Distance between two atoms exceeds box length

GROMACS version:2018.6
GROMACS modification: No
Hello all,
I am simulating two carbohydrate chains with box length of 59 Angstrom. The distance between two atoms (picked from two different carbohydrate polymers) are calculated using MDAnalysis (I have also verified with VMD). The distance varies with time as follows:

Clearly one can observe distance between two atoms exceeds the box length at some time frames.Here I have also attached snapshot of 990th frame and pointed out the two atoms.

At this particular frame the distance between two atoms is 81.17 Angstrom.
**The frame was corrected using trjconv with pbc mol and center flag.
Please let me know if I can provide any further information and please do comment on any technical or conceptual point I am missing here. That would be really great!! :)

Center on one chain to wrap them both into the central image. What you have is correct, in that the geometric center of the dimer is coincident with the geometric center of the box. That’s what trjconv is designed to do. Multimeric systems always require a bit more care on the user’s part to get the imaging right.

So, is this okay to have interatomic distance to be more than box length? But, in PBC, we cannot expect inter particle distance greater than BOX/2, right?
I think I am somehow making mistake in either understanding the concept or GROMACS technicality in handling PBC. Please correct me or suggest something if I am wrong. That will be appreciable :)

The distance you are getting is an artifact of incorrect imaging. Like is said, to get the actual inter atomic distance within the unit cell, you need to re-wrap the coordinates more carefully by centering on one chain only.

Sorry for my late response as I could not access my desktop for few days.
Thank you so much for pointing this out. I have wrapped the trajectory centering on one chain using an index file. And it looks far better than the previous one.
But still the second chain is coming out of the box at some time frame. Any suggestion on that?
Thanks again :)

Is your box of sufficient size to contain both chains without minimum image violations?

Hello Justin,
I have put both the chains inside the box using -d 1.0 flag to gmx editconf. From this can I say that there is no minimum image violations?
Or should I check this using other tools?

Do any of the molecules in your system substantially change their dimensions (e.g. elongate) during the simulation? You can check for violations with gmx mindist -pi.

I have attached the plot obtained from gmx mindist. My box length is 5.91nm, but the minimum distance between periodic images are much less than the box length :(
So I guess I have to redo the simulation with larger box, right?

Minimum image violations only occur if the minimum periodic distance between atoms is less than the longest nonbonded cutoff. You have a bunch of frames that may have come close, but I don’t know what settings you’re using.

The coulomb cutoff and VDW cutoff both are set to 1.2 nm in the mdp file.

Then it appears you have a fair number of frames that are minimum image violations. The box size needs to be set to account for possible conformational change; if the starting structure might elongate, you have to plan ahead for that. This is probably why you can’t successfully re-image the trajectory. If the box is too small, then the images can’t be resolved properly.

Thank you so much Justin for your time and suggestions.
I will change the box size so that it can easily accommodate both chains and can take care of elongated structure. Along that I will use gmx mindist to check if there is minimum image violations (this is a vital technical thing I was not aware of).
I will get back to this thread if I have further issues regarding this.
Thanks again :)