The implemented algorithm to apply a non-satisfied constraint (fixed distance in this case) is a very primitive one. There is a harmonic potential added along the fixed coordinate with a corresponding force constant that is increased with each iteration. You can beef up this force more quickly using, e.g.,
fixed_coord_force_constant = 2.0
However, in this case the problem may not be the arbitrary, additional force. This system has hydrogen bonds in a dimer - along with their angular low-energy motions. This will never be easy.
If you set
simple_step_scaling = True
then the restricted-step RFO will not be used. Instead the RFO step sizes will simply be scaled down. In my very limited testing this performed very marginally better than the ‘proper’ RS-RFO step in this case. Once the step sizes are small (<0.1au is initial default), neither restriction is in effect. I tried the Newton-Raphson step
step_type = NR
and it behaved very similarly. There is steepest-descent code which may be good as a last resort. I haven’t tried it out in awhile, and it will typically take many iterations.
In conclusion, you can try simple_step_scaling to see if the iterative solution for alpha in RS-RFO is the problem. And you can try N-R steps, to see if the RFO equations are the problem. And if you are still struggling to converge, then the problem is probably just searching a near planar subsurface with a quadratic method.