1. Introduction

Ever since their first demonstration at the turn of the century, carrier-envelope phase (CEP) stable femtosecond oscillators have brought on a revolution in metrology [1–3]. Phase-stable amplification of these pulses has, in turn, become a very important tool in attosecond science, especially in the production of single attosecond pulses [4–6]. By now, phase-stable amplification has been demonstrated in several types of laser systems based on chirped-pulse amplification (CPA), from systems having material, prism or transmission grating stretchers and compressors [7–9] to grating based systems that can be scaled to very high energies [10–16]. CEP stable multi-millijoule CPA systems have also recently become commercially available. These systems have been built with CEP stability in mind, contrary to equipment often present in many laboratories. Therefore, practical, uncomplicated modifications that can add CEP stability to existing systems are of great interest to a large scientific community.

This article reports on the successful CEP stabilization of a relatively large 1-kHz Ti:Sapphire-based CPA laser system consisting of a commercial CEP-stable seed oscillator, a large Öffner triplet grating stretcher, a regenerative amplifier, a cryogenically-cooledmultipass amplifier and a grating compressor. CEP stability for pulse energies up to 6 mJ is demonstrated. This upgrade has been done non-invasively, that is, without causing long interruptions to scientific experiments conducted with the laser.

2. Description of the laser system

The original CPA system (35 fs, 1.5 mJ) was delivered in 1998 (B.M. Industries/Thales). Prior to this work, a considerable overhaul was performed in 2006 when the amplifiers were upgraded to give 10 mJ per pulse before compression (Amplitude technologies). The present system layout is shown in Fig. 1(a). The seed oscillator is a CEP-stable Femtosource Rainbow, which provides very stable operation due to its rigid construction and collinear f-to-0 setup. The oscillator uses its broad bandwidth together with self-phase modulation in a periodically-poled lithium-niobate crystal to achieve the octave spanning spectrum required in the self-referencing f-to-0 interferometer without the need for a photonic crystal fibre [17]. Phase stabilization electronics (XPS800-E, Menlosystems) stabilizes the oscillator pulse-to-pulse phase slip to π/2 with a resulting in-loop phase jitter well below 100 mrad root-mean-square (RMS).

 

Fig. 1. (a) Simplified laser setup: AOM: acousto-optic modulator; OSC: oscillator; APD: avalanche photodiode; PD: photodiode; AOPDF: acousto-optic programmable dispersive filter; ISO: optical isolator; SP: spectrometer; PC: personal computer. Electrical connections are indicated with dashed lines and slow loop feedback paths in purple. (b) Cryocooler setup in the multipass amplifier. The vibrating cryocooler is mounted on a separate tripod completely isolated from the optical table.

Download Full Size | PDF

For fine control of system dispersion, an acousto-optic programmable dispersive filter (AOPDF, Fastlite Dazzler) is used after the oscillator. With the low-jitter option installed, factory timing jitter measurements on this particular unit point to an added optical CEP jitter of 50 mrad RMS. Besides being useful for dispersionmanagement, the AOPDF can also be used to fine tune the carrier frequency of amplified pulses simply by changing the spectral shape of the seed pulses [18]. The stretcher is a large, folded, triplet Öffner grating stretcher with a 5-ps/nm stretch factor. The size of this stretcher is considerable compared to stretchers in commercial CEP stable systems. The diameter of the concave mirror is 30 cm, the grating is 14 cm×12 cm, the beam height is 20 cm above the table surface, the footprint is 120 cm×60 cm, and the total beam path is 11 meters. Needless to say, this device is extremely sensitive to vibrations.

The first of the two amplifiers is regenerative with three KD*P Pockels cells (pulse picker, switch in/out, and cleaner). Saturation at an output energy of 0.5 mJ is obtained after roughly 13 cavity roundtrips. The second amplifier is a standard cryogenically-cooled five-pass bow-tie amplifier that is designed to boost the pulse energy up to 10 mJ. Both amplifiers are pumped by light from a 30-W diode-pumped solid-state laser (DM30, Photonics Industries). For pulse energies above 5 mJ or so, a second flashlamp-pumped 20-Wlaser (YLF20W, B.M. Industries) is used for themultipass amplifier. Finally, the beam is sent over to the adjacent optical table and into a standard double pass grating compressor. With large dielectric gratings (14 cm×12 cm) optimized for 800 nmthe throughput is approximately 74%. In order not to damage the gratings, the beam is expanded to 1.6 cm full width at half maximum (FWHM) in a telescope before entering the compressor. Frequency-resolved optical gating measurements indicate a minimum pulse length of 34 fs FWHM after compression. By tuning the AOPDF this can be reduced to 30 fs if minor satellites to the main pulse are allowed. For reference, the autocorrelation trace assuming a sech² pulse shape yields 30 fs and 26 fs, respectively.

