Roadmap for FDTDX

[ymahlau]

Since fdtdx is still quite a new framework a lot of features are still in development. I have been getting questions in different places what the plans for future development are. Therefore I want to give a list of features here that are planned and the rough timeframe that we deem realistic. That being said, we welcome any community contributions, so if you see a feature here that is planned in the far future but you need it now, then feel free to make a PR for that. If you want to implement any of these features let me know and I can provide help.

Additionally, if you dont see a feature in the list that you really need let us know and we will include it here.

Near Future development (upcoming months)

Feature	Status
GDSFactory Plugin (#1) / S-parameter calculation	In active development
Documentation (InvDes Tutorials, Component examples, Physics) (#6)	started
Dispersive Material Models (#4 )	started
Graphical User Interface (see here)	in active development
Non-uniform grid (#71)	planned, not started
Far field transformation utilities (#272)	planned, not started

Mid-Long Term future development (development might start in 6-12 months)

• Multi-GPU Scaling / Making fdtdx really fast (Make FDTDX extremely fast #104)
• Adjoint-based autodiff VJP implementation (Custom VJP for adjoint method #55)
• Simulation Symmetries to speed up simulations (Simulation Symmetries #267)
• Standalone Mode Solver

Long Term goals / goals without planned time frame

• Nonlinear material models (Plans for full anisotropic χ² tensor support? #271)
• Other simulation techniques: FDFD, RCWA, EME, ...

[michelfrising]

Hi Yannik and team,
thanks for this amazing project, I am really looking forward to use fdtdx more for my EM simulation projects! I was wondering if you have given any though to implementing the Eigenmode Expansion method in ftdx? I use it quite a lot to simulate waveguides and related structures where light is propagating into a preferential direction.
Curious to hear your thoughts on this!
Best,
Michel

[ymahlau]

Hi @michelfrising, thanks for the kind words! We are currently not planning to integrate EME into fdtdx, but rather focus on expanding the fdtd-specific features. That being said, I think for a lof of other simulation algorithms (FDFD, RCWA, EME, ...) there is a lack of good GPU-accelerated open-source tools right now. Therefore, we might start working on that at some point to fill that gap. Realistically, this will not start soon though.