[CP2K-user] Convergence problem- MD calculation for metal surface
Matt W
mattwa... at gmail.com
Mon May 13 10:07:01 UTC 2019
Hard identify the real problem.
You system runs sort of 'OK' with OT, i.e. it converges electronically, but
the MD explodes after a few steps.
I guess the initial guess it too poor for mixing to cope. It converges if
you start from the converged OT WFN. There seem to be further problems with
mixing and DFTB when moving atoms, doubt it has ever been tested.
I no very little about DFTB setups, so have no idea if the explosion is due
to bad parameters etc.
Matt
On Sunday, May 12, 2019 at 12:41:59 AM UTC+1, Kevin wrote:
>
> Dear Matt,
>
> Enclosed please find the necessary files for testing. Thank you in advance.
>
> Kind regards,
>
> Kevin
>
> On Friday, May 10, 2019 at 8:37:55 PM UTC+2, Matt W wrote:
>>
>> If you send me all files required to run (on here or to my email), I can
>> try and have a look.
>>
>> On Friday, May 10, 2019 at 3:02:59 PM UTC+1, Kevin wrote:
>>>
>>> Dear Matt,
>>>
>>> I would highly appreciate that if you let me know about any updates.
>>>
>>> Best,
>>>
>>> Kevin
>>>
>>> On Tuesday, April 2, 2019 at 12:49:53 AM UTC+2, Kevin wrote:
>>>>
>>>> Dear Matt,
>>>>
>>>> Thanks for the feedback. Actually I have used the same super-cell size
>>>> of the optimized slab and also whenever I use AIMD it works (but it's
>>>> expensive!). So I think the structure is OK.
>>>> I even tried once with OT method and again didn't get the mentioned
>>>> error. I have encountered the error while using DFTB or PM6.
>>>>
>>>> On Monday, April 1, 2019 at 8:04:28 PM UTC+2, Matt W wrote:
>>>>>
>>>>> I've not looked at your files, but are you sure you don't have atoms
>>>>> on top of each other, maybe across periodic images?
>>>>>
>>>>> On Sunday, March 31, 2019 at 11:27:50 PM UTC+1, Kevin wrote:
>>>>>>
>>>>>>
>>>>>> Dear all,
>>>>>> I'm encountering convergence problem for the MD simulation of single
>>>>>> molecule adsorption on Ni surface. This is a test case to see if I can
>>>>>> calculate my desired system. I'm using a GLE thermostat.
>>>>>> I've tried using DFTB and PM6 but in both cases I receive the error
>>>>>> below, mentioning:
>>>>>> *"Bad condition number R_COND = 0.000E+00 (smaller than the machine
>>>>>> working precision" common/mathlib.F:709.*
>>>>>> As you know, DFTB or PM6 are chosen because of their inexpensiveness!
>>>>>>
>>>>>> [image: Capture.PNG]
>>>>>>
>>>>>>
>>>>>> Here is my input file and I've attached the coordinate file:
>>>>>>
>>>>>>
>>>>>> &GLOBAL
>>>>>> PROJECT CH4-Ni
>>>>>> RUN_TYPE MD
>>>>>> PRINT_LEVEL LOW
>>>>>> &END GLOBAL
>>>>>>
>>>>>> &FORCE_EVAL
>>>>>> METHOD Quickstep
>>>>>> &DFT
>>>>>> &QS
>>>>>> METHOD PM6
>>>>>> &SE
>>>>>> RC_INTERACTION [angstrom] 8.0
>>>>>> RC_COULOMB [angstrom] 8.0
>>>>>> RC_RANGE [angstrom] 0.05
>>>>>> &END
>>>>>> &END QS
>>>>>> ! &QS
>>>>>> ! METHOD DFTB
>>>>>> ! &DFTB
>>>>>> ! SELF_CONSISTENT T
>>>>>> ! DO_EWALD T
>>>>>> ! DISPERSION T
>>>>>> ! &PARAMETER
>>>>>> ! PARAM_FILE_PATH /data/scc
>>>>>> ! PARAM_FILE_NAME scc_parameter
>>>>>> ! UFF_FORCE_FIELD uff_table
