Re: Simultaneous refinement of multiple atoms sharing same site

[Larry Finger]

On 8/31/19 9:44 AM, Ahmed Subrati wrote:

Dear all,

I wanted to ask how three atoms, sharing the same xyz position, could be refined 
in terms of their occupancies /via /FullProf. We know that for the case of two 
atoms, the first atom is set as 11.0 and the second one as -11.0 so that their 
sum is unity, but how would be the case for three atoms.


It does not matter what program you use. If you have X-ray data, and were to 
normalize the scattering curve for each type of atom by dividing by the atomic 
number, the curves would very nearly overlap. The minute differences would be 
the only information allowing you to determine those occupancies uniquely. If 
you merely constrain the sum of the occupancies, and the composition of the 
sample, the results will converge to nonsense results because the errors in the 
data would overwhelm those small differences.


If you had neutron scattering data, then you might be able to do it, depending 
on how different the scattering for the 3 types of nuclei.


Your only possibility is to add additional constraints. Perhaps average bond 
lengths will provide such information, or there may be other information 
available for your material. In any case, the diffraction data will not be 
sufficient for Reitveld, or single-crystal data.


Larry
++
Please do NOT attach files to the whole list 
Send commands to  eg: HELP as the subject with no body text
The Rietveld_L list archive is on http://www.mail-archive.com/rietveld_l@ill.fr/
++

Re: Rietveld website cancelled ?

[Larry Finger]

On 08/29/2018 02:01 PM, Le Bail Armel wrote:

OK, seems that the SDPD book preface looks unclear to some.

Western movies adepts well know that sentence :

"You see, in this world there's /two kinds/ of /people/, my friend: /those/ with 
loaded /guns/, and /those/ who /dig/. /You dig/. .."


Similarly, there are two kinds of people, those believing that the sense of the 
words "structure determination" and


"structure refinement" are the same, and the others. You dig.


I am happy to have been one of the "others". I also know the difference between 
lattice and structure. :)


Larry

++
Please do NOT attach files to the whole list 
Send commands to  eg: HELP as the subject with no body text
The Rietveld_L list archive is on http://www.mail-archive.com/rietveld_l@ill.fr/
++

Re: Rietveld website cancelled ?

[Larry Finger]

On 08/29/2018 04:15 AM, Le Bail Armel wrote:

Hi,

After the IUCr Monograph on Crystallography 13 (2002)

entitled "Stucture Determination from Powder Diffraction Data"

you may find in the preface :

"Although the Rietveld method of structure refinement from powder

diffraction data is often loosely considered to be synonymous with structure

determination, it is not. The Rietveld method only comes into play in the

final stage of the structure solution process when an approximate structural

model has been found."

WIF David, K Shankland, LB McCusker, C Baerlocher

A clear distinction having my complete agreement.

Armel


Armel,

Although the above statement may be generally true, I can provide a counter 
example. In Max W. Schmidt, Larry W. Finger, Ross J. Angel, and Robert E. 
Dinnabier, "Synthesis, crystal structure, and phase relations of AlSiO2OH, a 
high pressure hydrous phase", American Mineralogist, V. 83, 881-888, 1998, we 
only had a powder - no single crystals. From data measured at ESRF, we used GSAS 
to isolate the peaks of interest for the unknown phase, and used program TREOR 
to autoindex the unknown and identify the impurity phases. We then extracted 
intensities using the LeBail method, and determined the structure using direct 
methods with program SIRPOW. The structure was then refined with GSAS. We were 
even able to locate the H atom from Fourier maps!


Even 20 years ago, it was possible to determine the structure from X-ray powder 
patterns.


Larry


++
Please do NOT attach files to the whole list 
Send commands to  eg: HELP as the subject with no body text
The Rietveld_L list archive is on http://www.mail-archive.com/rietveld_l@ill.fr/
++

Re: Re:

[Larry Finger]

On 08/13/2018 03:43 PM, Toby, Brian H. wrote:

Dear Henk,

    I am not sure if this is what was intended in your commentary and e-mail, 
but I read from it the point that have heard often over the years, which is that 
someone is less than fully deserving of scientific credit because “s/he only 
developed the software” and in particular the term "the Rietveld Method” 
neglects the range of contributions from others in the scientific process. I 
think this needs a response.


