Вы находитесь на странице: 1из 10

Appl Phys A (2008) 93: 1726 DOI 10.

1007/s00339-008-4644-6

Lasing in femtosecond laser written optical waveguides


R. Osellame G. Della Valle N. Chiodo S. Taccheo P. Laporta O. Svelto G. Cerullo

Received: 12 October 2007 / Accepted: 4 March 2008 / Published online: 6 June 2008 Springer-Verlag 2008

Abstract The paper reviews our research efforts in femtosecond laser inscription of optical waveguides inside active Er:Yb-doped glass substrates and in the development of compact waveguide lasers. By optimising the fabrication parameters we obtain waveguides with coupling losses to standard telecom bers as low as 0.1 dB/facet and propagation losses lower than 0.4 dB/cm. The waveguides provide net gain, with a peak value of 7 dB, and laser action in the whole telecommunications C-band (15301565 nm). Single-frequency operation with output power up to 50 mW and passive mode-locking are demonstrated, making the systems useful as compact telecom transmitters. Our results show that femtosecond laser writing can produce highquality photonic devices, with performance comparable to or better than that obtainable with conventional waveguide fabrication techniques. PACS 42.82.Et 42.70.Hj 42.65.Re

around the focus [24]. This localized energy deposition induces, by a variety of mechanisms, a refractive index modication over a micrometer-sized volume of the material and enables to produce, by moving the laser focus inside the glass substrate, optical waveguides or more complex photonic circuits [5, 6]. Femtosecond laser waveguide writing shows, compared with traditional fabrication techniques, a number of advantages: (i) It requires a simpler and less expensive device production equipment, avoiding photolithography and cleanroom facilities. (ii) It enables rapid prototyping, since the device pattern can be easily changed by simple software control, with signicant cost reduction with respect to standard techniques using photolithographic steps and requiring the production of a mask. (iii) It is intrinsically a three-dimensional (3D) technique, since refractive index changes can be induced in any point in the bulk of the material within a given depth (100 m to 1 mm) from the surface; this characteristics can be exploited to implement novel device functionalities, which are impossible with standard fabrication methods. Despite these advantages, until recently femtosecond waveguide writing was considered mainly as a scientic curiosity, because the waveguide parameters (propagation and coupling losses, mode proles) were much worse than those of waveguides obtained with standard fabrication techniques. Recent dramatic progress has led to the demonstration of waveguides with quality comparable to those produced with standard techniques, while simultaneous advances in laser technology have produced much more com-

1 Introduction Femtosecond laser optical waveguide writing is a powerful microfabrication technique emerged in recent years [1]. When an ultrashort pulse is tightly focused in a transparent material, a combination of nonlinear absorption processes allows one to selectively deposit energy in a small volume

R. Osellame ( ) G. Della Valle N. Chiodo S. Taccheo P. Laporta O. Svelto G. Cerullo Istituto di Fotonica e NanotecnologieConsiglio Nazionale delle Ricerche, Dipartimento di FisicaPolitecnico di Milano, Piazza L. da Vinci, 32, 20133 Milano, Italy e-mail: roberto.osellame@polimi.it

18

R. Osellame et al.

pact and reliable writing lasers. The industrial application of femtosecond waveguide writing appears now to be feasible. Waveguide lasers and ampliers in active Er:Yb-doped glass substrates [7] have become increasingly interesting in the past few years for metropolitan and local optical network applications, due to the possibility of low-cost mass production, integration in optical subsystems, and small footprint. For such devices femtosecond laser writing, due to its above mentioned advantages, can compete with standard fabrication techniques, such as silica-on-silicon, plasma-enhanced chemical vapour deposition, ion-exchange, and solgel. This paper summarizes our research efforts in the femtosecond laser fabrication of active waveguide devices, focusing in particular on waveguide lasers. After an introduction to femtosecond waveguide writing (Sect. 2), in Sect. 3 we describe the optimization of the waveguide fabrication process in order to achieve mode-matching with standard single-mode bers (SMFs) and net gain. In Sect. 4 we report on advanced, femtosecond laser fabricated, waveguides lasers, such as single-frequency and mode-locked devices. Finally in Sect. 5 we draw some conclusions and perspectives for future work.

2 Optical waveguide writing with femtosecond lasers 2.1 Physical mechanisms of femtosecond refractive index modication When a femtosecond laser pulse crosses a transparent material, with energy gap Eg greater than the laser photon energy, there is no linear absorption through interband transitions from valence to conduction band. At high intensities, however, absorption can take place through nonlinear phenomena, such as multiphoton, tunnelling and avalanche ionization. Multiphoton ionization involves the simultaneous absorption of m photons of energy h, where m is the minimum integer such that mh > Eg . Tunnelling ionization occurs when the very high electric eld of the laser pulse lowers the Coulomb potential energy barrier and enables an electron to tunnel from the valence to the conduction band. At high intensities, the multiphoton and tunnelling ionization processes compete [24]; typically, in femtosecond laser interaction with dielectrics, multiphoton ionization dominates over tunnelling. In avalanche ionization, an electron, free at the bottom of the conduction band of the material and exposed to an intense light eld, is accelerated and acquires kinetic energy. When its total energy exceeds the conduction band minimum by more than the bandgap energy, it can ionize another electron from the valence band, resulting in two electrons near the conduction band minimum. These electrons can be in turn accelerated by the electric eld, causing an avalanche in which the free electron

