[CP2K-user] [CP2K:13327] SCCS Solvation Model Implementation - Molecule Radii Explodes
Krack Matthias (PSI)
matthi... at psi.ch
Mon May 18 06:52:48 UTC 2020
Hi Dev
Never tried such large dielectric constants of 100-1000. Values much larger than 80 for a solvent seem to me unusual.
Matthias
From: cp... at googlegroups.com <cp... at googlegroups.com> On Behalf Of Dev Rana
Sent: Sonntag, 17. Mai 2020 18:02
To: cp2k <cp... at googlegroups.com>
Subject: Re: [CP2K:13327] SCCS Solvation Model Implementation - Molecule Radii Explodes
Hi Matthias,
Thank you so much! This worked. However, when I go beyond a dieletric constant of 300 i.e. 300, 400, 500, 1000+, with no other changes to the input file. The convergence starts to go beyond 1 again. Do you have a reference that might help me understand what input parameters I should be using to offset my changing conditions? I'm going to try and print a cube file as you suggested and visualize it in VMD, I'm assuming with the newer higher dieletric constants the surface might be too close to the molecule and I will have to change my rho min/max to adjust for it. however I'm not sure by how much and I hate to keep troubling you.
Best Regards,
Dev
On Friday, May 15, 2020 at 1:16:45 PM UTC-4, Matthias Krack wrote:
Hi Dev
The attached revised input converged after a few steps using SCCS with a dielectric constant of 80. I changed the method and adapted the rho_min and rho_max values which define the surface of the cavity and thus they are system dependent to some extent. If rho_max is chosen too large for instance, the surface might be too close to the molecule. It is often helpful to plot the cavity, e.g. with VMD, using a cube file dump of the dielectric function<https://manual.cp2k.org/cp2k-7_1-branch/CP2K_INPUT/FORCE_EVAL/DFT/PRINT/SCCS/DIELECTRIC_FUNCTION.html>.
HTH
Matthias
From: c... at googlegroups.com<javascript:> <c... at googlegroups.com<javascript:>> On Behalf Of Dev Rana
Sent: Freitag, 15. Mai 2020 17:47
To: cp2k <c... at googlegroups.com<javascript:>>
Subject: Re: [CP2K:13314] SCCS Solvation Model Implementation - Molecule Radii Explodes
Hi Matthias,
Thank you again for all your help. I went a bit beyond your suggestion and changed EPS_SCF to 1E-12 and changed cell size to 10 x 10 x 10 A3. To answer your questions:
1.) I was able to run GEO_OPT without SCCS and the simulation completed.
2.) I was able to converge and complete a SCCS GEO_OPT simulation with a dielectric constant of 1.
I then took the dieletric constant back up to 80 (with no other changes), and the convergence fails - in fact the convergence Hartree's are trending greater than 1.
I'm going to do a small ramp up in dielectric constant: 2, 5, 10, etc. To see what happens. But I should be able to use 80+ as the dielectric constant without issues.
What do you think? Am I missing something?
Again input, output, and XYZ attached.
