From Papertape Input to

1
From Papertape Input to Forensic
CrystallographyA History of the Program PLATON

• Ton Spek,
• Bijvoet Center
• Utrecht University
• The Netherlands
• K.N.Trueblood Award Lecture
• Chicago, July 29, 2010.

2
Some History

• Back in 1966 I started crystallography as a
  student in the Laboratory for Crystal and
  Structural Chemistry at Utrecht University that
  was at that time headed by Prof. A.F. Peerdeman.
• Peerdeman (co-author of the famous Bijvoet,
  Peerdeman van Bommel paper on absolute
  configuration) was the successor of Prof.
  J.M.Bijvoet.
• Dorothy Hodgkin came over during that time to
  tell about the Vitamin B12 structure and her
  oversees collaboration with Ken Trueblood

3
After WWII, Bijvoet had Managed to start a new
lab in a stately house (used by the Gestapo
during WWII) close to the centre of the city of
Utrecht. Part of the house was his private
domain.
After his retirement, he still kept a
pied-a-terre for when he was in Utrecht. As a
student, I shared the family bedroom in its
double function as student room.
4
Former Crystal Palace and home of Prof. J.M.
Bijvoet
5
Computing-I

• The Crystal Palace was the home of the first two
  generations of computing platforms within the
  university of Utrecht (Zebra and X1
  respectively).
• In 1966 computing had moved to a University
  computing centre elsewhere in the city.
• Computing was done from then on with an Algol
  language specific X8 computer (From a Dutch
  company Electrologica, later part of Philips)
• Processing was essentially one job at a time.

6
16kW
1966, Electrologica X8 ALGOL60 Mainframe
(lt1MHz)
7
Computing-II

• Jobs were run by an operator during daytime
  shifts
• Most of our crystallographic work was done during
  the once-a-week 13 hour nightshift when we as
  crystallographers had the computer for ourselves.
  Half of the staff stayed overnight.
• We were during that nightshift the scientist, the
  software developer and the system operator in
  one.
• I/O was paper tape based. One job at a time. Very
  little memory. No stored binaries, thus
  recompilation everytime.

8
Computing-III

• Programs and data were on paper tape
• The preparation of programs and program input
  were done on the so called Flexowriter. This very
  noisy electical typewriter was also often used as
  output medium.
• Editing was done with a pair of sissors to cut
  out unwanted material from the source code and
  adhesive tape to glue a substitute in the paper
  tape.

9
Flexowriter for the creation and editing of
programs and input data
10
The Science

• My supervisor, Dr. J.A. Kanters, gave me an
  interesting assignment to work on.
• He handed me a batch of white crystals with
  unknown composition (code named M200).
• The assignment was to find out what was the
  structure, using single crystal X-ray techniques
  only.

11
Data Collection for M200

• Preliminary investigations done with film data
  pointed at space group P-1.
• A Patterson synthesis based on integrated
  Weissenberg projection data subsequently
  suggested a light atom structure.
• Eventually a three-dimensional data set was
  collected with an Enraf-Nonius AD3 diffractometer
  
• (two weeks of datacollection !).

12
Nonius AD3 Diffractometer
13
Structure Determination of M200

• It took half a year to finally find the
  structure.
• The laboratory had a tradition in Direct Methods
  (Beurskens, de Vries, Kroon, Krabbendam)
• However, all available software failed to solve
  my structure (these were pre-MULTAN days ..)
• In the end I had to write my own Direct Methods
  program (AUDICE) that solved the triclinic
  structure including many other unsolved
  structures that were hanging around in the lab.

14
The Structure
3-Methoxy-glutaconic acid
15
The Program
16
The Program AUDICE

• AUDICE was one of the Symbolic Addition programs
  that were developed in that period.
• Its specialty was that at the start of the
  evaluation of the strong triple product
  indications for a positive sign, 27 symbols were
  introduced for strong starting reflections rather
  than in the order of three by some other
  approaches. Eventually, 8 solutions were produced
  by eliminating 24 symbols based on multiple
  indications.
• In addition the correlation method was used to
  improve the reliability of triple phase relations.

