Fast pulsed mode-locked lasers 1
1
Erwin Bente , Martijn Heck1,2, Pascual Muñoz3, Amandine Renault2, Richard Nötzel1, Meint Smit1 COBRA Research Institute, Technische Universiteit Eindhoven, P.O. Box 513, 5600 MB Eindhoven, the Netherlands Phone: +31 40 2475106, Fax +31 40 2448375, E-mail:
[email protected] (2Laser Centre Vrije Universiteit, Amsterdam, the Netherlands) 3 ( Grupo de Comunicaciones Opticas, Universidad Politecnica de Valencia, Spain)
Abstract - InP integration technology and InAs/InP(100) quantum dot gain material are shown to be promising for realising fully integrated modelocked laser systems operating at 1.5µm. Fast switching of output pulse parameters is possible with such systems. Introduction Nowadays ultra-fast optical systems are available for many research and development laboratories in academia and industry. These lasers systems are typically based on titanium doped sapphire (Ti:S) laser material and Kerr lens modelocking, and more recently modelocked short pulse fibre lasers and amplifiers. This availability of ultra-fast pulses has driven many research areas from basic physics to biological and medical research and to mechanical engineering. Spectacular results have been achieved such as extremely accurate optical frequency combs [1] or coherent excitation applications in chemistry and biophotonics [2]. However the laser systems involved can still be bulky and vulnerable, and require skilled personnel to operate them. The cost involved prevents the application of pico- and femtosecond pulses that have been demonstrated in many fields, such as telecommunication (time domain multiplexed systems, synchronized pulse sources, multiwavelength lasers for wavelength-division multiplexed systems), photonic sampling, optical clock distribution, multi-photon microscopy, and sensing. Semiconductor modelocked lasers are much more compact and in principle robust, but until recently have turned out to have to narrow operating regimes for practical applications. Indium-phosphide (InP) based semiconductor laser technology and planer optical integration technology have the potential to change that for sources with a wavelength around 1550 nm. There are two developments that make that one can envision fully integrated laser systems on a single semiconductor chip. The first is planar active/passive integration. Optical amplifiers, passive waveguide devices and electro-optic phase modulators may be combined in this way [3]. The second is the advent of InAs/InP (100) quantum dot (QD) gain material [4]. In this paper we present two components with which such an integrated modelocked laser system may be put together: the mode-locked laser diode (MLLD), where we will focus on the latest results obtained using quantum dot gain material, followed by the pulse shaper. Obviously integrated InP MLLD systems will have limitations certainly when compared to the solid-state.
We end with discussing these and how some of them might be overcome. Modelocked laser diodes and quantum dots In our work, MLLs integrated on an InP chip have been developed that are passively modelocked using a saturable absorber (SA). The SA is a short Semiconductor Optical Amplifier (SOA) section that is reversely biased. The work has been largely concentrated on ring laser cavities [5] Indeed, much of the research effort that can be found in the literature is focused on monolithic modelocked lasers in Fabry-Pérot (FP) type configurations [6]. A ring configuration has however significant advantages. Firstly, the repetition rate of the laser can be controlled accurately by photolithography as opposed to a device that has cleaved facet mirrors. Secondly, a ring laser typically operates as a Colliding Pulse Modelocked (CPM) laser where two counterpropagating pulses collide in the SA [7]. This improves the modelocking performance, in particular the stability. Furthermore, when active-passive technology is used the laser can be directly integrated with other devices such as an all-optical switch or a pulse compressor. We have fabricated ring and linear MLLDs in the material system InP/InGaAsP with bulk gain material. Using either bulk or quantum well gain material the bandwidth of these MLLDs is however limited to between 1 nm – 5 nm [8, 5]. For many applications a broad coherent optical bandwidth of the output of the source is necessary. This bandwidth may well exceed 1 THz so pulses significantly shorter than 1 ps may be produced. Also the operating regime for the MLLD, that is the range of values for the current through the optical amplifier section and the reverse bias voltage on the absorber section, should be large for practical implementation. With quantum well and bulk gain material, these operating ranges have turned out to be so small that this hampers applications. The origin of these narrow operating ranges is for a large part related to the self-phase modulation effects originating from the effect of the carrier concentration on the refractive index. The carrier concentration in the amplifier and absorber do vary more and faster when the pulse inside the laser becomes shorter, and thus this leads to stronger frequency changes throughout the pulse (chirp). This also points a way to how to get to shorter pulses: minimise the dynamics in the laser cavity while maximising the coherent bandwidth.
