Pymatgen

pymatgen

Overview

pymatgen is the standard Python library for materials science computation. Its core data model — Structure (periodic crystalline materials) and Molecule (non-periodic) — provides a unified representation for input/output across 30+ file formats (CIF, POSCAR/CONTCAR, XYZ, PDB, Gaussian, VASP). The library integrates with the Materials Project REST API (mp_api) to retrieve 150,000+ DFT-computed structures with band gaps, formation energies, and elastic constants. pymatgen is the foundation of the atomate2 and Custodian workflow frameworks for high-throughput DFT.

When to Use

• Parsing and converting crystal structure files between CIF, POSCAR, XYZ, and other formats
• Querying the Materials Project API for computed band gaps, formation energies, and stability data
• Constructing and analyzing phase diagrams and Pourbaix diagrams for thermodynamic stability
• Generating VASP, Quantum ESPRESSO, or CP2K input files from structure objects
• Computing X-ray diffraction (XRD) and neutron diffraction patterns for comparison with experiment
• Analyzing symmetry, space groups, and Wyckoff positions of crystal structures
• Use ASE when running molecular dynamics or interfacing with multiple MD/DFT codes via a unified runner

Prerequisites

• Python packages: pymatgen, mp-api (Materials Project client)
• Data requirements: structure files (CIF, POSCAR) or Materials Project API key
• API key: free at materialsproject.org — set PMG_MAPI_KEY env var

pip install pymatgen mp-api
# Set API key
export PMG_MAPI_KEY="your_api_key_here"
# Or via pymatgen config
python -c "from pymatgen.core import SETTINGS; SETTINGS['PMG_MAPI_KEY'] = 'your_key'"

Quick Start

from pymatgen.core import Structure, Lattice, Species

# Build silicon diamond cubic structure from scratch
a = 5.431  # Angstroms
lattice = Lattice.cubic(a)
silicon = Structure(
    lattice=lattice,
    species=["Si", "Si"],
    coords=[[0, 0, 0], [0.25, 0.25, 0.25]],
)
print(f"Silicon: {silicon.formula}, {silicon.volume:.2f} ų")
print(f"Space group: {silicon.get_space_group_info()}")
# Silicon: Si2, 40.89 ų
# Space group: ('Fd-3m', 227)

Core API

Module 1: Structure and Lattice

Core data structures for periodic crystals.

from pymatgen.core import Structure, Lattice, Element, Species
import numpy as np

# From lattice parameters
lattice = Lattice.from_parameters(a=4.05, b=4.05, c=4.05,
                                   alpha=90, beta=90, gamma=90)

# Build FCC aluminum
al_fcc = Structure(lattice, ["Al", "Al", "Al", "Al"],
                   [[0, 0, 0], [0.5, 0.5, 0], [0.5, 0, 0.5], [0, 0.5, 0.5]])

print(f"Formula: {al_fcc.formula}")
print(f"Sites: {len(al_fcc)}")
print(f"Volume: {al_fcc.volume:.3f} ų")
print(f"Density: {al_fcc.density:.3f} g/cm³")

# Access sites
for site in al_fcc:
    print(f"  {site.species_string} at {site.frac_coords}")

# Load from file
from pymatgen.core import Structure

# From CIF (most common exchange format)
struct = Structure.from_file("material.cif")

# From POSCAR (VASP format)
struct_vasp = Structure.from_file("POSCAR")

# Get neighbors within cutoff
site = struct[0]
neighbors = struct.get_neighbors(site, r=3.0)
print(f"Neighbors within 3 Å: {len(neighbors)}")
for nn in neighbors[:3]:
    print(f"  {nn.species_string}: {nn.nn_distance:.3f} Å")

Module 2: Materials Project API Query

Retrieve DFT-computed properties for 150,000+ materials.

from mp_api.client import MPRester
import os

api_key = os.environ.get("PMG_MAPI_KEY", "your_key")

with MPRester(api_key) as mpr:
    # Search by chemical system
    docs = mpr.materials.summary.search(
        chemsys=["Li-Fe-O"],
        fields=["material_id", "formula_pretty", "energy_above_hull",
                "band_gap", "is_stable"]
    )
    print(f"Li-Fe-O materials: {len(docs)}")
    for d in docs[:5]:
        print(f"  {d.material_id}: {d.formula_pretty}, "
              f"Eg={d.band_gap:.2f} eV, above_hull={d.energy_above_hull:.3f} eV/atom")