    There is no doubt that pattern fitting /for extracting peak 
intensities/ predates the 1967 and 1969 papers from Hugo Rietveld, even though 
back then peak areas were most commonly estimated with a planimeter 
(https://en.wikipedia.org/wiki/Planimeter — I can remember my father around then 
using one for chromatography) or cutting and weighing. Likewise many people 
worked on developing the parameterization that allowed quantitative modeling of 
CW neutron diffraction peaks (though notably my understanding is that the 
well-known Cagliotti equation is a reformulation that Hugo first published, but 
credited to Cagliotti.) Van Laar sometime later contributed greatly to our 
understanding of low-angle asymmetric peak broadening (but again, fairly or not, 
most credit goes to Finger, Cox and Jephcoat who first released a general 
purpose program, even though Eddy and David first showed that it could be 
coded). I would guess that many people discussed the idea of determining 
crystallographic parameters by directly fitting to a diffractogram, but Hugo 
first developed an Algol code that actually implemented that concept, in a time 
when tackling such a large problem with the tiny computers available those days 
was an incredible achievement. If that was not enough, since Algol was not 
widely used, he then rewrote his code in Fortran to make his method more available.


     His Fortran code offered quite sophisticated crystallographic models, 
including magnetic scattering, and allowed complex groupings of parameters. 
Taking the idea of full pattern fitting from a concept to a method required 
considerable innovation. As one example, Hugo invented an intensity extraction 
algorithm that, as far as I am aware, is in every current program powder fitting 
in current use. Hugo’s extraction method was later incorporated into the LeBail 
method, when Armel came up with his smart idea to recycle those intensities.


     One measure of how far ahead of the curve Hugo’s work had been is to look 
at how long it took to see widespread acceptance. Around 1985, when I was first 
exposed to it, it was only just becoming be used outside of a few choice labs. 
Incidentally, what I used then was a code that had been passed through several 
hands and modified in each, but was based on Hugo’s original Fortran 
implementation. My understanding is that just about every Rietveld code that was 
available through to the 1990’s, with the one exception of GSAS, contained some 
of Hugo’s code.


     Hugo was one of many, many people who contributed to modern powder 
diffraction practice, but his solo work handed the world a tool which invented 
my field — powder diffraction crystallography — and I am most respectful of 
that. I cannot speak for others, but I personally will continue to use the term 
Rietveld analysis to honor an accomplishment that was well ahead of its time. It 
did not occur in a vacuum — little in science does, including the work of both 
Newton and Einstein — but still was an incredible step forward.


Brian,

My interpretation is not that the writer of the software should have been 
discredited, but that the other contributors should have been authors.


Your example of Finger, Cox, and Jephcoat is a good one. Yes, we did stand on 
the shoulders of others, but I think we gave them credit in the manuscript. For 
whatever reasons, none of their contributions had made it into code. By our 
re-deriving the equations in a computable form, and providing a sample 
implementation code, our contribution stuck. In fact, ResearchGate still reports 
3 or 4 citations of the paper per month, which is remarkable for a paper 
published 24 years ago.


There would have been a closer parallel if I had ignored Dave's and Andrew's 
contributions to the project, and published a single-author paper! That would 
not have been fair, nor was it the way I worked.


Larry
++
Please do NOT attach files to the whole list 
Send commands to  eg: HELP as the subject with no body text
The Rietveld_L list archive is on http://www.mail-archive.com/rietveld_l@ill.fr/
++

Re: The development of powder profile refinement...

[Larry Finger]

On 08/10/2018 10:55 AM, Alan Hewat wrote:
Thanks Henk (and Bob). It's a story that must be quite common in science. Humans 
like to simplify, and identify with heros. Humanity teaches us that we stand on 
the shoulders of others. Few discoveries, especially in science, are made by a 
single individual.


Profile Refinement was perhaps an idea "whose time had come" with the 
application of computers. But the stroke of genius was to refine the structure 
directly from the data points, rather than first refine structure factors. Even 
that may have been simply because computers at that time could only handle a 
limited number of parameters.


A fascinating story. Alan.


Henk,

I agree with Alan that you told a fascinating story, and not what I was 
expecting. It is always a temptation to give yourself more credit than an 
unbiased observer might allow. This case represents a most egregious instance.


To show the state of early computing, one of the graduate students in 
Crystallography in the Geology Department at the University of Minnesota when I 
started in 1962, was awarded a PhD for writing a computer program to calculate 
powder patterns! This was on a Univac 1103 with capabilities similar to that of 
the computer described in Bob and Henk's article. With such limited 
capabilities, you had to choose the problem with considerable care.