&GLOBAL
PROJECT SCCS
RUN_TYPE geo_opt
PRINT_LEVEL medium
SAVE_MEM TRUE
&END GLOBAL
&FORCE_EVAL
METHOD Quickstep
# STRESS_TENSOR analytical
&DFT
BASIS_SET_FILE_NAME /shared/centos7/cp2k/cp2k-6.1.0/data/BASIS_MOLOPT
POTENTIAL_FILE_NAME /shared/centos7/cp2k/cp2k-6.1.0/data/POTENTIAL
&QS
METHOD GPW
EPS_DEFAULT 1.0E-12
EXTRAPOLATION ASPC
&END QS
&MGRID
NGRIDS 5
CUTOFF [Ry] 600
&END MGRID
&SCCS
ALPHA [N*m^-1] 0.0
BETA [GPa] 0.0
DELTA_RHO 2.0E-5
DERIVATIVE_METHOD fft
DIELECTRIC_CONSTANT 80
GAMMA 0.0
EPS_SCCS 1.0E-12
! EPS_SCF 1.0E-10
MAX_ITER 100
! METHOD Andreussi
METHOD Fattebert-Gygi
MIXING 0.6
! &ANDREUSSI
! RHO_MAX 0.008
! RHO_MIN 0.00015
! &END ANDREUSSI
&FATTEBERT-GYGI
BETA 1.3
RHO_ZERO 0.00078
&END FATTEBERT-GYGI
&END SCCS
&SCF
SCF_GUESS ATOMIC
MAX_SCF 300
EPS_SCF 1.0E-12
&DIAGONALIZATION ON
ALGORITHM STANDARD
&END DIAGONALIZATION
! &OT on
! MINIMIZER DIIS
! PRECONDITIONER FULL_ALL
! ENERGY_GAP 0.002
! &END OT
&OUTER_SCF
EPS_SCF 1e-12
MAX_SCF 30
&END OUTER_SCF
&END SCF
&XC
&XC_FUNCTIONAL PBE
&END XC_FUNCTIONAL
&END XC
&PRINT
&PDOS
&EACH
GEO_OPT 1
&END EACH
APPEND TRUE
&END PDOS
&SCCS ON
&EACH
GEO_OPT 10
&END EACH
&DENSITY_GRADIENT off
&EACH
GEO_OPT 10
&END EACH
APPEND
&END DENSITY_GRADIENT
&DIELECTRIC_FUNCTION off
&EACH
GEO_OPT 10
&END EACH
APPEND
&END DIELECTRIC_FUNCTION
&POLARISATION_POTENTIAL off
&EACH
GEO_OPT 10
&END EACH
APPEND
&END POLARISATION_POTENTIAL
&END SCCS
&END PRINT
&END DFT
&SUBSYS
&CELL
ABC [angstrom] 10 10 10
! PERIODIC XYZ
! MULTIPLE_UNIT_CELL 1 1 1
&END CELL
&TOPOLOGY
COORD_FILE_NAME Al4C.xyz
COORD_FILE_FORMAT XYZ
! MULTIPLE_UNIT_CELL 1 1 1
&END
&KIND C
ELEMENT C
BASIS_SET DZVP-MOLOPT-GTH
POTENTIAL GTH-PBE-q4
&END KIND
&KIND Al
ELEMENT Al
BASIS_SET DZVP-MOLOPT-SR-GTH
POTENTIAL GTH-PBE-q3
&END KIND
&END SUBSYS
&END FORCE_EVAL
&MOTION
&GEO_OPT
TYPE MINIMIZATION
OPTIMIZER BFGS
MAX_ITER 100
&END GEO_OPT
&CONSTRAINT
&FIXED_ATOMS
COMPONENTS_TO_FIX XYZ
LIST 1
&END FIXED_ATOMS
&END CONSTRAINT
&PRINT
&STRUCTURE_DATA
DISTANCE 1 2
DISTANCE 1 3
DISTANCE 1 4
DISTANCE 1 5
DISTANCE 2 3
DISTANCE 2 4
DISTANCE 2 5
DISTANCE 3 4
DISTANCE 3 5
DISTANCE 4 5
ANGLE 1 2 3
ANGLE 1 2 4
ANGLE 1 2 5
ANGLE 1 3 2
ANGLE 1 3 4
ANGLE 1 3 5
ANGLE 1 4 2
ANGLE 1 4 3
ANGLE 1 4 5
ANGLE 1 5 2
ANGLE 1 5 3
ANGLE 1 5 4
&EACH
GEO_OPT 1
&END EACH
&END STRUCTURE_DATA
&TRAJECTORY
&EACH
GEO_OPT 1
&END EACH
&END TRAJECTORY
&VELOCITIES
&EACH
GEO_OPT 1
&END EACH
&END VELOCITIES
&FORCES
&EACH
GEO_OPT 1
&END EACH
&END FORCES
&CELL
&EACH
GEO_OPT 1
&END EACH
&END CELL
&RESTART
&EACH
GEO_OPT 1
&END EACH
&END RESTART
&END PRINT
&END MOTION
On Wednesday, May 13, 2020 at 8:08:47 PM UTC-4, Matthias Krack wrote:
Hi Dev
The settings in the regression tests are often physically not meaningful, since these test runs have to be completed after a few seconds.
