We report on an extensive characterization of a 15GHz integrated bulk InGaAsP passively modelocked ring laser at 1530 nm. The laser is modelocked for a wide range of amplifier currents and reverse bias voltages on the saturable absorber. We have measured a timing jitter of 7.1 ps (20 kHz – 80 MHz), which is low for an all-active device using bulk material and due to the ring configuration. Measured output pulses are highly chirped, a FWHM bandwidth is obtained of up to 4.5 nm. Such lasers with high bandwidth pulses and compatible with active-passive integration are of great interest for OCDMA applications.
©2006 Optical Society of America
Integrated modelocked semiconductor lasers are key components for high bit rate telecommunication and other optoelectronic systems . Much of the published research effort has been concentrated on monolithic modelocked lasers in Fabry-Pérot type configurations. Compared to these linear devices, a relatively small amount of results has been published on integrated modelocked ring lasers. A ring configuration has however significant advantages. Firstly, the repetition rate of the laser is mainly determined by the physical length of its cavity. For a ring laser this length is controlled accurately by photolithography as opposed to a device with cleaved facet mirrors where the cleaving process introduces significant fabrication uncertainties. A ring laser typically operates as a Colliding pulse Modelocked (CPM) laser where two counter-propagating pulses collide in the saturable absorber which improves the modelocking performance. Furthermore, if active-passive technology is used the laser can be directly integrated with other devices such as an all-optical switch.
The first monolithic modelocked semiconductor ring laser was reported by Hohimer and Vawter in 1993 . The lasing wavelength of around 874 nm and 1.3 ps transform limited pulses at 86 GHz repetition rate have been reported. The output coupling used was a Y junction. Unfortunately, no RF spectrum of the laser output was reported. In 1998, Yu et al.  have reported on realizations of a number of designs of passively modelocked ring lasers using one or two Saturable Absorbers (SAs). Lasers were fabricated using double quantum well (DQW) GaAs/AlGaAs material. Their symmetric laser cavity where a single Saturable Absorber (SA) is exactly opposite the output coupler, showed approximately a two times smaller modelocked range than a symmetric device with two SAs and an asymmetric device using a single SA. In this paper we mean by symmetric cavity, a cavity where the two counter-propagating pulses experience the same optical path (mirrored) at the same time. Yu et al. attributed the poor performance of the symmetrical ring laser to the pulses colliding in the (active) MMI coupler. No RF spectra had been measured at that time, and the streak camera used had a resolution of 12ps FWHM, which did not fully resolve the pulses. Avrutin et al. in 1999 reported the first results in passively modelocked InP/InGaAsP ring lasers . Their design included two SAs as well in order to double the pulse narrowing effect of the CPM and to avoid the pulse collision in the gain section. Despite applying a range of biases to the absorbers sections, no modelocking at the fundamental frequency was observed. More recently Ohno et al. reported  on a 30GHz hybridly modelocked ring laser with integrated passive waveguides. We have also reported in 2005 on a 27GHz modelocked ring laser using active-passive integration. This device was passively modelocked but its operation suffered from internal reflections .
In this paper we report on the most complete and detailed characterization of an integrated passively modelocked ring lasers reported up to now. All the modelocked states of the integrated modelocked ring laser are described, when only a limited number of operating points were reported in Refs. [2–6]. After a description of the design, the modelocking regions are presented on a map of the amplifier current and the reverse bias applied on the SA. The modelocking operation is defined on the basis of the recorded RF spectra. The laser shows modelocked operation for a relatively large range of current and reverse biases on the SA. The RF peak linewidth of the modelocking states has been measured as well. The lasers show a relatively good timing stability for a passively modelocked device compared to Fabry-Pérot devices fabricated on the same wafer. Optical spectra have been measured at all the studied operating points; the optical bandwidth of the laser is wide at a high reverse bias voltage i.e. a 4.5 nm FWHM has been measured from one of our devices. Pulses that we measured are wide (5–8 ps) with regard to the available bandwidth, which means that the pulses are highly chirped. The range of tunability of the modelocked repetition rate is also presented on a map. Besides, the evolution of the integrated modelocked ring laser through its different states is illustrated by two animation movies. The first movie shows the evolution of the RF spectrum when scanning the amplifier current and keeping the SA voltage at a fixed reverse voltage. The second one shows, in the same conditions, the evolution of the optical spectrum. Finally jitter measurements have been performed on the passively modelocked ring lasers. The value measured for the 20 kHz – 80 MHz range is compared with passively modelocked FP lasers as no measurements of timing jitter have been reported in literature for passively modelocked integrated ring lasers. Such lasers with high bandwidth pulses and compatible with active-passive integration are of great interest for Optical Code Division Multiple Access applications (O-CDMA) .
