[CP2K-user] INCORRECT TORSION ENERGY WHEN USING AMBER PARAMETERS

Ajay Muralidharan ajay... at gmail.com
Tue Jan 28 01:18:50 UTC 2020


I have made an ad-hoc fix by adjusting the amber.prmtop for my case. 
But for the sake of the unsuspecting researcher, TORSIONS still has to be 
fixed (for AMBER style forcefields) in the main version of the CP2K code.



On Thursday, January 23, 2020 at 5:01:35 PM UTC-6, Ajay Muralidharan wrote:
>
>  
> Hello all,
> I noticed that torsions evaluated by CP2K when using AMBER parameters are 
> incorrect (and exactly half of the correct answer). Here is the description 
> below:
>
> I do single point energies of a small dipeptide in vacuum using amber 
> forcefield. (attached input files for CP2K and gromacs)
> Gromacs and OPENMM give the same intramolecular energy components.
> _____________________________________________
> *CP2K RESULTS:* (using amber.prmtop from amber tools):
> more 1.out | grep -A 2 "FIST energy"
> BOND = 17.8009* 4.184 = 74.478 kJ/mol
> ANGLE = 34.4220 * 4.184 = 144.021 kJ/mol
> TORSION = 10.0356 * 4.184 = *41.988 kJ/mol* (Incorrect value)
> _____________________________________________________
>
> *GROMACS RESULTS:* (after converting amber.prmtop -> gromacs.top)
> BOND = 74.478 kJ/mol
> ANGLE = 144.021 kJ/mol
> TOTAL TORSIONS =  80.691(proper) + 3.2858 (improper) = *83.97 kJ/mol* (Correct 
> value)
>  *(Exactly double as that from CP2K!) *
> TOTAL PE = ~251 kJ/mol. (This matches exactly with OPENMM when using 
> amber.prmtop )
>
> After lots of testing, I concluded that cp2k uses a different pre-factor 
> (of K/2) for calculating torsions. Amber parameters were generated for a pre-factor 
> of K (not K/2). 
> ______________________________________________________
>
> Then saw this post describing a similar issue (7 years ago): 
> *https://groups.google.com/forum/#!searchin/cp2k/amber$20torsion%7Csort:date/cp2k/PeMpumD2oXU/1uWBed2IfhkJ 
> <https://groups.google.com/forum/#!searchin/cp2k/amber$20torsion%7Csort:date/cp2k/PeMpumD2oXU/1uWBed2IfhkJ>*
> *(Seems like TORSIONS is still broken in svn:18464)*
>
> Here is my version of cp2k:
> mpirun -n 1 cp2k.popt --version   
> CP2K version 6.1   
> SVN source code revision svn:18464   
> cp2kflags: libint fftw3 libxc parallel mpi3 scalapack xsmm 
> libderiv_max_am1=5 libint_max_am=6 plumed2 mkl  
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200127/0f1646e8/attachment.htm>


More information about the CP2K-user mailing list