Pith. sign in

REVIEW 3 major objections 6 minor 35 references

The effects of different cooling and heating function models on a simulated analog of NGC300

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The choice of how a simulation approximates gas cooling and heating—interpolation table versus machine learning—changes the simulated galaxy's thermal state and its C II emission rates by 10-20%.

desk verdict Controlled comparison of cooling-table vs ML emulator in a galaxy simulation, but the 5 Myr convergence check is too weak to support the headline phase-diagram offset. read the letter →

arxiv 2412.15324 v1 pith:ZZYJG5PG submitted 2024-12-19 astro-ph.GA

David Robinson , Camille Avestruz , Nickolay Y. Gnedin , Vadim A. Semenov This is my paper · ORCID
classification astro-ph.GA
keywords gascoolingandheatingfunctionsmachinelearningemulatorinterpolationtabletemperature-densityphasediagramCIIemissionisolatedgalaxysimulationNGC300analogXGBoost
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks whether the practical choice of how a simulation computes gas cooling and heating rates—rather than any change in astrophysical physics—can alter the galaxy that comes out of the simulation. It compares two approximations in an idealized isolated-galaxy simulation of an NGC300 analog: a standard polynomial interpolation table of Cloudy photoionization calculations, and a machine-learning (XGBoost) model trained on the same calculations, swapped one-for-one in the hydrodynamic code. The runs differ: the machine-learning run is systematically hotter in low-density gas, the two phase diagrams cross along a critical curve, and integrated C II emission rates shift by 10-20% for several collision channels. If these differences persist in more realistic settings, then approximation error alone is a source of uncertainty in simulated galaxy thermal states and in observable line-emission predictions.

What carries the argument

The load-bearing object is the one-to-one replacement of the Gnedin and Hollon (2012) polynomial interpolation table with XGBoost gradient-boosted tree models from Robinson et al. (2024), both fit to the same grid of Cloudy photoionization-equilibrium cooling and heating calculations at fixed metallicity $Z = 0.3\,Z_\odot$. The comparison machinery is the temperature-density phase diagram residual $\Delta = (m_{\rm GH12} - m_{\rm XGB})/(m_{\rm GH12} + m_{\rm XGB})$, whose zero set defines the critical curve, and the equilibrium-temperature crossings $\Gamma(T_{\rm equil}) = \Lambda(T_{\rm equil})$ that locate the curve relative to the two models' predicted cooling and heating functions.

What would settle it

Run the same two cooling-function models with radiative transfer and spatially varying metallicity for longer than 5 Myr; if the systematic temperature offset at low densities and the critical curve disappear or move dramatically, the claim that approximation choice alone reshapes the thermal state would not generalize.

Watch

Extended reading notes

Core claim

The paper establishes that the choice of cooling and heating function approximation, by itself, changes the thermal state of gas in a simulated galaxy. In the run using the machine-learning approximation, low-density gas with $-3 \lesssim \log(n_b/\mathrm{cm}^{-3}) \lesssim -1$ is systematically hotter than in the run using the interpolation table. The phase diagrams cross along a critical curve in temperature-density space where both runs hold equal gas mass, with the largest mass differences just above and below that curve. Integrated C II emission rates differ by 10-20% for some excitation channels. The authors state that the net cooling function determines gas temperature, and since these simulations tie star formation efficiency to velocity dispersion, the thermal differences could propagate to star formation and feedback.

Load-bearing premise

The comparison assumes that five million years is enough for the gas to reach its new steady state under the swapped cooling functions, and that turning off radiative transfer while fixing metallicity and photoionization rates isolates the approximation's effect.

