1. Introduction

High-energy ultrafast lasers with repetition rates in the hundreds of kHz range provide superior results in laser waveguide inscription because they can operate in a thermally-moderated regime, which is inaccessible to lower-repetition-rate systems [1]. Waveguide inscription in this high-repetition-rate regime is limited by the availability of suitable sources, which are generally restricted to systems operating around 1060 nm, in the form of cavity-dumped solid-state lasers [2] or pulse-picked Yb:fiber master-oscillator power-amplifier systems [3]. The inscription of infrared materials such as Si [4], Ge and certain chalcogenide glasses implies a need for ~1-µJ sources operating at wavelengths well above 1060 nm. In this context, we present results from a new class of synchronously-pumped optical parametric oscillator (OPO), providing sub-250-fs pulses at repetition frequencies from 101 kHz – 15.3 MHz, and whose energies are sufficient for waveguide inscription via refractive-index modification [1].

To date, the highest pulse energies from a femtosecond OPO were obtained from a fiber-feedback OPO, based on periodically-poled stoichiometric LiTaO₃, producing 339-nJ, 840-fs signal pulses at a wavelength of 1450 nm [5]. Recently, another fiber-feedback OPO based on MgO:PPLN and operating from 1.5 – 1.7 µm produced 490-nJ, 100-ps pulses [6]. Both of these systems realized energy scaling by operating at high average power, however cavity-dumping a synchronously-pumped OPO [7] offers an alternative route to high-energy pulses by using a pump laser with significantly lower average output power. In this paper, we describe a cavity-dumped system exceeding the performance of all previous femtosecond OPOs in both pulse energy and peak power.

2. Design of the cavity-dumped optical parametric oscillator

Our work extends earlier research in which a 10-W Yb:fiber laser was used to synchronously pump a MgO:PPLN OPO [8,9]. The pump laser was a commercial system (Fianium FP1060-1uJ) producing chirped 15.3-MHz pulses compressible to sub-400-fs durations. After compression, 6.6 W were available for pumping the OPO, which was resonant at 1.52 µm and was based on a 30-µm period MgO:PPLN crystal of length 1 mm. The OPO is shown in Fig. 1 and was an asymmetric, folded X-cavity, in which the shorter arm contained an SF10 prism pair for dispersion control, and the longer arm a TeO₂ acousto-optic modulator (AOM) situated between two focusing mirrors (M₆ and M₇). The resonant signal pulses were dumped as they propagated from mirror M₆ to M₇, and the dumped pulses were collimated by M₇ then collected by an extra-cavity mirror. The dumper was situated in the middle of the cavity to maximize the available switching time before the residual intra-cavity pulse returned after a reflection from the output coupler (OC). Mirror-pairs M_4/5 and M_8/9 formed a 4f-relay system, which achieved a stable 9.8-m long cavity that was insensitive to misalignment.

 

Fig. 1 Cavity layout: PG, pulse generator; LPF, low-pass filter; C, compressor; PD, photodiode; X, MgO:PPLN crystal; D, AOM cavity-dumper. See text for other definitions.

Download Full Size | PDF

Cavity dumping was implemented by locking the internal clock of a pulse generator (Quantum Composers 9534) to the 15.3-MHz pump laser repetition frequency (f _REP) to produce electrical pulses at a sub-harmonic of f _REP. These pulses entered the modulation input of a radio-frequency (RF) driver (Gooch and Housego, A35110) which supplied a 110-MHz carrier to the AOM (Gooch and Housego, M110-2H/GJ), and their delay (65.2 ns) and duration (35.4 ns) were optimized to ensure complete dumping of a single resonant pulse. To confirm that the AOM was only active on a single pulse traveling towards the OC, a pair of photodiodes was employed. One monitored the beam dumped as it traveled from M₆ to M₇, while the second was aligned to detect accidental dumping of light traveling from M₇ to M₆, which occurred if the arrival time or duration of the electrical pulse activating the AOM was wrong. This approach guaranteed the most efficient energy extraction.

The OPO included active wavelength stabilization as described in [9], enabling stable long-term operation and achieving a relative-intensity noise (RIN) for the output-coupled pulses approaching that of the Yb:fiber pump laser. The feedback loop sensed the OPO wavelength by detecting the non-phase-matched second-harmonic signal light (λ ~760 nm) leaking through mirror M₂. This beam was dispersed by grating G₁ and focused by lens L₂ onto a position-sensitive detector (PSD) to derive an error signal which was low-pass filtered (LPF) at 15 Hz before entering a proportional-integral controller (PI-C) and a high voltage amplifier (AMP), whose output actuated mirror M₉ via a piezoelectric transducer (PZT).

