Proposal of a micromagnetic standard problem for ferromagnetic resonance simulations

This repository accompanies the paper

"Proposal of a standard problem for ferromagnetic resonance simulations" by

Alexander Baker, Marijan Beg, Gregory Ashton, Maximilian Albert, Dmitri Chernyshenko, Weiwei Wang, Shilei Zhang, Marc-Antonio Bisotti, Matteo Franchin, Chun Lian Hu, Robert Stamps, Thorsten Hesjedal, and Hans Fangohr*

Journal of Magnetism and Magnetic Materials 421, 428-439 (2017), http://dx.doi.org/10.1016/j.jmmm.2016.08.009

Preprint available at http://arxiv.org/abs/1603.05419

*[email protected]

This repository provides data files and scripts which allow the reader to reproduce these results. Alternatively, you can use this code as a basis to apply it to their own micromagnetic problems.

Start by reading Quick start below to get an overview of the contents of this repository and the different ways in which you can use it.

---

Table of contents

• Quick start
• Prerequisites
• Repository structure
• Running the scripts and tests in this repository
• Detailed installation instructions for prerequisites

Quick start

Before you start, check the prerequisites to make sure you have the necessary software installed.

Depending on your interest and expertise, you can use this repository in different ways:

1. Download or browse the data files underlying the main figures 2-5 in the paper.

2. Re-produce figures 2-5 from our pre-computed reference data (no micromagnetic software needed):

   make reproduce-figures-from-oommf-reference-data

3. Run our micromagnetic simulation scripts to recompute the raw data files (OOMMF required):

   make recompute-oommf-data

   Then compare the recomputed data with our reference data (to verify that you obtain the same results on your computer):

   make compare-data

4. Produce figures 2-5 from the freshly computed micromagnetic simulation data from the previous step. This is useful to verify that you get the same output plots on your machine.

   make reproduce-figures-from-oommf-recomputed-data

5. If you have run the standard problem proposed in the paper with your own micromagnetic software, you can use our plotting code to visualise and compare the results. (Advanced.)

The remainder of this document describes the structure of this repository and gives a brief overview of its contents. Then we explain in more details the different ways in which you may want to use it.

Prerequisites

To run the code in this repository, the following software must be installed. For reference, we list the version numbers which we use for testing, but the code should work with most other versions as well.

• OOMMF (1.2 alpha 6 (30-Sep-2015), built from this tarball)
• Python (3.5.1)
• Python modules:
  • numpy (1.10.4)
  • scipy (0.17.0)
  • matplotlib (1.5.1)
  • py.test (2.8.5) (to run the automated tests)
• git (2.6.4) (optional, required to clone the repository. Download zip file as alternative.)
• Nmag (0.2.1) (optional)

The easiest and most convenient way of installing these prerequisites is by using conda as described in the detailed installation instructions below.

Using conda has multiple advantages: the installation works the same way for all operating systems, everything is installed locally in your home directory so that it does not interfere with your system installation, and you can easily remove everything should you wish to do so.

If you do not want to use conda then you can install the prerequisites manually, for example using pip (for Python modules) and/or using the package manager of your operating system.

Repository structure

The conceptual layout of this repository is as follows (we have omitted some files and directories that are not relevant to the end user).

.
├── micromagnetic_simulation_data/
│   ├── reference_data/
│   │   ├── oommf/
│   │   └── nmag/
│   └── recomputed_data/
│
├── figures/
│   ├── reference_plots_from_paper/
│   ├── generated_from_reference_data/
│   └── generated_from_recomputed_data/
│
├── src/
│   ├── micromagnetic_simulation_scripts/
│   │   ├── oommf/
│   │   └── nmag/
│   ├── postprocessing/
│   └── reproduce_figures.py
│
└── tests/

• micromagnetic_simulation_data/

  • reference_data/

    The "raw" micromagnetic simulation data that we used to produce figures 2-5 in the paper (it was generated using the scripts in the src/ folder). We provide reference data produced with both OOMMF and Nmag.

  • recomputed_data/

    Initially empty. When you run the micromagnetic simulation scripts in the src/ folder, they will place their output in this directory. This allows you to compare data computed on your machine to our reference data in order to check that you get the same results.

• figures/

  • reference_plots_from_paper/

    This folder contains the exact plots that were used for figures 2-5 in the paper (in .png and .pdf format). They were generated using the scripts in src/, applied to the OOMMF reference data.

  • generated_from_reference_data/

    generated_from_recomputed_data/

    Both of these folders are initially empty. When you run the plotting scripts in the src/ folder, they will place their output plots in these directories. This allows you to re-produce the plots from our paper (either using the reference data we provide, or using data that was re-computed on your own machine) to check that you get the same results.

• src/

  • micromagnetic_simulation_scripts/

    This folder contains scripts for both OOMMF and Nmag which implement the simulation setup for the proposed standard problem described in our paper. You can run these scripts to re-compute the "raw" data that serves as the basis for our figures.

  • postprocessing/

    This folder contains a small Python module to facilitate the reading of raw simulation data in various formats and plotting of the figures.

  • reproduce_figures.py

    A Python script which allows you to conveniently produce the plots for figures 2-5 from micromagnetic simulation data (both from our reference data and from re-computed data).

• tests/

  This directory contains our automated test suite. For the most part you can ignore the contents, but they provide an easy way to run the entire test pipeline and check that you can reproduce our results.

  Running the command

  make test
  

  will perform the following sub-steps.