>>>>>> ! &END PARAMETER
>>>>>> ! &END DFTB
>>>>>> ! &END QS
>>>>>> &SCF
>>>>>> MAX_SCF 30
>>>>>> EPS_SCF 5.0E-6
>>>>>> &OUTER_SCF
>>>>>> MAX_SCF 30
>>>>>> EPS_SCF 5.0E-6
>>>>>> &END
>>>>>> &DIAGONALIZATION ON
>>>>>> ALGORITHM STANDARD
>>>>>> &END DIAGONALIZATION
>>>>>> ADDED_MOS 300
>>>>>> &SMEAR ON
>>>>>> METHOD FERMI_DIRAC
>>>>>> ELECTRONIC_TEMPERATURE [K] 350
>>>>>> &END SMEAR
>>>>>> &MIXING
>>>>>> METHOD BROYDEN_MIXING
>>>>>> BETA 1.5
>>>>>> ALPHA 0.1
>>>>>> NBROYDEN 8
>>>>>> &END MIXING
>>>>>> &PRINT
>>>>>> &RESTART LOW
>>>>>> BACKUP_COPIES 1
>>>>>> &END RESTART
>>>>>> &END PRINT
>>>>>> SCF_GUESS ATOMIC
>>>>>> &END SCF
>>>>>> &POISSON
>>>>>> POISSON_SOLVER MT
>>>>>> PERIODIC XY
>>>>>> &EWALD
>>>>>> EWALD_TYPE SPME
>>>>>> GMAX 25
>>>>>> O_SPLINE 5
>>>>>> &END EWALD
>>>>>> &END POISSON
>>>>>> &END DFT
>>>>>>
>>>>>> &SUBSYS
>>>>>> &CELL
>>>>>> ABC 12.4597 12.9485 23.1387
>>>>>> PERIODIC XY
>>>>>> &END CELL
>>>>>> &COORD
>>>>>> @INCLUDE 'ch4-ni.coord'
>>>>>> &END COORD
>>>>>> &TOPOLOGY
>>>>>> &CENTER_COORDINATES
>>>>>> &END CENTER_COORDINATES
>>>>>> &END TOPOLOGY
>>>>>> &END SUBSYS
>>>>>> &END FORCE_EVAL
>>>>>>
>>>>>> &MOTION
>>>>>> &MD
>>>>>> ENSEMBLE NVT
>>>>>> STEPS 10000
>>>>>> TIMESTEP 0.5
>>>>>> TEMPERATURE 300
>>>>>> &THERMOSTAT
>>>>>> REGION MASSIVE
>>>>>> TYPE GLE
>>>>>> &GLE
>>>>>> NDIM 3
>>>>>> A_SCALE [ps^-1] 1.00
>>>>>> A_LIST 2.000000000000e+0
>>>>>> 1.365763041717e+1 0.000000000000e+0
>>>>>> A_LIST -1.365763041717e+1
>>>>>> 9.418257899376e-1 5.860040064992e+2
>>>>>> A_LIST 0.000000000000e+0
>>>>>> -5.860040064992e+2 0.000000000000e+0
>>>>>> C_LIST 3.071856358605e+2
>>>>>> 1.052252204958e+0 -3.083120039764e+2
>>>>>> C_LIST 1.052252204958e+0
>>>>>> 1.353606645969e+4 0.000000000000e+0
>>>>>> C_LIST -3.083120039764e+2
>>>>>> 0.000000000000e+0 1.352888082383e+4
>>>>>> &END GLE
>>>>>> &END THERMOSTAT
>>>>>> &PRINT
>>>>>> &ENERGY
>>>>>> &EACH
>>>>>> MD 10
>>>>>> &END EACH
>>>>>> &END ENERGY
>>>>>> &END PRINT
>>>>>> &END MD
>>>>>> &FREE_ENERGY
>>>>>> &FREE_ENERGY_INFO
>>>>>> &EACH
>>>>>> MD 1000
>>>>>> &END
>>>>>> &END
>>>>>> &METADYN
>>>>>> &PRINT
>>>>>> &COLVAR
>>>>>> COMMON_ITERATION_LEVELS 3
>>>>>> &EACH
>>>>>> MD 1
>>>>>> &END
>>>>>> &END
>>>>>> &HILLS
>>>>>> COMMON_ITERATION_LEVELS 3
>>>>>> &EACH
>>>>>> MD 1
>>>>>> &END
>>>>>> &END
>>>>>> &END
>>>>>> &END METADYN
>>>>>> &END FREE_ENERGY
>>>>>> &PRINT
>>>>>> &TRAJECTORY
>>>>>> FORMAT XYZ
>>>>>> &EACH
>>>>>> MD 20
>>>>>> &END EACH
>>>>>> &END TRAJECTORY
>>>>>> &RESTART_HISTORY
>>>>>> &EACH
>>>>>> MD 10000
>>>>>> &END EACH
>>>>>> &END RESTART_HISTORY
>>>>>> &RESTART OFF
>>>>>> &END RESTART
>>>>>> &END PRINT
>>>>>> &END MOTION
>>>>>>
>>>>>> I would be grateful if anyone could take a look to my input file and
>>>>>> let me know about the possible way to get the structure converged. I've
>>>>>> already relaxed the metal structure before using it for md simulation.
>>>>>>
>>>>>> Thanks in advance.
>>>>>>
>>>>>> Kevin
>>>>>>
>>>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190513/67bf6a9f/attachment.htm>
More information about the CP2K-user
mailing list