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

Welcome to the third 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 a well-deserved congratulations to Thuc Hoang and the ASC HQ team for receiving the Samuel J. Heyman Service to America Medal, in recognition of their role in developing the El Capitan supercomputer.  Other featured highlights in this edition include: 

• Lawrence Livermore National Laboratory’s (LLNL’s) use of El Capitan for rapid turnaround of simulations supporting the Annual Assessment Review certification.
• Selection of the Predictive Science Academic Alliance Program IV (PSAAP IV) Centers.
• Sandia National Laboratories (SNL) ASC and W93/Mk7 team partnership reduced system-level modeling time from six months down to six weeks.
• Major release 8, minor release 1 of the Evaluated Nuclear Data File (ENDF/B-VIII.1) with LANL contributions.
• Use of Los Alamos National Laboratory’s (LANL’s) advanced simulation tools to enhance prompt diagnostics at the Nation Ignition Facility (NIF).

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

---

Thuc Hoang and the ASC HQ team honored with the Samuel J. Heyman Service to America Medal. 

Thuc Hoang and the Office of Defense Programs’ ASC team received the Samuel J. Heyman Service to America Medal, in recognition of their role in developing the El Capitan supercomputer.  The “Sammies” award is among the most prestigious honors a federal civil servant can be given for contributions to the United States.  Acting NNSA Administrator Teresa Robbins noted in the June 20th NNSACAST, “Thuc and her team exemplify the qualities that the award represents: professional excellence, visionary leadership, and selflessness in service of the American people.”  
The launch of El Capitan at LLNL was the product of a multiyear collaboration between NNSA, the NNSA laboratories, and industry leaders.  As the world’s fastest and most powerful supercomputer and the first exascale system dedicated to national security, El Capitan enables unprecedented feats of computing in support of the nuclear stockpile.  As one of the crown jewels of the Stockpile Stewardship Program, it serves as a tangible example of the vast technological edge the U.S. maintains over adversaries in the realm of nuclear security.  NNSA’s strong partnership with LLNL, SNL, and LANL made this achievement possible.

Thuc and the ASC team forged new partnerships with industry to enable a revolutionary design of the chipset needed to make El Capitan work.  Serving as the project manager, organizer, and leader of this complex undertaking, Thuc combined her mastery of technical minutiae with a strategic understanding of the enterprise’s computing requirements.  In addition to honoring Thuc, the following ASC team members contributed immeasurably to this achievement: Simon (Si) Hammond, K. Michael Lang, James Peltz, David Etim, and Anthony Lewis.

---

LLNL Integrated Codes report early successes of El Capitan application in the Annual Assessment Review certification.

The LLNL design community has leveraged El Capitan to perform critical simulations supporting the Annual Assessment Review (AAR).  Since El Capitan’s transition to the closed side at the end of March, Integrated Codes (IC) teams have collaborated closely with Livermore Computing and the design community to enable relevant simulations on the system.  This capability has allowed for the completion of 3D ensembles at medium resolution, as well as multiple highly-resolved 3D simulations.  The rapid turnaround of these simulations has provided timely insights for the AAR team’s findings (LLNL-ABS-2010603).

---

Predictive Science Academic Alliance Program IV (PSAAP IV) Centers Selected

Nine PSAAP IV Centers were selected to support research in predictive science enabled by exascale computing technologies, breakthroughs in verification and validation and uncertainty quantification (V&V/UQ), and data-driven methodologies (including machine learning and data science) to expand the frontiers of large-scale simulation science.  PSAAP IV will also foster the development of the next generation of computational scientists and engineers.  Centers were awarded either as a Predictive Simulation Center (PSC) or a Focused Investigatory Center (FIC) via cooperative agreements as follows:

• University of Florida (PSC)
• University of Michigan (PSC)
• Massachusetts Institute of Technology (PSC)
• Oregon State University (PSC)
• University of Virginia (PSC)
• Brown University (FIC)
• University of California, San Diego (FIC)
• Michigan State University (FIC)
• University of New Mexico (FIC)

More information is available in the DOE/NNSA press release and on the official PSAAP website

---

SNL demonstrates capability to deliver system-level computational models in record time.

