Hello,
I am using compute pair/local to dump distances between pair of atoms. I created a group with atom type as follows:
group AB type 1 2
and use the computes (simplified version)
compute 1 AB property/local ptype1 ptype2
compute 2 AB pair/local dist
dump dump1 AB local 200000 AB.dat c_1[1] c_1[2] c_2
When inspecting the AB.dat file I see pairs of bonded atoms (1-2 interaction) listed. I thought this would not happen because the weighting factor for 1-2 pairwise interactions are set to zero by the special_bonds command in my script file:
special_bonds lj/coul 0.0 0.0 0.5
However I am using the hybrid option for pair style
pair_style hybrid lj/cut/coul/long 15.0 15.0
and I am wondering whether this is causing the special_bonds command to be ignored.
Other insight on this issue appreciated.
Thank you.
What else is part of your pair hybrid? It’s possible
that if you have other pair styles listed which use
different neighbor lists, that the 1-2 interactions
are coming from that.
Steve
That is the only pair style. Additional lines in the script file are as follows:
pair_style hybrid lj/cut/coul/long 15.0 15.0 # cutoffs for LJ and Coul
bond_style hybrid harmonic # stretching
angle_style hybrid harmonic # bending
dihedral_style hybrid opls # torsion
improper_style none # out of plane angle
pair_coeff 3 3 lj/cut/coul/long 0.07 3.55 # CA-CA
pair_coeff 5 5 lj/cut/coul/long 0.03 2.42 # HA-HA
pair_coeff 2 2 lj/cut/coul/long 0.07 3.55 # CA(O)-CA(O)
pair_coeff 1 1 lj/cut/coul/long 0.17 3.07 # OH-OH
pair_coeff 4 4 lj/cut/coul/long 0.0 0.0 # HO-HO
bond_coeff 1 harmonic 553.0 0.945 # OH-HO
bond_coeff 2 harmonic 450.0 1.364 # OH-CA(O)
bond_coeff 3 harmonic 469.0 1.4 # CA-CA(O)
bond_coeff 4 harmonic 367.0 1.08 # CA-HA
angle_coeff 1 harmonic 35.0 113.0 # HO-OH-CA(O)
angle_coeff 2 harmonic 70.0 120.0 # OH-CA(O)-CA
angle_coeff 3 harmonic 63.0 120.0 # CA-CA(O)-CA
angle_coeff 4 harmonic 63.0 120.0 # CA(O)-CA-CA(O)
angle_coeff 5 harmonic 35.0 120.0 # CA(O)-CA-HA
dihedral_coeff 1 opls 0.0 1.682 0.0 0.0 # HO-OH-CA(O)-CA
dihedral_coeff 2 opls 0.0 7.25 0.0 0.0 # OH-CA(O)-CA-CA(O)
dihedral_coeff 3 opls 0.0 7.25 0.0 0.0 # OH-CA(O)-CA-HA
dihedral_coeff 4 opls 0.0 7.25 0.0 0.0 # CA-CA(O)-CA-CA(O)
dihedral_coeff 5 opls 0.0 7.25 0.0 0.0 # CA-CA(O)-CA-HA
etc...
Also when I create RDF data, I see peaks coming from bonded atoms both
(1-2) and (1-3). The (1-4) interaction peaks should be there because
of
special_bonds lj/coul 0.0 0.0 0.5
Thanks in advance for inputs.
> What else is part of your pair hybrid? It's possible
> that if you have other pair styles listed which use
> different neighbor lists, that the 1-2 interactions
> are coming from that.
>
> Steve
That is the only pair style. Additional lines in the script file are as
follows:
there is no need to use a hybrid style for any of those interactions.
you'd make your life much easier without.
