[Pw_forum] General question on restarted calculation

[Cameron Foss]

Hello all,

I have had mixed results with restarted calculations when calculating
phonons in TMD MoSe2. Particularly, a dispersion obtained from an
interrupted simulation that was restarted (after an unplanned stop)
returned an 'ok' dispersion despite having some negative frequencies in the
ZA branch (the ZA being the characteristic quadratic flexural phonon mode
in 2D materials). This 'ok' dispersion was calculated on a 6x6x4 MP grid
size. In an attempt to refine the simulation and avoid numerical errors, I
increased the MP grid size to 17x17x1 (in literature a 2D grid size
definition has been used to generate the dispersion of other 2D materials
like MoS2) and implemented controlled stops within walltime limits to avoid
unplanned interruptions in the simulation. Another change between the old
and new simulation was the fact that I isolated the monolayer in the middle
of the unit cell rather than resting at the bottom (I have provided both
input scripts). The dispersion obtained from this latter simulation was
practically nonsense as if the input file defined a highly unstable
structure despite the convergence of relaxation calculations. In short the
first interrupted simulation on a more coarse MP grid yield far better
results than the isolated denser grid simulation. Is there any insight as
to what possible errors that may have occurred? or is it possible that I
did not trace the BZ appropriately as a result of the height offset? (I did
attempt to trace along the z-direction at the Gamma and K points and their
values were accurate and invariant of z despite the rest of the dispersion
being completely incorrect)

%% the first input file that gave an ok dispersion %%
Phonons in MoSe2
&control
    calculation='scf'
    restart_mode='from_scratch',
    !pseudo_dir='directory where pseudopotentials are stored/',
    !outdir='directory where large files are written/'
    pseudo_dir='/home/cjfoss/espresso-5.1/pseudo/',
    outdir='/oasis/scratch/cjfoss/temp_project/out'
    prefix='mose2PH',
 /
 &system
    ibrav=4, celldm(1)=6.2134195, celldm(3)=8,
    nat=3, ntyp=2, ecutwfc =140
 /
 &electrons
    conv_thr =  1.0d-14
    mixing_beta = 0.7
 /
ATOMIC_SPECIES
 Mo  95.94    Mo.pw-mt_fhi.UPF
 Se  78.96    Se.pw-mt_fhi.UPF
ATOMIC_POSITIONS alat
Se       0.000000000   0.000000000   0.000000000
Mo       0.500000000   0.288675135   0.493465193
Se       0.000000000   0.000000000   0.986930347
K_POINTS automatic
 6 6 4 1 1 1


%% the second input file that gave bad results %%
Phonons in MoSe2
&control
    calculation='scf'
    restart_mode='from_scratch',
    !pseudo_dir='directory where pseudopotentials are stored/',
    !outdir='directory where large files are written/'
    pseudo_dir='/home/cjfoss/espresso-5.1/pseudo/',
    outdir='/home/cjfoss/espresso-5.1/2dout'
    prefix='mose2PH_v7',
 /
 &system
    ibrav=4, celldm(1)=6.2134195, celldm(3)=8,
    nat=3, ntyp=2, ecutwfc =140
 /
 &electrons
    conv_thr =  1.0d-12
    mixing_beta = 0.7
 /
ATOMIC_SPECIES
 Mo  95.94    Mo.pw-mt_fhi.UPF
 Se  78.96    Se.pw-mt_fhi.UPF
ATOMIC_POSITIONS alat
Se       0.000000000   0.000000000   3.000000000
Mo       0.500000000   0.288675135   3.493465193
Se       0.000000000   0.000000000   3.986930347
K_POINTS automatic
 17 17 1 1 1 1

%% note the main differences are the MP grid size and the z-direction
offset.

Any insight is welcome!

Best regards,
Cameron
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20160322/39b0e2a9/attachment.html>