The laser system

Detailed descriptions of the laser source employed can be found in [36, 37]. This chapter gives a survey over the general characteristics and capitalizes on the obtained improvements in the course of this work. To achieve the few cycle NIR pulses needed for the experiments presented a modified version of the commercially available chirped-pulse-amplification sys- tem (Femtopower Compact Pro, Femtolasers Produktions GmbH) was employed. The Ampli- fier system is seeded by an ultra broadband Ti:Sapphire oscillator, emitting pulses of about 6 fs with an output power of 300 mW at 78 MHz [38]. To obtain the error signal needed for the phase stabilization (section 2.2) the oscillator pulses are focused into a periodically poled crystal that broadens the spectrum and generates the difference frequency simultaneously (section 2.2.2) [39]. For seeding the amplifier (seed power ~1 nJ) the oscillator pulses are dispersively stretched to 15 ps in bulk material and then amplified to 1.2 mJ / 3 kHz in 9 passes through a Ti:Sa crystal pumped with ~24 W / 3 kHz at 532 nm. Subsequent recompres- sion takes place in a hybrid prism- / positive dispersive mirror-compressor. The prism com- pressor is modified to overcompensate the positive chirp of the amplified pulses, therefore the pulses maintain their negative chirp throughout the final prism and are not object to self-phase modulation (SPM) which would cause a narrowing of the spectrum. The full compression takes place by reflection on positive dispersive chirped mirrors where SPM does not occur. Thereby, sub-25 fs pulses (full-width at half maximum FWHM) extending over a spectral width of 61 nm and energy of 900 µJ [40] are achieved. The carrier-envelope-phase stabiliza-

tion mandatory for the performed experiments is realized by branching of approximately 4% of the amplified pulses after the chirped mirror compressor that are sent into a collinear f-2f interferometer (chapter 2.2.3) that serves as phase-monitor and for the feedback loop of the phase-stabilisation. Fig. 2.1 shows the layout of the oscillator / amplifier system.

Fig. 2.1: Laser setup schematic. The Ti:Sapphire oscillator generates 6 fs pulses with 320 mW of average power at a repetition rate of 78 MHz. The pulses are stretched to 15 ps and amplified to 1 mJ / 3 kHz pulse energy and subsequently recompressed to 20 fs in a hybrid prism- / positive dispersion mirror compressor. Substantial spec- tral broadening is provided by a neon filled hollow fiber that is followed by a set of (negatively) chirped mirrors that compress the pulses down to <5 fs.

2.1.1 Pulse compression

After the amplifier, the pulses are focussed into a 1 m long hollow-core fiber with inner di-

ameter of 250 μm filled with neon at a pressure of ~ 2 bar. The intense pulses broaden their

spectrum (see Fig. 2.1) due to self-phase-modulation and the wider spectrum is recompressed to sub 5 fs using a set of low-loss chirped mirrors covering a wavelength range of 550 to 950 nm [41]. The transmission through the fiber is above 50 % resulting in pulse energies of 400 µJ after recompression.

2.1.2 Temporal and spatial pulse characterisation

For the continuous monitoring of the temporal pulse characteristics and dispersion adjust- ment, around 4 % of the power are split after the fiber and the chirped mirror compressor and sent to an interferometric second order autocorrelator. From the measured second-harmonic yield of the stepwise time-delayed interference of two pulse replica, the pulse duration can be calculated. Fig. 2.2 shows a typical measured autocorrelation trace. Fringe resolved autocorre- lation traces (ACF) can be calculated [42] and compared with the measured ones to get an approximation of the pulse length. In case of the ACF shown in Fig. 2.2, this comparison yielded a pulse length of 3.9 fs. Even though ACF is the most widely used technique in de- termining the duration of short laser pulses, it suffers certain weak points.

Fig. 2.2: Spectrum measured after the hollow-core fiber extending from 400 to 1000 nm (upper panel). A second order autocorrelation function (ACF) measured after compressing the pulses in the chirped mirror compressor is shown in the lower panel. Comparison with calculated ACF´s using the input spectrum leads to a pulse length estimation of 3.9 fs.

Laser pulses of this short duration comprise only ~1.5 cycles of the light electric field and since they are so sharply localized in the time domain, they necessarily cover a wide range of frequencies. As can be seen in Fig. 2.2, the spectrum of the light that composes the pulse is

spread over one octave requiring very precise control of the dispersion of the different col- ours. Since the fringe resolved ACF is not wavelength dispersive, inchoate compensation of the dispersion can only be observed in a change of the retrieved trace. Sign and order of a possible chirp can hardly be retrieved; in addition, bandwidth limitations and nonlinearities of the employed optics raised further questions on the reliability of the retrieved pulse character- istics.

Fig. 2.3: Measured (upper left) and calcu- lated (upper right) FROG trace of the compressed laser pulses and the retrieved temporal intensity profile and phase (lower left). A simple Gauss fit (grey line) of the retrieved pulse yields a pulse dura- tion of 4.1 fs FWHM, the drop in signal around -2.5 fs is an artefact and can be traced back to improper recording of the spectra for this time delay. Across the pulse the spectral phase is adequately flat, indicating the almost Fourier limited com- pression of the pulses.

In order to overcome these limitations, measurements with a self diffraction frequency re- solved optical gating device (FROG) [43] were performed. The fully reflective setup introduc- ing no additional dispersion in the beam path and a self-made quartz plate fire polished down to 5 µm thickness acting as nonlinear medium widened the supported bandwidth to cover the full spectrum contained in the pulses. Fig. 2.3 shows the results obtained using commercial retrieval software (Frog3, Femtosoft technologies) based on adaptive algorithms.

The comparison with the measured, frequency resolved trace is striking; even weak modula- tions in the background of the false colour representations are reproduced by the algorithm. For the measurement shown, the obtained pulse length amounts to 4.1 fs in good agreement with the ACF for almost perfectly dispersion compensated pulses. Despite certain ambiguities of the retrieval, the method is clearly predominant in comparison with ACF. Since the pulse still measures itself in this technique, the absolute timing is concealed as well as the absolute phase of the pulse; the insight into the spectral phase however will lead to major improve- ments in the design of particularly adapted chirped mirrors. In the obtained results for the ac- tual setup, the spectral phase already shows only small deviations of a flat function that can be contributed to little amount of residual higher order dispersion.

A good focusability and tight focussing are crucial for experiments where high intensities are required. As an example, only applications involving laser-plasma and multiphoton processes, laser ionization, or the discussed high harmonic generation shall be cited. The spatial quality of the compressed laser beam was characterized by means of the M² parameter. Using a CCD- Camera in combination with a microscope objective, the focal spot size was accurately deter- mined and the beam diameter as a function of the distance to the focus was obtained using the knife-edge method.

Fig. 2.4: Beam profile observed after the fiber (left) and beam profile in the focus (right) with a fit assum- ing a Gaussian profile with 19 µm width.

The measurement resulted in a very small M² value of M² = 1.8 +/- 0.1 with a Bessel-like beam profile caused by the hollow core fiber acting as spatial filter that improves the beam characteristics. Fig. 2.4 shows the beam profile observed after the hollow core fiber and the beam profile in the focus.