pair_style hybrid lj/cut/coul/long 15.0 15.0 # cutoffs for LJ and Coul
bond_style hybrid harmonic # stretching
angle_style hybrid harmonic # bending
dihedral_style hybrid opls # torsion
improper_style none # out of plane angle
pair_coeff 3 3 lj/cut/coul/long 0.07 3.55 # CA-CA
pair_coeff 5 5 lj/cut/coul/long 0.03 2.42 # HA-HA
pair_coeff 2 2 lj/cut/coul/long 0.07 3.55 # CA(O)-CA(O)
pair_coeff 1 1 lj/cut/coul/long 0.17 3.07 # OH-OH
pair_coeff 4 4 lj/cut/coul/long 0.0 0.0 # HO-HO
bond_coeff 1 harmonic 553.0 0.945 # OH-HO
bond_coeff 2 harmonic 450.0 1.364 # OH-CA(O)
bond_coeff 3 harmonic 469.0 1.4 # CA-CA(O)
bond_coeff 4 harmonic 367.0 1.08 # CA-HA
angle_coeff 1 harmonic 35.0 113.0 # HO-OH-CA(O)
angle_coeff 2 harmonic 70.0 120.0 # OH-CA(O)-CA
angle_coeff 3 harmonic 63.0 120.0 # CA-CA(O)-CA
angle_coeff 4 harmonic 63.0 120.0 # CA(O)-CA-CA(O)
angle_coeff 5 harmonic 35.0 120.0 # CA(O)-CA-HA
dihedral_coeff 1 opls 0.0 1.682 0.0 0.0 # HO-OH-CA(O)-CA
dihedral_coeff 2 opls 0.0 7.25 0.0 0.0 # OH-CA(O)-CA-CA(O)
dihedral_coeff 3 opls 0.0 7.25 0.0 0.0 # OH-CA(O)-CA-HA
dihedral_coeff 4 opls 0.0 7.25 0.0 0.0 # CA-CA(O)-CA-CA(O)
dihedral_coeff 5 opls 0.0 7.25 0.0 0.0 # CA-CA(O)-CA-HA
etc...
Also when I create RDF data, I see peaks coming from bonded atoms both
(1-2) and (1-3). The (1-4) interaction peaks should be there because
of
how do you know those are bonded atoms and not interactions between
different molecules.?
axel.
> What else is part of your pair hybrid? It's possible
> that if you have other pair styles listed which use
> different neighbor lists, that the 1-2 interactions
> are coming from that.
>
> Steve
That is the only pair style. Additional lines in the script file are as
follows:
there is no need to use a hybrid style for any of those interactions. you'd
make your life much easier without.
Right. This is in preparation for a more complicated model in the
future. But if this is giving me the present troubles, I will drop it.
pair_style hybrid lj/cut/coul/long 15.0 15.0 # cutoffs for LJ and Coul
bond_style hybrid harmonic # stretching
angle_style hybrid harmonic # bending
dihedral_style hybrid opls # torsion
improper_style none # out of plane angle
pair_coeff 3 3 lj/cut/coul/long 0.07 3.55 # CA-CA
pair_coeff 5 5 lj/cut/coul/long 0.03 2.42 # HA-HA
pair_coeff 2 2 lj/cut/coul/long 0.07 3.55 # CA(O)-CA(O)
pair_coeff 1 1 lj/cut/coul/long 0.17 3.07 # OH-OH
pair_coeff 4 4 lj/cut/coul/long 0.0 0.0 # HO-HO
bond_coeff 1 harmonic 553.0 0.945 # OH-HO
bond_coeff 2 harmonic 450.0 1.364 # OH-CA(O)
bond_coeff 3 harmonic 469.0 1.4 # CA-CA(O)
bond_coeff 4 harmonic 367.0 1.08 # CA-HA
angle_coeff 1 harmonic 35.0 113.0 # HO-OH-CA(O)
angle_coeff 2 harmonic 70.0 120.0 # OH-CA(O)-CA
angle_coeff 3 harmonic 63.0 120.0 # CA-CA(O)-CA
angle_coeff 4 harmonic 63.0 120.0 # CA(O)-CA-CA(O)
angle_coeff 5 harmonic 35.0 120.0 # CA(O)-CA-HA
dihedral_coeff 1 opls 0.0 1.682 0.0 0.0 # HO-OH-CA(O)-CA
dihedral_coeff 2 opls 0.0 7.25 0.0 0.0 # OH-CA(O)-CA-CA(O)
dihedral_coeff 3 opls 0.0 7.25 0.0 0.0 # OH-CA(O)-CA-HA
dihedral_coeff 4 opls 0.0 7.25 0.0 0.0 # CA-CA(O)-CA-CA(O)
dihedral_coeff 5 opls 0.0 7.25 0.0 0.0 # CA-CA(O)-CA-HA
etc...
Also when I create RDF data, I see peaks coming from bonded atoms both
(1-2) and (1-3). The (1-4) interaction peaks should be there because
of
how do you know those are bonded atoms and not interactions between
different molecules.?
I compare the RDF with the histogram of the data I dump with
property/local and pair/local commands. Also I dump custom data for
the molecule ids for each atom. After filtering and comparing data of
intermolecular interactions I can tell what peaks in the RDF are from
intramolecular interactions. The (1-2) and (1-3) intramolecular
interactions are included in the RDF.
I think I will run without the hybrid option and compare.
Thanks,
>
>
>
>>
>> > What else is part of your pair hybrid? It's possible
>> > that if you have other pair styles listed which use
>> > different neighbor lists, that the 1-2 interactions
>> > are coming from that.
>> >
>> > Steve
>>
>> That is the only pair style. Additional lines in the script file are as
>> follows:
>>
>
> there is no need to use a hybrid style for any of those interactions.
you'd
> make your life much easier without.
