Pymatgen (Python Materials Genomics) is a robust, open-source Python library for materials analysis. These are some of the main features:

  1. Highly flexible classes for the representation of Element, Site, Molecule, Structure objects.
  2. Extensive input/output support, including support for VASP (, ABINIT (, CIF, Gaussian, XYZ, and many other file formats.
  3. Powerful analysis tools, including generation of phase diagrams, Pourbaix diagrams, diffusion analyses, reactions, etc.
  4. Electronic structure analyses, such as density of states and band structure.
  5. Integration with the Materials Project REST API.

Pymatgen is free to use. However, we also welcome your help to improve this library by making your own contributions. These contributions can be in the form of additional tools or modules you develop, or feature requests and bug reports. Please report any bugs and issues at pymatgen’s Github Issues page. If you wish to be notified of pymatgen releases and need help with using pymatgen, please become a member of pymatgen’s Google Groups.

The code is mightier than the pen.

Matgenie & Examples

The Materials Virtual Lab has developed a matgenie web app which demonstrates some of the basic functionality of pymatgen.

Another good way to explore the functionality of pymatgen is to look at the examples in a series of ipython notebooks. Below are a quick look at some of the graphical output possible.


Top: (left) Phase and (right) Pourbaix diagram from the Materials API. Bottom left: Calculated bandstructure plot using pymatgen’s parsing and plotting utilities. Bottom right: Arrhenius plot using pymatgen’s DiffusionAnalyzer.

Why use pymatgen?

There are many materials analysis codes out there, both commerical and free. So you might ask - why should I use pymatgen over others? Pymatgen offer several advantages over other codes out there:

  1. It is (fairly) robust. Pymatgen is used by thousands of researchers, and is the analysis code powering the Materials Project. The analysis it produces survives rigorous scrutiny every single day. Bugs tend to be found and corrected quickly. Pymatgen also uses CircleCI and Appveyor for continuous integration on the Linux and Windows platforms, respectively, which ensures that every commit passes a comprehensive suite of unittests. The coverage of the unittests can be seen at here.
  2. It is well documented. A fairly comprehensive documentation has been written to help you get to grips with it quickly.
  3. It is open. You are free to use and contribute to pymatgen. It also means that pymatgen is continuously being improved. We have a policy of attributing any code you contribute to any publication you choose. Contributing to pymatgen means your research becomes more visible, which translates to greater impact.
  4. It is fast. Many of the core numerical methods in pymatgen have been optimized by vectorizing in numpy. This means that coordinate manipulations are extremely fast and are in fact comparable to codes written in other languages. Pymatgen also comes with a complete system for handling periodic boundary conditions.
  5. It will be around. Pymatgen is not a pet research project. It is used in the well-established Materials Project. It is also actively being developed and maintained by the Materials Virtual Lab, the ABINIT group and many other research groups. The plan is to make sure pymatgen will stand the test of time and be the de facto analysis code for most materials and structural analysis.

With effect from version 3.0, pymatgen now supports both Python 2.7 as well as Python 3.x. For developers working to add new features to pymatgen, this also means that all new code going forward has to be Python 2.7+ and 3 compatible. Our approach is to have a single codebase support Python 2.7 and 3.x, as per current best practices. Please review the coding guidelines.

Change log


  • Improve support for alternative settings in SpaceGroup.
  • Fix respect for user_incar_settings in MPNonSCFSet and MPSOCSet
  • Support for argcomplete in pmg script.
  • Speed ups to Ewald summation.
  • Add functionality to parse frequency dependent dielectric function.
  • Improvements to Bolztrap support.

Older versions

Getting pymatgen

Standard install



Before installing pymatgen, you may need to first install a few critical dependencies manually.

  1. It is highly recommended that you use Python>=2.7.9 or >=3.5+. In fact, unless you are very sure you have dependencies that require Python 2, it is highly recommended that you use Python >=3.5+.
  2. Installation has been tested to be most successful with gcc. Use gcc where possible and do “export CC=gcc” prior to installation.
  3. Numpy’s distutils is needed to compile the spglib and pyhull dependencies. Numpy should be the first thing you install.
  4. One of the best things you can do is to use conda to manage all this. The instructions for Shared cluster installation pretty much applies to any machine as well.

The version at the Python Package Index (PyPI) is always the latest stable release that is relatively bug-free. The easiest way to install pymatgen on any system is to use pip:

pip install pymatgen

Wheels for Mac and Windows have been built for convenience.

The bleeding edge developmental version is at the pymatgen Github repo. The developmental version is likely to be more buggy, but may contain new features. The Github version include complete test files. After cloning the source, you can type in the root of the repo:

pip install .

or to install the package in developmental mode:

pip install -e .

Detailed installation instructions, including installation of option dependencies, set up for POTCAR generation, Materials Project REST interface usage, setup for developers, etc.are given on this page.

