New Journal of The open–access journal for physics

Powerful 170-attosecond XUV pulses generated with few-cycle laser pulses and broadband multilayer

M Schultze1, E Goulielmakis1,4, M Uiberacker2, M Hofstetter2, J Kim3,DKim3, F Krausz1,2 and U Kleineberg2,4 1 Max-Planck-Institut für Quantenoptik, Hans-Kopfermannstrasse 1, D-85748 , 2 Department für Physik, Ludwig-Maximilians-Universität, Am Coulombwall 1, D-85748 Garching, Germany 3 Laser Science Laboratory, Department of Physics, POSTECH, Pohang, Kyungbuk 790-784, Korea E-mail: [email protected], [email protected]

New Journal of Physics 9 (2007) 243 Received 20 June 2007 Published 31 July 2007 Online at http://www.njp.org/ doi:10.1088/1367-2630/9/7/243

Abstract. Single 170-as extreme ultraviolet (XUV) pulses delivering more than 106 photons/pulse at ∼100 eV at a repetition rate of 3 kHz are produced by ionizing neon with waveform-controlled sub-5 fs near-infrared (NIR) laser pulses and spectrally filtering the emerging near-cutoff high-harmonic continuum with a broadband, chirped multilayer molybdenum–silicon (Mo/Si) mirror.

4 Authors to any whom correspondence should be addressed.

New Journal of Physics 9 (2007) 243 PII: S1367-2630(07)53049-1 1367-2630/07/010243+11$30.00 © IOP Publishing Ltd and Deutsche Physikalische Gesellschaft 2 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

Contents

1. Introduction 2 2. Generation of attosecond XUV pulses 3 3. Isolation of a single attosecond pulse via spectral filtering with multilayer optics 3 4. Measurement of attosecond XUV pulses 4 5. Experimental 5 6. Results and discussion 7 Acknowledgments 9 References 10

1. Introduction

Tracking and controlling electron dynamics on the atomic scale call for techniques that provide at least ten-to-hundred times improved temporal resolution as compared with the broadly proliferated femtosecond technology, for techniques that offer the potential for capturing processes evolving on an attosecond timescale (1 as = 10−18 s). Sub-femtosecond or attosecond pulses in the extreme ultraviolet (XUV) to soft x-ray regime [1]–[4] and sub-femtosecond electron pulses [5, 6] constitute key elements for developing these techniques. Advances in ultrafast optics, foremost intense few-cycle laser pulses comprising merely a few well-controlled oscillations of the light field [7, 8](Tpulse∼2.5 fs), have allowed—via high harmonic generation in an atomic gas—reproducible production of isolated XUV attosecond pulses of 250 as duration at ∼100 eV [9] and 130 as duration at 35 eV [10], which correspond to only 1/10 and 1/20 of the field oscillation of a laser pulse, respectively. These pulses permitted for the first time a direct and complete characterization of light waveforms [11]. High harmonic generation in rare gases was discovered to be a powerful means of providing coherent XUV radiation from a compact laser-driven source a long time ago [12, 13]. However, only few-cycle-laser pulses permitted efficient confinement of the emitted coherent, laser-like radiation to within the laser oscillation cycle and extension of the spectrum of this confined radiation to the water window [14] and well beyond [15]. With the advent of intense few- cycle near-infrared (NIR) pulses [16] the isolation of sub-laser-cycle duration XUV light bursts (one burst per laser pulse) with well-reproduced pulse parameters became feasible for the first time [9, 16]. Controlled isolation of a single energetic harmonic pulse requires control of the amplitude of spectral components of the emitted XUV radiation over a broad (several eV or more) spectral range. Furthermore, transform-limited XUV pulse production calls for precise control of the phase over the spectral band selected by band-pass filtering. Such a phase control has been demonstrated by utilizing the dispersion of materials near electronic resonances [17, 18]. However, the applicability of this approach is restricted to selected spectral domains. By analogy with broadband dispersion control of optical pulses in the visible and nearby spectral ranges with chirped multilayer dielectric mirrors [8], chirped multilayer metallic mirrors have been proposed and developed for the same purpose in the XUV spectral range [19, 20]. In this paper, we demonstrate the utilization of this technology for the generation of isolated 170 as XUV pulses. This result has been achieved by ionizing neon atoms with sub-5 fs NIR pulses without the need

New Journal of Physics 9 (2007) 243 (http://www.njp.org/) 3 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT for polarization gating [10, 21] resulting in a substantially improved NIR-to-XUV conversion efficiency and photon fluxes useful for time-resolved spectroscopy. Our work is motivated by the need for shorter and more powerful XUV pulses for interrogating a wide range of fundamental electronic processes in matter evolving on a hyperfast, sub-100 as timescale [22, 23].