density grows exponentially; for sufciently high densities, the dielectric becomes strongly absorbing. If the dielectric is illuminated by a long pulse (with several picoseconds or nanosecond duration), the peak intensity is too low to allow multiphoton or tunnelling ionization, even if the total pulse energy might be rather high. The only possible absorption mechanism is avalanche ionization starting from an initial seed of free electrons in the conduction band, which in an insulator are due to impurities and dislocations within the focal volume of the laser pulse. Since their number is subject to large uctuations, the absorption process is erratic and poorly reproducible. With femtosecond laser pulses, on the other hand, the peak intensities are much higher and multiphoton ionization becomes signicant. When the intensity exceeds a given threshold, some free electrons are generated in the focal volume by multiphoton ionization. These electrons act as a seed, produced in a fully deterministic fashion [4], for the avalanche ionization process. Only femtosecond pulses allow, by the unique combination of multiphoton and avalanche ionization, to depose energy, in a highly controlled and reproducible way, in a small volume inside the bulk of a transparent material. While the process of nonlinear absorption of femtosecond pulses in dielectrics is well assessed, the physical mechanisms by which it can induce permanent refractive index changes in glasses are not yet fully understood. If the absorbed energy is too high, catastrophic material damage occurs, leading to the formation of voidlike structures [8]; for a lower energy, there is a regime in which the material maintains its good optical quality, and there are permanent refractive index changes. Several mechanisms have been proposed to explain such changes, none of which seems to be fully general. A rst possible mechanism is color centre formation [9]: the femtosecond irradiation produces in the material a sufcient number of color centres which, while absorbing in the UV, modify by the KramersKroenig mechanism the refractive index of the material at the wavelengths of interest. An alternative mechanism is thermal: energy deposited by the laser melts the material in the focal volume, and the subsequent rapid resolidication dynamics freezes the glass in the high temperature state leading to density (and therefore refractive index) variations [10, 11]. A third possible mechanism is direct photostructural change induced by the femtosecond laser pulses, i.e., rearrangement of the network of chemical bonds in the glass matrix inducing a density increase [12]. In practical cases, all the mechanisms discussed above play a role in refractive index change, and it is difcult to disentangle their relative contributions. Finally it is worth noting that in crystalline materials, femtosecond laser irradiation generally produces a decrease in refractive index; this can be easily understood by considering that, in a crystal, the atoms are in the closest possible

Lasing in femtosecond laser written optical waveguides

19

arrangement and that any change in the lattice order will lead to a lower density. Optical waveguides can nevertheless be created on the sides of the modied region, where stresses induce a refractive index increase [13, 14]. 2.2 Optical waveguide writing with different femtosecond laser systems Nonlinear absorption in glasses takes place for intensities around 15 1013 W/cm2 which, for a pulse duration of 100 fs, correspond to uences of 15 J/cm2 . Such uences can be reached for pulse energies ranging from a few Js (for mild focusing) to a few tens of nJs (for tight, nearly diffraction-limited focusing) by suitably setting the scan speed and the repetition rate. Two different regimes of femtosecond micromaching can be distinguished, depending on whether the pulse period is longer or shorter than the time required for heat to diffuse away from the focal volume: the low-frequency regime, in which material modication is produced by the individual pulses, and the high-frequency regime, in which cumulative effects take place [15, 16]. Since the heat diffusion time out of the absorption volume in the glass is 1 s, the transition between the two regimes takes place at frequencies around 1 MHz. Two different writing geometries are possible, longitudinal and transverse, in which the sample is translated, respectively, along and perpendicularly to the beam propagation direction. In the longitudinal geometry, the waveguides are intrinsically symmetric, and their transverse size is determined by the focal spot size; however, the waveguide length is limited by the focal length of the focusing objective, and the possibility of writing complex structures is severely limited. The transverse geometry provides a much greater exibility and allows one to write waveguides or photonic circuits of arbitrary length and complexity. It has, however, the disadvantage of producing a strong asymmetry in the waveguide cross section, due to the difference between waveguide size perpendicularly to the beam propagation direction, given by the beam focal diameter, and along the propagation direction, given by the confocal parameter [17]. This problem can be solved in the low-frequency regime by the use of suitable beam shaping techniques (see Sect. 3) [1720] or in the high-frequency regime by exploiting isotropic heat diffusion out of the focal volume [10, 16]. Following the pioneering work by Hiraos group [5, 6], many studies were published on femtosecond laser optical waveguide writing. They included both fundamental investigations, aimed at the study of the optical characteristics of the waveguides as a function of the writing conditions and glass substrate composition, and applied research, aimed at the fabrication of photonic devices. In the rst class we can include the papers studying the structural modications that lead to refractive index variations in fused silica

