\(\renewcommand{\AA}{\text{Å}}\)

fix external command

Syntax

fix ID group-ID external mode args
  • ID, group-ID are documented in fix command

  • external = style name of this fix command

  • mode = pf/callback or pf/array

    pf/callback args = Ncall Napply
      Ncall = make callback every Ncall steps
      Napply = apply callback forces every Napply steps
    pf/array args = Napply
      Napply = apply array forces every Napply steps

Examples

fix 1 all external pf/callback 1 1
fix 1 all external pf/callback 100 1
fix 1 all external pf/array 10

Description

This fix allows external programs that are running LAMMPS through its library interface to modify certain LAMMPS properties on specific timesteps, similar to the way other fixes do. The external driver can be a C/C++ or Fortran program or a Python script.


If mode is pf/callback then the fix will make a callback every Ncall timesteps or minimization iterations to the external program. The external program computes forces on atoms by setting values in an array owned by the fix. The fix then adds these forces to each atom in the group, once every Napply steps, similar to the way the fix addforce command works. Note that if Ncall > Napply, the force values produced by one callback will persist, and be used multiple times to update atom forces.

The callback function “foo” is invoked by the fix as:

foo(void *ptr, bigint timestep, int nlocal, tagint *ids, double **x, double **fexternal);

The arguments are as follows:

  • ptr = pointer provided by and simply passed back to external driver

  • timestep = current LAMMPS timestep

  • nlocal = # of atoms on this processor

  • ids = list of atom IDs on this processor

  • x = coordinates of atoms on this processor

  • fexternal = forces to add to atoms on this processor

Note that timestep is a “bigint” which is defined in src/lmptype.h, typically as a 64-bit integer. And ids is a pointer to type “tagint” which is typically a 32-bit integer unless LAMMPS is compiled with -DLAMMPS_BIGBIG. For more info please see the build settings section of the manual. Finally, fexternal are the forces returned by the driver program.

The fix has a set_callback() method which the external driver can call to pass a pointer to its foo() function. See the couple/lammps_quest/lmpqst.cpp file in the LAMMPS distribution for an example of how this is done. This sample application performs classical MD using quantum forces computed by a density functional code Quest.


If mode is pf/array then the fix simply stores force values in an array. The fix adds these forces to each atom in the group, once every Napply steps, similar to the way the fix addforce command works.

The name of the public force array provided by the FixExternal class is

double **fexternal;

It is allocated by the FixExternal class as an (N,3) array where N is the number of atoms owned by a processor. The 3 corresponds to the fx, fy, fz components of force.

It is up to the external program to set the values in this array to the desired quantities, as often as desired. For example, the driver program might perform an MD run in stages of 1000 timesteps each. In between calls to the LAMMPS run command, it could retrieve atom coordinates from LAMMPS, compute forces, set values in fexternal, etc.


To use this fix during energy minimization, the energy corresponding to the added forces must also be set so as to be consistent with the added forces. Otherwise the minimization will not converge correctly. Correspondingly, the global virial needs to be updated to be use this fix with variable cell calculations (e.g. fix box/relax or fix npt).

This can be done from the external driver by calling these public methods of the FixExternal class:

void set_energy_global(double eng);
void set_virial_global(double *virial);

where eng is the potential energy, and virial an array of the 6 stress tensor components. Eng is an extensive quantity, meaning it should be the sum over per-atom energies of all affected atoms. It should also be provided in energy units consistent with the simulation. See the details below for how to ensure this energy setting is used appropriately in a minimization.

Additional public methods that the caller can use to update system properties are:

void set_energy_peratom(double *eng);
void set_virial_peratom(double **virial);
void set_vector_length(int n);
void set_vector(int idx, double val);

These enable setting per-atom energy and per-atom stress contributions, the length and individual values of a global vector of properties that the caller code may want to communicate to LAMMPS (e.g. for use in fix ave/time or in equal-style variables or for custom thermo output.


Restart, fix_modify, output, run start/stop, minimize info

No information about this fix is written to binary restart files.

The fix_modify energy option is supported by this fix to add the potential energy set by the external driver to both the global potential energy and peratom potential energies of the system as part of thermodynamic output or output by the compute pe/atom command. The default setting for this fix is fix_modify energy yes. Note that this energy may be a fictitious quantity but it is needed so that the minimize command can include the forces added by this fix in a consistent manner. I.e. there is a decrease in potential energy when atoms move in the direction of the added force.

The fix_modify virial option is supported by this fix to add the contribution computed by the external program to both the global pressure and per-atom stress of the system via the compute pressure and compute stress/atom commands. The former can be accessed by thermodynamic output. The default setting for this fix is fix_modify virial yes.

This fix computes a global scalar, a global vector, and a per-atom array which can be accessed by various output commands. The scalar is the potential energy discussed above. The scalar stored by this fix is “extensive”. The global vector has a custom length and needs to be set by the external program using the lammps_fix_external_set_vector() and lammps_fix_external_set_vector_length() calls of the LAMMPS library interface or the equivalent call of the Python or Fortran modules. The per-atom array has 3 values for each atom and is the applied external force.

No parameter of this fix can be used with the start/stop keywords of the run command.

The forces due to this fix are imposed during an energy minimization, invoked by the minimize command.

Note

If you want the fictitious potential energy associated with the added forces to be included in the total potential energy of the system (the quantity being minimized), you MUST not disable the fix_modify energy option for this fix.

Restrictions

none

Default

none