Click for next page ( 46


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 45
Lasers in Ct ~ Information Processing C. Kumar N. Pate! Since the invention of the laser in 1958, a tremendous amount of progress has been made in the field, both in the science and technology of lasers themselves and in the variety of applications of lasers (Schawlow and Townes, 19581. Lasers now cover the range of wavelengths from x rays to microwaves, where they merge with other coherent-radiation sources, such as klystrons. Uses of lasers also cover a broad spectrum: science, remote . ~ . . sensing, monitoring, detection, Information transmission and processing, industrial processing, defense, and medicine and surgery. In some fields, the introduction of lasers may be heralded as a "killer" technology that totally displaces an existing technol- ogy; in other fields, it may be a new-domain technology that has uncovered applications not thought of before; in still other fields, laser technology may turn out to fill a niche not occupied by any other technology. As John S. Mayo of AT&T Bell Laboratories has pointed out, the transistor should be consid- ered a killer technology because it displaced vacuum tubes (Mayo, 1985~; automatic speech recognition appears to be a new-domain technology; broadcast television ought to be viewed as occupying a niche in the information dissemination world without significantly changing the quality of information and coexisting with radio, newspapers, and other means of dissem- inating information. It is too early to decide how lasers will fit into various fields. This paper will describe accomplishments and future possibilities in communications and information processing and will allow the reader to decide how laser tech- nology will shape the existing technologies. What remains to be 1 ~1 45

OCR for page 45
46 C. KUMAR N. PATEL seen is whether lasers have contributed to a revolution or are part of the gradual evolution of the information age. Society relies on at least three distinct activities in the infor- mation age. The first is the creation of information; the second is the transmission of information; and the third is the manip- ulation of information. This paper will focus on accomplished and anticipated changes brought about by the exploitation of lasers and associated technology in information transmission and processing. These two areas share many properties but have many significant differences as well. LASERS IN COMMUNICATIONS The explosion in the use of lasers in communications has come about through a simultaneous improvement in the quality of the medium through which light energy is transmitted and the increased understanding of the laser sources, detectors, and associated phenomena that allow tailoring of the properties of materials and devices. For economic exploitation of fiber-based lightwave systems for information transmission, two parameters, sometimes combined, are very important. The first is the max- imum data transmission rate, which itself is limited by capabili- ties of the lasers, detectors, and associated electronics. The second is the maximum distance a bit stream can be transmitted over an optical fiber before a repeater is necessary. It is clear that the properties of the optical fiber contribute to the second parameter. Relevant optical fiber properties are the absorption losses and the chromatic dispersion. An appropriate standard of measurement for an information transmission system is the product of bit rate and distance, that is, the distance between repeaters at a prescribed bit rate. The impact of lightwaves on the capacity of a communication system is summarized in Figure 1, which shows the growth in system capacity since the construc- tion of the first telephone lines in 1890. The introduction of lightwave systems is causing a sharp change in the rate at which channel capacity has increased over the past 100 years. OPTICAL FIBERS Free space propagation never caught on for terrestrial lightwave communications because of the potentional interruptions aris- ing from fog, rain, and other natural phenomena. Lightwave transmission through guided media, optical fibers, is not a new 1 1

OCR for page 45
LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 47 Bow 107 ~5 HI 1 0 6 o 1000 By to] 100 > car J OPT I CAL FIBER - COMMUNICATION ,'~ SATELLITES at_ _~ L5 CARRIER ~ ~ COAXIAL SYSTEM COAXIAL ~ TODAY'S COAXIAL CABLE CABLE LINKS ~ AND MICROWAVE HIGHWAYS _ CARRYING 600 ~ CARRYING 32,000 VOICE VOICE CHANNELS,< CHANNELS : ~ 10 / CARRIER TELEPHONY FIRST USED: /.~ 12 VOICE CHANNELS ON ONE -- / WIRE PAIR -:TELEPHONE LINES FIRST CONSTRUCTED 1 ', 1 1 1 1 1 ~ 1 , 1 ~ 1 1 1 1 1 1890 1910 1930 1950 1970 1990 2010 2030 2054 ) YE AR FIGURE 1 Channel capacity improvement as a function of time. Note the discontinuous change in the average slope with the introduction of lightwave systems. phenomenon, but the use of fibers for lightwave communica- tions puts stringent requirements on the tolerable losses in the medium. More than 20 years ago, Kao and Hochkam (1966) proposed the use of clad glass fiber as a lightwave transmission medium. Initially, materials limitation due to absorption of impurities led to transmission losses of more than 100 dB/km in the 1960s. By the early 1970s, fiber losses were reduced to about 10 dB/km at 850 nm in silica fibers (see Figure 2) (S. R. Nagel, personal communication). Much of the improvement was brought about by a careful elimination of impurities. These losses were low enough that the early lightwave systems were designed to operate in the low-loss region of 850 nm (Figure 2) and use the available GaAs-GaAlAs heterostructure lasers. The next decade saw a continued elimination of impurities, such as OH. By 1976 this resulted in optical fibers with losses as small as 1.0 dB/km at 1.3 ,um (Figure 2~. The next generation of lightwave systems was designed to take advantage of the low-loss region near 1.3 ,um. This also has the additional advantage of being a region where the fiber dispersion is zero. Further improvements in fibers arose from reductions in OH and have shifted the minimum loss region to 1.55 ,um (Figure 2), where the losses are 0.15 dB/km (Nelson et al., 1985~. The current generation of lightwave systems is designed to take advantage of these low losses. See Appendix A for a detailed discussion of fiber loss and dispersion as they affect lightwave communica- tions. 1

