Learning Forum

[debin.meng]

Dear Lumerical Team

I am trying to use 3D FDTD to reproduce some of the results shown in the published paper as given in the attached figure below.

[Guilin Sun]

There are a couple of issues in simulating band structure:
1: the purpose of the simulation time is to get the Fourier Transform of the time signal. It does NOT need to be long. If it is too long, some low Q modes will be missing from the bands. Please refer online examples. Usually the autoshutoff level can be larger than 0.9, instead of 1e-5 the default. Thus band structure simulation is one of the few examples that use the set simulation time to terminate the simulation.
2: this structure has symmetry in x,y and z. therefor there are many different modes mixed, eg, the so called TE and TM modes. you can use specific symmetry BCs to isolate them:
Symmetric and anti-symmetric BCs in FDTD and MODE
Bandstructure of photonic crystal waveguide - Line defect
Photonic crystal cavity
Please make sure the fundamental mode has a clear definition and can be quantified or justified.
Of course you may also use the dipoles with specific polarization to isolate the modes/bands.
3: Actually I did not see light line from your result. You can either plot them (may need to modify the script from the examples), or place some time monitors in the free space to simulate it.
4: it seems you used a lot of dipoles and time monitors. To speed up the simulation, you can reduce their number, about 10 for each should be ok.
5: once you can get clean band structures, and you only want to find the band gap, you do not need to sweep all the k points. Using the band edge k should be enough.
Please try and see if the results can be improved.

[debin.meng]

Dear Õ¡ÖÞÇüÕ©ê
I have tried your setup. Now I have a bandgap result as shown in the figure below. By setting the frequency range from 0 to 3.5e14 Hz, I can see the light line for the structure. However, there are lots of additional lines in the results in comparison to the results shown in the paper. Is this due to the definition of the source mode or due to the "findresonance" function? It is really hard for me to read the results in the figure. At the end of the plot where k =20, many points seem to be very close to each other while not 100% the same, should I treat them as the same one?
In this simulation, my setting is shown in the figure below. I use PML for both the y and z-axis. In order to simulate fundamental TE mode from left to right, should I use the setting like this or should I use antisymmetric for y min?

[debin.meng]

[Guilin Sun]

AS mentioned previously, using source polarization is one way, but the best way is to use symmetry BCs. try to use symmetrical and anti-symmetrical BCs in xy plane. You will need to know what is TE and which is TM. Probably you can get more info from the paper. Currently the bands are degenerated.