
Back to top
Accelerating Multiphysics Simulations
Livermore has focused on preparing for the deployment of El Capitan, the National Nuclear Security Administration’s first exascale supercomputer. El Capitan’s heterogeneous computing architecture—running on central processing units (CPUs) and graphics processing units (GPUs)—requires next-generation multiphysics codes to perform efficiently across multiple computing architectures. Livermore researchers have extended the capabilities of the MARBL multiphysics code using GPUs to include additional physics crucial to high-energy-density (HED) physics and fusion modeling. The results are published in the April 2024 issue of Journal of Fluids Engineering.
MARBL’s simulations target HED physics in mission-relevant applications such as stockpile stewardship and inertial confinement fusion experiments. By advancing software abstractions and algorithmic developments to add features to MARBL performance on GPUs, the team verified that additional physics models were implemented accurately and ensured their efficient performance when running on the next generation of GPU-based machines. A new type of preconditioner (a processing technique to improve the convergence of iterative methods for solving linear systems), the RAJA Portability Suite, and the modular finite element discretization library (MFEM) proved instrumental in enabling MARBL to target a variety of GPU–CPU architectures.
Faster computation through GPUs increases the rate of scientific discovery by enabling researchers to run many simulations simultaneously. “The ability to rapidly iterate at full fidelity and high resolution in 3D is crucial for efficient discovery,” says Livermore computational physicist and principal investigator Rob Rieben.
Contact: Rob Rieben (925) 422-3783 (rieben1 [at] llnl.gov (rieben1[at]llnl[dot]gov)).
Explaining Recent Climate Modeling Discrepancies
Differences between climate observations and model simulations can be caused by model errors or natural fluctuations in the climate system. Determining the impact of natural climate variability can validate models and inform future climate projections. A Livermore scientist and collaborators from the University of Washington, the National Oceanic and Atmospheric Administration Pacific Marine Environmental Laboratory, and Pacific Northwest National Laboratory have identified a distinct pattern of surface temperature change associated with natural variability from 1980 to 2022. Their research is published in the June 16, 2024, issue of Geophysical Research Letters.
The team analyzed multidecadal trend patterns from Coupled Model Intercomparison Project 6 (CMIP6) simulations in which natural variability warms the arctic but has an overall, global cooling effect. They found that these simulations produce enhanced warming in the Barents and Kara Seas and cooling in the tropical eastern Pacific and Southern Ocean due to natural variability. The same features are imprinted on observed surface-temperature changes from 1980 to 2022 and help to understand model–observational differences in warming in key regions. While this pattern of temperature change is not seen in the average simulation over the sixth phase of the CMIP models, each of the climate models examined in the study infrequently produce a similar pattern.
“The relative role of different drivers of model–observational discrepancies in the pattern of warming has important implications for our understanding of climate sensitivity, as well as regional changes in climate,” says Stephen Po-Chedley, Livermore scientist and co-author. “This work shows that natural variations in Earth’s climate likely contribute to key differences in the simulated-versus-observed pattern of surface–air temperature changes.”
Contact: Stephen Po-Chedley (925) 422-3421 (pochedley1 [at] llnl.gov (pochedley1[at]llnl[dot]gov)).
Inflated Exoplanet a Result of Tidal Heating
WASP-107b is an exoplanet nearly the size of Jupiter with only a tenth of its mass. As one of the lowest-density planets known, conventional planet formation theories do not fully explain its inflated nature. A Livermore scientist and international collaborators have used transmission spectroscopy via the James Webb Space Telescope to characterize the exoplanet’s atmosphere, revealing a surprisingly small amount of methane and solving the mystery of its internal properties. Their results are published in the June 27, 2024, issue of Nature.
The small methane levels in WASP-107b’s atmosphere indicate that its core must be much more massive and its interior significantly hotter than previously estimated. To have such an inflated nature suggests that the exoplanet must be tidally heated by its eccentric orbit around its host star, in which tidal friction generated by orbital and rotational energy is dissipated as heat into the planet’s interior. In addition to explaining the inflated nature of the planet, the researchers also discovered an array of previously unseen—but expected—molecules in its atmosphere: carbon-, oxygen-, nitrogen-, and sulfur-bearing molecules. The molecules had previously not been detected simultaneously in a transiting exoplanet. “The study solves a long-standing challenge in exoplanet science,” says Livermore’s Peter McGill, a postdoctoral researcher in Astronomy and Astrophysics Analytics and co-author of the paper. “For the first time, we are linking the chemical composition of an exoplanet’s atmosphere to its internal properties.”
Contact: Peter McGill (925) 424-5304 (mcgill5 [at] llnl.gov (mcgill5[at]llnl[dot]gov)).