small semiconductor research group--some 10 scientists--in its telecommunications laboratory. That group hired Kroemer to answer any theoretical questions that arose, to give talks on any subject he thought relevant--and to keep his hands off the research equipment.
"I enjoyed this thoroughly," he recalls. For one, he had liked the role of teacher since high school, when his physics teacher asked him to prepare and deliver a lecture to the class. For another, being at the researchers' beck and call presented him with a wide variety of problems in diverse subjects.
In solving one of those problems, he went against the conventional wisdom of the time. Researchers were developing pn junctions of indium and germanium. They did this by depositing a layer of indium on a layer of germanium, then heating the structure to form the pn junction. Kroemer was trying to understand how exactly the junction formed.
Obviously the molten indium dissolved some of the germanium, and the belief was that it diffused into the germanium beyond the layer in which the germanium dissolved. But Kroemer concluded that the process was one of recrystallization--the heated indium dissolves some of the germanium, and then upon cooling the germanium precipitates out and recrystallizes, incorporating some of the indium atoms, which replace some of the germanium atoms in the lattice.
Name: Herbert Kroemer
Current job: professor of electrical and computer engineering and materials at the University of California, Santa Barbara
Date of birth: 25 August, 1928
Birthplace: Weimar, Germany
Height: 178 cm
Family: married to Marie Louise Kroemer, five children
Education: Ph.D. in physics, University of Göttingen, Germany, 1952
First electronics job: calibrating instruments at Siemens AG, Berlin (summer 1948)
Most unusual job: working nearly a kilometer underground in a coal mine (summer 1949).
Biggest surprise in career: winning the Nobel Prize for Physics in 2000
Patents: 10, more or less
Most recent books read: The Supreme Court, by William H. Rehnquist; The Politics of Excellence, by Robert Marc Friedman; Sylvia Nasar's A Beautiful Mind
Computer: Apple G4
Favorite Web site: Cond-mat (http://xxx.lanl.gov/archive/cond-mat)
Favorite saying: "If in discussing a semiconductor problem, you cannot draw an energy band diagram, then you don't know what you are talking about."
Favorite cities: Munich, Hamburg
Car: 14-year-old Toyota Celica
Languages spoken: English, German
Organizational memberships: IEEE, American Physical Society
Honors: IEEE Medal of Honor, Nobel in physics, IEEE Fellow award, eponymous asteroid
What he didn't know was that researchers in the United States, at General Electric Co. and RCA Corp., had simultaneously reached the same conclusion.
But what he did know was that to be at the research forefront, he needed to leave the German Postal Service and get to the United States. He started looking for a way to get there.
Researchers from other countries occasionally visited the lab in which he worked, curious about this small semiconductor research group. In 1953 one visitor was William Shockley, then at Bell Telephone Laboratories. "I spent about two hours with him," Kroemer said. "We were having a marvelous time. I told him about the work that I'd done for my Ph.D. dissertation, and about some of my ideas of how to make transistors fast by putting an electric field into the base. He seemed intrigued by that."
Kroemer asked him about coming to Bell Labs, but Shockley, as an official visitor, told Kroemer that he would have to go through official channels, starting with informing Postal Service management of his intentions to apply for a job in the United States. The young researcher was so grateful for the job he had at the Postal Service that he was "terribly squeamish about telling my management that I wanted to leave."
Later in 1953, the Darmstadt lab had another U.S. visitor: Ed Herold from RCA. Kroemer asked him whether RCA was working on npn transistors (back then pnp transistors dominated). Herold was guarded in his responses; but Kroemer guessed out loud what the RCA researchers were doing, what alloys they were using (lead-antimony), the percentage of the antimony, and the alloy temperatures. His guesses proved quite close to RCA's experiments, and the impressed Herold didn't hesitate to offer him a job. (All the same, it took a year for Kroemer to obtain a visa, even with RCA's help.)
