Hareesh Thuruthipilly

Context . Ultra-diffuse galaxies (UDGs) in the field, gravitationally isolated from galaxy clusters, are faint systems that remain poorly represented in the literature due to lack of spectroscopic confirmation and the difficulty of obtaining high signal-to-noise emission-line measurements. Aims . We present a spectroscopic study of 17 blue UDG candidates in the field using optical emission lines to confirm their ultra-diffuse nature and characterise their physical properties. Methods . Data were collected using the MODS instrument at the Large Binocular Telescope Observatory and the MISTRAL instrument at the Observatoire de Haute-Provence. Using the software pandora EZ , we derived spectroscopic redshifts ( z _spec ) for our field UDG candidates. We then computed their effective radii ( r _eff ) and central surface brightnesses (µ _0,g ). We measured the H α and H β emission line fluxes in the 17 spectra and derived star formation rates (SFRs) from the line-luminosity relation. We performed forced photometry on our sample to obtain photometric fluxes and applied an aperture correction on the H α integrated fluxes, propagating the correction to the derived SFRs. Using colour relations and estimated dust attenuation, we then computed stellar masses (M _∗ ) and corrected the SFRs accordingly. Finally, we placed our sample on the star-forming main sequence, comparing its distribution with that of local star-forming galaxies. Two sources were further examined as potential giant low surface brightness galaxies (GLSBGs). Results . In our samples, we identify nine confirmed UDGs, eight other low surface-brightness galaxies (LSBGs), including one GLSBG. The z _spec of our field UDGs span a range of 0.015–0.037. Their r _eff ranges from 1.69–4.99 kpc, and their µ _{0 ,g} range from 24.05–24.98 mag arcsec ⁻² . Consistent with previous studies, the galaxies exhibit low to moderate dust content, with an average V-band attenuation of 0.29 mag . The spectroscopically confirmed UDGs presented in this work, after aperture correction, exhibit −2.95 ≤ log (SFR) ≤−1.60 M _⊙ yr ⁻¹ as an upper limit and 7.20 ≤ log (M _∗ ) ≤ 8.08 M _⊙ . These values place them along the star-forming main sequence. Conclusions . Our results indicate that blue field UDGs are characterised by heterogeneous dust attenuation and occupy the same region of the star formation–stellar mass plane as dwarf LSBGs. This is consistent with the broader population of local star-forming galaxies and indicates no distinct global star-forming behaviour.

Low-surface-brightness galaxies (LSBGs) are vital for understanding galaxy formation, but their diffuse nature makes them challenging to detect. Upcoming large-scale surveys are expected to uncover a large number of LSBGs, which require robust automated methods to identify them across heterogeneous datasets. As a precursor to the Legacy Survey of Space and Time (LSST) and , we explore domain adaptation techniques to build homogeneous LSBG catalogues across current surveys. Euclid We investigate the use of computer vision models and domain adaptation for cross-survey LSBG identification. Using models trained on Dark Energy Survey (DES) data, we search for LSBGs in the Kilo-Degree Survey Data Release 5 (KiDS DR5). We examine their structural and stellar population properties to pave the way for large-scale LSBG studies with LSST and . Euclid We used an ensemble consisting of one convolutional neural network (CNN) and two transformer models trained on DES cutouts and applied to KiDS DR5 imaging with surface-brightness normalisation. Structural parameters were estimated with galfitm and the sample was further refined through visual inspection to produce the final candidate sample. Photometric redshift and stellar population properties were estimated through spectral energy distribution (SED) fitting with CIGALE. We identified 20,180 LSBGs and 434 ultra-diffuse galaxies (UDGs) in KiDS DR5. Their structural parameters are similar to the known LSBGs from DES and Hyper Suprime-Cam SSP Survey (HSC-SSP). The KiDS-LSBGs follow a continuous size–luminosity relation connecting classical dwarf galaxies and UDGs and their colours are bimodal (∼73% blue, ∼27% red). Cross-matching with spectroscopic and cluster catalogues provided redshifts for 4,913 systems, enabling a systematic characterisation of the star-forming main sequence of LSBGs. Photometric redshifts derived via SED fitting are mildly overestimated (≃ 0.024), leading to systematic offsets in stellar mass and star formation rate estimates. However, these biases induce only small shifts (∼0.13-0.22 dex) in terms of the specific star formation rate, thereby preserving the structure of the star-forming main sequence. Strong environmental trends are also evident, with cluster LSBGs and UDGs exhibiting redder colours and reduced star formation compared to non-cluster systems. This indicates efficient quenching driven by environmental processes. We demonstrate that with domain adaptation, cross-survey LSBG identification can be achieved with deep learning models. Thus, the methodology presented here provides a powerful and scalable pathway for constructing homogeneous LSBG catalogues across surveys. This framework is well-suited for the era of LSST and , when millions of diffuse galaxies will be discovered and consistent cross-survey classifications will become essential. Euclid