17
The Correlation Method P for triple H,K,HK
depends on E(H)E(K)E(HK) Correlation
Method ? Improved P on the basis of P of
three adjacent triples E(H)E(L)E(HL)
E(K)E(L-K)E(L) E(HK)E(L-K)E(HL) I.e
. Strengthening of P(E(H)E(K)E(HK) when in
addition E(HL),E(L-K),E(L) strong (Note
Theoretically formalized in terms of
neighbourhoods, Hauptman)
L
H
K
HK
18
Epilogue

• The structure of M200 has been published
• Unfortunately, attempts to publish AUDICE in Acta
  Cryst. stranded on the referee requirement to
  compare its performance on non ALGOL (real ..)
  platforms.
• Anyway AUDICE was superseded by the program
  MULTAN (Fortran) on the new CDC University
  Mainframe.
• The structure solves and refines in a matter of
  seconds on current hardware with SYSTEM S gt

19
Automatic Structure Solution of M200 in the
No-Questions-Asked Mode
20
Direct Methods Meetings

• Multiple meetings and schools were organized in
  the 70s with Direct Methods (software and
  theory) as its major subject.
• Examples are the NATO schools in Parma and York,
  the schools in Erice (1974 1978) and the
  meetings at the Medical Foundation (Buffalo)
  where I met Ken Trueblood.
• Important ones werealso the CECAM workshops on
  Direct Methods (5 weeks!, bringing together
  people working in the field to work on current
  issues) in the early 70s in Orsay (near Paris)
  around a big IBM-360 with lectures by Hauptman.
  (Participants Germain, Main, Destro, Viterbo).
  The program MULTAN was finalized there.
• Photo of the participants of the Parma 1973
  meeting and the 1978 Erice School next

21
Hauptman Lectures Parma Spring 1973
22
(No Transcript)
23
The National Facility

• In 1971, a national single crystal service
  facility was started, with me to make it all
  happen..
• I kept that position for 38 until my emeritus
  status in 2009.
• The project is now continued by my former
  co-worker Martin Lutz
• My last postdoc was Maxime Siegler, now staff
  crystallographer at the John Hopkins University.
• The program PLATON is a side product of the
  national facility (note never explicitly funded
  !)

24
PLATON

• Work on PLATON started in 1980.
• The idea was to produce with a single CALC ALL
  instruction an exhaustive listing of derived
  geometry to give to our clients.
• Over time numerous additional tools have been
  added on the basis or the needs in our service
  setting.
• PLATON is, in combination with SHELX, one of the
  major tools for our service.

25
PLATON Tools

• The available tools are shown as clickable
  options on the opening window of the program.
• Examples are ADDSYM for the detection of missed
  symmetry, TwinRotMat for automatic twinning
  detection and SYSTEM S for guided/automated
  structure determination)
• Here we will look in some detail at a few of the
  tools
• SQUEEZE for the handling of disordered solvents
• Structure Validation (used as part of the IUCr
  CheckCIF)
• FLIPPER, a new approach to structure determination

26
(No Transcript)
27
The Disordered Solvent Problem

• Molecules of interest often co-crystallize (only)
  with the inclusion of a suitable solvent
  molecule.
• Solvent molecules often fill voids in a structure
  with little interaction and located on symmetry
  sites and with population less than 1.0
• Often the nature of the (mixture) of included
  solvent(s) is unclear.
• Inclusion of the scattering contribution of the
  solvent can be done either with a disorder model
  or with SQUEEZE.

28
THE MOLECULE THAT INVOKED THE BYPASS/SQUEEZE TOOL
Salazopyrin from DMF R 0.096
29
Structure Modeling and Refinement Problem for the
Salazopyrin structure
Difference Fourier map shows disordered channels
rather than maxima How to handle this in the
Refinement ? SQUEEZE !
30
Looking down the Infinite Channels in the
Salazopyrin Structure
How to model this disorder in the L.S-Refinement ?
31
The SQUEEZE Tool

• The SQUEEZE tool offers an alternative to the
  refinement of a disorder model for a structure
  containing disordered solvent.
• The contribution of the disordered solvent to the
  calculated structure factors is taken into
  account by back-Fourier transformation of the
  electron density found in the solvent region of
  the difference map.
• This requires an iterative series of difference
  map improvements.
• Firstly, the solvent accessible region has to be
  indentified to be used as a mask over the
  difference density map.

32
Solvent Accessible Voids

• A typical crystal structure has only in the order
  of 65 of the available space filled.
• The remainder volume is in voids (cusps)
  in-between atoms (too small to accommodate an
  H-atom)
• Solvent accessible voids can be defined as
  regions in the structure that can accommodate at
  least a sphere with radius 1.2 Angstrom without
  intersecting with any of the van der Waals
  spheres assigned to each atom in the structure.
• Next Slide Void Algorithm Cartoon Style ?

