Usinga specially designed semiconductor junction diode with heavy doping concentration,semiconductor lasers have been used to provide the reliable optical source required in fibre-based
Trang 1The rapidly growing information era has been augmented by a global network of opticalfibre [1] By offering an enormous transmission bandwidth of about 1014Hz and a lowsignal attenuation, the low-cost, glass-based single-mode optical fibre (SMF) provides anideal transmission medium In order that information can be carried along the SMF,information at the transmitter side is first converted into a stream of coherent photons Using
a specially designed semiconductor junction diode with heavy doping concentration,semiconductor lasers have been used to provide the reliable optical source required in fibre-based lightwave communication With its miniature size compatible to the SMF, thesemiconductor laser diode has played a crucial role in the success of optical fibrecommunication systems
This chapter has been organised as follows: in section 1.2, the historical progress ofoptical communication is presented Before exploring the characteristics of semiconductorlasers, various configurations of optical fibre-based communication systems are discussed insection 1.3 Depending on the type of detection method used, both direct and coherentdetection schemes are discussed Based upon the characteristics of coherent opticalcommunication systems, the performance requirements of semiconductor lasers arepresented at the end of the chapter In particular, the significance of having an opticalsource that oscillates at a single frequency whilst having a narrow spectral linewidth isreviewed
1.2 HISTORICAL PROGRESS
In the early days of human civilisation, simple optical communication in terms of signal firesand smoke was used In those days, only limited information could be transferred within line
Distributed Feedback Laser Diodes and Optical Tunable Filters H Ghafouri–Shiraz
# 2003 John Wiley & Sons, Ltd ISBN: 0-470-85618-1
Trang 2of sight distances In addition the transmission quality was strongly restricted byatmospheric disturbances This form of visual communication was extended and used inthe form of flags and signal lamps until the early 1790s, when a French scientist, ClaudeChappe [2] suggested a system of semaphore stations Messages were first translated into asequence of visual telegraphs These were then transmitted between tall towers which could
be as far as 32 km apart These towers acted as regenerators or repeaters such that messagescould be transmitted over a longer distance However, this method was slow and costly sincemessages had to be verified between each tower
With the beginning of a modern understanding of electricity in the 19th century, scientistsstarted to investigate how electricity might be used in long distance communication Thetelegraph [3] and telephone [4] were two inventions best representing this early stage ofthe electrical communication era During that period of time, optical communication in theatmosphere received less attention and the systems developed were slow and inefficient.The lack of suitable optical sources and transmission media were two factors that hinderedthe development of optical communication It was not until the early 1960s when theinvention of laser [5] once again stimulated interest in optical communication A lasersource provides a highly directional light source in which photons generated are in phasewith one another By modulating the laser, the coherent, low divergence laser beam enablesthe development of optical communication Due to the atmospheric attenuation, however,laser use is restricted to short distance applications
Long distance communication employing laser sources became feasible after abreakthrough was reached in 1966 when Kao and Hockham [6] and Werts [7] discoveredthe use of glass-based optical waveguides By trapping light along the central core of thecylindrical waveguide, light confined along the optical fibre could travel a longer distance ascompared with atmospheric propagation Despite the fact that the attenuation of the opticalfibre used was so high, with virtually no practical application at that time, this new way ofcarrying optical signals received worldwide attention With improvements in manufacturingtechniques and intensive research, the attenuation of optical fibre continued to drop Fibreloss of about 4.2 db km1was reported [8] for wavelengths around 1 mm, whilst low-lossfibre jointing techniques also became available
In order to build an optical communication system based on optical fibres, researchers inthe 1960s started focusing on the development of other optical components including opticalsources and detectors [9–11] A new family of optical devices based on semiconductorjunction diodes was developed By converting electrical current directly into a stream ofcoherent photons, semiconductor lasers are considered to be reliable optical laser sources.Based on similar working principles, efficient photodetectors based on the junction diodewere developed By responding to optical power, rather than optical electromagnetic fields,optical signals received are converted back into electrical signals In this early phase ofdevelopment, semiconductor lasers used were restricted to pulse operation at a very lowtemperature It was not until the 1970s that practical devices operating in continuous wave atroom temperature became feasible [12]
The availability of both low-loss optical fibre and reliable semiconductor-based opticaldevices laid the cornerstone for modern lightwave communication systems In the late1970s, lightwave systems were operated at 0.8 mm [13] Semiconductor lasers and detectorsemployed in these systems were fabricated using alluminium gallium arsenide alloy AlGaAs[14] Optical fibres used had a large core of diameter between 50 and 400 mm whilst typicalattenuation was about 4 dB km1 At the receiver side of the system, direct detection was
Trang 3used in which optical signals were directly converted to baseband optical signals Theoverall system performance was limited by the relatively larger attenuation and inter-modaldispersion of the optical fibre used.
