The JTEC panel studied six applications-based categories of lasers: (1) transmission lasers, (2) pumping lasers for erbium (Er)-doped fiber amplifiers, (3) local-loop or access lasers, (4) analog lasers for CATV or other subcarrier multiplexing applications, (5) on-premises or interconnect lasers, and (6) visible, shorter-wavelength lasers for optical storage, sensing, or display. (The first three applications tend to be dominated by common carrier and telecommunications companies.)
Included in the transmission lasers category are photonic integrated circuits (PICs), wavelength-division multiplexed (WDM), and time-division multiplexed (TDM) laser sources, and 1.3 micron and 1.55 micron wavelength devices based on InP. The pumping lasers category contains a brief review of work on 1.48 micron and 0.98 micron pump lasers for erbium-doped fiber amplifiers (EDFAs). The local-loop and access lasers category focuses primarily on uncooled, unbiased 1.3 micron laser development for fiber-to-the-home (FTTH), fiber-in-the-loop, (FITL) and other "access" applications. The analog lasers category is dominated by the cable TV industry, although there is a segment of interest in satellite communications and phased-array radars. The on-premises or interconnect lasers category tends to be dominated by customer equipment encompassing the emerging markets of intercomputer and intracomputer interconnection driven by multimedia applications. However, the massively parallel architecture proposed for intracomputer interconnection is also being explored for optical switching applications. Wavelengths for this category tend to range in the visible to 0.98 micron area, based on gallium arsenide (GaAs) substrate technology, although, some would still like to use 1.3 micron devices, even for these short-link applications. Finally, the visible lasers category primarily includes work on red lasers for optical memory (e.g., in compact discs), pointer, and scanner applications.
Table 5.1 lists these various kinds of lasers and the Japanese companies that make them. Many of the leading results in these areas can be found in proceedings of recent conferences, such as the International Semiconductor Laser Conference recently held in Hawaii (IEEE 1994). No ordering of the companies indicates the level of effort.
Transmission lasers are used in the core of the telecommunications network in Japan and elsewhere worldwide, typically to transmit highly multiplexed data over relatively long distances. Such lasers are currently being developed for use in high-data-rate time-division multiplexed systems and in newer architectures using wavelength-division multiplexing. Much of the laser technology emerging for optical fiber communications in Japan is driven by its largest internal telecommunications equipment customer, Nippon Telephone and Telegraph (NTT). Although recently, Japan's domestic markets have been open to competitive ventures, and foreign markets are also important, NTT still creates most of Japan's demand for optical fiber transmission systems and components. Japan's overseas telecommunications company, KDD, is active in submarine cable, and it generates a market for long-haul transmission components, but NTT seems to dominate the general directions of new device development in the transmission area. Companies most actively involved in product and systems development are NEC, Fujitsu, Oki, Hitachi, and Toshiba, but any company is free to bid on an NTT "open tender," its public call for proposals for a given system. (See Chapter 2 for a fuller description of NTT's bidding system.)
Time-division multiplexing is the most common digital technique common carriers use to transmit many telephone conversations simultaneously. Data rates are continuing to rise, although the limitations of electronics and the desire to be able to tap off individual time slots inexpensively in loop architectures has introduced pressure to limit the data rate to
Japanese Diode Laser Efforts
* University efforts are shown in parentheses
some value. Additional capacity could be obtained by using wavelength-division multiplexing. Presently, the highest TDM data rate used is 2.5 Gbit/s, but there are plans in Japan to develop 10 Gbit/s TDM links. In the United States there seems to be more interest in using WDM above 2.5 Gbit/s. The reasons for this disparity partially derive from the fiber base already established in each country. In the United States good 1.3 micron fiber is in place. It is not very suitable for high-bit-rate, long-haul transmission at 1.55 microns. In Japan, much of the existing fiber is not of the highest quality; thus, Japanese firms contemplate adding new dispersion-shifted fiber that is designed for high-bit-rate transmission at 1.55 microns.
