I am trying to inject the mode of a plasmonic stripe WG(Au Palik) on top of a silica substrate(n = 1.44) with air as cladding. The mode source seems to find modes with negative refractive indices and when the simulation is run it diverges. I also used the COMSOL FEM software to check the modes of this structure and it finds the modes which have a different refractive index(the absolute value) than the ones found by Lumerical. I think the boundary conditions, mesh being not as fine as needed may be the possible culprits. What may I be doing wrong?

# Mode profile of a plasmonic stripe WG

**bkhanaliloo**#2

Hi @oarisev14

Can you please upload your simulation file for review? I think this would be the fastest way to hopefully troubleshoot the problem.

Thanks

**oarisev14**#3

Here is the simulation file attached.TaperedNanowireBasic.fsp (911.3 KB)

The modes calculated with Comsol FEM software is like the following:

**bkhanaliloo**#4

Dear @oarisev14

Many thanks to my colleague, @fgomez, it looks like the mode of interest interferes with the PML boundaries, and some nonphysical stuff happens. As a result the refractive index is negative. This is mainly because you have a silicon oxide substrate that conducts the light all the way down to to the PML layer.

To solve the problem I increase **both** simulation and ModeSource region, mainly in the -z direction, until the value behaves normal. I have the screenshot below with the value for *n* and also the mode profile for your reference. You can see that the mode has a very weak intensity close to PML layer in the screenshot.

I hope that this will solve the converging problem. However, if it wasn’t the case you can try convergence testing. Also, since in your simulation gold nanowire cuts through PML, you probably find this link interesting to read as we were dealing with the same diverging problem.

I hope this could answer your question.

**oarisev14**#5

Dear @bkhanaliloo

This workaround seems to tackle the issue, but since the mode source is always bounded by the simulation domain it will probably increase the simulation time. Apart from that will a finer simulation mesh improve this behavior? When I convert the plot to log scale I see that the fields still reach the boundaries. One more thing: is it possible to plot the fields at a particular z slice as a line plot before exporting the data?

**bkhanaliloo**#6

Dear @oarisev14

Yes, you are right. They will take longer as you need to have a larger simulation area so that the fields are weak enough when they reach the PML layer. As you said, even with the current simulation region, fields still reach the BCs, evident from log-scale plots, but are few orders of magnitude weak so we can ignore them.

I just used some finer mesh (for convergence purposes) and it doesn’t seem that results are changing much. So, the current mesh size looks to be reasonable.

Yes, please follow the steps as in the screenshots below. After you calculated the modes, select the mode of interest and then:

and then: