High harmonics gave the bandwidth, but not the timing
High harmonics generated in gases supply a broad extreme-ultraviolet spectrum, the bandwidth that very short pulses need. L’Huillier’s work helped establish this source and explain how it works. Bandwidth alone does not fix the duration: relative spectral phases determine how sharply the components add within each burst. An evenly spaced harmonic comb still repeats in time when those phases change. Isolating one burst also requires control of when the harmonics are emitted and which part of the spectrum is selected.
The 2023 scientific background treats making this light and measuring it as separate achievements. The measurements came in 2001. Agostini’s group characterized a train of bursts, each roughly 250 attoseconds long; Krausz’s group characterized a single isolated pulse. A train and a single pulse have different structures in time, so one result cannot stand in for the other.
RABBIT: two paths to the same electron energy interfere
An electron can reach the same final energy by two routes: by absorbing a lower XUV harmonic and an infrared photon, or by absorbing the next higher odd harmonic and emitting an infrared photon. The infrared step places electrons in a sideband between the harmonic peaks. The two routes interfere, so the sideband yield oscillates as the infrared delay is scanned. Its oscillation phase carries the phase difference between the two harmonics.
That phase has two parts: the spectral phase of the XUV light and a two-photon phase contributed by the atom. A measured phase therefore cannot be read directly as the moment the source emitted light. Reconstructing the pulse train also needs the harmonic amplitudes and a stated correction or approximation for the atomic part.
Streaking: the infrared field shifts each electron by its release time
For a single XUV burst, a synchronized infrared field changes each photoelectron’s final momentum by an amount that depends on when the electron was released. Recording electron energy against delay turns that timing into a spectrogram. In their abstract, Hentschel and coauthors report a soft-X-ray pulse lasting 650 attoseconds, with an uncertainty of 150 attoseconds.
In the simplest classical picture, integrate the electric force on an electron released at the stated time and assume the infrared vector potential vanishes after the pulse. With the electric-field convention below and the elementary charge positive, the momentum shift has a minus sign. This dipole approximation neglects the magnetic force, the residual ion’s Coulomb field and scattering.
The duration comes from a model of the measurement
Both methods obtain a duration by applying a model of the measurement to energy-versus-delay data. That model makes assumptions about the target atoms, the infrared dressing field and the retrieval method. Neither method films an electron orbit frame by frame. When comparing reported delays, differences in phase convention and atomic response matter.
Chirped-pulse amplification (CPA) supplies the intense, short driving pulse. It does not measure harmonic phases or reconstruct an attosecond pulse, so the site’s femtosecond-source Lab does not count as an attosecond measurement instrument.
Look at the measured trace before the reconstructed pulse
When you read a reported duration, keep the measured sideband oscillation or streaking trace beside it, and find out which atomic-phase correction or propagation approximation produced it. The linked CPA model covers only how the driving pulse is made. Measuring attosecond pulses is a separate task that it does not attempt.