Degradation and Reliability of Semicondcutor Lasers
This is a continuation from the previous tutorial - infrared and visible semiconductor lasers based on other material systems.
Introduction
The performance of semiconductor lasers can degrade during their operation. This degradation is usually characterized by an increase in the threshold current that is often accompanied by a decrease in the external differential quantum efficiency.
The dominant mechanism responsible for this degradation is determined by one or several of the fabrication processes including epitaxy, device processing, and bonding. In addition, the degradation rate of lasers processed from a given wafer depends on the operating conditions, namely, the operating temperature and injection current.
Although many of the degradation mechanisms are not fully understood, an extensive amount of empirical observations exists in the literature. These observations have allowed the fabrication of InGaAsP laser diodes with an extrapolated median lifetime in excess of 25 years at an operating temperature of 10°C.
Detailed studies of the degradation mechanisms in injection laser diodes have been motivated by the desire to have reasonably accurate estimates of the operating lifetime before using the diodes in practical systems. For many applications the laser sources are expected to operate reliably over a period in excess of ten years.
An appropriate reliability assurance procedure becomes necessary, especially for applications such as undersea lightwave transmission systems, where the laser replacement cost is very high.
Reliability assurance is usually carried out by operating the laser under a high stress (e.g., high temperature) that enhances the degradation rate so that a measurable value can be obtained in an operating time of a few hundred hours.
The degradation rate under normal operating conditions can then be obtained from the measured high-temperature degradation rate using the concept of an activation energy. The purpose of this tutorial is to discuss (i) degradation mechanisms and (ii) strategies for reliability assurance.
The degradation mechanisms can be separated into three categories. They are: (i) defect formation in the active region; (ii) catastrophic mirror damage at high power densities; and (iii) degradation of current-confining junctions.
It is important to mention at the outset that although reliable semiconductor lasers have been fabricated using the InGaAsP alloy system, the study of degradation mechanisms has been largely descriptive and the results in many cases tentatively interpretative.
Laser diodes are operated at high injected current densities, which create high-energy electrons and holes, thermal gradients, potential for strain fields, and a high nonradiative recombination rate inside the active region. These factors can promote the motion, multiplication, and growth of isolated defects into clusters, which can significantly degrade the performance of lasers.
Catastrophic degradation due to mirror damage usually occurs during pulsed, high-power operation. The surface of a crystal is an imperfect lattice with many dangling bonds that absorb impurities from the air. These absorbed impurities form defect sites that cause excess optical absorption. When the optical intensity at the laser facet exceeds a certain critical value, the localized heat at the mirror facet can be large enough to cause melting and hence destroy the facet.
As discussed in the strongly-index guided semiconductor lasers tutorial, index-guided buried-heterostructure lasers use reverse-biased junctions for confining the current to the active region. During the aging process under normal operating conditions, defects may be generated in the current-confining junctions, which reduces the ability of these junctions to confine the current to the active region. The observed increase in threshold current may then be due to increased leakage current.
Defect Formation in the Active Region
In this section we discuss observations on defect motion and the development of specific types of defect clusters in InGaAsP.
The high density of recombining electrons and holes and a possible presence of strain and thermal gradients can promote defect formation in the active region of the laser.
The defect structures that are generally observed are the dark spot defect (DSD) and the dark line defect (DLD). These defect structures were first observed in AIGaAs lasers and LEDs.
The DLD, as the name suggests, is a region of greatly reduced radiative efficiency of roughly linear form. It was first observed in the active region of an aged AIGaAs proton-stripe double-heterostructure laser.
The DLD appears as a dark linear feature crossing the luminescent stripe at 45° in degraded lasers. Since the active stripe is oriented along the (110) direction, the DLDs are oriented along the (100) direction.
An extensive literature exists on DLDs in AIGaAs lasers, from which the following conclusions can be drawn.
These defects are seen to form in regions that do not differ in radiative efficiency from the surrounding regions prior to the formation of the defect. Optically excited AIGaAs lasers exhibit similar DLD growth.
Transmission electron microscopy (TEM) studies have shown that the DLD structure observed in injection lasers and in optically pumped semiconductor lasers is identical and therefore results from the same process.
These observations suggest that the phenomenon of DLD growth is intrinsic to the material and is not related to (though may be influenced by) the migration of metals from the contact, and p-n junction, or the particular dopant used.
