[QE-users] Pseudopotential for SCAN functional

[Xin Zhong]

Dear all

I am testing SCAN (R2SCAN as well) functional for (Mg)-Al-Si-O system.
Currently I am testing with SCAN pseudopotential from Yi Yao's library
https://yaoyi92.github.io/scan-tm-pseudopotentials.html, and also the
SG15-ONCV-PBE PP. Both seem to work, and the results differ to some extend
(by ca. 0.1 ev relative energy). However, for Mg there is no SCAN-PP
currently available (in the future I also want to include Ca, also no
SCAN-PP found). I have been looking through the history of the threads and
see that people either use SCAN or PBE for PP. Based on some recent QE
lecture I found online (by Colonna and Timrov) it seems that SCAN PP
provides better results when combined with input_dft=scan, and pseudo-dojo
should be generally avoided.

My question is if it is ok to mix SCAN-PP with ONCV-PBE-PP (so we sort of
"maximize" the accuracy?), or one should only use PBE-PP for internal
consistency, as I am a bit concerned that mixing PP built with different
functionals can be dangerous, is this the case for SCAN? Or maybe there
exist some new SCAN PP.

Any suggestion or comment would be greatly appreciated!!

Thank you so much
Xin


Below is my input, to be complete.

&CONTROL
   calculation      = "vc-relax"
   tstress          = .true.
   tprnfor          = .true.
   outdir           = "./pw1"
   forc_conv_thr    =  1.0D-4,
   pseudo_dir       = "/path/PP"
/
&SYSTEM
   ecutwfc          = 160
   ecutrho          = 640
   ntyp             = 2
   nat              = 12
   ibrav            = 0
   input_dft        = "SCAN"  (or R2SCAN)
/
&ELECTRONS
   electron_maxstep = 80
   mixing_beta      = 0.7
/
&IONS
/
&CELL
   press            = 0
/

ATOMIC_SPECIES
Si 28.085   Si.SCAN.UPF
O  15.999    O.SCAN.UPF

or
Si 28.085  Si_ONCV_PBE-1.2.upf
O  15.999   O_ONCV_PBE-1.2.upf


The question is if it is a good idea to mix these two? For Si-O it is ok
but there is question for Al,Mg,Ca.