Johannes Voss has ported BEEF [1] (Bayesian Error Estimation Functional) to both Quantum Espresso and VASP. In both cases a small library called libbeef
must be built.
Version | Source Code | Precompiled Linux 64-bit binary |
---|---|---|
0.1.1 |
The latest development version can be obtained with this command:
git clone https://github.com/vossjo/libbeef |
Installation instructions for the different versions are contained in the README file in the above downloadable files.
Installation instructions (including an ASE python interface) can be found here:
https://github.com/vossjo/ase-espresso/wiki
Newer versions of VASP (V. 5.4.x and up) only need changes to the makefile to enable the BEEF functionals: add -Dlibbeef to the CPP flags and -Lproperpath -lbeef to the library linker flags.
For older versions of VASP, a patch file must now be applied to the appropriate unmodified version of the VASP source code. The patch files can be downloaded from here:
VASP Version | Patch File |
---|---|
5.3 |
Apply the patch within the VASP directory containing the fortran files by executing a command like:
patch -p1 < vasp.5.3.beef.diff |
Then add the following to your Makefile
line where the LIB
variable is defined:
-Lpathtobeeflibrary/lib -lbeef |
Then build VASP as usual. The following lines should be appended to your .INCAR
file to run a BEEF calculation (the last line is optional and triggers the calculation of ensemble energies for error estimation):
GGA = BF LUSE_VDW = .TRUE. Zab_VDW = -1.8867 LBEEFENS = .TRUE. |
Instructions for doing the error estimations in VASP can be found in the libbeef
README
file.
[1] Jess Wellendorff, Keld T. Lundgaard, Andreas Møgelhøj, Vivien Petzold, David D. Landis, Jens K. Nørskov, Thomas Bligaard, and Karsten W. Jacobsen, "Density functionals for surface science: Exchange-correlation model development with Bayesian error estimation", Phys. Rev. B 85, 235149 (2012).