I have a question about bond breaking during minimization.

First, I made protein structure connecting two protein domains manually.

Then only problem was steric clash occured between residues near the protein domain interfaces.

After I minimized the protein, some residues separated from the protein structure although the steric clash were solved.

For example bond distances of carbon increased from 0.15nm to 0.25 nm.

Second, to solve this problem I made protein structure again only removing steric hinderance between the two domains.

Using this protein structure I minimized the model again and the output structure was fine.

To figure out this difference I check topology of first case structure and second case structure.

However, there was no difference between two topology files.

Thus, my question is why bonds breaking happen after minimization using first case sturcture.

(Of course, I know bond breaking is not possible during minimization.)

I guess that this because of steric clash of residues but I don't know how it possible in principle.

Is this the problem only happening in GROMACS?