33
DEFINE SOLVENT ACCESSIBLE VOID
STEP 1 EXCLUDE VOLUME INSIDE THE VAN DER
WAALS SPHERE
34
DEFINE SOLVENT ACCESSIBLE VOID
White Area Ohashi Volume. Location of possible
Atom centers
STEP 2 EXCLUDE AN ACCESS RADIAL VOLUME TO
FIND THE LOCATION OF ATOMS WITH THEIR CENTRE AT
LEAST 1.2 ANGSTROM AWAY
35
The
DEFINE SOLVENT ACCESSIBLE VOID
STEP 3 EXTEND INNER VOLUME WITH POINTS
WITHIN 1.2 ANGSTROM FROM ITS OUTER BOUNDS
36
Listing of all voids in the unit cell
The numbers in refer to the Ohashi Volume
EXAMPLE OF A VOID ANALYSIS
37
VOID APPLICATIONS

• Detection of Solvent Accessible Voids in a
  Structure
• Calculation of Kitaigorodskii Packing Index
• Determination of the available space in solid
  state reactions (Ohashi)
• Determination of pore volumes, pore shapes and
  migration paths in microporous crystals
• As part of the SQUEEZE routine to handle the
  contribution of disordered solvents in a crystal
  structure.

38
SQUEEZE

• Takes the contribution of disordered solvents to
  the calculated structure factors into account by
  back-Fourier transformation of density found in
  the solvent accessible volume outside the
  ordered part of the structure (iterated).
• Refine with SHELXL using the solvent free .hkl
• Or CRYSTALS using the SQUEEZE solvent
  contribution and the the full Fobs
• NoteSHELXL lacks option for fixed contribution
  to Structure Factor Calculation.

39
SQUEEZE Algorithm

1. Calculate difference Fourier map (FFT)
2. Use the VOID-map as a mask on the FFT-map to set
   all density outside the VOIDs to zero.
3. FFT-1 this masked Difference map -gt contribution
   of the disordered solvent to the structure
   factors
4. Calculate an improved difference map with F(obs)
   phases based on F(calc) including the recovered
   solvent contribution and F(calc) without the
   solvent contribution.
5. Recycle to 2 until convergence.

40
SQUEEZE In the Complex Plane
Fc(solvent)
Fc(total)
Fc(model)
Fobs
Solvent Free Fobs
Black Split Fc into a discrete and solvent
contribution Red For SHELX refinement,
temporarily substract recovered solvent
contribution from Fobs.
41
Real World Example

• THF molecule disordered over a center of
  inversion
• Comparison of the result of a disorder model
  refinement with a SQUEEZE refinement

42
Disorder Model Refinement Final R 0.033
43
Comparison of the Results of the two Modeling
Procedures
Disorder Model R 0.033
SQUEEZE Model R 0.030
44
LISTING OF FINAL SQUEEZE CYCLE RESULTS
45
ANALYSIS OF R-VALUE IMPROVEMENT WITH RESOLUTION
AANALYSIS
46
Concluding Remarks

• The CSD includes in the order of 1000 entries
  where SQUEEZE was used.
• Care should be taken for issues such as charge
  balance

47
Charge Flipping

• Charge Flipping as an alternative for structure
  solution by Direct Methods was introduced by G.
  Oszlanyi A. Suto (2004). Acta Cryst. A60, 134.
• Similar to SQUEEZE it involves iterated forward
  and backward Fourier transforms.
• PLATON implements an experimental version of
  Charge Flipping named FLIPPER.
• Following is an example of the P21, Z2 structure
  of vitamin C solved by FLIPPER starting with all
  reflections assigned a phase of zero degrees.

48
(No Transcript)
49
FLIPPER

• Charge Flipping is done with data in space group
  P1.
• The space group is determined from the solution
• The methods can be used for automatic structure
  determination of non disordered structures
• Following is the real time display of the
  progress in the development of the structure
  after each Fourier cycle, followed a full
  refinement.

50
(No Transcript)
51
Automated Structure Validation

• It is easy to miss problems with a structure as a
  busy author or as a referee
• Increasingly Black-Box style analyses done by
  non-experts
• Limited number of referees experts available
• It is easy to hide problems with a ball-and-stick
  style illustration
• Sadly, fraudulous results and structures have now
  been identified in the literature thus
  contaminating the assumed solid information in
  the CSD.

52
Structure Validation with PLATON

• Automated Structure Validation was pioneered and
  pushed by Syd Hall as section editor of Acta
  Cryst C. by
• The creation of the CIF Standard for data
  archival and exchange (Hall et al., (1991) Acta
  Cryst., A47, 655-685.
• Having CIF adopted by Sheldrick for SHELXL93
• Making CIF the Acta Cryst. submission standard
• Setting up early CIF checking procedures for Acta
• Inviting me to include PLATON checking tools such
  as ADDSYM and VOID search.

