1. Introduction

Fiber optical parametric oscillators (FOPOs) exhibit a remarkable range of performance in terms of wavelength tunability and pulse duration. There are continuous wave and pulsed FOPOs using standard fibers or microstructure fibers (MFs) [1–14]. From a practical perspective, these instruments are becoming well suited for multiphoton diagnostic applications, such as multiphoton fluorescence imaging and coherent anti-Stokes Raman spectroscopy [15, 16] where there is a pressing need for compact portable sources for use in clinical research environments. The use of FOPOs promises to dramatically extend the functionality of ultrafast fiber lasers both in terms of short-pulse generation and wavelength agility. In turn, this will lead to devices that have similar performance to Ti-, Nd-, and Cr-doped solid state lasers, but are less cumbersome and have a dramatically smaller footprint.

While wavelength tunability has been the subject of numerous papers, the use of FOPOs for generating ultrafast pulses has not received much attention. In this paper we report on a FOPO system that delivers the shortest optical pulses yet reported for a FOPO of any type. We obtain nearly transform limited 70-fs pulses for a system incorporating a 4.2-cm MF. The system operates at a repetition rate of 50 MHz with pulse energies up to 1-nJ. Although wavelength tunability is not the focus of this work, intracavity optical spectra suggest the possibility of using FOPOs for the generation of ultrashort pulses over a wavelength range from 700 nm to 1400 nm. We describe the general operation of the system, its temporal stability, and the performance variations as functions of pump power.

Table 1 summarizes the current state of research into pulsed FOPOs which incorporate a wide variety of standard or microstructure fibers. During the past 10 years, pulse durations ranging from 450 fs to a few nanoseconds have been reported. The first FOPO [2] that included MF as the parametric gain medium operated in the 750 nm wavelength range, and Deng, et al. [3], implemented a pulsed oscillator centered around wavelengths of 1000 nm. A third notable body of work is that of Harvey, et al. [4]. These three works illustrate the FOPO scheme’s flexibility in terms of wavelength, fiber type, pulse duration, phase matching considerations and cavity designs.

An important aspect of our work is the use of very short gain fibers, which has been suggested by several scientists including Deng, et al., [3]. Other recent work by Sharping et al., [5] highlights the expected increase in tunable bandwidth and output power that can be achieved by using the shortest possible parametric gain fiber within a FOPO cavity. The use of short silica nanowires has drawn considerable attention even in single pass settings. Recent work has shown that short MFs can favor nonlinear pulse propagation for temporal compression by avoiding dispersive pulse broadening and/or Raman-shifted soliton effects. Foster, et al., [17] have used an optimized sub-wavelength diameter fiber to generate single-cycle pulses. By choosing a fiber of optimized length and core diameter one can combine spectral broadening from the enhanced effective nonlinearity with higher-order soliton compression. Domachuk, et al., [18] have also studied the novelty of nanoscale glass waveguides for a variety of applications, most notably the generation of a supercontinuum spectrum extending from 789 nm up to 4870 nm. They take advantage of the enhanced transmission and nonlinear properties of a MF fabricated from soft glasses. Other important work explores the generation of low-noise [19] and all-fiber supercontinuum sources [20].

Table 1. Comparison of FOPO devices

View Table

2. Methods

The MF used here is a commercially available MF (Crystal Fibre - SC-5.0-1040) that has been drawn down to a reduced core size in order to modify the fiber’s dispersion profile. We use a filament-heated tapering setup to modify the fiber and remove the transition regions which leaves just the narrowed waist region. The tapering setup can produce drawn-down fibers in excess of 10 cm in length. The dispersion-zero of the fiber is measured using spectral interferometry [22] in which the short MF under test is placed in one arm of a Michelson interferometer. The wavelength dependence of the pulse delay can be obtained by comparing the delay measured from the front fiber facet reflection and the output fiber facet reflection. The group-velocity dispersion (GVD) as a function of wavelength is shown in Fig. 1(a) for both the modified fiber and the unmodified (as purchased) fiber. The dispersion zero is 1010 nm for the modified fiber (core diameter of 4 µm, nonlinear coefficient γ~12 (W km)^-1) and 1060 nm for the unmodified fiber (core diameter of 5 µm, γ~10 (W km)^-1). The nonlinear coefficients foThe resulting dispersion profile for the modified fiber is nearly ideal for use in a FOPO which is pumped by a 1032 nm Ytterbium-doped fiber laser. This can be seen in Fig. 1(b) where we plot the phase matched “peak” location as a function of pump laser wavelength for both the modified and unmodified fibers. Note that the resulting fiber also has a relatively large glass-to-air fraction in the cladding of the MF. Our experiments suggest that fibers with large glass-to-air filling ratios are less susceptible to damage, presumably because coupling of the intense pump light into the interstitial strands of glass is less likely.