Shared cluster installation

If you are installing pymatgen on shared computing clusters, e.g., the XSEDE or NERSC resources in the US, the best way is to use conda to perform a local install. This guarantees the right version of python and all dependencies:

# Here, Python 3.5+ is assumed.
# You can change the following to Miniconda2 if you prefer Python 2.7
bash -b

# Reload bash profile.
source $HOME/.bashrc
source $HOME/.bash_profile

# Install numpy and other pydata stack packages via conda.
conda install --yes pymatgen

Using pymatgen

pymatgen overview

Overview of a typical workflow for pymatgen.

The figure above provides an overview of the functionality in pymatgen. A typical workflow would involve a user converting data (structure, calculations, etc.) from various sources (first principles calculations, crystallographic and molecule input files, Materials Project, etc.) into Python objects using pymatgen’s io packages, which are then used to perform further structure manipulation or analyses.

Quick start

Useful aliases for commonly used objects are now provided. Supported objects include Element, Composition, Structure, Molecule, Spin and Orbital. Here are some quick examples of the core capabilities and objects:

>>> import pymatgen as mg
>>> si = mg.Element("Si")
>>> si.atomic_mass
>>> print(si.melting_point)
1687.0 K
>>> comp = mg.Composition("Fe2O3")
>>> comp.weight
>>> # Note that Composition conveniently allows strings to be treated just
>>> # like an Element object.
>>> comp["Fe"]
>>> comp.get_atomic_fraction("Fe")
>>> lattice = mg.Lattice.cubic(4.2)
>>> structure = mg.Structure(lattice, ["Cs", "Cl"],
...                          [[0, 0, 0], [0.5, 0.5, 0.5]])
>>> structure.volume
>>> structure[0]
PeriodicSite: Cs (0.0000, 0.0000, 0.0000) [0.0000, 0.0000, 0.0000]
>>> # You can create a Structure using spacegroup symmetry as well.
>>> li2o = mg.Structure.from_spacegroup("Fm-3m", mg.Lattice.cubic(3),
                                        ["Li", "O"],
                                        [[0.25, 0.25, 0.25], [0, 0, 0]])
>>> # Integrated symmetry analysis tools from spglib.
>>> from pymatgen.symmetry.analyzer import SpacegroupAnalyzer
>>> finder = SpacegroupAnalyzer(structure)
>>> finder.get_spacegroup_symbol()
>>> # Convenient IO to various formats. You can specify various formats.
>>> # Without a filename, a string is returned. Otherwise,
>>> # the output is written to the file. If only the filenmae is provided,
>>> # the format is intelligently determined from a file.
>>> # Reading a structure is similarly easy.
>>> structure = mg.Structure.from_str(open("CsCl.cif").read(), fmt="cif")
>>> structure = mg.Structure.from_file("CsCl.cif")
>>> # Reading and writing a molecule from a file. Supports XYZ and
>>> # Gaussian input and output by default. Support for many other
>>> # formats via the optional openbabel dependency (if installed).
>>> methane = mg.Molecule.from_file("")
>>> # Pythonic API for editing Structures and Molecules (v2.9.1 onwards)
>>> # Changing the specie of a site.
>>> structure[1] = "F"
>>> print(structure)
Structure Summary (Cs1 F1)
Reduced Formula: CsF
abc   :   4.200000   4.200000   4.200000
angles:  90.000000  90.000000  90.000000
Sites (2)
1 Cs     0.000000     0.000000     0.000000
2 F     0.500000     0.500000     0.500000
>>> # Changes species and coordinates (fractional assumed for structures)
>>> structure[1] = "Cl", [0.51, 0.51, 0.51]
>>> print(structure)
Structure Summary (Cs1 Cl1)
Reduced Formula: CsCl
abc   :   4.200000   4.200000   4.200000
angles:  90.000000  90.000000  90.000000
Sites (2)
1 Cs     0.000000     0.000000     0.000000
2 Cl     0.510000     0.510000     0.510000
>>> # Replaces all Cs in the structure with K
>>> structure["Cs"] = "K"
>>> print(structure)
Structure Summary (K1 Cl1)
Reduced Formula: KCl
abc   :   4.200000   4.200000   4.200000
angles:  90.000000  90.000000  90.000000
Sites (2)
1 K     0.000000     0.000000     0.000000
2 Cl     0.510000     0.510000     0.510000
>>> # Replaces all K in the structure with K: 0.5, Na: 0.5, i.e.,
>>> # a disordered structure is created.
>>> structure["K"] = "K0.5Na0.5"
>>> print(structure)
Full Formula (K0.5 Na0.5 Cl1)
Reduced Formula: K0.5Na0.5Cl1
abc   :   4.209000   4.209000   4.209000
angles:  90.000000  90.000000  90.000000
Sites (2)
  #  SP                   a    b    c
---  -----------------  ---  ---  ---
  0  K:0.500, Na:0.500  0    0    0
  1  Cl                 0.5  0.5  0.5
>>> # Because structure is like a list, it supports most list-like methods
>>> # such as sort, reverse, etc.
>>> structure.reverse()
>>> print(structure)
Structure Summary (Cs1 Cl1)
Reduced Formula: CsCl
abc   :   4.200000   4.200000   4.200000
angles:  90.000000  90.000000  90.000000
Sites (2)
1 Cl     0.510000     0.510000     0.510000
2 Cs     0.000000     0.000000     0.000000
>>> # Molecules function similarly, but with Site and cartesian coords.
>>> # The following changes the C in CH4 to an N and displaces it by 0.01A
>>> # in the x-direction.
>>> methane[0] = "N", [0.01, 0, 0]

The above illustrates only the most basic capabilities of pymatgen. Users are strongly encouraged to explore the usage pages (toc given below).

