Chemical Crystallography Laboratory
University of Oxford

This short bit of text is rendered with the "display:none" style. If you can see it you may wish to upgrade to a web standards compliant web browser... Otherwise, don't worry, the content of this site is still accessible!

Frequently Asked Questions

Chapter 7: Refinement

7.1: What infomation does CRYSTALS need for least squares refinement?

7.2: Refinement instructions

7.3: How do I refine the occupancies of an atom or fragment split over two sites?

7.4: How do I add restraints?

7.5: How do you use Platon's Squeeze from CRYSTALS?

7.6: What are parts and groups?

List 1 - Cell
List 2 - Symmetry
List 3 - Scattering factors
List 4 - Weighting scheme (default - unit weights)
List 5 - Parameters (model)
List 6 - Observations (reflections) AND/OR List 16 - Observations (restraints)
List 12 - Refinement instructions
List 23 - Refinement options
List 28 - Reflection acceptance criteria (default - all)
List 29 - Asymmetric unit contents

List 12 can be entered in the following manner:

  \LIST 12
  FULL I(1,X'S,U'S) C(7,X'S) O(1,Y) O(1,U[ISO]) UNTIL C(6)
  GROUP C(1) UNTIL C(6)
  END
 

This example refines the position and Uij's of I(1), the position of C(7), the y-coordinate of O(1) and the isotropic thermal parameters of atoms O(1) until C(6) based on the order they appear in list 5. It also refines the positions of atoms C(1) to C(6) as a rigid-body.

Alternatively, choose Guided Automatic from the Refinement menu to run a script which automatically writes List 12 and refines the structure based on its assessment of the current state of progress.

Use constraints. Constraints are set up using List 12, so you must be prepared to eschew the GUIDE and write your own constraint instructions. Here is how:

Step one:
Before you do anything, ensure that the occupancies of the two sites add up to one. A constraint only affects the shifts applied to parameters: if the sum of the parameters is not sensible to begin with, it never will be.
There are several ways to do this. In these examples the occupancy of each fragment is set to 0.5:

   1. Use the graphical interface - right click an atom, choose Edit from
      the popup menu and change the occupancy in the dialog box. Click Apply.
      Repeat for each atom. 
   2. Use the \EDIT command (changing the atom names, of course):
           \EDIT
           CHANGE C(1,OCC) C(10,OCC) UNTIL C(15) 0.5
           CHANGE C(101,OCC) C(110,OCC) UNTIL C(115) 0.5
           END
      Please remember that UNTIL sequences are dependent on the order
      of atoms in LIST 5, don't use them if you're not sure.
   3. 'Punch out', edit, and read back in the parameters (LIST 5). There
      are menu and toolbar options to do this, or just type:
           \SCRIPT SYSED5
 

Step two:
Setup the constraints. If you like using the GUIDE, then it will already have produced a basic LIST 12 (constraints) for you.
To edit List 12, there are menu and toolbar options, or just type:

           \SCRIPT EDLIST12
 

Presumably, the first two lines of the LIST 12 instruction will be something like this:

        
           \LIST 12
           BLOCK SCALE X'S U'S
 

There may also be lots of RIDE constraints if you have chosen to ride the hydrogen atoms.
Add the following lines after those and before the 'END' statement (changing the atom names, of course):

           \   This is a comment.
           \   The EQUIVALENCE constraint, split over two lines here
           \   maps all the model parameters following it onto one
           \   least squares parameter. ie. for all the occupancies
           \   given here only one value will be refined.
           EQUIVALENCE C(1,OCC) C(10,OCC) UNTIL C(15)
           CONTINUE C(101,OCC) C(110,OCC) UNTIL C(115)
           \   This is a comment.
           \   The weight instruction means that the derivatives and
           \   shifts of the listed parameters will be multiplied by
           \   the given weight. This has the effect of ensuring that
           \   the sum of the occupancies of the two sites remains
           \   constant.
           WEIGHT -1 C(101,OCC) C(110,OCC) UNTIL C(115)
 

Save and close the file, CRYSTALS will automatically read it back in.

