Single-Mode Semiconductor Lasers
Semiconductor lasers oscillate in several longitudinal modes simultaneously because of a relatively small gain difference (~0.1 cm-1) between two neighboring modes of the cavity. The resulting spectral width (2-4 nm) is acceptable for some applications but becomes a concern for many others. This tutorial is devoted to techniques that can be used to design semiconductor lasers such that they emit light predominantly in a single longitudinal mode.
The basic idea is to design the laser such that losses are different for different longitudinal modes of the cavity, in contrast with FP lasers whose losses are mode-independent. The following figure shows the gain and loss profiles schematically for such a laser.
The longitudinal mode with the smallest cavity loss reaches threshold first and becomes the dominant mode. Other neighboring modes are discriminated by their higher losses. The power carried by these side modes is usually a small fraction (< 1%) of the total emitted power. The performance of a single-mode laser is often characterized by the mode-suppression ratio (MSR), defined as MSR = Pmm/Psm, where Pmm is the main-mode power and Psm is the power of the most dominant side mode. The MSR should exceed 1000 (or 30 dB) for a good single-mode laser.
1. Distributed Feedback Lasers
Distributed feedback (DFB) semiconductor lasers were developed during the 1980s and are used routinely for WDM lightwave systems. The feedback in DFB lasers, as the name implies, is not localized at the facets but is distributed throughout the cavity length. This is achieved through an internal build-in grating that leads to a periodic variation of the mode index. Feedback occurs by means of Bragg diffraction, a phenomenon that couples the waves propagating in the forward and backward directions. Mode selectivity of the DFB mechanism results from the Bragg condition: The coupling occurs only for wavelengths λB satisfying
where Λ is the grating period, 
is the average mode index, and the integer m represents the order of Bragg diffraction. The coupling between the forward and backward waves is strongest for the first-order Bragg diffraction (m = 1). For a DFB laser operating at λB = 1.55 μm, Λ is about 235 nm if we use m = 1 and = 3.3. Such gratings can be made by using a holographic technique.
From the standpoint of device operation, semiconductor lasers employing the DFB mechanism can be classified into two broad categories: DFB laser and distributed Bragg reflector (DBR) lasers. The following figure shows two kinds of laser structures.
Though the feedback occurs throughout the cavity length in DFB lasers, it does not take place inside the active region of a DBR laser. In effect, the end regions of a DBR laser act as mirrors whose reflectivity is maximum for a wavelength λB satisfying the above equation. The cavity losses are therefore minimum for the longitudinal mode closest to λB and increase substantially for other longitudinal modes. The MSR is determined by the gain margin defined as the excess gain required by the most dominant side mode to reach threshold. A gain margin of 3-5 cm-1 is generally enough to realize an MSR > 30 dB for DFB lasers operating continuously. However, a larger gain margin is needed (> 10 cm-1) when DFB lasers are modulated directly. Phase-Shifted DFB lasers, in which the grating is shifted by λB/4 in the middle of the laser to produce a π/2 phase shift, are often used, since they are capable of providing a much larger gain margin than that of conventional DFB lasers. Another design that has led to improvements in the device performance is known as the gain-coupled DFB laser. In these lasers, both the optical gain and the mode index vary periodically along the cavity length.
Fabrication of DFB semiconductor lasers requires advanced technology with multiple epitaxial growths. The principle difference from FP lasers is that a grating is etched onto one of the cladding layers surrounding the active layer. A thin n-type waveguide layer with a refractive index intermediate to that of the active layer and the substrate acts as a grating. The periodic variation of the thickness of the waveguide layer translates to a periodic variation of the mode index along the cavity length and leads to a coupling between the forward and backward propagating waves through Bragg diffraction.
A holographic technique is often used to form a grating with a pitch of ~0.2 μm. It works by forming a fringe pattern on a photoresist deposited on the wafer surface by interfering two optical beams and then etching the pattern chemically. In the alternative electron-beam lithographic technique, an electron beam writes the desired pattern on the electron-beam resist. Both methods use chemical etching to form grating corrugations, with the patterned resist acting as a mask. Once the grating has been etched onto the substrate, multiple layers are grown by using an epitaxial growth technique. A second epitaxial regrowth is needed to make a BH device. Despite the technological complexities, DFB lasers are routinely produced commercially. They are used in nearly all 1.55-μm optical communication systems operating at bit rates of 2.5 Gb/s or more. DFB lasers are reliable enough that they have been used since 1992 in all transoceanic lightwave systems.