At RCA in Princeton, N.J., Kroemer did theoretical research on an impurity diffusion process for building transistors. In the diffusion process, the doping of the base region was deliberately graded from a high value at the emitter to a lower value at the collector. Because this gradient introduced a built-in electric drift field into the base, the result was called a drift transistor. The first commercial product to come out of that research--the 2N247--had a high-frequency performance far beyond that of other commercially available transistors of its time. Its power gain cutoff frequency of 132 MHz made it suitable for use in FM radios.
While Kroemer was theorizing about how a drift field could make transistors switch faster, he had an idea about grading the basic semiconductor itself. If an alloy of two semiconductors replaced the single semiconductor, it could be given a continually varying composition to change its band gap, which is a measure of the amount of energy required to move an electron from a semiconductor's valence band to its conduction band. This varying band gap would be another way to introduce a drift field into the base, again in order to improve transistor frequency performance.
He had mentioned varying a material's band gap in a paper while still in Germany, but expanded the idea and in 1957 published two papers about it, one in the RCA Review, another in the Proceedings of the IEEE.
Theory into practice
While Kroemer trusted his theory, he didn't know how to build actual semiconductors using his principles. Building them would require either a base region consisting of a graded mix of different semiconductor materials with varying band gaps or else one material in the base but a different material in the emitter.
He tried to build a transistor with germanium-silicon alloy as the emitter on a germanium base. To this end, a gold-silicon blended mixture was alloyed onto germanium at 600 °C, hot enough for the melted mixture to begin eating up germanium, precipitating the germanium-silicon alloy emitter on cooling. Unfortunately, during the cooling, most of the devices cracked. "It was one of those technological blind alleys where you're not exactly embarrassed that you have tried it, but you're not surprised it didn't work," he says.
At the end of 1957, Kroemer decided to get out of transistor research. He had no interest in traditional transistors, and heterostructure transistors, with existing material technology, could not be built.
"I promised myself," he says, "that if a new technology for building heterostructures arose, I'd get back into it."
Kroemer left RCA in 1957 and returned to Germany; he, and more especially, his wife, was homesick. Becoming head of a semiconductor group at Philips Research Laboratory in Hamburg, he pushed for work on gallium arsenide, looking at what happens when you apply large electric fields to gallium arsenide semiconductors. "I thought GaAs was going to be an important material, so it was worthwhile studying it."
Kroemer feels he did little significant work at Philips and, since his wife quickly concluded she preferred the United States after all, in 1959 he went to Varian Associates (Palo Alto, Calif.), where he did a little research on tunnel diodes before turning to other problems.
Back in the heterostructure game
Then Kroemer's ideas about heterostructure devices, shelved for half a dozen years, came back to his attention with a vengeance.
It was March 1963. The previous summer, Kroemer and a Varian colleague, Sol Miller, had attended the Annual Device Research Conference, at which GaAs lasers had been introduced. Miller was interested and at Varian's weekly colloquium, he gave a talk about the new lasers. Though scientifically fascinating, he said, the devices could only work at very low temperatures and only for very short pulses, and so would never be truly practical. Asked why, Miller explained that the problem was the lack of charge-carrier confinement: at normal temperatures, electrons would diffuse out of one side of the device as quickly as they were supplied from the other side, as would the holes; therefore the electron-hole pair concentration would never become high enough to cause laser action by stimulated emission. Low temperatures suppressed the effect, but only for brief periods of time.
Kroemer disagreed. Based on his work in heterostructures, the solution, to him, seemed obvious--you just vary the device's band gap, putting a narrower gap in the center and a wider gap in the outer regions, so that the electrons and holes would concentrate in the center [see "Heterostructures Explained"]. "My reaction was instantaneous," he told Spectrum. "The moment somebody told me about the problem, it snapped."
He wrote up his idea as a paper and submitted it to Applied Physics Letters, where it was rejected. Rather than fighting the rejection, he was persuaded to submit it to the Proceedings of the IEEE. There it was accepted, but drew little attention. He also filed for a patent on the technology; issued in 1967, it expired in 1985.
