Via the Rietveld Users' Mailing List
Subject: hkl-dependent peak asymmetry? Date: Tue, 17 Dec 2002 10:51:46 -0500 To: [email protected] From: "Maxim V. Lobanov" [[email protected]] Dear Rietveld list subscribers, I would be interested if somebody could suggest a software (and, maybe, a corresponding example) which can handle hkl-dependent peak asymmetry. I mean, we are now trying to refine the structure from high-resolution synchrotron data, and clearly see different asymmetry for different types of reflections (specifically, "Lorentz tails" on either left or right side). It might indicate (of course) some kind of symmetry lowering, but of very low magnitude (it is actually high resolution data), and at present I would prefer to treat this as "hkl-dependent asymmetry". Thank you in advance. Sincerely, Maxim. __________________________________ Maxim V. Lobanov Department of Chemistry Rutgers University 610 Taylor Rd Piscataway, NJ 08854 Phone: (732) 445-3811
|
Subject: Re: hkl-dependent peak asymmetry? Date: Tue, 17 Dec 2002 10:12:52 -0600 From: Frank May [[email protected]] To: [email protected] Hello Maxim, Do you also see differences in peak width which you can relate to the various HKL classes? I've been looking since 1985 for a way to handle high resolution data with these effects. Keep me posted if you find anything. Regards from St. Louis, Frank May Department of Chemistry University of Missouri-St. Louis (314) 516-5098
|
Subject: Re: hkl-dependent peak asymmetry? Date: Tue, 17 Dec 2002 16:39:15 +0000 (GMT) To: [email protected] From: [email protected] (Irene Margiolaki) Hello, The following reference which describes anisotropic peak broadening in powder diffraction, might be of some help: P. Stephens, J. Appl. Cryst. Vol. 32, page: 281-289, 1999 In GSAS, there is a specific profile function (type 4) which involves this effect and calculates automatically the allowed anisotropic strain coefficients for different crystal structures. Of course, this could be an idea in the case that you observe difference in the broadening of the individual peaks in the profile. Rena.
|
Subject: Re: hkl-dependent peak asymmetry? Date: Tue, 17 Dec 2002 18:03:17 +0000 From: Andreas Leineweber [[email protected]] To: [email protected] Dear Maxim, Sample dependent ASYMMETRIES are treated only rarely quantitatively. According to my knowledge, no free Rietveld program can handle such a thing. There is the program Topas by Bruker, where you can program nearly anything you want, if you have a good phenomenological model for that. The origin of such effects lies probably in anisotropic micro-strain distributions, which may be, furthermore, anisotropic (size effect should be impossible). Such effects may, e.g. be caused by composition distributions. An example how to treat it in a relatively rude way (but it is working) is described in Ehrenberg, H., Theissmann, R., Gassenbauer, Y., Knapp, M., Wltschek, Weitzel, H. Fuess, H. Herrmannsd�rfer, T. & Sheptyakov, D. (2002). J. Phys.: Condens. Matter 14, 8573-8581: You just assume the presence of several phases of different lattice parameters. For that it is good (but not neccessary) to know the concentration dependence of the lattice parameters. You may try to look whether this effect occurs only for certain direction in the reciprocal space. Another example, which involves, however, a phenomenological treatment of the asymmetry, is found in Langford, J. I., Louer, D. (1991), Journal of Applied Crystallography 24, 149-155. You can also ask me directly if you have further questions. Yours sincerely Andreas Leineweber -- Dr. Andreas Leineweber please note the new address Max-Planck-Institut fuer Metallforschung Heisenbergstrasse 3 70569 Stuttgart Germany Tel. +49 711 689 3365 Fax. +49 711 689 3312 e-mail: [email protected] home page of department: http://finix.mpi-stuttgart.mpg.de/mittemeijer/english/index_english.htm
|
Subject: Re: hkl-dependent peak asymmetry? Date: Tue, 17 Dec 2002 18:14:24 +0000 From: Andreas Leineweber [[email protected]] To: [email protected] Sorry, Maxim, once more, Another important case of anisotropic line broadening occurs for stacking faults, e.g. in fcc materials. Even for the simplest cases this is relatively complicated (e.g. L. Velterop, R. Delhez, Th.H. de Keijser, E.J. Mittemeijer, D. Reefman: X-ray diffraction analysis of stacking and twin faults in f.c.c. metals: a revision and allowance for texture and non-uniform fault probabilities, J. Appl. Crystallogr. 33 (2000) 296-306.). In a Rietveld similar approach this was applied (Scardi, P. & Leoni, M. (2002). Acta Cryst. A58, 190-200.). But the simple, freely available programs (see last mail) cannot do that according to my knowledge. Best regards Andreas Leineweber "-- Dr. Andreas Leineweber please note the new address Max-Planck-Institut fuer Metallforschung Heisenbergstrasse 3 70569 Stuttgart Germany Tel. +49 711 689 3365 Fax. +49 711 689 3312 e-mail: [email protected] home page of department: http://finix.mpi-stuttgart.mpg.de/mittemeijer/english/index_english.htm
