OFFICE OF ADVANCED SIMULATION AND COMPUTING AND INSTITUTIONAL R&D PROGRAMS

Welcome to the November 2025 issue of the ASC newsletter - published quarterly to socialize the impactful work performed by the National Nuclear Security Administration (NNSA) laboratories and our other partners.  This edition begins with Los Alamos National Laboratory’s (LANL’s) October 28th announcement regarding ASC’s fifth Advanced Technology System (ATS-5), named “Mission,” to be operational at LANL as an NNSA Trilab resource in 2027.  Other featured highlights in this edition include:  

• Lawrence Livermore National Laboratory’s (LLNL’s) Marbl multi-physics code improving modeling accuracy through a benchmark study of a National Ignition Facility (NIF) experiment.
• The Tri-Lab Remote Computing Enablement (RCE) project delivering its final milestone – a classified, production high-performance computing continuous integration capability between computer centers at LLNL, LANL, and SNL.
• Sandia National Laboratories’ (SNL’s) partnership with SpiNNcloud to evaluate the large-scale neuromorphic computing (NMC) SpiNNaker2 testbed at SNL-NM, advancing research on the impact of NMC systems for national security.

Please join me in thanking the professionals who delivered the achievements highlighted in this newsletter and on an ongoing basis, all in support of our national security mission. 

Dr. Stephen Rinehart
Assistant Deputy Administrator, ASC

---

LANL announces ASC’s Advanced Technology System-5 “Mission” system focused on national security science.  

On October 28, 2025, LANL announced its selection of HPE and NVIDIA as partners for a new supercomputer, “Mission,” to be built, delivered, and installed in the coming years, with HPE selected as the prime contractor.  Mission will support the critical modeling and simulation that underpins national security science as well as fundamental science research and artificial intelligence (AI) applications across the NNSA complex. 

Mission will be built by HPE for ASC.  The system will be based on the new HPE Cray Supercomputing GX5000 equipped with the next-generation NVIDIA Vera Rubin platform, which combines the company’s NVIDIA Vera central processing units (CPUs) with NVIDIA Rubin graphics processing units (GPUs) and interconnected with NVIDIA Quantum-X800 InfiniBand networking, named in honor of pioneering American astronomer Vera Rubin.  Direct liquid-cooled and purpose-built for exascale systems, the NVIDIA Vera Rubin platform is designed for AI-era supercomputing.

When it will be operational in 2027, Mission will be running exclusively in the classified space.  Mission will replace LANL’s current Crossroads system, leveraging its HPC and AI capabilities to support modeling and simulation efforts.  The Mission supercomputer represents the first NNSA high-performance computing (HPC) system in the post-exascale era.  Mission’s time-to-solution capabilities, especially with the ability to run multiple simulations at one time, will be unleashed on the large and complex simulations.  Mission will feature multi-tenant capabilities and can accommodate multiple workloads and user groups at the same time.  This functionality will streamline operations and allow for more research to happen concurrently.  Details on project timelines and completion will be available as the initiative is finalized.  Read more in LANL’s October 28th news release (LA-UR-25-30515).

---

LLNL's Marbl code enhances inertial confinement fusion modeling for national security.

Researchers recently modeled the N210808 burning plasma inertial confinement fusion (ICF) shot at the NIF using Marbl, LLNL’s newest multi-physics code.  Accurate modeling of radiation transport and thermonuclear burn is essential for understanding and optimizing ICF performance, supporting the NNSA’s nuclear security mission.

The team conducted a comprehensive evaluation of Marbl’s built-in diffusion module and its interfaces with two radiation transport codes: Teton (deterministic) and Imp (stochastic).  Through a series of one- and two-dimensional simulations with varying mesh resolutions, angular quadrature, and particle counts, they systematically benchmarked yield and compression results.  Special attention was given to known challenges such as ray effects and spatial asymmetries in Teton, as well as noise and computational demands in Imp, with iterative adjustments made to models and algorithms to address these issues.

