Searcharxiv⌕ Search

arXiv subjects

Nicola C. Amorisco

Publications and source records attributed to Nicola C. Amorisco.

At least 19 recordsLinked to original sources
arXiv2026

Real-time virtual circuits for plasma shape control via neural network emulators: experimental demonstration on MAST Upgrade

Conventional plasma shape control in tokamaks relies on virtual circuits (VCs) that are computed offline from linearisations around a small, tailored number of reference equilibria, and deployed as expertly prepared schedules during the discharge. Here, we report on the first experimental deployment of real-time VCs. We replace pre-set look up tables with VCs updated in real time using surrogates of the plasma response. Both the existing control architecture and the interpretability of VC-based control are retained. Previous work showed that neural network emulators can produce accurate VCs, and validated their performance in closed-loop shape control simulations. Here, we report their first experimental validation on MAST Upgrade (MAST-U). Dedicated experiments spanning different scenarios, including prescribed shape perturbations, feedback-driven divertor-leg motion, and strongly evolving plasma configurations, show that real-time VCs can realise plasma shape control tasks within the MAST-U plasma control system. These results establish the experimental feasibility of real-time linearisations as a practical extension of conventional plasma shape control in tokamaks. The present implementation demonstrates a central step towards a simpler control workflow, in which manually constructed, phased VC schedules are replaced by VCs generated automatically online from a trained surrogate model, without scenario-specific retraining.

physics.plasm-ph↗
arXiv2026

Real-time virtual circuits for plasma shape control via neural network emulators: integration and testing in the MAST-U PCS

The deployment of advanced, AI-enabled control algorithms in tokamak experiments requires robust integration with existing plasma control system (PCS) architectures and extensive pre-experimental validation. In this contribution, we describe the integration and testing of neural-network-emulated virtual circuits for plasma shape control within the MAST Upgrade (MAST-U) PCS environment. The neural network models predict the plasma shape using the plasma current, poloidal field coil currents, and plasma profile parameters. In this paper, we explain how they are deployed via a real-time C++ inference server that interfaces with the PCS, returning the shape prediction and its Jacobian, and how, from the latter, virtual circuit matrices and updated coil current requests are computed for real-time actuation. Emphasis is placed on the validation workflow and best practices adopted to ensure confidence in the proposed control framework prior to experimental deployment. This work demonstrates practical AI-based shape control components for fusion control systems, with direct relevance for upcoming MAST-U experiments and future devices.

physics.plasm-ph↗
arXiv2026

Real-time virtual circuits for plasma shape control via neural network emulators

Reliable position and shape control in tokamak plasmas requires accurate real-time regulation of several strongly coupled shape parameters. The control vectors that disentangle these couplings, referred to as \textit{virtual circuits} (VCs), enable independent shape parameter control for a specific Grad--Shafranov (GS) equilibrium. Numerical calculation of VCs is not currently feasible in real time, therefore VCs are usually computed prior to each experiment, using a small number of reference GS equilibria sampled along the desired scenario trajectory, with each VC used to control the plasma within a preset time interval. While effective near the reference equilibrium, this approach can lead to degraded performance as the plasma departs from the reference equilibrium and/or from the desired trajectory, and it complicates the design of robust control strategies for rapidly evolving plasma configurations. In this paper, we construct neural-network-based emulators of plasma shape parameters from which VCs can be derived, to provide the MAST Upgrade (MAST-U) plasma control system with state-aware VCs in real-time. To do this, we develop an extensive library of over a million simulated GS equilibria, covering a substantial portion of the MAST-U operational space. These emulators provide differentiable functions whose gradients can be rapidly computed, enabling the derivation of accurate VCs for real-time shape control. We perform extensive verification of the emulated VCs by testing whether they disentangle the control problem. The neural-network-based approach delivers high accuracy and orthogonality across a diverse range of equilibria. This work establishes the physical validity of emulated VCs as a scalable and general alternative to schedules of precomputed VCs.

physics.plasm-ph↗
arXiv2026

Dwarf stellar haloes: a powerful probe of small-scale galaxy formation and the nature of dark matter

