arXiv subjects

Riccardo Betti

Publications and source records attributed to Riccardo Betti.

8 recordsLinked to original sources

arXiv2026

Decomposition-Guided Diffusion Language Models for Inertial Confinement Fusion Prediction

Inertial confinement fusion (ICF) is a leading pathway toward clean energy, but each shot at the National Ignition Facility costs on the order of one million dollars, making accurate AI surrogates a high-value target. We study exogenous-driven ICF waveform prediction, where a 512-step neutron-rate diagnostic must be inferred directly from a laser pulse and target design parameters, with no historical response observed. The regime stresses standard time-series predictors with temporal sparsity (picosecond peak in a nanosecond window), input-output scale mismatch (under 300 real shots), and peak sensitivity (picosecond timing). We propose ICF-DLM, to our knowledge the first LM-based ICF predictor, combining (i) a physics-typed decomposition into yield $Y_{DT}$, peak timing $t_{\mathrm{peak}}$, and local waveform $w_{\mathrm{local}}$; (ii) bidirectional denoising that defers commitment to peak location; and (iii) a physics-driven PPO reward re-injecting metric structure across numeric tokens. On ICFBench (50K simulations + 232 experimental shots), ICF-DLM cuts peak-timing error from 11.6 to 9.2 steps over a matched autoregressive LLaMA-3-8B and outperforms classical sequence models and LLM-based time-series predictors. Beyond ICF, the recipe shows potential to address science domains with low data and sparse events.

cs.LG↗

arXiv2026

Human-in-the-Loop Meta Bayesian Optimization for Fusion Energy and Scientific Applications

Inertial Confinement Fusion (ICF) holds transformative promise for sustainable, near-limitless clean energy, yet remains constrained by prohibitively high costs and limited experimental opportunities. This paper presents Human-in-the-Loop Meta Bayesian Optimization (HL-MBO), a framework that integrates expert knowledge with few-shot, uncertainty-aware machine learning to accelerate discovery in data-scarce, high-stakes scientific domains. HL-MBO introduces a meta-learned surrogate model with an expert-informed acquisition function to recommend candidate experiments. To foster trust and enable informed decisions, HL-MBO also provides interpretable explanations of its suggestions. We show HL-MBO outperforms current BO methods on ICF energy yield optimization, as well as benchmarks in molecular optimization and critical temperature maximization for superconducting materials.

cs.LG↗

arXiv2026

BayMOTH: Bayesian optiMizatiOn with meTa-lookahead -- a simple approacH

Bayesian optimization (BO) has for sequential optimization of expensive black-box functions demonstrated practicality and effectiveness in many real-world settings. Meta-Bayesian optimization (meta-BO) focuses on improving the sample efficiency of BO by making use of information from related tasks. Although meta-BO is sample-efficient when task structure transfers, poor alignment between meta-training and test tasks can cause suboptimal queries to be suggested during online optimization. To this end, we propose a simple meta-BO algorithm that utilizes related-task information when determined useful, falling back to lookahead otherwise, within a unified framework. We demonstrate competitiveness of our method with existing approaches on function optimization tasks, while retaining strong performance in low task-relatedness regimes where test tasks share limited structure with the meta-training set.

cs.LG↗

arXiv2025

Effects of External Magnetic Fields on the Multi-mode Rayleigh-Taylor Instability

The magneto-Rayleigh-Taylor instability (mRTI) is a key process in inertial confinement fusion and is thought to be widespread in the interstellar medium, where it can concentrate plasma into discrete structures. We present resistive MHD simulations of the nonlinear evolution of multi-mode mRTI in both two and three dimensions, examining the effects of uniform external magnetic fields oriented either parallel or perpendicular to the initial interface. In both 2-D and 3-D, weak parallel fields enhance mixing-zone growth, whereas stronger fields suppress it. For perpendicular fields, growth is initially inhibited but becomes enhanced at later times. These behaviors arise from magnetic tension, which modifies flow anisotropy, buoyancy, drag, and vortex dynamics. The interplay of these mechanisms governs the distinct ways in which magnetic fields influence mRTI evolution.

physics.flu-dyn↗

arXiv2025

Generation of Strong Fields with Subcritical Density Plasmas to Study the Phase Transitions of Magnetized Warm Dense Matter