OCR for page 45
4~3 C. KUMAR N. PATEL 100.0 E 10.0 - ~n An o J Cal 1.0 CL o 0.1 WaVELENGTH ('`m) FIGURE 2 Spectral loss data for silica fibers. LASERS _ V / 1976 _ THEORY _ /\1983 -_1 , ~ I , 1 , 1 , 1 0.8 1.0 1.2 1.4 1.6 The year 1970 was significant for lightwave communications from two points of view. First, the optical fiber loss dropped below 20 dB/km and, second, Hayashi and Panish achieved the first continuous wave operation of a semiconductor laser at room temperature (Hayashi et al., 19701. These two key break- throughs heralded the arrival of the age of lightwave commu- . . nlcatlons. Injection semiconductor lasers were first reported in 1962 (Hall et al., 1962; Holonyak and Bevacqua, 1962; Nathan et al., 1962; Quist et al., 1962~. They were homojunction devices whose threshold currents for laser action were so high that a practical lightwave communication that would require continu- ous wave operation at room temperature could not be envi- sioned. The following advances, important in the eventual continuous wave operation of semiconductor lasers at room temperatures, occurred in rapid succession: (a) the demonstra- tion of a GaAl-AlGaAs heterostructure growth by Woodall et al. (1967~; (b) the demonstration of heterojunction lasers by Ha- yashi, Panish, and their coworkers (Hayashi et al., 1969; Panish et al., 1969~; and (c) the demonstration of double heterostruc- ture for a dual-confinement active region by Alferov and col- leagues (Alferov et al., 1969) and by Panish et al. (1970~. By 1976 the double~heterostructure laser structure of Hayashi and Pan-

OCR for page 45
LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 49 ish had been developed to a point at which estimated life reached a million hours based on extrapolation from aging tests at elevated temperatures (Hartman et al., 19771. Today, the double heterostructure concept for carrier and photon confine- ment is used in all practical semiconductor lasers for communi- cations and other applications. To achieve continuous wave laser action at room temperature, even with the double heterostructure concept, lateral confine- ment (often called guiding) is necessary for both the injected current and the photons. In the GaAs-GaAlAs laser of the early 1 970s, this was achieved by using proton bombardment to define an active stripe of the laser (see Figure 3~. The proton bombardment makes the exposed material resistive, so that the injected current is confined to the narrow stripe shielded under the tungsten wire. The optical gain thus produced in the stripe provides gain-induced guiding in the lateral dimension. The first-generation lightwave systems, operating with multimode fibers and at approximately 840 nm, used the stripe geometry lasers. Early in 1975 it became clear to many people working in the field that the region of low fiber loss was going to shift to longer wavelengths as the concentration of OH impurities in the fibers was being reduced. Further, the zero dispersion wavelength could now match with a low-loss region at 1.3 ,um (see Appendix A). The use of zero dispersion necessitated the use of single- 12 lam TUNGSTEN WIRE PROTONS 1~' p+-GaAs A:/ P-AIxGO1 x i/ n OR p-GaAs ~ it/ N-AlX~1-xA ~_! SUBSTRATE / Do, / it/ By/ / DY .D ~~ ~ METAL FIGURE 3 Proton-bombarded stripe geometry laser. 1