53
WHAT ARE THE VALIDATION QUESTIONS ?

• Single Crystal Structure Validation addresses
  three simple but important questions
• 1 Is the reported information complete?
• 2 What is the quality of the analysis?
• 3 Is the Structure Correct?

54
How is Validation Currently Implemented ?

• Validation checks on CIF data can be executed at
  any time, both in-house (PLATON/CHECK) or through
  the WEB-based IUCr CHECKCIF server.
• A file, check.def, defines the issues that are
  tested (currently more than 400) with levels of
  severity and associated explanation and advise.
  (www.cryst.chem.uu.nl/platon/CIF-VALIDATION.pdf)
• Most non-trivial tests on the IUCr CheckCIF
  server are executed with routines in the program
  PLATON. (Identified as PLATxyz)

55
VALIDATION ALERT LEVELS

• CheckCIF/PLATON creates a report in the form of a
  list of ALERTS with the following ALERT levels
• ALERT A Serious Problem
• ALERT B Potentially Serious Problem
• ALERT C Check Explain
• ALERT G Verify or Take Notice

56
VALIDATION ALERT TYPES

• 1 - CIF Construction/Syntax errors,
• Missing or Inconsistent Data.
• 2 - Indicators that the Structure Model
• may be Wrong or Deficient.
• 3 - Indicators that the quality of the results
• may be low.
• 4 Info, Cosmetic Improvements, Queries and
• Suggestions.

57
PLATON/CHECK CIF FCF Results
58
Which Key Validation Issues are Addressed

• Missed Space Group symmetry (being Marshed)
• Wrong chemistry (Mis-assigned atom types).
• Too many, too few or misplaced H-atoms.
• Unusual displacement parameters.
• Hirshfeld Rigid Bond test violations.
• Missed solvent accessible voids in the structure.
• Missed Twinning.
• Absolute structure
• Data quality and completenes.

59
Evaluation and Performance

• The validation scheme has been very successful
  for Acta Cryst. C E in setting standards for
  quality and reliability.
• The missed symmetry problem has been solved for
  the IUCr journals (unfortunately not generally
  yet There are still numerous Marshable
  structures).
• Most major chemical journals currently have now
  some form of a validation scheme implemented.
• Recently included FCF validation

60
FCF-VALIDATION

• - Check of the CIF FCF data Consistency
  (including R-values, cell dimensions)
• - Check of Completeness of the reflection data
  set.
• - Automatic Detection of ignored twinning
• - Detection of Applied Twinning Correction
  without having been Reported in the paper.
• - Validity check of the reported Flack parameter
  value against the Hooft parameter value.
• - Analysis of the details of the Difference
  Density Fourier Map for unreported features.

61
Sloppy, Novice or Fraudulent ?

• Errors are easily made and unfortunately not
  always discernable from fraud.
• Wrong element type assignments can be caused as
  part of an incorrect analysis of an unintended
  reaction product.
• Alternative element types can be (and have been)
  substituted deliberately to create a new
  publishable structures.
• Reported and calculated R-values differing in the
  first relevant digit !?

62
Some Relevant ALERTS

• Wrong atom type assignments generally cause
• Serious Hirshfeld Rigid Bond Violation ALERTS
• Larger than expected difference map minima and
  maxima.
• wR2 gtgt 2 R1
• High values for the SHELXL refined weight
  parameter

63
Acta Cryst. (2007), E63, m1566.
Sn(IV)(NO3)4(C10H8N2)2
64
2.601 Ang.
Missing H in bridge Sn(IV) gt Lanthanide(III)
65
The Ultimate Shame

• Recently a whole series of isomorphous
  substitions was detected for an already published
  structure.
• Similar series have now been detected for
  coordination complexes (Transition metals and
  lanthanides)
• How could referees let those pass ?
• Over 100 structures now retracted
• Fraud detected by looking at all papers of the
  same authors of a strange structure (and their
  institutions)

66
BogusVariations (with Hirshfeld ALERTS) on the
Published Structure 2-hydroxy-3,5-nitrobenzoic
acid (ZAJGUM)
67
Comparison of the Observed data for two
isomorphous compounds.
Tool platon d name1.fcf name2.fcf
Conclusion The Same Data !
The Only Difference Is the SCALE !
SLOPPY Or FRAUD ?
68
Thanks !

• My former co-workers over 38 years and in
  particular my successor Dr. Martin Lutz
• Dr. Louis Farrugia for following my frequent
  updates with his MS-Windows implementation
• The users of the software for ideas and bug
  reports.
• Lachlan Cranswick for promoting my software and
  who is sadly no longer with us here.

69
IUCr Crystallographic Computing School 2005 Siena
70
(No Transcript)