Warm dense matter (WDM) is a regime where Fermi degenerate electrons play an important role in the macroscopic properties of a material. Recent experiments have brought us closer to understanding unmagnetized processes in WDM, but magnetized WDM remains unexplored because kilotesla magnetic fields are required. Although there are examples of field compression generating such fields by imploding pre-magnetized targets, these existing methods give no independent control over the parameters of the magnetized plasma and result in limited laser access for sample creation and diagnosis. In this paper, numerical simulations show that kilotesla magnetic fields can be obtained by shining laser beams onto the inner surface of a cylindrical target, rather than on the outer surface. This approach relies on field compression by a low density high-temperature plasma, rather than a high-density, low-temperature plasma, used in the more conventional approach. With this novel configuration, the region of peak magnetic field is mostly free of plasma, hence other beams can reach a sample placed in the region of the peak field to form WDM and diagnose it.

physics.plasm-ph↗

arXiv2024

Can Kans (re)discover predictive models for Direct-Drive Laser Fusion?

The domain of laser fusion presents a unique and challenging predictive modeling application landscape for machine learning methods due to high problem complexity and limited training data. Data-driven approaches utilizing prescribed functional forms, inductive biases and physics-informed learning (PIL) schemes have been successful in the past for achieving desired generalization ability and model interpretation that aligns with physics expectations. In complex multi-physics application domains, however, it is not always obvious how architectural biases or discriminative penalties can be formulated. In this work, focusing on nuclear fusion energy using high powered lasers, we present the use of Kolmogorov-Arnold Networks (KANs) as an alternative to PIL for developing a new type of data-driven predictive model which is able to achieve high prediction accuracy and physics interpretability. A KAN based model, a MLP with PIL, and a baseline MLP model are compared in generalization ability and interpretation with a domain expert-derived symbolic regression model. Through empirical studies in this high physics complexity domain, we show that KANs can potentially provide benefits when developing predictive models for data-starved physics applications.

cs.LG↗

arXiv2021

Scale interactions and anisotropy in Rayleigh-Taylor turbulence

We study energy scale-transfer in Rayleigh-Taylor (RT) flows by coarse-graining in physical space without Fourier transforms, allowing scale analysis along vertical direction. Two processes are responsible for kinetic energy flux across scales: baropycnal work $Λ$, due to large-scale pressure gradients acting on small-scales of density and velocity, and deformation work $Π$, due to multi-scale velocity. Our coarse-graining analysis shows how these fluxes exhibit self-similar evolution that is quadratic-in-time, similar to RT mixing layer. We find that $Λ$ is a conduit for potential energy, transferring energy non-locally from the largest scales to smaller scales in the inertial range where $Π$ takes over. In 3D, $Π$ continues a persistent cascade to smaller scales, whereas in 2D $Π$ re-channels the energy back to larger scales despite the lack of vorticity conservation in 2D variable density flows. This gives rise to a positive feedback loop in 2D-RT (absent in 3D) in which mixing layer growth and the associated potential energy release are enhanced relative to 3D, explaining the oft-observed larger $α$ values in 2D simulations. Despite higher bulk kinetic energy levels in 2D, small inertial scales are weaker than in 3D. Moreover, the net upscale cascade in 2D tends to isotropize the large-scale flow, in stark contrast to 3D. Our findings indicate the absence of net upscale energy transfer in 3D-RT as is often claimed; growth of large-scale bubbles and spikes is not due to "mergers" but solely due to baropycnal work $Λ$.

physics.flu-dyn↗

arXiv1999

Ambipolar Diffusion in YSO Jets

We address the issue of ambipolar diffusion in Herbig-Haro jets. The current consensus holds that these jets are launched and collimated via MHD forces. Observations have, however, shown that the jets can be mildly to weakly ionized. Beginning with a simple model for cylindrical equilibrium between neutral, plasma and magnetic pressures we calculate the characteristic time-scale for ambipolar diffusion. Our results show that a significant fraction of HH jets will have ambipolar diffusion time-scales equivalent to, or less than the dynamical time-scales. This implies that MHD equilibria established at the base of a HH jet may not be maintained as the jet propagates far from its source. For typical jet parameters one finds that the length scale where ambipolar diffusion should become significant corresponds to the typical size of large (parsec) scale jets. We discuss the significance of these results for the issue of magnetic fields in parsec-scale jets.

astro-ph↗