pith. sign in

arxiv: 2606.19447 · v1 · pith:7LVXIMOFnew · submitted 2026-06-17 · 🌌 astro-ph.GA · astro-ph.IM

Reimagining SED Fitting with Cosmological Galaxy Simulations and Machine Learning

Dhruv T. Zimmerman , Desika Narayanan This is my paper

Pith reviewed 2026-06-26 20:10 UTC · model grok-4.3

classification 🌌 astro-ph.GA astro-ph.IM
keywords SED fittingmachine learninggalaxy simulationsphotometrystellar massstar formation rateredshiftdust mass
0
0 comments X

The pith

Phot-Gal uses machine learning trained on simulated galaxy photometry to recover redshift, stellar mass, dust mass, and star formation rate more accurately than prospector on held-out test data.

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

The paper presents Phot-Gal as a machine learning alternative to traditional SED fitting, which collapses complex 3D galaxies into simple scalars and becomes computationally prohibitive for large surveys. Phot-Gal trains on photometry produced by 3D radiative transfer through cosmological simulations that include varied physics, then uses K-nearest neighbors imputation to handle any number of input bands and outputs predictions with uncertainties. On data held out from the same simulation suite, the model recovers the target properties with higher accuracy than the standard tool prospector across multiple metrics. The authors also break down which photometric bands drive the predictions, trace how each workflow step shapes the output posterior, and test generalization to data outside the training distribution.

Core claim

Phot-Gal solves the inverse SED fitting problem by training a machine learning model on photometry generated from 3D radiative transfer of simulated galaxies. The model accepts arbitrary input photometry via K-nearest neighbors imputation and predicts redshift, stellar mass, dust mass, and star formation rate together with uncertainties. When evaluated on a testing set drawn from the same simulation suite, Phot-Gal recovers the properties more accurately than prospector, though its uncertainty estimates are more likely to be miscalibrated when fewer photometric constraints are supplied. The workflow components are dissected to reveal physical motivations for the most influential inputs and t

What carries the argument

Phot-Gal, a machine learning model trained on radiative-transfer photometry from cosmological galaxy simulations that employs K-nearest neighbors imputation to process arbitrary numbers of input bands.

If this is right

  • Phot-Gal recovers redshift, stellar mass, dust mass, and star formation rate with higher accuracy than prospector on the simulation test set.
  • Uncertainty estimates from Phot-Gal are more likely to fail to reflect true offsets when fewer photometric bands are supplied.
  • Dissection of the model identifies physically justified photometric inputs that carry the most predictive weight.
  • Each step in the Phot-Gal workflow can be traced to its contribution to the output posterior.
  • The model demonstrates measurable generalization when applied to data drawn from outside the original training distribution.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the implemented simulation physics capture the dominant processes in real galaxies, Phot-Gal could be retrained on larger simulation volumes to serve upcoming wide-field surveys.
  • The imputation strategy suggests the same architecture could ingest heterogeneous multi-survey photometry without requiring uniform band coverage.
  • The observed degradation in uncertainty calibration with sparse inputs points to a natural next step of adding explicit uncertainty-aware loss terms during training.

Load-bearing premise

Performance measured on held-out galaxies from the same simulation suite reliably indicates how well the model will work on real observed galaxies.

What would settle it

A side-by-side comparison of Phot-Gal and prospector outputs against independent property measurements (for example, spectroscopic stellar masses or dust masses) on a sample of real galaxies with known redshifts.

Figures

Figures reproduced from arXiv: 2606.19447 by Desika Narayanan, Dhruv T. Zimmerman.

