Learning Forum

[smoa]

I am writing to inquire about the limitations of GPU-accelerated FDTD simulations as specified on your website: https://optics.ansys.com/hc/en-us/articles/17518942465811-Getting-started-with-running-FDTD-on-GPU 

Specifically, I am interested in simulating FDTD with library metals such as Aluminum (Al) using CRC data. I would like to know if the presence of negative permittivity materials will inherently cause the simulation to diverge.

Additionally, the website mentions the requirement for narrowband simulations. Could you please clarify if this refers to simulations with a single wavelength, or what constitutes a "narrowband" range in this context?

 

Thank you for your time and attention to this matter.

[Amrita Pati]

Hi,

Specifically, I am interested in simulating FDTD with library metals such as Aluminum (Al) using CRC data. I would like to know if the presence of negative permittivity materials will inherently cause the simulation to diverge.

The use of metals with negative permittivity doesn't inherently cause divergence. That said, usually these metals are highly dispersive and can sometimes cause instability in the simulation. In these cases, the first thing we can do is identify what is causing the divergence. The two main categories of divergence are linked to 1. PML Boundaries 2. dt stability factor. Once we figure out which of these is causing the issue we can take steps to address it. This article provides information in how to identify the type of divergence and then the steps to solve the issue: Troubleshooting diverging simulations in FDTD. Please have a look and feel free to let me know if there are any questions.

Additionally, the website mentions the requirement for narrowband simulations. Could you please clarify if this refers to simulations with a single wavelength, or what constitutes a "narrowband" range in this context?

The GPU-implementation of FDTD doesn't support the multi-coefficient material model at the moment. So, even if you use the Al (CRC) data, only the material data at a single wavelength is used. So, by narrowband we mean that as long as the material can be approximated by a single material data point within a wavelength range, it should be fine. In some cases, it could be a narrowband source like 100 nm, but if the material is highly dipersive in the wavelength range you are considering then it is not a good idea to use a single material data point over the entire source bandwidth, and in that case we should perform single frequency simulation instead. 

But the good news is that the GPU-implentation of FDTD will start supporting the multicoefficient model very soon (most likely in the next release in January), so we will no longer have this limitation, which means that you would be able to simulate a wideband source with a single simulation on GPU.

Regards,
Amrita