|
Subject: Re: hkl-dependent peak asymmetry? Date: Tue, 17 Dec 2002 09:32:44 -0800 To: [email protected] From: Luca Lutterotti [[email protected]] Dear Maxim, what compound are you analysing and what kind of symmetry it has (space group)? I am asking this because, the more common cause of hkl dependent symmetry are faults, either deformation, stacking, growth or twin faults. The problem for a phenomenological approach on it is that there is no actually a general model or treatment for them. Only for close packed cubics (like FCC, BCC) and hexagonal there is the well known treatment of Warren. Some people have tried to model (not refine) such faults in clay materials. So we have implemented the Warren (+recent modifications) treatment in Maud (Lutterotti and Gialanella, Acta Materialia, 46[1], 101-110, 1998); but as I said before it is only for close packed materials. Maud, I recall, is just a free Rietveld program; it's problem is that it is in Java, so it is not Microsoft certified ;-)) Best regards, Luca Lutterotti
|
Subject: Re: hkl-dependent peak asymmetry? From: "Joerg Bergmann" [[email protected]] To: "[email protected]" [[email protected]] Date: Tue, 17 Dec 2002 21:24:12 +0100 (CET) On Tue, 17 Dec 2002 10:51:46 -0500, Maxim V. Lobanov wrote: >Dear Rietveld list subscribers, >I would be interested if somebody could suggest a software (and, maybe, a >corresponding example) which can handle hkl-dependent peak asymmetry. I >mean, we are now trying to refine the structure from high-resolution >synchrotron data, and clearly see different asymmetry for different types >of reflections (specifically, "Lorentz tails" on either left or right >side). It might indicate (of course) some kind of symmetry lowering, but of >very low magnitude (it is actually high resolution data), and at present I >would prefer to treat this as "hkl-dependent asymmetry". You may handle this in BGMN by defining "multiplet" lines (RefMult=2 or 3...). In following, you may define some hkl-dependent expression for the shift/intensitiy/with parameters of the 2/3... particular lines. Giving different shifts/intensities/width parameters for iref=1 resp. iref=2... gives some anisotropy for a hkl line as representated by a non-resolved multiplet. We have done such for example for the description of hkl-dependent anisotropies of clays. Regards J. Bergmann, Dresden [email protected]
|
Date: Wed, 18 Dec 2002 08:23:21 +0100 From: "Kern, Arnt A." [[email protected]] To: [[email protected]] Dear Maxim, I already posted this about one month ago. Check it out - it works extremely well! Cheers, Arnt ----------------------- For the upcoming Topas V2.1 I am working on two tutorial examples (Pawley or LeBail fitting) for anisotropic line broadening, which are available on request (data and INP files): - Norbornane, taken from A. LeBail's PowBase (http://pcb4122.univ-lemans.fr/) - Pboxas, taken from the KoalaRiet distribution (http://www.ccp14.ac.uk/) For Pboxas the use of spherical harmonics is shown for "gauss_fhwhm" and / or "lor_fwhm" and / or asymmetry using a "circles convolution" (-> axial divergence). Its easy to use an "exponential" or "1/x convolution" instead of the circles convolution. "Just for fun" it is also shown how to accommodate hkl-dependent peak shifts by linking spherical harmonics to e.g. the zero point error. Please keep in mind: With spherical harmonics it is always extremely important to check parameter correlations and therefore also to avoid to use higher orders than necessary. When fitting second-rank tensors the situation is similar. -- By the way: By convoluting the sum of two exponentials with a Voigt or a pseudo-Voigt function you not only can get a nice description of e.g. ISIS TOF neutron data with downto 5 refineable profile parameters only; by using spherical harmonics to make the exponentials hkl-dependent you also get a wonderful fit of extreme profile anisotropy e.g. due to disorder as in clays. The physical meaning of the refined (profile) parameters is another story though..., at least the fit looks good ;-). This is meant to clearly illustrate my opinion, that in general such refinement models should be only used with great care! They should never be used to just simply get a better fit... This is why I like the example of Lachlan very much - he thoroughly checked the results. Cheers, Arnt
|
