MD error

[shyoun]

Hello,

I calculated a simulation of deposition of GeH4 on MoS2 with a ReaxFF potential parameter that I made.
And as soon as GeH4 got closer to MoS2, all of atom's information was deleted.

This is part of the structure information.
# Define elements
elements = [Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur,
            Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur,
            Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur,
            Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, Sulfur, .... Hydrogen, Hydrogen]

# Define coordinates
fractional_coordinates = [[ nan,  nan,  nan],
                          [ nan,  nan,  nan],
                          [ nan,  nan,  nan],
                          [ nan,  nan,  nan], ....  [ nan,  nan,  nan]]

# Define velocities
velocities = [[ nan,  nan,  nan],
              [ nan,  nan,  nan],
              [ nan,  nan,  nan], .... [ nan,  nan,  nan]]*Angstrom/fs

Is there any known cause of this kind of error? I wonder why this error is happening.
I'll put up the code and potential information together, just in case it helps.

[Anders Blom]

I looks like the force and then the velocities of the atoms became ill defined, probably because the atoms ended up too close to each other. Some potentials cannot handle when atoms have high energy/temperature/velocity and does not reject them well enough when they come close. See if you can run with lower initial velocities, or at lower temperature. Also, try first to run the ReaxFF model on a more static problem to test it and see that it behaves correctly.

[shyoun]

Thank you for your advice.

I tested it in low temperature. Unfortunately, I found that simulation with 10K(and also 1K) results in the same error.
The speed of atoms does not exceed 0.01 Angstrom/fs just before an error occurs.
Is this just a matter of potential parameter?
There are several combinations that is not among torsion term in the potential parameter. Is this the cause?

[Anders Blom]

Hard to say but yes it's likely to be an issue with the parameters. Was this a published ReaxFF potential, or something you made yourself?

Perhaps you want to try your hand at using our Moment Tensor Potentials instead - you can see some details about this in the release webinar linked above

[shyoun]

Thank you for your answer.

I made potential by referring to the parameters of several papers.
And the 'webinar' means the webinar held on 2021/6/22?
I haven't heard it yet, but I'm glad it's related to my problem.

Thank you so much. I'll listen to the webinar and try it.

[Anders Blom]

To start with, yes, but we are also planning a dedicated MTP webinar in August

[shyoun]

Thank you! I will take it, too