1. Introduction

Optical parametric chirped pulse amplification (OPCPA) has been the subject of much research over the past decade due to its large single-pass amplification, broad amplification bandwidth, high contrast, and the lack of stored energy in the amplifying crystal [1, 2]. These amplification qualities make OPCPA an ideal method for producing high repetition rate, high energy ultra-short pulse trains. The Extreme Light Infrastructure - Beamlines is currently constructing a 1 kHz repetition rate beamline with a design pulse energy of 100 mJ and pulse duration of 20 fs based entirely on OPCPA amplification, which will be used to produce high quality XUV and soft x-ray sources.

With the advent of fiber lasers and Yb-doped crystal-based lasers capable of producing high energy pulses with durations on the order of 1 ps and shorter, several groups have begun researching short pulse OPCPA pumped at 515 nm [3–5]. The advantage of amplification with ps pulses is the increased damage threshold of the OPCPA crystal [6] and the higher achievable fields and shorter crystals which result in larger amplification bandwidth. Additionally, signal pulses stretched to only a few picosecond can be compressed with chirped mirrors, eliminating the need for grating compressors. As demonstrated in the pioneering work at Max Planck Institute of Quantum Optics, Yb:YAG thin disk amplifiers are ideal for pumping picosecond OPCPA [3, 7]. Due to their low quantum defect and sub-millimeter thickness, Yb:YAG thin disk amplifiers are capable of handling high average powers and can amplify pulses to > 10 mJ at repetition rates above 1 kHz with minimal nonlinearity and excellent beam quality. Even after such amplification, the bandwidth of the Yb:YAG is broad enough to support pulses compressed to less than 2 ps.

Parametric amplification using short pulses puts strict requirements on the synchronization of the pump and seed pulses. Timing jitter between pulses on the order of mere hundreds of femtoseconds results in significant fluctuations of the OPCPA amplified output; the required level of synchronization cannot be achieved through electronic synchronization. To provide some degree of passive synchronization between the pump and broadband signal pulses both can be produced in the same oscillator [3, 5, 7, 8]; however, in most cases, this passive synchronization alone is not sufficient for stable amplification as the accumulated path length of pump pulses during amplification in a regenerative amplifier can be roughly 1 km. Considering the large number of roundtrips, thermal drift, air turbulence and vibrations cause significant synchronization instabilities.

For long-term, stable operation of the OPCPA, active stabilization of the relative delay between the pump and signal pulses is required. Active optical synchronization of laser pulses to relative jitter levels of σ = 1 fs and below has been achieved in the stabilization of fiber links [9,10] or in the synchronization of two femtosecond oscillators [11,12]. In the context of synchronizing pulses for picosecond OPCPA at kHz repetition rates, roughly 20 fs RMS jitter has been achieved and shown to be sufficient for stable, long-term OPCPA amplification [13]; for higher repetition rate (at 300 kHz), 2 fs RMS jitter has been achieved, as mentioned in [14].

Here we present an optical synchronization system which has comparable performance to the system described in [13], and the advantage of requiring signal energies over 4 orders of magnitude lower to produce a robust error signal. One key difference between previous schemes and the scheme described here is the use of parametric amplification rather than sum frequency generation (SFG) as the nonlinear process used to detect the relative delay of pulses. By using parametric amplification, it is straightforward to derive a strong error signal from mirror leakage of both the pump and signal pulses, so virtually all the broadband pulse energy from the oscillator, in our case 2 nJ, can go to seeding the OPCPA without any pre-amplification. Using the frequency doubled pump pulses at 515 nm for the stabilization, rather than the 1030 nm pulses from the regenerative amplifier, also allows the delay to be measured very close to the OPCPA stage.

2. Experimental setup

Figure 1 shows a schematic of the system. The octave-spanning Ti:sapphire oscillator (Femto-lasers Rainbow series) provides two pulse trains at a repetition rate of 80 MHz. The first output is broadband, centered at 800 nm with a pulse energy of 2 nJ; the second output is a seed for the pump beam, centered at 1030 nm with a pulse energy of 1 nJ (after amplification in fiber).

 

Fig. 1 Overal layout of the system setup. A femtosecond, octave-spanning Ti:sapphire oscillator provides a broadband seed for OPCPA as well as seed at 1030 nm for a regenerative amplifier. Mirror leakage of pump (green dashed line) and signal (red dash-dotted) beams is sent to jitter stabilization system. Coarse corrections are fed back to a motorized translation stage, while fine corrections are fed back to an intra-cavity piezo-mounted mirror in the regenerative amplifier. Performance of the jitter stabilization system was measured independently of the loop by the reference jitter measurement.