OCR for page 45
50 c. 1 KUMAR N. PATEL CAP ACTIVE ~ SUBSTRATE METAL CONTACT P- DIFFUSION I InP I.'.'- ' ' ' l I n GO Asp ( AcTlvE ) i I n GaAsp ( CLA4D!NG ) I I InGaAsP (CAP) ~1= MESA SUBSTRATE ~P P~_ ! ~ - ~ , . . ~ , ~ ~ 1 _ n n STRIPE CHANNEL SUBSTRATE I NSULATOR ~ I ~1 `\\\\\\ ......................... R I DGE n P n n n r ~ t~ .,., P ~-~ ~ n A A ~ ~ A ~ ~ P n BURIED HETERO BURIED CRESCENT p p n n FIGURE 4 Six structures for single transverse-mode lasers. . DOUBLE CHANNEL An _P n mode fibers, and the long wavelength required going to ternary and quaternary compounds for lasers. Along with the single- mode fibers came the need for single transverse-mode lasers for efficient coupling of the laser output into the fiber. The six new structures shown in Figure 4 accomplish this to various degrees. The amount of lateral optical guiding determines the size of the mode, that is, mode volume, and therefore determines the "single-modedness" of the laser. Where the optical guiding is provided either by the lateral loss or by lateral mismatch of dielectric constants, they are called strongly guided structures. Guiding achieved only by the lateral definition of the gain (gain guiding) is a weak guiding process. These latter structures are thus called weakly guiding structures. STATUS OF LASERS FOR COMMUNICATIONS During the past several years the number of optical fiber communications systems has grown spectacularly. Most have been high-capacity systems in long-distance communications networks Nearly all of the systems installed today use single- mode silica fiber and InGaAsP/InP-based semiconductor lasers operating at a wavelength of 1,300 nm. This section will describe the key parameters of these commercially available lasers and the potential for improved performance as suggested by labo- ratory results. It will also discuss the principal current areas of laser research to give a sense of the types of communications lasers that might be possible in the future.

OCR for page 45
LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 51 A variety of laser structures are used today in communications systems. Some of the most popular are shown in Figure 4. Nearly all are of the general class of buried heterostructures and are produced by various methods of liquid-phase epitaxy (LPE). In some, a planar active layer is grown first and subsequently patterned and covered by an LPE overgrowth to bury the laser stripe hence the name buried heterostructure. In another major design, the active layer is grown in a V-shaped groove to give lateral confinement of carriers and light. These various types are all characterized by a narrow active stripe (1.5-3 ,um) and strong index guiding. Despite differences in details of design, the performance of the various types is not very different. The major variables characterizing all communications lasers are optical power output, modulation bandwidth, reliability, fre- quency spectrum, and cost. The following is a summary of the current status and future prospects in each of these five catego- ries. Optica/ Power Output Nearly all lasers available today are capable of generating at least 10 mW of optical power at the facet of the laser chip. Some can be driven as high as 30 mW. Laboratory studies have reported outputs as high as 100 mW (and higher for laser arrays). However, for packaged commu- nications lasers, the output available at the fiber pigtail is typically 1 mW average during modulation (O dBm). This is lower than the maximum facet power due to a 4- to 5-dB laser-to- fiber coupling loss and the need to operate the laser at less than its maximum power to maintain good reliability. The O-dBm output is adequate for systems operating up to several hundred megabits per second with less than 30-km repeater spacings, but only marginal for bit rates over 1 Gbit/s at 30 km or for systems (e.g., undersea cables) requiring long repeater spacings. Modulation Bandwidth Nearly all lasers available today can be modulated at 500 MHz. This is easily adequate for most systems in use. With minor optimization, most laser designs can be stretched to the 1.7-Gbit/s rate that is the fastest system com- mercially available. Laboratory tests of specially optimized laser designs have been reported in excess of 5 GHz, whereas the world record (obtained at 77 K) is 36 GHz (Bowers, 1985~. Re/iabi/ity The most reliable lasers are used for undersea applications. Such lasers typically have a mean life in excess of 100,000 hours at room temperature. In terrestrial systems lasers with somewhat shorter lifetimes can often be used and are 1 1

OCR for page 45
52 C. KUMAR N. PATEL usually available for a significantly lower cost. Such lifetimes are nevertheless considerably less than those usually specified for most telephone equipment, for which a million hours is a typical desirable mean lifetime. Reliability is a severe problem for lasers that must withstand temperatures up to 70C. These tempera- tures typically are used for accelerated aging tests, and lifetimes of 1,000 hours or less are considered good by today's standards. Clearly, there is room for improvement in high-temperature reliability. Such conditions can be met today only with thermo- electric coolers in the laser package. Frequency Spectrum Nearly all lasers sold today operate in the fundamental transverse mode but with multiple longitudinal modes. The mean spectral width is typically about 5 nm and consists of several longitudinal modes spaced by roughly 1 nm. For 1,300-nm lasers, such a spectral width is acceptable in most systems because the fiber dispersion crosses zero at this wave- length (Appendix A). However, for 1,550-nm lasers that use the low-loss window in silica fibers, such a multifrequency spectrum is unacceptable for information systems operating faster than several tens of megabits per second because of the significant fiber dispersion at this wavelength. Therefore, single-frequency lasers have recently been devel- oped that meet system needs at 1,550 nm. A few vendors now offer distributed feedback (DFB) lasers for systems requiring a single-frequency laser (Kogelnik and Shank, 19711. DEB lasers are also useful at 1,300 nm for systems operating above 1 Gbit/s. The linewidth of a single longitudinal mode of a semiconductor is about 100 MHz at 1 mW of continuous wave output. This is determined by the laser cavity length of 300 ,um. However, under modulation, the linewidth broadens to roughly 10 GHz because of frequency chirp as the laser current is changed (Olsson et al., 1984~. This chirp effect will be a problem for future 1,550-nm long-haul systems operating beyond 1 Gbit/s even with single-frequency lasers. Cost Communications lasers currently cost about $1,000 or more, depending on their power, wavelength, bandwidth, spec- tral purity, and reliability. This cost is acceptable for high- capacity trunk systems but is at least an order of magnitude too high for use in local systems between homes and offices. The price is high because it must cover the sophisticated testing needed to ensure reliability and, to a lesser degree, the cost of the mechanical package. That the AlGaAs lasers for compact audio disc (CD) players now cost roughly $10 gives hope for a

