A different mirror.. .

can be controlled. Pulsed power output of layers are used to confine electrons and over 1 W has been demonstrated from a holes, while thicker layers serve as mirrors single array. Edge-emitters have delivered 10 or waveguides to confine photons. “Thin” is times that much, but their geometry restricts about 10 nm, and “thicker,” about 100 nm. them to one-dimensional arrays. After these layers are grown uniformly To be sure, engineering hurdles still loom across the full wafer, the next stage is to de- for surface-emitting lasers. The biggest is marcate individual laser devices with such the substantial elimination of the heat from standard fabrication processes as ion im- the light generation process. The other plantation or plasma etching. They may be emiconductor laser diode problems that must be resolved uiclude con- either separated for individual packaging or technology has at last gone trolling the light’s polarization and delivering left joined together to form an array. beyond the edge emission of the power in one mode. Rut since potential In surface-emitters, as with the edge- light. In that invention of the applications range over optical-fiber com- emitters, most of the pioneering work has early 196Os, photons shoot munications, printing and scanning, two-di- been done with heterostructures of indium out of one edge of the semi- mensional image recognition. and com- gallium arsenide and aluminum gallium ar- conductor wafer after re- puting, the debices are moving rapidly from senide alloys. Those 111-V materials are bounding off mirrors that research into manufacturing. ideal for complex surface-emitter devices. have been literally cleaved ATOMIC LAYERING. The surface-emitting The ciystal lattice of AlGaAs matches the out of the ci-ystalline substrate. In the new laser’s entire structure-mirrors and all- spacing of the lattice of gallium arsenide surface-emitting laser, though, photons is built up by “spray painting” crystalline substrates, so it can he grown epitaxially on bounce vertically between mirrors grown layers onto a substrate with atomic pre- GaAs wafers without fear of crystal defects into the structure, and then shoot straight cision. By means of techniques such as mol- and dislocations.Although InGaAs materials up (or down) from the wafer’s surface. ecular beam epitaxy and metal-organic are not naturally lattice-matched, they can ‘Those differences seem simple, but their po- vapor phase epitaxy, hundreds of layers of be grown on GaAs substrates if the layers tential consequences for manufacturing effi- semiconductor materials can he grown one are kept thin-and the elastic strain im- ciency and new applications are tremendous. on top of the other. By mixing and matching parted to the thin layers has beneficial elec- Most critically, laser devices that emit the tnaterials to create “designer” alloys, it tronic properties. With in situ diagnostics, light from their upper surface can be fah- is possible to grow a crystalline structure tolerances can be tightly controlled, and ricated side by side on a wafer in vast with all the electrical and optical properties processing technologies such as ion implan- numbers. Tens of millions can fit on a 100- desired for its various parts. tation and etching are well understood. millimeter-diameter wafer. Packing them in This method of tailoring semiconductor Finally, 111-V heterostructures give laser so densely enhances manufacturing effi- structures is called handgap engineering diodes their lowest lasing thresholds and ciency. Even more important, the way they [see “Quantum-tailored solid-state devices,” hest temperature characteristics. are made means they can be integrated on by Titnothy J. Drummond, Paul L. Gourley, STRUCTURAL DIFFERENCE. In the surface- chip with transistors and other devices. and ’Thomas E. Zipperian, IEEE Spectrum,, emitter, the semiconductor laser structure There is no need to wire each of them indi- June 1988, pp. 33-371. In essence, thin is swiveled by 90 degrees from the orien- vidually to a circuit, as is done with edge- emitters, and they can even be optically Index of refraction: the ratio of the speed of linked to overhead elements. ~ Defining terms

Coupling efficiency also benefits from a ~ light in a vacuum to its speed in a material surface-emitting geometry. The beam of Band gap energy: the minimum energy longitudinal optical modes: discrete optical light that issues from an edge-emitter, typ- required to break a valence electron from a crys- resonances occurring in a semiconductor laser ically through a 1-by-2-pm aperture, is tal bond and set it loose in the crystal lattice as a when twice the cavity length (distance between usually both elliptical and divergent-not at mobile carrier. mirrors) corresponds to an exact multiple of the all the round, collimated beam that couples Edge-emitting laser: the conventional semi- photon wavelength its entire energy into an optical fiber. In conductor laser, in which the gain medium is Quantum well: a thin layer of material of com- contrast, the surface-emitter aperture can grown epitaxially on a substrate and the two mir- paratively narrow band gap sandwiched between he shaped to give the beam the ideal rors are formed by cleaving the resulting structure wider-band-gap layers for the purpose of trap-

