arXiv subjects

S. Riggi

Publications and source records attributed to S. Riggi.

At least 19 recordsLinked to original sources

arXiv2026

Discovery of sulfur oxides in the ejecta of a B[e] supergiant

B[e] supergiants represent a rare class of luminous, evolved massive stars surrounded by dusty circumstellar disks. Since their intense UV fields were long thought to sterilize their surroundings, molecular detections beyond carbon monoxide have remained elusive, leaving their chemical reservoirs largely unexplored. Whether these environments can sustain a complex molecular chemistry is a fundamental question with significant astrochemical implications. Here we report the detection of chemically rich molecular gas surrounding the B[e] supergiant HD~87643, using ALMA observations. Our data reveal the presence of the sulfur oxides SO and SO$_2$ and other sulfur-bearing species, marking the first detection of these molecules in an early-type evolved massive star. We find a high fractional abundance of SO$_2$ relative to H$_2$, which our chemical modelling can reproduce in timescales as short as $\sim$10$^4$ yr in an oxygen-rich environment. These results indicate that the detected molecules trace a short-lived, rapidly evolving phase of out-of-equilibrium chemistry. Furthermore, we measure an anomalously low $^{32}$SO/$^{33}$SO, that we attribute to mass-independent fractionation driven by intense photochemistry. This mechanism mirrors processes proposed to explain the $^{33}$S excesses in the atmosphere of the Archaean Earth. Our findings suggest that B[e] supergiants could serve as unique laboratories for studying sulfur chemistry under extreme radiation conditions, opening potential avenues to investigate the fractionation processes that shaped the isotopic signatures found in the early geological record.

astro-ph.SR↗

arXiv2026

Evolved massive stars and their impact on their environment

The comprehension of the final stages of massive star evolution and their path toward the eventual supernova explosion necessarily involves the study of stellar winds and the circumstellar environment (CSE) surrounding them in the transitional phases, during which stellar winds and eruptive mass loss profoundly shape the surrounding environment. The study of the pre-supernova progenitors, from Red Supergiants, passing through the Luminous Blue Variable stage to Wolf-Rayet stars, is of key importance because, focusing on their nebulae, they directly prove the mass-loss activity of the star that, through wind and eruptive events, shapes the environment in which the supernova will explode. Such environment, interacting with the ejecta, will heavily affect supernovae spectrophotometric signatures. The Square Kilometre Array, with its extraordinary capabilities to combine high spatial resolution, sensitivity and wide frequency coverage will adress the most critical observational issues that currently prevent detailed characterization of CSE and, thus limit our ability to constrain its connections to supernova and remnants proprierties.

astro-ph.SR↗

arXiv2026

The 10-15 GHz radio continuum survey of the Galactic Plane with SKAO

Star formation emerges from the complex interplay between gravity, turbulence, magnetic fields, and stellar feedback, all of which vary across spatial scales and Galactic environments. Over the past decades, extensive multiwavelength surveys of the Galactic Plane have progressively unveiled this complexity. Far-infrared and sub-millimetre surveys have identified and characterized tens of thousands of star-forming regions, revealing their mass, temperature, and evolutionary stage. Complementary molecular-line surveys, spanning several CO transitions and isotopologues, have mapped the gas kinematics from giant molecular clouds down to sub-parsec structures. The advent of interferometers such as ALMA has revolutionized this field, enabling systematic studies of gas dynamics, fragmentation, and collapse in dense clumps at scales of a few thousand astronomical units. At the same time, mid-infrared and radio surveys at frequencies 0.8 <= nu <= 5 GHz have traced ionised gas associated with the earliest and latest phases of massive-star evolution, including thermal radio jets, hypercompact and ultracompact HII regions, supernova remnants, planetary nebulae, and evolved massive stars. Yet, a uniform, Galaxy-wide census of ionised structures and feedback processes remains elusive. A transformational leap forward requires a sensitive, high-resolution radio survey of the Galactic Plane at 10-15 GHz, capable of resolving physical scales smaller than 0.05 pc at distances up to 20 kpc. This is precisely the goal of the SKA-Mid Galactic Plane survey, which will, with its unprecedented sensitivity, angular resolution, and mapping speed, provide the first panoptic view of ionised gas and stellar feedback across the Milky Way.

