I have been running MD of certain CG systems with tabulated potentials. Both the bond and pair potentials in the tables look reasonable in my opinion - the bond potentials have a minimum say at around 2.0-2.5 A and then increase steeply toward both sides, while pair potentials drop from very high values at small distances to ~0 at larger distances. I am also sure based on visualizations that the mapping is correct in all situations.
Despite using the same algorithm and same parameters for obtaining the pair and bond CG potentials in all the systems that I want to investigate, some systems give successful complete MD runs while some run into the “Bond length > table outer cutoff” error at some early stage of simulation after minimization. A senior colleague said that this might be partly due to the instability of the pair potentials as they fluctuate between say -0.1 to 0.1 at longer distances - however this minor fluctuation appears in table files that give both the successful and unsuccessful simulations and I cannot tell without certainty before running MD whether it will work or fail. Do you know what could be the problem that causes this error and how to eliminate it?
This is more a physics question than a LAMMPS question. The error you get is pretty clear, some bonds extend further than the maximum value of your tabulated model. Period.
As to why this happens, there can be many reasons both in the way your model was developed (using Boltzmann Inversion? Force-Matching? Other methods?) or the way you run your simulations (timestep, temperature, pressure, initial configuration etc). Since you provide nearly no information related to these parameters, there is little help you can get. The best for you might be to check in the relevant literature to see how your model compare to other CG models developed for similar systems, find a way to correct your wiggling values and see if they give meaningful results[1] or fiddle around with parameter used to produce your models.
In the end this is the kind of stuff that you have to figure out to do the science behind your work. If you think your error is due to a bug in LAMMPS computations or if there are issue in some input commands you can pinpoint, then you might come back and ask for more relevant help with minimal working examples for people here to reproduce your issue.
You might be interested in numerical filters like Savitzky–Golay for such purpose. Look for more informations in the relevant numerical science literature. ↩︎
Thank you very much for your reply and explanation, I did not even expect to get it so quickly. After discussing with the senior colleague of mine I realized that the error can be avoided if I estimate harmonic bond potential coefficients from the tabulated bond potentials and use them for each bond in the MD simulation.