The study identified key factors influencing simulation accuracy, including the choice of Compton scattering models and mesh refinement strategies, and provided actionable insights for future improvements.  These findings will guide ongoing code development, such as enhanced smoothing algorithms and alternative transport models, ultimately enabling more robust and reliable ICF predictions.  This work strengthens the NNSA’s ability to model complex high-energy-density physics, supporting critical national security objectives (LLNL-ABS-2013677).

---

ASC Tri-Lab Remote Computing Enablement project provides classified, production-quality, cross-site continuous integration capability.

The Tri-lab HPC Remote Computing Enablement (RCE) Cross-site Continuous Integration (CI) project has delivered its final milestone: achieving classified, production, Tri-lab HPC continuous integration (CI) between the computer centers of LLNL, LANL, and SNL.  Users may now easily perform DevOps CI remote computing for codes simply by interacting with their local code repositories, resulting in remote runs on systems such as Crossroads and El Capitan.  This will further the efficiency of software development and increase quality of the HPC application codes developed to meet the NNSA Stockpile Stewardship mission.  This achievement brings the years-long CI project to a successful close, with ongoing operations support now turned over to the design agency HPC centers.  Full documentation is available on the centers’ documentation resources: https://hpc.llnl.gov/technical-bulletins/bulletin-592. 

The RCE project, initiated in 2019 under the ASC Facility Operations and User Support (FOUS) subprogram, is dedicated to expanding the broader NNSA HPC user base’s access to HPC resources.  This initiative aims to broaden mission-critical modeling and simulation activities through its networking, authentication, storage, and CI working groups (LLNL-ABS-2013679).

---

SNL achieves neuromorphic computing breakthrough.

SNL’s Neural Exploration & Research Lab (NERL) has partnered with SpiNNcloud to evaluate the large-scale neuromorphic SpiNNaker2 testbed, named NERL Braunfels, which arrived at SNL-NM in March 2025.  This 1,152 chip system enhances SNL's leadership in neuromorphic computing.  As SpiNNcloud’s first customer, SNL is positioned to design the testbed to advance research on the impact of neuromorphic computing on national security.  The system enables exploration of applications in scientific computing, optimization, graph algorithms, and large neural networks.  SNL's work with this advanced system will also inform future designs of HPC systems (SAND2025-12122M).
 

 

---

Hypersonic breakthroughs: SNL is accelerating ablation modeling for nuclear weapons systems.

Recent advancements powered by the ASC program have significantly improved our understanding of the thermochemical responses of vehicles at hypersonic speeds, which is vital for the performance and safety of thermal protection systems during reentry.  Collaborative efforts have led to the development of new ASC models that accurately govern the behavior of thermal protection materials, enhancing their credibility through rigorous verification and validation.  Innovative axisymmetric formulations for coupling flow to ablation have drastically reduced simulation times from days to minutes, while also improving code robustness and predictive confidence.  These advancements are being adopted by Nuclear Deterrence (ND) systems, playing a crucial role in stockpile management and modernization efforts (SAND2025-12122M).

---

El Capitan at LLNL retains title as world’s fastest supercomputer on latest TOP500 list. 

El Capitan has once again secured the top position on the TOP500 List of the world’s most powerful supercomputers, as announced at the 2025 International Conference for High Performance Computing, Networking, Storage and Analysis (SC’25).  Developed in partnership with HPE and AMD, El Capitan is the NNSA’s first exascale supercomputer, achieving a verified 1.809 exaFLOPS (quintillion calculations per second) on the High Performance Linpack (HPL) benchmark.  This accomplishment not only reaffirms El Capitan’s status as the fastest verified computer but also highlights its critical role in national security by performing essential calculations to ensure the safety and reliability of the U.S. nuclear deterrent.

El Capitan continued its dominance by earning the “triple crown” of performance, ranking first on the High Performance Conjugate Gradients (HPCG) and HPL Mixed Precision (MxP) benchmarks, while also demonstrating strong energy efficiency by placing 23rd on the Green500 list.  The system’s theoretical peak of 2.88 exaFLOPS underscores DOE’s leadership in exascale computing, with Oak Ridge National Laboratory’s Frontier and Argonne National Laboratory’s Aurora rounding out the top three.  El Capitan’s advanced architecture, featuring over 46,000 AMD Instinct APUs and HPE Cray EX technology, enables unprecedented speed and precision for high-resolution, full-scale simulations that are vital for national security applications.

Beyond El Capitan, LLNL’s supercomputing ecosystem remains unmatched, with 13 systems on the TOP500 list—the most of any facility worldwide.  Notably, Tuolumne, a sibling system to El Capitan, maintained its No. 12 ranking and brings next-generation technology to unclassified research.  Other notable systems at LLNL such as Sierra, Lassen, and several advanced computing clusters have supported a wide range of research areas, from AI to materials science.  This extensive fleet allows scientists and engineers to efficiently match workloads to resources, accelerating innovation from concept to large-scale simulation.  Read more in LLNL’s November 17th news release.

---

LANL uses quantum molecular dynamics for better simulations of ejecta.

Many systems critical to national security can generate particulate matter—known as ejecta—from material interfaces during normal operation.  In simulations of these systems, it is essential to accurately predict the behavior of ejecta once formed.  This requires models that capture the complex interactions between particles and the surrounding gas, including drag forces, heat exchange, and potential chemical reactions.  These processes directly influence the fate and state of ejecta transport.

Recent studies, motivated by experiments on cerium ejecta in reactive hydrogen environments, have led to the refinement of a point-particle reaction model for ejecta undergoing hydriding reactions.  As hydriding progresses, the material properties of the growing hydride layer can cause flaking, or shedding, of the outer shell.  This flaking can alter the particle and ejecta cloud dynamics in two major fashions.  First, flaking reduces the size of the ejecta particles themselves, which reduces their inherent Stokes number and makes them reach momentum equilibrium more rapidly with ambient fluid flow.  Secondly, the shed hydride mass is assumed to be small enough to instantaneously equilibrate to the surrounding fluid, leading to an altered spatial and velocity distribution of the total mass in the ejecta cloud compared to a non-shedding ejecta cloud, as shown in Figure 7 for various assumed hydride shell thicknesses.

The original model tracked the inner and outer boundaries of a hydriding shell as it formed on ejecta surfaces in hydrogen gas.  It accounted for the heat released by the hydriding reaction, its transfer to the surrounding gas, and hydride phase transitions.  The model also included the potential for mass loss due to hydride flaking from expansion or hydrodynamic stresses.  However, it assumed a constant diffusion coefficient for hydrogen through cerium hydride.  As shown in Figure 8, this assumption led to poor agreement with experimental temperature data.  Simulated ejecta temperatures were highly sensitive to the diffusion coefficient, and only by reducing it by an order of magnitude did the model approach experimental values.

To address this limitation, the model was updated to incorporate a temperature-dependent diffusion coefficient derived from quantum molecular dynamics simulations.  This improvement dramatically enhanced agreement with experiments: simulated ejecta temperatures now fall within the experimental uncertainty bounds, as shown in Figure 8, without the need for manual tuning.  This improvement provides more confidence that the hydriding reaction rates of ejecta particles are captured correctly and that the resulting distribution of shed hydride mass and ejecta particles represents an expected reality for flows in applications of interest.  (LA-UR-25-29355).

 

 

---

SNL is advancing nuclear survivability assessments with RAMSES.

In the ongoing pursuit of nuclear deterrence, it is critical to assess and ensure the survivability of weapon systems in extreme radiation environments.  The ASC RAMSES (Radiation Analysis Modeling and Simulation of Electrical Systems) code suite aims to provide integrated, end-to-end extreme radiation and electrical environment assessments for key ND use cases. 

