[gmx-developers] turning charges on/off

[Mark Abraham]

Jason de Joannis wrote:
> Quoting Mark Abraham <Mark.Abraham at anu.edu.au>:

I'm keeping the discussion on-list.

>> Jason de Joannis wrote:
>>
>>> I wonder if it is possible to turn charges on and off during the 
>>> course of a simulation. Will PME still work? (I want to do monte 
>>> carlo mutations between different molecular species)
>>
>>
>> So long as your system is well-defined at each time step, why wouldn't 
>> PME work? That algorithm isn't dependent on previous steps.
> 
> Well that is good news. I was thinking that there might be some kind of 
> pme array initializations etc that depend on number of charges.

:-) The algorithm doesn't, but the implementation might have something 
like the problem you suggest. I was thinking of higher-order problems!

>> Anyway, this sort of thing isn't implemented in gromacs, and I imagine 
>> would be a large amount of work to implement. Good luck!
>>
>> Mark
>>
> It shouldn't be that much work (hopefully). Just change the q-value and 
> hopefully Gromacs picks up on that... I will look into it some more today.

If it's just a change, then that might be painless, but if you're 
changing the number of charges, as you imply above, then things might be 
messy. Better would be to define all the sites originally, and zero the 
charges for whatever isn't needed at a given step.

Mark