Pith. sign in

REVIEW 3 major objections 5 minor 33 references

Using Neural Networks to Automate the Identification of Brightest Cluster Galaxies in Large Surveys

T0 review · 3 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read A neural network trained on simulated galaxy clusters can locate the brightest cluster galaxy (BCG) directly from multiband images, reaching near-perfect agreement on simulated tests, strong agreement on bright real clusters, and moderate…

desk verdict Honest proof-of-concept for image-only BCG finding; the WHL12 ground-truth caveat is real but acknowledged, and the method deserves refereeing. read the letter →

arxiv 2502.00104 v1 pith:JTZDGAQL submitted 2025-01-31 astro-ph.GA astro-ph.CO

Patrick Janulewicz , Tracy M.A. Webb , Laurence Perreault-Levasseur This is my paper · ORCID
classification astro-ph.GAastro-ph.CO
keywords brightestclustergalaxiesneuralnetworksBCGidentificationmultibandphotometrySDSSTheThreeHundredsimulationtransferlearninggalaxycenters
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 argues that a convolutional neural network can identify the centroid of the brightest cluster galaxy (BCG) directly from multiband photometric images, with no galaxy catalog, per-galaxy redshift, or foreground subtraction. Trained on mock images from The Three Hundred simulation, the network reaches R²≈0.94 on simulated test clusters, but falls to R²≈0.60 when applied to real SDSS clusters, reflecting gaps between simulation and observation. Restricting to real clusters with bright BCGs (r_BCG≤16.5) brings the correlation to R²≈0.99, and fine-tuning with only a few hundred real images closes most of the accuracy gap. The method also outperforms the simple baseline of selecting the brightest galaxy in the field, and it maintains usable accuracy out to z≈0.6 when trained over a wider redshift range. The picture that emerges is that simulation-trained networks can automate BCG finding at scale, provided the target population resembles the training data or a small amount of real data is used for adaptation.

What carries the argument

The central object is the ResNet18 convolutional network used as a position regressor: it consumes a five-channel (SDSS u,g,r,i,z) image of a 1 Mpc region around a cluster and returns a predicted (x,y) centroid for the BCG. Training images are deliberately off-centered with a Gaussian offset of 150 kpc standard deviation so the model learns to prefer central galaxies while tolerating the offsets between BCGs and cluster centers defined by other techniques. Accuracy is measured by R² on the x- and y-coordinates and by the fraction of predictions within 25 kpc of the truth.

What would settle it

Take a sample of SDSS clusters with r_BCG>16.5, obtain independent BCG positions from deep, high-resolution imaging or from a catalog using a different definition such as redMaPPer or X-ray peaks, and measure the network's offsets against those labels; if the network agrees no better with these independent labels than with WHL12, the accuracy drop is real, but if agreement improves, the gap is largely WHL12 label error.

Watch

Extended reading notes

Core claim

The central claim is that the BCG centroid can be regressed directly from pixel data: a ResNet18 takes a 192×192 five-band image as input and outputs the x,y position of the brightest cluster galaxy. The paper reports near-perfect performance when training and testing are both on simulations (R²≈0.94), strong performance on the brightest real clusters (R²≈0.99 for r_BCG≤16.5), and moderate performance on the full real sample (R²≈0.60), with the degradation attributed to the simulated clusters being more massive and hosting brighter BCGs than the typical real cluster. A model trained on real images transfers well to simulations, but not the reverse, suggesting the simulations cover a subset of the real diversity. The authors conclude the technique is a promising route to fast, automated BCG identification for upcoming wide surveys, with transfer learning as a pragmatic fix for dataset shift.

Load-bearing premise

The real-data evaluation takes the WHL12 catalog's BCG positions as correct, so any systematic errors in how WHL12 picks the brightest galaxy, particularly for dim or distant clusters, are counted as failures of the network rather than as label noise.