In order to reduce the cost associated with the installation and maintainence of electricalrepeaters used in the lightwave communication systems, it was clear that the repeaterspacing could be improved by extending the operating wavelength to a new region between1.1 and 1.6 mm where the attenuation of the optical fibre was found to be smaller Figure 1.1shows the relation between the attenuation of a typical SMF and optical wavelength Forsystems operating at a longer wavelength, semiconductor optical devices were fabricatedusing quantenary InGaAsP alloy In order to avoid inter-modal competition associated withhigh-order oscillation modes inside the optical fibre, optical fibres having a smaller corediameter of about 8 mm were used In this way, oscillation in an optical fibre was reduced tosingle mode For systems operating in such a longer wavelength region, both wavelengths at1.3 and 1.55 mm have received a lot of attention For systems operating near 1.3 mm, it wasfound that the single-mode fibre used had minimum dispersion, and hence maximumbandwidth could be achieved In the early 1980s, many systems were built using single-mode fibre at around 1.3 mm wavelength An even lower fibre attenuation of about0.2 dB km1is found at around 1.55 mm However, the deployment of lightwave systems inthe 1.55 mm region was delayed due to the intrinsic fibre dispersion which limits themaximum bit rate the system can support The problem was later alleviated by adoptingdispersion-shifted or dispersion-flattened fibre [15,16] Alternatively, semiconductor lasersoscillating in single longitudinal modes were developed [17,18] By limiting the spread ofthe laser spectrum, this type of laser is widely used in upgrading the 1.3 mm lightwave
Figure 1.1 Attenuation of silica-based optical fibre with wavelength (after [44])
Trang 4systems to 1.55 mm wavelengths in which conventional single-mode fibres were used Since
1988, field trial tests for coherent lightwave communication systems have been carried out[19–21]
In order to improve the bit rate of the present lightwave system whilst utilising availablefibre bandwidth in a better way, frequency division multiplexing (FDM) schemes [22] wereimplemented Before information is converted into optical signals, electronic multiplexing isoften applied in combining the signals Such a system is normally referred to as coherentoptical communication since heterodyne or homodyne detection is used at the receiver end
By mixing the incoming optical signal with an optical local oscillator, coherent detectionemploys a different technique as compared with the direct detection method In the 1980s,the development of coherent optical communications was hindered due to poor spectralpurity and frequency instability in semiconductor lasers Due to advances in fabricationtechniques, semiconductor lasers nowadays show improved performance
In long-haul optical fibre communication systems, fibre dispersion and intrinsicattenuation are two major obstacles that affect the system performance In the 1990s,optical fibre communication systems continued to develop in order to tackle these obstacles
To circumvent the fibre dispersion, the non-linear optical soliton able to travel extremelylong distances was proven both theoretically [23,24] and experimentally [25,26] By usingoptical amplifiers [27,28] as pre-amplifiers, post-amplifiers and optical repeaters, onewitnesses the deployment of these wideband amplifiers in optical communication networks
In the coming years, networks employing a densely spaced wavelength divisionmultiplexing (WDM) scheme [29] are expected As a result, more channels and hencemore information will be transmitted over a single optical fibre link There is no doubt that anew paradigm of communication comprising an optically transparent network is already onthe way [30]
1.3 OPTICAL FIBRE COMMUNICATION SYSTEMS
By transferring information in the form of light along an optical fibre, a communicationsystem based on optical fibres starts to grow rapidly This system, like many othercommunication systems, consists of many different components A simple block diagram
as shown in Fig 1.2 represents the various components required in an optical fibrecommunication system At the transmitter side, information is encoded, modulated and isthen converted into a stream of optical signals At the receiver side, optical signals receivedare filtered and demodulated into the original information For long distance applications,repeaters or regenerators have to be used to compensate the intrinsic attenuation of opticalfibre In order to maximise the amount of information that can be transferred over a singleoptical fibre link, various multiplexing schemes might also be applied