OCR for page 45
LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 53 considerable reduction of cost in future InGaAsP communica- tions lasers as well. The extreme cost reduction obtained for CD lasers has been due largely to a highly uniform materials technology, modest reliability needs, a simple package, and limited testing before final assembly. Some, but not all, of these factors may translate to communications lasers, so that a goal of lowering costs to $100 might not be unreasonable in the future. FUTURE TRENDS IN LASERS The future trends in communications lasers can be described by a brief summary of the current frontiers of semiconductor laser research. The major research areas today are materials, fre- quency control, linewidth, and integration. Materials Nearly all InGaAsP lasers sold today are produced by liquid-phase epitaxy (LPE) (Casey and Panish, 19781. This is a convenient method for laboratory research and has been successfully scaled up for production. However, the epitaxial layers grown in this way are not as uniform as those possible with newer growth techniques, such as metal-organic chemical vapor deposition (MOCVD) (Dupuis, 1984) or molecular beam epitaxy (MBE) (Cho, 1983; Cho and Arthur, 1975~. It is generally believed that MOCVD and MBE offer the potential of higher manufacturing yield and hence lower costs. Thus, there is considerable research into these growth methods. Perhaps the most exotic and promising of the new methods is gas-source MBE (Parish and Temkin, 1985~. This was devel- oped because the conventional MBE technique using solid elemental sources, such as gallium or arsenic, could not grow good-quality material that contained both arsenic and phospho- rus. The fact that gas-source MBE can produce atomically sharp interfaces between layers of different compositions and hence band gaps (see Figure 5) gives the potential for a rich array of novel structures. It is well known that AlGaAs lasers grown by MBE using tailored band gaps and multi-quantum wells (MQWs) give superior performance (Tsang, 1981~; hence, it is expected that such techniques will also be beneficial when applied to InGaAsP lasers. New materials are important for semiconductor lasers oper- ating at wavelengths beyond 1,550 nm. Fiber research is cur- rently focusing on new materials in the search for ultralow-loss fiber in the wavelength range between 2 and 5 ,um (see Appen- dix A). Lasers are being studied in this region as well. In general, the materials systems being studied are based on either GaSb or 1 1

OCR for page 45
54 C.KUMARN.PATEL InP _ Eg= 1.1 ,um 1.3pm 1.5 pm TERNARY a 1 BOA 50A -~ lDda8al a (a) (bl lc} ~1 02000A . ~InP ~ r]500A ~ r3000A E3=~.] Ilm Ll ~ ~L: t. ~ {e} If ~ FIGURE 5 Different types of multi-quantum well laser structures fab- ricated using molecular beam epitaxy. InAs, since InP-based materials cannot operate beyond about 1,650 nm. The best result to date for room-temperature oper- ation of a continuous wave laser is slightly beyond 2 ,um using LPE growth in a GaSb-based system (Caneau et al., 19851. Considerable effort in the 3- to 4-,am range in InAs-based sys- tems has produced lasers at 77 K, but none at room temperature. This trend toward poorer temperature performance with longer wavelengths is expected theoretically; it may ultimately limit the commercial appeal of this wavelength range for all but special situations in which cryogenic laser packages are acceptable. Frequency Contro/ The recent trend toward single-frequency lasers, such as DFB and cleaved-coupled-cavity (C3) lasers (Tsang et al., 1983), is the best example of this sort of research. In recent DFB laser research, unwanted longitudinal modes have been suppressed more than 40,000 to 1 relative to the main mode (Tsang et al., 1985~. Gratings for DFB lasers are typically fabricated by holographic photolithography. However, recent results (Temkin et al., 1985) with electron beam lithography promise even higher resolution and greater control over the detailed shape of the grating (see Figure 6~. In spite of the recent progress with single-frequency lasers, much still remains to be done. In particular, even though the longitudinal mode control is excellent, the fabrication process cannot yet be sufficiently controlled to set absolute frequency standards, or channels, with a precision and reproducibility anywhere near the linewidth of the laser.

