QUESTIONNAIRE FOR STRUCTURE DETERMINATION BY POWDER DIFFRACTOMETRY ROUND ROBIN Submitted by S. Wilke, M. A. Neumann, G. E. Engel, F. J. J. Leusen Molecular Simulations Ltd, 230/250 The Quorum, Barnwell Road, Cambridge CB5 8RE, England Only the solution of the second sample has been attempted. The quality of the experimental data is sufficient to solve the crystal structure. 1. Preliminary work 1.1 Did you obtain additional information from the chemical formula ? – No Remark: The molecular formula was taken from transparency 30 of A. Le Bail’s talk on trends in SDPD (http://www.cristal.org/iniref/ecm18/t30.gif) 1.2 Did you obtain additional information from the powder pattern ? – Yes The program Powder Fit was used to determine accurate lattice parameters, profile parameters and background coefficients from the synchrotron powder pattern. Powder Fit is part of the Cerius2 software package. 1.3 Did you extract the structure factors ? – No 2 Structure solution 2.1 Did you use direct methods ? – No 2.2 Did you use Patterson methods ? – No 2.3 Did you use another method ? – Yes 2.3.1 If yes, which method(s) ? The structure was solved using a direct space approach and a Monte Carlo/simulated annealing method. Molecules are treated as rigid bodies with a limited number of internal torsional degrees of freedom. The quality of the model structure is determined by a full profile comparison with the experimental powder diffraction pattern. Minima of the Rwp factor are located by rigid body Rietveld refinement, thus allowing for relatively high temperatures in the simulated annealing procedure. In a first step, the structure was solved with a total of 11 degrees of freedom. The initial temperature was chosen high enough to overcome all barriers on the Rwp hypersurface. The comparison with the synchrotron powder pattern was carried out in the angular range from 2.12° to 17°, covering a total of 133 reflections. Hydrogen atoms were ignored in the calculation of the peak intensities. The best Rwp factor obtained had a value of 6.04% (All cited Rwp values have been calculated without background subtraction.). In a second step, the solution was improved by searching the Rwp hypersurface in the vincinity of the minimum located in the first step. For this purpose, some of the intramolecular bonds were cut and the numbers of torsional degrees of freedom was increased. A Monte Carlo/simulated annealing run at low temperature with a small Monte Carlo step width and a total of 27 degrees of freedom resulted in an Rwp factor of 2.87%. The comparison with the experimental powder spectrum was performed for the full angular range from 2.12° to 40.0°. Hydrogen atoms were now taken into account in the calculation of the diffraction intensities. 2.3.2 Which program(s) did you use ? The structure was solved using the program Powder Solve, implemented in the Cerius2 software package. 2.4 Did you first locate the whole structure – Yes 2.4.3 Was the initial model derived from the molecular formula ? – Yes The initial molecular geometry was derived from the molecular formula by force field and semi-empirical calculations. Energy minimisations were carried out using the Dreiding 2.21 force field, the COMPASS force field and the PM3 hamiltonian, leading to three different starting geometries. The crystal structure was solved with the result of the Mopac PM3 calculation as a starting point. 