pith. machine review for the scientific record. sign in

arxiv: 2604.16539 · v1 · submitted 2026-04-16 · 🌌 astro-ph.GA · astro-ph.IM

Recognition: unknown

Classifying Supermassive Black Hole Growth Regimes to Observables Across Cosmological Simulations with Forecasts for LSST

Hitaishi Chillara
Authors on Pith no claims yet

Pith reviewed 2026-05-10 09:48 UTC · model grok-4.3

classification 🌌 astro-ph.GA astro-ph.IM
keywords supermassive black holesLSSTmachine learningcosmological simulationsphotometryblack hole growthgalaxy colorsSMBH feedback
0
0 comments X

The pith

Machine learning on LSST photometry distinguishes over-massive from under-massive supermassive black hole growth regimes at 91 to 94 percent accuracy in simulations.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests whether the broadband colors recorded by the Legacy Survey of Space and Time can reveal whether a galaxy's central black hole is growing faster or slower than expected from the mass of its host galaxy. The authors convert three cosmological simulations into LSST filter bands, train an ensemble classifier on the resulting photometry, and obtain 91 to 94 percent accuracy inside individual simulations. When the same model is trained on one simulation and evaluated on another, accuracy remains 83 to 89 percent, showing that the color ordering between the two growth regimes survives changes in sub-grid feedback physics. Because LSST will deliver photometry for millions of galaxies that lack direct black-hole mass measurements, the classifier supplies a practical route to mapping post-seeding growth histories across large volumes.

Core claim

Forward-modeling SIMBA, IllustrisTNG, and EAGLE into LSST bands produces an ensemble machine-learning classifier that separates over-massive and under-massive SMBH growth regimes at 91-94 percent accuracy within SIMBA and IllustrisTNG. Cross-simulation transfer with rank-normalized features reaches 83-89 percent accuracy, indicating that the relative photometric ordering of the regimes is preserved across different sub-grid SMBH feedback prescriptions. Signal decomposition shows the separation is driven mainly by host-galaxy colors and the shape of the accretion-state spectral energy distribution rather than any direct inversion of the luminosity prescription.

What carries the argument

Ensemble machine-learning classifier trained on rank-normalized broadband photometry obtained by forward-modeling cosmological simulations into LSST bands

Load-bearing premise

The relative ordering of photometric signatures for over-massive versus under-massive SMBH growth regimes remains consistent across different sub-grid feedback prescriptions and will hold for real galaxies.

What would settle it

A large sample of galaxies with independent black-hole mass measurements from reverberation mapping or stellar dynamics whose LSST photometry yields growth-regime assignments that disagree with the mass-based labels at a rate well above 10 percent.

Figures

Figures reproduced from arXiv: 2604.16539 by Hitaishi Chillara.