Editorial extensions

If this is right

  • In the machine-learning run, low-density gas ($-3 \lesssim \log(n_b/\mathrm{cm}^{-3}) \lesssim -1$) is systematically hotter than in the interpolation-table run.
  • The two runs' phase diagrams cross along a critical curve where both simulations contain equal gas mass, and the largest mass differences sit just above and below that curve.
  • At a given density, the critical curve lies at a temperature between the equilibrium temperatures of the two cooling/heating models, so the offset is traceable to where each model balances cooling against heating.
  • Integrated C II emission rates differ by 10-20% for several excitation channels, meaning observable line predictions inherit the approximation choice.
  • Because the simulations tie star formation efficiency to the gas velocity dispersion, the thermal differences can propagate into star formation and feedback in longer or more realistic runs.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the machine-learning models are the more accurate of the two at fixed metallicity, then the low-density temperature offset implies that the interpolation table's errors—including occasional negative cooling predictions—are moving gas to cooler phases than the underlying Cloudy rates would produce.
  • The critical curve is a portable diagnostic: any new cooling/heating approximation could be located relative to an existing run by computing its equilibrium-temperature curve and checking where its phase diagram crosses old ones.
  • Because C II is a target for line-intensity mapping at redshifts 3-9, a 10-20% systematic in emission efficiency from the cooling approximation could be a relevant systematic for survey forecasts, although the paper does not quantify this.
  • The 20x slowdown of direct ML replacement suggests practical hybrid routes—precomputing tables from the ML models or distilling them into faster emulators—could capture the accuracy gain without the cost.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. This paper compares two approximations of the gas cooling and heating function---the GH12 interpolation table and the XGBoost surrogate models of Robinson et al. (2024), both trained on the same Cloudy calculations---by running two otherwise identical hydrodynamics simulations of an isolated NGC300-like galaxy, started from the same t ≈ 600 Myr snapshot of the Semenov et al. (2021) runs, with radiative transfer switched off, constant photoionization rates, and fixed metallicity Z = 0.3 Z_sun. After 5 Myr, the authors report that (i) the gas in the XGB run is systematically hotter at −3 ≲ log n_b ≲ −1, (ii) the phase-diagram mass residual between the runs has a well-defined 'critical curve' of equal gas mass with opposite-signed residuals just above and below it, and (iii) C II excitation-rate ratios r_j deviate from unity by 10–20% at low densities and up to ~40% at high densities. They attribute these differences directly to the approximation choice and connect the critical curve to the different radiative equilibrium temperatures of the two approximations.

Significance. If the reported differences are robust, the paper provides a clean, controlled demonstration that the numerical approximation of the cooling/heating function---not any change in input physics---can shift the predicted thermal state and line-emission rates of simulated gas. The experimental design is a genuine strength: identical initial snapshot, same code, and a single-variable replacement make the causal attribution to the approximation clean. The paper is also commendably explicit about the in-sample nature of the Table 2 accuracy comparison and about the idealized setup (no radiative transfer, fixed metallicity, rate ratios rather than luminosities), and it ships analysis code. The main quantitative claims, however, rest on one pair of runs and one 5 Myr snapshot, and the convergence and significance support provided is not yet strong enough to establish that the reported differences are steady-state consequences of the approximation rather than transient or noise-dominated features.

