History of PEtot

Petot was initially developed by Lin-Wang Wang while he worked in National Renewable Energy Laboratory at Golden Colorado. It was then parallelized by Lin-Wang Wang and Andrew Canning after L.W. Wang moved to NERSC at Lawrence Berkeley National Laboratory in 1999. Especially, it uses Andrew Canning’s Parallel Fast Fourier Transformation (FFT) code designed for electronic structure calculations. L.W. Wang also developed the version_2 of the code (during 2003-2004), which incoporates the ultrasoft pseudopotentials. Further more, it has a k-point parallelization on top of the G-space parallelization, it can also calculate isolated cluster systems and slab systems without the periodic image potentials. The work by Lin-Wang Wang on this project at LBNL is funded by U.S. Department of Energy through NERSC.

Features in PEtot

• It is a plane-wave pseudopotential density functional ground state calculation package.
  
• It does LDA, LSDA, and GGA.
  
• It can relax the atoms to find minimum atomic positions of the total energy.
• It can calculate the optical matrix elements between different states.
• It can use both the norm conserving pseudopotentials and ultrasoft pseudopotentials, or a mixture of them.
  
• It can be used to calculate 3D isolated clusters or 2D slabs without the periodic image potentials.
  
• It has two levels of parallelization, the lower level parallizes on G-space, and the higher level parallizes on k-points. This is very useful for bulk and surface calculations when there are a large number of k-points.
  
• The package contains the main PEtot code, and sub-packages for norm conserving pseudopotential generation (from J.S. Martin) and for ultrasoft pseudopotential generation (from D. Vanderbilt's group).
• It has two algorithms for electronic state improvement: (1) conjugate gradient, (2) residual minimization. The residual minimization scheme reduce the need for orthogonalization, thus leads to an effective O(N^2) scaling for large systems.
• It has two schemes for self-consistent potential mixing: (1) Pulay-Kerker, (2) Pulay-ThomasFermi. This second method improves significantly the convergence speed for large inhomogeneous systems like surfaces and metallic bars.
  
• It has different schemes for state occupation finite temperature smearing for metallic calculations.
  
• It can use either the G-space, or real-space Kleinman-Bylander nonlocal pseudopotential implementations. For the real-space implementation, it uses a mask-function-scheme, without the need for preprocessing for the pseudopotentials.
• For multiple k point calculations, it stores the not-used k-point wavefunctions on the disk. This reduces the memory requirement significantly.
• It uses n1*n2*n3 grid to represent the wavefunction in real space, while it uses n1L*n2L*n3L to describe the charge density in real space.
  
• It is written in Fortran 90 and parallelized using MPI. On the first level of parallelization, in G-space, each processor possesses a bunch of n1-directioned columns of G coefficients. In real space, the n1*n2*n3 grid is divided in the n1 dimension into N-processor slices, and each processor possesses one slice of the real space data. The transformation between the G-space and real-space is done using the parallel FFT developed by Andrew Canning. On the second level of parallelization, a different group of processors handle different k-points. 

For more information, you can see README

Back-to-home-page