[gmx-users] Shell (Drude) model for polarization in GROMACS

[Eric Smoll]

Hello GROMACS users,

I am looking over the shell (Drude) model for polarization in GROMACS.
There isn't much information available in the manual (probably because this
feature is rarely used).  I was hoping someone knowledgeable about
polarizable simulations in GROMACS could help answer a few of my
questions:

(1) How exactly are Thole functions turned on? The 2018.1 manual does not
specify. I am guessing they are hard-coded into all shell molecular
dynamics simulations. Am I correct?

(2) To add a shell to a topology, I suspect I must specify a shell atomtype
(setting the particle type to "S" for shell) and list the shells in the
atoms directives.  Using the "[ atomtypes ]" format in oplsaa.ff for a
molecule with a res-name of "ABC", I assume the following will produce two
shell particles with a charge of +1 and -1. Am I correct?

[ atomtypes ]
; atom-type bond-type atomic-number mass charge particle-type sigma epsilon
opls_9999 Sh 1 0 0 S 0 0
; I am guessing that

[ atoms ]
; atom-number atom-type res-number res-name atom-name charge-grp charge
1 opls_9999 1 ABC Sh1 1 +1
2 opls_9999 1 ABC Sh2 2 -1
....

(3) I suppose connectivity of each shell to its core is indicated with the
"[ polarizability ]" directive. I am guessing "i" is the core atom and "j"
is the shell particle.

(4) Since the code carries out an SCF calculation to relax the position of
the shells at every timestep, I assume there is no need to specify shell
masses or to thermostat the shell DOF.  Does GROMACS omit shell particles
from thermostats?  The manual does not specify.

Thanks for the help!

Best,
Eric