ARTICLES Measuring and controlling the birth of attosecond XUV pulses N. DUDOVICH 1 *, O. SMIRNOVA 1 , J. LEVESQUE 1,2 , Y. MAIRESSE 1 , M. YU. IVANOV 1 , D. M. VILLENEUVE 1 AND P. B. CORKUM 1 * 1 National Research Council of Canada, Ottawa, Ontario K1A 0R6, Canada 2 INRS-EMT, 1650 boulevard Lionel-Boulet, CP 1020, Varennes, Qu´ ebec J3X 1S2, Canada *e-mail: Nirit.Dudovich@nrc-cnrc.gc.ca; Paul.Corkum@nrc-cnrc.gc.ca Published online: 15 October 2006; doi:10.1038/nphys434 Generating attosecond pulses has required a radically different approach from previous ultrafast optical methods. The technology of attosecond measurement, however, is built on established methods of characterizing femtosecond pulses: the pulse is measured after it has left the region where it was produced. We offer a completely different approach: in situ measurement. That is, we integrate attosecond-pulse production and measurement in a manner that can be applied to many high-order nonlinear interactions. To demonstrate this approach, we combine a low-intensity (<10 −3 ) second-harmonic beam with the fundamental beam, to gently perturb the production process without significantly modifying it. The attosecond-pulse duration is read from the modulation of the even-harmonic signal as a function of the two- field delay. Increasing the second-harmonic intensity slightly (<10 −2 ), we extend measurement to control. We demonstrate control by manipulating the high-harmonic spectrum with high efficiency. M easuring and controlling the birth of attosecond XUV pulses can be viewed as electron interferometry. This interpretation arises naturally from the three-step model 1,2 of high-harmonic and attosecond-pulse generation. In the first step, an intense laser field removes an electron from its parent atom, splitting the wavefunction into a coherent superposition of a bound state and a free-electron wavepacket. In the language of interferometry, ionization is a beam splitter. In the second step, the free-electron wavepacket moves in the oscillating laser field and returns to the parent atom. This is the delay line that we can adjust. The delay line is schematically described as one of the two electron trajectories in Fig. 1a. In the final step, the two portions of the wavefunction overlap. Their interference produces an oscillating dipole, leading to attosecond pulses. In general, interferometry allows us to characterize fully all aspects of both beams—the electronic orbital 3–6 and the re-collision electron. For our experiment we use a multicycle (30 fs) pulse as a fundamental beam. In each 1/2 cycle the three-step process is repeated so the output is a train of attosecond pulses 7 . The spectrum of a train of attosecond pulses is a comb of harmonics of the driving laser frequency. Figure 1a shows the electron trajectories for alternate half-cycles of the laser field. Because the positive and negative half cycles are symmetric (solid curve), the left and right arms of the ‘interferometer’ are balanced and only odd harmonics of the fundamental are produced. An electron typically accumulates a phase of ∼10π on each arm for our 10 14 W cm −2 800 nm pulse. In an interferometer, a small phase shift, δφ, can modify the interference, even if the electron accumulates a large phase along the delay line. We use a weak (perturbative) 400nm beam (the second harmonic of the fundamental beam) to induce δφ. In general, a combination of the fundamental field and its second harmonic breaks symmetry, as illustrated in Fig. 1b. In our case, the second-harmonic field (Fig. 1a, dashed curve) unbalances the interferometer, breaking the symmetry and producing even harmonics of the fundamental. The phase accumulated by the electron is enhanced in the half cycle where the fundamental and second-harmonic fields are appropriately phased, and suppressed in the adjacent half cycles where the two fields are oppositely nature physics VOL 2 NOVEMBER 2006 www.nature.com/naturephysics 781 Nature Publishing Group ©2006