Published May 30, 2023
| Version 1.0
Report
Open
Scaling up and coupling adaptive kinetic Monte Carlo with large-scale DFT
Authors/Creators
- 1. School of Chemistry, University of Southampton
- 2. School of Chemistry, University of Southampton; Scientific Computing Department, STFC-UKRI, Rutherford Appleton Laboratory
- 3. Scientific Computing Department, STFC-UKRI, Daresbury Laboratory
Description
An interface was written to combine the functionality of an adaptive kinetic Monte Carlo program (ACDC) and an ab initio electronic structure program (ONETEP). The merger of the two programs was motivated by the limitations in empirical potentials and allowed for the exploration of the potential energy surface of adatoms on the MgO(001) and Al(001). We examined two schemes for optimising the utilisation of the ARCHER2 platform. Due to the nature of the transition state searches (differing number of force evaluations), task farming was observed to be the most efficient. In addition, this approach is more fault tolerant and more extensible in the longer term.
Notes
This work was funded under the embedded CSE programme of the ARCHER2 UK National
Supercomputing Service (http://www.archer2.ac.uk).