CHARMM Development Project Forums Thread Like Summary - CHARMM constrained QM/MM dynamics crashes

I am trying to run a constrained walk simulation for committor analysis of transition path ensembles. I constrain the QM atoms as well as some of the residues
interacting with it using "cons harm force" and run the dynamics, and this calculation starts to become erratic within few steps and crashes. I have attached the
input and the output files here. I would appreciate any help that can make me understand this crash and fix it. Thanks!

I have to say that I've never used any of the QM/MM methods myself, so I cannot comment on the QM setup.

However, I've noticed a few things in the output log that may be relevant.

First, the CONS HARM command may need to follow the reading of the coordinates from ts254.crd, esp. if the restrained coordinates are significantly different from those in the initial coordinates read earlier in the script.

Second, at dynamics startup the message

**** Warning **** The following extraneous characters
were found while command processing in DCNTRL
TREF 300

indicates a failed attempt to establish a thermostat. That is the T setting for the weak coupling (Berendsen) thermostat, which is known to be flawed; it gives a T distribution which is significantly narrower than a proper Boltzmann distribution, and its usage is deprecated. The TCONS and TCOUPLE keywords are also required for that thermostat.

Third, I note there are no boundary conditions of any kind. While this may not a huge issue with the QM part, there are known artifacts from the particle:vacuum interface present for a system with no periodic boundary conditions. The more rigorous Hoover thermostat requires the use of periodic boundary conditions via CRYSTAL commands.