Figure 1
Figure 1. Figure 1: Label-flip removal diagnostics. Top row: A distribution of offsets from the simulation-fitted reference relation for retained (filled) and removed (dashed) populations that are split by class: (blue for under-massive, red for over-massive). In each simulation the removed population is denser towards the nearer classification boundary than the retained population: Simba (|∆|rem ≈ 0.3 vs. |∆|ret ≈ 0.6 dex), … view at source ↗
Figure 2
Figure 2. Figure 2: Analysis of Simba’s black hole population at z ≥ 4. Top Left: Mass function showing downsizing behavior. Top Right: MBH-z relation colored by the Eddington ratio. Bottom: Number density evolution. limit tend to produce flatter, bluer SEDs that are readily distinguished from sub-Eddington populations. 4.4. Signal Decomposition (Circularity Tests) As our forward model sequentially links MBH → M˙ → Lbol → mag… view at source ↗
Figure 3
Figure 3. Figure 3: Simba black hole scaling relations at z ≥ 4 that follow M. Habouzit et al. (2021), with fit slopes of IllustrisTNG overlaid and dashed lines of local scaling relations [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of median MBH-M∗ relations from Simba (red) and IllustrisTNG (green) against the Trinity empirical model (blue; H. Zhang et al. 2023) at z = 4–7. IllustrisTNG is shown at z = 4–5, where sufficient black hole growth has occurred for a reliable median relation. Gold stars mark JWST-discovered AGN (H. Zhang et al. 2024). limited sample distributions. The ∼ 4–11% accu￾racy gap between cross-simulati… view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of forward model predictions with LSST DP1. Top: color-color diagrams (Simba predictions in red, DP1 in blue). Bottom: cumulative magnitude distributions. 87% (Test B), while shuffling accretion rates drops ac￾curacy by only 2–4% (Test A), indicating the signal is distributed across the composite SED rather than con￾centrated in a single feature [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Band-by-band comparison of Simba forward model predictions against DP1 observations restricted to z < 2, with median offsets and KS statistics annotated. The DP1 sample consists of 1,737 AGN candidates at z < 2 (58 with spectroscopic redshifts from SIMBAD cross-matching, 1,679 with photometric redshifts). The signal decomposition and attention heatmap studies confirm that the classification captures physi￾… view at source ↗
Figure 7
Figure 7. Figure 7: Redshift distribution of LSST DP1 AGN candidates from SIMBAD cross-match (N = 439, median z = 1.38) [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FT-Transformer attention heatmaps for Simba. The absolute magnitude proxy and spectral range receive the highest weights in both modes, indicating a consistent reliance on SED shape regardless of selection effects [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Simba classification in magnitude-limited mode. Top Left: ROC curves. Bottom Left: Color-color distribution by class. Bottom Right: MBH-M∗ relation mapping [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Simba classification in intrinsic mode. Similar to [PITH_FULL_IMAGE:figures/full_fig_p014_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Color distribution comparison between magnitude-limited (solid) and intrinsic (dashed) modes. While magnitude limits truncate the faint population, the color-space separation between under-massive (blue) and over-massive (red) classes is robustly preserved [PITH_FULL_IMAGE:figures/full_fig_p015_11.png] view at source ↗
read the original abstract

The possibility of over-massive black holes suggested by James Webb Space Telescope photometric discoveries of 'little red dots', may disfavor light supermassive black hole (SMBH) seeds. However, what should constitute the mass (range) of 'heavy' seeds remains relatively unconstrained. Moreover, Vera C Rubin Observatory's Legacy Survey of Space and Time will photometrically characterize galaxies without direct black hole mass measurements. We forward-model the SIMBA, IllustrisTNG, and EAGLE cosmological simulations into the photometric bands of LSST to train an ensemble machine learning classifier. Our framework achieves $91\%$--$94\%$ accuracy across SIMBA and IllustrisTNG in distinguishing between over-massive and under-massive SMBH growth regimes under LSST magnitude limits, using only broadband photometry. Furthermore, cross-simulation transfer experiments (training on one cosmological simulation and evaluating on another using rank-normalized features) achieve $83\%$--$89\%$ accuracy. This suggests the relative photometric ordering of growth regimes is largely preserved even across fundamentally different sub-grid SMBH feedback prescriptions. Signal decomposition shows our classification is driven by host galaxy colors ($82\%$--$87\%$ accuracy) and, relatedly, the accretion-state's spectral energy distribution shape as opposed to an inversion of our forward model's analytical luminosity prescription. Given that the evaluated simulations employ heavy seed prescriptions ($\geq 10^{4}~M_\odot$), our methodology establishes a validated baseline for classifying post-seeding growth regimes.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The manuscript develops an ensemble machine learning classifier trained on forward-modeled LSST broadband photometry from the SIMBA, IllustrisTNG, and EAGLE cosmological simulations to distinguish over-massive versus under-massive SMBH growth regimes. It reports 91–94% accuracy within SIMBA and IllustrisTNG, 83–89% accuracy in cross-simulation transfer using rank-normalized features, and shows via signal decomposition that host galaxy colors drive the classification (82–87% accuracy) rather than direct inversion of the analytical luminosity model. All evaluated simulations employ heavy seeds (≥10^4 M_⊙), and the work positions itself as a validated baseline for post-seeding regime classification from photometry alone.