astro-ph.GA↗

arXiv2026

ASKAP EMU detection of an Odd Radio Circle (ORC) candidate: J094412-751016 (Anglerfish)

We report diffuse extended radio-continuum emission spatially coinciding with the IR source WISEA J094409.17-751012.8, and a semi-variable star, V687 Carinae. We use 944 MHz radio data from the large-scale Evolutionary Map of the Universe (EMU) survey to analyse this diffuse emission (EMU J094412-751016), which we nickname "Anglerfish". We investigate if the spatially correlated infrared (IR) source, WISEA J094409.17-751012.8, is physically related to Anglerfish. The IR colours of WISEA J094409.17-751012.8 are indicative of an elliptical galaxy, raising the possibility that Anglerfish may belong to the newly-discovered class of extragalactic radio sources known as Odd Radio Circles (ORCs) with WISEA J094409.17-751012.8 as the host galaxy. We also investigate the possibility that Anglerfish is physically related to the star, V687 Carinae, and whether it may be a remnant from a previous epoch of stellar mass-loss. We determine that a physical association between the radio emission and the star is unlikely due to the emission's non-thermal nature and the star's weak stellar winds compared to the theoretical expansion velocity of the 'shell'. It is possible that Anglerfish may be a Galactic high-latitude supernova remnant (SNR); however, we find that the observed size and luminosity are not consistent with this scenario. We also investigate the ORC scenario, which we deem the most likely scenario based on the Anglerfish's observed properties such as size, brightness, lack of other frequency detections, and spectral index. We therefore propose Anglerfish as an ORC candidate, but note that additional radio and optical observations are vital to further constrain the properties and confirm this classification.

astro-ph.GA↗

arXiv2026

Toward Vision-Language Assistants for Radio Astronomical Source Analysis

Vision-language models (VLMs) have recently shown promise in general-purpose reasoning tasks, yet their applicability to domain-specific scientific workflows remains largely unexplored. In this work, we evaluated a series of open-weight and commercial VLMs on six tasks relevant to radio astronomy, such as source morphology classification. We also introduced radio-llava, a fine-tuned multimodal assistant built on the LLaVA architecture and adapted for the radio domain through instruction fine-tuning. In zero-shot mode, commercial models like GPT-4.1 outperform open-weight VLMs on most radio benchmarks. However, radio-llava significantly improves upon both base LLaVA and commercial models across nearly all tasks. Despite these gains, specialized vision-only models still deliver substantially better performance across the board. Additionally, we observed that fine-tuning introduces catastrophic forgetting on general multimodal tasks, with performance drops up to 40% that can be partly mitigated with a more diverse training dataset or shallow fine-tuning.

astro-ph.IM↗

arXiv2025

A scaling relationship for non-thermal radio emission from ordered magnetospheres - II. Investigating the efficiency of relativistic electron production in magnetospheres of BA-type stars

Magnetic BA stars host dipole-like magnetospheres. When detected as radio sources, their luminosities correlate with the magnetic field and rotation. Rotation is crucial because the mechanism undergirding the relativistic electron production is powered by centrifugal breakouts. CBOs occur wherever magnetic tension does not balance centrifugal force; the resulting magnetic reconnection provides particle acceleration. To investigate how physical conditions at the site of the CBOs affect the efficiency of the acceleration mechanism, we broadly explore the parameter space governing radio emission by increasing the sample of radio-loud magnetic stars. High-sensitivity VLA observations of 32 stars were performed in the hope of identifying new centrifugal magnetospheres and associated CBOs. We calculated gyro-synchrotron spectra using 3D modeling of a dipole-shaped magnetosphere. We evaluated combinations of parameters. The number of relativistic electrons was constrained by the need to produce the emission level predicted by the scaling relationship for the radio emission from magnetic BA stars. About half of the observed stars were detected, with luminosities in agreement with the expected values, reinforcing the robust nature of the scaling relationship for CBO-powered radio emission. Comparing the competing centrifugal and magnetic effects on plasma locked in a rigidly rotating magnetosphere, we located the site of CBOs and inferred the local plasma density. We then estimated the efficiency of the acceleration mechanism needed to produce enough non-thermal electrons to support the radio emission level. Given a constant acceleration efficiency, relativistic electrons represent a fixed fraction of the local thermal plasma. Thus, dense magnetospheres host more energetic particles than less dense ones; consequently, with other parameters similar, they are intrinsically brighter radio sources.

