1. Introduction

The recent emphasis on developing ultrafast oscillators that can directly provide high-energy pulses at multi-MHz repetition rates has been driven by their potential to replace more complex and expensive ultrafast amplifier systems in certain applications. In particular, refractive index modification in common dielectric materials has been shown to be possible using pulse energies of only a few nJ, with a 25-MHz Ti:sapphire laser oscillator being used to write waveguides inside borosilicate glass, with a modification threshold of only 5 nJ [1]. Systematic studies now provide strong support for implementing laser waveguide writing at repetition rates above 1 MHz [2], where a reduction of the modification threshold energy from 750 nJ to 80 nJ has been shown to be possible simply by increasing the repetition rate from 200 kHz to 2 MHz. High-energy multi-MHz repetition-rate ultrafast oscillators, some including cavity-dumping elements [3], are the sources of choice for waveguide inscription in this regime, and their development has been technically enabled by designs incorporating an extended cavity scheme using an intracavity Herriott-cell [4,5] or relay-imaging mirror sequence [6]. Such laser systems have been limited to wavelengths below 1250 nm, relying either on Ti:sapphire [4,5], Nd-doped [7], Yb-doped [3] or Cr⁴⁺-doped [6] gain materials.

Many infrared materials, which are interesting candidates for ultrafast waveguide inscription, cannot be inscribed at short wavelengths, and require irradiation at wavelengths considerably longer than 1 μm. For example, inscription in silicon requires pulses longer than 1.1 μm, germanium 1.8 μm and chalcogenide glasses from 0.6 - 1.5 μm. High-energy near-infrared oscillators are therefore required for this application, particularly those with diffraction-limited outputs that permit the lowest modification threshold to be achieved by using high-numerical-aperture focusing [1]. Synchronously-pumped optical parametric oscillators (OPOs) offer a route to extending the wavelength coverage of high-energy ultrafast oscillators to the near and mid-infrared, and a cavity-dumped high-repetition-rate (82 MHz) OPO has been reported which produced 1 MHz, 90 nJ pulses at 1200 nm [8]. A fiberfeedback OPO, synchronously pumped by a 58 W thin-disc laser [9], achieved 56 MHz, 339 nJ at 1450 nm, although the weak cavity feedback associated with this geometry led to a non-diffraction-limited (M² = 1.6) output beam.

In the present work, we report an extended-cavity OPO operating at 15.3 MHz and pumped by a femtosecond Yb:fiber master-oscillator power-amplifier (MOPA). An Yb:fiber MOPA represents an inexpensive high-energy source, where the repetition rate can be readily reduced to frequencies as low as 10 MHz. Previous work reported an Yb-fiber-amplifier-pumped femtosecond OPO that produced 1.7 nJ pulses at 1550 nm [10]. By using intracavity relay imaging in a conventional free-space resonator we now demonstrate energy scaling up to 72 nJ, corresponding to the highest energies directly produced from a free-space ultrafast OPO and representing a promising new source for waveguide inscription in infrared materials.

2. Optical parametric oscillator configuration

The OPO was pumped by a commercial sub-picosecond Yb:fiber laser (Fianium FemtoPower 1060-10), which operated at a centre wavelength of 1064 nm with a pulse-repetition frequency of 15.3 MHz. This laser is a simple architecture based on a picosecond master oscillator followed by a power amplifier stage in which the picosecond pulses acquire a broad bandwidth through self-phase modulation. The resulting chirped pulses had durations of 7 ps with a bandwidth of approximately 25 nm and energies of around 700 nJ. We compressed the pulses using a variable compressor comprising a pair of fused-silica transmission gratings (Ibsen Photonics). Although the chirp on the output pulses was not fully compressible, the grating compressor achieved pulse durations of 437 fs (FWHM) with an efficiency of 63.5%.