Lasers for high-speed time-division multiplexing applications
1.3 micron and 1.55 micron FP and 1.55 micron DFB lasers. TDM lasers currently in production include 1.3 micron and 1.55 micron wavelength Fabry-Pérot (FP) lasers, as well as 1.55 micron distributed feedback (DFB) lasers. The 1.55 micron DFB lasers are used for the longest-distance links where both fiber loss and dispersion are factors, such as undersea links. The narrow linewidth of the DFB limits the effects of dispersion, and the fiber loss is minimal at 1.55 µm. Moreover, EDFAs operate at this wavelength.
Research in this area is directed at obtaining lasers with low parasitics, high intrinsic modulation bandwidth, and low spectral chirp (i.e., excess spectral bandwidth caused by modulation). In Japan, NEC, Oki, Fujitsu, and Hitachi are carrying out concentrated work on high-speed lasers. In this field, the United States holds the record for highest modulation bandwidth at 1.55 micron (Morton et al. 1994). However, all four of the Japanese companies just named have lasers that can be modulated to above 10 GHz, and this is more than sufficient for the proposed 10 Gbit/s systems. Because of chirp, many Japanese R&D efforts focus on integrating modulators with CW lasers, as will be outlined later. Experimental records for long-haul systems tend to go back and forth between AT&T and NTT, but it seems clear that Japan is somewhat ahead in developing 10 Gbit/s sources for the marketplace, as indicated by the high number of related publications by Japanese companies (OFC 1994 and 1995). Research in this area seems to be well poised to go into production within several Japanese companies.
Soliton laser sources. Some R&D work also focuses on developing mode-locking techniques to generate narrow-pulse trains, which must be gated in an external modulator to encode information. Narrow pulses are desired for soliton pulse transmission, in which the fiber dispersion is combined with nonlinear effects to generate a stable pulse width over very long distances. Solitons are being studied primarily for transmission over long uninterrupted paths, such as in undersea cables. NTT has been very active in this area. Recently, it propagated soliton pulses at a 80 Gbit/s rate over 500 km with periodic EDFAs and an EDF ring laser source (Nakazawa et al. 1994). AT&T's efforts in the United States have been sizable as well (primarily for undersea systems), and at present it would appear that the United States is slightly ahead of Japan in the development of soliton-based transmission systems. It seems clear that several Japanese companies will be able to provide 10 Gbit/s TDM systems in the near future. There is research on 40 Gbit/s TDM, and such systems might also be available within a few years. Here, the question is mainly one of architecture: it may be more desirable to use WDM to accomplish this net capacity level.
Lasers for wavelength-division multiplexing applications
In the United States, Bellcore and AT&T intend to develop WDM techniques to multiplex data where rates are higher than 2.5 Gbit/s. Several critical U.S. government-funded industry consortia have helped generate extensive efforts in recent years. In Japan there is also considerable activity in WDM sources for such systems. NTT has been carrying out several research efforts, and NEC, Hitachi, Fujitsu, and Toshiba have all developed some components. The key laser devices are tunable lasers or arrays of different-wavelength lasers. Issues are the control of the wavelength separation during manufacture and its subsequent absolute control under environmental variations. Equally important in these networks are tunable receivers or receiver arrays, as well as switches and fiber links that either re-encode the appropriate wavelength or are optically transparent.
Tunable lasers. In the area of tunable lasers, NEC has been active for many years in developing 3-section DBR devices, and Hitachi has recently demonstrated excellent results with a similar 3-section DBR structure. Many of the basic ideas were generated earlier at the Tokyo Institute of Technology under Professor Suematsu (1983, 161-176). Work at AT&T compares very favorably with the Japanese 3-section tunable DBR device (Koch et al. 1988).
NTT has recently experimented with a superstructure grating DBR laser that provides additional tuning range by using multiple reflection bands in two separate gratings (Tohmori 1992). The first work on this concept was done at the University of California at Santa Barbara (Jayaraman 1993), but no U.S. company seems to be pursuing it now. There is work being done in Europe on a variety of tunable lasers.
WDM arrays. Monolithically integrated WDM arrays of DFB or DBR lasers have been pursued by Toshiba, NTT, Hitachi, and NEC. In the United States, there has been leading work at Bellcore (Zah 1992), as well as at AT&T and some other places. Also, numerous WDM experiments have been carried out in Japan with discrete devices.