We use N-body cosmological simulations and empirical galaxy models to study the merger history of dwarf-mass galaxies (with M_halo~10^10 M_Sun). Our input galaxy models describe the stellar mass-halo mass relation, and the galaxy occupation fraction. The number of major and minor mergers depends on the type of dark matter; in particular, minor mergers are greatly suppressed in warm dark matter models. In addition, the number of mergers that bring in stars is strongly dependent on the galaxy occupation model. For example, minor mergers are negligible for stellar halo growth in models with a high mass threshold for galaxy formation (i.e. 10^9.3 M_Sun at z=0). Moreover, this threshold for galaxy formation can also determine the relative difference (if any) between the stellar haloes of satellite and field dwarfs. Using isolated simulations of dwarf-dwarf mergers, we show that the relative frequency of major and minor mergers predict very different stellar haloes: Typically, "intermediate" dark matter merger ratios (~1:5) maximise the growth of distant stellar haloes. We discuss the observability of dwarf stellar haloes and find that the surface brightness of these features are incredibly faint. However, when several dwarfs are stacked together models that form particularly rich stellar haloes could be detectable. Finally, we show that stellar streams in the Galactic halo overlapping in phase-space with known dwarf satellites are likely remnants of their stripped stellar haloes. The mere existence of dwarf stellar haloes can already put constraints on some small-scale models, and thus observational probes should be a high priority.

astro-ph.GA↗
arXiv2026

TokaMark: A Comprehensive Benchmark for MAST Tokamak Plasma Models

Development and operation of commercially viable fusion energy reactors such as tokamaks require accurate predictions of plasma dynamics from sparse, noisy, and incomplete sensors readings. The complexity of the underlying physics and the heterogeneity of experimental data pose formidable challenges for conventional numerical methods, and highlight the promise of modern data-native approaches. A major obstacle in realizing this potential is, however, the lack of curated, openly available datasets and standardized benchmarks. Existing fusion datasets are scarce, fragmented across institutions, facility-specific, and inconsistently annotated, which limits reproducibility and prevents a fair and scalable comparison of AI approaches. In this paper, we introduce TokaMark, a structured benchmark to evaluate AI models on real experimental data collected from the Mega Ampere Spherical Tokamak (MAST). TokaMark provides a comprehensive suite of tools designed to unify access to multi-modal fusion data and standardize evaluation protocols. The benchmark includes a curated list of 14 tasks spanning a range of physical mechanisms, exploiting a variety of diagnostics and covering multiple operational use cases. A baseline model is provided to facilitate transparent comparison and validation within a unified framework. By establishing a unified benchmark, TokaMark aims to accelerate progress in data-driven AI-based plasma modeling, contributing to the broader goal of achieving sustainable and stable fusion energy. The dataset, benchmark, documentation, and tooling are open-sourced under https://github.com/UKAEA-IBM-STFC-Fusion-FMs/tokamark_baseline.

physics.plasm-ph↗
arXiv2026

TokaMind: A Multi-Modal Transformer Foundation Model for Tokamak Plasma Dynamics

We present TokaMind, to our knowledge the first open-source foundation model for tokamak plasma dynamics, based on a Multi-Modal Transformer (MMT) and pretrained on heterogeneous diagnostics from the publicly available MAST dataset. TokaMind supports multiple data modalities (time-series, 2D profiles, and videos) with different sampling rates, robust missing-signal handling, and efficient task adaptation via selectively loading and freezing four model components. To represent multi-modal signals, we use a lightweight fixed-basis Discrete Cosine Transform embedding (DCT3D) and provide a clean interface for alternative embeddings (e.g., Variational Autoencoders). We evaluate TokaMind on the recently introduced MAST benchmark TokaMark, which comprises 14 tasks with heterogeneous reconstruction and forecasting objectives. Our results show that fine-tuned TokaMind outperforms the strongest benchmark baseline on all but one task. Compared with training the same architecture from scratch under a matched epoch budget, warm-start adaptation is most beneficial on demanding downstream settings, including long-horizon forecasting and high-dimensional equilibrium objectives. These findings highlight the value of multi-modal pretraining for tokamak plasma dynamics and provide a practical, extensible foundation for future fusion modeling tasks. Training code and model weights are publicly available at github.com/UKAEA-IBM-STFC-Fusion-FMs/tokamind and huggingface.co/UKAEA-IBM-STFC, respectively.