# Get specific material by MP ID
with MPRester(api_key) as mpr:
    doc = mpr.materials.summary.get_data_by_id(
        "mp-149",  # Silicon
        fields=["structure", "band_gap", "formation_energy_per_atom",
                "density", "is_stable", "symmetry"]
    )
    struct = doc.structure
    print(f"Si mp-149: band_gap={doc.band_gap:.3f} eV, "
          f"density={doc.density:.3f} g/cm³")
    print(f"Space group: {doc.symmetry.symbol}")

Module 3: Symmetry Analysis

from pymatgen.symmetry.analyzer import SpacegroupAnalyzer
from pymatgen.core import Structure

struct = Structure.from_file("material.cif")

# Symmetry analysis
sga = SpacegroupAnalyzer(struct, symprec=0.1)

print(f"Space group: {sga.get_space_group_symbol()} ({sga.get_space_group_number()})")
print(f"Crystal system: {sga.get_crystal_system()}")
print(f"Point group: {sga.get_point_group_symbol()}")

# Get conventional / primitive cell
primitive = sga.get_primitive_standard_structure()
conventional = sga.get_conventional_standard_structure()
print(f"Primitive: {len(primitive)} sites | Conventional: {len(conventional)} sites")

# Wyckoff positions
sym_dataset = sga.get_symmetry_dataset()
print(f"Wyckoff letters: {set(sym_dataset['wyckoffs'])}")

Module 4: Phase Diagrams

Thermodynamic stability and phase boundary analysis.

from pymatgen.analysis.phase_diagram import PhaseDiagram, PDPlotter
from mp_api.client import MPRester
import os

api_key = os.environ.get("PMG_MAPI_KEY", "your_key")

with MPRester(api_key) as mpr:
    # Get all entries in the Li-Fe-P-O chemical system
    entries = mpr.get_pourbaix_entries(["Li", "Fe"])

# For phase diagram, use computed entries
with MPRester(api_key) as mpr:
    entries = mpr.get_entries_in_chemsys(["Li", "Fe", "O"])

pd = PhaseDiagram(entries)
print(f"Stable phases: {len(pd.stable_entries)}")

# Check stability of a specific composition
from pymatgen.core import Composition
comp = Composition("LiFeO2")
e_hull = pd.get_e_above_hull(pd.qhull_entries[0])
print(f"E above hull: {e_hull:.3f} eV/atom")

# Plot (requires matplotlib)
plotter = PDPlotter(pd, backend="matplotlib")
plotter.show()

Module 5: XRD Pattern Simulation

from pymatgen.analysis.diffraction.xrd import XRDCalculator
from pymatgen.core import Structure
import matplotlib.pyplot as plt

struct = Structure.from_file("material.cif")  # or build programmatically

# Calculate XRD pattern (Cu Kα radiation, λ = 1.5406 Å)
calculator = XRDCalculator(wavelength="CuKa")
pattern = calculator.get_pattern(struct, two_theta_range=(10, 80))

print(f"Diffraction peaks: {len(pattern.x)}")
for two_theta, intensity, hkl in zip(pattern.x[:5], pattern.y[:5], pattern.hkls[:5]):
    print(f"  2θ={two_theta:.2f}°, I={intensity:.1f}, hkl={hkl}")

# Plot
fig, ax = plt.subplots(figsize=(10, 4))
ax.bar(pattern.x, pattern.y, width=0.3, color="black")
ax.set_xlabel("2θ (degrees)")
ax.set_ylabel("Intensity (arb. units)")
ax.set_title(f"XRD Pattern — {struct.formula}")
plt.tight_layout()
plt.savefig("xrd_pattern.pdf", bbox_inches="tight")

Module 6: DFT Input File Generation

Generate VASP input sets for DFT calculations.

from pymatgen.io.vasp.sets import MPRelaxSet, MPStaticSet
from pymatgen.core import Structure

struct = Structure.from_file("material.cif")

# Generate VASP relaxation input set (Materials Project standard)
relax_set = MPRelaxSet(struct)

# Write to directory
import os
os.makedirs("vasp_relax", exist_ok=True)
relax_set.write_input("vasp_relax")
print("Generated: POSCAR, INCAR, KPOINTS, POTCAR (requires VASP pseudopotentials)")

# Inspect key INCAR settings
incar = relax_set.incar
print(f"ENCUT: {incar.get('ENCUT')} eV")
print(f"KPOINTS: {relax_set.kpoints}")

# For static calculation after relaxation
static_set = MPStaticSet.from_prev_calc(prev_calc_dir="vasp_relax")
static_set.write_input("vasp_static")

Key Concepts

Fractional vs Cartesian Coordinates

pymatgen Structure stores atomic positions in fractional coordinates (relative to lattice vectors, range 0–1). Convert to/from Cartesian (Angstroms) using struct.lattice.get_cartesian_coords(frac) or struct.lattice.get_fractional_coords(cart). Most file formats use Cartesian; pymatgen converts automatically on read/write.