~ circular cross section and its diameter can to expose reflective crystal facets. Light generated ping mobile carriers and confining them quantum- be made large enough, at 5 pm, to minimize in the gain medium bounces horizontally between mechanically i the divergence of the light rays. the facet mirrors (there is usually some lateral Vertical-cavity surface-emitting laser: a

~ Last, if left attached together on the wafer, confinement), and some leaks out through one solid-state laser in which the two mirrors as well ’ surface-emitting lasers can form a two-di- mirror to become the laser beam. as the intervening gain region are grown epitaxi-

~ mensional array whose power and direction Epitaxy: the growth of crystal by depositing ally on a semiconductor substrate Light gener- atoms on a substrate a single layer at a time by ated in the gain region bounces vertically ~ Paul L Gourley, Kevin L Lear, arid techniques employing molecular beams or between the mirrors, and some leaks through the Richard P Schneider Jr gaseous chemicals, top surface mirror to form the emitted beam , Sandia National Laboratories

001 8-923.i/94/S4.00 :‘1994lEEE 31 tation of a cleaved-cavity edge-emitter. The laser light does not travel horizontally within a single layer inside the wafer, but instead bounces vertically through the epi- taxial layers between mirrors grown above and below the active layer. Because the active layer is only 10 pm thick, the length of the cavity between the mirrors is less than 10 ym-extremely short in com- parison with the 100 pm or more of the edge-emitter’s cavity. The mirrors in a surface-emitter are grown into the laser structure itself. Each is composed of paired layers of two dif- ferent materials, such as GaAs and AIAs: In this scanning electron micrograph, the “checkerboard”is in ac- one material has a high index of refraction tuality a two-dimensional phased-locked array of vertical-cavity and the other a low index. (The index of re- surface-emitting lasers. Molecular beam epitaxy was employed to fraction is the ratio of the speed of light in a grow the semiconductor wafer, complete with mirrors [light and vacuum to the speed of light in the ma- dark stripes], and the result was topped by a thick phase-shift layer terial.) Because the light waves from suc- [solid light layerl. Then reactive ion beam etching was used to cessive layers interfere constructively, to- define each laser individually as a square, measuring 5 ,um on a gether, the pairs of layers can reflect more side, and the thick phase-shift layer was etched off the top of every than 99 percent of the photons at the laser’s other laser in the array. The inset is an optical micrograph of light emission wavelength. emitted from the operating array, The mirrors are known as quarter-wave stacks because each layer’s thickness is equal tiplies rapidly, and some of the photons leak While he!ium-neon gas lasers of the type to one-quarter of the wavelength of the light out through the circular aperture in the used in supermarket bar-code scanners are inside the material. The wavelength there is contact on one mirror, producing an intense approximately 0.01 percent efficient, com- shorter than it is in a vacuum by a ratio equal beam of coherent light. mercial laser diodes are 30 percent efficient to the refractive index. For example, the re A COMPARATIVE CINCH. With today’s precise or better. fractive indices of GaAs and AlAs are 3.5 and epitaxial technology, surface-emitters are One serious obstacle to high efficiency in 2.9, respectively, at the emission wavelength expected to be much faster and simpler to surface-emitters, however, is high series re- of 980 nm, so that the respective layers need make in high volume than conventional sistance to the charge carriers (electrons to be 70 nm and 84 nm thick to create an ef- edge-emitting lasers can be. They need less and holes) transported through the laser fective mirror. special handling, above all because the mirrors. The high resistance reduces the On the top surface of the surface-emitter mirrors are an integral part of the laser’s efiiciency and generates unwanted and the underside of the substrate, metal is structure. (In contrast, the cleaved-facet heat, which limits the number of devices deposited to form electronic contacts. A mirrors of advanced edge-errdters may that can be integrated onto a chip. circular aperture is left in one contact as an have to be coated or ion-milled, or gratings ROLE CONFLICT. Each quarter-wave stack in exit for the laser light. When a voltage is may have to be formed by laborious etching fact serves two functions: as a highly re- applied to the contacts, electrons and