[Pw_forum] vc-relax for B-N-C cannot be truely converged

[Lei Cao]

Dear All,

I have been trying to optimize the structure of single layer hexagonal BNC.
The vc-relax run finished smoothly and gave me the optimized positions of
each atom and the lattice parameters.

However, if I use these optimized values for scf calculation, it gave me
large force between atom and large stress. If I use these values for
another vc-relax run, then it will take quite a few bfgs steps to get the
optimized atomic positions and lattice parameters. Surprisingly, they are
different from the values I got from the last vc-relax run.

Could you please help me with this? My code is pasted below.

Thanks,
Lei

------------------

 &control

    calculation = 'vc-relax'

    restart_mode='from_scratch',

    prefix='BNC',

    tstress = .true.

    tprnfor = .true.

    pseudo_dir = '/home/QE/pseudo/',

    outdir='./temp_dir/'

 /

 &system

    ibrav=  4, celldm(1)= 9.53616076609, celldm(3) = 3.99905580615, nat= 8,
ntyp= 3,

    ecutwfc = 120.0

    occupations='smearing', smearing = 'mv', degauss = 0.002

 /

 &electrons

    diagonalization='david'

    mixing_mode = 'plain'

    mixing_beta = 0.7

    conv_thr =  1.d-13,

 /

 &IONS

 /

 &CELL

    cell_dynamics = 'bfgs' ,

    press = 0.00,

    wmass =  0.00700000,

    press_conv_thr = 0.1,

    cell_dofree = '2Dxy'

 /

ATOMIC_SPECIES

 B   10.811     B.pbe-mt_fhi.UPF

 N   14.00674   N.pbe-mt_fhi.UPF

 C   12.0107   C.pbe-mt_fhi.UPF

ATOMIC_POSITIONS crystal

B        0.002466254   0.004932509  -0.000000000

N        0.170582689   0.341165379  -0.000000000

N        0.496083977  -0.007832045  -0.000000000

B        0.664200413   0.328400824   0.000000000

C        0.013913144   0.504233187   0.000000000

C        0.176346623   0.829100147  -0.000000000

C        0.490320044   0.504233187   0.000000000

C        0.652753523   0.829100147   0.000000000

K_POINTS automatic

 16 16 1 0 0 0
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20161026/e01222df/attachment.html>