OCR for page 45
LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 55 .... FIGURE 6 Distributed feedback grating. tin e width Although the single-longitudinal-mode, 1.5-,um la- sers have dramatically improved the performance of fiber-optic communication systems, they are not without problems. Under continuous wave operation, a typical DEB (300 ,um long) or C3 laser has a linewidth of about 100 MHz. Under direct amplitude modulation, however, the linewidth is broadened, or chirped,

OCR for page 45
90 C. KUMAR N. PATEL - [~M H~Ull.~ fop '~~ G80 47In 0 53 A' (a) c, J cat - - ~u l I . I I 111111 1' 1 bP: . I no | a+| `[ I _6aO 47lnO Il348~j \ \ \ \ o STANCE (b) (c) FIGURE B-2 Band diagram (a), device structure (b), and electric-field profile (c) of an SAM avalanche photodiode with chirped superlattice to eliminate carrier pileup at the InP/GaInAs interface. the electric field at the heterointerface is significantly smaller than in a conventional SAM APD, thus reducing the dark current to lower values. This effect was demonstrated in AlIn- As/GaInAs Hi-Lo SAM APDs grown by molecular beam epitaxy (MBE) and in InP/GaInAs Hi-Lo SAM APDs grown by liquid- phase epitaxy. SAM APDs have also been studied extensively in AlInAs/ GaInAs alloys. A potential advantage of this combination is that the band discontinuities are more favorably aligned for high- speed operation. In fact, AlInAs/GaInAs SAM APDs without intermediate grading layers have demonstrated a speed of response comparable to that of InP/GaInAs SAM APDs with grading layers (Capasso et al., 1984b). The quality of the AlInAs must be improved, however, before these detectors can become a real challenge for InP-based APDs. 11

OCR for page 45
LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 91 ADVANCED AVALANCHE PHOTODIODES AND SOLID-STA TE PHOTOMUL TIPLIERS The multiplication noise of an avalanche photodiode is known to increase strongly, at a given value of the gain M, as the ratio of the electron/hole ionization coefficients K = ~/p approaches unity. In fact, it can be shown that in this limit, the APD noise is proportional to M3, whereas in the opposite ideal limit, in which only one carrier can ionize, the noise increases as a function of M2 (McIntyre, 1966~. Most III-V alloys, including InP used in SAM APDs, have an ionization rate ratio (cz/,B or p/~) in the range from 1 to 3, and as such are unsuitable to achieve the low noise performance of Si APDs at shorter wavelengths (in Si, c~/,8 ' 20~. Research has concentrated on multilayer structures capable of artificially enhancing c}l,l3 using material systems with ~ approx- imately equal to ,l3. These efforts have led Capasso (1985) to the concept of a solid-state photomultiplier. In 1981 a group at Bell Laboratories showed that in an MBE-grown AlGaAs/GaAs quantum well APD the a/,8 ratio is enhanced by a factor of 4 over the bulk value for c~l,B GaAs (Capasso, 19851. This effect is partially due to the difference between the conduction and valence band discontinuities. At the University of Michigan, more recent extensive work on such structures has shown the above effect for a greater range of layer thicknesses Juang et al., 1985~. A potential problem in this structure is that the pileup of carriers in the quantum wells may deteriorate the pulse re- sponse. Recently, however, Mohammed and colleagues showed this is not a problem and demonstrated response times of less than 200 ps in AlInAs/GaInAs quantum well APDs (Mohammed et al., 1986~. This is due to hot electron effects and tunneling through the barriers. Another structure designed to enhance the o/,(3 ratio is a PIN APD with a graded gap in the i region. Electrons, which move toward lower gap regions than holes, "see" a lower ionization energy, and this effect enhances the c~l,8 ratio (Figure Bed. Ionization ratios of 5 to 7 have been demonstrated in an AlGaAs prototype structure (Capasso, 19851. Yet another approach is the channeling APD in which the ~/,8 ratio is enhanced by spatially separating electrons and holes in materials of different band gaps (Capasso, 1985~. This is done using an interdigitated npnp lateral structure (Figure B-4~. This new depletion and detection scheme has several other advan- tages (experimentally demonstrated), such as the extremely low capacitance that is independent of the sensitive area of the detector and the large volume of depleted material. Interest-

