I’m trying to perform energy minimization on a molecule (e.g., C₆H₁₄) using ReaxFF, but I’m encountering unusually large energy values, and the minimization completes within two seconds, only running two iterations. I’m not sure what’s going wrong or if I’m missing something critical in the setup.
Here is the input setup:
units real
atom_style charge
read_data C6H14new2.data
#Define ReaxFF Pair Style and Coefficients
pair_style reaxff NULL
pair_coeff * * ffield.reax C H
# Charge Equilibration Fix for ReaxFF
fix qeq all qeq/reax 1 0.0 10.0 1.0e-6 reaxff
#Basic Thermo Output
thermo 10
thermo_style custom step temp pe press
#Direct Minimization
minimize 1.0e-4 1.0e-6 1000 10000
Any insights or suggestions would be greatly appreciated!
There isn’t anything to suggest since you do not explain:
- what makes the energy values unusually large
- what is the problem about having few minimization steps and whether the resulting geometry is reasonable or not
Thank you for your reply.
- Unusually Large Energy Values: I’m seeing values like 1850 kcal/mol, which suggests a significant issue, but I need a clearer idea of what factors in my model could lead to this extreme.
- Few Minimization Steps: My code is performing only two minimization steps before stopping, which seems insufficient.
Why do beginners think it is a good idea to learn molecular dynamics using ReaxFF? I really have no idea of why this particular force field is so popular these days.
This kind of statement without any context (e.g. which energy is this?) is basically useless and not helping.
Please note that LAMMPS ReaxFF implementation allows to access the total values of all the individual energy components, so it should be straightforward to pinpoint the main contributor. You can also compare to simpler but similar molecules with the same or similar C-to-H ratio and compute their energy and see if their value scales. You cannot compare to force field calculations like Amber or CHARMM, since they use exclusions.
… and for all force field calculations (including ReaxFF) the fact remains that the absolute value of the energy is arbitrary; only relative values matter.
Why? Are you sure you are not in a metastable geometry due to symmetry? or that the initial geometry is not already quite optimal? A simple visualization can tell.
My suspicion is that you don’t need to do atom typing.
Of course this is ignoring the fact that compared to e.g. OPLS-AA or Amber and CHARMM, ReaxFF is rather poor quality, especially for systems it hasn’t been trained for and for compounds where conventional DFT is not a good choice. E.g. dispersion interactions are particularly poor unless you use complex meta-GGA functionals
ReaxFF is good at molecular structure and reactions. 
1 Like
Reaxff can do chemical reactions and almost no other force field can.
Sure, I am aware of that. But in this forum, I have seen many instances (like in this thread) of it being used to do all sorts of things: compute the dielectric constant of graphene, salty solutions with an external electric field, iron oxide, the diffusion coefficient of CO2 in CaO (okay, for this one a reaction may occur), just to quote the most recent ones.
If one is interested in modelling some physical observable depending on the microscopic distribution of particles, then there are much better choices.
The energy value in question is the potential energy of the system. The geometry was pre-minimized using Avogadro when the structure was built, so it should be close to an optimized state.
i will check with similar molecules and see . thank you very much
As @alphataubio mentioned, if you want to model a chemical reaction, then ReaxFF is what you need to learn.
But if you are interested in modelling the solid-state properties of materials (viscoelasticity, diffusion coefficients, thermal transport, order coefficient, etc.), then you better stick to type-I force fields, that is. models based on point particles with a mass and a fixed charge. Just the two cents from a random guy on a forum, clearly.