Discussion Forum

Hi

thats not my field but you should scan the papers on the main web site of COMSOL and your library crossing the Landau-Lifshitz-Gilbert equation name and the term COMSOL (SciVerse of Elsevier should be helpful for that but it depends if your library is linked to their data base, there are others too ;)

--
Good luck
Ivar

Hi Jim,

I know, this might be a little late but if you are still working on this type of equations, maybe this one is helpful for you, too:

www.comsol.com/community/forums/general/thread/22468/

Kind regards,
Alex

Alex,

Great papers! I will take a look into how to model these using COMSOL. I assume since there is no preset physics module that you used the mathematics module. Is this correct? If so, how did you make sure the structural properties of LLG dynamics were preserved over a specific time? Initially, I was going to write a numerical scheme, but ended up just using analytical results for the 2D linearized equations because there were so many stability issues.

Thanks,
Jm

Hi Jim,

I guess the mathematics module is the 4.2 version of the original PDE mode, right? I am actually stuck at 3.5 since all my orignal model generators were written using COMSOL script because they could not be properly handled in the GUI anymore (some of the model files on magnetic nanoparticles contained several 100,000 characters in the expressions tab, actually takes half a minute just to open it). I never had any performance issues so I left it the way it was. Anyway, all the equations were implemented using weak form modeling. I am not absolutely sure but I think this is actually the only way to do so due to the curl products in the LLG-equations.

I am not quite sure what you mean by structural properties, the length constraint of the vector field? If so, there are two possibilities to implement this. Either by a adding a point-wise constraint in the set equations or by reformulation of the entire equation system into a polar representation. I used the first approach since for whatever reason I find it a lot easier to add additional contributions this way and I see myself getting really confused once I start to add any tensor-like quantities in regards to magnetostrictive effects. However, that is just personal preference and I see the second approach to be numerically more efficient.

Hope that answers your questions.
Kind regards,
Alex

Hi Alex,

Hope I can still find you here.
I'm learning to use COMSOL to do similar things.
Having read your papers, I am still confused on how to apply the length constraint to the vector field.
Will weak form do this?
If I use a pointwise constraint, should I choose 'bidirectional, symmetric' or 'user defined'?
If I have to use 'user defined', what to fill in 'constraint equation' and 'constraint force equation'?
Suppose I have a vector field with three component (ux,uy,uz), will 'ux*conj(ux)+uy*conj(uy)+uz*conj(uz)' do?
How about the weak constraint?
I hope there is not so big difference between 4.2a and 3.5, so that you can clearly get my question.
Thank you in advance.
:-)

Regards,
Kai

Hi Kai,
well this reply might be a little late and, hopefully, you could solve your problem in the meanwhile.
It's been quite a while but as far as I remember, I employed a pointwise constraint which was implemented via 'ux*ux + uy*uy + uz*uz - 1' (at least in 3.5 constraint expressions were set to 0, that's where the -1 comes from). I don't know about birectional, symmetric but if you can enter an expression in the user-defined option, try this one and see if it works.
I didn't quite get what you meant by 'Will weak form do this?'. The weak formulation of the LLG-equations will not maintain the vector modulus by itself but I think you will need to implement the equation via weak form modelling to deal with cross products. Indeed, if you do not enter an additional equation your set of dependent variables should be under-determined and you should end up with singular system matrices.

Hope that helps,
Alex