3. Cavity-dumping performance

3.1 Pulse measurements

The OPO produced pulses, which were centered in wavelength around 1.52 µm, and tunable over only a limited range because of opposing constraints associated with the reflectivity profile of the available mirrors and the periods of the gratings on the MgO:PPLN crystal. The OPO, dispersion-compensated by a SF10 prism pair, operated with a net cavity group-delay dispersion (GDD) estimated to be ~-1000 fs², which ensured the generation of clean, near-transform-limited pulses.

An example of an autocorrelation measured at a dumping frequency of 3.06 MHz (f _REP/5) is shown in Fig. 2(a) , and its corresponding spectrum appears in Fig. 2(b). The pulse duration was estimated by adding quadratic, cubic and quartic spectral phase to the measured spectrum to find a pulse whose autocorrelation corresponded best with the experimental measurement. The red line in Fig. 2(a) shows the resulting fit produced by the pulse shown in Fig. 2(c), whose FWHM duration is 228 fs. The result shown was obtained for a pulse energy of 280 nJ, which implies a peak power of 0.94 MW.

 

Fig. 2 (a) Autocorrelation and (b) corresponding spectrum of the cavity-dumped signal pulses, recorded at a dumping frequency of 3.06 MHz. The best-fit autocorrelation envelope, shown as the dashed red line in (a), corresponds to the intensity (solid black lines) and phase (dashed green lines) profiles shown in (b) and (c). The results shown here were acquired from the dispersion-compensated OPO.

Download Full Size | PDF

3.2 Power, energy and efficiency measurements

Figure 3 shows the signal pulse sequence detected after the output-coupler (T = 1.75%) with an InGaAs photodiode (Thorlabs, DET10C/M). The blue pulses are the fundamental pulses at f _REP and the cavity was dumped at a frequency of 154 kHz (f _REP /99). The data indicate a dumping efficiency of 83%, and the red line shows an exponential fit corresponding to time of 1.38 µs for the intracavity pulses to recover to 90% of their steady-state energy. The data in Fig. 3 provide a convenient estimate for the unsaturated gain, implying a value of ~1.5, consistent with results from an earlier (unpublished) Rigrod analysis [10].

 

Fig. 3 Signal pulse sequence measured after the output coupler at a dumping frequency of 154 kHz. The dumping efficiency is determined from the change in the signal before and after dumping (red arrow).

Download Full Size | PDF

Cavity dumping was studied for dumping frequencies from 101 kHz (f _REP /151) to 3.83 MHz (f _REP /4) and data were recorded with the OPO operated with and without the dispersion-compensating SF10 prism pair. Figure 4(a) shows the average output power and pulse energy from the system, and Fig. 4(b) shows the same data re-plotted to show the peak power of the cavity-dumped pulses. Peak powers were derived from the estimated pulse profile (Fig. 2(c)) in proportion to their individual pulse energies for the OPO with SF10 prisms. The peak powers for the uncompensated cavity were derived from an estimated pulse duration of 410 fs (sech² temporal shape), inferred from the 775 fs FWHM of their interferometric autocorrelation (not shown here). Results with and without dispersion control are not directly comparable since the data from the OPO without prisms were obtained using a different pulse generator (Berkeley Nucleonics, BNC575) operated in trigger mode. Operated with the SF10 prisms the highest cavity-dumped pulse energy was 617 nJ, slightly lower than the highest energy of 650 nJ recorded from the OPO operated without the SF10 prisms. Pulse break-up in the absence of dispersion management meant that without prisms the OPO produced longer pulses with lower peak powers. The data in Fig. 4(b) show that including dispersion control resulted in substantially higher peak powers (2.07 MW) than the best that could be obtained without prisms (1.4 MW).

 

Fig. 4 (a) Cavity dumped average power (blue triangles) and pulse energies (red circles) for dumping frequencies from 101 kHz – 3.83 MHz. (b) Cavity dumped peak power (blue triangles) and pulse energies (red circles) for the same dumping frequencies. In both figures the solid and dashed lines indicate, respectively, results obtained from the OPO operated with and without SF10 prisms for dispersion control.