Quantum dot (QD) gain material is typically considered promising for the application in MLLDs due to its broad gain spectrum. Sub-picosecond pulse generation down to 0.4 ps having a bandwidth of 14 nm has been achieved with InAs-GaAs QD material operating at wavelengths around 1.3 µm [9]. These lasers are shown to have robust operating regimes [10]. One should note that the advantageous behaviour of the QD devices turns out to be related to the specific properties of the dynamics of the carriers in the material, their effect on the refractive index and the optical properties of the dots. Linear quantum dot modelocked lasers At the COBRA research institute we have fabricated QD material suitable for operating in the 1.5µm wavelength region. The QD laser structure is grown on n-type InP (100) substrates by metal-organic vapour-phase epitaxy (MOVPE), as presented in [4]. In the active region five InAs QD layers are stacked. These are placed in the centre of a 500 nm InGaAsP optical waveguiding core layer. The bottom cladding of this laser structure is a 500-nm thick n-InP buffer and the top cladding is a 1.5µm p-InP with a compositionally graded 300-nm pInGaAs(P) top contact layer. This layerstack is compatible with a butt-joint active-passive integration process for possible further integration. This integration has been demonstrated recently with the fabrication of integrated extended cavity Fabry-Perot lasers [11]. For our first work on modelocked QD lasers, two-section FP-type laser devices have been designed and realized. The ridge waveguides have a width of 2 µm and are etched 100 nm into the InGaAsP waveguiding layer. To create electrical isolation between the two sections, the most highly doped part of the p-cladding layer is etched away. The waveguide and isolation sections are etched using an optimized CH4 / H2 two-step reactive-ion dry etch process. The structures are planarised using polyimide. Two evaporated and plated metal pads contact the two sections to create two contacts. The backside of the n-InP substrate is metallised to create a common ground contact for the two sections. The processing is fully compatible with our active-passive integration scheme. The structures are then cleaved to create the mirrors for the FP cavity. No coating is applied. The two-section devices are operated by forward biasing the longer gain section, creating a semiconductor optical amplifier (SOA) and by reversely biasing the shorter gain section, creating a saturable absorber (SA). The devices are mounted on a copper chuck, p-side up, which is kept at a temperature of 10 ˚C. The total length of the devices is 9 mm. Here we present results from one of the two-section devices with an SA length of 270 µm. A 9-mm one-section device (i.e. an SOA with cleaved mirrors), fabricated on the same wafer, is used for reference purposes. To demonstrate the wide range of the current and absorber voltage for which QD show modelocking we present results from a study of these ranges for passive modelocked operation of 5 GHz monolithic two-section
InAs/InP (100) QD lasers. These devices operate at wavelengths of around 1.53 µm. Some first results have been published in [12] The laser with a 270-µm SA section has lasing threshold current values of 660 mA to 690 mA for SA reverse bias voltages of 0 V to -4 V respectively. Passive modelocking is first studied by recording the electrical power spectrum using a 50-GHz photodiode and a 50-GHz electrical spectrum analyzer. The RF-spectra obtained for this laser show clear peaks at the cavity roundtripfrequency of 4.6 GHz. In Fig. 1 the height of these RFpeaks over the noise floor is given as a function of the operation parameters, i.e. the SA bias voltage and the SOA injection current. A large, robust operating regime with RF-peak heights over 40 dB is found for values of the injection current of 750 mA up to 1.0 A and for values of the SA bias voltage of 0 V down to -3 V. The width of this RF peak is narrow, i.e. 0.57 MHz at -20 dB (for ISOA = 900 mA and VSA = -1 V). Also the position of this RF-peak, which is centred around 4.599 GHz, is stable within 3 MHz for the operating regime mentioned above. In MLLDs based on bulk gain material minimum RF-linewidths of 2.5 MHz at -20 dB have been reported, with a stability of the roundtrip frequency of about 50 MHz over their operating regime [5]. So a clear improvement of the laser stability is observed by using QD gain material instead of bulk gain material. For comparison we studied the 9-mm one-section QDlaser. The threshold current of this device is 380 mA. The electrical spectrum shows no distinct peak at the roundtrip frequency. This may seem an obvious observation since for the mechanism of passive modelocking in laser diodes with bulk or quantum-well gain material the SA is crucial [8, 5, 7]. However one-section quantum-dash lasers without an SA and emitting at 1.56 µm have been reported to show reliable passive modelocking [13]. A typical optical spectrum of a 9-mm two-section laser
Fig. 1 RF-peak heights (over the noise floor) for a 9-mm laser with a 270-µm SA (left) and a 540-µm SA (right). The electrical bandwidth used to obtain the RF-spectra is 50 kHz. An external SOA was used before the 50-GHz photodiode to boost the optical output power of the 270-µm laser by about 6 dB – 7 dB.
