In the rapidly evolving field of nuclear physics, the need for ultra-fast data analysis is more critical than ever. Traditional approaches to processing enormous datasets generated by nuclear experiments often require hours or even days, delaying scientific insight and impeding experimental flexibility. However, a groundbreaking initiative spearheaded by the Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) and Oak Ridge National Laboratory (ORNL) promises to revolutionize this landscape. By developing a cutting-edge software pipeline known as DELERIA—the Distributed Event-Level Experiment Readout and Integrated Analysis system—researchers aim to analyze massive nuclear physics data streams in mere seconds, enabling real-time experiment adjustments that optimize outcomes.
DELERA is designed to accompany the Gamma-Ray Energy Tracking Array (GRETA), a next-generation nuclear spectrometer currently under construction at Berkeley Lab. GRETA represents a leap forward in detector technology, encompassing a spherical assembly of 120 hyper-pure germanium crystals arranged to capture detailed data during nuclear reaction events. Scheduled for installation at the Facility for Rare Isotope Beams (FRIB), Michigan State University, in 2026, GRETA will investigate atomic nuclei by detecting the gamma rays released when samples are bombarded with charged particles accelerated within FRIB’s radioactive ion accelerator. This innovative combination of experimental hardware and software promises to unravel complex nuclear structures with unprecedented precision.
The software backbone supporting GRETA’s mission, DELERIA, enables the system to stream raw experimental data directly to some of the nation’s most powerful supercomputing facilities. By leveraging the Energy Sciences Network (ESnet), a DOE-supported infrastructure for high-speed data transport, DELERIA transmits experimental information from Berkeley Lab to Oak Ridge Leadership Computing Facility (OLCF) processors nearly 2,000 miles away. This direct pipeline allows the massive dataset from photon collision events to be processed instantaneously within sophisticated computing clusters, eliminating the need for costly and bulky local computing setups. The result? Scientists receive real-time feedback that can inform experimental settings dynamically, increasing the expedition and accuracy of nuclear research.
GRETA’s design hinges on its array of ultra-pure germanium detectors, each capable of recording the energy and positional coordinates of incoming gamma photons emitted during nuclear reactions. When the radioactive ion beam collides with atomic samples, the resultant nuclear transitions generate an intense flux of photons—on the order of hundreds of thousands per second—that strike the detectors. Each interaction generates electrical signals transmitted through multiple contact points on the crystals. Computing these signal patterns to determine the three-dimensional coordinates (X, Y, Z) and energies of photons requires immense computational throughput.
Here, DELERIA’s intricate data pipeline shines. The computational workload involves transforming complex raw data from the individual crystal contacts into meaningful spatial and energetic information about each photon event. This is performed under stringent latency constraints: each interaction must be analyzed and recorded within 10 seconds to maintain real-time feedback. The ORNL team, utilizing OLCF’s Defiant cluster—an AMD GPU-accelerated system consisting of 36 nodes—provides the necessary parallel processing power to tackle these calculations. By offloading data analysis to national computing centers, DELERIA drastically reduces GRETA’s local infrastructure demands, offering a scalable and flexible solution that can adapt to future upgrades or extensions to other experimental facilities.
A key innovation of DELERIA lies in its remarkable data compression capabilities. By intelligently filtering and interpreting collision events, the software reduces data storage needs by a factor of 40, achieving a 97.5% reduction in data volume without compromising informational fidelity. This compression allows scientists to preserve critical data for prolonged analysis or archival purposes without overwhelming storage systems. Simultaneously, the system maintains the high granularity needed for detailed nuclear modeling and interpretation.
To validate the feasibility of this ambitious pipeline, the research team constructed a virtual testbed simulating GRETA’s detector array at Berkeley Lab. This simulation generates synthetic photon collision events that mimic real experimental conditions and sends their data streams via ESnet over 4,000 miles round-trip to the OLCF computing cluster for processing. The system currently handles approximately 480,000 photon events per second, processing and returning analyzed data to Berkeley in under 10 seconds—demonstrating the capacity for true real-time nuclear event analysis across continental distances.
