1. The System-S Approach to Automated Structure Determination: Problems and Solutions Ton Spek National Single Crystal Service Utrecht University, The Netherlands

2. Is a Fully Automatic Structure Determination Possible ? Yes, in many cases given good quality low temperature reflection data and correct prior knowledge of the chemical composition. Good examples are light atom structures with no disorder and with no unknown solvent of crystallization. No, in all cases needing an experienced crystallographer (I.e. when no automated methods are available yet for the problem at hand) e.g.: many/most coordination complexes.

3. Current Setup in Utrecht KappaCCD/Rotating Anode/LNT I Collect, DENZO or EVAL-CCD (Not AUTO!) I import.cif (includes Formula & Unit Cell) I Silent System-S Auto-Mode: platon -F import.cif I Result files -> ORTEP/PLUTON Plots

4. Typical Import CIF Filter Job Unix Command Line Instruction: platon –F import.cif ….. Wait 0.5 minute ….. Standard Console Output: Tentative Result on: /mnt/spea/import_s.cif and: /mnt/spea/import_s.res Validation Report: /mnt/spea/import_s.chk Details may be found in: /mnt/s/import

5. 0.5 Min. from import.cif to ORTEP (3GHz Linux Box)

6. System-S Approach Crystallographic Shell around public domain tools (SHELX, SIR, PLATON etc.) Two Modes of operation: Guided Mode: Suggests/Executes next step in the structure determination (handles I/O) Automatic Mode: (NQA-Mode) Executes automatically all default steps Build-in Structure Validation System-S is included in the UNIX PLATON exe.

7. Standard System-S Example Two files: name.ins & name.hkl Content name.hkl : SHELX style HKLF 3/4 Content name.ins: TITL, CELL, CELLSD, SFAC, UNIT, HKLF 4 (or 3) Note: No Spacegroup Info Needed Run: s name.ins NQA. 18 Seconds later for C16H20NO3P ….

9. Problems and Solutions There are still many interesting ‘Real World’ problems to be solved for a full proof automatic replacement of an experienced crystallographer. Already the use of the Guided Mode to quickly try multiple approaches to solve a structure is a big help (even for the experienced crystallographer) System-S stores all relevant I/O files in a hierarchic directory structure for inspection when problems arise.

10. Problem #0 – Complete Failure Example: The chemist provides us with crystals labeled to be a Zn-coordination complex requiring human ‘Expert Forensic Crystallography’ to tell us that the inorganic salt MnK2(SO4)2 gives us a perfect fit with the experimental data. Partial Solution: Check with databases (I.e. CELL or simulated powder patterns) and stop.

11. Failure #1 – Space group The space group can not always be determined uniquely from systematic absences (e.g Cc or C2/c). Some structures solve only in lower than the actual space group symmetry (sometimes only in P1) Solution: loop over the various space group candidates to solve the structure and use ADDSYM to find out about the correct symmetry of the model (implemented in S).

12. Failure #2 – Phase Problem -Attempts to automatically solve a structure with a single solution package might fail. -Solution: loop over all available packages and approaches until a solution is found. -System-S: SHELX86, SHELXS97, SHELXD, SIR97, SIR2002, DIRDIF99 -‘Workup’ with EXOR – Example next …

13. Raw Output from SHELXS86

14. Result of the EXOR workup

15. Failure #3 – Incorrect Atom Type Assignment Automatic atom type assignment is not always straightforward. A large variation in displacement parameters often hampers a correct atom type assignment Solution: detailed analysis of intra- and intermolecular geometry and ADP’s Example next …

16. Misassigned Atom Types (S & P)

17. Atom Type Assignment Assignment Methods (can be) based on: 1 - An analysis of the Peak heights & Shapes 2 - the value of the displacement parameters 3 - Population parameter refinement (S) 4 - Refinement of each site with alternative scattering types (E.g. C,N,O) 5 - Chemistry (& CSD knowledge, Mogul)

18. Failure #4 - Disorder Type 1 – Orientational, Conformational or Substitutional disorder of (part of) the structure. Solution: (Expert) Crystallographer Type 2 – Voids filled with unknown (mixtures of) solvents. Solution: PLATON/SQUEEZE

19. Failure #5 - Twinning (Pseudo) Merohedral Twinning may lead to refined structures with unexpectedly high final R- values and high residual peaks in a difference Fourier Map. Solution: Software to detect the applicable twin law. E.g. Rotax (Simon Parsons & Bob Gould) available in WINGX and CRYSTALS or PLATON/TwinRotMat Example next ….

20. Space group P-3 – R = 20 % 3 independent molecules

23. Problem #6 – Wrong Structure Sometimes automatic procedures can come up with ‘reasonably looking’ but wrong structures. Structure validation software should send out proper ALERTS (e.g. IUCr Checkcif) Example: The chemist expects a Cu-complex, so that is what he gets … (confirmed by X-ray crystallography !) …. and tries to publish … and caught by a knowledgeable referee … on the bases of Checkcif ALERTS.

24. ORTEP of the False Structure R < 7 % Cu not Coordinated =>

25. Ortep of the Correct Structure R < 6 %

26. CONCLUDING REMARKS SYSTEM-S can be used easily for the re- examination of structures with CIF + FCF data taken from the Acta Cryst Archives or for refereeing purposes to investigate unclear details of the analysis. Future implementation: C,N,O etc. discrimination in difficult cases More chemical knowledge Extension of the already available procedures for structure determination without any Content Information.

27. End