In addition to the DLD, dark spot defects (DSDs) are also observed. The DSD is also a region of reduced radiative efficiency in the active region but lacks the linear aspect of DLDs.
DSDs are generally associated with regions of poor epitaxial growth that grow in size or localized defects in the substrate that propagate to the epitaxial layers during device operation.
1. Experimental Techniques
DLDs and DSDs can be easily observed by means of electroluminescence, photoluminescence, or cathodoluminescence.
However, a detailed study requires the use of TEM techniques where the sample preparation is a nontrivial exercise and the quality of the results obtained is determined exclusively by the sample preparation expertise.
The defects can also be identified by techniques such as deep-level transient spectroscopy (DLTS), scanning electron microscopy (SEM), and the electron-beam-induced current (EBIC) mode of SEM. We now briefly describe these techniques and the results obtained.
2. Electroluminescence
The observation of nonluminescent regions or regions of reduced radiative efficiency inside the active region of a semiconductor laser requires the fabrication of a special laser structure.
Figure 14-1 schematically shows this laser structure, commonly known as the window laser. Typically 20-25 μm wide, the window is formed on the substrate side using photolithographic techniques.
Since the InGaAsP laser emits its light with photon energies smaller than the band gap of the InP substrate, spontaneous emission from the active region can be observed directly through the window.
The window-laser structure allows continuous monitoring of the luminescence efficiency of the active region and is also compatible with the normal "p-down" bonding configuration. Since luminescence occurs as a result of current injection (electrical excitation), it is referred to as electroluminescence (EL).
Very often, degraded lasers show dark regions in the active-stripe EL when observed through the window, as shown in Fig. 14-2. These dark regions sometimes appear in a random fashion without any particular crystallographic orientation. Both DLDs and DSDs have been observed in LEDs and in InGaAsP double-heterostructure material.
Most of the work related to the observation and generation of DLDs and DSDs under current injection has been reported for LEDs. Although LEDs with extrapolated room-temperature lifetimes in excess of 109 hours have been reported, DLDs and DSDs do occur in these devices, and when they occur, they indeed limit the lifetime.
Yamakoshi et al. have observed generation of the <110>-type defects at operating temperatures higher than 170°C.
Temkin et al. observed a similar degradation mode under storage aging at 220°C without current injection; Fig. 14-3 shows their results.
The <110> DLDs observed in their samples originated from inclusion-like defects that were traced to the first-grown n-InP layer and/or to its interfaces with the substrate and active layers. No <100>-type DLDs were observed.
Additionally, under thermal aging without current injection, no change in the \(I-V\) characteristic was seen, whereas under aging with current injection, the \(I-V\) characteristics indicated the presence of leakage current.
Veda and coworkers have observed <110>-type DLDs and DSDs in aged LEDs. The DLDs corresponded to misfit dislocations and were identified by TEM.
X-ray analysis of the DSDs showed that these regions are precipitate-like defects with an excess of In and P. It was suggested that such regions may be formed during epitaxy; for example, thermal decomposition of the substrate just prior to the growth of the n-InP buffer layer or thermal decomposition at the interface between the n-InP layer and the quaternary active layer.
3. Photoluminescence
Photoluminescence is the luminescence occurring under photoexcitation of the semiconductor.
A scanning photoluminescence technique was used in the first report of DLDs observed in InGaAsp. Figure 14-4 shows a schematic diagram of the apparatus used in these studies.
Electron-hole pairs can be created at a suitable depth from the surface of the sample by choosing the right excitation wavelength. The apparatus also allows for the measurement of the catastrophic damage threshold. The defects may be created and observed by focusing the beam at a given point of the sample for a suitable length of time.
Johnston et al. observed DLDs in both single-layer and double-heterostructure InGaAsP. The sample chosen for the experiment had a lattice mismatch of less than 0.1%.
Figure 14-5 shows the dependence of the minimum photoexcitation energy density (intensity x exposure time) required for degradation on the active-layer thickness.
The observed lower threshold for thick layers is interpreted as resulting from strain in the epitaxial layers increasing with increasing thickness. Dislocations can propagate more easily in the presence of a strain field.
The TEM studies of optically degraded samples revealed small-scale defect clusters oriented along <110> in the defective regions. The characteristics of the degraded region were analyzed in detail.
