Expanding the boundaries of scientific knowledge and advancing the technological state of the art to solve problems of national and global importance
Through its science and technology capabilities, Livermore makes fundamental discoveries about nature, develops innovative technologies that improve life and drive the economy, and carries out its mission to improve national security.
Harnessing the Power of AI
With its S&T leadership responsibilities, exascale supercomputers, and expertise in high-performance computing (HPC), the DOE laboratories are focused on developing safe and trustworthy AI systems for scientific discovery, energy research, and national security. A White House National Security Memorandum on Artificial Intelligence in October 2024 reinforced the special responsibility of NNSA and its laboratories to harness AI for nuclear and national security missions. LLNL is providing leadership in these efforts. In May 2024, DOE and LLNL AI-powered research efforts and successes were presented center stage at the Special Competitive Studies Project (SCSP) AI Expo for National Competitiveness.
LLNL is exploring use of generative AI and advanced machine learning (ML) to make transformative breakthroughs for mission-critical applications. These efforts leverage the Laboratory’s world-leading high-performance computing (HPC) systems, cutting-edge modeling and simulation capabilities, large data sets, critical science and technology drivers, and multidisciplinary expertise. The overarching goals of expedited scientific discovery and rapid design-to-product delivery to meet mission needs have resulted in key advances, which are described throughout the Annual Report (see National Ignition Facility, Science and Technology, and Partnerships): achievement of ignition and progress in fusion research; rapid design of antibodies for vaccines and cancer-fighting pharmaceuticals; outcome-optimized product design; and robotic-driven experimentation. Many of the Laboratory’s efforts in AI are supported through its AI Innovation Incubator—part of LLNL’s Data Science Institute—which aims to advance AI for applied science at scale through industry collaboration, leadership, and strategic mission-driven investments.
Harnessing the power of AI in the national interest while safeguarding its misuse is of the utmost importance. In April 2024, the University of California and LLNL jointly hosted the AI Safety Workshop where industry, academic, and national laboratory experts explored the complexities surrounding AI, including policy issues and the need to build more secure and reliable models. LLNL is pursuing research to ensure that AI solutions are both safe and trustworthy. For example, the Laboratory participated in a study with external partners to examine 16 mainstream large language models (LLMs) across eight dimensions of trustworthiness, using 30 public data sets as benchmarks on a range of simple to complex tasks, to create a comprehensive LLM trustworthiness evaluation framework.
Progress in Quantum Computing
Quantum technologies offer new paradigms for computing, direct simulation of complex quantum mechanics, and unprecedented precision in sensing for detection and measurement. The Livermore Center for Quantum Science offers a state-of-the-art facility that connects researchers, capabilities, and development efforts, both internally and with external partners, to further quantum innovation and streamline collaboration. Recently, researchers installed a microchip containing 21 coupled superconducting qubits (basic units of information) inside the center’s Quantum Design and Integration Testbed. The microchip’s implementation was a significant step toward developing larger qubit arrays and longer, more complex algorithms needed to execute quantum operations. LLNL has also made breakthroughs in reducing quantum devices’ susceptibility to noise and improving the fabrication of precision components and hardware to achieve better performance. Using emerging topological materials, researchers successfully developed miniaturized circulators—devices used to protect qubits from noise sources—that are 1,000 times smaller than previously achieved.
Detecting Rare Events
Rare particle detection is at the heart of collaborative research efforts to better understand the fundamentals of the universe and improve nuclear security. Several innovative activities are underway. The LUX-ZEPLIN experiment uses the world’s most sensitive dark matter detector. It is located deep underground to shield it from cosmic rays. New experimental results have determined more precise constraints for detecting weakly interacting massive particles (WIMPS)—a leading dark matter candidate—finding no evidence of WIMPs above a mass of 9 gigaelectronvolts/c2. The Precision Oscillation and Spectrum Experiment (PROSPECT) revealed that higher energy neutrinos from both high- and low-enriched uranium reactors are more common than models predict, providing insight into a persistent difference that scientists have encountered between predictions and measurements of reactor neutrinos. In addition, a Livermore team leads the Intense Compact Muon Sources for Science and Security project for the Defense Advanced Research Projects Agency. Muons can penetrate far deeper than is possible with x-rays—through the equivalent of a 30-meter-thick concrete wall. The initiative aims to use a compact high-power laser to rapidly generate muons, significantly greater than the natural flux rate, and make imaging much faster. The first phase of the four-year program will focus on proof-of-principle experiments that clearly demonstrate laser-produced muons.
Advances in Additive Manufacturing
In FY 2024, Laboratory researchers reported in the Journal of Applied Physics great strides in accelerating the design and manufacture of systems whose performance depends on complex physics processes and intricate engineering details. In a three-year research project named DarkStar, the team pursued a new AI-aided approach, called inverse design, that focuses on the discovery of optimized design solutions according to the desired outcome and enables the study of time-dependent physics performance as a function of design parameters. First, the complex system is characterized by 10 or so design variables. Then a very large number of designs are sampled within the design-variable 10D space and the system performance of each is calculated with high-resolution 2D or 3D physics simulations. The AI/ML tool learns design sensitivities as it “fills in” the space, locating areas that merit search with more simulations and homing in on areas of optimal design consistent with all constraints. Visualization tools enable the quick study of the time evolution of options interpolated within the space of possible designs.