To ensure successful implementation of optical fibre communication links, carefulplanning and system consideration is necessary Apart from the performance characteristics
of every component used within the system, it is also necessary to consider interactions andcompatibility between various components Depending on the system requirements, the type
of transmission (analogue or digital), required transmission bandwidth, cost and reliability,may vary from one system to another According to the type of detection method used at thereceiver end, it is common to categorise an optical fibre system into either a direct detection
or a coherent detection scheme
Trang 51.3.1 Intensity Modulation with a Direct Detection Scheme
Simply by varying the biasing current injected into a semiconductor laser diode at thetransmitter, the so-called intensity modulation with direct detection (IM/DD) scheme waswidely adopted The expression ‘intensity modulation’ derives from the fact that theintensity of the light emitted at the transmitter side is linearly modulated with respect to theinput signal for either digital or analogue systems The expression ‘direct detection’ is usedbecause the optical detector at the receiver end responds to optical power, rather thanelectromagnetic fields as compared to radio or microwave links In other words, all opticalsignals received at the optical detector are demodulated into baseband electrical signals Due
to its simplicity and low cost, the IM/DD transmission scheme has had great success, inparticular in point-to-point transmission systems In order to explore the potential of theoptical spectrum, however, coherent detection has to be used
1.3.2 Coherent Detection Schemes
Compared to the IM/DD transmission scheme, coherent optical communication [31–33] ischaracterised by mixing the incoming optical signal with the local oscillator so that thebaseband signal (for homodyne detection) or an intermediate frequency (IF) signal (forheterodyne detection) is generated at the receiver Since spatial coherence of the carriers andlocal oscillators is exploited, the expression ‘coherent’ is used to describe such a systemconfiguration The advantages of coherent detection have long been investigated and wererecognised in the 1960s [34], but it was not until the late 1970s that single-modetransmission from an AlGaAs semiconductor laser was demonstrated [35,36] With a
Figure 1.2 Simple block diagram showing various components for optical fibre communicationsystems
Trang 6narrower spectral output, fibre-based lightwave systems employing coherent detectionbecame feasible.
Various digital modulation methods have been used in coherent optical communication,including the amplitude-shift keying (ASK), the frequency-shift keying (FSK) and thephase-shift keying (PSK) methods [37,38] They differ from one another in the way digitalmessages can be transmitted by variations in amplitude, frequency and phase, respectively.For any digital transmission scheme and receiver architecture, a bit error rate (BER) in theregion between 109 and 1010 must be achieved at the receiver side for a satisfactorytransmission
The coherent optical communication system using homodyne/heterodyne detection hasseveral advantages over the IM/DD transmission scheme [39,40] First of all, coherentdetection can improve the receiver sensitivity by about 15 to 20 dB, depending on themodulation scheme adopted As a result, spacing between repeaters is improved for longdistance communication, whilst transmission rates can be increased in existing long distancelinks without reducing the repeater distance Moreover, by using modulation like PSK orFSK, which are well known in communication theory, the receiver can push to reach theideal quantum noise detection limit In addition, by adopting densely spaced frequency-division multiplexing (FDM) or wavelength division multiplexing (WDM), a wider fibrebandwidth can be utilised In practice, however, the coherent optical system has a stringentrequirement for device performance In Fig 1.3, a general block diagram for the coherentoptical communication system is shown
As illustrated in Fig 1.3, two injection lasers are involved in the system One acts as atransmitter and the other as a local oscillator The laser transmitter which acts as an opticalfrequency oscillator can be used directly in the FSK transmission An external modulator isoptional for the ASK and the PSK transmission before the optical signals are launched intothe single-mode fibre (SMF) Optical amplifiers like semiconductor laser amplifiers (SLA)