physics.plasm-ph↗
arXiv2025

Real-Time Applicability of Emulated Virtual Circuits for Tokamak Plasma Shape Control

Machine learning has recently been adopted to emulate sensitivity matrices for real-time magnetic control of tokamak plasmas. However, these approaches would benefit from a quantification of possible inaccuracies. We report on two aspects of real-time applicability of emulators. First, we quantify the agreement of target displacement from VCs computed via Jacobians of the shape emulators with those from finite differences Jacobians on exact Grad-Shafranov solutions. Good agreement ($\approx$5-10%) can be achieved on a selection of geometric targets using combinations of neural network emulators with $\approx10^5$ parameters. A sample of $\approx10^{5}-10^{6}$ synthetic equilibria is essential to train emulators that are not over-regularised or overfitting. Smaller models trained on the shape targets may be further fine-tuned to better fit the Jacobians. Second, we address the effect of vessel currents that are not directly measured in real-time and are typically subsumed into effective "shaping currents" when designing virtual circuits. We demonstrate that shaping currents can be inferred via simple linear regression on a trailing window of active coil current measurements with residuals of only a few Ampères, enabling a choice for the most appropriate shaping currents at any point in a shot. While these results are based on historic shot data and simulations tailored to MAST-U, they indicate that emulators with few-millisecond latency can be developed for robust real-time plasma shape control in existing and upcoming tokamaks.

physics.plasm-ph↗
arXiv2023

Automated galaxy-galaxy strong lens modelling: no lens left behind

The distribution of dark and luminous matter can be mapped around galaxies that gravitationally lens background objects into arcs or Einstein rings. New surveys will soon observe hundreds of thousands of galaxy lenses, and current, labour-intensive analysis methods will not scale up to this challenge. We instead develop a fully automatic, Bayesian method which we use to fit a sample of 59 lenses imaged by the Hubble Space Telescope in uniform conditions. We set out to \textit{leave no lens behind} and focus on ways in which automated fits fail in a small handful of lenses, describing adjustments to the pipeline that allows us to infer accurate lens models. Our pipeline ultimately fits {\em all} 59 lenses in our sample, with a high success rate key because catastrophic outliers would bias large samples with small statistical errors. Machine Learning techniques might further improve the two most difficult steps: subtracting foreground lens light and initialising a first, approximate lens model. After that, increasing model complexity is straightforward. We find a mean $\sim1\%$ measurement precision on the measurement of the Einstein radius across the lens sample which {\em does not degrade with redshift} up to at least $z=0.7$ -- in stark contrast to other techniques used to study galaxy evolution, like stellar dynamics. Our \texttt{PyAutoLens} software is open source, and is also installed in the Science Data Centres of the ESA Euclid mission.

astro-ph.CO↗
arXiv2022

Testing strong lensing subhalo detection with a cosmological simulation

Strong gravitational lensing offers a compelling test of the cold dark matter paradigm, as it allows for subhaloes with masses of $\sim10^{9}$ M$_\odot$ and below to be detected. We test commonly-used techniques for detecting subhaloes superposed in images of strongly lensed galaxies. For the lens we take a simulated galaxy in a $\sim10^{13}$ M$_\odot$ halo grown in a high-resolution cosmological hydrodynamical simulation, which we view from two different directions. Though the resolution is high, we note the simulated galaxy still has an artificial core which adds additional complexity to the baryon dominated region. To remove particle noise, we represent the projected galaxy mass distribution by a series of Gaussian profiles which precisely capture the features of the projected galaxy. We first model the lens mass as a (broken) power-law density profile and then search for small haloes. Of the two projections, one has a regular elliptical shape, while the other has distinct deviations from an elliptical shape. For the former, the broken power-law model gives no false positives and correctly recovers the mass of the superposed small halo, but for the latter we find false positives and the inferred halo mass is overestimated by $\sim4-5$ times. We then use a more complex model in which the lens mass is decomposed into stellar and dark matter components. In this case, we show that we can capture the simulated galaxy's complex projected structures and correctly infer the input small halo.

