Openff (Charge Increment Model)

[Shahriar-Keshvari]

Hello:

I have been able to run ForceBalance using the given example in tutorial 4. In that example the parameters to change are the epsilon and sigma (Rmin). However, I am interested in changing the partial charges and try matching the physical properties.

In openff specification (https://docs.openforcefield.org/projects/toolkit/en/topology/users/smirnoff.html) the example given does not include the the keyword "parametrize" and "parameter_eval", but in ForceBalacne study guide, these key words are defined within the charge increment tag. Using the ForceBalance study guide the program run but no charge increment is done (after any iteration the changes to partial charges are zero). Using the openff guide, the program crashed after the first iteration.

Now my question is, what is the correct format? If possible to have a working "chargeIncrementmodel" tag, explained for a molecule like ethanol, that would be very helpful. Please let me know if you need more information.

Thanks
Shah

Here is the error message if I use the openff guidance on charge Increment:
Traceback (most recent call last):
File "/home/shahriar/anaconda3/bin/ForceBalance.py", line 45, in Run_ForceBalance
optimizer.Run()
File "/home/shahriar/anaconda3/lib/python3.10/site-packages/forcebalance/optimizer.py", line 322, in Run
xk = self.OptTabself.jobtype
File "/home/shahriar/anaconda3/lib/python3.10/site-packages/forcebalance/optimizer.py", line 951, in NewtonRaphson
return self.MainOptimizer(b_BFGS=0)
File "/home/shahriar/anaconda3/lib/python3.10/site-packages/forcebalance/optimizer.py", line 676, in MainOptimizer
dx, dX_expect, bump = self.step(xk, data, trust)
File "/home/shahriar/anaconda3/lib/python3.10/site-packages/forcebalance/optimizer.py", line 748, in step
Emin = min(Eig)
ValueError: min() arg is an empty sequence

[mattwthompson]

The SMIRNOFF specification defines everything included in SMIRNOFF force fields. The section on <ChargeIncrementModel> defines the contents of that tag

The OpenFF Toolkit supports so-called "cosmetic" attributes which some downstream tools hack for behavior specific to their use cases.

in ForceBalacne study guide, these key words are defined within the charge increment tag

Could you remind me where you're seeing this?

[Shahriar-Keshvari]

Hello, I'm just bringing this up again. I just want to fiddle with the partial charges and try to reproduce some physical properties.

[mattwthompson]

ForceBalance, while immensely valuable to OpenFF's fitting efforts for many years, is not an OpenFF product and we don't maintain examples in that project. It's unfortunate that some may have fallen out of date but the ForceBalance developers would be the people to contact about any errors while running those examples.

[Shahriar-Keshvari]

Hmm, does this means that OpenFF does not support any sort of fiddling with the charges for the sake of physical properties calculations? If so, how to do that using OpenFF?

[mattwthompson]

There are many ways define partial charges in a SMIRNOFF force field and OpenFF supports all of them: https://openforcefield.github.io/standards/standards/smirnoff/#partial-charge-and-electrostatics-models

[Shahriar-Keshvari]

So I have used the keyword "parametrize=charge" inside the .xml file under the library of charge tag to specify that only partial charges to be changed. The program (ForceBalance) runs successfully however, after the partial charges are changed, there is a non zero total charge (the molecule is not neutral) associated with the molecule. The error message is:

openff.interchange.exceptions.NonIntegralMoleculeChargeError: Molecule CCC#N has a net charge of 0.0174447 compared to a total formal charge of 0.0.

Is this issue related to OpenFF or ForceBalance? I should mention that the outcome is independent whether I include the "ChargeIncremet" tag or not.