2.4.4 Were the initial atomic coordinates taken from a known structure ? – No 3 Structure completion Since the whole structure was located in the structure solution step, structure completion was not necessary. 4 Final refinement Three alternations of rigid body Rietveld refinement and lattice energy minimisation were used in the final refinement step. The rigid body refinement was carried out with Powder Solve using 5 rigid bodies with a total of 11 internal torsional degrees of freedom. The lattice energy minimisations based on the COMPASS forcefield were limited to a small number of minimisation steps, thus assuring a compromise between a low Rwp factor and reasonable bond lengths and bond angles. The last Rietveld refinement step produced a Rwp factor of 2.42%. The subsequent lattice energy minimisation was stopped when the Rwp factor reached a value of 2.67%. Finally, the hydrogen bonded network was analysed. Some of the torsion angles defining the orientation of the OH groups were slightly adjusted to obtain a more favorable orientation for hydrogen bonding. This modification led to a final Rwp factor of 2.70% (Rp=2.0%). At the present stage, Powder Solve does not allow the refinement of thermal parameters and the calculation of standard deviations. All thermal parameters were set to zero. The following atomic coordinates were obtained: NAME X Y Z C1 0.66562 0.12218 -0.39185 C2 0.56719 0.19782 -0.37715 C3 0.46574 0.16839 -0.32059 C4 0.48365 0.07626 -0.26126 C5 0.52054 -0.01593 -0.31407 C6 0.62879 0.01030 -0.37581 C7 0.36954 0.05111 -0.20873 C8 0.29390 -0.03113 -0.22477 C9 0.31668 -0.11241 -0.29584 C10 0.41303 -0.07088 -0.35872 C11 0.18122 -0.04754 -0.17332 C12 0.10320 -0.13919 -0.19698 C13 0.10974 -0.18298 -0.27782 C14 0.19814 -0.13769 -0.34321 C15 0.01895 -0.17597 -0.13737 C16 -0.05442 -0.26014 -0.15697 C17 -0.04676 -0.30567 -0.23637 C18 0.03382 -0.26669 -0.29641 C19 0.21009 -0.21028 -0.42090 C20 0.56525 0.29162 -0.41612 C21 0.80850 -0.01680 -0.27748 C22 0.71608 -0.16433 -0.36146 O1 0.00132 -0.13085 -0.05839 O2 0.15712 -0.04022 -0.37577 O3 0.14787 0.01176 -0.11693 O4 0.34471 0.11668 -0.14195 O5 0.36794 0.21544 -0.31471 O6 0.58297 0.10232 -0.20556 O7 0.76187 0.13836 -0.42857 O8 0.46541 0.35105 -0.41288 N1 0.65508 0.33576 -0.46066 N2 0.74387 -0.05052 -0.35670 H1 0.21760 -0.29338 -0.40522 H2 0.28773 -0.19002 -0.46189 H3 0.12884 -0.20105 -0.46095 H4 0.08807 -0.05209 -0.41011 H5 -0.12044 -0.28934 -0.11047 H6 -0.10801 -0.37029 -0.25207 H7 0.40524 0.11026 -0.09822 H8 0.36758 -0.02116 -0.40665 H9 0.85330 0.05789 -0.29099 H10 0.87622 -0.07639 -0.26248 H11 0.74397 -0.00888 -0.22490 H12 0.80114 -0.20787 -0.35462 H13 0.67468 -0.17899 -0.42401 H14 0.65168 -0.18541 -0.31133 H15 0.47828 0.41115 -0.44744 H16 0.65055 0.40842 -0.48355 H17 0.55624 0.09247 -0.14730 H18 0.03362 -0.30295 -0.35968 H19 0.03105 -0.06157 -0.06244 H20 0.54873 -0.07041 -0.26385 H21 0.35283 -0.18457 -0.26691 H22 0.44976 -0.13789 -0.39438 H23 0.60596 -0.01311 -0.44167 H24 0.72902 0.29816 -0.48253 H25 0.80589 -0.03518 -0.40802 CL1 -0.42735 -0.92106 -0.00103 At this stage, the refined structure was compared to the structure shown on transparency 30 of A. Le Bail’s talk on trends in SDPD, and it was noticed that the (OH)C(NH2) group of the refined structure was rotated by 180° with respect to the depicted one. Indeed, turning the (OH)C(NH2) group by 180° significantly improved the Rwp factor, and after one step of rigid body Rietveld refinement, lattice energy