9
`
` Chapter 9
`
`A laser is a light
`osci a or, w ic
`qenerates its own
`I
`signal.
`
`The word laser was coined as an acronym for light am plification by the stimulated emission
`o f radiation, but that phrase glosses over the critical distinction between amplification and
`oscillation. A laser is a light oscillator, not a light amplifier.
`r r
`c
`.
`.
`.
`,
`.
`.
`,
`,
`.
`,
`.
`. ,
`An amplifier boosts the strength of an external signal, but doesnt generate a signal on its
`own. An oscillator generates a signal internally at a wavelength or frequency determined by
`its structure. The spark that starts laser oscillation is a spontaneously emitted photon, which
`stimulates emission of another photon, starting a cascade of other photons. A pair of mirrors
`on opposite ends of the device keeps the oscillation going. Light bounces back and forth
`between the mirrors, as shown in Figure 9.8, stimulating the emission of more photons on
`each pass. The pair of mirrors form a resonant cavity. One mirror reflects all light that strikes
`it, but the other mirror transmits a fraction of the light, which becomes the laser beam.
`Reflection of light back and forth between the mirrors makes it pass multiple times
`through more excited laser material, amplifying the light more than is possible on a single
`
`An initial spontaneous emission (black dot) stimulates emission of more photons.
`Mirrors at the ends of the laser cavity reflect light back and forth, building up
`stimulated emission. A fraction of the light leaks through a partly transparent
`mirror to form the laser beam.
`
`Stimulated emission generates coherent light, with all the waves
`lined up in phase.
`
`FIGURE 9.8
`Laser emission from a resonant cavity.
`
`MASIMO 2014
`PART 4
`Apple v. Masimo
`IPR2020-01526
`
`

`

`Light Sources m
`
`pass. The mirrors select light that bounces back and forth in a line between them, so only
`light aimed in that direction is amplified, which concentrates the light emission into a nar-
`row beam, as shown in Figure 9.8.
`Each laser material has its own characteristic gain, which varies with the wavelength
`and conditions in the laser medium. Stimulated emission produces the strongest ampli-
`fication at wavelengths where the gain is strongest, and laser oscillation further narrows
`the range of wavelengths. The laser structure determines how much the spectral width is
`narrowed.
`Optical Amplifiers in Fiber Optics
`Optical amplifiers deserve special attention here because they are very important in fiber-
`^
`optic communications. By amplifying a weak optical signal, an optical amplifier increases Optical^amplifiers
`the distance the signal can be transmitted. An optical amplifier is essentially a laser with-
`^
`k
`'
`I
`out mirrors at the ends. The light makes a single pass, instead of bouncing back and forth
`y
`between mirrors, and is amplified by the gain within the amplifier, as shown in Figure 9.7.
`Two types of optical amplifiers are particularly important in fiber optics.
`The semiconductor optical am plifier is a diode laser with its ends coated or integrated with
`semiconductor waveguides so they don’t reflect light. (Semiconductors have a high refrac-
`tive index, so they reflect some of the light trying to leave the crystal.) You’ll learn more
`about this amplifier in Chapter 12.
`The erbium-doped fib er am plifier is an optical fiber with erbium added to its core, as
`described in Chapter 7. It can amplify weak light signals that pass through it under the
`proper conditions. You’ll learn more about its operation in Chapter 12.
`We’ll turn now to lasers because this chapter is about signal sources. We focus first on
`semiconductor lasers, starting with the simplest common type, because they are the usual
`type used in fiber optics. Later in this chapter, you’ll learn about fiber lasers.
`
`Simple Semiconductor Lasers
`
`Like LEDs, semiconductor lasers are two-terminal devices called diodes in which holes flow
`through a p region and electrons flow through an n region to cause recombination in a junc-
`tion layer separating the regions. Generally, diode lasers and LEDs use the same materials.
`The key differences are in their manner of operation and in the internal structures that
`control their operation.
`LEDs produce spontaneous emission from electrons that release their surplus energy as
`they fall from the conduction band into the valence band. Diode lasers produce stimulated
`emission, which involves extracting light energy from the recombining electrons before
`they can spontaneously emit light. This extraction process requires concentrating the exci-
`tation energy to produce the population inversion required for laser action in the junction
`layer. This in turn requires a laser resonator, higher drive currents than those used in LEDs,
`and confinement of both the excitation and the generated light. We’ll start with the sim-
`plest type of laser, called a Fabry-Perot laser, and describe each of these factors, then move
`on to other types of lasers.
`
`Semiconductor
`lasers resemble
`LEDs in important
`ways.
`
`

