The simplest approach is to calculate the orientation of the dipolar moment for a set of molecules in your system:

http://manual.gromacs.org/programs/gmx-dipoles.html

This is a gross approaximation, but is you molecule set happens to be large, the statistical bias in dipolar orientation gives you information about the local polarizing forces at defined sites within your system.

But in MD you have limited information about true polarization, unless you use a polarizable force field, such as AMOEBA. As far as I know AMOEBA is yet to be implemented in GROMACS, as this requires rewriting the numerical integration code.

You may approximate the polarization effects using the "Drude oscillator", and this has been implemented in GROMACS. This basically consists of adding to each atom two dummy particles with opposing charges of the same magnitude, so its total charge is null. The particles are linked by a bond of the harmonic kind. You may look at this reference:

Article Implementation of Extended Lagrangian Dynamics in GROMACS fo...

You may read this portal for reference:

https://www.uni-muenster.de/Chemie.pc/heuer/simulationmethodology/polarizationmethods.html

Best wishes

Rogelio

Dear Rogelio,

Thank you so much for your detailed and useful answer.

That's just a water molecule and i used AMBER99 force field. So do you think gmx dipoles can do it efficiently?

Best regards

Roja

If you mean ONE water molecule in a box of other molecules it will be too inaccurate.

If you mean the system is a large number of pure water molecules, the answer is probably yes.

Finally, when you say you used AMBER99, it probably means you used the TIP3P, TIP4P or TIP5P as water model.

Be careful! - each one of this models has a different number of point charges to describe the water molecule and their behaviour is not going to be the same.

Best wishes,

Rogelio