A photograph of two realized 15 GHz mode-locked ring lasers is presented in Fig. 1. It shows the layout of the lasers. The ridge waveguides are 2 µm wide and are visible as dark lines. The ring is 5.291 mm long. In order to avoid any reflections that can occur, a number of measures have been taken in the design. The bends have their radius of curvature which decreases adiabatically down to a 500 µm in the middle of the curve, in order to avoid offsets between straight and curved waveguides which can give rise to reflections. The coupler is a 200 µm long directional coupler with a gap of 0.9 µm, this design was preferred to multimode interference coupler (MMI) also to avoid any reflection . The coupler was designed to give ~35% of output coupling. Facet reflectivity is reduced using angled facets at 7° and an AR-coating. The design uses three metal contacts as indicated in Fig. 1. The larger one contacts most of the ring and the directional coupler. The two output waveguides have their own electrical contact. Finally a short separate electrical contact connected to a larger pad is used for the SA. The SA lengths used are 30 or 60 µm. The devices are realized in material consisting of a 120 nm thick λ=1.55 µm bulk InGaAsP layer between two 190 nm thick λ=1.25 µm InGaAsP layers. The structure is clad by a 1500 nm thick p-InP layer with gradual doping levels and a 50 nm p-InGaAs contact layer. The electrical isolation sections that separate the different electrical contacts are 20 µm long and under a 45 degrees angle with the waveguide, furthermore part of the cladding (1200 µm) has been removed during the processing. The electrical isolation obtained between different contacts is 200 kΩ, which is more than sufficient.
3.1 Experimental setup
The InP chip is fixed on a copper mount that is temperature controlled using a TEC (Thermo Electric Controller). Light from both outputs of the laser is fiber coupled using two lensed fibers and led through two optical isolators before any instruments. No problem of reflection from the uncoated lensed fibers back to the ring laser has been observed. The optical spectra have been recorded with a high resolution OSA (APEX AP2041A) with a resolution of 0.8 pm. An EDFA has been used for amplification in order to have sufficient intensity to measure time series, RF spectra and autocorrelator traces. RF spectra were recorded using a 12 ps IR photodetector (New Focus model 1024) and a 50 GHz Agilent PSA E4448A spectrum analyzer. The Single Side Band (SSB) phase noise spectra have been measured directly on the spectrum analyzer using the phase noise measurements utility. Autocorrelator traces have been recorded from an APE “PulseCheck TC” autocorrelator. For all the results presented here, the current of the two output waveguides has been kept at 140 mA. It is above transparency so there is a small amount of amplification but low enough for the ASE or reflections not to influence the behavior of the modelocked ring laser.
3.2 Modelocking region and RF spectra
The modelocking region has been delimited from the measured RF spectra. Two detailed observed RF spectra from the ring laser are presented in Figs. 2(a) and 2(b). In theoretical literature proper modelocked operation is identified when the height of the peak in the RF spectrum at the fundamental frequency is 25dB above any other component in the RF spectrum below the fundamental frequency . In this paper we also use this convention to identify proper modelocked operation of the laser. In addition to this we identify the laser as being imperfectly modelocked when the peak at the fundamental frequency is 25 dB over the noise floor, but does not satisfy the requirement for proper modelocking. In the spectrum in Fig. 2(a) the peak at the fundamental frequency is 50 dB over the noise floor and 33 dB over the lower frequency signal intensity. In this situation the ring laser is thus properly modelocked. Next to the low frequency noise a small peak around 1 GHz is visible. It corresponds to the relaxation oscillation of the laser which is becoming dominant in self pulsation regimes at high current injection in the amplifier. A detailed view of the RF spectrum around the fundamental frequency (span of 250 MHz) is presented in Fig. 2(b). The linewidth of the peak in the RF spectrum is 2.8 MHz at −20 dB and less than 400 kHz at −3 dB. For each measured point, the RF spectra have been recorded on a large span and short span. The parameter ranges for both modelocking regimes have been determined by scanning the amplifier current and absorber voltage and recording the RF spectra from the laser output. The evolution of the spectrum when scanning the amplifier current and keeping the SA voltage fixed at V=-2.0V is illustrated by the animation linked to Fig. 2. The SA voltage value has been chosen for the current scan to cover every regime. When the current is increased, the peak at the fundamental frequency slowly increases in intensity and gradually narrows down to a constant value for the properly modelocked states. At higher current values the transition from a modelocking state to a self pulsating regime occurs abruptly.
