December 1, 2010 / Vol. 35, No. 23 / OPTICS LETTERS
3991
160 GHz harmonic mode-locked AlGaInAs 1:55 μm strained quantum-well compound-cavity laser Lianping Hou,1,* Mohsin Haji,1 Rafal Dylewicz,1 Piotr Stolarz,1 Bocang Qiu,2 Eugene A. Avrutin,3 and A. Catrina Bryce1 1
School of Engineering, University of Glasgow, Glasgow G12 8LT, UK 2
3
Intense Ltd., Glasgow G72 0BN, UK Department of Electronics, University of York, York Y010 5DD, UK *Corresponding author:
[email protected]
Received August 23, 2010; revised October 4, 2010; accepted October 18, 2010; posted November 2, 2010 (Doc. ID 133911); published November 23, 2010 We characterized the reflectivity and the modal discrimination of intracavity reflectors (ICRs) with different numbers of slots and presented harmonic mode-locking operation of a monolithic semiconductor laser comprising a compound cavity formed by a single deeply etched slot ICR fabricated from 1:55 μm AlGaInAs strained quantum well material. Gaussian pulses were generated at a 161:8 GHz repetition rate with a pulse duration of 1:67 ps and a time–bandwidth product of 0.81. © 2010 Optical Society of America OCIS codes: 140.4050, 140.7090, 190.4160.
Generation of picosecond optical pulses from semiconductor mode-locked lasers (MLLs) at high repetition rates has gained increasing attention due to its significance in next generation optical communications [1]. One of the distinct ways of achieving high-frequency repetition rates is to introduce harmonic mode-locking (ML) techniques. A harmonic MLL produces an optical pulse train at a harmonic of the fundamental round-trip frequency of the device, which can be achieved with subharmonic optical injection [2], colliding pulse ML (CPM) [1,3], and compound-cavity ML (CCM) [4,5]. Frequencies of up to 1:5 THz have been achieved with the use of CCM effects in distributed Bragg reflector lasers [4], but under a narrow range of operation conditions and with limited reproducibility. Later, purpose-built AlGaAs/GaAs CCM lasers showed operation at repetition frequencies of up to 2:1 THz, at the wavelength of 850 nm [5]. Theoretical analysis performed in [5] implied that, while the accurate position of the intracavity reflector(s) [ICR(s)] in the cavity was important for CCM operation, the stability of the CCM regime could be potentially higher than that of the rival CPM approach due to the strong linear modal selectivity of the compound cavity. In agreement with theory, the best results at very high harmonic numbers/repetition rates of ∼1000 GHz were obtained using lasers with more than one ICR positioned at integer fractions of the cavity length [5]. The twin-ICR approach allows more accurate selection of the desired harmonic number than the single-slot approach, as the relative position of the two ICRs is defined lithographically [5,6]. On the other hand, it lacks the flexibility of the single-ICR layout, which allows the harmonic number to be chosen at the cleaving stage, which makes the simpler single-ICR constructions also worth investigating, particularly at relatively modest repetition rates, such as ∼100–160 GHz. Because of the flexibility of the single-ICR layout, in this Letter, we analyze the dependence of the performance of such a laser on the reflectance of the ICR (which in our study is varied by changing the number of slots that form the single ICR) theoretically and experimentally. Our recent reports on the ML behavior of lasers fabricated in the AlGaInAs/InP material system indicate 0146-9592/10/233991-03$15.00/0
that it has potential for better intrinsic characteristics compared to the conventional InGaAsP/InP system [7]. We extend the approach in [5] from short wavelength (GaAs/AlGaAs 850 nm) to a longer wavelength (AlGaInAs/InP 1550 nm). Harmonic CCM devices are very demanding on the ICR reflectivity. Therefore, the key challenge was to obtain a high-quality mirror structure with low scattering loss and optimum reflectivity in order to minimize the threshold current and maximize the quantum efficiency while achieving pure ML at a target harmonic number. The epitaxial structure is described in [7]. The device cavity was designed to be 1070 μm long, with a 20-μmlong saturable absorber (SA), gain section 1 (long gain section), gain section 2 (short gain section), and a 3:2 μm deep ICR based on one to four slots (364 nm wide, ∼50% duty cycle) placed across the 2.5-μm-wide and 1.92-μm-high ridge waveguide (Fig. 1). The ICR was positioned at one-fourth of the cavity length relative to the gain section 2 side mirror, resulting in M ¼ 4 and, hence, an ML frequency of 160 GHz. The fabrication process is described in [6]; the only difference is that here, hydrogen silsesquioxane (HSQ) was used in the planarization of the ridge waveguide [Fig. 1(a)]. Our aim was to