wakis+ECHO3D matching

[AlVMakarov]

Hello!

I am currently studying the nonlinear effects in linear accelerators. To verify my results, I use several simulation tools, including ECHO3D(https://echo4d.de) and CST Particle Studio, with which I have practical experience. As a relatively new user, I successfully replicated the results from simple example geometries provided with ECHO3D using both programs and obtained meaningful results for a real accelerating structure.

I found the wakis code very appealing due to its user-friendliness and open-source nature. I have also successfully configured the solver for a GPU cluster, which was convenient and opened new possibilities for research. However, when modeling the example geometry from ECHO3D, I encountered a significant discrepancy in the results between wakis and the ECHO3D/CST. Interestingly, for a real accelerator section model, the results are roughly comparable, but I am not satisfied with the level of agreement.

Is there a known explanation for this discrepancy? My initial hypothesis is that the codes might use different integration methods for calculating the wake potential, although I have not delved deeply into this topic.

1. If the methods indeed differ, are there any plans to add a new solver or integration scheme to wakis?
2. Alternatively, could the root cause be an incorrect simulation setup on my part?

[elenafuengar]

Hello!!

Thanks a lot for sharing your experience with Wakis so far :D I am glad you found it accessible and user friendly, that's definitely one of our code principles. We have been mainly benchmarking Wakis against CST Wakefield/eigenmode for cavity-like structures. This benchmark effort is ongoing, expanding to more complex geometries and materials, and I am discovering and fixing issues along the way. This said, I haven't benchmarked with ECHO3D yet, so this is a very exciting opportunity! To answer your questions:

1. Absolutely, we could think of adding the indirect (Napoly) method or the indirect testbeams, if there is a case that could not be solved by the currently implemented direct method.

2. Could you share which case you were trying to simulate, the relevant parts of your simulation setup (in case I can spot some issues) and maybe include some plots showing the discrepancy? You can paste/attach them directly to the comment.

I hope we can find the root cause of the discrepancy, and it would greatly improve the robustness of the code 🦾 Thanks again for reporting this!

[AlVMakarov]

Apologies for the delayed response;
I needed time to prepare the materials. For the time being I have selected a simple geometry for which I am calculating wake fields in three codes. However, during this period computational resources are busy, I have not yet received the calculation results for this geometry in wakis, but I can share the results from ECHO3D or CST, as well as describe the initial conditions I have used.

I have also delved deeper into the topic and found an interesting article. Perhaps you have already seen it, but I decided to share it with you. This article provides a rather good description of the mathematics for wake field integration methods.
DOI: https://doi.org/10.1103/PhysRevSTAB.9.102002