holes and regrowing.) flecting mirror for the photons, and as a are injected through the opposing mirrors Thanks to its built-in mirrors, each current path for the electrons and holes. and collect in the laser’s active region (or surface-emitter can also be conveniently Choosing mirror materials for these two cavity). There they are quantum-mechan- screened at the wafer level, before being purposes is a balancing act. Material com- ically trapped by quantum wells-thin separated. Since wafer-level testing iden- binations with the largest possible dif- layers of material in which the electrons tifies defective devices before packaging, ference in refractive indices make the best and holes have lower potential energy than only good ones are packaged and costs are mirrors, since they give large reflections at in the surrounding material. again reduced. In contrast, a wafer of edge- each interface and so require fewer in- An electron and hole from the collections emitters must be cleaved at least into terfaces. For physical reasons, indices of in the quantum wells can recombine at a separate laser bars-effectively one-di- refraction and band gap energy are in- random time to generate a photon but at an mensional arrays-to create the mirrors versely correlated, so material pairs with energy just too low for it to be absorbed in before the devices can be tested. In some the material above and below the quantum cases, edge-emitting lasers must even be in band gap energy. wells. The quarter-wave mirrors reflect packaged to provide sufficient heat-sinking When the adjacent materials had very dif- many of those photons back to the quantum before they can be qualified at full power, ferent band gap energies, the electrons and well. Sometimes the photon is reabsorbed and packaging is one of the principal costs holes were compelled to “jump”up from the by the well, in which case it is converted of laser diode manufacture. lowenergy material to the high-energy ma- into a new electron and hole. But if enough Packaging itself is often simpler for a terial. Consequently, while materials with a current has been pushed through the device surface-emission geometry. The output large index difference yielded highly re- to create an accumulation of electrons and facet of an edge-emitting diode must be pre flective mirrors, the concomitant changes in 1 holes in the quantum wells, the photon is cisely aligned to the edge of its heat sink so band gap energy between the layers were not absorbed and converted. On the con- that the heat sink does not obscure the less than helpful, for they represented a trary, the quantum wells become trans- output beam. But since surface-emitters series of large energy barriers that the parent and the photon can bounce back and emit light perpendicular to the plane of the carriers had to surmount as they moved , forth between the mirrors hundreds of wafer and away from the heat sink, their from one mirror layer to the next. times before escaping. In fact, the photon alignment tolerances can be much looser. That impediment to carrier flow was ev- is likely to stimulate a new electron and hole In general, semiconductor lasers are the idenced in high resistances, which increased to recombine, issuing a new photon. The most efficient of all lasers at converting the voltage drop and lowered the power con- photon population in the active region mul- input electric into output optical power. version efficiency. The frst surface-emitters

32 IEEE SPECTRUM AUGUST 1994 ~ as a result had efficiencies of only crystal at various spots. Strain

~ a few percent as well as threshold in addition flattens thie gain voltages of 3 V or higher. The spectrum, making the la:ser less wasted power, worse yet, was not sensitive to changes in aimbient i lost; instead it went into heating temperature.

~ the laser device, shortening its ULTRASHORT CAVITY. In a twical i lifetime and limiting the number ’ that could be joined into a two-di-

~ mensional array. Many approaches have been used to decrease the series re- sistance of the mirrors while main- and switching characteristics. taining their high reflectivity. They The energy separation of the I have for the most part aimed at longitudinal optical modes (or

~ smoothing the change in bandgap primary lasing wavelengths) is i energy by grading the compo- inversely proportional to the

~ sition of the mirror layers. This cavity length. Accordingly, the i works well because electrons and modes are separated by approx- I holes have wavelengths (1 to 10 imately 0.1 electron volt, corre-

~ nni) that are much shorter than sponding to a wavelength of

~ the optical wavelengths (about about 100 rim.

