|
Number 13 - February 2001
ICSD USER'S GROUP NEWSLETTER
(Inorganic Crystal Structure Database)
edited by
I.David Brown
Brockhouse Institute for Materials Research
McMaster University
Hamilton, Ontario, Canada L8S 4M1
The following email exchange may help to clarify some of the
unexpected space group assignments that can be found in the ICSD.
On Fri, 12 May 2000, Michael O'Keeffe wrote:
I have been looking into the occurrence of certain space
groups and find that there are VERY many errors. My guess is that
MOST of the entries for the pure rotation classes are wrong (the
structures reported actually have higher symmetry). Sometimes
this is obvious, but sometimes less so. There should be some
automatic way of checking the symmetry of each entry. The problem
is not trivial as the data base is used to screen for certain
symmetry properties such as absence of a center of symmetry.
Example: look at F23. Most of the entries obviously have symmetry
Fm-3m (NaCl or CaF2 structures). Note that BeO is in there too
with the NaCl structure. Obviously wrong on two counts. All the
entries for I432 are actually Im-3m. Etc., etc.
David Brown replied:
Thank you for your message which I am copying to Vicky Karen
for information. She is the person at NIST who is preparing the
new software for the ICSD and she should be kept informed of the
kinds of problems people encounter.
The problem you raise about wrong space groups is probably
encountered mostly in the cubic space groups and is the result of
a policy we adopted early on of reporting the structure as the
author reported it. At that point we did not want to risk
putting more errors into the database by trying to correct an
author's interpretation which we thought might be wrong. (We did
try to correct obvious numerical errors). In the early days of
crystallography is was customary to report the space group with
the *minimum* symmetry necessary to describe the structure.
Nowadays we look for the space group with the *maximum* symmetry.
Thus early structure determinations of Fm-3m crystals are often
reported in F23 because there was no way of knowing for sure that
the center of symmetry was present. Another consequence of this
policy is that wrong structures are also included in the ICSD
(they may, after all, be correct structures obtained from a
non-standard sample). Such structures should, however, have
pointers to the correct determination (pointers should be
included to all determinations of the same compound, but this is
not trivial to do).
While I will defend the decision not to try to re-interpret
the author's original description, there is certainly an argument
(which increases as we get more effective database software) for
introducing some standardisation. Bergerhoff has introduced his
SICS (Standardised Inorganic Crystal Structures database) in
which he has selected the best determination of each structure
from the ICSD and has transformed it to a standard setting which
allows him to look for similar structures. This is one possible
way around the problem, but it does put the user one step away
from the original report which may contain important information
that might be lost in the standardisation. Another possibility
is to introduce into the ICSD new records which contain the
reinterpreted information. One could then ask the retrieval
programs to use either the standardised records or the original
records, but both would be available for inspection.
As you say, preparing the standardised form is not a trivial
job, but a couple of dedicated people (Bergerhoff and his team,
as well as Allmann as a retirement project) have worked through
much of the database trying to clean it up. However, any clean
up should leave a record of what has been done so that errors
introduced during the editorial reinterpretation can be spotted.
Mike O'Keeffe responed:
Thanks for the quick and detailed response to my note. At the
risk of exceeding your patience, may I have a few more words on
the question of symmetry and the use of symmetry groups in the
description of crystal structures.
Crystallography is in the enviable position of being able to rely
on exact mathematical results for its underpinnings, and I think
that all of us who care about the subject should treat these
basics with respect. The point I raised was not whether NaCl
might have symmetry F23 or P1 or no symmetry at all; after all
(to be pedantic) we can only prove the absence of symmetry, not
its presence. The question I raise is the different one of the
symmetry of the REPORTED structure. To say that NaCl with a cubic
cell and Z = 4 has symmetry F23 is indefensibly wrong and
recognizably wrong (one needs at least one set of general
positions for symmetry F23, i.e. Z >= 48). Yes I know that your
comuter will generate the NaCl structure using F23 (mine will
anyway) but surely that is not the whole point of using symmetry.
Some other reasons for caring: (a) I have in front of me an
analysis of the frequency of occurrence of space group symmetries
in ICSD, clearly its interpretation is open to question. (b)
Recently a list of all non-centrosymmetric oxides from ICSD was
published. It is likewise greatly corrupted. (c) I am interested
in discovering structures based on simple organising principles
that are non-centrosymmetric (packings of equivalent cylinders is
one), and also in finding if these are used in crystal structures
(hence my recent interest in I432). (d) Of course many important
physical effects such as second harmonic generation or
piezoelectricity are related to (the absence of) symmetry. If
software is being written to screen ICSD entries for errors of
the sort I refer to, may I call attention to Table 14.2 of the
International Tables which appears not to be well known, even to
crystallographers. What is needed is a listing of the symmetries
of all combinations of Wyckoff positions for all the space
groups. ICSD could then report both the space group used by the
author and the actual symmetry of the reported structure. I
suggest the latter be used as the default for searches.
Michael O'Keeffe
Regents' Professor
Department of Chemistry
Arizona State University
Tempe AZ 85287
Tel: (1) (480) 9653670
Fax: (1) (480) 9652747
http://www.asu.edu/clas/chemistry/faculty/okeefe.htm
Message from Alan Hewat.
The latest 15 Jan 2001 version of ICSD-for-WWW contains
several new features requested by users. The new features are
described on:
http://barns.ill.fr/dif/icsd/news.html and may be
tried on the ILL's own server
http://barns.ill.fr/dif/icsd/ In particular they include:
1) Acrobat PDF output of calculated and indexed powder patterns
(as an alternative to postscript)
2) Chime/PDB 3D-display of structures (as an alternative to VRML)
3) Username/Password access (as an alternative to IP names) see:
http://barns.ill.fr/dif/icsd/icsd-password.html
4) Various minor bug fixes (with thanks to all for reports and
suggestions)
The present ICSD-for-WWW database from FIZ-Karlsruhe contains
57,195 entries. If you run your own ICSD-for-WWW server, this
version for SGI, Sun and HP servers is available from:
ftp://ftp.ill.fr/pub/dif/icsd/
|