OCR for page 45
92 C. KUMAR N. PATEL FIGURE B-3 Band diagram of high-field region of a graded gap avalanche photo- diode. The 1-1' electron-hole pair initiates avalanche multi- plication as follows: Electron 1 creates by impact ionization the electron-hole pair 2-2' in the lower gap region. Hole 1' creates the electron-hole pair 3-3' in the higher gap region. Thus, the electron has a lower ionization energy than the hole. 3' \ \ 1\ ~1 ,~ 2' ingly, the channeling detector concept has found important applications in nuclear physics as a position-sensitive drift cham- ber to detect high-energy particles (Gatti and Rehak, 1984~. Probably the most promising of these structures for optical communications is the staircase avalanche photodiode, which is the solid-state analog of a photomultiplier (Capasso, 1985~. The structure consists of a graded gap superlattice low-doped layer sandwiched between a p+ and n+ layer. When a reverse bias is applied, the sawtooth potential profile is converted in a potential staircase. The materials are chosen in such a way that the magnitude of the step (conduction band discontinuity) is greater than the gap after the step, and the valence band step is negligible (Figure B-5~. Electrons impact-ionize only at the steps because only there does their kinetic energy exceed the band gap, whereas holes do not ionize. Capasso has shown theoreti- cally that the excess noise factor F for such a structure is practically unity, similar to a phototube (Capasso, 1985~. No other type of APD, including an ideal conventional one in which only one type of carrier can ionize, has this unique property. In

OCR for page 45
LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 93 fact, until recently, the lower theoretical limit for F at high gains was thought to be 2. One additional advantage of the staircase APD is the low-voltage operation (~5 V for a gain of ~30~. The materials that are investigated for this application are HgCdTe and AlGaAs/GaSb grown by MBE. Experimental demonstra- tions have not yet been reported. Clearly, the staircase detector has the potential for achieving unprecedented receiver sensitiv- ities at both high and low bit rates, provided one can minimize the dark current of the device. Another approach to the solid-state photomultiplier is based on a recently discovered avalanche multiplication mechanism (impact ionization across the band-edge discontinuity (Capasso et al., 19861. In suitably designed superlattice structures, hot carriers in the barrier layers can collide with carriers confined or dynamically stored in the wells and impact-ionize them out across the band-edge discontinuity (Figure Bob. In this ioniza- tion effect, only one type of carrier is created, so that the positive feedback is eliminated, leading to the possibility of a quiet avalanche with small excess noise. A multistage graded gap avalanche photodiode based on this concept has been demon- strated, and it exhibits a near single-carrier-type multiplication, similar to a photomultiplier (Allam et al., 1987~. \~ \ \~ \~\ - \~ W9, _ , 6,~e EC EV FIGURE B-4 Band diagram of the channeling avalanche photodiode.

OCR for page 45
94 C. KUMAR N. PATEL p he it, - ~ i. _ ~~."' (/`Er (a) 4^Ev - \\ ( b ) n FIGURE B-5 Band diagram of the staircase solid-state photomultiplier. (a) shows the unbiased graded multilayer region, and (b) shows the complete staircase detector under bias. The arrows in the valence band indicate that the holes do not impact-ionize; hole multiplication due to electron-initiated impact ionization is not shown for simplicity. Recently, an SAM APD with an Si-Ge superlattice absorbing layer and a Si multiplication region has been demonstrated (Temkin et al., 1986~. To maximize absorption without degrad- ing high-speed operation, the device has a waveguide geometry (lateral illumination) (Figure B-7~. The device has potential for achieving the low multiplication noise of silicon at long wave- lengths, but part of this advantage is offset by the relatively large coupling losses and other technological difficulties associated with the lateral illumination. One of the most interesting appli- cations of this device is for integrated optics. At compositions such that the spin-orbit splitting approxi- mately equals the band-gap in certain alloys (AlxGa~-xsb, 1 1

OCR for page 45
LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 95 - - FIGURE B-6 Band diagram of solid-state photomultiplier based on impact ionization across the band-edge discontinuity of carriers stored in the wells. Hg~_xCdxTe), the ionization rates ,`3/a ratio attains a large value (10-20) because of the near-zero momentum transfer in the ionizing collisions of holes (resonance impact-ionization) (Ca- passo, 1985~. These compositions correspond to band gaps Ag ~ 1.3-1.5 ,um and thus may be suitable for avalanche detectors for communication systems. HgCdTe appears particularly promis- ing from this point of view because the dark currents are much FIGURE B-7 Schematic diagram of Gex-si~-x long-wavelength wave- guide superlattice avalanche PIN photodiode.