Dislocations clusters oriented along (110) were found in both the thick and thin samples even though the threshold power for degradation differed by several orders of magnitude.
4. Cathodoluminescence
The cathodoluminescence imaging technique has been extensively used to study dark defects in semiconductors.
The method involves exciting the semiconductor material with an electron beam typically obtained from a scanning electron microscope. The created electron-hole pairs recombine to emit light. The cathodoluminescence imaging technique involves guiding this radiation (using a light pipe) from the sample to a photomuliplier tube located outside the SEM chamber.
Figure 14-6 shows a schematic of a cathodoluminescence imaging system.
Figure 14-7 shows examples of spatially resolved images of the light-emitting region in the p-InP confining layer of different LEDs. Dark, circular, or oblate features represents examples of DSDs. Chin et al. showed that the migration of gold from the p contact is responsible for the DSD formation in LEDs.
5. Dark Defects under Accelerated Aging
Accelerated aging techniques, which include high-temperature and high-power operation, are generally used to estimate the usable lifetime of injection lasers under normal operating conditions.
The generation rate of DSD and DLDs is enhanced under accelerated aging. When the accelerated aging is done at high temperature, an activation energy can be defined for the generation rate of defects in the active region.
The activation energy \(E_\text{a}\) is defined using
\[\tag{14-2-1}t_\text{d}=t_0\exp\left(\frac{E_\text{a}}{k_\text{B}T}\right)\]
where \(t_\text{d}\) is the generation time for the first defect, \(t_0\) is a constant, and \(T\) is the operating temperature.
Fukuda et al. have measured the generation time of DSDs and DLDs in InGaAsP gain-guided lasers operating at 1.3-μm and 1.55-μm wavelengths.
The generation of DSDs and DLDs was observed by electroluminescence using a window-laser configuration. The lasers that did not exhibit DSDs and DLDs operated for a long time without degradation.
Figure 14-8 shows the measured pulsed threshold current \(I_\text{th}^\text{p}(t)\) normalized to the initial value.
The increase in threshold current with time was associated with an increase in the number of DSDs and <100> DLDs. Furthermore, Ga- and As-rich regions in the active layer were correlated with the location of the DSDs.
A saturation in the number of DSDs and DLDs occurred in about 50 hours of aging at 250°C; beyond that time the increase in the threshold current caused by further aging also showed a tendency to saturate.
Moreover, samples without any DSDs or DLDs exhibited very little change in threshold even after 3,000 hours of aging.
It was concluded that dark defects were not created during aging and that their origins were already present in samples that exhibited DSDs and DLDs after a short-term aging (less than 100 hours) at 250°C.
Fukuda et al. found that the generation time of DSDs and the generation rate of <100> DLDs are weakly dependent on temperature, with activation energies of 0.16 eV and 0.2 eV, respectively.
However, they found that the generation time of the first DSD depends strongly on the operating current density. From their experimental result, they derived the following empirical relation:
\[\tag{14-2-2}t_\text{d}\propto\exp(-AJ^2)\]
where \(J\) is the current density and \(A\) is a constant.
The results on the aging characteristics of lasers differ considerably from those of where DSDs and <100> DLDs were observed in a surface-emitting LED. From the experimental observations, the generation time of the first DSD can be empirically expressed by the relation
\[\tag{14-2-3}t_\text{d}\propto{J}^{-2}\exp\left(\frac{E_\text{a}}{k_\text{B}T}\right)\]
where \(E_\text{a}\) = 1.2 eV.
The considerable difference in the activation energies of DSD generation reported by the two groups would suggest that different mechanisms of DSD generation are operative in the two structures. Which mechanism is actually dominant may relate to details of device fabrication, including processing, bonding, and epitaxy.
Catastrophic Degradation
For the purposes of this section, we define catastrophic degradation as the sudden degradation in the performance characteristics of laser diodes associated with the application of a large current pulse.
This phenomenon has been studied in detail in AIGaAs double-heterostructure lasers and epitaxial layers. For AIGsAs lasers the degradation is caused by strong absorption of the stimulated emission at the facet, which causes local heating and subsequent melting of the material near the facet.
The degraded-facet region typically exhibits dislocation networks and multiple dislocations loops, which are thought to be generated during the cooling of the molten region.
Catastrophic degradation in InGaAsP-InP double-heterostructure material under intense optical excitation has been studied. The threshold power for catastrophic damage is about one order of magnitude larger in InGaAsP material than in AIGaAs material.