A self-referencing f-to-2f interferometer (APS800, MenloSystems) is used as a CEP detector for the amplified pulses. During amplification a slow drift of the CEP usually occurs, and this must be compensated by another feedback loop, the so called slow loop [7]. Several different implementations of the slow loop have been demonstrated, including feedback to the oscillator locking electronics [7], grating position [14] or prism position [8] in the stretcher, and grating separation in the compressor [15]. CEP control using an AOPDF has also very recently been demonstrated [9, 19]. In this work the slow loop has been implemented in two different ways: voltage feedback to the oscillator locking electronics and phase control of the acoustic wave in the AOPDF. Feedback paths are indicated in purple in Fig. 1(a).

3. Route to carrier-envelope-phase stability

Early on it was clear that the first problem to be encountered on a route to a stable CEP would be the cryogenic cooler assembly in the multipass amplifier. Gas pressure fluctuations in the cryocooler (ARS DE-104) generate a considerable amount of vibrations, and these must not propagate to the optical table. For this reason the cryocooler assembly was mounted on a separate tripod and is hovering a few millimeters above the optical table surface completely isolated from the rest of the setup (Fig. 1(b)). Later, during first tests after installation of the CEP-stable oscillator and the phase detector for the amplified pulses, no CEP stability was observed even with the noisy cryocooler turned off. What follows in this section is a short summary of the path to a stable and controllable phase: the encountered problems and their solutions.

The second problem encountered was electromagnetic interference in the oscillator fast loop and in the pulse picking caused by the high-voltage switching of the three Pockels cells. In the fast loop this interference manifested itself as clear and regular spikes in the error signal, whereas the interference in the pulse picking was more subtle. These interference problems where removed by re-routing cables, putting aluminum foil for added shielding, using a bandpass filter in front of the pulse counter input and by using separate pulse pick-up photodiodes for different parts of the system in order to remove potential ground loops and noise propagation. As a result a total of four separate photodiodes are used in timing and control. With these modifications CEP stability on the few second time scale was observedwith short stable periods interrupted by long periods of very turbulent behavior.

By inspection, i.e., by tapping on different parts while monitoring the CEP, it became clear that the stretcher seemed by far the most sensitive part with respect to vibrations. To investigate the role of table vibrations further, a piezo-electric transducerwasmounted on the optical tables, and it turned out that even veryweak floor vibrations could cause severe turbulence of the phase. Thus it seemed that vibrations in the floor were exciting resonances in the table, in the rigid table legs, and in the optomechanical components themselves. Since the system stretches over two separate optical tables, floating the tables did not seem like a quick and easy solution. Instead, special anti-vibration rubber sheets (Novibra) were placed under the table legs as well as under the cryocooler tripod in order to isolate the optical tables from floor vibrations with excellent results. Even though the anti-vibration sheets are not designed for demanding applications, they do remove efficiently the part of the noise spectrum that can induce resonances in the table and the table legs (50–100 Hz or so). Note also that the laboratory is located in the basement of the building and that there is no heavy traffic nearby, which makes the environment already quiet by default.

At this point the system could be regarded as phase stable; however, to further reduce the phase jitter and to be able to operate the flashlamp-pumped Nd:YLF laser, vibration isolating sheets (Sorbothane) were placed under the pump lasers. Also, rotary pumps, cooling water and cryocooler pumps, and pump laser drivers are all located in separate rooms away from the laser setup. Vibrating cooling water hoses have also been clamped down by a large concrete block before going onto the optical table.