OCR for page 45
96 C. KUMAR N. PATEL lower than the corresponding AlGaSb alloy of the same gap. Societe Anonyme de Telecommunications in France has already developed HgCdTe PIN detectors at A = 1.3 ,um with dark currents of approximately 1 nA and plans to have a working low-noise 1.3-,um APD using the above resonance effect in the near future. PHOTOCONDUCTORS In recent years, GaO.47In0.53As photoconductors have attracted considerable attention as possible alternatives to PIN and ava- lanche detectors in the 1.3-1.6-,um wavelength regions. The best results so far obtained at bit rates of 1 Gbit/s are 1 or 2 dB lower in sensitivity than the best PIN-FET receivers (Chen et al., 1984~. Extensive theoretical analyses at the AT&T Bell Labora- tories have shown that in the above wavelength range and at bit rates ranging from 500 Mbit/s to 2 Gbit/s, the photoconductor can, at best, match the performance of a PIN in a receiver but never do better than an APD (Forrest, 1985~. On the other hand, the photoconductor has advantages of very low voltage operation and easy fabrication technology. These features may ~ n n n ~ ~ E] EF ;/~/7~///7/] U U ~ U ~~= ~ I_ ~ . ~ ,_ , ~ J U U UP ~ .' tang Fox, ma_ L_ I I he I ~ I - r ;rl 1 FIGURE B-8 Band diagram of superlattice photoconductor; shown the effective mass filtering mechanism. 1S

OCR for page 45
LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 97 WAVE LENGTH (~m) 10 4 1.8 1.6 1.4 1.2 103 z 1 0 IL Cad ~ 10 111 of a ~ 10-2 111 10-3 10 -4 10-5 1 .0 , , , , , , ~ , ~ - ~ 4V (BOOK) _ _ _ f i r ~ f f E/gOK (SUPERLATTICE) _ i 2x 10-5 V (COOK) - ~. ~ ~ ~ 5 x 10-3V (300K) _ tl f EgOOK (GaO 47lno S3AS) gK (SUPERLATTICE) / EgOK / (G0O.471nO: I 1 1 1 1 ~ 1 0.7 0.8 0.9 1.0 1.1 1.2 PHOTON ENERGY ( ev ) 2X10-5 V (70K) _ 1 1 1 FIGURE B-9 Current gain in an effective mass filter detector. be attractive for applications in which one is willing to trade performance against cost, such as local area networks. In addi- tion, the lateral geometry makes the photoconductor particu- larly attractive for monolithic integration with an FET. Recently, Capasso and coworkers (1985) have demonstrated a new type of photoconductor called an effective mass filter. This structure consists of a superlattice with thin layers (30-A barrier and 30-A wells) and achieves high gain and low noise at very low voltages (<0.2 V) using the large difference between the tun- neling rates of electrons and holes through the superlattice barriers (Figures B-8 and Bob. Although the most recently demonstrated device is slow, another variety of effective mass filter that uses miniband conduction of electrons rather than phonon-assisted tunneling has potential for high-speed, low- noise, and low-voltage operation. These features may make such detectors attractive for optical communication systems, particu- larly at wavelengths beyond 1.5 Am.

OCR for page 45
98 C. KUMAR N. PATEL REFERENCES Allam, I., F. Capasso, K. Alavi, and A. Y. Cho. In press. Proceedings of the GaAs Symposium. Las Vegas, Nev. Capasso, F. 1985. The physics of avalanche photodiodes. P. 1 in Semiconductors and Semimetals, Vol. 22, Lightwave Communications Technology: Part A, Material Growth Technologies, W. T. Tsang, ed. New York: Academic Press. Capasso, F., A. Y. Cho, and P. W. Foy. 1984a. Electron. Lett. 20:635. Capasso, F., B. Kasper, K. Alavi, A. Y. Cho, and I. M. Parsey. 1984b. Appl. Phys. Lett. 44:1027. Capasso, F., I. Allam, A. Y. Cho, K. Mohammed, R. l. Malik, A. L. Hutchinson, and D. Sivco. 1986. Appl. Phys. Lett. 48:1294. Chen, C. Y., B. L. Kasper, and H. M. Cox. 1984. Appl. Phys. Lett. 44:1142. Forrest, S. R. 1985. P. 329 in Semiconductors and Semimetals, Vol. 22, Lightwave Communications Technology: Part A, Material Growth Technolo- gies, W. T. Tsang, ed. New York: Academic Press. Gatti, E., and P. Rehak. 1984. Nuclear Instr. Methods 225:608. Holden, W. S., J. C. Campbell, I. F. Ferguson, A. G. Dental, and Y. K. thee. 1985. Electron. Lett. 22:886. Juang, F. Y., U. Das, Y. Nashimoto, and P. K. Bhattacharya. 1985. Appl. Phys. Lett. 47:972. Kasper, B. L. 1986. P. 119 in Technical Digest of the Optical Fiber Communi- cations Conference. Atlanta, Ga. McIntyre, R. I. 1966. IEEE Trans. Electron. Devices ED- 13: 164. Mohammed, K., F. Capasso, I. Allam, A. Y. Cho, and A. L. Hutchinson. 1986. Appl. Phys. Lett. 47:597. Pearsall, T. P., and M. A. Pollack. 1985. P. 174 in Semiconductors and Semimetals, Vol. 22, Lightwave Communications Technology: Part A, Mate- rial Growth Technologies, W. T. Tsang, ed. New York: Academic Press. Temkin, H., T. P. Pearsall, J. C. Bean, R. A. Logan, and S. Luryi. 1986. Appl. Phys. Lett. 48:963. APPENDIX C: COHERENT SYSTEM EXPERIMENT 1 Conventional direct-detection lightwave receivers are limited in their performance by thermal noise. The only way to circumvent this problem is to amplify the signal without adding excess noise. One way to achieve this amplification is by heterodyne gain: The incoming optical signal is mixed with a local oscillator, and the beat signal, which contains the information, is multiplied by the local oscillator. Such systems, using a local oscillator, are called coherent systems. The principle is illustrated in Figure C-1, which shows the improvement in receiver sensitivity for a 150-Mbit/s coherent system by increasing the local oscillator power (N. A. Olsson, personal communication). As the local oscillator power is increased, the receiver sensitivity approaches the fundamental shot noise limit. However, because of nonideal components, such as the quantum efficiency of the detectors, the ultimate shot noise limit is hard to reach.