The pump laser source is a 370-fs, 50-MHz mode-locked Ytterbium-doped fiber laser (PolarOnyx - Uranus) with a polarized output. The laser consists of a seed oscillator followed by an amplifier, and we measured the repetition rate stability to be +/- 20 Hz/hour, corresponding to a length fluctuation of about 4.8 µm/hour. This level of stability is important since the FOPO cavity must remain synchronous with that of the pump laser. Inclusion of all losses due to isolating the pump source, delivering the pump light into the system, and coupling into the fiber are included, results in a maximum of about 1.2 W of average power delivered through the fiber.

 

Fig. 1. (a) Before- and after-modification plots of measured GVD vs wavelength for the SC- 5.0–1040 MF used to provide parametric gain for the FOPO. (b) Calculated phase matched peak locations according to the measured GVD profile for the same two fibers.

Download Full Size | PDF

The FOPO setup with a 4.2-cm-long MF is shown schematically in Fig. 2(a). Pump light is coupled into the 3-m-long Fabry-Perot cavity through a dichroic mirror [provided by Precision Photonics, see Fig. 2(b)] and into the MF through an aspheric lens (Thorlabs, C230TM-B) with a maximum total coupling efficiency of ~60%. In a single pass, we observe spectral sidebands due to four-wave mixing. When the FOPO cavity is synchronized with the pump laser cavity, the long wavelengths oscillate while the short wavelengths are coupled out of the cavity via transmission through the high reflector. The fiber length of 4.2 cm was chosen based on our ability to mount it in a mechanically stable fashion. Oscillation has been achieved for fiber lengths as short as 1 cm, but a systematic comparison of the output has not yet been performed.

 

Fig. 2. (a) Schematic of the FOPO setup. SPD, short-pass dielectric; AL, aspheric lens; Mx, mirror; HR, high reflector; SPF, short-pass filter. (b) Measured reflectivity of the SPDs used as the end mirrors of the cavity.

Download Full Size | PDF

A desirable feature of this system is its relatively simple alignment. The cavity length of the oscillator is approximately established by measuring the repetition rate of the pump laser and setting the length of the cavity appropriately. The most challenging aspect is coupling the pump light efficiently into the fundamental mode of the MF. Once satisfactory coupling is achieved (~50%), the output from the fiber is collimated towards the output end mirror. An infrared viewer is used to overlap the retro-reflections on both sides of the cavity. If one has sufficient pump power to obtain about 10 π of nonlinear phase shift, then a visible spot due to supercontinuum generation is observed which aids in alignment and provides more than enough power to observe oscillation. Finally, the cavity length is fine-adjusted by translating the input pump coupler until the system flashes. Once oscillating, the FOPO cavity alignment is optimized by minimizing the pump power threshold of oscillation.

 

Fig. 3. Plots of output peak amplitude and peak variance as a function of cavity length delay as extracted from the pulse train observed using a fast oscilloscope.