Right. This is in preparation for a more complicated model in the
future. But if this is giving me the present troubles, I will drop it.
>
>
>>
>> pair_style hybrid lj/cut/coul/long 15.0 15.0 # cutoffs for LJ and
Coul
>>
>> bond_style hybrid harmonic # stretching
>> angle_style hybrid harmonic # bending
>> dihedral_style hybrid opls # torsion
>> improper_style none # out of plane angle
>>
>> pair_coeff 3 3 lj/cut/coul/long 0.07 3.55 # CA-CA
>> pair_coeff 5 5 lj/cut/coul/long 0.03 2.42 # HA-HA
>> pair_coeff 2 2 lj/cut/coul/long 0.07 3.55 # CA(O)-CA(O)
>> pair_coeff 1 1 lj/cut/coul/long 0.17 3.07 # OH-OH
>> pair_coeff 4 4 lj/cut/coul/long 0.0 0.0 # HO-HO
>>
>> bond_coeff 1 harmonic 553.0 0.945 # OH-HO
>> bond_coeff 2 harmonic 450.0 1.364 # OH-CA(O)
>> bond_coeff 3 harmonic 469.0 1.4 # CA-CA(O)
>> bond_coeff 4 harmonic 367.0 1.08 # CA-HA
>>
>> angle_coeff 1 harmonic 35.0 113.0 # HO-OH-CA(O)
>> angle_coeff 2 harmonic 70.0 120.0 # OH-CA(O)-CA
>> angle_coeff 3 harmonic 63.0 120.0 # CA-CA(O)-CA
>> angle_coeff 4 harmonic 63.0 120.0 # CA(O)-CA-CA(O)
>> angle_coeff 5 harmonic 35.0 120.0 # CA(O)-CA-HA
>>
>> dihedral_coeff 1 opls 0.0 1.682 0.0 0.0 # HO-OH-CA(O)-CA
>> dihedral_coeff 2 opls 0.0 7.25 0.0 0.0 # OH-CA(O)-CA-CA(O)
>> dihedral_coeff 3 opls 0.0 7.25 0.0 0.0 # OH-CA(O)-CA-HA
>> dihedral_coeff 4 opls 0.0 7.25 0.0 0.0 # CA-CA(O)-CA-CA(O)
>> dihedral_coeff 5 opls 0.0 7.25 0.0 0.0 # CA-CA(O)-CA-HA
>>
>> etc...
>>
>> Also when I create RDF data, I see peaks coming from bonded atoms both
>> (1-2) and (1-3). The (1-4) interaction peaks should be there because
>> of
>
>
>
> how do you know those are bonded atoms and not interactions between
> different molecules.?
I compare the RDF with the histogram of the data I dump with
property/local and pair/local commands. Also I dump custom data for
the molecule ids for each atom. After filtering and comparing data of
intermolecular interactions I can tell what peaks in the RDF are from
intramolecular interactions. The (1-2) and (1-3) intramolecular
interactions are included in the RDF.
not 100% convincing. depending on the size of your system, the molecule
ids may be those of periodic images.
what is the distance of the offending peaks in the RDF? and what pairs of
interactions are you looking at?
axel.
[snip]
not 100% convincing. depending on the size of your system, the molecule ids
may be those of periodic images.
what is the distance of the offending peaks in the RDF? and what pairs of
interactions are you looking at?
O-H interaction: 0.98 A in the data file from pair/local compute.
The atoms are in the same molecule, specifically in the O-H bond of
the molecule.
I think this is happening b/c you are using long-range Coulombics. In that
case the 1-2 interactions are still in the neighbor list, b/c there is a contribution
to the energy/force that is computed in the pair style to combine with
the long-range contribution correctly. So compute pair/local sees the pair interaction
as well. Nothing to do with pair hybrid.
This should be documented on the compute pair/local doc page. Note
that the IMPORTANT NOTE there says “may not appear in the pair list”,
not “will not”. This case is the reason why.
I think this is happening b/c you are using long-range Coulombics. In that
case the 1-2 interactions are still in the neighbor list, b/c there is a
contribution
to the energy/force that is computed in the pair style to combine with
the long-range contribution correctly. So compute pair/local sees the pair
interaction
as well. Nothing to do with pair hybrid.
Thanks. Yes I tested not using hybrid and I get identical results.
This should be documented on the compute pair/local doc page. Note
that the IMPORTANT NOTE there says "may not appear in the pair list",
not "will not". This case is the reason why.
Got it. The additional statement in the doc page: "More specifically,
this may be true of I,J pairs with a weighting factor of 0.0;"
My mistake taking it as a true assertion.