How can an ultrashort pulse be detected?

[Adel Makram]

How can an ultrashort pulse, such as in a time frame of femtosecond, be detected? Is it possible for a slow detector to pick up a signal in a time of femtosecond? How does an autocorrelation play a role here?

 

[blue_leaf77]

You mean be measured? There are a number of techniques that are able to do that task, some of the most popular: frequency resolved optical gating (FROG), SPIDER, and second order autocorrelation. Only the first two are able to reconstruct the electric field profile (the oscillation). 2nd order autocorrelation, owing to its symmetry property, can only give us the rough idea of how the pulse profile (the envelope only) looks like. As for the use of slow detector, generally slow detectors such as photodiode are too slow for ultrafast measurement, but I hear recently people are proposing the idea of using compressive sensing to accompany the slow detector, but I haven't gone into detai on this matter.

 

[Adel Makram]

You mean be measured? There are a number of techniques that are able to do that task, some of the most popular: frequency resolved optical gating (FROG), SPIDER, and second order autocorrelation. Only the first two are able to reconstruct the electric field profile (the oscillation). 2nd order autocorrelation, owing to its symmetry property, can only give us the rough idea of how the pulse profile (the envelope only) looks like. As for the use of slow detector, generally slow detectors such as photodiode are too slow for ultrafast measurement, but I hear recently people are proposing the idea of using compressive sensing to accompany the slow detector, but I haven't gone into detai on this matter.

What I know that autocorrelator (AC), can derive the power spectrum of the signal and tells the frequencies included within it ( as both AC and power spectrum are both Fourier pair). But how to measure or derive the pulse length of the signal? Does it use the frequency spectrum to derive the intensity as a function of time by inverse FT?

 

[blue_leaf77]

What I know that autocorrelator (AC), can derive the power spectrum of the signal and tells the frequencies included within it

That's the first order AC (aka field autocorrelation), but there is no demanding reason in calculating the power spectrum from the field AC as you already have a spectrometer.

But how to measure or derive the pulse length of the signal?

Using only the second order AC (aka intensity AC), it's theoretically not possible to derive the pulse length exactly. What is typically done is that we assume the actual pulse to have certain pulse shape, e.g. Gaussian, sech², or Lorentzian, then from the known relation between the pulse length of those respective pulse shape and the corresponding AC duration, one can derive the assumed pulse length. That's what you want to do when you only want to get the idea of how long the actual pulse spans in time and when the pulse haven't undergone too severe distortion such as temporal dispersion.

Does it use the frequency spectrum to derive the intensity as a function of time by inverse FT?

Blindly calculating the field from the IFT of the spectrum will only give you the transform-limited field profile, not the actual one in which various distortions may have modified the pulse.