1. Introduction

Distributed feedback (DFB) semiconductor lasers are of great interest and have been extensively utilized for their high-bit-rate transmission performance in long-haul optical fiber communication systems [1–3]. However, there are still some unsolved problems and improvements needed in their properties.

A DFB laser with uniform index-coupled gratings, which is normally realized using interference lithography (IL), intrinsically has two degenerate lasing modes [4]. At the Bragg wavelength, the optical mode is anti-resonant. There are two resonant modes that are symmetrically positioned on both sides of the Bragg wavelength. But due to the arbitrary phase at the cleaved facets, the effect of facet coating, non-uniform gain distribution, etc., the degeneracy may sometimes be destroyed, and the laser is able to lase in a single longitudinal mode. However, this uncertainty can result in problems such as low yield, an uncontrolled lasing wavelength, multi-mode lasing, mode-hopping, etc. One solution to the degenerate modes problem is to insert a quarter-wave phase shift [5] into the gratings at the center of the laser cavity by increasing the grating pitch by a half period. The resulting optical mode has a strong resonance that is pinned at the resonant wavelength, which is strongly selective and therefore stable. Although there are some approaches that can potentially realize this phase shift [6], a precise and controllable quarter-wave phase-shift is still not achievable by any means other than electron-beam lithography (EBL). Compared to the cost-ineffective and low-throughput EBL, the widely used optical IL approach is scalable to larger wafer areas, but it is only capable of providing uniform gratings. To resolve this problem, a specially designed sampled Bragg grating (SBG) method has been proposed [3,7], by which an equivalent phase shift (EPS) can be realized in uniform IL-produced gratings.

Additionally, DFB lasers are usually fabricated using either a ridge-waveguide (RWG) structure [1–3], or a buried heterostructure (BH) [8], both of which require at least two to three epitaxial growths. The regrowth process is expensive and complicated, placing critical demands on wafer cleanliness and surface morphology. To avoid epitaxial regrowth and the potential problems that it may bring, a laterally-coupled DFB (LC-DFB) laser has been proposed [9]. However, due to the non-planar topography on which the gratings are fabricated, the LC-DFB lasers are usually realized by using EBL. Recently, a LC-DFB laser with first-order gratings achieved by IL has been reported [10], in which the gratings are first patterned on a planar surface and then transferred onto the sidewalls during the ridge etching process. Good single-mode lasing performance and strong mode selection were achieved, with a SMSR of 37 dB.

In this paper we report the first laterally-coupled DFB laser with an equivalent quarter-wave phase shift achieved by means of interference lithography. The laser contains a specially designed sampled grating fabricated on each sidewall of a ridge-waveguide structure, which effect an equivalent quarter-wave phase-shift at the center of the laser cavity, pinning the lasing mode at a new, strongly resonant wavelength. This is achieved without requiring EBL or regrowth, using a scalable, wafer-scale process. LC-DFB lasers with good lasing performance and excellent wavelength stability are obtained. It is also demonstrated that the lasing wavelength can be tuned over a wide range by simply adjusting the design parameters of the sampled gratings.

2. Principle

The EPS method was proposed and demonstrated previously in Bragg gratings [11], fiber lasers [12], silicon waveguide [13], conventional RWG lasers [3]. The basic principle can be described by the following two mechanisms.

Firstly, by sampling the gratings with a period of length P, an additional wave-vector is provided to enable phase-matching to a new resonance wavelength, λ_p, which is shifted away from the Bragg wavelength. The phase-matching condition for λ_p can be expressed as:

Secondly, a quarter-wave EPS, θ_p ( = π), can be obtained at the new resonance wavelength by increasing the sampling period P at the center of the laser cavity by an amount ΔP = P/2:

The base grating is a uniform grating, which can be patterned by IL. Typically, P is of the order of micrometers, which can be defined using conventional photolithography. By applying the EPS to the LC-DFB lasers, the laser devices can be fabricated using an e-beam-free and regrowth-free method, which offers a batch processing capability that meets mass-production requirements.