is given in Fig. 2. The spectrum is wider (6 nm – 7 nm) than that from MLLD with quantum well or bulk gain material. As explained in [12], this output spectrum is coherent and one might hope to be able to compress this output well into the femtosecond range. The output pulses of these lasers however are well over 100 ps long, and heavily up-chirped, with a chirp value of approximately 20 ps/nm. This was shown by compressing the pulses to a minimum pulse width using standard single-mode optical fiber (SMF, second order dispersion of approximately 16 – 20 ps/(nm·km)).
Intensity (linear)
QD MLL output spectrum
1526
1528
1530
1532
1534
1536
1538
Wavelength (nm)
Fig. 2. Optical spectrum obtained for the 270 µm SA device. Injection current is 950mA and SA bias voltage is -0V. The optical bandwidth used to obtain the spectrum is 0.16 pm.
In Fig. 3 the autocorrelator traces are shown for different lengths of SMF after the 270 µm SA laser output. As can be seen the pulse is compressed to a minimum duration with 1500 m of SMF. The compression appears however to be partial which can either be caused by a non-linear part of the chirp (or higher order SMF-dispersion), or by the non-Gaussian shape of the optical spectrum (Fig. 2). No autocorrelator traces could be obtained without SMF. Recorded jitter values measurements on the laser in the 10 kHz - 80 MHz range give values over 35 ps. These values are ten times higher than those obtained from bulk devices [5]. To reduce this jitter we have planned to investigate hybrid modelocking using an electronic RF signal.
SH power (a.u.)
800 m 1200 m 1500 m
0
50
100
Fast arbitrary pulse shaping To be able to compress a pulse train as the one obtained from the QD-MLLD device one needs to address the compression more sophisticated than with just a single mode fibre. Although the modes in the laser spectrum are locked, they are not locked to a zero phase difference, but to a non-zero value which is different for each mode. Thus for each mode one would require a phase and amplitude correction to create an ‘ideal’ short pulse. Also the fibre solution is obviously not monolithically integrated. What is thus required is to build a pulse shaper. This can consist of an arrayed waveguide grating (AWG) to split the output spectrum from the MLLD over a number of waveguides. In each of the waveguides the phase may be controlled through an electro-optic modulator and the amplitude through an amplifier. Then the signals are combined again using an identical AWG. Such a device can also be used for dispersion (pre-) compensation in ultrafast time domain multiplexing systems and arbitrary waveform generation [14] and can be applied to e.g. create rectangular pulses to define switching windows, or in optical code-division multipleaccess (O-CDMA) [15]. Other important applications can be found in bio-imaging, using second harmonic pulses for multiple photon excitation [16]. These integrated pulse shapers have been fabricated and studied in our facilities and by others, e.g. [17]. Our work on pulse shapers has been recently reported in [18] and we just present the main result here. It concerns a 4 THz bandwidth InP-based integrated pulse shaper, consisting of an AWG-pair and electro-optical phase modulators and is schematically depicted in Fig. 4. The transmission spectra of the two AWGs should overlap, to optimize the transmission. Separate temperature controls for both AWGs allow for individual tuning of these transmission spectra. The array of phase modulators (PHMs) is integrated to modulate the phase of these spectral components and apply an arbitrary dispersive profile. Special attention in this device is paid to the suppression of ringing of the signal due to the non-flat AWG channel transmission [18]. The device fabrication is fully compatible with the fabrication of the modelocked lasers. The AWG pair is designed to have 20 channels with a 200 GHz spacing and a centre frequency around 1550 nm. The free spectral range is 4 THz, i.e. equal to 20 × 200 GHz. To minimize the appearance of satellite
150
Time (ps)
Fig. 3. Autocorrelator traces obtained with a 270-µm SA device. Injection current is 900 mA and SA bias voltage is 1V. The length of SMF used for compression is indicated.
