It has been nearly four decades since the demonstration of the first diode laser at GE (Schenectady, NY) in 1962. Turning the first laboratory lasers into practical devices was an agonizingly slow development, hindered time and again by distressingly brief diode lifetimes. It took another ten years to invent the double heterojunction, the breakthrough that extended the life of diode lasers by seven orders of magnitude. This development turned the laser into a consumer item, available for pocket money.
The inventors of diode lasers largely foresaw the revolution in communication and optical storage that has transformed the laser marketplace. Many observers predict that the true revolution in these applications is only just beginning. The demands of these markets are certainly driving the near-term development of diode lasers.
Taking the cost out of telecom
The need to reduce the cost of diode lasers has become a preeminent concern for developers. Uncooled lasers, for example, are expected to eliminate the expense and size of TE coolers used in long-haul modules. Today’s uncooled lasers meet commercial requirements up to case temperatures of 85°C.
The requirement that every DWDM channel have its own laser (and spare) results in large inventories of expensive devices. Distributed feedback lasers can be configured to four different wavelengths, reducing the spare inventory problem, but with hundreds of DWDM channels per fiber in the offing, more relief is needed. New tunable lasers are being developed that will operate at any of the wavelengths used in the long-haul band.
The ultimate in tunable lasers may prove to be quantum dots (QD)—hemispherical drops of alloy about 100 atomic layers in diameter arrayed on a substrate. Their submicroscopic size results in perhaps the lowest possible values of chirping, threshold current, and linewidth, along with high modulation frequency. Fabricating QDs is a somewhat random process that results in a variation in dot size, which allows a selection of wavelength by setting the length of an external cavity.
Lengthening the good
An array of QDs is the most extreme dimensional version of a vertical-cavity surface-emitting laser (VCSEL). Following the commercial introduction of VCSELs by Honeywell (Morristown, NJ), advantages in beam quality, modulation, and lower manufacturing costs led quickly to VCSELs dominating the short-distance multimode application at 850 nm. An intensive effort is underway to obtain the same benefits of VCSELS at longer telecom wavelengths.
The length of the active region in a VCSEL is short, a few tens of nanometers through the quantum wells, and the gain is correspondingly small. VCSELs depend on low-loss dielectric stacks with reflectivities that exceed 99% to maintain lasing. These mirrors are grown for 850-nm VCSELs using alternating layers of aluminum arsenide (AlAs) and gallium arsenide (GaAs), which have a natural lattice match.
The active layers of telecom sources for 1300 nm and 1550 nm, however, are based on alloys of indium phosphide (InP), not GaAs, and it has proven exceedingly difficult to grow mirror structures based on these alloys. Initially the best devices used “wafer fusion” to combine materials of different lattice constants into one structure. The wafer-fusion technique used the mirrors grown on GaAs substrates and gain structures on InP substrates and then essentially glued them together.
An additional challenge has been to pump the lasers using injection current when dielectric layers essentially surround the laser. Tackling this problem was deferred by using optical pumping to concentrate on the basic material problems, where progress has come from new alloys that allow the use of GaAs mirror stacks.
VCSELs at longer wavelengths
Researchers at Sandia National Laboratories (Albuquerque, NM) have added indium and nitrogen to form an active layer of InGaAsN for emission at 1300 nm, which still allows growth on a GaAs substrate and the use of reflecting stacks similar to conventional VCSELs. Cielo Communications (Broomfield, CO) plans to offer an electrically pumped version of this VCSEL. "Based on the VCSEL's inherent technical advantages, the 1.3-µm VCSEL will render obsolete the majority of 1.3-µm edge-emitting lasers used in telecommunications today," states Bob Mayer, Cielo's vice president of marketing
At 1550 nm, researchers at the University of California Santa Barbara have demonstrated operation of a VCSEL grown on an InP substrate using molecular beam epitaxy. This optically pumped device uses InGaAs in the active layer, as does a device that Bandwidth9 (Fremont, CA) has used in a demonstration at 1550 nm over 50 km of fiber at 2.5 Gbit/s without optical amplification.
