[CP2K-user] [CP2K:14617] Re: Struggling to get SCF Convergence in Cell_Opt run

Lenard Carroll lenardc... at gmail.com
Fri Jan 22 08:44:26 UTC 2021
Previous message (by thread): [CP2K-user] [CP2K:14614] Re: Struggling to get SCF Convergence in Cell_Opt run
Next message (by thread): [CP2K-user] Building for AVX512 while compiling on a AVX128 machine
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Fantastic, thanks for the info. I've only recently started my journey into
DFT, been doing MD simulations for years now.

On Fri, Jan 22, 2021 at 10:41 AM Travis <polla... at gmail.com> wrote:

> Hi,
>
> For small band gap materials, you *need* to apply a broadening or
> smearing function that smooths out the energy levels near the Fermi level
> so that you don't oscillate between filling a low lying unoccupied level on
> one SCF iteration, emptying it on the next, and so on and so on. There are
> multiple methods for this such (refer to VASP manual for example,
> https://www.vasp.at/wiki/index.php/ISMEAR), CP2K implements Fermi
> smearing. You apply an electronic temperature that describes the shape of
> the broadening function. *It isn't a physical temperature.*
>
> -T
>
> On Friday, January 22, 2021 at 3:19:27 AM UTC-5 ASSIDUO Network wrote:
>
>> Then I cannot use smearing, I guess I'll have to turn it off or will have
>> to talk to my supervisor about using a non-zero temperature.
>>
>> On Fri, Jan 22, 2021 at 10:17 AM Marcella Iannuzzi <ma... at gmail.com>
>> wrote:
>>
>>>
>>>
>>> No smearing with 0K does not help.
>>> Smearing needs a finite temperature, can be lower than 500, for instance
>>> 300K, not 0K
>>> On Friday, January 22, 2021 at 9:04:31 AM UTC+1 ASSIDUO Network wrote:
>>>
>>>> I'm going to follow your advice, but if I remember correctly, last time
>>>> I tried smearing with 0 K, the stimulation ended with an error. If it does
>>>> it again, I will ask for some feedback again.
>>>>
>>>> On Friday, January 22, 2021 at 9:59:40 AM UTC+2 Marcella Iannuzzi wrote:
>>>>
>>>>>
>>>>>
>>>>> Hi
>>>>>
>>>>> The smearing with finite electronic temperature is needed in order to
>>>>> allow convergence of the metallic electronic structure, even if the
>>>>> structure is kept a 0K.  The number of added MOS depends on the density of
>>>>> states around Fermi.  Too many states do not hurt, in terms of convergence.
>>>>> I just chose a safe number, probably less are also OK, as a lower smearing
>>>>> temperature would also work.
>>>>> As a functional I took one that I often use. Anyway, I think that
>>>>> adding VDW is a good idea.
>>>>> All these aspects are more general on electronic structure theory and
>>>>> not specific of CP2K. Please read in the literature for better
>>>>> understanding.
>>>>> Regards
>>>>> Marcella
>>>>>
>>>>>
>>>>> On Friday, January 22, 2021 at 8:37:46 AM UTC+1 ASSIDUO Network wrote:
>>>>>
>>>>>> I should have added this earlier, the simulation must be performed at
>>>>>> 0 K (my supervisor's request), it's a static run. Therefore, I cannot set
>>>>>> the temperature to 500 K. Also, why the need for ADDED_MOS=200? In another
>>>>>> thread, I was told that ADDED_MOS=100 was too much.
>>>>>>
>>>>>> Also, why the use of  FUNCTIONAL  XC_GGA_C_PBE and  FUNCTIONAL
>>>>>> XC_GGA_X_RPW86? I'm new to CP2K so just want to understand everything
>>>>>> better.
>>>>>> On Friday, January 22, 2021 at 9:24:10 AM UTC+2 Marcella Iannuzzi
>>>>>> wrote:
>>>>>>
>>>>>>> Hi Lenard,
>>>>>>>
>>>>>>> I got it converging in 10 iterations.
>>>>>>> The outer SCF with diagonalisation is useless, since there is no
>>>>>>> preconditioner.
>>>>>>> With metals you need to use smearing.
>>>>>>> Here are some settings I used:
>>>>>>>       ADDED_MOS 200
>>>>>>>       &DIAGONALIZATION T
>>>>>>>         ALGORITHM STANDARD
>>>>>>>       &END DIAGONALIZATION
>>>>>>>       &MIXING T
>>>>>>>         METHOD BROYDEN_MIXING
>>>>>>>         ALPHA 0.01
>>>>>>>         BETA 0.5
>>>>>>>         NBUFFER 8
>>>>>>>       &END MIXING
>>>>>>>       &SMEAR
>>>>>>>         METHOD FERMI_DIRAC
>>>>>>>         ELECTRONIC_TEMPERATURE 500
>>>>>>>       &END SMEAR
>>>>>>>
>>>>>>>     &XC
>>>>>>>       &XC_FUNCTIONAL
>>>>>>>           &LIBXC
>>>>>>>             FUNCTIONAL XC_GGA_X_RPW86
>>>>>>>           &END LIBXC
>>>>>>>           &LIBXC
>>>>>>>             FUNCTIONAL  XC_GGA_C_PBE
>>>>>>>           &END LIBXC
>>>>>>>       &END XC_FUNCTIONAL
>>>>>>>
>>>>>>>       &VDW_POTENTIAL
>>>>>>>        POTENTIAL_TYPE NON_LOCAL
>>>>>>>         &NON_LOCAL
>>>>>>>           CUTOFF 300
>>>>>>>           TYPE RVV10