2. Generation of attosecond XUV pulses

Emission of XUV radiation by laser pulses interacting with atoms can be described by a three-step model where the electron, driven by the electric field of the laser pulse (i) is set free around the maximum of the field (tunnel ionization), (ii) follows a quasi-classical motion driven by the electric field and (iii) emits its excess energy upon recombination with the parent ion as high energy photons [24]. The maximum energy of the emitted photons is given by Emax = Ip +3.17Up with Ip being the atomic ionization potential and Up being the ponderomotive energy of the electron in the oscillating electric field of the laser. The only parameter that can be controlled in this process by the laser pulse parameters is Up defining the highest energy photons emitted. In the case of multi-cycle laser pulses the electrons encounter the parent ion every half-cycle of the electric field which results in a series of emissions that manifest themselves in the spectral domain as high harmonic structure [12, 13, 25]. In contrast, when a few-cycle laser pulse comes into play the energy of the emitted light wavepacket from consecutive recombination events close to the peak of the pulse can vary substantially from the maximum emitted energy. By setting the carrier envelope phase (CEP) of a few-cycle laser pulse such that the field maximum coincides with the maximum of the pulse envelope (resulting in a substantial modification of Up for consecutive recombination events)—referred to as cosine pulse—it is possible to restrict the emission of the most energetic light wavepackets to take place only once per laser pulse [9]. By contrast, in a sinusoidally-shaped laser field a couple of re-colliding electron wavepackets lead to emission at the highest photon energies, resulting in two XUV bursts separated by half a period of the laser field. The extension of the XUV/x-ray spectrum associated with a single re-collision event and hence with the potential to form an isolated attosecond pulse is precisely linked to the intensity contrast between consecutive field maxima and thus the duration of the driving pulse. In the case of ∼4.5–5fs pulses utilized in this work, this contrast is expected to be approximately 20 eV [26] and can therefore potentially support isolated XUV pulses down to a duration of less than 150 as.

3. Isolation of a single attosecond pulse via spectral filtering with multilayer optics

Driver pulses with these parameters have been available for quite some time [10]. The generation of isolated XUV attosecond pulse generation down to the sub-200 as regime has been hitherto limited by the lack of XUV optics that could filter this continuum and sufficiently suppress spectral components that lead to emergence of a satellite pulse. This limitation has been overcome in this work by using a specially-designed aperiodic (or chirped)Mo/Si multilayer mirror. Multilayer XUV mirrors consist of periodic or aperiodic stacks of alternating ultra-thin layers of two or more different materials (typical layer thicknesses 1–10 nm, typical layer numbers 10–500). Varying the layer thicknesses allows for a controlled

New Journal of Physics 9 (2007) 243 (http://www.njp.org/) 4 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT design of the centre reflectivity, the reflected bandwidth as well as of the spectral phase. The latter parameters affect the time response of the mirror upon reflection of an attosecond XUV pulse. While the bandwidth of a periodic XUV multilayer in first Bragg order is mainly limited by the number of contributing bi-layers N to λ/λ = 1/N to be about 10% of the reflected (photon energy) or less, broader reflection profiles can be realized by aperiodic multilayer stacks. Furthermore, the evolution of the spectral phase, which controls the spectral dispersion of the pulse and by that its temporal evolution, can be optimized and shaped by the interference structure of the multilayer. Over many years small bandwidth, high reflectivity Mo/Si multilayers have been developed for the 80–100 eV photon energy range and proven their capability for spectrally filtering, dispersing and steering high harmonics. While 250 as isolated XUV pulses have been generated and filtered at 93 eV by using a periodic Mo/Si multilayer with approx. 8.6 eV full width at half maximum (FWHM) bandwidth [11]. Aperiodic multilayer mirrors can support a substantially broader bandwidth required for shorter XUV pulses. In this work we use a coating design with a reflection bandwidth of ∼16 eV FWHM and optimized dispersion characteristics holding the potential for producing isolated 150 as pulses. Even the compensation of an intrinsic spectral chirp of the incoming XUV pulse (resulting from the HH generation process or propagation effects) by the group delay dispersion (GDD) of the XUV multilayer mirror (‘chirped XUV mirror’) over the whole reflected bandwidth is feasible which will enable compression of the pulse to its Fourier limit [27, 28].