Editorial extensions

If this is right

  • Simulation-trained networks can pre-publish BCG catalogs for surveys before real calibration data arrive, with the caveat that accuracy is high only when the real clusters resemble the simulated ones.
  • A characteristic cut such as r_BCG ≤ 16.5 selects a regime where the network is reliable, giving a simple operational rule for users of large surveys.
  • Fine-tuning with a few hundred real images corrects most of the simulation bias, so early small data releases can be leveraged to adapt the model quickly.
  • The network outperforms choosing the brightest galaxy in the field, especially in the presence of foreground interlopers, and it degrades more slowly with redshift.
  • Performance drops sharply beyond z≈0.6, but it is unclear whether that is an SDSS depth or resolution limit or an intrinsic limit of the approach.

Reading between the lines

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

  • A natural extension the paper does not test is applying the same architecture to deeper, higher-resolution surveys like Euclid or Rubin using their filter sets; the pipeline is band-agnostic, so the redshift ceiling could shift upward.
  • The paper's evidence that simulations form a subset of real images suggests a direct improvement: enriching simulated training sets with dimmer and less massive clusters should improve generalization to the full real sample.
  • The confidence or uncertainty prediction suggested in the discussion could be implemented as a two-headed network that also outputs a variance, letting users flag low-confidence detections rather than relying on a magnitude cut.
  • Since WHL12 labels are themselves derived by picking the brightest nearby galaxy, the reported accuracy on real data conflates network errors with catalog errors; a cleaner test would compare against visually verified BCGs.
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 / 5 minor

Summary. The paper proposes a neural-network method to identify brightest cluster galaxy (BCG) centroids directly from multiband photometric images, with the goal of automating BCG identification in large surveys. The authors train a ResNet18 on mock observations derived from The Three Hundred simulations, adding SDSS sky noise to improve realism, and test it on simulated images and on SDSS cluster images from the WHL12 catalog. They report R^2 ≈ 0.94 between predictions and truth on simulations, R^2 ≈ 0.60 on the full real-data test set, and R^2 ≈ 0.99 for the bright subset with r_BCG ≤ 16.5. The paper also compares transfer learning from simulations to real data, benchmarks against a SExtractor-based brightest-galaxy baseline, and studies performance as a function of redshift. The central claim is that neural networks are a promising method for automating BCG identification, especially when trained on simulations and applied to bright, low-redshift clusters.

Significance. If the central claim is supported, the method would be a practical contribution for upcoming wide-field surveys such as Euclid and Rubin/LSST, since it requires only photometric images and no per-galaxy redshift or source catalog. The paper has several strengths: the simulation test uses a cluster-level split, avoiding leakage between training and test projections; the comparison against a SExtractor brightest-galaxy baseline is a useful sanity check; and the transfer-learning experiments quantify how much real data are needed to correct simulation bias. The main weakness is that the real-data evaluation uses WHL12 catalog entries as ground truth, so the reported real-data accuracy measures agreement with a heuristic rather than independent physical truth. This issue is acknowledged in the text but not quantified, and it bears directly on the paper's headline real-data results.

