large deviation in volume using SCAN for metallic Li

[bonan_zhu]

I am trying out SCAN meta GGA functional for a few Li-containing materials. For the purpose of testing, I performed a simple singlepoint calculation for metallic Li. The version of vasp is 5.4.4.

POSCAR

Code: Select all

# Compound:Li 
  1.0000000000
  2.7979845200   0.0000000000  -0.9892369100
 -1.3989922600   2.4231256700  -0.9892369100
  0.0000000000   0.0000000000   2.9677107400
Li
    1
Direct
  0.0000000000   0.0000000000   0.0000000000

INCAR

Code: Select all

ALGO = normal
EDIFF = 1e-06
EDIFFG = -0.03
ENCUT = 700
IBRION = 2
ISIF = 3
ISMEAR = 0
ISPIN = 1
LASPH = .TRUE.
LREAL = .FALSE.
METAGGA = SCAN
NCORE = 4
NEDOS = 2000
NELM = 200
NELMIN = 4
NSW = -1
NWRITE = 1
PREC = accurate
SIGMA = 0.05

and the k-point grid is set to 21x21x21 with gamma centering.

Using SCAN gives a pressure of -92.97 kB using the `Li_sv` POTCAR, if the PBE is used the pressure is -0.70 kB.
However, if I use the `Li` POTCAR, SCAN gives 5.10 kB, and 1.64 kB with PBE.

If I use SCAN to perform a relaxation using the `Li_sv` POTCAR, the equilibrium volume goes down to 10 A^3 rather than 20 A^3 per formula unit obtained by PBE.

Note that `Li_sv` does not have core electrons, and here is there no `kinetic energy (partial)` entries in the POTCAR, but Li does have core electrons. Could this be what the problem is?

I would really appreciate it if the developers and people with more experience with SCAN can look into this issue.
Thanks.

[bonan_zhu]

This is probably fixed already in a patch - I was using 5.4.4 apr17: https://cms.mpi.univie.ac.at/marsweb/in ... -4-18apr17.

I don't find this bug in VASP 6.2.0.