4. Measurement of attosecond XUV pulses

The basic idea behind techniques developed for the characterization of XUV bursts, is the generation of a photoelectron replica of the attosecond pulse the temporal structure of which is probed by the ultrafast oscillation of the electric field of a laser pulse [29, 30]. Assume ionization of atoms by XUV attosecond pulses in the presence of a laser field E(t) and detection of the generated electrons along the polarization direction of the field: an electron released at the moment t undergoes a momentum shift p which is given by ∞   p(t) = e EL(t )dt = eA(t) t where e is the electron charge and A(t) the vector potential of the field. Therefore, the ultrafast variation of the field amplitude maps the temporal evolution of the attosecond electron pulse to a corresponding momentum and electron distribution. We have put this idea into practice in a concept dubbed the ‘atomic transient recorder’ [9]. In practice a series of electron spectra acquired in different delay settings between the laser and the electron pulse (a spectrogram), offer the possibility of an accurate retrieval of the temporal structure of the electron and consequently the XUV attosecond pulse, at the same time [31]. The technique in that case is closely related to frequency resolved optical gating (FROG), where the gate for the characterization of the attosecond electron pulse is provided by the field oscillation. To this end, various reconstruction algorithms have been developed [32]–[34] and allow retrieval not only of the attosecond pulse but also simultaneously of the laser pulse (gate) enabling metrology of optical fields. The reconstruction of the experimental results discussed in this work is also based on the above mentioned principles.

New Journal of Physics 9 (2007) 243 (http://www.njp.org/) 5 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

Figure 1. Experimental set-up: few-cycle phase-stabilized NIR laser pulses are focused into a noble gas target (Ne) to produce the XUV radiation used in the experiment. The remaining NIR pulse and the XUV light propagate through differential pumping stages to prevent the gas used in the generation process from entering the experimental chamber. The ‘HH-Meter’ tracks the ionization of the background gas and therefore provides a direct measure of the XUV flux. The NIR intensity is adjusted by an aperture. A thin Zr foil mounted at the centre of a 5 µm thin pellicle acts as bandpass filter for the lower harmonics and the fundamental. The fundamental is transmitted through the pellicle in form of an annular beam around this filter. The two concentric beams are then focused by a double mirror assembly (inset) consisting of a (15.6 eV) aperiodic multilayer mirror in the inner part and a silver-coated perforated mirror of the same focal length. XUV and NIR pulses are spatially and temporally overlapped in the Fourier plane while their delay can be adjusted over a range of ±150 fs by translating the inner mirror along the beam axis using a closed loop piezo stage.

5. Experimental

The experiments were performed at theAS-1 attosecond beamline of our laboratory.The beamline is equipped with a FemtoPower Pro laser system (Femtolasers Produktions GmbH) that delivers pulses with 25 fs duration and 1 mJ energy at 3 kHz. These pulses are spectrally broadened upon propagation through a neon-gas-filled capillary [7] and compressed down to an FWHM duration between 4.5 and 5 fs with chirped dielectric mirrors. The compressed pulses have energy of 400 µJ and are carried at a wavelength of ∼750 nm. The CEP is controlled utilizing a monolithic stabilization scheme for the oscillator [35] while slow phase drifts imparted to the pulses along the amplification process are compensated for by an f–2f interferometer installed at the output of the amplifier [26]. The resulting phase stability of our sub-5 fs pulse has been characterized with a stereo ATI phase detector [36]–[38] and was found to be better than 100 mrad over a period of several hours.

New Journal of Physics 9 (2007) 243 (http://www.njp.org/) 6 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

Figure 2. XUV spectra emanating from the interaction of few-cycle phase stabilized laser pulses with Ne atoms after being transmitted through 1.5 µm of a free standing Zr foil. The spectra are integrated over ∼300 shots of the laser. Red curve: spectrum of the XUV pulses emitted by a sinusoidal few-cycle pulse while black curve is the spectrum of XUV pulses generated by a cosine pulse. The highly modulated spectrum associated to a sinusoidal waveform constitutes the spectra manifestation of a substantial amount of satellite in the time domain. The reflectivity of the 15.6 eV (FWHM) aperiodic multilayer mirror employed in the experiments presented here, is marked by a blue line. Grey curve shows the transmission of the 150 nm Zr foil used in the experiment (scale shown on left axis).