Bandwidth9 plans to offer a tunable VCSEL using microelectromechanical systems (MEMS) to change the cavity length (see figure). The expected tuning range of 40 nm will cover the full L or C bands of DWDM defined by the International Telecommunication Union.
Progress in nitride
Two separate fields merge in the development of VCSEL arrays of gallium nitride (GaN) lasers. This material has been a major news story since the invention of blue LEDs and diode lasers in the late 1990s by Shuji Nakamura, then at Nichia Chemical Corp (Tokushima, Japan). VCSEL designs for GaN devices may prove helpful in overcoming problems that persist with blue lasers.
The threshold current for the best nitride lasers is an order of magnitude higher than that for GaAs lasers of comparable design. While GaN can withstand levels of defects that would be fatal in other laser materials, over time the high operating current exacerbates imperfections in the crystal. High-quality epitaxial layers of GaN are difficult to grow because of the significant differences between its lattice constant and that of available substrate materials. Lateral epitaxial overgrowth (LEO), developed at North Carolina State University (Raleigh, NC) may extend the lifetime of nitride lasers.
The LEO technique interrupts the growth of the GaN layer to deposit a SiO2 film. This film leaves pockets of exposed GaN that serve to seed high-quality crystal formation when the nitride growth is resumed. The dielectric film forms a barrier to the dislocations that originate at the interface of the nitride and substrate, resulting in the nearly total absence of these extended defects in the active layer.
Trapping excitons
An entirely different approach may lead to reduced thresholds and higher power. The attraction of electrons for the vacated holes in the valence band leads to bound electron-hole pairs in a semiconductor, called excitons. Exciton pairs do not remain bound during the typical operation of GaAs lasers. However, theoretical work suggests that exciton-binding energies in GaN are much greater than those in GaAs, and should result, with fabrication of the proper device structure, in significant enhancements in the gain of a nitride laser.
If the energy of the exciton resonance is coupled to the optical modes in a cavity, as can happen in a sufficiently restricted quantum well, the resulting resonance is called a microcavity polariton. The total energy produced in a GaN VCSEL gain region, where excitons are expected to survive, may be a superposition of the optical and exciton energies. At present, optically pumped GaN VCSELs have been demonstrated in the lab, but a significant challenge is to find a means providing injection current through the dielectric mirrors.
Indium uprising
Gallium nitride lasers containing indium in the active layer (to lengthen the output to 390 to 420 nm) are available with specified cw room temperature operation of 10,000 hours. Assuming their price falls considerably, their first major commercial application is expected to be digital versatile disks (DVDs). Other potential applications such as full-color projection displays and laser printers will require a longer wavelength than is available at present.
Adding more indium increases the wavelength, but also increases the threshold current. The GaN lasers have been demonstrated at 450 nm, but with greatly reduced lifetimes. Evidence indicates that as the fraction of indium in the alloy exceeds a few percent, there is a tendency for it to form random-sized clusters. This disorder creates additional conduction-band states that greatly reduce the efficiency of the laser.
The first diode lasers at 480 to 520 nm were demonstrated nearly a decade ago using II-VI alloys such as zinc selenide, but these devices have yet to last more than a few hundred hours. It also may prove that the indium limitation is a fundamental problem preventing nitride lasers from operating at these wavelengths, leaving mid-blue to green lasers still to be discovered. In the meantime, there remain opportunities in optical storage and biotechnology at the shorter wavelengths.
A more common future
Although the growth of optoelectronics has been phenomenal, it still is dwarfed by the worldwide silicon-based electronics industry. Development and manufacturing using silicon is much easier and cheaper than any diode-laser alloy—but as an indirect bandgap material (electron-hole recombination heats the crystal rather than producing photons), silicon has limited use for optoelectronics. Now there are the first indications that this may change.
A research team from universities at Catania and Povo, Italy has obtained amplification of 800-nm light from quantum dots of silicon pumped at 390 nm. Integrating lasers directly onto silicon chips could truly produce a technological revolution. Optical computers, for example, which have long been the subject of speculation but fraught with difficulties, might become a realistic possibility, enabling computing speeds almost a million times faster than present day-machines.