OCR for page 45
LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 99 -60 ~ -55 z -50 LLI (n -45 > Cal ~ 40 -35 _ . _ _ ~ /~ _ _ ~ ~ _ = i~ = ~ _ _ 4r -25 - 20 -15 -10 -5 0 LOCAL OSCILLATOR POWER (dBm) SHOT NOISE LIMIT Dl RECT DET ECT 10N FIGURE C-1 Sensitivity improvement with coherent detection at a bit rate of 150 Mbit/s. The coherent lightwave system experiment described here used differential phase shift keying (DPSK) at data rates of 400 Mbit/s and 1 Gbit/s and a transmission distance of 150 km. The system is depicted in Figure C-2 (Linke et al., 19861. The trans- mitter and local oscillator lasers are external cavity lasers. Phase modulation of the optical carrier was achieved with a titanium- diffused LiNbO3 waveguide phase modulator (Schmidt and Cross, 1978~. The modulator had an insertion loss of 1.8 dB and required a modulation voltage of 8.5 V peak to peak for a 180-degree phase shift. After transmission through 150 km of fiber with a transmission loss of 39.6 dB, the transmitted signal is mixed with the local oscillator in a 3-dB fiber coupler. The balanced-mixer dual-detector receiver efficiently uses the avail- able local oscillator and signal power and suppresses any excess amplitude noise in the local oscillator. The equalized bandwidth of the high-impedance front end was more than 3.5 GHz. The intermediate-frequency (IF) signal was processed in a delay line discriminator, and part of the IF signal was used in a feedback circuit that frequency-locked the local oscillator laser to the incoming data signal. The system was evaluated by measuring the bit-error rate as a function of the received power. In both cases the error rate could be decreased to arbitrarily low levels (measured down to 1 x 10-~) by increasing the received power. The absence of an error floor is the absolute proof of the spectral purity and low phase noise of the external cavity lasers

OCR for page 45
00 C. KUMAR N. PATEL EXTERNAL GRATING LASER PHASE MODULATORS_ ' - ' T 150km DIFFERENTIAL ~ RECTOR ENCODER ~ DATA IN RF ~ ~ F ~ 'BALANCED Y VARIABLE 't DELAY DATA OUT ~ - ] ?~ ADJUSTER ~ r 3dB OPTICAL | Em PER EQUAL IZER ~ FIGURE C-2 Experimental setup for coherent detection lightwave demonstration. used. The measured receiver sensitivity at 400 Mbit/s and 1 Gbit/s was55.3 and44 dBm, respectively. In the ideal shot noise limited case, DESK modulation requires 21 photons per bit for a 1 x 10-9 error rate. The measured system performance was 6.4 and 11 dB from this theoretical limit at 400 Mbit/s and 1 Gbit/s, respectively. Part of the discrepancy is accounted for by the thermal noise of the receiver, and by the less-than-unity quantum efficiency of the photodetectors. In spite of the devia- tion from the ideal shot noise limit, the measured receiver sensi- tivities are the best reported for the respective data rate and are about six times better than the best reported direct detection sensitivities. This coherent system experiment is the first gigabit- per-second system and the first time a coherent system has outper- formed the direct detection counterpart in transmission distance. REFERENCES Linke, R. A., B. L. Kasper, N. A. Olsson, and R. C. Alferness. 1986. Electron. Lett. 22:3~31. Schmidt, R. V., and P. C. Cross. 1978. Opt. Lett. 2:45-47.