cutoff convergence problems
hanno....@kit.edu
Dear cp2k users,
For a convergence study of the cutoff/rel cutoff I wrote the input attached to this post. When running I face two challenges: independent of the system under investigation (for e.g. a single water molecule or single Si atom) there is an enormous jump in the total energy, when increasing the cutoff from 249 to 250 (for e.g. H2O -30.83018512 a.u. to 0.160674035 a.u. or for single Si from -8.282002659 a.u. to 0.428439103 a.u.). In addition, an increase of the cutoff to 350 or higher produces a segmentation fault. The input and the xyz file for the water example and the segmentation error report are printed at the bottom. The calculations were performed with CP2K version 6.1 on 16 parallelized Intel(R) Xeon(R) E5-2630 v4 @ 2.20GHz CPUs.
All ideas and suggestions concerning these observations are welcome. Thanks for your help in advance.
Hanno Schütz
PhD student
HIU-Helmholtz Institute Ulm
--------------------------------------------------------------------------------------------------------------------------------------
Input file
--------------------------------------------------------------------------------------------------------------------------------------
!Defines filename of geometry input file in xyz format.If not in same folder use complete path.
!Be carefull input must only contain element symbol x y z, no additional text or characters.
@SET COORD_INP '~/cp2k/H2O_cutoff_test_16np/H2O_test.xyz'
!Sets Base set for DFT
@SET BASE DZVP-GTH
!Sets effective core potential for DFT
@SET POT GTH-BLYP
&GLOBAL
PROJECT H2O_CUTOFF_16np_r40_249
RUN_TYPE ENERGY
PRINT_LEVEL MEDIUM
&END GLOBAL
&FORCE_EVAL
&DFT
BASIS_SET_FILE_NAME ~/cp2k/BASIS_SETS
POTENTIAL_FILE_NAME ~/cp2k/POTENTIALS
&MGRID
CUTOFF 249
REL_CUTOFF 40
NGRIDS 4
&END
&SCF
SCF_GUESS ATOMIC
MAX_SCF 1
&OT
MINIMIZER DIIS
PRECONDITIONER FULL_SINGLE_INVERSE
&END
&END
&XC
&XC_FUNCTIONAL BLYP
&END
&XC_GRID
XC_DERIV NN10_SMOOTH
XC_SMOOTH_RHO NN10
&END
&END
&END
&SUBSYS
&CELL
ABC 21.069647 21.069647 21.069647
&END
&COORD
@INCLUDE ${COORD_INP}
&END
&KIND H
BASIS_SET ${BASE}
POTENTIAL ${POT}-q1
&END
&KIND C
BASIS_SET ${BASE}
POTENTIAL ${POT}-q4
&END
&KIND N
BASIS_SET ${BASE}
POTENTIAL ${POT}-q5
&END
&KIND O
BASIS_SET ${BASE}
POTENTIAL ${POT}-q6
&END
&KIND F
BASIS_SET ${BASE}
POTENTIAL ${POT}-q7
&END
&KIND S
BASIS_SET ${BASE}
POTENTIAL ${POT}-q6
&END
&END
&END
--------------------------------------------------------------------------------------------------------------------------------------
xyz water
--------------------------------------------------------------------------------------------------------------------------------------
O -1.59084 1.93421 0.01717
H -0.60151 1.90292 -0.01212
H -1.88020 1.24282 -0.63002
--------------------------------------------------------------------------------------------------------------------------------------
Segmentation fault for each CPU
--------------------------------------------------------------------------------------------------------------------------------------
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
cp2k.popt 00000000075BAA41 Unknown Unknown Unknown
cp2k.popt 00000000075B8B7B Unknown Unknown Unknown
cp2k.popt 0000000007572EE4 Unknown Unknown Unknown
cp2k.popt 0000000007572CF6 Unknown Unknown Unknown
cp2k.popt 0000000007511129 Unknown Unknown Unknown
cp2k.popt 0000000007519136 Unknown Unknown Unknown
libpthread-2.22.s 00007F6232995C10 Unknown Unknown Unknown
cp2k.popt 00000000023D96B4 pw_poisson_method 300 pw_poisson_methods.F
cp2k.popt 0000000000A3D0E5 qs_ks_methods_mp_ 435 qs_ks_methods.F
cp2k.popt 0000000000A3BAEB qs_ks_methods_mp_ 1204 qs_ks_methods.F
cp2k.popt 000000000157DD50 qs_scf_mp_init_sc 747 qs_scf.F
cp2k.popt 00000000015784BE qs_scf_mp_scf_env 421 qs_scf.F
cp2k.popt 000000000156E540 qs_scf_mp_scf_ 231 qs_scf.F
cp2k.popt 00000000012087B9 qs_energy_mp_qs_e 86 qs_energy.F
cp2k.popt 00000000013551B1 qs_force_mp_qs_ca 115 qs_force.F
cp2k.popt 00000000010B00DA force_env_methods 242 force_env_methods.F
cp2k.popt 0000000000411F7C cp2k_runs_mp_run_ 323 cp2k_runs.F
cp2k.popt 0000000000408B14 MAIN__ 281 cp2k.F
cp2k.popt 000000000040781E Unknown Unknown Unknown
libc-2.22.so 00007F6230E3E725 __libc_start_main Unknown Unknown
cp2k.popt 0000000000407729 Unknown Unknown Unknown
hut...@chem.uzh.ch
there is something fundamentally wrong with your results.
With your setting I get an energy for the water molecule of -17.157697 a.u.
Check you implementation and input file (why only 1 SCF iteration?).
regards
Juerg
--------------------------------------------------------------
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: hanno....@kit.edu
Sent by: cp...@googlegroups.com
Date: 11/05/2018 04:55PM
Subject: [CP2K:10901] cutoff convergence problems
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.
hanno....@kit.edu
Hi Juerg,
thanks for your answer. The reason why the number of SCF iterations was set to 1 is, that I generated this input to converge the cutoff and rel_cutoff of large systems (several hundrets of atoms, H2O is just a test system). This approach was described in the HOWTO on cp2k homepage https://www.cp2k.org/howto:converging_cutoff . The probleme with my input is the large energy jump when increasing the cutoff from 249 to 250 that appears with all systems I tested so far. To your energy value: did you increase the number of SCF cycles or change anything else?
Again thank you for your help.
Best,
hut...@chem.uzh.ch
my energy was with converged SCF, initial guess (SCF=1) is about 16.5 a.u.
The change in energy from 249->250 is about 0.005 a.u.
This is because of the discrete FFT grids. With 249 Ry you get a grid with 200 points,
with 250 Ry you get 216 points.
regards
Juerg
--------------------------------------------------------------
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: hanno....@kit.edu
Sent by: cp...@googlegroups.com
Subject: [CP2K:10911] Re: cutoff convergence problems