|
The Diode Laser—the First Thirty Days Forty Years Ago |
|
Original source: A little over forty years ago, on Sunday, September 16, 1962, Dr. Gunther Fenner, a member of the team headed by Dr. Robert N. Hall at the General Electric Research Development Center in Schenectady, New York operated the first semiconductor diode lasers [1], [2], [3]. Within about thirty days, workers in three other laboratories in the USA soon independently demonstrated their own versions of the injection laser. Dr. Nick Holonyak, Jr. [4], [5], [6], at General Electric’s Syracuse, New York facility; Dr. Marshall I. Nathan [7], [8], at IBM Research Laboratory, Yorktown Heights, New York, and Dr. Robert Rediker [9], [10], at MIT Lincoln Laboratory, Lexington Massachusetts, led the efforts of these three other groups that also succeeded in making a semiconductor laser in late 1962 [11]. While three of these early p-n junction lasers were fabricated by Zn diffusion into conventional (and commercially available) n-type GaAs “bulk” crystals, one of these “first” laser diodes (Holonyak’s) was created from a small piece of single-crystal ternary GaAsP alloy material grown by vapor-phase transport, making it the first alloy compound semiconductor device ever demonstrated. It was also to become the first alloy semiconductor device to become of commercial importance. From this humble beginning, the vast materials and technologies of modern compound semiconductor alloy devices have emerged. What are GaAs p-n Junctions Good For? While Ge and Si were well-known semiconductors by the late 1950’s, the III-V compound semiconductors (sometimes referred to at that time as “intermetallic compounds”), were not well studied and had no obvious unique application. By 1960, Si had become the dominant semiconductor due to its large bandgap relative to Ge and to the beneficial features of its native oxide, SiO2 discovered by Carl Frosch in 1955. It was not clear exactly what benefits could be obtained by replacing Si by GaAs since GaAs had no stable native oxide and was much more difficult to make in high-purity form. Consequently, in as late as 1958, few device researchers considered that GaAs was worth much effort. In fact, it had been only recently, in 1952, that Heinrich Welker, working at Siemens in West Germany, had identified that GaAs was, in fact, a semiconductor [12].
While the basic elements of the light-emitting properties
of GaAs p-n junctions were known for several years prior
to this 1962 report, some in attendance at the DRC complained that these
high-efficiency electroluminescence results violated the Second Law
of Thermodynamics! [10]. What made these announcements so electrifying
to many in the audience attending this session of the SSDRC was that
it suddenly became clear that a semiconductor diode could be a very
efficient generator of photons, perhaps the most efficient “converter”
of electrical energy into optical energy ever demonstrated [16]. This
revelation was an important motivation for some of those working in
this field to pursue their dreams of making a semiconductor laser. The
race for the semiconductor injection laser was on.
Of course, in some respects, the Lincoln Labs group had a sizable lead in this work since they had a relatively large research group already organized and experienced in the study of GaAs diodes. Earlier in 1961, and even before, various other groups, including the IBM Research Labs, had considered the concept of a semiconductor laser. The IBM group even had a US Army-sponsored research contract to make such a device [8], [17]. Research groups in the United Kingdom also had joined the chase for a semiconductor laser with well-organized GaAs p-n junction research activities at Royal Signals and Radar Establishment under Cyril Hilsum. In the Soviet Union, Nikolay G. Basov and co-workers in the Lebedev Institute in Moscow and D. N. Nasledov and co-workers at The Ioffe Physico-Technical Institute in Leningrad were also considering how to achieve population inversion in a semiconductor [18], [19]. In France, Pierre Aigrain at École Normale Suprieure had proposed that laser operation of a semiconductor could occur and was reportedly (in 1961) planning to visit the USA with a working semiconductor laser in his pocket! [20] Also in 1961, Maurice Bernard and Guillaume Duraffourg, working at CNET in France, had published a paper analyzing the possibility of laser operation in semiconductors [21]. How Do You Make a Semiconductor Laser?
Of course, the basic quantum mechanics behind the
radiative recombination and laser operation of chromium ions in ruby
and the earlier demonstration of the microwave amplification by the
emission of radiation—i.e., the MASER—using ammonia molecules,
were fundamentally different from that occurring in a direct-bandgap
semiconductor. There was no clear path to take what was known about
stimulated emission as demonstrated for discrete atomic states like
those found in ruby, and to translate this to the broad density of states
found in the band structure of a typical semiconductor. In fact, some
approaches being explored in 1962 involved indirect semiconductors,
e.g., Si and Ge, as well as putting rare-earth atoms into the semiconductor
“host” crystal, e.g., U or Nd, to create “atomic-like”
energy transitions (like the Cr+ ions in a ruby crystal) within the
band structure of the semiconductor crystal (an approach that has not
borne fruit to this date).