Catastrophic degradation of InGaAsP lasers following a large current pulse has also been studied. Generally, no facet degradation, which is characteristic of melting at the active region (similar to AIGaAs lasers), has been observed. This may be partly due
to a smaller surface recombination rate in InGaAsP than in AIGaAs.
In cases where facet degradation has been seen, it extended from the top contact to the p-InP cladding layer, implying that heating was caused by a large amount of current passing near the facet.
In many cases the degradation may not be catastrophic. Then the increase in the threshold current is accompanied by a decrease in the external differential quantum efficiency and by a "soft" turn-on in the \(I-V\) characteristics that is characteristic of a leakage path.
This is especially the case for buried-heterostructure lasers that utilize reverse-biased junctions for lateral current confinement.
Veda et al. have measured the current needed to achieve a certain level of degradation as a function of the pulse width for V-grooved substrate buried-heterostructure lasers. Figure 14-9 shows their results.
The lasers did not degrade below the current levels denoted by circles; above those points (neighboring crosses) catastrophic degradation did occur. For 0.1-μs current pulses, degradation did not occur up to the maximum available current of 20 A.
Figure 14.9 shows that the current required for degradation decreases with increasing pulse width.
The degraded regions of catastrophically degraded diodes were analyzed using selective defect etching and energy-dispersive x-ray spectroscopy; contact metals (e.g., Au) in the core of degraded regions were found.
It is suggested that catastrophic degradation may be caused by local degradation of some internal current-confining region that can result in a large localized leakage current and subsequent melting.
Degradation of Current-Confining Junctions
As discussed in the strongly index-guided semiconductor lasers tutorial, many high-performance laser structures utilize current-restrictive layers so that most of the injected current flows through the active region. Effective current injection to the active region is necessary in order to obtain low threshold currents and high output powers.
A mode of degradation associated with buried-heterostructure lasers is an increase in the leakage current (the current flowing outside the active region) under accelerated aging. This in turn increases the device threshold and decreases the external differential quantum efficiency.
An increase in leakage current usually appears as a soft turn-on in the \(I-V\) characteristics of the laser. An easily recognizable signature is a "bump" in the \(I\text{d}V/\text{d}I-I\) characteristics, as shown in Fig. 14-10.
We have seen in the strongly index-guided semiconductor lasers tutorial that such a bump is indicative of a resistive shunt path. An increase in leakage current not only increases the device threshold but also generally decreases the external differential quantum efficiency.
The observations of electron-beam-induced current (EBIC) at laser facets is a useful technique in detecting defects at current-confining junctions.
In the buried-heterostructure laser (Fig. 5-20 in the strongly index-guided semiconductor lasers tutorial), the p-n junctions help in confining the current to the active region.
Figure 14-11 shows the aging behavior of three buried-heterostructure lasers along with their EBIC images. The bright regions indicate EBIC signals from the p-n junctions, while the dark regions show the presence of defects. The defect can be either the absence of a junction (growth defect) or an electron (or hole) trap generated during aging.
Reliability Assurance
In some semiconductor laser applications the system design lifetime is long (~ 20-25 years), and the replacement of components (e.g., lasers) can be prohibitively expensive.
An example is provided by the repeaters of an undersea lightwave transmission system where the failure of just a few lasers can cripple the system. Thus it is important to have a strategy for establishing the expected operating lifetime of a semiconductor laser.
Laser failure can occur by several means, including (i) infant mortality (lasers with grown-in defects that degrade very quickly); (ii) chance failures, these could be catastrophic damage due to external factors; and (iii) a gradual degradation mechanism in which some characteristic of laser operation (e.g., the threshold current) changes slowly with time.
In the last case an estimate of the operating lifetime can be obtained by monitoring that characteristic parameter.
Examples of infant failures are lasers with DSDs or DLDs already present in the active region at the time of fabrication or lasers with defects generated in the buffer layers during epitaxy that quickly propagate to the active region, resulting in a decrease in luminescence or an increase in threshold current.
1. Stress Aging
Some lasers inhibit an initial rapid degradation after which the operating characteristics of the lasers become stable. Given a population of lasers, it is possible to quickly identify the "stable" lasers using a high-stress test (also known as the purge test).