[9, 12] and those exploring the possibility to write waveguides in different materials, such as doped silica, borosilicate, phosphate, and chalcogenide glasses [2124], and recently linear [13, 14] and nonlinear [2528] crystals. Among the demonstrated device functionalities, apart from straight waveguides, there are splitters [29], directional couplers [30, 31], MachZehnder interferometers [32], and Bragg gratings [33, 34]. While some device architectures are similar to those used with standard planar waveguide fabrication methods, others exploit the unique 3D capabilities of femtosecond waveguide writing, such as a 1-to-3 coupler [35, 36], 3D microring resonator [37], and 3D waveguide arrays [38]. Currently, femtosecond laser waveguide writing has been applied to a wide variety of glasses and to some crystals, such as lithium niobate. The typical refractive index change is of the order of 5 10 103 , allowing waveguides that are nearly perfectly mode-matched to standard telecom bers; the propagation losses can be as low as 0.2 dB/cm and fabrication speeds exceed 10 mm/s. Different types of femtosecond laser systems have been employed for waveguide fabrication. The most commonly used systems, because of their broad availability in many research laboratories, are regeneratively amplied Ti:sapphire lasers, with 1200 kHz repetition rate, a few Js pulse energy, and 50200 fs pulsewidth. These systems work in the low-frequency regime and allow processing a wide variety of materials but suffer from a number of disadvantages: they provide low refractive index changes, typically 1 2 103 ; they allow only low processing speeds around 20 100 m/s; they require complex and expensive setups. It is also possible to write waveguides directly with a Ti:sapphire oscillator if the pulse energy is increased to 20100 nJ by decreasing the repetition rate to 525 MHz by means of a telescope or a multipass cell [10, 16]. In this case waveguides are written in the high-frequency regime, and the writing speeds are dramatically increased up to a few cm/s. However only a limited range of glass substrates turn out to be processable with this system [39]. The best results, in terms of both fabrication speed and waveguide quality, have so far been obtained using ytterbium-based lasers [15, 40]. Examples of such systems are a bulk cavity-dumped oscillator and a ber-based master oscillator/power amplier system. Such systems produce 1 J 300-fs pulses at repetition rate up to 1 MHz at 1030 nm wavelength. Such parameters are ideal for processing a wide range of different materials. In addition, diodepumping makes these systems particularly compact and efcient, which is important for industrial applications of the technique requiring maintenance-free, turn-key operation. All the results reported in Sect. 4 have indeed been obtained with an Yb-based system.

20

R. Osellame et al.

3 Towards net gain and laser action 3.1 Writing with an amplied Ti:sapphire laser The active waveguides were written in phosphate glass bases (QX, Kigre Inc.) doped with Er (1.1 1020 ions/cm3 ) and Yb (1.5 1021 ions/cm3 ). The rst experiments were performed with a regeneratively amplied Ti:sapphire laser, generating 150-fs, 500-J pulses at a 1-kHz repetition rate at 790-nm wavelength. The laser pulse energy used for waveguide writing, controlled by a variable attenuator, ranged from 0.5 to 10 J. The optical setup is schematically reported in Fig. 1. At rst, waveguides were written using the 20 (numerical aperture NA = 0.3) objective and focusing the beam to a radius of w0 = 3 m at the waist; the focus was located about 200 m below the surface of the sample which was moved perpendicularly to the beam propagation direction by a motorized translation stage at speeds of order 20 m/s. A microscope image of the exit face of the waveguide is shown in Fig. 2a; it shows a strongly elliptical cross section extending over 70 m along the depth direction and only 5 m along the lateral direction. This is a well-known problem for transversal writing [17]. In fact, the waveguide cross-section has a width equal to about twice the beam waist w0 , while it has a dimension in depth of 2 the order of the confocal parameter (b = 2w0 /), which is typically much larger. This problem can be overcome [18, 19] by introducing a focusing geometry in which the femtosecond writing beam is astigmatically shaped by changing both the spot sizes in the tangential and sagittal planes and the relative positions of the beam waists. This shaping allows one to modify the interaction volume in such a way that the waveguide cross section can be made circular and with arbitrary size. As shown in Fig. 1, the astigmatic beam shaping can be easily obtained by a cylindrical telescope that decreases the spot size in the lateral direction while keeping the same spot size in the translation direction [18, 19]. An alternative technique recently introduced to astigmatically shape the writing laser beam uses a rectangular slit before the microscope objective [20]. Figure 2b shows a microscope image of the exit face of a waveguide written with astigmatic beam shaping for the optimized relative position z0 of the two beam waists. Using these focusing conditions and pulse energy of 5 J, it was possible to obtain symmetric waveguides with a 18 m diameter. To obtain a smaller waveguide cross section, we used a 50 (NA = 0.6) objective and decreased the beam spot size before the cylindrical telescope; in this way we obtained smaller focused beam spot sizes, thus decreasing the Rayleigh range and the waveguide dimension along the depth direction; suitable adjustment of z0 allowed obtaining a symmetric waveguide with smaller size. Figure 2c shows an image of the exit face of such a waveguide with 8 m diameter.

