[CP2K:1031] Re: input coordinates & charges

Csilla Varnai cv... at cam.ac.uk
Tue May 13 12:31:37 UTC 2008

Dear All,

I am trying to run a program which invokes CP2K at each time step.  
Basically, I will need to run CP2K to calculate the forces during my  
simulations, so I am trying to find the best way to do that.
I have the atomic coordinates and connectivities stored during the  
run, from which I can create CHARMM format to pass it to CP2K.

I have generated a modified forcefield file of the original  
distribution of CHARMM from the www.charmm.org home page to have a 1- 
to-1 mapping between the atomic types and atomic charges. Having  
revised the way of generating the modified forcefield I have a  
forcefield of reasonable size, and the forcefield can be read by CP2K  
fine. I seem to have no problem with the forcefield any longer, thanks  
for the help!

I thought of relying on the internal topology builder of CP2K.
But I am getting the following error message with the default  
BONDLENGTH_MAX = 3 Angstroms:

     GENERATE| ERROR in connectivity generation!
     GENERATE| The THRESHOLD to select possible bonds is larger than  
the max. bondlength
     GENERATE| used to build the neighbors lists. Increase the  
     GENERATE| Present THRESHOLD (   7.233872 ).  Present  
BONDLENGTH_MAX (   5.669178 )

Why does it say 5.669178 instead of 6? Is its unit bohr?
If I increase the BONDLENGTH_MAX, it does not help. If I set 4.5  
Angstroms then I am getting the following:

    GENERATE|  Preliminary Number of Bonds  
generated:                         15757

     GENERATE| WARNING in connectivity generation!
     GENERATE| Two molecules/residues named (HSE) have different   
number of atoms.
     GENERATE| Molecule starting at position (238) has Nr. <4> of atoms.
     GENERATE| while the other same molecules have Nr. <11> of atoms!
     GENERATE| Increasing bondparm_factor by 1.05.. An error was found  
in the generated
     GENERATE| connectivity. Retry...
     GENERATE| Present value of BONDPARM_FACTOR (   1.155000 ).

     GENERATE| ERROR in connectivity generation!
     GENERATE| The THRESHOLD to select possible bonds is bigger than  
the MAX bondlength
     GENERATE| used to build the neighbors lists. Increase the  
     GENERATE| Present THRESHOLD (   8.792816 ).  Present  
BONDLENGTH_MAX (   8.503768 )

Can the hydrogen bonds cause problems in this case?


>> I have redefined the atomic types, and regenerated the CHARMM22
>> potential file according to the new atomic types, at first sight in a
>> very simple-minded way, substituting all the possible new atomic  
>> types
>> (say, CC1, CC2) into the old ones (CC). But in this case, due to the
>> large potential file, running the program is hopeless.
> what program? cp2k? hopeless in which sense?
> it would be much easier to help you, if you'd explain
> in more detail what you are actually trying to achieve
> and what kind of systems you are planning to model.
>> I must confess I have not tried the VMD psfgen, since I would like to
>> invoke CP2K from another program, and that is the reason why I would
>> like to avoid the use of other programs like the VMD psfgen. If I
> whatever you do you _have_ to generate a "topology".
> using a .psf and .par file is a very convenient way.
> if you don't like using psfgen, you could consider
> creating your topology directly. have a look at
> http://www.ks.uiuc.edu/Training/Tutorials/namd/namd-tutorial-unix-html/node21.html
> and the following pages for a minimal description of the format.
> more details are in the parametrization tutorial on the same
> server and - of course - on http://www.charmm.org/
> unless you have to generate a large number of topologies
> on the fly, i would guess that actually using psfgen with
> some smart VMD/tcl scripting will be much faster than
> writing your own tool.
>> the cp2k f77_interface, would it make any difference concerning to
>> this problem? Should I also have a potential file containing atomic
>> types with unambiguously defined charges? If that is the case then I
>> run into the same memory problem. Does the psfgen seem to be the only
>> solution?
> what memory problem?
> cheers,
>    axel.
>> Best regards,
>> Csilla
>> On 7 May 2008, at 17:04, Teodoro Laino wrote:
>>> Dear Csilla,
>>> you got exactly the point.
>>> In CP2K we take the CHARMM convention where a type has associated a
>>> charge.
>>> This is different from AMBER where the same type can have different
>>> charges.
>>> So the best thing to do is to redefine your types according the
>>> convention that a type
>>> has a well defined charge (this most of the times is terribly easy..
>>> just append a letter or a number to
>>> the existing type).
>>> Cheers
>>> Teo
>>> p.s.: even with PDB this won't necessarily work. I'm not TOTALLY
>>> sure but  in case you define two different
>>> charges for the same type, only the first one may be assigned to
>>> that type. The result is that the
>>> charges of your system will be wrong. I will add a warning for
>>> handling this exception, though reading
>>> the charges from the beta column of the PDB is not something really
>>> standard.
>>> On 7 May 2008, at 17:55, Csilla Varnai wrote:
>>>> in case of peptides. Hence, I need to read in the charges. I have
>>>> found only 1 way to do it so far, namely reading it from a PDB  
>>>> file.
>>>> My problem is that in this case I loose the accuracy of the
>>>> coordinates, since this format is fairly restricted and I cannot  
>>>> read
>>>> the coordinates only at a +/-0.001 level.
>>>> Could anyone please help me, if there is a better way I've missed  
>>>> to
>>>> get both the charges and the coordinates at a given accuracy?
> >

More information about the CP2K-user mailing list