Schematic of an autocorrelator.  A pulse is split into two, one is variably delayed with respect to the other, and the two pulses are then crossed in a nonlinear-optical crystal, yielding a new light pulse whose intensity is given by the product of the two pulse intensities. If the two pulses overlap in time, some light is generated; if not, then no light is generated. The expression for the autocorrelation is shown at the upper right. In this and later figures, E(t) is the pulse electric field vs. time (indicative of the intensity and phase), and I(t) is the intensity vs. time.

Autocorrelations of stable and unstable trains of pulses, showing the coherent artifact.   Red indicates the pulse intensity, while blue indicates the phase. Notice that the autocorrelation does not depend at all on the phase (by design).  Also, structure in the intensity washes out in the autocorrelation. The relatively narrow feature in the autocorrelation of the unstable pulse train is the coherent artifact. Note that it's considerably shorter than the actual typical pulse in the train but is often confused for the actual pulse length.

Invented in the 1980s, interferometric autocorrelation was the second technique used to measure ultrashort laser pulses. Like intensity autocorrelation, it uses a laser pulse to measure itself, but it also interferes the light generated in the nonlinear-optical medium with light generated by the individual beams.  It yields some phase information, but not enough to determine the pulse intensity and phase.  This tutorial examines the characteristics and shortcomings of this also obsolete method.  Basically, it yields information essentially equivalent to the intensity autocorrelation and the spectrum.