1. Introduction

Microstructured fibers (MSFs) have considerably enhanced the possibilities of tailoring linear and nonlinear optical fiber properties. Shifting the zero-dispersion wavelength (ZDW) into the visible region and increasing the effective nonlinearity has enabled supercontinuum generation at the oscillator pulse energy level, see e. g. Refs. [1, 2]. Design of the dispersion properties of MSFs [3–5] yet enables other promising applications. While conventional fiber geometries only allow for soliton operation at above ~ 1.5 μm wavelength [6], MSFs can be designed to support solitons in the one-micron range. This allows for zero or anomalous dispersion operation within the gain bandwidth of neodymium or ytterbium, transferring the utility of erbium-based mode-locked fiber laser concepts to the nearer infrared range. Therefore, it appears possible to integrate laser gain and dispersion compensation into one single microstructured fiber-optical element.

However, so far only isolated reports on integration of doped MSFs into mode-locked fiber cavities have been published. Furusawa et al. reported the first actively mode-locked holey-fiber laser [7]. This laser employed an ytterbium-doped MSF as the gain medium and produced pulses of > 15 ps duration at 1030 nm wavelength. Intrinsic dispersion compensation has been demonstrated in air-clad large-core ytterbium-doped photonic crystal fibers for pulse amplification up to several microjoules [8]. Several other authors reported on hybrid schemes for short-pulse generation in the one-micron region, combining undoped MSFs for dispersion compensation and doped standard fibers [9, 10]. Here we report the first passive mode-locking of a Nd-doped MSF laser, using only a semiconductor saturable absorber mirror (SESAM, [11]) as the mode-locking mechanism. Preliminary results dealing with this topic have already been presented [12].

 

Fig. 1. Set-up of the mode-locked Nd:MSF laser. The solid beam represents the laser signal whereas the red outlined beam indicates the pump signal. M1, dichroic mirror (butt-coupled); M2, output coupler, ports A and B; L1, L2, L3 focusing lenses. The inset shows an scanning-electron micrograph of the Nd:MSF cross-section. d air hole diameter, Λ pitch.

Download Full Size | PDF

2. Experimental setup

The setup of our mode-locked MSF laser is shown in Fig. 1. A linear cavity configuration is used in the experiments. The Nd:MSF has a length of 1.25 m and is butt-coupled to a dichroic mirror, which allows for pumping the fiber laser either by a fiber pigtailed diode laser or by a cw Ti:sapphire laser, tuned to 803 nm. Launched power into the fiber core is 110 and 480 mW, respectively. On the opposing side, the fiber is coupled to an external air-filled resonator section (lens L2). The total optical cavity length amounts to 2.7 m, corresponding to a pulse repetition rate of 56MHz. A dielectric folding mirror (M2, T = 7%@1060 nm) is used for output coupling into two separate arms (ports A and B). A SESAM serves as the cavity end mirror and is used for self-starting and sustaining the passive mode-locking process. The saturable absorber exhibits a saturation fluence F _sat = 100 μJ/cm², an upperstate lifetime τ₂₁ = 32ps, and a modulation depth ΔR ~ 8%. The laser beam is focused onto the SESAM by a lens (L3, f=15 mm), resulting in an effective spot size of A _eff ~ 130 μm². This spot size on the SESAM was chosen carefully to avoid Q-switched mode-locked operation according to Ref. [13]. With this set-up, we observed a slope efficiency of 9% in either of the two ports relative to the launched pump power, corresponding to mode-locked output powers of up to 40 mW in one port. All measurements shown in the following were obtained using the Ti:sapphire pump source and port A as the output port.

 

Fig. 2. Calculated group velocity dispersion vs. wavelength. At the lasing wavelength, which is represented by the dashed line, the MSF still operates in the normal GVD regime. The calculation was carried out based on the measured MSF geometry (cf. inset in Fig. 1).

Download Full Size | PDF

A scanning-electron micrograph of the Nd:MSF cross-section is shown as an inset in Fig. 1. The solid silica fiber core, which exhibits a diameter of 7 μm, is doped with 1300 ppm Nd₂O₃. The core is embedded in a hexagonal air hole lattice with an air hole diameter of d = 4 μm and a pitch of Λ = 5 μm. Based on the measured geometry of the fiber, we calculated the group velocity dispersion (GVD) of this fiber shown in Fig. 2. The calculation reveals a ZDW of ~ 1.1 μm. This means that the dispersion is strongly reduced compared to standard fibers, but the fiber is still weakly normally dispersive (~ -10ps/nm/km) at the lasing wavelength of 1.06 μm. The MSF is birefringent (beat length 4 mm). The birefringence did not cause any noticeable degradation of the laser performance.

