Is G-SSNEB available now?

[yjhao]

Dear Prof. Graeme and others,

I would like to find a transition state with structures having different unit cell. I have found that your group published a paper about this one. Is it available in VTST tool? I found this address (http://theory.cm.utexas.edu/vtsttools/neb/), but it in not available. I am looking forward to your prompt reply! Thank you very much!

Sincerelys,
Yanjun Hao

[graeme]

Yes, the GSSNEB is included in our vtstcode. You can turn it on with LNEBCELL=.TRUE.

[yjhao]

which version is CINEB? How to set? Thank you very much!

[graeme]

LCLIMB=.TRUE.

[shilendra]

Dear All,

I am learning ssNEB on Fe system from hexagonal to BCC phase. I am doing calculation with VTST tools installed with VASP. I am facing two problem my system is running only on IOPT=3 tag if I am giving IOPT=2 and 1 then job is killing suddenly. Where is problem ?

Other thing is that I did ssNEB calculation with IOPT=3 with 5 images. I am getting saddle point but my second images geometry is much more stable than final geometry (less energy than final geometry). I have calculated initial and final geometry calculation separately.

Than You
Shilendra

[graeme]

1. We have not implemented a variable cell routine for IOPT=2, which is why that is not running. You should see a message in the OUTCARs: Must set IOPT = 3 or 7 when using LNEBCELL=.TRUE.

2. You need to have the KPOINT mesh when calculating the endpoints and the band. For your initial state you have 12x12x10 and for the band it is 12x12x12.

3. Run the calculation again starting from where you left off. For changes in the cell, you have effectively changed your plane wave cutoff. Once the band converges in a single iteration, you will have reached self-consistency.

4. Try relaxing (and/or looking closely) at image 02 to see if this is a stable state and perhaps equivalent to an endpoint. This can indicate a different reaction mechanism from the one that you anticipated in your initial guess.

[shilendra]

Dear Prof. Graeme and Penghao Xiao thank you for your help and support always !

I have some queries regarding ssNEB.

1. If initial and final geometries do not have equal number of atoms still we can do ssNEB calculation. How will do ssNEB calculation in that case ?

2. How does choose KPOINTS in ssNEB calculation as we have to give only 1 KPOINTS file for all geometries ?

3. Can we give ENCUT 1.3 time for better accuracy in ssNEB calculation. I have seen for Na system for 12*12*12 KPOINTS if I am giving default ENCUT then final geometry (Na BCC) is more stable and if I give 1.3 times ENCUT for same KPOINTS 12*12*12 then initial geometry (Hexagonal Na) is more stable. Why I do not know ?

4. I was reading other methods of ssNEB calculation I have query about DIMER method. How will prepare input in case of DIMER method if we know only one state.

5. Always we have to calculate initial and final geometry separately with usual tags in VASP ssNEB calculate only for number of images we give.
Any other necessary thing for doing ssNEB calculation

I sent you earlier Fe ssNEB calculation files my results were not right you gave me some suggestions. I am correcting calculations meanwhile I have done calculations for Na systems. I have attached all my results here. As per my understanding it seems right.
Please look at that and comment !

Thank You
Shilendra

[graeme]

That calculation does look reasonable. You might want to freeze one atom in the cell which should prevent any translation and give you a smoother spline interpolation.

And your questions:

1. No, the initial and final states must have the same number of atoms.

2. All images must have the same k-point mesh.

3. I'm a little suspicious about a switch in stability between these two states which you change from the 10^3 k-point grid and your ENCUT x1.3. If you are correct, however, you must choose settings for which the energy difference between the initial and final states is converged. Remember, you should not compare calculations with different settings. Systematically increase your K-point mesh and cutoff and look at the energy difference between the initial and final states and look for convergence. That will let you know which settings are appropriate for the ssNEB.

4. Penghao will be able to help with the ssDimer. We have not yet implemented the ssDimer directly into vasp and so this has to be done through the ASE interface.

5. Correct.

[xph]

SSDimer is only available in our TSASE package, not in the VTSTCODE. TSASE can invoke VASP as calculator too, but I strongly recommend trying an empirical potential first because SSDimer searches are really expensive.

See the page here for more details:
http://theory.cm.utexas.edu/tsase/ssdimer.html

Instructions on setting VASP calculator can be found here:
https://wiki.fysik.dtu.dk/ase/ase/calculators/vasp.html

[shilendra]

Dear Prof. Graeme thank you !

As you said if two geometries do not have equal number of atoms then we can not do ssNEB calculation.

1. My query is that If we able to make number of atoms equal in two geometries by making supercell of different dimensions in that case can we do ssNEB calculation ?

2. How to freeze one atom in the cell for smoother spline interpolation. Is there any tag or selective dynamics by doing true and false of atomic positions.

Thank You
Shilendra

[graeme]

Yes, you can make supercells of crystal structures with (for example) difference numbers of atoms per unit cell so that the supercells have the same number of atoms.

Freezing one atom will prevent translation, which increases the distance along the path and leads to oscillations in the spline interpolation. This is not a problem for the energies along the path.

[shilendra]

Dear Prof. Graeme in my earlier posts I sent you my ssNEB results for Na system. You looked that and told it is reasonable. In that calculation I took standard PBE Na pseudopotential (PBE_Na).

Now same calculation I tried with PBE Na_pv pseudopotential but I am not getting same trend of energy. Results are totally opposite and wrong from earlier calculation with standard Na pseudopotential.

I did earlier calculation on 12*12*12 KPOINTS mesh with standard PBE_Na pseudopotential. With same KPOINTS I did calculation for PBE_Na_pv pseudopotential. But I was getting opposite results then I increased KPOINTS mesh 20*20*20 still I am getting same results pattern.

I am not clear why results are coming different on changing pseodopotential type.
Na having ENCUT=100 eV (approx.)
Na_pv having ENCUT=260 eV (approx.

I have attached here all simulation files and graph.

Shilendra

[graeme]

A couple of suggestions. First, a small one, turn symmetry off. This will not change anything but it is much safer for these calculations.

More important, try restarting your calculation and even just do a single iteration from where it left off and see what energy profile you get. It appears that your cell has changed significantly during the optimization and you need to restart to get a plane wave basis that is consistent with the new cells. See if the intermediate minimum is still present and if so, minimize to see what it looks like. If the energy profile changes significantly, you can increase the plane wave cutoff to avoid such problems.

[shilendra]

Thank You Dear Prof. Graeme,

I will implement your suggestions. For restarting calculation I have to do CONTCAR to POSACR inside each image folder and ISTART=1.
Is this right way as usual we do for restart calculation.

Shilendra

[graeme]

Of you can run the "vfin.pl target_dir" which will move all the information from you calculation into the target_dir, copy CONTCAR to POSCAR and have thing set up for a continued calculation. This script will detect if you are doing an NEB or regular calculation and proceed accordingly.