Significance. If the reported accuracies and cross-simulation robustness hold, the result is significant for providing an observationally practical route to classify SMBH growth regimes in the large LSST photometric sample without requiring direct black-hole mass measurements. The preservation of relative photometric ordering across differing sub-grid feedback prescriptions is a useful finding given current uncertainties in SMBH modeling, and the color-driven decomposition aligns with known observational degeneracies between AGN and host properties. The use of multiple simulations and explicit focus on LSST magnitude limits are strengths that enhance the forecast value.

major comments (2)
  1. [§3] §3 (label definition): The over-massive and under-massive labels are defined internally via deviations from each simulation’s M_BH–M_* relation; the manuscript must state the precise threshold (e.g., number of sigma or percentile cut) and demonstrate that the reported accuracies are insensitive to reasonable variations in this cut, because the training labels are load-bearing for all accuracy claims.
  2. [§4.2] §4.2 (cross-simulation transfer): The 83–89% transfer accuracies are obtained with rank-normalized features, yet no table or figure shows the per-feature importance or the distribution of the dominant color features for over- versus under-massive samples across the three simulations; without this, the claim that “relative photometric ordering is largely preserved” lacks direct quantitative support.
minor comments (3)
  1. [Abstract] Abstract: EAGLE is listed among the simulations used for forward modeling, but accuracy numbers are quoted only for SIMBA and IllustrisTNG; the role of EAGLE (training, testing, or only forward modeling) should be stated explicitly.
  2. [Figures] Figure captions (e.g., Figure 2 or 3): Add the exact LSST bands employed and the magnitude limit applied when reporting the color-only accuracies.
  3. [Methods] Methods: The ensemble classifier details (base learners, hyper-parameter search, and cross-validation scheme) are only summarized; a short table of the final hyper-parameters and the number of independent runs used for the quoted accuracies would improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive review and recommendation for minor revision. We address each major comment point by point below.

read point-by-point responses

  1. Referee: [§3] §3 (label definition): The over-massive and under-massive labels are defined internally via deviations from each simulation’s M_BH–M_* relation; the manuscript must state the precise threshold (e.g., number of sigma or percentile cut) and demonstrate that the reported accuracies are insensitive to reasonable variations in this cut, because the training labels are load-bearing for all accuracy claims.

    Authors: We agree that the exact label threshold must be stated explicitly for reproducibility. In the revised manuscript we will add the precise definition used (deviation from the simulation-specific M_BH–M_* relation at fixed stellar mass) together with a sensitivity test showing that the quoted accuracies remain stable under reasonable variations of the cut (e.g., ±0.3 dex). revision: yes

  2. Referee: [§4.2] §4.2 (cross-simulation transfer): The 83–89% transfer accuracies are obtained with rank-normalized features, yet no table or figure shows the per-feature importance or the distribution of the dominant color features for over- versus under-massive samples across the three simulations; without this, the claim that “relative photometric ordering is largely preserved” lacks direct quantitative support.

    Authors: We acknowledge that direct quantitative support for the preservation of relative photometric ordering would strengthen the cross-simulation claim. In the revised manuscript we will add a table of ensemble feature importances for each simulation and a figure displaying the distributions of the dominant color features for the over-massive versus under-massive populations across SIMBA, IllustrisTNG, and EAGLE. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper defines over-massive and under-massive SMBH growth regimes from internal simulation quantities (e.g., BH mass relative to host properties) in SIMBA, IllustrisTNG, and EAGLE. It then forward-models broadband LSST photometry from the same simulations and trains an ensemble ML classifier whose reported accuracies (91-94% intra-simulation, 83-89% cross-simulation with rank-normalized features) are measured on held-out data or transferred simulations. These are standard empirical performance metrics, not reductions by construction. Cross-simulation transfer explicitly tests generalization across independent sub-grid feedback prescriptions rather than re-deriving the labels. Signal decomposition further isolates the role of host colors versus direct luminosity inversion, providing an internal control. No self-definitional loop, fitted parameter renamed as prediction, load-bearing self-citation, or smuggled ansatz appears in the methodology; the derivation chain remains self-contained against the simulation ground truth.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the assumption that the three cosmological simulations faithfully capture the photometric consequences of different SMBH growth regimes. No new physical entities are introduced; the work relies on existing simulation outputs and standard ML techniques.

