very long calculation time

Amin

unread,

Mar 13, 2017, 5:28:21 AM3/13/17

to cp2k

Hi all,

I am trying to calculate the adsorption energy of co2 on Tio2 surface (with 192 atoms in my supercell). but the problem is that it takes too long time to be calculated. I'm using 80 processors (2.8 GHz - each) but as I set the wall time for 7 days, it reached that wall time and calculations stopped. I have done similar calculations for fewer atoms (48,98) in super cell but in those cases (even with fewer CPUs) it takes at most 1 day to calculate everything.

Because I'm a beginner in CP2K, I don't have any idea weather this amount of time for calculation is reasonable. I have attached critical parts of my input file. I will be grateful if someone gives a comment regarding my question.

- Any suggestion for decreasing the calculation time with the same number of atoms is warmly appreciated.. :)

&GLOBAL

PROJECT Ch

RUN_TYPE GEO_OPT

PRINT_LEVEL LOW

EXTENDED_FFT_LENGTHS

&END GLOBAL

&FORCE_EVAL

METHOD Quickstep

&DFT

BASIS_SET_FILE_NAME BASIS_MOLOPT

POTENTIAL_FILE_NAME GTH_POTENTIALS

LSD

CHARGE 0

MULTIPLICITY 1

&MGRID

CUTOFF 800

REL_CUTOFF 60

&END MGRID

&QS

METHOD GPW

EPS_DEFAULT 1.0E-12

&END QS

&POISSON

PERIODIC XYZ

&END POISSON

&SCF

SCF_GUESS ATOMIC

EPS_SCF 1.0E-6

MAX_SCF 100

&OT

PRECONDITIONER FULL_ALL

MINIMIZER CG

&END OT

&OUTER_SCF

MAX_SCF 20

EPS_SCF 1.0E-6

&END OUTER_SCF

&END SCF

&XC

&XC_FUNCTIONAL PBE

&END XC_FUNCTIONAL

&VDW_POTENTIAL

DISPERSION_FUNCTIONAL PAIR_POTENTIAL

&PAIR_POTENTIAL

TYPE DFTD3(BJ)

REFERENCE_FUNCTIONAL PBE

CALCULATE_C9_TERM .FALSE.

PARAMETER_FILE_NAME dftd3.dat

R_CUTOFF 15.0

&END PAIR_POTENTIAL

&END VDW_POTENTIAL

&END XC

&END DFT

&SUBSYS

&CELL

ABC 15.138 15.138 30.0000

PERIODIC XYZ

&END CELL

&TOPOLOGY

COORD_FILE_FORMAT XYZ

COORD_FILE_NAME ./Anatase001-192.xyz

&END

&END FORCE_EVAL

&MOTION

&GEO_OPT

OPTIMIZER BFGS

MAX_FORCE 4.50000000E-004

&END

&CONSTRAINT

&FIXED_ATOMS

LIST 0..95

&END FIXED_ATOMS

&END CONSTRAINT

&END MOTION

hut...@chem.uzh.ch

unread,

Mar 13, 2017, 6:35:58 AM3/13/17

to cp...@googlegroups.com

Hi

is your 'long' simulation time caused by:

- slow convergence of geometry optimization
- slow convergence of SCF
- high CPU time per SCF cycle?

To improve CPU time per SCF I would suggest (guessing without
knowing your full input)
- use only MOLOPT-SR basis sets
- reduce EPS_DEFAULT to 1.e-10
- Preconditioner -> FULL_SINGLE_INVERSE

Do you really need LSD?

regards

Juerg Hutter
--------------------------------------------------------------
Juerg Hutter                         Phone : ++41 44 635 4491
Institut für Chemie C                FAX   : ++41 44 635 6838
Universität Zürich                   E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----To: cp2k <cp...@googlegroups.com>
From: Amin
Sent by: cp...@googlegroups.com
Date: 03/13/2017 10:28AM
Subject: [CP2K:8806] very long calculation time

