Softwares

Pylada: This is a suite of python computational framework for first-principles computation, including that for defect properties with the supercell approach. I am a co-author for this code. I am using Pylada for preparing, managing, and processing ab-initio, mainly VASP, calculations.

rVV10: Implemented within VASP, will be included within next versions, and is available for versions 5.3.5 and 5.4.1 on request by email (Please make sure you have legal access to VASP). The implementation was done by me and Dr. Zeng-Hui Yang.

VV10 denotes for the non-local van der Waals correlation functional devised by O.A. Vydrov and T. van Voorhis in 2010. rVV10 denotes for the revised VV10  method allowing for fast implementation. They are very similar with each other.

VV10 has been available in many quantum chemistry codes like Q-CHEM and ORCA, and rVV10 has been implemented in codes such as Quantum Espresso, SIESTA, CP2K, ONETEP, and VASP.

SCAN+rVV10 is a versatile van der Waals density functional by combing the Strongly Constrained Appropriately Normed (SCAN) meta-GGA semilocal exchange-correlation functional with the rVV10 non-local correlation functional. It is the best dispersion-corrected semilocal functional for layered materials to date.

Combing PBE-GGA with rVV10 with a judicious choice of the damping parameter also works very well for layered materials, and I coined it as PBE+rVV10L with the "L" denoting for layered materials. Of course, this combination is not as versatile as the meta-GGA-level SCAN+rVV10, however it does not suffer from numerical instability and divergence.

One conceptual difference between PBE+rVV10L/SCAN+rVV10 and the popular Chalmers-Rutgers vdW-DF family is that here a small vdW correction with minimal empiricism is added to a well-behavior non-empirical semilocal functional, while the vdW-DFs by design require an ideally vdW-free exchange which may result in less accurate description of other properties especially the electronic structure. (Note there is different opinion on the non-empiricism of the SCAN meta-GGA, e.g., PRL 117, 073001.)

You are welcome to freely use the rVV10 within VASP for testing and application (please cite the SCAN+rVV10 and PBE+rVV10(L) papers), and also to collaborate on developing new functionals.

--- Update (04/23/2017):
In VASP 5.4.4, the SCAN, and the rVV10 implementations have been included.
To enable SCAN, set:
METAGGA = SCAN
                               
For meta-GGA calculations, make sure the POTCAR including the following lines (vasp wiki):
# grep kinetic POTCAR                    
kinetic energy-density
mkinetic energy-density pseudized
kinetic energy density (partial)
               
Besides, the following two lines are better to be included in the INCAR file for meta-GGA calculations (I add them also for GGA ones).
LASPH = T
ADDGRID = T
                               
To enable SCAN+rVV10, further add the following two lines to the INCAR file:
LUSE_VDW = T
BPARAM = 15.7
 
In the official version, use LUSE_VDW=T combines rVV10 to any meta-GGAs, and combines vdW-DFs with any GGAs.
To use PBE+rVV10 or PBE+rVV10(L), you need slightly change xcgrad.F and xcgrad_spin.F. I will release a patch when I have the official version and time.

Besides, in the makefile.include, I suggest not to include -DnoAugXCmeta in CPP_OPTIONS.

Have fun!
---- End of Update (04/23/2017) ----

---- (08/05/2017) ----

Converging the SCAN meta-GGA calculations:

I am not a co-author for the SCAN meta-GGA, but I have received requests for help on the convergence problem with SCAN. As a meta-GGA. SCAN has a much more complicated mathematical form than GGAs, especially PBE that may be the mathematically simplest GGA. The difficulty to reach convergence with SCAN happens occasionally, especially for some magnetic systems, and systems with vacuum surrounded. Below is my handling:

1,     Restart the calculation with a pre-converged PBE WAVECAR (ISTART = 1, the default value when WAVECAR exists)    
2,     Decrease the mixing parameter AMIX from the default 0.4 to 0.2 and to 0.1, increase NELM to 120 or more.
3,     Try the Anderson mixing by setting IMIX = 1, AMIX = 0.1, BMIX = 0.0001, AMIX_MAG = 2*AMIX, BMIX_MAG = 0.0001
4,     The above methods are usually enough, and I would pray for a while if they are not ...

Hope this helps.

p.s., the first three tips can be found in the vasp wiki.

---- End (08/05/2017) ----

GGA exchange functionals within VASP for WC06, vdW-DF-cx, and etc., available on request by email (Please make sure you have legal access to VASP).
---- (02/28/2019) ---- 
You can use vdW-DF-cx in VASP since version 5.4.4 (not my contribution), and what you need is to set in the INCAR:

        GGA           =  CX
        AGGAC       =  0.0
        LUSE_VDW = .TRUE.

         (This seems not been included in the VASP wiki yet.)
---- End (02/28/2019) ----


Email: Haowei.Peng@gmail.com
Comments