ABINIT package: context of development

The Corning code for electronic structure calculations began to be developed in the late eighties by D.C. Allan, Corning Inc. The fundamental algorithm of this code, the band-by-band conjugate gradient algorithm, was proposed by M.P. Teter, M. Payne and D.C.Allan, and described thoroughly in the review paper [Payne1992] where other references can be found. It was written in Fortran 77. Some technical features were: a plane-wave representation of wavefunctions, use of pseudopotentials, local-density approximation (LDA) within the Density-Functional Theory (DFT).

In July 1990, X. Gonze joined Cornell University, where M.P.Teter was professor. Building upon the Corning code, the Respfn (Response Function code) was written. See [Gonze1997a] for a complete description of the algorithms. While the Corning and Respfn code were consistent at the end of X. Gonze stay in Cornell University (in september 1992), they began to diverge afterwards. Corning Inc. had agreed to allow further development based on the version 920813 in Louvain-la-Neuve. Some features of Respfn were developed by J.-C. Charlier and C. Lee both before the separation of codes as well as after it.

In 1993 Corning Inc. agreed with Biosym Inc. that the Corning code, renamed Plane_Wave, would be commercialized by Biosym Inc. In 1995, Biosym Inc. merged with MSI (Molecular Simulation Inc.), and shortly afterwards, the decision was taken by MSI not to continue the development of the Plane_Wave code.

In 1996, D.C. Allan and X. Gonze explored the possibility to write a new code, that should not be commercialized, but rendered available to the scientific community as a freeware. It would be written in Fortran 90, would include parallel features, would be based on a SCF (Self-Consistency Field) algorithm. Its writing would be facilitated by the availability of the Respfn codes and Corning 920813 codes, as well as later developments of the Corning code (then renamed Plane_Wave), excluding the features developed by Biosym computer scientists, or tightly bound with Biosym proprietary features.

The project was first named DFT2000, but the name ABINIT was definitely adopted in September 1998. Version 1.9 of ABINIT was made available in March 1999 to beta testers (for some it was done a bit earlier), outside Louvain-la-Neuve or Corning. It had only “ground-state” features.

Version 2.0 was released in July 1999. The possibility to compute response-functions (phonons, Born effective charges, dielectric constant) was available.

Version 3.0, the first version to be proposed on the Web under GNU GPL licence, has features that considerably improve upon the existing Respfn and the 920813 version of Corning (or even the latest available version of Plane_Wave), i.e. much faster, better parallelisation … ABINIT v3.0 was made available in December 2000.

An advisory commitee was set up in June 2000, and served for the whole lifetime of the version 3.

Version 4.0 was delivered after the first international ABINIT developer workshop, in January 2003, with a renewed advisory committee, and the aim to bridge the speed gap with respect to other plane-wave based first-principles codes: PAW, parallel FFT, better geometry optimization, better molecular dynamics.

The second international ABINIT workshop, in May 2005 was held in Paris.

Version 5.0 was delivered in Autumn 2005, with a completely renewed build system.