major comments (3)
  1. [Section 2.1 / Appendix A (Eq. A1, Fig. 5)] The convergence argument in Appendix A does not establish that the 5 Myr snapshot reflects a settled response to the replaced cooling and heating functions, so the central attribution claim in Sections 3.1–3.2 is not yet supported. Three specific problems: (i) the statement that the denominator of Eq. (A1) 'ensure[s] that −1 ≤ δ ≤ 1' is false whenever Δ_a and Δ_b have opposite signs; for example, Δ_a = 0.2 with Δ_b = −0.1 gives δ = −3, and Δ_b = −Δ_a makes the denominator vanish, so bins in Fig. 5 that flip sign between snapshots would saturate or overflow the color scale. (ii) An unstructured map of δ measures only the absence of spatial correlation in the incremental change between consecutive snapshots; a slowly drifting residual with uncorrelated per-bin increments would look the same, so 'no structure' is not a stationarity test. (iii) The paper offers no physical timescale argument: for the gas at −3 ≲ log n_b ≲ −1 and T ~ 10^3–10^4 K where the headline temperature offset is claimed, the radiative relaxation time t_therm = (3/2)k_B T / (n_b |Γ − Λ|), with net cooling rates of order 10^-26–10^-24 erg cm^3 s^-1 implied by the Cloudy-based models in this regime, is roughly 10^6–10^9 yr, comparable to or far longer than the 5 Myr elapsed. The 'systematically hotter' XGB gas and the critical curve at low density may therefore be transient features carried over from the common starting snapshot. The authors should either extend the runs to several ISM dynamical times and show the residual map stops evolving, tabulate per-bin t_therm estimates demonstrating 5 Myr ≫ t_therm in the affected density range, or explicitly rescope the conclusions to short-term differences.
  2. [Section 3.1 / Figs. 1–2] The phase-diagram structures that carry the paper's claims, namely the low-density median-temperature offset, the critical curve, and the residual bands around it, are presented without any estimate of their statistical significance. The analysis uses a single pair of runs and one snapshot per run; regions of Fig. 2 are described as 'noise with no clear structure,' but no information is given about how many resolution elements or how much gas mass sit in the bins that define the critical curve, so the zero-crossing contour could be partly a sampling artifact. Similarly, the claim that the XGB run is 'systematically hotter' for −3 ≲ log n_b ≲ −1 is based on the median curves, yet the paper does not report whether the median offset exceeds the inter-run percentile overlap in that density range, a check it does apply to the higher-density range (−1 ≲ log n_b ≲ 1). I recommend reporting the gas mass or cell count per phase-diagram bin and/or bootstrapping the residual maps over cells, so the reader can assess whether the critical curve and the temperature offset are significant rather than consistent with Poisson fluctuations.
  3. [Section 3.2 / Eq. (3)] The interpretation of the C II comparison is slightly stronger than the calculation supports. The ratio r_j in Eq. (3) is a ratio of emission rates per C II ion, computed with the same temperature-dependent rate coefficients in both runs and with the C II abundance implicitly held fixed. Because the two runs differ in thermal state, and because the cooling/heating models encode different ionization states from the same Cloudy training data, the C II abundance distribution is not guaranteed to be the same in both runs. The measured 10–20% variations in r_j are well defined as stated, but the closing sentence of Section 3.2 ('we would expect the actual C II luminosity to be different') should be qualified: an actual luminosity comparison requires the ionic abundance field, which is not tracked in these runs.
minor comments (6)
  1. [Eq. (3), Section 2.3] The integrals in Eq. (3) are written over dT alone, but r_j is plotted as a function of n_b in Fig. 4; the density conditioning should be written explicitly (for example, the phase-diagram distribution evaluated at fixed n_b and integrated over T).
  2. [Abstract and Section 2.2] There are several wording and typographical issues, including 'uses machine learning for the interpolation instead on an analytic function' in the abstract and 'The machine learning models in of Robinson et al. (2024)' in Section 2.2; these should be corrected.
  3. [Table 2, Section 2.2] The in-sample caveat for the mean-squared-error comparison appears in the text but not in the Table 2 caption; moving it to the caption would prevent readers from treating Table 2 as an out-of-sample accuracy claim.
  4. [Appendix A, Fig. 5] Bins in Fig. 5 where Δ_a and Δ_b have opposite signs can produce |δ| > 1 or divergent values (see major comment 1); the paper should state explicitly how such bins are handled in the color scale of the δ maps.
  5. [Section 3.1, Fig. 3] The equilibrium-temperature explanation for the critical curve assumes that gas at a given density is radiatively relaxed; in a turbulent disk most gas is not at radiative equilibrium, so this mechanism should be labeled as a heuristic interpretation rather than a derived result.
  6. [Acknowledgments] The repository URL in the Acknowledgments ('ngc300 analysis') contains a space and is not a valid URL; the correct link should be provided.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the simulation outputs are emergent, and the methods compared are independent emulators of a common Cloudy training set.

