Non-Bonded parameters definition not working

I am running equilibration MD on a prototypical system. It is rectangular block of particles which are closely spaced. Earlier I ran the MD simulation of a ligand movement through this block using gromacs 2018 and calculated the interactions on cpu, which resulted in slow performance. The significant mdp settings that I used are the following:

energygrps = FUL HCP SOL LIG
energygrp_excl = HCP HCP LIG LIG
rlist = 1.20
cutoff-scheme = group

; OPTIONS FOR ELECTROSTATICS AND VDW
; Method for doing electrostatics
coulombtype = PME-Switch
rcoulomb-switch = 0.9
rcoulomb = 1.0
; Dielectric constant (DC) for cut-off or DC of reaction field
epsilon-r = 1
; Method for doing Van der Waals
vdw-type = Switch
; cut-off lengths
rvdw-switch = 0.9
rvdw = 1.0
freezegrps = HCP LIG
freezedim = Y Y Y Y Y N

Using this setting I could disable non-bonding interaction calculations between the rectangular block atoms (HCP group) and ligand atoms too as I did not want the ligand to change conformation

topol.top file:
[atomtypes]
OW OW 8 15.9994 0.000 A 2.5519e-03 2.5104e-06
HW HW 1 1.0079 +0.520 A 0.0 0.0
MW MW 0 0.0000 -1.040 D 0.0 0.0
CF CF 6 12.0110 0.000 A 0.0 3.7326e-06 ;2.03e-03 3.7326e-06;crowder fullerene
CX CX 6 12.0110 0.000 A 2.03e-03 3.7326e-06 ;ligand fullerene
CW CW 6 12.0110 0.000 A 4.9183e-05 2.5197e-07
CP CP 6 12.0110 0.000 A 1.6394e-04 8.3992e-07

[nonbond_params]
; i j func c6 c12
OW CF 1 2.276e-03 3.061e-06

CF-atomtypes belong to a crowder species where I want it to interact repulsively with everything else except water, so I added a nonbond_params for water-crowder interaction

Now I am trying to run the same in gpu, I am using the following mdp:

;energygrps = FUL HCP SOL LIG
;energygrp_excl = HCP HCP LIG LIG
rlist = 1.20
cutoff-scheme = group

; OPTIONS FOR ELECTROSTATICS AND VDW
; Method for doing electrostatics
coulombtype = PME-Switch
rcoulomb-switch = 0.9
rcoulomb = 1.0
; Dielectric constant (DC) for cut-off or DC of reaction field
epsilon-r = 1
; Method for doing Van der Waals
vdw-type = Switch
; cut-off lengths
rvdw-switch = 0.9
rvdw = 1.0
freezegrps = HCP LIG
freezedim = Y Y Y Y Y N

I commented the energygrp and energygrp-excl as the exclusion is not supported in gpu, so I changed the topol.top file to get rid of non-bonding interactions between HCP atoms which are CW and CP and LIG atoms which are CX:

topol.top file:
[atomtypes]
OW OW 8 15.9994 0.000 A 2.5519e-03 2.5104e-06
HW HW 1 1.0079 +0.520 A 0.0 0.0
MW MW 0 0.0000 -1.040 D 0.0 0.0
CF CF 6 12.0110 0.000 A 0.0 3.7326e-06 ;2.03e-03 3.7326e-06;crowder fullerene
CX CX 6 12.0110 0.000 A 2.03e-03 3.7326e-06 ;ligand fullerene
CW CW 6 12.0110 0.000 A 4.9183e-05 2.5197e-07
CP CP 6 12.0110 0.000 A 1.6394e-04 8.3992e-07

[nonbond_params]
; i j func c6 c12
OW CF 1 2.276e-03 3.061e-06
CW CW 1 0.000 0.000
CP CP 1 0.000 0.000
CX CX 1 0.000 0.000

Now when I am running a NVT equilibration I see that the total and potential energy is in the order of +10^7, extremely high, I checked with mdrun -rerun and found that HCP-HCP energy is in the same order during the simulation with gpu and new settings

Now I am not understanding that even after using the nonbond_params 0 for C6 and C12 for CP, CW, CX. The non-bonding interactions between these pairs is still non-zero. What should I do to get rid of this issue and run the MD such that these atoms are non-interacting like I expected.

Any help would be much appreciated, thank you

With you nonbond_params section you only set the CP-CP and CX-CX interactions to zero. CP and CX with anything else is not zero. All combinations not specified in nonbond_params are computed using C6 and C12 in atomtypes.

Thank you for your reply Dr. Hess, I checked and now after setting other individual interactions like CP-CW, etc. I don’t see that large positive energy building up any more.