A Focus on Crystallography

The assignment of atom types is one of the first stepsafter solving a single crystal X-ray structure analysis. It isoften difficult or even impossible to assign correct atomtypes if there are atoms with similar atomic number pres-ent in the structure solution. The Cambridge StructuralDatabase (CSD) allows to perform statistical analysesof bond lengths and delivers an elegant solution for theproblem of ambiguous atom types. Mean bond lengthsof the structural units SiO4, SiO6, PO4, and PO6were cal-culated from CSD data. These were compared with bondlengths from the structure solution of a hitherto unknownsilicophosphate. This procedure allowed for an unambig-uous assignment of atom types in the structure solution. The Cambridge Structural Database The Cambridge Structural Database (CSD) is a comple-mentary database to ICSD. The CSD is compiled anddistributed by the Cambridge Crystallographic Data Cen-tre (CCDC). This is a non-profit company and has been aregistered charity since 1987. The producers of the CSDdescribe it as the world’s repository of experimentallydetermined small-molecule organic and metal-organiccrystal structures. The database contains the results ofover half a million x-ray and neutron diffraction analy-ses. This unique database of accurate crystal structureshas become an essential resource to scientists aroundthe world. Each crystal structure undergoes extensivevalidation and cross-checking by expert chemists andcrystallographers to ensure that the CSD is maintainedto the highest possible standards. Also, each databaseentry is enriched with bibliographic, chemical and physi-cal property information, adding further value to the rawstructural data. These editorial processes are vital forenabling scientists to interpret structures in a chemicallyBond lengths statistics as a tool for crystalstructure analysis – a case study with acrystalline silicophosphateUsing the Cambridge Structural Database to assign atom types meaningful way. The CSD is continually updated with newstructures (>40,000 new structures each year) and withimprovements to existing entries [1]. We use the in-house version of the CSD. Searches areperformed with the software ConQuest. Geometric para-meters obtained from a search with ConQuest are export-ed to the software module VISTA for statistical analysis.Further analysis of the data is possible with Mercury. Thissoftware offers a comprehensive range of tools for struc-ture visualization, the exploration of crystal packing andfurther statistical analyses. The problem with the silicophosphate structure Silicophosphates (aka SiPOs) are promising materialsfor the use as lasers, optical fibers and orthopaedicimplants. That is because of their special material andoptical properties such as low melting points and highrefractive indices. Furthermore, they contribute to theunderstanding of geological processes. Until now, manydifferent silicophosphate structures were discovered. Re-markably, therein the silicon atoms are octrahedrally-co-ordinated to six oxygen atoms. This is in contrast to mostsilicates in nature where silicon occurs in tetrahedral co-ordination. All known SiPOs are composed of SiO4andPO4tetrahedra as well as SiO6octahedra. These struc-tural building blocks are connected differently to eachother and yield a multitude of various silicophosphatestructures [2]. The classical synthesis of such silicophos-phates is performed by melting processes, chemicaltransport reactions or sol-gel-techniques. Recently, wediscovered another way to prepare silicophosphate ma-terials with octahedral silicon structures. The syntheseswere performed at room temperature and normal pres-sure in diethylether solution. The reaction of tetraethox-ysilane (TEOS) and anhydrous H3PO4gives amorphousSi-O-P materials with SiO4, PO4and SiO6structural units.On the other hand, chloroethoxysilanes and anhydrousphosphoric acid yielded oligomeric silicophosphates. * TU Bergakademie Freiberg, Institut für Anorganische Chemie 24 A Focus on Crystallography From the latter reactions one crystalline product was ob-tained. The X-ray structure analysis of this product leadsto several problems which will be discussed in the follow-ing sections. The structure solution delivered a large complex ionwhich consists of silicate and phosphate subunits. Fur-thermore, tetraethylammonium and solvent moleculeswere identified. After several refinement cycles it becameobvious that the unambiguous assignment of the atomtypes phosphorous and silicon in the structure wouldbe no simple task (Fig. 1). Both elements have similaratomic radii, comparable electron densities, and are ableto create the same coordination geometry with oxygen.Especially the subunits SiO4/ PO4and SiO6/ PO6couldboth be present in the structure. It would be most helpfulto discriminate these subunits on the basis of differentbond lengths. Therefore we were in need of reliable bondlengths for these four substructures. Bond lengths statistics from the CSD A search for the substructures SiO4/ PO4and SiO6/ PO6was performed with the CSD in order to obtain statistical-ly reliable bond lengths derived from experimental data.For that purpose the structural elements were drawnwithin the module ConQuest and the number of bondedatoms was fixed to 4 or 6. The option “ADD 3D” was usedto extract the bond lengths from the structures in thedatabase (Fig. 2). The search for the SiO4unit gave 587hits. These were analysed in the software module VISTA.The mean bond lengths for this hit list is 1.618 Å withminimal and maximal values of about 1.53 and 1.71 Å, re-spectively (see Fig. 3). The outliers were analyzed. Theseoutliers could be eliminated from the statistical analysiswithin the VISTA software. But this should be done onlyfor good reason. The same procedure was applied to thestructural units PO6, PO4, and SiO6. The result of this sta-tistical analysis is summarized in Table 1.Fig. 1: Structure of the silicophosphate anion after isotropic refinement(dark grey – carbon, red – oxygen, blue – silicon or phosphorous?) Fig. 2: Graphical user interface of the CSD 25A Focus on Crystallography Tab. 1: Mean bond lengths of structural units derived from the CSDstructural unit mean bond lengths in Å hits standard deviation lower quartile upper quartile PO41.5353217 0.0161.5231.546 SiO41.618587 0.0201.5931.639 PO61.71758 0.0201.7051.719 SiO61.78253 0.0391.7661.788((Figure 3.ps oder Figure 3.pdf)) wo ist die GrafiK???????? Fig. 3: Bond lengths statistics for the SiO4unitPlot DataFile=/home/aoch/aschvTest=1Tot.Obs.=1036Obs.=1036Supp.=0 X-axisMin.=1.527Max.=1.709Range=0.181Mean=1.618Mean SE=0.001Sample SD=0.020 HistogramMedian=1.620Skew=0.190Quantile=10.000LQ=1.593HQ=1.639Bin Width=0.005Max. Bin=149.000TRN07