Hello all LAMMPS users,
I tried to add rotational motion for molecules with the fix deposit command. The command allows us to give translational energy in x/y/z directions, but as far as I know, it does not have the option to add rotational motion.
Naturally, a molecule has these motions (vibrational motion is not considered here). Do you know how to add it?
I’m using the following version of LAMMPS: 23 Jun 2022 - Update 1.
Thank you in advance.
Naoya
Well, if you want fix deposit to add a rotation, you will have a modify the source code for that command accordingly and recompile.
However, where would be the benefit? If you intend to do some kind of sputtering operation the rotational energy would be negligible, and for other modes, the molecules will pick up rotation, once they interact with existing molecules or atoms and while they are in free flight, it doesn’t matter.
Dear Dr. Axel,
Thank you for giving me advice. I understand it is needed to modify the source code and recompile it.
For answering your question, if we run the sputtering simulation using an atom, rotation can be negligible because incident energy is relatively high. But we are using this command for chemical vapor deposition (CVD) simulation using molecules, and we found that adding rotation affects the surface chemistry because the incident energy is lower than the sputtering case. That is the benefit.
Does it make sense? I would suggest this addition due to a good expansion for LAMMPS software in the future.
Best regards,
Naoya
Can you elaborate on that? In which way does it change? Can this be quantified?
How would that change differ from using, e.g., a different random number seed.
Pleases feel free to make this change and then submit it for inclusion into the LAMMPS distribution. Large parts of LAMMPS were written by people like you, that had a specific need and then went ahead and implemented it and were generous enough to share it with the community. 3.2. Submitting new features for inclusion in LAMMPS — LAMMPS documentation
It’s not just generosity @naoyauene – if your method is a significant improvement on older methods and your code works well and you write it up well, and other people adopt your method, then it’s going to be a fairly well-cited publication, if you need to keep track of that sort of thing.
(Having said that, I too don’t intuitively see why depositing a molecule with angular momentum would be that different from, say, depositing it with translational velocity at a grazing angle to the surface in question.)
Can you elaborate on that? In which way does it change? Can this be quantified?
A SiH3 molecule collided on the Si surface and the surface event is classified: as reflection, desorption, chemisorption, and physisorption depending on the behavior. After thousands of collisions, the probability of those events can be quantified. The detail is described in the published paper.
https://www.researchgate.net/publication/357779062_Reactive_force-field_molecular_dynamics_simulation_for_the_surface_reaction_of_SiH_x_2-4_species_on_Si1_0_0-2_1H_surfaces_in_chemical_vapor_deposition_processes
In this paper, I have not considered rotational motion. But, I added rotational motion for checking the influence and found that the chemisorption probability decreased by over 20%. I have not changed the random seeds yet, but I guess 20% is not in the range of randomness. Did I answer your question?
Pleases feel free to make this change and then submit it for inclusion into the LAMMPS distribution. Large parts of LAMMPS were written by people like you, that had a specific need and then went ahead and implemented it and were generous enough to share it with the community. 3.2. Submitting new features for inclusion in LAMMPS — LAMMPS documentation
Good suggestion. Thank you.
Dr. Srtee,
Thank you for your comment.
(Having said that, I too don’t intuitively see why depositing a molecule with angular momentum would be that different from, say, depositing it with translational velocity at a grazing angle to the surface in question.)
Actually, I also wonder why this changed. Before checking that effect, I expected that rotational motion does not affect the surface chemistry well. One note is that it depends on the reactivity of the gas molecule. For example, a molecule with high reactivity such as SiH2 does not have the rotational effect because it chemisorbs immediately. But, a molecule with low reactivity such as SiH4 seems to have the rotational effect.
This just prompts a few more questions:
Dear Dr. Axel,
I answer your questions.
When you added rotational motion, did you reduce the translation velocity accordingly so that the total kinetic energy per molecule remains the same?
No, I did not reduce the translation velocity. I tried to add it by using the following commands;
velocity pre create $T 39849 mom no dist gaussian
fix 1 pre momentum 1 linear 1 1 0 # cancel momentum of COM in x and y directions
They are not fix deposit commands but based on this work, I would like to add the rotational motion.
Did you use fixed deposit velocities or ranges (and thus a range of kinetic energy at impact)? And did you assign velocities only in the z-direction or others as well?
I used fixed velocity only in the z-direction. Detailed information is described in the published proceedings.
https://www.researchgate.net/publication/356056295_Reactive_Force-Field_Molecular_Dynamics_Study_of_the_Effect_of_Gaseous_Species_on_SiliconGermanium_Alloy_Growth_by_PECVD_Techniques
How close in succession did you deposit molecules? i.e. did you have isolated incidents or could it be that deposited molecules could interact while in flight?
Did you deposit with the “target” option? Does the insertion region cover the entire box?
The molecule has been added to the system every 5.5 ps in the above paper. In my work, the molecule has basically finished its events (desorption/chemisorption) within 2.0 ps. Therefore, molecules do not interact in the gas phase. I did not use the “target” option. Added molecules can interact on the entire growing surface.
Then you are comparing apples and oranges. I think the results should be classified based on the total kinetic energy molecules have on impact.
We have already established that this requires modification of the source code, and that I and @srtee do not believe that including the rotation makes a significant difference, if impact kinetic energy is accounted for properly.
Sorry, but this is too much information for me. This is not my area of research interest. I don’t have time to study whole papers just because you are unable to convince me to program the feature you want on your behalf.
I am really not sure how this corresponds to rotational motion instead of grazing translational motion.
In any case, perhaps you could emulate your desired effect by imposing a torque on molecules mid-flight, using a carefully specified fix addforce on a dynamic group of one type in the flight region. I know this might seem like a lot of work, but it is even more work to write a custom fix only to run the risk of undetected bugs and problems – LAMMPS is flexible enough that its existing, well-documented behaviour already covers an astonishing multitude of sins.