Figure 1
Figure 1. Figure 1: — Summary of fiducial physics Simba and IllustrisTNG galaxies’ physical relations at z = 0. Top left: The stellar mass - halo mass relation for simba and IllustrisTNG galaxies compared to Behroozi et al. (2013). Top middle: The stellar mass - gas mass relation for galaxies. Top right: The stellar mass - black hole mass relation for galaxies. Bottom left: the stellar mass - star formation rate relation for … view at source ↗
Figure 2
Figure 2. Figure 2: — Distribution of spectra for Simba and IllustrisTNG galaxies generated by 3D radiative transfer at z = 0. Solid lines represent the median emission at a particular wavelength amongst all the z = 0 spectra and the shaded regions represent the 16th and 84th percentiles of the distribution of the spectra. 10 4 10 2 10 1 10 0 10 1 flux (Jy) 10 5 10 6 z=0 spectrum GALEX Hubble SDSS JWST WISE Spitzer Hershel wa… view at source ↗
Figure 3
Figure 3. Figure 3: — Visual representation of the photometric filters we train Phot-Gal with for an example z = 0 spectrum. The left half plot is at a different scale than the right half to better present the filters visually up to ∼ 2 µm [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: — Demonstration of the performance of the KNN imputer that we include in Phot-Gal to handle missing data. This test evaluates whether the KNN imputer is successfully reconstructing the shape of the galaxy SEDs based on the available photometry. The KNN imputer is trained with access to the complete set of photometry, and then we take the ratio of the SED the imputer typically recovers with access to only t… view at source ↗
Figure 5
Figure 5. Figure 5: — Comparison of Phot-Gal and traditional SED fitting in recovering galaxy stellar mass and the output confidence intervals when all training filters and redshift are available as inputs. Our model recovers galaxy stellar mass better than traditional SED fitting and is also largely returning appropriate uncertainties. Top left: simple 1D comparison for accuracy for stellar mass with all trained filters avai… view at source ↗
Figure 6
Figure 6. Figure 6: — Similar to [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: — Similar to [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: — Comparison of the performance of Phot-Gal with EAZY-PY and prospector for recovering the true redshift with only JWST MIRI and NIRCam photometry available. 1.0 0.5 0.0 0.5 1.0 log (M*, inferred/M*,true) 0.0 0.5 1.0 1.5 2.0 Density Model Type SBI Model Prospector 8 9 10 11 12 log (M* )(M ) 1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00 lo g (M *, in fe rre d / M *,tr u e) Model Type SBI Model Prospector 0.0… view at source ↗
Figure 9
Figure 9. Figure 9: — Similar to [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: — Similar to [PITH_FULL_IMAGE:figures/full_fig_p017_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: — Similar to [PITH_FULL_IMAGE:figures/full_fig_p018_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: — A SHAP value plot for the NGBoost stellar mass inference model in Phot-Gal indicates that the rest-frame K band is primarily what Phot-Gal uses to predict stellar mass. Input features are ordered from highest to lowest for how much the ML model relies on the feature to recover the stellar mass. High absolute SHAP values mean that the property is more important to the output prediction. The shading for a… view at source ↗
Figure 13
Figure 13. Figure 13: — Similar to [PITH_FULL_IMAGE:figures/full_fig_p019_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: — Similar to [PITH_FULL_IMAGE:figures/full_fig_p020_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: — The NGBoost model output standard deviation dominates the contributions to the output posterior. We present a set of violin plots explaining the breakdown of uncertainty source contributions to Phot-Gal ’s final CI output for the stellar mass inference. From left to right, another source of uncertainty is added to the test to show the contribution of each of the components of Phot-Gal to the size of the… view at source ↗
Figure 16
Figure 16. Figure 16: — Phot-Gal performs the worst in recovering stellar mass on galaxies with stellar ages and masses in areas of parameter space that are sparsely populated in the training set. Red crosses indicate that Phot-Gal overestimates the true stellar mass and magenta crosses indicate that it underestimates the true stellar mass by at least 0.5 dex. The histogram reflects the properties of the galaxies that make up … view at source ↗
Figure 17
Figure 17. Figure 17: — Comparison of performance of fiducial models in Phot-Gal on the testing set against models only trained on Simba galaxies and tested on IllustrisTNG galaxies for recovering stellar mass. There are only minor differences in stellar mass and SFR in this case as those are results that these galaxy models are tuned on, whereas the assumptions for dust mass cause in the distributions in the top right panel t… view at source ↗
Figure 18
Figure 18. Figure 18: — Phot-Gal shows mixed results in recovering the ground truth of a dataset constructed on the spectra of Simba-C galaxies. Top left: 2D KDE plot of accuracy in recovering ground truth stellar mass as a function of stellar mass. Top right: 2D KDE plot of accuracy in recovering ground truth SFR as a function of stellar mass. Bottom left: 2D KDE plot of accuracy in recovering ground truth dust mass as a func… view at source ↗
Figure 19
Figure 19. Figure 19: — Comparison of the performance of the Phot-Gal with eazy-py and Prospector for recovering the true redshift with HST photometry. the performance is still reasonable. Additionally, the bias where the lowest mass galaxies are overpredicted and the highest mass galaxies are underpredicted has worsened. In the limiting ideal case in Section 4.1, we noted that there is a characteristic stellar mass for each r… view at source ↗
Figure 20
Figure 20. Figure 20: — Similar to [PITH_FULL_IMAGE:figures/full_fig_p031_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: — Similar to [PITH_FULL_IMAGE:figures/full_fig_p032_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: — Similar to [PITH_FULL_IMAGE:figures/full_fig_p033_22.png] view at source ↗
read the original abstract

