Thanks Jason
>> Hello all
>>
>> Is there a way to calculate the electrostatic potential at a given
>> atom position or point in space in cpptraj?
>
>
>cpptraj cannot (yet) do PME electrostatics. There are a couple actions that do electrostatic
>calculations, but they do direct space pair-pair interactions only using minimum image convention.
>
What are the methods to calculate through the direct space sum?
>
>>
>> Or any other analysis program?
>>
>
>None in Amber that I know about.
>
>
>>
>> I can calculate it using my own programs, but I want to use PME to get
>> long range electrostatics that would go beyond where periodic
>> boundary conditions end. This might seem silly, but I have a
>> surfactant system that has a high ionic strength.
>>
>>
>
>
>>
>>
>> I have also tried to use
>>
>> image xoffset 0 yoffset 0 zoffset 0
>>
>> image xoffset 0 yoffset 0 zoffset 1
>>
>> ... etc
>>
>> To make the system more like and 'infinite' one, but this is quite
>> cumbersome.
>>
>
>It will be _very_ hard to make the system large enough that you can compute an electrostatic potential
>using just a direct sum-over-pairs (and it will get _very_ slow very quickly). Your best bet really is
>to use an Ewald-based method. There are a couple options:
I have done some tests and I think I can use the direct space sum if I can get about 20 nm between my structure and the boundary. The potentials are different but the fields are about the same.
>
>1) Write your own PME or Ewald code. The electrostatic potential at any given point is given by the sum
>of the direct space ESP and the reciprocal space ESP. But I believe that the convolution of the charge
>grid with the electrostatic potential is done in reciprocal space and the result is back-transformed to
>real space. So I think you would want to back-transform the reciprocal space pair potential before
>convoluting it with the charge density (if what you're after is the ESP, that is).
>
>2) Use an existing PME or Ewald code to extract what you need. The electric field is easy enough to get
>by dividing the electrostatic force by the charge of that particle. The ESP seems to me to be a little
>bit more challenging. Since I don't think you can pull out the reciprocal space pair potential from any
>of the existing PME/Ewald codes, you would effectively need to find the electrostatic energy of putting
>a unitary point charge at a particular location. So compute the full electrostatic energy (and only the
>electrostatic energy), put a unit charge at the location you want the ESP at, and compute another
>electrostatic energy. I believe the difference is the electrostatic energy of that point charge in the
>field generated by all of the particles in all periodic boxes. Divide by that charge and you get the
>ESP.
>
>For #2, the easiest way I can think of doing this (and the way I would use if I had to do it) would be
>to use the OpenMM capabilities in ParmEd and write a Python script to compute the ESPs. The general
>steps here would be:
>
>1) Use createSystem to create an OpenMM system with all of the relevant forces
>
>2) Create two new, separate System objects with only a single NonbondedForce object. Populate the new
>NonbondedForce objects from the NonbondedForce object in the System you created in step 1, making sure
>to set all of the Lennard-Jones parameters to 0 so the only energies computed come from electrostatics.
>Add a new particle to the second system (and second NonbondedForce) that will be the unitary charge at
>the point in space you want to compute the ESP.
>
>3) Create Context objects for each of these systems and compute their total potential energies.
>
>4) Take the difference between the energies computed in (3) and divide by the charge you assigned to the
>particle you added.
>
>The advantage of (1) is that you'll almost certainly learn a lot more about PME and Ewald. The
>advantage of (2) is that ParmEd provides Python classes for most of Amber's input and output files
>(restarts, trajectories, NetCDF restarts and trajectories, prmtop files, etc.) which should make your
>job easier, and you'll add a powerful tool to your repertoire (and you won't have to program and
>validate your own PME implementation, of course).
>
>This was quite lengthy, but hopefully it was helpful, Jason
>
Thanks for the suggestions, I also wondered if I could somehow use amber's PME code somehow?
Cody
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Aug 16 2014 - 13:30:02 PDT