Hi all,I am trying to calculate the adsorption energy of co2 on Tio2 surface (with 192 atoms in my supercell). but the problem is that it takes too long time to be calculated. I'm using 80 processors (2.8 GHz - each) but as I set the wall time for 7 days, it reached that wall time and calculations stopped. I have done similar calculations for fewer atoms (48,98) in super cell but in those cases (even with fewer CPUs) it takes at most 1 day to calculate everything. Because I'm a beginner in CP2K, I don't have any idea weather this amount of time for calculation is reasonable. I have attached critical parts of my input file. I will be grateful if someone gives a comment regarding my question.- Any suggestion for decreasing the calculation time with the same number of atoms is warmly appreciated.. :)
&GLOBAL PROJECT Ch RUN_TYPE GEO_OPT PRINT_LEVEL LOW EXTENDED_FFT_LENGTHS&END GLOBAL

&FORCE_EVAL METHOD Quickstep &DFT BASIS_SET_FILE_NAME BASIS_MOLOPT POTENTIAL_FILE_NAME GTH_POTENTIALS LSD CHARGE 0 MULTIPLICITY 1 &MGRID CUTOFF 800 REL_CUTOFF 60 &END MGRID &QS METHOD GPW EPS_DEFAULT 1.0E-12 &END QS &POISSON PERIODIC XYZ &END POISSON &SCF SCF_GUESS ATOMIC EPS_SCF 1.0E-6 MAX_SCF 100 &OT PRECONDITIONER FULL_ALL MINIMIZER CG &END OT &OUTER_SCF MAX_SCF 20 EPS_SCF 1.0E-6 &END OUTER_SCF &END SCF &XC &XC_FUNCTIONAL PBE &END XC_FUNCTIONAL &VDW_POTENTIAL DISPERSION_FUNCTIONAL PAIR_POTENTIAL &PAIR_POTENTIAL TYPE DFTD3(BJ) REFERENCE_FUNCTIONAL PBE CALCULATE_C9_TERM .FALSE. PARAMETER_FILE_NAME dftd3.dat R_CUTOFF 15.0 &END PAIR_POTENTIAL &END VDW_POTENTIAL &END XC &END DFT
&SUBSYS &CELL ABC 15.138 15.138 30.0000 PERIODIC XYZ &END CELL &TOPOLOGY COORD_FILE_FORMAT XYZ COORD_FILE_NAME ./Anatase001-192.xyz &END

&END FORCE_EVAL&MOTION &GEO_OPT OPTIMIZER BFGS MAX_FORCE 4.50000000E-004 &END &CONSTRAINT &FIXED_ATOMS LIST 0..95 &END FIXED_ATOMS &END CONSTRAINT&END MOTION

--

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 cp2k+uns...@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.

Visit this group at https://groups.google.com/group/cp2k.

For more options, visit https://groups.google.com/d/optout.

Amin

unread,

Mar 13, 2017, 6:57:14 AM3/13/17

to cp2k

Thank you very much for your swift reply.

I am replicating a calculation that has been done previously by VASP. at first I tried to calculate adsorption energy with 48 atoms in my supercell, but there was a major difference between my results and the previous one.

Then I modified my structure to include more atoms. I'll try to apply your suggestions but I also attach my complete input file in case that it might help to give a better perspective.

-Because of this point that calculation stopped after reaching the wall-time, I set the new input file to use the Restart wave-function. But I'm not sure that it's been done correctly! Sorry if my questions are so rudimentary!

I sincerely appreciate your time and considerations.

