Coulombic Interactions in MB-pol with Conventional Force Fields

[Jackie Jiang]

Dear Developers,

Thank you for your outstanding work on MBX. As someone new to the MBX framework, I have a couple of questions about hybridizing the MB-pol water model with conventional force fields in LAMMPS.

In the provided example (examples/mof_ff_lammps+6_h2o_6_co2_mbx), the metal-organic framework (MOF) includes charged particles, yet no pair style and kspace_style is specified for Coulombic interactions . Could you explain how the charged atoms in the MOF affect the dynamics of water molecules in the MB-pol model, and vice versa?

Additionally, I am interested in simulating a system involving graphene oxide (GO) with non-rigid functional groups interacting with water using MB-pol. Could you provide guidance on the proper LAMMPS input settings for such a system, including any necessary pair styles or parameters to account for the interactions?

Thank you in advance for your assistance.

Best regards,

[Philip Zhu]

Hi Jackie,

I can answer your first question and I’ll leave other questions to other group members.

In lammps data file, you specifies charges on MOFs, and those point charges are passed into MBX, which handles all electrostatics. Therefore, to avoid double-counting the electrostatic energies, you don’t specify columb pair styles in lammps.

MBX will take the MOF atoms as external charges, using particle mesh eward (PME) if the box is periodic, to calculate permenent and induced electrostatics. Parameters regarding the cutoff of kspace, grid densities of PME can be specified in mbx.json.

best,

Xuanyu

[Jackie Jiang]

Hi Xuanyu,

Thank you for your prompt response! I see. Will the electrostatic force hanlded by MBX be passed in reverse to the external charged atoms (such as those in a MOF)  or are they considered only for  the water molecules?

Best,

[Philip Zhu]

I may not understand what you mean by passing in reverse; MBX receives information about external charges (all MOF/GO atoms), and MBX also calculates position-dependent charges for MB-pol water molecules. MBX then will calculate the electrostatics for everything, including water-water, MOF-MOF, and MOF-water electrostatics interactions.

Xuanyu

[Richa Rashmi]

Dear Jackie,

In response to your second question, please find attached test input files for MBX/LAMMPS simulations of a graphene/water slab system using MB-pol for water and the Tersoff potential for graphene. Most of the setup follows a standard MBX/LAMMPS simulation for water. The key differences are:

1. The use of dp1 in fix mbx to pass external charges (in this case graphene carbon) to MBX.
   
2. The use of pair_style hybrid/overlay to incorporate both MBX and Tersoff potentials.
   
3. The use of coul/exclude to remove Coulombic interactions between 1-2 and 1-3 bonded atoms in graphene. Since MBX calculates the pairwise Coulombic interactions for all external charges (i.e. Coulombic interaction between bonded C-C will be calculated by MBX), if a certain potential (e.g. OPLS-AA) does not require Coulombic interactions between 1-2 and 1-3 bonded pairs, then we need to use coul/exclude for those pairs.
   

You can modify the input files as needed to convert the system from graphene to graphene oxide.
Please feel free to reach out if you have any further questions.

Best regards,
The MBX Team

[Jackie Jiang]

Dear Richa,

Thank you for the detailed tutorial on modeling a graphene/water slab system. It has been extremely helpful!

I have a few questions regarding the applicable pressure and temperature ranges for such a nanoconfined water system. In the previous MB-pol work (NC 2023), a phase diagram of bulk water was reported for pressures between 0 and 1 GPa. Can MB-pol accurately describe extreme confinement systems, such as those at sub-nanometer scales (e.g., monolayer or bilayer), under high pressure? If so, what are the limitations or considerations for these conditions?

Additionally, I conducted some tests and noticed that the output pressure of water using MB-pol differs significantly from TIP4P at the same water density. Below is the LAMMPS command I used to compute the pressure of water. Could you please confirm if this approach is appropriate?

"

compute wstress water stress/atom NULL
compute p water reduce sum c_wstress[1] c_wstress[2] c_wstress[3]
variable wpxx equal -(c_p[1])/vol
variable wpyy equal -(c_p[2])/vol
variable wpzz equal -(c_p[3])/vol

"

Best,

[Jackie Jiang]

I have an additional question regarding the MBX fix command for a graphene oxide (GO)/water slab system in LAMMPS.

For the graphene/water slab system, the atom IDs were assigned as GRA(1:C) Water(2:O 3:H), and the external charge was specified as:

fix fxmbx all mbx 2 dp1 1 1 1 1 h2o 2 3 3 2 3 3 json mbx.json

For a GO system with atom IDs GO(1:C 2:O 3:H) Water(4:O 5:H), which of the following MBX fix commands is correct?

(1) fix fxmbx all mbx 2 dp1 1 3 1 1 h2o 4 5 3 4 5 5 json mbx.json
(2) fix fxmbx all mbx 2 dp1 1 3 3 1 2 3 h2o 4 5 3 4 5 5 json mbx.json

Could you clarify which one accurately reflects the atom types and their interactions for the GO/water system?

Best,

[Richa Rashmi]

Dear Jackie

1. There are two sources of limitations: (a) theoretical framework, (b) numerical. The theoretical framework assumes water as molecules, so MB-pol should work as long as the water molecules do not dissociate. Note that due to numerical errors, MB-pol water may dissociate before real water molecules do at extremely high pressures. We recently reported that numerical errors can sometimes lead to instabilities in MB-pol around 2.5GPa and above (https://doi.org/10.26434/chemrxiv-2025-76p6j). As long as the simulation is stable, there should not be an intrinsic problem of using MB-pol in sub-nanoconfined systems.
2. Since the density of MB-pol and TIP4P water are different at a given temperature, the pressure is expected to be different for both models at the same density. So, we believe what you are seeing is normal.
3. The following fix mbx line for graphene oxide/water simulation: fix fxmbx all mbx 2 dp1 1 3 1 1 h2o 4 5 3 4 5 5 json mbx.json is correct. Since MBX treats all atoms in graphene oxide as part of the dp1 monomer, assigning them the same atom type is appropriate.

Please feel free to reach out if you have any further questions.

Best regards,
The MBX Team

[Jackie Jiang]

Dear Richa,

Thank you for your prompt and detailed response. Your guidance and the provided reference are very helpful and greatly appreciated!

Best,