Introduction to semiconductor lasers

Breck Hitz

Semiconductor lasers, or laser diodes as they often are called, are by far the most ubiquitous of all lasers. Every year, laser-diode sales are a hundred-fold greater than the unit sales of any other type of laser. At the beginning of the 21st century, laser-diode sales exceed 100 million units per year. They are used in an enormous variety of applications, from CD players to laser printers to telecommunications systems. Very simple semiconductor lasers are much smaller and lighter, and much more rugged, than other lasers (see Fig. 1).

Modern laser diodes
The first laser diodes, which were developed in the early 1960s, looked something like the device illustrated in Figure 1. But semiconductor lasers have evolved a long way since then. The device shown would require very high current flow to maintain a population inversion, and the heat generated by the steady-state current would quickly destroy the device.

To reduce the current and heat while maintaining a population inversion, modern laser diodes pack the stimulated emission into a small region. In this way, the current density is great enough to maintain a population inversion, but the total current does not overheat the laser. Two approaches are used to increase the density of stimulated emission: increasing the density of charge carriers, and increasing the density of intracavity optical power.

Both techniques invoke the sophisticated semiconductor fabrication techniques that have evolved over the past 30 years. These allow the more complex structures to be grown, literally one molecule at a time, from the basic raw materials. Today, methods such as molecular-beam epitaxy and metal-organic chemical vapor deposition allow the creation of semiconductor structures that are only several atoms thick.

One way to increase the density of charge carriers is to use a stripe electrode (see Fig. 2). Instead of injecting the current over a wide area of the diode's surface, current is injected only along a narrow stripe, resulting in a much higher concentration of charge carriers inside the diode.

The laser confines the current to a small region with its stripe electrode. This is called "current confinement" in the plane of the junction and it also confines the generated photons perpendicular to the plane of the junction with its "double heterostructure" design. Several junctions of dissimilar material are in this design, not just a single junction as shown in Figure 1. The electrons and holes combine in the narrow region shown as having thickness "d" in the figure, and the material here has a higher refractive index than the material above it or below it. That means that photons are reflected off the interface between the materials and therefore are confined to the region.

This is a common technique of confining charge carriers to a small region, the stripe electrode. It also illustrates one technique for confining the photons—total internal reflection from the interface above and below the active region. But the photons are still free to spread out sideways. "Index guiding" structures put a stop to this. Here, the laser has been fabricated with a low-index material on both sides of the active regions, as well as above and below the active region (see Fig. 3). Now the photons cannot spread out in any direction.

Confining photons in frequency can also be beneficial. The best way to reduce the bandwidth of a laser is usually to reduce the bandwidth of the laser's feedback, and small, built-in gratings often replace the resonator mirrors in semiconductor lasers (see Fig. 4). Such lasers are sometimes called "distributed feedback" (DFB) lasers because the feedback—or reflectivity—is distributed over the length of the grating, rather than occurring all at once at a mirror. The wavelength that is fed back is determined by the period of the grating. Usually, a DFB laser has a grating fabricated into the entire length of the laser. A variation referred to as a "distributed Bragg reflector" has a distinct grating fabricated into the substrate on each side of the active area.

The end-faces of an "external resonator" laser diode are not fabricated into mirrors. Rather, the laser diode is placed in a separate resonator like a conventional gas or solid-state laser. Such lasers are capable of very narrow bandwidth and good frequency stability.

Arrays
Another important development was the evolution of laser-diode arrays—multiple lasers in a single device. Although the beam produced by this one-dimensional diode array device is of dubious quality for many applications, the output power is several times greater than can be obtained from a single laser. And many applications of high-power arrays—such as optical pumping of solid-state lasers—do not require high beam quality. In some cases, the optical phase of adjacent lasers can be locked, resulting in improved beam quality and stability.

Even greater power can be obtained by stacking one-dimensional arrays on top of each other, in essence creating a two-dimensional array. Several manufacturers produce high-power laser-diode "stacks" from which the output exceeds 100 W. Although the beam quality from such stacks is relatively poor, these devices are ideally suited for applications that require efficient delivery of high power, such as optical pumping of other lasers, and for many industrial and medical applications.

Vertical-cavity surface-emitting lasers
One of the most fundamental problems with the laser diodes discussed so far is the highly divergent, elliptical beam. The divergence of a laser beam is inversely proportional to the beam size at the source—the smaller the source, the larger the divergence. The "edge-emitting" laser illustrated in Figure 2, for example, is an extremely small light source—perhaps 1-µm wide—and it is much smaller in the vertical direction than in the horizontal. So its output beam diverges much more rapidly in the vertical direction than in the horizontal. The vertical divergence is typically tens of degrees, and the horizontal divergence is several degrees. The highly divergent, elliptical beam can be corrected, to an extent, with a cylindrical lens, but the inherent problem of a small, elliptical source can never be completely rectified.

Furthermore, laser diodes have other fundamental limitations. Although they are extremely small, their resonators are still hundreds of micrometers in length, long enough to support multiple longitudinal modes—so unless the laser bandwidth is artificially reduced (by fabricating a DFB structure, for example), mode hopping among these modes produces an instability in both the amplitude and the frequency of the laser output. Another limitation is that, because the output beam emerges from the edge of the cleaved crystal and the crystal is not cleaved until the end of the manufacturing process, it is not possible to test the devices optically during manufacture. This limitation tends to drive up the price of manufacturing. Lastly, monolithic, two-dimensional arrays cannot be created with edge-emitting devices.

The vertical-cavity surface-emitting laser (VCSEL) avoids these shortcomings. In the conventional laser diodes we've discussed so far, the cavity—or resonator—is in the horizontal plane. In a vertical-cavity laser, the cavity is along the vertical direction. In the vertical-cavity laser, the mirrors are located above and below the population inversion, instead of on either side. The horizontal-cavity laser is an edge emitter, while the vertical cavity is a surface emitter. The advantage of such a design is that it eliminates the problem of a divergent, elliptical beam caused by a small, irregular emitting surface of an edge emitter. The emitting area of a surface emitter is round, and many times larger.

The resonator of a VCSEL is shorter than that of a conventional laser. In fact, it is so short that the spacing between longitudinal modes is too great for more than one mode to oscillate. So the mode-hopping instability of conventional (edge-emitting) lasers is eliminated. Moreover, the manufacturing difficulties are reduced because it is possible to subject a wafer-full of VCSELs to optical testing during the manufacturing process. Very high densities, tens of millions of diodes per wafer, can be achieved, further driving down the cost of individual diodes. And it seems much more straightforward to fabricate two-dimensional arrays.

For more information about semiconductor lasers, see Laser Focus World, Matthews, April 2000 to December 2000.

ACKNOWLEDGMENT
Adapted with permission from Introduction to Laser Technology, B. Hitz, J. J. Ewing and J. Hecht, publisher IEEE, Piscataway, NJ.

BRECK HITZ is executive director of the Laser and Electro Optic Equipment Manufacturers Assn. in Pacifica, CA; e-mail: [email protected].