I studied on BN simulations in the literature very much and saw no charge

and coulombic interactions that is why I exclude the charges but if you

consider it as a possible deficiency of the model, maybe I should go over

and check if I missed something about that.Besides charge considerations, one thing I cannot understand clearly is

which of the following statements you are exactly saying:a)Using LJ parameters in addition to tersoff and SPC/E is mistake

i am not saying that.

b)Using combining rules for 0-B and 0-N is a mistake

yes, as well as for H-B/H-N.

c)Both of a and b.

to really explain the reason, i would have to give you a lecture on

force field parameterization strategies and methods. this is text book

stuff and this is not the place and neither do i have the time.

Also I want to add that I tried LJ interaction just for cross relations but

not for B-B and N-N interactions because I got inconsistent results for pure

solid BN model with LJ; however, tersoff did well.

i still can't make sense of that. lets see what you have:

- a LJ based model to do BN (but it also likely has bonded interactions)

- a tersoff based model to do BN

- a LJ + charge based model to do water (SPC/E)

what you do *not* have is a model describing the interaction of water

(the SPC/E version, that is) with BN. a) and c) were parameterized

independently and

what you also have not provided is an explanation of how you determine

when your simulation is giving you the right and when the wrong

answer.

As you say parameterization for BN with SPC/E is exaclty what I am looking

for.

but you don't have it and applying mixing rules to two independent

models is not the way to obtain them. besides, why using an inferior

method, when you have access to the real thing: i just did a quick

search on google for molecular dynamics simulations on boron nitride

and water and my first two hits show parameterizations that *do*

include significant partial charges on B and N and, in fact, the

presence of water - as my chemical intuition tells me it should -

enhances those partial charges. the same kind of polarization happens

with CNTs, however, there you can often ignore it if you keep the CNT

structure rigid and include an averaged version of the polarization in

the LJ parameters. since BN is polar, this cannot be done as easily.

now, mixing tersoff with lj/cut/coul/long for this kind of system

consistently is not as simple using pair_style hybrid because of the

long-range coulomb interaction. what you would need is a so-called

ONIOM model approach:

B: compute the BN system only with lj/cut/coul/long + kspace

C: compute the BN system with tersoff w/o kspace

the total forces and energies are then computed as: A - B + C

axel.

p.s.: to pre-empt your next question. the doi for the two papers i

found are: 10.1049/mnl.2009.011 10.1021/jp076747u