~ 300 nnd within the materials. In A 0.1-eV separation is wider

~ other words, grading the tran- than the spectrum of gain re- , sitions between the materials eases the motion of the charge carriers on the scale of the carrier wavelengths while at the same time retaining the large differ- ences in refractive index and the high degree of reflectivity at the The rnirrors qf a vertzcul-cuaity surface-emitting laser are pazrvd phi)ton’s wavelength. layers qf semiconductor materzals, which together reflect more than Grading approaches developed 99 percent of photons at the lasing iuuvelength. The uctzve region by one of the authors (Lear) at sandwzched bptwecn the nizrrors zncludas seceral qiunturn wells- Sandia National Laboratories in ulfmfhznluyers that tra) elec trons [dot31 und holes lopen circles1 New Mexico and by Larry Coldren znjected from metal conturts upplied to the substrute rind tot and his colleagues at the Uni- mirror In the quuntuni icells fhc iurrzers rccombzne to produce versity of California at Santa photons IZLUQ c~rrozusl.Photons (,f the right zuwelength niultzfily us Barbara have been tolerably suc- they bounce betzc?een tht niirrovs Sonic ofthe photons leak out of cessful. They have lowered the re- the top mirror through an upc~uri~dctined in the top metul contuct sistance of quarter-wave mirrors to produce un inteiiae bemi of colicrent lzght enough to yield surface-emitter threshold voltages of less than 1.5 V and devices cvith efficiencies of 21 percent. the wave crests. In surface-emitters, pe- heating will alter the alignment, but will not Overall operating efficiencies are now riodic gain can be realized by growing substantially change the laser output if the within a factor of three of the best conven- quantum wells at positions that coincide gain spectrum is broad enough. And in fact, tional edge-emitting lasers. with the antinodes in the optical standing the laser can be so designed that optimum In addition, Jack Jewel1 of Photonics Re- wave. The lasing wavelength is thus sta- alignment occurs near the highest specified search Inc. , Broomfield, Colo., and his col- bilized by being phase-locked to the gain operating temperature. The improved align- leagues have demonstrated a low-voltage regions. Moreover. the gain material in the nient then compensates for decreased gain vertical-cavity method that allows the active region is more effectively utilized as at high temperature. In this way, a nearly carriers to bypass the mirrors altogether. a source of electrons and holes that re- constant operating current can be main- Their design is siniilar to one pioneered by combine to create photons in the lasing tained over a wide range of temperatures Kenichi Iga and his colleagues at the Tokyo niodc. Furthcr evidence suggests that pe- (approximately 100 K). Such a result has Institute of Technology for indium-phos- riodic gain can enhance spontaneous not been achieved with conventional edge- phide-based surface-emitting lasers. They emission, which starts the lasing process. emitting lasers; the reason is that their inject current into the active region from Another means of increasing the laser’s wavelength hops to match the peak of the the sides of the device, underneath mirrors efficiency is to lard the active regions with gain spectrum, and the peak’s height de- that can even be made from insulating ma- thin layers whose crystal lattices are mis- creases with temperature. terials and that can be applied like coatings matched just cmough to give rise to elastic A laser’s stability is related to the cavity instead of being grown epitaxially. Although strain. Holes in strained layers move more lifetime-that is, the average time a photon such an approach offers flexibility, it as yet easily to match the motion of electrons. remains in the cavity. For stable lasing plus does not offer the same performance as Consequently. the threshold current for optimum output efficiency, the mirrors have devices based on epitaxially layered con- lasing is reduced. heating effects are niin- to reflect enough light to sustain the lasing ducting mirrors. imized, and the laser lifetime is extended. process, but leak enough light to maintain a

~ GAIN IN GAIN. The active region has also As a bonus. t;trained-layer quantum wells strong output beam. benefited from bandgap engineering. The are ideally suited to periodic gain. Strained The laser’s switching speed is also laser’s gain can be increased by a phe- laycm can be kept thin (about 5 nm) and be related to cavity lifetime, in inverse pro-