axioms (2)
  • domain assumption The cosmological simulations (SIMBA, IllustrisTNG, EAGLE) produce realistic photometric signatures for over-massive and under-massive SMBH regimes under their respective sub-grid prescriptions.
    All training and transfer results depend on this assumption; the paper treats the simulations as ground truth for labeling.
  • domain assumption Forward modeling of simulation outputs into LSST bands introduces no systematic biases that would alter the relative ordering of growth regimes.
    The training data are created via this forward model.

pith-pipeline@v0.9.0 · 5578 in / 1496 out tokens · 66135 ms · 2026-05-10T09:48:50.333685+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

61 extracted references · 60 canonical work pages · 5 internal anchors

  1. [1]

    2017, MNRAS, 472, L109, doi: 10.1093/mnrasl/slx161

    Angl´ es-Alc´ azar, D., Faucher-Gigu` ere, C.-A., Quataert, E., et al. 2017, MNRAS, 472, L109, doi: 10.1093/mnrasl/slx161

    work page doi:10.1093/mnrasl/slx161 2017

  2. [2]

    T., Mellema, G., Pen, U.-L., et al

    Begelman, M. C., Volonteri, M., & Rees, M. J. 2006, MNRAS, 370, 289, doi: 10.1111/j.1365-2966.2006.10467.x

    work page doi:10.1111/j.1365-2966.2006.10467.x 2006

  3. [3]

    F., & de Jong , R

    Bell, E. F., & de Jong, R. S. 2001, ApJ, 550, 212, doi: 10.1086/319728

    work page doi:10.1086/319728 2001

  4. [4]

    K., Blecha, L., Torrey, P., et al

    Bhowmick, A. K., Blecha, L., Torrey, P., et al. 2025, MNRAS, 538, 518, doi: 10.1093/mnras/staf269

    work page doi:10.1093/mnras/staf269 2025

  5. [5]

    Bluck, A. F. L., Piotrowska, J. M., & Maiolino, R. 2023, ApJ, 944, 108, doi: 10.3847/1538-4357/acac7c Bogd´ an, A., Goulding, A. D., Natarajan, P., et al. 2024, Nature Astronomy, 8, 126, doi: 10.1038/s41550-023-02111-9

    work page doi:10.3847/1538-4357/acac7c 2023

  6. [6]

    Bromm, V., & Larson, R. B. 2004, ARA&A, 42, 79, doi: 10.1146/annurev.astro.42.053102.134034

    work page doi:10.1146/annurev.astro.42.053102.134034 2004

  7. [7]

    2003, ApJ, 596, 34, doi: 10.1086/377529

    Bromm, V., & Loeb, A. 2003, ApJ, 596, 34, doi: 10.1086/377529

    work page doi:10.1086/377529 2003

  8. [8]

    C., et al

    Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682, doi: 10.1086/308692

    work page internal anchor Pith review doi:10.1086/308692 2000

  9. [9]

    A., Schaye, J., Bower, R

    Crain, R. A., Schaye, J., Bower, R. G., et al. 2015, MNRAS, 450, 1937, doi: 10.1093/mnras/stv725 Dav´ e, R., Angl´ es-Alc´ azar, D., Narayanan, D., et al. 2019, MNRAS, 486, 2827, doi: 10.1093/mnras/stz937

    work page doi:10.1093/mnras/stv725 2015

  10. [10]

    and Volonteri, M

    Devecchi, B., & Volonteri, M. 2009, ApJ, 694, 302, doi: 10.1088/0004-637X/694/1/302

    work page doi:10.1088/0004-637x/694/1/302 2009

  11. [11]

