LCPQ/dft_tools
3
0
Fork 0
mirror of https://github.com/triqs/dft_tools synced 2025-04-07 17:23:24 +02:00
Code Issues Releases Wiki Activity

[doc] update VASP interface doc for new interface features

Browse Source
This commit is contained in:
the-hampel 2025-02-19 14:18:51 +01:00
parent 7b8b3dd055
commit dbbbf020fa
2 changed files with 23 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

41
doc/guide/conv_vasp.rst
View File

@ -8,7 +8,7 @@ The VASP interface relies on new options introduced since version 5.4.x In
particular, a new INCAR-option `LOCPROJ
<https://cms.mpi.univie.ac.at/wiki/index.php/LOCPROJ>`_, the new `LORBIT` modes
13 and 14 have been added, and the new `ICHARG` mode 5 for charge
self-consistent DFT+DMFT calculations have been added.
self-consistent DFT+DMFT calculations have been added. The VASP interface for charge self-consistent calculations is officially supported as of VASP version 6.5.0 (see `VASP ICHARG=5 documentation <https://www.vasp.at/wiki/index.php/ICHARG>`_). It is highly recommended to compile VASP with hdf5 support enabled (`-DVASP_HDF5`) to enable features of the interface. This allows to use the interface while symmetries are switched on in VASP, and enables spin-polarized feedback in charge self-consistent calculations to VASP.
The VASP interface methodologically builds on the so called projection on
localized orbitals (PLO) scheme, where the resulting KS states from DFT are
@ -18,26 +18,16 @@ The implementation is presented in `M. Schüler et al. 2018 J. Phys.: Condens.
Matter 30 475901 <https://doi.org/10.1088/1361-648X/aae80a>`_.
The interface consists of two parts, :py:mod:`PLOVASP<triqs_dft_tools.converters.plovasp>`, a collection of
python classes and functions converting the raw VASP output to proper projector
python classes and functions converting the VASP output to proper projector
functions, and the python based :py:mod:`VaspConverter<triqs_dft_tools.converters.vasp>`, which
creates a h5 archive from the :py:mod:`PLOVASP<triqs_dft_tools.converters.plovasp>` output readable by
`SumkDFT`. Therefore, the conversion consist always of two steps.
Here, we will present a guide how the interface `can` be used to create input for a DMFT calculation, using SrVO3 as an example. Full examples can be found in the :ref:`tutorial section of DFTTools<tutorials>`.
Limitations of the interface
============================
For VASP version older than 6.5.0 there are a few limitation of the interface as it was not officially supported and only text file based. See `Remarks for VASP older than 6.5.0`_ for details.
* The interface works correctly only if the k-point symmetries
are turned off during the VASP run (ISYM=-1).
* Generation of projectors for k-point lines (option `Lines` in KPOINTS)
needed for Bloch spectral function calculations is not possible at the moment.
* The interface currently supports only collinear-magnetism calculation
(this implies no spin-orbit coupling) and spin-polarized projectors have not
been tested.
* The converter needs the correct Fermi energy from VASP, which is read from
the LOCPROJ file. However, VASP by default does not output this information.
Please see `Remarks on the VASP version`_.
Generation of projectors for k-point lines (option `Lines` in KPOINTS) needed for Bloch spectral function calculations is not possible at the moment.
VASP: generating raw projectors
===============================
@ -383,16 +373,20 @@ For two correlated sites, one can define the file as follows:
0.0 1.0 0.0 0.0 0.0
0.0 0.0 0.0 1.0 0.0
Remarks on the VASP version
===============================
Remarks for VASP older than 6.5.0
=================================
In the current version of the interface the Fermi energy is extracted from the
DOSCAR. However, if one pursues to do charge self-consistent calculations one
needs to write the Fermi energy to the projectors (`LOCPROJ` file), as the DOSCAR
is only updated after a full SCF/NSCF run. The file should contain the Fermi energy
in the header. One can either copy the Fermi energy manually there after a successful
VASP run, or modify the VASP source code slightly, by replacing the following line in
`locproj.F` (around line 695):
The above mentioned interface has some limitations for VASP versions older than 6.5.0:
* The interface works correctly only if the k-point symmetries
are turned off during the VASP run (ISYM=-1).
* The interface currently supports only collinear-magnetism calculation
(this implies no spin-orbit coupling) and spin-polarized projectors have not
been tested.
* The converter needs the correct Fermi energy from VASP, which is read from
the LOCPROJ file. However, VASP by default does not output this information.
For VASP older than 6.5.0 the Fermi level cannot be extracted from the `vaspout.h5` file, then interface falls back to extract the Fermi level from the DOSCAR. However, if one pursues to do charge self-consistent calculations one needs to write the Fermi energy to the projectors (`LOCPROJ` file), as the DOSCAR is only updated after a full SCF/NSCF run. The file should contain the Fermi energy in the header. One can either copy the Fermi energy manually there after a successful VASP run, or modify the VASP source code slightly, by replacing the following line in `locproj.F` (around line 695):
::
< WRITE(99,'(4I6," # of spin, # of k-points, # of bands, # of proj" )') NS,NK,NB,NF
@ -417,7 +411,6 @@ Next, we need to pass this option when calling from `electron.F` and `main.F`
Now Vasp should print in the header of the `LOCPROJ` file additionally the Fermi energy.
Another critical point for CSC calculations is the function call of
`LPRJ_LDApU` in VASP. This function is not needed, and was left there for debug
purposes, but is called every iteration. Removing the call to this function in `electron.F` in line 644 speeds up the calculation significantly in the `ICHARG=5` mode. Moreover, this prevents VASP from generating the `GAMMA` file, which should ideally only be done by the DMFT code after a successful DMFT step, and then be read by VASP.

