Question Regarding CI-NEB Convergence Issue Using VTST Tools
Posted: Fri Jun 13, 2025 8:35 pm
Hi, I'm running a CI-NEB calculation using VASP and VTST tools, but I'm encountering convergence issues.
As shown in the attached plot, the ionic steps remain almost unchanged, and the maximum force is stuck at around 0.353 eV/Å. Moreover, instead of tracing a reaction path with decreasing energy, the image energies are increasing monotonically from the initial state (image 0) to the final image (image 6), which seems unphysical.
My setup:
IBRION = 3, IOPT = 3, IMAGES = 5
SPRING = -5, LCLIMB = .TRUE.
ENCUT = 500, EDIFF = 1E-5, EDIFFG not explicitly set
NSW = 1200, ISIF = 2, NELM = 100
I used linear interpolation for image generation and ran the standard VTST NEB workflow. nebef.pl shows little structural change between images and a rising energy trend. The climbing image seems not to be helping at this stage.
What might be the cause of this behavior? Should I regenerate the images with greater initial displacements, change the spring constant, or disable climbing until forces converge better?
Any advice would be appreciated!
As shown in the attached plot, the ionic steps remain almost unchanged, and the maximum force is stuck at around 0.353 eV/Å. Moreover, instead of tracing a reaction path with decreasing energy, the image energies are increasing monotonically from the initial state (image 0) to the final image (image 6), which seems unphysical.
My setup:
IBRION = 3, IOPT = 3, IMAGES = 5
SPRING = -5, LCLIMB = .TRUE.
ENCUT = 500, EDIFF = 1E-5, EDIFFG not explicitly set
NSW = 1200, ISIF = 2, NELM = 100
I used linear interpolation for image generation and ran the standard VTST NEB workflow. nebef.pl shows little structural change between images and a rising energy trend. The climbing image seems not to be helping at this stage.
What might be the cause of this behavior? Should I regenerate the images with greater initial displacements, change the spring constant, or disable climbing until forces converge better?
Any advice would be appreciated!