just my 2 cents... > Could I be so stupid to say that such kind of works, including mine, are > nothing? following Nicolae, I should also add to the list myself as well as most people participating to the four editions of the size-strain conference/meeting/workshop and all participants to Davor's size-strain round-robin.
I bet people should spend more time in the library... this is the point. This is also a self criticism as I'm not the best library addict (though, online resources has simplified life enormously)... We should not try to use line profile analysis methods as a black box: it is easy to obtain numbers from measured data (with a proper software a computer can do it automatically), but then it is in the ability of the scientist to attach them a proper physical meaning. What it is difficult (perhaps impossible?) is willing and pretending to do it in the general case as we're dealing with something that has no precise rules (domain size, shape and their distributions are not properties of the materials, nor they can be easily predicted in advance). Some simple cases have been studied and some references already posted by several people in here, and in most of them the agreement between diffraction and alternative techniques is quite good: just in few cases, though, enough information is available to interpret the strain broadening fully in terms of physical defects present in the material, or to model the size term using a more or less complex distribution of (iso-shape) domains. But also in those cases the result is the one compatible with the model assumptions and does not pretend to be "God's truth". So welcome the round robin on a more complex sample to test the maturity of the algorithms (they should be even tested on simpler examples, as concluded on the latest size-strain conference, but that's another story..), but beware that without any a priori info (or with a wrong one!), a vast set of odd results can be obtained. As a comparison, it would be like pretending to do a search match, a structure solution or, even worse, a Rietveld refinement on a material for which we don't know any chemical information... Going back to Leonid's question, well the answer is easy: check the premises... the assumptions behind the use of the TCH function are not compatible with he presence of a lognormal distribution of domains. It can be proven mathematically that the Fourier coefficients for a profile describing a lognormal distribution of domains have a hook at low Fourier number, hook that cannot be reproduced by any whatsoever voigtian or voigtian-like curve. This is a common problem in the use of Voigt and voigt-like curves in describing the peak profiles from nanocrystalline powders and is also the main source of the "superlorentzian" peak tails (they are a trick to get rid of the physical information contained in the profile ;) we are a bit masochist, aren't we?) Best regards Mat -- Matteo Leoni Department of Materials Engineering and Industrial Technologies University of Trento 38050 Mesiano (TN) ITALY Tel +39 0461 882416 e-mail: [EMAIL PROTECTED] Fax +39 0461 881977 Web: www.matteoleoni.ing.unitn.it