Shifting in absorbance curve


Dear Lumerical team,

I made a validation of solar cells paper. The structure of this paper is rectangular silicon with Al under it and ARC (Si3N4) on the top.

In the paper, there is a curve between wavelength and absorbance. My simulated curve by using FDTD is shifted and I don’t know what is the problem.

Can you help me in this problem, please?



Hi @yara_elsebaai,

It’s difficult to say without seeing the simulation file. That said, there’s quite a few parameters that can affect the simulation. Typically, once you know your simulation is correctly set up, it can be useful to perform a convergence testing to see how these parameters affect your results.

As a first step, maybe you could share your simulation file (in layout mode) so we can check it. Feel free to share the reference of the paper too.


Thank you for your reply.
These are the file and the paper.


Hi @yara_elsebaai,

I had a look at the file and I think the main issue is the dimensions of the FDTD region. If we look at your simulation:

The way it is set, you are modelling a periodic distribution of pillars, as there is a gap between the object and the FDTD periodic boundaries. To model the structure as intended (different layers extending in the xy plane), you just need to have the FDTD region to be the size of the unit cell you want to model. Without particle in the Si, the width of the FDTD region can be arbitrary, but when you add the particle, then the size of the FDTD region will define the period of your structure.

Note that you can extend the structure through the limit of the simulation region, it doesn’t matter.

Another point, you are using the “solar_generation” analysis group to calculate the absorption. The group covers the Si3N4 and Al layers. To get the absorption in Si only, you should set its dimension to the thickness of the Si layer only.

Also, as a start, you can use a coarse mesh so the simulations are faster.


Thank you for your reply.

I worked the modifications but the problem is still existing.


I think there’s 2 points that are not specified in the paper that can affect the simulation results:

  1. The mesh size
  2. The material properties: they don’t specify what data they are using for the different materials.

Moreover, the silicon nitride data you are using in the simulation is only defined for wavelength over 1.5um. It means there’s no valid data in the wavelength range of the simulation to create an accurate fit.

For these reason, it might be difficult to fully replicate the results presented in the paper.