The OPO was based on a MgO-doped periodically-poled LiNbO₃ (MgO:PPLN) crystal consisting of four uniform gratings with periods ranging from 28 - 31 μm and with a length of 1 mm. The crystal was fabricated with plane-parallel faces which were anti-reflection coated over a wavelength range from 0.8 - 1.55 μm. To permit temperature tuning, and as an added precaution against photorefractive damage, the crystal was mounted in an oven and its temperature was maintained at 85°C. Several factors influenced the choice of focusing into the crystal. To minimize the risk of surface damage to the MgO:PPLN crystal we chose a beam radius of 40 μm (defined at 1/e² intensity point) which represented a compromise between parametric gain and damage considerations. This focal size was also readily compatible with the minimum spot size to which the pump pulses could be focused. The pump laser exhibited a beam quality factor of M² ~ 1.3, which originates from its use of 50 μm-diameter large-mode-area fiber within the Yb:fiber amplifier, and is poorer than that obtained from solid-state lasers such as Ti:sapphire that are more commonly used as pump sources. By characterizing the pump beam we achieved the required focal diameter by using a single 100 mm-focal-length lens placed 750 mm after the laser head. The pump beam was introduced into the OPO cavity by pumping through one of the OPO focusing mirrors, as illustrated in Fig. 1. The OPO resonator was based on an asymmetric “V-cavity” design, with the focusing section around the crystal comprising a concave end-mirror with a radius of 150 mm and a turning mirror with a radius of 200 mm, which collimated the intracavity beam into an arm of length 9.55 m. Synchronous pumping required the OPO cavity length to be matched to the pulse repetition rate of 15.3 MHz, resulting in a total length of 9.8 m, measured between the curved end-mirror (M1) and the output coupler (M8). Following the extended-cavity strategies used successfully to scale the pulse energies obtained from Ti:sapphire laser oscillators [4,5], we used relay imaging – implemented using two pairs of 2000 mm radius concave mirrors – to achieve a long resonator which was stable and produced an intracavity beam with a diameter less than 3.4 mm at all points in the cavity. The M² of the beam was characterized and found to be 1.07 (1.11) in the horizontal (vertical) directions, which compares very well with the lower beam quality from a competitive high-energy OPO based on a low-finesse cavity [9]. All mirror-folding angles were small (< 2°), minimizing astigmatism caused by the curved mirrors.

 

Fig. 1. The OPO and pump optics, including a compressor (C), a variable attenuator, comprising a half-wave plate (λ/2) and polarizing beam splitter (PBS), and a focusing lens (L). X, MgO:PPLN crystal; M1 and M2, concave focusing mirrors of radii 150 mm and 200 mm respectively; M3 - M6, concave relay-imaging mirrors with radii of 2000 mm; M7 and M8, plane high-reflectivity and output coupling mirrors respectively.

Download Full Size | PDF

3. Performance characterizations

3.1 Pump depletion and conversion efficiency

The pump depletion was investigated under maximum output power, and a comparison of the pump spectra measured after the crystal with the OPO oscillating and blocked is shown in Fig. 2. Strong depletion was observed between 1057 nm and 1067 nm, with evidence of back-conversion from the signal to the pump at 1056 nm. The pump acceptance bandwidth of our 1 mm-long MgO:PPLN crystal is calculated to be around 80 nm, so does not explain the confinement of the pump depletion to only a 10 nm wide band; instead the limited conversion is attributed to incompressible spectral phase distortion arising from the self-phase-modulation of the pump pulses in the Yb:fibre amplifier. This phase distortion is also responsible for the back-conversion at 1056 nm. A conversion efficiency based on the difference in the areas under both spectra was calculated to be 36%.

 

Fig. 2. Spectra of the depleted pump (gray fill) and undepleted pump (yellow fill). The intensity scale is normalized to the undepleted spectrum.

Download Full Size | PDF

When the OPO was operated with a 22% output coupler the maximum output power at a centre wavelength of 1535 nm was 1.09 W for a pump power of 6.4 W, implying a signal extraction efficiency of 17.0% and an idler extraction efficiency of 7.6%. The idler efficiency was inferred using the Manley-Rowe relations, taking an idler wavelength of 3.42 μm. The difference in the pump depletion and the total extraction efficiency indicates the parasitic loss in the cavity. By taking the reflectivity of the cavity mirrors to be 99.9% we obtained a reflectivity loss for each cavity roundtrip of 1.5%, leaving a 10% loss at the MgO:PPLN crystal which can be explained by a 2% residual reflectivity at the antireflection-coated faces.

The slope efficiency, measured with a 22% output coupler was determined to 22.2% with an estimated pump threshold of 1.25 W, as indicated in Fig. 3.

 

Fig. 3. Signal output power (solid circles) as a function of pump power for a 22% output coupler, and a linear fit through the data (blue line), extended to cross the abscissa. The slope efficiency was determined to be 22% and the pump threshold was estimated to be 1.25 W.

Download Full Size | PDF

The optimal output coupling was examined by recording the maximum extractable average power for output coupler transmissions of 10 %, 22 %, 35 %, and 40 %, which led to output powers of 980 mW, 1080 mW, 1040 mW, and 982 mW, respectively. Following, the analysis of Siegman [11], which provides an analytic expression relating the total output power to the oscillator gain and the cavity losses, we inferred an optimum output coupling of 24 %, close to the value used for the measurements presented in this paper. The corresponding 7.5% cavity losses inferred from this analysis are in satisfactory agreement with the estimated 10% losses implied by the pump depletion and extraction efficiency results presented above.

