Transmittance of circularly polarized light calculated with different wavelength ranges

Hi, everyone.

I’m trying to calculate transmittance of circularly polarized light passing through helical gold structures by referring the paper published in Science (fig. 1).

When I set the wavelength range as “3um to 10um”, the calculated transmittance looks similar to the figure in the Science paper.
However, when I expanded the wavelength range to “1um to 10um”, the transmittance for right-handed circularly polarized light exceeded 1.0, which is obviously strange.

I didn’t change any settings except for the wavelength range. Why does such invalid result occur?

3to10um_L.fsp (404.8 KB)
1to10um_L.fsp (406.1 KB)
3to10um_R.fsp (404.8 KB)
1to10um_R.fsp (406.1 KB)

Hi @Sho

Many thanks to @fgomez, I guess we finally could find the reason why this happens.

Here is the screenshot of the Material Explorer and I purposely include the longer wavelength response here:

As you can see, gold becomes very lossy at higher wavelength. For such a broad pulse, even if you set your wavelength to be from 1um to 10 um, injected pulse includes higher wavelengths component. On the other hand, the energy in your simulation region is calculated by summing over all the wavelength which directly effects the early termination (set by auto shutoff level). As a result the longer wavelength causes a fast decay on the energy and auto shutoff to trigger earlier. Thus your simulation does not run to completion (where all the frequency components decays completely) and you see transmissions larger than 1 at some wavelength.

Therefore, a careful attention to the injected pulse behavior is required. You can either change the pulse bandwidth or uncheck the optimize for short pulses from source Edit tab. Unchecking the optimize for short pulses will use a slightly longer pulse in time domain (narrower bandwidth) and will exclude the longer wavelength that causes the problem.

Please let me know if you still have questions.


Hi, @Sho

If you increase the frequency points of the monitor, the problem will be solved.

Hi, @bkhanaliloo
Thank you very much for your assistance.
Unchecking the optimize for short pulses worked well. In the case I unchecked that option, the results calculated with the wavelength range from 1 to 10um seemed to be reasonable and agreed with the results calculated with the range from 3 to 10um.

Well, is there any situation to avoid uncheking the optimize for short pulses option? I’m wondering in what case the option is appropriate, and in what case not appropriate.

Hi @shihwen

Thank you for your comment. Unfortunately, the increase in the frequency points didn’t improve the result. The transmission curves were basically same as before, though the calculated curves got smooth.

Thank you, @Sho,

I simulated different trials.
In the final trial, I increase the frequency points and accidentally uncheck the optimize for short pulses.
Sorry for leading the wrong solution.

Hi @Sho

Generally for cases that transmission plots are greater than 1, PML is the primary source (and first guess) of the problem. In this case I had doubts why changing the lower end of the bandwidth (from 3 um to 1um) results in transmissions greater than 1 at higher wavelengths while we deal with linear material in this case.

Material fit was the the second place to check. While you have set it properly, fit shows a significant increase in material loss at higher wavelengths. With some background knowledge about FDTD and knowing that we might have transmissions greater than for early shutoff, I guessed that the injected light (which has significant amplitude in higher wavelengths) is causing nonphysical behaviour. I hope this gives you some intuition, though it was the first case of its kind and took me a while to figure it out!

I generally don’t recommend to do very broadband simulations unless you do convergence testing. With such a broadband pulse you will need to have a close attention to material fit and use techniques to inject the most proper pulse (and sometime use shorter pulse).

Thank you very much for providing me with the 3-10 um pulse spectra, it was very useful to troubleshoot the problem.

Best of luck!