DOE Argonne National Laboratory simulations model fusion reactor blanket at record detail

Researchers have completed the largest and most detailed simulations to date of a water-cooled lead-lithium (WCLL) blanket for ITER, the fusion experiment under construction in France. The multi-institutional team captured six interacting processes, liquid-metal flow, heat transfer, buoyancy, tritium transport, water cooling and magnetic fields, at a resolution of two trillion grid points.

“We’re tackling a very challenging multiphysics, multiscale problem involving fluid systems with very different properties,” said project lead Elia Merzari, a professor of nuclear engineering and mechanical engineering at Pennsylvania State University with a joint appointment at DOE’s Argonne National Laboratory. The team’s work is a finalist for the 2026 Gordon Bell Prize, recognizing outstanding achievement in high-performance computing.

WCLL Blanket Simulations Advance Fusion Energy Modeling

The fidelity of the simulations achieved by the team hinged on capturing the complex interplay of magnetohydrodynamic (MHD) effects within the liquid metal, requiring resolutions previously unattainable. Researchers observed that MHD actively suppresses turbulence and reorganizes the flow of liquid lead-lithium, directly impacting pressure distribution, temperature gradients and overall heat transfer characteristics within the blanket module. This suppression isn’t merely a detail. It fundamentally alters how the blanket functions, demanding accurate modeling to predict performance.

The team’s simulations, run on the Aurora exascale supercomputer, used two trillion grid points to resolve these complex fluid dynamics. Understanding the nuances of this flow is critical because temperature variations within the liquid metal itself generate buoyancy forces, further influencing the flow patterns, particularly near the wall facing the plasma.

These buoyancy-driven currents add another layer of complexity, and the simulations revealed how they interact with the magnetically-controlled flow, creating a dynamic system that demands detailed analysis. The ability to accurately model these coupled phenomena represents a step forward in fusion reactor design. The high-fidelity data generated can also be used to validate and refine lower-resolution models, accelerating the design process and reducing the need for costly physical experiments. The scale of the simulation, two trillion grid points, was not chosen arbitrarily.

According to Merzari, “Modeling two trillion grid points is something you can only do with exascale computing.” This resolution is essential to capture the thin boundary layers that form in the fluid flow, layers that significantly impact heat transfer and overall blanket performance. The implications extend beyond the immediate challenges of blanket design.

“There is such interesting physics in simulations at this scale and so much we can learn as we continue working to develop better fusion energy technologies,” Merzari said. The team’s work, detailed in the paper “Exascale Multiphysics Simulations for Liquid Metal Fusion Blankets,” represents a step toward realizing the promise of fusion power.

There is such interesting physics in simulations at this scale and so much we can learn as we continue working to develop better fusion energy technologies.

Elia Merzari, a professor of nuclear engineering and mechanical engineering at Pennsylvania State University with a joint appointment at DOE’s Argonne National Laboratory

Exascale Computing Drives Two-Trillion Grid Point Resolution

Using two trillion grid points, the team’s simulations represent the first time a fusion blanket has been modeled with such fidelity, including all relevant physics. This level of detail was enabled by access to the Aurora and Frontier exascale supercomputers, facilities of the U.S. Department of Energy. Capturing these effects at such high resolution required significant code development and optimization. Researchers expanded the NekRS computational fluid dynamics code, originally built on the Nek5000 framework, to model magnetohydrodynamic (MHD) effects and adapt it for GPU-accelerated exascale computing.

Doctoral student Yu-Hsiang Lan of the University of Illinois Urbana-Champaign led this effort, a critical step in preparing for the massive simulations. “It really helped us get this work done in a fraction of the time,” Merzari said.

This finding is significant because it provides a more accurate understanding of how the blanket will behave under real-world conditions. The team also employed the open-source OpenMC code to calculate neutron transport, providing essential data on heating, tritium production and water activation as inputs to the NekRS simulations. The integration of these codes allowed for a comprehensive, multiphysics model of the blanket’s behavior. In-situ visualization, using the Ascent software, aided in analyzing the transient flow behavior captured in the simulations.

This capability allowed researchers to observe the evolving flow in detail, generating high-quality movies that aided in understanding the complex interactions within the blanket. The team used Aurora’s substantial memory capacity to run the complete multiphysics model, including MHD, at the two-trillion grid point resolution.

“The need for that resolution is really driven by the physics.” Simulation, the researchers note, is an important tool for design and can help reduce the cost of actual experiments. The team is already exploring how the simulation data can be used for further analysis and optimization, outlining a potential pathway for accelerating fusion energy development.

My role was to develop the solver that brings MHD into NekRS and validate it against similar problems in the literature to make sure everything was working correctly before we pushed it to larger scales.

Yu-Hsiang Lan, a doctoral student at the University of Illinois Urbana-Champaign

NekRS Code Development Enables Multiphysics MHD Simulations

Yu-Hsiang Lan, a doctoral student at Urbana-Champaign, spearheaded this effort, adapting and optimizing algorithms for graphics processing units (GPUs) to handle the immense computational demands. The resulting code allowed for a progressive refinement of the model, starting with lower resolutions and incrementally adding detail, a strategy that proved helpful for managing the complexity of the simulation. This approach was essential for tackling the WCLL blanket design for ITER, the international fusion reactor currently under construction in France. The fidelity achieved allows for observation of detailed flow evolution, generating data useful for validating lower-resolution models.

The team used the capabilities of both the Aurora and Frontier supercomputers, running simulations on over 9,000 nodes on each system and using nearly their full capacity. This demonstrated strong scaling, meaning that adding more computing nodes continued to reduce simulation time. The ability to perform these simulations on different GPU architectures further enhanced the team’s flexibility and efficiency.

Lan is already contributing to a separate project that aims to use the simulation data to develop simplified system-analysis tools for fusion applications. The team received computing time on Aurora and Frontier through DOE’s Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program, highlighting the importance of national investment in high-performance computing infrastructure for advancing scientific research.

Development of NekRS builds on decades of advances in the Nek5000 code, which previously supported Gordon Bell Prize-winning research and demonstrated high performance and scalability for incompressible flow simulations on CPU-based supercomputers. The ongoing development of NekRS and its application to fusion energy research highlight the potential of computational modeling to accelerate the transition to a sustainable energy future.

Now that we’ve shown this is doable, people can use the results or the code as a tool to validate a low-fidelity model.

Yu-Hsiang Lan, a doctoral student at the University of Illinois Urbana-Champaign
Stay current

See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.

Avatar of Ivy Delaney

Ivy Delaney

Ivy Delaney has been working with neural networks and machine learning since the mid-nineties, back when a couple of hidden layers and a long afternoon of training counted as ambitious. She has watched the field go from academic curiosity to the thing quietly running underneath everything, and she brings that long view to quantum computing. For Quantum Zeitgeist she covers the ground where the two fields meet. That means quantum machine learning and the variational algorithms it leans on, and it also means the less glamorous but more interesting story of classical machine learning already doing real work inside quantum machines, decoding error-correcting codes, calibrating noisy hardware and learning the error models that simulators depend on. She writes about the hardware those algorithms have to run on too, and about the post-quantum cryptography scramble that the same hardware has set off. Her stories typically start with the paper, whether that is peer-reviewed work, conference proceedings or an arXiv preprint, with the source linked so you can hold a claim up against the research it came from. She is unimpressed by benchmarks that will not say what they beat, and by demonstrations that only work in the press release.

Latest Posts by Ivy Delaney: