FixDt greatly affecting results

[Reiley Dorrian]

Hi everyone, I'm currently having trouble implementing temperature into my simulations, but the problem isn't with the temperature function itself. Rather, I have found that as soon as I declare FixDt to some value (which I understand is required in order to use finite temp) the results of my simulations change drastically, even when I've changed nothing else about the system. For instance, below is an output after running with adaptive timestep,

and here is an output for the exact same system (still at Temp = 0K) only with a fixed Dt=1e-12:

As you can see the results are just random noise, even when the temp is still set to 0. I believe the specifics of my system aren't relevant since I've been able to reproduce this problem on even very simple test scripts, so I think it may be something to do with my installation. I'm running mumax3.10 on Windows with CUDA V9.1.85.

If anyone has come across this issue before, I would love some advice as to how this can be fixed. I'm happy to share more specific information upon request. Thanks!

[Felipe Garcia]

Hi,

That is not an issue. Your timestep is rather large for a normal set of micromagnetic parameters. The fixdt has to be smaller than the one used in the adaptive method because it is lower order. If you note the dt in the web interface during the running with adaptive time step then normally it is between 1e-13 s and 1e-14 s. There is a test about this in the mumax3 paper. For some specific problems (large Aex) even smaller than that range. For a single moment maybe it works with 1 ps. But that is unlikely in the general case due to the presence of exchange. The test is that if noise appears for fixdt and T=0 the timestep is too large. So you should reduce the time step.

You can also try if one gets similar results with the newer time adaptive scheme with temperature. This does not require fixdt. I have no experience with it.

Best regards,

Felipe

--
You received this message because you are subscribed to the Google Groups "mumax2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mumax2+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/mumax2/48a42843-fc5b-4285-89f0-a4a367f7d85bn%40googlegroups.com.

[Sulaiman Alrisi]

any suggestion to add temp in finite time.