`

`Fabry-Perot Laser Cavities
`As you learned earlier, laser action occurs when light bounces back and forth within a
`resonant cavity, stimulating emission from excited atoms. The simplest type of resonant
`cavity is a pair of parallel mirrors, as shown in Figure 9.8. This is called a Fabry-Perot
`cavity after the two men who first used this arrangement to observe the interference of
`light waves.
`In a gas laser, the two mirrors are on opposite ends of a tube containing the laser gas. In
`a semiconductor laser, the two mirrors are opposite edges of the semiconductor chip. One
`edge has a coating that reflects most of the light back into the semiconductor. (Typically a
`small amount of the light is transmitted so laser power can be monitored.) The opposite
`edge transmits more light, which emerges as the laser beam. (Semiconductors have high
`refractive indexes, which means they naturally reflect much of the light back into the solid
`and so may not require special reflective coatings.)
`Some semiconductor lasers are called edge emitters because the light emerges from the
`edge of the chip, not from the surface. As with LEDs, the light is generated in the junc-
`tion layer. Recall that edge-emitting LEDs are used in communications, as shown in
`Figure 9.5.
`Stripe-Geometry Lasers
`A stripe-geometry laser, shown in Figure 9.9, confines light both vertically and horizon-
`tally within the junction layer. The junction layer itself is thin, typically a fraction of one
`micrometer.
`The vertical confinement is created by making the p-n junction— called the active layer
`of the laser— of a semiconductor compound that has a refractive index slightly higher than
`
`A Chapter 9
`
`Fabry-Perot diode
`lasers emit from
`the edge of the
`chip.
`
`A stripe-geometry
`laser confines
`light in the
`junction layer.
`
`FIGURE 9.9
`A double-
`heterojunction
`stripe-geometry
`laser
`
`3 0 0 -5 0 0 |xm T yp ica l Len gth
`
`

`

`that of the p and n layers above and below it. This is done by changing its composition
`slightly, such as by adding a small amount of aluminum to gallium arsenide, which cre-
`ates a boundary with different refractive index called a heterojunction between the two
`layers. The refractive-index difference between the junction layer and the layers above
`and below it creates a waveguide effect, which confines light in the junction plane, just
`as the lower-index cladding confines light in the core of an optical fiber. This layered
`structure is called a double heterojunction or double heterostructure and was a crucial step
`in the development of semiconductor devices, for which Herbert Kroemer and Zhores
`Alferov shared the 2000 Nobel Prize in Physics.
`A stripe-geometry laser also confines laser action within a narrow stripe, typically only a few
`micrometers wide, in the junction layer. In telecommunications lasers, this area is usually a nar-
`row high-index stripe in the junction layer so the difference in refractive index guides light in
`the horizontal plane just as the heterojunction guides light vertically. This index-guiding
`limits the laser to oscillating in a single mode and matches the core size of single-mode fibers.
`Another approach is to limit current flow to a narrow stripe in the junction layer by
`depositing insulating layers that block current flow in other regions. This limits the popu-
`lation inversion to the narrow stripe where current flows, thus confining laser gain to the
`stripe. Lasers that only use gain-guiding do not confine light as well as index-guided lasers,
`but they are adequate for some purposes, and the two types of guiding can be combined in
`the same device.
`There are many variations in the internal design of stripe-geometry diode lasers. Several
`layers of various compositions may be used to control the flow of current and light. In gen-
`eral, the more tightly the layers confine light, the more efficiently they produce a popula-
`tion inversion in the junction layer, and the more efficient the laser. Quantum well
`structures fabricated in the junction layer constrain where electrons and holes can
`recombine, improving light confinement and enhancing the performance of diode lasers.
`We won’t go into depth on these designs because this book focuses on fiber optics.
`Instead, we will concentrate on differences in design that affect the function and perform-
`ance of diode lasers.
`
`Laser and LED Performance
`A first step in understanding diode laser performance is to compare the operation of lasers
`and LEDs. Like an LED, a laser requires a drive voltage greater than the bandgap voltage
`in order to generate light. Both diode lasers and LEDs must be forward-biased, with posi-
`tive bias applied to the p-type material and negative bias to the n material.
`A profound difference between lasers and LEDs is their behavior as the drive current
`increases from zero. At low currents, both devices generate some light by spontaneous emis-
`sion from recombining carriers, although lasers in general are inefficient. However, once
`the drive current exceeds a threshold value, the laser begins to generate stimulated emis-
`sion, which increases much faster with drive current, as shown in Figure 9.10. Above this
`threshold current, a diode laser generates light much more efficiently than an LED. No such
`threshold exists for an LED.
`The threshold is the point where the optical gain in the laser cavity exceeds the loss. As
`drive current increases, more carriers recombine, and are available for stimulated emission.
`
`Light Sources V
`
`A double
`heterojunction
`confines light in
`the junction plane.
`
`Laser operation
`begins above
`a threshold
`current.
`
`

