1. Introduction

The state of the art for sub-200 fs (herein referred to as “femtosecond”) pulsed light generation is the Ti:Sapphire [1] laser which can be tuned over the range of 700-1100 nm. This system, in combination with crystal-based parametric devices, can deliver femtosecond pulses with wavelength tunability throughout the visible and well into the infrared optical spectrum. Femtosecond pulsed laser systems have enabled groundbreaking studies on microscopy, time-resolved photochemistry, and frequency metrology [2–4]. The principal drawbacks of these systems are their size, complexity and cost.

An alternative architecture for wavelength-tunable femtosecond pulse generation is the mode-locked fiber laser combined with a fiber-optical parametric oscillator (FOPO) [5]. Within this architecture the fiber laser delivers a stable high-power pulse train at a fixed wavelength and the FOPO enables conversion to other wavelengths [6–10]. Most early FOPOs operated in the 1.5 μm wavelength range and took advantage of optical fiber communication technology such as low loss optical fibers, fiber-integrated filters, and amplified pulsed lasers. Microstructure fibers (MFs), also called photonic crystal fibers, have extended the functionality of FOPOs to visible wavelengths and further into the infrared.

In this paper we present detailed studies of the pulses generated by an ultrashort pulsed FOPO incorporating a microstructure fiber (MF) for parametric gain. The FOPO generates a 50 MHz pulse train in the 1210 nm to 1340 nm wavelength range. The output pulses are as short as 130 fs in duration with average output powers of 10 mW to 100 mW corresponding to peak powers as high as 12 kW. This is the first ultrafast FOPO with greater than 50 mW average power to operate in this wavelength range, and the first study of the full temporal phase and amplitude of the output from a FOPO. The measurements reported here provide further experimental insight into the behavior of several recently discussed pulsed FOPO systems [9, 10].

2. Experiment

Our experimental setup is depicted in Fig. 1(a) . The pump laser is a mode-locked Ytterbium fiber laser (PolarOnyx, Uranus) that emits a 50 MHz train of pulses. The laser generates 2.1W of average power at a wavelength of 1032 nm. The average power that is routinely coupled through the MF is 900 mW. The FOPO is a Fabry-Perot cavity formed between a glass window (W) and a metallic mirror (M4). Other mirrors serve to fold the cavity whose round trip time must be synchronous with the pump pulse period. Light is coupled through the MF and collimated within the cavity using aspheric lenses (Thorlabs, Inc. C230TMeB). The output coupler is a short-pass dielectric mirror placed near L2. When the system is well aligned, the output mode is about 3 mm in diameter, and is well collimated for the output infrared wavelength. Note that the cavity oscillates as a result of gain and feedback at the short wavelength sideband while the long wavelength is coupled out as the useful output. Interference filters (not shown) remove any pump light from the output coupled infrared pulse train. Starting from very good alignment, the system will typically deliver stable output for about two hours. The system is open to air currents in the lab and has no active stabilization. The primary cause of instability is imperfect coupling of the pump laser into the optical fiber. Coupling losses lead to temperature induced mechanical changes in the fiber clamp.

 

Fig. 1 (a) Schematic of the FOPO. The output coupler is a short-pass dielectric (SPD) that reflects wavelengths longer than 1100 nm and transmits shorter wavelengths. Reconstructed temporal amplitude and phase profiles (b) error=0.007 and (c) error=0.01 of pump laser pulses recorded on different days.

Download Full Size | PDF

The fiber is a 2-cm long piece of MF (SC 5.0 1040) fiber that was modified in a linear flame-brush tapering apparatus to have a zero dispersion wavelength near 1030 nm. The tapering apparatus consists of a hydrogen-oxygen torch that was set to burn as cool and with as small of a flame size as possible. The fiber is clamped between two translation stages pulling at 0.01 mm/s each for a distance of 6 mm each. The flame brushed back and forth at a rate of 2 mm/s for a length of 4.6 cm. The tapering recipe results in an estimated core diameter reduction from 5.0 to 4.7 μm. The two ends of the MF were spliced to standard fiber (SMF 28) allowing the optical transmission to be monitored throughout the tapering process. The insertion loss of the spliced SMF/MF combination measured before and after modification changed by less than 1 dB as a result of the flame-brushing process.

There are two key differences between this FOPO system compared with our earlier versions [5,10]. The first is that the pump combining optic, which also serves as one end of the cavity, is an uncoated glass microscope slide. This reduces the cost of building such a device dramatically because one no longer needs a specialty optic for coupling the pump into the cavity. As a consequence of using a microscope slide as the reflector the cavity experiences a loss of roughly 99%. Fortunately, the parametric gain can easily overcome this loss in order to support oscillation above threshold in the cavity. The second key difference from earlier implementations is the placement of the dichroic output coupler mid-cavity rather than as one of the end mirrors. The advantage of this approach is that the transverse beam size, divergence, and propagation direction can more easily be modified using lenses outside of the cavity. In the previous implementation, the output coupler was in place of M4 and the output beam would have propagated about 2 m inside of the cavity. We found experimentally that our measurement system was much easier to align from week to week because the output coupler could be adjusted slightly without having an effect on the oscillation of the cavity.