full rationale

The paper's central comparison—GH12 polynomial table interpolation versus XGBoost machine-learning emulators—is a controlled numerical experiment. Both approximations are evaluated on the same Cloudy calculations, but the downstream quantities (temperature-density phase diagrams, residual maps, critical curves, and CII emission ratios) are outputs of full hydrodynamic simulations, not fitted parameters. The only in-sample accuracy metric (Table 2) is explicitly acknowledged by the authors as not predictive of performance on new data, and it does not enter the simulation analysis. The interpretation in Fig. 3, relating the critical curve to equilibrium temperatures of the two approximation functions, is a post-hoc explanation rather than an input to the simulation. Self-citations to Gnedin & Hollon (2012), Robinson et al. (2024), and Semenov et al. (2021) supply the compared methods and the initial snapshot, but they are not used to force the result; the comparison would be meaningful even if those works were by other authors. The Appendix A convergence check is an empirical assessment, not a circular argument. No equation reduces to its own input, and no fitted parameter is renamed as a prediction. Accordingly, no circular step is present.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

No new scalar constants are fitted in this paper. The XGBoost and GH12 models are fixed inputs from the cited literature, both derived from the same Cloudy calculations. The main chosen inputs are the constant photoionization rates, the fixed metallicity, and the short run duration, which are simplifications rather than fitted parameters. The critical curve is a descriptive feature of the comparison, not an invented physical entity.

free parameters (2)
  • Photoionization and photodissociation rates Q_LW, Q_HI, Q_HeI, Q_CVI = 2e-11, 2e-17, 3e-16, 9e-18 s^-1
    Spatially constant ISM averages taken from the fiducial Semenov et al. (2021) run with radiative transfer; chosen for computational convenience and fixed in both simulation runs (Table 1).
  • Gas metallicity Z = 0.3 Z_sun
    Chosen because it is one of the metallicities with exact Cloudy calculations, avoiding the shared quadratic metallicity interpolation between the two approximations (Section 2.2).
assumptions (5)
  • domain assumption Cloudy photoionization calculations provide the ground truth cooling and heating functions for both approximations.
    Section 2.2 states that GH12 interpolates exact Cloudy calculations and Robinson et al. (2024) trained on the same Cloudy calculations; this paper compares approximation schemes, not the accuracy of Cloudy physics.
  • domain assumption Replacing the GH12 table with XGBoost models is a valid one-to-one substitution at fixed metallicity.
    Section 2.2 asserts the inputs (T, nH, Q rates) and outputs (cooling and heating functions) at fixed Z are identical, so differences are attributed to the approximation method rather than to implementation details.
  • domain assumption Running without radiative transfer and with spatially constant photoionization rates isolates the effect of the cooling and heating function approximation.
    Section 2.2 and Table 1 turn off radiative transfer and use ISM-averaged rates; a spatially varying radiation field could change where and how the two approximations differ.
  • domain assumption Five megayears is sufficient for the temperature-density phase diagram to converge after the cooling function swap.
    Appendix A compares 1-2 Myr and 4-5 Myr residual maps and concludes convergence; longer evolution or a denser snapshot cadence could reveal additional evolution.
  • domain assumption The t=600 Myr snapshot is a settled common starting point for both runs.
    Section 2.1 initializes both runs from the same fiducial snapshot after physical processes were added in stages; if the gas retains thermal memory of the GH12 cooling history, the first few Myr may not be purely representative.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The effects of different cooling and heating function models on a simulated analog of NGC300." pith.science (2026). https://pith.science/paper/ZZYJG5PG