XUV light is generated by gently focusing (f = 500 mm) the sub-5 fs laser pulses into a quasi-static gas cell filled with Neon at a pressure of ∼350 mbar. The length of the interaction region amounts to ∼2.5mm. For the characterization of the emanating XUV light we employ a home-built spectrometer consisting of a flat-field grating (Hitachi-Flat-field [39]) and a back-illuminated CCD camera, (Roper scientific, Princeton Research). An adjustable aperture (down to 0.5 mm) installed 1 m behind the source allows spatial filtering of the XUV beam before entering the spectrometer. With this apparatus we can record high-quality spectra with an integration time of merely 50 ms (150 laser shots), permitting convenient real-time monitoring and optimization of the high-harmonic radiation. A quasi-monolithic double-mirror assembly consisting of a XUV Mo/Si multilayer mirror (diameter: 3 mm) with a bandwidth of 15.6 eV centred at 93 eV and an outer concentric perforated mirror (diameter: 1 cm, silver coated) (see inset figure 1) is placed approximately 1.5 m downstream and used to focus the XUV and the laser beam respectively into a second Ne target. The delay between laser and XUV pulse can be varied by means of a precise piezo- translation unit (Physik Instrumente: resolution ∼5nm). Electrons ejected in the polarization direction of the laser are collected by a time-of-flight (TOF) electron spectrometer having a resolution of better than 0.5 eV in the range 60–90 eV.

New Journal of Physics 9 (2007) 243 (http://www.njp.org/) 7 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

Figure 3. Atomic transient recorder spectrogram of the XUV attosecond pulse shown in 3D contour representation. It comprises a series of ∼40 electron spectra generated in Ne and recorded as a function of the delay between the XUV and the few-cycle laser pulse in steps of 200 as.

6. Results and discussion

Figure 2 depicts XUV spectra integrated over 300 shots of the laser pulse for a CEP of the driving field resulting in a >20 eV broad near-cutoff continuum (black line). A CEP shift by π/2 with respect to the previous setting results in a highly modulated spectrum (red line). To fully utilize this extended continuum we employ a broadband aperiodic multilayer mirror, the reflectivity curve of which is shown in blue in figure 2. A secondary reflectance fringe (Kiessig fringe) in the lower energy part of the spectrum at around 70 eV can be substantially suppressed by means of a 150 nm thin Zr foil (see figure 1 and its caption). For the characterization of the XUV pulses with the atomic transient recorder the XUV continuum is reflected, filtered and focused by the multilayer mirror into a Ne gas nozzle along with the IR field. Series of electron spectra as a function of the delay between the XUV and laser pulse are shown in false colour representation in figures 3 and 4. Figure 3 depicts a high resolution (200 as step in delay) high streaking field (W > 10 eV shift of the electron peak) scan. Figure 4 shows a long range scan in delay with a scanning step of 300 as. For the retrieval of the duration of the attosecond pulse we have developed software (dubbed: ATTOSTREAK) based on the theoretical and retrieval principles discussed and referred to in the

New Journal of Physics 9 (2007) 243 (http://www.njp.org/) 8 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

Figure 4. Long-range atomic transient recorder spectrogram of a 170 XUV attosecond pulse. A series of electron spectra generated in Ne have been recorded as a function of the delay between the XUV and the few-cycle laser pulse in steps of 300 as. The electron spectrometer is oriented such as to collect electrons emitted only in the direction of the polarization vector of the laser field utilizing the geometry presented in [9]. To ensure high resolving power the streaking field is set to shift the electron distribution in energy by more than the half of the initial width, which corresponds to ∼8eV. previous section. The algorithm accounts for possible satellite pulses, while all key parameters including the chirp of the attosecond pulse and the temporal shape of the laser field are left free to be reconstructed invoking the power of a genetic algorithm. The retrieved pulse duration by our analysis was 170 ± 10 as approximately 5% longer compared with the transform limited pulse supported by the reflected spectral bandwidth (approximately 160 as). The satellite pulse was found to contain about 17% of the total photon number. As pointed out above, the electric field of the XUV attosecond pulse and that of the laser pulse can be reconstructed from the same spectrogram. However the prerequisites regarding the spectrogram are not the same. The former can be characterized with sufficient accuracy utilizing solely spectra recorded over a delay range that spans a field oscillation; the latter can be accessed only if the attosecond pulse samples at least 5–10 field oscillations for a few-cycle laser pulse, necessary to probe its built-up and disappearance. The retrieved duration (FWHM) of the NIR laser pulse was 4.7 fs. To demonstrate the degree of control and reproducibility of the waveform of our few-cycle pulses, we shift the phase of the driving field by π (as to switch from a cosine to an anti-cosine driving waveform) and repeat a similar scan as in figure 5 but limited in delay within ±7fs around the peak of the pulse. Both series of spectra are depicted in figure 4. A complete characterization of a new source entails always a photon yield measurement. We employed our yield-calibrated x-ray CCD and calculated the number of photons in the focus

