[gmx-users] Van der Waals: switch attractive term off

[Mark Abraham]

> Hi Mark,
>
> On Jul 1, 2006, at 3:24 AM, Mark Abraham wrote:
>
>>> It seems that you can modify the following line in the function
>>> *mk_nbfp() in force.c and achieve what you want.
>>>
>>> 	C6(nbfp,atnr,i,j)   = idef->iparams[k].lj.c6;      (it is in line 117
>>> of force.c in version 3.3)
>>>
>>> and change it to
>>>
>>>          C6(nbfp,atnr,i,j)   =  0;
>>>
>>> This will set the attractive interaction between atom i & j to 0.
>>
>> More correctly, it will set the coefficient of the r^6 term to zero.
>
> That's the goal.

Good - obviously it's actually the r^(-6) term :-) There's a semantic
difference between setting the attractive interaction to zero (i.e. zero
coefficient for the additive term with a negative sign), and setting the
LJ force to zero in the attractive region (force normal before the LJ
potential minimum, zero thereafter). These two require quite different
mathematical treatments, hence the concern raised by Berk.

>> This
>> alters the shape of the function for *all* r, both in the "attractive"
>> and
>> "repulsive" regions,
>
> I could not get that. How could the setting C6(nbfp,atnr,i.j) = 0 can
> affect the repulsive interaction term, keeping in mind that even if
> you  are using a ff that uses 'sigma' or 'epsilon'  to construct C6 and
> C12 and explicit the minimum of the potential, you are not affecting
> them?

Per equation 4.3 in the gromacs manual, V_LJ(r_ij) = C12_ij * r_ij^(-12) -
C6_ij * r_ij^(-6). Now for any C12_ij > 0 and C6_ij > 0 this potential has
a local minimum and asymptotic behaviour as r_ij -> 0 and r_ij -> infinity
(Figure 4.1). You can speak of the "repulsive" and "attractive" regions to
either side of this minimum because the sign of the derivative (i.e. the
force) changes here. The repulsive region is dominated by the C12 term
because r_ij is small and the r^(-12) term is larger, and the attractive
region is dominated by the C6 term likewise. However both contribute to
both regions, so zeroing one affects both regions, which might not be what
was wanted by someone who was not describing precisely what they wanted
:-)

It doesn't matter how C6 or C12 is being constructed if you're zeroing it
afterwards and using this formula, too.

>> such that the potential function no longer has a
>> local minimum.
>
> That's true and that's possibly be the concern of the user who wanted
> only to switch-off  the attractive (c6) vdw without asking for its
> consequences.

Yes - they're not getting zero force in the formerly-attractive region,
they're getting a small repulsive force.

Mark