@misc{pith2026241215324,
  author       = {Pith},
  title        = {Pith review of: The effects of different cooling and heating function models on a simulated analog of NGC300},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZZYJG5PG}},
  note         = {Machine review of arXiv:2412.15324}
}
abstract

Gas cooling and heating rates are vital components of hydrodynamic simulations. However, they are computationally expensive to evaluate exactly with chemical networks or photoionization codes. We compare two different approximation schemes for gas cooling and heating in an idealized simulation of an isolated galaxy. One approximation is based on a polynomial interpolation of a table of Cloudy calculations, as is commonly done in galaxy formation simulations. The other approximation scheme uses machine learning for the interpolation instead on an analytic function, with improved accuracy. We compare the temperature-density phase diagrams of gas from each simulation run to assess how much the two simulation runs differ. Gas in the simulation using the machine learning approximation is systematically hotter for low-density gas with $-3 \lesssim \log{(n_b/\mathrm{cm}^{-3})} \lesssim -1$. We find a critical curve in the phase diagram where the two simulations have equal amounts of gas. The phase diagrams differ most strongly at temperatures just above and below this critical curve. We compare CII emission rates for collisions with various particles (integrated over the gas distribution function), and find slight differences between the two simulations. Future comparisons with simulations including radiative transfer will be necessary to compare observable quantities like the total CII luminosity.

Figures

Figures reproduced from arXiv: 2412.15324 by the authors.

Figure 2
Figure 2. shows that there is a ‘critical curve’ where ∆ = 0, indicating that the two simulation runs have identical gas mass in the bins along this curve. The critical curve runs from temperatures of T ∼ 104 K at a density of 5.0 2.5 0.0 2.5 log (nb [cm 3 ]) 2 4 6 8 lo g (T [K]) 1.0 0.5 0.0 0.5 1.0 mGH12 mXGB mGH12 + mXGB [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. — Predicted cooling (blue curves) and heating functions (red curves) for the GH12 interpolation table (Gnedin and Hollon 2012, solid curves) and XGBoost machine learning models (Robin￾son et al. 2024, dashed curves) at nb = 1 cm−3 . The cooling and heating curves for a given model intersect at the equilibrium tem￾perature. In the lower panel, we show the overall density profile for the GH12 simulation run (the profi… view at source ↗
Figure 4
Figure 4. — The C II emission ratio rj , as defined in Equation (3) as a function of baryon number density nb (upper panel) exicted by interactions with electrons (solid blue), atomic hydrogen or helium (dashed orange), CMB photons (dash-dotted green), and molecular hydrogen with para (dashed red) and ortho (solid purple) spins. For comparison, the bottom panel shows the overall baryon number density profile for the GH12 simu… view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: — Residual between residuals δ1 Myr,2 Myr (left, see Equation (A1)) and δ4 Myr,5 Myr (right) [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 15 canonical work pages

  1. [1]

    Spitzer , Lyman ,\ 10.1086/146200 journal journal \ volume 124 ,\ pages 20 ( year 1956 ) NoStop

    author author J. Spitzer , Lyman ,\ 10.1086/146200 journal journal \ volume 124 ,\ pages 20 ( year 1956 ) NoStop

    doi:10.1086/146200 1956

  2. [2]

    author author H. I. \ Ewen \ and\ author E. M. \ Purcell ,\ 10.1038/168356a0 journal journal \ volume 168 ,\ pages 356 ( year 1951 ) NoStop

    doi:10.1038/168356a0 1951

  3. [3]

    author author R. M. \ Hjellming , author C. P. \ Gordon , \ and\ author K. J. \ Gordon ,\ @noop journal journal \ volume 2 ,\ pages 202 ( year 1969 ) NoStop

    work page 1969

  4. [4]

    author author A. C. \ Cheung , author D. M. \ Rank , author C. H. \ Townes , author D. D. \ Thornton , \ and\ author W. J. \ Welch ,\ 10.1103/PhysRevLett.21.1701 journal journal \ volume 21 ,\ pages 1701 ( year 1968 ) NoStop