This quarter, the RAMSES team has made significant progress in enhancing simulation capabilities and refining workflows to support faster, more comprehensive assessments for our ND partners.  The RAMSES/Gemma team has successfully executed a 3,500-node simulation on the El Capitan system, testing its ability to deliver analysis for the W87-1 and W80-4 systems efficiently with large-scale simulation.  The RAMSES/ITS team has enhanced radiation transport simulations through the implementation of the new MeshTallyServer capability.  Preliminary results show a 75% decrease in the number of preprocessing steps that would normally necessitate substantial analyst involvement in the modeling and simulation process.  The RAMSES/CHEETAH-MC team has also adopted the MeshTallyServer capability.  As part of the Accelerated Digital Engineering (ADE) initiative, the RAMSES/Xyce team integrated dose rate-aware analysis into the application-specific integrated circuit (ASIC) design flow, establishing a new standard for radiation-hardened requirements and developing a suite of compact models that enhance simulation accuracy and accelerate project timelines. The Q Framework team did significant refactoring of its internal representation of a state machine model to facilitate extensions of the abstraction/refinement-checking functionality that formally verifies connections between system- and component-level models.  That and other related improvements will support application of Q Framework beyond software to formally verify ND digital hardware components going forward.  All of these activities will enable RAMSES to provide more comprehensive answers faster for our ND partners.

Looking ahead to FY26, the RAMSES team will focus on enhancing radiation survivability assessments for the W87-1 and W93 programs.  Key objectives include developing a production-ready system generated electromagnetic pulse (SGEMP) capability to accelerate cable assessments and analyzing SGEMP effects in weapon cavities under higher pressures.  The team is also working on assessing system susceptibility to electromagnetic environments and developing capabilities for thermal-mechanical shock responses (SAND2025-12122M).

---

SNL is enhancing accuracy and credibility in ballistic weapon system modeling.

Hypersonic flight requires modeling of complex, gas-phase physics that cannot be completely validated with experiments on the ground, increasing uncertainty in any flow modeling of weapon systems.  Recent research conducted by SNL significantly enhances the credibility of models used for ballistic weapon systems during reentry.  Notably, this is the first thorough validation of high-temperature, hypersonic flow, which is crucial for accurate modeling.  As a result, analysts can now more accurately measure errors and uncertainties in their models, which is essential for evaluating weapon system performance in flight.  These findings lead to more precise weapon system models, ensuring safety margins in the design of new systems, and ultimately support quicker and more reliable calculations for developing new weapon systems (SAND2025-10702M).

 

---

LANL researchers pioneer benchmarks to assess large language model reasoning under uncertainty. 

 

LANL’s critical national security mission demands advanced tools for weapons-design optimization that can reason under uncertainty, make accurate predictions, and provide reliable decision support.  Large language models (LLMs) have the potential to revolutionize this approach by automating complex workflows and enabling rapid discovery of new insights.  However, before deploying these systems in high-stakes applications, researchers need ways to evaluate their fundamental reasoning capabilities; a challenge that can be addressed through benchmark development.

A LANL team pioneered two novel benchmark suites specifically designed to test LLMs’ capacity to reason under uncertainty in mission-relevant contexts.  The first suite assesses models’ ability to make probabilistic comparisons between variables, which is a fundamental capability for weapon design optimization.  LANL’s second, more complex suite assesses models’ ability to analyze frequency data tables and identify causal relationships through intermediate calculations and probabilistic reasoning.  These benchmarks collectively assess a model’s skill in tackling a fundamental uncertainty quantification (UQ) challenge: judging inequalities between uncertain variables.  In doing so, they provide a measure of the foundational probabilistic reasoning crucial for weapon design optimization.

Tests of thirteen leading LLMs revealed insights about capabilities and limitations.  Although some models demonstrated promising performance on basic tasks, LANL’s benchmarks exposed significant gaps in their ability to handle more complex forms of reasoning under uncertainty—reasoning that is essential for weapon design.  The largest models did well on the simpler benchmark, with accuracy above 90% (Table 1).  However, only OpenAI’s models managed to get passing grades on the more complex benchmark (Table 2).  The results suggest that strong performance in one type of task can hide weaknesses in others, underscoring the need to rigorously benchmark performance across a broad spectrum of tasks.  Comprehensive benchmarks are required to ensure that AI systems can reliably meet LANL’s exacting mission requirements (LA-UR-25-29752).

