THURSDAY AFTERNOON / CLEO'99 / 4 41 output CThL2 Fig. 1. Schematics of the erbium doped fiber laser. Notatio ns: HR = highreflector, S A = saturable absorber, TPA = two-photon absorption material (placed either on SA or on opposite HR). Ref. = reference beam used to measure the reflectio n fro m SA, BP = beam split- ter. EDF = erbium doped fiber, WD M = wave- length division multiplexer, FR = Faraday rota- tor, WP = wave plate. -2W -l W 0 Record Time (Microseconds) CThL 2 Fig. 2. Illustration of the start -up pro- cess of the erbium doped fiber. The signal strength is normalized to that of CW mode- locked signal. Inset: Illustrative trace of a typical Q-switched envelope with mode-locked pulses. minimize the amount of two-photon absorp- tion. We report here a new mode-locking mechanism, where the use of a two-photon absorber in addition to a saturable absorber allows for CW mode-locki ng o reliably ev olve from Q-switched mode locking. An optimal design of the two-photon absorber limits the peak power of the Q- switched pulses v ia opti - ca l limiting and thus aids in preventing dam- age to the saturable absorber. Moreover, the threshold for optical limiting is chosen to set an optimum stability point for CW mode- locked operation. The environmentally stable fiber lase r3 con- figuration is sketched in Fig. 1. When mode- locked, the laser provides -350 fs pulses at a wavelength of -1.56 pm a nd -5 0 MHz rep- etition rate. Pumped with a single-mode lase r diode operating at 980 nm, the output power can be var ied betwe en 3-15 mW depending on the pump level and on the output coupling, which may be adjusted with the ori entation of the waveplate. The saturable absorbers used for modelocking are InGaAsP epilayers with an insertio n loss that can vary from 20 to 80%. The carrier lifetime i s -5 ps. The start-up process of the fiber laser was recorded using a gigahertz detector and oscil- loscope. As shown in Fig. 2, the fiber laser first generates a series of Q-switched, mode-locked spikes. This behavior persist s for several milli- seconds. The laser makes a transition to CW mode-lock ing after a couple of relatively small, but broader Q-switched, mode-locked spikes (Fig. 2). The Q-switched envelopes have a width of approx. 1 ps, and contain a number of short mode-locked pulses a s shown in the inset TPALoss " I L 4 OUOOl 0.01 0.1 I ____-- - -_ I lo loo Intensity (GW/C~') CThL2 Fig. 3. Intensity dependen ce o f the nonlinear losses due to satur able absorber, two- photon absorption and the sum of the two com- ponents. of Fig. 2. Using an intensity autocorrelation, the pulse width of the Q-switched, mode- locked pulses was measured to be less than 3 ps. This data indicates that the initial pulse shortening process i s completed in very few round trips. Af ter the transition to CW mode- locking, the pulses then shorten further to 300 fs, requiring several psec to d o so. During modelocked operati on, the absorber is driven heavily into saturation. An accurate theoreti cal description for the above phenomenon is outside the range of standard laser stability mode ls based on per- turb ation theo ry becau se of the presence o f the large, heavily saturated gain in the cavity (20 dB). B y mea suring th e rele vant laser pa- rameters, conventional pe rturbation theories 1.4 predict the laser to be well into the Q-switched, mode-locked region. This predic- tion qualitatively agrees with the observed Q-switched, mode-locked pulses. A descrip- tion for the transition from Q-switched, mode-loc king to CW mode-lockin g, however, remains to be provided. In the present laser, the transition from Q-switched mode-locking to CW mode- locking is facilitated by the incorporation of a bulk InP two-photon absorber5 nto the cavity . Two-photon absorption in the InP provides an optical limiting mechanism, which suppresses the Q-switched, mo de-locked pulse intensities and stretches the Q-switch pulse envelopes to eventually achieve CW mode-l ocking. The InP two-photon absorber is designed to have a small insertion loss (-5%) at CW mode- locked power leve ls but to have signif icant ab- sorption (>50%) at the power levels of the Q-switched, mode-locked pulses. The two- photon absorber material can be placed on the saturable absorber or at the opposite end of the cavity. With the saturable absorber/two-photon absorber combinatio n, the total nonlinear loss can be self-adjusted to provide an optimum stability egion for CW mode-locking. (Fig. 3). As experimentall y verified , the in sertion of a two-photon absorber has no measurable det- rimental effec ts on the attainable output power and achievable pulse widths from the fiber la- ser cavity. Rather, the power limitation from the saturable absorber allo ws tighter focusing on the saturable absorber without incurring optical damage, allowing the exploitation of largernonlinear loss needed for the most rapid self-starting. In conclusion we present here a mode- locked laser which sel f starts by a means diff er- theories do not predict self-starting through Q-switch mode-l ocking. Differe nces with per- turbative theories are understandable . These predicted conditions for self-starting of CW mode-locking use static numbers, while the key parameters for stable modelocking are ,varying dynamically during the Q-switched mode-locking of this laser. It is very conceiv- able that these numbers vary appropriately so that CW modelocking c an initiate. More spe- cifically, he laser ma kes a transiti on fro m fast saturable absorber modelocking during Q-switching to slow saturabl e absorber mo d- elocking during CW mode-locked operation. A more complete dynamical analysis is re- quired to understand these transitions. 1. H.A. Haus et al., J. Quantum Electronics, 2. A.T. Obeidat, W.H. Knoxand J.B. Khur- gin, Conf. on Lasers and Electro-Optics, CLEO, Baltimor e, paper CtuP28 (1997). M. E. Fer mann , L.-M. Yang, M. L. Sto ck and M.J. Andrejco, Opt. Lett., 19, 43 (1994). F. Kartner, L. Brovelli, D. Kopf, M. Kamp, I. Calasso, U. Keller, Opt. Eng., 34, 2024 (1995). A. Agnesi et al., Opt. Lett., 18,637 (1993). QE-12,169, (1976). 3. 4. 5. CThW 3:OO pm Saturable absorber modelocking uslng nonepitaxially grown serniconductor- doped films I.P. Bilinsky, R.P. Prasankumar, J.N. Walpole,* L.J. Missaggia,, J.G. Fujimoto, Department o f Physics, Departm ent o f Electrical Engineerin g and Com pute r Science and Research Laboratory o f Electronics, Massachusetts Institute of Technology, Cambridge, Masachusetts 02139 US A Semiconductor saturable absorbers are useful for many applications in ultrafast optics, in- cluding the generation of femtosecond pulses in solidstate The most common satu- rableabsorbertechnologies aresemiconductor saturable absorber mirrors3 and saturable Bragg reflecto rs,4 which have bee n extensive ly used for both saturable absorber modelocking and initiation of Kerr lens modelocking (KLM). However, these devices require epi- taxial growth, which imposes lattice-match ing constraints on the absorber materials and also requires co mplex and expensive s ystems. We demonstrate novel, non-epitaxially grown semiconductor saturable absorber de- vices for laser modelocki ng. These devices con- sist of semiconductor nano crystallit es doped into silica films using R F ~puttering.~ hese films can be deposited on virtually any sub- strate, including oxides such as glass and di- electric coatings as well as metal mirrors. By varying he doping density of the semiconduc- tor q uantum dots, one can adjust the absorp- tion coefficient of the device. A wide range of semiconductor materials can be doped in to the