By the time I graduated 5 years later, the University had a CDC 6600 
"supercomputer", which had the blazing speed of 3 megaFLOPS (double precision)! 
For comparison, this is roughly the performance of an x86 machine with an 0.7 
GHz clock!!


Larry

++
Please do NOT attach files to the whole list 
Send commands to  eg: HELP as the subject with no body text
The Rietveld_L list archive is on http://www.mail-archive.com/rietveld_l@ill.fr/
++

Re: Stoichiometry and occupancy fractions of solid solutions

[Larry Finger]

On 12/19/2016 09:29 AM, Cline, James Dr. (Fed) wrote:

Your data is from where?  I wouldn’t trust numbers from refinements as these
from lab data.



Jim





James P. Cline
Materials Measurement Science Division
National Institute of Standards and Technology
100 Bureau Dr. stop 8520 [ B113 / Bldg 217 ]
Gaithersburg, MD 20899-8523USA
jcl...@nist.gov 
(301) 975 5793
FAX (301) 975 5334



*From:*rietveld_l-requ...@ill.fr [mailto:rietveld_l-requ...@ill.fr] *On Behalf
Of *Othman Al Bahri
*Sent:* Monday, December 19, 2016 3:54 AM
*To:* rietveld_l@ill.fr
*Subject:* Stoichiometry and occupancy fractions of solid solutions



Dear all,



I've made a series of solid solution powders using a solid state reaction in the
form A_2 B_3-x C_x O_12 at_x= 0.5 steps. A_2 B_3 O_12 is orthorhombic while A_2
C_3 O_12 is monoclinic. I'm refining the XRD data to find the atomic
distribution of the solute.



I've constrained the sum of the occupancy fractions for each relevant site to
equal 1. At low concentrations of the solute, I initially set the solute's
occupancy fractions to 0 and keep the solvent's occupancy at 1 then refine the
fractions (after following the usual Rietveld refinement steps). This seems to
give reasonable occupancy fraction values (no big numbers or negative values)
but the stoichiometry is way off. This is probably because each site has
different Wykoff multiplicities so constraining the sum of each site's fractions
to 1 is insufficient.



Let's assume that I knew the stoichiometry from Mass Spectroscopy or XPS - is
there a way to constrain the stiochiometry in a Rietveld refinement? I'm using
GSAS-II and comfortable with FullProf but feel free to give advice for any other
open-source software.



I've seen a few papers where the authors mention, typically in the supplementary
info, that their refinements' stoichiometry was off and that it should be
ignored. However I'm not comfortable with this approach and would appreciate
your advice.



This is my first time working with solid solutions so please feel free to offer
any general advice on what I should be careful with. I've tested for phase
mixtures (insolubility) by visually comparing my XRD patterns with the sum of
simulated XRD patterns of molar mixtures and through Rietveld refinements with
two phases. The system I'm working with has been reported but the original
authors didn't do Rietveld refinements - they were interested in physical
property measurements.



Kind Regards,



Othman


I am showing my age, but in my PhD thesis, I demonstrated that even when using 
single-crystal data to refine distributions of such favorable species as Fe vs 
Mg, constraining full occupancy was not sufficient. A stoichiometry constraint 
was also needed. The reference is Finger, Larry W (1969) The crystal structure 
and cation distribution of a grunerite, Mineral. Soc. Am. Spec. Paper, 2, 95-100.


If you know the overall chemical composition from electron microprobe data as in 
my case, or from the chemicals used in the synthesis, use that information to 
constrain the refinement!


Larry

++
Please do NOT attach files to the whole list 
Send commands to  eg: HELP as the subject with no body text
The Rietveld_L list archive is on http://www.mail-archive.com/rietveld_l@ill.fr/
++

Re: Powder Diffraction Discussion Group on Facebook

[Larry Finger]

I belong to Facebook, but I almost never post there. My wife uses one of our 
Linux machines to read it so that she can keep up with the postings of our 
children and grandchildren.As it is very easy to get malware from Facebook, she 
never uses her own computer with Windows 7 as OS.


Anything posted to a powder diffraction group at Facebook will be unknown to me 
unless it is also posted here.


Troglodyte Larry

++
Please do NOT attach files to the whole list alan.he...@neutronoptics.com
Send commands to lists...@ill.fr eg: HELP as the subject with no body text
The Rietveld_L list archive is on http://www.mail-archive.com/rietveld_l@ill.fr/
++

Re: regress in crystallographic good practices and knowledge

[Larry Finger]

On 05/08/2015 11:16 AM, Kurt Leinenweber wrote:

Hi Alan,

I still like the Rietveld list and do read all the posts.

