Issue with Convergence Criteria upon Job Restart in DL-FIND/ORCA

1

Hi everyone,

I encountered an issue while running an optimization using DL-FIND with the ORCA interface. My job was terminated due to the time limit on the HPC. I restarted my job using the restart option in the QM theory part and saved my orbital file as save.gbw.

However, after restarting, the job is converging based only on the Max Grad and RMS Grad criteria from the first cycle, and it is not checking the energy convergence criteria. In my previous run, the last step did not converge in terms of energy, but now, upon restarting, it is not considering the energy, Max step, and RMS step criteria.

Is there a reason for this behavior, and how can I ensure that all convergence criteria (Energy, Max step, and RMS step) are considered upon restarting?

Regards,
Ram

2

Hi Ram,

This happens because the restart for the QM theory part is not related to restarting the geometry optimisation. To do this you need to use the DL-FIND restart option (making sure the dump option is set in the first run so that restart checkpoint data is saved). Without it DL-FIND is effectively restarting from scratch, so won’t have any previous step or energy data to consider.

Best wishes,

Tom