3. Design and fabrication

Based on the principle described above, the sampled grating structure of the LC-DFB laser with EPS can be designed. The grating pitch, Λ, is set to be 254.8 nm. The interference lithography tool is set up in a Lloyd’s mirror configuration, using a 325 nm He-Cd UV laser. The sampling pitch, P, is designed to be 6.5 µm, and an EPS region of 3.25 µm is inserted at the center of the cavity (by extending one period to 9.75 µm), leading to a phase-matched resonant wavelength at λ_p = 1,571 nm. The Bragg wavelength will be at ~1,630 nm, which is sufficiently far away from the gain peak to eliminate the possibility of lasing at the Bragg wavelength. Here the “+” sign in Eq. (1) has been chosen mainly for a larger grating pitch, which is relatively easier to fabricate. The fabrication of the lateral gratings for the proposed laser has an etching aspect ratio of 1:16. If the “–” sign in Eq. (1) is chosen, the grating pitch will be smaller (236.3 nm), and in order to etch to the same depth, an even higher etching aspect ratio is needed. In this report, the EPS section with an additional length of 3.25 µm at the center of the cavity is designed to contain gratings, although the presence or absence of gratings in the EPS region will not affect the equivalent phase shift.

The wafer, which contains an AlGaInAs/InP epistructure designed for ridge-waveguide (RWG) lasers, was grown by Metal-Organic Vapor Phase Epitaxy (MOVPE) in a single growth run on a (100) n-InP substrate. The gain peak occurs at 1,560 nm. The epilayer structure contains an n-InP buffer layer, an n-InP cladding layer, an active region, a p-InP cladding layer and a contact layer. The active region consists of a Graded-Index (GRIN) separate confinement heterostructure (SCH), in which five AlGaInAs QWs are bounded between p-doped and n-doped GRIN layers, each of which contains an InAlAs carrier confinement layer.

In the fabrication process, a SiO₂ mask layer is first deposited on the surface of the epiwafer by plasma-enhanced chemical vapor deposition (PECVD). Then a 20 nm thick Cr mask layer is evaporated onto the SiO₂ surface. The sampling pattern containing the specially designed EPS region is first transferred from the photomask to the Cr layer by conventional photolithography followed by a Cl₂/O₂ Inductively-Coupled Plasma (ICP) etch. Regions in the SiO₂ layer covered by Cr define the non-grating regions, while the exposed SiO₂ will define the patterned grating regions. Optical interference lithography is used to pattern the gratings on top of the entire planar SiO₂/Cr surface. Since the Cr layer is much thinner than the photoresist layer (160 nm), the surface flatness variation (20 nm) has a negligible effect on the IL grating patterning. The grating quality in the photoresist is equally good with or without the Cr sampling layer. The IL gratings are then transferred to the SiO₂ mask by reactive-ion etching (RIE). The SiO₂ that is covered by Cr is totally protected, giving rise to the non-grating regions. Another photolithography step is then performed to pattern the 2.2 µm wide ridge waveguides. Cl₂/O₂ ICP and RIE etching processes are used to remove the Cr and the SiO₂ masks from regions where the ridge waveguide trenches will be etched. Before etching the ridge trenches, an oxygen RIE process [10] is used to remove a 200 nm wide region of photoresist from each edge of the ridge photoresist mask, thereby exposing a 200 nm wide region of sampled SiO₂ gratings along each sidewall of the ridge. The remaining photoresist now acts as a mask for etching the ridge in regions without gratings (W_r = 1.8 µm), while the narrow but exposed 200 nm wide sampled SiO₂ gratings provide a mask for etching the deep lateral sampled grating (W_g = 200 nm) along both sidewalls as the trenches are etched. Cl₂/CH₄/H₂ ICP etching is used for anisotropic etching of the III-V semiconductor materials. The grating aspect ratio is controlled to 0.5 for a maximized coupling efficiency.

The ridge and the gratings are both etched to a depth D = 2 µm, which is ~150 nm above the quantum well active region. Then the photoresist and Cr masks are removed, using O₂ RIE for the photoresist, and a Buffered Oxide Etch (BOE) wet etch for the SiO₂, leaving the etched ridge and sampled gratings along both sidewalls of the ridge. Figure 1 shows a scanning electron microscope (SEM) image of the ridge at the center of the laser cavity, where the quarter-wave EPS exists. It can be seen that the non-grating ridge sidewalls are very smooth, without any vestiges of gratings. It is due to the ultra-high etching selection ratio of SiO₂ to Cr. The sampling pattern is still visible on the bottom of the etched ridge trench. The pattern was partially transferred into the epiwafer contact layer during etching and removal of the SiO₂ sampling mask, and propagates to the bottom of the ridge trench during ICP etching of the trench regions.

 

Fig. 1 SEM image of the LC-DFB laser with the EPS region at the center of the laser cavity after etching the ridge trench.