9
doc/guide/dftdmft_selfcons.rst
View File

@ -133,8 +133,8 @@ Once VASP reaches the point where the projectors are generated
it creates a lock file `vasp.lock` and waits until the lock file is
removed. The shell script, in turn, waits for the VASP process and once
the lock file is created it starts a DMFT iteration. The DMFT iteration
must finish by generating a Kohn-Sham (KS) density matrix (file `GAMMA`)
and removing the lock file. The VASP process then reads in `GAMMA`
must finish by generating a Kohn-Sham (KS) density matrix (file `GAMMA` or `vaspgamma.h5`)
and removing the lock file. The VASP process then reads in `GAMMA`/`vaspgamma.h5`
and proceeds with the next iteration. PLOVasp interface provides a shell-script :program:`vasp_dmft` (in the triqs bin directory)::
vasp_dmft [-n <number of cores>] -i <number of iterations> -j <number of VASP iterations with fixed charge density> [-v <VASP version>] [-p <path to VASP directory>] [<dmft_script.py>]
@ -162,7 +162,7 @@ calculations with the main difference that its functionality (apart from the
lines importing other modules) should be placed inside a function `dmft_cycle()`
which will be called every DMFT cycle and returns both the correlation energy and the SumK object.
VASP has a special INCAR `ICHARG=5` mode, that has to be switched on to make VASP wait for the `vasp.lock` file, and read the updated charge density after each step. One should add the following lines to the `INCAR` file::
VASP has a special INCAR `ICHARG=5 <https://www.vasp.at/wiki/index.php/ICHARG>`_ mode, that has to be switched on to make VASP wait for the `vasp.lock` file, and read the updated charge density after each step. One should add the following lines to the `INCAR` file::
ICHARG = 5
NELM = 1000
@ -170,6 +170,7 @@ VASP has a special INCAR `ICHARG=5` mode, that has to be switched on to make VAS
IMIX=1
BMIX=0.5
AMIX=0.02
LSYNCH5=True
Technically, VASP runs with `ICHARG=5` in a SCF mode, and adding the DMFT
changes to the DFT density in each step, so that the full DFT+DMFT charge
@ -187,6 +188,8 @@ To understand the difference please make sure to read `ISTART flag VASP wiki
`NELMIN` ensure that VASP does not terminate after the default number of
iterations of 60.
The `LSYNCH5` flag is set to `True` to ensure that the VASP writte vaspout.h5 file can be read while VASP is running. Starting from VASP 6.5.0 all communication between VASP and TRIQS is performed through two h5 files: `vaspout.h5 <https://www.vasp.at/wiki/index.php/Vaspout.h5>`_ and `vaspgamma.h5 <https://www.vasp.at/wiki/index.php/Vaspgamma.h5>`_ when VASP is compiled with hdf5 support.
For more detailed and fine grained methods to run Vasp in CSC also on clusters see the methods implemented in `solid dmft <https://triqs.github.io/solid_dmft/_ref/dft_managers.html>`_.