In preparation for the upcoming Frontiers in Optics conference, I’ve been experimenting with the cross-correlation frequency-resolved optical gating (XFROG) algorithm. We’d like to prove that it can be used to measure extremely complex laser pulses with time-bandwidth products of ~1000. Our goal is to use XFROG to measure supercontinua. Supercontinua are the result of a laser pulse that has undergone a dizzying array of both linear and nonlinear processes and, ultimately, forms a very broadband source of light. We believe that using XFROG to measure supercontinua is the first step towards measure it with double-blind XFROG for which two pulses are simultaneously measured. This would allow for the measurement of the reference pulse and supercontinua simultaneously with a fairly simple device.
Currently, I am running some tests on how the reference pulse width effects the performance of the retrieval algorithm. I’ve generated simulated pulses with time-bandwidth products of ~1000. Using these pulses along with transform-limited Gaussian references pulses of varying widths, I’ve generated FROG traces to be run through the algorithm. I’ve placed some examples of how the reference pulse width effects the polarization gating XFROG trace given by the equation below.
1/10th standard deviation of unknown I(t)
1/5th Std. Dev. I(t)
1/2 Std. Dev. I(t)
These FROG traces show intensity vs. angular frequency (vertical axis) and delay. It is clear that increasing the reference pulse width in time with respected to the unknown pulse washes out the temporal features as expected. However, the reverse will be the case for transform-limited references pulses that are broad in frequency compared to the unknown pulse. It is important to optimize the reference pulse for the best resolution is both domains.