Overview#

MDPy is a freely distributed Molecular Dynamics (MD) package based on Python, which is GPU-accelerated and licensed under the BSD License. It is designed to enable researchers with no prior experience in coding with C++ or Fortran to implement new theoretical models with GPU acceleration efficiently.

Zhenyu Wei (zhenyuwei99@gmail.com), from Yunfei Chen’s group in Southeast University, developed and maintains MDPy. As a newly developed program, we always welcome developers to contact us or discuss in our Github site.

The document is organized in four parts:

  1. Quick start: This section provides an introduction to our program and installation suggestions.

  2. Theory behind MDPy: This section introduces the physical models and mathematical tricks implemented in MDPy.

  3. Tutorial: This section provides basic example codes to give users a clear understanding of MDPy’s features and workflow.

  4. API reference: This section gives a detailed description of all packages and classes of MDPy.

Tip

We highly recommend that beginners of MD simulation read the “Theory behind MDPy” before utilizing MDPy.

Tip

Once the user becomes familiar with MDPy, the “API reference” will be a useful resource for quickly checking their code.

Framework#

../_images/framework.png

Fig. 1 Framework of MDPy#

In MDPy, the main task of conducting a simulation is creating a well-defined Ensemble object. An Ensemble object includes all the information required for the MD simulation, such as:

  • Topology: This contains the particles’ properties, such as atom type, mass, and charge, as well as other topological information like bonds, angles, and dihedrals.

  • State: This contains the spatial coordinates and velocities of each particle in the Topology. This information describes the coordinate of the Ensemble in the phase space. Additionally, State also includes information about the Periodic Boundary Condition (PBC) and Neighbor list.

  • Constraints: This is a list of Constraint objects, each of which describes a specific interaction between particles. It returns the force on each particle and the potential energy of the Ensemble.

Tip

MDPy also provides a higher-level API for creating typical Ensemble, which is included in the mdpy.recipe package.

The user can then perform operations on the Ensemble object. MDPy supports two types of operators:

  • Minimizer: Takes the Ensemble object as input and minimizes the total potential energy of the Ensemble.

  • Integrator: Takes the Ensemble object as input and integrates the forces acting on each particle in a specific manner to sample the phase space.

Installation#

Beta version

We strongly recommend installing MDPy from the source code for the beta version. First, you need to clone this repository to your local device:

git clone https://github.com/mdpy-dev/mdpy.git

Then go to the path of mdpy, running:

conda install --file requirements.txt -c defaults -c conda-forge
conda develop .

Now, any changes made to the MDPy source code can be reflected in your code without needing to reinstall the MDPy package.

Performance#

Cite MDPy#