Photonics

Photonics

Simulation of periodic arrays of nanoparticles with TFSF source

    • mj.mendes
      Subscriber

      Following your suggestion of tips and best practices when using the FDTD TFSF source, we tried to implement the simulation in a way only the PML top and bottom borders (z axis) are outside the TFSF source and played with the boundary conditions, namely the number of layers, pml alfa, kappa, sigma… However, we were not successful in solving the converging issues towards the 1e-5 auto-shutoff level when the distance between particles is above 200 nm.

      In the pictures below and corresponding .fsp files we’re sending (identified with the corresponding figure number as in this post), you can see the different validation tests we performed: first a finite array of 10x10 particles (Fig. 1); second, a x, y periodic array, with all the DFT monitors surrounding the TFSF source, which in turn is surrounded by the FDTD region (Fig. 2); third, the suggested configuration (that also allows to precisely define the interparticle distance), where the x and y limits of the TFSF source are outside the DFT monitors and the FDTD region (Fig. 3). In the first case, the finite 10x10 array simulation shows no issues for either short (Fig. 1a) or large interparticle distance (Fig. 1b).

       

      Figure 1. Simulations for a 10x10 finite array: a) 50 nm interparticle distance; b) 500 nm interparticle distance

       

      However, when performing the simulations with periodic boundaries, either with the FDTD region surrounding all DFT monitors and the TFSF source (Fig. 2), or where the x and y limits of the TFSF source are outside the DFT monitors and the FDTD region (Fig. 3), the overall profile seems to agree with the 10x10 finite array for short interparticle distances (Fig. 2a and Fig. 3a), but some interference peaks appear, which are more evident the larger the interparticle distance is (Fig. 2b and Fig. 3b) and are probably a consequence of the lack of convergence. On the accompanying .fsp files you can also see we employed a high number of frequency points, and we tested all the way to the maximum number of PML layers. The scattering cross section calculations can be performed with the accompanying script (calcScatt.lsf).

      Figure 2. Simulations for a x, y periodic array (PML boundaries for z min and z max) and all DFT monitors placed inside the TFSF source: a) 50 nm interparticle distance; b) 500 nm interparticle distance

       

      Figure 3. Simulations for a x, y periodic array (PML boundaries for z min and z max) and the z-top and z-bottom DFT monitors placed outside the TFSF source: a) 50 nm interparticle distance; b) 500 nm interparticle distance

       

      We ask for your assistance and advise on how we can improve our periodic simulations to overcome the lack of convergence.

       

      Table 1. Correspondence between Figures and the provided simulation files

      # Figure

      Lumerical .fsp file

      Figure 1a

      10x10_50nm_(Fig1a).fsp

      Figure 1b

      10x10_500nm_(Fig1b).fsp

      Figure 2a

      periodicBC_TFSF IN_50nm_(Fig2a).fsp

      Figure 2b

      periodicBC_TFSF IN_500nm_(Fig2b).fsp

      Figure 3a

      periodicBC_TFSF x,y OUT_50nm_(Fig3a).fsp

      Figure 3b

      periodicBC_TFSF x,y OUT_500nm_(Fig3b).fsp

    • Guilin Sun
      Ansys Employee

      We are not allowed to download custmer files in this forum. If the problem persists, please send us emails if you have premium support.

      First, we should realize that, a 10X10 simulation with PML is different from perioidc structure with periodic BCs. This may seem not intutitive, think the diffraction of 10 slits and infinite number of slits. The intenisty of 10X10 has continuous intenisty in the far field, say 1m away. However, the periodic one gives only grating orders, with isolated diffraction angles.

       

      So the first thing first is to make sure you want to simulate finite-sized or infinite number of periods. If there is large angle radiation, you may need to use SteepAngle PML with more PML layers, and move PML at least half away from the scatter.

       

Viewing 1 reply thread
  • You must be logged in to reply to this topic.