New Journal of Physics 9 (2007) 243 (http://www.njp.org/) 9 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

Figure 5. Light wave oscillations of a phase stabilized few-cycle laser pulse recorded over a delay interval of ±7 fs around the envelope maximum of the laser field for two different phase settings of its CEP that differ by π. utilizing the tabulated transmission of the Zr filter, (48–62% @ 80–100 eV, see figure 2) and the measured peak reflectivity (by synchrotron radiation reflectometry) of our 15.6 eV FWHM bandwidth multilayer ∼5%. The analysis yielded ∼3.6 × 108 photons s−1 for the attosecond pulse delivered into the target. The flux at the HHG source was more than an order of magnitude higher for the same spectral range suggesting that further improvements afford promise for substantially higher photon flux delivered to experiments. This photon flux is comparable to, even slightly higher than, its predecessor source (250 as @ 95 eV, 8.6 eV FWHM) at our laboratory and suggests that generating attosecond pulses with considerably shorter duration is attainable while maintaining other key parameters. In conclusion, we have demonstrated that pulses as short as 170 as can be reliably generated in the range of ∼100 eV utilizing an at least 20 eV broad near-cutoff XUV continuum and broadband multilayer optics. The transition to shorter isolated attosecond pulses (approximately 70% shorter than previously generated without polarization control) maintaining all important parameters required in potential experiments, including high energy photons for triggering inner-shell dynamics, as well as high photon flux, affords the promise that sub-100 as isolated pulses will be efficiently generated in the near future by driving pulses not shorter than 4 fs. Controlled light fields and isolated attosecond pulses precisely synchronized to their cause will open the door to accessing and controlling yet unexplored processes in atoms, molecules, solids and nanostructures.

Acknowledgments

This work was supported by the Deutsche Forschungsgemeinschaft (DFG) in the frame of the excellence cluster ‘ Centre for Advanced Photonics’ (MAP), the Marie Curie

New Journal of Physics 9 (2007) 243 (http://www.njp.org/) 10 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

Research Training Network XTRA (MRTN-CT-2003 505138) and by the European Commission Integrated Infrastructure Initiative LASERLAB-Europe. EG acknowledges an Intra-European Marie-Curie fellowship MEIF-CT-2005-024440. The authors acknowledge scientific discussion and collaboration with N Kabachnik, U Heinzmann and A Wonisch (University of Bielefeld, Germany) on the development of chirped XUV multilayer optics.

