1. Introduction

Tunable lasers have been researched extensively in the last few decades. Due to the flexibility of wavelength tunability, they have been widely applied in many fields, such as optical fiber communication, optical sensing and optical measurement [1]. Having advantages of manufacturability, compactness and reliability, monolithic widely tunable semiconductor lasers are more preferable in optical fiber communication systems [2]. Commercially available monolithic tunable semiconductor lasers are usually based on gratings to realize mode selection, such as the distributed feedback laser array and the distributed Bragg reflector type tunable lasers [3–6]. However, fabrication of complex gratings needs high-resolution lithography technology like electron beam lithography (EBL). Moreover, buried gratings also increase the requirement on the quality of regrowth. Lots of efforts have been tried to simplify the fabrication of monolithic tunable semiconductor lasers. Instead of using gratings, many other elements, such as double ring resonators [7,8] and coupled cavities [9,10], are utilized to achieve mode selection. However, performance of these new kinds of tunable semiconductor lasers is not comparable to that of grating-based tunable semiconductor lasers. The ring-resonator filters exhibit superior filtering characteristics and have a compact structure. A double-ring-resonator tunable laser realized a tuning range of 50 nm, a maximum output power of 3 dBm, threshold currents around 40 mA and SMSRs larger than 30 dB [7]. Also, a parallel-ring-resonator tunable laser was reported to have stable output powers with power variation of less than 1 dB across a 35 nm tuning range [8]. To avoid regrowth, an all-active V-cavity tunable laser was proposed and demonstrated [9]. The V-cavity laser is a low-cost tunable laser scheme, but performance still can’t compare with the grating-based tunable semiconductor lasers. The tuning range of the V-cavity laser can reach 40 nm with 50 GHz frequency spacing by changing the working temperature, SMSRs are below 40 dB and fiber output powers are less than 6.6 dBm [10]. Recently, a tunable Quantum dot laser based on two all-active ring resonators was demonstrated [11]. This laser was grown on Si and regrowth-free, but it only realized a tuning range of 16 nm with output powers of over 2.7 mW.

The MCI laser proposed and demonstrated recently are promising devices, which realize a quasi-continuous tuning range of more than 50 nm and good single mode performance with SMSRs higher than 40 dB (typically higher than 48 dB) [12,13]. The MCI laser is based on eight coupled FP cavities, so it can be easily fabricated by standard photolithography. In our first run, all the waveguides of the MCI laser were deeply etched with a single ICP dry etching so as to reduce the complexity of fabrication. Because the active layers of the gain section were etched through, the performance of the MCI laser were deteriorated. The threshold currents were more than 30 mA and the total output powers were less than 5 mW with a current of 100 mA injected into the gain section [13]. To improve the laser performance, we made the gain and phase sections to be surface ridge waveguides and the other waveguides to be deep ridge waveguides.

Mode stabilization is a critical issue for tunable lasers. Wavelength must keep stable in a long lifetime without mode hopping. However, any change in the laser and the ambient condition may cause wavelength shift or even mode hopping, such as aging, temperature change. We used an optimization algorithm based characterization scheme to make the MCI laser output any desired wavelengths [14]. Injection current settings can be directly obtained by the characterization scheme, but information on the relationship between the tuning currents and the output characteristics around these settings can’t be acquired in the process of characterization. Therefore, before using the MCI laser in practical applications, it is necessary to figure out the tuning characteristics so as to realize reliable and stable wavelength control in long-term usage. To the best of our knowledge, there are no reports on the tuning characteristics of coupled cavity lasers with more than three cavities. In this paper, we report the tuning characteristics of the MCI laser with improved performance for the first time. This paper is organized as follows: in section 2, the operation principle of the MCI laser is explained; in section 3, fabrication of the MCI laser is described briefly; in section 4, characterization results of the MCI laser packaged into a standard 14-pin butterfly package are given; in section 5, tuning characteristics of the MCI laser are discussed; and finally, a brief conclusion is given in section 6.

