National Academies Press: OpenBook

Lasers: Invention to Application (1987)

Chapter: Lasers in Communications and Information Processing

« Previous: Lasers in Modern Industries
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 45
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 46
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 47
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 48
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 49
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 50
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 51
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 52
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 53
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 54
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 55
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 56
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 57
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 58
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 59
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 60
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 61
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 62
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 63
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 64
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 65
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 66
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 67
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 68
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 69
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 70
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 71
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 72
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 73
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 74
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 75
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 76
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 77
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 78
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 79
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 80
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 81
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 82
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 83
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 84
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 85
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 86
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 87
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 88
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 89
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 90
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 91
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 92
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 93
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 94
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 95
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 96
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 97
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 98
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 99
Suggested Citation:"Lasers in Communications and Information Processing." National Academy of Engineering. 1987. Lasers: Invention to Application. Washington, DC: The National Academies Press. doi: 10.17226/1003.
×
Page 100

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

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

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

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

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-

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

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.

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

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 70°C. 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

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

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.

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,

56 C. KUMAR N. PATEL (a) (b) FIGURE 7 Chirp-induced broadening of laser linewidth (l-A/div hori- zontal scale): with modulation (a); without modulation (b). as a result of the carrier density dynamics during modulation of the laser (Olsson et al., 1984~. Typically, this chirp gives a modulated linewidth of approximately 10 GHz (1 A), as shown in Figure 7. For data rates below l Gbit/s, the chirping is of little consequence to communication systems. For the very high bit-rate systems, however, the wavelength chirping of the lasers, coupled with the dispersion of the fiber, can give a substantial penalty. For example, in the terabit-km/s experiment described later in this paper, the 10 channels incurred chirp penalties between 1 and 3.5 dB. That is, the receiver needed between 25 and 225 percent more power to achieve a given error rate than would have been required if the laser did not chirp. Two methods have been demonstrated to solve the chirp problem: injection locking and external modulation (Olsson et al., 1985~. Injection locking has been used in a 2-Gbit/s system, and external modulation has been used in a 4-Gbit/s experiment. Both experiments demonstrated the elimination of any chirp- related penalties. With these techniques, injection locking and external modulation, the optical communication systems have become so sophisticated that, for the first time, the linewidth of the transmitted signal is given by the information bandwidth. The coding of the information, however, is still the most primitive (and simplest) possible; only the energy in the trans- mitted pulses is detected. The next step in refinement would be a system in which the phase, frequency, and polarization of the optical wave are significant. This leads us to coherent optical communication

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 57 systems that put even greater demands on the lasers. For successful operation of a coherent optical communication sys- tem, the phase noise of the lasers must be minimal. This severely restricts the linewidth of both the transmitter and the local oscillator laser. In direct detection systems, the chirp penalty is eliminated for laser linewidths equal to or less than the in- formation bandwidth. Coherent systems, in contrast, require linewidths that are only 1/30 to 1/1,000 of the information bandwidth. Phase-shift keyed homodyne systems require the narrowest linewidth, and amplitude-shift keyed heterodyne sys- tems are the most linewidth tolerant. DEB and C3 lasers, although operating in a single longitudinal mode, do not have sufficiently narrow linewidths or good frequency stability for coherent applications. By operating the semiconductor laser in an external cavity, the Q. or "quality factor," of the longitudinal modes is increased, and the linewidth of the laser is dramatically decreased. Figure 8 shows the beat spectrum of two 1.5-,um external cavity lasers. The width of the beat spectrum 60 dB down from the peak is 4 MHz, which indicates full width at half-maximum (FWHM) spectral width of 2 kHz for each laser (Olsson and van der Ziel, 1987~. i 1 ~ ~ 1 1 1 1 J ~ _. __ 1 1 1 1 1 1 FREQUENCY 2 MHZ /d iv FIGURE 8 Beat spectrum of two 1.5-,um external cavity lasers. 11

58 C. KUMAR N. PATEL DETECTORS FIGURE 9 Optical electronic integrated circuit. (a) Laser; (b) photode- tector; (c) GaAs; (d) GaAlAs; (e) GaAlAs (active layer); (f) GaAlAs; (g) GaAs (conducting layer); and (h) semi-insulating GaAs. Integration Many investigators believe that the ultimate scheme to reduce cost and improve performance is opto-elec- tronic integrated circuits (OEICs) (Matsueda et al., 1984J. Con- siderable work is being done on this subject in Japan. Figure 9 shows an example of an optical electronic circuit of this type a laser monolithically integrated with a small amplifier circuit. Such OEICs range from simple laser arrays to proposals for integrated regenerators and complete integrated optical elec- tronic wideband switching modules. Such complex circuits will probably not be actually used for some time, but the potential for significant cost reduction in high-volume products is a major force behind such work. In the last few years, along with the parallel efforts in reducing the optical fiber losses and improving semiconductor laser performance, there has been an intense research and develop- ment effort on photodetectors for optical communications. Several important technological and material developments have driven and motivated this effort: (1) the development of low-loss, low-dispersion fibers at 1.3- to 1.55-,um wavelengths; (2) the availability of high-quality and low-doped epitaxial alloys (InGaAsP) and InP grown by liquid-phase and vapor-phase

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 59 epitaxy (Stillman et al., 1983~; and (3) the impressive advances in heterojunction research. These advances include the growth of high-quality heterojunctions and superlattices by MBE (Cho, 1983; Cho and Arthur, 1975) and MOCVD (Dupuis, 1984~; better experimental (Margaritondo, 1986) and theoretical (Har- rison, 1985) understanding of band discontinuities and perpen- dicular transport in heterostructures (Capasso et al., 1986~; and the realization of the enormous degrees of freedom afforded by heterojunction and superlattices in device design (band-gap engi- neering) (Capasso, 1985~. The detectors used in optical communi- cations are PIN photodiodes, avalanche detectors, and photocon- ductors (see Appendix B). STATUS OF LIGH1WAVE COMMUNICATIONS The light sources used in today's optical communication sys- tems, regardless of type, differ little in wavelength, output power, modulation bandwidth, or electrical power require- ments. These quantities vary at most by 2 orders of magnitude between the different types of light sources. However, a varia- tion of 10 orders of magnitude is seen in the emission linewidth of the light sources. A typical light-emitting diode (LED) has an emission linewidth of 1,000 A (2 x 10~3 Hz) at a 1.3-,um wave- length. A 1.5-,um external cavity semiconductor laser, on the other hand, has a linewidth of a few kilohertz (Olsson and van der Ziel, 19871. Table 1 lists the most commonly used light sources and their respective linewidths. Not surprisingly, this large variation in emission linewidth has a profound effect on the performance of optical communication systems. In effect, the evolution of the optical communication technology has been the record of a quest to reduce and control the linewidth of the light source. One example of this is a 20-Mbit/s data rate system using an LED with a linewidth of 1,000 A. In this case, the linewidth is a million times larger than the information bandwidth, hardly an efficient ap- TABLE ~ Linewi~ths of Important Light Sources Used in Lightwave Communications Device Linewidth (Hz) 013 l2 108 103 Light-emitting diode Multimode laser Single-mode laser External cavity laser _~

60 C. KUMAR N. PATEL proach and not much different from Hertz's first radio transmis- sion experiments using an open-air spark as the transmitter. The following section describes the first lightwave communi- cation system to reach an information-carrying capacity of more than l terabit-km/s. This is followed by a discussion of the system limitations imposed by the linewidth of the light sources listed in Table 1 and by examples of current state-of-the-art perform- ance. See Appendix C for a description of a coherent lightwave communication system approaching the theoretical limit of performance. TERABIT-KM/S EXPERIMENT In this system demonstration (Hegarty et al., 1985) of close- spaced wavelength division multiplexing with ultrahigh capac- ity, 10 single-frequency DEB lasers were multiplexed into a single 68-km transmission fiber. The experimental details are shown in Figure 10. L I DE MUX 1 ~ DET L7~ L 8~L~ l L 10 ANRITSU ~ 2 TE ST SET r Gb Is MUX V ~ FIGURE 10 Setup for a 10-laser wavelength division multiplex light- wave system.

