ABINIT

ABINIT
Name	ABINIT
Author	Xavier Gonze
Developer	European Theoretical Spectroscopy Facility
Released	1997
Programming language	Fortran
Operating system	Linux, Unix, macOS
Genre	Electronic structure
License	GNU General Public License

ABINIT

ABINIT is an open-source suite for computing the electronic properties of materials using first-principles methods. It performs self-consistent calculations of total energies, charge densities, and electronic structures using plane-wave pseudopotential approaches and many-body perturbation theory, widely used across computational physics, materials science, and chemistry. Major user communities include researchers associated with CNRS, University of Liège, CEA, École Polytechnique, and national laboratories such as Lawrence Berkeley National Laboratory, Argonne National Laboratory, and Brookhaven National Laboratory.

Overview

ABINIT implements methods based on Kohn–Sham equations, Density Functional Theory, and extensions including Hartree–Fock method, GW approximation, and Bethe–Salpeter equation. The package supports norm-conserving pseudopotentials and projector augmented-wave datasets provided by groups at P. J. Hay, David Vanderbilt, and institutions like Quantum ESPRESSO partner projects. Users often compare ABINIT results with codes from VASP, SIESTA, CASTEP, GPAW, and WIEN2k. Funding and collaborations have involved agencies such as the European Commission, ANR, NIH, and national research councils including EPSRC and NSF.

Features and Capabilities

ABINIT provides ground-state total-energy calculations, structural relaxations, phonon spectra via Density-Functional Perturbation Theory, and response functions for dielectric tensors and Born effective charges. It includes many-body techniques like one-shot G0W0 and self-consistent GW and optical spectra via the Bethe–Salpeter equation formalism. ABINIT supports molecular dynamics (Born–Oppenheimer and Car–Parrinello flavors), non-collinear magnetism with spin–orbit coupling, and constrained-DFT schemes used in studies linked to Hubbard U corrections and DFT+U workflows. Integration with pseudopotential libraries and interoperability with formats used by Materials Project, AFLOW, NOMAD and databases at Harvard University enhances high-throughput screening, machine-learning workflows with groups at DeepMind and Google research, and materials informatics initiatives led by MIT and Stanford University.

History and Development

Initial development began under leaders such as Xavier Gonze at Université catholique de Louvain and collaborators from Université de Liège and Université Libre de Bruxelles. Early milestones paralleled advances at CEA and collaborations with the European Theoretical Spectroscopy Facility network. The code evolved through contributions from researchers affiliated with CNRS, Université Paris-Saclay, and international partners in the United States Department of Energy complex. Major releases incorporated community contributions from groups including those at Princeton University, University of California, Berkeley, University of Cambridge, and ETH Zurich, reflecting cross-institutional support spanning Max Planck Society and national research laboratories such as Oak Ridge National Laboratory.

Architecture and Implementation

ABINIT is written primarily in Fortran, structured around modular libraries for pseudopotential handling, plane-wave basis operations, and linear-response solvers. It interfaces with numerical libraries such as BLAS, LAPACK, ScaLAPACK, and FFTW for dense linear algebra and fast Fourier transforms, and leverages parallelization frameworks like MPI and OpenMP to scale on supercomputers including Titan (supercomputer), Blue Gene systems, and clusters at CERN. Workflow orchestration is facilitated by input/output conventions compatible with tools from ASE, pymatgen, and AiiDA, enabling provenance tracking akin to systems used by Zenodo and data repositories at EMBL-EBI.

Usage and Applications

Researchers apply ABINIT to compute band structures, phonon-mediated superconductivity parameters, optical absorption spectra, and defect energetics in semiconductors and oxides studied at institutions like Sandia National Laboratories, Los Alamos National Laboratory, and NREL. Case studies include materials investigated at IBM Research, Microsoft Research, Toyota Research Institute, and consortia such as Graphene Flagship and Battery 500 Consortium. ABINIT supports simulations of perovskites, transition-metal oxides, two-dimensional materials investigated by groups at Bell Labs, and molecular crystals of interest to researchers at Pfizer and Roche in computational chemistry collaborations.

Licensing and Community

Distributed under the GNU General Public License, ABINIT encourages community contributions from academic groups at Università di Milano, Peking University, Tsinghua University, Seoul National University, and national institutes including RIKEN and INRIM. Governance includes steering committees with members from European Theoretical Spectroscopy Facility and contributors affiliated with CNR, CSIC, and Forschungszentrum Jülich. Training workshops and summer schools are held in partnership with ICAM-I2CAM, Lausanne workshops, and conferences such as ICPEAC and MRS Fall Meeting.

Performance and Benchmarks

Performance optimizations target scalability on petascale and emerging exascale systems, with benchmark comparisons against VASP and Quantum ESPRESSO using testcases derived from Materials Project datasets and community benchmark suites developed at NIST and NLBL. Profiling uses tools like Valgrind, TAU, Intel VTune, and monitoring on clusters managed by SLURM and Torque schedulers. Reported scaling demonstrates efficient parallelization across thousands of cores on machines at NERSC and PRACE centers, with GPU-accelerated pathways under exploration in collaborations with NVIDIA and the Oak Ridge Leadership Computing Facility.

Category:Computational physics software