Attosecond science is based on steering electrons with the electric field of well controlled femtosecond laser pulses1. It has led to the generation of extreme-ultraviolet pulses2 with a duration of less than 100 attoseconds (ref. 3; 1 as = 10−18 s), to the measurement of intramolecular dynamics (by diffraction of an electron taken from the molecule under scrutiny4, 5) and to ultrafast electron holography6. All these effects have been observed with atoms or molecules in the gas phase. Electrons liberated from solids by few-cycle laser pulses are also predicted7, 8 to show a strong light-phase sensitivity, but only very small effects have been observed14. Here we report that the spectra of electrons undergoing photoemission from a nanometre-scale tungsten tip show a dependence on the carrier-envelope phase of the laser, with a current modulation of up to 100 per cent. Depending on the carrier-envelope phase, electrons are emitted either from a single sub-500-attosecond interval of the 6-femtosecond laser pulse, or from two such intervals; the latter case leads to spectral interference. We also show that coherent elastic re-scattering of liberated electrons takes place at the metal surface. Owing to field enhancement at the tip, a simple laser oscillator reaches the peak electric field strengths required for attosecond experiments at 100-megahertz repetition rates, rendering complex amplified laser systems dispensable. Practically, this work represents a simple, extremely sensitive carrier-envelope phase sensor, which could be shrunk in volume to about one cubic centimetre. Our results indicate that the attosecond techniques developed with (and for) atoms and molecules can also be used with solids. In particular, we foresee subfemtosecond, subnanometre probing of collective electron dynamics (such as plasmon polaritons9) in solid-state systems ranging in scale from mesoscopic solids to clusters and to single protruding atoms.
Figures at a glance
Figure 1: Overview of the experiment. 
Electrons, indicated in blue, are emitted (large blue arrow pointing away from surface) from a sharp tungsten tip (left) by, and interact with, a few-cycle laser electric field (red waveform). Controlled by the carrier-envelope phase, the liberated electron may be driven back into the tip, where it can scatter elastically and gain more energy in the laser field before it reaches the detector (not shown).
Figure 2: Carrier-envelope phase modulation in photoelectron spectra. 
a, Carrier-envelope phase-averaged electron count rate as function of energy (blue solid curve). About three photon orders, indicated by encircled numbers, are visible in the direct part (that is, when the electrons do not re-collide with the tip; E = 0 to ~4.3 eV). For E > 4.5 eV, the plateau region starts with five more photon orders visible. The green points depict the modulation depth of the count rate when varying the carrier-envelope phase (error bars, fit errors of modulation curves; see Supplementary Information). Insets, carrier-envelope phase modulation in the photocurrent with the spectrometer acting as an energy high-pass filter at 3 eV (left inset) and 11 eV (right inset; both with the carrier-envelope offset frequency set to
). a.u., arbitrary units. b, Contour plot of the electron count rate as function of carrier-envelope phase offset and energy. φCE was increased in steps of π/8. At 4.3 eV, the plot is split into two regions for better visibility (see colour scales under). In each region, we plot the normalized count rate (Supplementary Information). Measured data range over 2π and are extended over ~4π for better visibility. Yellow circles show the position of the cut-off for a given carrier-envelope phase offset (red curve, sinusoidal fit). c, Individual electron spectra extracted from the contour plot in b for four different carrier-envelope phase offsets separated by π/2. Only the plateau region is shown. Fringes are clearly visible for 0.6π and −0.9π, but almost no fringes are visible for 0.1π and −0.4π. d, Blue diamonds, average peak visibility in the plateau region (red curve, sinusoidal fit). The peaks used to determine the visibility are marked with grey arrows in a. Note that peak visibility and cut-off position are nearly maximally out of phase.
Figure 3: Theoretical modelling of the experimental data. 