4. Results

Fig. 2(a) shows single-shot, in-loop phase jitter measured at an output energy of 6 mJ with slow loop feedback applied to the AOPDF (red dots in the background) and the oscillator locking electronics (black dots). Here, approximately every fourth pulse has been recorded. In both cases the resulting noise has a gaussian distribution with a standard deviation of 510 mrad and 470 mrad, respectively (Fig. 2(b)). The electronic feedback has the advantage of providing a tight lock. There are, however, a number of disadvantages, including feedback range limitations and, especially, added noise to the oscillator locking electronics.

 

Fig. 2. (a) In-loop, single shot CEP jitter at a pulse energy of 6 mJ with slow loop feedback applied to the AOPDF (red dots in the background) and to the oscillator locking electronics (black dots). (b) Histograms of the data in (a) show gaussian distributions with a standard deviation of 510 mrad and 470 mrad. (c) CEP control using the AOPDF at an output energy of 4.2 mJ is demonstrated by spelling ”Lund”.

Download Full Size | PDF

The Dazzler feedback scheme, on the other hand, will never run out of range due to its cyclic nature, the feedback being just the phase of the acoustic wave that ranges from 0 to 2π. Furthermore, this technique will not disturb the oscillator in any way. In this work, the disadvantage was the slow update rate of the acoustic waveform (around 6 Hz or so), and the fact that only 32 different waveforms could be preloaded into the memory of the driver electronics, which resulted in a feedback step size of 2π/32=200 mrad. Therefore, the feedback value could differ from the correct one by up to 100 mrad. These disadvantages do not exist in newer models with much higher update rates and in which several hundred waveforms can be preloaded into memory. Nevertheless, suprisingly good results were obtained also with our device as can be seen in Fig. 2(c), where CEP control at an output energy of 4.2 mJ is demonstrated by spelling ”Lund”.

Other possible implementations of the slow loop include control of various grating, prism or wedge positions in the beam path. While all of these would probably work, they are based on mechanically moving parts and will sooner or later run out of range forcing a reset of the feedback and momentary instabilities in the phase; therefore, an AOPDF based approach that will neither disturb the oscillator nor will it ever run out of range is probably the method of choice for large laser systems such as the one discussed here where the slow loop needs to correct large drifts. A summary of CEP noise measured at different pumping conditions is given in Table 1. Again, the data is single-shot with approximately every fourth pulse recorded over a time span of 15 minutes, which is the typical time scale of an attosecond experiment.

Table 1. Measured in-loop phase jitter (mrad) for different feedback schemes and at different pumping conditions. The data is single-shot with approximately every fourth pulse recorder over a time span of 15 minutes.

View Table

5. Discussion

The results presented in the previous section are encouraging and would presumably result in observable CEP dependent effects in many experiments, but for certain applications, such as the production of single attosecond pulses using high-order harmonic generation, reduced CEP jitter would be desirable. One big contributor to CEP noise in the present system is the cryogenic cooler setup in the multipass amplifier (Fig. 1(b)). Even though it is hovering above the optical table surface, vibrations propagate via the tripod to the floor and to the table. But more importantly, the entire vacuum chamber vibrates, including the Ti:sapphire crystal and the windows. This has two consequences: first, the overall beam pointing stability is reduced and second, since the windows are tilted, the movement of the vacuum chamber results in a changing path length in vacuum (and air), and hence a change in phase. By turning off the cryocooler for a few seconds, a clear reduction in CEP jitter can be seen: the RMS jitter drops from the 400 mrad regime to the 300 mrad range. Thus, in order to increase the stability of the laser, a new design incorporating vibration dampers for the cryocooler is needed.

To further reduce the CEP jitter, additional layers of vibration damping material might be put under the pump lasers, the stretcher assembly, and also under the table legs. Active stabilization of pulse energy fluctuations [20] and an update of the AOPDF drive circuitry are other minimally invasive steps that can be taken for additional noise reduction.

6. Conclusions

In conclusion, it was shown that CEP stability can be added to a relatively large, and partly old, CPA based laser system consisting of a large grating stretcher, a regenerative amplifier, a cryogenically-cooledmultipass amplifier and a grating compressor. The necessary steps needed to achieve CEP stable operation at an output energy of 6 mJ were described and additional steps to further reduce the CEP jitter were proposed. Clearly, by progressing step by step CEP stability can be added to existing CPA laser systems without causing long interruptions to scientific experiments.