SED fitting is the most common technique to recover galaxy physical properties from observed photometry. However, SED fitting requires many assumptions that essentially collapse a galaxy from a three-dimensional spatially varying object with complex structure into a scalar point. Moreover, modern inference techniques are computationally intensive, which presents a unique challenge in the era of extremely large datasets. We present \textsc{Phot-Gal}, a new galaxy SED modeling tool that solves the inverse problem of SED fitting by training a machine learning model on photometry generated from 3D radiative transfer of simulated galaxies with a wide range of implemented physics. \textsc{Phot-Gal} is designed to accept an arbitrary amount of input photometry by utilizing a $K$-nearest neighbors imputation strategy. Our fiducial model predicts redshift, stellar mass, dust mass, and star formation rate with uncertainties based on the provided input photometry. We evaluate the performance of \textsc{Phot-Gal} relative to the commonly-used SED fitting tool \textsc{prospector} in successfully recovering each of these properties with several metrics for the inferred values and uncertainties and find that it outperforms the accuracy of standard SED fitting software on the testing set. However, with fewer photometric constraints, \textsc{Phot-Gal} is more likely to have output uncertainties that do not reflect the offset from the ground truth. We dissect the components of \textsc{Phot-Gal} to find reasonable physical justifications for the photometry it relies on most, understand how each step in its workflow contributes to the eventual output posterior, and evaluate its ability to generalize to novel data.

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 / 2 minor

Summary. The paper introduces Phot-Gal, an ML-based SED fitting tool trained on photometry generated via 3D radiative transfer from cosmological galaxy simulations. It uses K-nearest neighbors imputation to handle arbitrary photometric inputs and predicts redshift, stellar mass, dust mass, and SFR with associated uncertainties. The central claim is that Phot-Gal outperforms the standard tool prospector in accuracy on a held-out test set drawn from the same simulation suite, while also examining component contributions, physical justifications, and generalization to novel data; the abstract notes that output uncertainties frequently fail to reflect true offsets under sparse photometric constraints.

Significance. If validated beyond simulations, the approach could enable faster inference on large photometric catalogs by directly learning from simulated physics rather than parametric assumptions. The work ships a reproducible ML pipeline and dissects feature importance, but the simulation-only test set means any accuracy gain is currently demonstrated only as fidelity to the training distribution rather than transfer to observations.

major comments (2)
  1. [Abstract] Abstract and testing-set evaluation: the claim that Phot-Gal 'outperforms the accuracy of standard SED fitting software' is supported only on held-out data from the identical cosmological simulation suite used for training; because both the ground-truth labels and the photometry-to-property mapping derive from the same implemented physics and radiative transfer, the reported metrics reduce to how well the model reproduces the simulation distribution rather than an external benchmark.
  2. [Abstract] Abstract: the paper states that generalization to novel data was examined, yet supplies no quantitative metrics (e.g., bias, scatter, or uncertainty calibration) on actual observed galaxies; this leaves the transferability of the reported accuracy advantage and uncertainty calibration untested, which is load-bearing for any claim of practical superiority over prospector for real SED fitting.
minor comments (2)
  1. The description of the K-nearest neighbors imputation strategy lacks detail on the choice of K and distance metric; these free parameters should be stated explicitly with sensitivity tests.
  2. Figure captions and axis labels should clarify whether error bars represent the model's reported uncertainties or the offset from ground truth.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their constructive comments. We address each major point below, with revisions where appropriate to improve clarity without altering the manuscript's core scope.

