Exascale Day Audio: Ariana Peck

Impact Area: Free Electron Laser Research


Ariana Peck, SLAC National Accelerator Laboratory


Hi. I’m Ariana Peck from SLAC National Accelerator Laboratory and a researcher contributing to the Exascale Computing Project. Exascale computing serves as a powerful lens that allows us to investigate structure and dynamics at a molecular scale from diffraction images collected at the SLAC Linac Coherent Light Source (LCLS for short), which is one of the world’s premier x-ray Free Electron Laser facilities. We expect that exascale capabilities will have a transformative impact on our science by enabling us to analyze terabytes of imaging data from our free electron laser in quasi real time. This rapid analysis in turn provides critical feedback for our users, who can adjust data collection on-the-fly to maximize the insights derived from each experiment.

The x-ray light pulses from LCLS are extremely short and bright, sort of like a photographic strobe. In serial femtosecond crystallography, we can use these pulses to create movies of enzymatic reactions that are critical for life on Earth, like the capture of sunlight to produce oxygen. However, an ambitious experiment like this requires us to sequentially expose hundreds of thousands of crystal samples to x-rays, record the diffraction patterns, and then to look at the data with the fastest supercomputers available to calculate the molecular structure. Faster computing allows us to run more accurate algorithms, which in turn permit us to observe finer molecular details. Ultimately, we are interested in the motion of a single electron, as it transfers energy from one atom to another.

Single-particle imaging is a second method that we’re transforming with exascale computing. In these experiments, we collect terabytes of diffraction data of randomly oriented particles with the goal of solving the 3-dimensional structure of the particle in a near-physiological state, unaffected by crystallization or freezing. By leveraging exascale computing, in mere minutes we can process hundreds of thousands of noisy images to determine the structure at nanoscale resolution, in some cases resolving the particle in distinct biological states. Exascale resources will enable us to scale this data processing workflow to many nodes and assess its performance using high-throughput simulations, anticipating the massive volumes of data that will be routinely collected once ongoing upgrades to the LCLS facility are complete.

These examples are just two of many experimental techniques that we’re excited to accelerate, transform, and automate with exascale computing resources. Serial femtosecond crystallography and single-particle imaging have provided invaluable test beds to demonstrate the speed and scale of analysis that are possible at exascale. In the future, we’re excited to harness what we’ve learned from ExaFEL to advance algorithms used by other challenging experiments performed at LCLS.