When Si/Al = 1 there is no problem, because there is an obvious 50%-50% configuration that respect the Lowenstin rule.
When Si/Al > 1, I start from the 50:50 configuration and change back Al to Si: it should not be a problem
When Si/Al >> 1, I may start from the pure silica and add a few Al randomly: it should not be a problem
As for the cations, I start from the crystallographic positions and try to keep the ones with high partial occupancy first and then chose randomly some position with lower partial occupancy until the stochiometry is fine.
THE BIGGEST PROBLEM I see, is when I start:
from Si/Al = 1 randomly replacing Al and deleting cations
the +1 cations (e.g., Na) to get +2 cationic structure (e.g., Ca), and I remove randomly half of the cations and replace the others
The problem stem with the fact that doing this operation randomly may create larger charges because of a non-equilibrium structure.
To solve this I can:
run some NVT for the cations to equilibrate them: but I need a FF, and they may be kinetically trapped in experimental reality, not in their thermodynamic minimum position
don't go for a random deletion but delete cations that were close to replaced Al, or follow experimental evidences.[1]
[1] For example, regarding zeolite LTA, the Handbook of Zeolite Science and Technology (pag 108) says:
In zeolite NaA (|Na96| [Al96Si96O384]—LTA) (99), the 96 cations go first to the 64 6- rings where they are coordinated to three oxygen atoms. Then they go to 24 8-rings where they coordinate to two oxygen atoms and where they partially block the 8-ring windows. The final 8 sodium cations are in the front of the 4-rings also coordinated to two oxygen atoms. Figure 17 illustrates these coordination environments. When the Na+ cations are completely exchanged by Ca2+ to form CaA (|Ca48|
[Al96Si96O384]—LTA), the now 48 cations per unit cell go only to the 6-rings (100,101). There is a modest change in the framework structure that is observed in the change of the unit cell dimensions (from 24.555 Å to 24.47 Å) and in the geometry of the 6- and 8- rings
When Si/Al = 1 there is no problem, because there is an obvious 50%-50% configuration that respect the Lowenstin rule. When Si/Al > 1, I start from the 50:50 configuration and change back Al to Si: it should not be a problem When Si/Al >> 1, I may start from the pure silica and add a few Al randomly: it should not be a problem
As for the cations, I start from the crystallographic positions and try to keep the ones with high partial occupancy first and then chose randomly some position with lower partial occupancy until the stochiometry is fine.
THE BIGGEST PROBLEM I see, is when I start:
To solve this I can:
[1] For example, regarding zeolite LTA, the
Handbook of Zeolite Science and Technology(pag 108) says: