Pavel: In my original email I very carefully gave credit for IMPLEMENTING the TLS concept. Of course the ideas and some programs had been around long before, but it was the IMPLEMENTATION IN REFMAC that resulted in TLS becoming widely used. I had actually considered putting it into SHELXL but had not done so for two reasons (a) I was too lazy and (b) I missed an essential trick that REFMAC introduced, namely the combination of TLS with an additive isotropic B-value for each atom.
Dale: You are quite correct that AFIX 137 breaks my argument about not depositing (SHELX) hydrogen atoms because they can be recalculated with no loss of experimental information. However to be fair, if you generate the first .ins file using SHELXPRO (the recommended procedure) you will get AFIX 33 that doesn't have this problem. For Pavel and others unfamiliar with SHELXL, AFIX 33 is a pure riding model with a staggered methyl group but AFIX 137 assumes local threefold symmetry, finds the initial torsion angle by a three-fold averaged fit to the difference density and then refines the torsion angle in the following cycles. Since this torsion angle is not given explicitly in the output files, if AFIX 137 hydrogens are not deposited, they cannot be regenerated except by a full refinement against the experimental data. George Prof. George M. Sheldrick FRS Dept. Structural Chemistry, University of Goettingen, Tammannstr. 4, D37077 Goettingen, Germany Tel. +49-551-39-3021 or -3068 Fax. +49-551-39-22582 On Wed, 15 Sep 2010, George M. Sheldrick wrote: > Dear Pavel, > > May I suggest that you take a look at the SHELX manual: > http://shelx.uni-ac.gwdg.de/SHELX/shelx.pdf > before sending your SHELX questions to CCP4bb? You might > even find some good ideas for implementing in phenix_refine! > > For example, > if you look up 'non-crystallographic symmetry' in the index > you will discover that SHELXL applies NCS in the form of > restraints, not constraints, which has the advantage that it > can be applied locally and in combination with all other > restraints and constraints involving the same atoms. However > you will not find TLS in the index, because the credit for > implementing this very useful concept should be given to > Martin Winn, Garib and Ethan, long after the current version > of SHELXL (and its manual) were released in 1997. And because > SHELXL only reads one instruction file (*.ins) and one > reflection file (*.hkl) but no other data files or libraries, > and FORTRAN will always be FORTRAN, the deposition of these > two files would be sufficient to define the refinement for > posterity. > > Best wishes, George > > Prof. George M. Sheldrick FRS > Dept. Structural Chemistry, > University of Goettingen, > Tammannstr. 4, > D37077 Goettingen, Germany > Tel. +49-551-39-3021 or -3068 > Fax. +49-551-39-22582 > > > On Wed, 15 Sep 2010, Pavel Afonine wrote: > > > Dear Felix, > > > > > Shortly Shelxl knows crystallography best. > > > > I have no doubts about this. May questions were, though: > > > > > > - how .ins file encodes the information about NCS groups used in > > > > refinement (atom selection for NCS groups, restraint weights for > > > > different > > > > groups, etc? > > > > > > > > - how .ins file encodes the information about TLS (again, atom > > > > selections > > > > for TLS groups, TLS matrices, etc)? Related, does it have a concept of > > > > having TLS and other components to the total atomic displacement > > > > parameter > > > > (ADP)? > > > > > > > > - If I recall it correctly, to fix (=not refine) a certain parameter > > > > (say > > > > occupancy or B-factor) in Shelxl you need to add a number 10 to it. Is > > > > it > > > > true? IMHO, this might lead to confusion if such a file gets deposited > > > > to > > > > PDB. > > > > Best, > > Pavel. > > > > > >