`

`m Chapter 9
`
`FIGURE 9.10
`LED and laser
`power/current
`curves.
`
`Diode lasers are
`much more
`efficient than LEDs.
`
`This increases the gain within the laser resonator. Below threshold, the gain that light
`makes in a round trip of the laser cavity is lower than the losses it suffers from absorp-
`tion and light escaping through the end mirrors. At threshold, the gain exceeds the loss,
`and above threshold stimulated emission increases very rapidly with drive current, as
`shown in Figure 9.10. Although the curve looks steep, the increment in output power
`does not exceed the increment in input power. Laser efficiency can be measured in two
`ways, either overall efficiency comparing the output power to input power, or the slope
`efficiency, which measures the extra power generated per increment in drive current.
`Diode lasers are much more efficient than LEDs, with slope efficiencies that can reach
`tens of percent.
`LEDs don’t have resonant cavities, because their light-emitting surfaces are made to sup-
`press reflection, so they don’t produce stimulated emission or have a threshold. That means
`their output increases steadily as drive current increases from zero, but the rate of increase
`is much less.
`The threshold current is an important figure of merit for diode lasers. Below the thresh-
`old, most of the input energy must be dissipated as heat; above the threshold, much of the
`input energy emerges as light. In general, the lower the threshold, the better the laser’s
`efficiency and performance. Reducing the laser threshold also tends to increase laser lifetime
`because it reduces the heat dissipation and operating temperature.
`Another important difference between lasers and LEDs is that the laser emits a much
`narrower range of wavelengths. Figure 9.10 shows that spontaneous emission from an
`LED varies across a range of wavelengths. Stimulated emission varies in the same way,
`but the amplification process builds up the difference because photons at the peak
`
`

`

`Light Sources
`
`VCSELs emit from
`their surface,
`perpendicular to
`the junction layer.
`
`wavelength are more likely to stimulate emission than those away from the peak. You’ll
`learn more about how this works when we describe laser wavelengths later in this
`chapter.
`Vertical Cavity Diode Lasers
`The edge-emitting laser is a tried-and-true design, but its emission from the edge of the
`chip can be a significant practical disadvantage. Many edge-emitting lasers can be fabri-
`cated at one time on a single wafer, but the entire wafer must be cleaved before each
`individual laser can be mounted and tested. Lasers that emit from the surface of the chip
`can be tested while still on the wafer, making them easier to package and produce, and
`thus lower in cost. This has led to the development of the vertical-cavity surface-emitting
`laser or VCSEL.
`The resonant cavity in a VCSEL is perpendicular to the junction layer and vertical in
`the wafer, so the laser output emerges from the surface, as shown in Figure 9.11. This is
`a device rather different from an edge-emitting laser. Amplification in a VCSEL occurs
`only in the thin junction layer, so the light can be amplified by only a small amount in
`each pass through the laser cavity. In contrast, light in an edge-emitting laser passes
`through the entire length of the junction layer (a few hundred micrometers) on each pass,
`so it is amplified much more. To compensate for the lower gain within the VCSEL cavity,
`the mirrors must reflect more light than those used in edge-emitting lasers.
`
`FIGURE 9.11
`A vertical-cavity
`suface-em itting
`laser.
`
`L a se r O u tp u t
`
`M eta l C o n ta ct
`
`M etal C o n ta ct
`
`n-Type S u b stra te
`(tra n sp a re n t to
`la se r light)
`
`M u ltila ye r M irro r
`(p a rtly tra n sp a re n t)
`
`S p a ce rs
`Ju n ctio n
`
`Layer
`
`p- Type
`M u ltila ye r M irro r
`(to ta lly re fle ctin g )
`
`03o
`
`

