vmd-l: NPT/ Missing bonds / atoms & atom overlap

Hello everyone,

I have a question which is connected to both LAMMPS and VMD

: I am trying to simulate some long molecules ( consisting of C,O and H) in
an NPT ensemble, to observe some conformational changes. However, I get the
message of missing bonds or atoms in dihedrals and my simulation crashes in
maximum 100.000 femtoseconds.

this is almost always caused by either a bad initial structure,
bad/incorrect force field parameters, incorrect topology, bad
simulation settings or a combination of them.

I have looked up the relevant solutions ( fix/nve, fix langevin, and I use a
small timestep of 0.1). The missing atoms are due to atom overlap , which
causes an abnormally high temperature and eventually one of the atoms gets
blown out of the box.

is this overlap present in the initial configuration or does it happen
over time. in the first case, your initial geometry is bad, in the
second case, your force field parameters or topology assignment.

My coefficients are correct - I have double-checked them. I use packmol to

i cannot recall the number of times that people have claimed to have
double-, triple, whatever-multiple checked their input, yet there were
still mistakes, bad unit conversions, omissions or typos.

pack around 30 molecules and then VMD with the topo tools to assign charges

When I first view the molecules in the VMD (prior to lammps), the structure
is correct. However when I see in VMD the trajectory of the lammps file ,
until it crashes, the molecules seem to not have some bonds and have created
others, despite the fact that the VMD displayed them correctly in the first

I know it is not a LAMMPS issue because i did the exact same simulations
with a similar molecule and it worked perfectly. So I think it is a VMD

the fault is most likely in your force field parameter assignment
and/or topology data.

issue that I am having. Is it possible that the radii of the atoms are not
correct, even though the display is correct? For example, for carbon, a

VMD doesn't know *anything* about your force field and what VMD shows
is based on some very simple heuristics. it is irrelevant whether VMD
shows the "right" or "wrong" radii.

radius of either 1.4 or 1.3 Angstrom leads to a correct display, but maybe
the lower value can lead to overlap...?


Thanks for the tips Axel,

Turned out I had to perform a good energy optimization of the initial molecule ( I am new to VMD and LAMMPS) and I had done just a simple geometry optimization! That solved the problem and everything’s fine now!