02-219er 244.249

Subtleties in Crystal Structure Solution from PowderDiffraction Data Using Simulated Annealing:Ranitidine Hydrochloride Department of Physics and Astronomy, State University of New York at Stony Brook, Stony Brook, New York 11794-3800 Received 3 June 2002; revised 24 July 2002; accepted 9 August 2002 ABSTRACT: Recent advances in crystallographic computing and availability of high-resolution diffraction data have made it relatively easy to solve crystal structures frompowders that would have traditionally required single crystal samples. The success ofdirect space methods depends heavily on starting with an accurate molecular model. Inthis paper we address the applicability of using these methods in finding subtleties suchas disorder in the molecular conformation that might not be known a priori. We useranitidine HCl as our test sample as it is known to have a conformational disorder fromsingle crystal structural work. We redetermine the structure from powder data usingsimulated annealing and show that the conformational disorder is clearly revealed by thismethod. ß 2003 Wiley-Liss, Inc. and the American Pharmaceutical Association J Pharm Sci 92:244-249, 2003Keywords: ranitidine HCl; simulated annealing; powder diffraction; structure to understand their limitations. In particular,they depend on constructing a parameterized Powder diffraction techniques have traditionally model of the molecule and so it is possible to been used for identification and quantification of encounter problems that have subtleties that are polycrystalline material and solving simple crys- not embodied in the model. In this paper we tal structures. The information contained in a address such a problem by using simulated powder diffraction pattern is intrinsically less annealing to determine the structure of a com- than that obtained from a single crystal, as the pound that is known to have site disorder, so that three-dimensional intensity distribution is com- the molecule does not fit into the unit cell in a pressed to one dimension. Recently, methods have single configuration. We use the well-known ulcer been developed to solve increasingly complicated medication ranitidine HCl (N-(2-{[5-(dimethyla- organic molecular structures from powder data by minomethyl)-2-furanyl]methylthio}ethyl)-N0- modeling the structure in direct space, using methods such as random searches,1 Monte Carlo,2 ide, C13H23N4O3Sþ Á ClÀ). Ranitidine HCl is an genetic algorithms,3,4 and simulated annealing.5,6 H2-receptor antagonist used for treatment of Because these methods are being improved to peptic ulcers and related disorders. The crystal solve more complicated structures, it is important structure of form II ranitidine HCl is known fromsingle crystal data,7 and the N-ethyl-N0-methyl-2-nitro-1,1-ethenediamine moiety takes two confor- Correspondence to: Ashfia Huq (Telephone: 631-632-8157; mations, so that there is 50% occupancy in each of Fax: 631-632-4977; E-mail: ashfia.huq@sunysb.edu) two sites for two nitrogen, one carbon, and two Journal of Pharmaceutical Sciences, Vol. 92, 244-249 (2003)ß 2003 Wiley-Liss, Inc. and the American Pharmaceutical Association JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 92, NO. 2, FEBRUARY 2003 RANITIDINE CRYSTAL STRUCTURE DETERMINATION WITH POWDER DIFFRACTION DATA Nobs data points in the profile, and Nvar param-eters varied in the fit.
Ranitidine HCl powder was purchased from US The premise of direct space structure solution Pharmacopoeia and was dried for 1 h in a vacu- is that most bond distances and angles can be um at 608C. The powder diffraction pattern was predicted by molecular mechanics or other means collected on beamline X3B1 of the National Syn- to required accuracy. However, torsions around chrotron Light Source at Brookhaven National single bonds cannot be predicted by such methods, so the task of direct space structure solution is selected by a double crystal Si(111) monochroma- essentially to twist up the molecule and locate it tor. The sample was loaded in a 1.5-mm thin- within the crystallographic unit cell to produce walled quartz capillary and mounted on the the best agreement with experimental data. The horizontal axis of the diffractometer. The dif- initial configuration of the molecule in this pro- fracted X-rays were selected by a Ge(111) analy- blem was obtained from the molecular modeling zer crystal on the detector arm to obtain angular program CS Chem3D, where the energy of the resolution of $0.018 full width at half-maximum molecule was minimized using semi-empirical (fwhm). Diffracted X-rays were detected by a com- quantum mechanical methods (MOPAC).10 For mercial NaI scintillation detector, and the mea- ranitidine HCl, this leaves 20 parameters to solve sured X-ray counts were normalized to the signal the structure. Eleven of these parameters are the from an ionization chamber between the mono- torsion angles shown in Figure 1. Three param- chromator and sample to correct for decay and eters (Euler angles) give the orientation of the fluctuations of the incident beam intensity.
ranitidine molecule, and the remaining six param-eters are fractional coordinates that locate theranitidine molecule and the ClÀ in the cell.