3.2 Intensity noise

The use of an extended cavity for a synchronously-pumped OPO raises the question of whether the system may be vulnerable to environmental noise because of the greater number of mirror mounts and longer free-space path travelled by the intracavity beam. By using silicon and InGaAs photodiodes, we monitored the intensity noise present on the pump pulses (Si) and the OPO signal pulses (InGaAs). Figure 4 shows the power spectral densities of the intensity noise present on both outputs, with the frequency range from 1 Hz to 100 kHz being sufficient to include noise arising from acoustic sources and atmospheric turbulence along the beam path in the cavity. The data in Fig. 4 were recorded without any active cavity-length stabilization, and three independent measurements led to very similar results.

 

Fig. 4. Power spectral densities of the intensity noise on the pump (black) and the OPO signal (green) outputs, measured with Si and InGaAs photodiodes respectively. The right axis shows the cumulative intensity noise for the pump (red) and the OPO signal (blue). The mean output levels from the pump and OPO were normalized to 1 V for the analysis.

Download Full Size | PDF

While the noise measurement shows that the OPO pulses possess greater intensity noise than those from the pump laser, the absolute value of the noise is at a low level (<0.1%), and indistinguishable from the pump laser at frequencies higher than those typically associated with acoustic vibrations. Comparison of the cumulative phase noise shows a division at 2 kHz, increasing at lower frequencies. An implementation of active cavity-length stabilization [12] with a bandwidth of at least 2 kHz is expected to reduce the noise in the acoustic region significantly. In a boxed configuration, the output power of the OPO was very stable and only exhibited a slow drift as the lab temperature changed.

3.3 Signal pulse characterizations

We characterized the signal pulses using interferometric autocorrelation performed using twophoton absorption in a silicon photodiode [13], and recorded the signal spectrum simultaneously with the autocorrelation traces. The spectra of the OPO signal pulses showed significant structure, indicating that their shapes could not be described by the Gaussian or sech²(t) intensity profiles commonly used to infer pulse durations from autocorrelation measurements. For this reason, we estimated the signal pulse durations from a fit to the interferometric autocorrelation, based on adding quadratic and cubic spectral phase to the measured spectrum. Fig. 5(a) presents an interferometric autocorrelation recorded at a signal energy of 72 nJ and corresponding to the spectrum shown in Fig. 5(d). The autocorrelation shown in Fig. 5(c) represents the best fit to the experimental autocorrelation obtained by optimizing the amount of quadratic and cubic phase added to the spectrum in Fig. 5(d). The temporal intensity of the corresponding pulse is shown in Fig. 5(b), and the green line in Fig. 5(d) shows the spectral phase of this pulse. The duration of the signal pulses inferred using this method was 1.50 ps, which was 1.8 times the transform-limited duration of 840 fs. Although this approach is not fully unambiguous, it allows a better estimate of the pulse duration than can be obtained by a naive assumption of a standard analytic pulse shape.

 

Fig. 5. Experimental blue (a) and fitted red (b) interferometric autocorrelation, indicating a pulse duration of 1.5 ps. The temporal intensity of the pulse (c) calculated from the measured spectral intensity and fitted phase in (d). Data were obtained at 1080 mW signal power.

Download Full Size | PDF

The operating regime in which the results of Fig. 5 were recorded corresponded to one that produced narrowband pulses with a modest amount of chirp, indicated by the fringe coherence prevailing into the wings of the autocorrelation. Tuning in the highest energy region was limited, however tuning from 1420 - 1560 nm was obtained by using grating and cavity-length adjustment. A second regime was accessible at longer wavelengths by adjusting the cavity length by a small amount. In this configuration, the OPO produced signal pulse energies of 62 nJ, and the corresponding signal autocorrelation and spectrum are shown in Fig. 6. A broadening of the spectrum is visible (Fig. 6(d)) with an intensity drop at 1533 nm between two local maxima around 1528 nm and 1537 nm. A similar analysis to that used for the shorter-wavelength signal data implied signal durations of 1.56 ps, a factor of 3.9 times the transform-limited duration of 400 fs.

 

Fig. 6. Experimental blue (a) and fitted red (b) interferometric autocorrelation, indicating a pulse duration of 1.56 ps. The temporal intensity of the pulse (c) calculated from the measured spectral intensity and fitted phase in (d). Data were obtained at 980 mW signal power.

Download Full Size | PDF

4. Summary and conclusions

We have reported a synchronously-pumped MgO:PPLN OPO operating at, to our knowledge, the lowest repetition-rate (15.3 MHz) of any ultrafast OPO, and producing the highest pulse energies (72 nJ) obtained from a free-space resonator design. Although uncharacterized in this work, the 3.5 μm idler pulses from the OPO are expected to have energies of up to 30 nJ. The OPO is expected to be well suited for ultrafast refractive index modification of infrared materials, including chalcogenide glasses and semiconductors, and other applications including multi-photon microscopy and coherent anti-stokes Raman spectroscopy.