    Dancing in the dark: galactic properties trace spin swings along the cosmic web

    Dubois, Y., Pichon, C., Welker, C., et al. 2014, MNRAS, 444, 1453, doi: 10.1093/mnras/stu1227

    work page Pith review doi:10.1093/mnras/stu1227 2014

  12. [12]

    Fan, X., Ba˜ nados, E., & Simcoe, R. A. 2023, ARA&A, 61, 373, doi: 10.1146/annurev-astro-052920-102455

    work page doi:10.1146/annurev-astro-052920-102455 2023

  13. [13]

    A., Becker, R

    Fan, X., Strauss, M. A., Becker, R. H., et al. 2006, AJ, 132, 117, doi: 10.1086/504836

    work page doi:10.1086/504836 2006

  14. [14]

    and Clayton, Geoffrey C

    Gordon, K. D., Clayton, G. C., Misselt, K. A., Landolt, A. U., & Wolff, M. J. 2003, ApJ, 594, 279, doi: 10.1086/376774

    work page doi:10.1086/376774 2003

  15. [15]

    2021, in Advances in Neural Information Processing

    Gorishniy, Y., Rubachev, I., Khrulkov, V., & Babenko, A. 2021, in Advances in Neural Information Processing

    2021

  16. [16]

    S., et al

    Habouzit, M., Li, Y., Somerville, R. S., et al. 2021, MNRAS, 503, 1940, doi: 10.1093/mnras/stab496

    work page doi:10.1093/mnras/stab496 2021

  17. [17]

    S., Li, Y., et al

    Habouzit, M., Somerville, R. S., Li, Y., et al. 2022, MNRAS, 509, 3015, doi: 10.1093/mnras/stab3147

    work page doi:10.1093/mnras/stab3147 2022

  18. [18]

    2023, ApJ, 959, 39, doi:10.3847/1538-4357/ad029e

    Harikane, Y., Zhang, Y., Nakajima, K., et al. 2023, ApJ, 959, 39, doi: 10.3847/1538-4357/ad029e

    work page doi:10.3847/1538-4357/ad029e 2023

  19. [19]

    2014, ApJ, 781, 60, doi: 10.1088/0004-637X/781/2/60

    Hirano, S., Hosokawa, T., Yoshida, N., et al. 2014, ApJ, 781, 60, doi: 10.1088/0004-637X/781/2/60

    work page doi:10.1088/0004-637x/781/2/60 2014

  20. [20]

    Hopkins, P. F. 2015, MNRAS, 450, 53, doi: 10.1093/mnras/stv195

    work page doi:10.1093/mnras/stv195 2015

  21. [21]

    M., Mellema, G., Alvarez, M

    Hopkins, P. F., & Quataert, E. 2011, MNRAS, 415, 1027, doi: 10.1111/j.1365-2966.2011.18542.x

    work page doi:10.1111/j.1365-2966.2011.18542.x 2011

  22. [22]

    2020, ARA&A, 58, 27, doi: 10.1146/annurev-astro-120419-014455 IRSA, & SSC

    Inayoshi, K., Visbal, E., & Haiman, Z. 2020, ARA&A, 58, 27, doi: 10.1146/annurev-astro-120419-014455 Ivezi´ c, v., Kahn, S. M., Tyson, J. A., et al. 2019, ApJ, 873, 111, doi: 10.3847/1538-4357/ab042c

    work page doi:10.1146/annurev-astro-120419-014455 2020

  23. [23]

    D., Onoue, M., Inayoshi, K., et al

    Kocevski, D. D., Onoue, M., Inayoshi, K., et al. 2023, ApJL, 954, L4, doi: 10.3847/2041-8213/ace5a0

    work page doi:10.3847/2041-8213/ace5a0 2023

  24. [24]