SCCS%EPS_SCF delays the start of the inner SCCS convergence loop only for values greater zero. Otherwise that value is not considered and the inner SCCS cycle is activated immediately.
SCCS%EPS_SCCS, the convergence threshold of the inner SCCS loop for the polarization charge should be smaller (tighter) than 1.0E-6, at least 1.0E-8. As soon as the inner SCCS loop is activated, the timings for each SCF iteration step will increase significantly. Also the implicit solvent comes not for free. You can reduce the box size to save computer resources, because your system is small. Does the setup for your system CAl4 work/converge properly for a geo_opt run in gas phase, i.e. without SCCS? If that is the case then I would start with a smaller dielectric constant than 80 for the SCCS and see how the system evolves. The print level medium provides a detailed SCCS output.
HTH
Matthias
From: c... at googlegroups.com<mailto:c... at googlegroups.com> <c... at googlegroups.com<mailto:c... at googlegroups.com>> On Behalf Of Dev Rana
Sent: Donnerstag, 14. Mai 2020 00:29
To: cp2k <c... at googlegroups.com<mailto:c... at googlegroups.com>>
Subject: Re: [CP2K:13298] SCCS Solvation Model Implementation - Molecule Radii Explodes
I take back what I previously said. The 0th geo opt step finishes quickly with SCF steps taking 20-25s each. The 2st geo opt step does not finish and each SCF step takes 5+ minutes. Here's my input file and output for that. I'm also running this on GPU.
What do you advise? I've attached the input output files and paste input below. The convergence values keep increases, rather than decreases. And Total Energy keep lowering.
&GLOBAL
PROJECT SCCS
RUN_TYPE geo_opt
PRINT_LEVEL low
SAVE_MEM TRUE
&END GLOBAL
&FORCE_EVAL
METHOD Quickstep
# STRESS_TENSOR analytical
&DFT
BASIS_SET_FILE_NAME /shared/centos7/cp2k/cp2k-6.1.0/data/BASIS_MOLOPT
POTENTIAL_FILE_NAME /shared/centos7/cp2k/cp2k-6.1.0/data/POTENTIAL
&QS
METHOD GPW
EPS_DEFAULT 1.0E-12
EXTRAPOLATION ASPC
&END QS
&MGRID
NGRIDS 5
CUTOFF [Ry] 600
&END MGRID
&SCCS
ALPHA [N*m^-1] 0.0
BETA [GPa] 0.0
DELTA_RHO 2.0E-5
DERIVATIVE_METHOD cd5
DIELECTRIC_CONSTANT 80
GAMMA [mN/m] 0.0
EPS_SCCS 1.0E-6
EPS_SCF 1.0E-7
MAX_ITER 100
! METHOD Andreussi
METHOD Fattebert-Gygi
MIXING 0.6
! &ANDREUSSI
! RHO_MAX 0.001
! RHO_MIN 0.0001
! &END ANDREUSSI
&FATTEBERT-GYGI
BETA 1.3
RHO_ZERO 0.00078
&END FATTEBERT-GYGI
&END SCCS
&SCF
SCF_GUESS ATOMIC
MAX_SCF 300
EPS_SCF 1.0E-7
&DIAGONALIZATION ON
ALGORITHM STANDARD
&END DIAGONALIZATION
! &OT on
! MINIMIZER DIIS
! PRECONDITIONER FULL_ALL
! ENERGY_GAP 0.002
! &END OT
&OUTER_SCF
