Femtosecond fiber chirped- and divided-pulse amplification system

International audienceWe implement both chirped pulse amplification and divided pulse amplification in the same femtosecond fiber amplifier setup. This scheme allows an equivalent stretched pulse duration of 2.4 ns in a compact tabletop system. The generation of 77 W of compressed average power at 4.8 MHz, together with 320 fs and 430 μJ pulses at a repetition rate of 96 kHz, is demonstrated using a distributed mode-filtering, rod-type, ytterbium-doped fiber. Limitations in the temporal recombining efficiency due to gain saturation inside the fiber amplifier are identified. A very efficient way of scaling the energy of amplified femtosecond pulses was provided by the chirped-pulse amplification (CPA) method [1]. This technique consists in temporally spreading the ultrashort pulse prior to amplification using a highly dispersive optical setup, thereby reducing the peak power by up to four orders of magnitude. This process can be reversed after the amplification to reveal the true amplified peak power. Although CPA is used in virtually every amplified femtosecond system, its implementation is increasingly difficult and costly when the desired stretched pulse duration is greater than 1 ns, due to the increasing size of the gratings and overall optical setup. Such long stretched pulse durations are therefore not used for compact tabletop or industrial systems. The divided-pulse amplification (DPA) idea was proposed more recently [2] and is similar to CPA in the sense that it also consists in redistributing the pulse energy over a time interval larger than the initial pulse width to reduce the peak power. In this scheme, a train of several orthogonally polarized delayed pulse replicas is generated and amplified before final recombination. The technique has been successfully implemented to amplify picosecond pulses [3,4], for which it is difficult to use the CPA method because of the small optical bandwidth available. It has also been used to amplify femtosecond pulses in the parabolic regime [5], which necessitates un-chirped pulses at the input of the amplifier. In both cases the delay between two successive replicas was of the order of a few picoseconds and was therefore advantageously realized using centimeter-long, highly birefrin-gent vanadate crystals. In this Letter, we propose and demonstrate for the first time a femtosecond fiber amplifier that uses both concepts simultaneously to scale the output energy of a ta-bletop femtosecond fiber system. Because the CPA stretched-pulse duration is of the order of hundreds of picoseconds, the DPA delays are implemented using free-space interferometers with reasonable arm lengths of up to 1.32 m. Because the energy of the amplified pulses is not negligible compared to the saturation energy of the fiber amplifier, a decrease of the temporal combining efficiency appears, related to the different levels of saturation of the gain medium seen by the replicas in the train, leading to different phase accumulations due to self-phase modulation (SPM) and Kramers–Krönig (K–K) effects. Despite this limitation, our experiment allows the generation of 320 fs, 430 μJ pulses at a repetition rate of 96 kHz, corresponding to 42 W average power, using a distributed mode filtering (DMF) rod-type large-mode-area fiber [6] with a combining efficiency of 82%. The 1.2 m × 0.4 m footprint experimental arrangement is depicted in Fig. 1. The DPA setup is included inside a moderately nonlinear fiber CPA in which the impact of nonlinearities can be partially compensated by the dispersion mismatch of the stretcher and compressor units [7]. The front end is composed of a passively mode-locked ultrafast oscillator, a pulse picker, a stretcher, and a low-power fiber preamplifier. The oscillator generates pulses at the central wavelength of 1030 nm and pulse duration of 200 fs. The repetition rate of the oscil-lator of 25 MHz is divided by an acousto-optic modulator pulse picker. In this experiment, two repetition rates of 4.8 MHz and 96 kHz are selected to investigate the linear and nonlinear regimes, respectively. The stretching ratio is designed to preserve a pulse duration above 500 ps after the power amplifier when operated at maximum gain. The low-power fiber preamplifier is used to boost the average power to a few tens of mW to ensure efficient operation of the power amplifier. An optical isolator is placed after the front end to prevent any optical feedbac