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 61 FIGURE 11 Wavelength division multiplexer. The lasers employed in this experiment were heteroepitaxi- ally ridge overgrown (HRO) DEB lasers (Tsang et al., 1985~. The pure single-mode operation of these lasers, even under high-speed modulation, was essential for achieving the narrow channel spacing, low cross talk, and error-free operation of the system. The lasers used a second-order grating, and the mode rejection ratio—the ratio of the dominant mode power to the power in the next largest mode was between 400:1 and 10,000:1 at a 2-Gbit/s modulation. The laser wavelengths were between 1.53 and 1.56 ,um. The multiplexer (Figure 11) had 22 channels, consisting of 23 single-mode fibers brought together in a linear array. The core-to-core spacing in the array was 24 ,um, and the free ends of the 22 input fibers were pigtailed with microlenses for coupling to the lasers (Plate 3~. The remaining fiber was the output channel, and its free end was spliced to the transmission fiber. Coupling between the input fibers and the output fiber was achieved with a 2.5-cm lens and a 600 1/mm grating. The resulting channel spacing was 13.5 A, and the average coupling loss between the input and output fiber was 3 dB. Demultiplex- ing at the end of the 68-km transmission fiber was achieved with a grating and with the receiver photodetector acting as the spatial filter. The overall cross talk between adjacent channels

62 C. 1 KUMAR N. PATEl 10-4 10-5 10-6 1 fir o ~ 10-7 m lo-8 10-9 10- 10 T ~it/s683!m ; at: - 31 -29 -27 - 25 -23 RECEIVED POWER (dBm) FIGURE 12 Received bit error rates for channels 5 and 9 of the wavelength division multiplex system. was less than -23 dB, with most of the cross talk originating in the demultiplexer. The almost pure silica-core transmission fiber had an average loss (including splice losses) of 0.22 dB/km for the 10 lasers, and the dispersion at 1.55 ,um was 19 ps/km nm. The receiver used an avalanche photodetector, and the receiver sensitivity at 2 Gbit/s was—32 dBm. The system performance was evaluated with all 10 lasers providing full power into the fiber and by sequentially applying a 2-Gbit/s pseudorandom nonreturn to zero modulation to each laser. The received bit pattern was

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 63 compared with the transmitted pattern, and the bit error rate (BER) was recorded as a function of received power. For all channels, a BER of less than 1 x 10-9 was achieved, and the BER was independent of the presence of the other channels. BER curves, a recording of the system error rate versus the received power, for the channels requiring the least (channel 5) and the most (channel 9) power for obtaining a 109 BER is shown in Figure 12. The difference between the two channels is mainly due to the laser chirping effect. To measure the cross talk, the laser corresponding to the selected channel was turned off, and the photocurrent resulting from all the other nine channels was measured. The ratio of this photocurrent to the signal photocurrent, when the laser for this channel was oper- ated, was less than—23 dB (0.5 percent). This level of cross talk is negligible as confirmed with the BER measurements. This system demonstration is significant in several respects. It is the first optical communication system that is even attempting to get close to the possible capacity of fiber-optic systems. This system uses a third of the bandwidth of the best LED system but has more than 200 times the capacity. It is also the first demonstra- tion of close-spaced wavelength division multiplexing, a neces- sary technique if the ultimate capacity of the fiber is going to be reached. The demonstrated capacity corresponds to 21 million voice channels-kilometers. When buying a 300-page book for $10, the price is approximately $1.5 x 10-6 per bit of informa- tion. At this price per bit, 20 Gbit/s for 1 year corresponds to $1 trillion. The first optical communication systems to be developed, LED-based systems using multimode fiber, are in general lim- ited by the modal dispersion of the fiber. A typical system is the SLIC-96 loop feeder operating at a data rate of 90 Mbit/s and a repeater spacing of 20 km. The problem of modal dispersion was solved by introducing single-mode fiber. Because of the large spot size of regular surface-emitting LEDs, only a small fraction of the light can be coupled into the fiber. This, plus the increased output power and modulation bandwidth available from lasers, spurred the development of 1.3-,um wavelength lasers. However, with special edge-emitting LEDs with small spot size, as much as 30 ,uW of power has been coupled into a single-mode fiber, and a transmission distance of 35 km at 180 Mbit/s has been demonstrated. Conventional semiconductor lasers emit light in a few (3-10) longitudinal modes spaced about 11

64 C. KUMAR N. PATEL 30 20 10 by ~ ~ o He o In ~ -10 _ In -20 _ _ -3O' 1 Or _ _ I ~ ~ r N;~ ....... \ QUADRUPLE- CLAD CONVENTIONAL 1 1 1 1 ~ 1.7 1.8 1.2 1.3 1.4 1.5 1.6 WAVELENGTH (~m) FIGURE 13 Chromatic dispersion in single-mode silica fibers as a function of wavelength. The solid line is for conventional step index fiber. Note the zero dispersion region near 1.3 ,um, which permits the use of multimode laser sources without sacrificing bit rate or distance between repeaters. The dotted curve is the measured dispersion of an experimental quadruply clad fiber, with an index profile shown in the inset. 10 A apart. The linewidth of a laser with five modes is 10~2 Hz at 1.3 ,um, which is the first loss minimum of silica optical fibers, as seen in Figure 2. Fortunately, at a 1.3-,um wavelength, the chromatic and waveguide dispersion in the fiber cancels (Figure 13 i. While the broad linewidth of the source is a terrible waste of available bandwidth, the large linewidth does not impose a dispersion penalty at 1.3 ,um, and the transmission systems are loss limited. For example, the transatlantic undersea cable TAT-8 (see Figure 14) uses 1.3-am lasers at a data rate of 296 Mbit/s and a repeater spacing of about 50 km. At 1.5 ,um, the transmission loss of the optical fiber is only about half of that at 1.3 ,um; therefore, in principle, the repeater spacing can be doubled by switching to a 1.5~ m light source in the transmitter. The dispersion, however, which was minimal at 1.3 ,um, is substantial typically, 18 ps/km nm at a 1.5-,um wavelength. The hard-sought breakthrough was the achieve- ment of single-longitudinal-mode operation of 1.5-,um wave- length lasers. Some conventional semiconductor lasers emit

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 65 predominantly in a single longitudinal mode. Under continuous wave operation, as much as 99 percent of the output power may be concentrated in one mode. Continuous wave measurements, however, are poor indicators of how the laser will perform in a communication system. The requirement for optical communi- cation systems is that the laser emits in one longitudinal mode all the time and under amplitude modulation. No conventional Fabry-Perot semiconductor laser fulfills this requirement. As described above, the two most successful laser structures for achieving dynamic single-mode operation are the distributed feedback (DFB) (Kogelnik and Shank, 1971) and cleaved- coupled-cavity (C3) (Tsang et al., 1983) lasers. With the advent of these narrow linewidth lasers, the improvement in capacity, transmission distance, and bit rate of fiber communication systems was dramatic. System experiments over 103 km at 420 Mbit/s, 130 km at 2 Gbit/s, and 117 km at 4 Gbit/s were demonstrated in rapid succession. Further, the use of single- longitudinat-mode lasers opens up the possibility of close-spaced wavelength division multiplexing (WDM) and thereby makes possible more efficient use of the available fiber bandwidth. The state of the art in WDM is represented by the recent 10-channel, LANDS END ~ _ ~1 my VILLAGE GREEN H~. TUCKERTON C ` - ~ 009 THE TAT 8 CONFIGURATION PROPOSED FOR A SIX FIBER SL SYSTEM To to I' / FIGURE 14 Layout of transatlantic lightwave system (TATTY.

