At the end of WWII I studied and learned vacuum-tube electronics in Urbana (U of I) with GI-bill WWII veterans. I was already familiar with the electronics of free-electron devices (vacuum tubes), but not with the new dual-conduction (bipolar) minus-plus (−/+) electron-hole (n-p or p-n) semiconductor electronics, the transistor, that John Bardeen brought from Bell Labs to Urbana in 1951---- the electronics of the active device and conducting charge all contained and manipulated inside the solid. Why put the charge, the electron, in vacuum if not necessary? Furthermore, is it possible a better device and electronics geometry can be rendered totally in the solid than in vacuum? The transistor was revolutionary. It was to lead to a new electronics “geometry” (p-n, ΔEg, etc.) fully embedded in the semiconductor, eventually patterned in many forms and shrunk ultra-small, and then still smaller, approaching now atomic size, today’s integrated circuit. The semiconductor became and remains the unique substance of electronics. Because of John Bardeen and what I was learning from him (1951–1954), I had reason and the opportunity to switch from vacuum tube research (a microwave multipactor) to transistor and semiconductor research (now 60 years).
The transistor changed electronics, and the world. It taught us about the hole, in the beginning just a quantum construct, and made it real. It taught us (Bardeen’s statement often repeated, but not often enough, and for the last time at an NHK Urbana office interview, June, 1990) that current, as revealed by the transistor the essential base current (IB) could generate excess e-h pairs, a non-equilibrium population. It taught us about excess e-h pairs and recombination, a vital part of the transistor, and as a result in time how to generate light electronically (recombination radiation), in fact, how to realize an “ultimate lamp,” basically a p-n junction (Am. J. Phys. 68, 864 (2000)). After the transistor, in fact, because of the transistor, solar cell and LED research made sense. Earlier it was “Edisonian”, trial and error, guessing. The transistor changed everything. The transistor, based so solidly on quantum science, marked the beginning of the modern era in electronics. The semiconductor came alive. It suddenly (1947) had purpose. Electronics as known today began with the transistor, including the diode laser and LED. Earlier work didn’t matter. It didn’t lead step-by-step (causally) and connect scientifically and technically to today’s LED. It couldn’t, not before quantum science revealed the semiconductor had an energy gap.
But why leave Si and transistor research (transistors, p-n-p-n switches, thyristors, IC’s, etc.) and turn to III-V’s? Why speculate and risk failure with more complicated, less available man-made materials? Bob Noyce (Philco, Shockley, Fairchild, Intel) and I argued about this at the 1962 IRE (IEEE) DRC, he that Si offered still more, and I that it was time to look at III-Vs, at new devices, LED’s, heterojunctions, etc.—and who cared if a less knowledgeable boss (my boss in Syracuse) told me if I couldn’t get funding support from the Air Force and the manager of GE rectifier department, then “there’s the road”? In other words, work on Si or else go elsewhere, leave or get fired.
There is more to this, a long story, but cut short, everything changed after the 1962 DRC and Rediker’s Lincoln Lab group’s report (Keyes and Quist) that they had broadcast an infra-red (IR) signal 100’s of yards from a simple Zn-diffused GaAs diode, meaning that a GaAs p-n junction was a good recombination radiation generator. How good, good enough perhaps to become a laser? Could a semiconductor, in fact, be a laser? Most speculation, mostly wrong, was that somehow “discreteness” must be introduced into the GaAs p-n junction to emulate, somehow copy the narrow spectral line behavior of the 1960–1962 Maiman-style of laser. A small number of us were not misled by these speculations, knowing direct-gap band-to-band recombination in an atomically dense continuous high gain system, a semiconductor, should not be compromised with discrete deep levels (to somehow make it behave like a dilute system laser, narrow line width but low gain). All that was needed was some form of augmentation to help the autocatalytic behavior, the spatial self-adjusting and self-tuning in energy, of the electron-hole-photon recombination interaction in a direct-gap semiconductor, the GaAs class of material. Maybe all that was needed was higher level excitation of the semiconductor and the help of a cavity, a resonator as generally true of oscillators, something I knew from my student days.