Composition and Oxidation States

Composition("LiFePO4") parses chemical formulas. Structure.add_oxidation_state_by_guess() uses bond valence to assign formal charges (+Li, -O, etc.) needed for Pourbaix diagrams and some property calculations.

Common Workflows

Workflow 1: High-Throughput Stability Screen

from mp_api.client import MPRester
from pymatgen.analysis.phase_diagram import PhaseDiagram
import pandas as pd, os

api_key = os.environ.get("PMG_MAPI_KEY", "your_key")

# Screen lithium-transition-metal oxides for stability
systems = [f"Li-{m}-O" for m in ["Mn", "Co", "Ni", "Fe", "V"]]
results = []

with MPRester(api_key) as mpr:
    for system in systems:
        docs = mpr.materials.summary.search(
            chemsys=[system],
            fields=["material_id", "formula_pretty", "energy_above_hull",
                    "band_gap", "is_stable", "formation_energy_per_atom"]
        )
        for d in docs:
            results.append({
                "system": system,
                "mpid": d.material_id,
                "formula": d.formula_pretty,
                "e_above_hull": d.energy_above_hull,
                "band_gap": d.band_gap,
                "stable": d.is_stable,
            })

df = pd.DataFrame(results)
stable = df[df["stable"] == True].sort_values("band_gap")
print(f"Stable phases: {len(stable)}/{len(df)}")
print(stable[["formula", "system", "band_gap", "e_above_hull"]].head(10))
stable.to_csv("stability_screen.csv", index=False)

Workflow 2: Structure Manipulation and Export

from pymatgen.core import Structure
from pymatgen.transformations.standard_transformations import (
    SupercellTransformation, SubstitutionTransformation
)

# Load and analyze structure
struct = Structure.from_file("material.cif")
print(f"Original: {struct.formula}, {len(struct)} sites")

# Create 2×2×2 supercell
sc_matrix = [[2, 0, 0], [0, 2, 0], [0, 0, 2]]
supercell = SupercellTransformation(sc_matrix).apply_transformation(struct)
print(f"Supercell: {len(supercell)} sites")

# Substitute element (e.g., 10% Fe doping on Mn sites)
sub = SubstitutionTransformation({"Mn": {"Mn": 0.9, "Fe": 0.1}})
doped = sub.apply_transformation(struct)
print(f"Doped composition: {doped.composition.reduced_formula}")

# Export in multiple formats
struct.to(filename="output.cif")            # CIF
struct.to(filename="POSCAR")               # VASP POSCAR
struct.to(filename="output.xyz")           # XYZ
print("Exported CIF, POSCAR, XYZ")

Key Parameters

Common Recipes

Recipe: Batch CIF to POSCAR Conversion

from pymatgen.core import Structure
from pathlib import Path

cif_dir = Path("cif_files")
poscar_dir = Path("poscar_files")
poscar_dir.mkdir(exist_ok=True)

for cif_path in cif_dir.glob("*.cif"):
    try:
        struct = Structure.from_file(str(cif_path))
        out_path = poscar_dir / f"{cif_path.stem}_POSCAR"
        struct.to(filename=str(out_path))
        print(f"Converted: {cif_path.name} → {out_path.name}")
    except Exception as e:
        print(f"FAILED {cif_path.name}: {e}")

Recipe: Get Band Gap for List of Materials

from mp_api.client import MPRester
import pandas as pd, os

mp_ids = ["mp-149", "mp-2815", "mp-1265"]  # Si, GaAs, TiO2

api_key = os.environ.get("PMG_MAPI_KEY", "your_key")
with MPRester(api_key) as mpr:
    docs = mpr.materials.summary.get_data_by_ids(
        mp_ids, fields=["material_id", "formula_pretty", "band_gap", "is_gap_direct"]
    )

df = pd.DataFrame([{
    "mpid": d.material_id,
    "formula": d.formula_pretty,
    "band_gap_eV": d.band_gap,
    "direct_gap": d.is_gap_direct,
} for d in docs])
print(df.to_string(index=False))

Troubleshooting

Related Skills

• pymoo — multi-objective optimization for materials property screening using pymatgen descriptors
• autodock-vina-docking — structure preparation workflow analogous to pymatgen for molecular docking
• zarr-python — efficient storage of large arrays from MD trajectories or property databases

References

• pymatgen documentation — full API reference, tutorials, and compatibility matrix
• Materials Project API docs — REST API and mp-api client reference
• pymatgen paper: Ong et al. (2013), CMS — original publication and design philosophy
• Materials Project paper: Jain et al. (2013), APL Materials — database description and computed properties
• pymatgen GitHub — source code and examples