warnings with pair style lj/long/tip4p/long

Dear all,

I tried to use the nice water slab input scripts of Mario Orsi
(http://www.orsi.sems.qmul.ac.uk/downloads.html). In particular, I
tried to include the long-range part of the LJ potential, using the
following modified forcefield.TIP4P-2005 file (full input script
enclosed):

orsi_long.tgz (2.24 KB)

Dear all,

I tried to use the nice water slab input scripts of Mario Orsi
(http://www.orsi.sems.qmul.ac.uk/downloads.html). In particular, I
tried to include the long-range part of the LJ potential, using the
following modified forcefield.TIP4P-2005 file (full input script
enclosed):

========
mass 1 1.00794 # H
mass 2 15.9994 # O

pair_style lj/long/tip4p/long long long 2 1 1 1 0.1546 10.0 8.5
kspace_style pppm/disp/tip4p 1.0e-5

pair_coeff 1 1 0.0 0.0
pair_coeff 1 2 0.0 0.0
pair_coeff 2 2 0.18520 3.1589

bond_style harmonic
bond_coeff 1 500.0 0.9572

angle_style harmonic
angle_coeff 1 50.0 104.52

The simulation runs but I receive 2 warnings which I'm not sure to understand:

* WARNING: Mixing forced for lj coefficients
(../pair_lj_long_tip4p_long.cpp:1408)

There are only O and H atoms in the system, and all pair coefficients
(1 1, 1 2, 2 2) are defined in the input script. What is the problem?

to have consistent treatment of long-range LJ interactions, mixed
coefficients will have to be overridden. in reciprocal space, you sum
over the contributions of individual atoms and not pairs, so you have
to enforce how mixed terms are computed since they have to be included
into the per atom loops.

* WARNING: Using largest cutoff for pair_style lj/long/tip4p/long
(../pair_lj_long_tip4p_long.cpp:1410)

long-range processing is done with the same kspace settings for both
LJ and Coulomb, thus there can only be one global cutoff.

I couldn't find in the doc page for pair_style lj/long/tip4p/long a
mention that a single "cutoff" had to be used for both the LJ and
Coulomb interactions. Maybe it should be added somewhere?

as will my previous comment on the same type of question. steve will
be happy to include corrections or improvements to the documentation.
specifically the lj/* pair style documentation has become quite
convoluted over time, so some cleaning up and refactoring would be a
very useful contribution.

axel.

Dear all,

I would like to add some brief comments to this discussion:

1.) Whenever using the pppm/disp, one should definitely
include the

kspace_modify force/disp/real ...
kspace_mdofiy force/disp/kspace ...

commands in the input script. Otherwise the simulations will
either be too inaccurate or too slow. You should check out
the (Isele-Holder) source from the LAMMPS doc page and
my other paper that is referenced therein.

2.) The issue about the warning is somewhat tricky:

In principle, it is possible to use lj coefficients that do not
mix. This is however only possible when using the pppm/disp
and not the ewald/disp. To enable this feature, one has to use the

kspace_modify mix/disp ...

command. The default for the pppm/disp is that
some mixing will be assumed for the parameters.
This is mainly because the modification for unequal
mixing was added much later. Using the enforced
mixing for the dispersion coefficients seemed to be
the best way to avoid inconsistencies between the
pppm/disp and the ewald/disp and also with previous
versions of LAMMPS before this feature was enabled.

In your case, it does not matter anyway, because the
dispersion coefficients (the C6 coefficients in your LJ
potential) follow mixing rules, which results from
that the dispersion coefficient of the hydrogen atom is zero.

I hope this was helpful and not confusing.

Best, Rolf

Thanks Axel and Rolf for your answers. I had seen your 2013 paper, but
now I read it (and the J. Chem. Phys. 2012) in detail. I'm making some
tests on the water slab simulation, with the updated force field
command lines:

pair_style lj/long/tip4p/long long long 2 1 1 1 0.1546 7.0 # 10.0
pair_modify mix arithmetic
kspace_style pppm/disp/tip4p 1.e-4
kspace_modify force/disp/real 0.0002
kspace_modify force/disp/kspace 0.004

I will try to prepare an update for the related doc pages with what I
understood from my reading and tests, but I will welcome your expert
feedback.

Best regards,
Laurent

2015-01-27 11:28 GMT+00:00 Rolf Isele-Holder <[email protected]...>:

Thanks Axel and Rolf for your answers. I had seen your 2013 paper, but
now I read it (and the J. Chem. Phys. 2012) in detail. I'm making some
tests on the water slab simulation, with the updated force field
command lines:

pair_style lj/long/tip4p/long long long 2 1 1 1 0.1546 7.0 # 10.0
pair_modify mix arithmetic
kspace_style pppm/disp/tip4p 1.e-4
kspace_modify force/disp/real 0.0002
kspace_modify force/disp/kspace 0.004

I will try to prepare an update for the related doc pages with what I
understood from my reading and tests, but I will welcome your expert
feedback.

that is the right spirit, and the perfect way of improving the
documentation. :wink:
steve (almost) always reviews additions to the documentation, and if
you post them here, others get a chance to look at it as well and
comment.

rolf,
do you think you could also put together a few paragraphs for the
HOW-TO section that explains a little bit how parameters can tuned for
optimal performance and accuracy. while there is a detailed discussion
in your paper, it would be nice to have the most important conclusions
presented in explicit LAMMPS syntax. right now a lot is scattered
across multiple doc pages. without knowing what to look for, it can be
quite tedious to collect this and it is possible to overlook important
details.

axel.

Dear all,

rolf,
do you think you could also put together a few paragraphs for the
HOW-TO section that explains a little bit how parameters can tuned for
optimal performance and accuracy. while there is a detailed discussion
in your paper, it would be nice to have the most important conclusions
presented in explicit LAMMPS syntax. right now a lot is scattered
across multiple doc pages. without knowing what to look for, it can be
quite tedious to collect this and it is possible to overlook important
details.

Sorry for the late reply. I had planned to send the reply +
the text for the HOW-TO section at the same time, but then
I moved writing it several times...
Anyway: I will add some info to the HOW-TO section such that
a correct usage of the pppm/disp will be aleviated in the future.
I will possibly be able to send it out before the end of this
week.

Best, Rolf

Dear all,

2015-02-02 12:48 GMT+00:00 Rolf Isele-Holder <[email protected]...>:

Anyway: I will add some info to the HOW-TO section such that
a correct usage of the pppm/disp will be aleviated in the future.
I will possibly be able to send it out before the end of this
week.

That sounds great!

Sorry for my long silence, I'm still running some tests on the effect
of the different parameters on the surface tension of different water
models. It seems I get a small but systematic difference with
published results, I will keep you informed of my progress...

Best regards,
Laurent

Dear all,

here are finally the updates to the how-to section regarding
how to use the pppm/disp. I also made some updates to the
code such that an error message will show up if the user does
not make any specifications regarding the pppm/disp parameters.
The new version of the code and the related documentation will
hopefully prevent non-optimal usage of the pppm/disp in the
future.

Best, Rolf

submit.tar (1.15 MB)

thanks a lot rolf.

i'm currently merging it into LAMMPS-ICMS and will do some tests and
forward the result to steve. there are some conflicts with very recent
changes in the development head, so it won't go in smoothly as is.

axel.