AKMC

Adaptive kinetic Monte Carlo (AKMC) is an established method for simulating kinetic evolution through transitions between states. The implementation available through SODA uses EON-derived components and provides core AKMC functionality only, primarily for benchmarking and framework testing. Advanced features such as basining (superbasin handling) are not implemented. These tutorials cover controlled example runs, execution checks, and performance comparisons using SODA’s parallel execution and data-collection infrastructure.

For AKMC simulation runs, consult the current EON documentation and its AKMC guide.

Start with installation and SODA basics.

Examples

Begin with the single-adatom EAM example. It uses the same small system as the shared run-preparation walkthrough.

Example

What you will do

Requirements

First diffusion run with EAM

Run the lightweight Pd adatom benchmark with its supplied settings and check the expected 36 states and four processes per state. Also introduces the larger island example.

AKMC-enabled build, CPU cores, and the bundled Pd EAM potential.

AKMC with GAP

Prepare the bundled Pd/O GAP model and run a short example with threaded CPU workers.

AKMC build with LAMMPS/QUIP and OPENMP support.

AKMC with MACE

Configure native MACE forces for two oxygen atoms on Pd(110) and launch GPU workers.

Native MACE-enabled AKMC build, a compatible GPU/APU, and an external model file.

Inputs live in examples/akmc/. Each walkthrough copies them into a separate run directory so you can adapt the calculation without editing the examples.

Tip

Optimize your starting structure with the potential or model selected for the run. Preserve its cell, periodic boundaries, and fixed atoms, and save it as pos.con or pos.xyz. Re-optimize when changing the potential.

The EAM walkthrough retains the supplied 20,000 KMC steps. The more expensive GAP and MACE walkthroughs start with a short setup check; each KMC step can require multiple saddle searches before the confidence target is reached.

Settings and analysis

  • AKMC settings: concise explanations of confidence, thermal windows, searches, and SODA runtime options, including differences from native EON behavior.

  • Analyzing AKMC output: inspect states and processes, read the KMC transition history and simulated time, and export structures.