ASC teams have partnered with the W93/Mk7 analyst team to cut system-level modeling time from six months down to six weeks.  ASC seeks to develop reliable simulation capabilities for National Defense partners, with the W93/Mk7 program marking a significant advancement in life extension efforts.  In a recent sprint for the W93/Mk7 Reference Design 5 Drop (RD5), the team built and tested a system-level model in record time.  Updates from RD5 were quickly integrated into a previous model using advanced AI/ML tools and discretization techniques, achieving simulation times that were 35 times faster on the new Cascade graphics processing unit (GPU) platform.  This effort demonstrates ModSim's potential to enhance design processes and facilitate effective tradeoff studies.  Key outcomes include rapid development of a structural dynamics model for RD5, improvements for ASC tools, and increased credibility of artificial intelligence/machine learning (AI/ML) within the analyst community (SAND2025-07972M).

---

Newly released nuclear data (with LANL contributions) gives users access to experimental discoveries.

The National Nuclear Data Center formally released a new nuclear data library in August 2024, designated ENDF/B-VIII.1 (major release 8, minor release 1 of the Evaluated Nuclear Data File).  This data is the culmination of seven years of new evaluation work.  LANL was an indisputable leader as part of an international collaborative effort between many partnering institutions, including highly influential measurements of Pu-239 at LANSCE (Chi/Nu experiment measuring the prompt fission neutron spectrum, DANCE experiment measuring neutron capture), a multitude of new evaluations (Pu-239, Ta-181, Pt-all, and Li-6), and a careful validation process.  The ASC program was the most important sponsor for LANL-based evaluation, processing, and validation efforts contributing to the delivery of this library.

Following numerous improvements to the NJOY nuclear data–processing code, the ENDF/B-VIII.1 library was processed into continuous energy and multigroup data for use in neutron transport codes such as MCNP and PARTISN.  It is available to LANL users on high-performance computing (HPC) platforms through the MT81 and the Lib81 application libraries.  ENDF/B-VIII.1 also features the richest and best-tested uncertainty information to date, enabling uncertainty-aware workflows such as forward propagation and nuclear data adjustment.  This library continues a trend of improving performance on validation benchmark suites, one example of which is shown in Figure 3 (LA-UR-25-26176).

---

LANL’s advanced simulation tools enhance prompt diagnostics at the NIF.

The NIF is a crucial part of the U.S. Stockpile Stewardship Program.  To ensure the safety, security, and effectiveness of the nuclear stockpile, experiments at NIF provide information regarding the behavior of materials at high temperature and pressure.  Prompt diagnostics, such as neutron pinhole radiography that is fielded on NIF experiments, provide measures of the performance and outputs of imploding capsules.

To better understand the resulting signals from these diagnostics, a forward-modeling capability was developed by LANL using the Eulerian Application Project’s xRAGE code and the Radiation Transport Project’s Monte Carlo Applications Tool Kit (MCATK).  This forward-modeling capability relies on a linking process through which xRAGE communicates hydrodynamic state and the number of thermonuclear reaction sources coming from a NIF experiment.  xRAGE generates a set of specialized link files that can be post-processed using MCATK to generate simulated prompt diagnostic signals (Figure 4).

These forward-modeled prompt diagnostic signals can be compared with experimental signals to improve understanding of the underlying physics and to validate simulation models.  As this linking capability and the prompt-diagnostics modeling capability of MCATK improve, this feature could also help modelers and experimentalists work together to improve the diagnostics fielded on NIF and other experimental platforms.

(LA-UR-25-26176)

---

SNL enhances SGEMP analysis through user-centric tool development.

User experience (UX) research was conducted in FY25 Q2 to identify the needs of users, addressing the cable system-generated electromagnetic pulse (SGEMP) use case.  The Accelerated Model Development (AMD) initiative, in partnership with the Next Generation Simulation (NGS) team, delivered a minimum viable product to the RAMSES user community enabling numerous interrogations of a model to extract 1D stack-ups for analysis.  The deployed tool helps analysts identify hotspots and sensitivities and aims to inform design decisions in days (not months).  Usability testing prior to the release of the tool validated the tool’s value proposition and identified needed features for production use.  Development and delivery of the tool will be ongoing into FY26, working toward delivering a capability that supports the end-to-end workflow and eliminates the need for analysts to write workflow scripts.  User experience studies will continue in FY25 Q4 to help inform development prioritization (SAND2025-09696M).

---

LANL is improving simulation performance with optimized implicit Monte Carlo decompositions.