It appears that U.S. companies are somewhat ahead of the Japanese in development of tunable lasers, due to recent efforts. However, this position derives from a clear difference in focus. The Japanese seem to be moving directly to 10 Gbit/s before strongly considering WDM systems, whereas the U.S. companies seem to feel that WDM should dominate transmission above 2.5 Gbit/s. (The possible exceptions are MCI and Sprint, which appear more open to moving directly to 10 Gbit/s for intercity links.) As noted above, there are some bonafide reasons for the departure in strategy between the United States and Japan, due to differences in the embedded fiber base for most common carriers.
Photonic integrated circuits for TDM and WDM applications
The PICs in this source section have already been suggested above. In the WDM area, PICs are used in integrated arrays of laser sources and in multisection tunable laser structures. In the TDM area, the most widely researched PIC is a DFB laser integrated with an electroabsorption modulator. This is desired to reduce the chirp that normally accompanies the turn-on and turn-off transients. Some of the best device results have been reported in a recent publication from NTT that showed 20 GHz bandwidth modulation and 22 dB on/off with only 2 V peak-to-peak drive (Wakita 1994). Monolithic devices with 2.5 mW out modulated at 2.5 Gbit/s with ± 2.5 V drive on the modulator section have been used in recent AT&T systems experiments with EDFAs to transmit error-free data over 517 km (Johnson 1994). KDD, Hitachi, Fujitsu, Oki, and NEC have developed similar integrated laser-modulators, and they appear to be in a position to manufacture them in the near future. KDD is perhaps the biggest driver of this technology, and Hitachi seems to be in the best manufacturing position, owing to its outstanding selective area growth technology. The need for this kind of device seems to be well defined worldwide. (The majority of PICs are being researched for receiver applications, and these will be discussed below.)
The capabilities in the United States are equal to or stronger than those in Japan in this PIC area. However, the once-strong U.S. efforts at Bellcore seem to be fading, and the efforts in Japan seem to be more widespread and more closely connected to potential production facilities. In the future, the JTEC panel expects to see an increasing effort in Japan to make low-cost, reliable PICs for both WDM and TDM applications.
Both 1.48 micron and 0.98 micron FP lasers are being pursued at a number of Japanese companies for pumping EDFAs. Those companies working on 1.48 micron include NEC, Oki, Mitsubishi, and Furukawa. Those working on 0.98 micron include Fujitsu, Hitachi, KDD, and NEC. In both cases, the powers out are sufficient, but the lifetimes of 0.98 micron pumps are still being questioned. The shorter wavelength has some efficiency advantage, both in the diode and in the pumping process; thus, newer efforts seem to be focusing on this. The state-of-the-art performance for 0.98 micron in Japan is illustrated by some recent high-power, high-temperature results from Hitachi. The highest CW power reported was 466 mW, and a T 0 of 156 K was observed between 20 and 80 degrees C (Sagawa 1994). At 1.48 microns, Furukawa's result was about the best the panel found: it has obtained CW powers of ~360 mW, with threshold currents of ~30 mA (Kasukawa 1994).
The 1.48 micron diodes being produced in Japan are generally better than those of the United States. In the 0.98 micron area the choice is less clear, with at least one U.S. company (SDL) appearing to offer a somewhat better product (Gignac 1994). The panel expects to see a continuing effort to produce pump lasers with high reliability at lower cost. Many future all-optical network architectures will depend upon the presence of low-cost amplifiers.
1.3 micron FP lasers
"Access" lasers are those being developed for use in the final mile in accessing homes and businesses with high-bandwidth optical fibers; thus, they are the ones to be used in the widely discussed fiber-to-the-home (FTTH) and fiber-in-the-loop (FITL) architectures. These lasers must be low in cost and must be useable with simple circuits. Thus, it is generally assumed that they must operate without coolers and without any prebias. Low-threshold 1.3 micron Fabry-Pérot structures are being developed by numerous companies for this application. Fujitsu, NEC, Furukawa, Sumitomo, Toshiba, Hitachi, and Mitsubishi are heavily involved (see Yamamoto et al. 1994; Uomi 1994; Ae 1994). It is expected to be a big market worldwide, and some housing and business developments with fiber are already being constructed in Japan. It is believed there that demand for high-bandwidth optical fibers will come from multimedia applications. In fact, the multimedia market is expected to become one of the largest in Japan in the next century. The telecommunications industry is clearly counting on this to provide a big boost to its businesses in the coming years.