Modelocking is achieved for a range of reverse bias voltages on the SA between 0.9 and 2.2 V and a range of amplifier currents between 355 and 410 mA (3.3 kA/cm2 and 4.9 kA/cm2). The parameter range for modelocked operation is presented in Fig. 3. Here the intensity of the peak at the modelocking frequency in the RF spectrum is presented as a function of amplifier current and SA reverse voltage (device Labs=40 µm). Operating points that show proper modelocking are represented in a color scale in the graph and those that show imperfect modelocking are represented in a grey scale. For most of the detected modelocked states the peak at the fundamental frequency in the RF spectrum is higher than 40 dB above the noise floor. Fig. 4 shows the width of the peak in the RF signal at 20 dB below its maximum for all recorded current and reverse bias voltage values. We have chosen a threshold of 20 dB, as we used a 16 kHz resolution, the discretization of the data would be too much visible at the bandwidth at −3dB. For reverse bias voltages between 1.2 and 2.2 and for operating points where the laser is properly modelocked, most of the measured -20dB linewidths are between 2.5 and 5.5 MHz (dashed lines on the graph).
3.3 Output power of the RMLL
The power of both outputs of the ring laser (corrected for coupling and isolator losses) are plotted in Figs. 5(a) and 5(b) as function of amplifier current and SA reverse bias voltage. In Fig. 5(c), the sum of the power from the two outputs is plotted. The common current supply of the two output waveguides is kept at 140 mA. No hysteresis has been observed at the laser threshold. The laser starts lasing at 338 mA and at 347 mA for an applied reverse bias voltage on the SA of respectively −0.8 V and −2.5 V. Figure 5(d) shows the output power levels in the two directions at −1.8 V on the SA in detail. The ring laser operates CW in both directions below and within the current range where the laser is modelocked. Increasing the current further from the modelocking regime, depending on the reverse bias voltage on the SA, the ring laser enters a self-pulsating regime. The transition into the modelocked state from the lower current values is almost not visible on the output power curves. Knowing the transition from the modelocking to the self-pulsating state from the evolution of the RF spectra, one can notice a small asymmetry in both output power levels from the laser.
The larger current range, for which the ring laser stays operating with the same output power in both directions, is at a reverse bias voltage of -1.9 V on the SA. For all measured SA voltages, the transition where the symmetry in lasing direction is broken is very clear in the graphs 5(a), 5(b) and 5(c). Actually this transition might be explained by differences in either in feedback from the end facets  or Amplified Spontaneous Emission (ASE) from the two pumped output waveguides injected into the laser cavity, or a combination of these effects. These output waveguides differ in length by a factor of four (see Fig. 1). From -0.8 V to -1.9V reverse bias voltage on the SA, a possible explanation would be that the ASE of the long output waveguide coupled into the ring laser cavity is more intense and makes the laser operating almost uni-directionally. For reverse bias voltages of -1.9 V to -2.5V on the SA, the situation is more complex and no obvious explanation is available. For any reverse bias voltage on the SA, above 445 mA laser injection current, the situation becomes more complicated; intermittent switching of the lasing direction in the ring is observed.
