Practical considerations for transition state finding calculations
This page details a few practical considerations for obtaining accurate and reliable transition state energies using VASP.
NEB calculations
Restart with a better guess for the initial and final state configurations
Problem: While performing an NEB calculation, it might appear that there is a point on the band which has lower energy than that of the initial and final state configurations (corresponding to the structures that were placed in the 00 and 0x, where x is the number of IMAGES plus one).
Possible solutions
- . Relax the configurations with lower energy. These structures would then correspond to new initial and final configurations. Restart the NEB with these new configurations as the endpoints of the band.
- Make sure that the NEB was run with sufficient (and commensurate to the initial and final structures) accuracy. An important requirement in EDIFF, which governs the accuracy of the forces used in the NEB method.
Illustration TODO
Band becomes "floppy"
Problem The band between the IMAGES of the NEB calculation is non-monotonic and appears "floppy". This problem typically appears when one or more images are added to an already computed band.
Possible solutions
- This problem is likely caused by the value of SPRING being too small.
- Check if EDIFF is accurate enough for computation of the forces.
Band does not converge
Problem The calculation does not converge, i.e., the NEB method does not find a reasonable path between an initial and final state configuration. Sometimes a completely different path is found than the desired path.
Possible solutions
- Choose a different interpolation strategy to generate your initial guessed path
- Try the Improved dimer method instead
- Check if the numbering of the atoms is the initial and final configurations. Different ordering of atoms will lead to an unexpected path. Visualizing the initial interpolated path should help with diagnosing any issues related to incorrectly ordered atoms.
Checking convergence
Problem Unsure if the highest energy point in the NEB is the first order saddle point between the initial and final state configurations.
Possible solutions
- Check if there is exactly one imaginary mode by computing the second derivatives of the energy as implemented in IBRION.