&GLOBAL

PROJECT Chemisorption-newconf

RUN_TYPE GEO_OPT

PRINT_LEVEL LOW

EXTENDED_FFT_LENGTHS

&END GLOBAL

&FORCE_EVAL

METHOD Quickstep

&DFT

WFN_RESTART_FILE_NAME Chemisorption-newconf-RESTART.wfn

BASIS_SET_FILE_NAME BASIS_MOLOPT

POTENTIAL_FILE_NAME GTH_POTENTIALS

LSD

CHARGE 0

MULTIPLICITY 1

&MGRID

CUTOFF 800

REL_CUTOFF 60

&END MGRID

&QS

METHOD GPW

EPS_DEFAULT 1.0E-12

&END QS

&POISSON

PERIODIC XYZ

&END POISSON

&SCF

SCF_GUESS RESTART

&KIND C

ELEMENT C

BASIS_SET DZVP-MOLOPT-SR-GTH

POTENTIAL GTH-PBE-q4

&END KIND

&KIND Ti

ELEMENT Ti

BASIS_SET DZVP-MOLOPT-SR-GTH

POTENTIAL GTH-PBE-q12

&END KIND

&KIND O

ELEMENT O

BASIS_SET DZVP-MOLOPT-SR-GTH

POTENTIAL GTH-PBE-q6

&END KIND

&END SUBSYS

&PRINT

&TOTAL_NUMBERS ON

&END TOTAL_NUMBERS

&END PRINT

&END FORCE_EVAL

&MOTION

&GEO_OPT

OPTIMIZER BFGS

MAX_FORCE 4.50000000E-004

&END

&CONSTRAINT

&FIXED_ATOMS

LIST 0..95

&END FIXED_ATOMS

&END CONSTRAINT

&END MOTION

Amin

unread,

Mar 13, 2017, 7:07:09 AM3/13/17

to cp2k

To add up some thing: I think that is because of the slow convergence of the SCF.

The interesting point is that with 48 atoms in my supercell, SCF converged in less than one hour! but now with 192 atoms, it takes more than a week (even more!, because it hasn't converged yet!)

hut...@chem.uzh.ch

unread,

Mar 13, 2017, 7:11:02 AM3/13/17

to cp...@googlegroups.com

Hi

you still didn't answer my question:

is your 'long' simulation time caused by:

- slow convergence of geometry optimization

==> Numbers please.

- slow convergence of SCF

==> How many iterations? Did it converge at all?

- high CPU time per SCF cycle?

==> Give us some numbers. It is impossible to say something
without having any information.

Do you really need LSD?

regards

--------------------------------------------------------------
Juerg Hutter                         Phone : ++41 44 635 4491
Institut für Chemie C                FAX   : ++41 44 635 6838
Universität Zürich                   E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----To: cp2k <cp...@googlegroups.com>
From: Amin
Sent by: cp...@googlegroups.com

Date: 03/13/2017 12:07PM
Subject: Re: [CP2K:8808] very long calculation time

To add up some thing: I think that is because of the slow convergence of the SCF.The interesting point is that with 48 atoms in my supercell, SCF converged in less than one hour! but now with 192 atoms, it takes more than a week (even more!, because it hasn't converged yet!)

On Monday, March 13, 2017 at 11:57:14 AM UTC+1, Amin wrote:Thank you very much for your swift reply.I am replicating a calculation that has been done previously by VASP. at first I tried to calculate adsorption energy with 48 atoms in my supercell, but there was a major difference between my results and the previous one. Then I modified my structure to include more atoms. I'll try to apply your suggestions but I also attach my complete input file in case that it might help to give a better perspective. -Because of this point that calculation stopped after reaching the wall-time, I set the new input file to use the Restart wave-function. But I'm not sure that it's been done correctly! Sorry if my questions are so rudimentary!

I sincerely appreciate your time and considerations.

