Force field potential for PVDF, PVDF-TrFE and PVDF-TeFE in LAMMPS

Hi,

Is there any available force field potential for the polyvinylidene fluoride(PVDF), PVDF-TrFE and PVDF-TeFE? Or I should ask how to implement MSXX and MSXXS potential on PVDF system?

Before I post it, I notice there is a related thread in 2010(http://lammps.sandia.gov/threads/msg10326.html). Dr. Kohlmeyer had provided the detailed work flow on it (Thanks!). I also notice that there are some information regarding this question(http://lammps.sandia.gov/doc/pair_hbond_dreiding.html). At the end of the page, it mentioned two papers, one is Mayo, Olfason, Goddard III, J Phys Chem, 94, 8897-8909 (1990) and the other one is Liu, Bryantsev, Diallo, Goddard III, J. Am. Chem. Soc 131 (8) 2798 (2009). I am curious are these all the information I need to build up the potential by myself? Or there is the potential I can use right away to test my desired initial configuration?

Deeply appreciate your help.

Cheers,
F.S.

Hi,

Is there any available force field potential for the polyvinylidene
fluoride(PVDF), PVDF-TrFE and PVDF-TeFE? Or I should ask how to implement
MSXX and MSXXS potential on PVDF system?

force fields for classical MD are usually independent of the
simulation software (for as long as the necessary functional forms are
available). the best way to find out about them is to search the
literature. for reasonably common materials, there often are multiple
alternatives. once you have identified the publications and the
description of the necessary functional forms, you can compare them
with what is available in LAMMPS and then begin to set up your
simulation.

axel.

Forgot to respond a while back. The advise was very helpful. Thanks a lot.

Cheers,
F.S.

Dear LAMMPS users,

In order to implement of force field parameters for PVDF system, I have identified the literature and tried to build the force field potential in the corresponding format. I still cannot get the right, or say the same unit cell values as the literature claimed.

In order to validate my potential, I am trying to find the minimal energy at 0K. Since I have a positive Coulomb potential, which is the dominating term in the potential energy, the box will expand (forever) in one specific direction to lower its energy.

I also notice that I get a positive Coulomb potential energy but negative elong potential energy. (Some discussion in this thread http://lammps.sandia.gov/threads/msg09816.html) Is it possible?

For instance if we set up Coulomb cut-off distance equals 10 angstroms, does it mean

ecoul -> Energy term when r = 0 ~ 10 angstroms which is calculated in real space.
elong -> Additional energy term when r > 10 angstroms which is calculated in kspace.

Here is part of my input file. I am struggling with this question for a while. Please generously shed some light. Thanks in advance.

clear
units real
dimension 3
boundary p p p
atom_style full
kspace_style ewald 0.0001
pair_style born/coul/long 7.6
neighbor 1.0 bin
neigh_modify every 10 one 10000
bond_style harmonic
angle_style class2
dihedral_style hybrid harmonic class2cosine
improper_style class2cosine

born style (http://lammps.sandia.gov/doc/pair_born.html)

pair_coeff 1 1 0.08440 0.323642 3.88370 579.22 0
pair_coeff 2 2 0.04200 0.218750 3.5 205.89 0
pair_coeff 3 3 0.08440 0.323642 3.88370 579.22 0
pair_coeff 4 4 0.01655 0.233042 2.74990 14.08 0
pair_coeff 1 2 0.05896 0.266076 3.68686 341.97 0
pair_coeff 1 3 0.08440 0.323642 3.88370 579.22 0
pair_coeff 1 4 0.05105 0.274631 3.26799 123.34 0
pair_coeff 2 3 0.05896 0.266076 3.68686 341.97 0
pair_coeff 2 4 0.03333 0.225783 3.10236 68.05 0
pair_coeff 3 4 0.05105 0.274631 3.26799 123.34 0

Cheers,
F.-C.

Comments below.

Steve

Comments below.

Steve

Yes, my system is charge neutral.

Understood. One related question. Since absolute energy has no physical meaning (http://lammps.sandia.gov/threads/msg42445.html), we only consider the energy difference. Is there a common sense that which energy term should be dominated? Say delta_ecoul > delta_bond > delta_angle, or it mainly depends on the system?

Thanks for the prompt response and Happy Easter.

Cheers,
F.-C.