Fig. 1 Optical setup used for femtosecond laser waveguide writing

Fig. 2 Optical microscope images of the exit face of waveguides written with: a a circularly symmetric beam; b a focused astigmatic beam with a 20 objectives; c a focused astigmatic beam with a 50 objectives

The refractive index changes n in the waveguides were measured by a commercial refractive index prolometer [41]. Typical peak values are n = 2 3 103 . The waveguides turned out to be single-mode at 1.5 m. A mode waist of radius 10 m for the 18 m diameter waveguide (Fig. 3a) and a comparable mode waist for the 8 m diameter waveguide were measured. The mode waist did not scale with the waveguide dimension due to the rather low index change. In fact, for small waveguide diameter, the guided mode quickly approaches the cut-off condition, thus being less conned in the waveguide region. The insertion losses (IL) were measured by coupling the waveguide with two standard telecom SMFs, with index matching uid to minimize Fresnel losses, and by comparing the transmitted power to that measured with the same bers spliced. The measurement was performed at the wavelength of 1600 nm, chosen because it falls outside the absorption band of the Er ions. IL of 55.5 dB for a propa-

Lasing in femtosecond laser written optical waveguides

21

Fig. 3 Experimental intensity near-eld proles along the depth direction (dot-dashed curve) and along the lateral direction (solid line) of the waveguide mode compared to a bre mode (dashed curve). a Waveguide written with the 1-kHz Ti:sapphire laser; b waveguide written with the Yb:KYW laser at 885 kHz

Fig. 5 a Absorption and internal gain as a function of wavelength for a 9-mm-long waveguide written with the 1-kHz Ti:sapphire laser. b Internal gain for a 22-mm long waveguide written with the Yb:KYW laser at 885 kHz; dashed line represents the waveguide insertion losses

Fig. 4 Optical setup used for characterizing the gain in active waveguides

gation length of 25 mm were typically measured. In order to distinguish the contributions arising from coupling losses (CL) and propagation losses (PL), we compared the near elds of the waveguide and ber modes (Fig. 3a). The theoretical maximum coupling efciency was estimated from the overlapping integral of the eld proles in the waveguide (E g ) and ber (E f ), according to the formula =
[ Eg Ef dx dy]2 Eg Eg dx dy Ef Ef dx dy

(1)

giving a value for the CL of 22.4 dB/facet. The PL over 25 mm can be determined, by subtraction, to be 0.8 dB, corresponding to 0.3 dB/cm. The gain properties of the waveguides were characterized in a standard optical amplier conguration (Fig. 4). We used a dual-pumping setup with two GaAlAs laser diodes

at 980 and 976 nm, providing up to 240 mW total pump power. A tunable laser with an in-line variable attenuator generated a seed signal spanning the 1530 to 1565 nm wavelength interval; the signal power was kept constant at 30 dBm. Pump and signal were multiplexed by wavelength division multiplexers (WDMs), and the output SMFs were butt coupled to the waveguides using index-matching uid. Waveguides with different length were tested; a peak enhancement of 3.9 dB at the wavelength of 1534 nm was obtained in a 9 mm long waveguide (Fig. 5a); longer waveguides provided worse results due to the very high Yb concentration that caused a strong absorption of the pump light, leading to population inversion only on pathlengths shorter than 1 cm. The measured absorption of the unpumped waveguide is also shown; the peak absorption was 2 dB. In this case, we were able to demonstrate, for the rst time, internal gain over the whole telecom C-band (1530 1565 nm), with a peak of 1.4 dB, consistent with a 75% of the Er population inverted. This value of internal gain was however not sufcient to overcome the IL of 5 dB. To achieve net gain and thus real amplication, a twofold improvement was needed: on one side, optimisation of substrate doping, allowing the use of longer samples with higher gain; on the other side, and mostly important, reduction of

22

R. Osellame et al.

the CL. The latter was to be accomplished by increasing the refractive index change; this however was not possible by writing waveguides with the amplied laser system. 3.2 Writing with a cavity-dumped Yb-based laser In the following we report on optical waveguide writing experiments [40, 42, 43] using a cavity-dumped Yb:KYW femtosecond laser oscillator [44, 45]. The waveguide writing set-up was the same as that reported in Fig. 1 apart from the fact that the astigmatic beam shaping was not employed. We tried different focusing objectives and obtained the best results in terms of circular symmetry of the waveguide mode with a very high NA objective (100oil-immersion Zeiss Plan-Apochromat, 1.4 NA) [40]. The waveguides were written inside the glass at depth of 170 m from the surface. This is the nominal depth to minimize the aberrations with this kind of objectives. However, the same objective provided uniform results at least for depths in a 100 m range around the nominal one. The same phosphate glass described in the previous section was used but doped with 1.8 1020 Er ions/cm3 and 3.6 1020 Yb ions/cm3 . Such doping levels have been optimized to obtain higher gain per unit length and longer devices. The versatility of this cavity-dumped system drove the research activity towards the denition of an optimized set of writing parameters, namely optimal pulse energy, repetition rate, and writing speed [43]. As an example, Fig. 6 shows the refractive index prole, measured with the refractive index prolometer, for a waveguide written at 885 kHz repetition rate and 270 nJ, with 300 m/s translation speed. The actual spot size of the writing laser beam inside the sample was below 1 m, due to the very tight focusing used. The size of the modied region was indeed much larger ( 20 m), indicating the occurrence of thermal diffusion. However, while the isotropy of thermal diffusion should give almost symmetric proles, here we observed a clearly asymmetric prole consisting of two lobes, one positive and one negative. We believe that this asymmetry should not be ascribed to some misalignment of the focusing objective, since it changed by varying the repetition rate or pulse energy. It seems to be related to the glass characteristics and requires further study to be fully understood. This refractive index prole however was very promising, since the positive part was single peaked with a very high index change n = 8 103 . Typical results for IL in these waveguides were signicantly improved, with values around 1.11.2 dB for 22-mm-long waveguides. Figure 3b shows the near eld of the single-mode guided at 1600 nm by a typical waveguide, together with the mode prole of the coupling bre, along the two transversal axes. Notwithstanding a slight ellipticity