`

`Chapter 9
`
`VCSELs can be
`made in two-
`dimensional
`arrays. They are
`easy to test and
`mount.
`
`LEI s are more
`reliable than
`edge-emitting
`lasers.
`
`Laser output
`declines with age.
`
`VCSEL mirrors normally are made with standard semiconductor processing techniques.
`They are composed of a series of layers with alternating compositions, so they selectively
`reflect a narrow range of wavelengths. Which wavelengths the mirrors reflect depends on
`the thicknesses and refractive indexes of the layers. (You may remember the concept from
`the discussion of fiber Bragg gratings in Chapter 7. It’s a basic principle of optics that was
`developed decades ago.)
`A VCSEL emits from a round spot typically 5 to 30 pm across on the surface of the
`wafer. This spot is larger than the core of a single-mode fiber, but smaller than the core of
`a multimode fiber. It is also larger than the output spot of an edge-emitting laser, so
`diffractive effects do not make a VCSEL beam spread out as rapidly. The beam is also
`circular, unlike the oval beam from an edge emitter.
`In addition to being easy to manufacture and package, VCSELs have low threshold cur-
`rents and are quite efficient in converting input electrical power into light. This means they
`consume less power and dissipate less heat than edge-emitters. They also have a longer life-
`time and can be directly modulated at data rates well above 1 Gbit/s.
`Unlike other diode lasers, VCSELs can be made in two-dimensional arrays covering the
`surface of a wafer, and these individual lasers can be modulated separately. Such arrays are
`attractive for optical switching and signal processing, to produce beams for transmission
`through optical fibers or free space.
`Currently, VCSELs are the favorite fiber-optic lasers for wavelengths of 750 to 1000 nm.
`Longer wavelengths have proved more difficult because materials that emit light at those wave-
`lengths don’t work well for the multilayer mirrors used in VCSELs.
`
`Laser Reliability
`Early GaAs lasers were unreliable, but great improvements have been made. Nonetheless,
`LEDs are more reliable than edge-emitting diode lasers because the lasers have higher
`current densities and optical power outputs. Threshold current can be an index of laser
`reliability; the lower the threshold, the longer-lived the laser. VCSELs, which have very
`low thresholds, are the most reliable lasers.
`Operating temperature is a major factor in laser reliability; increasing temperature short-
`ens lifetimes, and elevated temperatures are used in accelerated-aging tests. Threshold cur-
`rents increase with operating temperature, increasing the waste heat generated within the
`laser, which further increases temperature and degrades efficiency. Gallium arsenide is more
`vulnerable than InGaAsP to this problem, which can lead to thermal runaway. To control
`heat buildup and efficiency decreases, many laser transmitters are built with active temper-
`ature stabilization, such as thermoelectric coolers. Most lasers are packaged with heat sinks,
`even if active cooling is not required.
`Output power of diode lasers tends to decline slowly with age. To compensate for this
`decline, the transmitter can be designed to slowly increase drive current so the output
`power remains constant. A laser operated in this way is said to fail when it no longer
`delivers the required output power.
`Diode lasers are particularly vulnerable to damage from electrostatic discharges. Careful
`handling and proper packaging can overcome this problem, but you should be aware of its
`potential and always ground yourself when handling lasers.
`
`