Understood. One related question. Since absolute energy has no physical
meaning (http://lammps.sandia.gov/threads/msg42445.html), we only consider
the energy difference. Is there a common sense that which energy term should
be dominated? Say delta_ecoul > delta_bond > delta_angle, or it mainly
depends on the system?

it depends on the system, the boundary conditions, the force field
parameterization strategy, whether you have errors in your input.
everything.

axel.

So in general, how can we validate the force field parameters?

In this thread (http://lammps.sandia.gov/threads/msg20393.html), Dr. Kohlmeyer had mentioned that “force fields are always compromises”. I assume it means we can obtain some experimental quantities from simulation, but some we can’t. If we construct the force field from the literature, should we know what “physical” properties we can reproduce (if they didn’t mention from the literature)? What is the universality of it? Thanks for the insight.

Cheers,
F.S.

So in general, how can we validate the force field parameters?

run test calculations for your compounds of interest and determine how
close relevant properties are to what you are looking for. mind you,
while it may be difficult to get a good match with absolute values,
trends are often in better agreement due to error cancellation.

In this thread (http://lammps.sandia.gov/threads/msg20393.html), Dr.
Kohlmeyer had mentioned that "force fields are always compromises". I assume
it means we can obtain some experimental quantities from simulation, but
some we can't. If we construct the force field from the literature, should
we know what "physical" properties we can reproduce (if they didn't mention

a paper that describes the parameterization of a force field without
describing how it is validated and how you can reproduce that
validation is a useless paper.

from the literature)? What is the universality of it? Thanks for the

nobody can know a priori how transferable a force field is. full stop.

since in a classical force field you are in essence an integrating out
details, usually parameters are more likely to be transferred to
similar compounds (i.e. where the part you integrate out and represent
with a non-varying average value) and similar physical states
(gas/liquid/solid, temperature, pressure). in addition, things are
usually easier to control for pure substances than for mixtures.

axel.

So in general, how can we validate the force field parameters?

Here you better go and do some textbook (original publication) reading as
the answer to your Q will not be short. One pioneering paper that deals
with the topic can be found here:
doi:10.1209/0295-5075/26/8/005<http://dx.doi.org/10.1209/0295-5075/26/8/005>
And so there are many more...

In this thread (http://lammps.sandia.gov/threads/msg20393.html), Dr.
Kohlmeyer had mentioned that "force fields are always compromises". I
assume it means we can obtain some experimental quantities from simulation,
but some we can't.

What Axel probably meant is that, in the majority of cases, the expressions
describing the functional form of the force field cannot be strictly
derived from a more fundamental (first-principles) approach (such as
quantum mechanics) and thus will be inherently limited in their prediction
power by this mere fact. Never mind the additional source of noise derived
from the fitting procedures employed when generating specific
parametrizations for such forcefields.

If we construct the force field from the literature, should we know what

"physical" properties we can

reproduce (if they didn't mention from the literature)? What is the

universality of it? Thanks for the insight.

"If we construct the force field from the literature" ??? Did you mean: *if
we take the parameters from a known paper?* If yes, I will reply to your
second Q with another Q: Shouldn't you know what a tool does before trying
to use it?
Carlos

So in general, how can we validate the force field parameters?

Here you better go and do some textbook (original publication) reading as
the answer to your Q will not be short. One pioneering paper that deals with
the topic can be found here: doi:10.1209/0295-5075/26/8/005
And so there are many more...

In this thread (http://lammps.sandia.gov/threads/msg20393.html), Dr.
Kohlmeyer had mentioned that "force fields are always compromises". I assume
it means we can obtain some experimental quantities from simulation, but
some we can't.