major comments (3)
  1. [§4.3 (also §2.2, §5)] The real-data evaluation uses WHL12 catalog entries as ground truth. WHL12 defines the BCG as the brightest galaxy within a fixed distance from the cluster center, so the reported R^2 ≈ 0.60 on real data and R^2 ≈ 0.99 on the bright subset measure agreement with that heuristic, not independent physical identification. The sentence in §4.3 that dim BCGs are also challenging for WHL12 acknowledges the issue but does not quantify label noise. For the central claim that the network identifies BCGs in real surveys, the authors should validate on an independent label set (e.g., redMaPPer candidates, visual inspection, or spectroscopically confirmed members) or explicitly reframe the real-data results as emulation of the WHL12 algorithm.
  2. [§4.3, Figure 6] The bright-subset result R^2 ≈ 0.99 rests on only 60 test clusters, and no uncertainties or confidence intervals are reported for the R^2 or A_T statistics anywhere in the paper. With n = 60, a handful of outliers can move R^2 substantially, so the comparison among 0.94, 0.60, and 0.99 is not established. Please provide bootstrap or jackknife uncertainties for all reported R^2 and accuracy values. In addition, because the r_BCG ≤ 16.5 cut was chosen after inspecting Figure 5, either justify it a priori or confirm it on an independent held-out sample to avoid post-hoc selection inflating the apparent performance.
  3. [§4.6, Figure 10, and Abstract] The abstract and Section 5 state that the method performs well up to z ≈ 0.6, but the redshift test in §4.6 uses networks trained on real SDSS images, not the simulation-trained network that is the paper's main proposal. The simulation-trained model is only tested at z = 0.15–0.25. Please state explicitly which model supports the z ≈ 0.6 claim, or add a corresponding test of the simulation-trained model if that is the intended claim.
minor comments (5)
  1. [Abstract and §1] There are typographical errors: 'such those' should be 'such as those' in the abstract, and 'advantanges' should be 'advantages' in the introduction.
  2. [§3.1] The sentence 'We select a ResNet18 architecture and pass the x and y pixel values of the BCG centroid as learned parameters' is confusing; the x and y values are the network outputs, not learned parameters, and the wording should be corrected.
  3. [§3.2, Eq. (4)] The indicator notation '1≤ T(di)' is hard to parse; it should be typeset as an indicator function, e.g., \mathbb{1}[d_i \le T], and the definition should be stated in words as well.
  4. [Table 1 and §4.6] The 'Real (full redshift)' row in Table 1 uses dashes in a way that is unclear; specify the training and test sample sizes explicitly in the table or caption. The redshift experiment in §4.6 should also clarify whether the 'full redshift' training set and the test set are disjoint at the cluster level, not merely at the image level.
  5. [§3.1] The off-centering distribution is truncated at roughly 167 kpc because of the 1 Mpc cutout, so the model never sees BCG offsets beyond that value. If target surveys have larger BCG-to-center offsets, performance may degrade; please discuss this limitation explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the neural network predictions are evaluated against independently defined labels and no derivation reduces to its own inputs.

full rationale

This paper is an empirical machine-learning study, not a derivation chain, and its claims are not circular. The network is trained on simulated cluster images from The Three Hundred project with BCG positions defined by the simulation, and tested on held-out simulated images plus real SDSS images whose BCG labels come from the external WHL12 catalog. The reported statistics R^2 ≈ 0.94 on simulations, R^2 ≈ 0.60 on real data, and R^2 ≈ 0.99 on the bright r_BCG ≤ 16.5 subset are measured post-hoc on held-out data rather than being produced by fitting the same quantity that is later predicted. The only potentially load-bearing external input, the WHL12 BCG label, is cited as an independent catalog and the paper explicitly concedes that dim BCGs may be misidentified in that catalog, which is an honest caveat rather than a circular step. The selection of the bright subset is a soundness and generalization concern, not a circularity concern, because the model parameters and the subset definition do not depend on the evaluation labels in any way that forces the reported result. No equation is defined in terms of its own conclusion, no fitted parameter is renamed as a prediction, and no self-citation is used to justify a core premise. Verdict: no significant circularity.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

Everything the central claim rests on is either the simulation-to-reality fidelity, the WHL12 ground-truth labels, or the hand-chosen offset/threshold/subset choices. There are no invented physical entities. The biggest burden is domain fidelity: the sim-trained model only reaches R^2 ≈ 0.99 on real data after restricting to a bright-subset that matches the simulation's BCG population.

free parameters (4)
  • offset_std_dev = 150 kpc (truncated at ~167 kpc)
    Chosen by hand to mimic BCG-X-ray peak offsets from literature (Seppi et al. 2023); the network learns a centering prior from this distribution, so the choice affects all reported accuracies.
  • accuracy_threshold_T = 25 kpc
    Defines what counts as a 'correct' prediction in A_T; chosen as a conservative BCG-dominated radius (Brough et al. 2024). Different thresholds would change the reported accuracy percentages.
  • bright_subset_cut = r_BCG ≤ 16.5
    Selected post hoc after observing poor performance on dimmer BCGs; the headline R^2 ≈ 0.99 is measured only on this subset of 60 test clusters.
  • training_epochs = 100
    Chosen from the learning-curve plateau to avoid overfitting; the transfer-learning runs also use 100 fine-tuning epochs.
assumptions (3)
  • domain assumption The Three Hundred gizmo-simba simulations produce BCG photometry and large-scale environment representative of real SDSS clusters.
    The simulation-trained model is the paper's core contribution. Section 4.1 shows simulated images are systematically brighter and more powerful than real WHL12 images, which directly explains the drop from R^2 ≈ 0.94 to 0.60 when transferring to real data.
  • domain assumption WHL12 BCG positions, Petrosian r_BCG, and the mass-richness scaling relation are accurate enough for ground truth and subsetting.
    All real-data evaluations and the bright-subset analysis use WHL12 (§2.2, Eq. 1). The paper itself concedes in §4.3 that dim BCGs are challenging for WHL12, which means label noise could bias the reported real-data accuracies.
  • ad hoc to paper The synthetic off-centering Gaussian (σ = 150 kpc, truncated at ~167 kpc) covers the real BCG offset distribution in surveys.
    This offset distribution is imposed on both the simulated and real images during training and testing (§3.1). It is justified only by a reference to X-ray peak offsets (Seppi et al. 2023); if real BCG offsets are larger or more variable in the target surveys, the model will perform worse than reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Using Neural Networks to Automate the Identification of Brightest Cluster Galaxies in Large Surveys." pith.science (2026). https://pith.science/paper/JTZDGAQL