obtain the optimum reflectivity for the ICR while reducing the mirror loss to a minimum. Figure 2(a) shows the simulated wavelength-dependent reflection spectra using the finite-difference time-domain method. By using the Hakki–Paoli technique [8], we experimentally measured the reflectivity of ICRs with different numbers of slots. We injected 7 mA into gain section 2 (below the threshold current of 13 mA of a laser with one slot, and 9 mA with two, three, and four slots) and measured the amplified spontaneous emission (ASE) spectra from the output of the cleaved facet at the gain section 2 side of the laser. For comparison, we have also captured the ASE spectrum from an Fabry–Perot laser of the same cavity length (267:5 μm), with the same injection current as that of the gain section 2 on the same wafer. Based on Rcleave ¼ 0:32, we obtained the RG (R being the reflectivity, and G the modal gain [8]) product and the reflectivity as a function of the wavelength for the © 2010 Optical Society of America
3992
OPTICS LETTERS / Vol. 35, No. 23 / December 1, 2010
Fig. 1. (Color online) (a) Scanning electron microscope (SEM) pictures of ridge waveguide with HSQ planarization, (b) laser wafer cross section with a deeply etched single-slot ICR, and (c) schematic of monolithic CCM device with one ICR.
ICRs with different numbers of slots [see Fig. 2(b)]. At the wavelength of 1550 nm, R1-slot ¼ 37%, R2-slot ¼ 70%, R3-slot ¼ 84%, and R4-slot ¼ 90%, which is in good agreement with the simulation results presented in Fig. 2(a). Based on the methods outlined in [5], we have calculated the mode discrimination, i.e., the difference Δamod between the threshold losses of the favored mode and the mode with the nearest threshold, for M ¼ 4 versus the ICR reflectance RICR in a cold (passive) laser cavity. The amplitudes and phases of the ICR reflectances and transmittances were obtained from the threedimensional simulations outlined above. The results are shown in Fig. 3, which indicates that, at RICR ≪ 1, the selectivity of the cavity rapidly increases with an increase in RICR , in agreement with the asymptotic Δamod ∼ RICR 1=2 obtained in [5]. On the other hand, at large RICR , the shorter subcavity begins to dominate the cavity
modes until, with RICR → 1, the shorter subcavity is independent, and there is once again no mode selectivity. The optimum ICR reflectance, giving the maximum modal seðoptimumÞ ≈ 0:32 for the cavity lectivity, is predicted to be RICR geometry used. The experimentally used one-slot ICR has the closest reflectance to this value of all the constructions studied, if slightly higher at R1-slot ≈ 0:37. The characterization of the fabricated lasers showed that the laser with a single-slot ICR did indeed produce the best ML operation at the harmonic number M ¼ 4, in agreement with the simulations. Thus, all the following results are based on lasers with a single-slot ICR. The value of R1-slot could have been slightly decreased to get closer to ðoptimumÞ RICR by tuning the width of the slot; however, the improvement in the selectivity would have been only marginal owing to the rather broad maximum in Fig. 3. For the same experimental conditions as that of Fig. 5, a slight Q-switching envelope was observed for twoslot-ICR, at M ¼ 4. For three-slot-ICR and four-slot-ICR, unstable and incomplete (less than 100% modulation) ML was found, at a repetition frequency equal to the fundamental round-trip frequency of the cavity of length L1 (ML frequency ∼53:33 GHz). The insets of Fig. 3 show the stable and 100% modulation ML for the thrree-slotICR and four-slot-ICR cases with the following bias conditions: I LG ¼ 36 mA, I SG ¼ 2:5 mA, and V SA ¼ −3:0 V. The pulse widths are 1.1 and 0:9 ps, respectively. This solution can be used to fabricate high-quality two-section MLLs without need for an additional high-reflectivity coating applied to the cleaved facet. Pure ML of the device at the harmonic number of M ¼ 4 was observed by forward biasing the two gain sections (60 mA < I LG < 80 mA, 0 mA ≤ I SG ≤ 20 mA) and reverse biasing the absorber section (jV SA j ≥ 2:0 V). Figure 4 shows the typical average output power from the facet adjacent to the SA for different reverse bias levels applied to the SA, with gain section 2 injection current I SG ¼ 5 mA. It can be seen that the threshold current with an unbiased absorber section is 20 mA, and increasing the reverse voltage on the SA leads to an increase in the threshold current. Increasing I SG reduces the threshold current, due to the reduction in the absorption loss/increase in the gain within the short gain section. Kinks were observed on some L–I (light–current) curves, which are likely to be caused by a thermal detuning of the long and short gain sections and are connected with a mode hop [9].