Context. The James Webb Space Telescope (JWST) enables the identification of quiescent galaxies out to early epochs, offering a transformative view of their evolution. However, the photometric selection of quiescent galaxy candidates (QGCs) and the derivation of their key physical quantities, such as stellar masses (M _★ ) and dust attenuation, remain highly sensitive to the assumed star formation histories (SFHs), where dust–age degeneracies and modelling choices continue to stand as a major source of uncertainty. Aims. We aim to quantify how the inclusion of JWST/MIRI data and different SFH models impacts the selection and characterisation of QGCs. We test the robustness of the physical properties inferred from the spectral energy distribution (SED) fitting, such as M _★ , age, star formation rate (SFR), and dust attenuation (A _V ). We study how these parameters impact the quiescence criteria of the galaxies across cosmic time. Methods. We performed SED fittings for ∼13 000 galaxies at z ≤ 6 from the CEERS/MIRI fields with a ≤20 optical to mid-infrared broadband coverage. We implemented three SFH prescriptions: a flexible delayed model, a non-parametric model, and an extended regulator model. For each SFH, we compared the results obtained with and without MIRI photometry and dust emission models. We evaluated the impact of these configurations on both the number of QGCs, selected on the basis of their rest-frame UVJ colours, specific SFR, and main-sequence offset, as well as on their key physical properties, such as M _★ , A _V , and stellar ages. Results. The number of selected QGCs varied significantly, with the SFH ranging from 70 to 100 out of a mass-complete sample of ∼5000 galaxies, depending on the model. This number increased to 103–180 when MIRI data were included, driven by improved constraints on both dust attenuation and M _★ . We found a strong correlation between A _V and M _★ of QGCs at z ≤ 2.5, with massive galaxies (M _★ ∼ 10 ¹¹ M _⊙ ) shown to be ∼ 1.5 − 4 times more attenuated than low-mass galaxies (M _★ ∼ 10 ⁹ M _⊙ ). Regardless of the SFH, ∼13% of QGCs exhibit significant attenuation (A _V > 0.5) in support of recent JWST results on dust-rich quiescent galaxy candidates.

We employed deep learning to improve the photometric redshifts (photo-zs) in the Kilo-Degree Survey Data Release 4 bright galaxy sample (KiDS-DR4 Bright). This dataset, used as foreground for KiDS lensing and clustering studies, is flux-limited to r < 20 mag with mean z = 0.23 and covers 1000 deg². Its photo-zs were previously derived with artificial neural networks from the ANNz2 package trained on the Galaxy And Mass Assembly (GAMA) spectroscopy. Here, we considerably improve on these previous redshift estimations by building a deep learning model, Hybrid-z, that combines an inception-based convolutional neural network operating on four-band KiDS images with an artificial neural network using nine-band magnitudes from KiDS+VIKING. The Hybrid-z framework provides state-of-the-art photo-zs for KiDS-Bright with negligible mean residuals of O(10⁻⁴) and scatter at a level of 0.014(1 + z) – representing a reduction of 20% compared to the previous nine-band derivations with ANNz2. Our photo-zs are robust and stable independently of galaxy magnitude, redshift, and color. In fact, for blue galaxies, which typically have more pronounced morphological features, Hybrid-z provides a larger improvement over ANNz2 than for red galaxies. We checked our photo-z model performance on test data drawn from GAMA as well as from other KiDS-overlapping wide-angle spectroscopic surveys, namely SDSS, 2dFLenS, and 2dFGRS. We found stable behavior and consistent improvement over ANNz2 throughout. Finally, we applied Hybrid-z trained on GAMA to the entire KiDS-Bright DR4 sample of 1.2 million galaxies. For these final predictions, we designed a method of smoothing the input redshift distribution of the training set in order to avoid propagation of features present in GAMA related to its small sky area and large-scale structure imprint in its fields. Our work paves the way toward the best-possible photo-zs achievable with machine learning for any galaxy type for both the final KiDS-Bright DR5 data and for future deeper imaging, such as from the Legacy Survey of Space and Time.

