Re: [sdpd] RIET:STX:Clay:CCP14: Minor Crysfire 2000 powder inde

[Robin Shirley <R.Shirley...@surrey.ac.uk>]

Jon and fellow Rietvelders/sdpd-ers

> Does this just mean dividing the observed d-spacings by 2 (or 3,4..)
> to get more "observed" values, or am I missing a more cunning trick?

In essence that's exactly what lies behind it, though perhaps a bit more
systematically than most humans would consider worth grinding through.

For example, it works iteratively, making overlap-avoidance tests at each
step, and produces the minimum extension set using the lowest available
orders.

But the bottom line is that no more information is injected into the
pattern by this process - it would be impossible for things to be
otherwise, since to win more degrees of freedom you have to make more
actual observations (which is sometimes not experimentally possible).

At least it would be impossible to conjure up more information (i.e.
degrees of freedom) just through working from a fixed set of observed
peak positions - in principle there might be more information to be
found buried beneath the profile, though in practice it is usually very
low-grade.

In fact, not only does pattern extension *not* add more information, but
if the original data contained systematic errors, then the extra
higher-order lines will not reproduce these in a self-consistent way -
all available correction and calibration should always be done *before*
pattern extension.

Having said all that, it is still a great convenience to be able to
extend unavoidably sparse patterns from High-P/High-T experiments in a
few seconds to give a format that none of the indexing programs will
object to.  For good (though sparse) data, this can open the door to
getting cell (or at least sub-cell) information that can otherwise be
hard to lock onto.

The authentic pattern that Lachlan supplied me for testing these issues
(yes, it was Lachlan's turbo-charged lobbying on behalf of his
High-P/High-T clientele that lay behind this) originally contained only
7 lines.  As you can imagine, it was tough to interpret in that form,
but becomes significantly more stable after pattern-extension, yielding 
some quite interesting possible solutions (or partial solutions).

But it's certainly a case of caveat emptor, and is intended specifically 
for folk who know what they're doing.

I'll try to put together a short paper on this when a bit more experience
has been gained on how it usefully it performs.

Robin Shirley

P.S.  "All very fine, but when will it be available?", I expect you're
muttering.  If nothing bad emerges from the current round of tests, the
initial Crysfire 2002 release looks like being a matter of days rather
than weeks.  In fact I'm considering putting up the current academic beta
on CCP14 tomorrow, for comment.

--------------------------------------------------------------

To:            sdpd...@yahoogroups.com
From:          Jonathan WRIGHT <wright...@esrf.fr>
Date:          Wed, 10 Jul 2002 15:50:16 +0200
Subject:       Re: [sdpd] RIET:STX:Clay:CCP14: Minor Crysfire 2000 powder
               indexing suite update
Reply-to:      sdpd...@yahoogroups.com

>For those who try to index on the limit of the abyss of impossibility
>(by having less than 20 peaks) - the new Crysfire 2002 will have "pattern
>extension" to trick various programs that insist of having not less than
>20 peaks into still running.

Lachlan/Robin,

Does this just mean dividing the observed d-spacings by 2 (or 3,4..) to get 
more "observed" values, or am I missing a more cunning trick?

Jon

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Free $5 Love Reading
Risk Free!
http://us.click.yahoo.com/TPvn8A/PfREAA/Ey.GAA/UIYolB/TM
---------------------------------------------------------------------~->

 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/