or erbium-doped fibre amplifiers (EDFA) are used in long distance transmission for boostingthe signal
Under the heterodyne receiver category with non-zero IF frequency, two different types ofpostdetection process have been adopted The name heterodyne receiver with coherentpostdetection processing (HE/CP) is usually given to one that has IF carrier recovered at thereceiver Similarly, heterodyne receiver with incoherent postdetection processing (HE/IP)describes the system that has no IF carrier recovered Comparatively, the HE/IP receiver
Figure 1.3 Schematic diagram for the coherent optical communication system
Trang 7configuration is the simplest as IF carrier reconstruction is unnecessary However, it showsthe weakest receiver sensitivity among the three receiver designs The incoherentpostdetection process could be used in conjunction with several modulation schemes such
as ASK, FSK and differential phase-shift keying (DPSK) In the HE/CP receiver design, IFsignals are recovered at the receiver stage for further signal processing The coherentpostdetection process can improve the receiver performance and so it is applicable to anymodulation method However, it is substantially more complicated than the incoherentmethod and stringent device performance is required
For zero IF frequency, the homodyne receiver has the best receiver sensitivity as data isrecovered directly from the optical mixing process at the receiver A narrower receiverbandwidth and only baseband electronic processing are required These offer significantadvantages to the homodyne receivers In practice, however, the technologies required inachieving these advantages in the homodyne receiver are demanding An effectivesynchronous demodulation process is essential in phase locking the local oscillator andthe received optical signal Phase jitters caused by phase noise and shot noise could impairthe system performance easily It has been evaluated [41] that the phase variance must belimited to within 10 to ensure a lower power penalty for a BER 109 This sets anupper limit on the permissible laser spectral linewidth and other laser performancecharacteristics In the coming sections, we are going to discuss some fundamental devicecharacteristics and their impact on system performance
1.4 SYSTEM REQUIREMENTS FOR HIGH-SPEED OPTICAL COHERENT COMMUNICATION
1.4.1 Spectral Purity Requirements
An ideal monochromatic laser source has been needed for some time As a result, thespectral purity of the laser source has often been the first issue confronting users ofsemiconductor lasers in coherent optical communication Due to the dispersive nature of theoptical fibres used, digital pulses are broadened whilst propagating along the optical fibre.Such pulse spreading causes adjacent pulses to overlap so that errors occur as a result ofinter-symbolic interference (ISI) Thus, apart from the power limitation due to the intrinsicfibre attenuation, the transmission distance is also limited by dispersion
The use of single-mode fibres has eliminated the severe inter-modal dispersion of mode fibres However, because of the finite spectral width of the optical sources, single-mode fibre is limited by chromatic dispersion (or intra-modal dispersion) Since the lasersources do not emit a single frequency but a band of frequency, each frequency component
multi-of the field propagates with a different time delay in the single-mode fibre, causing abroadening of the initial pulse width and hence intra-modal dispersion The delay differences
in single mode fibre may be caused by the dispersive properties of material through variation
in the cladding refractive index (material dispersion) and also the guidance effects within thestructure (waveguide dispersion) In order to minimise the effect of dispersion in single-mode fibre and hence improve the transmission distance, there are two different approaches.The first method involves the use of a dispersion-shifted or dispersion-flattened fibre With adistinctive refractive index profile, these fibres can reduce the effect of dispersion at the
Trang 81.55 mm wavelength significantly Another possible way involves the improvement ofsemiconductor laser sources The ability to lase in single mode with a narrow linewidth cancircumvent the effect of dispersion In the rest of this section, the concept of single-modeoperation, especially the possibility of a single longitudinal mode, will be discussed, whilstthe impact and the control of spectral linewidth will be left for later sections.