    doi:10.1103/physrevlett.21.1701 1968

  5. [5]

    author author C. F. \ McKee \ and\ author J. P. \ Ostriker ,\ 10.1086/155667 journal journal \ volume 218 ,\ pages 148 ( year 1977 ) NoStop

    doi:10.1086/155667 1977

  6. [6]

    author author C. Dobbs ,\ 10.3389/fspas.2023.1272771 journal journal Frontiers in Astronomy and Space Sciences \ volume 10 ,\ eid 1272771 ( year 2023 ) ,\ http://arxiv.org/abs/2312.01854 arXiv:2312.01854 [astro-ph.GA] NoStop

    work page arXiv 2023

  7. [7]

    Hennebelle \ and\ author M

    author author P. Hennebelle \ and\ author M. Y. \ Grudi \'c ,\ 10.48550/arXiv.2404.07301 journal journal arXiv e-prints \ ,\ eid arXiv:2404.07301 ( year 2024 ) ,\ http://arxiv.org/abs/2404.07301 arXiv:2404.07301 [astro-ph.GA] NoStop

    arXiv doi:10.48550/arxiv.2404.07301 2024

  8. [8]

    Photochemistry and Heating/Cooling of the Multiphase Interstellar Medium with UV Radiative Transfer for Magnetohydrodynamic Simulations

    author author J.-G. \ Kim , author M. Gong , author C.-G. \ Kim , \ and\ author E. C. \ Ostriker ,\ 10.3847/1538-4365/ac9b1d journal journal \ volume 264 ,\ eid 10 ( year 2023 ) ,\ http://arxiv.org/abs/2210.08024 arXiv:2210.08024 [astro-ph.GA] NoStop

    work page Pith review arXiv 2023

Show all 35 references
  1. [9]

    Scannapieco , author M

    author author C. Scannapieco , author M. Wadepuhl , author O. H. \ Parry , author J. F. \ Navarro , author A. Jenkins , author V. Springel , author R. Teyssier , author E. Carlson , author H. M. P. \ Couchman , author R. A. \ Crain , author C. Dalla Vecchia , author C. S. \ Fr...

  2. [10]

    \ Kim , author T

    author author J.-h. \ Kim , author T. Abel , author O. Agertz , author G. L. \ Bryan , author D. Ceverino , author C. Christensen , author C. Conroy , author A. Dekel , author N. Y. \ Gnedin , author N. J. \ Goldbaum , author J. Guedes , author O. Hahn , author A. Hobbs , auth...

  3. [11]

    Monaco , author A

    author author P. Monaco , author A. J. \ Benson , author G. De Lucia , author F. Fontanot , author S. Borgani , \ and\ author M. Boylan-Kolchin ,\ 10.1093/mnras/stu655 journal journal \ volume 441 ,\ pages 2058 ( year 2014 ) ,\ http://arxiv.org/abs/1404.0811 arXiv:1404.0811 [a...

    2014 arXiv

  4. [12]

    Hou , author C

    author author J. Hou , author C. G. \ Lacey , \ and\ author C. S. \ Frenk ,\ 10.1093/mnras/stx3218 journal journal \ volume 475 ,\ pages 543 ( year 2018 ) ,\ http://arxiv.org/abs/1708.02950 arXiv:1708.02950 [astro-ph.GA] NoStop

    2018 arXiv

  5. [13]

    Hou , author C

    author author J. Hou , author C. G. \ Lacey , \ and\ author C. S. \ Frenk ,\ 10.1093/mnras/stz730 journal journal \ volume 486 ,\ pages 1691 ( year 2019 ) ,\ http://arxiv.org/abs/1803.01923 arXiv:1803.01923 [astro-ph.GA] NoStop

    2019 arXiv

  6. [14]

    author author A. J. \ Richings \ and\ author J. Schaye ,\ 10.1093/mnras/stw327 journal journal \ volume 458 ,\ pages 270 ( year 2016 ) ,\ http://arxiv.org/abs/1506.08829 arXiv:1506.08829 [astro-ph.GA] NoStop

    2016 arXiv

  7. [15]

    author author P. R. \ Capelo , author S. Bovino , author A. Lupi , author D. R. G. \ Schleicher , \ and\ author T. Grassi ,\ 10.1093/mnras/stx3355 journal journal \ volume 475 ,\ pages 3283 ( year 2018 ) ,\ http://arxiv.org/abs/1710.01302 arXiv:1710.01302 [astro-ph.GA] NoStop

    2018 arXiv

  8. [16]

    author author V. A. \ Semenov , author A. V. \ Kravtsov , \ and\ author N. Y. \ Gnedin ,\ 10.3847/1538-4357/ac0a77 journal journal \ volume 918 ,\ eid 13 ( year 2021 ) ,\ http://arxiv.org/abs/2103.13406 arXiv:2103.13406 [astro-ph.GA] NoStop

    2021 arXiv

  9. [17]

    author author N. Y. \ Gnedin \ and\ author N. Hollon ,\ 10.1088/0067-0049/202/2/13 journal journal \ volume 202 ,\ eid 13 ( year 2012 ) ,\ http://arxiv.org/abs/1201.5116 arXiv:1201.5116 [astro-ph.CO] NoStop

    2012 arXiv

  10. [18]

    Robinson , author C

    author author D. Robinson , author C. Avestruz , \ and\ author N. Y. \ Gnedin ,\ 10.1093/mnras/stad3880 journal journal \ volume 528 ,\ pages 255 ( year 2024 ) ,\ http://arxiv.org/abs/2310.09328 arXiv:2310.09328 [astro-ph.CO] NoStop

    2024 arXiv

  11. [19]

    Zhao , author J

    author author Y. Zhao , author J. Liu , author Z.-Y. \ Zhang , \ and\ author T. G. \ Bisbas ,\ 10.48550/arXiv.2410.20684 journal journal arXiv e-prints \ ,\ eid arXiv:2410.20684 ( year 2024 ) ,\ http://arxiv.org/abs/2410.20684 arXiv:2410.20684 [astro-ph.GA] NoStop

    2024 doi arXiv

  12. [20]

    Casavecchia , author U

    author author B. Casavecchia , author U. Maio , author C. P \'e roux , \ and\ author B. Ciardi ,\ 10.48550/arXiv.2410.14284 journal journal arXiv e-prints \ ,\ eid arXiv:2410.14284 ( year 2024 ) ,\ http://arxiv.org/abs/2410.14284 arXiv:2410.14284 [astro-ph.GA] NoStop

    2024 doi arXiv

  13. [21]

    author author E. D. \ Kovetz , author M. P. \ Viero , author A. Lidz , author L. Newburgh , author M. Rahman , author E. Switzer , author M. Kamionkowski , author J. Aguirre , author M. Alvarez , author J. Bock , author J. R. \ Bond , author G. Bower , author C. M. \ Bradford ...

    doi arXiv

  14. [22]

    author author A. V. \ Kravtsov ,\ title High-resolution simulations of structure formation in the universe ,\ @noop Ph.D. thesis ,\ school New Mexico State University ( year 1999 ) NoStop

    1999

  15. [23]

    author author A. V. \ Kravtsov , author A. Klypin , \ and\ author Y. Hoffman ,\ 10.1086/340046 journal journal \ volume 571 ,\ pages 563 ( year 2002 ) ,\ http://arxiv.org/abs/astro-ph/0109077 arXiv:astro-ph/0109077 [astro-ph] NoStop

