trouble fitting pattern that should be straightforward...

matsmcfly1

Hi all, I am posting up here as i am having trouble fitting a pattern that I think should be very straightforward. The pattern is pretty clean (only a few low intensity peaks attributed to minor phase), the pattern matches the calculated pattern (from cif) very well.

Pawley fit of the pattern is no problem at all, but a simple rietveld (rigid body, fixed coords) really struggles with intensities:

I cant fathom out why this is being so tricky ad its something I have encountered before in the past so thought I would use this instance as an excuse to ask the experts here.

P.S. I am happy to share the cif and pattern with anyone who is interested.

johnsoevans

There's obviously something significantly wrong with your structural model. This can be that it's just "a bad model", but it can often be relatively trivial things. Some examples are:

Wrong space group setting (e.g. you have the wrong origin choice specified in topas for the space group you're using). Perhaps specifiying P21/c when it's P21/n.
For trigonal structures having coordinates like 0.3333 instead of =1/3; (TOPAS interprets 0.3333 as being a general position).
Extreme preferred orientation (but looks unlikely here).

My first check is always to use "view_structure" to see if the model looks vaguely sensible. If it does I then click to show the atomic labels. If you have multiple atoms close to a special position then the labels become illegible (they get superimposed on each other).
It's also useful to calculate e.g. density in the INP file to make sure it's sensible.
You could also use MVW(0,0,0) to check molecular mass (in cell) to make sure it's what you expect.
Adding "num_posns 0" to the site line is also useful as it shows the multiplicity of sites occupied so it's easy to check your formula is correct. Syntax is "site W1 num_posns 0 x 0.25 y 0.25 z 0.75.....". The zero gets updated after refinement.

matsmcfly1

dear Prof Evans - thanks for your reply.

I believe space group to be correct (Fdd2).

MVW output looks OK: (MVW( 4675.947, 4524.199, 100.000))
View structure looks OK, but when I apply labels (as you sugggest) I think there is a problem with one component of the cocrystal, it seems like there are 2 labels for each atomic position:

num_positions shows 16 for all apart from C10 position (num_positions = 8) and this atom lies on a 2-fold axis

If half my malonic acid molecule is symmetry generated, what is the best way to deal with this in Topas for the refinement? I have changed the occupancy of each position to 0.5 - is this the most appropriate way to sort it out?

Fit is now far better:

johnsoevans

Good that you've found the problem.
There's been some discussion on this before (try searching "rigid body symmetry". If you've got a rigid body that sits on a symmetry element then you need to think a little bit about what degrees of freedom are possible. e.g. if you have a benzene ring lying flat on a mirror plane, you could rotate about an axis perpendicular to the ring (call it z) but not about x or y.
If the benzene ring is perpendicular to a mirror then you only need to define some (3 or 4 depending on orientation) of the six atoms. Or if you define all of them, then some need to be half occupancy. In this case you'd again be restricted to rotating about a single axis. It sounds like this is the situation you have.
Hope that's some help.

Ramon13

johnsoevans Interesting! Seeing as this had such a profound effect in @matsmcfly1's fit, I'm wondering if there are any general symptoms to look out for, apart from intensity misalignment that's not due to the scale factor, and corresponding rules or solutions to apply to the site information to minimise these issues? I wouldn't have known this could be an issue with my fit until I read this thread and I'm interested to add this to my QPA toolkit.

rowlesmr

Calculated density will be entirely wrong.

alan-coelho

Its more accurate to determine the correct symmetry as John has explained. However, occ_merge can be used to merge atoms on a site if one is unsure of the site symmetry.

Ramon13

Is there a general rule for min/max constraints on site positions? I usually am refining it "by sight" in the sense that I view the structure, and if there's a minor change that doesn't completely destroy the structure and it improves the fit, I'll go with it. But I really should have something a bit more quantitative. For example, for unit cell parameters I'll usually do min/max +/- 0.05.

johnsoevans

You have to be a bit careful about using min/max in this way - i.e. just to hold a structure together.
The real power of min/max limits in TOPAS comes from their use when solving the normal equations of least squares. This improves the refinement stability enormously (see Alan's paper ont eh BCCG method).
If a parameter hits its min/max limits at the end of a refinement you are best to remove it from the model as it may influence other parameters and/or prevent full convergence. If you use jEdit you'll see a coloured warning that flags this.
So you could set min/max values so that atoms shift by no more then e.g. 0.3 A from starting position during refinement, but you still need to consider if the data contain sufficient information to refine these parameters. What value to choose would depend on your starting model and how close you expect it to be to the compound you are studying.