Discussion List Archives

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

Re: core CIF discussion #6

  • To: Distribution list of the IUCr COMCIFS Core Dictionary Maintenance Group <coredmg@iucr.org>
  • Subject: Re: core CIF discussion #6
  • From: Howard Flack <crystal@flack.ch>
  • Date: Fri, 18 Jun 2004 14:09:06 +0200
  • In-reply-to: <408D320B.8010209@mcmaster.ca>
  • References: <408D320B.8010209@mcmaster.ca>
> I will start working on the next round after June 30 which
> is therefore the deadline for receiving comments.

   I made it before the deadline even if they are digging up the street 
just outside my office. The whole building is vibrating. I only hope I 
have been thinking clearly with all this noise going on. I went into 
another office on the other side of the building to read and think about 
the text but I have to come back to my own office to type my comments. 
Its so noisy I think I'm going home now.

H

H.D. Flack's comments on Core CIF Discussion #6



####################
#
# Items included below                     Recommendation        HDF's opinion
# ----------------------------------------------------------------
#
# Item 6 Atom_site_refinement_flag_        Editorial changes     Accept Editorial changes
#
# Item 7 diffrn_refln_status               Drop                  Drop
#
# Item 8 _geom_bond_count                  Approve               Approve
#
# Item 9 _refine_scattering_density        Drop                  Drop
#
# Item 10 _atomic_sites_solution_hydrogen  Approve               Approve
#
# NEW ITEM
#
# Item 11  Distributed scattering density  Discussion
#

  HDF thinks it is a good idea but found the text confusing. I hope that DJW has read the text of item 11.  I look forward to hearing whether the proposed CIF encoding corresponds to his current needs and is also flexible enough to allow future developements.

  The title 'distributed scattering density' is a bit misleading, I find. One expects to be able to represent p orbitals or d orbitals or third-order vibration effects under such a title. The objective of item 11 is however more limited. It really sets out to provide a means of representing some forms of disordered atomic centres. The convolution with the distribution of electrons around the disordered atomic centres is not really part of its objective. May be another name would illustrate more readily this goal. When the text talks about 'density' it is the nuclear density of the atomic centres which is implied and not (or usually not) the electronic density of the atoms.

> data_distributed_density_[]
>   _name               '_distributed_density_[]
>   _category           category_overview
>   _type               null
>   _definition
>;              Items in the distributed density category describe the
>              geometric arrangement of an atom or atoms when they are
>              distributed uniformly over a line or surface such as a ring,
>              cylinder or sphere.

  I don't see why you want to limit the scope of this item to UNIFORM distributions. I think the items should be expressed to allow non-uniform distributions even if you only have particular uniform distribution in mind at the moment for implementation.

  Moreover throughout the text there are confusions concerning the geometric objects and distributions which are presented. In the last sentence of the _definition, neither a ring nor a cylinder nor a sphere is either a line or a surface. The distributions of the nuclear density described in item 11 are always 'fat' or 'thick' as you specifically allow for this through the Uiso of the distribution. There are no nuclear-density distributions here which are spread along a line or over a surface. You can not really even call your thick lines as cylinders because the Gaussian distribution associated with Uiso gives them a fuzzy nature. How do you square this with 'they are distributed uniformly'. Its no good trying to use the mean nuclear position either because, for the ring, this is the centre of its defining circle (where in practice the atoms don't go). 

>  _distributed_density_Uiso

  Why limit yourself to isotropic?

>   _example  ;  The distribution was modelled using a disk of scattering
              density of the given radius.

   I don't think this is a very good example. You only need to include 'disk' as one of the possibilities in _distributed_density_shape .
   Is it really a disk? A disk for me is infinitely thin and of a well defined radius. With Uiso .ne. 0 the atoms are fuzzy and the disk has blurred edges and a definite thickness.

>data_distributed_density_direction_
>loop_    _name                                   >'_distributed_density_direction_h'
>                     '_distributed_density_direction_k'
>                     '_distributed_density_direction_l'
>   _category         distributed_density
>   _type             numb
>   _type_conditions  esd
>   _list             both
>   _list_reference   '_distributed_density_id'
>   _units            rlu
>   _units_detail     reciprocal lattice units
>   _example          ?
>   _definition
>            The direction ratios (in reciprocal lattice units) of the line of
>            distributed density or of the axis of the cylindrical shell or
>            ring of distributed density.             Note that in a >non-orthogonal coordinate system the direction
>            ratios of a line are expressed in terms of the reciprocal
>            rather than the real lattice.