Fig. 6 Refractive index prole, measured with a near-eld refractometer, of a waveguide written with the Yb:KYW laser at 885 kHz

of the waveguide mode, the excellent mode matching between the two structures provided estimated coupling losses as low as 0.1 dB. Thanks to the strongly reduced IL and to the improved doping levels, net gain in femtosecond laser written waveguides was achieved for the rst time. Figure 5b shows the small signal internal gain of the 22-mm long waveguide as a function of wavelength, measured with an accuracy of about 0.5 dB with a pump power level of 490 mW. An internal gain higher than 7 dB and 3 dB was measured at 1535 and 1565 nm, respectively. Due to the very low insertion losses (1.1 dB), this waveguide was able to provide net gain in the whole C-Band, ranging from 6 dB at the peak to 2 dB at the edge. Therefore, it could act as an optical amplier, or it could be used as an active medium in a laser cavity, as shown in the next section.

4 Femtosecond laser written waveguide lasers The net gain observed in active waveguides fabricated by using the Yb:KYW oscillator (see Sect. 3) enabled us to demonstrate, for the rst time, femtosecond laser written advanced waveguide lasers able to provide tunability in the whole C-band, single longitudinal mode and stable modelocking operation. In particular, we report here on two optimized active waveguides: a 20-mm long sample fabricated at 505 kHz repetition rate, 436 nJ energy per pulse; and a 36-mm long one fabricated at 885 kHz repetition rate, 250 nJ energy per pulse (writing speed of 50100 m/s for both). The shortest sample was particularly suitable to develop very compact FabryPerot laser cavities, while the longer one, due to the higher net gain, was employed in the demonstration of a mode-locked laser. After laser inscription, the end-facets of the samples were polished, and the two waveguides were

Lasing in femtosecond laser written optical waveguides

23

fully characterized at 1590 nm (outside the erbium absorption band) in terms of passive guiding properties, demonstrating almost the same features. In particular, from the overlap integral between the experimental intensity proles of the waveguides and ber modes, the theoretical CL to standard SMF was estimated to be about 0.1 dB/facet in the 20-mm long sample and 0.22 dB/facet in the 36-mm long one. Following the method reported in the previous section (see also [46] for details), we measured IL of 1 dB in the 20-mm long sample and 1.9 dB in the 36-mm long one, thus the PL was estimated to be at worst PL = (IL 2CL)/L = 0.4 dB/cm in both samples. 4.1 Tunable and single-frequency lasers Figure 7 shows a schematic of the laser cavity employing the 20-mm-long active waveguide in a linear conguration where the waveguide is butt-coupled on both sides to ber Bragg gratings (FBGs) fabricated in a standard SMF. A broad-band at top FBG with 1-nm bandwidth (FWHM) provided high reectivity (99.8%) at one side, and a 0.25-nm bandwidth FBG was used as output coupler. An indexmatching uid able to support high power density at 980 nm was inserted between waveguide and ber ends. Two berpigtailed InGaAs laser diodes with a bi-propagating pumping scheme supplied up to 510 mW incident pump power (260 mW from pump 1 and 250 mW from pump 2). In order to remove the parasitic interaction between the counter-propagating pumping beams, a single-stage optical isolator was connected to pump 1, and, in addition, a halfwave bre polarization controller was inserted along the bre connecting pump 2 to the waveguide. By rotating the axis of the polarization controller, the polarizations of the counter propagating pumping beams were set orthogonal to each other, thus avoiding any interference which was detrimental for the pump efciency. Output power was measured by an optical spectrum analyzer. The latter conguration, with respect to discrete mirrors or dielectric coatings, provides the output power already

coupled into a standard SMF with no additional losses and allows for a simple change of the laser wavelength and output coupling by substitution of the gratings. In Fig. 8, the output power of the waveguide laser at 1534 nm (circles) is reported as a function of the incident pump power for an output coupling of 32%. A pump power threshold of about 110 mW and a slope efciency of 8.4% were measured, with a maximum output power exceeding 30 mW at 500 mW incident pump power. For telecom applications, it is extremely important to fabricate a compact laser source able to generate in parallel all the channels of a WDM transmission. The femtosecond laser writing technology is able to provide such a device in a very easy and cheap way. Indeed, the waveguides here presented showed net gain in the whole C-band and can thus be used to form a waveguide laser array emitting on the ITU grid. To demonstrate this possibility, we made the waveguides lase close to the C-band upper edge (1560 nm). With an output coupling of 15%, we obtained a pump power threshold of 90 mW and a slope efciency of 6.6%, with a maximum output power of 23 mW (see Fig. 8, squares). The inset in Fig. 8 shows a typical relative intensity noise (RIN) trace of the laser operating at 1534 nm with 30 mW output power, recorded by means of a fast photodiode connected to an electrical spectrum analyzer. The RIN was dominated by a peak of about 105 dB/Hz, located at 200 kHz, corresponding to the relaxation oscillation frequency of the cavity. In the previous experiment, the laser cavity was longer than 50 cm, and the output coupler had a bandwidth of 0.25 nm, thus causing a spectral broadening of the source due to multimode oscillation. To achieve single mode operation, the output coupler was replaced by a narrow bandwidth FBG and the cavity length was reduced by cutting the FBG bers. Among the available FBGs, we selected the one providing the narrowest bandwidth and an output coupling