astro-ph.CO↗
arXiv2022

Galaxy-galaxy strong lens perturbations: line-of-sight haloes versus lens subhaloes

We rederive the number density of intervening line-of-sight haloes relative to lens subhaloes in galaxy-galaxy strong lensing observations, where these perturbers can generate detectable image fluctuations. Previous studies have calculated the detection limit of a line-of-sight small-mass dark halo by comparing the lensing deflection angles it would cause, to those caused by a subhalo within the lens. However, this overly simplifies the difference in observational consequences between a subhalo and a line-of-sight halo. Furthermore, it does not take into account degeneracies between an extra subhalo and the uncertain properties of the main lens. More in keeping with analyses of real-world observations, we regard a line-of-sight halo as detectable only if adding it to a smooth model generates a statistically significant improvement in the reconstructed image. We find that the number density of detectable line-of-sight perturbers has been overestimated by as much as a factor of two in the previous literature. For typical lensing geometries and configurations, very deep imaging is sensitive to twice as many line-of-sight perturbers as subhaloes, but moderate depth imaging is sensitive to only slightly more line-of-sight perturbers than subhaloes.

astro-ph.CO↗
arXiv2021

Systematic errors induced by the elliptical power-law model in galaxy-galaxy strong lens modeling

The elliptical power-law (EPL) model of the mass in a galaxy is widely used in strong gravitational lensing analyses. However, the distribution of mass in real galaxies is more complex. We quantify the biases due to this model mismatch by simulating and then analysing mock {\it Hubble Space Telescope} imaging of lenses with mass distributions inferred from SDSS-MaNGA stellar dynamics data. We find accurate recovery of source galaxy morphology, except for a slight tendency to infer sources to be more compact than their true size. The Einstein radius of the lens is also robustly recovered with 0.1% accuracy, as is the global density slope, with 2.5% relative systematic error, compared to the 3.4% intrinsic dispersion. However, asymmetry in real lenses also leads to a spurious fitted `external shear' with typical strength, $γ_{\rm ext}=0.015$. Furthermore, time delays inferred from lens modelling without measurements of stellar dynamics are typically underestimated by $\sim$5%. Using such measurements from a sub-sample of 37 lenses would bias measurements of the Hubble constant $H_0$ by $\sim$9%. Although this work is based on a particular set of MaNGA galaxies, and the specific value of the detected biases may change for another set of strong lenses, our results strongly suggest the next generation cosmography needs to use more complex lens mass models.

astro-ph.GA↗
arXiv2021

Cold dark matter subhaloes at arbitrarily low masses

A defining prediction of the cold dark matter (CDM) cosmological model is the existence of a very large population of low mass haloes, down to planet-size masses. However, their fate as they are accreted onto haloes many orders of magnitude more massive remains fundamentally uncertain. A number of numerical explorations have found subhaloes to be very resilient to tides, but resolution limits make it difficult to explore tidal evolution at arbitrarily low masses. What are the structural properties of heavily stripped subhaloes? Do tidal effects destroy low-mass CDM subhaloes? Here we focus on cosmologically motivated subhaloes orbiting CDM hosts, and show that subhaloes of any initial mass can be stripped to arbitrarily small mass fractions. We show that previous numerical results can be reproduced by a simple model that describes tidal evolution as a progressive `peeling' in energy space and subsequent re-virialization. Tidal heating can effectively be neglected at all masses, because its importance i) does not increase for subhaloes of decreasing mass at accretion, and ii) it decreases as stripping proceeds. This allows us to predict analytically the structural properties of subhaloes with any initial mass and arbitrary degrees of mass loss. Under the hypotheses that CDM haloes have centrally divergent density profiles and approximately isotropic phase space distributions, our results prove that subhaloes of very low masses are a robust prediction of the CDM model: as for haloes, CDM subhalo populations extend and are abundant down to very low masses.

