[QE-users] Problem with 2D cutoff and gamma calculations

[Mattia Angeli]

Dear QE users,

I'm trying to relax a rather large 2D heterostructure. Given the large 
volume of the supercell, I'm forced to perform the relaxation using the 
K_POINTS {gamma} card to save memory.
The relaxation is performed using the 2D cutoff of ref. PRB, 96(7), 
75448 and a large vacuum in the z-direction.


Apparently, these two options seem to not work jointly. Is this a 
well-known issue?
The scf accuracy oscillates wildly and the scf cycle doesn't converge. I 
tested this on different 2D heterostructures and the problem persists.


I've prepared an input file (posted below), on a much smaller unit cell 
that reproduces the problem.
If the option 'assume_isolated='2D'' is removed, the calculation works 
fine.
On the other end, if instead the K_POINTS {gamma} option is replaced by 
K_POINTS {automatic} 1 1 1 0 0 0,  the calculation works fine.


&CONTROL
   calculation  = 'relax',
   prefix       = "test",
   pseudo_dir   = "./pseudo",
   outdir       = "./out",
   verbosity='high',
   disk_io='high',
/

&SYSTEM
   ibrav     = 4,
   A  =      3.298035242552
   C  =      30.0
   nat       = 6,
   ntyp      = 2,
   ecutwfc   = 60,
   ecutrho   = 480,
   occupations = 'fixed'
   input_dft='VDW-DF2-B86R'
   assume_isolated = '2D'
/

&ELECTRONS
   conv_thr    = 1.D-7,
   mixing_mode='local-TF',
   diagonalization= 'david',
/

&IONS
  ion_dynamics='bfgs',
/


ATOMIC_SPECIES
W  183.84  W.pbe-spn-kjpaw_psl.1.0.0.UPF
Se 78.96   Se.pbe-n-kjpaw_psl.1.0.0.UPF

ATOMIC_POSITIONS (crystal)
W        0.000000000   0.000000000   0.110023209
Se       0.333333333   0.666666666   0.166071077
Se       0.333333333   0.666666666   0.053868439
W        0.666666666   0.333333333  -0.110001069
Se       0.000000000   0.000000000  -0.053886514
Se       0.000000000   0.000000000  -0.166075142


K_POINTS {gamma}

! K_POINTS {automatic}
! 1 1 1 0 0 0



Thank you,


Mattia Angeli,
PhD student at SISSA (Trieste), Italy.