HMM structure FDTD simulation


#1

I am trying to simulate a HMM structure using FDTD solution.Here is my file.HMM2.fsp (1.4 MB).I get the idea of this device from this paper http://www.nature.com/nmat/journal/v15/n6/full/nmat4609.html
But ny result doesn’t resemble the reflectance spectrum.
Thanks.


仿真和文献差异很大,找不出原因
#2

Dear @1205067.ass

I took a look at your simulation file and found a couple of critical issues which without addressing them comparing the results will not be useful.

  • Plane wave with PML boundary: This will cause edge effects. If your device is periodic you need to change BCs to periodic, otherwise we need to use TFSF source.

  • You need to extend geometry into PML layer to avoid artificial reflections. Profile monitor needs to be extended as well so that you are capturing all the reflected light.

  • Geometry is not set properly. You need to carefully modify it so that it matches the paper.

I strongly recommend you to spend time with simulation file (specifically the geometry) and also take a look at our getting started and metamaterial application examples. Please modify your file and run the recommended tests and I will be glad to be of a help.

Thanks


#3

Thank you for your consideration.I modified the simulation file as HMM_biosensor.fsp (475.3 KB)
For using the wavelength sweep under optimization and sweep tab I get the reflectance spectrum as

this.I don’t understand why I don’t get the peak at (nearly)800nm and at 1200nm.


#4

Dear @1205067.ass

Thank you for modifying the simulation file. There is a big improvement.

There are a few things that still need to be modified:

  • Mesh: it is very coarse at the moment. You can start with coarse mesh for initial simulations, but you will need to use finer mesh to make sure that objects are resolved properly. You will also need to perform a convergence test to make sure that your results are converging and thus are reliable.
  • If you are using a broadband source at an angle, you need to use BFAST source
  • Make sure that material parameters agree with the paper ones. You can check the fitting from Materials tab.

Please keep me updated with the results and let me know if you had further questions.

Thanks


#5

Thank you for your advice.I have got some issues for the modification that you propose.After making the source type as BFAST and using finer mesh(keeping mesh accuracy 2 and setting a mesh override/increasing the mesh accuracy to 6 or 8),It shows that around 30-40 hours is needed for the simulation.Here is my simulation file , HMM_biosensor.fsp (357.8 KB)
My computer’s configuration is ,

Do I do any mistake or is it normal that my simulation will take this time?


#6

Dear @1205067.ass

Thanks for the modifications and updates. It looks like your geometry is not the same as the paper as they are not using a plane wave. This means that we will not be able to compare the results with the ones in the paer. Please correct me if I am wrong.

Regarding your simulation file: Since you are using a very broadband BFAST source with very fine meshes at a relatively steep angle of 50 degrees, as well as plasmonic effects with stacks that will trap light, simulations will take a long time. Here are a few things that I modified in your simulation file:

  1. use FDTD mesh accuracy of two, but introduced a mesh override region. I reduced PML to 8 for initial simulations.
  2. brought the source closer to the geometry to save light travel time in free space (a few mesh cells is enough)
  3. reduce the source bandwidth to be 500 - 700 nm. You can study it for different bandwidths, but this should be a good start.
  4. based on the maximum wavelength (700nm in this case) I left half wavelength gap (i.e. 350nm) between PML and critical objects. When you change the source bandwidth, you will need to update the FDTD region span.
  5. I modified monitor to capture only transmission, and recording data for 20 wavelengths

Now time for running simulations is ~ 7 hours (which I expect to finish earlier than that). The other option will be to use a single frequency plane wave source with Bloch BCs, and then perform a sweep over wavelength. Please note that due to nature of your geometry, you will need to run these simulations for a long time. I have the simulation file attached.

HMM_biosensor_Modified.fsp (359.2 KB)

Please keep me updated should you have any results of had any questions.

Thanks