Fig. 7 Scheme of the linear laser cavity setup employing a bi-directional pumping scheme. ISO: Optical Isolator. PC: Polarization Controller at /2. WDM: Wavelength Division Multiplexer

Fig. 8 Inputoutput characteristics of the laser at 1534 nm and at 1560 nm. Inset shows a typical RIN spectrum of the laser at 1534 nm

24

R. Osellame et al.

as close as possible to the measured optimum value [47]. Our choice was a 57% output coupler with 64 pm FWHM bandwidth (corresponding to about 8 GHz). After cutting the FBG bers, the cavity length was L = 5.5 cm. Such a compact cavity resulted in a large free spectral range (FSR) of 1.82 GHz, thus consistently reducing the number of modes falling within the FWHM bandwidth of the FBG. Figure 9 shows the input-output characteristic of the 5.5-cm long laser cavity. The threshold pump power was 124 mW with 21% slope efciency. To ascertain single mode operation, we monitored the laser power spectrum by means of a scanning confocal FabryPerot interferometer (see inset in Fig. 9). The interferometer had a xed FSR of 8 GHz, a nesse of 100, and the signal to noise ratio on the photodiode was about 26 dB. Single mode operation was maintained till a maximum pump power level of about 400 mW, corresponding to an output power of about 55 mW. At higher pump power, a slightly multimodal regime appeared, with weak side peaks 1.8-GHz apart from the central mode, thus corresponding to adjacent longitudinal modes (indicated by arrows in the inset of Fig. 9). The very low intensity of these modes was conrmed by the fact that no change in the laser slope efciency between the single-mode and multi-mode regime was observed, as reported in Fig. 9. As compared to previously reported waveguide lasers fabricated with the same technology, this cavity suffers from a higher intensity noise. Actually, the use of a narrow bandwidth FBG make the single-mode laser RIN quite sensitive to cavity length uctuations induced either by mechanical vibrations at the ber-waveguide butt-coupling interface or by environmental temperature variations. An improvement to power stability could be achieved by using ber-pigtailing technique to permanently connect the waveguide to FBGs and by employing a Peltier cell to stabilize the length of the cavity against thermal variations.

4.2 Mode-locked laser Classically, the technique of passive mode-locking employing semiconductor saturable absorber mirrors has been widely adopted to generate ps and fs laser pulses [48], but recently a new technology based on carbon nanotubes (CNTs) has emerged as an alternative for the realization of saturable absorbers. CNTs show in fact strong and tunable saturable absorption in the near infrared, ultrashort recovery time, and can be cheaply assembled into polymer composites and easily integrated into optical ber communication systems. This enabled the rst demonstration of a mode-locked laser source based on carbon nanotubes technology and a femtosecond laser written waveguide. CNTs specially designed to have efcient saturable absorption at 1.5 m were produced by laser ablation [49] and embedded in free standing polyvinyl alcohol (PVA) lms of 50 m thickness [50]. The lm had a broad band absorbance spectrum around 1.5 m (see Fig. 10a), with 0.36 peak (corresponding to 1.52 dB absorption, of which 0.6 dB are saturable), a laser damage threshold exceeding 600 MW/cm2 , a saturation intensity of about 80 MW/cm2 , and a recovery time shorter than 1 ps. By sandwiching such a lm between a couple of ferrule plane connectors (FC/PC), with index matching uid at both ber ends, a ber-pigtailed CNT-PVA mode-locker was packaged (see Fig. 10b). Figure 11 shows a schematic of the mode-locked waveguide laser in a ring cavity conguration integrating the 36-mm long active waveguide fabricated by femtosecond laser writing and the ber-pigtailed CNT-PVA mode-locker. Two 976-nm pump laser diodes, providing 480 mW total incident power, were coupled to the waveguide by means of WDMs. Under this bi-directional pumping condition, the waveguide was able to provide up to 7.3 dB net gain at

Fig. 9 Inputoutput characteristic of the optimized ultra-compact waveguide laser. Inset shows the FabryPerot spectra at maximum output power in single mode (blue) and slightly multimodal regime (red)

Fig. 10 Schematic of the ring laser cavity for mode-locking operation

Lasing in femtosecond laser written optical waveguides Fig. 11 a Absorbance spectrum of the CNT-PVA free standing lm of 50 m thickness. b Schematic of the ber-pigtailed CNT-PVA mode-locker

25

Fig. 12 a Laser power spectra in continuous-wave and passive mode-locking regimes. b Autocorrelation trace of the laser pulses

1535 nm [46]. A bre coupler was used to couple 5% of the intracavity radiation out of the ring resonator. Unidirectionality of the ring was imposed by an optical isolator. Figure 12a shows the laser output spectrum recorded by an optical spectrum analyzer. Continuous-wave laser action started at 450 mW incident pump power, and self-starting single-pulse stable mode-locking was observed just above laser threshold. The laser central wavelength in the mode-locking regime was 1535 nm, with an oscillating bandwidth (at 3 dB) of 1.6 nm. By means of standard autocorrelation techniques (see trace in Fig. 12b) the pulse duration was estimated to be about 1.6 ps, resulting in a time bandwidth product of 0.329, in fairly good agreement with the 0.315 value for transform limited sech2 pulses. The repetition rate of this cavity was 16.7 MHz, and the timing jitter was estimated to be about 19 ps, corresponding to a remarkably low relative timing jitter of 3 107 [50]. A much higher repetition rate can be obtained using a linear cavity where the saturable absorber is directly placed on the waveguide facet. As an example, according to the saturation intensity of the nanotube absorber, 1 GHz operation could be feasible with an intracavity average power of about 100 mW. Such a power level was already demonstrated in similar waveguide lasers operated in cw [43]. A linear cav-

ity should also provide a strong reduction of the intracavity losses with respect to the ring cavity and thus potentially allows for much higher average output power (now limited to about 0.5 mW) as well as larger bandwidth of the pulses.