An example of the state of the art is the low-threshold laser recently developed at NEC-Ohtsu. This laser has been operated at 1 Gbit/s in data links over a temperature range of 20-100 deg. C without any prebias. Thresholds as low as 0.4 mA have been measured, and a very high-yield process has been developed (Ae 1994). Fujitsu has a similar result (Nobuhara 1994). This is the kind of performance required for the desired access laser. Of course, still other companies have competitive efforts, and it is expected that this kind of performance will be commonplace in a year or two.
In this area the Japanese seem to have a clear lead over the United States. Several companies are ready to move into production with viable structures. The JTEC panel is not aware of any U.S. company being in a similar state of readiness for production. In the next decade, the JTEC panel agrees that the multimedia market will generate a large demand for high-bandwidth fiber to homes and businesses. The multimedia business in Japan is estimated to reach over 5% of its GNP by 2010; thus, it would surpass the automotive industry in terms of both sales and employment. Of course, this would require a high- bandwidth backbone, so the transmission business discussed above would also be supported by the demand for multimedia services.
There are also research efforts exploring the use of vertical-cavity surface-emitting lasers (VCSELs) for access applications in several companies, including NEC, NTT, and Furukawa, as well as a university effort at the Tokyo Institute of Technology (TIT). Since the access network seems to be rigidly focused on using 1.3 micron and 1.55 micron wavelengths, however, the VCSEL effort for access applications is also limited to this wavelength range. Thus, until these long-wavelength VCSELs look practical, the level of effort will remain small.
1.3 micron DFB lasers for CATV
The biggest use of analog lasers currently is for cable television (CATV). Companies such as NEC, Mitsubishi, and Matsushita are producing 1.3 micron DFB lasers for CATV applications. The required high degree of linearity with associated low distortion is only possible in a small fraction of conventional cleaved-facet DFB lasers. This is because of the randomness of the facet location relative to the grating lines. NEC has developed a novel partially corrugated design that appears to improve the yield substantially (Okuda 1994). NEC results represent the state of the art in Japan.
The other major application of analog lasers is satellite links. Both military and commercial applications exist. Coaxial cable is relatively lossy and bulky, so optical fiber permits much more efficient remoting of the antennas. Here, analog TV information may be the most difficult information to transmit. Other wireless applications are also envisioned for the future. A cell-like architecture requires many land-based stations. These can be much more simple if analog encoding is used; thus, again, it may be desirable to use analog optical fiber links to connect to these remote cell-stations in a wireless environment.
Because of Ortel and AT&T, the United States is competitive in this area. If demand increases in the future, the United States should still be able to remain competitive. The panel does expect demand to remain high in the coming years, due to both CATV and antenna-remoting applications.
Interconnect lasers encompass devices used over relatively short distances, usually under the control of the user (rather than a common carrier) for more or less dedicated and specialized traffic (although multiuser local area networks would be common). On-premises interconnections of telecommunications switches would also provide a common carrier market.
The vertical-cavity surface-emitting laser was first pursued by Professor Iga of TIT in Japan, and several companies have begun small efforts in recent years. As the name implies, the vertical-cavity laser is a very short-cavity laser with a pair of mirrors on each side of an active layer, all grown sequentially on the surface of a semiconductor, so that the light recirculates and emits in the vertical direction. The key difference between the vertical-cavity and the conventional in-plane edge-emitting laser is that the VCSEL can be entirely fabricated and tested at the wafer level, whereas the edge-emitter is only completed once the wafer is cleaved into individual devices, thereby forming facet mirrors. Thus, the handling involved in making and testing the edge-emitters tends to be much higher, which increases the cost. Of course, edge-emitters can be made at the wafer level with grating mirrors or etched facets, but the processing is more complex. Again, the promise of the VCSEL is that it should be a low-cost, easily-packaged device for use in high-volume applications. Because VCSELs are naturally formed in arrays, they would seem to have advantages in this format. This is especially true for 2-D arrays, which can only be done with surface emitters.