One primary practical problem faced by everyone interested in making a diode laser was how to determine that it was in fact lasing. While it may seem somewhat difficult to understand with the benefit of hindsight, many of those thinking about diode lasers in 1960-62 were not exactly certain as to what specifically to look for to determine the laser “threshold” or that lasing was occurring. Since 1960, it was obvious that coherent light should produce some type of interference pattern but this was not perhaps uniquely created by stimulated emission originating from a small aperture typical of a GaAs diode. As Robert Hall has written: “It seems strange now, but at that time, one of our big uncertainties was to know what to look for as evidence that the diode was lasing” [3].
For Hall, who already had extensive experience with
GaAs alloy junctions, tunnel diodes, and light-emitting diodes, the
project to make a laser diode was an extension of his prior research
work. It also coupled well with his optical experience in his earlier
youthful hobbyist efforts to build telescopes and to polish lenses and
mirrors [3]. Hall’s laser project team included Dick Carlson, Gunther
Fenner, Jack Kingsley, and Ted Soltys. Whereas other groups thinking
about semiconductor lasers had proposed to use a macroscopic “external
cavity” into which a GaAs diode was placed, Hall decided to polish
parallel faces onto his GaAs diodes so that the Fabry- Perot optical
cavity geometry was built into the device. This approach was not universally
applied and, in fact, the importance of optical feedback into the diode
“active region” was not fully appreciated by many workers.
Hall’s team operated their first successful GaAs laser diodes under
pulsed conditions at 77K on September 16, 1962 [1]. A schematic diagram
of Hall’s early concept for an injection laser is shown in Figure
1. The first verification of laser operation was made through the observation
of the near and far-field interference patterns using an infrared up
converting “snooper scope” which was being used in Hall’s
lab to study the emission from such infrared light emitters.
As noted above, IBM also had an early effort in the
development of the diode laser. In 1961, Rolf Landauer, recently appointed
a Department Director at the relatively new IBM Research Lab at Yorktown
Heights, NY, initiated the IBM team effort in the development of a GaAs
laser diode [8]. Like Hall’s Group and the Lincoln Labs Group,
the IBM team concentrated on Zn- diffused GaAs p-n junction diodes
with which they had previous experience in the development of GaAs bipolar
transistors. The team included Rick Dill, Walter Dumke, Bob Keyes (not
the same person as was on the Lincoln Labs team), Gordon Lasher, John
Marinace, Marshall Nathan, Michael Stern, and others [8]. The IBM GaAs
laser team was also inspired to accelerate their research effort by
the Lincoln Labs presentations at the 1962 DRC, although no one from
IBM was present at the DRC presentation. They had, however, heard of
the results from an article appearing in The New York Times on
July 10, 1962, and they had also heard in a seminar given at IBM in
January 1962 by Sumner Mayburg of the GTE Sylvania Laboratory about
his recent observations of high-efficiency radiative recombination from
GaAs p-n junction diodes [26]. Thus, Landauer, Dumke, and Keyes
had been thinking about the semiconductor laser problem for some time
prior to the June 1962 SSDRC conference, and Dumke in particular, had
calculated the requirements for population inversion in a semiconductor
[27]. In the first sentence of Dumke’s paper (submitted April 3,
1962), he stated “Since the initial operation of the ruby maser,
there has been considerable speculation concerning the possibility of
observing maser action in semiconductors such as Ge and Si.” After
rejecting the possibility of laser operation in the indirect semiconductor
Ge due to excessive free-carrier absorption, near the end of
this paper, he concluded “At present, it is not clear whether or
not one would obtain anything like typical maser action from a device
utilizing direct transitions as in GaAs” [27].
After receiving the news about the 1962 SSDRC reports
on efficient GaAs diodes, the IBM team was spurred to an increased level
of activity. Several of the newly “expanded” laser team dedicated
their efforts to the making diodes and analyzing their performance.
Gordon Lasher considered what theoretically would be observed in the
stimulated emission spectrum of a GaAs diode. Marshall Nathan concentrated
on studying the photoluminescence and electroluminescence from various
GaAs structures operated under pulsed conditions at low temperatures.
While the IBM team realized that spectral line narrowing should occur
for a diode laser, and that the light output versus current should be
superlinear, because the IBM team could not figure out how to provide
for proper “cavity mirrors” for the diodes, no provision was
made for optical feedback in this early work [8]. During Nathan’s
studies of these devices on September 29, 1962, he observed narrowing
of the electroluminescence to a FWHM ~3nm—and a few days later
he found that the line width measured for this first successful diode
was the limit of the spectrometer ~0.2nm. In fact, the GaAs diodes showing
“stimulated emission” first reported by the IBM group did
not employ an optical cavity and no attempt to provide optical feedback
was made [7], [8]. The IBM paper describing their first stimulated emission
results was submitted to Applied Physics Letters (received on
October 6, 1962) and was published on November 1, 1962 [7]. Later, Rick
Dill and Dick Rutz developed a cleaving process for formation of the
optical cavity by cleaving all four sides of the laser diode [28].
