November 1, 2006 / Vol. 31, No. 21 / OPTICS LETTERS 3107

Simultaneous achievement of narrow pulse widthand low pulse-to-pulse timing jitter in 1.3 �m

passively mode-locked quantum-dot lasers

Maria Teresa Todaro,* Jean-Philippe Tourrenc, Stephen P. Hegarty, Carmel Kelleher, Brian Corbett,Guillaume Huyet, and John G. McInerney

Physics Department/Tyndall Institute, Lee Maltings, National University of Ireland, University College, Cork, Ireland

Received June 12, 2006; revised August 10, 2006; accepted August 28, 2006;posted August 30, 2006 (Doc. ID 71906); published October 11, 2006

We have analyzed pulse width and timing jitter in passively mode-locked two-section InAs quantum-dot la-sers emitting at 1310 nm and have identified two distinct, extensive mode-locked regions with robust shortpulses and low timing jitter. A record combination of 2 ps pulses and 25 fs/cycle timing jitter (500 fs,1–100 MHz), with 1 mW average output power per facet, is demonstrated. © 2006 Optical Society ofAmerica

OCIS codes: 160.6000, 140.4050.

There is increasing interest in the development ofmonolithic mode-locked lasers (MLLs) for the genera-tion of coherent, stable, high-repetition-rate picosec-ond and subpicosecond optical pulses. Because oftheir compactness, mechanical stability, low powerconsumption, low cost, and possibility of integrationwith high-speed electronics, they are ideal sources inoptical communication, in data storage and samplingsystems, for imaging in complex scattering media, forproducing THz radiation, and for the generation andextraction of clock signals. For such applicationsgood-quality locking evidenced by low timing jitter isrequired, as fluctuations in the arrival time of thepulses degrade the bit error rate and the time reso-lution. Among the different MLL configurations, themonolithic two-section passively MLL is the most ef-fective design, as it allows high volume productionand low cost and simplicity of fabrication and oper-ates independently of an electrical modulationsource. However, passively mode-locked semiconduc-tor lasers have been reported to suffer from highertiming jitter than do actively mode-locked devices1

owing to the absence of an active retiming source.MLLs based on quantum dots (QDs) are very at-

tractive ultrashort pulse sources because of the inho-mogeneous spectral broadening offered by the disper-sion in QD size. In addition, QD MLLs are expectedto exhibit lower timing jitter and more robust operat-ing regimes compared with quantum well devices asa result of the reduced amplified spontaneous emis-sion in three-dimensional quantum confinement2 anda lower linewidth enhancement factor with respect toless confined structures.3

Ultrashort high-power pulses4,5 at a high repeti-tion rate6 from QD passively MLLs have been al-ready reported, even though the jitter characteristicsof such devices have received relatively littleattention.7 Only recently ultralow jitter of an activelymodulated mode-locked QD laser inserted into a ringconfiguration has been demonstrated.8 In any case,no simultaneous achievement of both narrow pulse

width and timing jitter has ever been reported.

0146-9592/06/213107-3/$15.00 ©

In this Letter we report the characterization ofpulse width and timing jitter properties of 1.3 �mpassively mode-locked QD two-section lasers in awide range of operating conditions. We experimen-tally demonstrate the generation of robust and stablepicosecond pulse trains together with very low timingjitter in two distinct spectral regions. We also showthe best combination reported to date (to our knowl-edge) of ultrashort pulses, as short as 2 ps, with anultralow timing jitter of 25 fs/cycle in such devices.

The molecular-beam-epitaxy-grown QD wafer wassupplied by NL Nanosemiconductor for room-temperature lasing at 1.3 �m. The active region con-sisted of 10 layers of InAs QDs separated by 33 nm ofGaAs and bounded by Al0.35Ga0.65As cladding layers.3 �m wide ridge waveguide lasers were fabricated bystandard optical lithography and dry etching. Two-section lasers were formed by wet chemical etching ofa 20 �m wide gap in the heavily doped top contactlayer, to provide an electrical isolation �1 k� be-tween the absorber and the gain sections. Thelengths of the gain and absorber sections were Lg=4.5 mm and La=0.5 mm, respectively, to provide arepetition rate of about 8 GHz. No coatings were ap-plied to the cleaved facets. Devices were mounted pside up on a copper heat sink and were measured at20°C under thermoelectric cooler control.

Mapping of devices and frequency domain jittercharacterization were performed by using a 45 GHzphotodiode, followed by a low-noise 40 GHz amplifierand a 50 GHz electrical spectrum analyzer. Charac-terization of pulse widths was carried out by usingstandard optical second-harmonic generation auto-correlation techniques.