Download Full Size | PDF

The maximum average power extracted in the cavity-dumped beam was 1.1 W at a frequency of 3.06 MHz. This power is only slightly lower than the highest obtained through the OPO output coupler (1.2 W), implying that the cavity dumper introduced only minor parasitic losses. With an improved pulse generator, dumping at f _REP would be possible, presenting the possibility of using the dumper as a variable output coupler. The calculated pulse energies were corrected to a lower value than that inferred directly from the diffracted average power because any pulse traveling from M₇ to M₆ experiences parasitic reflections from both faces of the AOM that contribute to the power measured in the diffraction direction. Acquiring simultanously the power transmitted through the output coupler, as well as the reflected power from the AOM in steady state (AOM inactive) and while in operation, allowed us to subtract the residual power based on the output coupled power. All the results shown are based on corrected values and therfore represent the real single-pulse energy and peak power respectively. This correction is only necessary if a linear cavity is employed since a ring cavity would not show any residual power while the AOM was inactive.

3.3 Beam quality

Diffraction from an AOM occurs throughout the bulk of the modulator, not at a single plane as in a conventional diffraction grating. For this reason, it is important to quantify the effect of cavity-dumping on the beam quality of the dumped pulses. We measured the beam-quality parameters of the output-coupled and cavity-dumped beams in both the horizontal (X) and vertical (Y) directions. The horizontal direction corresponds to the diffraction direction and is therefore expected to be more severely affected by cavity dumping. For the output-coupled beam we recorded beam quality factors of $M_X^2=1.12$ and $M_Y^2=1.06$ (data not shown here), while for the cavity-dumped output we measured slightly higher values of $M_X^2=1.26$ and $M_Y^2=1.16$. In each case, the data were acquired by using the scanning knife-edge technique to measure the 1/e ² beam radius at regular intervals after a 50-mm focal-length lens. The beam radius at each position was determined from the average of 3 independent knife-edge scans, and Fig. 5 presents the experimental data fitted using a procedure similar to that described in [11]. The measurements were recorded at a dumping rate of 3.06 MHz and correspond to pulse energies of 280 nJ.

 

Fig. 5 (a) Horizontal and (b) vertical beam radius measurements (circles) and fit to an M ²-corrected Gaussian-beam propagation equation (solid lines), with M ² = 1.26 (horizontal) and M ² = 1.16 (vertical). The insets show horizontal and vertical waist radii of 11.3 µm and 10.9 µm respectively.

Download Full Size | PDF

3.4 Relative intensity noise

We recorded RIN performance over a frequency range of 4 mHz - 500 Hz for the pump laser, the output-coupled OPO signal pulses (cavity-dumper turned off) and the cavity-dumped OPO signal pulses at a dumping frequency of 3.06 MHz (f _REP /5). For each case, Fig. 6 shows the RIN data (left axes) and the total noise power integrated from high to low frequencies (right axes), whose maximum value is equivalent to the time-domain root-mean squared (RMS) noise. The acquisition time for each measurement was 250 seconds, which provided sensitivity to medium- and long-term variations of the kind associated with acoustic and thermal noise. Comparing the measurements, we found that the RMS noise on the output-coupled OPO pulses was 1.11 times the pump RMS noise, while the RMS noise of the cavity-dumped pulses was 2.13 times the RMS noise of the pump. In absolute terms, the RMS noise of the cavity-dumped OPO was still very low, with a value of 8 mdBc corresponding to 0.2% RMS noise.

 

Fig. 6 (a) Comparison of the RIN measured for the pump laser (blue) and the output coupled (T = 22%) OPO signal pulses (red) at the maximum average output power of 1.2 W. (b) Independent RIN measurement of the pump laser (blue) and comparison with the cavity-dumped (3.06 MHz) OPO signal pulses (red) at the maximum average output power of 1.1 W. On both graphs, the right axis shows the RIN integrated from high to low frequencies.

Download Full Size | PDF

4. Summary and conclusions

We have demonstrated sub-250-fs cavity-dumped pulses from an ultrafast OPO operating around 1.52 µm and with maximum energies of 650 nJ and maximum peak-powers of 2.07 MW. The system maintained pulse energies ≥ 392 nJ and peak powers ≥ 1.3 MW over a repetition range from 101 kHz – 3.06 MHz. The high pulse repetition frequency, excellent beam quality (M ² ~1.2) and RIN characteristics (8 mdBc) of the system make it immediately applicable in the fabrication of ultrafast-laser-inscribed waveguides in materials whose optical transmission properties are incompatible with inscription using 800-nm Ti:sapphire or 1040-nm Yb-doped lasers.