astro-ph.CO↗
arXiv2021

Halo concentration strengthens dark matter constraints in galaxy-galaxy strong lensing analyses

A defining prediction of the cold dark matter (CDM) cosmological model is the existence of a very large population of low-mass haloes. This population is absent in models in which the dark matter particle is warm (WDM). These alternatives can, in principle, be distinguished observationally because halos along the line-of-sight can perturb galaxy-galaxy strong gravitational lenses. Furthermore, the WDM particle mass could be deduced because the cut-off in their halo mass function depends on the mass of the particle. We systematically explore the detectability of low-mass haloes in WDM models by simulating and fitting mock lensed images. Contrary to previous studies, we find that halos are harder to detect when they are either behind or in front of the lens. Furthermore, we find that the perturbing effect of haloes increases with their concentration: detectable haloes are systematically high-concentration haloes, and accounting for the scatter in the mass-concentration relation boosts the expected number of detections by as much as an order of magnitude. Haloes have lower concentration for lower particle masses and this further suppresses the number of detectable haloes beyond the reduction arising from the lower halo abundances alone. Taking these effects into account can make lensing constraints on the value of the mass function cut-off at least an order of magnitude more stringent than previously appreciated.

astro-ph.CO↗
arXiv2019

Giant cold satellites from low-concentration haloes

The dwarf satellite galaxies of the Milky Way Crater II and Antlia II have uncommonly low dynamical mass densities, due to their large size and low velocity dispersion. Previous work have failed to identify formation scenarios within the $Λ$CDM framework and have invoked cored dark matter haloes, processed by tides. I show that the tidal evolution of $Λ$CDM NFW haloes is richer than previously recognised: tidal heating causes the innermost regions of haloes that fall short of the mass-concentration relation to expand significantly, resulting in the formation of giant, kinematically cold satellites like Crater II and Antlia II. Furthermore, while the satellite is reaching apocenter, extra-tidal material can cause an even more inflated appearance. When present, as likely for the larger Antlia II, nominally unbound material can be recognised thanks to its somewhat hotter kinematics and line-of-sight velocity gradient. Contrary to other formation scenarios, Crater II and Antlia II may well have experienced very little mass loss, as in fact hinted by their observed metallicity. If indeed a satellite of NGC1052, tidal evolution of a low-concentration halo may similarly have led to the formation of NGC1052-DF2.

astro-ph.GA↗
arXiv2018

Globular cluster populations and the kinematical fingerprints of minor mergers

We use Monte Carlo $Λ$CDM assembly histories, minor-merger N-body simulations and empirical relations between halo mass and the globular cluster (GC) abundance to study the kinematical properties of halo GCs in massive galaxies, with $M_{vir}(z=0)=10^{13.5}M_\odot$. While the accreted stellar halo is dominated by the contributions of massive satellites, we show that satellites with low virial mass (i.e. low satellite-to-host virial mass ratio, VMRs) are important contributors to the population of accreted GCs. The relative contribution of accretion events with low VMRs is highest for the halo population of blue GCs and gradually decreases for red GCs and accreted stars. As a consequence of the reduced efficiency of dynamical friction on minor mergers, our populations of cosmologically accreted blue GCs are systematically more spatially extended and have higher velocity dispersions than for red GCs, in agreement with observations. For the same reason, assembly histories including a higher fraction of minor mergers result in more spatially extended GC populations. GC line-of-sight velocity distributions featuring negative values of the kurtosis $κ$, as recently observed, are ubiquitous in our models. Therefore, $κ<0$ is not at odds with an accretion scenario, and in fact a fingerprint of the important contribution of minor mergers. However, our populations of accreted GCs remain mostly radially biased, with profiles of the anisotropy parameter $β$ that are mildly radial in the center ($β(r<10~{\rm kpc})\sim0.2$) and strongly radially anisotropic at large galactocentric distances ($β(r>30~{\rm kpc})\gtrsim0.6$), for both red and blue populations.

