-
Notifications
You must be signed in to change notification settings - Fork 4
Description
Hi,
I have a few questions/suggestions regarding the use of the dictionaries to specify the spacegroup sampling/constraints used structure generation/ranking methods. I am coming at this from the area of Organic CSP, but am unsure if the same queries/concerns would be applicable to other areas of CSP.
The first question regards specifying different extents of sampling in different spacegroups. Currently, when looking at the structure generation category it appears that you can provide a list of multiple sampled spacegroups via space_group_number_list, but when it comes to specifying a stopping criterion, stopping_criteria_max_structures_evaluated or _csp.random.number of samples are intended only to take a single number as their values, meaning it is difficult to simply specify a sampling scheme in which differnet numbers of structures are sampled in each spacegroup.
One solution could be to write a loop specifying each spacegroup(s) using followed by its corresponding max structures sampling, for example:
loop_
_csp.structure_generation.space_group_number_list
_csp.structure_generation.stopping_criteria
_csp.structure_generation.stopping_criteria_max_structures_evaluated
[1 61 19] "Max Stuctures Reached" 1000
[4 33] "Max Stuctures Reached" 2000
However, its not quite clear to me from the examples/documentation if this would be accepted within the conventions the community is seeking to develop. Is this a suitable way of specifying the sampling? Similarly, its perhaps not 100% clear from the descriptions whether (if a loop was not used) a single provided stopping_criteria_max_structures_evaluated would refer to the total across the spacegroups or the total in each spacegroup?
Secondly, in the structure ranking or geometry optimisation category it would be useful to be able to specify if symmetry is contrained during optimisation. This would affect whether the spacegroup can change during the optimisation. For example, specifying something like:
_geometry_optimisation.constrained_symmetry True
Lastly, when using workflows that do not constrain spacegroup symmetry at all ranking stages, it could be useful in a structure .cif to be able to specify the original spacegroup in which the structure was generated during the structure generation stage. E.g
_predicted_structure.initial_spacegroup 15
Many Thanks. Please let me know if anything is unclear or if you'd prefer me to pose these questions elsewhere