Dear colleagues,

Let us consider the system consisting of two parts complying one to

ReaxFF ('ReaxFF subsystem') and and the other to some non-reactive force

field ('non-ReaxFF subsystem'). We achieve this, say, with command

pair_style hybrid reax/c controlfile lj/cut/coul/long 10.0

Atomic charges in ReaxFF subsystem must be equilibrated with fix

qeq/reax, while charges in non-ReaxFF subsystem are immutable. 'The QEq

method minimizes the electrostatic energy of the system by adjusting

the partial charge on individual atoms based on interactions with their

neighbors.' (from http://lammps.sandia.gov/doc/fix_qeq_reax.html)

The question is:

does qeq/reax algorithm takes into account static charges of

neighbor atoms from non-ReaxFF subsystem during minimization of

electrostatic energy via adjusting atomic charges within ReaxFF

subsystem?

Best regards,

Vadim

Dear colleagues,

Let us consider the system consisting of two parts complying one to

ReaxFF ('ReaxFF subsystem') and and the other to some non-reactive force

field ('non-ReaxFF subsystem'). We achieve this, say, with command

pair_style hybrid reax/c controlfile lj/cut/coul/long 10.0

Atomic charges in ReaxFF subsystem must be equilibrated with fix

qeq/reax, while charges in non-ReaxFF subsystem are immutable. 'The QEq

method minimizes the electrostatic energy of the system by adjusting

the partial charge on individual atoms based on interactions with their

neighbors.' (from http://lammps.sandia.gov/doc/fix_qeq_reax.html)

The question is:

does qeq/reax algorithm takes into account static charges of

neighbor atoms from non-ReaxFF subsystem during minimization of

electrostatic energy via adjusting atomic charges within ReaxFF

subsystem?

you are worrying about the wrong problem. this is a bogus model. using

pair style hybrid causes more problems beyond charge equilibration.

1) how are you going to handle the "mixed" interactions between the reaxff

part and the class 1 force field part?

you cannot use the lj/cut lennard-jones parameters, since they are not

consistent with the reaxff charges.

2) also, how are you going to handle the long-range coulomb? kspace

interactions are not computed for pairs of atoms, but considers the

(long-range) coulomb interactions of each atom with *all* atoms in the

system and their (infinite) periodic replica. reaxff instead uses a

wolf-sum approximation to long-range coulomb, if i remember correctly.

to have a more consistent treatment of the interactions, you have to

follow an approach similar to what is done in QM/MM calculations. at the

very least you should implement an ONIOM-style mechanical coupling:

- you do a simulation of the *entire* system with lj/cut/coul/long and

kspace

- you do a simulation of the reaxff subsystem without the embedding

environment

- you do a simulation of the exact same subsystem but with lj/cut/coul/cut

now your total hamiltonian would be: U_total = U_all - U_lj/cut/coul/cut +

U_reaxff

this will still have some inconsistencies (e.g. the embedding environment

cannot polarize the reaxff subsystem - hence the term "mechanical

coupling").

if you want to handle than, you'd have to add an electrostatic embedding,

similar to QM/MM couplings, and even that is not fully consistent, since

your static system is non-polarizable, while the reaxff subsystem is, and

thus there is an non-physical unbalanced energy transfer during MD.

axel.