ALB Crystallography
CCP14 Homepage 
SDPDRR Homepage
What's New
Known Methods
The Samples
Download Data 
Poster 

Structure Determination by Powder Diffractometry Round Robin

The SDPDRR Homepage is at http://www.cristal.org/SDPDRR/ 

Results for Sample 2

C22H24N2O8HCl 

Searching for an unknown organic compound, the tetracycline hydrochloride was first believed to be adequate from a recent publication (F.N. Blanchard, Powder Diffraction, 4 (1989) 103-105). This work states that the previous data were limited to an unindexed powder pattern (JCPDS-ICDD 13-898). The new work indexed the pattern and  suggested two possible space groups (P212121 and P21212). Searching for structural data in the CSD databank produced a link to the Powder Diffraction reference and also shown that a crystal structure was determined (K. Kamiya, M. Asai, Y. Wada and M. Nishikawa, Experientia, 27 (1971) 363) under the name achromycin hydrochloride. Nevertheless, no 3D atomic coordinates were available (which is amazing). Thus it was concluded, that tetracycline hydrochloride finally represented a good sample for test, after all, given that so many already known structures served previously for testing methods. A search match procedure with the PDF-2 ICDD-JCPDS database would be positive (39-1987) and reveal the name of the compound. The molecular formula would be then very easy to find from a great variety of sources. For instance in the online Aldrich catalog. Typing "tetracycline" as keyword with the Yahoo search engine revealed 10613 links ! One can even found a .mol file and a .pdb one ! 

We thus tried to solve it, since the atomic coordinates were unavailable, first looking if an isotypic compound could not be located. The CSD databank presented two possible data sets with similar cell parameters and same space group (P212121). These are the beta-6-deoxy-oxytetracycline hydrochloride C22H24N2O8HCl (DXXTEC) with a = 11.638 A, b = 12.177 A and c = 15.840 A (J. Bordner, Acta Crystallogr. B35 (1979) 219) and the oxytetracycline hydrochloride C22H24N2O9HCl (TERMYC) with a = 11.19 A, b = 12.49 A and c = 15.68 A (H. Cid-Dresdner, Z. Kristallogr. 121 (1965) 170). Let us recall the tetracycline hydrochoride cell parameters here : a = 10.983 A, b = 12.861A  and c = 15.728 A. Both above data were used as starting coordinates for a Rietveld refinement on the aldxconv data (conventional X-rays). Only the non-hydrogen atoms were kept in the 2 starting models, and some sligth modifications were done for being consistent with the molecular formula (the O5 atom in TERMYC was suppressed and the same O5 atom in DXXTEC was moved to the O6 position). Trying to refine without constraint did not led to any satisfying result. Using soft constraints on interatomic distances, the refinement from the DXXTEC data blocked to high reliability values, whereas the TERMYC data started at RB = 35% after scale factor refinement, and rapidly converged to Rp = 9.32, Rwp = 12.5 (background subtracted), RB = 4.6 % (see the plot by DMPLOT and the Ortep structure drawing). Then, the constraints were removed and the reliabilities dropped to Rp = 8.73, Rwp = 11.7 and RB = 3.7. The problem was that interatomic distances tended to be out of usual values : it was concluded that the resolution was not sufficient. 
 
Then the synchrotron pattern was recorded, expecting a better resolution (which is not really the case due to the short wavelength retained (0.692 A) by the operator who did not wanted to select a higher wavelength for reasons of possible harmonic contamination). Here are the test results for the application of conventional direct methods (i.e. from raw extracted factors, possibly on reduced datasets). The background was estimated manually from the aldsync pattern with the DMPLOT program. The program FULLPROF was then used (Le Bail method). Links are given to files renamed as .txt files : 

R-Factors (%) were : Rp =  0.91, Rwp = 1.30    (background included) and Rp = 6.10, Rwp: 5.47 (background subtracted). The fit, as shown by DMPLOT, seems acceptable. 

Then the OVERLAP software was used in order to suppress those peaks overlaping too much. Four data files are prepared for use by SHELXS-97.  The OVERLAP program eliminated reflections having a neighbouring one at less than : 

Note that OVERLAP does not care with the estimated standard deviations and lowers the values in order to keep the weak reflections in the data set prepared for SHELXS. The weight of the Cl atom is not enough for expecting a Patterson map to disclose it easily together with sufficient C, N and O atoms for starting a refinement. Trying with direct methods was preferred in a first step. SHELXS-97 produces 4 output files from the 4 data sets above in P212121 space groups :  All these files were examined having in mind the fact that only one independent Cl atom should be found in the cell (in general position), with not neighbour at less than 3 Angstroms, and that distances between C-C and C-O atoms could be between 1.25-1.55 Angstroms. Only the M33005 proposition fulfilled these requirements provided that the Fourier map peak N°2 was discarded as a ghost of the Cl atom. 
We were expecting 33 non-hydrogen atoms in this structure so that some propositions among the 43 were also discarded on distance criteria : peaks N°21, 26 and 34. Data were used in a SHELXL-97 refinement, the number of atom was progressively reduced to 30 with a R1 reliability factor ~35% (result in tetra.lst). Then a first Rietveld fit was attempted which gave Rp = 4.9%, Rwp = 8.7% (background included) ; Rp=32.6%, Rwp=36.0% (background subtracted) and RB = 28.5%, RF= 17.6%, just refining the scale factor and zeropoint, keeping other profile parameters at their values obtained previously for the Le Bail fit (summary of FULLPROF results in tetra.sum). Then, the atomic coordinates were also refined (C/N/O undifferenciated with C diffusion factor) and the new reliabilities were Rp = 2.28%, Rwp=3.14% (background included), Rp=15.3%, Rwp = 13.2% (background subtracted) and RB=10.08 and RF=13.82 (summary of FULLPROF results in tetra2.sum). At this stage, identification of atoms was attempted, but a mixture of atoms had been obtained from which even a carbon cycle was not located ! As well as the conventional X-ray pattern, the synchrotron one was not enough resolved... Nevertheless, it was thought that methods using special treatment and non-equipartition of the intensity of overlapping reflections could possibly succeed. At least methods of molecule location should succeed. In order to verify this assumption, the PATSEE software was used with the TERMYC coordinates. The molecule was located from both 0.005 and 0.01° OVERLAP reduced dataset. Of course, the rotation and translation corresponding to the best propositions were very small :  The refinement of the structure from the coordinates of the isotypical compound (TERMYC in CSD) with synchrotron data is also available here (and see the plot) Rp =9.11 , Rwp = 7.40 % (background subtracted), RB = 4.97 and RF = 8.13%. It was also not possible to refine the tetracycline hydrochloride structure without soft constraints from the synchrotron data (interatomic distances stabilized to unusual values). Of course, no attempt was made for locating hydrogen atoms. 

Finally, a very small single crystal was selected (40x30x20 microns) and was studied at Daresbury. The structure was determined up to the H atoms location and refined without constraint (W. Clegg, to be published). The .cif file will be sent exclusively to the participants. 

Thanks for your participation. 
 

Copyright © 1998 L.M.D Cranswick & A. Le Bail

[SDPDRR Homepage | What's New | Download Data]
[CCP14 Homepage ALB Crystallography]
Please feel free to email any queries to Lachlan Cranswick at: l.cranswick@dl.ac.uk