Bridging Simulations and Observations: Towards Next-Generation Studies of Dwarf Galaxy Structure with LSST

Hubble image of compact dwarf galaxy SBS 1415+437. Bridging Simulations and Observations: Towards Next-Generation Studies of Dwarf Galaxy Structure with LSST LSST@Europe7 2025/09/15 Garreth Martin*,Aaron Watkins, Yohan Dubois, Sugata Kaviraj, Duho Kim, Katarina Krajic, Ilin Lazar, Frazer Pearce, Sebastien Peirani, Christophe Pichon, Sukyoung Yi, Julien Devriendt, Adrianne Slyz *University of Nottingham ([email protected]) Martin et al 2025, NRAS 541 1831 [email protected] Lazar+(2024 a,b) Low-mass galaxies as Laboratories for Galaxy Evolution • Dwarfs do not appear to be solely an extension of high-mass populations • Some morphological features present in the high-mass regime extend to dwarf galaxies • But we also observe dwarfs with morphologies and structural properties only found in the low-mass regime (Lazar+2024a, 2024b) • Cosmological simulations are tuned to reproduce high mass galaxy populations, but not the low mass Universe, which is observationally incomplete Martin et al 2025, NRAS 541 1831 [email protected] Low-mass galaxies as Laboratories for Galaxy Evolution Watkins+2025 • Dwarfs do not appear to be solely an extension of high-mass populations • Some morphological features present in the high-mass regime extend to dwarf galaxies • But we also observe dwarfs with morphologies and structural properties only found in the low-mass regime (Lazar+2024a, 2024b) • Cosmological simulations are tuned to reproduce high mass galaxy populations, but not the low mass Universe, which is observationally incomplete Martin et al 2025, NRAS 541 1831 [email protected] Low-mass galaxies as Laboratories for Galaxy Evolution • Deep-wide imaging reveals that previous wide-area surveys (e.g. SDSS) missed many low-mass galaxies due to surface brightness limits. • e.g. “Ultra-diffuse galaxies” (van Dokkum+2015) highlight significant selection biases in past observations. • Biased towards the most star-forming objects (Kaviraj+2025) • Dwarf galaxies are very sensitive to feedback and environmental processes due to shallow potential wells • Stellar feedback • Indication that AGN play some role (e.g. Reines+2013,Kaviraj+2019,Bichang'a+2024) • Interactions with environment Martin et al 2025, NRAS 541 1831 [email protected] A New Era of Observations and Simulations Credit: Princeton University/HSC Project • Next-generation surveys (Rubin, Euclid, JWST) are revolutionizing our view of lowmass galaxies with unprecedented depth. • The COSMOS field (HSC-SSP) provides one of the deepest current datasets for studying faint dwarfs (μi(3σ, 10″×10″) > 31 mag arcsec-2). • Cosmological simulations (e.g. NewHorizon, TNG50, FIREbox) now resolve low-mass galaxies over relatively large volumes. • Forward modelling allows direct comparison between real and simulated galaxies. Martin et al 2025, NRAS 541 1831 [email protected] A New Era of Observations and Simulations Credit: Y. Dubois, H. Choi • Next-generation surveys (Rubin, Euclid, JWST) are revolutionizing our view of lowmass galaxies with unprecedented depth. • The COSMOS field (HSC-SSP) provides one of the deepest current datasets for studying faint dwarfs (μi(3σ, 10″×10″) > 31 mag arcsec-2). • Cosmological simulations (e.g. NewHorizon, TNG50, FIREbox) now resolve low-mass galaxies over relatively large volumes. • Forward modelling allows direct comparison between real and simulated galaxies. Martin et al 2025, NRAS 541 1831 [email protected] Simulations NewHorizon (Dubois+2021) TNG50 (Nelson+2019, Pillepich+2019) Code RAMSES (AMR) AREPO (moving mesh) Volume Zoom-in of 20 Mpc spherical region from Horizon-AGN 50 Mpc box Resolution ~1.3×10⁴ M☉(stars), ~34 pc (spatial) ~8.5×10⁴ M☉(stars/gas), 100–140 pc (spatial) Environment Coverage Field and group (max halo ~10¹³ M☉) Field, group & poor clusters (~10¹⁴ M☉) Star Formation Turbulence-regulated Schmidt law ISM Physics Multiphase ISM Idealised two-phase model SN Feedback Mechanical feedback from SN Type II (Kimm & Cen 2014) Direct heating + delayed kinetic winds (Springel & Hernquist 2003) Extras MHD Martin et al 2025, NRAS 541 1831 [email protected] Observations We use data from the COSMOS field, supported by deep Hyper SuprimeCam (HSC) imaging: •COSMOS2020 (Weaver+2021) • Provides stellar masses, redshifts, and rest-frame properties via comprehensive (40band) multiwavelength photometry. • Photometric redshifts reach <1% precision for bright sources. •HSC-SSP Imaging (Aihara+2019) • Deep i-band imaging (μ ≈ 31 mag arcsec⁻²) over the central 1.5° of COSMOS. • We use the DR2 deepCoadd images to preserve extended flux. • COSMOS probes relatively average environments, with a galaxy number density similar to TNG50 and NewHorizon volumes at 0.05 < z < 0.3. Martin et al 2025, NRAS 541 1831 [email protected] Connecting Observations and Theory • Galaxy morphology encodes key information about formation history, feedback, and environment. • Morphological comparisons between observations and simulations help test physical prescriptions. • In this work, we: • Generate realistic synthetic HSC-like images from TNG50 and NewHorizon. • Measure structural properties of COSMOS dwarf galaxies. • Compare structural diversity across observed and simulated samples. • Our aim: to assess how well current simulations reproduce the diversity of dwarf galaxy structure, and what this reveals about feedback and ISM physics. Martin et al 2025, NRAS 541 1831 [email protected] Martin et al 2025, NRAS 541 1831 [email protected] Mass Evolution Trends Differ • At high mass end (~109.5 M☉), simulations