Doc updates for the new VASP interface in version 6.5.0 (#269)

* [doc] update VASP interface doc for new interface features
* [doc] update VASP SVO tutorial for new interface features
* [doc] NiO VASP CSC tutorial updates for new interface
* [fix] fixes for latest dfttools version
* [doc] updates to CSC NiO example with VASP

bin/vasp_dmft.in

@@ -5,7 +5,7 @@ MPIRUN_CMD=mpirun
 show_help()
 {
 echo "
-Usage: 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>]
+Usage: vasp_dmft [-n <number of cores>] -i <number of iterations>  -j <number of VASP iterations with fixed charge density> [-p <path to VASP directory>] [<dmft_script.py>]
 
        If the number of cores is not specified it is set to 1 by default.
 
@@ -14,9 +14,6 @@ Usage: vasp_dmft [-n <number of cores>] -i <number of iterations>  -j <number of
        Set the number of VASP iteration with a fixed charge density update
        inbetween the dmft runs with -j <number of VASP iterations with fixed charge density>
 
-       Set the version of VASP by -v standard(default)/no_gamma_write to
-       specify if VASP writes the GAMMA file or not.
-
        If the path to VASP directory is not specified it must be provided by a
        variable VASP_DIR.
 
@@ -56,12 +53,6 @@ while getopts ":n:i:j:v:p:h" opt; do
            VASP_DIR=$OPTARG
         fi
         ;;
-     v)
-        if [ -n "$OPTARG" ]; then
-           VASP_VERSION=$OPTARG
-#           echo "Version of VASP (writing GAMMA file (standard) or not (no_gamma_write): $VASP_VERSION"
-        fi
-        ;;
       h)
         show_help
         exit 1
@@ -99,11 +90,6 @@ if [ -z "$NDFTITER" ]; then
   NDFTITER=1
 fi
 
-if [ -z "$VASP_VERSION" ]; then
-  echo "  VASP version not specified, setting to standard"
-  VASP_VERSION="standard"
-fi
-
 shift $((OPTIND-1))
 
 if [ -z "$1" ]; then
@@ -115,13 +101,12 @@ fi
 echo "  Number of cores: $NPROC"
 echo "  Number of iterations: $NITER"
 echo "  Number of iterations with fixed density: $NDFTITER"
-echo "  VASP version: $VASP_VERSION"
 echo "  Script name: $DMFT_SCRIPT"
 
 rm -f vasp.lock STOPCAR
 # run in serial and use OMP_NUM_THREADS here for vasp >=6.2 in case there are problems
 stdbuf -o 0 $MPIRUN_CMD -np $NPROC "$VASP_DIR" &
 
 
-$MPIRUN_CMD -np $NPROC @TRIQS_PYTHON_EXECUTABLE@ -m triqs_dft_tools.converters.plovasp.sc_dmft $(jobs -p) $NITER $NDFTITER $DMFT_SCRIPT 'plo.cfg' $VASP_VERSION || kill %1
+$MPIRUN_CMD -np $NPROC @TRIQS_PYTHON_EXECUTABLE@ -m triqs_dft_tools.converters.plovasp.sc_dmft $(jobs -p) $NITER $NDFTITER $DMFT_SCRIPT 'plo.cfg' || kill %1
 

doc/guide/conv_vasp.rst

@@ -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 all 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
+=================================
+
+The above mentioned interface has some limitations for VASP versions 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 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.

doc/guide/dftdmft_selfcons.rst

@@ -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>]
@@ -162,14 +162,15 @@ 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
   NELMIN = 1000
   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
@@ -187,7 +188,9 @@ 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.
 
-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>`_.
+The `LSYNCH5 <https://www.vasp.at/wiki/index.php/LSYNCH5>`_ flag is set to `True` to ensure that the `vaspout.h5 <https://www.vasp.at/wiki/index.php/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 hdf5 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>`_.
 
