Lachlan,
a most interesting letter that I'm going to reply to piecemeal. Comment is
purely my own and you should check with others.
Tony Raftery.
>
>Not sure if this has been brought up before or it may be common
>knowledge in some quarters?
>
>I have been giving some commercial Powder XRD training seminars -
>including search-match on PDF-1 database using commercial software
>the laboratories have purchased (in this case Philips PC-Identify). =20
>I am curious about the following and am interested in other know=20
>of similar phases and/or want to trade some phase ID pitfall stories.
>
>----
>
>Of the many and various issues on the pitfalls of search-match:
>
>1) One of the points I like to mention about Phase ID is that you
> can't trust the person who gives you the sample. (That "Calcium
> containing phase" may really be an Aluminium phase?)
>
*** True but some information is better than none. I tend to make it
standard 'policy' to require the client to tell me what is in the sample or
why the sample is of interest.
=20
>2) The other is not placing 100% trust in reference data. Which you
> may not realize/appreciate unless you encounter this=20
> the hard way. =20
>
>I am curious if there are other examples of the following??
>
>-----
>
>One of the pitfalls I see people fall into (having beaten
>the addiction myself after much pain) is failing=20
>to identify a trace peak - that turns out to be due to
>a major phase - as the ICDD data does not include the trace=20
>peak in the card. In many cases, this only comes to light when=20
>doing a quantitative Rietveld analysis on the sample. A spurious,
>problem trace peak that stood out like (to use the common=20
>Australian expression) "dog's balls" disappears into the Rietveld=20
>fit, which has been "correctly" modeled as part of one of the=20
>major phases.
>
>
**** I tend if I can to run the pattern through Siroquant (if I can, and
mostly I can as they tend to be geological), to look at the residual and the
truely unidentified.
>
>An example of this is to do a search match on a lead-acid battery
>sample supposedly containing (PbO2) ~50% plattnerite, ~50 Scrutinyite. =20
>Search on the PDF-1 database and you identify this=20
>as 45-1416 - Scrutinyite and 35-1222 - Plattnerite - Syn.
>
>But this leaves a peak at 2.436=C5 (36.87=B0 2-theta with CuKa). If you=20
>are not used to this, trying to identify this peak can put you=20
>in a world of pain - or perhaps more hair loss - or perhaps=20
>matching the peak with an incorrect phase (perhaps resulting
>in an expensive goose chase trying to eliminate a phantom phase
>from a product - or rejecting a million dollar shipment as impure?)
>As the saying goes - send lawyers, guns and money - though in the=20
>above it is likely that only lawyers would be involved?
>However, do a quant Rietveld refinement on the sample and=20
>this fits the trace peak as:
> h k l 2-theta
> 1 1 1 36.988 =20
>
>This also matches a very minor hump at 2.218=C5 (40.68=B0).
> 1 2 0 40.691 =20
>
>(In this case, and it hindsight, it is interesting that is you=20
>do a PDF-1 search on deleted patterns, you can match the=20
>deleted 25-447 Plattnerite which does match all the most
>obvious trace peaks - including the minor hump at 2.218=C5 (40.691=B0))
>
>----
>
>Another looks like magnetite?
>
>A sample with magnetite can be matched with ICDD/JCPDS card=20
>19-629 Magnetite. This leaves a trace peak at 1.929=C5 (47.19=B0)
>
>Do a Rietveld quant analysis and this is found to be the:
> h k l 2-theta
> 1 3 3 47.156
>
>----
>
>I'm not sure if people are used to "experience/gut feeling"=20
>when determining the sensitivity of Powder XRDs to trace peaks.
>When helping out with a Rietveld Quant Analysis course I had
>a query from two people who were refreshing their knowledge.
>They were very curious about a trace impurity peak in the fluorite
>standard we were using for this course at 1.57=C5 (58.6=B02theta CuKa). =20
>They did XRD since the "good old days" and never saw a fluorite=20
>standard with an impurity peak in that position. Clicking on the=20
>"run Rietveld" button showed it to be the 2 2 2 reflection.
