Increasing performance with exclude building neighbor list

Dear All,

I have a question about incresing calculation speed when using two different pairwise interaction potential with hybrid/overlay. In my example, I have some molecules , say, anthracene (C,H) on the surface of hBN (B,N) monolayer. For hBN-anthracene interaction I want to use ILP potentials and for anthracene-anthracene interaction (for intramolecular interactions also), LJ parameters from GAFF. I can set to zero for C-C, C-H in ILP parameters to neglect this interaction, but it always calculate normal vectors for each atom, eventually computational time is the same with one that I used only ILP potential for all interactions. Is there any trick which makes this process faster? I was going to use neigh_modify exclude molecule/inter command, but it neglects all interactions (which are calculated by LJ too) as I understood.
I appreciate any suggestions.

Here is a short example how I used:

pair_style hybrid/overlay ilp/graphene/hbn 15.0 1 lj/cut/coul/long 9.0 9.0
pair_modify tail yes
kspace_style pppm 0.0001

pair_coeff * * ilp/graphene/hbn BNCHO1.ILP B C H N
pair_coeff 3 3 lj/cut/coul/long  0.0150000   2.5996425
pair_coeff 2 2 lj/cut/coul/long  0.0860000   3.3996695
pair_modify mix arithmetic

Best Regards,

Please see this note from the ilp pair style:

This potential (ILP) is intended for interlayer interactions between two different layers of graphene, hexagonal boron nitride (h-BN) and their hetero-junction. To perform a realistic simulation, this potential must be used in combination with intralayer potential, such as AIREBO or Tersoff potential. To keep the intralayer properties unaffected, the interlayer interaction within the same layers should be avoided. Hence, each atom has to have a layer identifier such that atoms residing on the same layer interact via the appropriate intralayer potential and atoms residing on different layers interact via the ILP. Here, the molecule id is chosen as the layer identifier, thus a data file with the “full” atom style is required to use this potential.

Based on that note, I would conclude that your entire model is flawed and the pair style is not compatible with how you intend to use it. Thus, any discussions about performance are irrelevant.

Thank you for your quick reply.

I saw this example of ILP for water-graphene system.

pair_style  hybrid/overlay aip/water/2dm 16.0 lj/cut/tip4p/long 2 3 1 1 0.1546 10 8.5
pair_coeff  2 2   lj/cut/tip4p/long    8.0313e-3  3.1589      # O-O
pair_coeff  2 3   lj/cut/tip4p/long    0.0        0.0         # O-H
pair_coeff  3 3   lj/cut/tip4p/long    0.0        0.0         # H-H
pair_coeff  * *   aip/water/2dm        CBNOH.aip.water.2dm    C Ow Hw

This is the same potential but only difference is the way how it calculates normal vectors. It uses LJ for water-water interaction and others with ILP by excluding this interaction (set zero for these parameters in potential file). That was the reason that I thought it works.