  1. make unit-tests

     Runs a set of tests which check that our own code implementation works correctly. These should pass if you have all the prerequisites installed correctly. Therefore, an errors in this step probably indicates that something is wrong with your setup.

  2. make reproduce-figures-from-oommf-reference-data

     Runs the script src/reproduce_figures.py using our OOMMF reference data as input. Produces the plots for figures 2-5 and places them in the output directory figures/generated_from_reference_data/oommf/.

  3. make recompute-oommf-data

     Runs the simulation scripts in src/micromagnetic_simulation_scripts/oommf/ to re-compute the "raw" data using OOMMF. The resulting data files are placed in micromagnetic_simulation_data/recomputed_data/oommf/.

  4. make compare-data

     Compares the freshly computed data from step (iii) with our reference data to ensure that both coincide.

  5. make reproduce-figures-from-oommf-recomputed-data

     Runs the script src/reproduce_figures.py using the data computed in step (iii) as input. Produces the plots for figures 2-5 and places them in the output directory figures/generated_from_recomputed_data/oommf/.

Running the scripts and tests in this repository

Before you start, check the prerequisites to make sure you have the necessary software installed.

If you are using conda (see instructions below), make sure that your conda environment for this repository is activated:

source activate fmr-stdproblem

Clone this repository and change into the newly created directory.

git clone https://github.com/fangohr/micromagnetic-standard-problem-ferromagnetic-resonance.git
cd micromagnetic-standard-problem-ferromagnetic-resonance

If you want to run all the steps described below at once, simply run:

make all

This runs the following sub-steps, which you can also perform individually.

• Run the unit tests to check that everything is installed correctly. (This step is optional but recommended.)

  make unit-tests
  

• Reproduce the plots for figures 2-5 using our pre-computed reference data:

  make reproduce-figures-from-oommf-reference-data
  

  If the unit tests passed then this step should also work. Because this step uses pre-computed data it does not require any micromagnetic software to be installed. The resulting plots are placed in the directory figures/generated_from_reference_data/oommf/.

• Recompute the raw data by running the OOMMF simulation scripts:

  make recompute-oommf-data
  

  This will produce four "raw" data files (dynamic_txyz.txt, mxs.npy, mys.npy, mzs.npy) in the directory micromagnetic_simulation_data/recomputed_data/oommf/.

• Compare the freshly computed data from the previous step with our reference data to ensure that both coincide.

  make compare-data
  

  This reads the data files in the two directories micromagnetic_simulation_data/reference_data/ and micromagnetic_simulation_data/recomputed_data/ and compares them numerically. If the difference is above a small threshold (close to machine precision), the test fails.

• Reproduce the plots for figures 2-5 using the freshly computed reference data.

  make reproduce-figures-from-oommf-recomputed-data
  

  The resulting plots are placed in the directory figures/generated_from_recomputed_data/oommf/.

Detailed installation instructions for prerequisites

These instructions assume that you are on some kind of Linux/Unix system. While the code should certainly work on Windows, we have not tested this and some of the instructions below may need tweaking. Also, unfortunately we do not currently provide a conda package for OOMMF so that you need to install OOMMF yourself (see instructions on the OOMMF homepage). If you use Windows and find any missing steps then feel free to contact us, or even better submit a pull request (PR) for this repository.

The easiest way of installing all the prerequisites is using the package manager conda. The conda installer allows you to create dedicated Python environments very easily (similar to Python's virtualenv, but in a much cleaner and more powerful way). It also allows to install non-Python packages and thus provides an easy way of making OOMMF available.

Since conda does not touch your system installation at all and installs everything in a local directory (in a subfolder of your home directory by default), you can even use conda temporarily to test this repository. Afterwards you can delete the conda installation folder again, which will bring your system back into the original state.

Use the following steps to install conda and create a conda environment containing all required dependencies. (If you do not want to use conda then you will need to install these manually or via the package manager of your operating system.)

1. Install conda. There are two options for this:

• Install the full Anaconda Python distribution. This is almost 300 MB in size but comes bundled with a lot of Python packages useful for scientific computing.

• Install Miniconda. This is much smaller (ca. 30 MB) because it only makes the conda command available and leaves the installation of additional packages to you.

Either choice is fine. The installation works by simply downloading the installer from one of the links above and running it. The installer will not touch your system but install everything into a local folder (for example, ~/miniconda3 in your home directory). If you wish to get rid of your conda installation, simply delete this folder.

2. Make sure that your ~/.bashrc file contains a line similar to the following. The conda installer will typically offer to add this for you automatically.

export PATH=~/miniconda3/bin:$PATH

Note that the exact path may depend on whether you installed Miniconda or the full Anaconda distribution, so if you add this manually then make sure it points to the correct location of your installation (conda will print this information at the end of the installation procedure).

3. To activate the conda installation, run

source ~/.bashrc

Alternatively, opening a new terminal window is likely to achieve the same. This step makes the conda executable available in the terminal.

4. Create a new conda environment called `fmr-stdproblem` which contains all necessary packages (these are specified in the file `environment.yml`).

conda env create --name fmr-stdproblem -f environment.yml

5. Activate the newly created environment.

source activate fmr-stdproblem

This should provide all the necessary prerequisites. If you ever want to delete the conda installation,
simply remove the folder where conda was installed (for example, `~/miniconda3`) and remove the line
from your `~/.bashrc` file that was added in step 2 above.

----------

Acknowledgements: This work has been supported by the EPSRC through a
Doctoral Training Grant, and the Centres for Doctoral Training
EP/G03690X/1 and EP/L015382/1 .