Hi, the follows are my questions, and I am thankful for the answers. I skip some irrelevant information, but please inform me if it needs to be furthering specified to be answered .
Description: I am running a system for short surfactant chains (6 dpd beads) dissolved in the water using DPD simulations. Style of bonded potential it is harmonic. I tried to apply additional non-bonded 1-3 body interactions as molecule rigidity, which are also in harmonic (50% weaker and 2 times longer than 1-2 bond). Simulations ran sufficiently long in nvt with temperature set to 1. Without this 1-3 bond applied, the temperature and pressure outputs are stabilized around reasonable values. While additional 1-3 bonds applied, there are strong fluctuations in T, P, Epair, and Emol, and the average value of T and P is much higher than those without 1-3 bonds.
Question 1: After debugging the problem, I feel the error is more from the bad communicate cutoff. Based on the manual, one needs to assign “appropriate” ghost cutoff if bonded interactions extend further than the pairwise cutoff. For DPD, the pairwise cutoff is Rc, and say it is one in lj unit. I have equilibrium length of 0.8 for 1-2 bond and 1.6 for 1-3 bond. Should I set my ghost cutoff about 2 or even larger? Should ghost cutoff be upper-limited by the size of periodic boundary? What is the consequence if I set the cutoff too short or too long as it should be?
Question 2: I know LAMMPS handle the minimum image well, but may I know whether there is a paper describing the details of the algorithm for potential calculations while the molecules are across the periodic boundary? Please kindly let me know if it has been already described somewhere in the manual.
Comments below.
Steve
Hi, the follows are my questions, and I am thankful for the answers. I skip
some irrelevant information, but please inform me if it needs to be
furthering specified to be answered .
Description: I am running a system for short surfactant chains (6 dpd beads)
dissolved in the water using DPD simulations. Style of bonded potential it
is harmonic. I tried to apply additional non-bonded 1-3 body interactions as
molecule rigidity, which are also in harmonic (50% weaker and 2 times longer
than 1-2 bond). Simulations ran sufficiently long in nvt with temperature
set to 1. Without this 1-3 bond applied, the temperature and pressure
outputs are stabilized around reasonable values. While additional 1-3 bonds
applied, there are strong fluctuations in T, P, Epair, and Emol, and the
average value of T and P is much higher than those without 1-3 bonds.
Question 1: After debugging the problem, I feel the error is more from the
bad communicate cutoff. Based on the manual, one needs to assign
"appropriate" ghost cutoff if bonded interactions extend further than the
pairwise cutoff. For DPD, the pairwise cutoff is Rc, and say it is one in lj
unit. I have equilibrium length of 0.8 for 1-2 bond and 1.6 for 1-3 bond.
Should I set my ghost cutoff about 2 or even larger? Should ghost cutoff be
upper-limited by the size of periodic boundary? What is the consequence if I
set the cutoff too short or too long as it should be?
If the cutoff is too short, LAMMPS will throw an error when it tries
to find an atom in a 3-body angle (for example) and it can't find it as
a ghost atom. If you're not getting such as error, you're fine. Note that
it is likely the center atom that is trying to find the other 2. So you don't
need to find atoms that are 2 hops away.
Question 2: I know LAMMPS handle the minimum image well, but may I know
whether there is a paper describing the details of the algorithm for
potential calculations while the molecules are across the periodic boundary?
Please kindly let me know if it has been already described somewhere in the
manual.
Minimum image is that algorithm. It's a basic MD concept, so not described
in the LAMMPS manual. Try Allen & Tildesley if you want a text-book
explanation.