 

Implicit Monte Carlo (IMC) thermal radiation transport (TRT) is often the dominant computational cost in large, complex multiphysics simulations of high-energy density physics (HEDP) and inertial confinement fusion (ICF).  Recent improvements to optimize the parallel decomposition dynamically have resulted in significant runtime performance improvements—as high as 4x for some problems.

In a multiphysics code, an established parallel decomposition by a “host” program may not be ideal for all individual algorithms.  In particular, IMC methods enable dynamic phase-space resolution, allowing users to focus computational effort on regions of interest.  This dynamic resolution, while powerful, also creates highly load-imbalanced simulations.  Additionally, the simulations evolve dynamically in time, and the optimal parallel IMC decomposition often changes substantially.  This behavior leads to underutilization of computational resources, which limits overall performance.

To address this long-standing issue, the Jayenne IMC TRT team at LANL recently released a fully automated mesh redecomposition feature for particle load-balancing.  The new algorithm allows the Jayenne package to generate and store its own mesh decomposition.  Jayenne now can create a near-optimal work distribution without negatively affecting the performance of other coupled physics packages.  To do this, Jayenne stores the number of particle path segments per cell from the previous time step.  This information is passed to the ParMETIS graph partitioning library to determine a spatial mesh decomposition (Figure 6).

A major challenge of this work involves balancing the cost of remapping cell quantities and particles between decompositions with the improved computational performance of the IMC cycle.  Jayenne uses efficiency metrics to ensure that a new mesh decomposition is created only when particle load-balance outweighs remapping costs.  Jayenne does not remap portions of the problem with no particles to track, which further reduces the cost of this scheme (Figure 7).  This new approach to optimize parallel decomposition dynamically improves the computational efficiency and throughput of HEDP and ICF users’ daily workflows.  Redecomposition achieves faster execution time and more balanced memory needs for IMC.  It thus enables larger and higher-fidelity simulations previously untenable due to suboptimal IMC decomposition (LA-UR-25-26176).

---

SNL AI agent automates model defeaturing tasks in seconds.

Simplifying complex features in a computational model to make simulations run more smoothly, a process called defeaturing, can be a very time-consuming part of developing simulation models. Researchers at SNL, part of the Multi-Agent Design Assistant (MADA) Geometry Agent team, have leveraged an AI agent to automate defeaturing tasks using simple language commands resulting in meshable models.  For example, users can instruct the AI agent with natural language to “remove all holes and fillets with a radius less than 0.4 cm,” and the agent completes the task in under a second.  This innovation cuts down the time needed for defeaturing from hours (or even days) to just seconds, greatly speeding up the entire digital design process (SAND2025-09696M).

---

Canary: SNL's common software testing tool improves integrated code developer productivity.

Canary is an open-source tool developed at SNL that efficiently executes software test suites, specifically targeting scientific applications running on HPC systems.  It employs a plug-in architecture to facilitate widespread adoption without necessitating changes to existing tests.  Historically, the use of many software testing tools leads to duplication of maintenance and feature development efforts across multiple organizations.  As the number and diversity of HPC platforms and architectures continue to grow, these duplicated efforts have become increasingly problematic.  Throughout FY25, several projects at SNL have adopted Canary as their testing tool, which has reduced duplication, improved test turnaround times, and promoted transferable developer skills through shared tools and processes.  The open-source publication of Canary on GitHub has also attracted interest from the broader research community and from projects at other nuclear security enterprise sites (SAND2025-09696M).

---

LANL’s Crosslink integration enhances 3D modeling capabilities in NNSA simulations.

The NNSA relies on advanced computational modeling and simulation capabilities to maintain the safety, security, and effectiveness of the nuclear deterrent without nuclear testing.  These capabilities are critical for understanding complex physical phenomena, assessing the performance of nuclear weapons systems, and supporting stockpile stewardship activities.  The Common Modeling Framework (CMF) at LANL plays a crucial role in this mission by providing computational infrastructure that enables modelers to set up, run, and analyze simulations; compare results to experimental data; and archive workflows for future studies.  CMF’s ability to accurately model complex 3D geometries and effects is particularly important for reproducing real-world scenarios and enhancing the fidelity of nuclear-weapons simulations.