Despite this success, the team faces formidable challenges associated with data latency over long-distance transmission. While the computational processing time for a single photon event is an impressively low 5 milliseconds, the light-speed-limited data transmission between Berkeley Lab and OLCF takes approximately 120 milliseconds round-trip. This disparity highlights inherent physics-imposed limits that cannot be bypassed, necessitating novel computational strategies to maintain efficient throughput.
To overcome latency bottlenecks, researchers have devised a method of parallel event processing capable of “tricking” the delay caused by data travel time. By concurrently analyzing multiple photon events in a pipeline fashion, where some events are being processed while others are still in transmission, the system keeps computing resources fully occupied. This concurrency-driven approach ultimately accelerates overall analysis timelines by roughly an order of magnitude, ensuring the supercomputer cluster remains busy and avoids idle wait times during transmission delays.
This DELERIA-powered infrastructure is part of a broader vision materializing under the Advanced Computing Ecosystem (ACE) program at OLCF. ACE serves as a testbed infrastructure providing experimental capability across a spectrum of computing architectures, fully integrating DOE’s leadership-class computational centers with national scientific facilities under the Integrated Research Infrastructure (IRI) initiative. By aligning high-performance computing assets, cutting-edge networking, and user-driven research, ACE targets transformative scientific capabilities across disciplines—from nuclear physics to materials science and beyond.
This pioneering collaboration not only pioneers handling extreme-scale nuclear data streams but also sets the stage for scaling similar data pipelines across diverse DOE research platforms. Among the core team leading this endeavor are Gustav Jansen from ORNL, Mario Cromaz from Berkeley Lab, and experts across the ESnet and OLCF teams. Their collective expertise demonstrates a successful proof of concept for streaming and processing big data in real-time—an imperative advancement as experimental science delves into increasingly complex phenomena demanding rapid analytic turnaround.
In the long term, DELERIA and GRETA’s synergy could significantly expand the frontiers of nuclear physics, allowing scientists to investigate fine-grained nuclear structure and reaction dynamics faster and more reliably than ever before. This capability is essential for answering fundamental questions about atomic nuclei and their behavior, which has implications ranging from nuclear energy to astrophysics. As the GRETA spectrometer comes online and DELERIA matures, the scientific community anticipates a new era where data is more than just collected—it is immediately harnessed, understood, and applied.
Ultimately, the successful deployment of DELERIA stands as a testament to how combining innovative software design, powerful supercomputing resources, and ultra-high-speed networking infrastructure can transform scientific discovery. By pushing the boundaries of data processing speed and efficiency, this project not only accelerates nuclear research but also lays down scalable frameworks for future scientific Big Data challenges across disciplines and national laboratories.
—
Subject of Research: Nuclear physics data analysis and high-speed data pipeline development for gamma-ray spectroscopy
Article Title: Accelerating Nuclear Discovery: Real-Time Data Streaming and Analysis for the GRETA Spectrometer
News Publication Date: 2025-05-19
Web References:
– GRETA Technology: https://greta.lbl.gov/technology
– Facility for Rare Isotope Beams (FRIB): https://frib.msu.edu/
– Energy Sciences Network (ESnet): https://www.es.net/
– OLCF Defiant: https://www.olcf.ornl.gov/olcf-resources/compute-systems/wombat/
– OLCF Integrated Research Infrastructure: https://www.olcf.ornl.gov/2023/11/10/integrated-research-infrastructure-at-ornl/
– Lawrence Berkeley National Laboratory News: https://newscenter.lbl.gov/2025/05/19/building-a-data-pipeline-to-accelerate-discovery/
– DOE Office of Science: https://energy.gov/science
Image Credits: Jason Smith/ORNL, Berkeley Lab, U.S. Department of Energy
Keywords
Supercomputing, National laboratories, Nuclear energy, High-speed data streaming, Gamma-ray spectroscopy, Real-time data analysis, Nuclear physics instrumentation, DOE research facilities, Data pipeline, Distributed computing