read point-by-point responses

  1. Referee: [Abstract] Abstract and testing-set evaluation: the claim that Phot-Gal 'outperforms the accuracy of standard SED fitting software' is supported only on held-out data from the identical cosmological simulation suite used for training; because both the ground-truth labels and the photometry-to-property mapping derive from the same implemented physics and radiative transfer, the reported metrics reduce to how well the model reproduces the simulation distribution rather than an external benchmark.

    Authors: We agree the evaluation uses held-out data from the training simulation suite, so metrics reflect fidelity to that distribution's physics and radiative transfer. This is explicitly a controlled test with known ground truth, allowing direct comparison to prospector on identical inputs. The abstract already qualifies the result as 'on the testing set.' We will revise the abstract and discussion to emphasize that outperformance is demonstrated within the simulation framework rather than claiming broader external validity. revision: yes

  2. Referee: [Abstract] Abstract: the paper states that generalization to novel data was examined, yet supplies no quantitative metrics (e.g., bias, scatter, or uncertainty calibration) on actual observed galaxies; this leaves the transferability of the reported accuracy advantage and uncertainty calibration untested, which is load-bearing for any claim of practical superiority over prospector for real SED fitting.

    Authors: The 'novel data' tests in the manuscript refer to held-out simulation galaxies with varied photometric coverage or subpopulations, not real observations. We acknowledge that no quantitative metrics (bias, scatter, calibration) are supplied for actual observed galaxies. This is a genuine limitation of the current work, which prioritizes validation where ground truth is known. We will add explicit clarification in the abstract and conclusions that transfer to observations remains unquantified and is left for future work. revision: yes

standing simulated objections not resolved
  • Provision of quantitative metrics on real observed galaxies, which would require new analysis outside the simulation-based scope of the present manuscript.

Circularity Check

0 steps flagged

No circularity: standard ML training and held-out simulation testing

full rationale

The paper trains Phot-Gal on photometry from cosmological simulations and evaluates recovery of redshift, stellar mass, dust mass, and SFR on a held-out test set drawn from the same simulation suite, comparing metrics against prospector. This is a conventional supervised learning workflow with no self-definitional reduction, no parameter fitted to a subset then renamed as a prediction of the same quantity, and no load-bearing self-citation or imported uniqueness theorem. The reported outperformance is an empirical result internal to the simulation framework; it does not reduce by construction to the inputs via any equation or ansatz. Generalization to real observations is noted as unquantified in the abstract but is an applicability question, not a circularity in the derivation or evaluation chain presented.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that the chosen cosmological simulations plus radiative transfer produce photometry sufficiently representative for training an invertible model; no free parameters or invented entities are explicitly introduced in the abstract.

free parameters (1)
  • K in KNN imputation
    Hyperparameter controlling how missing photometric bands are handled; value not stated in abstract.
axioms (1)
  • domain assumption Cosmological galaxy simulations with implemented physics and 3D radiative transfer generate photometry that can be used to train a model capable of recovering real galaxy properties from observations.
    This premise underpins the entire training data generation step described in the abstract.

pith-pipeline@v0.9.1-grok · 5810 in / 1357 out tokens · 32994 ms · 2026-06-26T20:10:31.412838+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

16 extracted references · 1 linked inside Pith

  1. [1]

    B. M. Tinsley, The Astrophysical Journal151, 547 (1968). H. Spinrad and B. J. Taylor, The Astrophysical Journal Supplement Series22, 445 (1971). S. M. Faber, Astronomy and Astrophysics20, 361 (1972). J. Walcher, B. Groves, T. Budav´ ari, and D. Dale, Astrophysics and Space Science331, 1 (2011). 26 C. Conroy, Annual Review of Astronomy and Astrophysics51, ...

    Pith/arXiv arXiv 1968

  2. [2]

    Speagle, The Astrophysical Journal876, 3 (2019). S. Lower, D. Narayanan, J. Leja, B. D. Johnson, C. Conroy, and R. Dav´ e, The Astrophysical Journal904, 33 (2020). B. Wang, J. Leja, H. Atek, R. Bezanson, E. Burnham, P. Dayal, R. Feldmann, J. E. Greene, B. D. Johnson, I. Labbe, M. V

    2019

  3. [3]

    Nanayakkara, S

    Maseda, T. Nanayakkara, S. H. Price, K. A. Suess, J. R. Weaver, and K. E. Whitaker, arXiv e-prints , arXiv:2504.15255 (2025). D. Narayanan, C. Conroy, R. Dav´ e, B. D. Johnson, and G. Popping, The Astrophysical Journal869, 70 (2018). S. Lower, D. Narayanan, J. Leja, B. D. Johnson, C. Conroy, and R. Dav´ e, The Astrophysical Journal931, 14 (2022). C. Hahn ...

    arXiv 2025

  4. [4]

    Ferrero, and A

    Ferreira, I. Ferrero, and A. Finoguenov, ˚ ap691, A175 (2024). 27 Euclid Collaboration, A. Humphrey, P. A. C. Cunha, L. Bisigello, C. Tortora, M. Bolzonella, L. Pozzetti, M. Baes, B. R. Granett, A. Amara, S. Andreon, N. Auricchio, C. Baccigalupi, M. Baldi, S. Bardelli, C. Bodendorf, D. Bonino, E. Branchini, M. Brescia, J. Brinchmann, S. Camera, V. Capobia...

    2024

  5. [5]

    Conselice, A. Enia, L. K. Hunt, P. Iglesias-Navarro, E. Iodice, J. H. Knapen, F. R. Marleau, O. Muller, R. F. Peletier, J. Roman, R. Ragusa, P. Salucci, T. Saifollahi, M. Scodeggio, M. Siudek, T. de Waele, A. Amara, S. Andreon, N. Auricchio, C. Baccigalupi, M. Baldi, S. Bardelli, P. Battaglia, R. Bender, C. Bodendorf, D. Bonino, W. Bon, E. Branchini, M. B...

    arXiv 2024

  6. [6]

    Bryan, The Open Journal of Astrophysics7, 54 (2024). C. C. Lovell, W. J. Roper, A. P. Vijayan, S. M. Wilkins, S. Newman, and L. Seeyave, The Open Journal of Astrophysics8, 152 (2025). W. Roper, C. Lovell, A. Vijayan, S. Wilkins, H. Akins, S. Berger, C. Sant Fournier, T. Harvey, K. Iyer, M. Leonardi, S. Newman, B. Pautasso, A. Perry, L. Seeyave, L. Sommovi...

    2024

  7. [7]

    Spergel, R. S. Somerville, R. Dave, A. Pillepich, L. Hernquist, D. Nelson, P. Torrey, D. Narayanan, Y. Li, O. Philcox, V. La Torre, A. Maria Delgado, S. Ho, S. Hassan, B. Burkhart, D. Wadekar, N. Battaglia, G. Contardo, and G. L. Bryan, The Astrophysical Journal915, 71 (2021). F. Villaescusa-Navarro, S. Genel, D. Angl´ es-Alc´ azar, L. A. Perez, P. Villan...

    2021

  8. [8]

    Hassan, E

    Mohammad, S. Hassan, E. Moser, E. T. Lau, L. F. Machado Poletti Valle, A. Nicola, L. Thiele, Y. Jo, O. H. E. Philcox, B. D. Oppenheimer, M. Tillman, C. Hahn, N. Kaushal, A. Pisani, M. Gebhardt, A. M. Delgado, J. Caliendo, C. Kreisch, K. W. K. Wong, W. R. Coulton, M. Eickenberg, G. Parimbelli, Y. Ni, U. P. Steinwandel, V. La Torre, R. Dave, N. Battaglia, D...

