DFTB+ Snapshot 081217
This is a snapshot of the DFTB+ developer version (taken at 17th of December, 2008). It has release quality, but lacks a few features, which we would like to inlcude in the final 1.1 release. The code can be downloaded by registered users.
Release notes
Compilation
DFTB+ Snapshot 081217 had been successfully compiled and tested (using the tests in the autotest system) on the following architectures:
Machine | System | Compiler | BLAS/LAPACK | Notes |
|---|---|---|---|---|
i686 | Linux | Intel Fortran (32-bit) 10.1 | MKL 10.1 | |
x86_64 | Linux | Intel Fortran(em64t) 10.1 | MKL 10.1 |
|
DFTB+ is written in Fortran 95, so any Fortran 95 compliant compiler should be able to compile the code. Please note, that due to bugs in the compiler, only version 4.4 or above of gfortran is able to compile the code.
Changes
Apart from minor bug fixes the following changes had been made since release 1.0.1:
Atoms can be fixed during MD:
The MovedAtoms option can be set during MD to only allow a subset of atoms to move.
Option for minimising memory consumption:
The MinimiseMemoryUsage option can be used to store the eigenvectors generated during multiple k-point and spin polarised calculations on disc, so that the overall memory consumption is reduced to the consumption necessary for just one k-point and one spin component.
MD bug fixed:
Version 1.0.1 contained a bug in the MD driver, causing occasional strange behaviour for light atoms during an MD run. The bug is now fixed.
Initial velocities in MD simulations:
Initial velocities can be now specified for MD simulations. This can also be used to restart MD calculations, by starting with the last geometry, velocities and charges from a previous run.
Progam stops if unprocessed nodes exist:
The default behaviour of the parser is now to stop if any unprocessed nodes exist in the input file dftb_in.hsd or dftb_in.xml, which probably indicates an error in the input. This default behaviour can be turned off by using the IgnoreUnprocessedNodes option. The off the check is, however, not recommended.
Net charges of the atoms printed:
The net charges of the atoms are written to the files detailed.out (user readible output file) and results.tag (tagged output file).
Polynomial repulsives possible:
Slater-Koster files using polynomial repulsives instead of splines can be used now, if the PolynomialRepulsive option is set accordingly. SK files which such repulsives have been generated by the groups at TU Dresden or Jacobs University Bremen.
Writing out of Hamiltonian and overlap matrices:
Using the WriteHS and WriteRealHS options, the folded (square, dense) and the real space (sparse) Hamiltonian and overlap can be written to specific files.
State resolved Mulliken populations:
If the WriteEigenvectors option is used, the eigenvectors and (now in the new code) the Mulliken populations of the individual states are written to eigenvec.out.