Spurious Fresnel reflections of the fiber end faces represent a major obstacle towards usage of MSFs in mode-locked laser cavities. These reflections cause stray Fabry-Perot effects, which pose a substantial scattering mechanism from the mode-locked pulse into the temporal continuum. Depending on its strength, this mechanism can completely prevent start-up of the mode-locking process [14, 15]. In conventional fiber cavities these parasitic effects can easily be suppressed by angle-cleaving or angle-polishing of the end faces. However, neither of these methods is directly applicable to MSFs, because mechanical stress unavoidably damages the fragile lattice structure, especially at high air-fill fractions. We resolved this problem by ruggedizing the MSF end face prior to polishing. We used an UV curable liquid resin, which is readily pulled into the air holes by capillary forces and has an index of refraction of 1.53, i.e., slightly higher than the 1.45 index of silica. After UV curing, a solid composite section of reproducible length is formed. The solidified fiber end face is then easily processed. We angle-polished the end face and removed most of the resin, leaving only a small cap layer that seals the air holes. The cap layer is sufficiently thin (one millimeter or less) not to affect the waveguiding properties of the MSF endface. Note that this processing is only required for the fiber end face close to L2. With the rather moderate intracavity power levels below 0.5 W we did not record indications of optical damage in the cap layer.

3. Results and discussion

With the angled end face the mode-locked behavior is strongly contrasted to the behavior without subcavity suppression. In the latter case, we observed irregular spiking without indications for continuous mode-locked operation. The observed short pulse trains exhibited multiple periodicities due to the coupled cavities. However, employing the angle-polished fiber, we can greatly suppress these effects and observe cw mode-locking operation at only the fundamental cavity roundtrip frequency of 56 MHz. We convinced ourselves that we have in fact achieved clean cw mode-locked operation by measuring the radio frequency spectrum of the pulse train (see Fig. 3). This measurement indicates suppression of self-Q-switching sidebands [13] by more than 38dB.

 

Fig. 3. Radio frequency spectrum of the first intermode beatnote at f _R = 56 MHz. Resolution bandwidth is 1 kHz. The linewidth of the beatnote is ≤ 500Hz (FWHM) at the resolution limit. The modulation sidebands at ± 1 MHz are ≥ 38 dB below the carrier.

Download Full Size | PDF

We further characterized the mode-locked operation by measuring autocorrelation and optical spectrum. Figure 4 shows the measured intensity autocorrelation function of the mode-locked pulses. The duration of the autocorrelation is about 30 ps (FWHM). The shape of the autocorrelation appears to be close to a triangular shape, indicative for a steep slope in at least one of the wings of the laser pulse shape. To provide further insight we decorrelated the autocorrelation by application of the TIVI algorithm [16]. For the data in Fig. 4, the algorithm converges to the nearly single-side exponential pulse shape shown in the inset. The reconstructed pulse profile exhibits a pulse duration of 26 ps (i. e. a near-unity deconvolution factor) and reproduces remarkably well the response function of the SESAM with its 32-ps relaxation time. The measured optical spectrum of the laser pulses is depicted in Fig. 5. The measured spectral width is 1.85 THz (FWHM). The calculated time-bandwidth product of ~ 50 reveals that the pulses are strongly chirped. It should be pointed out that no systematic dispersion compensation has yet been achieved. Ideal compression of these chirped pulses should result in a 200-fs pulse duration. In the current configuration, however, pulse shaping appears to be ruled by an interplay between saturable absorption of the SESAM and dispersive broadening. Self-phase modulation must also play an important role, as indicated by the rather high time-bandwidth product. This is confirmed by a calculation of the nonlinear length, which yields an estimate of about 1 m.

 

Fig. 4. Intensity autocorrelation function of the mode-locked pulse train with a duration of 30ps (FWHM). The inset shows the decorrelated intensity profile of the pulse [16]. The reconstructed pulse width is 26ps (FWHM).

Download Full Size | PDF

From the pulse characterization, we estimate that the pulse carries a group delay dispersion of 15 ps² [6], an amount that can only be removed by grating pairs of high groove densities and at large grating separation in an extracavity compressor. It is therefore desirable to further optimize the MSF dispersion for generation of pulses closer to the bandwidth limit, preferably by shifting the dispersion into the anomalous regime. Nevertheless this first demonstration of passive mode-locking of a MSF laser already indicates the potential of transferring pulse generation methods that are based in the 1.5 μm erbium band. An advantage of the 1 micron regime is the mature GaAs/AlGaAs SESAM technology, which can readily be exploited to provide self-starting and generation of picosecond pulses. As demonstrated, self-Q-Switching and sub-cavity effects can be removed by proper design and fiber preparation. In this first demonstration, we already achieved pulse durations of about 26 ps at a pulse energy of 0.7 nJ. We expect that specifically designed fibers will allow for a significant reduction of intrafiber dispersion and therefore allow for a much shorter pulse duration. In conclusion, we believe that microstructure geometries are very promising to greatly extend the operation regime of mode-locked fiber lasers to well outside the 1.5 μm window.

 

Fig. 5. Mode-locked spectrum measured at a resolution of 0.2 nm. The center wavelength is ~ 1060 nm. The spectral width is ~ 7 nm (FWHM).

Download Full Size | PDF