3.4 Optical spectra of the RMLL
A typical example of an optical spectrum of a ring laser in a modelocked state is plotted in Fig. 6. The spectrum is smooth and wide (4 nm FWHM). The shape is not symmetric, more modes are active at the lower wavelength side of the peak. The linewidth of a lasing mode is typically around 600 MHz. The animation of the evolution of the optical spectrum versus the amplifier current is linked to the Fig. 6. Again here the reverse voltage on the absorber is fixed to -2.0 V and the output waveguide current is 140 mA. The laser starts lasing at 340 mA. The ring laser operates CW in both directions up to 364 mA before modelocking. The transition is clearly visible in the spectrum, the top of the optical spectrum becomes more rounded. If the current is increased further, the laser stays modelocked until 408 mA in the case presented in Fig. 5. This is not visible from this animation; one should watch the evolution of the RF spectra from Fig. 2. Between 408 and 428 mA the modelocking deteriorates and shows oscillations around 1 GHz. Above 428 mA the evolution of the spectrum can be interpreted as that the gain curves of the amplifier and absorber shift away from each other at high current, making the laser to become unstable. Intermittent switching of the lasing direction in the ring is observed above 445 mA.
The FWHM of the optical spectrum of the laser is plotted in Fig. 7 as a function of the amplifier current and SA reverse bias voltages. Only the operating points that show good or imperfect modelocking, delimited in Fig. 3, are plotted. One can clearly see in Fig. 7 that the bandwidth increases with the reverse bias voltage and thus with absorption, but only up to a certain point. Above that point the modelocking deteriorates and the bandwidth reduces as well. This can be attributed to a change in absorption or a mismatch in spectrum between the absorber and amplifier due to the increasing electric field strength in the absorber. In the case of the 40µm long SA, for V=-2.3 V the bandwidth does not exceed 3.5 nm. The largest bandwidth recorded was 4.5 nm from the device with the 50 µm long SA.
3.5 Autocorrelator results
The measured optical spectrum would support pulses down to 700 fs if the laser would produce transform limited pulses. However, the autocorrelator reveals long pulses (5–8 ps) with sharper peaks on top and a clear zero level in between. An example of an autocorrelator trace is plotted in Fig. 8 for the device with a longer SA (50µm). The shape of the autocorrelator trace indicates there is an issue with the dispersion in the fibers used in the setup and/or a chirp on the pulses. The peaks on the top of the pulses are not coherence peaks but they are due to a partial compression of the pulse . The output pulses are highly and nonlinearly chirped. This was shown by filtering the laser output with a 1 nm bandwidth filter before leading it to the autocorrelator. In this situation the autocorrelation showed 7.5 ps pulses without sharp peaks on top of the pulses. Adding 8 meters of standard single mode fiber the pulse could be compressed down to just over 4 ps which is approximately two times over the bandwidth limit. We suspect that the non-linearity of the chirp is linked to the colliding pulse regime and the very long all active laser cavity with a high confinement factor (0.28) in the bulk gain material. The transition from the modelocking regime and the self-pulsating one is almost not visible on the autocorrelator traces.
3.6 Repetition rate
The repetition rates have been recorded for all the modelocked states (good and imperfect). The repetition rate as a function of amplifier current and SA reverse bias voltage is plotted in Fig. 9. The total range of tuning is 60 MHz. As described by Arahira and Ogawa in Ref.  the tuning is led by the gain/absorption saturation effects. Changes in refractive index are not sufficient to cover such a range. According to the explanation of Arahira and Ogawa, the presented device operates with a low unsaturated gain and in such a case the repetition rate decreases with increasing pulse energy. The pulse energy increases with the amplifier current thus the repetition rate decreases. Increasing the reverse bias voltage on the SA decreases the pulse energy, as a consequence the repetition rate increases.
3.6 Jitter measurements
No measurements of the timing jitter are reported in literature for modelocked ring lasers. In literature only a limited number of measured jitter values are available on passively and hybridly modelocked Fabry-Pérot lasers. Jitter values reported for passively modelocked lasers are typically higher than 8 ps [14–19]. Details of the measured values are listed in Table 1. The range of integration over the phase noise signal is not always the same or even indicated in many publications. A reduction of the timing jitter value by a factor greater than two has been demonstrated in Ref.  by using an integrated External Cavity Laser (ECL). The reduction in length of active waveguides has two positive effects. It first reduces the nonlinear phenomena in the amplifier (dispersive effects and total amount of carriers) thus the pulses width. It decreases as well the total amount of Amplified Spontaneous Emission (ASE) and carrier dynamics which are the main sources of timing jitter in a passively modelocked laser .
