SAXS speeds ahead with real-time visualizations

You’re back at your home institute analysing gigabytes of raw data collected at the ESRF during several days of experiment. You spot something unusual. To properly investigate, you need to change the sample’s concentration or view it at a smaller scale. But your beam time is up. You wish you could turn back the clock to have just another few minutes in Grenoble.

Thanks to the online computing resources in the ESRF’s structure-of-soft-matter group, users can visualise the results of small-angle scattering (SAXS) measurements in real time. SAXS probes the nanostructure and non-equilibrium dynamics of materials in which thermal fluctuations are sufficient to alter their properties on time-scales of a millisecond or shorter. Examples range from liquid crystals and polymers to granular materials and self-assembling amphiphilic molecules.

Online processing

The SAXS detectors produce hundreds of gigabytes of raw data per day from X-rays scattered at small angles due to nanoscale order in the material. While online preprocessing is a routine part of many experiments at the ESRF, the ID02 beamline is uniquely equipped to reduce data from multiple 2D time-resolved detectors online. The first step is to correct for detector imperfections, such as geometric distortion. Next, the 2D spatial coordinates have to be transformed to polar coordinates and then azimuthally integrated to obtain 1D scattering profiles as a function of scattering vector. Finally, the results must be fitted with functions using a MATLAB-based program, allowing researchers to derive or infer real-space structure.

With data acquisition times in the millisecond range, a lot of processing power is needed for the online data reduction to match the high frame rate. “We put a lot of effort into the software to enable this,” says ID02 scientist Michael Sztucki. Staff in the structure-of-soft-matter group have developed a data analysis package that can operate simultaneously with several detectors and has a graphical interface. The software is written in Python, and the data from different detectors can be processed in parallel.

“You can produce a very large amount of data but their worthiness need to be cross-checked,” explains ID02 scientist in charge, Theyencheri Narayanan. “The online processing of data allows us to detect and eliminate many pitfalls in typical scattering experiments at an early stage.”

Online processing of raw data is also vital in ID02’s anomalous scattering experiments (ASAXS). By recording data at various energies close to the X-ray absorption edge of a certain element, ASAXS experiments can determine the position of that element in the structure. But changes in the intensity due to this anomalous scattering are tiny compared to artefacts arising from the measurements at different energies. Michael Sztucki and co-workers have used ASAXS to deduce the spatial distribution of counterions, providing a crucial input to theoretical models of complex soft-matter systems dominated by electrostatic interactions. And in “kinetic” experiments that investigate the spontaneous self-assembly of soft-matter systems, the data-reduction software allows researchers to evaluate the robustness of pathways followed in the self-assembly process.

Other beamlines in the ESRF structure-of-soft-matter group will soon have online data-analysis schemes. ID10 is one of them. The goal is to allow intensity–intensity autocorrelation functions of 2D scattering patterns to be calculated online at the rate of 1000 frames per second to derive or quantify the dynamics of the system, which is currently performed offline.

Structural simulation

In ID13, users need to study the orientation of the nanostructure as a function of position, for which 2D patterns are the most useful because the systems are not isotropic. And at ID09, where photosensitive systems are probed by X-rays after a pump laser pulse to see how they change the structure, users rely heavily on molecular dynamics simulations. Here, the challenge is to calculate the time-dependent atom–atom correlation functions, which requires a priori knowledge of the system, and users then extract a solution iteratively by simulating tens of thousands of 3D structures. This modelling is especially intensive computationally when, for example, extracting low-resolution structure from non-crystallisable proteins in solution.

“The workstations are all pretty new so we don’t need any major hardware at this time,” says Sztucki. “One issue for the group is finding the time to develop analysis routines for simulations so that users can extract even more from scattering experiments.”

Matthew Chalmers

This article appeared in ESRFnews, March 2011. 

To register for a free subscription and to rapidly receive the current issue, please go to: