Regarding the different optimizers in CI-NEB

[Kester]

Dear all,

I would like to use the CI-NEB method to obtain the activation barrier for a molecule dissociation.
From the converged CI-NEB calculation (2 images), the two images I obtained are rather similar to my reactant (image-01) and product (image-02). Thus, I could not obtain the saddle.

Before the CI-NEB calculation, I used EDIFF = 1E-05 and EDIFFG = -0.013 to relax my reactant and product structures.
Could this be due to the reduced convergence criteria in my CI-NEB INCAR (EDIFF = 1E-04 and EDIFFG = -0.02)?

I have attached the output files for reference.

Based on the previous post(s) in this forum, if I understand correctly, IOPT=3 and IOPT=7 are efficient for saddle identification, whereas IOPT=1 and IOPT=2 give more accurate representation of the MEP? For this reason, and also the limited computational resources, I used IOPT=7 in my calculation. However, using the same IOPT=7 optimizer, my other structures could not converge after some 900 NSW steps.

Thanks in advance.

[graeme]

The problem here is unrelated to the optimizer. You have a long path which is no longer continuous.

Try making a new band with 5 or 7 images. Use ibrion 3 or 7 until the forces drop to about 1 eV/Ang. Check that you have a continuous path (i.e. that the atoms do not jump around discontinuously between images). Then you can switch to IOPT=1 or 2 for more rapid convergence.

The choice of optimizer does not determine the accuracy of the minimum energy path.

[Kester]

Dear Prof. Graeme,

Thank you for the feedback.
I have shortened the path with a new set of 5 images, and I believe the structures/atoms are continuous.
The images were computed with IOPT7 with IBRION3 until the forces converge to 1eV/Ang, no hiccups up to this point.

Following the initial convergence, I then switched to IOPT1 with IBRION3 for the second CI-NEB run.
However, the calculation did not converge, after some 300 NSW steps.

Is the path not continuous enough? Or is there problem with the INCAR parameter settings?
Attached is it calculation for reference.

Regards,
Kester