Context. Low-surface-brightness galaxies (LSBGs) are important for understanding galaxy evolution and cosmological models. Nevertheless, the physical properties of these objects remain unknown, as even the detection of LSBGs can be challenging. Upcoming large-scale surveys are expected to uncover a large number of LSBGs, which will require accurate automated or machine learningbased methods for their detection. Aims. We study the scope of transfer learning for the identification of LSBGs. We used transformer models trained on Dark Energy Survey (DES) data to identify LSBGs from dedicated Hyper Suprime-Cam (HSC) observations of the Abell 194 cluster, which are two magnitudes deeper than DES. A new sample of LSBGs and ultra-diffuse galaxies (UDGs) around Abell 194 was compiled, and their properties were investigated. Methods. We used eight models, divided into two categories: LSBG Detection Transformer (LSBG DETR) and LSBG Vision Transformer (LSBG ViT). The data from DES and HSC were standardised based on the pixel-level surface brightness. We used an ensemble of four LSBG DETR models and another ensemble of four LSBG ViT models to detect LSBGs. This was followed by a singlecomponent Sérsic model fit and a final visual inspection to filter out potential false positives and improve sample purity. Results. We present a sample of 171 LSBGs in the Abell 194 cluster using HSC data, including 87 new discoveries. Of these, 159 were identified using transformer models, and 12 additional LSBGs were found through visual inspection. The transformer model achieves a true positive rate of 93% in HSC data without any fine-tuning. Among the LSBGs, 28 were classified as UDGs. The number of UDGs and the radial UDG number density suggests a linear relationship between UDG numbers and cluster mass on a log scale. The UDGs share similar Sérsic parameters with dwarf galaxies and occupy the extended end of the R_eff − M_g plane, suggesting they might be an extended sub-population of dwarf galaxies. We also found that LSBGs and UDGs near the cluster centre are brighter and redder than those in outer regions. Conclusions. We have demonstrated that transformer models trained on shallower surveys can be successfully applied to deeper surveys with appropriate data normalisation. This approach allows us to use existing data and apply the knowledge to upcoming and ongoing surveys, such as the Rubin Observatory Legacy Survey of Space and Time (LSST) and Euclid.

Context. Strong-lensing systems are expected to be discovered in great numbers by next-generation surveys. They provide a powerful tool for studying cosmology and the galaxy evolution. The coupling of the galaxy structure and cosmology through distance ratios means that it is essential for advancing both fields to examine the evolution of the lensing galaxy mass density profiles. Aims. We introduce a novel method that is independent of the dark energy assumed in the model to investigate the mass density slopes of lensing galaxies and their redshift evolution using an extended power-law (EPL) model. Methods. We adopted a nonparametric approach based on artificial neural networks trained on type Ia supernovae data to reconstruct the distance ratios of strong-lensing systems. These reconstructed ratios were compared with theoretical predictions to estimate the evolution of EPL model parameters. Results. A negative evolutionary trend of the mass density power-law exponent with increasing redshift is observed across different analysis levels. Assuming a triangular prior for the anisotropy of lensing galaxies, we find evidence for a redshift evolution of the mass density slope, quantified as ∂γ/∂z = −0.20 ± 0.12. Conclusions. This study confirms that the redshift evolution of the matter density slopes in lensing galaxies can be determined independent of dark energy models at the population level. The Legacy Survey of Space and Time (LSST) Rubin Observatory forecasts are expected to identify 100 000 strongly lensed galaxies. We show based on simulations with data from the LSST that spectroscopic follow up of just 10% of these systems can reduce the uncertainty in the redshift evolution coefficient of the total mass density slope (Δ∂γ/∂z) to 0.021. This precision would be able to distinguish between evolving and nonevolving scenarios for lensing galaxies.

