Au nanoparticle array partially embedded in dielectric slab

I simulated a Au nanodisk (D=165 nm) hex array (P=600 nm) partially embedded in dielectric slab (n=1.56). The result is overall same as my experiment, but has somewhat different in transmission as shown in the following capture. There is one extra peak at the edge of the valley at about 840nm in simulation. But no such peak happened in my experiment. Could you please help to figure out if the simulation has some setup problems that induce this peak? Many thanks!

The simulation file:nanodisk array buried in dielectric.fsp (6.0 MB)

The simulated and experimental transmissions:

The experimental transmissions:

Hi, I’ve had a look at your simulation file and nothing seems immediately obvious to me. I have also done some brief convergence tests on your original file but the peak @810nm seems pretty consistent.

This is the index xz prfile and you can see that the gold disk is sticking into the substrate:

This 810nm peak is very consistent in my tests.

This is the same xz index profile if you choose not to stick the gold disk in the susbtrate (by setting the mesh order).

Now the small peak @810nm is gone. File attached.nanodisk array buried in dielectric (1).fsp (1011.1 KB)

I wonder which design is original planned and what is being fabricated in experiment? Is it possible the fabricated device is somewhere halfway?

Hi Kchow,

Thanks for your quick reply. The gold disk should be embedded a little bit in the substrate.
I cannot confirm it but I believe it is the case, since there are valleys at 560 nm in both exp and simulation,.
In your simulation, no embedding and no valley appear. So I tend to accept that particles embedded in the substrate.

I changed the embedding depth to 15 nm. The attached is its spectrum. The 810nm peak disappeared, but the 840 nm valley’s shape has somewhat change. Is it possible to further improve the simulation to approach the experiment? Thanks!


I think so far it seems to suggest that disk is somewhat embedding. Your new plot from simulation looks pretty good, compared with experimental data. I guess you could play around with the depth embedded in the substrate and try to account for fabrication tolerance.