`

`Light Sources
`
`Diode lasers are
`made on GaAs or
`InP substrates.
`
`Emission
`wavelength
`depends on the
`composition of the
`junction layer.
`
`Laser Wavelength
`
`The output wavelength of diode lasers is central to their use in fiber-optic systems. Both
`the peak wavelength emitted and the range of wavelengths are important for system per-
`formance. The composition of the semiconductor in the junction layer determines the
`wavelengths where a laser (or LED) can emit light. The device structure determines which
`wavelengths in that range the laser can emit. We’ll start by looking at the materials, then
`turn to the structures used to produce particular effects.
`
`Semiconductor Laser Materials
`Earlier in this chapter, you learned that the LEDs and diode lasers used in fiber-optic sys-
`tems are made of III-V semiconductor compounds. The laser structure consists of a series of
`layers on a substrate wafer. The composition of the substrate is chosen to be compatible with
`the composition of the other layers, which in turn are chosen to work with the composi-
`tion of the junction layer. In practice, substrates are made of two-element compounds, gal-
`lium arsenide or indium phosphide, which are easier to produce in bulk than the three- or
`four-element compounds used for the active layers.
`The principal compounds used are:
`
`9 GaQ^Al*As on GaAs for 780 to 850 nm
`•
`Iri(i_x)GaxAs on GaAs for 980 nm
`•
`In(l-x)GaJ(As(i_jr)Pjl on InP for 1100 to 1700 nm
`
`The subscripts indicate the relative fractions of each element. Indium, gallium, and alu-
`minum are all Group III elements and can be interchanged with each other. Arsenic and
`phosphorus are Group V elements and can be interchanged with each other. Compounds
`containing three elements are called ternary, and those containing four elements are called
`quaternary. As you will learn later, 980- and 1480-nm lasers are used to pump optical
`amplifiers, while others are used as signal sources.
`The composition of the active layer determines the peak gain wavelength. For example,
`the gain of InGaAsP peaks at 1310 nm for a composition of In0 73Ga0.27As0.58P0.42- As
`you learned earlier, the process of stimulated emission amplifies the peak wavelength
`more than other wavelengths, narrowing the range of output wavelengths from the broad
`spectrum seen in an LED to the narrower range of a diode laser, shown in Figure 9.1.
`Other layers may have slightly different composition, and the whole structure is
`deposited on a substrate of either GaAs or InP, which are much easier to make in the
`large volumes needed for substrates.
`
`Laser Spectral Range
`The spectral range of diode lasers depends on their structure as well as their composition. The
`simple edge-emitting Fabry-Perot laser described earlier, with one mirror at each end, has a
`bandwidth of 1 to 3 nm concentrated on multiple narrow lines, as shown in Figure 9.12.
`These multiple lines arise from the nature of the resonant cavity.
`
`Fabry-Perot lasers
`have spectral
`widths of 1 to
`3 nm.
`
`

`

`Chapter 9
`
`FIGURE 9.12
`Wavelengths in
`multiple
`longitudinal
`modes.
`
`For light to resonate within the laser cavity, the round-trip distance between the mirrors
`must equal an integral number of wavelengths. Thus, a laser cavity with length L in a laser
`material with refractive index n can resonate at wavelengths X defined by
`
`2 nL = NX
`
`where TV is an integer and where the laser material has large enough gain. Each wavelength
`spike in Figure 9.12 corresponds to a different value of N. The spikes span the range of
`wavelengths where the gain is highest.
`Each spike in Figure 9.12 is a separate longitudinal mode of the laser, which means a
`resonance along the length of the laser cavity. (The modes across the width of a laser or
`an optical fiber are transverse modes, defined by the width of the laser or fiber; narrow-
`stripe diode lasers operate in a single transverse mode.) Each of these longitudinal
`modes has much narrower spectral width than the entire envelope of modes emitted by
`the laser. The spacing between longitudinal modes depends on the cavity length and
`wavelength. The longer the cavity length (measured in wavelengths), the closer the
`modes are spaced. Edge-emitting Fabry-Perot diode lasers have short cavities, only
`about 500 pm long, and their modes are about 0.6 nm apart at 1300 nm or about 0.7 nm
`apart at 1550 nm.
`Minor fluctuations during operation can make edge-emitting Fabry-Perot lasers
`“hop” between modes, shifting the emission wavelength suddenly. The emission peak in
`Figure 9.12 moves from one longitudinal mode to another. This and other problems
`typically limit Fabry-Perot edge-emitting lasers to transmission rates less than 1 Gbit/s
`and to coarse versions of wavelength-division multiplexing with widely separated optical
`channels.
`The same principles apply to VCSELs, but their cavities are much shorter than those of
`edge-emitters, so their longitudinal modes are tens of nanometers apart instead of a frac-
`tion of a nanometer. This means that in practice VCSELs emit a single longitudinal mode
`and can transmit at much higher speeds than edge-emitting Fabry-Perot lasers.
`
`