Fig. 4. Schematic overview of the realised AWG-based pulse shaper configuration with both Gaussian and flat-top transmission in/outputs.
pulses at 5 ps time spacing, due to the non-flat AWG channel transmission, we have added waveguides with MMI couplers for flat-top transmission. Pulse shaping was investigated for pulses at 80 MHz and with a duration of 0.3 ps. Our flat-top AWG inputs strongly reduce the pulse ringing as compared to AWGs with Gaussian inputs by from 50% - 60% down to 12% 16% and are well able to reconstruct the pulses to duration of 0.3 ps. For the application of the device as a dispersion compensator it was shown that it is well suited to compensate up to 0.2 ps/nm dispersion and a pulse peak power reconstruction of over 95%. In Fig.5 measurements of the phase control and transmission spectrum of the device are presented. The reason that in this section we are talking about fast pulse shaping is the fact that the integrated electro-optic phase modulators have a bandwidth that is easily 100MHz and can be optimised further to have bandwidths of even tens of GHz if a travelling wave design is chosen. The modulation bandwidth of the phase modulators can in principle be as large as the frequency spacing between the channels. But even at the lower frequencies one can e.g. in multi-photon microscopy optimise the pulse profile of a laser system to compensate for dispersion in the sample on the fly at video-rates. This should be compared with the capabilities of more traditional pulse shapers often used with Ti:S laser systems that consist of bulk gratings and LCD spatial light modulators where pulse shape control is much slower.
Fig. 5. Measured phase of the spectral components of the transmitted pulse spectrum using Gaussian AWG channels. The PHMs in the channels are separately biased at -2 V (dark grey) and -4 V (light grey) and compared to the unbiased case (black). The transmitted spectrum is also shown.
Conclusions and future direction Both the QD-MLL and the pulse shapers show very promising results that show a way towards fully integrated femtosecond laser systems with unique fast pulse shaping capabilities. It is shown that using activepassive integration technology one can realise such systems. However we are still a number of issues to be tackled. For the QD-MLL improvement of the jitter performance through hybrid modelocking is required as well as research into a deeper understanding of the observed modelocking. A serious limitation is the amount of power that can be transported through the standard ridge waveguide in the
semiconductor material (10 to 20mW average) because of two-photon absorption and the creation of hot carriers by the light in the passive waveguides [19]. This may become an issue at the output of a pulse shaper that contains a series of optical amplifiers. Also for a most accurate system the AWG channel spacing should be equal to the MLL mode spacing so that every mode can be controlled. This is demonstrated in InP technology [20] but the AWGs do become quite large when the repetition rate of the laser is less the say 50GHz. One solution may be to investigate hybrid integration technologies where InP active components are combined with silicon type of passive optical components. There the spectral filtering can be performed in high resolution and the power handling in such materials is considerably higher. Acknowledgements This work was supported by the Netherlands Foundation of Scientific Research (NWO) through the NRC Photonics Grant, by the Smartmix program MEMPHIS the Dutch Ministry of Economic Affairs, the European Network of Excellence ePIXnet , and COST action 288. References 1. R. Holzwarth et al., Phys. Rev. Lett. 85 (2000) p.2264 2. J.L. Herek, J. Photochem. Photiobio. A. 180 (2006) p. 225 3. E. Bente et al, Photonics West, Optoelectronic Integrated Circuits VIII, Proc. of SPIE Vol. 6124, (2006) 612419 4. R. Nötzel et al., Jap. J. of Appl. Phys. 45 (2006) p. 6544 5. Y. Barbarin et al., Optics Express 14 (2006) p. 9716 6. K.A. Williams et al., New J. Phys. 6 (2004) p. 179 7. E. Bente et al., Opt. Quant Electron (2008) DOI 10.1007/s11082-008-9184-y 8. R. Kaiser et al., IEEE J. of Sel. Top. Q.E. 13 (2007) p. 125 9. E.U. Rafailov et al., Appl. Phys. Lett. 87 (2005) 081107 10. J.-P. Tourrenc et al., 15th International Conference on Ultrafast Phenomena, paper WC9, 2006 11. H.Wang et al., Electr. Lett. 44 (2008) 10th April, No. 8 12. M.J.R. Heck et al., Optics Express 15 (2007) p. 16292 13. C. Gosset et al., Appl. Phys. Lett. 88 (2006) 241105 14. K. Takiguchi et al., Electron. Lett. 40 (2004) p. 537 15. Chen Ji et al., IEEE J. Sel. Top. Q. E. 11 (2005) p. 66 16. H.C. Guo et al., 15th International Conference on Ultrafast Phenomena, paper TuE8, August 1. 2006 17. R. P. Scott et al., Optics Express 15 (2007) p. 9977 18. Heck et al., IEEE J. of Q.E. 44 (2008) p. 370 19. van Thourhout et al., Phot. Tech. Lett. 13 (2001) p. 457 20. N. K. Fontaine et al., OFC 2008 paper OTuC7