NEB image has lower energy than the fixed initial point

[Dada]

I am doing NEB calculation on water splitting on certain surfaces, and after NEB convergence and looking at the results, I found out that the 1st image has a lower energy than the 0th image which is the fixed initial point. Is this reasonable or does this make sense, considering that the fixed points (0th and 7th points in a 6-image NEB calculation) should be the local minima of the path and should therefore have the lowest energy?

Thanks!

[graeme]

Check (carefully) that you are using the same settings for the NEB calculations as for the calculation of the endpoints. Then, yes, if any image shows an intermediate minimum, it is good practice to minimize that image. Also, make sure that ISYM=0 in case your initial state is in a high-symmetry state (although unlikely with water on surface).

[Dada]

Thank you very much for your reply Dr. Graeme! Can I clarify something, what do you exactly mean by minimizing that image? Does it mean that I need to relax the image? I checked my INCAR file and the only difference that I made was to lower the convergence criterion for the NEB run (I used EDIFFG = -0.01 for the relaxation of the fixed points, and then I used EDIFFG = -0.03 for the NEB run). By the way, what I did was I just replaced my fixed zeroth image with that image that has a lower energy (I relaxed it first before using it). To my surprise, I am getting a non-zero force on that new zeroth image. Is that reasonable, since I am thinking that forces on fixed images should be zero? I am currently using the LBFGS optimizer (I am also having a problem with regards to convergence).

[enola]

Hi Dada, I also have the similar problem, I find that some of the images even have lower energies than the 0th one (barrier is negative), but my 0th was fully relaxated before i do the NEB,and should be the minimum point.
is it because of the accuracy is not high enough, since i use EDGIGGF=-0.5, and meanwhile my initial state and final stateare quite similar.