I see that in your simulation files you have 2x1 tilings of your unit cell to create your simulation box. This results in small lattices with sites that neighbour with themselves. There are warnings about this printed in your general_output.txt files, e.g.:
Code:
***************
Warning code 806103 from lattice_builder_module: self-neighboring detected when constructing a lattice following the instructions in lattice_input.dat.
This can be normal for small lattices but manual checking is recommended.
More information:
Site 1 neighbors with itself.
***************
***************
Warning code 806104 from lattice_builder_module: possibly invalid lattice neighbor list in the lattice constructed following the instructions in lattice_input.dat.
This can be normal for small lattices but manual checking is recommended.
More information:
Encountered neighboring site 1 twice for site 1.
***************
These lead to problems when detecting, adding or removing patters (energetic clusters or elementary events). The algorithms are not designed to treat such cases, and one can get random behaviour, e.g. I was getting the following in the latest version (not yet released), no matter how large I was making the process participation array:
Code:
***************
Internal error code 801009 from kmc_simulation_handle_module: process participation list capacity (for dimension 2) reached when trying to add a lattice process in the list. Increase the capacity using the keyword override_array_bounds in the simulation input file.
More information:
Process participation list size (dimension 2): max_adsorprocparticipc_dim2 = 100000.
***************
After digging in a bit into how this happened, it appears that, at some point when species and processes are added to the lattice upon initialisation, an adsorbate appears with label 5 (which is invalid, since there are only 4 lattice sites), and then the expression nadsorprocparticip(p) gives a random result, since the index p = 5 exceeds the array bounds. Again, the root cause is the inability of the algorithms to treat very small lattice constructs (which are essentially not valid for KMC simulations).
I've tried the zacros301 simulation with a 2x2 lattice and it works fine; can you please redo the simulations with larger lattices and let me know if you see any issues?
As a remedy to this issue, we could consider upgrading the warning about the self-neighboring to an error in the next version of
Zacros. This way, the user will have to make the lattice bigger to proceed.
