CIF (S.R. Hall, F.H. Allen and I.D. Brown, Acta Crystallogr., A47 (1991) 655-685) is an extremely flexible format for deposition of crystallographic data and is already the method of choice for the transmission of data to the Cambridge (organic) and Inorganic structural databases. At the end of a structure refinement with SHELXL-93 two archive files are produced: *.cif contains the crystal data, atomic coordinates, bond lengths etc., and *.fcf contains observed and calculated structure factors, both in CIF format. The .fcf file requires no further processing but the .cif file must be edited by the user to include items such as the crystal color that even the most sophisticated program cannot deduce from the diffraction data; this editing takes the form of replacing a question mark with the appropriate information. With a little practice it is perfectly possible for humans to read CIF files, but SHELX users are encouraged to use the program CIFTAB (supplied with SHELXL-93) to produce more tasteful tables of crystal data, bonds and angles, structure factors etc. for referees.
All refinements with SHELXL-93 are performed with F2 rather than F. This enables ALL data to be used rather than only data with F greater than a specified threshold, with the result that the experimental information is more fully exploited. For weakly scattering crystals this can appreciably improve the precision of the structure determination. However the R-index:
wR2 = (Sum[w(Fo2 - Fc2)2] / Sum[wFo4])1/2
that (in the absence of restraints) is minimized during the refinement is for statistical reasons about twice as large as the conventional index R1 (based on F):
R1 = Sum | |Fo| - |Fc| | / Sum |Fo|
and to make it worse an R-index based on ALL data is inevitably larger than one based only on data with F greater than a given threshold. It is rumored that a leading journal (that comes out in both German and English editions) has already rejected several papers reporting structures refined with the new program because "the R-factor was too high"!
Note that wR2 should not be confused with 'wR' or 'Rw' (both usually based on F). Since weighting schemes for F-squared and F refinements are quite different, it may not even be possible to calculate a meaningful Rw if the structure has been refined against F-squared.
For comparison with other (older) structures it is however very desirable to quote a conventional R-index (i.e. R1) calculated with a threshold of F2 > 2sigma(F2) [that effectively corresponds to F > 4sigma(F)]. R1 also has the advantage that it is relatively insensitive to manipulation of the weighting scheme. For a SHELXL refinement the rigorously defined but perhaps not always intuitively obvious CIF rules require wR2 (for all data) to be called '_refine_ls_wR_factor_all' and R1 (based on F for F2 > 2sigma(F2)) to be called '_refine_ls_R_factor_obs'.
Some very observant referees have drawn authors' attention to the fact that they had not fixed any coordinates to define the origin in polar space groups. This is because SHELXL-93 automatically uses the mathematically superior 'polar axis restraints' proposed by H.D. Flack and D. Schwarzenbach (Acta Crystallogr., A44 (1988) 499-506) that restrains a suitable weighted sum of atomic coordinates to be constant. Similarly the Flack 'racemic twinning parameter' is always estimated by the program where appropriate and reported in the .cif file (but not used in the calculation of Fc2 unless specified by the user), so that it is unlikely that the user will fail to notice when it is necessary to determine the correct 'absolute structure' (H.D. Flack, Acta Crystallogr., A39 (1983) 876-881).
SHELXL-93 uses scattering factors and absorption coefficients from the new Volume C of International Tables for Crystallography (1992), so there will be small discrepancies with values of mu etc. calculated by programs that still use older values.
SHELXL-93 estimates esds in bond lengths, angles and torsion angles from the full covariance matrix. The contributions of the cell esds are also included rigorously, except that the (usually unknown) correlations between the cell parameters are ignored unless defined by the crystal symmetry (e.g. the error in a cubic cell dimension affects the esd of a bond length but not of an angle). The esds in the equations of least-squares planes and in the distances of atoms from such planes are also calculated from the full matrix, but the (small) contributions of the cell esds to these esds involve some approximations. Thus there will be discrepancies with esds calculated for checking purposes by programs that do not have access to the full covariance matrix etc.
SHELX is supplied free of charge to academics on condition that they reference the program in any published work in which it was used. The correct literature reference to the structure solution program SHELXS-86 is:
G.M. Sheldrick, Acta Crystallogr., A46 (1990) 467-473.
A suitable reference for the structure refinement program SHELXL-93 will be:
G.M. Sheldrick, J. Appl. Cryst., (1993) in preparation.
Back to Appendices
Back to Table of Contents