VAMP - Vienna Ab initio Molecular dynamics Package

"VAMP" is a full featured ab-initio molecular dynamics package running on various platforms (RISC-workstations, vector processors, ...). It has been written (almost) from scratch by Georg Kresse in 1992 and some extensions/improvements have been made by Jürgen Furthmüller (and of course also Georg Kresse) in 1993/1994 (and the coding is still going on :-)

• The electronic degrees of freedom are treated on the basis of density functional theory in the local density approximation or the generalized gradient approximation (various LDA- and GGA-XC-functionals implemented) using pseudopotentials in a plane-wave basis -- selfconsistent or non-selfconsistent using the Harris functional.
• It is possible to use nonlocal norm-conserving or "ultrasoft" (Vanderbilt-) pseudopotentials in the (generalized) Kleinman-Bylander-representation with an (in principle) arbitrary number of projectors (although more than two projectors are usually not recommended) with treatment of s-, p- and d-nonlocalities.
• The Kohn-Sham equations are solved iteratively using various band-by-band or "all band simultaneous" update schemes. The most important schemes are a preconditioned conjugate gradient technique and a residual minimization technique (whereof the latter minimizes the number of order N^3 operations and allows an independent treatment of all bands what makes it very easy to parallelize this algorithm). Furthermore some block Davidson-like scheme and a selfconsistent update of all degrees of freedom (including Fermi weights ...) is available (but both schemes will not give the optimum performance though they are still useful in many cases ...). All schemes use a (modified) Broyden-mixing (or alternatively also available a Kerker-mixing) for stabilizing the convergence of the charge density towards self-consistency.
• First attempts to parallelize the code on a SGI Power Challenge have already been made in 1993/1994 and with some few further work on the code it should be possible to run this code also on parallel machines quite efficiently, at least on shared-memory system with an intermediate number of processors ...)
• The program uses fractional occupation numbers for the Kohn-Sham orbitals according to the finite-temperature extension of the density functional theory formulated by Mermin. The Brillouin-zone samplings base on regular k-meshes (Monkhorst-Pack special k-points) and the band occupations are calculated using either a Fermi-smearing method, a (generalized) Gaussian smearing method (according to Methfessel and Paxton) or the tetrahedron method ("classical" linear tetrahedron method or some improved tetrahedron method including "Blöchl corrections"). For all smearing methods the occupation entropies are calculated and included into the total (free) energy to make the (free) energy and the (Hellmann-Feynman) forces consistent. For the Gaussian smearing method and the Fermi smearing method an energy extrapolated towards zero smearing width will also be evaluated. Some remark: The improved tetrahedron method (though showing an absolutely extraordinary convergence of total energies) gives energies being not compatible with the (Hellmann-Feynman) forces and is hence not recommended for molecular dynamics simulations.
• A classical treatment of the ionic degrees of freedom is used (according to Newton's classical equations of motion) with ability to perform microcanonical or canonical molecular dynamics (with temperature control by a Nose-thermostat), structural optimizations by means of a conjugate gradient minimization or a quasi-Newton approach (which allows in principle also a search for maxima or saddle points on the ionic total energy surface) or static calculations.
• The forces on the atoms and (only usuable for structural optimizations) also stresses (full stress tensor!) are calculated according to the Hellman-Feynman theorem including so-called "Harris-corrections" for the case of non-selfconsistency (corrects for errors in the Hellman-Feynman forces/stresses to first order in the deviation of the total charge density from selfconsistency).
• The code includes a full featured symmetry analysis (automatic recognition of the lattice types and all symmetry elements, application of symmetry to Brillouin-zone sampling with ability to set up symmetry-reduced k-meshes and tetrahedra-networks automatically, symmetrization of charge densities and forces/stresses).
• Several analytical informations are also provided by VAMP like radial pair distribution functions for liquids (and amorphous phases), density of states (with decomposition into l-projected local density of states and general information on l-projections at each atomic site of all bands), eigenvalues, charges, positions (with complete history during molecular dynamics runs). There exist also some miscellaneous tools for visualization of output data, pre-analysis of structural input data, "split-and-merge" tools for splitting (non-selfconsistent) calculations (for band structures or density of states) involving a large number of k-points (which would exceed the available memory capacity) into a series of single k-point jobs.
• One significant aim of the coding done by G. Kresse was (from the very beginning) not only to establish some general purpose tool for ab-initio atomic-scale simulations but also to get a high performance package coming close to peak performances. So our code is not only universal but also very fast (maybe the fastest code in universe? :-). Many of our people here at the institute don't believe this because we usually still need thousands and tenthousands of CPU-hours for each project but other codes will surely do no better and faster job ... (?).

A more detailed description ("handbook") can be found here. Warning: we have still some few problems the local installation(s) of our LaTeX2HTML translator (some formulas look really UGLY!). So maybe you want to try this Postscript file (but you have to be patient - it's a quite large file!).