The Lincoln Labs work on GaAs was initiated in 1958
to study the possibility of making III-V high-speed microwave devices
[10], [29]. The III-V program was driven by the vision of Robert Rediker
who championed the study of GaAs when most other groups were concentrating
on Si. Rediker had visited Prof. Henrich Welker in Erlangen, Germany
in 1958 to learn more about GaAs and related materials. Welker, while
working at Siemens in Germany, was the first person in the West to identify
the “intermetallic” III-V materials as semiconductors. The
early Lincoln Labs GaAs program was quite small and involved only two
technical staff members and one technician. This team developed a diffusion
technique for producing p-n junctions. The Lincoln laser effort
in fact evolved from work in Rediker’s Group on the comparison
of the electroluminescence characteristics of GaAs diffused and alloy
diodes. Keyes was invited to join the team since “he owned the
spectrometer” [30]. As noted above, the work of Keyes and Quist
on the study of the recombination radiation from diffused GaAs p-n
junctions led to the observation and report of high internal quantum
efficiencies, which “sparked” the post-SSDRC efforts at many
research labs to develop a semiconductor laser. The Lincoln Lab semiconductor
laser effort presumably benefited in various ways from earlier theoretical
papers on semiconductor lasers written by B. Lax, H. Zeiger, and W.
Krag in 1959 [31], [32]. As noted, above, this work was devoted to development
of efficient emitters and included several key contributors to the theory
and the experimental characterization of the recombination radiation
from GaAs p-n junctions. In 1962, this effort centered in Rediker’s
group included Bob Keyes, Bill Krag, Ted Quist, Al McWhorter, Herb Zeiger,
and others. The search for laser operation from GaAs diodes fabricated
at Lincoln Labs was successful on Friday October 12, 1962 [33], with
laser operation first confirmed by the observation of “filaments”
in the near-field pattern of a GaAs diode when examined with an infrared
converter [9]. The Lincoln Labs device employed a Fabry-Perot cavity
having polished parallel facets, as suggested earlier by Zeiger [34].
The paper describing these laser results was submitted to the Editor
of Appl. Phys. Lett. on Oct. 23, 1962 and it was published (after changes)
on December 1, 1962. A photograph of one of the early Lincoln Labs GaAs
laser diode is shown in Figure 3.
Do Alloys Work? It is interesting to note that at this time (1960-62), many workers in the semiconductor field were convinced that an alloy was not worth pursuing and, if such materials were successfully produced, that random alloy “disorder” would render them useless due to the stochastic nature of the distribution of atoms in the lattice [35]. Furthermore, there were various ideas about how to produce such alloys. For example, one idea (a poor idea which has not succeeded even to date) involved the diffusion of P into GaAs to create GaAsP layers. Epitaxial growth of semiconductors was still a very new concept and while the creation of alloyed p-n junctions using metal “preforms” was a well-known process, this process lacked the control necessary to create true alloy semiconductors.
Holonyak, like Hall and others, was excited by the possibilities described in the Lincoln Labs paper at the July 1962 DRC. Working with GaAsP diodes in his Syracuse lab, Holonyak believed he had an important advantage over others working on the diode laser problem using GaAs—he could SEE the light coming from his diodes and could therefore see the near and far- field patterns. This meant that he could determine quickly whether the diodes were lasing or not by looking at the diode spatial emission characteristics and also for the same well known “laser speckle” produced by the visible ruby laser [36]. At the SSDRC, Holonyak thought about forming an external cavity for the optical feedback. His ideas on cavity formation changed after discussions with Hall. Holonyak and Hall had a joint contract with the Air Force for semiconductor device research and under this funding agreement, they shared some ideas, but they were geographically separated and were still “competitors” and operated essentially independently as far as making the first laser is concerned [37], [38]. After the 1962 SSDRC, Holonyak and Hall did discuss the problem of the formation Fabry-Perot cavities and they also shared their ideas on how to make a laser cavity as early as August 31, 1962 [5]. Hall proposed using the semiconductor diode faces as mirrors. Hall’s approach was to lap and polish the diode facets and Holonyak decided to try to cleave the Fabry-Perot mirrors for his optical cavity (an idea apparently initially overlooked by other groups trying to make a laser diode). This approach, while subsequently nearly universally used to create diode laser facets, actually delayed Holonyak’s first GaAsP laser demonstration due to the difficulty he experienced in identifying the appropriate cleavage planes in his small “bulk” GaAsP crystallites. (Unfortunately, the GE lawyers decided not to file for a patent on this idea but the IBM team did ultimately file on this concept!)