However, meeting the national security need for advanced 3D-modeling capabilities presents significant challenges. CMF’s most widely used geometry and meshing tool, Ingen, was primarily designed for 2D simulations, and requires 2D-to-3D transformations and additional scripting to build 3D simulations.  This approach limits the efficiency and accuracy of 3D modeling efforts, potentially impacting the quality and reliability of some 3D simulation results.  Furthermore, the increasing complexity of modern weapons systems and the demand for higher-fidelity simulations require more intuitive and scalable tools to create and manipulate 3D geometries and meshes.  These challenges necessitate developing new tools and workflows that can handle the complexities of 3D modeling while maintaining the reproducibility and customization capabilities that are essential for rigorous scientific analysis.

To address these challenges and advance the state of the art in 3D modeling for national security applications, LANL's ASC/IC PUMA project has developed Crosslink, a block topology-based meshing tool.  Crosslink provides a more intuitive approach to building block-structured meshes in 3D, offers a graphical user interface for enhanced usability, includes some specific meshing features for LANL multi-physics simulations, and is intended to scale to larger problems on NNSA’s advanced computing platforms.  Crosslink was recently integrated into CMF workflows, a significant step forward in enabling modelers to design simulations directly in 3D space, accelerating the setup of 3D Lagrangian simulations within CMF.  This integration allows users to load pedigreed geometries, build block topologies, and automatically generate high-quality meshes for simulations.  By streamlining the 3D modeling process and improving the efficiency of workflow execution on HPC systems, these advancements contribute to the modernization of verification and validation capabilities, ultimately enhancing confidence in the simulation tools used for critical national security assessments (LA-UR-25-25440).

---

Dakota 6.22 released: Advancements in Dakota accelerate analyses of computational simulations for SNL Nuclear Deterrence (ND) designers.

Dakota is essential for nuclear deterrence and other DOE applications.  The latest version, Dakota 6.22, enhances multifidelity (MF) sampling methods, which significantly improve the speed and accuracy of uncertainty quantification (UQ) calculations.  These advancements make UQ feasible for complex, high-fidelity simulations and simplify verification and validation during early nuclear weapon design phases.  Traditional Monte Carlo methods are often impractical for high-fidelity models, but MF methods can be much faster, enabling effective statistical assessments.  Dakota 6.22 introduces new numerical approaches that boost the performance, stability, and flexibility of MF estimators.  Recent studies show performance improvements of up to two orders of magnitude for UQ analysis, demonstrating the impact of these enhancements (SAND2025-07972M).

---

In the latest TOP500 List of the world’s most powerful computing systems, El Capitan at LLNL reigns supreme across three major supercomputing benchmarks.

In the 65th edition of the TOP500 List, released June 10th at the ISC High Performance conference in Hamburg, Germany, El Capitan reasserted its position as the world’s most powerful supercomputer, repeating its 1.742 exaFLOP performance on the industry-standard High-Performance Linpack (HPL) benchmark.  For the first time, El Capitan also topped the High-Performance Conjugate Gradient (HPCG) benchmark, achieving 17.41 petaFLOPS — a complementary performance metric that reflects the complex, memory-intensive workloads typical in real-world science and engineering applications. Additionally, El Capitan debuted at No. 1 on the HPL-MxP (formerly HPL-AI) benchmark, with a stunning 16.7 exaFLOPS of performance using mixed-precision AI techniques.  Boasting more than 11 million cores and over 44,000 accelerated processing units (APUs), El Capitan also delivers 58.9 gigaFLOPS (GFs)/watt of energy efficiency — earning the 26th spot on the latest GREEN500 List of most energy-efficient systems worldwide and a TOP500 “honorable mention.” For more information, see LLNL's June 16th press release.

---

Welcome Aboard...

LANL ASC program

Juston Moore is the new Aritifical Intelligence for Nuclear Security (AI4NS) Program Manager at LANL.  Juston is a national leader in AI security and brings 10 years of experience in the LANL Global Security program.  In his new role, Juston looks forward to developing trustworthy and reliable systems that leverage frontier AI to enhance the NNSA deterrence mission.  He co-chairs the LANL AI Risk Technical Assessment Group (AIRTAG) and leads an LDRD-DR project on “Efficient and Scalable Guardrail Certification for Science and Security.”  Juston holds a Bachelor of Science degree from New Mexico Tech and a Master of Science degree from the University of Massachusetts Amherst, both in Computer Science.  In his free time Juston enjoys traveling, cooking, kayaking, and hiking with his wife and shepsky.

