Indirect Umbrella Sampling INDUS and pressure control fix npt

LAMMPS LAMMPS Mailing List Mirror
1

Dear all,

I would like to run Indirect Umbrella Sampling (INDUS) in NPT simulation following the article of Patel Varilly Chandler and Garde (JStatPhys 2011). It is therefore necessary to include the forces coming from biaising potential to virial expression of internal pressure (Eq.17) in order to correctly simulate the NPT ensemble using Parrinello Rahman method.

In LAMMPS manual, for compute pressure command, it is written “If no extra keywords are listed, the entire equations above are calculated. This includes a kinetic energy (temperature) term and the virial as the sum of pair, bond, angle, dihedral, improper, kspace (long-range), and fix contributions to the force on each atom.” But I think that forces coming from an addforce fix are not accounted for in pressure computation, and that “fix” refers to SHAKE-like commands.

I have tested it on a simple example, of a single LJ atom in a cubic box, without any initial velocity. Compute pressure command prints a pressure equal to 0, even if I add a force through addforce fix. My version of LAMMPS has been downloaded the 22nd July 2016. Would you have any idea of how to implement this modification without rewriting the fix NPT command? In particular in the case of GPU package?

Thanks a lot,

Romain

Hereafter are my input and initial configuration files:

variable temperature equal 0.8

variable Ftype equal 1

variable Feps equal 1.0

variable Fsig equal 1.0

variable dumptime equal 1

variable thermotime equal 1

variable damp equal 100

processors * * *

units lj

boundary p p p

atom_style full

bond_style harmonic

angle_style harmonic

pair_style lj/cut 4.0

read_data data.input.txt

neighbor 2.0 bin

pair_coeff {Ftype} {Ftype} {Feps} {Fsig}

pair_modify mix arithmetic

group fluid type ${Ftype}

neigh_modify delay 5

neigh_modify one 4000

fix 0 fluid addforce 1.0 1.0 1.0

fix 1 fluid nvt temp {temperature} {temperature} ${damp}

compute temp_pres fluid temp

compute Pres all pressure temp_pres

dump dmp all custom ${dumptime} dump1.lammpstrj id type x y z vx vy vz

dump_modify dmp flush yes sort id format “%d d .6f .6f .6f .6f .6f %.6f”

thermo ${thermotime}

thermo_style custom step temp epair c_Pres c_Pres[1] c_Pres[2] c_Pres[3]

thermo_modify flush yes

run 10

write_data data.end1.txt

And initial configuration:

1 atoms

1 atom types

0 10 xlo xhi

0 10 ylo yhi

0 10 zlo zhi

Masses

1 1.0

Atoms

1 1 1 0 1 1 1

2

Dear all,

I would like to run Indirect Umbrella Sampling (INDUS) in NPT simulation
following the article of Patel Varilly Chandler and Garde (JStatPhys 2011).
It is therefore necessary to include the forces coming from biaising
potential to virial expression of internal pressure (Eq.17) in order to
correctly simulate the NPT ensemble using Parrinello Rahman method.

In LAMMPS manual, for compute pressure command, it is written “If no extra
keywords are listed, the entire equations above are calculated. This
includes a kinetic energy (temperature) term and the virial as the sum of
pair, bond, angle, dihedral, improper, kspace (long-range), and fix
contributions to the force on each atom.” But I think that forces coming
from an addforce fix are not accounted for in pressure computation, and
that “fix” refers to SHAKE-like commands.

I have tested it on a simple example, of a single LJ atom in a cubic box,
without any initial velocity. Compute pressure command prints a pressure
equal to 0, even if I add a force through addforce fix. My version of
LAMMPS has been downloaded the 22nd July 2016. Would you have any idea of
how to implement this modification without rewriting the fix NPT command?
In particular in the case of GPU package?

​a) if you want to modify LAMMPS, you should upgrade to the latest
development version (and keep it up-to-date), so that submitting your
extended code back to the developers is easier. also, you are more likely
to get help, if you are using the latest development version. no developer
will want to spend time doing testing with old versions.​

b) there is no need to worry about the GPU package. the GPU code only
applies to time consuming parts of the time step, i.e. select pair and
kspace styles. the rest of LAMMPS is used without any modification and is
run on the CPU (and in part run concurrently to the GPU styles).

c) there is no need to change fix npt. as you already noted, there is
already an infrastructure in place to tally virial data from fixes into the
global virial. what is missing is the code in fix addforce (and related
fixes) to support this and code added for fix_modify, so that it can be
requested to include the added force from given fixes to the virial. the
default, however, needs to remain to not do it, as there are many cases,
where it is undesired to include external force contributions in the virial.

axel.