Generating more power from an individual diode laser is not necessarily a matter of enlarging the gain region—in fact, the gain region generally is made as thin as possible. Power from these salt-grain-sized devices increases linearly with current,
P = (Iop - Ith) nd
where Iop is the operating current, Ith is the threshold current, and nd is the differential quantum efficiency; quantum efficiency of diode lasers can be as high as 90%. The key to generating higher power, therefore, is to make the laser capable of operating reliably at high current.
Life equals power
Optical power and current densities within a diode laser are high enough to melt many materials. Just as with early efforts to improve reliability, high-power operation requires controlling the mechanisms that cause damage, including controlling defects in the alloys, reducing the threshold current, and removing heat from the device (as well as forming reliable electrical contacts). Limiting damage to the output facets of the laser has been especially critical.
Commercial lasers operating between 700 and 870 nm are based most commonly on alloys of gallium aluminum arsenide (Ga0.93Al0.07As, for example, provides the gain in pump lasers for diode-pumped solid-state lasers—DPSSLs). Output from the laser passes through the facets cleaved on the <110> crystal plane. The optical power density at the cleaved facet can be as high as 10 MW/cm2. This power heats the facet, which may oxidize the material and rapidly increase light absorption, leading to runaway heating and causing the facet to melt. High-power GaAlAs lasers with uncoated facets have very short lifetimes.
Saving face
Coating the laser facets is done to control their transparency and to encapsulate the facets against the degrading effects of oxygen and water vapor. A facet coating is not just a l/4 layer deposition. Considerable effort has been devoted to research into so-called nonabsorbing mirrors.
Examples of facet treatments include the etch and regrowth of the facet region to obtain finer control of impurities or to alter the alloy at the output of the device, or cleaving the facets in ultrahigh vacuum and depositing a protective film of silicon that is thin enough (<50 Å) to be transparent. Treated facets have damage thresholds at least two to four times higher than untreated facets.
A different approach to protecting the facets is to widen the active layer so as to spread the beam and decrease the electric-field intensity. It was this principle that allowed gain-guided lasers to live on after the invention of index-guided devices, with their lower threshold current and superior beam quality. The "mushiness" of the confinement due to gain guiding results in a more diffuse output, allowing higher-power operation.
Of course, the wider output region results in multitransverse-mode operation. High-power diode lasers are "broad-stripe" devices, with output apertures about 100 µm or more, typically several times wider than single-mode lasers. A variety of designs spread the beam near the facet without resorting to gain guiding, resulting in active layers that have "V" or "Y" (or more complex) shapes when viewed in cross section.
Several Japanese firms, including Matsushita (Sunnyvale, CA), have made use of a characteristic of liquid-phase epitaxy to thin the active region near the facet and so diffract and broaden the output beam. Liquid-phase epitaxy tends to deposit thinner layers over narrow features, so narrowing the base layer under the active region at the ends of the device produces the desired effect.
Going to the well
The invention of the quantum-well structure was a milestone in the development of diode lasers (see “Quantum well is essential to high-power devices”). To achieve high power, single quantum wells can be stacked in a single device to form a multiple quantum well (MQW). The higher-bandgap confinement layers sandwiched between the thin active layers must be thick enough to prevent the escape of charges by tunneling.
Multiple quantum wells have excellent carrier confinement. The multiple-layer structure provides superior optical confinement as well, so MQWs are used in both gain-guided and index-guided lasers and form the active region in most high-power devices.
The atoms in the crystals of the GaAs system are arranged ideally in what is called a zinc blend lattice. In a real crystal, imperfections disturb the ideal arrangement of this lattice. Such imperfections include point and line defects. Typical point defects are atoms missing from the lattice or extra atoms spaced between the crystal lattice. Line defects are locations where the lattice is severely displaced. There are four main causes of defects: imperfections in the substrate; contaminants, in particular oxygen, in the growth process; lattice mismatch; and mechanical damage, such as arises from cleaving the processed wafer.
Bulk can be degrading
All defects increase absorption and heating of the crystal. Furthermore, heating the crystal (either by absorption or from the device current) causes the defects to spread through the lattice. Defects can propagate to the active (gain) region of the laser where they cause a dark spot or a line (see figure). Dark line defects are an especially serious problem for GaAs based material, which is soft compared to some other laser alloys, and so allows defects to propagate more readily.
One method of limiting the propagation of defects is to place barriers in the material around sensitive areas of the device. Special layers of harder material, purposely introduced impurities, or material with a controlled mismatch in the lattice constant have all been used to refract propagating defects away from the active layer.