Figure 1 maps laser behavior versus gain and ab-sorber section bias voltage. The absorber reversevoltage and gain current were scanned between 1and 4 V and between 150 and 220 mA, respectively.The threshold current increased with increasing re-verse voltage, owing to higher loss in the absorber re-

gion. Above lasing threshold, a small region of un-

2006 Optical Society of America

3108 OPTICS LETTERS / Vol. 31, No. 21 / November 1, 2006

stable mode locking was observed. At higher currentsseveral distinct mode-locked regions were observed,with a difference of approximately 10 MHz in theirrepetition rates. Two wide-current-range regionswere observed, denoted regions 1 and 2 in Fig. 1,each being associated with a distinct group of longi-tudinal modes. Between these two regions we ob-served a transition region where switching occurredbetween the two mode groups, at approximatelymegahertz rates. The group separation was 2–3 nm,similar to the mode grouping phenomenon alreadyobserved in cw QD lasers9 and attributed to leakageof the optical field through the substrate, modulatingthe gain spectrum.10 The low aluminum content ofthe cladding layers, and hence the small index stepbetween the cladding and the waveguide, would sup-port this hypothesis. For a given reverse bias, chang-ing the gain current would change the gain spectrumand temperature, initiating mode group switching.

The mode-locked regions of most significant extent,regions 1 and 2, were studied in terms of pulse widthand root mean square (RMS) timing jitter. The blacksymbols correspond to operative conditions for mea-surements of pulse width and timing jitter reportedbelow. The autocorrelation traces were deconvolvedby assuming a sech2 profile. Figure 2 shows pulsewidths at different bias conditions in regions 1 and 2.In each region, pulse width increased with increasingcurrent at a given reverse voltage, and the minimumpulse width decreased with the reverse voltage. Theformer trend is attributed to increased self-phasemodulation, while the latter is attributed to reduc-tion of the absorption recovery time. The lowest mea-sured pulse width, shown in the inset of Fig. 2, was2 ps at 190 mA gain section current and −3.6 V ab-sorber bias. The traces being recorded thanks to abackground-free intensity autocorrelator, the offseton the displayed trace is induced by the really tightalignment requirements of the autocorrelator. To de-termine the RMS timing jitter, we evaluated the

Fig. 1. Observed mode-locking regions versus absorber re-verse voltage and gain current. The black symbols corre-spond to operative conditions for measurements of pulsewidths and RMS timing jitter.

single-sideband phase noise (SSB-PN) by analysis of

the fourth-harmonic electrical spectrum of the de-tected signal, to minimize any amplitude jittereffect.11 Figure 3 shows the wide-span rf power spec-trum of the two-section passively mode-locked QD la-ser under bias conditions of 190 mA and −3.6 V. Thisplot confirms passive mode-locking with a repetitionfrequency of 8.25 GHz as defined by the cavitylength. In the inset we include the SSB-PN spectraldensity L�f� extracted from the fourth harmonic inthe 10 kHz to 100 MHz range.

The RMS timing jitter was calculated by using thefollowing expression:

� =1

2�nfR�2�

fmin

fmax

L�f�df, �1�

where n is the number of the harmonic, fR is the rep-etition frequency, and fmin and fmax determine the off-set frequency range over which the spectral densityL�f� is integrated. Figure 4 shows RMS timing jitterintegrated in the 1–100 MHz range in both signifi-

Fig. 2. Pulse width versus gain current at different biasconditions in the mode-locking regions (1 and 2). Inset,overlapping of autocorrelation trace and sech2 profile fit ofthe minimum pulse width under 190 mA current and−3.6 V bias.

Fig. 3. Rf power spectrum from a QD passively MLL un-der 190 mA current and −3.6 V bias. Inset, SSB PN spec-

tral density of the fourth harmonic.

November 1, 2006 / Vol. 31, No. 21 / OPTICS LETTERS 3109

cant mode-locked regions (black symbols). We can seefor both ML regions that, for a given reverse voltage,there exists an optimum current that minimizes tim-ing jitter. Indeed, close to the common border of eachregion, switching between mode-locking regimes cre-ates strong undesirable amplitude modulation of thepulses, which does not allows a correct timing jitterdetection, inducing an overestimation even at thefourth harmonic. Opposite to the pulse width trend,for each region the minimum jitter tends to increasewith reverse bias, the minimum in region 1 alwaysbeing smaller than that in region 2. The lowest RMStiming jitter, measured in region 1 at low reversebias, was 400 fs. As the timing jitter of passivelyMLLs is cumulative,12 it is more appropriate to de-fine a jitter per pulse cycle for the devices understudy as