    2023

  9. [9]

    Regan, J

    Turk, J. Regan, J. H. Wise, H.-Y. Schive, T. Abel, A. Emerick, B. W. O’Shea, P. Anninos, C. B. Hummels, and S. Khochfar, Monthly Notices of the Royal Astronomical Society466, 2217 (2017), aDS Bibcode: 2017MNRAS.466.2217S. A. Rahmati, A. H. Pawlik, M. Raiˇ cevi´ c, and J. Schaye, Monthly Notices of the Royal Astronomical Society430, 2427 (2013). F. Haardt ...

    2017

  10. [10]

    Quataert, and N

    Hopkins, E. Quataert, and N. Murray, Monthly Notices of the Royal Astronomical Society470, 4698 (2017a). D. Angl´ es-Alc´ azar, R. Dav´ e, C.-A. Faucher-Gigu` ere, F.¨Ozel, and P. F. Hopkins, Monthly Notices of the Royal Astronomical Society464, 2840 (2017b). H. Bondi, Monthly Notices of the Royal Astronomical Society 112, 195 (1952). E. Choi, J. P. Ostri...

    1952

  11. [11]

    Peacock, S. Cole, P. Thomas, H. Couchman, A. Evrard, J. Colberg, and F. Pearce, Nature435, 629 (2005). D. Narayanan, M. J. Turk, T. Robitaille, A. J. Kelly, B. C

    2005

  12. [12]

    McClellan, R. S. Sharma, P. Garg, M. Abruzzo, E. Choi, C. Conroy, B. D. Johnson, B. Kimock, Q. Li, C. C. Lovell, S. Lower, G. C. Privon, J. Roberts, S. Sethuram, G. F. Snyder, R. Thompson, and J. H. Wise, The Astrophysical Journal Supplement Series252, 12 (2021), aDS Bibcode: 2021ApJS..252...12N. B. Paxton, L. Bildsten, A. Dotter, F. Herwig, P. Lesaffre, ...

    2021

  13. [13]

    Peletier, Monthly Notices of the Royal Astronomical Society 404, 1639 (2010), aDS Bibcode: 2010MNRAS.404.1639V. J. Falc´ on-Barroso, P. S´ anchez-Bl´ azquez, A. Vazdekis, E. Ricciardelli, N. Cardiel, A. J. Cenarro, J. Gorgas, and R. F

    2010

  14. [14]

    bd-j/sedpy: sedpy v0.2.0,

    Peletier, Astronomy and Astrophysics532, A95 (2011), aDS Bibcode: 2011A&A...532A..95F. A. Vazdekis, P. Coelho, S. Cassisi, E. Ricciardelli, J. Falc´ on-Barroso, P. S´ anchez-Bl´ azquez, F. La Barbera, M. A. Beasley, and A. Pietrinferni, Monthly Notices of the Royal Astronomical Society449, 1177 (2015), aDS Bibcode: 2015MNRAS.449.1177V. T. P. Robitaille, A...

    arXiv 2011

  15. [15]

    joshspeagle/dynesty: v3.0.0,

    Sheehan, Matt Pitkin, Matthew Kirk, Lu Xu, Joel Leja, and joezuntz, “joshspeagle/dynesty: v3.0.0,” (2025). G. B. Brammer, P. G. van Dokkum, and P. Coppi, The Astrophysical Journal686, 1503 (2008). L. S. Shapley, Proceedings of the National Academy of Sciences 39, 1095 (1953). B. T. Draine, D. A. Dale, G. Bendo, K. D. Gordon, J. D. T

    2025

  16. [16]

    Sheth, and H

    Hollenbach, K. Sheth, and H. I. Teplitz,\apj663, 866 (2007). R. C. Kennicutt, Jr.,\araa36, 189 (1998b). K. E. Whitaker, A. Pope, R. Cybulski, C. M. Casey, G. Popping, and M. S. Yun, The Astrophysical Journal850, 208 (2017), aDS Bibcode: 2017ApJ...850..208W. D. T. Zimmerman, D. Narayanan, K. E. Whitaker, and R. Dav´ e, \apj973, 146 (2024). R. T. Hough, D. ...

    2007