[gmx-users] DPPC lipid model in Gromos 53a5

[Mark Abraham]

> Mark Abraham wrote:
>>> Dear all,
>>>
>>> In the definition of the DPPC lipid molecule in ffG53a5.rtp I found,
>>> that
>>> some
>>> dihedral angles are defined twice:
>>>
>>>   [ dihedrals ]
>>> ;  ai    aj    ak    al   gromos type
>>>    C33     N   C32   C31     gd_29
>>>      N   C32   C31   O32     gd_4
>>>      N   C32   C31   O32     gd_36   <-
>>>    C32   C31   O32     P     gd_29
>>>    C31   O32     P   O31     gd_20
>>>    C31   O32     P   O31     gd_27   <-
>>>    O32     P   O31    C3     gd_20
>>>    O32     P   O31    C3     gd_27   <-
>>>
>>> Is this a mistake or on purpose ?
>>>
>>> Does the second entry overwrite the first entry, i.e. the values of the
>>> second
>>> entry are used ?
>>
>> It depends on the functional form - you'll need to look up these gd_??
>> macros in the forcefield .itp file. A second function with the same
>> function type will over-write the first. I think it does so silently. A
>> different function type on the same atoms is accepted silently.
>>
>
> I'm not sure this is correct: check your top file. If you have two
> dihedrals in the top file both are evaluated and the effective
> interaction is the sum of these. This is a way to fake a more
> complicated dihedral term.

Ah, yes, David is right (as usual). I tested this, and found that a
gmxdump of a grompp of a topology that included a second bonded
interaction of the same type on the same atoms reveals that both functions
are used.

Please accept my apologies for the mistake - I believe I extrapolated
incorrectly from the fact that multiple [ dihedraltypes ] entries with the
same atom types and function type do not add. In the original example (and
my test case) the functions are explicitly defined, and thus the [
dihedraltypes ] lookup is not used.

To return to the issue in the original post - one could well wish to have
periodic dihedrals with different periodicities to add to give a more
complex functional form than is possible with a single periodic dihedral.
More perverse applications could be imagined :-)

Mark