A recent paper  of Ji et al. reports a low value of 3.6ps (range of integration: 20 kHz -80 MHz) in free running mode for a colliding pulse modelocked laser (MQW) with two integrated extended passive waveguides. Colliding pulse modelocked (CPM) lasers provide a deeper saturation of the SA and a more effective pulse narrowing compared to a conventional modelocked laser design with the absorber at one side of the FP cavity. An important technological issue with a linear CPM design is that the absorber needs to be very precisely in the middle of the cavity which is difficult to achieve with cleaved facets. Our symmetric ring MLL is a CPM due to its bidirectional operation. The advantage of the ring is that the alignment of the SA in the middle of cavity is lithographically controlled and therefore much easier compared with the cleaved devices. The Single Side Band (SSB) phase noise spectrum is plotted in Fig. 10. Timing jitter values obtained from our ring lasers for a number of ranges of integration of the phase noise are listed in Table 1 for a good operating point (Iamp=369 mA, VSA=-1.8V). There is no ITU recommendation for 15 GHz, the integration ranges of the phase noise used are the ones for 10 GHz. We have measured for the 20 kHz -80 MHz range a timing jitter of 7.1 ps, which is low for an all-active device using bulk material. Furthermore a sizeable 3 to 4.5 nm bandwidth is available.
In this paper we have reported on the complete characterization of a 15GHz integrated bulk InGaAsP passively modelocked ring laser at 1530 nm. The laser is modelocked for a wide range of current values in the optical amplifier and reverse bias voltages on the SA. The lasers show a relatively good timing stability for a passively modelocked device i.e. a 50 dB RF peak power with a linewidth<400 KHz at -3 dB. We have measured for the 20 kHz-80 MHz range a timing jitter of 7.1 ps, which is low for an all-active device using bulk material and in our opinion due to the ring configuration. Measured output pulses are highly chirped, a FWHM bandwidth is obtained of up to 4.0 nm for the device with a 40 µm long SA and even 4.5 nm for a 50 µm long SA. The repetition rate has a tuning range of 60 MHz. Such lasers with high bandwidth pulses and compatible with active-passive integration are of great interest for OCDMA applications  where information is coded in the spectrum.
This research is supported by the NRC Photonics program and the Towards Freeband Communication Impulse program of the Dutch Ministry of Economic Affairs. The authors would like also to thank Kresten Yvind for his useful discussions.
References and links
1. S. Arahira and Y. Ogawa, “Retiming and reshaping function of all-optical clock extraction at 160 Gb/s in monolithic mode-locked laser diode,” IEEE J. Quantum Electron. 41, 937–944 (2005. [CrossRef]
2. J. P. Hohimer and G. A. Vawter, “Passive mode locking of monolithic semiconductor ring laser at 86 GHz,” Appl. Phys. Lett. 63, 1598–1600,(1993). [CrossRef]
3. S. Yu, T. F. Krauss, and P. J. R. Laybourn, “Modelocking in large monolithic semiconductor ring lasers,” Opt. Eng. 37, 1164–1168 (1998). [CrossRef]
4. E. A. Avrutin, J. H. Marsh, J. M. Arnold, T. F. Krauss, H. Pottinger, and R. M. De la Rue, “Analysis of harmonic (sub) THz passive mode-locking in monolithic compound cavity Fabry-Perot and ring laser diodes,” IEE Proc. Optoelectron. 146, 55–61 (1999). [CrossRef]