��pp = �2D2TR, �2�

where TR is the repetition rate and D is a diffusionconstant calculated as reported in Ref. 13, thus prov-ing a pulse-to-pulse jitter as low as 20 fs/cycle. Sig-nificantly, the simultaneous achievement of narrowpulse width and low jitter is relatively robust in theselasers, especially in region 2: good mode locking oc-curred over broad ranges of current and reverse volt-age (�30 mA, 2 V), where the pulse-to-pulse timingjitter ranges from 37 to 110 fs/cycle in the exploredbias range. This surpasses typical quantum-well de-vices, which should allow QD MLLs to be manufac-turable, stable, controllable, and reliable in real ap-plications, including those where low cost and highrobustness are paramount.

In summary, we have simultaneously achievedstable, very narrow pulse widths and very low timingjitter in passively mode-locked two-section quantum-

Fig. 4. RMS timing jitter versus gain current at differentbias conditions in the mode-locking regions (1 and 2).

dot (QD) lasers under different operating conditions.We found two distinct, extensive mode-locked regionswith different characteristics. Both regions haveshown narrow pulse widths ��2–3 ps�, and the firstregion exhibited timing jitter as low as 20 fs/cycle.For reverse bias of 4 V, the combination of a 2 pspulse width with 25 fs/cycle timing jitter (500 fs,1–100 MHz) is the best combination of pulse widthand jitter reported in mode-locked QD lasers to date.

This study was supported by the Science Founda-tion Ireland under the Research Frontier, PrincipalInvestigator and National Access Programs, the IrishHigher Education Authority under the PRTLI pro-gram, and the European Space Agency.

*Permanent address, National NanotechnologyLaboratory of CNR-INFM, Università di Lecce, Dis-tretto Tecnologico-ISUFI, Via Arnesano, 73100 Lecce,Italy; e-mail, mariateresa.todaro@unile.it.References

1. D. J. Derickson, R. J. Helkey, A. Mar, J. R. Karin, J. G.Wasserbauer, and J. E. Bowers, IEEE J. QuantumElectron. 28, 2186 (2002).

2. T. W. Berg and J. Mork, IEEE J. Quantum Electron.40, 1527 (2004).

3. J. Muszalski, J. Houlihan, J. Huyet, and B. Corbett,Electron. Lett. 40, 428 (2004).

4. E. U. Rafailov, M. A. Cataluna, W. Sibbett, N. D.II’inskaya, Yu. M. Zadiranov, A. E. Zhukov, V. M.Ustinov, D. A. Livshits, A. R. Kovsh, and N. N.Ledentsov, Appl. Phys. Lett. 87, 081107 (2005).

5. M. G. Thompson, A. Rae, R. L. Sellin, C. Marinelli, R.V. Penty, I. H. White, A. R. Kovsh, S. S. Mikhrin, D. A.Livshits, and I. L. Krestnikov, Appl. Phys. Lett. 88,133119 (2006).

6. M. Kuntz, G. Fiol, M. Lämmlin, D. Bimberg, M. G.Thompson, K. T. Tan, C. Marinelli, R. V. Penty, I. H.White, V. M. Ustinov, A. E. Zhukov, Yu. M.Shernyakov, and A. R. Kovsh, Appl. Phys. Lett. 85, 843(2004).

7. M. G. Thompson, K. T. Tan, C. Marinelli, K. A.Williams, R. V. Penty, I. H. White, M. Kuntz, D.Ouyang, D. Bimberg, V. M. Ustinov, A. E. Zhukov, A.R. Kovsh, N. N. Ledentsov, D.-J. Kang, and M. G.Blamire, Electron. Lett. 40, 346 (2004).

8. M. T. Choi, J. M. Kim, W. Lee, and P. J. Delfyett, Appl.Phys. Lett. 88, 131106 (2006).

9. D. Bimberg and N. Ledentsov, J. Phys. Condens.Matter 15, R1063 (2003).

10. E. P. O’Reilly, A. I. Onischenko, E. A. Avrutin, D.Bhattacharyya, and J. H. Marsh, Electron. Lett. 34,2035 (1998).

11. D. Von der Linde, Appl. Phys. B: Photophys. LaserChem. 39, 201 (1986).

12. D. Eliyahu, R. A. Salvatore, and A. Yariv, J. Opt. Soc.Am. B 14, 167 (1997).

13. L. A. Jiang, M. E. Grein, H. A. Haus, and E. P. Ippen,IEEE J. Sel. Top. Quantum Electron. 7, 159 (2001).