Dimer method using IOPT 1 breaks the adsorbate molecule

[sxd375]

Dr. Graeme

I have been running dimer for a transition state with LBFGS optimiser, with default values other than INVCURV = 0.005 (I read in the forum, it makes the optimiser rather conservative in taking steps). However, the adsorbate molecule breaks off and drifts apart in some 200th iteration leading to very high forces. I donot understand why. I have attached the dimer calculation here.

Please let me know what I am doing wrong in here.

P.S. also as I pointed out about LBFGS optimiser during dimer, the DIMCAR doesnot update/rotate for the second time even after 300 steps and as ever I get following DIMCAR(for other cases even though similar happened, but dimer converged still.)

Step Force Torque Energy Curvature Angle
1 0.18801 1.99353 -802.98099 -4.85432 1.72500
1 0.18801 1.15474 -802.98099 -4.95597 3.34210
1 0.18801 0.95404 -802.98099 -4.98347 1.63867

Thanks