Mitchell Institute

Although light nuclear clusters are known to form abundantly in warm and dilute nuclear matter, their role in hot and dense nuclear matter remains unclear due to the lack of experimental indication for their modifications by the Mott effect under such conditions. To address this issue, we resort to intermediate-energy heavy-ion collisions, where light clusters are mainly produced in the transiently formed hot and dense matter. A kinetic approach, which includes dynamically the formation and dissociation of light clusters, is employed to deduce the strength of the Mott effect and the α -particle fraction in hot and dense nuclear matter from the light-nuclei yields measured by the FOPI Collaboration in central Au + Au collisions at energies of 0.25 A to 0.6 A GeV . We find an unexpectedly abundant α clustering in this environment, which will have profound implications for modeling the nuclear equation of state and describing supernovae and neutron star mergers.

The marginal likelihood or evidence in Bayesian statistics contains an intrinsic penalty for larger model sizes and is a fundamental quantity in Bayesian model comparison. Over the past two decades, there has been steadily increasing activity to understand the nature of this penalty in singular statistical models, building on pioneering work by Sumio Watanabe. Unlike regular models where the Bayesian information criterion (BIC) encapsulates a first-order expansion of the logarithm of the marginal likelihood, parameter counting gets trickier in singular models where a quantity called the real log canonical threshold (RLCT) summarizes the effective model dimensionality. In this article, we offer a probabilistic treatment to recover non-asymptotic versions of established evidence bounds as well as prove a new result based on the Gibbs variational inequality. In particular, we show that mean-field variational inference correctly recovers the RLCT for any singular model in its canonical or normal form. We additionally exhibit sharpness of our bound by analyzing the dynamics of a general purpose coordinate ascent algorithm (CAVI) popularly employed in variational inference.

As part of a National Oceanographic Partnership Program (NOPP) project, seven teams-comprising investigators from universities, federal laboratories, and industry-are collaboratively investigating the generation, propagation, and dissipation of internal waves in the global ocean using complementary, state-of-the-art observations and model simulations. Internal waves, generated by the interaction of tides, winds, and mean flows, permeate the ocean and influence its physical state. Internal waves transport scalar and vector properties-both geographically and across scales-and contribute to irreversible mixing, modulate acoustic propagation, and complicate the identification of subinertial (e.g., geostrophic) flows in observations. For these reasons, accurately representing internal waves in global ocean forecast models is a high priority. The collaborations reported here are improving the understanding of the internal wave life cycle and enhancing model skill in simulating it. Three observational teams are collecting in situ data using 1) redeployable moored arrays that resolve internal waves from multiple directions, 2) global deployments of profiling floats that measure internal wave energy fluxes, shear, and mixing, and 3) high-resolution arrays that focus on bottom boundary layer processes. Four modeling teams are guiding the design and placement of these observation platforms and are using the collected observations to 1) improve internal wave representation and dissipation in ocean models, 2) conduct high-resolution process studies, and 3) implement data assimilation in idealized, regional, and global simulations. These efforts are further supported by high-resolution sea surface height measurements from the new Surface Water and Ocean Topography (SWOT) satellite, which provide context for in situ observations and improve ocean forecasting systems. SIGNIFICANCE STATEMENT: A collaboration among scientists from U.S. universities, national laboratories, and industry is advancing our understanding and prediction of internal waves in the global ocean. These waves-characterized by vertical scales of tens to hundreds of meters and horizontal scales of tens to hundreds of kilometers-play a critical role in maritime commerce, naval operations, and ocean circulation. The team integrates novel observational approaches, including internal wave-resolving moored arrays, ship-of-opportunity float deployments, bottom boundary layer-distributed sensor networks, and satellite wide-swath altimetry, with cutting-edge global, regional, and process-model simulations. Together, these efforts are improving the repre-sentation of internal wave processes in ocean models and enhancing their predictive capabilities for operational forecasts.

Abstract Sea surface height (SSH) reflects a superposition of balanced motions (BM) and unbalanced motions (UBM), which play distinct roles in ocean dynamics and energetics. The Surface Water and Ocean Topography (SWOT) satellite provides two‐dimensional SSH maps at spatial resolutions where BM and UBM coexist, but its 21‐day repeat cycle precludes conventional temporal filtering for signal separation. We present a probabilistic machine learning framework for decomposing instantaneous SSH snapshots into balanced and unbalanced components. To address the spectral bias inherent in standard mean‐squared‐error (MSE) loss functions, we apply zero‐phase component analysis (ZCA) whitening to the target fields, equalizing the contribution of all spatial scales during training. To provide calibrated uncertainty estimates, we model the ZCA‐transformed outputs as independent Gaussians and minimize the negative log‐likelihood as the loss function. Trained on Lagrangian‐filtered LLC4320 simulations in the Agulhas region, our model outperforms standard MSE baselines in both accuracy and training efficiency, achieving threefold faster convergence. Evaluation of the uncertainty estimates shows that ML‐based predictions are often overconfident, and that prediction uncertainty correlates with the relative strength of BM versus UBM: high uncertainty occurs when UBM is weak (decomposition unnecessary). Applied to SWOT observations, the model produces physically plausible decompositions consistent with theoretical expectations for balanced and wave‐dominated flows.

Self-interacting dark matter (SIDM) theories predict that dark matter halos experience core-collapse in late-stage evolution, a process where the halo's inner region rapidly increases in density and decreases in size. This process can be modeled by treating the dark matter as a gravothermal fluid, and solving the fluid equations to predict the density profile evolution. This model is incomplete without calibration to N-body simulations, through a constant factor /3 included in the thermal conductivity for the long-mean-free-path limit. The value of /3 employed in the gravothermal fluid formalism has varied between studies, with no clear universal value in the literature. In this work, we use the N-body code Arepo to conduct a series of isolated core-collapse simulations across a range of scattering cross sections, halo concentrations, and halo masses to calibrate the heat transfer parameter /3. We find that /3 is independent of cross section, halo concentration, and halo mass for velocity independent elastic scattering cross sections. We present a model for an effective /3 as a function of a dimensionless cross section, to describe halo evolution in the long mean free path limit, and show that it accurately captures halo evolution as long as the cross section is not too large. This effective model facilitates comparisons between simulations and the gravothermal model and enables fast predictions of the dark matter density profile at any given time without running N-body simulations.