API documentation

For detailed documentation of all modules and classes, please refer to the API docs.

More resources

The founder and maintainer of pymatgen, Shyue Ping Ong, has conducted several workshops (together with Anubhav Jain) on how to effectively use pymatgen (as well as the extremely useful custodian error management and FireWorks workflow software. The slides for these workshops are available on the Materials Virtual Lab.

pmg - Command line tool

To demonstrate the capabilities of pymatgen and to make it easy for users to quickly use the functionality, pymatgen comes with a set of useful scripts that utilize the library to perform all kinds of analyses. These are installed to your path by default when you install pymatgen through the typical installation routes.

Here, we will discuss the most versatile of these scripts, known as pmg. The typical usage of pmg is:

pmg {setup, config, analyze, plotdos, plotchgint, convert, symm, view, compare} additional_arguments

At any time, you can use "pmg --help" or "pmg subcommand --help" to bring up a useful help message on how to use these subcommands. With effect from v4.6.0, pmg also supports bash completion using argcomplete, which is useful given the many options available in the cli tool. To enable argcomplete, pip install argcomplete and either follow argcomplete’s instructions for enabling global completion, or add the following line to your .bash_profile (this method usually works more reliably):

eval "$(register-python-argcomplete pmg)"

Here are a few examples of typical usages:

#Parses all vasp runs in a directory and display the basic energy
#information. Saves the data in a file called vasp_data.gz for subsequent

pmg analyze .

#Plot the dos from the vasprun.xml file.

pmg plotdos vasprun.xml

#Convert between file formats. The script attempts to intelligently
#determine the file type. Input file types supported include CIF,
#vasprun.xml, POSCAR, CSSR. You can force the script to assume certain file
#types by specifying additional arguments. See pmg convert -h.

pmg convert input_filename output_filename.

#Obtain spacegroup information.

pmg symm -s filename1 filename2

#Visualize a structure. Requires VTK to be installed.

pmg view filename

#Compare two structures for similarity

pmg compare filename1 filename2

#Generate a POTCAR with symbols Li_sv O and the PBE functional

pmg generate --potcar Li_sv O --functional PBE


Some add-ons are available for pymatgen today:

  1. The pymatgen-db add-on provides tools to create databases of calculated run data using pymatgen.
  2. The custodian package provides a JIT job management and error correction for calculations.
  3. The pymatgen-diffusion by the Materials Virtual Lab provides additional useful analyses for diffusion in materials.


Pymatgen is developed by a team of volunteers. It is started by a team comprising of MIT and Lawrence Berkeley National Laboratory staff to be a robust toolkit for materials researchers to perform advanced manipulations of structures and analyses.

For pymatgen to continue to grow in functionality and robustness, we rely on other volunteers to develop new analyses and report and fix bugs. We welcome anyone to use our code as-is, but if you could take a few moment to give back to pymatgen in some small way, it would be greatly appreciated. A benefit of contributing is that your code will now be used by other researchers who use pymatgen, and we will include an acknowledgement to you (and any related publications) in pymatgen.

Reporting bugs

A simple way that anyone can contribute is simply to report bugs and issues to the developing team. You can either send an email to the `pymatgen’s Google Groups page`_ or even better, submit an Issue in our Github page.

Developing new functionality

Another way to contribute is to submit new code/bugfixes to pymatgen. The best way for anyone to develop pymatgen is by adopting the collaborative Github workflow (see contributing page).

How to cite pymatgen

If you use pymatgen in your research, please consider citing the following work:

Shyue Ping Ong, William Davidson Richards, Anubhav Jain, Geoffroy Hautier, Michael Kocher, Shreyas Cholia, Dan Gunter, Vincent Chevrier, Kristin A. Persson, Gerbrand Ceder. Python Materials Genomics (pymatgen) : A Robust, Open-Source Python Library for Materials Analysis. Computational Materials Science, 2013, 68, 314–319. doi:10.1016/j.commatsci.2012.10.028

In addition, some of pymatgen’s functionality is based on scientific advances / principles developed by various scientists. Please refer to the references page for citation info.


Pymatgen is released under the MIT License. The terms of the license are as follows:

The MIT License (MIT)
Copyright (c) 2011-2012 MIT & LBNL

Permission is hereby granted, free of charge, to any person obtaining a copy of
this software and associated documentation files (the "Software"), to deal in
the Software without restriction, including without limitation the rights to
use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of
the Software, and to permit persons to whom the Software is furnished to do so,
subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.


About the Pymatgen Development Team

Shyue Ping Ong of the Materials Virtual Lab started Pymatgen in 2011, and is still the project lead.

The Pymatgen Development Team is the set of all contributors to the pymatgen project, including all subprojects.

The full list of contributors are listed in the team page.

Indices and tables