[Pw_forum] Hybrid Calculation
Giuseppe Mattioli
giuseppe.mattioli at ism.cnr.it
Fri Dec 29 23:42:01 CET 2017
Dear Brendan
Let me add something more specific:
> 2) When running this calculation, I notice that my code crashes with the
> error "dexx is negative!". After doing some searching through the archives,
> The advice suggests to try different exxdiv_treatments - but this gives me
> a host of new problems, such as knowing which ecutvcut to use (and why),
> etc.
Sometimes there is a bit of instability in the algorithm, specially
when the occupations of KS states oscillate during the scf iterations.
The strict requirement on dexx has been therefore "softened" in
version 6.2 (look into electrons.f90) to check the differences. Thus,
if you are using QE < 6.2 my advice is to update your version. If you
are already using 6.2 you can try to suppress the call to 'dexx is
negative' and to see where your calculation is going, but i don't
recommend it if you don't know what you're doing!
> 1) In general, does my input file for such a calculation look correct?
> (input = below)
well...
> ecutrho = 720
useless. you are using norm-conserving pseudopotentials. ecutrho
should be automatically fixed at ecutwfc*4, and you generally don't
need more than this. Use ecutfock instead, it saves a lot of time. It
is generally safe to use ecutfock=ecutwfc for geometry optimizations
and ecutfock=2*ecutwfc for energy/properties calculations.
adaptive_thr=.true. often provides further speedup.
> nbnd = 235,
> occupations = 'smearing',
you do not need to specify both. The latter generally provides a
suitable amount of unoccupied states to work well with noninteger
occupations, if needed.
> degauss = 0.005,
tight. you should start with a larger value (0.01 Ry = 0.13 eV), which
generally leads to stable results in the case of
insulators/semiconductors.
> nosym = .true.,
why?
> K_POINTS automatic
> 1 1 1 0 0 0
if you want to use gamma then *use* gamma. You save a lot of time
K_POINTS {gamma}
but your structure probably requires a converged (or an *affordably*
converged...) k-points mesh. In principle you should also check that
the q-mesh is converged (nqx1=1, nqx2=1, nqx3=1 is the default, but is
very expensive to use a richer mesh).
HTH
Giuseppe
Quoting Manu Hegde <mhegde at uwaterloo.ca>:
> I am not that expert in hybrid density functional theory. But looking into
> your input seems like you have large cut-off. First start with some simple
> assumptions. Just follow the example in QE libraray. Make sure it works.
> Also you need lot of memory. Hope it helps.
> Manu
> (University of Waterloo)
>
> On Dec 28, 2017 1:39 PM, "Chemist" <mrbsmith58 at gmail.com> wrote:
>
> Hello,
>
> I am attempting to run a hybrid scf calculation on a Pb doped V-oxide
> structure, and I have two questions.
>
> 1) In general, does my input file for such a calculation look correct?
> (input = below)
>
> 2) When running this calculation, I notice that my code crashes with the
> error "dexx is negative!". After doing some searching through the archives,
> The advice suggests to try different exxdiv_treatments - but this gives me
> a host of new problems, such as knowing which ecutvcut to use (and why),
> etc.
>
> I am hoping that someone with expertise can help steer me in the correct
> direction. My input is below. The value for ecut was found by doing a
> convergence study at the gamma point.
>
> Thank you for your time,
> Best,
>
> Brendan Smith
> State University of New York at Buffalo, Ph.D Student, 2nd Year
>
>
> &CONTROL
> calculation = 'scf',
> nstep = 50,
> pseudo_dir = '/projects/academic/alexeyak/brendan/PSEDUO/norm_
> conserving/non_rel/',
> outdir = './',
> prefix = 'x',
> disk_io = 'low',
> /
>
> &SYSTEM
> ibrav = 0,
> nat = 46,
> ntyp = 3,
> nspin = 1,
> nbnd = 235,
> ecutwfc = 90,
> ecutrho = 720,
> occupations = 'smearing',
> smearing = 'gaussian',
> degauss = 0.005,
> nosym = .true.,
> input_dft = 'PBE0'
> /
>
> &ELECTRONS
> electron_maxstep = 300,
> conv_thr = 1.D-7,
> mixing_beta = 0.35,
> /
>
> ATOMIC_SPECIES
> Pb 207.2 Pb.pbe-mt_fhi.UPF
> V 50.9415 V.pbe-mt_fhi.UPF
> O 15.9994 O.pbe-mt_fhi.UPF
>
> K_POINTS automatic
> 1 1 1 0 0 0
>
> CELL_PARAMETERS angstrom
> 15.4779996872 0.0000000000 0.0000000000
> 0.0000000000 3.6440000534 0.0000000000
> -3.3441298425 0.0000000000 9.5546809225
>
> ATOMIC_POSITIONS angstrom
> Pb -1.252373291 0.000000000 3.769703925
> Pb 13.386243452 0.000000000 5.784976998
> Pb 6.486626236 1.822000027 3.769703925
> Pb 5.647243147 1.822000027 5.784976998
> V 4.861534709 0.000000000 0.938269674
> V 7.272335672 0.000000000 8.616411034
> V 12.600534091 1.822000027 0.938269674
> V -0.466664633 1.822000027 8.616411034
> V 1.427260535 0.000000000 1.105476574
> V 10.706608873 0.000000000 8.449204278
> V 9.166260840 1.822000027 1.105476574
> V 2.967609491 1.822000027 8.449204278
> V 3.038442017 0.000000000 3.899265370
> V 9.095428288 0.000000000 5.655415552
> V 10.777441400 1.822000027 3.899265370
> V 1.356427984 1.822000027 5.655415552
> O 0.000000000 0.000000000 0.000000000
> O 7.738999844 1.822000027 0.000000000
> O 12.419713211 0.000000000 0.481555919
> O -0.285843304 0.000000000 9.073124825
> O 4.680713367 1.822000027 0.481555919
> O 7.453156540 1.822000027 9.073124825
> O 9.548904122 0.000000000 0.741443261
> O 2.584965723 0.000000000 8.813237661
> O 1.809904278 1.822000027 0.741443261
> O 10.323965566 1.822000027 8.813237661
> O 6.011811013 0.000000000 2.064766617
> O 6.122059193 0.000000000 7.489914590
> O 13.750810395 1.822000027 2.064766617
> O -1.616941112 1.822000027 7.489914590
> O 3.327103622 0.000000000 2.120183708
> O 8.806766173 0.000000000 7.434497357
> O 11.066103466 1.822000027 2.120183708
> O 1.067766329 1.822000027 7.434497357
> O 0.784110151 0.000000000 2.606516942
> O 11.349759479 0.000000000 6.948164266
> O 8.523110110 1.822000027 2.606516942
> O 3.610759635 1.822000027 6.948164266
> O 10.287975715 0.000000000 4.055961942
> O 1.845894129 0.000000000 5.498718980
> O 2.548975872 1.822000027 4.055961942
> O 9.584893973 1.822000027 5.498718980
> O 4.533116432 0.000000000 4.489744680
> O 7.600753412 0.000000000 5.064936242
> O 12.272116276 1.822000027 4.489744680
> O -0.138246431 1.822000027 5.064936242
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
GIUSEPPE MATTIOLI
CNR - ISTITUTO DI STRUTTURA DELLA MATERIA
Via Salaria Km 29,300 - C.P. 10
I-00015 - Monterotondo Scalo (RM)
Mob (*preferred*) +39 373 7305625
Tel + 39 06 90672342 - Fax +39 06 90672316
E-mail: <giuseppe.mattioli at ism.cnr.it>
More information about the users
mailing list