Sometime after the successful operation of the first
semiconductor lasers at the GE Schenectady Lab, Hall’s boss, Roy
Apker, called Holonyak at Syracuse to tell him that Hall’s group
was running a diode laser and suggested that he should stop trying to
cleave facets. Holonyak then decided to polish facets and quickly developed
his own “home-made” process to do just that. This process
was successful on the first try in producing a “good” set
of diodes on October 9 that were tested on October 10, 1962, using the
diode testing apparatus at the Schenectady laboratory since Hall’s
lab was better equipped for this study [35]. The first of Holonyak’s
GaAsP diodes that were tested operated as lasers under pulsed conditions
at 77K. Holonyak, thinking that the game was over since Hall had beaten
him to a laser, did not think it was urgent to write up his results.
Consequently, it was days later in October that Holonyak finally got
around to writing up the GaAsP laser results and to submit them to a
new journal, Applied Physics Letters [4]. These results (received
at Appl. Phys. Lett. on October 17, 1962) demonstrating a compact and
efficient source of visible coherent light, would ultimately be the
basis of the first commercially available visible semiconductor light
emitters and was the genesis of the now ubiquitous “light-emitting
diode” (LED) which is almost universally understood to imply a
device that creates photons “visible to the human eye” based
upon minority carrier injection and radiative recombination of excess
carriers. The demonstration of an alloy semiconductor laser at
essentially the same time as the GaAs laser provided dramatic and ample
proof that alloys were good for something after all. Sometime after
the first demonstration of the GaAsP laser, Holonyak arranged for a
photograph to be taken of the emission from one of his diodes. This
photo, reproduced in Figure 4 was the first photograph of a diode laser
made from it’s own light. Who Wants to Buy Lasers or LEDs? Not long after the first laser demonstrations, on
November 28, 1962, GE held a semiconductor laser conference at Schenectady
for invited representatives of the Department of Defense. Some in GE
had already recognized that this was a very “disruptive” technology
that could have important defense and commercial applications. In addition,
before the end of 1962, GE offered both GaAs and GaAsP lasers for sale,
becoming the first company to offer such devices commercially [39].
Since these devices only operated pulsed at low temperatures, they were
obviously useful only for research purposes or some special defense
applications. The price for one of Hall’s IR-emitting GaAs laser
diodes was initially $1,600—a price somewhat arbitrarily set by
the GE marketing group at ten times that of a currently available Texas
Instruments IR incoherent “LED”. This price was later reduced
to $800 [39]. The “visible red” GaAsP laser diode price was
initially priced at $3,200 each and then reduced to $1,600 [39]. This
was based upon the fact that GE’s marketing managers decided that
Holonyak’s visible laser was “twice as valuable” as Hall’s
IR-emitting laser device. Incoherent GaAsP visible LEDs were also offered
for sale. Interestingly, GaAsP red-emitting LEDs are still sold 40 years
later. While some researchers did not appreciate the fundamental distinction
between the “infrared” GaAs lasers and the “visible”
GaAsP lasers, Holonyak was acutely aware of the significance of visible-spectrum
LEDs and lasers and this feeling was shared by others working in the
field [40]. These first visible injection lasers made of the ternary
alloy GaAsP foreshadowed the future when virtually all semiconductor
lasers would employ alloy materials. In addition, the GaAsP first compound
semiconductor alloy light emitters were the earliest progenitors of
the now-ubiquitous LED and they have spawned related devices in a variety
of III-V ternary and quaternary materials systems, leading to the continued
development of an “ultimate light source”—the high-efficiency
injection luminescence source available today in the form of advanced
“high-brightness” LED products in the InAlGaP and InAlGaN
alloy materials systems.
Today, over forty years after the first demonstration
of the injection laser in September, 1962, and the first demonstration
of a compound semiconductor alloy device in October, 1962, advances
in materials and device design has made the injection laser a fundamental
and essential device in many important systems. For example, high-performance
red-emitting InAlGaP heterojunction quantum-well DVD lasers are currently
primarily grown by metalorganic chemical vapor deposition and cost less
than $1US in packaged form! It is also interesting to note that GaAsP
LEDs, closely related to those first diffused GaAsP diodes, are still
produced commercially. Furthermore, the diode laser has become the dominant
form in the commercial laser market with millions being produced each
year! We enjoy multitudinous benefits of the early research and development
of semiconductor diode lasers, multi-element semiconductor alloys and
their derivative heterojunctions and the future seems to hold an even
greater variety of applications for these devices and their progeny.
References: [1] Robert N. Hall, private communication, October
2002.
|
