Discussion List Archives

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: comments on coreCIF.dic 2.1

  • To: Multiple recipients of list <coredmg@iucr.org>
  • Subject: Re: comments on coreCIF.dic 2.1
  • From: Howard Flack <Howard.Flack@cryst.unige.ch>
  • Date: Wed, 20 Jan 1999 16:46:40 GMT
Dear All but especially David,

>10. _diffrn_radiation_polarisn_norm
>11. _diffrn_radiation_polarisn_ratio

>         I am willing to wait for this advice.  This does not prevent
> making a clarification in the _*_norm definition, however.

  I looked up Azaroff's book and one of his papers. My memory had played
tricks on me. According to the book, (Elements of x-ray
crystallography), he used the perpendicular/parallel nomenclature in the
description of a very old experiment to prove that X-rays were
electromagnetic radiation! In the paper "Polarization correction for
Crystal-monochromated X-radiation" Acta Cryst (1955). 8, 701-704 he uses
a sigma and pi instead of perpendicular and parallel. But in any case
the publications are as old as the hills and the people needing these
items may very well use something quite different.
  The definitions are bad or incomplete, I am very well prepared to
admit but both need input from people that have need of them now. In
order to allow Brian to finish this release, I suggest that we leave
them as such at the moment and ask the experts (during the Glasgow
congress) what they really need and how they do things.

> > 
> > (3) data_diffrn_measured_fraction_theta_full
> I agree with Brian here that the dictionary definitions should reflect
> what is possible, not what is desirable. 


>  (4) data_diffrn_refln_[]
>   By all means add parentheses.


> (6)  data_diffrn_reflns_[] 
>         Same as above.
  OK but note the change of order as well. Everything on _diffrn_reflns
first followed by the note on _diffrn_refln in parentheses.

> (17)  data_refln_symmetry_multiplicity
>        The Fs may be different but their MAGNITUDES (as mentioned in the
> definition) are the same if one ignores the imaginary components of the
> scattering factors.
  The general, and nowadays usual, case is that the imaginary component
is significant, even if small. That is the one I always have in mind.
Vanishingly small f''s are a special case.

>   I think some clarification is needed in the case of
>   non-centrosymmetric structures,

  Yes I agree. The mention of structure-factor magnitudes as a criterion
makes things very murky indeed.  The criterion must be only one of
symmetry. I would write the existing definition as follows:

Definition L:    The number of reflections symmetry-equivalent under the
Laue symmetry
                 to the present reflection. In the Laue symmetry, 
                 Friedel opposites (h k l and -h -k -l) are equivalent.
Tables of
                 symmetry-equivalent reflections are available in
International Tables
                 for Crystallography, Volume A, section 10.2 .

  This mentions the Laue symmetry in which Friedel opposites are
equivalent. A definition of something related would be:

Definition C:    The number of reflections symmetry-equivalent under the
crystal point
                 group to the present reflection. Only in
centrosymmetric point 
                 groups are the Friedel opposites (h k l and -h -k -l)
                 Tables of symmetry-equivalent reflections are available
                 International Tables for Crystallography, Volume A,
section 10.2 .
  This uses the crystal point group. The two values (L and C) are
identical in the case of a centrosymmetric structure.
  As in pure mathematical form, the count includes the basic reflection
because it is always equivalent to itself.
  If both of these definitions are useful, they would need to be given
separate names. Which one is useful or more pertinently is in use? The
multiplicity does not enter in single-crystal least-squares calculations
(as far as I can see). The multiplicity corresponding to definition L
always enters into electron-density calculations (Fourier maps) for
single-crystal and powder work. (The Fourier transform of the electron
density gives the Fs without dispersive contribution, both f' and f''.
Doing the inverse transformation, the 'Fobs' need to be 'corrected' for
the effect of dispersion, f' and f''.). What about Rietveld refinement
for powders)? hkl and -h -k -l always overlap but in the
non-centrosymmetric case do not have the same intensity so it looks as
though one should use definition C for intensity calculation and L for
the electron density. In direct methods calculations and normalisation,
it is the parameter epsilon which appears most of the time. Does
multiplicity appear? I'm not sure. Probably one needs L if one is not
exploiting differences between Friedel opposites and C if one is. The
choice of using L or C seems much more to depend on the calculation to
be undertaken than on the symmetry of the structure itself.  

  Conclusion: The existing definition of _refln_symmetry_multiplicity is
confusing but it needs input all round to find something useful to
correct it.

> >13. _diffrn_refln_intensity_sigma
> >    _diffrn_reflns_av_sigmaI/netI
> >    _diffrn_reflns_class_av_sgI/I
> >    _diffrn_standards_scale_sigma
> >    _refine_ls_weighting_scheme 
> >    _reflns_threshold_expression
> >    _reflns_shell_meanI_over_sigI_ (two data items)

  It seems to me that there is no implicit or explicit requirement in
any of the above definitions that the uncertainties there referenced
should contain only contributions estimated from the Poisson counting
statistics. I always understood these uncertainties to be much more
general and capable of including whatever one thought wise and expedient
to put there. I do not think that there is any specific data item in
which one can describe how e.g. _diffrn_refln_intensity_sigma is
calculated. For us it defintely includes contributions from the
uncertainties of the attenuator, the scales, the stability etc and is
never pure Poisson.

  The agreed symbol for standard uncertainty is u.  So:

     _reflns_shell_meanI_over_uI_ (two data items)
    would be fine for me. I expect that there are other sigma, sig, sg,
etc in the core dictionary. 

Best wishes,
Howard Flack        http://www.unige.ch/crystal/ahdf/Howard.Flack.html
Laboratoire de Cristallographie               Phone: 41 (22) 702 62 49
24 quai Ernest-Ansermet             mailto:Howard.Flack@cryst.unige.ch
CH-1211 Geneva 4, Switzerland                   Fax: 41 (22) 702 61 08

[Send comment to list secretary]
[Reply to list (subscribers only)]