[EXTERNAL] requesting permission to augment pair_coeff command

Since the IJ coeffs = JI coeffs, we want to avoid the user
thinking that he can set them differently.
Or accidentally override an earlier setting with a new one.

It's not hard to modify my preprocessor to check for I<=J and swap the
two tokens if necessary.

From a user perspective, the only time I get a little temporarily

confused is at times like this:
pair_coeff 3 * 0.15 3.6
However everything is well documented.

If you set the cutoff at 2^(1/6) that is repulsive.

   I suppose shifted-truncated lennard jones potentials are pretty
common. But for whatever reason, both 1/r^12 and 1/r^6 repulsive
potentials have become established in the protein-folding community.
Not that one is better than the other but it's easier to submit a
paper borrowing one of the well-characterized models, like the "BLN
model", for example (Proc. Natl. Acad. Sci. Vol 87, p. 3526, 1990),
which use an ad-hoc combination of +1/r^12 and +1/r^6 repulsion.
People in our lab use tabular pair potentials to accomplish this.
This is a bit slower, but we can do that. Things start to get messy
when the number of particle types gets large.

   One area that's exciting to us, is the simulation of bigger things.
Eventually, things like cytoskeletal components or small organnelles.
This would require a large number of different particle types.
Whether we use shifted-truncated Lennard-Jones, or 1/r^12 repulsion,
or something else, we'd like to be able to determine the repulsive
parameters using automatic mixing rules. Currently, we specify a new
table for every pair of particle types. That's not scalable.

Are you saying you want some more automated way to
get the exact values for the cutoffs?

   I confess I was not aware how popular shifted-truncated Lennard
Jones interactions are in the polymer community. It's true that the
current mixing rules won't work well when we implement repulsion this
way. However I was proposing to represent pure repusion using a
positive or zero 1/r^6 term instead of shifted truncation.

Also not clear on how the mixing
is any different than the normal mixing rules.

   The rules I was going to propose are not different, aside from the
sign of the 1/r^6 term. The only thing I do is keep track of the sign
separately. (It's nothing clever. I'll explain more when I submit
the new pair style. Unfortunately the ordinary Lorentz/Berthelot
mixing rules effectively cause the Aij and Bij [Lennard-Jones AB
parameters] explode whenever either Bi or Bj approaches zero or
becomes positive.)

Alternately I can submit a new pair style which has these changes
already. You can take a look at it, and we can decide where we
want to go.

   In retrospect I think this is what I should do. There is no
standard "mixing rule" for repulsive interactions, as much as I might
like to impose my own rule.

   For our purposes, all we really need is one or two pair styles that
support our own proposal for repulsive mixing rules, which we can
implement ourselves. If they are seam reasonable enough to you, then
perhaps you'd consider including theses pair styles in the
distribution. (I was worried you would not consider this, so in my
last email I proposed editing the existing pair styles in a way which
would not alter their normal behavior. This was a hack using negative
sigma. I don't know why I thought this would be more acceptable to
you. I think I need more sleep.)



You can submit new pair styles if you like.
Send the formulas/doc page ahead of time
if you want me to take a look. I didn't follow
the details of your email.