2. Operation principle of the MCI laser

Figure 1 shows a microscope image of the fabricated MCI laser without the common phase section. The chip is 1.7 mm long and 0.45 mm wide. The MCI laser is composed of a gain section, a 1 × 8 splitter and eight arms with unequal length difference. The 1 × 8 splitter can be realized by cascaded 1 × 2 multi-mode interferometers (MMIs) or a 1 × 8 MMI. Both the 1 × 2 MMI and 1 × 8 MMI are commonly used in photonic integrated circuits. The fabrication tolerance of the 1 × 2 MMI is better than that of the 1 × 8 MMI, so the 1 × 8 splitter based on cascaded 1 × 2 MMIs is chosen in our first laser design. The front mirror of the MCI laser is a cleaved facet, which outputs light. In order to reflect the light of each arm back into the gain section, a one-port MIR is integrated at the end of each arm, which is an on-chip reflector based on multi-mode interference [15]. Mode selection of the MCI laser is realized by the addition of the reflections from the eight arms. There is an arm phase section on each arm to tune the phase of each arm independently. When the round-trip phase difference of any two arms is integral multiples of 2π at wavelength λ₀, the reflections of the eight arms can achieve constructive interference at wavelength λ₀. Then a narrow strong reflection peak can be generated at wavelength λ₀, as shown in Fig. 2. Nevertheless, the other wavelengths will not be in phase like λ₀, which generate a lot of suppressed reflection peaks. The shape of the whole reflection spectrum is decided by the arm length difference [12]. To ensure the MCI laser having a large tuning range and good single mode performance, the arm length difference should be carefully optimized so that the reflection spectrum has a narrow main reflection peak and the other random reflection peaks are suppressed as much as possible. Increasing the arm number, the main reflection peak can be narrower and the suppression of the other reflection peaks becomes better. However, more arms mean more arm phase sections, which makes the control of the MCI laser more complicated. Besides, the size of the laser also increases due to a larger 1 × N splitter (N>8) and longer arms. On the contrary, reducing the arm number, suppression of the adjacent cavity modes around the main reflection peak and the other reflection peaks of the reflection spectrum can’t be optimized at the same time: If making the main reflection peak to be narrower, the suppression of the other reflection peaks will be worse; If making the suppression of the other reflection peaks to be better, the main reflection peak will be wider and the suppression of the adjacent cavity modes around the main reflection peak will be worse. Therefore, a tradeoff between the arm number and the laser performance has to be made. Through simulation, we chose eight arms. By optimizing the arm length difference, eight arms can provide enough suppression on the adjacent cavity modes around the main reflection peak and the other reflection peaks of the reflection spectrum. Arranging the lengths of the eight arms in ascending order, the length difference is 78.31 µm, 4.86 µm, 8.02 µm, 140.32 µm, 18.24 µm, 27.48 µm and 179.47 µm, respectively.

 

Fig. 1. Microscope image of the fabricated MCI laser without the common phase section. The two inserts are scanning electron microscope images of the fabricated shallow-deep transitions.

Download Full Size | PDF

 

Fig. 2. Calculated reflection spectrum from the right side of the gain section with a main reflection peak at wavelength 1570 nm.

Download Full Size | PDF

In Ref. [13], a common phase section was also fabricated between the gain section and the 1 × 8 splitter to finely tune the cavity modes. For the MCI laser with the common phase section, wavelength tuning can be realized in two ways: One way is adjusting the phases of any seven arm phase sections to tune the reflection peak (coarse tuning) and the phase of the common phase section to align a cavity mode with the reflection peak (fine tuning); The other way is adjusting the phases of the eight arm phase sections at the same time. In this way, the main reflection peak and the cavity modes are tuned simultaneously [12]. Therefore, the common phase section is not necessary for the MCI laser. Without the common phase section, length of the laser is shorter and mode spacing between two adjacent cavity modes is larger. Free spectral ranges (FSRs) of The MCI lasers with and without the common phase section are about 0.23 nm and 0.27 nm, respectively. In this paper, the MCI laser without the common phase section was packaged and tested, of which wavelength tuning can only be realized by tuning the phases of the eight arm phase sections.

3. Fabrication

The MCI laser is an integrated device including active and passive components. Offset-quantum-well scheme was used for the active-passive integration [13]. InGaAsP/InP multiple-quantum-well (MQW) layers with seven compressively strained InGaAsP quantum wells (${\lambda _g} = 1540\; \textrm{nm}$) were used for the gain section. A 300-nm-thick 1.3Q InGaAsP passive layer was grown below the MQWs layers. MQWs in the passive regions were selectively removed by wet etching, which leaves the 1.3Q InGaAsP layer. A p-InP upper cladding layer, an InGaAs contact layer and an InP capping layer were re-grown by MOCVD over the entire wafer after the wet etching process. Compared with the butt-joint regrowth, active-passive integration based on offset-quantum-well scheme is much easier to realize. In the first run, all the waveguides of the MCI laser were deeply etched with a single dry etch step to simplify the fabrication process. Because the active layers were etched through, threshold currents increased and output powers decreased due to increased surface recombination of carriers and optical losses. In order to improve the performance of the MCI laser, we made the gain and all the phase sections to be surface ridge waveguides and the other passive waveguides to be deep ridge waveguides. Due to the difference of optical confinement, mode mismatch between the surface and deep ridge waveguides can cause a calculated loss of about 0.3 dB in our design. Mode of the guided light has to transform 6 times between the surface and deep ridge waveguides in a single round trip, which results in a high loss. Therefore, to reduce the loss resulted from mode mismatch, we designed and fabricated a tapered shallow-deep transition structure to connect the surface and deep ridge waveguides, as the two SEM images shown in Fig. 1. The mode conversion efficiency of the tapered shallow-deep transition is simulated to be more than 99% in theory and was experimentally demonstrated to have a loss less than 0.1 dB [16]. Fabrication processes of the surface and deep ridge waveguides are as follows: Firstly, all the waveguides were defined by an i-line stepper and were etched into the upper cladding InP layer through an ICP dry etching; Secondly, the surface ridge waveguides were selectively wet etched by HCl/H3PO4 solution and precisely stopped at the InGaAsP MQWs layers. The deep ridge waveguide regions were protected by photoresist in the process of wet etching; Then, we fabricated a SiO2 hard mask to protect the surface ridge waveguide regions and define the tapered shallow-deep transition structures; Finally, the deep ridge waveguides were etched again by an ICP dry etching. The other fabrication processes are introduced in Ref. [13].

4. Butterfly-packaged MCI laser

We proposed an optimization algorithm based characterization method to make the MCI laser output any desired wavelength accurately and fast [11]. An optical bandpass filter was used to get the power of the desired wavelength. Then currents injected into the eight arm phase sections were changed to maximize the output power from the optical bandpass filter. To improve the efficiency of the characterization process, particle swarm optimization (PSO) method was utilized to search for the appropriate current settings making the MCI laser output the designated wavelength. Because power measurement is very fast, the characterization time for a single wavelength can be as short as several seconds with high-speed data acquisition and processing. After the optimization process completes, we use an optical spectrum analyzer (OSA) to measure the laser spectrum. If the lasing wavelength is deviated from the set wavelength or the SMSR is not high, the optimization process will be executed again with random initial current settings. Repeating the optimization process, we can characterize the MCI laser with a number of given wavelengths across the tuning range efficiently.