5 Conclusions In this paper we have reviewed our activity on femtosecond laser fabrication of active optical waveguides in Er:Ybdoped glass substrates. The optimized waveguides display high optical quality, with coupling losses as low as 0.1 dB/facet with standard SMFs and propagation losses lower than 0.4 dB/cm; they provide net gain up to 7 dB and laser action in the whole telecom C-Band. Advanced systems, such as single-frequency lasers with output power exceeding 50 mW and mode-locked lasers, have been demonstrated. The relevance of these results is twofold. On the one hand, they demonstrate that the femtosecond laser waveguide writing technology has matured to a point that it can be used to fabricate practical devices, with quality comparable to that obtainable with standard techniques. On the other hand, the low-cost compact active waveguide devices can nd applications in the telecom metro and access networks.

26

R. Osellame et al. 23. A. Zoubir, M. Richardson, C. Rivero, A. Schulte, C. Lopez, K. Richardson, N. Ho, R. Valle, Opt. Lett. 29, 748 (2004) 24. V.R. Bhardwaj, E. Simova, P.B. Corkum, D.M. Rayner, C. Hnatovsky, R.S. Taylor, B. Schreder, M. Kluge, J. Zimmer, J. Appl. Phys. 97, 083102 (2005) 25. R.R. Thomson, S. Campbell, I.J. Blewett, A.K. Kar, D.T. Reid, Appl. Phys. Lett. 88, 111109 (2006) 26. J. Burghoff, C. Grebing, S. Nolte, A. Tnnermann, Appl. Phys. Lett. 89, 081108 (2006) 27. A.H. Nejadmalayeri, P.R. Herman, Opt. Lett. 31, 2987 (2006) 28. R. Osellame, M. Lobino, N. Chiodo, M. Marangoni, G. Cerullo, R. Ramponi, H.T. Bookey, R.R. Thomson, N.D. Psaila, A.K. Kar, Appl. Phys. Lett. 90, 241107 (2007) 29. D. Homoelle, S. Wielandy, A.L. Gaeta, N.F. Borrelli, C. Smith, Opt. Lett. 24, 1311 (1999) 30. W. Watanabe, T. Asano, K. Yamada, K. Itoh, J. Nishii, Opt. Lett. 28, 2491 (2003) 31. S.M. Eaton, W. Chen, L. Zhang, H. Zhang, R. Iyer, J.S. Aitchison, P.R. Herman, IEEE Photonics Technol. Lett. 18, 2174 (2006) 32. C. Florea, K. Winick, J. Lightwave Technol. 21, 246 (2003) 33. G.D. Marshall, M. Ams, M.J. Withford, Opt. Lett. 31, 2690 (2006) 34. H. Zhang, S.M. Eaton, P.R. Herman, Opt. Lett. 32, 2559 (2007) 35. S. Nolte, M. Will, J. Burghoff, A. Tuennermann, Appl. Phys. A 77, 109 (2003) 36. R.R. Thomson, H.T. Bookey, N.D. Psaila, A. Fender, S. Campbell, W.N. MacPherson, J.S. Barton, D.T. Reid, A.K. Kar, Opt. Express 15, 11691 (2007) 37. A.M. Kowalevicz, V. Sharma, E.P. Ippen, J.G. Fujimoto, K. Minoshima, Opt. Lett. 30, 1060 (2005) 38. T. Pertsch, U. Peschel, F. Lederer, J. Burghoff, M. Will, S. Nolte, A. Tnnermann, Opt. Lett. 29, 468 (2004) 39. R. Osellame, N. Chiodo, V. Maselli, A. Yin, M. Zavelani-Rossi, G. Cerullo, P. Laporta, L. Aiello, S. De Nicola, P. Ferraro, A. Finizio, G. Pierattini, Opt. Express 13, 612 (2005) 40. R. Osellame, N. Chiodo, G. Della Valle, S. Taccheo, R. Ramponi, G. Cerullo, A. Killi, U. Morgner, M. Lederer, D. Kopf, Opt. Lett. 29, 1900 (2004) 41. P. Oberson, B. Gisin, B. Huttner, N. Gisin, Appl. Opt. 37, 7268 (1998) 42. S. Taccheo, G. Della Valle, R. Osellame, G. Cerullo, N. Chiodo, P. Laporta, O. Svelto, A. Killi, U. Morgner, M. Lederer, D. Kopf, Opt. Lett. 29, 2626 (2004) 43. R. Osellame, N. Chiodo, G. Della Valle, G. Cerullo, R. Ramponi, P. Laporta, A. Killi, U. Morgner, M. Lederer, D. Kopf, O. Svelto, IEEE J. Sel. Top. Quantum Electron. 12, 277 (2006) 44. A. Killi, U. Morgner, M.J. Lederer, D. Kopf, Opt. Lett. 29, 1288 (2004) 45. A. Killi, A. Steinmann, J. Drring, U. Morgner, M.J. Lederer, D. Kopf, C. Fallnich, Opt. Lett. 30, 1811 (2005) 46. G. Della Valle, R. Osellame, N. Chiodo, S. Taccheo, G. Cerullo, P. Laporta, A. Killi, U. Morgner, M. Lederer, D. Kopf, Opt. Express 13, 5976 (2005) 47. G. Della Valle, S. Taccheo, R. Osellame, A. Festa, G. Cerullo, P. Laporta, Opt. Express 15, 3190 (2007) 48. U. Keller, Nature (Lond.) 424, 831 (2003) 49. S. Lebedkin, P. Schweiss, B. Renker, S. Malik, F. Hennrich, M. Neumaier, C. Stoermer, M.M. Kappes, Carbon 40, 417 (2000) 50. G. Della Valle, R. Osellame, G. Galzerano, N. Chiodo, G. Cerullo, P. Laporta, O. Svelto, U. Morgner, A.G. Rozhin, V. Scardaci, A.C. Ferrari, Appl. Phys. Lett. 89, 231115 (2006)