Context.With the current and upcoming generation of surveys, such as the Legacy Survey of Space and Time (LSST) on theVera C. RubinObservatory and theEuclidmission, tens of billions of galaxies will be observed, with a significant portion (~10⁵) exhibiting lensing features. To effectively detect these rare objects amidst the vast number of galaxies, automated techniques such as machine learning are indispensable.Aims.We applied a state-of-the-art transformer algorithm to the 221 deg²of the Kilo Degree Survey (KiDS) to search for new strong gravitational lenses (SGLs).Methods.We tested four transformer encoders trained on simulated data from the Strong Lens Finding Challenge on KiDS data. The best performing model was fine-tuned on real images of SGL candidates identified in previous searches. To expand the dataset for fine-tuning, data augmentation techniques were employed, including rotation, flipping, transposition, and white noise injection. The network fine-tuned with rotated, flipped, and transposed images exhibited the best performance and was used to hunt for SGLs in the overlapping region of the Galaxy And Mass Assembly (GAMA) and KiDS surveys on galaxies up toz= 0.8. Candidate SGLs were matched with those from other surveys and examined using GAMA data to identify blended spectra resulting from the signal from multiple objects in a GAMA fiber.Results.Fine-tuning the transformer encoder to the KiDS data reduced the number of false positives by 70%. Additionally, applying the fine-tuned model to a sample of ~5 000 000 galaxies resulted in a list of ~51 000 SGL candidates. Upon visual inspection, this list was narrowed down to 231 candidates. Combined with the SGL candidates identified in the model testing, our final sample comprises 264 candidates, including 71 high-confidence SGLs; of these 71, 44 are new discoveries.Conclusions.We propose fine-tuning via real augmented images as a viable approach to mitigating false positives when transitioning from simulated lenses to real surveys. While our model shows improvement, it still does not achieve the same accuracy as previously proposed models trained directly on galaxy images from KiDS with added simulated lensing arcs. This suggests that a larger fine-tuning set is necessary for a competitive performance. Additionally, we provide a list of 121 false positives that exhibit features similar to lensed objects, which can be used in the training of future machine learning models in this field.

Context.Low-surface-brightness galaxies (LSBGs), which are defined as galaxies that are fainter than the night sky, play a crucial role in our understanding of galaxy evolution and in cosmological models. Upcoming large-scale surveys, such asRubinObservatory Legacy Survey of Space and Time andEuclid, are expected to observe billions of astronomical objects. In this context, using semiautomatic methods to identify LSBGs would be a highly challenging and time-consuming process, and automated or machine learning-based methods are needed to overcome this challenge.Aims.We study the use of transformer models in separating LSBGs from artefacts in the data from the Dark Energy Survey (DES) Data Release 1. Using the transformer models, we then search for new LSBGs from the DES that the previous searches may have missed. Properties of the newly found LSBGs are investigated, along with an analysis of the properties of the total LSBG sample in DES.Methods.We created eight different transformer models and used an ensemble of these eight models to identify LSBGs. This was followed by a single-component Sérsic model fit and a final visual inspection to filter out false positives.Results.Transformer models achieved an accuracy of ~94% in separating the LSBGs from artefacts. In addition, we identified 4083 new LSBGs in DES, adding an additional ~17% to the LSBGs already known in DES. This also increased the number density of LSBGs in DES to 5.5 deg⁻². The new LSBG sample consists of mainly blue and compact galaxies. We performed a clustering analysis of the LSBGs in DES using an angular two-point auto-correlation function and found that LSBGs cluster more strongly than their high-surface-brightness counterparts. This effect is driven by the red LSBG. We associated 1310 LSBGs with galaxy clusters and identified 317 ultradiffuse galaxies among them. We found that these cluster LSBGs are getting bluer and larger in size towards the edge of the clusters when compared with those in the centre.Conclusions.Transformer models have the potential to be equivalent to convolutional neural networks as state-of-the-art algorithms in analysing astronomical data. The significant number of LSBGs identified from the same dataset using a different algorithm highlights the substantial impact of our methodology on our capacity to discover LSBGs. The reported number density of LSBGs is only a lower estimate and can be expected to increase with the advent of surveys with better image quality and more advanced methodologies.