begin to converge toward observed values but still differ systematically •TNG50: • Strong increase in concentration and Sérsic index with mass. Large discrepancy with both NewHorizon and COSMOS •NewHorizon: • Remains somewhat too diffuse even at higher masses •COSMOS dwarfs: • Weak trends with mass — structural properties are relatively stable • Much better agreement has been shown at higher masses (e.g. Dubois 2021, Wang & Lilly 2023) •Highlights limitations in how feedback and star formation scale in simulations. Martin et al 2025, NRAS 541 1831 Feedback & ISM Physics Drive Divergence •TNG50: • Smooth ISM, continuous star formation, SN feedback, and MHD processes → retention of low AM gas • Promotes central gas accumulation → compact, concentrated structures • Insufficient angular momentum redistribution leads to overly compact galaxies •NewHorizon: • Multiphase ISM, bursty star formation, local SN feedback → low angular momentum gas ejected from central regions • Efficient redistribution of gas results in more diffuse, irregular galaxies •Impact of Feedback: •NewHorizon’s bursty SF leads to irregular morphologies and less compact structures •TNG50’s continuous SF results in smoother, more compact galaxies. Differences also influenced by resolution, PSF biases, and environment •Feedback and ISM models, not resolution or observational bias, are primary drivers. Martin et al 2025, NRAS 541 1831 [email protected] Summary Martin et al 2025, NRAS 541 1831 •Structural Differences: •Low-mass galaxies are highly sensitive to ISM, star formation, and feedback implementations • Reproducing global observables (e.g. stellar mass functions) isn't sufficient— resolved morphology adds crucial constraints especially given aparrent degeneracies between models in reproducing integrated properties like stellar mass (Wright+2024). •NewHorizon: Produces diffuse, extended galaxies with low concentration,burstier star formation •TNG50: Produces compact, concentrated galaxies with high central density,smoother star formation •Feedback and ISM Physics: •NewHorizon: Burstier SF, dynamic ISM leads to more asymmetric and less compact galaxies as low AM gas ejected efficiently •TNG50: Continuous SF with feedback uncoupled from the central parts of galaxies, smooth ISM results in more compact, concentrated structures •Discrepancy with Observations: • Both simulations show divergent trends compared to observed COSMOS dwarf galaxies, with neither fully capturing the observed structural diversity • Structural mismatch in dwarfs is a powerful diagnostic of sub-grid physics in simulations •Future Insights: •Next-generation surveys like LSST and Euclid will provide larger, deeper and higher-resolution datasets to constrain and refine simulations and better understand the physical mechanisms driving dwarf galaxy evolution. [email protected] Summary •Structural Differences: •Low-mass galaxies are highly sensitive to ISM, star formation, and feedback implementations • Reproducing global observables (e.g. stellar mass functions) isn't sufficient— resolved morphology adds crucial constraints especially given aparrent degeneracies between models in reproducing integrated properties like stellar mass (Wright+2024). •NewHorizon: Produces diffuse, extended galaxies with low concentration,burstier star formation •TNG50: Produces compact, concentrated galaxies with high central density,smoother star formation •Feedback and ISM Physics: •NewHorizon: Burstier SF, dynamic ISM leads to more asymmetric and less compact galaxies as low AM gas ejected efficiently •TNG50: Continuous SF with feedback uncoupled from the central parts of galaxies, smooth ISM results in more compact, concentrated structures •Discrepancy with Observations: • Both simulations show divergent trends compared to observed COSMOS dwarf galaxies, with neither fully capturing the observed structural diversity • Structural mismatch in dwarfs is a powerful diagnostic of sub-grid physics in simulations •Future Insights: • This pilot study is very limited by survey area. LSST will provide far wider datasets down to similar depth to constrain and refine simulations and better understand the physical mechanisms driving dwarf galaxy evolution. Martin et al 2025, NRAS 541 1831 [email protected] Additional slides Parameterize galaxy star formation history according to their formation time and level of burstiness NewHorizon galaxies have more bursty SFHs and formed earlier No evolution in SFH observed as a function of mass for TNG-50 dwarf galaxies Bursty SF Constant SF Formed early Formed late Star formation history Parameterize galaxy star formation history according to their formation time and level of burstiness NewHorizon galaxies have more bursty SFHs and formed earlier No evolution in SFH observed as a function of mass for TNG-50 dwarf galaxies We can measure the level of correlation between SFH and visual similarity Star formation history Partial correlations Correlation between the visual appearance of simulated galaxies and their star-formation histories is seen, even controlling for mass and environment. Understanding this link is key to understanding the differing dwarf galaxy properties between the two simulations Observe general correlation between more visually similar galaxies are more likely have similar star-formation histories When controlling for environment only TNG-50 shows a decrease in strength of correlation Correlation of morphology and SFH with local density disappears when restricted to less dense environments – dominated by internal processes in the field Star formation history vs visual similarity