66 C. KUMAR N. PATEL OPTICAL PROCESSI NO 11 20-Gbit/s, 68-km transmission experiment with the highest ca- pacity of any optical communication system, 1.37 terabit-km/s (see the box on p. 60~. The possibility of optical information processing is tantalizing for several reasons. All-optical systems would have multiple intrinsic advantages over electronic systems in speed, parallel- ism, and low cross talk. In addition, light propagates in free space with negligible dispersion and loss; light beams can intersect in space without cross talk; and a number of beams can be simultaneously operated on by one element, then separated. However, several difficult problems must be solved to use these advantages. In particular, methods of switching, storage, regen- eration, and performing logic operations with light have to be devised in analogy with their electronic counterparts. These functions are routinely obtained with electronics today. How- ever, it is clear that the optical versions of these and related operations are now in their infancy. The requirements for such a computer or information processor have yet to be clearly defined, and in particular, it is not clear what architecture would be suitable for these kinds of problems: either a limited number of devices operating at extremely high speeds or a massive parallel network of devices operating at lower speeds. The particular characteristics of the optical devices will help to determine a suitable architecture. In fact, even if suitable devices existed today for high-speed optical processing in parallel archi- tectures, algorithms for programming and operating massively parallel machines have yet to be devised. Although it may not be necessary to use lasers for optical information processing and computing, laser light has several advantages over conventional light sources, some of which can and have been exploited to make new devices and components. In particular, the highly focusable intensity that is obtainable with lasers makes nonlinear optics possible for use in optical logic and switching applications, and the generation of pico- second optical pulses with mode-locked semiconductor lasers makes high bit rates feasible in optical systems. OPTICAL SWITCHING Optical switching serves the function of routing a signal into various alternative paths. Several methods have been devised for

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 67 ~` FIBER ARRAY FIGURE 15 Schematic representation of a 4 x 4 Ti:LiNbO3 directional coupler switch array. such switching of light, notably acousto-optic and electro-optic beam deflectors. One necessary component of an optical proces- sor or computer is a device that can interchange the connections in order to reroute the paths of the individual output beams into other inputs. In principle, a switching device should be capable of interconnecting any output channel to any input channel with low cross talk and completely within a single machine cycle. At this time, we do not know what kind of cycle times might be used or what data rates in each channel might be necessary. Optical crossbars have been constructed using electro-optical materials such as lithium niobate in waveguide structures (Schmidt and Kaminow, 1974; Schmidt and Kogelnik, 1976; Alferness, 19811. Details of a 4 x 4 switch (see Figures 15 and 16) have already been published (McCaughan and Bogert, 1985~. Recently, as FIGURE 16 Packaged and pigtailed 4 x 4 optical switch.

68 C. KUMAR N. PATEL 1~1 many as eight channels have been switched at rates of several hundred megahertz with cross talk levels of - 30 dB per channel (Granesrand et al., 1986~. For optical processing applications as many as 10,000 switching elements are envisioned. This seems well beyond the range of existing fabrication technology. OPTICAL MODULATION Electro-optic modulation (Reinhart and Miller, 1972) and opti- cal switching (Shelton et al., 1978) have been demonstrated in GaAs/GaAlAs heterostructure waveguides. This low-loss wave- guide geometry creates a close coupling between the applied modulating (or switching) field and the propagating optical field, thus allowing for efficient modulation (or switching) by means of the linear electro-optic effect. Several types of modu- lators, including electroabsorption, phase, and polarization modulators, have shown good performance. For example, a polarization modulator has been fabricated that requires less than 10 V to produce an extinction ratio of 20 dB. One advantage of these semiconductor devices is the possibility of integrating them with lasers, detectors, and transistors to create integrated optical circuits (see Figure 9~. Efficient optical mod- ulation can also be obtained at room temperature with excitonic electroabsorptive effects in quantum well devices (Miller et al., 1985a). By integrating single quantum wells into waveguide structures, modulations as large as 10 dB have been obtained with logic-level drive voltages (Wiener et al., 1985), and switch- ing times of 100 ps have been demonstrated (Wood et al., 1985~. The intrinsic response time of the excitonic electroabsorption has been found (Knox et al., 1986) to be at least as short as 330 fs. The use of quantum well modulators at the board-and-chip level is considered feasible, since the dimensions, electrical drive requirements and operating temperatures, and wavelengths are consistent with those of existing and experimental high-speed GaAs logic chips. OPTICAL STORAGE An optical computer will need an optical memory, with optical inputs and outputs. The present generation of optical storage discs uses electronic drivers and readout devices, and therefore is not included in this discussion. Optical fibers have been suggested as a means of storing optical information. However, the speed of light and the rate at which photons are lost to dissipative processes would seem to militate against the storage

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 69 of optical information in fibers except for short times. To store information for only 1 Us, a fiber 300 m long is required. In certain types of optical computers, there will nevertheless prob- ably be a need for relatively short-term memories, but with high capacity and rapid access. To store information for longer periods, longer fibers are required; however, signal losses even- tually dominate the storage time. A method to counteract the loss is required. Many kilometers of single-mode optical fiber can be coiled onto small spools a few centimeters in diameter without severe loss due to bending (J. W. Simpson, personal communication). With ends coupled together to form a reentrant loop, and with Raman amplification used to compensate for recirculating signal losses, a large amount of information can be stored for a sur- prisingly long time (Mollenauer et al., 1985~. For example, a loop 40 km long would have an access time of about 200 ms, and assuming a 100-ps spacing between adjacent pulses, its capacity would be about 2 Mbits. Furthermore, with wavelength multiplex- ing, this capacity could be increased by 10 times or more. Storage times of tens of milliseconds should be possible, limited by noise effects. Erasure is provided by temporarily turning off the optical pump power and allowing the signal to dissipate naturally. Another promising technique of optical data storage is that of spectral hole burning (Gutierrez et al., 1982~. By using a relatively intense monochromatic pump laser to induce a pho- tochemical reaction, a spectrally narrow homogeneously broad- ened "hole" (i.e., frequency gap) can be written in an inhomo- geneously broadened optical transition. This information can then be read by using a weak optical probe beam. This approach offers the possibility of greatly increasing the storage density by the ratio of the inhomogeneous to homogeneous linewidths, which is typically approximately 103. Since the conventional planar geometry optical memories (for example, the optical discs) are limited by diffraction to a maximum of 108 bits/cm2 (corresponding to 1 bit per square wavelength), the spectral hole burning technique can increase the density to 10i ~ bits/cm2. The information can be stored for hours at 4 K, since the optical saturation is based on photochemical processes that have negli- gible thermal reversibility at low temperature. At present, up to 30 bits of information have been written in a doped polymer film (see Figure 17) with spectral holes as narrow as 100 MHz. Research aimed at making this a practical technology is directed toward increasing the sensitivity of the recording medium by several orders of magnitude to allow nanosecond read/write times. 111 111

70 C. KUMAR N. PATEL oh ~ ~ ~ 10 l 1 1 001 coo 1 1 1 1 o 1 01 1 1 Go 1 1 1 o l 1 1 ~ ~ _ _ Am ~ WAVELENGTH 1'n ~ ~ tar FIGURE 17 Sequence of 30 bits written onto a doped polymer film at 4 K. The presence or absence of spectral holes is designated by ones and zeros, respectively (Gutierrez et al., 19824. OPTlCAl l OGlC AND BISTABILITY The ability to perform logic functions using optical inputs and outputs must be considered a basic requirement of an optical computer or information processor. Of course, it would be possible to build hybrid devices wherein optical inputs are detected and fed into the inputs of conventional electronic logic elements, and the output signal is then converted to an optical signal by an LED or laser diode. The losses due to inefficiency at every conversion stage cannot be tolerated if we are to construct a system of high-density, fast-switching devices operating at high data rates. If light itself could be used to control other light beams through optical logic devices, a significant simplification would be possible. Interactions with light are generally weaker than is desirable; therefore, considerable effort is required to produce optical logic devices at present. Bistability, which is essential to optical logic, has been shown in a variety of optical devices. In particular, bistability as a result of absorption and dispersion changes, electroabsorption (self- electro-optic effect devices ESEED]), and index variations (op- tical logic etalons) have been shown. The availability of optically bistable devices of high speed, low dissipation, good stability, and small size will make it possible to implement true optical logic devices such as gates, inverters, and saturable amplifiers in large arrays that can be combined to perform optical processing and computation functions. At present, no specific devices satisfying these requirements have been implemented. The first optical bistable devices used nonlinear Fabry-Perot cavities and optical beams of sufficient intensity to change the