&GLOBAL PROJECT Chemisorption-newconf RUN_TYPE GEO_OPT PRINT_LEVEL LOW EXTENDED_FFT_LENGTHS&END GLOBAL
&FORCE_EVAL METHOD Quickstep &DFT WFN_RESTART_FILE_NAME Chemisorption-newconf-RESTART.wfn BASIS_SET_FILE_NAME BASIS_MOLOPT POTENTIAL_FILE_NAME GTH_POTENTIALS LSD CHARGE 0 MULTIPLICITY 1 &MGRID CUTOFF 800 REL_CUTOFF 60 &END MGRID &QS METHOD GPW EPS_DEFAULT 1.0E-12 &END QS &POISSON PERIODIC XYZ &END POISSON &SCF SCF_GUESS RESTART EPS_SCF 1.0E-6 MAX_SCF 100 &OT PRECONDITIONER FULL_ALL MINIMIZER CG &END OT &OUTER_SCF MAX_SCF 20 EPS_SCF 1.0E-6 &END OUTER_SCF &END SCF &XC &XC_FUNCTIONAL PBE &END XC_FUNCTIONAL &VDW_POTENTIAL DISPERSION_FUNCTIONAL PAIR_POTENTIAL &PAIR_POTENTIAL TYPE DFTD3(BJ) REFERENCE_FUNCTIONAL PBE CALCULATE_C9_TERM .FALSE. PARAMETER_FILE_NAME dftd3.dat R_CUTOFF 15.0 &END PAIR_POTENTIAL &END VDW_POTENTIAL &END XC &END DFT
&SUBSYS &CELL ABC 15.138 15.138 30.0000 PERIODIC XYZ &END CELL &TOPOLOGY COORD_FILE_FORMAT XYZ COORD_FILE_NAME ./Anatase001-192.xyz &END &KIND C ELEMENT C BASIS_SET DZVP-MOLOPT-SR-GTH POTENTIAL GTH-PBE-q4 &END KIND &KIND Ti ELEMENT Ti BASIS_SET DZVP-MOLOPT-SR-GTH POTENTIAL GTH-PBE-q12 &END KIND &KIND O ELEMENT O BASIS_SET DZVP-MOLOPT-SR-GTH POTENTIAL GTH-PBE-q6 &END KIND &END SUBSYS &PRINT &TOTAL_NUMBERS ON &END TOTAL_NUMBERS &END PRINT&END FORCE_EVAL&MOTION &GEO_OPT OPTIMIZER BFGS MAX_FORCE 4.50000000E-004 &END &CONSTRAINT &FIXED_ATOMS LIST 0..95 &END FIXED_ATOMS &END CONSTRAINT&END MOTION

Amin Jafarzadeh

unread,

Mar 13, 2017, 8:13:55 AM3/13/17

to cp...@googlegroups.com

Frankly, I don't know how to extract these information! I attached output file here.
I need LSD because the calculation done by VASP was spin polarized.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.

Visit this group at https://groups.google.com/group/cp2k.

For more options, visit https://groups.google.com/d/optout.

--

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/EFggHvw_Rdw/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cp2k+unsubscribe@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--

------------------------------------------------------------

Amin Jafarzadeh

PhD Candidate,

Research Group PLASMANT, University of Antwerp
Universiteitplein 1, BE-2610 Antwerp, Belgium

Tel: +32-32652364

Fax: +32-32652343

Chemisorption001-Res.out

Amin

unread,

Mar 13, 2017, 8:43:10 AM3/13/17

to cp...@googlegroups.com

I couldn't be able to attach the output file here.
I need LSD, because the calculation done by VASP was spin polarized.

Here is the information about the last step of optimization- after that calculations reached wall-time and stopped:

-------- Informations at step = 18 ------------

Optimization Method = BFGS

Total Energy = -5827.0471607579

Real energy change = 0.0858127047

Predicted change in energy = -14.5174588205

Scaling factor = 0.0077495614

Step size = 0.4724315332

Trust radius = 0.4724315332

Decrease in energy = NO

Used time = 93.234

Convergence check :

