UT theoretical chemistry code forum

Dear everyone,

I am new to NEB calculation.
I have a question about the initial and final configurations (two end points).

Say if I want to calculate the migration barrier of diffusion.
The two end structures with the diffusion species had been optimized.
In the initial and final structures, not only the coordinates of the diffusion species are different, but the structures of the network are also slightly different.

When generating the images, should I use "selective dynamics" to fix the diffusion atom(s) (one or more coordinates)?
And, does it make sense if I fix some of the network (F F F) atoms which have slightly different coordinates in the initial and final structures?

I have tried this way, if I am not mistaken, the barrier I got is simply too high.
0 0.000000 0.000000 -0.008545 0
1 1.928959 2.029859 -1.548028 1
2 3.901888 2.179661 -1.652672 2
3 5.931448 45.611082 -1.929207 3
4 7.167782 2.126470 0.825865 4
5 8.319538 0.593718 0.016019 5
6 9.469748 -0.078990 0.032283 6

I have also tried by replacing the network of final structure with initial one, except the diffusion molecule.
This yield a more reasonable barrier height, but the energy of the final point is much higher than it supposes to be.
0 0.000000 0.000000 -0.009639 0
1 1.083094 0.676768 -0.292703 1
2 2.280405 2.275954 0.316236 2
3 4.105170 2.210667 0.265655 3
4 5.864108 2.219394 -0.085779 4
5 7.745210 2.215853 -1.443539 5
6 9.585642 1.050677 0.003718 6

Can someone comment on my approach, or make some suggestions?
Thank you very much.

Bin

You can let everything relax in the band calculation.

If you see any net translation, you can also free one atom in the same location in your initial and final structures. This can be done without any loss of generality, and it will prevent translation along the calculated minimum energy path.

The problem with the high energy most likely has nothing to do with this. Check that your initial path makes sense (that atoms are not running into each other). If there are high initial forces, use a conservative optimizer such as IBRION=3 and POTIM=0.1 until the force drop below ~1eV/Ang before using a more aggressive optimizer that will converge more quickly.

Hi Dr. Henkelman,

Thank you for your reply. It helped a lot.

I tried the NEB calculation by relax everything (still fix the unit cell), the barrier height becomes more reasonable.
However, I found a local minimum for the middle image (I used 5 images). The energy of this local minimum is only slightly above the two end points, so that the neb.dat plot looks like twin peaks. I think this could be physical, and now I am doing another calculation with 9 images to see if there is any fine structure. Do you have any comments on this result?

Another question I have is about the OUTCAR for the two end points.
Should I use the same accuracy for NEB calculation and the two end points? (I used PREC=normal for NEB)
It is a bit strange that when I used PREC=normal for the two end points, they ended up with energies that are higher than some intermediate images.
When I used the results of PREC=high, the energies look much more reasonable.

My last question is:
If I got large forces by using even POTIM=0.1, should I just reduce it further? or if there is any other things that I can do?

Thank you.