Is it possible to block attachments so that people can’t attach files any more?

-Kurt


Thus writes the guy whose mailer blindly attaches a message footer rather than 
in-lining a signature. :)


Although I no longer practice crystallography, and no longer read the 
literature, I do find the discussions here of interest. At least I can follow 
how little crystallography the average solid-state researcher actually knows. 
Sad to know that bugs in a few programs could completely derail the science. All 
the greats of the early part of the 20th century must be really spinning in 
their graves!


Larry

++
Please do NOT attach files to the whole list alan.he...@neutronoptics.com
Send commands to lists...@ill.fr eg: HELP as the subject with no body text
The Rietveld_L list archive is on http://www.mail-archive.com/rietveld_l@ill.fr/
++

Re: Different atomic coordinates

[Larry Finger]

Stanislav Ferdov wrote:

Dear All,

I am trying to refine a phase that has a structure analogue with different 
chemical composition. The problem is that using the published atomic 
coordinates one cannot get a proper refinement (mismatch in the diffraction 
intensities). When I process the data and solve the structure by direct methods 
I get atomic coordinates that are suitable for a good a Rietveld refinement 
(the structure is the same). If somebody has explanation of this problem I will 
be very grateful to hear it.

Many thanks in advance.


You really should give some details. Does the space group have two 
possible origins? Etc, etc


Larry

Re: Question on Rietveld with x-ray

[Larry Finger]

Jae-Ho Chung wrote:

Dear experts on Rietveld with x-ray,

 

When it comes to Rietveld, I have exclusively been working with 
neutrons. Now I am just trying Rietveld with x-rays, and immediately 
came up with a question. Since x-ray scatters from electrons, it seems 
to me that we always have to use, for instance, O-2 instead of O, and so 
on, for ionic compounds. But I have an impression that O is still used 
in many examples, including certain example files given in GSAS, such as 
Y2O3.EXP. Why is it?


I suggest that you plot the scattering curves for O and O-2, and then 
decide how much difference it makes. In addition, you should also 
postulate a theory concerning the differences. Hint: The electron 
cloud is not a point object.


As a naturally following question, since no material is perfectly ionic, 
the oxidation numbers may become another “refinable” parameters. But I 
suppose they are never refinable in GSAS or in FullProf. Therefore, I 
suppose the fractional occupation may be compromised instead? Is this in 
any way discussed among x-ray experts?


See above. What experiment would you propose to make the oxidation 
number be refinable?


Larry

Re: Absolute structure from powder data?

[Larry Finger]

Franz Werner wrote:
 Dear Rietvelders
 
 Is it in principle impossible to determine the absolute structure from powder 
 data due to reflection overlap or is there a way via multiple wavelength 
 diffraction experiments?
 
 Thanks for your advice.

To determine absolute structure, you need three things: (1) a 
non-centrosymmetric structure, (2) at
least one anomalous scatterer, and (3) the differences in intensities between 
Friedel pairs of
reflections, i.e. reflection hkl and reflection -h-k-l.

In any diffraction experiment with the wavelength chosen carefully, you could 
satisfy conditions 1
and 2. Number 3 always nails you in a powder experiment due to the systematic, 
exact, overlap of hkl
and -h-k-l.

Larry

Re: stresses-crystal structure changes

[Larry Finger]

[EMAIL PROTECTED] wrote:
 Dear all:
 
 Let me make you the following question.
 
 Suppose a polycrystalline coating deposited onto a substrate in such a way
 that the coating is subjected to in-plane compressive residual stresses
 after the deposition process. The coating has a cubic fcc crystal
 structure before the deposition process, and also appears to be cubic fcc
 with increased out-of-plane lattice parameter after the deposition. I
 understand the increased out-of-plane lattice parameter because the
 stresses are compressive in nature and the XRD experiment collects
 information on the diffracting planes that are parallel to the sample
 surface. However, I have the following question:
 
 Since the residual stresses are compressive in nature, the fcc crystal
 structure will become distorted after the deposition process, and strictly
 will be no longer cubic. The change in the shape of the unit cell is most
 likely a function of the orientation of the unit cell with respect to the
 direction of the residual stresses. Thus, the cube can change to
 tetrahedron, rhombohedral, etc. If the stresses are large enough in
 magnitude (bout 1000 MPa), why the XRD pattern still shows a cubic crystal
 structure (there are no additional peaks, peak splitting, etc).

