Rutgers University Department of Physics and Astronomy

Vanderbilt Ultra-Soft Pseudopotential Site

This is the primary site for the distribution of the ``UltraSoft PseudoPotential'' (USPP) generation code from the group of David Vanderbilt at Rutgers University.

Translations of this page into other languages

Table of Contents

Browsing and Downloading the USPP Package

To take a first look at this package and get familiar with its contents, you can browse the latest version in the form of a DIRECTORY TREE.

However, to use the package, you should download it as described below.

Fortran sources for pseudopotential generation

This code is freely distributed under the terms of the GNU GPL public license. You may use it for your own research purposes, or pass it on to others for similar use. However, note the following restrictions:

The Fortran sources in the USPP package start with Release 7.3.3 (uspp-733). Previous versions were distributed by email request only; the last of these was "a7.3.2". As far as Fortran sources are concerned, uspp-733 is essentially functionally equivalent to a7.3.2 (only very minor fixes and cosmetic improvements were made in going from a7.3.2 to uspp-733).

The main changes in going from a7.3.2 to uspp-733 are a major renovation of the directory structure, makefiles, and documentation, and the addition of a library of contributed pseudopotentials.

See Source/README for a summary of improvements that were made to the source codes from one release to the next, including those leading up to the most recent uspp-736. The most detailed information about the release history can be found in the header comments making up the first few hundred lines main program runatom.f.

Conversion to PAW format using USPP2PAW

Thanks to Marc Torrent and Francois Jollet, it is possible to patch the USPP code so that it generates PAW (projector augmented wave) potentials readable by ABINIT. See How to build PAW atomic datasets for ABINIT using the USPP generator for instructions and download links.

Pseudopotentials Libraries

Users looking for a library of already-generated pseudopotentials are suggested to use the Garrity-Bennett-Rabe-Vanderbilt (GBRV) Pseudopotential Library that was developed in 2013 by Kevin F. Garrity, Joseph W. Bennett, Karin M. Rabe and David Vanderbilt, as described in "Pseudopotentials for high-throughput DFT calculations," Comp. Mater. Sci. 81, 446 (2014) ( local preprint). This library was designed and tested for use in high-throughput calculations which require an accurate, comprehensive, and computationally inexpensive pseudopotential library, and it has been tested in many bonding environments.

The library was generated in an attempt to be 1) comprehensive, 2) robustly accurate, and 3) computationally inexpensive. We have designed the library for a single low plane-wave cutoff and tested it by comparing to all-electron calculations for a variety of metallic, ionic, and covalently bonded materials. We find that the GBRV library gives lattice constants, bulk moduli, and magnetic moments which are of greater overall accuracy than other comprehensive pseudopotential libraries, and at comparable or lower computational cost.

While optimized for high-throughput calculations, the pseudopotentials should be appropriate for many applications.

Legacy library: Previously I had posted a "Legacy Library" of pseudopotentials that were assembled over time, mainly be members of the Vanderbilt group, and mainly in the late 1990s and early 2000s. This library is incomplete, ad-hoc, and not consistently tested. I no longer recommend using these, but for backward compatibility the details are still posted here.


Each of the links below points to a tarred-and-gzipped archive of the distribution, listed in reverse chronological order. The first part of the name (e.g., "uspp-736") refers to the release of the fortran sources in the Bin/ directory. The last digit (e.g., "-0") refers to the sequence of updates to such things as the library of pseudopotentials, the documentation, or the makefiles.

On most Unix systems, you can unpack the archive by doing something like

   gunzip uspp-736-0.tgz
   tar xf uspp-736-0.tar
   rm uspp-736-0.tar

Getting started

If you just want to generate one or more of the pseudopotentials that are already included in the library, you should find most of the information you need in AA-QUICK-START, which explains how to compile and run the program to generate the pseudopotential from the library parameter files.

If you are more ambitious and would like to modify an existing pseudopotential or generate an entirely new one, see Doc/TUTORIAL.

Note that, by default, the pseudopotential is generated as an unformatted binary data file as described in Doc/FORMAT. If necessary, it may be converted to formatted form using the utility program reform.f in the Utils directory.


The following files may be of some help in getting oriented:

The literature references on ultrasoft pseudopotentials are primarily these:

User community

I am aware of the following open-source codes that can make use of the ultrasoft pseudopotentials generated by this package:

Conversion routines have now been developed for:

In addition, the following proprietary packages are also compatible:

Another well-known proprietary code that uses ultrasoft potentials is the VASP package. However, VASP uses ultrasoft potentials generated using a somewhat different scheme from the one implemented here. Thus, pseudopotentials taken from the library distributed as part of VASP cannot be imported into the above code packages, and vice versa.

Archive of earlier versions

An archive of previous versions of the uspp package is available here. This could possible be useful in dealing with compatibility issues associated with potentials generated with earlier versions of the code.


Please send comments or suggestions for improvement to

Acknowledgments and Disclaimer

This Web page is based in part upon work supported by the US National Science Foundation under Grants No. 9981193, 0233925, 0549198, and 1005838. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the National Science Foundation.

Back to Vanderbilt Home Page.

Please send any comments on this page to