    Kormendy, J., & Ho, L. C. 2013, ARA&A, 51, 511, doi: 10.1146/annurev-astro-082708-101811

    work page internal anchor Pith review doi:10.1146/annurev-astro-082708-101811 2013

  25. [25]

    T., Mellema, G., Pen, U.-L., et al

    Lodato, G., & Natarajan, P. 2006, MNRAS, 371, 1813, doi: 10.1111/j.1365-2966.2006.10801.x

    work page doi:10.1111/j.1365-2966.2006.10801.x 2006

  26. [26]

    and Colpi, M

    Lupi, A., Colpi, M., Devecchi, B., Galanti, G., & Volonteri, M. 2014, MNRAS, 442, 3616, doi: 10.1093/mnras/stu1120

    work page doi:10.1093/mnras/stu1120 2014

  27. [27]

    2012, MNRAS, 427, 127, doi: 10.1111/j.1365-2966.2012.21948.x

    Lusso, E., Comastri, A., Simmons, B. D., et al. 2012, MNRAS, 425, 623, doi: 10.1111/j.1365-2966.2012.21513.x

    work page doi:10.1111/j.1365-2966.2012.21513.x 2012

  28. [28]

    2025, A&A, 702, A221, doi: 10.1051/0004-6361/202555571

    Ma, W., Cui, W., Dav´ e, R., Angl´ es-Alc´ azar, D., & Guo, H. 2025, A&A, 702, A221, doi: 10.1051/0004-6361/202555571

    work page doi:10.1051/0004-6361/202555571 2025

  29. [29]

    1995, ApJ, 441, 18, doi: 10.1086/175332 Ma´ ız Apell´ aniz, J

    Madau, P. 1995, ApJ, 441, 18, doi: 10.1086/175332

    work page doi:10.1086/175332 1995

  30. [30]

    Madau, P., & Rees, M. J. 2001, ApJ, 551, L27, doi: 10.1086/319848

    work page doi:10.1086/319848 2001

  31. [31]

    , keywords =

    Maiolino, R., Scholtz, J., Curtis-Lake, E., et al. 2024, A&A, 691, A145, doi: 10.1051/0004-6361/202347640

    work page doi:10.1051/0004-6361/202347640 2024

  32. [32]

    2018, MNRAS, 480, 5113, doi: 10.1093/mnras/sty2206

    Marinacci, F., Vogelsberger, M., Pakmor, R., et al. 2018, MNRAS, 480, 5113, doi: 10.1093/mnras/sty2206

    work page Pith review doi:10.1093/mnras/sty2206 2018

  33. [33]

    and Helly, J

    McAlpine, S., Helly, J., Schaller, M., et al. 2016, Astronomy and Computing, 15, 72, doi: 10.1016/j.ascom.2016.02.004 Milosavljevi´ c, M., Couch, S. M., & Bromm, V. 2009, ApJL, 696, L146, doi: 10.1088/0004-637X/696/2/L146

    work page doi:10.1016/j.ascom.2016.02.004 2016

  34. [34]

    P., Pillepich, A., Springel, V., et al

    Naiman, J. P., Pillepich, A., Springel, V., et al. 2018, MNRAS, 477, 1206, doi: 10.1093/mnras/sty618

    work page Pith review doi:10.1093/mnras/sty618 2018

  35. [35]

    2024, ApJL, 960, L1, doi:10.3847/2041-8213/ad0e76

    Natarajan, P., Pacucci, F., Ricarte, A., et al. 2024, ApJL, 960, L1, doi: 10.3847/2041-8213/ad0e76

    work page doi:10.3847/2041-8213/ad0e76 2024

  36. [36]

    First results from the IllustrisTNG simulations: the galaxy color bimodality

    Nelson, D., Pillepich, A., Springel, V., et al. 2018, MNRAS, 475, 624, doi: 10.1093/mnras/stx3040

    work page internal anchor Pith review doi:10.1093/mnras/stx3040 2018

  37. [37]