Subject: Re: hkl-dependent peak asymmetry? Date: Wed, 18 Dec 2002 12:01:04 +0100 To: [email protected] From: Armel Le Bail [[email protected]] Hi, I remember Davor Balzar giving account of the Size/Strain Round Robin at IUCr XIX Geneva : the Rietveld approach of these problems was clearly not recommended... I disagree when structures are complex enough to give more than a few Bragg peaks typical for a cubic sample with small cell parameter (CeO2 for instance). In complex cases, with huge overlapping, nothing else than a global pattern modelling is applicable, and this includes the Rietveld approach. However, if experts who will be the manuscript reviewers, think that Rietveld is not recommended in size/strain studies, then the publication of your results will be difficult, maybe. Can we expect the next size/strain round robin step with some samples more complex than a cubic size-only one ? With all these computer programs claiming to cope with hkl-dependent peak broadening and asymmetry, there is a gulf between expert recommendations and software possibilities (though these programs often produce only phenomenological fits that do not give any size/strain parameters). The cases of Norbornane and Pb oxalate are quite nice examples with data available on the Internet. There are even more bizarre data in Powbase than PbC2O4 and Norbornane. A query with broadening as a keyword returns 25 entries : http://sdpd.univ-lemans.fr/powbase/ Enough non-cubic cases to frighten experts... So, even if we know that no high accuracy in size/strain parameters can be expected by using the Rietveld approach in complex cases, should we do nothing instead ? Best, Armel
|
Subject: RE: hkl-dependent peak asymmetry? Date: Fri, 20 Dec 2002 10:20:33 +0100 To: [email protected] From: Alan Hewat [[email protected]] Forwarded from Davor Balzar: >>I remember Davor Balzar giving account of the Size/Strain >>Round Robin at IUCr XIX Geneva : the Rietveld approach of >>these problems was clearly not recommended... >Well, to set the record straight about my Geneva lecture and before the full >paper with some recommendations and recipes in regard to Rietveld-based >determination of size/strain, based on the evaluations of ceria round-robin >sample, is out of review (Armel is one of the paper co-authors and I thought >we all agreed on this:-): > >- Results obtained through Rietveld refinement agreed extremely well with >other, more traditional techniques. This, of course, for cynics would >implicate that for simple cases such as this one, any method would do. >However, the round-robin indicated that the results could very up to 100 or so >%, if the refinement is not carried out properly. The most important mistake >is not to refine all profile-related parameters that are pertinent to the >size/strain effects -- typical example is overlooking the Gaussian size term >(P). > >- Rietveld programs become more sophisticated and some of them can handle a >variety of line-broadening sources. However, this implies there are more >parameters to refine and more possibility of correlation with other refined >parameters especially if a limited theta (Q) space was covered in an >experiment. > >- As a consequence, size/strain related broadening is compensated by another >parameter with similar effect on line widths -- pattern fit looks excellent, >Rwp is low, but physical interpretation is wrong (as it was pointed out by >Armel and Arnt). > >- As this is directed toward newcomers in this field, one should always be >reminded that Rietveld and similar approaches are based on solving a >linearized system of nonlinear equations -- therefore, unless a guessed >parameter value is not close to the "true" one, the refinement may never reach >the "true" value and there is no way of knowing when or whether that value was >reached. > >- Therefore, one should try different methods to check the validity of results >obtained. However, most of the traditional line-broadening methods cannot be >applied or also carry large systematic errors if applied to low-symmetry >materials. Rietveld of similar full-powder pattern approach might be your only >method available. > >Thus, my full summary in Geneva (and our paper) was/is something like: >- Rietveld is good and getting better for size/strain IF you are aware of >pitfalls. >- In case of complicated structures with overlapping lines, multiphase >mixtures, etc.., the results should be taken cautiously, but it maybe your >only method available. >- Always check obtained results for physical soundness and compare to other >methods if possible. > >The bottom line: I think we all pretty much agree that the Rietveld route >should be used, and will be used more often in future. However, very carefully >because Rietveld programs are getting more user-friendly but far from being >laymen user-proof. > >Davor Balzar > >P.S: In this field, there are no experts, only different opinions.
|