The assumption behind the test is that operating the laser under a set of high-stress conditions (e.g., high current, high temperature, or high power) causes weak lasers to fail and possible winners to stabilize.
The concept is substantiated by the observations of Fig. 14-8, where it was found that in some samples the DLDs and DSDs grew rapidly under stress aging (less than ~ 50 hours at 250°C). Associated with DLD and DSD generation was a rapid increase in the threshold current. It was also observed that after an initial increase, the number of DLDs and DSDs saturated.
Similar degradation characteristics have been observed for 1.3-μm InGaAsP buried-heterostructure lasers; the results are shown in Fig. 14-12.
The lasers were aged at a constant power output of 5 mW/facet at 60°C. The change in the operating current with the aging time was initially rapid and thereafter it slowed down considerably, similar to the observations of Figure 14-8.
The initial rapid degradation ("incubation period") took place in the first 500-1,000 hours. A high-stress aging test can considerably reduce the aging time needed for the completion of this initial rapid degradation mode.
The increase in the operating current after stress aging has also been measured. Figure 14-13 shows the data. The stress conditions required CW operation of the laser at 100°C with a 250-mA current.
After 0, 20, and 40 hours of stressed aging the operating current for obtaining 3-mW/facet power at 60°C was measured. The lasers were of the buried-heterostructure type.
As Fig. 14-13 shows, some lasers exhibited an increase in operating current before stabilizing (similar to results of Figure 14-8) while others exhibited stable characteristics without a significant increase in the opening current.
Figures 14-8 and 14-13 show that the high-stress test can be used as a screening procedure to identify robust devices.
Determining the duration and specific conditions for stress aging is critical to the success of this screening procedure.
The aim is
- To identify the weak lasers, which would fail in a short time.
- To stabilize the initial high-degradation mode that may be present in some lasers and may give misleading indications of long-term reliability.
- To select lasers whose degradation is governed by a slow acting, long-term mechanism that can be thermally accelerated to allow for a determination of the expected operating lifetime.
2. Activation Energy
Central to the determination of the expected operating lifetime is the concept of thermally accelerated aging, the validity of which for AIGaAs injection lasers was shown by Hartman and Dixon.
The lifetime, at a temperature T is empirically found to vary as
\[\tag{14-5-1}\tau=\tau_0\exp\left(\frac{E_\text{a}}{k_\text{B}T}\right)\]
where \(E_\text{a}\) is the activation energy and \(\tau_0\) is a constant.
A similar relation also holds for the degradation rate.
Equation (14-5-1) shows that by operating the laser at high temperatures (60-80°C), it is possible to establish the expected operating lifetime under normal operating temperatures (typically in the range of 10-25°C). This allows the determination of expected lifetimes of 10-25 years using a high-temperature aging time of ~ 500-1,000 hours.
It has been pointed out that the simple Arrhenius-type relationship [Eq. (14-5-1)] may not be observed under all aging conditions and that the measured activation energy may differ under different aging conditions and temperature ranges.
Measurements of the lifetimes of InGaAsP-InP buried-heterostructure lasers have been reported. Figure 14-14 shows the lifetime values plotted on a log normal chart.
The aging rates were measured at 50 and 70°C under a constant power output of 5 mW/facet. The horizontal bars indicate the lifetime estimated from the measured increase in the drive current after 8,500 h of operation.
The failure criterion (for Fig. 14-14) was defined as a change \(\Delta{I}\) in the drive current of 30-50 mA. The median lifetimes \(\tau_\text{m}\) for 50°C and 70°C are \(1.4\times10^5\) hours and \(1.7\times10^4\) hours, respectively.
Assuming an Arrhenius-type relationship given by Eq. (14-5-1), the activation energy is estimated to be 0.9 eV.
Hakki et al. have estimated the activation energy of 1.3-μm InGaAsP buried-heterostructure lasers from a measurement of the degradation rate of lasers operating at the same output power (3-5 mW/facet) at two different temperatures.
The degradation rates were measured after the initial degradation mode was completed, i.e., beyond the "knee" in Fig. 14-12. The activation energy is obtained assuming an Arrhenius-type relationship for the degradation rate \(R\); i.e.,
\[\tag{14-5-2}R=R_0\exp\left(\frac{-E_\text{a}}{k_\text{B}T}\right)\]
where \(R\) equals the rate of change of the operating current for 5 mW/facet output power.