Step three:
Carry out the refinement. Don't use the GUIDE, it will overwrite all your hard work typing in those constraints. Instead use one of the following methods to initiate some cycles of least squares:

     1. Just type:
       \SFLS
       REFINE
       REFINE
       REFINE
       END
  or 2. Type \SCRIPT XREFINE to get a dialog with a few options.
     3. Choose "Refinement"->"Refine" from the menus, or the 
        equivalent toolbar button ("R").
 

Verify that the occupancy values after refinement make sense. Check that they add up to one, and confirm that only the occupancies that you intended to refine have in fact been refined.

Restraints are set up using list 16.

For easy generation of simple restraints select two, three or more atoms in the model window. Right click on one of the group and choose an option from the pop-up restraints sub-menu. Any restraints generated in this way will be appended to the existing restraint instructions in list 16.

An example List 16 could be:

  \LIST 16
  DIST  1.39 , .01 =      C(1) to C(2)
  DIST  0.0  , .01 = MEAN C(3) to C(4), C(4) to C(5)
  VIBR  0.0  , .01 =      C(6) to C(7)
  U(IJ) 0.0  , .02 =      C(8) to C(9)
  PLANAR                  C(1) until C(6)
  SUM                     K(1,OCC), K(2,OCC), K(3,OCC)
  LIMIT                   U[11] U[22] U[33]
  END
 

The first instruction restrains the bond C(1) to C(2) to be 1.39 angstroms with an e.s.d. of .01 angstroms.
 
The second instruction is a similarity restraint which restrains the difference between two bonds (C3-C4 and C4-C5) to be zero with an e.s.d of .01 angstroms.
 
The third restrains the difference between the mean square displacement of each atom along the direction of a bond to be zero, with an e.s.d of .01. Isotropically refined atoms will also be dealt with correctly.
 
The U(IJ) restraint between C(8) and C(9) restrains the difference between the corresponding Uij components to be zero with an e.s.d of .02.
 
The planar restraint restrains atoms C(1) to C(6) (based on the list 5 order) to be planar. The e.s.d defaults to 0.01.
 
The sum restraint holds the sum of the specified parameters (in this case the occupancy of three postassium atoms) to be constant during the refinement. The default e.s.d is 0.0001. This is useful where disordered atoms occupy more than two sites, since in these cases the total occupancy cannot be held constant using constraints.
 
The limit restraint attempts to limit the shift of the specified parameters to zero (in this case all the Uii anisotropic parameters). This is useful for controlling refinements of poor starting models. The default e.s.d is 0.01. Lowering it to 0.001 will effectively fix the parameter, and raising it to 10 will have almost no effect unless the shift due to the X-ray observations is huge.

I gave it a try but doesn't seem to make much difference. I am not sure that I am using it in the right way.
(1) You must make sure there is a void in the structure. (ie. delete the solvent that you want squeeze to take care of) - this is the a common misunderstanding about running squeeze.
(2) Run Squeeze from the Refinement menu. Three things might happen:
(a) CRYSTALS asks where platon.exe is. Tell CRYSTALS and carry on. If you don't have a copy of PLATON, them download one.
(b) Something flickers quickly onto the screen and then disappears, and CRYSTALS asks whether you want to use the results from Squeeze. Don't. It didn't work. You could try downloading a new version of Platon, and testing that Platon can be started in interactive mode by choosing 'Platon' from the Tools menu.
(c) Platon starts and grinds through some calculations for a while. If it ends in error, the don't apply the results when CRYSTALS asks. If something went wrong, then maybe your trying to squeeze to large or small a void or something.
(3) Look at the summary in the Platon dialog window. Close the Platon dialog window.
(4) CRYSTALS will ask 'Do you want to apply the results?'. Click Yes.
Instead of doing the usual 'Squeeze thing' of altering your FObs values (generally considered to be a BAD THING), CRYSTALS and PLATON together do something much nicer: the A and B components of the structure factor which are due to scattering from the 'void' are stored in your list of reflections; the F-obs value is unchanged. When calculating structure factors, the program adds together the scattering from the atomic electron density as usual, and then adds in the pre-computed A and B parts aswell. Not only does this avoid meddling with the FObs values, but it also (a) keeps the phase information from the scattering from the 'void', and (b) allows you to iteratively reapply squeeze if you think you've improved the model at all and (c) ditch the squeeze correction if you don't like it.