Single crystal X-ray diffraction (XRD) is the basis for the field of crystallography, a technique that has been so helpful in characterizing molecules, large and small, from DNA and proteins to pharmaceuticals. These methods rely on ‘single’ crystals of a large enough size to study in an X-ray beam, typically 100 μm on all sides. Large molecules in crystals produce data with thousands of diffraction spots, enough spots that you can resolve the structure with a little hard work. However, crystal growing is a dark art, and sometimes nature puts a variety of impediments in the way of growing nice, large crystals for XRD. These impediments give experimental scientists a major headache, and sometimes causing years of delay between the discovery of a new material or molecule and its eventual structural characterization.
One solution to this problem is Serial Femtosecond Crystallography (SFX), a technique that involves firing a jet of smaller crystals, suspended in a liquid, into the path of an X-ray free-electron laser (XFEL), an extremely bright coherent X-ray source generated from free electrons produced by a particle accelerator. This method packs incredible brightness into an extremely short (40 fs) pulse. Since the time of exposure of any one crystal is so short, it is a good choice for any crystal that is too small or too sensitive to ionizing radiation to study by traditional methods.
The SFX technique was first developed to solve crystal structures of big molecules: proteins. Hitting these crystals with the XFEL beam produces rich datasets with thousands of spots, or as Nate Hohman from the University of Connecticut put it, “You have spots for days”. But what Nate is interested in is metal-organic chalcogenolates (MOChas), hybrid semiconductors that self-assemble into small crystals that are just a few microns across. The crystals are too small for traditional characterization; single crystals that are large enough to be characterized using standard X-ray diffraction techniques can be difficult to produce and create a lot of chemical waste, each example consuming many long months of effort. What Nate wants is a method to solve structures from an easysynthesis method that produces small molecules.
Nate’s graduate student, Elyse Schriber, suggested using SFX to characterize their tiny MOCha crystals, however, there was a major conundrum: any single hit of a crystal might only produce about twelve diffraction spots, too few to use existing methods to interpret the data.
Schematic of the Serial Femtosecond X-ray crystallography set-up, taken from the original publication*
a) Tiny crystals in solution are passed through the collimated X-ray beam (XFEL). b) When an X-ray strikes a crystal it diffracts. c) The diffraction pattern hits the detector and diffraction spots are seen in different locations depending on the orientation of the crystal.
When beamtime allocation for Nate and Elyse to use XFEL SFX came up at the Linac Coherent Light Source in California, they had another problem: the crystals wouldn’t stay in solution. Pressed for time and needing a quick solution, Nate had a nifty trick up his sleeve. Dish detergent helps stabilize the chalcogenolates in solution. “Go for the one with the duck on it”, he told his team. With their small molecules now in the duck sauce solution, they could start collecting data.
Data analysis (big time!)
Although SFX requires careful calibration and sensitive equipment, it sounds like this might have been the easy part of the experiment. The data still need to be analyzed; there are only a few diffraction spots for each small molecule, and the random orientation of each molecule as it passes through the beam isn’t known. They also have a huge volume of data to wade through.
Aaron Brewster, research scientist at the Lawrence Berkeley National Laboratory, came up with an algorithm based on graph theory to determine the orientation of each crystal and then aggregate this information to compile diffraction patterns from crystals. Further data analysis can reveal the crystal structure of these small molecules (sm) so the technique as a whole is given the acronym smSFX.
Aaron developed his novel algorithm way back in 2015, but it stayed largely on the shelf until Nate’s team came along with the perfect problem: samples where there are enough crystals of the right size, a size too small for usual crystallographic methods.
Right now SFX is only available at five facilities in the world. Factor in Aaron’s novel algorithm and it seems that this new technique isn’t very accessible. Nate wants to change that too. With further work he wants to turn smSFX into a mail-in service. One reason that the technique generates large volumes of data, of the order of terabytes, is that for every shot of the x-ray beam only about 10% of the jets of liquid would contain a crystal (even after the concentration of the duck sauce has been fine-tuned) and of that 10%, only 8-10% of these successful shots could capture enough information to be useful to the algorithm. Since just 1% of the data that are collected contain truly useful information, this means you have to collect data for longer. Some SFX facilities deliver 120 shots per second while others deliver 3600 shots per second. Since the small molecules are easy to make, Nate knows that it’s straightforward to “squirt more duck sauce through there” and use a high rate of shots to minimize instrument time.
The terabytes of data were analyzed on a supercomputer, and it took a few years to perfect the algorithm using this set of samples that were the perfect problem for Aaron to solve. Now that the technique has been developed and proven to work, it can be made more accessible, and now structural solutions can be solved in real time during the experiments.
Materials synthesis and some cool photo-physical properties
The chalcogenolates self-assemble to create 2D materials, in this case inorganic polymers that are supported by organic ligands. They also don’t exhibit nano-scale size effects: their properties are the same as for the 3D material. Self-assembly is a much more straight forward way of creating 2D materials than techniques, such as exfoliation, which are used to create graphene. This could be a more cost effective way of creating materials that don’t exhibit nanoscale effects. The high exciton binding energy of these chalcogenolates means that these materials have applications for semiconductors, photovoltaic cells that might be more efficient that silicon-based ones, and for photo-catalysis. They may even have applications in quantum computing.
Scanning electron micrographs of the crystals, taken from the original publication*.
Scale bars are 5 μm. Left: Thiorene. Center: Mithrene. Right: Tethrene.
Nate’s long-term interest has been to create these layered materials with organic spacers. With the calcogenolates there are three components that can be varied: the metal atom; the organic molecule, or; the chalcogen atom (group 16 in the periodic table – Nate mainly works with sulphur, selenium and tellurium).
Nate very quickly hit on a silver-benzene-selenolate compound known as mithrene (AgSePh) with unusual properties: it glows blue which indicates a high energy photo-physical state. In contrast, swapping in sulfur (thiorene AgSPh) led to a compound without any photo-physical properties while using tellurium (tethrene: AgTePh) led to the emission of light at broader wavelengths when compared to mithrene.
Suspended microcrystals, taken from the original publication*
a) Thiorene, b) mithrene, c) Thethrene
The reasons for these differences were unknown until now and there wasn’t any agreement of the structure of thiorene. The new analytical technique, smSFX (recently published in Nature), showed that the 2D nature of the bonds in mithrene is broken in thiorene i.e. the silver atoms in thiorene form 1D chains while in mithrene they form 2D chains.
Side and top views of crystal structures from smSFX, taken from the original publication*
d) Thiorene, e) Mithrene, f) Tethrene.
Thermal ellipsoids for Ag (blue), S (yellow), Se (orange) and Te (magenta) are drawn at the 50% probability level.
Hydrogen atoms and one position of disordered C6H5 (for mithrene) are omitted for clarity.
g-i are models of thiorene (g), mithrene (h) and tethrene (i) with the view oriented down the c axis of the unit cell.
Final thought
This breakthrough in understanding the photo-physical properties of chalcogenolates came about through the passion and hard work of many people. Although Nate has been working on small molecules for years, and Aaron developed a novel algorithm, they may not have come together if it weren’t for the inspiration of Elyse who was working on her PhD and had an idea that led to a leap forward in materials science.
* All images from Elyse, Nate and Aaron's (and the rest of the team) recent article: "Chemical crystallography by serial femtosecond X-ray diffraction", Nature 601, pp 360-365 (2022). Published under a creative commons licence CC BY 4.0.