 

---

A Legacy of Leadership: LLNL’s Terri Quinn retires after four decades of service and innovation.

Terri Quinn, an icon of HPC leadership in the ASC program, retired from her position as LLNL’s Livermore Computing (LC) Systems and Environments Associate Program Director as of the end of October 2025. 

Terri’s distinguished career began after earning her BA in Mathematics from University of California, Irvine and serving three and a half years in the Naval Reserve, where she worked under Admiral Hyman G. Rickover at Naval Reactors.  Following her master’s degree at UC Davis, she joined LLNL in 1984 as a software engineer.  Terri contributed to major national security programs, developing real-time seismic monitoring software and leading the Nuclear Waste Management Group, where she helped create computer models for the Yucca Mountain Project.

Terri’s management journey at LLNL started in 1992 when she became leader of the Computer Systems and Applications Division, supporting multiple directorates and expanding her roles over the next three decades.  She served as Principal Deputy Department Head for Integrated Computing and Communications, Department Head for Computing, and Deputy Associate Director for High Performance Computing.  As LC’s program leader, Terri played a pivotal role in fielding top-tier HPC systems, including El Capitan, Sequoia, BlueGene, and Sierra, and led the Hardware and Integration area of the DOE Exascale Computing Project.

Throughout her career, Terri has been recognized for her expertise, mentorship, integrity, and leadership, earning numerous awards and international recognition, such as the 2023 HPC People to Watch designation.  She championed open-source software initiatives and fostered strong partnerships across academia, industry, and the national labs.  Terri’s legacy includes mentoring future HPC leaders and driving impactful projects that support the national interest, leaving an enduring mark on LLNL, the NNSA ASC program, and the broader computing community (LLNL-ABS-2013678).

---

Welcome Aboard...

LANL ASC program

Angela Herring is the acting Verification and Validation (V&V) program manager at LANL.  Angela’s career at LANL has been broad—she has been on staff since 2006 and has served as a project leader, deputy project leader, and technical contributor for a range of efforts within the Global Security and ASC programs.  She has served as the Group Leader for XCP-4 since 2020, where she leads efforts in workflow strategy, code linking, and the modernization of simulation setup tools for multiphysics codes.  Angela earned her BS in Aerospace Engineering from Mississippi State University and an MS in Mechanical and Aeronautical Engineering from the University of California, Davis.  In her free time, she enjoys scuba diving, swimming, and snowboarding.

Ben Santos is the acting Platforms program manager at LANL.  Ben joined LANL as an undergraduate student intern in 2005.  After completing his MS in Computer Science, he worked as a software engineer for FIM Photobucket.  Ben rejoined LANL and the HPC team as a staff member in 2009, joining the HPC-ENV consulting group.  He became the team leader in 2011, leading efforts in workload management, HPC scheduling, account processing, remote computing enablement (RCE), and user consulting.  Since 2022, Ben has been serving as the HPC-ENV group leader and earlier this year he completed a rotation as acting HPC Deputy Division Leader.  In his free time, he enjoys spending time outdoors and with his family.

SNL ASC program

Lori Basilio is serving as the new ASC Deputy Executive for SNL, replacing Justin Newcomer.  Lori holds a PhD in Electrical Engineering from the University of Houston and has a proven record of delivering exceptional results in complex technical environments.  In addition to her professional pursuits, Lori enjoys rock climbing and various outdoor activities, as well as reading and listening to podcasts on topics ranging from leadership to national security.

Shaunak Shende’s primary focus within the ASC program has been as a developer for CTH (Shock Code) where he is enhancing the code's ability to leverage GPU architectures across multiple platforms, as well as developing a coupling between a rigid body solver and SABLE (Shock Code).  Shaunak received his PhD in Computational Mechanics from Brown University in May 2025, concentrating on new computational methods for extreme events.  Outside of work, he enjoys spending time with his three cats, as well as traveling, cooking, and running.

LLNL ASC program

Britton J. Olson has been selected as the new Lead for Turbulence Modeling and Simulation in the Computational Physics group, effective May 16, 2025.  Britton brings over 15 years of experience in computational physics, specializing in turbulence modeling, direct numerical simulation, and HPC at LLNL.  He led one of the largest high-order turbulence simulations ever performed on the El Capitan system, demonstrating technical leadership and advancing HPC capabilities.  Britton manages large-scale simulation projects, mentors PhD students in a DOE fellowship program, and teaches computational physics.  He holds a PhD in Aeronautics and Astronautics from Stanford University and a BS in Mechanical Engineering from Brigham Young University.  Outside of work, Britton enjoys woodworking, hiking, and spending time with his family. 

---

NNSA LDRD/SDRD Quarterly Highlights

LLNL LDRD: Team develops new material that bends, bounces, and absorbs energy on demand.

Scientists at LLNL and their collaborators have created a new class of programmable soft materials that can absorb impacts like never before, while also changing shape when heated.  The research — which includes collaborators from Harvard University, the California Institute of Technology (Caltech), SNL, and Oregon State University — opens the door to smarter, lighter and more resilient materials that respond to the world around them.  Read more in the LLNL news highlight.

 

---

LANL LDRD: New approach detects adversarial attacks in multimodal AI systems.

New vulnerabilities have emerged with the rapid advancement and adoption of multimodal foundational AI models, significantly expanding the potential for cybersecurity attacks.  Researchers at LANL have put forward a novel framework that identifies adversarial threats to foundation models — AI approaches that seamlessly integrate and process text and image data.  This work empowers system developers and security experts to better understand model vulnerabilities and reinforce resilience against ever more sophisticated attacks.  Read more in the LANL news highlight.

 

---

Nevada National Security Site (NNSS) Hosts LDRD/SDRD Biannual Working Group.

Twice per year, members of LDRD programs at the national labs and the NNSS’ SDRD team meet in person to discuss national security research innovation and standards as part of their regular working group.  These meetings promote collaboration and alignment with NNSA R&D efforts between the NNSS’ SDRD program and the LDRD programs at LANL, LLNL, and SNL.  On April 2-3, 2025 the NNSS had the opportunity to host this working group at both the Site and in North Las Vegas.  

These discussions ensure that SDRD projects are on par with the larger national laboratories by establishing appropriate standards of R&D to facilitate important national security research.  By participating in this and other biannual working groups, the NNSS can leverage the SDRD program as an important piece of the larger national security puzzle and enact a productive future in research alongside its laboratory partners.  Read more in the NNSS news highlight.

 

---

Questions? Comments? Contact Us.

ASC Assistant Deputy Administrator: stephen.rinehart [at] nnsa.doe.gov (Dr. Stephen Rinehart)

ASC Deputy Assistant Deputy Administrator: thuc.hoang [at] nnsa.doe.gov (Thuc Hoang)

Program Director for Computing: simon.hammond [at] nnsa.doe.gov (Dr. Si Hammond)

Program Director for Simulation: anthony.lewis [at] nnsa.doe.gov (Anthony Lewis)

• Integrated Codes: james.peltz [at] nnsa.doe.gov (Dr. Jim Peltz)
• Physics and Engineering Models: robert.spencer [at] nnsa.doe.gov (Robert Spencer)
• Verification and Validation/PSAAP/CSGF: david.etim [at] nnsa.doe.gov (David Etim)
• Capabilities for Nuclear Intelligence: anthony.lewis [at] nnsa.doe.gov (Anthony Lewis)
• Computational Systems and Software Environment: simon.hammond [at] nnsa.doe.gov (Dr. Si Hammond), sara.campbell [at] nnsa.doe.gov (Sara Campbell), cheri.hautala-bateman [at] nnsa.doe.gov (Dr. Cheri Hautala-Bateman)
• Facility Operations and User Support: michael.lang [at] nnsa.doe.gov (K. Mike Lang)
• LDRD/SDRD: anthony.lewis [at] nnsa.doe.gov (Anthony Lewis)