astro-ph.SR↗

arXiv2025

Solar flare forecasting with foundational transformer models across image, video, and time-series modalities

We present a comparative study of transformer-based architectures for solar flare forecasting using heterogeneous data modalities, including images, video sequences, and time-series observations. Our analysis evaluates three recent foundational models - SigLIP2 for image encoding, VideoMAE for spatio-temporal video representation, and Moirai2 for multivariate time-series forecasting - applied to publicly available datasets of solar magnetograms from the SDO/HMI mission and soft X-ray fluxes acquired by GOES satellites. All models are trained and validated under consistent data splits and evaluation criteria, with the goal of assessing the strengths and limitations of transformer backbones across spatial and temporal representations of solar activity. We investigate multiple loss formulations (weighted BCE, focal, and score-oriented) and training balance strategies to mitigate class imbalance typical of flare datasets. Results show that while both SigLIP2 and VideoMAE achieve typical performance on image and video data (True Skill Statistic TSS~0.60-0.65), the time-series model Moirai2 reaches superior forecasting skill (TSS~0.74) using irradiance-based temporal evolution alone. These findings highlight the potential of pretrained transformer architectures and cross-modal learning for advancing operational space weather forecasting, paving the way toward unified multimodal models that integrate visual and temporal information.

astro-ph.IM↗

arXiv2025

MeerKAT discovery of 164 compact radio rings toward the Galactic Plane

We report the discovery of 164 compact (radius < 1 arcmin) radio rings using MeerKAT 1.3 GHz data from the SARAO MeerKAT Galactic Plane Survey (l=2-60deg, 252-358deg, |b|<1.5deg) and the Galactic Centre mosaic, from a search aimed at identifying previously uncatalogued radio sources. Within this sample, approximately 19 per cent of the rings contain a central point radio source. A multiwavelength analysis reveals a striking diversity: about 40 per cent of the rings enclose an isolated infrared point source, 50 per cent exhibit an extended counterpart in the mid- or far-infrared, and several are only detected in the radio band. We found that 17 per cent of the rings in the sample are positionally coincident (within 5 arcsec) with known entries in SIMBAD, including unclassified infrared sources, spiral galaxies, young stellar objects and long-period variable candidates. Based on these matches and exploiting ancillary multiwavelength data and catalogues, we explore several formation scenarios for the rings, such as HII regions, planetary nebulae, mass-loss relics from evolved massive stars, supernova remnants, nova shells, galaxies, galaxy cluster lenses and odd radio circles. Tentative classifications are proposed for nearly 60 per cent of the sample. These results highlight the potential of MeerKAT to uncover previously undetected compact radio structures and, particularly, recover missing Galactic radio-emitting objects.

