SQUEEZE: An effective cure for the disordered solvent syndrome in
crystal structure refinement.
The current version of SQUEEZE has been designed, dimensioned and tested
for small moiety structures containing disordered solvent molecules of the
type toluene, CH2Cl2, tetrahydrofurane, water, methanol etc..
Anions may be treated in the same way. However, take care of the problem
of uncertainty of charge balance.
Large voids currently require significant computing in the stage where the
size and shape of the solvent accessible void is determined. All
calculations are done in the triclinic system (data are expanded
automatically when appropriate).
Reflection data and FFT-array are stored in memory i.e. large structures
(and high symmetry) may require large amounts of memory
(change parameter NP21, that defines the available scratch area, globally
to a larger value).
Implementation and Use:
SQUEEZE has been implemented as the 'SQUEEZE option' in the program
PLATON/SQUEEZE should be compatible with small-moiety structure
refinement usage of the popular program SHELXL-97 (or related incarnations).
The program is used as a filter. Input files are:
- shelxl.hkl - (HKLF type 4)
- shelxl.res - (complete set of refined model parameters, including
hydrogen atoms but excluding any dummy atoms used to
describe the disorder region)
invoke the program with:
give on the prompt >>
or click on the main PLATON-menu option SQUEEZE.
The result will be in two files
- shelxl.lis: a listing file giving details of the calculations
- shelxl.hkp: a modified reflection file against which the ordered
structure parameters can be refined (the solvent
contribution has been eliminated from the reflection
In order to run SHELXL-97 on the 'solvent-free' Fo^2 data:
(Note: save all files you want to keep)
- cp shelxl.res shelxl.ins
- cp shelxl.hkp shelxl.hkl
- run: shelxl
In order to get an .fcf style file (Fo^2 + Fc^2 (model + solvent))
you will need shelxl.hkl (= shelxl.hkp) and shelxl.res
run: platon shelxl.res
with the interactive option: CALC FCF
Final R-values are reported on the basis of the WGHT parameters in the
There will be a difference in reflection count as compaired to the
SHELXL-run due to the differing number of surviving
The procedure (starting from the original reflection data) can be
repeated using the newly refined parameters when desired (This may
define a 'refined' void area. However, there will be rarely a need to
repeat the procedure).
The general procedure (based on a preliminary implementation of the
technique) has been described in more detail in:
Acta Cryst. (1990), A46, 194 as the 'BYPASS procedure' P. v.d. Sluis
& A.L. Spek)
The 'difference-map' improvement potential of this technique has been
demonstrated for small molecule structures. The technique should also
work for protein data. However, this has not been tested by us as yet with
PLATON/SQUEEZE. Current design features may cause problems when tried.
A directory containing test-examples is in
- The record length of the '.hkp' file has been increased above 80
to accommodate additional data (the exact details are not fixed)
including the original intensity and calculated solvent contribution
to the structure factors.
- The exact numbers on the SHELXL UNIT instruction are irrelevant
for the SQUEEZE procedure.
- The SHELXL-TWIN instruction is not available as yet in PLATON.
- PLATON internally calculates structure factors (using the same
scattering factors as used by SHELXL97) for the model given
in the shelxl.ins file.
- The solvent contribution to the structure factors is taken as
'error-free'. This means that the 'solvent-free' Fo^2 keep their
original su's (esd's).
- The method relies heavily on the quality of the low-order reflections.
The dataset should be as complete as possible. Systematic errors may
hamper the quality of the results.
- The Contour-Map option in PLATON can be used to inspect the
inproved difference-map (i.e. calculated with phases including the
disordered solvent contribution).
- Current conditions for applicability are:
- Reasonable data-resolution (say 25 degrees Mo)
- Structure of the known part completed with H-atoms
- Disorder density should be well outside the vanderWaals
surface of the known structure.
- The area to be 'SQUEEZE' should not be too large (say
less than in the order of 30% of the unit-cell volume.
Interpretation of the results
- A successfull application of SQUEEZE will show the following results:
- A new hkl-file against which a satisfactory refinement of the
discrete model can be done (purpose: good geometry, good R-value)
- Smooth convergence of the SQUEEZE iteration.
- Significant improvement of the R-value in low resolution data. (see
table at the end of the listing file).
- The number of electrons reported to be found in a void is close to
that expected for the assumed solvent.
- The difference map peaklist should not contain significant peaks
outside the VOID areas. Peaklist on .sqz file.
- Problems are indicated when
- Convergence is unstable
- A large number of reflections left out during the iteration
process (This may be well indicative for problems with the data,
and should be checked for).
- Large residual density excursions in the ordered part of the
- A report on the use of SQUEEZE should always report for each
(significant) independent void:
- Where the void is (i.e. x,y,z)
- Its volume in Ang**3 and multiplicity.
- The number of electrons recovered.
- Fo/Fc-listing with Original Fo and Fc including the solvent
Potential Problems and Pitfalls
Be aware of charge balance problems: SQUEEZED density in the disordered
solvent area might contain a charge that can have consequences for
the charge, valence and interpretation of the ordered structure part.
PLATON/SQUEEZE can take care of redundancy of reflection data on the
the input '.hkl' file. However, with high symmetry space groups this
can lead to some inefficiency. It helps in such cases to supply an
averaged, unique dataset (Unfortunately, direction cosines will
be unavailable in the latter case for post-absorption correction;
of-course a preceeding numerical correction for absorption is to be
The number of recovered electrons in the solvent area is strongly
dependent on the quality of the low-angle reflections. Supply COMPLETE
data sets !