Letting computers take the lead in determining the crystal structure could lead to less precise structures than previously thought. After discovering discrepancies in the published crystal structures of two explosive iodine azide modifications, German researchers re-analyzed the old diffraction data and corrected both structures.1
“The reported structures show a suspicious disorder – others call it disorder – with half-occupied atomic positions and partially colliding atoms,” says Ulrich Müller, professor emeritus at the University of Marburg, who carried out the study together with Stephan Schulz from the University of Duisburg-Essen.
Müller is convinced that there are many more imprecise structures and points out that too much trust in computer-aided structure determination routines is not good. In this case, checking the original X-ray data revealed the presence of superstructure reflections – very faint reflections found between the main reflections – that the computer had missed, he says.
The crystal structure of iodine azide (IN3rd) was first determined in 1993 by X-ray diffraction.2 Almost 20 years later, a group around Schulz found a second form of connection.3rd The researchers have now determined corrected structures for both phases, called α-IN3rd and β-IN3rd. “The new structures paint an improved picture of atomic positions and will likely allow more accurate and precise interatomic bond distances and angles,” comments inorganic chemist Douglas Keszler of Oregon State University in the USA.
Andrew Beale, a functional materials researcher at University College London (UCL) in the UK, adds that the c-axis in the corrected crystal structure is twice longer, resulting in the formation of a supercell. “This eliminates the problem of some of the nitrogen atoms being too close together, which shows that the original structures must have been wrong,” he says. “In addition, instead of 50% of the time, two nitrogen atoms are in one position only 100% of the time, which is compatible with a superstructure.”
“This work is a textbook example of why the automated data processing of modern software packages can only support, but not replace, human know-how in determining crystal structures,” says Alfred Amon, who researches metallic and inorganic substances at the UCL. “The automated algorithms with default settings act as a black box and hide details in the data that would otherwise have been easily discovered by a human observer.” Although programs are easier to use and new algorithms have been developed to deal with more complex structures, automated tools are not flexible enough to adapt to every problem.
Beale mentions that the type of connection also plays a role. “Difficult to prepare and difficult to handle samples like these will only increase the challenges of getting a reliable structural solution,” he says.
According to Müller, the new results show that computers are still unable to compete with humans when it comes to analyzing X-ray data. ” The default settings of the automatic algorithms can be adjusted in such a way that the error described is avoided. However, this only works when done by a person with crystallographic expertise. Always check the primary X-ray data carefully before trusting a computer! ‘
But Beale thinks things could get better. “Great strides have been made in applying AI-based methods to query diffraction data, and it can be imagined that incorporating these and other problems into a learning process will improve automated structure determination.”