Fig. 2. (Color online) (a) Simulated reflectivity. Inset, SEM micrographs of different ICRs. (b) Measured RG product. Inset, reflectivity as a function of the wavelength for the ICR with different number of slots.
December 1, 2010 / Vol. 35, No. 23 / OPTICS LETTERS
Fig. 3. (Color online) Calculated modal discrimination versus ICR reflectance and the corresponding autocorrelation trace of the pulse train for the ICR with different slots.
Figure 5 presents the optical spectrum using optical spectrum analyzer Agilent 86140B with the resolution bandwidth of 0:06 nm obtained with 80 and 5 mA current for gain 1 and gain 2 sections, respectively, with −2:0 V bias applied to the SA. The resulting optical spectrum is centered at 1564 nm with a 3 dB bandwidth of 3:95 nm and 161:8 GHz mode spacing corresponding to the calculated ML frequency. The autocorrelation trace obtained using an FR-103 XL autocorrelator by second-harmonic generation autocorrelation is also shown in Fig. 5, for the same experimental conditions (gain section current, bias, etc.) as used for that of optical spectrum. The pulse autocorrelation trace has a width of 2:36 ps, which deconvolves to 1:67 ps pulse duration assuming a Gaussian pulse shape. The pulse repetition rate was 161:8 GHz, which is in accordance with the optical spectrum mode spacing. The time–bandwidth product (TBP) of the pulse is equal to 0.81, which is somewhat larger than the
Fig. 4. (Color online) Typical L–I characteristics under SA applied bias from 0 to −3 V, in −0:5 V steps (from top to bottom) with I SG ¼ 5 mA. Inset, output power versus I LG for different I SG settings with VSA ¼ 0 V at a heat-sink temperature of 20 °C.
3993
Fig. 5. (Color online) Optical spectrum and corresponding autocorrelation trace of the pulse train of the mode-locked device for I LG ¼ 80 mA, I SG ¼ 5 mA, and V SA ¼ −2:0 V measured at the temperature of 20 °C.
transform-limited value (≈0:441). This is likely to be due to self-phase modulation in the gain section [10]. In conclusion, we have reported the mode-locked operation of a 160 GHz passively mode-locked AlGaInAs laser using a compound cavity formed by a single deeply etched ICR. A pulse width of 1:67 ps is deduced from intensity autocorrelation measurements. The TBP of the pulse is 0.81, which is somewhat larger than the transform limit (≈0:441) of a pulse with Gaussian profile. Because of the high quality of the ICR and the planarization using HSQ, the device exhibits a relatively low threshold current and high slope efficiency compared with the results in [5,6]. This work was supported by the Engineering and Physical Sciences Research Council (EPSRC) (project EP/E065112/1). References 1. S. Arahira, S. Sasaki, K. Tachibana, and Y. Ogawa, IEEE Photon. Technol. Lett. 16, 1558 (2004). 2. Y. Wen, D. Novak, H. Liu, and A. Nirmalathus, Electron. Lett. 37, 581 (2001). 3. T. Shimizu, I. Ogura, and H. Yokoyama, Electron. Lett. 33, 1868 (1997). 4. S. Arahira, Y. Matsui, and Y. Ogawa, IEEE J. Quantum Electron. 32, 1211 (1996). 5. D. A. Yanson, M. W. Street, S. D. McDougall, I. G. Thayne, J. H. Marsh, and E. A. Avrutin, IEEE J. Quantum Electron. 38, 1 (2002). 6. L. Hou, P. Stolarz, R. Dylewicz, M. Haji, J. Javaloyes, B. Qiu, and A. C. Bryce, IEEE Photon. Technol. Lett. 22, 727 (2010). 7. L. Hou, P. Stolarz, J. Javaloyes, R. Green, C. Ironside, M. Sorel, and A. C. Bryce, IEEE Photon. Technol. Lett. 21, 1731 (2009). 8. S. A. Merritt, C. Dauga, S. Fox, I.-F. Wu, and M. Dagenais, J. Lightwave Technol. 13, 430 (1995). 9. J. Fricke, H. Wenzel, M. Matalla, A. Klehr, and G. Erbert, Semicond. Sci. Technol. 20, 1149 (2005). 10. G. P. Agrawal and N. A. Olsson, IEEE J. Quantum Electron. 25, 2297 (1989).