The activation energy is then given by
\[\tag{14-5-3}E_\text{a}=\frac{k_\text{B}T_1T_2}{T_2-T_1}\ln\left(\frac{R(T_2)}{R(T_1)}\right)\]
Figure 14-15 shows the activation energies obtained from 26 sets of measurements. The median value of the activation energy is 1 eV, and the standard deviation \(\sigma\) is 0.13 eV.
Figure 14.15 also indicates that 96% of the values of the activation energy fall between 0.76 and 1.28 eV.
Table 14-1 gives the means, medians, and standard deviations of the measured values of \(E_\text{a}\) in different temperature ranges. The mean activation energy is relatively invariant in the temperature range of 40-80°C. This observation shows that it is reasonable to estimate the actual operating lifetime from the measured high-temperature degradation rate.
The mean time to failure (MTTF) is an important measure of the reliability of semiconductor lasers. From the data in Fig. 14-14, the MTTF extrapolated at 25°C exceeds \(10^7\) hours.
Experimental data on the reliability of buried-heterostructure lasers are shown in Fig. 14-16. Normalized operating currents are shown for 14 lasers screened for ~ 7,000 hours at 60°C and 3 mW/facet. The maximum degradation rate observed is 3.1% per 1,000 hours.
The measured degradation rate can be used to obtain an expected MTTF using a 50% change in \(I/I_0\) as a failure criterion. The MTTF at an operating temperature of 10°C (the ocean-bottom temperature) is then obtained using Eq. (14-5-1) and an activation energy of 0.9 eV.
The distance between arrows in Fig. 14-16 (1,200 h of burn-in time at 60°C and 3 mW/facet) represents 25 years of equivalent operating lifetime at 10°C, which is the expected cable lifetime and temperature of submarine lightwave systems.
It is evident from Fig. 14-16 that state-of-the-art 1.3-μm InGaAsP lasers can meet the system lifetime requirement (25 years at 10°C) if high-stress aging tests are used to eliminate the poor devices.
DFB Laser Reliability
A DFB laser should emit light in a single longitudinal mode during its entire lifetime. Reliability of DFB lasers is governed not only by the MTTF but also by the spectral stability.
A parameter that determines the performance of a DFB laser is the mode suppression ratio (MSR), i.e., the ratio of the intensity of the dominant lasing mode to that of the next intense side mode.
DFB lasers have been fabricated that have aging rates comparable to that shown in Fig. 14-16. An example of the spectrum before and after aging is shown in Fig. 14-17. Similar measurements have been done as a function of current.
These measurements show that the MSR of good DFB lasers with low light-emitting aging rates does not change significantly after aging, confirming spectral stability of the emission.
For some applications such as coherent lightwave systems, the absolute wavelength stability of the laser is important. Measured changes in the emission wavelength at 100 mA before and after aging of several devices are shown in Fig. 14-18.
Most of the devices do not exhibit any change in wavelength, and the standard deviation of the change is less than 0.2 nm. The absolute wavelength stability of these devices is adequate for coherent transmission applications.
Another parameter of interest in certifying the spectral stability of a DFB laser is the change in the dynamic line width (or frequency chirp) with aging. Since the chirp depends on data rate and the bias level, a certification of the chirp stability is, in general, tied to the requirements of a specific transmission system.
A measure of the effect of chirp on the performance of a lightwave system is the dispersion penalty. Measurements of dispersion penalty at 600 Mb/s for a total dispersion of 1,700 ps/nm have been done to study the effect of aging on DFB lasers.
The median change in dispersion penalty was less than 0.1 dB, and not a single device showed a change larger than 0.3 dB. This suggests that the dynamic line width or the chirp under modulation is stable.
DFB lasers are currently being used in many commercial systems. The lasers have proven to be adequately reliable even for the stringent requirements of undersea transmission.
The lifetime for single-wavelength operation is assured by measuring the optical spectrum of the device at fixed intervals during the certification process. For applications where replacement costs are very high, the development of a certification process for the assurance of single-wavelength operation over the lifetime of the system is very valuable.
The DFB lasers now available are robust enough for submarine cable applications. A transatlantic lightwave system (TAT-9), employing DFB lasers modulated at 590 Mb/s, became operational in 1992.
Since their initial demonstration in 1962, semiconductor lasers have come a long way in terms of their reliability.
The next tutorial introduces the history of digital coherent optical systems.