One could easily
imagine an efficient
GaAs laser that
could couple a
milliwatt of optical
power into a fiber
with a core diameter
of about 50 µm. Thus
was defined the first
generation of fiber-optic systems.
Important clues to improvements came
in early 1973 from an experiment in
which “windows” were fabricated on the
substrate side of stripe-geometry lasers
in such a way that spontaneous emission (and scattered stimulated emission
if present) from the stripe region of the
laser could be observed with an infrared
optical microscope. This important work,
published in Proceedings of the IEEE in
1973 (61, 1042), focused attention on the
elimination of so-called dark-line defects
(DLDs), which grew in a laser’s active
region during operation. These DLDs
were determined to be the principal failure mechanism in devices that stopped
working in the first 100 hours or so.
The published paper galvanized a large
technical community. It correctly stated
that: “The combination of low-strain processes and extreme
cleanliness in materials growth should provide a dramatic
increase in laser life.” It seemed that everyone in the world with
an electron microscope then decided to investigate this area.
A picture was, in this case, worth many thousands of words!
We subsequently worked hard to understand and eliminate
localized modes of degradation, including those associated
with dark-line defects in the long narrow lasing region of the
laser and those associated with mirror surfaces. Subsequent
experiments showed that DLDs identical to those seen in lasers
could be generated by optical pumping of undoped and unprocessed laser material, thus confirming that DLD initiation and
growth could result from properties of laser material that were
not associated with proton bombardment, p-n junction dopants
or contact metallization technology.
Many improvements in LPE growth technology and its
automation were also made during this period. Fundamental
difficulties with this “batch” process made it stubbornly difficult to reproducibly control but it was greatly improved in the
skilled hands of our crystal growers.
By early 1977, with continued work
on growth and process improvements,
screening techniques, and protocols for
accelerated aging, we felt that, for a set of
randomly selected lasers, we could confidently predict a median lifetime at 22 °C
at 34 years and a mean-time-to-failure at
22° C at 1. 3 million hours (>100 years).
The so-called “million hour paper,” which
was published in Applied Physics Letters in
1977 ( 31, 756), demonstrated to us and
to the world’s laser community that it was
possible to construct semiconductor laser
devices with very long lifetimes.
The secret of our success
Soon after achieving and publishing these
results, I attended a conference in England
on the general subject of light emission from
semiconductors. During the Q&A, the head of the laser devel-
opment program at the Standard Telecommunications Labora-
tory asked, publically and rather pointedly, “Dick, would you
please tell us the secret of your reliability success?”
I puzzled for a moment and then blurted into the micro-
phone, “We do everything very carefully.” This brought a good
deal of laughter from the audience, but it was not intended as a
joke. It took some years to convince skeptics that our develop-
ment program’s success required the solution of hundreds of
problems, innovation by scores of outstanding well-motivated
people, millions of dollars, systematic iteration and a good
deal of time. Perhaps our key achievement was the “proof of
principle” that semiconductor lasers with long lifetimes were
possible—a little like Roger Bannister’s four-minute mile. I’ve
found in years since that most business and political leaders,
Reliability milestones
By late 1974, with continuing work on many technology
fronts, the reliability situation had improved considerably, and
selected lasers had been operating continuously for more than
a year at room temperature (typically 30 °C). On the basis of
our data, we were able to conclude that “continuous room-temperature operation of these devices as lasers with power outputs
exceeding 1 mW per laser face for times in excess of 100,000 h
is possible.” This was an important feasibility demonstration,
but it reinforced the urgency of finding ways to confidently
“accelerate” this aging so that lasers tested for short periods
could be installed in the field with the expectation that they
would last for decades.
Viewing direction Mirror 75 μm 12 μm
C o ntact
n GaAs (substrate)
N GaAlAs (hole barrier)
p Ga As (active layer)
P GaAlAs (electron barrier)
p Ga As (contact layer)
Contact
Proton bombarded
semi-insulating
region
Proton-bombardment-delineated strip geometry GaAs/
Ga AlAs semiconductor laser with a “window” on the substrate
side. Note the four epitaxial layers of different composition
grown by liquid phase epitaxy.