Is it intended that the length of this vector have any meaning? Is so state what it is. If not define the values to represent a vector of unit length or of arbitrary length according to what is the most suitable. How about the following the text for _definition:
  The (covariant) components on a reciprocal-lattice basis of a vector of unit/arbitrary length used to indicate e.g. the axis of a cylinder or the axis of cylindrical shell or the normal to the plane of a circle.

>data__distributed_density_length
>   _name             '_distributed_density_length'
>   _category         distributed_density
>   _type             numb
>   _type_conditions  esd
>   _list             both
>   _list_reference   '_distributed_density_id'
>   _enumeration_range 0.0:
>   _units            A
>   _units_detail     Angstrom units
>   _example          ?
>   _definition
>;              The length of the line or cylinder of distributed density in
>              Angstrom units.  If this number is set to the translation
>              repeat in the direction of the axis of the distribution,
>              it describes an infinite line or cylinder of scattering
>              density running through the crystal.
>;

  These 'distributed densities' do not really occur as either a line or a cylinder as they are always fuzzy due to Uiso. If you drop the 'of distributed density' from the first line of _definition it makes sense. It seems to me that from an operational point of view it is most unwise to use a translation repeat to indicate infinity. It's better either to use a negative value for this purpose (lengths by definition are >= 0) or  to have different 'shapes' in _distributed_density_shape for the 'line segment' and 'line' {i.e. the one of infinite length} 'cylinder'/'cylindrical shell' and 'infinite cylinder'/'infinite cylindrical shell' {does it have a particular name}. The Fourier transforms of the finite/infinite objects have different functional forms. What does DJW think about all of this?

>data__distributed_density_shape
>   _name             '_distributed_density_shape'
>   _category         'distributed_density
>   _type             char
>   _list             both
>   _list_reference   '_distributed_density_id'
>loop_      _enumeration
>   _enumeration_detals line       ?
>                       ring       ?
>                       cylinder   ?
>                       sphere     ?
>                       other 'Give details in _distributed_density_details'
>   _definition
>;              A flag that indicates the type of distributed density.  The
>              line and ring are one dimensional arrangements of charge and
>              the cylinder and sphere are two dimensional charge shells.  In
>              each case the thickness of the distribution is determined by
>              the scattering factor of the element and the atomic
>              displacement parameter defined in _distributed_density_Uiso.
>;
  
  
  I'm happy with 'ring' as it implies a 'thick' or 'fuzzy' circle.
  I'm most unhappy with 'line' as it has no thickness. As from above you really need something inspired from 'line segment'. Likewise I'm unhappy with cylinder and sphere as they have such a precise meaning in mathematics and I rather suspect that what you really wanted was something based on 'cylindrical shell' and 'spherical shell'. May be also 'disk' or something implying 'fuzzy disk'. I think also that in 'In each case the thickness of the distribution' you need to include 'root-mean-square' in front of thickness. Also, again DJW, is there any need for Uaniso.

>data__distributed_density_Uiso
>   _name             '_distributed_density_Uiso'
>   _category         distributed_density          _type             numb
>   _type_conditions  esd
>   _list             both
>   _list_reference   '_distributed_density_id'
>   _enumeration_range  0.0:
>   _units            A2^
>   _units_detail     'Angstrom units squared'
>   _example          0.018(3)
>   _definition
>;              The value of the isotropic atomic displacement parameter
>              applied to each portion of the distributed density.  This
>              parameter together with the scattering factor of the atoms
>              whose density is distributed, indicates the thickness of the
>              line, ring or shell of density.
>;

   In _definition I think it most inappropriate to talk about the atomic scattering factor as contributing to the thickness. All of item 11 is only concerned with nuclear distributions. Also it is not the VALUE of the isotropic atomic displacement which is applied to each portion of the distributed density. The Uiso is a parameter of the Gaussian distribution describing the spread of each volume or line element of the underlying nuclear density distribution described by the chosen geometric object i.e. a line segment, a cylindrical shell, a spherical shell (or segments thereof), a circle, a circular segment. Again I'm convinced there is a use for non uniform distribution of atomic centres on these geometric objects even if the current programmes can not handle such complexity.


  Essentially it's good and useful stuff but it needs to be defined with much more clarity.

H. 

_______________________________________________
coreDMG mailing list
coreDMG@iucr.org
http://scripts.iucr.org/mailman/listinfo/coredmg

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