Because the diffracting planes are all parallel to the sample surface, you are 
only measuring d
spacings perpendicular to the surface. As a result, every such plane sees the 
same stress and you
get a cubic pattern with no splitting. If you could measure the diffraction 
pattern in transmission
you would see the other component of the stress (assuming that your stress 
distribution is an
ellipsoid of revolution). If the normal to the planes of diffraction were in 
the plane of the
coating, the pattern would again be cubic with a different lattice parameter. 
Only if the normal
were neither parallel nor perpendicular to the plane of the coating would you 
see splitting.

Larry

New Versions of DRAWxtl available

[Larry Finger]

The authors of DRAWxtl are pleased to announce the release of V4.3 and V5.3 of 
our program. As
before, V5.3 has a full GUI, whereas V4.3 is command-line oriented with an 
openGL window displaying
the active structure. Note: we skipped V5.2 so that our two series would have 
the same minor release
number. The new software versions, including source code and binaries for 
Windows, Linux (i386 only)
and Macintosh (universal binary), can be downloaded from 
http://www.lwfinger.net/drawxtl.

We have also published a description of the program in J. Appl. Cryst. 40, 
188-192, 2007.

There are too many changes in the new versions to list; however, the most 
important are as follows:

 1. DRAWxtl can plot incommensurately modulated structures up to (3 + 3)
dimensions. In addition composite crystals can be plotted, and the unit-cell
for each component is drawn.

 2. The unit-cell axis labels and the vector triple can be repositioned
using the mouse (V5.3 only).

 3. The program can now draw a best fitting plane through a list of atoms.

 4. An RGB triple may be used to specify color as well as named colors.

 5. The polyhedral sections now calculate the volume of the polyhedron, and for
tetrahedra and octahedra, distortion parameters (quadratic elongation, 
angle variance)
are calculated.

 6. The program can generate stereo POV files for the unofficial version called 
StereoPOV.

 7. Input of dn6 (O format) Fourier maps added.

 8. Reading of electron density and electron localisation function files from 
EXCITING added.

 9. Reading of DISCUS and FullProf input files added.

10. The slab cutout function now correctly handles non-rectangular slabs.

11. There is now a general file viewer under the 'View' menu (V5.3 only).

12. The loookat command now is honored by the GUI version (V5.3 only).

13. JANA-style M80 Fo/Fc files can be used to generate a Fourier map.

14. DRAWxtl can be called with a CIF file as its command-line argument,
allowing its use as graphics viewer in JANA and probably other systems

15. Fixed addition of lone pairs through the GUI of V5.3.

16. Many problems in reading 'foreign' input files, particularly CIF, have been 
fixed.

Re: how to find Polarization

[Larry Finger]

[EMAIL PROTECTED] wrote:

Dear All,
if I well understood JFC correction is perfect in case of
parallel incident beam; so, in case of conventional
Bragg-Brentano diffractometer, shouldn't it work well only
in case of use of Goebel Mirrors, that get incident beam
exactly parallel?
And is it true that using Goebel Mirrors and sample in
capillary (Debye-Scherrer) gets intensity values more
realistic than on a conventional Bragg-Brentano geometry?
Thanks in advance,
marco
  
Not really. The FCJ correction (note the authorship please) was derived 
for the parallel incident beam case, but it works for divergent optics. 
The main difference is that for the parallel beam case, one can measure 
the slit sizes perpendicular to the plane formed by the incident and 
diffracted beams, directly calculate the values for S/L and H/L, and get 
values very close to the best-fit results. For divergent beam optics, 
the effective width is greater than the apparent width. As discussed 
earlier in this list, in the divergent beam case, a set of 0.02 (radian) 
slits will yield refined values of 0.027 for S/L and H/L, not 0.02 as 
predicted from the geometry. In the extreme case, I removed the Soller 
slits on my conventional B-B diffractometer, and could still fit the 
resulting profiles, which were greatly affected by axial divergence. As 
I recall, S/L and H/L were on the order of 0.2! BTW, the intensities 
were increased by roughly a factor of 10. I had to cut the tube power to 
avoid saturating the detector.


Putting your sample in a capillary avoids a lot of sample problems that 
occur with a flat plate; however I'm not sure that I would make the 
blanket statement that you do. That topic should be addressed by someone 
with experience with mirrors.


Larry