Status of magnum.fe

Since our last post, magnum.fe has evolved quite a bit. However, we currently do not have the resources the maintain the open source version, so we decided to officially pause the development of the open-source branch.

The closed-source version works with the latest version of FEniCS (2016.2) and has some exciting new features, e.g.

  • Self-consistent spin diffusion that solves for electric potential and spin accumulation
  • Spin-Hall and inverse spin-Hall effect
  • Oersted field
  • String method solver for energy-barrier calculations
  • Slonczewski spin-torque
  • Zhang and Li spin-torque
  • Dzyaloshinskii-Moriya interaction

If you are interested in the full-featured version of magnum.fe, please feel free to contact us so we can provide you with an evaluation copy. Note that we updated the manual that now includes the documentation of the commercial version of magnum.fe.

magnum.fe, major update and more

Although being quiet for some time, we were very busy improving magnum.fe.

We introduced some major changes to the API including cached virtual attributes for intermediate results and we added a bunch of new algorithms and models.

Another big change is, that we decided to split up magnum.fe into an open source and a closed source branch. While the improved API was merged into the open source branch, algorithmic improvements and new models are currently only available in the closed source version.

Highlights of the closed source version include:

  • Fast preconditioned time integrator (up to 100x faster depending on the problem)
  • Self-consistent solution of spin diffusion and current distribution
  • Synthetic antiferromagnetic layers
  • Equilibrium treatment of spin diffusion

If you are interested in using the closed source version just drop us a line.

magnum.fe update for FEniCS 1.5

We updated magnum.fe to work with FEniCS 1.5 which was released this week. Most notable changes of the new release are:

  • The exchange-field computation is pulled in the ExchangeField class
  • The newly introduced point-integral measure is now used for nodewise operations
  • A Timer class is introduced for basic profiling

See the documentation for more details.

70 lines of NumPy

We present a very concise but complete micromagnetic finite-difference code in only 70 lines of NumPy (without comments). The code can be found on our GitHub page. A description of the project is available on arXiv (arXiv:1411.7188).

Publication on spin-diffusion algorithm

The preprint of our manuscript entitled “Self-consistent micromagnetic simulations including spin-diffusion effects” is now available on arXiv (arXiv:1410.6067). In this work present the spin-diffusion algorithm implemented in magnum.fe and perform numerical experiments comparing the model to the spin-torque model by Slonczewski as well as the model by Zhang and Li.