In particular erbium doped waveguide ampliers (EDWAs) with net gain of 1020 dB can nd application in passive optical networks; an array of single-frequency waveguide lasers, spaced on an ITU grid, can be used in dense WDM systems; the mode-locked waveguide laser, with repetition rate scaled to the 12 GHz range, could serve as a transmitter in high-speed optical time division multiplexing systems.
Acknowledgement The authors acknowledge partial nancial support by Fondazione Politecnico di Milano (Research Program Advanced materials and devices for Photonics).

References
1. K. Itoh, W. Watanabe, S. Nolte, C. Schaffer, MRS Bull. 31, 620 (2006) 2. D. Du, X. Liu, G. Korn, J. Squier, G. Mourou, Appl. Phys. Lett. 64, 3071 (1994) 3. B.C. Stuart, M.D. Feit, A.M. Rubenchik, B.W. Shore, M.D. Perry, Phys. Rev. Lett. 74, 2248 (1995) 4. A.P. Joglekar, H. Liu, E. Meyhofer, G. Mourou, A.J. Hunt, Proc. Natl. Acad. Sci. USA 101, 5856 (2004) 5. K.M. Davis, K. Miura, N. Sugimoto, K. Hirao, Opt. Lett. 21, 1729 (1996) 6. K. Miura, J. Qiu, H. Inouye, T. Mitsuyu, K. Hirao, Appl. Phys. Lett. 71, 3329 (1997) 7. X. Orignac, D. Barbier, X.M. Du, R.M. Almeida, O. McCarthy, E. Yeatman, Opt. Mater. 12, 1 (1999) 8. E.N. Glezer, M. Milosavljevic, L. Huang, R.J. Finlay, T.-H. Her, J.P. Callan, E. Mazur, Opt. Lett. 21, 2023 (1996) 9. M. Streltsov, N.F. Borrelli, J. Opt. Soc. Am. B 19, 2496 (2002) 10. C.B. Schaffer, A. Brodeur, J.F. Garcia, E. Mazur, Opt. Lett. 26, 93 (2001) 11. C.B. Schaffer, J.F. Garca, E. Mazur, Appl. Phys. A 76, 351 (2003) 12. W. Chan, T. Huser, S. Risbud, D.M. Krol, Opt. Lett. 26, 1726 (2001) 13. T. Gorelik, M. Will, S. Nolte, A. Tnnermann, U. Glatzel, Appl. Phys. A 76, 309 (2003) 14. V. Apostolopoulos, L. Laversenne, T. Colomb, C. Depeursinge, R.P. Salathe, M. Pollnau, R. Osellame, G. Cerullo, P. Laporta, Appl. Phys. Lett. 85, 1122 (2004) 15. S. Eaton, H. Zhang, P. Herman, F. Yoshino, L. Shah, J. Bovatsek, A. Arai, Opt. Express 13, 4708 (2005) 16. K. Minoshima, A.M. Kowalevicz, I. Hartl, E.P. Ippen, J.G. Fujimoto, Opt. Lett. 26, 1516 (2001) 17. M. Will, S. Nolte, B.N. Chichkov, A. Tnnermann, Appl. Opt. 41, 4360 (2002) 18. G. Cerullo, R. Osellame, S. Taccheo, M. Marangoni, D. Polli, R. Ramponi, P. Laporta, S. De Silvestri, Opt. Lett. 27, 1938 (2002) 19. R. Osellame, S. Taccheo, M. Marangoni, R. Ramponi, P. Laporta, D. Polli, S. De Silvestri, G. Cerullo, J. Opt. Soc. Am. B 20, 1559 (2003) 20. M. Ams, G. Marshall, D. Spence, M. Withford, Opt. Express 13, 5676 (2005) 21. K. Hirao, K. Miura, J. Non-Cryst. Solids 239, 91 (1998) 22. J.W. Chan, T.R. Huser, S.H. Risbud, J.S. Hayden, D.M. Krol, Appl. Phys. Lett. 82, 2371 (2003)

Вам также может понравиться