|
Kinetic Monte Carlo is a method for simulating the state-to-state dynamics of a
rare event system. For dynamics of atomic systems, the rare events correspond to
the thermal activation of atoms from one energy basin to another on the potential
energy surface. If the rates of these transitions are known, KMC can be used to
simulate dynamics over long time scales. The strength of KMC is that the fast
vibrational motion of atoms is averaged out, using transition state theory, so that
the dynamics can be modelled on the time scale of the thermally activated events.
KMC has some important limitations. The first is that the rates of thermal events
needs to be calculated before the simulation begins. This can be done using density
functional theory, for example, with the VASP code. For modeling solid systems, rates
can be found using harmonic transition state theory. This is a very computationally
efficient approximation because only the energy and frequencies of minima and saddle
points are needed to find rates of reactions.
Adaptive kinetic Monte Carlo is a method for simulating uses saddle point searches
to find processes available to the system. Kinetic Monte Carlo is
then used to advance the system from state to state.
Here, we present a package that can used to do adaptive kinetic Monte Carlo simulations in VASP.
Setup
Put the necessary scripts in place:
- Download the vtstscripts to get the scripts: akmc.pl, dimmins.pl and diminit.pl
Please report any bugs to xulijun-at-mail.utexas.edu.
- Three job handling scripts in the same folder: akmc_submit.pl, akmc_check.pl, and akmc_kill.pl
Note: Those file names can't be changed, and you need to write your own scripts since they are specific to the queuing system.
Start with a directory containing:
- config, the parameter file (see below for details)
- POSCAR, initial configuration
- KPOINTS, k-point mesh to sample the Brillouin zone
- POTCAR, pseudopotential file
- INCAR_min, parameters for minimization calculations
- INCAR_sp, parameters for saddle point finding calculations
- INCAR_dymmat, parameters for dynamical matrix calculations
- DISPLACECAR_sp, displacements for initializing dimer searches
- DISPLACECAR, displacements for dynamical matrix calculations
- akmc_min.sub, submission script for minimization jobs
- akmc_sp.sub, submission script for saddle point finding jobs
- akmc_dymmat.sub, submission script for dynamical matrix jobs
Run the script in this directory by typing the command: akmc.pl
Example job, and job related scripts for the SGE queueing system:
Configuration parameters
| Parameter |
Default Value |
Description |
| MaxJobs | 8 | Maximum number of job to run |
| NumSearches | 20 | Number of good saddle points to be found at each step or the number of searches in the dynamic stopping criteria |
| AkmcSteps | 100000 | Number of kMC steps |
| SimR | 0.1 | in Angstrom, to set up the dimer quenching |
| NumKT | 20 | in Kelvin, the allowed zone in the dynamic stopping criteria |
| Temperature | 300.0 | Temperature in Kelvin |
| SearchesAlgo | dimer | Saddle point finding algorithm: dimer or lanzcos |
| BarrierMAX | 10 | the highest barrier allowed (in eV) |
| JobFile | jobs.dat | a file storing jobs information at each step |
| StFile | st.dat | a file storing information for the state "st000x" |
| AkmcFile | akmc.dat | a file recording each KMC step information |
| DynMatFile | normalfreqs.dat | output file for DymMatrix.pl screen output |
| Prefactor | 1.0e12 | dynmat calculculation will be turned off if this line is not commented out |
| RateTableFile | RateTableFile.dat | the rate table file for each state |
| OldRunDirFile | stpool.dat | directories where those done states are located; if the is empty, the current akmc folder will be the only directory in which to check for possible repeated states |
| StEnergyFile | StEnergyFile.dat | a file containing energy for each initial state |
| Population | 0.2 | percentage of the lowest energy saddle point among all the saddles found |
| PrRecycleAll | 1 | if true, before starting a new state, check for quenched processes in previous states to recover good saddles (if any) |
| PrRecycle | 1 | if true, recycle good saddle points from the previous state |
| PrRecycleShift | 1 | if true, vector-shift good saddle points from the previous state |
| GrabQuenched | 1 | if true, check if a saddle has been quenched before in previous states |
| SurfaceRecShift | 1 | if true, check if the last atomic group is equivalent to the counterpart in previous states. if so, followed by vector-shift |
| ConvergenceAlgo | 1 | 1: dynamic criteria 0: fixed number of saddle searches 2: percentage control |
| equivalency | 1 | if true, using indistinguishablity in comparing the two states.
|
|