Kierstyn Paschke is the new LANL Facilities, Operations, and User Support (FOUS) Program Manager.  Kierstyn Paschke earned a Bachelor of Science degree in Computer Science from Montana State University and joined LANL in 2018 as a post-baccalaureate student within the HPC division.  Following her conversion to staff, she gained expertise in HPC, specializing in system administration and file systems.  In 2022, Kierstyn transitioned to program management, initially taking on the role of Institutional Computing Program Manager.  She later assumed the position of Deputy Program Manager for the ASC FOUS program in 2023.  Outside of her professional endeavors, Kierstyn enjoys a range of activities, including running, reading, hockey, and collaborating on projects with her husband.

SNL ASC program

Emily Stein joined the Physics & Engineering Models (P&EM) Subprogram in June.  Previously, she managed Applied Systems Analysis and Research, where she assembled and led technical teams to deliver an integrated simulation and analysis framework for the DOE’s Office of Spent Fuel and Waste Disposition.  Emily earned a Ph.D. in Earth Sciences from the University of California, Santa Cruz.  She lives in Albuquerque, NM with her husband and enjoys spending time outside, traveling, and the occasional sewing, knitting, or baking project.

---

Spotlight at SNL: A little more about Dan Turner…

Dan Turner has been involved with the ASC program since 2009, when he was hired by SNL as a SIERRA/Aria software developer.  Dan served as the Verification & Validation subelement lead for Tools and Software from 2017-2023. 

Dan has spent the last few years as the Artificial Intelligence for Nuclear Deterrence (AI4ND) Initiative lead.  In this capacity, he continues to learn of key AI/ML-related problems within the SNL ND program and has been working to come up with solutions to those problems. 

In his personal life, Dan spends a lot of time coaching youth mountain biking.  The photo (on right) shows Dan with the LaCueva Mountain Biking team.

---

NNSA LDRD/SDRD Quarterly Highlights

LLNL LDRD: Researchers explore the future of responsive 3D-architected materials.

In the evolving fields of materials science and 3D printing, engineers at LLNL are exploring novel ways to create materials and structures that adapt and respond to their environments.  A recent study featured on the cover of Science, conducted in collaboration with the California Institute of Technology (CalTech) and Princeton University, has introduced a revolutionary class of materials known as 3D polycatenated architected materials (PAMs).  These intricate structures can behave with both solid and liquid-like properties and have the potential to impact industries ranging from engineering to medicine.  Read more in the LLNL article available online.

 

---

LANL LDRD: New design aims to make more fuel-efficient and higher performance modular nuclear microreactors.

In 2021, a team of LANL scientists, led by nuclear engineer Topher Matthews, posed these questions to themselves, "What would happen if we developed a reactor with real-world, off-the-shelf materials?  How could we build a prototype reactor in three years?"  Through an LDRD project, Matthews and his team came up with a new reactor design, called ZiaCore, which is based on low-enriched uranium dioxide fuel used in the existing fleet of nuclear reactors that already power about 20% of the U.S. power grid.  The team believes the reactor design will result in a reactor concept that will use less fuel and have higher performance characteristics than current advanced reactor concepts.  A proof-of-principle experiment is scheduled for January 2026 at Nevada National Security Site (NNSS) in the Lab's National Criticality Experiments Research Center, where the Deimos experiment established an advanced reactor testbed.  The materials-driven approach to ZiaCore’s design could change how reactors are designed and operated. Read more in the LANL news highlight.

 

---

LLNL probes deflagration to better understand detonation.

In a new study, researchers at LLNL conducted laser ignition experiments in a diamond anvil cell and employed large-scale quantum molecular dynamics simulations to investigate the products of deflagration at high pressures.  The results could improve models of deflagration and high explosives (HE) overall.  These experiments and models aim to determine the products, or resulting materials, of a deflagration.  The composition of deflagration products, especially the solids, influences the amount of energy and pressure released in the reaction and whether it transitions to a detonation.  Typically, deflagration is studied at relatively low pressures.  But by using laser ignition in a diamond anvil cell, the team was able to acquire data at high pressures that are comparable to the detonation pressure of HE LLM-105.  Read more in the LLNL 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. James 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)