>This is routinely visible as a trace peak when run on 1050 based=20
>Philips XRD with Graphite monochromator - And I assume with any=20
>other good diffractometer(?) - Rietveld style data collection.
>
>----
>
>Modern Powder XRDs are easily this sensitive to detect trace
>peaks on routine, quick phase identification scans - definitely=20
>with Scrutinyite - maybe with Magnetite (at last with standard 1050
>base Philips gear I have used). These are not what I would call=20
>rare or obscure minerals(?).
>
*** As you know, there are other sources of 'trace' peaks such as lambda/2
and K-beta for 'strong' . These can be verified if you are aware of the
possibility.=20
>----
>
>While discussed before, I think a case can be made for having
>multiple data types (single crystal calculated, rietveld,=20
>transmission, reflection) for phases in the ICDD/JCPDS database.
*** I could not agree most strongly, there should be a calculated pattern
for every common or mineral for which there is a structure. Calculated
patterns are tricky, particularly for low level lines, my personal choice is
to 0.1% relative.=20
>
>Could a future for Rietveld Analysis be as part of a future,
>reliable, black box, expert, phase identification system? =20
>Press a button, it not only does the phase ID but then the=20
>quant analysis?
*** As I said before, the difference pattern is worth having. Another reason
I tend to put them through is that there are difficulties, say with the
feldspars, it is easy to say there is a feldspar but there usually are two
(sometimes three and sometimes four in sedimentary samples). You can't on
occasion decide which is the greater component. Also, I don't carry relative
scale factors in my head. The halite in a sample is intense but is it a
major (> 10%), a minor (<10%, >1%) or trace (<1%)?
>
>----
>
>Hope this is legible at this time of the morning! :-)
>
>Cheers,
>
>Lachlan.
>
>PS: On another point. Would anyone feel it safe to be performing=20
>search match with only the PDF-1 data? As an example, and from memory,=20
>it is easy to match various Lead Oxides to some samples. However, it is
>the information in the PDF-2 that tells you that some of the matches
>are high pressure high-temp phases - and would not be expected in
>a room temperature sample without some good explaining to do.
>
>If someone has purchased a PDF-1 database with their search-match
>software, does anyone know of a cheap upgrade path to a PDF-2 database?
>Having to pay US$6,100 for the CD-ROM from scratch sounds rather cruel
>when you have already forked out a major (possibly excessive) amount
>of money for a PDF-1 database?
* I don't know the policy. It has been some time since it has ceased being
put in the catalog. As I recall, most people looking to upgrade have a
current inorganic license but not a organic one. I would expect to pay no
more than half the current 'new' price. But the people with PDF-1 licenses
are those usually tied to manufactures software. One would need to ask its
price (from manufacturer) and compare it to the ICDD license upgrade cost.
I would not want to be in a position to do S/M without PDF2 because of the
supplementary info.
>I can't see evidence of an PDF-1 to PDF-2 upgrade path in=20
>the ICDD Web page at http://www.icdd.com???
>(I have already given one PDF-1 database user a near heart attack
>saying they need the PDF-2 CD-ROM. Hopefully this is not considered=20
>improper/dodgy advice for doing phase ID on complete unknowns?)
>
>Another query, what are the recommended counting statistics or
>data collection times for running a Grant-in-Aid pattern?
>Using Rietveld to calculate a pattern could help you determine
>the appropriate data collection conditions?
>
>Lachlan Cranswick - Melbourne, Australia _--_|\ =20
>Phone/Fax : (613) 9455-1345 / \
>E-mail : lachlan@melbpc.org.au \_.--._/
>Mobile Phone/Voice Mail : 0412-1141-31 v
>Crystallographic WWW : http://www.unige.ch/crystal/stxnews/stx/volnteer.htm=
=20
>
>