Installation

Requirements

• Python 3.10 or newer

• A working Quantum ESPRESSO or VASP installation (for DFT input generation; not required to import PAOFLOW itself)

• Core Python dependencies are installed automatically: numpy, scipy, mpi4py, pandas

---

Recommended: pip install

The simplest way to install PAOFLOW is from PyPI:

pip install PAOFLOW

This installs PAOFLOW and its mandatory runtime dependencies.

Optional dependency groups

PAOFLOW ships several optional extras. Install only what you need:

# Matplotlib + VTK plotting support
pip install "PAOFLOW[graphics]"

# Boltzmann transport extras (pydantic, pyyaml, psutil)
pip install "PAOFLOW[transport]"

# Fermi surface orbit analysis (pyskeaf)
pip install "PAOFLOW[pyskeaf]"

# Weyl point search (z2pack, tbmodels)
pip install "PAOFLOW[weyl_search]"

# JIT compilation via Numba (speed-up for dense k-grids)
pip install "PAOFLOW[fast]"

# Sparse matrix solvers (requires PETSc/SLEPc system libraries)
pip install "PAOFLOW[sparse]"

# All common extras at once
pip install "PAOFLOW[graphics,transport,pyskeaf,fast]"

---

Developer installation

To work on PAOFLOW itself, or to run the latest unreleased code from the repository:

git clone https://github.com/marcobn/PAOFLOW.git
cd PAOFLOW
pip install -e ".[dev,graphics,transport]"

The -e flag installs the package in editable mode — changes to src/PAOFLOW/ take effect immediately without reinstalling. The dev extra adds pre-commit and pytest.

---

Conda / virtual environment (recommended for HPC)

It is strongly recommended to work inside an isolated environment, especially on shared HPC clusters where MPI libraries must match the system’s MPI installation.

conda create -n paoflow python=3.12
conda activate paoflow

# Install MPI-compatible mpi4py via conda (links against the system MPI)
conda install -c conda-forge mpi4py

# Then install PAOFLOW from PyPI
pip install "PAOFLOW[graphics,transport]"

Warning

mpi4py and system MPI On HPC systems, pip install mpi4py may build against the wrong MPI library. Use conda install -c conda-forge mpi4py or load the correct MPI module and install mpi4py from source to ensure ABI compatibility.

---

Verifying the installation

Open a Python interpreter and run:

import PAOFLOW
print(PAOFLOW.__version__)

You should see the current version string (e.g. 2.9.3).

---