We used a locally developed simulated anneal- ing algorithm, PSSP, to find the best agreement between calculated and observed diffractionpatterns.6,11 We define a parameter S, which is The cell was first indexed using the program related to the weighted R factor of powder diffrac- TREOR.8 Indexing indicated a monoclinic cell tion, and seek to find the solution that minimizes indexed, we refined the powder pattern using only the lattice and profile parameters to describe the position and shape of all Bragg peaks, iteratively adjusting the intensity of each peak; this is li is the calculated profile of the LeBail fit at the ith point. The minimum value of S is sought commonly known as a LeBail fit. We performeda LeBail fit to the measured powder diffractionprofile using the program FULLPROF.9 Theprofile fit gave us figures of merit Rwp ¼ 6.08%and w2 ¼ 2.92, where In eqs. 1 and 2, Ioi and Ici are observed and cal-culated intensities of the ith profile point, respec-tively, and w Sketch of the ranitidine molecule showing the ith profile point, which is the inverse of the the 11 torsion angles that are used as internal degrees variance of that observation. There are a total of JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 92, NO. 2, FEBRUARY 2003 (a) The two solutions obtained from PSSP with S ¼ 0.046 and 0.048. (b) Two solutions for which S ¼ 0.057 and 0.058, having a different conformation. (c) All foursolutions obtained from PSSP superimposed. The views are along the crystallographic a*direction.
by simulated annealing, where we hypothesize PSSP starts out by performing Monte Carlo that S represents the energy of an imaginary searches to sample the configuration space at some physical system that is minimized by raising its high temperature. Random starting parameters temperature to some high value and gradually give S in the range 20-50; starting temperature lowering it, allowing it to seek the configuration(s) (dimensionless) was chosen as 50, so that essen- of lowest energy. A description of how S can be tially all moves would be accepted. We used economically calculated from integrated intensi- Nobs ¼ 100 reflections in the simulated annealing ties, without loss of information by overlapping algorithm computed 106 structures (requiring JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 92, NO. 2, FEBRUARY 2003 RANITIDINE CRYSTAL STRUCTURE DETERMINATION WITH POWDER DIFFRACTION DATA (a) Refined structure of the molecule obtained from ab initio powder structure solution for a single molecular configuration. Spheres are 50% density contoursof isotropic thermal parameters. Arrows indicate the five atoms that are configurationallydisordered. (b) Refined structure with both positions of disordered atoms N14, C16, C18,O20, and O21 indicated. (c) Single crystal solution (coordinates from ref. 7). All views arealong the crystallographic a* direction.
$1 h on a 650 MHz Pentium) before repeatedly The structures of the four solutions that came lowering the temperature by 20% until a final from PSSP, without refinement, are shown in temperature of 0.001 (dimensionless) was reached.
Figure 2. It is immediately seen that there are two We carried out 50 such calculations and obtained pairs of very similar solutions. In all cases, the four solutions, for which S ¼ 0.048, 0.046, 0.057, backbone from C8 and C9 to C13 is essentially and 0.058. The unsuccessful trial runs typically identical, but there are two different locations found for atoms N14, N16, C18, O20, and O21, JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 92, NO. 2, FEBRUARY 2003 ˚ ) intensity from ranitidine HCl as a function of diffraction angle 2Y. Shown are (a) the observed pattern (dots) and the best Rietveld-fitprofile (line), (b) peak positions, and (c) the difference curve between observed andcalculated profiles.
which are precisely the atoms that had 50% fragment of the molecule containing the atoms occupancy in two different sites in the single that are known to be disordered from the single crystal solution. This result suggests that PSSP crystal study.7 It is unrealistic that adjacent has found the two distinct conformations known bonded atoms would have vastly different thermal from the single crystal structure, without any parameters in the absence of disorder, which is also a strong indication that the model being used We obtained Rietveld refinements from each is not a correct description of the crystal structure.
