Discussion Forum

Hi Gleb.

Well, it sounds like you are exactly where I was about 3 years ago. Unfortunately, there is no direct way (by means of predefined application modes) to model magnetoresistive devices because up to date, there is no module which offers the governing equations for non-linear magnetism most probably because there is no COMSOL solver that can deal well with FEM-BEM methods which is the most common approach here. So what you need to do is to implement the governing equations for ferromagnetic systems. In case that the electric current is linearly coupled to the magnetic distribution, you should be able to obtain magnetoresistance ratios via some sort of postprocessing step (such as integration of the angle between the magnetization vectors for double layers in TMR-devices, see e.g. doi:10.1088/1367-2630/11/11/113027) directly from the solution. If you are trying to study spin torque transfer phenomena, you will need an additional set of equations that are non-linearly coupled to your ferromagnetic application mode.

To give you some more background information, I attached two papers from the coming COMSOL conference 2011. The first one is about the implementation of the micromagnetic equations (Landau-Lifshitz-Gilbert) within a hybrid 2D/3D-frame (you will also find an explanation why this is helpful). Works fine as far as I can tell but acts rather stiff sometimes. The second one gives an introduction to FEM-BEM approaches and their implementation into COMSOL. However, I will not recommend this at the current stage. From my experience, unless you really succeed in setting up your problem perfectly, it will most probably decrease the performance rather than increasing it due to the increased density of the system matrix. Also, chances are high you will end up running into some numerical instabilities originating from singularities along the boundary integrals. On the other hand, it really helps to reduce DOFs and maybe allows for a total reformulation into a two-dimensional model which may be necessary if you need another set of equations for spin torque effects.

For the beginning, I would recommend to implement your thin film systems by the employment of two geometry frames, a two-dimensional frame for the calculation of the magnetic degrees of freedom, and a three-dimensional one for the solution of the magnetostatic equation. While you may use various modes for the magnetic field (magnetostatics, anything elliptic or simply weak: test(phix)*(phix-Ms*mx)+ … ), I am pretty convinced, weak form modeling is the only option you have for the micromagnetic equations due to the curl products in the LLG-equations. The coupling between the two frames can be done by extrusion coupling variables.

So far, this method works fine as long as you restrict yourself to reasonable size scales (~ 1 to 2 µm lateral, ~ 10nm perpendicular) and material parameters. Also, a standard guess of the magnetic initial values will result in inconsistent initial values. A good strategy of how to reach valid initial guesses by employment of a sequential presolver is described in the papers.

I hope this gives you something to start with. Let me know if you get stuck somewhere.

Kind regards,
Alex

Hello Gleb

I'm a beginner for COMSOL Multiphysics, and I need to simulate the 2D magnetoresistance of a PN Junction.I'm using the comsol 5.0, but I don't know which physics I should select in the "Add Physics" section in comsol.Fortunately,I found your post in the comsol discussion forum,so I decide to send the message to you,I hope that you can help me to solve the problem. If it's not a problem for you ,please describe in detail these steps of modelling. I will also very much appreciate for your corresponding recommendations in this modelling question.Thank you very much!

Sincerely,Huaa Wen