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 7] characteristic of the medium within the cavity to perform the desired optical logic (Szoke et al., 1969; Seidel, 197 1; Gibbs et al., 1976, 1978~. Bistability has been demonstrated in several exper- iments. In particular, such devices constructed with MBE-grown GaAs/GaAlAs superlattice structure (Gibbs et al., 1982) (see Figure 18) or other semiconductors have a number of advan- tages, including a fast operating speed (10-ns switching times) and operating at a wavelength near 0.9 ,um and at room temperature (see Figure 19 for input/output characteristics of such a device). So far, the power requirements are large about 10 ni for a 1 OO-,um2 device. Such bistable devices are suitable for continuous wave processing and memory applications. SEED uses the electroabsorption in a GaAs MQW in combi- nation with a feedback loop to produce bistability with a single input wavelength (Miller et al., 1985b) (see Figure 201. A bias field is applied with a constant current source perpendicular to the plane of the quantum well, which shifts the excitonic resonance. When light is incident on the device, the induced photocurrent reduces the voltage across the quantum well, which shifts the resonance back in the direction of the unbiased excitonic absorption. This in turn reduces the absorption, which increases the intensity, and the device becomes optically bistable. Switching speed is limited by the RC time constant of the device, and speeds of nanoseconds to seconds have been demonstrated. Large transmission modulation (>10 dB) can be obtained with these devices. -'OO',m 0.2~`m GaAs UGLY BEAM \ I GaAs ~ ~ ~ SUBSTRATE jIto2mmj) Off _ _ 4 am ~~' 0.2l.m ~ DIELECTRIC COATINGS Go O.73Al0.27As 401A FIGURE 18 GaAs-GaAlAs superlattice structure used for room temper- ature optical bistability.

72 C. KUMAR N. PATEL ~ ~ ~ .~. ~ 4/1 1 1 ~1 1 1 1 Tl.ll,—- I _- ~4 I I I lil ll~ll~l ill IT _£: _ .... _~ .... _ FIGURE 19 Output versus input characteristics for the device shown in Figure 18. Upper trace shows the entire output. Lower trace shows output corresponding to the central part of the beam. Optical logic etalons (OLE) are nonlinear Fabry-Perot devices Jewell et al., 1986; Lee et al., 1986) (see Figure 21), similar in construction to the bistable devices discussed above. OLE de- vices, however, need not be bistable. An intense pump beam induces a change in the index of refraction of the nonlinear material in the etalon and shifts the resonance of the etalon, causing a change in the transmission of a probe beam. Cycle times of 30 ps have been obtained with 20-pJ input energy. Further, a 2 x ~ array of NOR gates (see Figure 22) has been operated at 82 MHz. The OLE is a three-port, all-optical gate that operates with two different wavelengths and is complemen- tary to other approaches. In normal operation, the control beam wavelength is shorter than the probe beam wavelength in order to provide the third port. Were such devices the only ones available for optical logic, the total number of sequential logic operations would be limited, because the probe wavelength would continue to shift down monotonically in each operation. Recently, complementary OLE devices have been demonstrated 0. L. iewell, personal communication) for which the control beam has a wavelength longer than that of the probe beam. Cascading now need not cause a monotonic shift in the operat-

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 73 EQUIVALENT CIRCUIT "RED" INfRARED CONTROL LIGHT IN LIGHT ~Dlm 1 - "RED" ~ CONTROLS ~ INFRARED ~. LIGHT IN ~ ~ LIGHT IN pa GaAs OPTICAL TRANSMITTED p GaAs~ WINDOW INFRARED LIGHT ~ L, ,' ~ ~ n MOW P 1 -| MQW I l ; ALL LAYERS | GaAs-AC GO AS SUPERLATTICE (SL) EXCEPT AS MARKED GOLD _. POLYIMIDE _ -1 cm~ _ n HIGH At SL ... ~ TRANSM ITTED INFRARED L IGHT FIGURE 20 Details of construction of a self-electro-optic effect device. ing wavelength. Operation near a 1.55-,um wavelength has been recently demonstrated (K. Tai, personal communication). An important requirement of optical logic devices is that they be cascadable, or that the output of one device can drive the input of at least the next device, if not several inputs. In optical terms, this means that there must be enough optical gain in the system to offset losses. This is not easily obtained in large arrays. However, various approaches have been suggested. One that is particularly interesting is the use of optically bistable elements PROBE INPUT 1 ~~g \1 INPUT 2 ~ NONLINEAR E TALON FIGURE 21 Schematic of an optical logic etalon. ~ 1 Em=> OUTPUT l 111 It

74 C. KUMAR N. PATEL 1 FIGURE 22 A 2 x 2 array of optical logic etalons. Left frame: two devices turned on. Right frame: all four devices turned on. (SOURCE: Tewell et al., 19861. that are biased near their switching transition by a strong beam. A weak signal beam falls on the elements and switches the devices into high output mode, thus providing net differential gain. This scheme removes the coherence between the signal and amplified signal beams, but provides the desirable modal isolation and saturable output characteristics for regenerative systems. Again, however, the problems of scaling too many elements for massively parallel systems operating at high data rates are formidable. DI RECTI ONS I N OPTl CA l PRO CESSI NO in conclusion, the requirements for optical processing and computing are not yet clearly defined. Some of the basic elements that will be required for a full-scale operational pro- cessor have, in fact, been demonstrated in a limited way, that is, single devices at moderate performance levels; however, no existing technology seems directly scalable to the large number of elements and high operating speeds that would allow us to take full advantage of the properties of light. The physical processes that have been exploited to date are either too slow or not efficient enough to allow close packing and high-speed operation. In fact, most technologies are so far from meeting the

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 75 requirements envisioned at this time that a breakthrough is probably needed. Lightwave communications, optical switching, and computation are only three areas of the many in which lasers are playing a pivotal role. But even in these areas, we can see that lasers, together with a number of significant achievements in related fields, have caused a revolution in communications. In informa- tion processing, the impact of laser technology is not yet visible, but if the past is any guide, information processing also will see its capabilities significantly enhanced through a variety of laser applications, many of which are still in the conceptual stage. CONCLUSION ACKNOWLEDGMENTS I would like to thank Drs. N. A. Olsson, D. V. Lang, E. Capasso, S. L. McCall, and D. S. Chemla for assistance in putting this manuscript together, and Dr. A. M. Glass for critical comments. Alferness, R. C. 1981. IEEE J. Quantum Electron. QE-17:946 - 959. Alferov, Zh. I., V. M. Andreev, V. I. Korol'kov, E. L. Portnoi, and D. N. Tretyakov. 1969. Sov. Phys. Semicond. 2:843. [Translated from Fiz. Tekh. Poluprovodn. 2: 1016 (1968).] Bowers, I. E. 1985. Electron. Lett. 21:1195. Caneau, C., A. K. Srivastava, A. G. Dentai, I. L. Zyskind, and M. A. Pollack. 1985. Electron. Lett. 21 :815. Capasso, F. 1985. Physica 129B :92. Capasso, F., K. Mohammed, and A. Y. Cho. 1986. IEEE I. Quantum Electron. QE-22(September): 1853. Casey, H. C., Jr., and M. B. Panish. 1978. Heterostructure Lasers, Parts A and B. New York: Academic Press. Cho, A. Y. 1983. Thin Solid Films 100:291. Cho, A. Y., and T. R. Arthur. 1975. Prog. Solid-State Chem. 10:157. Dupuis, R. D. 1984. Science 226:623. Gibbs, H. M., S. L. McCall, and T. N. C. Venkatesan. 1976. Phys. Rev. Lett. 36:1135. Gibbs, H. M., S. L. McCall, and T. N. C. Venkatesan. 1978. U.S. Patent 4,071,831. Gibbs, H. M., S. L. McCall, {. L. Newell, D. A. Weinberger, K. Tai, A. C. Gossard, A. Passner, and W. Wiegmann. 1982. Appl. Phys. Lett. 41:221. Granesrand, P., L. Thylen, B. Stoltz, K. Bergvall, W. Doldissen, H. Heidrich, and D. Hoffmann. 1986. Integrated and Guided Wave Optics Conference, Atlanta, Gal, February 26 - 28, Paper WAA-3. Gutierrez, A. R., J. Friedrich, D. Haarer, and H. Wolfrum. 1982. IBM J. Res. Dev. 26: 198-208. Hall, R. N., G. E. Fenner, J. D. Kingsley, T. l. Foltys, and R. O. Carlson. 1962. Phys. Rev. Lett. 9:366-368. Harrison, W. A. 1985. I. Vacuum Sci. Technol. B3:1231. REFERENCES _~

