I am a beginner in LAMMPS. The LAMMPS version I am using is “LAMMPS (31 Aug 2021)”.
I tried stretching the metal-ceramic interface using LAMMPS, using eam/alloy for metal, meam potential for ceramic, and morse potential for C-Cu and Ti-Cu for the interface. But after using fix1 all npt temp 298.15 298.15 ${tdamp} y 0 0 ${pdamp} z 0 0 ${pdamp} (s p p as boundary condition, stretch in x direction), but when I modify the cutoff of morse, And when increasing neighbor to make it larger than cutoff, the error “ERROR: Lost atoms: original 5338 current 5337(src/thermo.cpp:439)” appeared.
I looked at the running process and found that the box became very large in the x-direction during the npt process, but I found that the atomic positions did not change unexpectedly. I found that when I change the cutoff of morse to 2.5, there is no error, but when I change it to 6, the above error will appear. I can’t think of where his mistake lies.
“Lost atoms” errors are an indirect indication of a problem (same as an elevated body temperature is an indication of being sick, but it is not sufficient an indication for diagnosing the cause).
From your description it looks as if you have extremely high forces for some atoms and those accelerate a lot and thus are expelled from your system. This is an indication of either a bad initial geometry or bad potential parameters or a bad overall model. This is confirmed by the fact that the reduction the morse cutoff (which will largely disable the morse potential) changes the behavior.
There is not enough information here to make a more specific assessment.
Just one more recommendation, as a LAMMPS beginner, you have chosen a very complex pair style setup. This is very easy to get wrong. It is strongly recommended to practice your MD simulation skills with some simpler systems (using just one pair style) before trying more complex systems.
dear professor
First of all thanks for your answer. For the X-direction stretch I use “velocity mobile ramp vx 0.0 0.5 x 4 46 sum yes units box” timestep 0.001fs My above error occurs when I try to use npt to relax the system, the initial model metal- The spacing between the ceramic interfaces was set to 2 Angstroms, so I tried the npt above to relax the system so that the interface positions are at the optimum distance.
You made it very clear that my use and understanding of cutoff was not complete in my previous work, so there would be many mistakes and problems. Also apologies for not describing the problem in more detail.
Based on your suggestion I am going to increase the spacing of the interface in the initial model, wondering if this will solve the problem I am facing.
Just making (random?) modifications to your model based on a guess is neither a very scientific nor a very efficient approach to solve a problem. To solve a problem, you have to first understand the cause. To understand the cause, you need to carefully observe. That can be done with a careful visualization of your trajectory (i.e. with frequent output for a short run). Since you have observed a large change in box dimension in x direction you need to ask yourself if that would be self-inflicted by your model (and the velocity command you used, which is as far as I can tell not suitable for your purpose anyway). Also, using a timestep of 0.001fs seems very unrealistic and extremely wasteful.
Overall, my impression from your two posts is that you urgently need tutoring by somebody with experience in these kind of simulations. At the very least you need to spend (more) time studying input files and available tutorials for the same kind of studies for other, similar materials (more carefully).
Performing meaningful MD simulations is in many ways more a craft than a science and and only a genius can learn that on his or her own with help from text books or manuals. Online forums are not a sufficient substitute.
First of all thank you sir for your reply.
Yes, as you said I was in a bit of a rush, I think I should look for papers and books on stretching of dissimilar materials. Learn and learn from the experience of the predecessors to solve their own problems.
At the same time, it is a pity that my teachers and brothers could not give me MD-related guidance, because they did not understand or said they had not used MD for computational simulation. Therefore, I put too much hope on the forums for help. Of course, if you know relevant books or literature, you can recommend them to me, I would be very grateful.
I don’t know (and don’t care to know) enough about your specific area of research to make any specific recommendations, but - as I already mentioned - you need in person tutoring by an MD practitioner. If your adviser is not sufficiently experienced, finding an experienced collaborator is crucial anyway. As I already mentioned, it is highly unlikely that you can teach yourself what you need to know for doing meaningful research. Otherwise you are liable to repeat many of the mistakes others have done (with the best intentions) over the many years that MD is used for research.
