[snip]
>I'm very familiar with the magnetite peak at 1.929 =C5 - I spent a couple=
of
>weeks convincing myself it was really a magnetite peak, when I first began
>work on this project. My sample was 'home-made' and pure as proved by=
other
>techniques, and I wouldn't accept the possibility of an impurity, in spite
>of what I was being told by more experienced heads (at the time) than mine.
>I started out as a single crystal person and suffered with powder patterns,
>until I realized that the reference data base was neither accurate nor
complete.
[snip]
I remember after giving a talk at a Rietveld workshop (on the
lines of what "would it be like if you tried to do quantitative=20
Rietveld analysis" - on the Rutile/Ilmenite phase system)=20
there were some personal comments bemoaning that it was a pity=20
there are not more "Rietveld Failures" or=20
"I Tried to do Rietveld Analysis and it Ruined my Life" type=20
of review articles in the literature. Or perhaps just balanced
reviews. (The above didn't ruin my life entirely as shown by=20
results published in:
R. R. Merritt and L. M. D. Cranswick, Proceedings of the=20
6th AusIMM Extractive Metallurgy Conference, Brisbane,=20
3-6 July, 1994, pp 171-180)
Reading much of the literature on Rietveld analysis can give=20
an unrealistic picture on how easy it is to perform. =20
Whereas in reality it can be quite tricky, problematic, frustrating
and very easy to get wrong results without realizing it
if you don't keep your wits about you.
---
Phase Identification can be of similar frustration if youare not informed of the pitfalls in advance.
It would be nice if there was a web or internet database people=20could submit problems that occurred - and their solutions. Though Iexpect time constraints and people being a bit shy on revealingthe problems they had might make this suggestion somewhat of afantasy. "Failure databases" could be one of the interestingapplications of the internet for science and education.Though most likely they would eventually become failure andsolution databases(?).
----
My present Phase ID course notes are around 19 pages worth of=20graphical examples with the aim of giving participants a feel=20for what it is really like to do phase ID so they can help=20avoid the pitfalls of future samples. =20(Possibly a web version can appear soon with the possibility ofpeople making further contributions?).Though it would be a hefty(though highly educational) manual that went through lots ofcase examples of where Phase ID can go wrong.
Morals include - Search-match algorithms only crunch numbers and make suggestions on possible phases - the user has to match the=20 pattern. Don't be surprised if most of the suggestions given in the match list are rubbish. The user has to choose the suggestions that are not rubbish. Don't trust what the supplier says about the sample Don't trust reference databases (bit of a worry and somewhat=20 of an oxymoron(?)) Be aware of spurious peaks such as CukBeta (~0.13% in a Philips 1050 based system with Graphite monochromator=20 due to lack of pre-detector slit. Ron Jenkins mention this at a workshop at AXAA 96 in Sydney) Be aware of spurious peaks cause by Tungsten satellites due=20 to aging tubes. Learn how to identify spikes (single point) Mis-set detectors (and scintillation counters) can let high order wavelengths through the monochromator to the detector. Be aware of phases with high intensity low angle peaks=20 that have a habit of sticking to the sample holder irrespective of how well you wash the sample with solvent (first use an abrasive kitchen cleaner). Sharp single peaks can be due to single flakes of sample such as quartz or zircon. These can sometimes be quite intense. Don't smooth your data as this can make background look like peaks. If using a binder or sticking agent, always run the blank. When in doubt, add silicon as an internal position standard and=20 calibrate on it. The diffractometer or sample holder could be=20 out of alignment and the silicon will tell you about it and give you the absolute peak positions. Low element samples (i.e., graphite) can give aluminium (or alloy) peaks due to the sample holder. Run your sample holders, glass plates, low background plates;=20 both spun and unspun to determine if the sample holder generates spurious peaks. This does occur with most spun low background quartz plates. Experience counts. People with 10 to 20 years of experience can confidently identify trace peaks in complex phase systems. (Can be quite amazing seeing someone identify a complex phase system from a chart paper scan in around 5 seconds. Including=20 poorly crystalline phases that are not in the ICDD, and concluding=20 reactor and reaction conditions based on the presence of trace peaks,==20 peak separations and relative peak intensities) Extra information counts - SEM, XRF, Optical, Process Chemistry,=20 Don't be afraid to discuss problems with colleagues as possible=20 solutions and advice from their XRD experience may help. Example data and manuals from vendors do not seem much practical help as they try to show how wonderful/perfect their software is rather=20 than guiding people to perform phase ID. Thus much of their example can be very contrived to the extend of being near totally useless as educational tools. =20 (though people may be able to recommend vendor manuals that do not=20 follow this rule?)
-----
In terms of what I would add to a Phase ID, failure database, the=20following would be high up on the list as obtaining the correct answerin time was more luck than skill.
This involved analysing a sample from a commercial customer who wassure this was just iron oxide of some description. This particulardiffractometer was not 100% in alignment an was due for a full=20servicing - thus small peaks shifts did occur.Running the scan showed the sample was clearly magnetite/iron oxide. =20There was a couple of trace peaks but the overall=20result was rather clear. Before faxing off the result, another=20attempt at identifying the trace phase implied zincite (ZnO).This lead to nasty throbbing at back of head as to whether themagnetite was really magnetite. More search match showed itcould also be franklinite (iron zinc oxide) which has peaksvery close and of similar intensity to hematite. Adding siliconinternal standard and re-running implied franklinite. =20Sending it next door for an elemental XRF showed a decent amount ofzinc in the sample - thus franklinite was assigned.
----
It should be noted that the Franklinite JCPDS card (21-1012) does notlack the equivalent "missing magnetite" 1 3 3 peak at ~1.94=C5. Whichleads to the point on isostructural phases.
Do a search match on the calculated magnetite pattern.I don't have the structures for the following but they seem=20isostructural with magnetite and the PDF-1 is limited in this regard. =20Should the following minerals also have a visible 1 3 3 peak(?):- chromite (34-140) magnesioferrite, ordered, syn (17-464) brunogeierite, syn (25-359) magnesioferrite, disordered, syn (17-464) cuprospinel (25-283)
Trevorite seems to have the 1 3 3.
In the case of inorganic phases that appear isostructural withmagnetite=20but also lack a visible 1 3 3 peak : 39-1158 40-1119 38-1109 40-1120 42-1443 31-1477 6- 532 38- 182 35-1373 29- 881 =20 1-1110 37-1471 34-1400
44-1338 seems to have the 1 3 3 =2022-1086 seems to have the 1 3 340-975 seems to have the 1 3 3
----
Doing the above search match on a Scrutinyite does not seem tohave the same type of hits. =20
Perhaps someone with access to an ICSD database could assist hereto determine which of the above have a significant 1 3 3=20peak? Having cross validation between calculated and rawdata could be quite interesting and assist in the quality ofthe ICDD patterns? Not only direct comparison but also withisostructural phases? Are peaks lacking due to real sample=20effects, poor counting statistics or poor equipment?
-----
Cheers,
Lachlan.
Lachlan Cranswick - Melbourne, Australia _--_|\ =20Phone/Fax : (613) 9455-1345 / \E-mail : lachlan@melbpc.org.au \_.--._/Mobile Phone/Voice Mail : 0412-1141-31 vCrystallographic WWW : http://www.unige.ch/crystal/stxnews/stx/volnteer.htm