The main effort has been aimed at eliminating the sources of defects. The purity of source materials, the condition of the growth chamber, and control of growth temperature and alloy ratio continue to be refined. To reduce the stress from even the smallest lattice mismatch between layers of different ratios of GaAlAs alloys, the ratios can be changed gradually. Channels can be etched in the substrate to reduce the thickness of the material to be cleaved, limiting the likelihood of damage.
Aluminum-free and proud
Early research on diode lasers indicated that adding aluminum to the active layer extended lifetime by gettering oxygen impurities. This worked most probably because of high levels of oxygen impurities in these early devices. For the past several years, evidence has been mounting that aluminum is detrimental to device performance at high current, particularly near the facets.
So-called “Al-free” devices, based on indium gallium arsenide phosphide (InGaAsP), have several advantages. Alloys using indium have a low reactivity with oxygen. The speed with which charge carriers recombine near the surface of the alloy is much slower than in alloys containing aluminum—a significant effect in limiting heating near the facets. In addition, InGaAsP is a harder alloy than GaAlAs, inhibiting the spread of dark-line defects. cant problem for the Al-free compounds has been the difficulty in fabricating device structures. Varying the alloy ratio in layers of InGaAsP structures results in smaller bandgap differences than for layers containing aluminum. This allows greater leakage of current carriers and results in lower efficiency and higher device temperatures.
The solution is to use Al-free compounds in the active layer and use high-bandgap aluminum compounds in cladding layers for carrier containment. Using this concept, researchers at the University of Wisconsin (Madison, WI) reported in 1998 the development of a 100-µm-wide device operating at 805 nm with a CW output power greater than 8 W. Coherent Semiconductor (Santa Clara, CA) has adopted Al-free designs for all of its high-power diodes.
Not all manufacturers are convinced that the added complexity of aluminum-free devices is worth the effort, especially because all high-power diode lasers have coated facets for protection. Polaroid (Norwood, MA), for example, produces diode lasers based on GaAlAs that rely on high levels of control in sources and growth to meet lifetime requirements.
The meaning of the phrase "high-power diode laser" has changed dramatically in the past 20 years. At one time it referred to devices with more than 20-mW output. Today, essentially all diode lasers with less than 100-mW output are individual narrow-stripe devices operating in single-mode. Individual broad-stripe multimode lasers have exceeded 10 W.
Perhaps the obvious way to obtain high power, particularly in view of the small size of the devices, is to group numerous individual lasers together. Such diode arrays or bars are capable of output in the kilowatt range. With such high power densities come problemsthe multimode beam of a broad stripe device becomes incoherent output unless advances designs are employed, and a practical consideration, cooling the device, becomes critical.
ACKNOWLEDGMENTS
We would like to thank the following individuals and organizations for providing images used in the Semiconductor Lasers 2000 series timeline: Michael W. Davidson, Florida State University; GE Research and Development Center; Zhores I. Alferov; Nick Holonyak, University of Illinois; Dan Botez, University of Wisconsin-Madison; Lucent Technologies, Bell Labs; Connie Chang-Hasnain, University of California Berkeley; Nichia Chemical Corporation; Jack Jewell, Picolight Corp.Ed.
Next month the series will summarize the development of high-power arrays of diode lasers.
Quantum well is essential to high-power devices
Developed by Holonyak and coworkers at the University of Illinois (Urbana, IL), the quantum-well (QW) structure is essential to all modern diode-laser development. Using metal-organic-chemical-vapor-deposition or molecular-beam-epitaxy growth techniques, it is possible to make the active layer of the laser on the order of 50 nm thick. This constrains the motion of electrons and holes in the narrow dimension and limits their energy and momentum. The conduction band and the valence band each break into a series of discrete sub-bands, which look more like the orbitals of individual atoms than the continuous energy bands of a solid.
One consequence is that the distribution of charge carriers in the energy bands changes from a continuum to a stair-step-type distribution. It becomes easier for a charge carrier to hop from a conduction sub-band to a valence sub-band (or vice versa), with an increased likelihood of electron-hole recombination.
In the separate-confinement-heterostructure design (see figure), layers outside of the QW serve to independently confine photons. The confinement layers have a refractive index lower than the active layers, forming a waveguide. (The waveguide also may be formed by a gradual change of the indices, a so-called graded-index separate-confinement heterostructure, or GRINSCH.)
The true benefits of the QW are only realized when the structure is combined with "strained-layer" designs, commonly used for the longer-wavelength lasers of telecommunications. Without the strained layer, the lower threshold current and other benefits of the QW are in truth mostly a geometrical effect, resulting from the thinness of the active layer.