ReaxFF-lg input

GULP
1

I tried to use the reaxff-lg with GULP, but my structures collapsed. To start, I inspected the library provided with 6.1.2 version, reaxff_lg.lib, where the -lg part is included as

Dispersion corrections

becke_johnson_c6 kcal
C core C core 0.0 3.8510 0.0 10.0
C core O core 632.0 3.6755 0.0 10.0
C core N core 650.0 3.7555 0.0 10.0
C core H core 0.0 3.3685 0.0 10.0
O core O core 624.0 3.5000 0.0 10.0
O core N core 880.0 3.5800 0.0 10.0
O core H core 0.0 3.1940 0.0 10.0
N core N core 1239.0 3.6600 0.0 10.0
N core H core 295.0 3.2730 0.0 10.0
cutp 10.0 p7
<<<<<<<<<<

The GULP manual/help entry for becke_johnson_c6 is:

becke_johnson_c6
Type Option
Format becke_johnson_c6 <intra/inter> <bond/x12/x13/x14/mol/o14/g14> <kcal/kjmol>
atom1 atom2 C6 r0 rmax <2flags>
Units C6 in eVAngs6, r0 in Angs
Default None
Use Specifies potential parameters for the Becke-Johnson form of damped C6 term.
The energy is of the form:
E = - C6/(r6 + r0**6)
<<<<<<<<

In other words, in the example above, for N-N interaction we have the C6=1239.0 and r0=3.6600 Angs.
The value of r0 in other sources is significantly lower, i.e. around 1.8 (see for example
dx.doi.org/10.1021/jp201599t | J. Phys. Chem. A 2011, 115, 11016–11022 ) where the values are C6=1239 but R0 is 1.83.

Can you please tell me why the R0 values are so different, and which one is correct for N?

Best regards
Luiza

2

Hi Luiza,
Thanks for your post. The ReaxFF-lg paper is actually somewhat ambiguous in regard to the C6 damping terms. Based on the symbols used you would be right to expect the r0 to be 1.83 Ang for N. However, the wording says that R_eij is the distance between the atoms at equilibrium & then says that the parameters R_e are the radii for the elements. Given that the distance is the sum of the radii for the pair of atoms this would give rise to the higher values. The GULP library goes with the words rather than the symbols.
As a test I ran the structure of alpha-N2 to see how the choices work. If you use the GULP library as it is then the agreement with experiment is reasonable. If you reduce the r0 to 1.83 Ang then the cell shrinks by over 20% & the agreement is poor. This makes me think that the GULP choice is the right one, though I admit that without seeing the original code it’s hard to be certain.
What I can say with reasonable certainty is that halving all the r0 parameters will make your structures more likely to collapse than with the current values & so this isn’t likely to be the source of your problems.
Regards,
Julian

3

Hi Julian,
thanks for the answer.
So, long story short, I need to use the sum of two atomic radii, not the equilibrium distance. Indeed, this answers all my questions.
Thank you, and best regards,
Luiza