Download Full Size | PDF

The 1030 nm output is stretched by 250 ps/nm by a chirped fiber Bragg grating (TeraXion) and collimated. After collimation, the repetition rate is reduced to 1 kHz with a KD*P Pockels cell and the pulses are amplified to 28 mJ in a home-built regenerative amplifier using an Yb:YAG thin disk head from TRUMPF Scientific Lasers (TruMicro 5000 series). The pulses are then compressed to 1.8 ps and frequency doubled to 515 nm in a 1.5 mm length LBO crystal. A relative energy stability of the pulse train at 515 nm was measured to be 1.5 %. A mere 14 μJ of 515 nm light, which is leaked through one of the turning mirrors, is sufficient to seed the jitter stabilization system. 80 μJ of the remaining 515 nm were used for reference jitter measurements.

The broadband pulses from the oscillator would be normally directed toward the OPCPA with low-GDD dielectric mirrors which are transmissive below 700 nm, as we are interested in amplifying of the spectral region only between 700 nm and 900 nm in the OPCPA. The excess energy in the 650 nm – 700 nm range (∼ 30 pJ), which would otherwise be wasted, is instead sent to the jitter stabilization system. In this case the broadband pulses were directed towards the reference jitter measurement setup.

While it is based on a different nonlinear process and the geometry is quite different, the principle of the jitter stabilization scheme is quite similar to that described in [9] for stabilizing lengths of fiber. The concept of the jitter measurement is shown in Fig. 2(a) and is similar to that described in our presentation of the preliminary performance of the OPCPA pump laser [15]. A pump pulse at 515 nm is converted into two orthogonally polarized pulses, relatively delayed by a fixed time τ. The delay τ is introduced either by increasing the length in one arm of a Mach-Zehnder interferometer (jitter stabilization system) or by inserting a 1.9 mm long calcite crystal (reference jitter measurement). The pump pulses are collinearly overlapped with a diagonally polarized signal pulse at 675 nm and focused into two orthogonally oriented BBO crystals (3 mm length). The nonlinear interaction of the pulses is shown in Fig. 2(b). The crystals are cut for type I collinear parametric amplification at these wavelengths and the horizontal and vertical components of the signal are amplified independently in different crystals. The parametric gain of each polarization component of the signal depends on the relative delay Δt between the pump and signal pulses. Separating the horizontal and vertical components of the signal (or idler) and registering their difference in amplitude on a balanced photodetector, we can extract a dispersive signal which can be used for locking the delay between signal and pump pulses. The detection of the idler beam benefits from the absence of the background, however, the its wavelength range (around 2 – 2.1μm) is less convenient to work with in terms of detection. The dependence of the difference signal on the delay Δt is shown in Fig. 2(c).

 

Fig. 2 (a) Scheme of the optical part of the jitter stabilization system. In the reference jitter measurement setup, the Mach-Zehnder interferometer was replaced by the calcite crystal. Green dashed line, pump beam; red solid line, signal beam; λ/2, half wave plate; PBS, polarizing beamsplitter; τ, temporal delay introduced by a path difference in the interferometer; DM, dichroic mirror; NLCs, nonlinear crystals; FLT, color glass filter (OG550); PD, photodiode. (b) A concept of the nonlinear interaction. (c) A calibration curve – the dependence of the difference signal on the delay Δt.

Download Full Size | PDF

The signals from the balanced photodiodes are captured using a 50 MS/s digitizer (NI 5751) using a sampling clock generated by the electronic timing system (MRF EVR-300), which is synchronized with the laser repetition rate. Signal processing is done via FPGA (NI PXIe-7961R) with monitoring and supervision by a LabVIEW Real-Time based controller. The digitized balanced photodiode pulse is gated and numerically integrated to provide a single-shot error measurement. A digital PI loop feeds the corrections back to a piezo holding a cavity mirror of the regenerative amplifier. Mounting the piezo within the cavity allows small spatial displacements of the mirror to be amplified by the number of round trips in the amplifier (roughly 100 roundtrips) increasing the range of delays between pump and signal the piezo can correct. The voltage on the piezo is monitored and, if it wanders out of a specified range, a motorized actuator (Newport, Picomotor) on a translation stage before the regenerative amplifier engages and allows to bring the piezo voltage back to the middle of its operational range. The combined travel range of the piezo and picomotor corresponds to about ±85 ps.