EPS_SCF 1e-7
MAX_SCF 30
&END OUTER_SCF
&END SCF
&XC
&XC_FUNCTIONAL PBE
&END XC_FUNCTIONAL
&END XC
&PRINT
&PDOS
&EACH
GEO_OPT 1
&END EACH
APPEND TRUE
&END PDOS
&SCCS ON
&EACH
GEO_OPT 10
&END EACH
&DENSITY_GRADIENT off
&EACH
GEO_OPT 10
&END EACH
APPEND
&END DENSITY_GRADIENT
&DIELECTRIC_FUNCTION off
&EACH
GEO_OPT 10
&END EACH
APPEND
&END DIELECTRIC_FUNCTION
&POLARISATION_POTENTIAL off
&EACH
GEO_OPT 10
&END EACH
APPEND
&END POLARISATION_POTENTIAL
&END SCCS
&END PRINT
&END DFT
&SUBSYS
&CELL
ABC [angstrom] 30 30 30
! PERIODIC XYZ
! MULTIPLE_UNIT_CELL 1 1 1
&END CELL
&TOPOLOGY
COORD_FILE_NAME Al4C.xyz
COORD_FILE_FORMAT XYZ
! MULTIPLE_UNIT_CELL 1 1 1
&END
&KIND C
ELEMENT C
BASIS_SET DZVP-MOLOPT-GTH
POTENTIAL GTH-PBE-q4
&END KIND
&KIND Al
ELEMENT Al
BASIS_SET DZVP-MOLOPT-SR-GTH
POTENTIAL GTH-PBE-q3
&END KIND
&END SUBSYS
&END FORCE_EVAL
&MOTION
&GEO_OPT
TYPE MINIMIZATION
OPTIMIZER BFGS
MAX_ITER 100
&END GEO_OPT
&CONSTRAINT
&FIXED_ATOMS
COMPONENTS_TO_FIX XYZ
LIST 1
&END FIXED_ATOMS
&END CONSTRAINT
&PRINT
&STRUCTURE_DATA
DISTANCE 1 2
DISTANCE 1 3
DISTANCE 1 4
DISTANCE 1 5
DISTANCE 2 3
DISTANCE 2 4
DISTANCE 2 5
DISTANCE 3 4
DISTANCE 3 5
DISTANCE 4 5
ANGLE 1 2 3
ANGLE 1 2 4
ANGLE 1 2 5
ANGLE 1 3 2
ANGLE 1 3 4
ANGLE 1 3 5
ANGLE 1 4 2
ANGLE 1 4 3
ANGLE 1 4 5
ANGLE 1 5 2
ANGLE 1 5 3
ANGLE 1 5 4
&EACH
GEO_OPT 1
&END EACH
&END STRUCTURE_DATA
&TRAJECTORY
&EACH
GEO_OPT 1
&END EACH
&END TRAJECTORY
&VELOCITIES
&EACH
GEO_OPT 1
&END EACH
&END VELOCITIES
&FORCES
&EACH
GEO_OPT 1
&END EACH
&END FORCES
&CELL
&EACH
GEO_OPT 1
&END EACH
&END CELL
&RESTART
&EACH
GEO_OPT 1
&END EACH
&END RESTART
&END PRINT
&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 c... at googlegroups.com<mailto:c... at googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/b0c10452-d528-48da-9f83-07d9b01d5dfb%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/b0c10452-d528-48da-9f83-07d9b01d5dfb%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 c... at googlegroups.com<javascript:>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/cc63d43c-380a-4575-ac41-5b981251d294%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/cc63d43c-380a-4575-ac41-5b981251d294%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<mailto:cp... at googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/99722c46-cf0e-427f-a7f5-bd6eba017cd3%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/99722c46-cf0e-427f-a7f5-bd6eba017cd3%40googlegroups.com?utm_medium=email&utm_source=footer>.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200518/436e37c2/attachment.htm>
More information about the CP2K-user
mailing list