Max. step size = 0.4724315332

Conv. limit for step size = 0.0030000000

Convergence in step size = NO

RMS step size = 0.0495642754

Conv. limit for RMS step = 0.0015000000

Convergence in RMS step = NO

Max. gradient = 0.1337540443

Conv. limit for gradients = 0.0004500000

Conv. for gradients = NO

RMS gradient = 0.0130474061

Conv. limit for RMS grad. = 0.0003000000

Conv. for gradients = NO

---------------------------------------------------

----------------------------------- OT ---------------------------------------

Minimizer : CG : conjugate gradient

Preconditioner : FULL_ALL : diagonalization, state selective

Precond_solver : DEFAULT

Line search : 2PNT : 2 energies, one gradient

stepsize : 0.15000000 energy_gap : 0.08000000

eps_taylor : 0.10000E-15 max_taylor : 4

----------------------------------- OT ---------------------------------------

Step Update method Time Convergence Total energy Change

------------------------------------------------------------------------------

1 OT CG 0.15E+00 8.8 0.00180844 -5825.0697567570 -5.83E+03

Marcella Iannuzzi

unread,

Mar 13, 2017, 9:35:47 AM3/13/17

to cp2k

Dear Amin,

if this is your whole output, only 18 steps of the geometry optimisation have been performed,

each taking between 40 and 100 seconds. The SCF cycle requires actually a small number of iterations, each taking few seconds.

From these information, one can only conclude that your job was hanging for a week, after the calculation of these 18 steps.

Did you check when the output file has been modified for the last time? Did you check every time to time ow the calculation was going on during that week?

Anyway, I would guess that the problem is related to the machine where the job is running.

regards

Marcella

On Monday, March 13, 2017 at 1:43:10 PM UTC+1, Amin wrote:

I couldn't be able to attach the output file here.
I need LSD, because the calculation done by VASP was spin polarized.

Here are the information about the last step of optimization- after that calculations reached wall-time and stopped:

Amin

unread,

Mar 13, 2017, 10:10:02 AM3/13/17

to cp2k

Dear Marcella,

Your guess is very likely to be correct. because just a couple of minutes ago an other job that I had submitted before (same structure without co2 molecule) stopped (because of wall-time) after a week of calculation and as I checked the output file for last modification time, I strangely found out that the last modification time was about yesterday!! That says there has been nothing written since yesterday on the output file while the program just stopped right now!

Apparently, I need to contact technical staff.

Thank you very much.

Regards,

Amin

unread,

Mar 16, 2017, 5:38:19 AM3/16/17

to cp2k

Hi,

I noticed that the main problem for me is that in some cases CP2K stops updating the output file after a few minutes from the beginning of the calculation.

For example: I put 2 jobs in the cluster (one for Slab + Co2 and the other was just Tio2 Slab), after almost 20 minutes the calculation for slab completed very nicely, but for slab + co2 it's still running (after several hours!) and there is no update on the output file!

It also happened for me during an other calculation. in both cases, program stops to write on the output file and it continues to be in this situation until the wall-time is reached! and then job is being killed by the machine.

Now here is the question: what can cause this problem? to gain some straight, what kind of errors in input file, structure,... can stop the program from generating the output file?? And why isn't there any error in the output file?

Once that I found out that the reason can not be about the input file or structure then I can be sure that it is some kind of technical problem with the machine that I'm using.

I wound be very grateful if someone shares her/his experiences with me in this case.

Amin

unread,

Mar 21, 2017, 8:52:24 AM3/21/17

to cp2k

To close this thread, I want to mention that as I changed the version of CP2K from 3.0 to 2.6 the problem solved! I did it after changing every single parameter with my input file and structure. I have no idea that was a problem with CP2K itself or just the version 3.0 package was not installed correctly on our machine.

Stefan Becuwe

unread,

Mar 23, 2017, 6:02:42 AM3/23/17

to cp2k

I further investigated Amin's case on the cluster he's using and on another cluster too. The difference between 2.6 and 3.0 he noticed, is related to dependencies. Both versions are installed correctly, as far as we can see. It turns out that the computation does finish when I use CP2K without any dependencies (both 2.6 and 3.0). However, when Libint (1.1.4) and/or libxc (2.2.3) are dependencies, the computation seems to hang after 10 minutes, meaning no output is being written any longer (tested with CP2K 2.7, 3.0 and 4.1). Are any issues known with Libint or libxc?

Regards,

Stefan

Op dinsdag 21 maart 2017 13:52:24 UTC+1 schreef Amin:

Reply all

Reply to author

Forward