astro-ph.GA↗

arXiv2025

Broadband radio study of the supernova remnant Kes 73

Strong shocks occurring in supernova remnants (SNRs), and their interaction with an often anisotropic surrounding medium, make SNRs ideal laboratories for studying the production and acceleration of cosmic rays (CRs). Due to their complex morphology and phenomenology, different CR populations are expected to exist throughout the remnants, each characterized by its own energy spectrum. A comprehensive understanding of particle acceleration mechanisms and energetics in SNRs requires spatially resolved spectral and morphological studies. We want to highlight the crucial role of high-resolution radio images at high frequencies (> 10 GHz) for studying the spectral properties of different remnant regions and better constraining the models that describe their non-thermal emission from radio to $\gamma$-ray wavelengths. We studied the integrated radio spectrum of the SNR Kes 73 using single-dish observations performed with the Sardinia Radio Telescope (SRT) between 6.9 and 24.8 GHz, complemented by published data. The high-resolution map at 24.8 GHz was used to search for spatial variations in the spectral index across the remnant. We present the SRT images of Kes 73, providing the highest-frequency morphological and spectral characterization ever obtained for this source. By combining our 18.7 and 24.8 GHz maps with previously published interferometric images at 1.4 and 5 GHz, we identify a flatter spectrum in the western bright region compared to the rest of the shell. In the same region, we detect overlapping $^12$CO molecular emission and $\gamma$-ray radiation, providing strong evidence of SNR-molecular cloud interaction and enhanced CR production. We modelled the non-thermal radio to $\gamma$-ray emission from this region, favouring a lepto-hadronic scenario with a maximum electron energy of 1.1 TeV and a magnetic field strength of 25 $\mu$G.

astro-ph.HE↗

arXiv2025

The Evolutionary Map of the Universe: A new radio atlas for the southern hemisphere sky

We present the Evolutionary Map of the Universe (EMU) survey conducted with the Australian Square Kilometre Array Pathfinder (ASKAP). EMU aims to deliver the touchstone radio atlas of the southern hemisphere. We introduce EMU and review its science drivers and key science goals, updated and tailored to the current ASKAP five-year survey plan. The development of the survey strategy and planned sky coverage is presented, along with the operational aspects of the survey and associated data analysis, together with a selection of diagnostics demonstrating the imaging quality and data characteristics. We give a general description of the value-added data pipeline and data products before concluding with a discussion of links to other surveys and projects and an outline of EMU's legacy value.

astro-ph.GA↗

arXiv2025

radio-llava: Advancing Vision-Language Models for Radio Astronomical Source Analysis

The advent of next-generation radio telescopes is set to transform radio astronomy by producing massive data volumes that challenge traditional processing methods. Deep learning techniques have shown strong potential in automating radio analysis tasks, yet are often constrained by the limited availability of large annotated datasets. Recent progress in self-supervised learning has led to foundational radio vision models, but adapting them for new tasks typically requires coding expertise, limiting their accessibility to a broader astronomical community. Text-based AI interfaces offer a promising alternative by enabling task-specific queries and example-driven learning. In this context, Large Language Models (LLMs), with their remarkable zero-shot capabilities, are increasingly used in scientific domains. However, deploying large-scale models remains resource-intensive, and there is a growing demand for AI systems that can reason over both visual and textual data in astronomical analysis. This study explores small-scale Vision-Language Models (VLMs) as AI assistants for radio astronomy, combining LLM capabilities with vision transformers. We fine-tuned the LLaVA VLM on a dataset of 59k radio images from multiple surveys, enriched with 38k image-caption pairs from the literature. The fine-tuned models show clear improvements over base models in radio-specific tasks, achieving ~30% F1-score gains in extended source detection, but they underperform vision-only classifiers and exhibit ~20% drop on general multimodal tasks. Inclusion of caption data and LoRA fine-tuning enhances instruction-following and helps recover ~10% accuracy on multimodal benchmarks. This work lays the foundation for future advancements in radio VLMs, highlighting their potential and limitations, such as the need for better multimodal alignment, higher-quality datasets, and mitigation of catastrophic forgetting.

