Diffraction artefacts from periodic structure?



I’m trying to simulate a periodic array of dielectric square patterns and experience spikes at random wavelengths. The spikes show themselves as dips in the reflection spectrum and are usually only 1 datapoint wide. Following suggestions in this forum, I increased the distance of the PML layers from the grating which did not change the shape of the spectrum in any way. The spikes also do not extend below 0 and the simulation converges well.

I was wondering if this might be an artefact from the FFT, but could not find any documentation for such a case. I was hoping someone here could help me. Unfortunately my simulation is part of a confidential project, so I cannot attach the model or the resulting spectrum. I can say however that apart from those spikes the experimental spectrum is well matched.

Many thanks,

Metal nano-strip을 이용한 시뮬레이션 상의 문제점

I might need to see the shape of your spectrum to get a better idea, but it looks like what you are observing might be the Wood’s anomaly, judging from the mention of ‘spikes’. Have a look at the following link for further information:

The field contributing to the Wood’s anomaly travels almost parallel to the grating, resulting in a quite poor absorption by PML. You can use ‘steep angle’ PML profile and increase the number of PML layers to improve the absorption, but you will never get a 100 % absorption for such a steep-angled incidence

Wood’s anomaly occurs at a wavelength where the number of orders the grating support changes. This abrupt changes in the number of orders occurs since the grating should meet the following relation:

If the grating period (=simulation span) is 1 um and the medium is air (n = 1), then

If you consider (+1,0) order,

So you will always see this abrupt changes in the number of order at a wavelength that is exactly the same as the period.


Thanks for your reply. My PML settings are already steep angle with 64 layers, otherwise it wouldn’t converge. I can match a few peaks to Wood’s anomaly, but there are many more spikes in the spectrum. I’ll see if I can post some of the data.
Is there any other possible reason for the spikes or is the steep angle diffraction the explanation? That way I could be sure it’s an intrinsic problem of the simulation, not something I missed.


This is the spectrum as returned from my reflection monitor. Assuming the peak labelled one is the first Wood anomaly, the others follow as p²+q² being the number written. I am not sure if the number 7 makes sense in this case, since it cannot come from square numbers.
But you can see that there are more random spikes that do not fit the physical explanation.


Thanks @cr2a13 for the spectrum data.

There can be so many factors that can affect the spectrum, resulting in spikes and ripples over the spectrum. To name a few, mesh size, pml-to-structure distance, number of pml layers, PMl types can all affect the results.The relatively large bandwidth and amplitude of the spikes as well as their positions seems to suggest that they cannot be attributed solely to the Wood’s anomaly.
To get to the bottom of the problem, the best way is to have a look at your simulation file. Would it be possible for you to remove some of your simulation objects that are confidential and make the simulation a bit more generic? As long as the result shows some of the spikes, it might be possible to find out the cause of the problem from the simplified version.