Frequency-resolved optical gating characterisation of passively modelocked quantum-dot laser Y.-C. Xin, D.J. Kane and L.F. Lester A unique, ultra-sensitive frequency-resolved optical gating system was constructed, and the pulse shape and chirp of a quantum-dot mode- locked laser was unambiguously measured using this technique for the first time. A clear pulse asymmetry was detected, and evidence that the pulses are recompressible to sub-picosecond lengths is presented. Introduction: Interest in quantum-dot modelocked lasers (QDMLLs) has grown substantially in recent years since their first demonstration in 2001 [1], as applications for optical-time-domain multiplexing, arbi- trary waveform generation, and optical clocking are anticipated [2, 3]. Ultrafast pulses from QDMLLs have been reported as short as 390 fs using intensity autocorrelation techniques [4], but so far detailed charac- terisation examining the pulse shape, duration, chirp and degree of coherence spiking in these lasers has not been carried out. This Letter describes the first direct frequency-resolved optical gating (FROG) measurements on a QDMLL, which operates at a nominal repetition rate of 5 GHz. Even though FROG devices are available commercially, they normally lack the sensitivity required to measure the output pulses from modelocked semiconductor lasers. A unique, ultra-sensitive FROG system was assembled to realise the results presented here. With the FROG system, the pulsewidth and chirp are measured accurately, clear pulse asymmetry is observed, and evidence that pulse break-up in the QDMLLs is caused by interplay between the homogeneous linewidth of the individual quantum dots and the inhomogeneous broadening of the ensemble is described. Device and experimental setup: The two-section passive QDMLL devices were processed using an optimised six-stack dots-in-a-well (DWELL) laser structure, which was grown by elemental source molecular beam epitaxy (MBE) on a (001) GaAs substrate [5], following standard ridge waveguide laser processing [6]. The passive QDMLLs have a 1.0 mm absorber and 7.3 mm gain section. The cleaved facet near the absorber was high-reflectivity coated (R ’ 95%) and the other facet was low-reflectivity coated (R ’ 5%). The two-section laser was mounted on an AlN substrate and then on a copper heatsink. The temperature was kept at 208C with a TEC controller. The optical output of the laser was collected with an optical head, which integrates a lens, an isolator and a short 1 m single-mode polarisation-maintaining (PM) fibre pigtail, and then was coupled into the FROG system through the PM fibre. The FROG system includes a Femtochrome autocorrelator and a scanning monochromator, as shown in Fig. 1. The autocorrelator generates a background-free second harmonic generation (SHG) signal using a 1 mm-thick LiIO 3 crystal. The SHG signal was guided out of the autocorrelator and coupled into the monochromator with a set of mirrors and lenses. The mirror inside the autocorrelator is removable, which permits intensity autocorrelation to be evaluated also. The SHG signal in the monochromator was spectrally gated and detected by a highly sensitive photomultiplier tube (PMT), which is placed at the output slit. The resulting electronic signal from the PMT was amplified with a low-noise current preamplifier and then recorded with a digital oscilloscope. By scanning the grating in the monochromator, SHG FROG traces were obtained with a computer. digital oscilloscope aperture lens lens lens PMT low noise current preamplifier prism polarizer isolator trigger signal multimeter monochromator PMF autocorrelator CW current source second harmonic fundamental beams reference/delayed SHG crystal M2 M1 M3 + × – Fig. 1 Schematic diagram of FROG device Experiment results and discussion: The QDMLLs have a repetition rate of 5 GHz. The pulse shapes obtained from the autocorrelator are shown in Fig. 2a. The data shows symmetric pulses of 7.2–9.0 ps at gain currents of 100– 110 mA and a reverse bias of 24 V, which correspond to 5.0 – 6.3 ps pulses assuming a Gaussian pulse shape. The optical spectrum of the MLL has a 4.3 nm FWHM. The time–bandwidth products are about 4.2—about 12 times higher than their transform limit. Thus, with the intensity autocorrelation, we strongly suspect that the pulse is chirped, but it cannot be determined whether this is due to dispersion or nonlinear effects. time, ps –20 0.14 0.12 0.10 0.08 0.06 0.04 0.02 20 0 time, ps intensity, a.u. –20 20 0 100 mA 105 mA 110 mA 115 mA 100 mA 105 mA 110 mA 1.0 0.8 0.6 0.4 0.2 0 a b Fig. 2 Optical pulse shape of device with different gain currents and reverse bias of 4 V applied on 1 mm absorber a Pulse shapes obtained from intensity autocorrelations b Time-domain intensity profiles of retrieved pulses from series of FROG traces with drive currents from 100 to 115 mA FROG traces were obtained with different gain-currents and reverse biases. With a reverse bias of 3.5–4 V on the saturable absorber, good pulse retrievals were obtained at drive currents of 100–115 mA on the gain section of the QDMLL. Time-domain pulse intensity profiles derived from the pulse retrievals for the FROG traces are shown in Fig. 2b at four different drive currents with an absorber bias of 4 V. As the drive current increases, the length of the pulse stays constant until 115 mA drive. The lower drive current pulses are about 5–6 ps long while the 115 mA drive case is about 7–8 ps long. In all cases, a clear pulse asymmetry is observed. Since there is a time direction ambi- guity in SHG FROG, the FROG measurements alone cannot determine which edge is the leading edge of the pulse. However, the differential absorption is much higher than the differential gain under these operat- ing conditions. Hence the pulse should have a faster leading edge and a slower trailing edge. Fig. 3 shows the time-domain phase of the retrieved pulses. They are relatively consistent from pulse to pulse. The data indicates that the pulses from the QDMLL are mainly linearly-chirped. These measure- ments indicate that operational conditions exist for the QDMLL for which the pulses are well-behaved and completely recompressible. The transform-limited pulsewidth was determined to be about 600 fs by mathematically removing the spectral phase. While the bandwidth would support 300 fs pulses, the spectral structure limits the ultimate pulse duration. time, ps temporal phase, rad –20 0 20 40 60 80 100 120 –15 –10 –5 0 5 10 15 20 4 V 100 mA 4 V 105 mA 4 V 110 mA 4 V 115 mA Fig. 3 Time-domain phase of retrieved pulses shown in Fig. 2b An SHG FROG trace from the monolithic modelocked QD laser with a gain current of 105 mA and a reverse bias of 3.5 V on the absorber is shown in Fig. 4a. (The SHG FROG traces are symmetrised about t ¼ 0 ELECTRONICS LETTERS 9th October 2008 Vol. 44 No. 21