MicroMagnetic.jl

A Julia package for classical spin dynamics and micromagnetic simulations with GPU support.

Features

• Supports classical spin dynamics and micromagnetic simulations.
• Compatible with CPU and multiple GPU platforms, including NVIDIA, AMD, Intel, and Apple GPUs.
• Supports both double and single precision.
• Supports Monte Carlo simulations for atomistic models.
• Implements the Nudged-Elastic-Band method for energy barrier computations.
• Supports Spin-transfer torques, including Zhang-Li and Slonczewski models.
• Incorporates various energy terms and thermal fluctuations.
• Supports constructive solid geometry.
• Supports periodic boundary conditions.
• Easily extensible to add new features.

Installation

Install MicroMagnetic is straightforward as long as Julia (http://julialang.org/downloads/) is installed, and it is equally easy in Windows, Linux and Mac.

In Julia, packages can be easily installed with the Julia package manager. From the Julia REPL, type ] to enter the Pkg REPL mode and run:

pkg> add MicroMagnetic

Or, equivalently:

julia> using Pkg;
julia> Pkg.add("MicroMagnetic")

To install the latest development version:

pkg> add MicroMagnetic#master

To enable GPU support, one has to install one of the following packages:

For example, we can install CUDA for NVIDIA GPUs:

pkg> add CUDA

Now we will see similar messages if we type using MicroMagnetic

julia> using MicroMagnetic
julia> using CUDA
Precompiling CUDAExt
  1 dependency successfully precompiled in 8 seconds. 383 already precompiled.
[ Info: Switch the backend to CUDA.CUDAKernels.CUDABackend(false, false)

Quick start

Assuming we have a cylindrical FeG sample with a diameter of 100 nm and a height of 40 nm, we want to know its magnetization distribution and the stray field around it. We can use the following script:

using MicroMagnetic
@using_gpu() # Import available GPU packages such as CUDA, AMDGPU, oneAPI, or Metal

# Define simulation parameters
args = (
    task = "Relax", # Specify the type of simulation task (e.g., relaxation)
    mesh = FDMesh(nx=80, ny=80, nz=30, dx=2e-9, dy=2e-9, dz=2e-9), # Define the mesh grid for the simulation with 80x80x30 cells and 2 nm cell size
    shape = Cylinder(radius=50e-9, height=40e-9), # Define the shape of the magnetic structure as a cylinder with 50 nm radius and 40 nm height
    Ms = 3.87e5, # Set the saturation magnetization (A/m)
    A = 8.78e-12, # Set the exchange stiffness constant (J/m)
    D = 1.58e-3, # Set the Dzyaloshinskii-Moriya interaction constant (J/m^2)
    demag = true, # Enable demagnetization effects in the simulation
    m0 = (1,1,1), # Set the initial magnetization direction
    stopping_dmdt = 0.1 # Set the stopping criterion for the simulation based on the rate of change of magnetization dynamics
);

# Run the simulation with the specified parameters
sim = sim_with(args); 

# Save the magnetization and the stray field into vtk.
save_vtk(sim, "m_demag", fields=["demag"]) 

The magnetization and the stray field around the cylindrical sample are stored in m_demag.vts, which can be opened using Paraview.

Running MicroMagnetic.jl from Python

Thanks to PythonCall.jl, running MicroMagnetic.jl from Python is seamless.

Below is an example setup for Standard Problem #4:

from juliacall import Main as jl
jl.seval("using MicroMagnetic")

# Define simulation parameters
args = {
    "name": "std4",
    "task_s": ["relax", "dynamics"],                   # List of tasks to perform
    "mesh": jl.FDMesh(nx=200, ny=50, nz=1, dx=2.5e-9, dy=2.5e-9, dz=3e-9),  # Julia FDMesh object
    "Ms": 8e5,                                         # Saturation magnetization (A/m)
    "A": 1.3e-11,                                      # Exchange stiffness constant (J/m)
    "demag": True,                                     # Enable demagnetization
    "m0": (1, 0.25, 0.1),                              # Initial magnetization vector
    "alpha": 0.02,                                     # Gilbert damping coefficient
    "steps": 100,                                      # Number of dynamic simulation steps
    "dt": 0.01 * jl.ns,                                # Time step size (0.01 ns)
    "stopping_dmdt": 0.01,                             # Stopping criterion for relaxation
    "dynamic_m_interval": 1,                           # Save magnetization at each step
    "H_s": [(0, 0, 0), (-24.6 * jl.mT, 4.3 * jl.mT, 0)]  # Sequence of applied magnetic fields
}

# Run the simulation
sim = jl.sim_with(**args)

Structure of MicroMagnetic.jl