GCMC Zeolite Simulation. Overlap of atoms crashing simulation.

  • mikewelsch
  • mikewelsch's Avatar Topic Author
  • Offline
  • New Member
  • New Member
3 years 10 months ago #602 by mikewelsch

I am trying to simulate the adsorption of methane into a zeolite structure containing sodium cations and I get inconsistently timed errors stating that a deletion was attempted but molecule energy was too high and the simulation is then aborted.

From my knowledge of Monte Carlo, shouldn't an overlap be rejected in the first place? Is there something I'm missing?

Attached are my input file, log file containing the error at the end, xyz of the final frame that contains the overlap, and a screenshot pointing out where the overlap is occurring.

Thanks for any and all help!
Micah Welsch
The following user(s) said Thank You: Diogets

Please Log in to join the conversation.

3 years 10 months ago #603 by ryangmullen
Hi Micah,

I would recommend using an NVT simulation to relax a configuration generated by add_to_config. You can then use the final configuration from your NVT simulation as the starting configuration for a GCMC simulation to add methane.

The maximum change in energy, delta.U, that can be evaluated by Cassandra is specified by the variable max_kBT. In global_variables.f90, max_kBT is set to 708. If deleting a particle lowers the energy by more than max_kBT, then the probability of the reverse move—which is in the denominator of the acceptance criterion—is zero and the simulation will throw an error.

add_to_config can result in these high energy clusters a couple ways because adding particles to a configuration via add_to_config doesn't follow detailed balance. (It can't, since add_to_config doesn't require a chemical potential of the particles being added.) add_to_config will reject additions that place atoms closer than Rcutoff_Low (1.0 per your input file), but this can still result in high energy configurations, if:

(a) Rcutoff_Low is too small. This will be the case if the interaction energy between two atoms at distance of Rcutoff_Low is greater than max_kBT.
(b) adding a large number of particles, even if the energy of atoms at Rcutoff_Low is less than max_kBT. To illustrate, consider the scenario where atom 20 is added very close to atom 1 (but still greater than Rcutoff_Low) and the interaction energy is 0.6*max_kBT. On a later step, atom 83 is also added near to atom 1, but on the other side from atom 20 and the interaction energy is also 0.6*max_kBT. If, in a gcmc simulation, it is attempted to delete atom 1, delta.U will be 1.2*max_kBT.

NVT simulations generally aren't as sensitive to these high energy clusters. In the example considered above, atoms 20 and 83 can be slowly moved away from atom 1 such that delta.U for any attempted translation is less than max_kBT.

Here's a few other points to consider:

(1) With a Charge_Style of NONE your simulation is not computing electrostatic contributions to the energy and the sodium atoms being added are not really ions.
(2) The zeolite is rigid, but you are still calculating Si-Si, O-O, and Si-O energies. You can set these interactions to zero in your clinopt.mcf but still preserve your Si-CH4, Si-Na, O-CH4, and O-Na interactions by using a custom mixing rule. This will eliminate a large constant energy from your simulation (543829205.019).

Hope this helps,

Please Log in to join the conversation.

Time to create page: 0.102 seconds