While others were concerned with GaAs I was concerned with direct-gap GaAs1–xPx (x ≲ 0.45), which for larger energy gaps I had learned to grow (1960–1962) and make into p-n junctions. I wanted light I could see, not GaAs infra-red (IR) radiation. Although “correct” that GaAs1–xPx was stochastic in As-P arrangement on column V lattice sites, randomness, as such, did not make the crystal “rough and lumpy”. Random As-P arrangement was not a crystal defect in the usual sense. I knew this because my red GaAsP diodes were well behaved and electrically as good in performance as GaAs diodes. When our GE Syracuse crystal growers, laughing and swearing at me and claiming because I was not a chemist and didn’t know better, and got away with using presumably an impossible method to make GaAs1–xPx (x ≲ 0.45), I returned their swears in kind (maybe stronger), and made bright red III-V alloy diode light emitters. I knew enough about GaAs1–xPx to know I could compete with GaAs. Also I beat back arguments of competing big-lab managers convinced (so wrong!) that LEDs would eventually all be made in indirect-gap GaP (2.26 eV energy gap, green).
When DRC ended (July 1962), Hall didn’t know I planned to extend (direct-gap) diode performance and make a red GaAs1–xPx (x ≲ 0.45) laser, nor did I know he planned to make an IR GaAs diode laser. We were busy at DRC and did not discuss the possibility (questions) of a semiconductor laser: Big project vs. little project? The question (specious) of “discreteness”? How to overcome the broad spectral width of e-h recombination radiation? The current path itself across the gap the light generator? The goal of an “ultimate lamp”, ~100% conversion of electrical to optical energy, realistic or not? What else? Hall didn’t know what I was thinking and doing, and I didn’t know what he was thinking and doing till I talked with him in Aug (1962). I was going to use an external cavity on my III-V red alloy diode laser, and Hall had decided to use his diode crystal itself as the cavity by polishing the crystal edges into a Fabry-Perot configuration. I found out later he was a telescope builder in the past and liked polishing. I immediately suggested (GaAs) crystal cleaving to Hall (and to a lackadaisical GE patent attorney), and at home tried unsuccessfully to cleave my large-grain “poly” GaAs1–xPx. Who knows how much time (and crystal) I wasted because of the random orientation of my GaAs1–xPx? In the meantime Hall was successful with polishing, and could see his diode laser diffraction pattern with a “snooperscope”, which I had overlooked until Hall’s boss (Apker) called me one fall day to tell me of their success and urged me to hurry-up and polish the edges of my diodes into the form of a Fabry-Perot resonator. Apker earlier on a GE visit to Syracuse had seen my red alloy diodes and knew I must be close to operating a visible-spectrum III-V
GaAs1–xPx alloy laser. We did not know who the competition might be and felt the need to hurry. I immediately devised a polishing method, as it happened simpler than Hall’s but as good, and immediately had red III-V alloy GaAs1–xPx diode lasers.
Besides Hall proving a semiconductor (GaAs) could operate as a diode laser, I had gone further showing with my “homemade” crystal, not “off-the-shelf” GaAs, that the visible-spectrum (red) III-V alloy GaAs1–xPx could operate as a diode laser. This proved the important point that III-V alloys were viable device materials, no longer to be dismissed, in fact, necessary for many purposes. Both forms of diode lasers (GaAs and GaAsP 1962) established at once, because of their high external photon extraction efficiency and thus easily measurable external quantum efficiency (→ 100 %), that the internal direct-gap band-to-band recombination radiation efficiency was equally as high (or higher). It was no longer a DRC guess but a fact. The 1962 GaAs1–xPx laser proved III-V alloys were electrically and optically smooth enough to be useful visible-spectrum p-n “ultimate lamp” diode lasers and LEDs, and that the future of indirect-gap materials, say, GaP, was limited, maybe just useful for substrate purposes. It should have been clear that the III-V alloy was needed not just for bandgap and wavelength “tunability” but also for index steps for photon as well as carrier confinement. Also, since the composition, x, of GaAs1–xPx could be varied, we quickly established in crossing the direct-indirect transition,
x ~ 0.45, that indirect-gap crystal was a poor choice for light emitters. It no longer made sense to keep emphasizing work on GaP (indirect gap) as an LED material, and nonsense for proponents of GaP to be arguing with us for a long time (as happened) that alloys were hopeless and a poor choice for devices.
Our 1962 lasers pointed the direction, the path to the p-n “ultimate lamp” (laser or LED). And after Russell Dupuis (1977 DRC) proved the point of his powerful (electronically controlled) cold-wall MOCVD crystal growth reactor and success in universal Al-Ga substitution in III-V alloys, and then helped us make the quantum well useful in all III-V alloy lasers and LEDs, the future of the “ultimate lamp” was assured, indeed, is happening. Everything in science and technology does not just erupt, happen suddenly, happen by political decree (even with “politiking” among scientists), but might take 50 years, even if properly in sight maybe from the beginning, but still might take more time than 50 years. We have known about a p-n “ultimate lamp” for 50 years and now it is happening. The semiconductor, transistor, and p-n “ultimate lamp” LED are here to stay.