What Axel probably meant is that, in the majority of cases, the expressions
describing the functional form of the force field cannot be strictly derived
from a more fundamental (first-principles) approach (such as quantum
mechanics) and thus will be inherently limited in their prediction power by
this mere fact. Never mind the additional source of noise derived from the
fitting procedures employed when generating specific parametrizations for
such forcefields.

not quite. referring to first principles calculations doesn't
automatically guarantee good force field parameters. force fields are
by their very nature empirical. while many force fields have some kind
of "recipe" or general strategy (which often includes quantum chemical
calculations), it is not the only source to arrive at the final
parameters. based on the initial guess from those parameters, one
usually does a lot of test calculations and then varies parameters to
see, if a better match or better compromise can be achieved. usually,
there has to be a certain balance between the charge distribution, the
excluded volume and how "soft" a compound has to be, as well as how it
mixes with other compounds from the same force field. it can be very
tedious work since you have too few degrees of freedom and lots of
hidden correlations between properties and parameters.

axel

>
>>
>> So in general, how can we validate the force field parameters?
>
>
> Here you better go and do some textbook (original publication) reading as
> the answer to your Q will not be short. One pioneering paper that deals
with
> the topic can be found here: doi:10.1209/0295-5075/26/8/005
> And so there are many more...
>
>
>>
>> In this thread (http://lammps.sandia.gov/threads/msg20393.html), Dr.
>> Kohlmeyer had mentioned that "force fields are always compromises". I
assume
>> it means we can obtain some experimental quantities from simulation, but
>> some we can't.
>
>
> What Axel probably meant is that, in the majority of cases, the
expressions
> describing the functional form of the force field cannot be strictly
derived
> from a more fundamental (first-principles) approach (such as quantum
> mechanics) and thus will be inherently limited in their prediction power
by
> this mere fact. Never mind the additional source of noise derived from
the
> fitting procedures employed when generating specific parametrizations for
> such forcefields.

not quite. referring to first principles calculations doesn't
automatically guarantee good force field parameters.

A bit of a misunderstanding here. I meant to say that the functional form
could not be derived from quantum theory and thus ensure that "in
principle" its level of prediction could emulate that of the original
theory.
I did not mean to say that first principles numerical calculations (with
all their inherent approximations)
had to be the reference point.
Carlos

Thanks a lot. It is a fruitful discussion here.

Maybe I should also clarify my description as well. I indeed take the known force field parameters from literatures, which is not implemented in LAMMPS, so I rewrite it in LAMMPS format. In order to validate these input, (making sure I didn’t mistype a number or use the wrong unit), I want to reproduce the reported quantities (say property A) to be the same as this reference. I got the reasonable closed value (property A). Then I want to calculate some other quantities (say properties B, C, D, …), which aren’t reported in this reference, and to compare with others sources (calculations/experiments). Like Dr. Kohlmeyer mentioned, people use different strategy to generate force field parameters. They may or may not use B, C, D to generate this potential. That is why I asked “how can we validate the force field parameters?” Or say does it also “good” to calculate properties B, C, D, …?

The bottom line here seems to be we should be able to capture the trend, but not the absolute values. (the same as only energy difference has physical meaning using force field, but not its value)

PS. Thanks Carlos. (doi:10.1209/0295-5075/26/8/005) It is a very nice reference, (and classic, I believe).

Cheers,
F.S.

Thanks a lot. It is a fruitful discussion here.

Maybe I should also clarify my description as well. I indeed take the
known force field parameters from literatures, which is not implemented in
LAMMPS, so I rewrite it in LAMMPS format. In order to validate these input,
(making sure I didn't mistype a number or use the wrong unit), I want to
reproduce the reported quantities (say property A) to be the same as this
reference. I got the reasonable closed value (property A). Then I want to
calculate some other quantities (say properties B, C, D, ...), which aren't
reported in this reference, and to compare with others sources
(calculations/experiments). Like Dr. Kohlmeyer mentioned, people use
different strategy to generate force field parameters. They may or may not
use B, C, D to generate this potential. That is why I asked "how can we
validate the force field parameters?" Or say does it also "good" to
calculate properties B, C, D, …?

If you can reproduce the values of the properties reported in the original
paper for the FF then you should assume that your Lammps translation of the
params is operational (unless there is another reason for you to fear a
mistake in your procedure). With such params, then you can proceed to
predict B,C,D, etc and obtain whatever values the current FF is bound to
give. If you want further improvement that is another story.

The bottom line here seems to be we should be able to capture the trend,
but not the absolute values. (the same as only energy difference has
physical meaning using force field, but not its value)

PS. Thanks Carlos. (doi:10.1209/0295-5075/26/8/005) It is a very nice
reference, (and classic, I believe).

That reference has certainly received much attention. I leave the "classic"
reference solely to you in case the original authors don't see it as a
compliment but as a way to be labeled "old" :wink:
Carlos