2. Coupled-Cavity Semiconductor Lasers
In a coupled-cavity semiconductor laser, single-mode operation is realized by coupling the laser cavity to an external cavity, which feeds a portion of the exiting light back into the laser cavity. The feedback from the external cavity is not necessarily in phase with the field inside the laser cavity because of the phase shift occurring in the external cavity. The in-phase feedback occurs only for those laser modes whose wavelength nearly coincides with one of the longitudinal modes of the external cavity. In effect, the effective reflectivity os the laser facet facing the external cavity becomes wavelength-dependent and leads to low losses for certain wavelengths. The longitudinal mode that is closest to the gain peak and has the lowest cavity loss becomes the dominant mode.
Several kinds of coupled-cavity schemes have been developed for making single-mode laser; the following figure shows three of them.
A simple scheme couples the light from a semiconductor laser to an external grating (figure (a)). It is necessary to reduce the natural reflectivity of the cleaved facet facing the grating through an antireflection coating to provide a strong coupling. Such lasers are called external-cavity semiconductor lasers and have attracted considerable attention because of their tunability. The wavelength of the single mode selected by the coupled-cavity mechanism can be tuned over a wide range (typically 50 nm) simply by rotating the grating. Wavelength tunability is a desirable feature for lasers used in WDM lightwave systems. A drawback of the laser shown here fro the system standpoint is its nonmonolithic nature, which makes it difficult to realize the mechanical stability required of optical transmitters.
A monolithic design for coupled-cavity lasers is offered by the cleaved-coupled-cavity laser shown in figure (b) above. Such lasers are made by cleaving a conventional multimode semiconductor laser in the middle so that the laser is divided into two sections of about the same length but separated by a narrow air gap (with a width of ~1 μm). The reflectivity of cleaved facets (~30%) allows enough coupling between the two sections as long as the gap is not too wide. It is even possible to tune the wavelength of such a laser over a tuning range ~20 nm by varying the current injected into one of the cavity sections acting as a mode controller. However, tuning is not continuous, since it corresponds to successive mode hops of about 2 nm.
3. Tunable Semiconductor Lasers
Modern WDM lightwave systems require single-mode, narrow-linewidth lasers whose wavelength remains fixed over time. DFB lasers satisfy this requirement but their wavelength stability comes at the expense of tunability. The large number of DFB lasers used inside a WDM transmitter make the design and maintenance of such a lightwave system expensive and impractical. The availability of semiconductor lasers whose wavelength can be tuned over a wide range solves this problem.
Multisection DFB and DBR lasers were developed during the 1990s to meet the somewhat conflicting requirements of stability and tunability. Figure (c) above shows a typical laser structure. It consists of three sections, referred to as the active section, the phase-control section, and the Bragg section. Each section can be biased independently by injecting different amounts of currents. The current injected into the Bragg section is used to change the Bragg wavelength (λB = 2nΛ) through carrier-induced changes in the refractive index n. The current injected into the phase-control section is used to change the phase of the feedback from the DBR through carrier-induced index changes in that section. The laser wavelength can be tuned almost continuously over the range 10-15 nm by controlling the currents in the phase and Bragg sections. By 1997, such lasers exhibited a tuning range of 17 nm and output powers of up to 100 mW with high reliability.
Several other designs of tunable DFB lasers have been developed in recent years. In one scheme, the built-in grating inside a DBR laser is chirped by varying the grating period Λ or the mode index along the cavity length. The Bragg wavelength itself then changes along the cavity length. Since the laser wavelength is determined by the Bragg condition, such a laser can be tuned over a wavelength range determined by the grating chirp. In a simple implementation of the basic idea, the grating period remains uniform, but the waveguide is bent to change the effective mode index . Such multisection DFB lasers can be tuned over 5-6 nm while maintaining a single longitudinal mode with high side-mode suppression.