`

`Single-Frequency Lasers
`For high performance, low dispersion, and closer spacing of optical channels, laser emis-
`sion must be limited to a single longitudinal mode or, equivalently, to a single frequency.
`This has led to development of more elaborate laser resonators. Figure 9.13 shows three
`leading approaches.
`The distributed-feedback (DFB) laser, in Figure 9.13(a), has a series of corrugated ridges
`on the semiconductor substrate, which scatter light back into the active layer. This provides
`feedback like the cavity mirrors on a Fabry-Perot laser, although the details of the physics
`are different. The distributed Bragg reflection (DBR) laser shown in Figure 9.13(b) works in
`much the same way, but the grating is etched in a region outside the zone that is pumped
`by electric current. In both cases, the grating ridges are spaced evenly so they scatter only
`a narrow range of wavelengths back into the active layer of the laser. The active layer of the
`laser amplifies only this selected range of wavelengths, producing very narrow spectral
`bandwidths at a nominal “single frequency.” The wavelength depends on the line spacing
`in the grating and the refractive index of the semiconductor. Recall that Bragg reflection is
`the same effect that selects the wavelengths reflected by a fiber Bragg grating.
`
`Light Sources
`
`Single-frequency
`lasers are needed
`for high-speed
`transmission.
`
`Distributed-
`feedback and
`distributed Bragg
`reflection lasers
`emit only a single
`frequency.
`
`Grating limits emission
`to one frequency.
`
`No current here
`
`Drive current only
`through this region
`
`Facet coated to
`tilts back and forth prevenf reflection
`to change
`wavelength
`
`(size e x a g g e ra te d )
`
`S e le cte d \
`
`O u tp u t Beam
`
`' O u tp u t F ace t
`
`O th e r W a ve le n g th s
`
`— ■
`
`Len gth o f L a se r C a v ity ,
`
`c. External Cavity Tunable Laser.
`
`FIGURE 9.13
`Three single-frequency lasers.
`
`

`

`Chapter 9
`
`Tunable lasers can
`simplify logistics.
`
`Changing the
`length of a VCSEL
`cavity can tune its
`wavelength.
`
`A different way to stabilize laser wavelength is by placing an edge-emitting semiconduc-
`tor laser within an external cavity, which selects the emission wavelength. This requires
`coating one or both facets to suppress reflection back into the semiconductor, and
`adding one or two external mirrors to extend the resonator cavity beyond the laser chip.
`A wavelength-selective element also is added to the laser cavity. In the simple design of
`Figure 9.13(c), the tuning element is a diffraction grating, which serves as one external
`mirror, reflecting light at an angle that depends on its wavelength. (You can get the same
`effect by inserting a prism or some other wavelength-selective component into the laser
`cavity, but diffraction gratings are easier to use.) The laser chip emits a range of wave-
`lengths, but when they strike the grating, most wavelengths are reflected at angles that take
`them away from the laser chip. Only a very narrow range of wavelengths are at the right
`angle to be reflected back into the laser chip for further amplification. This limits output
`to a single frequency.
`Distributed-feedback and distributed Bragg reflection lasers are the types most often
`used to generate a single, fixed wavelength with very narrow spectral width. However, there
`is growing interest in lasers with output that can be tuned to emit at precise wavelengths.
`
`Tunable Lasers
`An external cavity laser is a good starting point for discussing tunable lasers, which can be
`changed in wavelength. You have already seen how a diffraction grating can reflect a single
`wavelength back into the laser chip for amplification, building up emission at a single wave-
`length. Turning the grating changes what wavelengths are reflected back to the laser chip,
`which changes the lasers output wavelength.
`As you will learn later, wavelength tunability is an attractive property for lasers used in
`WDM systems and measurement instruments. Standard lasers emit only a fixed wave-
`length, so a system with 80 different wavelengths requires 80 different models of laser.
`Moreover, the service department needs spares for every one of those 80 different laser
`models. If a telephone company wants to install 80-channel systems, its maintenance
`department would need to stock every site with spares for each of the 80 wavelengths. The
`logistics could become a nightmare.
`Tunability also can enhance the flexibility of optical networking. For example, it some-
`times may be necessary to move an optical channel from one wavelength to another,
`because the same wavelength isn’t available along its entire route. Using a tunable laser to
`generate the new signal would allow the wavelength to be changed without switching
`lasers. The laser might be tunable continuously across the spectrum, but system design
`would be easier if lasers were preset to emit precisely at standard wavelengths. Users would
`then select an optical channel, just as viewers select a channel on a modern television set,
`without having to adjust the laser to match the desired frequency. The technology is still
`young, but several approaches have been developed.
`We saw earlier that the resonant wavelength depends on the length of the laser cavity.
`Changing the cavity length has little effect on edge-emitting lasers, because their longitu-
`dinal modes are less than a nanometer apart. However, it can tune the wavelength signifi-
`cantly because a VCSEL cavity can be made very short, micrometers long rather than
`hundreds of micrometers long for edge-emitters. With a short enough cavity, only a single
`
`

