Ab initio crystal structure determination from microcrystalline samples

A 5-minute read

More than three centuries of use the term “diffringere” and one of utilizing “diffraction” to solve crystal structures have taught us that the single crystal is well worth obtaining. Indeed, chemists across the field are not shy about investing their time (and solvents) into producing this holy grail of products: the vast majority of 50 000 structures published per year are derived from the single crystal X-ray diffraction data [1]. Crystal structure determination from powder X-ray diffraction (PXRD, aka XRPD) or electron diffraction (ED) data has often been described as a last resort, when single crystals are not available. However, as microcrystalline powders outnumber single crystal compounds by a factor of 3–4 [1], new diffraction equipment prompts two basic questions: 

• What level of accuracy can we obtain from PXRD data? 
• How close are we to being able to routinely determine crystal structures from ED data?

Powder and electron diffractionists will readily claim that there are a million reasons to solve a structure from microcrystals or polycrystalline materials. PXRD and ED approaches can allow for a fast and environmentally friendly way to analyze materials, by reducing the time and amount of solvents necessary to grow single crystals. Another argument is that technological advancements have made instruments more available and improved the accuracy of structures determined from PXRD and electron diffraction data.

Concerns in structure determination from PXRD and electron diffraction data, however, are complex and many: availability of a structural model, description of fine structural details, biases caused by geometric restraints, insufficient scattering power and sample damage caused by X-rays or electrons. Resolving these concerns is possible, if we can collect high-quality data in a short time span. The following examples focus on new diffraction equipment and present structure determination of weakly scattering and sensitive small molecules from PXRD and ED data collected with hybrid pixel detectors (HPD).

What level of accuracy can we obtain from PXRD data?

The accuracy of crystal structures obtained from PXRD data can approach the accuracy of structures determined from single crystal diffraction data, even for small organic molecules measured in a laboratory diffractometer. 

Solving a crystal structure of an organic molecule from PXRD data does not necessarily require a molecular model. In many cases, the information on chemical composition or molecular fragment is sufficient to derive the crystal structure of a material [2,3]. Rietveld refinement can confirm the backbone of the structure, while difference Fourier maps can reveal fine structural details, such as disorder (Figure 1), [3]. In some cases, the refinement can even be carried out without geometrical restrains and result in clean and interpretable residual electron densities (Figure 2).      

Figure 1. Crystal structure of D-ribose, retrieved from wrong or unknown molecular structure [2, 3]. Disorder in one of the molecules was found in the difference Fourier maps (dmin ≈ 1 Å) and confirmed by nuclear magnetic resonance measurements. Data were collected at a synchrotron source, using a large HPC X-ray detector, covering the complete 2θ angle span in a single shot.

Figure 2. D-mannose: structure solved from single crystal diffraction data (red) compared to the structure solved from PXRD data (yellow) and its corresponding residual electron densities (blue). Rietveld refinement was performed without geometrical restraints (dmin ≈ 1 Å). Data were collected in a laboratory diffractometer, using a wide HPC detector in a step-scan mode.

Some of the prerequisites for this level of structure accuracy are the high statistics of the collected data, relatively short measurement times, and minimal temporal inconsistencies between collecting the data on low and high 2θ angles. Detector features necessary to meet these requirements are efficiency, size and resolution. Ideally, a detector would cover a complete 2θ range in a single step, whereas the high statistics can be achieved by repeating the measurements and averaging the obtained patterns (if they prove to be the same).

Although the structure determination from PXRD data offers a wealth of information, its success may be limited in cases of mixtures, inhomogeneous or very small samples, or salt-cocrystal disambiguation. The new key to surmounting these limits is electron diffraction.

How close are we to being able to routinely determine crystal structures from ED data?

As the crystalline materials got to be notoriously small and the level of desired structural details finer and finer, many researchers and scientists turned their eyes on electron diffraction techniques. Virtually unrestricted by crystal size and sensitive to hydrogen atoms, electron diffraction is predicted to expand the reach of structural analysis to samples down to nanometer scale. On the way to its applicability to a vast amount of materials, there were many obstacles, including a dedicated electron diffractometer. 

The first attempt to tackle this problem consisted of fitting an EIGER hybrid pixel detector to a Transmission Electron Microscope [1]. The first crystal structure reconstructions demonstrated two admirable facts:

• It is possible to determine the structure of a small organic compound, present in a mixture of microcrystalline and non-crystalline materials. 
• Structural analysis of a large and complex organic structure can reveal fine structural details, such as disorder.

Figure 3. Crystal structure of paracetamol. A crystal from a powder mixture was used to determine the structure of an active pharmaceutical ingredient (paracetamol). Data completeness of 40% was sufficient for the ab initio crystal structure solution, yielding a structural model down to atomic level.      

Figure 4. Crystal structure of methylene blue derivative. The data were collected by scanning the needle-shaped crystal along its long dimension. Structure solution and refinement gave insights in disorder of the BF4- counterion. The structure was found to be identical to the structure solved from single crystal X-ray diffraction data.

One of the prerequisites for this kind of structural analysis relies on the availability of a dedicated experimental apparatus. The setup used in the cited work is based on the integration of an electron-counting detector in a commercial Transmission Electron Microscope. The design of dedicated instruments is currently in progress, and represents a promising option for the future diffusion of the technique to a wider scientific audience. [4]

Conclusion

For decades, the crystal structure determination from PXRD and electron diffraction data was usually taken with a great care not to over interpret the collected data. New technologies enable high throughput measurements at laboratory diffractometers and synchrotron beamlines, yielding high-quality data. Looking deeper into these data can reveal fine structural details and achieve accuracy comparable to the single crystal equivalent. Our next challenge is to make sure we don’t underestimate the quality of PXRD and ED data, particularly if they are collected with advanced diffraction equipment.

References

1. Gruene, T. et al. (2018) Angew. Chem. Int. Ed. 57(50), 16313-16317
2. Šišak, D. et al. (2014) J. Appl. Cryst. 47, 1569-1575
3. Šišak, D. et al. (2010) Angew. Chem. Int. Ed. 49(26), 4503-4505
4. Interview with Sacha De Carlo: Future of Electron Diffraction

More on the subject

1. Synchrotron PXRD for pharma R&D
2. MicroED with DECTRIS: FAQs
3. The many faces of powder diffraction
4. High resolution in situ PXRD studies in a lab
5. Electron diffraction of submicron-sized 3D crystals