Download Full Size | PDF

Following definition of the ridge waveguides and the sampled gratings, another SiO₂ layer is deposited by PECVD for electrical isolation, followed by opening of the contact window on top of each ridge. Ti-Pt-Au is used as the p-contact metal, and Ti-Au is used as the n-contact metal, deposited on the top and back side of the device, respectively. The contacts are annealed by a rapid thermal annealing (RTA) process at 450 °C. After the wafer is cleaved into 900 µm long laser bars, a single-layer anti-reflection optical coating is deposited onto laser’s output facet to suppress lasing in the Fabry-Perot (FP) modes.

The coupling coefficient κ of such a device is strongly related to W_g, D, and W_r. κ is very small when the etch is stopped at over 300 nm above the QWs, but increases dramatically as etching approaches the QWs. The value of κ drops rapidly as W_r becomes wider [14], and goes up as W_g is increased, but the rate of increase slows significantly when the grating width exceeds 500 nm. The estimated value of κL for this device is 1.03.

The SEM images in Figs. 2(a) and 2(b) show the side-view and top-view of the LC-DFB laser with the EPS section after metallization.

 

Fig. 2 SEM image showing the side view (a) and top view (b) of a LC-DFB laser with an EPS region after metallization.

Download Full Size | PDF

4. Performance

Continuous-wave (CW) lasing is achieved over a wide range of temperatures and injection currents. At room temperature (~25 °C), the threshold current is 27 mA, corresponding to a threshold current density of 1.5 kA/cm². The output power is 14 mW at a 100 mA injection current, and is 23.5 mW at a 200 mA injection current.

Figure 3 shows the lasing spectrum at 100 mA bias current under room temperature, CW operation. Single longitudinal mode lasing is obtained at 1,571.2 nm, with a side mode suppression ratio (SMSR) of 41.1 dB. The suppressed Bragg wavelength at ~1,630 nm can be seen clearly in an extended spectrum.

 

Fig. 3 Lasing spectrum of the LC-DFB laser with a quarter-wave EPS region under an injection current of 100 mA. A SMSR of 41.1 dB is obtained.

Download Full Size | PDF

For lasers with an EPS, the possibility of lasing at the Bragg wavelength needs to be examined. If the Bragg wavelength is too close to the gain peak, as the injection current increases, and the gain spectrum shifts towards longer wavelengths, lasing at the Bragg wavelength becomes competitive [3]. As for the device design reported here, the Bragg wavelength does not lase at any temperature and injection current. The device exhibits a SMSR above 35 dB at up to 240 mA injection current.

The lasing mode shifts towards longer wavelengths with increasing bias current at the rate of ~10 pm/mA, under temperature control by a Thermoelectric Cooler (TEC). As the injection current is increased, even with TEC, the junction temperature still rises, causing the refractive index to increase [15]. However, for phase-shifted lasers the optical field intensity peaks at the phase shift region, where the higher carrier density under high injection current causes the effective refractive index to decrease [16], which partially reduces the temperature-induced red-shift, although the overall shift is still towards a longer wavelength.

The light-current-voltage (LIV) characteristics of the LC-DFB laser under different ambient temperatures has been measured and are shown in Fig. 4.

 

Fig. 4 LIV characteristics of the LC-DFB laser with EPS at different ambient temperatures.

Download Full Size | PDF

To demonstrate the effectiveness of the equivalent phase shift method in pinning the lasing wavelength and its ability to adjust the wavelength over a wide range, another LC-DFB laser with a different sampling period has been designed and tested. The sampling period P is designed to be 4.0 µm with an EPS region of 2.0 µm in the center of the cavity. The measured lasing wavelength is at 1,535.6 nm, which is very close to the designed wavelength of ~1,536 nm, and represents a wavelength blue-shift of 35 nm from the previous DFB laser. The lasing spectrum at 40 mA is shown in Fig. 5.

 

Fig. 5 Lasing spectrum of another LC-DFB laser with EPS design for 1536 nm operation.

Download Full Size | PDF

6. Conclusion

In this paper we report the first LC-DFB lasers with a quarter-wave equivalent phase shift. The lasers operate in a single longitudinal mode over a very wide range of injection currents. A SMSR of 41.1 dB is achieved under a 100 mA DC bias. At room temperature (~25 °C), and under 100/200 mA bias current, a 14.0/23.5 mW optical output power is obtained. This method provides a scalable and low-cost approach to fabricating quarter-wave phase-shifted DFB lasers using e-beam-free and regrowth-free processes, which may be useful for the fabrication of DFB lasers with stable and precise wavelengths, as single devices or as arrays in photonic integrated circuits both in academic research and for industrial production.