In the case of hard SCF and Relaxation process for periodic systems we use different tricks using Quantum Espresso. I think that two of them would be good if it has not already implemented in FF.
It would be easy to implement in FF code a parameter to be read as electron convergence threshold at the beginning of any relaxation process which is difficult to get convergence. Then when the structure becomes near its local minimum the code automatically uses the tighter SCF convergence threshold.
As an example suppose that the Relaxation steps, SCF convergence threshold, and UPSCALE were 40, 10**-6, and 100 respectively, the code would use (10**-6)/100 as SCF convergence threshold til say 38th step and would use the (10**-6) for the last 2 steps.
2. MIXING value
In DFT calculations at each relaxation step Quantum Espresso mixes the electron density of a structure with the corresponding one of the previous step to find a new guess for the next step. Sometimes it would be useful that we manually set the value of each part in such mixing.
International School for Advanced Studies (SISSA)
Sharif University of Technology