(a) Single mode along the transverse plane
It was shown in the previous section that the coherent optical communication systemrequires semiconductor lasers that can emit at a monochromatic frequency in order toachieve the required system BER As a result, it is necessary to achieve a single-modeoscillation in each of the orthogonal directions inside the laser cavity
To understand the transverse waveguiding problem of semiconductor lasers, one muststart with electromagnetic wave theory the basis for the study of electromagnetic wavepropagation is provided by Maxwell’s equations [42] For a medium with zero conductivitythe vector relationships may be written in terms of the electric field ~E and magnetic field ~Has
where " and are the permittivity and permeability of the medium The above equations areexpressed in the time harmonic form (with time variation term as ej!t) and are true forsource free and lossless media By using the vector identity and taking the curl on both sides
of eqn (1.1), one can arrive at the scalar wave equation for the electric field E such that
r2~E¼ k2~E¼ !2" ~E
¼ k2
where k is the propagation constant in the medium with the refractive index distribution of
n (x,y) and k0 is the free space propagation constant
Similarly, by taking the curl on both sides of eqn (1.2), one ends up with the scalar waveequation for the magnetic field H
r2~þ k2
Either eqn (1.3) or (1.4) can be used to determine the field components as they are related toone another by the Maxwell equations Nevertheless, the scalar wave equation for theelectric field is often used as the electric field is responsible for most physical processes and
it is the principal field used by photodetectors
To determine the transverse modal field of the semiconductor laser, one must first find thethickness and the refractive indices of materials used in the fabrication process Depending
on the specific laser structure, it is quite possible to have three or four epitaxial layers lying
on top of and below the active layer of the semiconductor laser These laser structures maylook complicated at first glance In fact, their waveguiding properties can be explained withthe use of a three-layer dielectric slab (or planar) waveguide As shown in Fig 1.4, the
Trang 9asymmetric waveguide consists of three layers The active layer, having refractive index n1
and thickness d, is sandwiched between the substrate and the cladding of the waveguide.Without loss of generality, it is assumed that the refractive indices of the slab waveguideobey the following inequality
where the equal sign implies a symmetrical waveguiding structure With such a planarstructure, the field variation along the y-axis can be ignored and so @=@y¼ 0 By separatingthe Maxwell equations into different field components, the following equations are obtained[42]
Figure 1.4 Schematic cross-section of a slab dielectric waveguide Refractive indices of differentregions are shown
Trang 10For a travelling wave propagating along the z direction, the electric field takes the form
where ~r is the radial vector in spaceðx; y; zÞ and the time harmonic term is omitted here forthe sake of simplicity bz is the propagation constant at a fixed angular frequency ! whichcan also be written as
with neff being the effective refractive index
The electric field component Eyfor different layers in the slab dielectric can be obtained
by substituting eqn (1.7) into eqn (1.3) and putting @=@y¼ 0 This is
Trang 11If p2and q2are negative whilst h2is positive (i.e 0 < bz< k0n3), the propagation is said
to be in the continuous radiation regime and the field solutions show sinusoidal behaviour inall three layers For a guided mode to occur, the constants p2, q2and h2must all be positive
In other words, the inequality k0n2<bz< k0n1 holds so that sinusoidal oscillation isrestricted to the central active layer while the electric field is decaying exponentially in otherlayers For the TE mode, a general formulation of EyðxÞ takes the form
A½h sinðhdÞ p cosðhdÞ þ B½p sinðhdÞ þ h cosðhdÞ ¼ 0 ð1:13bÞ
On combining the above equations and eliminating the arbitrary constants A and B, wearrive at the eigenvalue equation for the guided mode [43]
tanðhdÞ ¼ðp þ qÞh
From the above equation, it is clear that the active layer thickness d is decisive indetermining the guided mode By replacing n3 with n2 (hence q¼ p), the slab waveguidebecomes symmetrical and the above eigenequation becomes
¼p
tan hd2
¼ h
Trang 12Equations (1.16a) and (1.16b) represent the even and odd TE modes of the slab dielectricwaveguide In order to solve the eigenequation for the TE mode, an extra equation isnecessary By equating the propagating constants in eqns (1.10), one could obtain [44]
pd2
d2
2
where D is defined as the normalised waveguide thickness
Therefore, for a guided mode, the constants p and h must satisfy both eqns (1.14b) and(1.17) Using a graphical method, these equations can be solved Clearly, eqn (1.17)represents a circle with radius D on theðhd=2Þ ðpd=2Þ plane By putting in the odd and theeven TE mode eigenequations in the sameðhd=2Þ ðpd=2Þ plane, we obtain the plot shown
in Fig 1.6 Each intersection (or solution) with p > 0 and h > 0 between the circles and thetangent function corresponds to a guided mode, from which the propagation constant could
