Stabliization of simulation of a small period structure

I am try to repeat a simple simulaiton in a publication ( Effective medium properties and photonic crystal superstructures of metallic nanoparticle arrays).
The sphere diameter is 9.5 nm, and the lattice constant is 20 nm.
the materil follows the Drude model, where omega_p=1.37e16 (1/s), tao=4e-14s.
the surrounding is air. the mesh is 0.5n/grid. The following is the simulation configuration and the reported result
However, in my simulation, I got a lot of nosie (0.5nm). When I used a finer mesh (0.25nm), the noise became worse.

mesh = 0.5nm
I also tried to use stabilized PML layers, the result was as follows

this should be simple simulation. However, I cannot make it work well. I am not sure what is wrong here.
thanks very much for the help.
R_T_Sphere_a_20_d_9.5_n_1_JAP_01_08_2020.fsp (292.7 KB)

The problem for this simulation is, you used too fine mesh that PML is very thin! Thus PML reflection causes many ripples. When I use 1nm override mesh, mesh accuracy 3 and 64 PML layer for fast testing, I got

there are still many other ripples because even with the above settings, the total PML is only about 1/3 of the wavelength. TO improve, you will need

1: using thick PML through Custom PML with the PML parameters from Standard PML, but 128 layers or more, and make sure it has at least half wavelength;

2: When you use smaller mesh size for the particle, do not use very high back ground mesh accuracy, such as 8. usually 4 to 5 should be enough, as there are some other factors that will affect your simulation results compared with others, such as material data. More information can be found here Why my simulaitonm result is different from published paper or experiment?

3: When you use finer mesh, you may also need to increase zspan of the FDTD region, in order to let the PML located in uniform mesh region.

4: since the light interaction is mainly in the particle, you may need smaller autoshutoff min, in order to get correct peak value.

I attach the modified file for your reference, and please test, and then make some modifications to improve the result. If result gets worse, that means the modification is inappropriate.

One more thing about Ag material data: in our database, we have 3 sets of data from different labs. They are all from measurement. Do you know what the Ag refractive index the paper used? if you do not know, it will be difficult to duplicate their results.

thanks very much. this help me a lot. The rafractive index of Ag is given as a Drude model. so that should be fine. I think the unexpected peak my due to the starecase structure resulting from the mesh.

1 Like