[CP2K:4708] Re: Why simulation with PBE functional is too slow?
Matt W
MattWa... at gmail.com
Wed Dec 11 19:31:59 UTC 2013
Hi,
I can't comment on exactly what is happening, but you are having problems
with SCF convergence. This is not the same as one functional running much
slower than another, as such - i.e. it is a more or less physical problem
not that the code doesn't / can't run the two funcitonals with a similar
speed.
It appears that you get huge charge oscillations between the two centres in
your system - try
> grep ' 5 C 1'
PBE must place HOMO/SOMO, or similar, of both systems at very similar
energies. The slow down occurs when your system spin polarizes and you get
a fluctuating biradical solution.
At this point you need to wonder why this is happening. Maybe hybrid
functionals are needed?, maybe it is true that charge transfer between the
two things can occur, but this is a difficult problem for DFT.
> I don't know what happened when calculation comes to the step 27. Last
> time you mentioned that there are some usual keywords in my input file.
> What are they? Would you please help me to fix this problem? Thank you very
> much.
>
> I seem to remember (this is going back some time so likely I am wrong)
that you were using self interaction corrections - I have no experience
with them. Perhaps what works for BLYP doesn't do good things to PBE.
Again, it is the SCF convergence and possible problems with your
system/hamiltonian that gives the speed differential, not PBE being slower.
Matt
> Best wishes,
> Huan
>
>
>
>
>
> On Mon, Oct 7, 2013 at 8:27 PM, Matt W <Matt... at gmail.com <javascript:>>wrote:
>
>> Sorry, please don't try to post everything!!!!
>>
>> But what happens during a couple of SCF cycles, during localization, of a
>> single MD step.
>>
>> Matt
>>
>> On Monday, October 7, 2013 1:56:34 PM UTC+1, huan... at mail.huji.ac.ilwrote:
>>>
>>> Dear All,
>>>
>>> I am now using CP2K to run MD simulations on the benzen-Zundel cation
>>> with BLYP and PBE functional, respectively. I use the same number of CPU,
>>> but for PBE functional, the speed is 3 times slower than BLYP. In addition,
>>> the calculation with PBE functional takes nearly whole the memory, shown
>>> below.
>>>
>>> For BLYP functional-- Mem: 64380M total, 16753M used, 47626M free
>>>
>>> For PBE functional-- Mem: 64380M total, 63976M used, 403M free
>>>
>>> Is that common for PBE functional?
>>>
>>> Actually, I just copied the BLYP input file, and then changed the
>>> functional into PBE and added long range correction, C9 term. The rest of
>>> options, parameters, coordinates and basis sets are the same with BLYP
>>> input file.
>>>
>>> I pasted my input file below. Can anyone give me some suggestion? I
>>> appreciate it very much.
>>>
>>> Best wishes,
>>> Huan
>>>
>>> ======== INPUT FILE ===========
>>>
>>> &GLOBAL
>>> PREFERRED_FFT_LIBRARY FFTW #FFTSG # FFTESSL
>>> # FFT_POOL_SCRATCH_LIMIT 10
>>> # PROGRAM Quickstep
>>> PROJECT nve_50K # #
>>> RUN_TYPE MD # # MD ## GEO_OPT ## ENERGY ----------
>>> PRINT_LEVEL LOW
>>> # SEED -6
>>> &END GLOBAL
>>> &FORCE_EVAL
>>> &DFT
>>> CHARGE 1
>>> # UKS # ROKS with self-interaction correction (SIC)
>>> UKS
>>> &SIC
>>> SIC_METHOD MAURI_SPZ
>>> SIC_SCALING_A 0.2
>>> SIC_SCALING_B 0.0
>>> &END
>>> # non-periodic Poisson (cluster boundary condition)
>>> &POISSON
>>> POISSON_SOLVER MT
>>> &MT
>>> ALPHA 7.0
>>> REL_CUTOFF 2.0
>>> &END MT
>>> PERIODIC NONE
>>> &END POISSON
>>> BASIS_SET_FILE_NAME GTH_BASIS_SETS
>>> POTENTIAL_FILE_NAME GTH_POTENTIALS
>>> &MGRID
>>> CUTOFF 300
>>> &END MGRID
>>> &PRINT
>>> &MOMENTS ON
>>> COMMON_ITERATION_LEVELS 20000
>>> FILENAME ./MOMENTS
>>> ADD_LAST NUMERIC
>>> PERIODIC FALSE
>>> REFERENCE COAC
>>> &EACH 1
>>> &END
>>> &END MOMENTS
>>> &LOCALIZATION ON
>>> &TOTAL_DIPOLE
>>> COMMON_ITERATION_LEVELS 20000
>>> FILENAME ./TOTAL_DIPOLE.dat
>>> ADD_LAST NUMERIC
>>> PERIODIC FALSE
>>> REFERENCE COAC
>>> &EACH 1
>>> &END
>>> &END TOTAL_DIPOLE
>>> # &MOLECULAR_DIPOLES
>>> # COMMON_ITERATION_LEVELS 20000
>>> # FILENAME ./MOLECULAR_DIPOLES.dat
>>> # ADD_LAST NUMERIC
>>> # PERIODIC FALSE
>>> # REFERENCE COAC
>>> # &EACH 1
>>> # &END
>>> # &END MOLECULAR_DIPOLES
>>> &END LOCALIZATION
>>> # &MO_CUBES
>>> # NLUMO=1
>>> # NHOMO=1
>>> # &END MO_CUBES
>>> # &MO
>>> # EIGENVALUES T
>>> # OCCUPATION_NUMBERS T
>>> # COMMON_ITERATION_LEVELS 0
>>> # EACH 1
>>> # &END MO
>>> # &E_DENSITY_CUBE
>>> # &END E_DENSITY_CUBE
>>> &MULLIKEN
>>> &END MULLIKEN
>>> &LOWDIN
>>> &END LOWDIN
>>> &END PRINT
>>> &QS
>>> EPS_DEFAULT 1.0E-12
>>> EXTRAPOLATION PS
>>> EXTRAPOLATION_ORDER 3
>>> &END QS
>>> &LOCALIZE T
>>> EPS_LOCALIZATION 1.0E-4
>>> EPS_OCCUPATION 1.E-6
>>> OPERATOR BERRY
>>> METHOD JACOBI
>>> MAX_ITER 2000
>>> MAX_CRAZY_ANGLE 0.05
>>> &END LOCALIZE
>>> &SCF
>>> MAX_SCF 150
>>> SCF_GUESS ATOMIC # ATOMIC # RESTART ---cont use RESTART
>>> &MIXING
>>> &END MIXING
>>> &OT
>>> MINIMIZER DIIS
>>> #ROTATION TRUE #needed for ROKS+SIC!!
>>> &END OT
>>> &OUTER_SCF
>>> MAX_SCF 20
>>> EPS_SCF 1.0E-5
>>> &END OUTER_SCF
>>> &END SCF
>>> &XC
>>> &XC_FUNCTIONAL PBE #specified the functional
>>> &END XC_FUNCTIONAL
>>> &vdw_POTENTIAL
>>> DISPERSION_FUNCTIONAL PAIR_POTENTIAL
>>> &PAIR_POTENTIAL
>>> TYPE DFTD3
>>> REFERENCE_FUNCTIONAL PBE
>>> CALCULATE_C9_TERM .TRUE.
>>> REFERENCE_C9_TERM .TRUE.
>>> LONG_RANGE_CORRECTION .TRUE.
>>> PARAMETER_FILE_NAME dftd3.dat
>>> R_CUTOFF 15.0
>>> &END PAIR_POTENTIAL
>>> &END vdW_POTENTIAL
>>> &XC_GRID
>>> XC_DERIV SPLINE2_SMOOTH
>>> XC_SMOOTH_RHO NN10
>>> &END XC_GRID
>>> &END XC
>>> &END DFT
>>> &SUBSYS
>>> &CELL
>>> PERIODIC NONE
>>> ABC 20.000 20.000 20.000
>>> &END CELL
>>> &COORD
>>> C
>>> C
>>> C
>>> C
>>> C
>>> C
>>> H
>>> H
>>> H
>>> H
>>> H
>>> H
>>> H
>>> H
>>> O
>>> H
>>> O
>>> H
>>> H
>>> Ar
>>> &END COORD
>>> &KIND H
>>> BASIS_SET aug-TZVP-GTH
>>> POTENTIAL GTH-PBE-q1
>>> &END KIND
>>> &KIND C
>>> BASIS_SET aug-TZVP-GTH
>>> POTENTIAL GTH-PBE-q4
>>> &END KIND
>>> &KIND O
>>> BASIS_SET aug-TZVP-GTH
>>> POTENTIAL GTH-PBE-q6
>>> &END KIND
>>> &KIND Ar
>>> BASIS_SET DZVP-GTH
>>> POTENTIAL GTH-PBE-q8
>>> &END KIND
>>> &END SUBSYS
>>> &END FORCE_EVAL
>>> &MOTION
>>> &MD
>>> ENSEMBLE NVE
>>> STEPS 50000 #10ps total
>>> TIMESTEP 0.2 # 0.2fs
>>> TEMPERATURE 50.0
>>> #TEMP_TOL 40.0
>>> # &THERMOSTAT
>>> # REGION MOLECULE
>>> # &NOSE
>>> # LENGTH 3
>>> # YOSHIDA 3
>>> # TIMECON 50.0
>>> # MTS 2
>>> # &END NOSE
>>> # &END THERMOSTAT
>>> &END MD
>>> &PRINT
>>> &RESTART
>>> LOG_PRINT_KEY T
>>> &EACH
>>> MD 100
>>> &END EACH
>>> ADD_LAST NUMERIC
>>> &END RESTART
>>> &TRAJECTORY
>>> LOG_PRINT_KEY T
>>> FORMAT XYZ
>>> UNIT angstrom
>>> &EACH
>>> MD 1
>>> &END EACH
>>> ADD_LAST NUMERIC
>>> &END TRAJECTORY
>>> &VELOCITIES
>>> LOG_PRINT_KEY T
>>> FORMAT XYZ
>>> UNIT angstrom
>>> &EACH
>>> MD 1
>>> &END EACH
>>> ADD_LAST NUMERIC
>>> &END VELOCITIES
>>> &END PRINT
>>> &END MOTION
>>> ================================
>>>
>>> --
>> You received this message because you are subscribed to a topic in the
>> Google Groups "cp2k" group.
>> To unsubscribe from this topic, visit
>> https://groups.google.com/d/topic/cp2k/ivi2a2ScZwY/unsubscribe.
>> To unsubscribe from this group and all its topics, send an email to
>> cp2k+... at googlegroups.com <javascript:>.
>> To post to this group, send email to cp... at googlegroups.com <javascript:>
>> .
>> Visit this group at http://groups.google.com/group/cp2k.
>> For more options, visit https://groups.google.com/groups/opt_out.
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20131211/9cb37b80/attachment.htm>
More information about the CP2K-user
mailing list