Figure 3 shows the experimental characterization setups for the butterfly packaged MCI laser. To get stable characterization results, we packaged the MCI laser without the common phase section into a standard 14-pin butterfly package. The output light of the MCI laser is collimated by an aspherical lens. Then a free space isolator is inserted to prevent reflection light influencing the lasing characteristics of the laser. Finally, the light is coupled into a single mode fiber through a collimating lens. The overall fiber coupling efficiency is estimated to be about 60%. Isolator, thermistor and thermoelectric cooler (TEC) were also packaged into the butterfly package. Due to the limitation of the pin number, wavelength locker is not included in the package. The 14 pins include gain, ground, 8 arm phase sections anodes, thermistor +, thermistor -, TEC + and TEC -. Then the butterfly-packaged MCI laser was fixed in a custom-made mount. The output light was split by a 10:90 coupler. 10 percent of the light was coupled into an OSA to measure the optical spectrum of the characterization result after an optimization process finished. The other 90 percent of the light was coupled into an optical bandpass filter. The output light from the optical bandpass filter is detected by a photodetector. In the following test, working temperature was set at room temperature and the current injected into the gain section was fixed at 100 mA. Figure 4 shows the characterization results of wavelength tuning. We characterized the laser with a frequency spacing of 25 GHz. The characterization process spent several hours, which is mainly limited by the slow control of the Keithley source meters through GPIB. The superimposed lasing spectra are shown in Fig. 4(a). The laser longitudinal mode spacing is 0.272 nm. Frequency errors are within ± 2 GHz. The maximal tuning range that the MCI laser can achieve is more than 52 nm at a fixed temperature [13]. However, SMSRs on both sides of the tuning range decrease dramatically due to decreased modal gain and mode competition around the gain peak. Mode jump is easier to happen at these wavelengths. Only in a range of less than 45 nm, SMSRs can maintain high values and the lasing wavelength can keep stable. Here, we only characterized the device in the stable tuning range, which is more than 40 nm, as shown in Fig. 4. The tuning range is red-shifted from the C band, because the bandgap of the MQWs deviated from the designed value, which will be adjusted in our next run. The corresponding SMSRs measured by the OSA are shown in Fig. 4(b). SMSRs are above 45 dB across the tuning range. As the red asterisks shown in Fig. 4(b), SMSRs of some lasing wavelengths are relatively low, which is resulted from the imperfect current settings found by the PSO optimization processes. LIV curves are shown in Fig. 5. Kinks of the LI curves are mode jumps between adjacent cavity modes. Increasing the injection current of the gain section causes phase change of the laser cavity due to heating of the gain section. Threshold currents of six lasing wavelengths are less than 18 mA. Fiber powers of the device are around 5 mW at an injection currents of 120 mA. The overall fiber coupling efficiency is about 0.6, so the chip output powers are supposed to be more than 8 mW.

 

Fig. 3. Experimental characterization setups for the MCI laser.

Download Full Size | PDF

 

Fig. 4. (a) Superimposed lasing spectra of the butterfly-packaged MCI laser with a frequency spacing of 25 GHz; (b) Corresponding SMSRs characterized by PSO algorithm and after aligning the cavity modes of the eight cavities with the reflection peak.

Download Full Size | PDF

 

Fig. 5. Measured LIV curves of the device.

Download Full Size | PDF

5. Tuning characteristics of the MCI laser

For the MCI laser, eight phase sections have to be controlled simultaneously, so it’s difficult to generate tuning maps by sweeping control currents. First of all, we measured the impact of each arm phase section on the output power and lasing wavelength of the MCI laser. After finding the injection currents needed to output the desired lasing wavelength through the PSO algorithm, we swept the injection current of each arm phase section one by one and measured the output power and lasing wavelength of the MCI laser. In the process of sweeping, injection current of one arm was changed from 1 mA to 16 mA, while injection currents of the other seven arm phase sections were kept constant and set to be the values found by the optimization algorithm based characterization scheme. The sweeping process was repeated for all the eight arm phase sections. Wavelength was measured by a multi-wavelength meter (Keysight 86122C).

