CPD News Index

Workshop on Using Powder Data to Solve Crystal Structures, May 18th, 1998, Brookhaven National Laboratory, Upton, USA.

This workshop was organized as part of the 1998 Annual NSLS Users' Meeting, and was motivated by the rapid advances in the application of powder diffraction techniques to the ab-initio solution of unknown crystal structures during the past few years. In contrast to the well-known Rietveld profile method for structure refinement, this is a relatively new application which has been driven mainly by the development of high-resolution synchrotron x-ray techniques, especially for framework structures such as zeolites, fullerene derivatives and other molecular compounds, and small organic molecules of pharmaceutical significance. A major goal of the workshop was to review advances in the field since the landmark 1995 Oxford meeting on "Structure Determination from Powder Data" which followed the EPDIC IV meeting in Chester and was organized by Bill David and colleagues, the proceeedings of which will appear shortly as a sequel to the popular book on "The Rietveld Method" edited by Ray Young. Through the generosity of a number of sponsors who are acknowledged at the end of this article, it was possible to invite several speakers from Europe, and the 80 or so attendees were treated to a state-of-the-art exposition by an international cast of speakers from the USA and Europe, where much of the cutting-edge work in this field is currently taking place. Eight posters were also submitted for viewing during the lunchtime break.

The proceedings were led off by Henk Schenk (U. of Amsterdam) with a brief and lucid introduction to direct methods (which to date has been the most important tool for structure solution from powder data), followed by a description of the POWSIM program, including the requirements for data collection, the decomposition of the powder pattern into integrated intensities by an iterative procedure which allows a more reliable estimate of overlapping intensities, and finally structure solution. Over 20 unknown structures have been solved in this way, the largest having 28 atoms in the asymmetric unit.

Lynne McCusker (ETH, Zurich) next opened up her structural "toolbox" and pulled out two new tools, texture and structure envelopes, as an aid to solving zeolite structures. She showed that texture (preferred orientation), normally abhorred by powder diffractionists, could in fact be put to good use to extract more reliable intensities from heavily overlapped clusters of peaks from pole figure data obtained with a texture gonimeter, resulting in a dramatic improvement in the number of framework Si and O atoms found directly. She also described how to generate a "structure envelope" (periodic nodal surfaces separating regions of high and low electron density) which can be used to facilitate computer-assisted model building of zeolite structures.

A third new tool for zeolite structure determination was discussed by Ralf Grosse-Kunstleve (Yale U.), a recent graduate of the ETH school of toolmakers. This is the so-called "FOCUS" method, which incorporates crystal chemical information into the structure determination and combines automatic Fourier recycling with a topology search specific to zeolites. Zeolites also figured prominently in the following talk by John Newsam (Molecular Simulations Corp.), who described approaches to structure determination based on model construction and simulated annealing which have had considerable success in predicting and solving new structures. Recent improvements in the algorithm have overcome some of the previous limitations and improved the performance significantly. Several examples were given, including recent applications to molecular structures.

However, the growing sense of euphoria over all these achievements was somewhat dampened in the final talk of the morning session by Dick Harlow (DuPont Corp.), who reviewed the current status of the $1000 DuPont Challenge, which is to be awarded to the first person to provide a satisfactory solution of the structure of HAlF4 (an intermediate product in the industrial synthesis of AlF3 catalysts) from powder data collected at the NSLS. In order to win the prize, the model must also account for neutron data collected from the same material, and the three "reasonable" solutions submitted to date have all failed this test so the challenge is still open! Prospective contestants should check the Web at http://www.pitt.edu/~geib/powder.html.

In the opening talk of the afternoon session, Ken Shankland (Rutherford-Appleton Laboratory) gave a high-tech presentation of real-space techniques that are suitable for organic molecules, in particular pharmaceutical compounds. He emphasized the importance of appropriate data-collection strategies, such as using longer counting times at higher angles and the use of differential thermal expansion to separate overlapping peaks. Questioning claims of progress extrapolated from "one-off solutions," he showed structures of cimetidine, promazine HCl, C24H8F10, capsaicin, and ibuprofin, solved by simulated annealing and genetic algorithms. These techniques are complementary to direct methods, in that they require detailed knowledge of the molecule's expected bond lengths and angles, so that its internal degrees of freedom can be described by a collection of several torsion angles (up to ten for the cases shown), the only optimization criterion being the agreement with the experimentally determined structure factors.

Jim Kaduk (Amoco) described the powder structure solutions of several moderately large molecules (dimethyl 2,6- naphthalenedicarboxylate, dimethyl 2,7- naphthalenedicarboxylate, tremellitic anhydride, and diammonium terephthalate). He drew attention to the fact that trial solutions with significant conformational differences can produce nearly identical fits, and that one must frequently look to subtle features of the refinement to judge the correct structure.

Robert Dinnebier (U. of Bayreuth) compared standard tools (such as direct methods) and some unconventional algorithms (extended use of chemical constraints and rigid bodies, grid searches, the use of pseudo-atoms for compact units such as phenyl rings) to a variety of structure solutions: LiC6H5, RbC5H5, yellow pigment 14 (C40H30Cl2O4), C(Si(CH3)3)4, Si(Si(CH3)3)4, and C6H5OK. He also described the use of maximum entropy to study details of disorder and anharmonic displacement in LiC5(CH3)5.

Peter Stephens (SUNY, Stony Brook) presented no new structure solutions, but instead discussed several aspects of lineshape management, such as the influence of wavelength on resolution and intensity for analyzer crystal and parallel-blade collimator setups, and the importance of using the correct geometry (rather than an empirical approximation) for the asymmetry of low-angle peaks. The bulk of his talk was devoted to a description of a newly-developed algorithm for handling anisotropic broadening due to lattice strains in whole-pattern fits. This technique, based on a multi-dimensional description of the correlations between lattice metric parameters, permits vastly improved Rietveld fits. One outcome of this work was the ability to locate a hydrogen bond in the structure of sodium- parahydroxy- benzoate, which was recently solved by ab-initio methods from powder data.

In the final talk, Bob Von Dreele (Los Alamos National Lab.) described some first attempts at protein crystallography with powder samples, motivated by the rhetorical question, "how many atoms can be included in a Rietveld refinement?" In numerical test experiments on a small protein, he found that a free refinement was unstable, but by constraining bond lengths, he could get smooth convergence. With high- resolution data collected at the NSLS on a powder sample of whale myoglobin, which has 1401 atoms in a 65,000 A**3 unit cell, it was possible to index the cell, refine lattice parameters and profile coefficients, and, with appropriate constraints, to refine the atomic structure, in only 30 minutes per cycle on a 133 MHz pentium laptop!

On this impressive note, Peter Stephens closed the meeting by urging the audience to make use of the several high-resolution synchrotron beamlines and laboratory instruments available. A wide variety of sophisticated tools required for the ab-initio solution of complex structures have now been demonstrated, and the next phase is to work on problems where scientifically or commercially important information results from the application of these new techniques.

The support of the following corporate and institutional sponsors is gratefully acknowledged: International Centre for Diffraction Data, Air Products, Amoco, Chevron, Clariant, Mobil, UOP, SUNY X3B1 Powder Diffraction Facility, X7A PRT.

David Cox and Peter Stephens
(Workshop organizers)