VCSEL research is being carried out at NEC, Oki, Sanyo, Furukawa, Hitachi, and NTT. Japan's state of the art in the short wavelength region (0.8-1.0 µm) is a little behind that of the United States, but Japanese researchers have put more effort into exploring the long-wavelength regime. Nevertheless, U.S. efforts are comparable here also. Some of the better work in the short-wavelength area in Japan has been done by NEC. For example, NEC has recently reported a simple etched mesa structure with a threshold current of 190 µA (Numai et al. 1993). In the United States, Sandia has recently demonstrated devices with not only low thresholds, but also high wall-plug efficiencies (~ 50%) in the important 1-2 mW range (Lear et al. 1995). In the long-wavelength (1.3-1.6 µm) area, Japan's state of the art is comparable to or ahead of that in the United States. At TIT, Iga's group demonstrated the first room-temperature CW operation more than a year ago (Baba et al. 1993), but U.S. efforts have obtained significantly better threshold results using wafer bonding (Babi( et al. 1995). In the visible lasers area, the United States is clearly ahead, since only the efforts at Sandia have led to good results (Schneider et al. 1994).
The JTEC panel expects to see increased efforts in the future, once applications are identified. Hewlett-Packard is planning in the near future to introduce a parallel data link product to interconnect workstations using VCSELs. Motorola has a sample 10-channel parallel data link for sale. Vixel has VCSEL arrays for sale, and Honeywell plans to introduce some products soon. Parallel data links using VCSEL arrays have also been developed under government sponsorship in the United States. Thus, it would seem that numerous applications might soon emerge. It is also likely that long-wavelength VCSELs will become competitive; however, the work on access lasers will also continue to push progress in edge-emitters. If the market becomes large enough and if the advantages of wafer-scale processing can be utilized, then the VCSEL geometry may win. However, barring these events, the momentum is still on the side of the edge-emitters.
Edge-emitting array lasers
Edge-emitting arrays are also being proposed for use in parallel data links. Both GaAs- and InP-based arrays are being explored. At present the edge-emitters have a clear advantage in the long-wavelength regime. They can piggyback on the successes of the access lasers discussed above, and the VCSELs still are primitive. In the short-wavelength GaAs-based system, the contest is closer. Again, edge-emitters are well developed, but VCSELs are also performing extremely well and seem to possess clear advantages in manufacturability.
The state of the art in Japan in the long-wavelength regime is represented by recent work at Hitachi. Uomi et al. (1994, 2037) have recently reported very low threshold (0.56 mA) 1.3 micron devices with reasonable output powers, and already mentioned are similar results at NEC for access applications (Ae et al. 1994). This would appear to surpass U.S. activities. In the shorter-wavelength arena, there are a variety of data link activities. For example, NEC is developing 980-nm lasers for use with plastic clad glass fiber and 650 nm lasers for use with the new graded-index plastic optical fibers. Interconnect applications related to multimedia as well as automotive applications are perceived. U.S. efforts at 980 nm are competitive, but some of the best work is represented by university efforts (Zhao 1994), and technology transfer efforts must still take place. In the visible wavelength regime, the United States has fallen behind because of the thrust in Japan to develop sources for optical storage (e.g., CD lasers), which effort has no parallel in the United States.
In the future, edge-emitting 1.3 micron lasers will be developed for access and interconnect applications in parallel. The dual application will aid in making this longer wavelength more competitive for interconnect applications, where the shorter wavelengths might otherwise seem more natural. However, the continued development of shorter-wavelength visible sources for optical recording will also have a similar consequence for development of low-cost, reliable interconnect sources in the visible spectrum. Moreover, improvements in low-loss and low-dispersion (e.g., graded-index) plastic fiber will also push development of visible laser sources in the appropriate windows for interconnect applications, and some companies such as Deutsche Bundeposte are considering using 780 nm CD lasers for FTTH access applications. Thus, the argument for 1.3 micron might be used in a different way for the shorter wavelengths.