5. T. Ohno, H. Ishii, S. Matsuo, H. Okamoto, Y. Kawaguchi, Y. Kondo, I. Furuta, H. Ito, and Y. Yoshikuni, “Hybrid modelocking of semiconductor ring lasers incorporating passive deep-ridge waveguides,” Electron. Lett. 38, 884–885 (2002). [CrossRef]
6. Y. Barbarin, E. A. J. M. Bente, M. J. R. Heck, J. H. den Besten, G. Guidi, Y. S. Oei, J. J. M. Binsma, and M. K. Smit, “Realization and modeling of a 27-GHz integrated passively mode-locked ring laser,” IEEE Photon. Technol. Lett. 17, 2277–2279 (2005). [CrossRef]
7. C. Ji, R. G. Broeke, Y. Du, J. Cao, N. Chubun, P. Bjeletich, F. Olsson, S. Lourdudoss, R. Welty, C. Reinhardt, P. L. Stephan, and S. J. B. Yoo, “Monolithically integrated InP-based photonic chip development for O-CDMA systems,” IEEE J. Sel. Top. Quantum Electron. 11, 66–77 (2005). [CrossRef]
8. E. C. M. Pennings, R. van Roijen, M. J. N. van Stralen, P. J. de Waard, R. G. M. P. Koumans, and B. H. Verbeek “Reflection properties of multimode interference devices,” IEEE Photon. Technol. Lett. 6, 715–718 (1994). [CrossRef]
9. U. Bandelow, M. Radziunas, A. Vladimirov, B. Hüttl, and R. Kaiser, “40GHz modelocked semiconductor lasers: Theory, simulations and experiments,” Opt. and Quantum Electron. 38, 495–512 (2006). [CrossRef]
10. M. Sorel, G. Giuliani, A. Scire, R. Miglierina, S. Donati, and P. J. R. Laybourn, “Operating regimes of GaAs-AlGaAs semiconductor ring lasers: experiment and model,” IEEE J. Quantum Electron. 39, 1187–1195 (2003). [CrossRef]
11. S. Gee, G. A. Alphonse, J. C. Connolly, C. Barty, and P. J. Delfyett, “Ultrashort pulse generation by intracavity spectral shaping and phase compensation of external-cavity modelocked semiconductor lasers,” IEEE J. Quantum Electron. 36, 1035–1040 (2000). [CrossRef]
12. S. Arahira and Y. Ogawa, “Repetition-frequency tuning of monolithic passively mode-locked semiconductor lasers with integrated extended cavities,” IEEE J. Quantum Electron. 33, 255–257 (1997). [CrossRef]
13. L. A. Jiang, M. E. Grein, H. A. Haus, and E. P. Ippen, “Noise of mode-locked semiconductor lasers,” IEEE J. Sel. Top. Quantum Electron. 7, 159–167 (2001). [CrossRef]
14. J. D. Derickson, R. J. Helkey, A. Mar, J. R. Karin, J. G. Wasserbauer, and J. E. Bowers, “Short pulse generation using Multisegment Mode-Locked Semiconductor Lasers,” IEEE J. Quantum Electron. 28, 2186–2202 (1992). [CrossRef]
15. A. G. Weber, M. Schell, G. Fischbeck, and D. Bimberg, “Generation of single femtosecond pulses by hybrid mode-locking of a semiconductor laser,” IEEE J. Quantum Electron. 28, 2220–2229 (1992). [CrossRef]
16. H. Yokoyama, T. Shimizu, T. Ono, and Y. Yano “Synchronous injection locking operation of monolithic mode-locked diode lasers,” Opt. Rev. 2, 85–88 (1995). [CrossRef]
17. X. Wang, H. Yokoyama, and T. Shimizu, “Synchronized harmonic frequency mode-locking with laser diodes through optical pulse train injection,” IEEE Photon. Technol. Lett. 8, 617–619 (1996). [CrossRef]
18. F. Camacho, E. A. Avrutin, P. Cusumano, A. Saher Helmy, A. C. Bryce, and J. H. Marsh, “Improvements in mode-locked semiconductor diode lasers using monolithically integrated passive waveguides made by quantum-well intermixing,” IEEE Photon. Technol. Lett. 9, 1208–1210 (1997). [CrossRef]
19. C. Ji, N. Chubun, R. G. Broeke, J. Cao, Y. Du, S. J. B Yoo, K.Y. Liou, J. R. Lothian, S. Vatanapradit, S. N. G. Chu, B. Patel, W. S. Hobson, and W. T. Tsang, “Synchronized transform-limited operation of 10-GHz colliding pulse mode-locked laser,” IEEE Photon. Technol. Lett. 18, 625–627 (2006). [CrossRef]