minimisation and adjustment of the OH torsion angles a final Rwp factor of 2.15% was obtained (Rw=1.66%). The corresponding atomic coordinates are presented below. NAME X Y Z C1 0.66040 0.12422 -0.39107 C2 0.56103 0.19937 -0.37554 C3 0.46084 0.16846 -0.31900 C4 0.48227 0.07579 -0.26131 C5 0.52133 -0.01644 -0.31387 C6 0.62902 0.01130 -0.37525 C7 0.36966 0.04916 -0.20876 C8 0.29366 -0.03209 -0.22552 C9 0.31567 -0.11177 -0.29736 C10 0.41359 -0.07079 -0.35934 C11 0.18172 -0.04821 -0.17355 C12 0.10422 -0.13973 -0.19782 C13 0.11044 -0.18326 -0.27875 C14 0.19694 -0.13715 -0.34475 C15 0.02189 -0.17756 -0.13731 C16 -0.05019 -0.26283 -0.15643 C17 -0.04325 -0.30772 -0.23617 C18 0.03542 -0.26758 -0.29694 C19 0.20933 -0.20914 -0.42263 C20 0.56051 0.29254 -0.41541 C21 0.81095 -0.01770 -0.27809 C22 0.71765 -0.16355 -0.36357 O1 0.00432 -0.13289 -0.05847 O2 0.15176 -0.03886 -0.37321 O3 0.15017 0.00849 -0.11507 O4 0.34318 0.11461 -0.14261 O5 0.36221 0.21452 -0.31193 O6 0.58327 0.10140 -0.20677 O7 0.75457 0.14346 -0.42925 O8 0.65587 0.32826 -0.45965 N1 0.46572 0.35784 -0.41500 N2 0.74533 -0.04954 -0.35751 H1 0.21521 -0.29231 -0.40687 H2 0.28783 -0.18927 -0.46295 H3 0.12878 -0.19885 -0.46316 H4 0.08932 -0.05138 -0.41518 H5 -0.11447 -0.29351 -0.10915 H6 -0.10352 -0.37334 -0.25114 H7 0.39976 0.10598 -0.09727 H8 0.36867 -0.01992 -0.40665 H9 0.86470 0.05200 -0.29319 H10 0.87105 -0.08193 -0.25995 H11 0.74580 -0.00184 -0.22708 H12 0.79839 -0.20772 -0.34461 H13 0.69354 -0.17988 -0.43008 H14 0.64048 -0.18277 -0.32247 H15 0.72088 0.27963 -0.45540 H16 0.46502 0.42475 -0.44853 H17 0.55797 0.09855 -0.14837 H18 0.03466 -0.30361 -0.36024 H19 0.03762 -0.06457 -0.06170 H20 0.55070 -0.07199 -0.26470 H21 0.35190 -0.18326 -0.26730 H22 0.44877 -0.13808 -0.39541 H23 0.60666 -0.01096 -0.44139 H24 0.38740 0.34826 -0.38386 H25 0.80691 -0.03343 -0.40901 CL1 -0.42807 -0.92052 -0.00195 It has to be pointed out that the right orientation of the (OH)C(NH2) group has been missed due to the small amount of time spend on the final structure refinement. The hydroxy group and the amino group carry the same number of electrons and have comparable distances to the central carbon atom. Therefore it was an oversight not to calculate the Rwp factor as a function of the torsion angle of the (OH)C(NH2) group, and the right orientation could have easily been found. The Monte Carlo / Simulated annealing run that solved the structure did not find the correct orientation, because the end temperature was significantly higher than the difference of the Rwp factors of both orientations. The subsequent low temperature Monte Carlo / Simulated annealing run missed out the right orientation, because the temperature was not high enough to pass the barrier between the two conformations on the Rwp hypersurface. This example demonstrates the power of fully automated and integrated software for structure solution and refinement, but it also shows the importance of chemical and crystallographic expertise in applying such software. 5 CPU requirements The high temperature simulated annealing run was performed on a Silicon graphics O2 workstation with a R5000 processor running at 180 MHz. With the molecular geometry determined by a MOPAC calculation, the structure solution took a couple of hours. The low temperature simulated annealing run was carried out on a Silicon graphics Indigo2 workstation with a R10000 processor running at 195 MHz. The run took several hours.