We use second harmonic generation, frequency resolved optical gating (SHG-FROG) to fully characterize the input and output of our FOPO. This technique permits the determination of the temporal and spectral amplitude and phase of the pulses under test. One weakness of SHG-FROG is a time-reversal ambiguity. Despite this ambiguity, SHG-FROG gives us enormous insight about the quality of the pulses, both pump and FOPO output, participating in the experiments. Our SHG-FROG apparatus is a Michelson interferometer with offset beams coupled through a BBO crystal in order to obtain background-free SHG. The spectrum of the second harmonic is collected for a sequence of pulse delays using a 2-inch focal length lens and coupled into an optical spectrum analyzer (OceanOptics, HR2000 + ) via a multimode fiber. The scanning arm of the interferometer is actuated with a motion controlled translation stage (Newport, Inc., MFA-CC). LabView software acquires the spectrogram (a 2-dimensional map of spectrum as a function of delay) and controls the entire experiment. Analysis of the spectrogram including phase retrieval is performed using commercially available software (FemtoSoft Technologies v3.2.2).

3. Results

We recorded SHG-FROG traces of the pump pulses and find that the pulse shape is not a simple Gaussian or Sech² shape. The shape varies from week to week depending on laboratory conditions. Figures 1(b) and (c) depict the retrieved temporal amplitude and phase of pulses emerging from the pump laser. Well-formed pulses, as shown in Fig. 1(b), consist of a pronounced main peak and two small peaks on either side. The main peak is slightly chirped with less than 1 radian of phase variation across the pulse. The equivalent autocorrelation measurement assuming a Gaussian pulse shape suggests a pulse duration of about 360 fs FWHM. More complicated pulse shapes, as shown in Fig. 1(c), are observed when the system polarization is not properly adjusted. For poor adjustment, the pump pulses feature two peaks, and slightly more than 1 radian of phase variation. The laser performance is weakly coupled with temperature fluctuations in the lab and the pulse shape can change from day to day. The pump laser output can be corrected using fiber polarization controllers and a grating pulse compressor which are part of the laser system. With enough pump power or a long enough optical fiber, the FOPO will still function for sub-optimal pulse shapes.

The optimized output of the FOPO is shown in Fig. 2 . Figure 2(a) is the measured SHG-FROG spectrogram which is symmetric and has very little structure other than the main peak. The reconstructed temporal amplitude and phase reveal a pulse-duration of 120 fs, a spectral full-width at half maximum (FWHM) of 17 nm, and a time-bandwidth product of 0.4. The average output power for the data presented in Fig. 2 is 70 mW, and the center wavelength of the output is 1260 nm. Optimization of the FOPO is achieved by carefully adjusting the pump laser, setting the FOPO cavity length to the middle of its range of synchronization, and collimating the infrared output from the output coupler.

 

Fig. 2 FROG measurements of the output of the FOPO. (a) The spectrogram recorded when the FOPO was aligned for short pulse generation. (b) Temporal amplitude (right, markers) and phase (right, line) retrieved from the measured spectrogram.

Download Full Size | PDF

The data in Fig. 2 corresponds to the FOPO generating an output wavelength of 1260 nm, which can be adjusted due to dispersion in the cavity as the input window, W (see Fig. 1(a)), is translated. A composite of optical spectra recorded at the output of the FOPO are shown in Fig. 3(a) . The output wavelength ranges from 1210 nm up to 1340 nm as the end mirror is moved so that the cavity length is reduced.

 

Fig. 3 (a) a composite of optical spectra recorded as the cavity length of the FOPO is varied. (b) FROG spectrograms recorded for several different FOPO output wavelengths. Vertical shifts in wavelength are not shown here because each spectrogram has been re-centered vertically. (c) Reconstructed amplitude and phase obtained from the spectrograms.

Download Full Size | PDF

Spectrograms and retrieved pulse shapes for several settings of the end mirror are presented in Fig. 3 (b) and (c). The data reveals that dispersion tuning not only changes the center wavelength of the output, but it has a significant effect on the temporal shape of the pulses. Although the spectral shape of the pulses is well behaved throughout the range of synchronization, full pulse characterization measurements for different output wavelengths reveal that the temporal pulse shape is somewhat more complicated at the edges of the range of synchronization compared with the center. The least distorted pulses are obtained in the center of the tuning range. This is consistent with similar pulse shaping effects which have been studied in χ⁽²⁾ OPOs [11].

Figure 4 illustrates the variation in pulse quality as a function of pump power. The spectrograms and reconstructed pulse shapes are for coupled pump powers ranging from 950 mW down to near oscillation threshold at 850 mW. The pulse shape and phase variation can be seen to be more well-behaved near threshold compared to when the system is pumped with too much power.