be determined according to eqn (1.15)
Due to the periodicity of the trigonometric function, multiple modes may occur as thefrequency, and hence the radius of the circle, keeps increasing From Fig 1.6, there is onlyone intersection point between the circle and the tangent functions when the normalisedwaveguide thickness D has value
0 < D <
Figure 1.6 Graphical method to solve the eigenequation for a symmetrical three-layer dielectricwaveguide (after [44])
Trang 13This mode, designated as the TE1, is the fundamental mode excited in the slab dielectricwaveguide By expanding D according to eqn (1.17), the inequality shown above becomes
2 1
n2
tan hd2
2 2
n2
h
Using the same graphical technique, the same limitation for the active layer thickness isobtained for single transverse mode excitation in the TM mode The only difference betweenthe TE and the TM mode is the refractive index correction term added in the TM modeeigenequations for the symmetrical slab waveguide
For most semiconductor lasers having lateral confinement, the width of the active layer W
is finite With the width dimension found comparable to the active layer thickness, arectangular waveguide is formed Nevertheless, one can still follow a similar procedure tothat applied to the slab waveguide in solving the field solution of the rectangular waveguide.Figure 1.7 shows the schematic cross-section of a rectangular waveguide The central core
Figure 1.7 Schematic cross-section of a rectangular waveguide Refractive indices of the differentregions are indicated
Trang 14region, which is surrounded by four cladding regions, has the highest refractive indexðn1Þ.With the propagation mode mainly confined to the central core, the field penetration in thecorners (shaded regions in the figure) of the structure can be neglected [45].
An exact analytical solution for the propagation characteristics of the strip waveguide isnot possible and a certain degree of approximation is necessary A very convenient techniqueknown as the effective index method (EIM) can be used in providing an accurate analyticalsolution for the rectangular waveguide In this method, the equivalent slab waveguide in onedirection is solved first (see Fig 1.8) so that the effective refractive index of the central slabcan be generated This effective refractive index is then used to solve for the other slabwaveguide which is found perpendicular to the original one In this way, the solutions for thetwo slab waveguides are coupled In a similar way to the slab waveguide analysis, thenumber of excited transverse modes is determined by both the active thickness (d) and width(w) of the central core region Therefore, it can be shown that proper controls over the activethickness and width are necessary for single transverse mode operation
To show how the EIM works, a symmetrical rectangular waveguide is used instead With
n3¼ n2and n5¼ n4, we can solve for the central slab first by using the graphical approachshown earlier in the TE mode of the slab waveguide For a single transverse mode along thex-axis, one ends up with the following inequality [44]
W¼ k0
w
2ðn2 eff n2
4Þ1=2<
Figure 1.8 Procedures in analysing rectangular waveguides using EIM (a) The first slabapproximation along the x-axis; (b) the second slab approximation along the y-axis with neff used
Trang 15(b) Single longitudinal mode (SLM)
In semiconductor lasers, electron movements occur between two energy bands that consist
of a finite number of discrete energy levels Rather than a discrete energy transfer like thegaseous laser, semiconductor lasers are characterised by a wider gain spectrum In aninhomogeneously broadened laser, the gain spectrum may be found several times wider thanthe longitudinal mode spacing In the simplest type of optical resonator, better described asthe Fabry–Perot (FP) cavity, photons escape from a cavity which has two partially reflectedmirrors facing one another Without any frequency discrimination, photons at any frequencycould escape from the cavity As a result, it is common for several oscillation modes, asdepicted in Fig 1.9, to be observed; this is particularly true when it is under rapidly pulsedoperation
As will be shown in detail in chapter 2, the longitudinal mode spacing between eachfrequency component in the FP semiconductor laser cavity is given as
f ¼ c
where c is the free space velocity, ng is the group refractive index and L is the laser cavitylength In order to have a better control over the longitudinal modes in the FP cavity, severalmethods have been proposed By using a shorter cavity one can increase the mode spacing.However, a higher threshold current is expected and several longitudinal modes may occur atsevere modulation Besides, as we will discuss later, such a shorter cavity will lead to awider spectral width [17] A different method involves an external light injection fromanother laser [47] In this way, single longitudinal mode (SLM) oscillation can be achieved.However, careful tuning is essential in order that the lasing modes of the two lasers bematched [48] By adding a reflecting surface outside the FP cavity, an external coupledcavity has also been used [49] Due to the interference with the external cavity, the overallmodal loss becomes frequency dependent and hence limits the oscillation mode from the FP