Free-energy on virtual sites only (or vice versa)

User discussions
, ,
1

GROMACS version: 2025.3
GROMACS modification: No

Hi,

I am trying to perform a free energy calculation to obtain the zirconia/water interfacial free energy, following the dry wall approach by Leroy and Müller-Plathe (https://pubs.acs.org/doi/10.1021/acs.langmuir.5b01394).

Since the method requires that only attractive interactions are switched off during the free energy calculation, I figured out I could use virtual sites for attractive interactions and real atoms for repulsive ones:

[ atomtypes ]
; name  at.num   mass     charge  ptype    V(c6)        W(c12)
; Zirconium atoms (real)
ZR      40       91.22     0.0    A        0.0          1.631502e-06
; Oxygen atoms (real)
ZO	    8        15.9994   0.0    A        0.0          5.666209e-07
; Zirconium atoms (virtual)
ZRv     40       91.22     0.7952 V        2.7984e-05   0.0
; Oxygen atoms (virtual)
ZOv	    8        15.9994  -0.3976 V        0.00163      0.0

The other way around should also be ok (real->attractive, virtual->repulsive). It is also worth mentioning that I am freezing all zirconia atoms.

According to the documentation I should be able to use virtual sites in free-energy calculation:

On top of an atom

  • This allows giving an atom multiple atom types and with that also assigned multiple, different bonded interactions. This can especially be of use in free-energy calculations.

However, I haven’t found any .mdp option that allows for the λ-coupling of only virtual sites (or vice versa: anything but virtual sites). Is this actually possible?

EDIT: I am also trying transforming “normal” atoms into “purely repulsive” atoms (i.e. via an alchemical tranformation), maybe it’s a easier way. It would be interesting to see if the virtual sites approach could work too.

2

There is nothing special about virtual site atoms, except for that their coordinates are generated and forces on them redistributed. You lambda-couple (only) virtual sites by (only) adding B-state parameters for them.

3

Ah, I see. I think I misunderstood how B-state topologies work. Then virtual sites are not really needed, the alternative approach should work fine.