Setting correct cutoff/convergence criteria

ashish dabral ashishd... at gmail.com
Mon Jul 10 19:28:41 CEST 2017


Hi Matt,

Thanks for your suggestions. After playing around with the parameters, I 
figured the Extrapolation issue and cut-off issue (so fixed that). 
I already see improvement in time after some trials. I am wondering though 
which method would be best for extrapolation: ASPC etc.
I haven't switched to GAPW though. I will try that. Can you please 
elaborate on "Mixing and broyden optimizer will be moderately inefficient 
for insulators"
and a possible alternative.

Thanks

On Monday, 10 July 2017 18:48:55 UTC+2, Matt W wrote:
>
> Hi,
>
> there are quite a few things that could be improved in your input, but the 
> most significant problem is probably
>
>    EXTRAPOLATION USE_GUESS
>
> This means that after every step it starts again from an atomic guess, 
> which is very inefficient. By default CP2K extrapolates from the density 
> matrix of the previous steps.
>
> Also, 900 Ry is a very expensive cutoff (might be needed for some elements 
> with GPW). Change to METHOD GAPW in &QS and reduce the cutoff to ~300 Ry.
>
> Mixing and broyden optimizer will be moderately inefficient for 
> insulators, but the above are probably the biggest things.
>
> Hope that helps,
>
> Matt
>
> On Thursday, July 6, 2017 at 6:31:57 PM UTC+2, ashish dabral wrote:
>>
>> Hi,
>>
>> I am using cp2k for geometry optimization (GEO_OPT only). I have a system 
>> with almost 300 atoms(40x25x11 CELL). Even though cp2k is really fast, it 
>> seems
>> like here it is taking too much time as I have to wait almost 2 weeks to 
>> get it to converge with 12 cores (it does converge!) and a gamma point 
>> calculation.
>> I am wondering if my cut-off/convergence values are ok. I realized that 
>> the GEO_OPT parameter section was missing in the MOTION section but I assume
>> then that the default values were used. There is a variety of atoms in 
>> the complex (Si, C, Mo, Br etc.). Is it normal then, given the below setup, 
>> to consume 
>> so much time? A complete input is not possible to attach but I am 
>> attaching the file showing different parameters: 
>>
>> &FORCE_EVAL
>>   METHOD QS
>>    &DFT
>>     MULTIPLICITY  1
>>     CHARGE  0
>>     BASIS_SET_FILE_NAME /BASIS_MOLOPT
>>     POTENTIAL_FILE_NAME /GTH_POTENTIALS
>> *    SURFACE_DIPOLE_CORRECTION TRUE*
>> *    SURF_DIP_DIR Z*
>>     &QS
>>       METHOD GPW
>>       EXTRAPOLATION USE_GUESS
>>     &END QS
>>     &MGRID
>>       CUTOFF 900
>>       REL_CUTOFF 60
>>     &END MGRID
>>     &PRINT
>>       &MO OFF
>>       &END MO
>>        &PDOS DEBUG 
>>           NLUMO 5
>>        &END PDOS
>>     &END PRINT
>>     &SCF
>>       SCF_GUESS ATOMIC
>>       MAX_SCF 400
>>       ADDED_MOS 200
>>       &DIAGONALIZATION T
>>         ALGORITHM STANDARD
>>       &END DIAGONALIZATION
>>       &MIXING T
>>         NBUFFER 4
>>         BETA 1.25
>>         ALPHA 0.2
>>         METHOD BROYDEN_MIXING
>>       &END MIXING
>>       &PRINT
>>         &RESTART ON
>>         &END RESTART
>>       &END PRINT
>>       &SMEAR ON
>>         METHOD FERMI_DIRAC
>>         ELECTRONIC_TEMPERATURE [K] 300
>>       &END SMEAR
>>     &END SCF
>>     &XC
>>       &XC_FUNCTIONAL PBE
>>        * &PBE*
>> *          PARAMETRIZATION ORIG*
>> *        &END PBE*
>>       &END XC_FUNCTIONAL
>>       *&VDW_POTENTIAL                                                   
>>                                   *
>> *        POTENTIAL_TYPE PAIR_POTENTIAL                                   
>>                                     *
>> *        &PAIR_POTENTIAL                                                 
>>                                                    *
>> *            TYPE DFTD3                                                   
>>                                *
>> *            REFERENCE_FUNCTIONAL PBE                                     
>>                                   *
>> *            R_CUTOFF  10.0                                               
>>                                     *
>> *            PARAMETER_FILE_NAME /dftd3.dat                       *
>> *            SCALING 0.0                                                 
>>                                         *
>> *        &END PAIR_POTENTIAL                                             
>>                                          *
>> *      &END VDW_POTENTIAL *
>>     &END XC
>>   &END DFT
>>   &PRINT
>>       &STRESS_TENSOR ON
>>       &END STRESS_TENSOR
>>       &FORCES ON
>>       &END FORCES
>>   &END PRINT 
>> &END FORCE_EVAL
>> &MOTION
>>     &CELL_OPT
>>       OPTIMIZER BFGS
>>       MAX_DR 0.005
>>       MAX_ITER 100
>>       EXTERNAL_PRESSURE 0
>>       MAX_FORCE 0.0004
>>       RMS_DR 0.002
>>       PRESSURE_TOLERANCE 100
>>       RMS_FORCE 0.001
>>       TYPE DIRECT_CELL_OPT
>>     &END CELL_OPT                  
>>   &PRINT 
>>       &TRAJECTORY
>>           FORMAT PDB
>>       &END TRAJECTORY
>>   &END PRINT
>> &END MOTION               
>>
>> Thanks for any inputs.
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170710/cd2925d8/attachment.html>


More information about the CP2K-user mailing list