NEW RELEASE: VASP.6.4.0

A new release of VASP is available for download now!

This release differs from VASP.6.3.2 by the following changes/additions (amongst others):

  • DFT and hybrid functionals:
    • IVDW_NL: A new tag to choose the version of the nonlocal vdW-DF functional.
    • Spin-polarized formulation for nonlocal vdW functionals.
    • Hybrid functionals that combine MGGA and Hartree-Fock.
    • Laplacian-dependent MGGA.
    • Local mBJ potential.
  • ACFDT and GW:
    • Added EVGW0R algorithm (NATURALO=4).
    • all-in-one GW mode improved (LALL_IN_ONE).
    • OpenACC port to GPU of ALGO = ACFDTR.
  • Machine learned force fields:
    • New fast-prediction mode (typically 20-100 times faster MD trajectories are possible).
    • Improved neighbor list algorithm (cell lists + Verlet lists, supports fast-prediction mode).
    • New “super”-tag ML_MODE for easy training and application of ML force fields.
    • New human-readable header for ML_FF file containing most important parameters, corresponding time stamp in ML_LOGFILE.
    • Regularized SVD for refitting with optimized default parameters (ML_IALGO_LINREG = 4).
    • Additional ML tags presented in ML_LOGFILE.
  • Miscellaneous:
    • Added Nose-hoover chain thermostat
    • Added option to set Fermi energy in the middle of the band gap (EFERMI = MIDGAP).
    • Created new command line argument “–dry-run” to check the setup.
    • Allow faster wavefunction initialization with parallel random numbers (RANDOM_GENERATOR = PCG_32).
    • Write additional output to vaspout.h5 for extraction with py4vasp (pair correlation function, dielectric function, velocities).
  • Various improvements and bugfixes.

Go to the VASP Portal to get VASP.6.4.0. This requires a VASP.6 license issued by the VASP Software GmbH.
Get a license!