`

`Light Sources
`
`FIGURE 9.14
`Tunable VCSEL
`relies on a
`moving external
`micromirror to
`change cavity
`length and thus
`wavelength.
`
`Output Beam
`
`External partly transparent
`micromirror moves vertically,
`changing cavity length and
`wavelength.
`
`Mirror Support
`
`Active
`Layer
`
`Short Resonant
`Cavity
`
`longitudinal mode falls within the laser’s gain band. This makes possible the sort of tun-
`able VCSEL shown in Figure 9.14. A thin, partly transparent mirror is held above the
`VCSEL by a movable micro-electromechanical system (MEMS) device. Vertical motion of
`the MEMS mirror changes the resonant wavelength in the VCSEL cavity, tuning the wave-
`length by over 30 nm in the laboratory.
`Distributed-feedback and distributed Bragg reflection lasers can be tuned in other ways.
`As you saw earlier, the wavelengths selected depend both on the grating period and on the
`refractive index of the material. Changing the temperature of the material can change both,
`by causing thermal expansion (or contraction) of the laser material as well as by directly
`affecting refractive index. Passing a current through a material also affects the refractive
`index. Generally these changes are relatively small, and allow turning over only several
`nanometers.
`Tuning ranges can be extended to tens of nanometers by using more elaborate distrib-
`uted Bragg reflectors. One example is the sampled-grating distributed Bragg reflector (SG-DBR),
`which contains regions with different grating spaces that reflect a comb-like series of regu-
`larly spaced wavelengths. To make a tunable laser, slightly different separate sampled-grating
`reflectors are fabricated on each end of the active region of the laser. The laser can oscillate
`only at a wavelength reflected by the gratings on both ends, which can be tuned inde-
`pendently by changing current level or temperature. These changes shift the reflection
`peaks of the individual grating only slightly, but this small shift causes a much larger shift
`
`

`

`Chapter 9
`
`FIGURE 9.15
`Tuning ofsam pled-
`grating distributed
`Bragg reflector
`laser.
`
`This peak reflects.
`I
`
`R e fle ctio n Peaks
`fro m Each G ratin g
`
`Small shift in the peak
`reflection wavelengths of
`the grating at one end of
`the laser cavity causes a
`large shift in the wavelength
`where both ends reflect to
`produce laser oscillation.
`
`This peak reflects.
`* ------------
`
`Large Total W ave le n g th S hift
`
`in the wavelength at which both gratings reflect to allow laser oscillation, as shown in
`Figure 9.15. This is sometimes called a vernier effect, because a vernier scale works in the
`same way to amplify the size of a small change. A related approach is the grating-assisted
`coupler and sampled reflector (GCSR) laser, which combines a sampled-grating reflector
`with other elements to tune the output wavelength.
`Another approach to tuning is selecting one laser stripe from an array of several on a sin-
`gle semiconductor substrate. For example, if each of 12 stripes had a tuning range of 3 nm,
`and their center wavelengths were 3 nm apart, they could combine to cover a 36-nm range.
`The device could tune across one laser’s 3-nm range, then switch to the next laser and tune
`over its range.
`Other tunable lasers also are in development. They face a number of practical challenges.
`Tunable lasers must be locked to the right wavelengths so they don’t drift during operation.
`They also must be affordable, reliable, and compatible with standard telecommunications
`equipment. The spread of tunable lasers was stalled by the telecommunications downturn.
`Modulation and Wavelength Chirp
`Direct modulation via changing the drive current is the simplest and cheapest wa