Water vapor absorption in THz range

[Mikhail_MR]

Hi everyone!

Terahertz time-domain spectroscopy (THz-TDS) is a spectroscopic technique that is using the radiation in the THz range. https://en.wikipedia.org/wiki/Terahertz_time-domain_spectroscopy

A THz-TDS setup consists of an emitter and detector. The emitter creates a short broadband THz-pulse which propagates through a sample. The detector records this short pulse in the time domain (figure 1a in https://royalsocietypublishing.org/doi/full/10.1098/rspa.2007.0294). If we compare a reference signal (without sample) with a sample signal, we can obtain spectroscopic properties of a sample. We can also transform the signal into the frequency domain using the Fourier transform (figure 1b).

When a THz pulse propagates through the air, a part of it is absorbed by water vapor. It results in sharp absorption peaks in the frequency domain as it can be seen in figure 1b (solid line shows a signal with water vapor, the dashed line shows a signal without water vapor between emitter and detector). Sharp peaks in the frequency domain are translated to monochromatic oscillations (fluctuations) in the time domain.

The signal is recorded in the time domain. What is the physical reason for these fluctuations after the main pulse? I can think of two explanations, but I have arguments for and against both of them.

1. The absorbed radiation is re-emitted by water molecules after some time. Since water molecules absorb the radiation only in specific frequency regions, the re-emitted radiation consists of few sine oscillations. But how can I assume that the radiation is re-emitted towards the detector and not in any other direction? Can I somehow estimate the part that is radiated towards the detector?
2. The fluctuations are due to missing signal components at these frequencies. But why do the fluctuations after the main pulse stop after some time? Please take a look at figure 6a in the same paper. I would expect them to have the same amplitude until the end of the signal.

I would like to know that is the physical reason for the fluctuations after the main pulse in the time domain.

I appreciate any help.

Kind regards

 

[trurle]

I think the explanation (2) is closer to reality - frequency domain to time domain deconvolution error is to blame. The paper you references also use Gaussian window in time domain, therefore the response is shaped by window (resulting in response smoothing at high delay), not by a physical process.

 

[Mikhail_MR]

Thank you for your response. Could you please explain what deconvolution you have in mind. The figure 6a is a little bit misleading. The solid line shows a signal before Gaussian window was applied. As you can see, the fluctuations become less after some time.