Question about the NEB calculation

GULP
,
1

Dear Professor,

We have tried to construct GULP input for an NEB calculation (input file attached) using a simple two-atom system.

However, the energy profile looks like not what we expected to see and the output contains numerous NAN.

Please could you advice how to correct the input or better example?

opti conv neb

Options:

cartesian
Al core -1.000000 0.000000 0.000000 0.0000000 1.0000 0.0000
Al core 1.000000 0.000000 0.000000 0.0000000 1.0000 0.0000

cartesian
Al core -2.000000 0.000000 0.000000 0.0000000 1.0000 0.0000
Al core 2.000000 0.000000 0.000000 0.0000000 1.0000 0.0000

Nudged elastic band data

nebspring 0.0005
nebreplica 100

species 1
Al core 0.000000
polynomial
4
Al core Al core 65.000000 -96.000000 52.000000 -12.000000 &
1.000000 0.000 0.000 10.000
nebtangent 2
dump every 1 test_neb

Thank you.
Kind Regrads,
DG Kang.

2

Hi DG Kang,

The first thing to point out that while NEB is very popular in other codes, it’s not the best way to find a transition state in GULP. Given that GULP has analytic second derivatives for many force fields, it’s more efficient to use RFO and transition state optimisation (see example files). If you want to do something more in the NEB style to find a rough transition state (that can be improved with RFO/trans keyword) then I recommend using “sync” instead (which approaches the transition state from both sides) as in example79. Here is an example for your input that works:

opti conv sync
cartesian
Al core -1.000000 0.000000 0.000000 0.0000000 1.0000 0.0000
Al core 1.000000 0.000000 0.000000 0.0000000 1.0000 0.0000
fcartesian
Al core -2.000000 0.000000 0.000000 0.0000000 1.0000 0.0000
Al core 2.000000 0.000000 0.000000 0.0000000 1.0000 0.0000

nebreplica 2

species 1
Al core 0.000000

polynomial
4
Al core Al core 65.000000 -96.000000 52.000000 -12.000000 &
1.000000 0.000 0.000 10.000

NB: There is a change above relative to your input, which is why you probably get the NANs from NEB, in that you give 2 configurations, but sync and neb expect an initial configuration and a final one for the same configuration (note the use of “fcart” rather than “cart” to specify the final coordinates for the replica).
Hope that helps,
Julian