Crystal Structure Determination from Experimental Powder Data
The Powder Solve method, available within Reflex Plus from Accelrys, allows the determination of crystal structures of complex molecular solids directly from high-quality experimental powder diffraction data.
Here are the Powder Solve results for cimetidine. On the left is the comparison of powder patterns, on the right is the structure. Click on either for a more detailed picture.
The ideal method for solving crystal structures is single-crystal X-ray diffraction. However, attempts to grow quality crystals for single-crystal analysis can prove difficult. Often these crystallization experiments produce powders that are analyzed using X-ray powder diffraction. Subsequently, the ability to determine the crystal structure from powder data is highly desirable.
In recent years, a variety of methods have been developed to determine crystal structure directly from experimental powder data. A comprehensive review of these methods has been published.  This application note examines Accelrys' Powder Solve, the first commercial software method to determine structures from powder patterns [2-4]. An indirect method, Powder Solve applies a Monte Carlo simulated annealing approach partially based on the Structure Solve algorithm. Initial validation studies include the successful application of Powder Solve to 14 organic compounds of differing complexity.
Structure Determination from Powder Data
Given an indexable high-quality powder pattern, and the molecular fragments comprising the asymmetric unit, Powder Solve employs a three-step method for structure solution: (1) indexing the powder pattern; (2) refining the cell parameters, background coefficients, and peak intensities; and (3) solving the crystal structure.
The first step involves indexing the powder pattern to determine the cell parameters and crystal lattice class. Typically the most challenging step, it is important that the experimental powder pattern represents a single phase, displays little to no preferred orientation, and contains narrow peaks that have little overlap with neighboring peaks.
The second step involves a new technique called Powder Fit, a modified Pawley procedure to optimize cell parameters, peak profiles, integrated Bragg intensities, and background coefficients. These parameters are refined in order to minimize the weighted R-factor, Rwp, used to measure the similarity between the simulated and experimental powder patterns. The lower the Rwp value, the better the agreement.
The third step is Powder Solve itself. Before starting this step, the user must define the contents of the asymmetric unit as a series of non-overlapping rigid bodies where the junction of two rigid bodies defines a flexible torsion. Powder Solve then combines a Monte Carlo simulated annealing procedure with rigid-body Rietveld refinement with respect to all degrees of freedom in the system (i.e. torsional, translational, and rotational.) During each Monte Carlo step, the powder pattern is simulated and its agreement to the experimental powder data is calculated using Rwp. Subsequent structures are generated by changing one of the degrees of freedom in order to optimize the agreement between the powder patterns. For a problem with 10 degrees of freedom, the simulation typically takes 300,000 Monte Carlo steps.
As an initial validation study, Powder Solve successfully determined the crystal structures of 14 different organic compounds. These examples varied in complexity, from a simple structure with no torsional flexibility (1-methylfluorene, 6 degrees of freedom) to a complex structure with 12 flexible torsions (heptamethylene-bis(diphenylphosphine oxide), 18 degrees of freedom):
Synchrotron powder data was only available for three examples: fluorescein, sodium chloroacetate, and cimetidine. Powder data for the remaining samples was collected using conventional laboratory diffractometers. The Rwp values indicate that the crystal structure was determined in each case. The time column represents the time required for one Powder Solve run on an SGI O2 workstation with a single R5000 180 MHz processor. Except for the two most difficult examples, the structure was determined in just a few minutes. Multiple Powder Solve runs should be performed to validate the results.
Here are the Powder Solve results for heptamethylene-bis(diphenylphosphine oxide). On the left is the comparison of powder patterns, on the right is the structure. Click on either for a more detailed picture.
Powder Solve provides an advanced method for determining crystal structure when high-quality powder diffraction data is available. Since the method does not rely on force fields, it is applicable to a wide range of compounds, including salts, solvates, and flexible molecules. Once the crystal structure is known, additional characteristics can be examined. Such information will help researchers gather a better understanding of the compounds they work with.
K.D.M. Harris and M. Tremayne, Chemistry of Materials, 1996, 8 :2554.
F.J.J. Leusen and G.E. Engel, Journal of Pharmacy and Pharmacology, 1999, Suppl. 51:1.
F.J.J. Leusen, S. Wilke and G.E. Engel, Proceedings of the 14th International Symposium on Industrial Crystallization, Cambridge, UK, paper 249, pp. 1-6 (1999).
G.E. Engel, S. Wilke, O. Konig, K.D.M. Harris, F.J.J. Leusen, Journal of Applied Crystallography, 1999, in press.
J.M. Newsam, M.W. Deem, C.M Freeman, Accuracy in Powder Diffraction II: NIST Special Publication, 1992, 846 :80-91.