Abstract

Nonlinear pulse characterization techniques such as FROG and SPIDER^1,2 have a large advantage over linear techniques (such as spectral interferometry³) in that they can be self-referenced, i.e. no well-characterized reference pulse is needed to obtain the phase and amplitude of an ultrafast laser pulse. Many different self-referenced methods are now in common use, and pulse shapes extracted by these methods are commonly assumed to be fully characterized. However, we show here that for an important and readily generated class of pulses-pulses that consist of several components that are well-separated in frequency-none of the variants of these nonlinear characterization techniques are capable of determining the relative phase between the components of the pulse. If the components of the pulse are separated in time as well as in frequency, then even the cross-correlated methods (e.g. XFROG) fail to determine the relative phase. However, as long as a well-characterized reference pulse with a bandwidth that covers all components of the test pulse is available, linear spectral interference can readily determine this relative phase.

© 2002 Optical Society of America