Failure to recreate plasmonic heating simulation in a paper

[g.w.castellanosgonza]

Hi all,

I am trying to benchmark plasmonic heating with dimers by recreating the result of this paper: Nanoscale control of optical heating in complex plasmonic systems (Baffou, Quidant, García De Abajo. ACS Nano, 2010). I extracted this 2 figures

If I run FDTD for the figure 5e then I obtain the same near field and the absorption cross-section is identical to figure 5a for cero degrees. However, if I calculate the absorbed power at the given intensity and run DEVICE (with water modelled via conduction no convection because they also model it via conduction), I obtain this temperature profile:

In my simulation the illumination is from beneath instead of from above but that just changes the orientation. If we look at figure 5a and 4b and 4c we see that the absorption cross-sections at 45 degrees are lower but nevertheless they report and increase in temperature of 80 degrees, which is in the order of my result for the cero degrees.

In summary I am not sure if my simulation is well or not as this paper seems to be ambiguous in their heat simulations. They don’t use lumerical but that should not be the cause of this difference. I would appreciate any help and insight with this paper. I also uploaded my optical and thermal simulation:

Test_Cabs.fsp (312.3 KB)

tre.ldev (6.2 MB)

Thank you very much!

Kind regards,

Gabriel

[aalam]

Hi @g.w.castellanosgonza, I have taken a look at your files and wanted to ask a few questions,

1. How did you get the source power value in your Pabs analysis group in the FDTD project file?

2. In your DEVICE project file how did you decide on the thermal conductivity value of water and how did you decide on the size of the simulation region? Does the paper give any information about these? (the size of the simulation region will have an impact on your final temperature).

[g.w.castellanosgonza]

The way I find the power value is as follows:

  Pabs = I * Area

where I = 1mW/um2 taken from the paper and Area = area of the TFSF perpendicular to the direction of the light propagation

That’s how I calculate the power value. Regarding the size, I am aware that it influences the thermal temperature. However, I found that beyond some value (typically a size span of 5 um the final temperature distribution does not change significantly

[aalam]

Hi @g.w.castellanosgonza, Thanks for the clarifications. How about the value of ‘k’ for water? Are you using the same value as they did in the paper? The reason I ask this is because if you increase the value of k the temperature will decrease (for example when I changed k from 0.6 to 1, the temperature dropped from 395 K to 358 K).

[g.w.castellanosgonza]

Hi @aalam, I have been doing an exhaustive check of that paper and I think that they made a typo in the scale of those graphs. There is figure 4b and 4c and when I simulate the same conditions the temperature obtained is of the same order.

However, regarding this thermal simulations I have another question: I am trying to simulate the temperature and obtain a temperature map with high detail, in particular close to the structure, but as I go for higher detail the simulation takes more and more time. What is the optimal/smart way to obtain high detail? So far I have played only with the min edge length and max edge length of the triangles. I am reading now about the mesh constraint. I am assuming this works in similar fashion than with FDTD and I can get a fine detail in the region I want. Is that right?

Thanks a lot for your help.

Best regards,

Gabriel

[aalam]

Hi @g.w.castellanosgonza, I am glad that you were able to verify your results. Regarding getting a finer mesh in a region of interest you have the right idea. The mesh constraint in DEVICE works in a manner similar to the one in FDTD. The main difference is that instead of defining the mesh size in all directions like FDTD, you will have to define a single parameter (max edge length) in the DEVICE mesh constraint which will limit the edge length of all the elements inside that volume. For your purpose you can therefore place the mesh constraint in the volume surrounding the spheres and restrict the size of the elements to a small value.

[g.w.castellanosgonza]

Hi @aalam, I have been doing simulations with the mesh constraint but I think it was not really doing anything as the resulting temperature map was still not smooth. Next figure illustrates the case of best resolution I could get. As you can say there are some irregularities:

The simulation I used to obtain this image is also uploaded:
Heat_GapSweep10_30_1.ldev (6.3 MB)

Is that the best way to obtain a fine detail of this simulation? Because if I reduce the numbers of the max edge length then the tessellating process takes forever and the simulation actually breaks.

Thanks for the help in advance!

Kind regards,

Gabriel

[aalam]

Hi @g.w.castellanosgonza, using the mesh override region is the correct approach here to refine the mesh around the spheres. I believe the problem in your case is coming from the fact that you are using a very small number for the deflection tolerance (1e-6 um) which somehow is affecting the overall meshing process in an adverse way. When I changed this deflection tolerance to 1e-4 um the mesh seems to be refining as expected by the mesh override object (screenshot below).

Reducing the max edge length is not a good approach here because that will make the mesh finer everywhere which will make the simulation very very large unnecessarily. When you want to refine the mesh in a selected region the mesh override region is the better option.

[g.w.castellanosgonza]

Hi @aalam,

Indeed, you are right. I tried with your suggestion and it worked very well. Thanks a lot!

Kind regards,

Gabriel