~ nomenon known as periodic gain- spaced wid? apart (130 inn). a configuration portion. Adding to the cavity length and in- ’ tially, pumping the optical standing waie at that prevents the strain from dislocating the creasing the mirror reflectivity extend the 33 __ ~ ~~ Lattice Mismatch on GaAs Substrates (%)

2.8

2.6

2.4

2.2

2 .o

1 .e

1-6 --I 1.4

1.2

1.o Current, mA L-~- ~- .~-’ 0.8 The performance of surface-emitting lasers has 0.6 been improved dramatically by reducing the re- sistance of their mirrors. For two d$erent lasers, 0.4 the graph above plots the intensity of light [redl 0.2 and the voltage drop [blue] as functions of the -4% -3% -2% -1% 0 +1% +2% +3% +4% drive current. The lasers had a similar structure, Lattice Mismatch on InP Substrates (%I except that the energy barriers between the layers in the mirrors of the newer devices [solid lines1 Bandgap enerewhich cOmeSpOndS to emission wavelength [vertical axes, abovekz’@ends were not as abrupt as those of the older ones on the composition of the semiconductor alloy. Few III-Vsemiconductoralloys have lattice [dashed lined The effect of those graded barriers spacing that matches the lattice spacing of gallium arsenide [top hmizontal axis1 or indium in the new lasers was to reduce the voltage drop phosphide [bottom horizontal axis1 the two most common III-Vsubstrate materials. The solid and associated joule heating in the mirrors as lines represent continuous temly (three-element)alloy ComposiriOns that connect points cor- current was driven through them. As a result, responding to binary compositions. For example, the line linking aluminum arsenide and each newer laser could work to higher currents gallium arsenide represents alloys of AlGuAs-used for most vertical-cavity lasers. To emit and output powers than its predecessor. [The in the visible spectrum, however, phospb-bused materials are needed. The bandgap energy arrows in the diagram point to the vertical axis to Ifsilicon, not used for suflace-emitting lasers, is shown for reference. which each curve is referred.]

cavity lifetime. For a cavity 2 pm long and 20 elements. In all cases, though, the ob- Visible wavelengths, including red, with a mirror loss of 1.5 percent, lifetime is served beam width is about double what it orange, yellow and green, are required or 5 picoseconds. Thus, the laser’s intrinsic would be if limited by diffraction, indicating desirable in such applications as display frequency response is up to 200 GHz. Other that the phase-locking is not yet perfect. systems, laser printing, scanning systems factors affecting laser switching speed Prototype arrays have emitted just over 1W (such as bar-code scanners), short-haul include the collection of carriers in the output power in pulsed operation, or an optical communications systems, and optical quantum wells, the carrier concentration in order of magnitude less than the peak memory. One recent breakthrough used the wells, and optical loss mechanisms. To output of one-dimensional edge-emitting AlGaInP for visible-wavelength surface- date, parasitic elements have limited the re- arrays. The figure should improve as the emitters having continuous-wave operation sponse of surface-emitters to 14 GHz. mirror-resistance problem is solved. at room temperature, as shown by a group PLANAR ARRAYS. Perhaps the crowning To shape and control the output beam, led by one of the authors (Schneider) at glory of surface-emitters is their capacity the arrays can be integrated with diffractive Sandia. The devices are rapidly reaching a for integration en masse into two-dimen- or binary optics-lenses of semiconductor commercial level of performance, with sional arrays. Since the placement of indi- or dielectric materials that change abruptly threshold currents on the order of 1mA and vidual lasers on a wafer can be defined in shape, as opposed to the smooth curve of power outputs exceeding 2 mW continuous almost at will by standard microlithographic the usual lens. It may even be possible to wave at the red wavelength of 670 nm. processing, designers have plenty of free- modulate the optics electronically, to adjust The longer infrared wavelengths, partic- ularly 1300-1550 nm, feature in communi- ~ dom to shape the light beam in the far field, the relative phase of the lasers so that the 1 remote from the wafer surface. beam could be steered at very high speed. cation systems based on optical fiber (such 1 If dispersed on the wafer, the surface- Further issues will also need to be ad- as long-haul phone and data transmission emitters can be addressed either indepen- dressed as these lasers move toward com- lines). Here, InGaAsP-based surface-emit- dently or as a matrix. If crowded together, mercialization. For example, the reliability of ters operating at 1300 nm have been they can be coupled in a phase-locked array. surface-emitting lasers is not yet well quan- demonstrated by Iga’s group in Tokyo and When phase-locked, all the lasers in the tified, but early reports are encouraging. by John Bowers and his group at the Uni- array operate in unison to create a narrow, CHEMISTRY OF COLOR. The InGaAs/AlGaAs- versity of California at Santa Barbara in col- coherent beam. based heterostructure system, on which laboration with Hewlett-Packard Labor- Preliminary experiments with prototype most surface-emitters have so far been atories in Palo Alto, Calif. phase-locked arrays have shown that the based, is limited to wavelengths in the near For the mirrors in its surface-emitters, output beam narrows as the array size in- infrared, from 750 to 1100 nm. Much recent Iga’s group employed pairs of dielectric creases from a single element to two by attention has focused on extending their coatings such as silicon plus either silicon I two, three by three, and so forth up to 20 by materials and wavelength ranges. dioxide or magnesium oxide. Because the 36 EEE SPECTRUM AUGUST 1994 pairs differ greatly in refractive index, the optical-fiber communication systems. In formation systems could benefit from mirrors are highly reflective. The layers can these systems, the laser sources must emit surface-emitter-based improvements in also be deposited in a relatively simple light in a single mode within a narrow fre- memory systems. Optical discs have diode process. But as the mirrors conduct no elec- quency range in order to minimize dis- laser light focused to a spot on their surface. tricity, charge carriers must be injected lat- persion and noise. Now the ultrashort cavity Here again, the fine beam and integrated erally around and under both mirrors-a of surface-emitters makes them lase in a micro-optics of surface-emitters should requirement that complicates device fabri- single longitudinal mode unless the lasers veld low-cost miniature laser systems. cation. Also, because the dielectric mirrors are too big or driven too hard, whereupon To look farther into the future, holo- conduct heat poorly, the density of the they emit multiple transverse modes. Thus, graphic memory systems may be read and current in the small gain region may heat improving the single-mode performance of written by an array of lasers, each accessing up the devices and perhaps limit their peak surface-emitters is under intense research. a different piece of information stored in output power. That said, all indications are At visible wavelengths, the red surface- the hologram. These and other futuristic that further refinement of this structure emitter is an excellent candidate for short- applications will continue to steer surface- should yield truly practical and efficient haul fiber communications. The 650-nm emitter technology to the marketplace. 