Solar spectrum reflectance of insulating nanoparticles


I am new to FDTD and have been using the free trial version of Lumerical 2020 to do the online course and try to replicate this result from the attached paper (A. Zhou et al. Energy and Buildings 138 (2017) 641–647):
Enhanced solar spectral reflectance of thermal coatings through inorganic additives.pdf (1.6 MB)

The authors used Meep and observed that TiO2 nanoparticles (100 nm radius, dielectric constant of 9) embedded in acrylic (dielectric constant of 2.5) paint increase the reflectance for the solar spectrum from around 15% up to 80%.

Unfortunately, I was not able to replicate this result and I have not been able to identify what I am doing wrong (.fsp file attached).
Mie_FDTD_2D.fsp (2.5 MB)
Mie_FDTD_2D_p0.log (10.3 KB)

I would like to know if these simulations are possible and viable to consider acquiring a license.

Thanks in advance.

Obviously this is a random structure. However FDTD solves Maxwell Equations deterministically, meaning, a given structure has a theoretically defined result (when delta_x and delta_t approach zero). Thus without detailed knowledge of the published structure, it will be hard to duplicate, if not possible. Please refer this link for some information: Why my simulaitonm result is different from published paper or experiment?

I would say that FDTD can simulate such devices.

To evaluate the capacity of a software, I would suggest that you do some simple testing. For sophisticated devices, you will need experience, knowledge and refine the settings in order to get reasonable result.

For example, normal simulation will require the PML some distance away from device:
such as half wavelength.

and log file shows simulation terminates too early:

the simulation time should be long enough to make sure the frequency domain result is accurate:

here is why we need longer simulation time
Simulation time and Frequency domain monitors

I would suggest that you first check all the online solar examples:

In order to know why I point out the above 3 issues, you may also need to learn some basic settings from various resources such as edu videos:

Thanks for your reply.
I had already checked your online solar examples, as well as some regarding solar cells and nanoparticles. Also, I completed the FDTD course. Everything works fine when I implement your exercises, but if I try to adapt to this system, I do not get similar reflectances.

Initially I did put the PML away from the device, but then close to the bottom of the material was the way I find to replicate the ~15% reflectance without nanoparticles reported in the paper. With the PML in air or in SiO2 I get 0%.
(without nanoparticles and PML in air or SiO2 below)

I tried some of the things that you suggested and observed no improvements (images and logs below).
Mie_FDTD_2D_p0_1000fs.log (10.7 KB)
Mie_FDTD_2D_p0_5000fs.log (11.9 KB)
Mie_FDTD_2D_p0_1000fs_shutoff1e-10.log (10.3 KB)
Mie_FDTD_2D_p0_1000fs_substrate_SiO2originaldatabase.log (10.3 KB)
Regarding the materials, I am using a TiO2 imported database and a copy of your SiO2 with tolerance of 0.01 for a better fit.



Unfortunately, my extended free license ends today, so I will not be able to do more tests.
I still do not understand if the problem is with the simulation regions and parameters or with the results analysis (or both).

I found somehow your file corrupted, as it terminates very soon. You can copy all the objects but FDTD, and then open a new project file, paste the objects, and then add FDTD and modify it.

Such simulation will need more than 1000fs as light will bounce back and forth in order to get correct frequency-domain result. A fast test is to disable the particles, and only test the slab, which has analytical solution (stackrt Then you can modify the settings until get reasonable result.Next simulate the actual device.

Please contact your account manager to request trial license extension to continue the test.