Via the Sci.techniques.xtallography Newsgroup

From: "Natale Perchiazzi" [natale@dst.unipi.it] To: "rietveld" [rietveld_l@ill.fr] Subject: overall temperature factor refinement Date: Mon, 19 Nov 2001 12:36:45 +0100 Organization: Dipartimento di Scienze della Terra Pisa Dear All, from what I understood, in QPA refinements is correct to leave untouched the literature structural (positional and thermal) parameters (assuming I have correct structural models for my analysis). About thermal parameters, is this always true or maybe correct in some case (what?) to refine an overall temperature factor for each phase of the mixture? Best regards Natale Perchiazzi

Date: Tue, 20 Nov 2001 11:21:15 +0000 To: rietveld_l@ill.fr From: Lachlan Cranswick [l.m.d.cranswick@dl.ac.uk] Subject: Re: overall temperature factor refinement At 12:36 19/11/01 +0100, you wrote: >Dear All, > >from what I understood, in QPA refinements is correct to leave untouched >the literature structural (positional and thermal) parameters (assuming >I have correct structural models for my analysis). About thermal parameters, >is this always true or maybe correct in some case (what?) to refine an >overall temperature factor for each phase of the mixture? > >Best regards Natale Perchiazzi > (forwarded on behalf of Ian Madsen as the Rietveld listserver does not love him at this point in time - and refuses to let him post out) --+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ --+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ There are many and varied approaches being applied for QPA using Rietveld based techniques (as evidenced in the recent IUCr CPD round robin on QPA). Some seem to work better than others !! The following is the approach that we take when accurate QPA is required (but is not claimed to be the only valid procedure):- (1) Start with the published values of crystal structure (including thermal parameters). However, these values may only approximate the structure found in the phase in your sample. This is especially true in mineralogical materials where there may be variable composition (and hence SOF's in specific sites). (2) If some pure phase (or maybe a sample with very high concentration of this phase) is available, then collect a "structure refinement" quality data set (i.e. to very high upper limit of 2theta, good resolution, good counting statistics etc...) and refine as much of the structure as can be supported by the data (thermals, SOF's atom coordinates as appropriate). Take these values and fix them for the analysis of the 'real' sample - some parameters can still be refined at lower concentrations, but in general we only refine the unit cell, peak width/shape etc. At very low concentrations, even these values should be fixed at the values determined from the high concentration sample (or at least limited in the range over which they can vary) . (3) For phases with variable composition (e.g. minerals) obtain independent estimates of phase composition (e.g. electron probe microanalysis) and adjust the SOF's accordingly. (4) We have used the approach you suggested (refine an overall thermal parameter), but have largely abandoned it in favour of the approach in (2) above. Variations in the overall thermals can account for small differences between the structure model used and the 'real' phase BUT, the total thermal parameter should be checked for physical meaning - it should be positive and have a value which is 'reasonable' for the type of material being analysed. Large -ve or +ve values may indicate other problems in the Rietveld model, or simply indicate you do not have enough data to ensure stability. As with the refinement of any parameter, you must be sure that the data you have supports the parameter you are varying - a typical QPA data set (with high 2theta cut-off say at 80-90 degrees) is unlikely to have enough data to support refinement of thermal parameters, especially if the phase is not a major component. Note also that the thermal parameters have a strong correlation with the Rietveld scale factor - and hence on the determined phase abundances - so do not vary them without good reason for doing so. (5) Check your QPA against an independent measurement. The one we use regularly is to calculate the expected elemental composition from the QPA & the individual phase compositions and then compare this with chemical analysis (e.g. XRF). (6) In the recent QPA round robin, we used the approach in (2) above to obtain our own estimates of QPA for verification of mixing Sample 1a (95wt% fluorite) was used to refine details of the fluorite structure; similarly, Sample 1b (95wt% corundum) for the corundum structure, Sample 1c (95wt% zincite) for the zincite structure - and then fixed these values for the analysis of all 8 mixtures. I hope this helps - if not, feel free to contact me. Cheers o----------------------oo0oo---------------------------o Ian Madsen CSIRO Minerals Box 312 Clayton South 3169 Victoria AUSTRALIA Phone +61 3 9545 8785 direct +61 3 9545 8500 switch 0417 554 935 mobile FAX +61 3 9562 8919 Email Ian.Madsen@minerals.csiro.au o----------------------oo0oo---------------------------o