SYSTEM-SThe Challenge of Automated Structure

Determination

Ton SpekNational Single Crystal Service

Utrecht University, The Netherlands

The Challenge

• Early 1990’s, with a serial detector TurboCAD4 on rotating anode producing about one data set per day ==> need for:

• 1-Automated data collection• 2-Automated structure determination• Ad1: Argus (Schreurs/Duisenberg)• Ad2: System-S (Spek, 1991-)

Routine Structure Determination

Single CrystalI

Data collection (CAD4/Argus: GO)I

FormulaI

Autosolve, Refine and Validate (S: NQA)I

ORTEP

Current Setup

KappaCCD/Rotating Anode/LNTI

Collect, DENZO or EVAL-CCD (Not AUTO!)I

Import.cif (includes Formula)I

System-S Auto-Mode: platon -F import.cifI

ORTEP/PLUTON Plots

Standard System-S Example

• Two files: name.ins & name.hkl• Content name.hkl : SHELX style HKLF 3/4• Content name.ins: TITL, CELL, SFAC,

UNIT, HKLF 3 or 4• Note: No Spacegroup Info Needed• Run: s name.ins NQA.

Example: with shelx.ins & shelx.hkl

Structure solved/refined in 18 seconds

Critical Points of Failure• Formula - Can be either incomplete or completely

wrong (Example later). • Space group - Often not uniquely determined

when not P-1, P21/c or Pbca (e.g. Cc or C2/c) • Phase Problem - Multiple methods (Shelxs/Sir etc)• Atom type assignment• Non-ideal crystal: Disorder, Twinning• Voids with disordered (unknown) content (solvent

mixture, charges)

Misassigned Atom Types (S & P)

Atom Type Assignment

Assignment Methods (can be) based on:1 - the analysis of Peak heights2 - the value of the displacement parameters3 - Population parameter refinement (S)4 - Refinement of each site with alternative scattering types (E.g. C,N,O)5 - Chemistry (& CSD knowledge, Mogul)

Raw Output from SHELXS86

Result of the EXOR workup

Space-group Determination

• Use LePage Algorithm for the Metrical Symmetry Determination of the Lattice

• Get Laue averages for each of the proposed (Sub)Lattices and Laue group.

• Select tentative Lattice (e.g. tP, 4/m)• Get systematic absences• Select tentative space-group

Phase Determination

Primary:• SHELXS86 & SHELXS97• SIR97 & SIR2002• DIRDIF PATTY & ORIENT

‘WORKUP’ of Approximate Solutions:• EXOR(cise)

SYSTEM-S DESIGN FEATURES• EASY USE OF THE BEST PUBLIC DOMAIN

TOOLS (NO LOCAL MODIFICATION)• DIRECTORY TREE STRUCTURE FOR THE

ARCHIVAL OF INTERMEDIATE FILES• MANAGEMENT OF MULTIPLE SPACE

GROUP & STRUCTURE DETERMINATION ATTEMPTS (SHELXS, SIR, DIRDIF)

• AUTOMATIC, GUIDED AND COMMAND LINE MODES OF OPERATION

• BUILD-IN VALIDATION

Special Features• Automatic Spacegroup determination• EXOR: Automatic workup of raw Direct

Methods results (population refinement)• Automatic H-atom assignments• Automatic numbering scheme• Missed symmetry handling• Automatic (weighted) refinement• Validation

Horror Story

• Example: Paper submitted for publication in Acta Cryst. C

• Supposedly coordination complex with composition Cu(2+) Ligand(2-), at least that was what the chemist wanted to see confirmed with a routine X-ray study

• That is what he got …. With R < 7% using SHELXTL software

False Structure, R < 7%

Cu NOT coordinated !

Correct Structure

• Cu(2+) --> Br(1-) [HBr was used during the synthesis/crystallisation]

• Additional H-atoms needed/found on sp3-N and -CO2 (short intermolecular acid bridge.

• Thus NOT Cu(2+) L(2-) but Br(1-)L(1+)• Problems clearly indicated with validation

ALERTS

Correct Structure, R < 6%

What about the ORTEPS ?

• Surprisingly, the ORTEPS of the false and correct structure are very similar

• Ellipsoids are as expected for room-temperature data

Ortep of the False Structure

Ortep of the Correct Structure

Redetermination with System-S

• System-S / Guided Mode• Using original Room-temperature data and

expected formula (CuL)• Result: Direct Methods (SIR etc) reproduce

results largely consistent with the expected composition apart from some C,N & O miss-assignments, not uncommon for RT data and easy to correct.

Raw Result with SIR97

O O

CN N

Br

EXPERIENCE• AUTOMATIC STRUCTURE DETERMINATION

SYSTEM-S WORKS CURRENTLY BEST FOR :• - LIGHT ATOM STRUCTURES• - BASED ON LOW TEMPERTURE DATA• - WITH NO DISORDER• - WITH CORRECT FORMULE

Otherwise:• The Manual GUIDED-Mode appears to be the

more appropriate protocol.

CONCLUDING REMARKS• SYSTEM-S can be used easily for the re-

examination of structures with CIF + FCF data taken from the Acta Cryst Archives.

• Future implementation:• - (C,N,O) etc. L.S. Refinement• - More chemical knowledge• - Extention of the already available procedures

for Structure determination without any Content Information.