3. Performance of jitter stabilization system

We have evaluated the performance of this jitter stabilization scheme from both in-loop and reference signals, independently calibrated by curve in Fig. 2(c). Figure 3(a) shows the jitter between pump and signal, as measured in the jitter stabilization system. For this measurement the feedback loop was on and the error signal was fed back into the PI loop (in loop operation). Figure 3(b) shows the corresponding jitter, as measured in the reference jitter measurement, which was operating out-of-loop. In both cases while the loop on, the relative delay between pulses remains fixed.

 

Fig. 3 Relative delay between the pump and the signal pulse in time. (a) Jitter measured in loop, (b) jitter measured out-of-loop, (c) jitter measured with the loop disabled.

Download Full Size | PDF

Due to a drift of the beam pointing and pump power, the slope of the calibration curve was not precisely constant during the measurement. Based on a repeated measurement of the calibration curve before and after the measurement, we estimated a relative error of the calibration for measurement to be ∼20 %. This error should be substantially suppressed with an implementation of active beampointing stabilization. However, the lock is quite robust and does not fall for hours; while locked, the system is resistant against external perturbations. We calculate the jitter during this test to be σ = 17 ± 3 fs, which is <1% of the pump pulse duration. We measure the baseline noise to be ∼ 3 fs RMS (∼ −25 dB·fs/Hz), and the high gain-bandwidth and low common-mode rejection ratio of the detector contribute about ∼ 1.6 fs to the measured jitter. The performance of our setup is comparable to the jitter level achieved by the system reported in [13]. Similar results were reported in [15] with the similar pump pulse duration. Since then, the noise spectrum has been substantially improved.

To evaluate the degree to which active jitter stabilization is required for high quality synchronization we measure the free running relative drift of pump and signal pulses. When the loop is off, as shown in Fig. 3(c), the relative delay fluctuates by hundreds of femtoseconds over just several minutes. Because the pump and signal pulses we intend to synchronize are < 2 ps in duration, such delays would influence the spectral stability and efficiency of OPCPA.

Inspection of the relative delay power spectral density (PSD) of the unlocked system in Fig. 4(a) shows several small noise spikes across the spectrum. These spikes are present with the feedback both on and off, so we don’t expect this to be due to piezo resonance. The origin of those spikes is probably either acoustic or electronic. Based on PSD of the timing measurements which are presented in Figs. 4(b), 4(c) and 4(d), the bandwidth of the lock is roughly 30 Hz, which is about one decade below the Nyquist frequency (500 Hz) for this 1 kHz system. By comparing Figs. 4(c) and 4(d) the reduction of the PSD within the loop is attributed to the reduction of the signal noise and not the actual jitter. This is more evident for frequencies below 15 Hz. Since the current mechanical setup limits the frequency of the feedback to be below 100 Hz, higher acoustic frequencies remains uncompensated; our system is primarily compensating for slow thermal fluctuations or delays resulting from air currents. A different method of mounting the mirror to the piezo or a lighter mirror could be used to increase the bandwidth of the loop, but the current performance of the system is suitable for its purpose.

 

Fig. 4 (a) Power spectral density (PSD) of the pulse delay jitter when the feedback loop is off. (b) Low frequency part of PSD, loop off. (c) Low frequency part of PSD while the feedback is on, measurement out of loop. (d) Low frequency part of PSD while the feedback is on, measurement within loop. Peaks in (c) and (d) around 5 Hz are caused by the motion of the picomotor.

Download Full Size | PDF

4. Conclusion

We have demonstrated a jitter stabilization system capable of stabilizing relative delay between pump and signal pulses to less than 1% of the pump pulse duration. The lock is stable against normal environmental perturbations and has been tested to remain reliably locked. The performance is on par with previously reported systems used for picosecond OPCPA, but the system described here has the advantage of requiring broadband input pulse energies over four orders of magnitude lower.

The effective bandwidth of the system is far below the Nyquist frequency; this means that there is still room for improvement in the implementation of the jitter stabilization system. The next steps would be testing of a piezo with larger bandwidth and changing the mechanical design, so that we could increase the locking bandwidth.

Technical improvements aside, the principle of a small signal parametric amplification-based jitter stabilization system has been shown to be a simple, stable, and effective means of stabilizing relative pulse delays between picosecond pulses.