Kroemer wanted to start working on the creation of room- temperature lasers at once, but his superiors at Varian told him that such a device would never have any applications.
"This is the classic mistake--judging something not by what applications it might create, but by how it could fit into applications we've already thought of," Kroemer says. The applications it was useful for turned out to include fiber-optic communications, CD and DVD players, LED traffic lights, and laser pointers--none of which were around at the time.
Though Kroemer wasn't pleased by Varian's decision, the Gunn effect, which had just been discovered, interested him. This is a phenomenon in which microwave oscillations are produced when a certain voltage is applied to opposite faces of a semiconductor. For the next decade and more, Kroemer explored theories of why this occurred, three of those years at Varian, two at Fairchild Semiconductor Corp. (Palo Alto, Calif.), and nearly eight at the University of Colorado in Boulder.
Halls of academia
Kroemer was happy to move from industry to academia. Things at Fairchild had not gone well, because the company was dedicated to silicon technology and Kroemer's interests had long been elsewhere. He looked forward to the research freedom and also to teaching.
But he became dissatisfied. "We had hoped to set up a good solid-state engineering graduate program at Boulder, " he says. During the Vietnam War, many students went on to graduate school to reduce their chances of getting drafted. Stanford University typically recruited the academically top 5 percent of graduate students interested in solid-state research, and Boulder drew on the next 5 percent, who were still extremely good. But when graduate enrollments fell after the war's end, that source dried up. "Our ambitious graduate program would not fly--it was clear to me that I would be professionally dead if I stayed there," Kroemer recalls.
Word went out that he was open to a change, and in the fall of 1975, the University of California at Santa Barbara, in the person of Edward Stear, then head of its electrical engineering and computer sciences department, came calling. Santa Barbara at the time didn't have a very good academic reputation; what it did have was a well-equipped semiconductor device teaching laboratory.
"So, Herb, you know about our laboratory," Stear opened. "What would you do with it?"
Kroemer momentarily forgot that this visit was actually a job interview. "Sure as hell not what you're doing!"
"It was a rather unfriendly and hostile discussion, and Stear eventually snapped, 'Shut up,' " Kroemer recalls. He figured he had blown any chance of being hired. But then Stear told him, "I'm looking for someone to rock the boat; it looks like you're my man."
Kroemer, Stear tells Spectrum, "speaks very directly. He is honest, but can be sharp with people, too. He is intense and demanding. He can be a difficult person at times to work with, but people have ended up loving him." In any case, Stear knew that Kroemer could build the kind of program that Santa Barbara needed, and Kroemer was hired.
By the sea
Kroemer left for Santa Barbara in the summer of 1976. He had persuaded Stear not to compete with Stanford, Berkeley, and other top engineering schools in silicon technology, but instead to focus on compound semiconductors such as GaAs. He gave Santa Barbara even odds for making an impact in that technology.
"You want to be first-rate or absent," Kroemer says.
Kroemer convinced a few former colleagues that they should join him at Santa Barbara, and he also convinced the U.S. Army it should buy him a molecular beam epitaxy machine. He said at the time that he wanted it for making transistors with a gallium phosphide emitter on a silicon base, a crazy project if there ever was one. It was not enough to put Santa Barbara's engineering school on the map.
In the mid-1980s, however, the chancellor of the university, Bob Huttenback, decided to put all available money into improving the College of Engineering. A new dean was hired, and 15 faculty were added. "We raided Bell Labs," Kroemer recalls. "Today we have one of the best materials departments in the country--and we still don't have any silicon technology."
