Symmetry mode analysis - min max

ross-colman

Hi Topas Users,

I'm attempting a refinement using symmetry mode analysis to try to solve a distorted structure, including site occupancy ordering (i.e. going from mixed A/B occupancy of all cation sites, to having A or B occupation of specific sites).

Whilst the symmetry mode description block exported from ISODISTORT has built-in min/max constraints on the specific mode amplitudes, the final occupation of each atomic site is a linear combination of modes and is therefore a dependent parameter. Whilst unconstrained some sites can end up having negative occupancy based on the linear combination of amplitudes.

Is it possible to assign min/max values to dependent parameters? My attempts so far haven't worked:

e.g. in the final occupation parameter definition going from this:
"prm Zr_1_occ = 0.5 + Zr_1_do;: 0.50000"
to this:
"prm Zr_1_occ = 0.5 + Zr_1_do min 0 max 1;: 0.50000"

I assume the problem is due to the parameter Zr_1_occ being a dependent parameter (unlike the mode amplitudes which are independent parameters).

Is there a way around this?

I can imagine a complex set of penalties that switch on IF I go below 0 or above 1 could work but was hoping for something simpler to implement.

Thanks in advance for any advice anyone has.
Kind regards,
Ross

johnsoevans

Dear Ross,

As you say, you can't put penalties on dependent parameters to my knowledge.

When we did similar things in the past (e.g. some of the Ln2O2ZnSe2 type materials) we used penalties of the type you suggest. I think we did a combination of conditional penalties on individual sites (e.g. if occ<0.25 set to zero) and/or penalties on the overall composition. There will obviously be some instabilities associated with conditional penalties. One we had the right answer we could remove the penalties.

John

ross-colman

Dear John,

Thank you kindly for your very rapid response.

I hadn't thought of penalties on the overall composition, despite it seeming obvious now!
I'll start with that and move on to more complex conditional penalties if necessary.

Many thanks,
Ross