>>>>>>> ##          VERBOSE_OUTPUT
>>>>>>>           KERNEL_FILE_NAME ${data}/rVV10_kernel_table.dat
>>>>>>>         &END NON_LOCAL
>>>>>>>       &END VDW_POTENTIAL
>>>>>>>     &END XC
>>>>>>>
>>>>>>> The results:
>>>>>>>
>>>>>>>  Step     Update method      Time    Convergence         Total
>>>>>>> energy    Change
>>>>>>>
>>>>>>> ------------------------------------------------------------------------------
>>>>>>>      1 NoMix/Diag. 0.10E-01   20.0     0.51067755
>>>>>>>  -133.2796876462 -1.33E+02
>>>>>>>      2 Broy./Diag. 0.10E-01   17.6     0.00064724
>>>>>>>  -136.0334243526 -2.75E+00
>>>>>>>      3 Broy./Diag. 0.10E-01   17.5     0.03257808
>>>>>>>  -134.7316158415  1.30E+00
>>>>>>>      4 Broy./Diag. 0.10E-01   17.7     0.00019866
>>>>>>>  -133.0666478987  1.66E+00
>>>>>>>      5 Broy./Diag. 0.10E-01   17.6     0.00228816
>>>>>>>  -133.1462861174 -7.96E-02
>>>>>>>      6 Broy./Diag. 0.10E-01   17.6     0.00032933
>>>>>>>  -133.1654553845 -1.92E-02
>>>>>>>      7 Broy./Diag. 0.10E-01   17.6     0.00000406
>>>>>>>  -133.1816175525 -1.62E-02
>>>>>>>      8 Broy./Diag. 0.10E-01   17.7     0.00009047
>>>>>>>  -133.1825852315 -9.68E-04
>>>>>>>      9 Broy./Diag. 0.10E-01   17.6     0.00000504
>>>>>>>  -133.1830490186 -4.64E-04
>>>>>>>     10 Broy./Diag. 0.10E-01   17.6     0.00000031
>>>>>>>  -133.1828944498  1.55E-04
>>>>>>>
>>>>>>>   *** SCF run converged in    10 steps ***
>>>>>>>
>>>>>>>
>>>>>>>   Electronic density on regular grids:        -44.0000000000
>>>>>>>  0.0000000000
>>>>>>>   Core density on regular grids:               43.9999999999
>>>>>>> -0.0000000001
>>>>>>>   Total charge density on r-space grids:       -0.0000000001
>>>>>>>   Total charge density g-space grids:          -0.0000000001
>>>>>>>
>>>>>>>   Overlap energy of the core charge distribution:
>>>>>>> 0.00000001219968
>>>>>>>   Self energy of the core charge distribution:
>>>>>>> -231.41335460772382
>>>>>>>   Core Hamiltonian energy:
>>>>>>> 74.19344628639691
>>>>>>>   Hartree energy:
>>>>>>>  45.27318657026385
>>>>>>>   Exchange-correlation energy:
>>>>>>>  -21.30000506759340
>>>>>>>   Dispersion energy:
>>>>>>>  0.06400174824714
>>>>>>>   Electronic entropic energy:
>>>>>>>  -0.00016939092888
>>>>>>>   Fermi energy:
>>>>>>> 0.34714684334798
>>>>>>>
>>>>>>>   Total energy:
>>>>>>>  -133.18289444982969
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Regards
>>>>>>> Marcella
>>>>>>>
>>>>>>> On Friday, January 22, 2021 at 6:48:12 AM UTC+1 ASSIDUO Network
>>>>>>> wrote:
>>>>>>>
>>>>>>>> Hi there everyone, hope you've had a great week.
>>>>>>>>
>>>>>>>> I've been trying to run a cell optimization of bulk Au, and I am
>>>>>>>> using the attached input file, but I'm not getting an inner loop SCF
>>>>>>>> convergence. I've made many small changes, such as including/excluding
>>>>>>>> OUTER_SCF,  changing the SCF convergence criterion, changing the number of
>>>>>>>> cell optimization steps, changing the number of KPoints and changing the
>>>>>>>> mixing method. Nothing has worked. I haven't tried a combination of the
>>>>>>>> above though.
>>>>>>>>
>>>>>>>> Do you perhaps have any suggestions to me on how to get
>>>>>>>> convergence? Furthermore, I would also appreciate some tips to speed up my
>>>>>>>> simulations (settings/flags) wise.
>>>>>>>>
>>>>>>>> Thanks in advance,
>>>>>>>> Lenard
>>>>>>>>
>>>>>>>>
>>>>>>>> --
>>> You received this message because you are subscribed to the Google
>>> Groups "cp2k" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to cp... at googlegroups.com.
>>> To view this discussion on the web visit
>>> https://groups.google.com/d/msgid/cp2k/a48a92ce-995f-489b-a822-67e2e83e95e2n%40googlegroups.com
>>> <https://groups.google.com/d/msgid/cp2k/a48a92ce-995f-489b-a822-67e2e83e95e2n%40googlegroups.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>> --
> You received this message because you are subscribed to the Google Groups
> "cp2k" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to cp... at googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/cp2k/c8e010ca-db0d-459a-8051-678bfb7f4ae0n%40googlegroups.com
> <https://groups.google.com/d/msgid/cp2k/c8e010ca-db0d-459a-8051-678bfb7f4ae0n%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210122/3e23e203/attachment.htm>
Previous message (by thread): [CP2K-user] [CP2K:14614] Re: Struggling to get SCF Convergence in Cell_Opt run
Next message (by thread): [CP2K-user] Building for AVX512 while compiling on a AVX128 machine
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the CP2K-user mailing list