At 73, Kroemer remains a full-time member of the faculty. One problem he is working on concerns the influence of high electric fields on electron transport in semiconductor superlattices (alternating thin layers of two or more materials with different band gaps but similar crystal structures and lattice constants). More specifically, he is focused on a concept, called a Bloch oscillator, which can in theory generate oscillations up into the terahertz range, potentially opening up that frequency range for numerous applications. So far, it has never been satisfactorily demonstrated as a continuously running device. "I have some ideas, which may or may not be correct, of what to do about it," Kroemer tells Spectrum.
He is also looking at the phenomenon of induced superconductivity in semiconductors, created when superconducting materials are deposited on semiconductors and operated at low temperatures.
Tuesday morning, 3 a.m.
Of all the honors Kroemer has received over the years, the strangest was the naming of an asteroid after him. (Asteroid Kroemer orbits between Mars and Jupiter.) One honor that he thought beyond the grasp of a physicist who dealt in such a down-to-earth area as semiconductors (compared to those who grapple with invisible particles) was the Nobel Prize.
"Oh, my name had been mentioned over the years," Kroemer told Spectrum. But the Nobel Prize is almost invariably awarded for fundamental discoveries, not for applied research, and so I never believed the rumors."
The rumors grew stronger in 1996, when Kroemer was invited to give a talk at a Nobel symposium. "I still didn't catch on," he said. "I looked around at the attendees and saw Horst Stormer, and thought he was the most likely candidate of the group. When he received the prize in 1998, I was enthusiastic and didn't envy him at all--after all, my work was applied." (Stormer and two colleagues received the Nobel Prize for discovering that electrons acting together in strong magnetic fields can form new types of particles with charges that are fractions of the electronic charge.)
Although Kroemer never believed the Nobel would come to him, he did continue to pay attention to it. On 9 October 2000, the Nobel Prize in Physics was to be announced the next day. He went to bed that evening thinking, "Wouldn't it be funny if I would get a 3 a.m. phone call? But then I said to myself, Stop being silly, go to sleep!" (Noon in Stockholm, when Nobel announcements are typically made, is 3 a.m. in California.)
But when the phone did ring shortly before 3 a.m., his first response was panic--were his children all right? Had something happened to his grandson? His wife answered, and passed him the phone, saying "It's Stockholm."
"If my life depended on it, I could not reconstruct the next two or three sentences," Kroemer says. Then the caller put a friend of Kroemer's on the phone, to reassure him that it was not a joke, warning him that he had about 15 minutes before the public announcement was made and the media circus started.
"At that point all hell did break loose and the phone was ringing off the hook, starting with German news agencies, since I'm German and it was already midday there. I literally couldn't put the phone down."
The Nobel Prize in Physics that year was shared by three people--Jack S. Kilby, also an IEEE Fellow, for his part in developing the IC, and Kroemer and IEEE member Zhores I. Alferov, for "developing semiconductor heterostructures used in high-speed- and opto-electronics." Alferov, working in Russia, had made similar discoveries in parallel with, but separately from, Kroemer; the two first met in 1972 and have since become friends, even though they are, in a sense, competitors.
Kroemer's Nobel citation emphasizes the general principle of the heterostructure, not the individual devices. And that suits him just fine, because he has routinely deferred the question of applications. "Certainly, when I thought of the heterostructure laser, I did not intend to invent compact disc players," he says. "I could not have anticipated the tremendous impact of fiber-optic communications. I really didn't give a damn about what the uses were."
"That's not what I do. The person who comes up with applications thinks differently than the scientist who lays the foundation."
And Kroemer laid one fine foundation.
To Probe Further
View Herbert Kroemer delivering his Nobel Lecture at Stockholm University and hear an interview with the 2000 Nobel Laureates by checking out http://www.nobel.se/physics/laureates/2000/kroemer-video.html.
Two seminal papers by Kroemer on heterostructure devices are "A Proposed Class of Heterojunction Lasers," Proceedings of the IEEE, Vol. 51, 1963, pp. 1782-83, and "Heterostructure Bipolar Transistors and Integrated Circuits," Proceedings of the IEEE, Vol. 70, 1982, pp. 13-25.