DarkStar researchers developed the tool by examining design modifications to linear shaped charges, devices that have wide-ranging military and civilian applications. Detonation forms a high velocity jet out of material at the face of the high explosive (HE). The researchers focused on finding unimagined redesigns that would control the nonlinear growth of Richtmyer–Meshkov instabilities (RMI), a fluid-dynamics phenomena that arises as the jet starts to form. They succeeded. In a several-week design cycle, the ML tool would discover an intriguing design possibility, and additive manufacturing (AM) would enable rapid manufacture of the devise for testing. The team conducted 14 HE detonation experiments at Livermore’s High Explosive Applications Facility, comparing different methods that mitigated RMI growth and gaining greater insight into the phenomena.
LLNL also continues to make significant advances in AM, improving performance and enabling new applications. A Laboratory research team is developing a new process called Microwave Volumetric Additive Manufacturing (MVAM) that could revolutionize 3D printing. VAM techniques allow for rapid printing of complex 3D shapes in a single operation and eliminate the need for support structures; however, lasers are only effective at curing specific materials, primarily transparent and low-absorbing resins. Microwaves reach deeper into materials, are effective in curing a wide variety of resins, and can create complex, functional, and potentially meter-size parts. The researchers validated the MVAM method using a proof-of-concept experimental system. Other advances in AM at the Laboratory in FY 2024 include development of direct ink writing resins specially developed to provide electrostatic discharge protection (and cushioning) to sensitive electronic components and the printing of a novel soft material that can change shape in response to light and be used in “soft machines.”
Jupiter Laser Facility Reopens
Fifty years ago, Janus was installed in Building 174 (renamed the Jupiter Laser Facility in 2006) and produced the first neutron-yielding laser fusion reactions at the Laboratory. Additional lasers and use by more than 100 Ph.D. students and thousands of international researchers later, the Jupiter Laser Facility (JLF) celebrated its grand reopening in May 2024 after a four-year refurbishment of aging equipment. As the fifth-highest energy research laser in the United States, JLF is dedicated to high-energy-density science exploration and is one of the founding members of DOE’s LaserNetUS, a network of high-intensity laser systems across North America. JLF consists of three operating laser systems and target areas—Janus, Titan, and COMET—and a main laser bay. Every year, the facility welcomes international collaborators, academia, the private sector, and other national laboratories to conduct research.
Insight into Planetary Cores
LLNL scientists are investigating how iron—the primary element in Earth’s core— behaves under extreme conditions to better understand planetary geodynamic processes and more accurately simulate the evolution of planets. One principal effort in FY 2024 was the development of new equation of state (EOS) models, specifically designed to address the high pressures and temperatures that are present in the Earth’s core and those of super-Earths, which harbor more extreme conditions. The team carefully analyzed the extensive existing literature and executed first-principles quantum simulations for conditions that are relatively devoid of data. Importantly, based on their analyses, the researchers produced not just one EOS, but a family of models to address uncertainties. The resulting EOS models cover a broad range of conditions and applications, from simulating Earth and Earthlike planets to constructing models for other iron-rich materials of industrial relevance.
Experimental activities include one LLNL team’s use of the Dynamic Compression Sector beamline at Argonne National Laboratory to apply a nanosecond laser shock to iron. They found that when iron is subjected to high pressures and temperatures, it undergoes a phase transition from its usual body-centered cubic structure to a hexagonal close-packed crystal structure—and as pressure increases, the hexagonal structure eventually melts into a liquid. Another team used the Omega-EP laser at the University of Rochester’s Laboratory for Laser Energetics to compress magnesium oxide (MgO), a crucial component of Earth’s lower mantle, to ultra-high pressures. They characterized a phase transition that significantly alters MgO’s properties in a way that can drastically affect a super-Earth’s internal dynamics.
Bioscience Breakthroughs
BridgeBio Oncology Therapeutics has begun clinical trials of a new cancer drug developed through LLNL’s collaboration with BridgeBio and the National Cancer Institute at the Frederick National Laboratory for Cancer Research. The drug breakthrough—called BBO-8520—was found through a combination of HPC and an LLNL-developed platform that integrates AI and traditional physics-based drug discovery. BBO-8520 has shown promise in laboratory testing for inhibiting mutations of KRAS proteins—targets long considered “undruggable” by cancer researchers. KRAS mutations are linked to about 30 percent of cancers. Cleared by the U.S. Food and Drug Administration for human trials in December 2023, BBO-8520 reached this stage of development in record time, demonstrating the potential of HPC and AI-assisted design to save time and expense on new drug discovery.
Through a combination of computational modeling and in vivo efficacy screening, Livermore researchers have also discovered a promising new treatment to counteract the effects of fentanyl and related opioids. Called subetadex, the compound has a longer half-life than naloxone (the main drug used to treat fentanyl overdoses) and acts to encapsulate the opioid, preventing it from binding to receptors in the body. The new treatment is a major step toward neutralizing the toxic effects of fentanyl, which has killed more than 200,000 Americans in the last three years.
Tougher than Diamond
Through quantum-accurate, multi-million atom molecular dynamics simulations on Frontier, DOE’s first exascale system, a Livermore team has uncovered the extreme metastability of diamond at very high pressures, significantly exceeding its range of thermodynamic stability. The eight-atom body-centered cubic (BC8) crystal of carbon is predicted to exist on carbon-rich exoplanets and has potential as a new superhard material (harder than diamond) for scientific applications, if it can be produced on Earth. Key to this work was the team’s implementation of a highly accurate ML interatomic potential on on graphics processing units to accurately simulate the time evolution of billions of atoms at a wide range of high pressure and temperature conditions and experimental time and length scales. Results indicate that BC8 can only be synthesized within a narrow range of pressures and temperatures, providing insight into viable compression pathways for BC8 synthesis in a laboratory setting.