Optoelectronic integrated circuits (OEICs)
Monolithically integrated structures in this emitter laser category include integrated laser-drivers and "smart-pixel" circuits. (Most OEICs are associated with receivers.) In recent years, there has been a move away from purely monolithic integration in favor of hybrid integration using flip-chip bonding techniques; many of the applications once targeted for OEICs have gone this route. Only where very large numbers are perceived does the panel envision monolithic integration as being cost-effective. Fujitsu and NEC have been active in this area for some time.
In Japan there is some work, such as that by NTT and NEC, in integrating VCSELs with FETs and detectors to create smart pixels that could be useful in board-to-board interconnects for massively parallel computing or switching applications. Japanese efforts are at the basic research level and would seem to parallel similar efforts in the United States. However, the efforts in Japan appear to be better poised to move into production.
In the future, the panel expects that there will be continued modest development of OEICs, with applications coming only in the high-volume cases. These may appear in the interconnect area or in the optical storage area discussed below. For source integration, the panel expects to see continued use of flip-chip and MCM technologies to achieve the desired functionality with low-cost production.
780 nm compact optical discs
The use of lasers in compact optical disk players has brought about a revolution in the diode laser industry. For the first time, there is an application that requires production of more than a few thousand units per month. Thus, for the first time, the advantages of real mass production can be employed. These advantages are perhaps best illustrated by Rohm, a resistor manufacturing company that expanded into lasers. Through heavy use of automation generally foreign to the laser diode business, Rohm was able to capture half of the CD market, and in 1994 produced about 60 million laser diodes. This kind of demand has pushed the price down to about $2/laser (as of late 1994), even with the conventional cleaved-facet technology. Some argue that VCSELs would be much more manufacturable, but the market has to be large to justify this new technology at such small unit prices.
Besides Rohm, other major players in the CD market include Sony, Sanyo, and Sharp. Smaller efforts exist at Fujitsu, Matsushita, Mitsubishi, NEC, and Hitachi. Most of these look forward to the magneto-optic read/write technology being a new market requiring somewhat more demanding specifications (e.g., higher power) from the laser diodes. Matsushita has developed a phase-change reversible (PCR) media that it claims has the advantage of being more compatible with conventional CD technology. In fact, it announced a PD/CD-ROM product for 1995 that handles both PCR and CD disks.
Shorter-wavelength (630-680 nm) optical disc lasers
There is an evolution toward shorter wavelengths for optical storage. A new standard at 650 nm has already been established. The panel expects to see products using this standard in the next year or two. Together with some other improvements, about a factor of two in storage density is promised. This technology is based upon the AlGaInP/GaAs system, which is superior to the AlGaAs/GaAs system for wavelengths shorter than 780 nm. The ZnSe-like II-VI compounds and the GaN-like III-V compounds promise to move the wavelength into the blue range. These will be discussed in the materials section below.
Visible lasers for sensors and displays
Besides optical storage, applications for lasers in the 600-800 nm range lie in optical pointers, bar-code scanners, printers, data links, and display. The production of pointers relies upon visible light emission, and wavelengths in the 630-650 nm range are much better than ones at ~670 nm, even though higher power tends to be available there. This is due to the vast improvement in eye sensitivity. For the print heads and bar-code scanners an analogous argument holds, but here the shorter-wavelength lasers are preferable because of the sensitivity of the detecting medium. As already mentioned for data links, compatibility with plastic fiber (650 nm) is important. For display, again, 670 nm is a bit too long. Toshiba is a major player in these applications. In the future, the panel expects to see continued effort with AlGaInP red lasers, leading to reduced threshold currents and increased powers out in the 630 to 650 nm range. The Japanese are clearly ahead of the United States, but efforts such as those at Hewlett-Packard and SDL promise to make the United States competitive if the markets can be made available to the U.S. suppliers.
Integrated CD optical heads
There is also a significant amount of work in integrating CD lasers with other optical parts to make a more compact, robust and manufacturable optical head. Sony is working on a "laser coupler" that consists of a laser on a silicon submount containing a photodetector and circuit elements for preamplification and generation of servo signals. A molded microprism is also attached (Miyaoka 1994). Sharp uses a hologram half mirror and "OPIC," consisting of a photodiode, Si circuit, and CD laser in one element (Miyauchi et al. 1994).