    2002 arXiv

  16. [24]

    author author D. H. \ Rudd , author A. R. \ Zentner , \ and\ author A. V. \ Kravtsov ,\ 10.1086/523836 journal journal \ volume 672 ,\ pages 19 ( year 2008 ) ,\ http://arxiv.org/abs/astro-ph/0703741 arXiv:astro-ph/0703741 [astro-ph] NoStop

    2008 arXiv

  17. [25]

    Padoan , author T

    author author P. Padoan , author T. Haugb lle , \ and\ author A . Nordlund ,\ 10.1088/2041-8205/759/2/L27 journal journal \ volume 759 ,\ eid L27 ( year 2012 ) ,\ http://arxiv.org/abs/1208.3758 arXiv:1208.3758 [astro-ph.GA] NoStop

    2012 arXiv

  18. [26]

    author author V. A. \ Semenov , author A. V. \ Kravtsov , \ and\ author N. Y. \ Gnedin ,\ 10.3847/0004-637X/826/2/200 journal journal \ volume 826 ,\ eid 200 ( year 2016 ) ,\ http://arxiv.org/abs/1512.03101 arXiv:1512.03101 [astro-ph.GA] NoStop

    2016 arXiv

  19. [27]

    author author G. J. \ Ferland , author K. T. \ Korista , author D. A. \ Verner , author J. W. \ Ferguson , author J. B. \ Kingdon , \ and\ author E. M. \ Verner ,\ 10.1086/316190 journal journal \ volume 110 ,\ pages 761 ( year 1998 ) NoStop

    doi

  20. [28]

    author author N. Y. \ Gnedin \ and\ author A. V. \ Kravtsov ,\ 10.1088/0004-637X/728/2/88 journal journal \ volume 728 ,\ eid 88 ( year 2011 ) ,\ http://arxiv.org/abs/1004.0003 arXiv:1004.0003 [astro-ph.CO] NoStop

    2011 arXiv

  21. [29]

    Westmeier , author R

    author author T. Westmeier , author R. Braun , \ and\ author B. S. \ Koribalski ,\ 10.1111/j.1365-2966.2010.17596.x journal journal \ volume 410 ,\ pages 2217 ( year 2011 ) ,\ http://arxiv.org/abs/1009.0317 arXiv:1009.0317 [astro-ph.CO] NoStop

    2010

  22. [30]

    Chen \ and\ author C

    author author T. Chen \ and\ author C. Guestrin ,\ https://api.semanticscholar.org/CorpusID:4650265 journal journal Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining \ ( year 2016 ) NoStop

    2016

  23. [31]

    author author B. T. \ Draine ,\ @noop title Physics of the Interstellar and Intergalactic Medium \ ( publisher Princeton University Press ,\ year 2011 ) NoStop

    2011

  24. [32]

    Barinovs , author M

    author author G . Barinovs , author M. C. \ van Hemert , author R. Krems , \ and\ author A. Dalgarno ,\ 10.1086/426860 journal journal \ volume 620 ,\ pages 537 ( year 2005 ) NoStop

    2005 doi

  25. [33]

    author author S. S. \ Tayal ,\ 10.1051/0004-6361/200810055e journal journal \ volume 501 ,\ pages 381 ( year 2009 ) NoStop

    2009 doi

  26. [34]

    author author M. J. \ Turk , author B. D. \ Smith , author J. S. \ Oishi , author S. Skory , author S. W. \ Skillman , author T. Abel , \ and\ author M. L. \ Norman ,\ 10.1088/0067-0049/192/1/9 journal journal \ volume 192 ,\ eid 9 ( year 2011 ) ,\ http://arxiv.org/abs/1011.35...

    arXiv

  27. [35]

    author author C. R. \ Harris , author K. J. \ Millman , author S. J. \ van der Walt , author R. Gommers , author P. Virtanen , author D. Cournapeau , author E. Wieser , author J. Taylor , author S. Berg , author N. J. \ Smith , author R. Kern , author M. Picus , author S. Hoye...

    arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.