76 C. KUMAR N. PATEL Hartman, R. L., N. E. Schumaker, and R. W. Dixon. 1977. Appl. Phys. Lett. 31:756. Hayashi, I., M. B. Panish, and P. W. Foy. 1969. IEEE I. Quantum Electron. QE-5:211. Hayashi, I., M. B. Panish, P. W. Foy, and S. Sumski. 1970. Appl. Phys. Lett. 17:109. Hegarty, J., N. A. Olsson, and L. Goldner. 1985. Electron. Lett. 21:290-292. Holonyak, N., Jr., and S. F. Bevacqua. 1962. Appl. Phys. Lett. 1:82-83. Newell, I. L., Y. H. Lee, l. F. Duffy, A. C. Gossard, W. Wiegmann, and I. H. English. 1986. P. 32 in Optical Bistability III, Springer Proceedings in Physics, Vol. 8. New York: Springer-Verlag. Kao, K. C., and G. A. Hochkam. 1966. Proc. IKE 113:1151-1158. Knox, W. H., D. A. B. Miller, T. C. Damen, D. S. Chemla, C. V. Shank, and A. C. Gossard. 1986. Appl. Phys. Lett. 48:864. Kogelnik, H., and C. V. Shank. 1971. Appl. Phys. Lett. 18:152. Lee, Y. H., H. M. Gibbs, J. L. Newell, I. F. Duffy, T. N. C. Venkatesan, A. C. Gossard, W. Wiegmann, and l. H. English. 1986. Appl. Phys. Lett. 49: 486-488. Margaritondo, G. 1986. Solid-State Electron. 29:123. Matsueda, H., T. P. Tanaka, and H. Nakano. 1984. Proc. IKE 131(5):299-303. Mayo, J. S. 1985. The evolution of information technologies. Pp. 7 - 34 in Information Technologies and Social Transformation, B. R. Guile, ed. Wash- ington, D.C.: National Academy Press. McCaughan, L., and G. A. Bogert. 1985. Appl. Phys. Lett. 47:348-350. Miller, D. A. B., D. S. Chemla, T. C. Damen, A. C. Gossard, W. Wiegmann, T. H. Wood, and C. A. Burrus. 1985a. Phys. Rev. B32:1043. Miller, D. A. B., D. S. Chemla, T. C. Damen, T. H. Wood, C. A. Burrus, A. C. Gossard, and W. Wiegmann. 1 985b. IEEE l. Quantum Electron. QE-2 1: 1462. Mollenauer, L. F., R. H. Stolen, and M. N. Islam. 1985. Opt. Lett. 10:229. Nathan, M. I., W. P. Dumke, G. Burns, F. H. Dill, fir., and G. Lasher. 1962. Appl. Phys. Lett. 1:62-64. Nelson, K. C., D. L. Brownlow, L. G. Cohen, F. D. DiMarcello, R. G. Huff, I. T. Krause, P. J. Lemaire, W. A. Reed, D. S. Shenk, E. A. Sigety, J. R. Simpson, A. Tomita, and K. L. Walker. 1985. I. Lightwave Technol. LT-3(5):935-941. Olsson, N. A. 1985. Electron. Lett. 21:1085-1087. Olsson, N. A., and I. P. van der Ziel. 1987. l. Lightwave Technol. (Special Issue on Coherent Communications) LT-5:509 - 515. Olsson, N. A., N. K. Dutta, and K.-Y. Liou. 1984. Electron. Lett. 20:121. Olsson, N. A., H. Temkin, R. A. Logan, L. F. Johnson, G. J. Dolan, J. P. van der Ziel, and J. C. Campbell. 1985. J. Lightwave Technol. LT-3:63-67. Panish, M. B., and H. Temkin. 1985. J. Vacuum Sci. Technol. B3:657. Panish, M. B., I. Hayashi, and S. Sumski. 1969. IEEE I. Quantum Electron. QE-5:210. Panish, M. B., I. Hayashi, and S. Sumski. 1970. Appl. Phys. Lett. 16:326. Quist, T. M., R. H. Rediker, R. J. Keyes, W. E. Krag, B. Lax, A. L. McWhorter, and H. I. Zeiger. 1962. Appl. Phys. Lett. 1:91-92. Reinhart, F. K., and B. I. Miller. 1972. Appl. Phys. Lett. 20:36-38. Schawlow, A. L., and C. H. Townes. 1958. Phys. Rev. 112:1940. Schmidt, R. V., and I. P. Kaminow. 1974. Appl. Phys. Lett. 25:458 - 460. Schmidt, R. V., and H. Kogelnik. 1976. Appl. Phys. Lett. 29:503-505. Seidel, H. 1971. U.S. Patent 3,610,731. Shelton, J. C., F. K. Reinhart, and R. A. Logan. 1978. Appl. Opt. 17:2548 - 2555. Stillman, G. E., L. W. Cook, N. Tabatabaie, G. E. Bulman, and V. M. Robbins. 1983. IEEE Trans. Electron. Devices ED-30:364. Szoke, A., l. Goldhar, and N. A. Kurnit. 1969. Appl. Phys. Lett. 15:376. Temkin, H., G. J. Dolan, R. A. Logan, R. F. Kazarinov, N. A. Olsson, and C. H. Henry. 1985. Appl. Phys. Lett. 46:105. Tsang, W. T. 1981. Appl. Phys. Lett. 39:786.

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 77 Tsang, W. T., N. A. Olsson, and R. A. Logan. 1983. Appl. Phys. Lett. 42:650. Tsang, W. T., N. A. Olsson, R. A. Logan, C. H. Henry, L. F. Johnson, I. E. Bowers, and I. Long. 1985. IEEE l. Quantum Electron. QE-21(6):519 - 526. Wiener, T. S., D. A. B. Miller, D. S. Chemla, T. C. Damen, C. A. Burrus, T. H. Wood, A. C. Gossard, and W. Wiegmann. 1985. Appl. Phys. Lett. 47:1148. Wood, T. H., C. A. Burrus, R. S. Tucker, I. S. Weiner, D. A. B. Miller, D. S. Chemla, T. C. Damen, A. C. Gossard, and W. Wiegmann.1985. Electron. Lett. 21:693. Woodall, I. M., H. Rupprecht, and G. D. Pettit. 1967. Solid-State Device Conf., tune 19, 1967, Santa Barbara, Calif. (Abstracts reported in IEEE Trans. Electron. Devices ED-14:630.) FIBER lOSS Two parameters determine the usefulness of optical fibers in lightwave transmission. Both limit the maximum distance infor- mation can travel before it has to be regenerated using repeat- ers. The first is the propagation loss, which makes the signals weaker as they propagate down the fiber until they reach a point at which detection with a reasonable signal-to-noise ratio be- comes a problem. The second is the dispersion of light in the fibers, which leads to a broadening of the lightwave pulses during propagation. This puts a combined limit on the data rate and distance of propagation. The linewidth of the laser source plays an important role in the dispersion-limited propagation, while the laser power launched in the fiber plays a crucial role in the absorption-limited regime. However, the input power to an optical fiber cannot be increased arbitrarily, even by staying below obvious limits of materials damage. In single-mode fibers especially, nonlinear interactions such as stimulated Brillouin and Raman scattering begin to become important at high laser intensities. These processes, which act to limit the maximum useful power injection into the fiber, can be used advantageously for in-line amplification of lightwave signals. At present, the silica fiber losses have been reduced to minimum levels predicted by Rayleigh scattering, which, be- cause of its A-4 dependence, suggests that operation at longer wavelengths can yield even lower losses than the smallest loss seen in Figure 2. To exploit this possibility, a considerable amount of work is in progress to determine the best material systems that do not have intrinsic loss limits imposed by molec- ular vibrational frequencies of the constituents of the materials APPENDIX A: CHARACTERISTICS OF OPTICAL FIBERS