References

[1] Drescher M, Hentschel M, Kienberger R, Tempea G, Spielmann Ch, Reider G, Corkum P and Krausz F 2001 Science 291 1923–7 [2] Paul P, Toma E, Breger P, Mullot G, Auge F, Balcou P, Muller H and Agostini P 2001 Science 292 1689–92 [3] Hentschel M, Kienberger R, Spielmann C, Reider G, Milosevic N, Brabec T, Corkum P, Heinzmann U, Drescher M and Krausz F 2001 Nature 414 509–13 [4] Tzallas P, Charalambidis D, Papadogiannis N, Witte K, Tsakiris G 2003 Nature 426 267–71 [5] Niikura H, Legare F, Hasbani R, Bandrauk A, Ivanov M, Villeneuve D and Corkum P 2002 Nature 417 917–22 [6] Niikura H, Legare F, Hasbani R, Ivanov M, Villeneuve D and Corkum P 2003 Nature 421 826–9 [7] Nisoli M, DeSilvestri S, Svelto O, Szipocs R, Ferencz K, Spielmann C, Sartania S and Krausz F 1997 Opt. Lett. 22 522–4 [8] Szipocs R, Spielmann C and Krausz F 1994 Opt. Lett. 19 201–3 [9] Kienberger R et al 2004 Nature 427 817–21 [10] Sansone G et al 2006 Science 314 443–6 [11] Goulielmakis E et al 2004 Science 305 1267–9 [12] Macklin J, Kmetec J and Gordon C 1993 Phys. Rev. Lett. 70 766–9 [13] Huillier A L and Balcou P 1993 Phys. Rev. Lett. 70 774–7 [14] Spielmann C, Kan C, Burnett N, Brabec T, Geissler M, Scrinzi A, Schnurer M and Krausz F 1998 IEEE J. Sel. Top. Quantum Electron. 4 249–65 [15] Seres J, Seres E, Verhoef A, Tempea G, Strelill C, Wobrauschek P, Yakovlev V, Scrinzi A, Spielmann C and Krausz F 2005 Nature 433 596 [16] Baltuska A 2003 Nature 422 189 [17] Johnsson P 2005 Phys. Rev. Lett. 95 013001 [18] Henke B, Gullikson E and Davis J 1993 At. Data Nucl. Data Tables 55 349 [19] WonischA, Westerwalbesloh T, Hachmann W, Kabachnik N, Kleineberg U and Heinzmann U 2004 Thin Solid Films 464–5 473–7 [20] Morlens A et al 2006 Opt. Lett. 28 1558–60 [21] Sola I et al 2006 Nature Phys. 2 319–22 [22] Morishita T, Watanabe S and Lin C 2007 Phys. Rev. Lett. 98 083003 [23] Breidbach J and Cederbaum L 2005 Phys. Rev. Lett. 94 033901 [24] Corkum P 1993 Phys. Rev. Lett. 71 1994–7 [25] Papadogiannis N A, Kalpouzos C, Goulielmakis E, Nersisyan G, Charalambidis D, Auge F, Weihe F and Balcou P 2001 Appl. Phys. B 73 687–92 [26] Baltuska A, Uiberacker M, Goulielmakis E, Kienberger R, Yakovlev V S, Udem T, Haensch T W and Krausz F 2003 IEEE J. Sel. Top. Quantum Electron. 9 972–89 [27] Wonisch A, Neuhausler U, Kabachnik N, Uphues T, Uiberacker M, Yakovlev V, Krausz F, Drescher M, Kleineberg U and Heinzmann U 2006 Appl. Opt. 45 4147–56 [28] Kleineberg U 2006 X-ray Lasers Springer Proceedings in Physics 115 ed P V Nickles and K A Janulewiscz (Springer: Berlin) [29] Itatani J, Quere F, Yudin G, Ivanov M, Krausz F and Corkum P 2002 Phys. Rev. Lett. 88 173903 [30] Kitzler M, Milosevic N, Scrinzi A, Krausz F and Brabec T 2002 Phys. Rev. Lett. 88 173904 [31] Trebino R and Kane D 1993 J. Opt. Soc. Am. A 10 1101–11

New Journal of Physics 9 (2007) 243 (http://www.njp.org/) 11 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

[32] Yakovlev V, Bammer F and Scrinzi A 2005 J. Mod. Opt. 52 395–410 [33] Mairesse Y and Quere F 2005 Phys. Rev. A 71 011401 [34] Cormier E, Walmsley I, Kosik E, Wyatt A, Corner L and DiMauro L 2005 Phys. Rev. Lett. 94 033905 [35] Fuji T 2005 Opt. Lett. 30 332–4 [36] Paulus G G, Lindner F, Walther H, Baltuska A, Goulielmakis E, Lezius M and Krausz F 2003 Phys. Rev. Lett. 91 4 [37] Verhoef A, Fernandez A, Lezius M, O’Keeffe K, Uiberacker M and Krausz F 2006 Opt. Lett. 31 3520–2 [38] Schatzel M, Lindner F, Paulus G, Walther H, Goulielmakis E, Baltuska A, Lezius M and Krausz F 2004 Appl. Phys. B 79 1021–5 [39] Kita T, Harada T, Nakono N and Kuroda H 1983 Appl. Opt. 36 512–3

New Journal of Physics 9 (2007) 243 (http://www.njp.org/)