astro-ph.IM↗

arXiv2025

The SARAO MeerKAT Galactic Plane Survey extended source catalogue

We present a catalogue of extended radio sources from the SARAO MeerKAT Galactic Plane Survey (SMGPS). Compiled from 56 survey tiles and covering approximately 500 deg$^2$ across the first, third, and fourth Galactic quadrants, the catalogue includes 16534 extended and diffuse sources with areas larger than 5 synthesised beams. Of them, 3891 (24\% of the total) are confidently associated with known Galactic radio-emitting objects in the literature, such as HII regions, supernova remnants, planetary nebulae, luminous blue variables, and Wolf-Rayet stars. A significant fraction of the remaining sources, 5462 (33\%), are candidate extragalactic sources, while 7181 (43\%) remain unclassified. Isolated radio filaments are excluded from the catalogue. The diversity of extended sources underscores MeerKAT's contribution to the completeness of censuses of Galactic radio emitters, and its potential for new scientific discoveries. For the catalogued sources, we derived basic positional and morphological parameters, as well as flux density estimates, using standard aperture photometry. This paper describes the methods followed to generate the catalogue from the original SMGPS tiles, detailing the source extraction, characterisation, and crossmatching procedures. Additionally, we analyse the statistical properties of the catalogued populations

astro-ph.GA↗

arXiv2024

Diprotodon on the sky. The Large Galactic Supernova Remnant (SNR) G278.94+1.35

We present a re-discovery of G278.94+1.35 as possibly one of the largest known Galactic supernova remnants (SNR) - that we name Diprotodon. While previously established as a Galactic SNR, Diprotodon is visible in our new EMU and GLEAM radio continuum images at an angular size of 3.33x3.23 deg, much larger than previously measured. At the previously suggested distance of 2.7 kpc, this implies a diameter of 157x152 pc. This size would qualify Diprotodon as the largest known SNR and pushes our estimates of SNR sizes to the upper limits. We investigate the environment in which the SNR is located and examine various scenarios that might explain such a large and relatively bright SNR appearance. We find that Diprotodon is most likely at a much closer distance of $\sim$1 kpc, implying its diameter is 58x56 pc and it is in the radiative evolutionary phase. We also present a new Fermi-LAT data analysis that confirms the angular extent of the SNR in gamma-rays. The origin of the high-energy emission remains somewhat puzzling, and the scenarios we explore reveal new puzzles, given this unexpected and unique observation of a seemingly evolved SNR having a hard GeV spectrum with no breaks. We explore both leptonic and hadronic scenarios, as well as the possibility that the high-energy emission arises from the leftover particle population of a historic pulsar wind nebula.

astro-ph.HE↗

arXiv2024

Detection and classification of radio sources with deep learning

In this paper we present three different applications, based on deep learning methodologies, that we are developing to support the scientific analysis conducted within the ASKAP-EMU and MeerKAT radio surveys. One employs instance segmentation frameworks to detect compact and extended radio sources and imaging artefacts from radio continuum images. Another application uses gradient boosting decision trees and convolutional neural networks to classify compact sources into different astronomical classes using combined radio and infrared multi-band images. Finally, we discuss how self-supervised learning can be used to obtain valuable radio data representations for source detection, and classification studies.

astro-ph.IM↗

arXiv2024

Supernova Remnant Candidates Discovered by the SARAO MeerKAT Galactic Plane Survey

Context. Sensitive radio continuum data could remove the difference between the number of known supernova remnants (SNRs) in the Galaxy compared to that expected, but due to confusion in the Galactic plane, faint SNRs can be challenging to distinguish from brighter HII regions and filamentary radio emission. Aims. We wish to exploit new SARAO MeerKAT 1.3 GHz Galactic Plane Survey (SMGPS) radio continuum data, which covers $251\deg \le l \le 358\deg$ and $2\deg \le l \le 61\deg$ at $|b|\le 1.5\deg$, to search for SNR candidates in the Milky Way disk. Methods. We also use MIR data from the Spitzer GLIMPSE, Spitzer MIPSGAL, and WISE surveys to help identify SNR candidates. The identified SNR candidate are sources of extended radio continuum emission that lack MIR counterparts, are not known as HII regions in the WISE Catalog of Galactic HII Regions, and are not known previously as SNRs Results. We locate 237 new Galactic SNR candidates in the SMGPS data. We also identify and confirm the expected radio morphology for 201 objects listed in the literature as being SNRs and 130 previously-identified SNR candidates. The known and candidate SNRs have similar spatial distributions and angular sizes. Conclusions. The SMGPS data allowed us to identify a large population of SNR candidates that can be confirmed as true SNRs using radio polarization measurements or by deriving radio spectral indices. If the 237 candidates are confirmed as true SNRs, it would approximately double the number of known Galactic SNRs in the survey area, alleviating much of the difference between the known and expected populations.