78 C. KUMAR N. PATEl ll . 105 lop 103 E 10 m C,, 1 o 1 o J 1 0 0.1 0.01 0.001 (BecaKAl)FLuoRlDE SIO2/ (Zr BaLa Al)FLUORIDE / it\ Ce3 | / / I ZnCI 2 l //\OH(1ppb) IOH l(1ppb) 1 , ,,, 1 0.5 \(1ppb) / \Fe2~\ / (1ppb) ~ 1 ~ I, 1 1 , , ~ 5 10 WAVELENGTH (palm) FIGURE A- 1 Calculated spectral loss data for silica, sulfide, and fluoride fibers. Experimental data are also indicated. Or the impurities. Figure A-1 shows theoretical loss spectra for the silica, sulfide, and fluoride glasses (Lines, 1984~. Silica fibers, as mentioned above, are already at or near their theoretical minimum loss limits of approximately 0.1-0.2 dB/km at 1.55 ,um. Fluoride materials, for example, should show minimum loss approaching 10-2 dB/km at about 2.3 ,um. The current losses (Tran et al., 1986; Yoshida, 1986), however, are approximately 0.7 dB/km at about 2.3 ,um. The discrepancy of almost 2 orders of magnitude between the expected and the observed loss at 2.3 ,um is probably the result of impurities, and can probably be reduced, much as the loss in silica fiber was reduced by more than 3 orders of magnitude between 1973 and 1983. Other materials, notably ZnCl2, could have losses as small as 10-3 dB/km at 6.0 ,um. Although it is impossible to predict when the measured losses In practical fibers will come close to the expected values, it is interesting to speculate (Patel, 1981) about the impact of a 10-3-dB/km loss for lightwave fibers. This could make possible a

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 79 transatlantic lightwave communication channel with no repeat- ers between the continental United States and Europe. If the present division between the cost of the fiber cable and that of the repeaters continues to hold for the longer wavelength fibers and detectors and lasers, the cost of the transatlantic lightwave system and all other long-distance systems can drop signifi- cantly. This reduction might further affect the balance between satellite and lightwave communication for large bandwidth information exchange across the Atlantic Ocean. FIBER DISPERSION Although loss characteristics of the current silica fibers are relatively constant over a reasonably broad spectral range—for example, covering a few hundred angstroms around 1.3 or 1.55 ,um the propagation group velocity changes significantly over even such a narrow range of wavelengths. This dispersion in group velocity has little or no effect on low-bit-rate transmission. However, at high bit rates, the group velocity difference for frequencies comprising the Fourier component of the light pulse leads to a pulse broadening that limits the maximum fiber span between repeaters. Chromatic dispersion, D, is a measure of the pulse spreading and is defined as a differential of the group delay tg, L do 1 dng c do A den c dA2' (1) where tg is the propagation delay for a fiber of length L given by: L t = g Vg (2) where vg is the group velocity and ng is the group index given by: ng = vg (3) and ng is a function of wavelength, in general, which leads to nonzero chromatic dispersion, D. At approximately 1.3 ,um in

80 C. KUMAR N. PATEL 1 1 silica fibers, D = 0 (see Figure 13), giving rise to a desirable situation of zero pulse broadening for a narrow linewidth laser input. In general, however, D is nonzero. By changing the dopants or the index profile of the fiber core and cladding, the zero dispersion point can be shifted to a limited extent (Gloge, 1971; Cohen et al., 1982) without significantly increasing the absorption losses. The spectral width of the light pulse has two primary compo- nents. The first is the fundamental one that has its origin in the Fourier spectrum of the pulse, and this cannot be avoided. However, in practical situations, the lasers used for lightwave communication themselves possess spectral widths arising from multilongitudinal or other nonoptimal operation. For a typical multimode laser operating in the 1.3- or 1.55-,um region, line- widths as broad as 70-100 A are not unusual, corresponding to laser oscillation on three longitudinal modes of the optical cavity of the semiconductor laser. For a single longitudinal-mode laser with a distributed feedback (DEB) grating integral to the laser, linewidths of the order of 1 A are routine. Compared to these linewidths, the Fourier transform-limited spectral width for a 2-GHz/s bit rate is approximately 0.5 A. Thus, the propagation of 2-GHz/s bit rate pulses from multimode and single-mode lasers will experience different dispersion-related limitations. Figure 13 shows group velocity dispersion data for a single transverse-mode silica fiber plotted as a function of wavelength. The dispersion has been plotted in picoseconds of pulse broad- ening per kilometer of propagation and per nanometer of spectral width. The 1.3-,um low-loss window seen in Figure 13 for the silica fibers corresponds to a zero dispersion point of the fiber. Operating at the zero dispersion point avoids the prob- lems of pulse broadening due to dispersion. The transatlantic lightwave system TAT-8—due to be installed in the near future takes advantage of the zero dispersion point (Runge and Trischitta, 1984~. As seen from the spectral variation of the fiber loss, however, the lowest loss for the silica fiber occurs at a wavelength of 1.55 ,um, where the dispersion is approximately 20 ps/km nm. At this wavelength, we would expect that at a 2-GHz/s bit rate, a practical single-mode distributed feedback laser pulse (linewidth ~1 it) would broaden from an initial pulse width of 500 ps to about 750 ps when propagating through a 35-km fiber. The pulse from the multimode laser, on the other hand, would, for the same initial pulse width of 500 ps, broaden to about 7.5 ns going through the 35-km span, thus making communication impossible at this bit rate with a 35-km repeater span.

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 81 FIGURE A-2 Propagation of 1-ns laser pulses from a single-mode laser (linewidth, ~ 1 \) at 1.55 ,um through a single-mode fiber 35 km long. No detectable dispersion-induced broadening of the output pulse (top) is seen. (Bottom: input pulse.) Figures A-2 and A-3 show (N. A;. Olsson, personal communi- cation) experimental demonstration of single-longitudinal ver- sus multilongitudinal laser source injection into a fiber 35 km long. The pulse width is 1 ns at input, corresponding to a data rate of approximately 500 Mbit/s. Notice that the single-mode source pulse has no significant broadening at the output of the fiber. On the other hand, the multimode source pulse has broadened to about 10 us, and the output pulse has three distinct peaks that correspond to each of the three longitudinal modes of the laser propagating at different velocities. In these cases, the loss-limited fiber span would have been ' 200 km. Thus, in general, dispersion tends to limit the repeater spacing at high data rates. The data representing records for longest path propagation as a function of the bit rate for a wavelength of 1.55 ,um are shown (T. Li, personal communication) in Figure A-4. The loss-limited straight line corresponds to direct detection with a -45-dBm capability detector. The additional straight lines with larger negative slopes are drawn for 2-, 1-, and 0-A laser

82 C. KUMAR N. PATEL FIGURE A-3 Propagation of 1-ns pulses from a multilongitudinal mode laser at 1.55 ,um through a single-mode fiber 35 km long (other experimental conditions are identical to those in Figure A-2. Laser line width is ~70 A. Significant dispersion-induced broadening of the output pulse is seen. The breakup of the output pulse into three distinguishable components reflects the laser operation on three longi- tudinal modes, each one of which has a slightly different group velocity. 600 ~ ' ' ' ~ ' " ~ ~ ' ' ~ ' ' ' '_ 400 _ _ 200 100 80 60 4C 20 10 0.4 LOSS LIMIT (0.2 dB/ km. 1 mw) _ ~ · ATT o NTT · BTRL · NEC 1 Q86 ·°F~ DISPERSION - L I M IT - I 1 1 1 1,, 1 , 1 1 1. 1 1 1 1 1 0.2 0.5 1 2 Bl T RATE (Gb/S ) 5 10 FIGURE A-4 Experimental and theoretical bit rate versus distance data.