    2019, Computational Astrophysics and Cosmology, 6, 2, doi: 10.1186/s40668-019-0028-x

    Nelson, D., Springel, V., Pillepich, A., et al. 2019a, Computational Astrophysics and Cosmology, 6, 2, doi: 10.1186/s40668-019-0028-x

    work page doi:10.1186/s40668-019-0028-x

  38. [38]

    2019, MNRAS, 490, 3234, doi: 10.1093/mnras/stz2306

    Nelson, D., Pillepich, A., Springel, V., et al. 2019b, MNRAS, 490, 3234, doi: 10.1093/mnras/stz2306 19 NSF-DOE Vera C. Rubin Observatory. 2025, Legacy Survey of Space and Time Data Preview 1, doi: 10.71929/rubin/2570308

    work page doi:10.1093/mnras/stz2306 2025

  39. [39]

    2006, ApJ, 648, L5, doi: 10.1086/507678

    Franx, M. 2006, ApJ, 648, L5, doi: 10.1086/507678

    work page doi:10.1086/507678 2006

  40. [40]

    2024, ApJ, 964, 154, doi: 10.3847/1538-4357/ad3044

    Pacucci, F., & Loeb, A. 2024, ApJ, 964, 154, doi: 10.3847/1538-4357/ad3044

    work page doi:10.3847/1538-4357/ad3044 2024

  41. [41]

    2023, ApJL, 957, L3, doi:10.3847/2041-8213/ad0158

    Pacucci, F., Nguyen, B., Carniani, S., Maiolino, R., & Fan, X. 2023, ApJL, 957, L3, doi: 10.3847/2041-8213/ad0158

    work page doi:10.3847/2041-8213/ad0158 2023

  42. [42]

    2018b, MNRAS, 473, 4077, doi: 10.1093/mnras/stx2656

    Pillepich, A., Springel, V., Nelson, D., et al. 2018, MNRAS, 473, 4077, doi: 10.1093/mnras/stx2656

    work page internal anchor Pith review doi:10.1093/mnras/stx2656 2018

  43. [43]

    2019, MNRAS, 490, 3196, doi: 10.1093/mnras/stz2338

    Pillepich, A., Nelson, D., Springel, V., et al. 2019, MNRAS, 490, 3196, doi: 10.1093/mnras/stz2338

    work page doi:10.1093/mnras/stz2338 2019

  44. [44]

    2024, The Open Journal of Astrophysics, 7, doi: 10.33232/001c.123239

    Regan, J., & Volonteri, M. 2024, The Open Journal of Astrophysics, 7, doi: 10.33232/001c.123239

    work page doi:10.33232/001c.123239 2024

  45. [45]

    T., Fan, X., Newberg, H

    Richards, G. T., Fan, X., Newberg, H. J., et al. 2002, AJ, 123, 2945, doi: 10.1086/340187

    work page doi:10.1086/340187 2002

  46. [46]

    T., Lacy, M., Storrie-Lombardi, L

    Richards, G. T., Lacy, M., Storrie-Lombardi, L. J., et al. 2006, ApJS, 166, 470, doi: 10.1086/506525

    work page doi:10.1086/506525 2006

  47. [47]

    2021, MNRAS, 506, 613, doi:10.1093/mnras/stab1737

    Sassano, F., Schneider, R., Valiante, R., et al. 2021, MNRAS, 506, 613, doi: 10.1093/mnras/stab1737

    work page doi:10.1093/mnras/stab1737 2021

  48. [48]

    A., Bower, R

    Schaye, J., Crain, R. A., Bower, R. G., et al. 2015, MNRAS, 446, 521, doi: 10.1093/mnras/stu2058

    work page doi:10.1093/mnras/stu2058 2015

  49. [49]

    and Finkbeiner, Douglas P

    Schlafly, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103, doi: 10.1088/0004-637X/737/2/103

    work page internal anchor Pith review doi:10.1088/0004-637x/737/2/103 2011

  50. [50]