of the four PSSP solutions using GSAS.12 We Considering that the four solutions can be refined 96 variables, including the lattice param- superimposed with two distinct conformations of eters, profile parameters, fractional coordinates, the fragment containing atoms N11, N16, C18, and individual isotropic thermal parameters for O20, and O21, it is natural to try a Rietveld each nonhydrogen atom, and typically obtained refinement using a starting model where both of Rwp ¼ 11.12 and w2 ¼ 10.56. These refinements re- these conformations are considered to have half quired application of soft restraints on certain occupancy. This method gave a significantly better bond distances and angles to obtain a stable solu- fit, yielding Rwp ¼ 8.39% and w2 ¼ 5.88. (We also tion. In the powder solution we notice unusually carried out refinements starting from the known large thermal parameters for the atoms N11, N16, model of disorder from single crystal results and C18, N19, and O21 (Figure 3a) which are in the obtained an essentially identical fit, with Rwp ¼ JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 92, NO. 2, FEBRUARY 2003 RANITIDINE CRYSTAL STRUCTURE DETERMINATION WITH POWDER DIFFRACTION DATA 8.43 and w2 ¼ 5.61.) The molecule obtained from Sciences of the US Department of Energy under our model and the published structure from single crystal work are shown in Figures 3b and 3c,respectively. If we put in the hydrogen atoms in themodel, we get an even better agreement with the acquired data, with Rwp ¼ 7.41 and w2 ¼ 4.51; thisis the Rietveld refinement shown in Figure 4.
1. Masciocchi N, Bianchi R, Cairati P, Mezza G, Pilati Neither this work nor the single crystal diffrac- T, Sironi A. 1994. P-RISCON: A real-space scaven- tion can reveal whether this disorder is static or ger for crystal structure determination from powder dynamic, the magnitude of the reorientation time, diffraction data. J Appl Crystallogr 27:426-429.
or the short-range correlations that may occur 2. Tremayne M, Kariuki BM, Harris KDM. 1996. The development of Monte Carlo methods for crystal structure solution from powder n data: Simulta-neous translation and rotation of a structural frag-ment within the unit cell. J Appl Crystallogr 29:211-214.
3. Kariuki BM, Serrano-Gonza'lez H, Johnston RL, Harris KDM. 1997. The application of a genetic We have shown that it is possible to find a stable, algorithm for solving crystal structures from refinable structure of ranitidine HCl using pow- powder diffraction data. Chem Phys Lett 280: der data. The molecular disorder in the crystal structure is clearly seen in two different ways: 4. Shankland K, David WIF, Csoka T. 1997. Crystal distinct solutions have nearly identical figures of structure determination from powder diffraction merit, and Rietveld refinements that do not data by the application of a genetic algorithm. Z incorporate the conformational disorder lead to unrealistic thermal parameters. The solutions 5. David WIF, Shankland K, Shankland N. 1998.
Routine determination of molecular crystal struc- can be combined to give a disordered structure tures from powder diffraction data. Chem Comm with acceptable thermal parameters, which is identical to the previously known structure deter- 6. Pagola S, Stephens PW, Bohle DS, Kosar AD, mined with single crystal data. We have also Madsen SK. 2000. The structure of malaria pig- shown in such cases it is crucial to analyze several ment b-haematin. Nature 404:307-310.
solutions obtained from simulated annealing cal- 7. Ishida T, In Y, Inoue M. 1990. Structure of rani- culations to obtain the detailed crystal structure.
tidine HCl. Acta Crystallogr C46:1893-1896.
8. Werner PE, Eriksson L, Westdahl M. 1985. TREOR, a semi-exhaustive trial-and-error powder indexingprogram for all symmetries. J Appl Crystallogr 9. Rodriquez-Carvajal J. 1990. Program FULLPROF, We are very grateful to referees for encour- Abstracts of the Satellite Meeting on Powder Dif- fraction of the XV Congress of the IUCr, Toulouse, manuscript. Research was carried out at the 10. http://home.att.net/$mrmopac; http://www.camsoft.
National Synchrotron Light Source at Brookha- ven National Laboratory, which is supported 11. Pagola S, Stephens PW. 2002. Submitted to J Appl by the US Department of Energy, Division of Crystallogr; also http://powder.physics.sunysb.edu.
Materials Sciences and Division of Chemical 12. Larson AC, Von Dreele RB. 1987. Program GSAS, Sciences. The SUNY X3 beamline at NSLS is General Structure Analysis System (Los Alamos supported by the Division of Basic Energy Laboratory Report No. LA-UR-86-748, Los Alamos).
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 92, NO. 2, FEBRUARY 2003

Source: http://reactos.ccp14.ac.uk/ccp/web-mirrors/pssp/pdf/jps.pdf