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING (33 1000 100 or ~ 10 o 11 .0 0.1 . . . 0 · ° 0.8 Pam O · 1.3 Elm · 1.5~m · 1.5 Am WDM 1 1 1 1 1 1 6 1978 1980 1982 1984 1986 YEAR FIGURE A-5 Lightwave system performance (Gbit/s x km) as a func- tion of time. (WDM corresponds to a 10-channel multiplex study.) linewidths, which represent the dispersion-limited span max- ima. Now we see that a loss of 10-3 dB/km in future fibers does not immediately portend 10,000-km repeater spacing if the fibers have finite dispersion. Nonlinear phenomena such as soliton propagation have to be invoked to compensate for the unacceptable pulse broadening that will accompany the longer repeater spans promised by ever-decreasing fiber losses. Finally, another way of measuring the figure of merit of a lightwave system is to calculate the product of the data rate (bits per second) and the maximum distance between repeaters (con- sistent with acceptable received bit error rate). Figure A-5 shows the performance (bit rate multiplied by the distance) of light- wave systems as a function of time, beginning in 1975. Rapid improvement has occurred since that time. (The WDM data point is for a 10-laser, 1.55-,um wavelength multiplex system described elsewhere in this paper.) SOLITON PROPAGATION Hasegawa and Tappert (1973) pointed out that intensity-depen- dent nonlinear effects in the index of refraction of an optical fiber, given by: n = nO + 1n2I, (4) where no is the refractive index of the fiber at low intensities, I is the optical field intensity, and n2 is the nonlinear index, can be used for compensating the effect of chromatic dispersion, D.

84 C. KUMAR N. PATEL t a ) LASER A _~ t be_ (d) stow ( e ) 11.4 FIGURE A-6 Soliton propagation in a single-mode fiber of length equal to a half-soliton period. (a) This shows the laser pulse as it is launched into the fiber. (b) P = 0.3 W. Negligible nonlinear effects seen; only dispersive broadening takes place. (c) P = 1.2 W. Output pulse width narrows to the same as that for input pulse corresponding to the fundamental soliton propagation. (d) P = 5.0 W. Output pulse nar- rowed to minimum width corresponding to half period behavior of n = 2 solitons. (e) P = 11.4 W. First well-resolved splitting of the output pulse corresponding to n = 3 solitons. (Note that the threefold splitting in the these autocorrelation traces correspond to a twofold splitting of the pulse itself.) Even though n2 has a very small numeric value, given by n2 ~ 3.2 x 1o-~6 cm2 W-i, the long propagation lengths possible in low-loss fibers and the small core diameter of the single-mode fibers that yield high optical field intensities make it possible to see the effect of the nonlinear index. For example, a laser pulse with a power of 1 mW in a typical single-mode fiber produces an intensity I. - 103 Wcm-2. Hasegawa and Tappert's studies showed that under suitable conditions, stable pulses that main- tain their shape and width can be propagated in low-loss, single-mode fibers. These shape-maintaining pulses, called en- velope solitons, result from an interplay between the normal dispersion of the fiber, which would tend to broaden the pulses, and the nonlinear index, which tends to sharpen them. A stable soliton pulse in a fiber has a hyperbolic secant shape. In several elegant experiments, Mollenauer and his colleagues have shown the existence of such soliton propagation at modest values of injected laser power (Mollenauer et al., 1980~. Figure A-6 shows the results of one of the early experiments to demonstrate the propagation of solitons in approximately 700- m-long single-mode fibers. The lowest input power autocorre- lation trace of the output is for Pin ~ 0.3 W. for which a pulse broadening occurs because of the chromatic dispersion of ap- proximately 16 ps/km nm. For input power above this value, a

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 85 monotonic narrowing of the output pulse was observed; at Pin ~ 1.2 W the output pulse width mirrored the input pulse width and reached a minimum of 2 ps at Pin ~ 5.0 W. At higher powers, the output pulse shows splitting caused by higher order solitons. For this experiment, the length of the fiber was half a soliton characteristic period, zO, defined by: c r2 Z0 = 0.322~2 A2 D ' (5) where ~ is the full width at the half-intensity maxima of the pulse. For longer fiber lengths, soliton shape is no longer preserved at all points along the fiber but is recovered at the end of each soliton period. This assumes that the characteristic absorption length, 1/a, of the fiber is much larger than the soliton period. In a practical system for optical fiber lengths (i.e., repeater spacings) much greater than the absorption length, the soliton can no longer be preserved without periodic "in-line" amplifi- cation of the soliton intensity (Hasegawa and Kodama, 1982; Hasegawa, 1984~. Such amplification can take advantage of Raman gain in silica fibers when laser radiation at high fre- quency and relatively high power is injected (see Figure A-71. The Raman gain overcomes the distributed loss of the fiber, as well as local loss due to insertion of the directional coupler into the fiber path. Such in-line amplification may have advantages even when soliton propagation is not used (Hegarty et al., 1985~. Further, the use of semiconductor lasers as amplifiers has also been explored as a way to compensate for the fiber loss in a coherent lightwave communication system. Olsson has shown that in such a configuration, there is essentially no penalty incurred in the bit error rate (Hegarty et al., 19851. ~s~ 1.56- 1.59~m I r ~ ~ I ~ ~ ...~l I lL ~ ;1 1 tV kn~1.46~1.48``m ;1 1 lit ;1 i lL . I— L —~ L FIGURE A-7 In-line Raman amplification for an all-optical lightwave system. At "repeaters" spaced by L, continuous wave laser diodes inject power (at Ap) into the optical fiber in both directions (through wave- length-dependent directional couplers) for the Raman gain to compen- sate for the loss experienced by data pulses (As). 1 111

86 C. KUMAR N. PATEL 1 1 TABLE A- ~ Design Parameters for a Sing~e- Channe~ So~iton-Based Lightwave System Parameter Value Input power In-line amplifier separation Pulse width Data rate Total distance 3.0 mW 50 km 22.6 ps 4.4 Gbit/s 6,600 km The in-line optical amplification avoids the multitude of complexities associated with the conventional repeaters that rely on optical-to-electronic-to-optical signal conversions. These var- ious conversions and the electronic paths themselves can limit the bit rate. Further, for the wavelength division multiplexing scheme (Olsson et al., 1985), the conventional repeaters become unduly complex when we try to take advantage of the enormous bandwidth of lightwave communications by multiplexing tens of individual wavelengths on a single fiber. For example, each of the repeaters would need the optical demultiplexer and the optical multiplexer together with one electronic channel for each of the wavelength division multiplexed lightwave channels. A parametric investigation of the usefulness of soliton prop- agation systems in conjunction with in-line Raman amplification in the fiber itself yields (Mollenauer et al., 1986) products of bit rate and distance approximating 30,000 GHz-km for a single- fiber system and 300,000 GHz-km for a 24-channel WDM system, as shown in Table A-1. Signal laser powers of the order of '10 mW are sufficient with pump powers of the order of a few hundred milliwatts at in-line amplifier "repeater" lengths of approximately 50 km. Although these numbers represent ex- trapolations from observed soliton propagation, they could nonetheless be possible options for future ultrahigh-perform- ance lightwave systems. REFERENCES Cohen, L. G., W. L. Mammel, and S. I. fig. 1982. Electron. Lett. 18:1023— 1024. Gloge, D. 1971. Appl. Opt. 10:2252 and 2442. Hasegawa, A. 1984. Appl. Opt. 23:3302-3309. Hasegawa, A., and Y. Kodama. 1982. Opt. Lett. 7:287. Hasegawa, A., and F. Tappert. 1973. Appl. Phys. Lett. 23:142. Hegarty, J., N. A. Olsson, and L. Goldner. 1985. Electron. Lett. 21:290-292. Lines, M. E. 1984. J. Appl. Phys. 55:4058. Mollenauer, L. F., R. H. Stolen, and J. P. Gordon. 1980. Phys. Rev. Lett. 45:1095.

