Multi dataset multi phase weight fraction refinement

cjb

I'm refining a two-phase structure using (to start with) two data sets. Using various examples and tutorials I've got it to work except that TOPAS gives different weight percentages for the different data sets. My input file is of the form:

[...]
xdd 1...
     [...]
     str
          [...]
          weight_percent wt1 100
     str
          [...]
          weight_percent wt2 0
xdd 2...
     [...]
     str
          [...]
          weight_percent =wt1;
     str
          [...]
          weight_percent =wt2;
for xdds {
for strs 1 to 1 {
     [...]
}
for strs 2 to 2 {
     [...]
}
}

But still the weight percentages for xdd 2 aren't equal to those for xdd 1. How do I get TOPAS to recognise that it's the same mixture for both datasets?

Thanks

rowlesmr

First of all, I'm assuming that having the same weight fractions is a valid assumption.

THe weight fractions are calculated from the scale factors, unit cell mass and unit cell volume of all the phases that you have in your refinement, and are so scaled to a sum of 100%.

In order to set (or refine) the weight fractions, you need to do some maths on the Hill-Howard quantification equation.

To read up on the maths, read this page.

For a macro that can help you with up to 9 phases, have a look at this one.

cjb

Thanks for your reply! The two data sets are taken simultaneously using the same sample, from two data banks, so yes, it would be nonsensical for the two to give different phase fractions!

I had a look at those pages but couldn't figure out how to apply them. I've had another attempt, but still can't get it to work.

Using the macro I have an input of the form:

[...]

prm s1  1`
prm w1 = 100 - w2;:  90`
prm w2  10`
prm w3 =w1;
prm w4 =w2;

xdd 1...
     [...]
     str
          [...]
          wt_dets(1,4)
     str
          [...]
          wt_dets(2,4)
xdd 2...
     [...]
     str
          [...]
          wt_dets(3,4)
     str
          [...]
          wt_dets(4,4)
for xdds {
for strs 1 to 1 {
     [...]
}
for strs 2 to 2 {
     [...]
}
}

(N.B. I had to change instances of e.g. [m]#m_if p == 2;[/m] to [m]#m_ifarg p "2"[/m] to get the macro to run - that might be a version issue?)

This keeps giving an exception after a couple of iterations. Is there something obviously wrong with what I've done?

The weight fractions don't seem to be initialised at the values I set (and the scale factors seem to start negative) - do I have to set the weight fractions as w1,w2, etc. by hand or is the idea that that happens automatically by setting the scale factors? How do I get TOPAS to start with the values I set?

Thanks!

alancoelho

cjb

Have you ran test_examples\zro2-wt.inp. It works from Version 4.1 upwards and it is a 4 phase example.
alan

cjb

Thanks Alan. Yes, I used that as an example to try to set up my input file, but to no avail.

I've now achieved a measure of success though! I think the fact that all phases in the macro at http://topas.dur.ac.uk/topaswiki/doku.php?id=refine_weight_percentage are assumed to be different means it can't be used 'as is' for duplicated phases as I'm trying to do. The various equations would include things twice, once for each data set, and it would lock the scales for the second data set relative to those of the first, which there's no physical reason for.

I overcame this by duplicating everything in the macro with a third parameter, representing the data set number, so there are now two sets of every parameter (w1_1, w1_2, s1_1, s1_2, etc. for the first data set and w2_1, w2_2, s2_1, s2_2 etc. for the second). Now I can run the macro independently within each data set with just two phases, and simply set w2_1=w1_1 and w2_2=w1_2 while letting s1_1 and s2_1 both refine independently.

However, while this works perfectly for X-ray spectra, it's still resulting in an exception when I try it on neutron TOF data (both for the same sample). I've based my TOF inputs on the example here: http://topas.dur.ac.uk/topaswiki/doku.php?id=time_of_flight_tof_isis_instrument_standard_files#hrpd. Will the various scale_pks lines be interfering? Is there another obvious problem?

Thanks!

alancoelho

cjb

Send me the inp/data files and I will have a look; try and remove anything from the inp file that is not relevant
alan

cjb

Thanks to Alan I now have a solution - I'll leave it here for anyone else with the same query.

It's simply a matter of setting the ratio of the scale factors to be equal for each data set, while letting them refine as a group for each data set, i.e.

[...]
xdd 1...
     [...]
     str
          [...]
          scale s1_1 100
     str
          [...]
          scale s1_2 0
xdd 2...
     [...]
     local s2 10
     str
          [...]
          scale = s2 s1_1;
   str
          [...]
          scale = s2 s1_2;
for xdds {
for strs 1 to 1 {
     [...]
}
for strs 2 to 2 {
     [...]
}
}