In another scheme, a superstructure or sampled grating is used for the DBR section of a multisection laser. In such gratings, the amplitude or the phase of the coupling coefficient is modulated in a periodic fashion along the grating length. As a result, the reflectivity peaks at several wavelengths whose interval is determined by the modulation period. Such multisection DBR lasers can be tuned discretely over a wavelength range exceeding 100 nm. By controlling the current in the phase-control section, a quasi-continuous tuning range of 40 nm was realized in 1995 with a superstructure grating. The tuning range can be extended considerably by using a four-section device in which another DBR section is added to the left side of the device shown in (c) figure above. Each DBR section supports its own comb of wavelengths but the spacing in each comb is not the same. The coinciding wavelength in the two combs becomes the output wavelength that can be turning over a wide range (analogous to the Vernier effect).
In a different design, referred to as a tunable twin-guide laser, a tuning layer is added vertically within a standard DFB structure and two different sampled gratings are employed for tuning, as shown schematically in the figure below.
Such a device is much simpler to fabricate and operate than the three- and four-section DBR designs. The resulting laser can be tuned over a 40-nm wavelength range while maintaining a relatively large output power (~10 mW) and a high side-mode suppression ratio (>30 dB). Since the active and the tuning layers are separated by a passive n-type InP layer, such a device consists of two vertically stacked p-i-n diodes that can be biased independently. At the same time, the active and tuning layers act as cladding layers for the middle InP layer (with a higher refractive index) that forms an optical waveguide such that the intensity of the optical mode peaks in this middle layer. Because of a good fraction of the optical mode resides in both the active and tuning layers, the two electrically isolated diodes are optically coupled in the vertical direction, thus allowing a wide tuning of the mode wavelength through the Vernier effect implemented with the two sampled gratings.
4. Vertical-Cavity Surface-Emitting Lasers (VCSELs)
A new class of semiconductor lasers, known as vertical-cavity surface-emitting lasers (VCSELs), emerged during the 1990s with many potential applications. VCSELs operate in a single longitudinal mode by virtue of an extremely small cavity (~1 μm), for which the mode spacing exceeds the gain bandwidth. They emit light in a direction normal to the active-layer plane, in a manner analogous to that of a surface-emitting LED. Moreover, the emitted light is in the form of a circular beam that can be coupled into a single-mode fiber with high efficiency. These properties result in a number fo advantages that are leading to rapid adoption of VCSELs for lightwave communications.
As seen in the figure above, fabrication of VCSELs requires growth of multiple thin layers on a substrate. The active region, in the form of one or several quantum wells, is surrounded by two high-reflectivity (>99.5%) DBR mirrors that are grown epitaxially on both sides of the active region to form a high-Q microcavity. Each DBR mirror is made by growing many pairs of alternating GaAs and AlAs layers, each λ/4 thick, where λ is the wavelength emitted by the VCSEL. A wafer-bonding technique is sometimes used for VCSELs operating in the 1.55-μm wavelength region to accommodate the InGaAsP active region. Chemical etching or a related technique is used to form individual circular disks (each corresponding to one VCSEL) whose diameter can be varied over a wide range (typically 5-20 μm). The entire two-dimensional array of VCSELs can be tested without requiring separation of lasers because of the vertical nature of light emission. As a result, the cost of a VCSEL can be much lower than that of an edge-emitting laser. VCSELs also exhibit a relatively low threshold (~1 mW or less). Their only disadvantage is that they cannot emit more than a few milliwatts of power because of a small active volume. For this reason, they are mostly used for local-area and data-communication applications and have virtually replaced LEDs. Early VCSELs were designed to emit near 0.8 μm and operated in multiple transverse modes because of their relatively large diameters (~10 μm).
In recent years, VCSEL technology has advanced enough that VCSELs can be designed to operate in a wide wavelength range extending from 650 to 1,600 nm. Their applications in the 1.3- and 1.55-μm wavelength windows require that VCSELs operate in a single transverse mode. By 2001, several techniques had emerged for controlling the transverse modes of a VCSEL, the most common being the oxide-confinement technique in which an insulating aluminum-oxide layer, acting as a dielectric aperture, confines both the current and the optical mode to a <3-μm-diameter region. Such VCSELs operate in a single mode with narrow linewidth and can replace a DFB laser in many lightwave applications as long as their low output power is acceptable. They are especially useful for data transfer and local-loop applications because of their low-cost packaging. VCSELs are also well suited for WDM applications for two reasons. First, one can make two-dimensional VCSEL arrays such that each laser operates at a different wavelength. Second, VCSEL wavelengths can be tuned over a wide range (>50 nm) using the micro-electro-mechanical system (MEMS) technology.