 
 Elk

doc/tutorials/nio_csc_vasp/INCAR

@@ -1,24 +1,19 @@
 System  = NiO
 
-# for PLOT DOS
-# ISMEAR = -5
 # better convergence for small kpt grids
 ISMEAR = 2
-SIGMA = 0.05
+SIGMA = 0.04
 
 # converge wave functions
 EDIFF = 1.E-7
 NELMIN = 35
 
 # the energy window to optimize projector channels (absolute)
 EMIN = -3
-EMAX = 10
+EMAX = 12
 
 LMAXMIX = 6
 
-# switch off all symmetries
-ISYM = -1
-
 # project to Ni d and O p states
 LORBIT = 14
 LOCPROJ = "1 : d : Pr

doc/tutorials/nio_csc_vasp/INCAR.CSC

@@ -0,0 +1,29 @@
+System  = NiO
+
+# better convergence for small kpt grids
+ISMEAR = 2
+SIGMA = 0.05
+
+# converge wave functions
+EDIFF = 1.E-7
+
+# the energy window to optimize projector channels (absolute)
+EMIN = -3
+EMAX = 10
+
+LMAXMIX = 6
+
+# project to Ni d and O p states
+LORBIT = 14
+LOCPROJ = "1 : d : Pr
+           2 : p : Pr"
+
+# CSC flags
+ICHARG = 5
+NELM = 1000
+NELMIN = 1000
+NELMDL = -2
+IMIX=1
+BMIX=0.5
+AMIX=0.02
+LSYNCH5=True

doc/tutorials/nio_csc_vasp/INCAR.CSC.rst

@@ -0,0 +1,7 @@
+.. _INCAR.CSC:
+
+INCAR.CSC
+-----------
+
+.. literalinclude:: INCAR.CSC
+   :language: bash

doc/tutorials/nio_csc_vasp/KPOINTS

@@ -1,4 +1,4 @@
 Automatically generated mesh
        0
 Gamma
- 6 6 6
+ 11 11 11

doc/tutorials/nio_csc_vasp/NiO_local_lattice_GF.py

@@ -13,17 +13,11 @@
 
 filename = 'vasp'
 beta = 5.0
-mesh = MeshImFreq(beta=beta, S='Fermion', n_iw=1000)
+mesh = MeshImFreq(beta=beta, S='Fermion', n_iw=500)
 
 SK = SumkDFT(hdf_file = filename+'.h5', use_dft_blocks = False, mesh=mesh)
 
 
-# We analyze the block structure of the Hamiltonian
-Sigma = SK.block_structure.create_gf(mesh=mesh)
-
-SK.put_Sigma([Sigma])
-
-
 # Setup CTQMC Solver
 n_orb = SK.corr_shells[0]['dim']
 spin_names = ['up','down']
@@ -56,7 +50,7 @@
 if block_structure:
     SK.block_structure = block_structure
 else:
-    G = SK.extract_G_loc()
+    G = SK.extract_G_loc(transform_to_solver_blocks=False, with_Sigma=False)
     SK.analyse_block_structure_from_gf(G, threshold = 1e-3)
 
 SK.put_Sigma(Sigma_imp = [Sigma_iw])

doc/tutorials/nio_csc_vasp/maxent.py

@@ -29,12 +29,12 @@
         gf = gf + G_latt['up'][iO,iO]
     tm.set_G_iw(gf)
     tm.omega =LinearOmegaMesh(omega_min=-20, omega_max=20, n_points=201)
-    tm.alpha_mesh = LogAlphaMesh(alpha_min=0.01, alpha_max=20000, n_points=60)
+    tm.alpha_mesh = LogAlphaMesh(alpha_min=0.01, alpha_max=20000, n_points=30)
 
     tm.set_error(1.e-3)
     result=tm.run()
     result.get_A_out('LineFitAnalyzer')
-    
+
     if 'iteration_count' in ar['DMFT_results']:
         iteration_offset = ar['DMFT_results']['iteration_count']+1
         for oo in orb: