The current voxelization strategy determines the occupation of a cell based on whether any point of the structure's atom point cloud appears in it.
Though this is fine for most cases, it is possible for atoms to be placed in a single cell, even though its effective radius intersects one or more neighboring cells.
From this, the risk arises that two structures are placed next to each other in such a manner than when their placements are rendered out, two or more of their atoms are very close to each other.
In other words: each point is currently voxelized to precisely one cell, but a more cautious---perhaps more correct---voxelization would mark one to at most four cells as occupied for a single point.
Such a voxelization procedure can be considered to be more conservative.
Is this necessary?
Quite possibly, a subsequent energy minimization is sufficient to resolve these cases.
Since this is a ubiquitous part of any molecular dynamics procedure, we may be content with our current voxelization strategy.
In order to decide whether we need to consider a conservative voxelization scheme, we should test the performance current procedure.
To this end, we can come up with the following experiment:
A number of different systems are packed according to a range of parameters:
Packing density
Resolution
Size
A collision check is performed on the rendered structures.
This should point to any atoms that are too close for comfort directly after the packing.
It will also provide insight into the prevalence of the supposed issue for the different packing parameters.
An energy minimization is performed on each of the systems.
Another collision check is done to count the remaining collisions.
From these results, we can make a more informed decision on this matter.
If we end up deciding that this is necessary, it will also give us information toward choosing a conservative voxelization strategy and what parameters give appropriate results.
Finally, the experiment described above may live on as an integration test.
The current voxelization strategy determines the occupation of a cell based on whether any point of the structure's atom point cloud appears in it. Though this is fine for most cases, it is possible for atoms to be placed in a single cell, even though its effective radius intersects one or more neighboring cells. From this, the risk arises that two structures are placed next to each other in such a manner than when their placements are rendered out, two or more of their atoms are very close to each other. In other words: each point is currently voxelized to precisely one cell, but a more cautious---perhaps more correct---voxelization would mark one to at most four cells as occupied for a single point. Such a voxelization procedure can be considered to be more conservative.
Is this necessary?
Quite possibly, a subsequent energy minimization is sufficient to resolve these cases. Since this is a ubiquitous part of any molecular dynamics procedure, we may be content with our current voxelization strategy.
In order to decide whether we need to consider a conservative voxelization scheme, we should test the performance current procedure. To this end, we can come up with the following experiment:
From these results, we can make a more informed decision on this matter. If we end up deciding that this is necessary, it will also give us information toward choosing a conservative voxelization strategy and what parameters give appropriate results. Finally, the experiment described above may live on as an integration test.