[gmx-developers] Storing the system dipole for predicting infrared spectra

[Berk Hess]

On 2020-02-14 14:43 , David van der Spoel wrote:
> Den 2020-02-14 kl. 14:31, skrev Berk Hess:
>> Hi,
>>
>> In general one can not uniquely define a dipole in a periodic system. 
>> One can only define it when all molecules have no net charge. grompp 
>> warns about this when using the surface correction for PME.
>
> That is correct, but it does not answer my question. This is of course 
> relevant to liquids mainly.
>
> It could also be used for storing the electric current in a simulation 
> without saving the velocities at every time step (and this also works 
> for charged systems).
Then we would need to add an mdp option to compute and output the system 
dipole.
Note that this can be done now by setting a very large but not infinite 
surface epsilon, but that's of course not elegant.
>
>>
>> Cheers,
>>
>> Berk
>>
>> On 2020-02-14 14:22 , David van der Spoel wrote:
>>> Hi,
>>>
>>> it would be useful to be able to store the system dipole for whole 
>>> molecules from an MD simulation for computing infrared spectra. It 
>>> seems that the dipole that is calculated when using the PME surface 
>>> correction is based on broken molecules (but that may be correct for 
>>> that purpose).
>>>
>>> I have implemented an "applied force" that acts to store the (still 
>>> incorrect) dipole and then writes it to a file at the end of the run.
>>>
>>> However, I wonder whether it would be better to have the option to 
>>> store the dipole in the energy file again (with a warning that it 
>>> will slow down the simulation)?
>>>
>>> Any clues?
>>
>
>