Figure 6 gives the results of a lasing wavelength at 1565 nm. Lengths of the eight arms increase from Fig. 6(a) to Fig. 6(h). The eight red asterisks are original values of the injection currents found by the optimization algorithm based characterization scheme. We can see that the eight red asterisks are located around a maximum of each power curve, which correspond to the wavelength of 1565 nm. In Fig. 6(a), as the square root of the injection current increases, the output power changes like a cosine curve. There are two minima and one maximum. Absolute round-trip phase difference between the two minima and the maximum should be π. Lasing wavelength also changes with the output power. However, the lasing wavelength shows a trend of red shift, which is resulted from the heat generated by the increasing injection current. In Figs. 6(b)–6(h), mode hopping around the minima of the power curves occurs. Mode hopping around the minima of the power curves is easier to happen as the lengths of the arms increase. Longer the length of the cavity is, smaller the FSR is. Also, full-width-half-maximum (FWHM) of the cavity modes become narrower. Therefore, phase changes of the long arms influence the shape of the whole reflection spectrum more, especially when the round-trip phase difference is around π. As a result, mode hopping is easier to happen around the minima of the power curves. In Figs. 6(e) and 6(f), we can see the output power curves have two maxima, which both correspond to the wavelength of 1565 nm. This means that the two maxima have a round-trip phase difference of 2π. Besides, output power of the second maximum is less than that of the first maximum, which is resulted from the increased loss of free carrier absorption. The optimization algorithm based characterization scheme may find working point at the second maximum, as shown in Fig. 6(e). Normally, we tend to choose the first maximum, because it generates less heat and has higher output power.

 

Fig. 6. Output power and lasing wavelength versus square root of the injection currents of the eight arm phase sections. The lengths of the eight arms increase from (a) to (h). The blue line is output power, the orange dash line is lasing wavelength and the red asterisk is injection current found by the optimization algorithm based characterization scheme.

Download Full Size | PDF

In some cases, injection currents of the eight arm phase sections found by the optimization algorithm based characterization scheme are not all at the maxima of the power curves but around the maxima. Some working points are a little deviated from the maxima. That is why SMSRs measured after the PSO optimization processes show large variation, as shown in Fig. 4(b). To solve this problem, we can adjust the injection currents to make them locate at the maxima of power curves after the optimization algorithm based characterization being executed. In the process of adjustment, lasing wavelength also changes slightly. However, change of wavelength around the maxima of the power curve is within several pm, which has little influence. We show SMSRs measured after aligning the operating currents at the maxima of power curves in Fig. 4(b), which have a small variation across the tuning range. If the wavelength after adjustment isn’t in the range allowed, we can adjust the phase of the common phase section to compensate the wavelength shift and then align the operating currents at the maxima again.

To figure out the relationship between the lasing wavelength and tuning currents, we characterized the MCI laser from 1565 nm to 1585 nm with the laser longitudinal mode spacing of 0.272 nm. Frequency errors are within 2 GHz and SMSRs are above 50 dB. Tuning currents of the eight arm phase sections are shown in Figs. 7(a)–7(h). The asterisks are values of tuning currents acquired by characterization and the red lines are tuning currents fitted with a parabolic curve. We can see that as the output wavelength changes, the tuning currents change periodically. Periods of the tuning currents of the eight arm phase sections are about 1.87 nm, 3.3 nm, 3.5 nm, 3.8 nm, 6.6 nm, 4.8 nm, 3.5 nm, 1.22 nm, respectively.

 

Fig. 7. (a-h) Tuning currents of the eight arm phase sections versus lasing wavelength. The lengths of the eight arms increase from (a) to (h).

