[Debichem-devel] Bug#706608: cp2k: wrong amber torsion energy

[Michael Banck]

Package: cp2k
Version: 2.2.426-1
Severity: normal
Tags: upstream,patch

A bug has been found in the AMBER force field topology input regarding
torsion energies.  It has been fixed in upstream revision 12867.


Michael

On Thu, May 02, 2013 at 11:18:03AM +0200, Teodoro Laino wrote:
> Thanks Letif and apologies for the delay.
>
> The issue was indeed a bug and the fix is in the SVN. Should you
> encounter other discrepancies, please, do not hesitate to report them.
> best,
>
> Teo
> 
> On Apr 22, 2013, at 8:54 PM, Letif Mones <letif.mones at gmail.com> wrote:
> 
> > Hi Teo,
> > 
> > I am sending the following files: amber and cp2k inputs, amber
> > restart file and amber topology file (dmpoh_vacuum.atop). I also
> > attached another topology file (dmpoh_vacuum_mod.atop) in which I
> > halved the dihedral force constants. With this latter cp2k gives the
> > right answer.
> > 
> > Regards,
> > 
> > Letif
> > 
> > 
> > 
> > 
> > On 22 Apr 2013, at 17:53, Teodoro Laino wrote:
> > 
> >> Dear Letif,
> >> 
> >> thanks for reporting this issue. All parameters should be converted
> >> at the moment they are read. We also had few examples to check the
> >> numerics and I remember they were matching with Sander they were
> >> OK.  Do you have an input file (including all relevant files) that
> >> shows this problem?
> >> 
> >> Thanks,
> >> Teo
> >> 
> >> On Apr 22, 2013, at 5:23 PM, Letif Mones <letif.mones at gmail.com> wrote:
> >> 
> >>> Hi,
> >>> 
> >>> I did a molecular mechanical calculation in cp2k using an amber
> >>> topology file (force field: amber99). When I checked the bonded
> >>> energy terms I found that the bond and angle terms are correct but
> >>> the dihedral energy is twice as the energy calculated by sander. I
> >>> think the reason is that in the amber FF the form of the torsion
> >>> potential is: K / 2 * [ 1 + cos[ M*PHI - PHI0 ] ] while cp2k uses
> >>> the form of  K * [ 1 + cos[ M*PHI - PHI0 ] ].  So should not the
> >>> torsion force constant read from an amber topology file be
> >>> multiplied by 0.5 (and the improper one too as it has the same
> >>> form)?
> >>> 
> >>> Letif