1.3-pm surface-emitter sources. InAlGaP laser is well matched to the atten- (This work was sponsored by the US. Overall, prospects for the development of uation minimum of the rugged plastic fibers. Department of Energy under contract these technologies are bright, despite their The same device looks good for printing number DE-AC0494AL85000.) immaturity. In the long run, newer materials and scanning because its circular, single- TO PROBE FURTHER. “Vertical Cavity Surface- may be used to broaden their wavelength mode, low-divergence beam simplifies the emitting Lasers,” by Kenichi Iga, Fumio ranges still further. For example, the ma- Koyama, and Susumu Kinoshita, appeared terials best at emitting in the green, blue- in the September 1988Journal of Quantum green, blue and ultraviolet wavelength ranges Electronics, vol. 24, pp. 1845-55. It reviews are the 11-VI alloys and the 111-V nitrides. the early work on surface-emitting laser These wide-bandgap semiconductors rep- diodes done by Iga and his co-workers in resent the future for laser diodes issuing the Japan. The June 1991 issue of the same pub- shorter wavelenc;rths of the visible swctrum. lication reviewed semiconductor lasers with tnt they are currently at a much earlier stage a special section on the vertical-cavity type. c)f development than the phosphides. At the In “Quantum-tailored solid-state devices” c)ther end of the spectrum, laser diodes based UEEE Spectrum, June 1988, pp. 33-371, Cin the alloys of gallium indium arsenide anti- Timothy J. Drummond, Paul L. Gourley,and I:nonide and emitting at the wavelengths of Thomas E. Zipperian discuss band gap engi- 2!-5 pm have stirred intense interest; but to In this scannzng electron mzcrograph, a neering with strained layers and all-epitaxy dlate there have been no attempts at using photonic lattice structure features holes that mirrors and make mention of vertical-cavity (;aInAsSb for vertical-cavity lasers. are 250 nm zn dzameter etched znto an surface-emittinglasers. 1.IKELY USES. Closer at hand, the most pro- AlGaAs multilnyei wufer. Such photonac “Microlasers” (ScientificAmerican, Nov- rnising application is multichannel optical- lnttzces, whzch are fabrzcated with electron ember 1391, pp. 86-94), by Jack L. Jewell, f iber communications. The technology is bemi lzthography and reactaue ion beam James P. Harbison, and Axel Scherer, goes 1: eing used over short distances to link etching, are currently being znuestigated for into epitaxy and the surface processing of s,ystems in local-area networks, and is also possable use zn future generuttons (flow- vertical-cavity wafers for microlasers. teing pushed to replace electronic buses or threshold uertzcal-cuuzty lasers. Finally, “Optical Processes in Micro- backplanes. Present surface-emitter per- cavities,” by Yoshihisa Yamamoto and formance and attributes are well matched to design of the optics needed to focus it on Richart E. Slusher, which appeared in the multichannel systems of this kind. They can paper or film. Print bars based on light- June 1993 issue of Physics Today,pp. 66-73, be easily fabricated into multi-emitter arrays, emitting diodes may well be usurped by reviews upcoming generations of micro- have the power and modal properties to laser arrays, whose greater brightness resonators for quantum electrodynamics propagate over a few hundred meters, and should add to printer speed. And a large flat and microlasers. + can be modulated at rates of up to a gigabit array of lasers that can be turned on and off ~~ ~~~~ per second, while their circular beams are like pixels in a computer screen is distinctly ABOUT THEAUTHORS. Paul I.Gourley is program easily coupled into fibers. faster than a scanning system that relies on coordinator for photonics in the semiconductoi For starters, the more mature AlGaAs rotating mirrors to scan a single beam in LJhYSkS department of Sandia National Laboratories surface-emitters operating at wavelengths two dimensions. in Albuquerque, N M. His current interests includt near 850 nm will be used with GaAs or Visible wavelengths help images, whose photobiology, the invention of the biological micro- silicon detector technology. That is the plan contrast often suffers in infrared light. So cavity iaser, and studies of artificialiy structureG in the project being pursued by a protkge of visible-wavelength surface-emitters could semiconductors for lasers and broadband emitters the Advanced Research Projects Agency unseat the helium-neon lasers used in su- He received Department of Energy Awards in 1985 (ARPA) called the Optoelectronics Tech- permarket bar code scanners. and 1993 for outstanding research on strained-layei nology Consortium (OETC); the members Two-dimensional arrays are obviously and mirror heterosiructures. include AT&T, General Electric, IBM, and well suited to image processing and optical Kevin L Lear (M) is a senior member of the Honeqwell. OETC is working toward a 32- computing, especially if individual control technical staff in the Photonics Research Depart- channel, 500-Mbis-per-channel optical data electronics turn each laser into a “smart ment. He is at present investigating surface- link for board-to-board communications. pixel.” If detectors are then added to the emitting lasers and mirror transport issues. His The transmitter module, which is being de- pixels, the array can both receive multiple other interests include quantum-effect devices and veloped by AT&T Corp., couples a- 32- optical inputs and process them into optical compound semiconductor electronics. element AlGaAs surface-emitter array to 32 outputs. Arrays of this type are being de- Richard P. Schneider Jr. is a senior member of the optical fibers. veloped for photonic switching systems that technical staff in Sandia 3 Semiconductor Materiais Subsequently,as surface-emitters appear would interpret and route optical signals, as Research Department. His research is directed al the with 1.3- and 1.55-pm wavelengths, they ~’ well as for processors in optical computing epitaxial design and growth of verfical-cavity surface-

may replace the distributed-feedback edge- ~ systems. emiifing lasers, including pioneering work on some

1 emitting lasers used today in long-haul Finally, even electronic computing and in- operating in the red region ofthe visible spectrum. 3