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 87 Mollenauer, L. F., I. P. Gordon, and M. N. Islam. 1986. IEEE I. Quantum Electron. QE-22:157-173. Olsson, N. A., J. Hegarty, R. A. Logan, L. F. Johnson, K. L. Walker, L. G. Cohen, B. L. Kasper, and I. C. Campbell. 1985. Electron. Lett. 31:105. Patel, C. K. N. 1981. Soc. Photo-Opt. Inst. Eng. 266:22. Runge, P. K., and P. R. Trischitta. 1984. IEEE Selected Areas Commun. SAC-2: 78~793. Tran, D. C., K. Levin, M. Burk, C. Fister, and W. Broer. 1986. Proc. SPIE Symp. Infrared Optical Material and Fiber 618:48. Yoshida, S. 1986. Paper presented at the North Atlantic Treaty Organization Advanced Research Workshop on Halide Glasses for Infrared Fibers, March. APPENDIX B: DETECTORS IN OPTICAL COMMU N ICATIONS PIN PHOTODETECTORS AND PIN-FET RECEIVERS State-of-the-art PIN photodiodes in the A = 1.3-1.6-,um range consist of mesa or planar InO.53Ga0.47As pn junctions, grown lattice matched to InP substrates, with very low doped n layers (<5 x 10~5 cm-3) to achieve the required low capacitance ('0.5 pF) (Pearsall and Pollack, 19851. These devices are typically operated at low reverse bias voltage (~10 V) with dark currents of a few nanoamperes and external quantum efficiencies of 60-70 percent at A = 1.55 and 1.3 ,um. These detectors, in combination with a GaAs field-effect transistor (FET) front-end amplifier, have found wide use in fiber-optic receivers at data rates up to 400 Mbit/s. Experimental tests and theoretical evaluations have indeed shown that in this bit rate range, the use of state-of-the-art InP/GaInAs avalanche photodiodes (APDs) in place of the PIN photodiode in receivers yields a typical im- provement in sensitivity of 1 or 2 dB. On the other hand, the APD technology is much more demanding and costly than the PIN technology, and high-reliability planar 1.3- to 1.6-,um APDs have not yet been demonstrated. These considerations explain why the PIN-FET combination has been the most widely used in receivers at data rates of less than about 400 Mbit/s (Forrest, 1985). The minimum noise that an FET front-end amplifier can achieve in a photoreceiver is proportional to CT/gm CT is the total capacitance, which contains contributions from the detector, the FET, and parasitics (interconnects); and gm iS the FET transcon- ductance. CT should be minimized to maximize the receiver sensitivity. The latter quantity is defined as the optical power required to achieve a bit error rate of 10-9. The best PIN- receiver sensitivities have been achieved using a hybrid combi- 1

88 C. KUMAR N. PATEL 111 nation of a mesa GaInAs PIN and a GaAs semiconductor field-effect transistor with CT = 0.5 pF. This sensitivity is—42 dB/m at A = 1.3 ,um for a bit rate of 500 Mbit/s (Forrest, 1985~. HETEROJUNCTION AVALANCHE PHOTODIODES Simple homojunction GaO.47InO.53As photodiodes cannot be op- erated as low-noise avalanche detectors because the dark current becomes prohibitively large because of Zener tunneling across the band gap at voltages at which impact ionization sets in (Pearsall and Pollack, 1985~. This important fact, completely overlooked by proponents of the homojunction approach, is a consequence of the small gap (0.73 eV) and small electron effective mass in this alloy. This finding, first reported in 1980, clearly demonstrated the need for suitable heterojunction APDs (Pearsall and Pollack, 1985~. An important step in this direction was the proposal and 1979 demonstration of an APD that consisted of an InP pn junction grown next to a GaInAs(P) layer. To minimize the Zener tunneling in this device, the maximum electric field of the pn junction is located in the InP wide gap layers where the avalanche takes place, and the GaInAs(P) layer absorbs the infrared photons (Figure B-1~. Hence, the name SAM (separate \ —'VW~ he - \ \ - \ \ ~ ~\ FIGURE B- 1 Band diagram of an SAM (separate absorption and multiplication) avalanche photodiode

LASERS IN COMMUNICATIONS AND INFORMATION PROCESSING 89 absorption and multiplication) APD is given to this structure (Pearsall and Pollack, 19851. This device achieves good gains (10-20) and low dark currents (nearly 50 nA), provided the doping and thickness of the avalanche region are carefully tailored and controlled. However, a serious problem of this structure was discovered shortly after the device was demonstrated. When a GaO 47InO.53As absorbing layer is used, photogenerated holes, in the process of drifting into the InP layer, tend to pile up at the heterojunction interface. This effect produces a long tail (tens of microseconds in the worst cases) in the pulse response, making it impossible to use these devices at high bit rates. Researchers at the AT&T Bell Laboratories recently solved this problem when they introduced an intermediate quaternary GaInAsP grading layer between the ab- sorbing GaInAs layer and the InP avalanche region (Pearsall and Pollack, 1985) and, alternatively, introduced a chirped InP/GaO.47InO 53As superlattice that simulates a graded gap GaInAsP layer (Figure B-2) (Capasso, 1985~. Such pseudoquater- nary alloys represent a good example of band-gap engineering and should find applications in other optoelectronic devices as well, such as graded-index separate confinement heterostructure (GRINSCH) InP/GaInAs lasers. High-performance InP/GaInAs SAM APDs with one or two intermediate grading layers have recently been developed by Holden and others (Holder et al., 19851. These devices have dark current gains of 60 and gain-bandwidth products of 60 GHz and have captured all of the world records in receiver sensitivity experiments at bit rates exceeding 400 Mbit/s (Kas- per, 1986~. For example, recent tests at 4 and 8 Gbit/s at A = 1.5 ,um with a GaAs FET front-end have yielded receiver sensi- tivities of—31.2 and -26 dBm, respectively. The best values obtained at 420 and 2 Gbit/s are instead -44 and -36.6 dBm. These sensitivities are 5-10 dB better than those achieved with PIN photodiodes. To obtain an even higher sensitivity, the dark current should be further reduced, and the electron/hole ionization rate ratio should be increased (in InP SAM APDs, ,l~la = 31. The reduction of the dark current should produce a large sensitivity increase compared to PIN-FET receivers at low bit rates (<400 Mbits) as well. Such a reduction can be achieved by means of the recently demonstrated Hi-Lo SAM APD (Capasso et al., 1984a). In this device, a thin, heavily doped region is introduced in the lightly doped InP avalanche layer so that the electric field drops to a low value in the InP and GaInAs regions immediately adjacent to the heterointerface. In this structure,

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

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-

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

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.

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

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.

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

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/gO°K (SUPERLATTICE) _ i 2x 10-5 V (COOK) - ~. ~ ~ · · · ~ 5 x 10-3V (300K) _ tl f EgOOK (GaO 47lno S3AS) g°K (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.

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.

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

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 was—55.3 and—44 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.

Next: Lasers in Medicine »
Lasers: Invention to Application Get This Book
×
Buy Paperback | $50.00
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

Since the initial laser beam in 1960, use of lasers has mushroomed, opening new frontiers in medicine, manufacturing, communications, defense, and information storage and retrieval. Lasers: Invention to Application brings together a series of chapters by eminent scientists spanning the broad range of today's laser technology.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!