astro-ph.GA↗
arXiv2018

Formation and Incidence of Shell Galaxies in the Illustris Simulation

Shells are low surface brightness tidal debris that appear as interleaved caustics with large opening angles, often situated on both sides of the galaxy center. In this paper, we study the incidence and formation processes of shell galaxies in the cosmological gravity+hydrodynamics Illustris simulation. We identify shells at redshift z=0 using stellar surface density maps, and we use stellar history catalogs to trace the birth, trajectory and progenitors of each individual star particle contributing to the tidal feature. Out of a sample of the 220 most massive galaxies in Illustris ($\mathrm{M}_{\mathrm{200crit}}>6\times10^{12}\,\mathrm{M}_{\odot}$), $18\%\pm3\%$ of the galaxies exhibit shells. This fraction increases with increasing mass cut: higher mass galaxies are more likely to have stellar shells. Furthermore, the fraction of massive galaxies that exhibit shells decreases with increasing redshift. We find that shell galaxies observed at redshift $z=0$ form preferentially through relatively major mergers ($\gtrsim$1:10 in stellar mass ratio). Progenitors are accreted on low angular momentum orbits, in a preferred time-window between $\sim$4 and 8 Gyrs ago. Our study indicates that, due to dynamical friction, more massive satellites are allowed to probe a wider range of impact parameters at accretion time, while small companions need almost purely radial infall trajectories in order to produce shells. We also find a number of special cases, as a consequence of the additional complexity introduced by the cosmological setting. These include galaxies with multiple shell-forming progenitors, satellite-of-satellites also forming shells, or satellites that fail to produce shells due to multiple major mergers happening in quick succession.

astro-ph.GA↗
arXiv2018

First constraints on Fuzzy Dark Matter from the dynamics of stellar streams in the Milky Way

We present a novel method to constrain the mass of ultra-light bosons as the dark matter using stellar streams formed by disrupting Globular Clusters in the Milky Way. The turbulent density field of Fuzzy Dark Matter (FDM) haloes results in perturbations and dynamical heating of thin streams. Using numerical simulations based on an effective model, we explore the magnitude of this phenomenon and show that this is observable for the range of axion masses $m_a$ that is interesting for alleviating the `small-scale problems' of $Λ$CDM. We derive an analytical model for the thickening of thin stellar streams and obtain an early conservative lower limit for the boson mass of $m_a > 1.5 \times 10^{-22}~$eV, using pre-Gaia literature data for six Milky Way streams and after marginalizing over physical parameters. This demonstrates the great promise for using this novel dynamical method as a fully independent probe of FDM, to complement results based on Lyman-$α$ forest data.

astro-ph.GA↗
arXiv2018

The virial mass distribution of ultra-diffuse galaxies in clusters and groups

We use the observed abundances of ultra-diffuse galaxies (UDGs) in clusters and groups and ΛCDM subhalo mass functions to put constraints on the distribution {of present-day halo masses of satellite} UDGs. If all of the most massive subhaloes in the cluster host a UDG, UDGs occupy all subhaloes with \log M_{sub}/M_\odot\gtrsim11. For a model in which the efficiency of UDG formation is higher around some characteristic halo mass, higher fractions of massive UDGs require larger spreads in the UDG mass distribution. In a cluster with a virial mass of 10^{15}M_\odot, the 90% upper limit for the fraction of UDGs with \log M_{sub}/M_\odot>12 is 7%, occupying 70% of all cluster subhaloes above the same mass. To reproduce the observed abundances, however, the mass distribution of satellite UDGs has to be broad, with >30% having \log M_{sub}/M_\odot<10.9. This strongly supports that UDGs are part of a continuous distribution in which a majority are hosted by low mass haloes. The abundance of satellite UDGs may fall short of the linear relation with the cluster/group mass M_{host} in low-mass hosts, \log M_{host}/M_\odot\sim 12. Characterising these deviations -- or the lack thereof -- will allow for stringent constraints on the UDG virial mass distribution.

astro-ph.GA↗