Download Full Size | PDF

The passage from not oscillating to oscillating and back can be seen in the Fig. 3 where we plot the amplitude and relative amplitude variance as a function of end-mirror position (represented in picoseconds as a change in the cavity period). Following the curve from left to right, one sees that the amplitude is small and the relative variance is large when the cavity length is not synchronous with the pump laser. Adjusted in this manner, only a small amount of supercontinuum is present at the output. The amplitude increases and the variance decreases dramatically when the cavity becomes synchronous (between about 0.2 ps and 0.6 ps) and the system oscillates above threshold. In the middle of the range of oscillation the stability of the output of the FOPO is the same as that of the pump laser as observed on an oscilloscope. Beyond 0.6 ps the FOPO cavity is no longer synchronous, so the amplitude decreases and the variance increases. Careful inspection of the data reveals a slight increase in the amplitude in the range of 1.4 ps to 1.5 ps. This is believed to be the result of the system oscillating at the other polarization mode of the MF. The peak at 1.4 ps to 1.5 ps can be reduced by careful adjustment of the input pump polarization. The reduced variance at 0.4 ps in Fig. 3 illustrates the cavity-enhanced short-term stability. Pulse to pulse amplitude measurements on an single-shot oscilloscope trace reveal a system that is as stable as the mode-locked pump source. Long-term stability has not been characterized, but the system has been operated at least weekly for several months at maximum pump power without damage to the MF.

3. Results

The results described below are for a system oscillating at 1250 nm with the conjugate field at 880 nm coupled out of the cavity. Typical measurements of the optical spectrum and pulse autocorrelation are given in Fig. 4 where the autocorrelation for minimized pulse duration implies a Gaussian pulse width of 70 fs. The system has a threshold peak pump power of 19 kW. Adjusted for a minimum pulse duration the FOPO produces 0.4 nJ pulses with an output peak power of 5.0 kW for a pump peak power of 22 kW. This represents greater than 20% peak power conversion into useful output. The system can be adjusted to produce pulses with energies as large as 1.2 nJ, but the temporal and spectral profiles become distorted as shown in the bottom plots in Fig. 4. Many FOPOs reported in the literature include wavelength tunability by either including a spectral filter or by introducing an adjustable dispersive element to permit cavity length tuning. This system has been constructed with a minimum of filtering and dispersion in hopes of generating pulses with a broad spectral content. The center wavelength can, however, be tuned slightly (+/- 5 nm) by translating the aspheric fiber coupling lenses. The actual wavelength of operation differs from the predicted peak location given in Fig. 1. We are not certain why there is a difference but believe that it is due to the nonlinear phase shift introduced by the high power pump pulses.

The resulting output pulse duration is dependent on a combination of several factors: the achievable parametric gain bandwidth, the chirp of the input pulses, saturation of the gain above a certain peak pump power, and the amount of dispersion in the cavity. Firstly, we expect a lower limit to pulse duration that is given by the parametric gain bandwidth. Secondly, once pulses are generated, they accumulate a spectral phase contribution from GVD that is present in the cavity. The third factor comes from the fact that the oscillating signal and the output conjugate field are phase-conjugates of each other. This means that the output is created within the MF having the same magnitude but opposite sign chirp as the oscillating signal. The interplay of these pulse evolution factors is the subject of ongoing investigation. The data shows that the shortest pulses are nearly transform limited with a time-bandwidth product of 0.67 for a Gaussian pulse shape (the transform limited pulse duration is 50 fs).

The dependence of the output on pump power is illustrated in the plots in Fig. 4 in which the spectra and autocorrelations of the output of the system are shown for three different coupled powers ranging from near oscillation threshold (420 mW average, 19 kW peak) to the onset of saturation (500 mW average, 23 kW peak). Observe that the spectral width increases with power and develops additional structure. Eventually, a dip forms in the center of the spectrum giving clear evidence of the onset of gain saturation. The autocorrelations are consistent with the spectral data in which sidebands form on the traces which suggest the presence of saturation near the peak of the pulse. Although oscillation can occur for an extremely wide range of pump powers, these data reveal that the shortest pulse durations are achieved for a fairly narrow range of pump powers.

 

Fig. 4. Optical spectra (left) and autocorrelation traces (right) of the FOPO output for various coupled pump powers.

Download Full Size | PDF

4. Conclusions and future work

We describe the first FOPO capable of producing sub-100 fs pulses. The FOPO incorporates a reduced diameter 4-cm-long MF within a Fabry-Perot cavity. The presence of the cavity provides enhancement of the output power for a particular range of output wavelengths and for excellent pulse-to-pulse amplitude stability. This system is rich with opportunities for future exploration. In particular it will be critical to understand the mechanisms that limit the output power, wavelength, and pulse duration. In addition to fiber length, core size and dispersion properties, other cavity design choices such as coupling optics, mirrors, and the amount of feedback are critical as well.