astro-ph.GA↗

arXiv2024

Sifting the debris: Patterns in the SNR population with unsupervised ML methods

Supernova remnants (SNRs) carry vast amounts of mechanical and radiative energy that heavily influence the structural, dynamical, and chemical evolution of galaxies. To this day, more than 300 SNRs have been discovered in the Milky Way, exhibiting a wide variety of observational features. However, existing classification schemes are mainly based on their radio morphology. In this work, we introduce a novel unsupervised deep learning pipeline to analyse a representative subsample of the Galactic SNR population ($\sim$ 50% of the total) with the aim of finding a connection between their multi-wavelength features and their physical properties. The pipeline involves two stages: (1) a representation learning stage, consisting of a convolutional autoencoder that feeds on imagery from infrared and radio continuum surveys (WISE 22$\mu$m, Hi-GAL 70 $\mu$m and SMGPS 30 cm) and produces a compact representation in a lower-dimensionality latent space; and (2) a clustering stage that seeks meaningful clusters in the latent space that can be linked to the physical properties of the SNRs and their surroundings. Our results suggest that this approach, when combined with an intermediate uniform manifold approximation and projection (UMAP) reprojection of the autoencoded embeddings into a more clusterable manifold, enables us to find reliable clusters. Despite a large number of sources being classified as outliers, most clusters relate to the presence of distinctive features, such as the distribution of infrared emission, the presence of radio shells and pulsar wind nebulae, and the existence of dust filaments.

astro-ph.GA↗

arXiv2024

MeerKAT reveals a ghostly thermal radio ring towards the Galactic Centre

We present the serendipitous discovery of a new radio-continuum ring-like object nicknamed Kyklos (J1802-3353), with MeerKAT UHF and L-band observations. The radio ring, which resembles the recently discovered odd radio circles (ORCs), has a diameter of 80 arcsec and is located just 6 deg from the Galactic plane. However, Kyklos exhibits an atypical thermal radio-continuum spectrum ({\alpha} = -0.1 +/- 0.3), which led us to explore different possible formation scenarios. We concluded that a circumstellar shell around an evolved massive star, possibly a Wolf-Rayet, is the most convincing explanation with the present data.

astro-ph.GA↗

arXiv2024

Self-supervised contrastive learning of radio data for source detection, classification and peculiar object discovery

New advancements in radio data post-processing are underway within the SKA precursor community, aiming to facilitate the extraction of scientific results from survey images through a semi-automated approach. Several of these developments leverage deep learning (DL) methodologies for diverse tasks, including source detection, object or morphology classification, and anomaly detection. Despite substantial progress, the full potential of these methods often remains untapped due to challenges associated with training large supervised models, particularly in the presence of small and class-unbalanced labelled datasets. Self-supervised learning has recently established itself as a powerful methodology to deal with some of the aforementioned challenges, by directly learning a lower-dimensional representation from large samples of unlabelled data. The resulting model and data representation can then be used for data inspection and various downstream tasks if a small subset of labelled data is available. In this work, we explored contrastive learning methods to learn suitable radio data representation from unlabelled images taken from the ASKAP EMU and SARAO MeerKAT GPS surveys. We evaluated trained models and the obtained data representation over smaller labelled datasets, also taken from different radio surveys, in selected analysis tasks: source detection and classification, and search for objects with peculiar morphology. For all explored downstream tasks, we reported and discussed the benefits brought by self-supervised foundational models built on radio data.

astro-ph.IM↗