    2025, MNRAS, 541, 2070, doi: 10.1093/mnras/staf747

    Shankar, F., Bernardi, M., Roberts, D., et al. 2025, MNRAS, 541, 2070, doi: 10.1093/mnras/staf747

    work page doi:10.1093/mnras/staf747 2025

  51. [51]

    T., Strauss, M

    Shen, Y., Richards, G. T., Strauss, M. A., et al. 2011, ApJS, 194, 45, doi: 10.1088/0067-0049/194/2/45

    work page doi:10.1088/0067-0049/194/2/45 2011

  52. [52]

    2009, MNRAS, 398, 2122, doi: 10.1111/j.1365-2966.2009.15261.x

    Springel, V. 2010, MNRAS, 401, 791, doi: 10.1111/j.1365-2966.2009.15715.x

    work page doi:10.1111/j.1365-2966.2009.15715.x 2010

  53. [53]

    2018, MNRAS, 475, 676, doi: 10.1093/mnras/stx3304

    Springel, V., Pakmor, R., Pillepich, A., et al. 2018, MNRAS, 475, 676, doi: 10.1093/mnras/stx3304

    work page doi:10.1093/mnras/stx3304 2018

  54. [54]

    T., Strateva, I., Brandt, W

    Steffen, A. T., Strateva, I., Brandt, W. N., et al. 2006, AJ, 131, 2826, doi: 10.1086/503627

    work page doi:10.1086/503627 2006

  55. [55]

    2009, ApJ, 696, 1798, doi: 10.1088/0004-637X/696/2/1798 The Astropy Collaboration, Price-Whelan, A

    Tanaka, T., & Haiman, Z. 2009, ApJ, 696, 1798, doi: 10.1088/0004-637X/696/2/1798 The Astropy Collaboration, Price-Whelan, A. M., Lim, P. L., et al. 2022, ApJ, 935, 167, doi: 10.3847/1538-4357/ac7c74 Vanden Berk, D. E., Richards, G. T., Bauer, A., et al. 2001, AJ, 122, 549, doi: 10.1086/321167

    work page doi:10.1088/0004-637x/696/2/1798 2009

  56. [56]

    Vestergaard, M., & Peterson, B. M. 2006, ApJ, 641, 689, doi: 10.1086/500572

    work page doi:10.1086/500572 2006

  57. [57]

    , month = jul, year =

    Volonteri, M. 2010, Astronomy and Astrophysics Reviews, 18, 279, doi: 10.1007/s00159-010-0029-x

    work page doi:10.1007/s00159-010-0029-x 2010

  58. [58]

    2021, Nature Reviews Physics, 3, 732, doi: 10.1038/s42254-021-00364-9

    Volonteri, M., Habouzit, M., & Colpi, M. 2021, Nature Reviews Physics, 3, 732, doi: 10.1038/s42254-021-00364-9

    work page doi:10.1038/s42254-021-00364-9 2021

  59. [59]

    MNRAS , author =

    Weinberger, R., Springel, V., Pakmor, R., et al. 2018, MNRAS, 479, 4056, doi: 10.1093/mnras/sty1733

    work page doi:10.1093/mnras/sty1733 2018

  60. [60]

    Wolpert, D. H. 1992, Neural Networks, 5, 241, doi: 10.1016/S0893-6080(05)80023-1 Zhang , H., Behroozi, P., Volonteri, M., et al. 2023, MNRAS, 518, 2123, doi: 10.1093/mnras/stac2633

    work page doi:10.1016/s0893-6080(05)80023-1 1992

  61. [61]

    2024, MNRAS, 531, 4974, doi: 10.1093/mnras/stae1447

    Zhang, H., Behroozi, P., Volonteri, M., et al. 2024, MNRAS, 531, 4974, doi: 10.1093/mnras/stae1447

    work page doi:10.1093/mnras/stae1447 2024