Download Full Size | PDF

The MCI laser can be regarded as a combination of eight FP cavities. Each cavity has their own free FSR $\Delta {\mathrm{\lambda }_i}$ (i=1, 2, 3…8), as shown in Fig. 8(a). Due to different arm lengths used, these FSRs are slightly different. For the packaged MCI laser, the calculated FSRs of the eight cavities are 0.315 nm, 0.294 nm, 0.293 nm, 0.291 nm, 0.259 nm, 0.255 nm, 0.25nmand 0.221 nm respectively. The targeted output wavelength spacing is close to these FSRs. The optimization process makes each FP cavity have one of their modes align with the target wavelength, as shown in Fig. 8(a). Consequently, each FP cavity needs to shift their cavity mode slightly through current injected into the arm phase section in order to align with the targeted output wavelength. Due to Vernier effect, for each FP cavity the alignment situation between its cavity modes and the target output wavelengths will show a periodic behavior. This means that the tuning currents injected into the phase section will also show a periodic behavior as demonstrated in Fig. 7. The period in wavelength can be simply calculated as

 

Fig. 8. (a) Schematic drawing of the cavity modes of the eight FP cavities and the targeted output wavelengths; (b) periods of the tuning currents of the eight arm phase sections. The red squares are values calculated by Eq. (1) and the asterisks are values extracted from the tuning currents.

Download Full Size | PDF

Besides, we also characterized the packaged MCI laser with a much smaller wavelength spacing by aligning the operating currents at the maxima of the power curves. In the process of characterization, the longest arm was chosen as the reference arm. First of all, we found the operating currents of a desired wavelength by the optimization algorithm based characterization scheme and adjusted the operating currents to the maxima of the power curves. Then, we increased the injection current of the longest arm with a small step and kept it constant, while injection currents of the other seven arm were adjusted to make them locate at the maxima of the power curves. Finally, the lasing spectrum was measured by the OSA. Repeating this process, the lasing wavelength can be decreased continuously with a small step of about 20 pm. The characterization results are shown in Fig. 9. The injection currents change periodically and the periods of the injection currents are consistent with the calculated FSRs of the eight cavities. As the lasing wavelength decreases, all the injection currents increase. Because increasing the injection current of the reference arm means to blue-shift the cavity mode of the longest cavity. To align the other seven cavity modes with the reference cavity mode, injection currents of the other seven arms also need to increase so as to blue-shift their own cavity mode. Although wavelength decreases of the eight cavity modes are equal, round-trip phase changes of the eight arms are different due to different FSRs. Round-trip phase changes of the eight cavities can be calculated as

 

Fig. 9. (a-h) Injection currents of the eight arm phase sections versus lasing wavelength. The lengths of the eight arms increase from (a) to (h). The blue asterisks are injection currents acquired by making the phases of the other seven arm align to the phase of the longest arm, the red lines are fitted with a parabolic curve, and the black circles are injection currents shown in Fig. 7.

Download Full Size | PDF

6. Conclusion

In this paper, we report the latest development of the MCI widely tunable semiconductor laser with improved performance. Mode selection of the MCI laser is based on constructive interference of eight coupled FP cavities. Therefore, they can be fabricated by conventional photolithography. In our first run, all the waveguides of the lasers were deeply etched with a single ICP dry etching so as to reduce the complexity of fabrication. The laser had high threshold currents of more than 30 mA and low total output powers. To improve the performance of the MCI laser, we made the gain and phase sections to be surface ridge waveguides, while the other waveguides were deeply etched. We packaged the MCI laser without the common phase section into a standard 14-pin butterfly package. The butterfly-packaged MCI laser shows a tuning range of more than 40 nm with SMSRs higher than 48 dB and threshold currents below 18 mA at room temperature.

Besides, we also investigated the relationship between the lasing wavelength and tuning currents. By making the MCI laser output light at a wavelength spacing of the laser cavity mode spacing, we found that the injection currents of the eight arm phase sections change periodically and periods of the tuning currents of the eight arm phase sections are different. The MCI laser can be treated as a combination of eight FP cavities. Because the laser cavity mode spacing is close to the FSRs of the eight cavities of the MCI laser, the tuning currents show periodic behaviors due to Vernier effect. Actually, the injection currents of the eight arms change with periods equal to their own FSRs. Thus, using the relationship between the lasing wavelength and the injection currents, we can make the laser output any desired wavelength by fitting and interpolation.