anyuzx / lammps_brownian Goto Github PK

custom LAMMPS fix used for Brownian Dynamics Simulation (Overdamped Langevin)

C++ 100.00%

lammps_brownian's Introduction

🟢 Latest version of LAMMPS now contains official fix command for overdamped BD. See here.

This is the custom LAMMPS fix for overdamped Langevin dynamics(Brownian dynamics). LAMMPS has fix langevin for Langevin dynamics simulation. However, the algorithm LAMMPS use to integrate Langevin dynamics equation is Velocity-Verlet. This is not suitable for high friction(overdamped) simulation. Here are several files I modified from original fix_langevin.cpp and some other files to give LAMMPS a fix for overdamped Langevin dynamics simulation. This fix is only suitable for high friction case since velocity is overdamped. For more information and derivation of equation of motions, you can read this blog post as well https://www.guangshi.io/posts/simulating-brownian/

How to use

Build LAMMPS

Put fix_bd.cpp/fix_bd_baoab.cpp and their header files into src folder of LAMMPS.
Build LAMMPS.

Use it in input file

fix ID group-ID bd temperature damp seed

ID, group-ID
bd = style name of this fix command
temperature = desired temperature of run (temperature units)
damp = damping parameter (time units)
seed = random number seed to use for white noise (positive integer)

Examples

fix 1 all bd 1.00 0.01 324231

Remarks

For low friction, please use original fix langevin. As for what value is appropriate for high and low friction, one should do experiments themselves.
damp should be much smaller than one. This is to ensure the parameters are indeed for overdamped/high friction/low inertia regime. For instance (with LJ unit for simplicity), damp = 0.01 is sufficiently small for Brownian dynamics simulation. damp = 100 is too large to use this fix. Instead, just use fix langevin for a large value of damp constant.
❗ Only fix bd and fix bd/baoab works. Do not use OMP versions.

Other Options

fix bd/baoab: BAOAB algorithm for Brownian Dynamics simulation (see https://arxiv.org/abs/1203.5428).
~~fix bd/srk: SRK algorithm for Brownian Dynamics simulation.~~ (discarded)
~~fix bd/omp/fix bd/srk/omp/fix bd/baoab/omp: OMP version of Brownian Dynamics simulation.~~ (discarded)

Reference

Leimkuhler, Benedict, and Charles Matthews. "Rational construction of stochastic numerical methods for molecular sampling." Applied Mathematics Research eXpress 2013.1 (2012): 34-56.

lammps_brownian's People

Contributors

Stargazers

Watchers

Forkers

mdluffy666 jluchemlib noisyoscillator sottcap siyuchen94 aortiza pmadhikar jordanshivers chejunwei2

lammps_brownian's Issues

Mass dependence

I have a question about BD. As long as I know, Brownian Dynamics should be mass-independent. But, when I went through your codes, I found it seems your implementation is mass-dependent (See line 103 of fix_bd.cpp

        randf = sqrt(rmass[i]) * gfactor;
        x[i][0] += dtv * dtfm * (f[i][0]+randf*random->gaussian());
        x[i][1] += dtv * dtfm * (f[i][1]+randf*random->gaussian());
        x[i][2] += dtv * dtfm * (f[i][2]+randf*random->gaussian());

line 112

        dtfm = dtf / mass[type[i]];
        randf = sqrt(mass[type[i]]) * gfactor;
        x[i][0] += dtv * dtfm * (f[i][0]+randf*random->gaussian());
        x[i][1] += dtv * dtfm * (f[i][1]+randf*random->gaussian());
        x[i][2] += dtv * dtfm * (f[i][2]+randf*random->gaussian());

).
Have you checked your program with masses other than 1.0? I am not sure if this will work.

Issue with command unfix for the fix bd baoab

I had some issue (segmentation fault) when trying to use the unfix command on the fix bd baoab.
It seems to come from the fact that the memory associated to the quantity rnum cannot be freed.

I managed to solve the issue by adding to the cpp:

FixBDBAOAB::~FixBDBAOAB()
{
  delete random;
  memory->destroy(rnum);
  atom->delete_callback(id,0);
}

In the header file I also changedvirtual ~FixBDBAOAB() {} to virtual ~FixBDBAOAB()

compatibility with enforce2d

have you tested this with enforce2d? I'm finding that my atoms are moving in the z direction with the bd fix, even when I've got enforce2d

fix_bd.o error

HI, I am trying to simulate a chromatin system in the overdamped regime using BD simulation. I have uploaded the fix_bd and fix_bd_baob.cpp and the corresponding header files in the src folder. While building Lammps I am getting the following error:
Can you help me where it goes wrong .It seems force -> numeric is the issue. How to fix this and didnt get much help in google for this

../fix_bd.cpp:45:21: error: ‘class LAMMPS_NS::Force’ has no member named ‘numeric’
t_target = force->numeric(FLERR,arg[3]); // set temperature
^~~~~~~
../fix_bd.cpp:46:21: error: ‘class LAMMPS_NS::Force’ has no member named ‘numeric’
t_period = force->numeric(FLERR,arg[4]); // same as t_period in fix_langevin_overdamp.cpp
^~~~~~~
../fix_bd.cpp:47:17: error: ‘class LAMMPS_NS::Force’ has no member named ‘inumeric’
seed = force->inumeric(FLERR,arg[5]); //seed for random number generator. integer
^~~~~~~~
Makefile:109: recipe for target 'fix_bd.o' failed
make[1]: *** [fix_bd.o] Error 1
make[1]: Leaving directory '/home/vanitha/Downloads/lammps-29oct20/src/Obj_serial'
Makefile:195: recipe for target 'serial' failed
make: *** [serial] Error 2