 

Fig. 4 (a) FROG spectrograms recorded at the output of the FOPO for several different pump power settings. (b) Reconstructed amplitude and phase obtained from the spectrograms.

Download Full Size | PDF

3. Analysis

The FOPO is tunable in wavelength as the end mirror of the cavity is scanned. The delay introduced by scanning is typically 670 ± 70 fs. There are several mechanisms which might result in this wavelength change. Recall that the output wavelength decreases as the cavity length increases. As such, the cavity is experiencing normal net dispersion at the oscillating wavelength. We consider four potential contributors to GVD: cavity folding mirrors, the dichroic output coupler, the optical fiber, and the fiber-coupling lenses. The cavity folding mirrors are silver and gold mirrors with protective coatings and we assume that they introduce a negligible group delay. The dichroic output coupler mounted at a 45 degree angle might introduce significant GVD. One can observe above threshold operation of the FOPO without that optic present. A comparison of the scanning range with and without the output coupler implies that the output coupler introduces 120 ± 50 fs of group delay. The GVD of the fiber averages −55 ps/nm km over the range of wavelengths of oscillation. For two passes through the 2-cm long fiber we expect a normal group delay of about 150 ± 30 fs over the entire range of tunability. The fiber-coupling lenses have a thickness of 2.79 mm at the axis. Using the refractive index data published by the manufacturer one can estimate that there should be a normal group delay of 60 ± 5 fs per pass through the lenses. Four passes through fiber coupling lenses leads to 240 ± 20 fs of normal group delay. Totaling these contributions one obtains 510 ± 60 fs of normal GVD. Given that the pump pulse duration is 370 ± 30 fs, these four contributions explain the dispersion tuning behavior we see from this FOPO.

Naturally, these same physical mechanisms limit the minimum pulse duration we can generate at the FOPO output, which was 130 fs for the system described in this report. An interesting feature of this system is the ability to configure the system to oscillate at long wavelengths and output couple the short wavelengths. We reported on such a system earlier [10] where the pulse duration was as short as 70 fs. The reduced pulse duration is consistent with the fact that the oscillating component travels in the anomalous dispersion wavelength range of the MF. As such, there is some anomalous dispersion to offset the normal dispersion introduced by the other components.

As shown in Fig. 3 (c), the quality of the output also changes as the wavelength is tuned. Spectrograms recorded at four different wavelength settings reveal that near the edge of the tunable range at 1200 nm, the pulses are clearly asymmetric and have a large leading or trailing peak (SHG-FROG cannot distinguish which part of the pulse is the front and which is the back). The fact that there is such a peak is consistent with the assumption that as the cavity length is changed, only the leading or trailing edge of the oscillating pulse overlaps well with the subsequent pump pulse and is amplified above threshold. As the cavity is tuned towards 1245 nm, closer to the center of its tuning range, the pulses remain complex but are much more symmetric. Furthermore, when optimized for a short pulse duration (see Fig. 2), the output wavelength is 1260 nm and the system delivers a symmetric, nearly transform-limited pulse. The temporal phase reveals that the residual chirp on the pulses appears to be higher order (cubic or quartic), although the magnitude of these variations is less than 0.1 radians.

The quality of the output varies as the pump power changes. The oscillator reaches a threshold where the gain exceeds the cavity losses at about 825 mW. The output power increases somewhat as the pump power is increased and then saturates at about 900 mW. Spectrograms recorded for different pump powers reveal the impact of instantaneous gain saturation on the output pulses. Well above threshold (far left in Fig. 4 (a)) the FOPO output is complicated with multiple peaks and phase fluctuations exceeding 2 radians across the pulse. This is the case because gain saturation occurs near the peak of the pump pulses and energy can be coupled back into the pump. Data recorded at intermediate pump powers shows that there is a transition as the pump power is reduced to a pulse shape having most of its power in one peak. Near threshold (far right in Fig. 4) one observes that the pulse takes on a much more symmetric shape, and the phase variations decrease to about 1.5 radians across the pulse. Optimization of the pulse duration with respect to pump power consists of operating at a pump power which is above the oscillation threshold by only about 5%. When optimized (see Fig. 2), the pulse is symmetric and exhibits a minimum of phase variations across the pulse.

4. Conclusions and future work

The system reported here uses simplified optics and delivers nearly 150 nm of tunability in the near infrared. We observe pulse durations as short as 130 fs and output powers approaching 100 mW average (12 kW peak). Detailed studies of the pulse quality reveal that nearly transform-limited 130 fs pulses can be generated, but the pulse quality becomes asymmetric and chirped when tuned away from the optimal setting. Pulse quality is also dependent on the pump power relative to the threshold for oscillation where there is a small dynamic range of powers (~10% of the threshold value) over which optimal pulses are generated.

We anticipate that these studies will inform future work in designing FOPOs with a broader range of wavelength tunability, shorter output pulses, greater power, better efficiency, and enhanced ease of use.