Discussion Forum

I am currently attempting to simulate a multi-layered magnetic shield in 3D. The shielding consists of concentric, cylindrical layers of Mu Metal and Superconducting Niobium. I have approached the simulation with the magnetic fields ACDC module. So far, I have been able to simulate the behavior of the superconductor correctly using a frequency domain study with gauge-fixing. The mu metal also behaves correctly when I simulate it alone in a stationary study.

However, much to my dismay, combining the two materials in a single study produces one of the following outcomes:

1) Stationary Study: In this case the study converges in a reasonable amount of time, but the behavior of the superconductor is no longer modeled appropriately. In fact, the simulation seems to ignore the presence of any material at all in the domains with superconductor.

2) Frequency Domain Study: Here, the study simply fails to converge. The progress log of the GMRES error exhibits strange behavior, in that it repeatedly goes to a very small value (on the order of 10e-5) and then shoots up (order 100). The error in the non-linear solver slowly decreases, but eventually plateaus around 2. Halting the study mid-way when the GMRES is very low produces sensible results (both mu metal and superconductor acting appropriately) but I am not sure how accurate these results are.

Playing with the mesh in either case does not seems to help. The mesh always converges fine. In all cases, I am simply placing the materials in a uniform magnetic field. Any explanation would be very much appreciated.