@misc{pith2026250200104,
  author       = {Pith},
  title        = {Pith review of: Using Neural Networks to Automate the Identification of Brightest Cluster Galaxies in Large Surveys},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JTZDGAQL}},
  note         = {Machine review of arXiv:2502.00104}
}
abstract

Brightest cluster galaxies (BCGs) lie deep within the largest gravitationally bound structures in existence. Though some cluster finding techniques identify the position of the BCG and use it as the cluster center, other techniques may not automatically include these coordinates. This can make studying BCGs in such surveys difficult, forcing researchers to either adopt oversimplified algorithms or perform cumbersome visual identification. For large surveys, there is a need for a fast and reliable way of obtaining BCG coordinates. We propose machine learning to accomplish this task and train a neural network to identify positions of candidate BCGs given no more information than multiband photometric images. We use both mock observations from The Three Hundred project and real ones from the Sloan Digital Sky Survey (SDSS), and we quantify the performance. Training on simulations yields a squared correlation coefficient, R$^2$, between predictions and ground truth of R$^2 \approx 0.94$ when testing on simulations, which decreases to R$^2 \approx 0.60$ when testing on real data due to discrepancies between datasets. Limiting the application of this method to real clusters more representative of the training data, such those with a BCG r-band magnitude $r_{\text{BCG}} \leq 16.5$, yields R$^2 \approx 0.99$. The method performs well up to a redshift of at least $z\approx 0.6$. We find this technique to be a promising method to automate and accelerate the identification of BCGs in large datasets.

Figures

Figures reproduced from arXiv: 2502.00104 by the authors.

Figure 1
Figure 1. The five columns on the left represent photometric bands from SDSS. Simulated galaxy clusters are shown in the first row along with random SDSS noise in the second row. The combination of the two is shown in the third row, and a real WHL12 cluster is shown in the final row for comparison. In the rightmost column, we give an example of a cluster being off-centered by illustrating the process on the simulated r-band. … view at source ↗
Figure 2
Figure 2. The first row contains the 1D PDF of flux values averaged over 5000 images for each dataset. The second row shows the power spectrum averaged over the same images for each dataset. Columns indicate SDSS bands. While the two datasets appear to be quite similar, the simulated images show higher flux and power. in each of the five bands. We then average the resulting power spectra across all images and report the resul… view at source ↗
Figure 3
Figure 3. The redshift and mass distribution of galaxy clus￾ters. Real clusters are represented by blue triangles, while simulated ones are represented by black circles. Note that the horizontal lines from simulations are due to the four dis￾crete snapshots. Overall, the real clusters cover a wider mass range and tend to be less massive as a whole compared to simulated ones. As expected, the real clusters are found to occupy … view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Learning curve when training on simulated galaxy clusters. We compare the training and validation loss for simulated clusters. We also add the validation loss for real clusters. Both validation curves plateau after relatively few epochs, with the validation loss for re…
Figure 5
Figure 5. Figure 5: Performance for different values of rBCG and M200. Accuracy is defined as the proportion of predictions within 25 kpc of the true position. The number of samples in each bin is indicated in parentheses. Lower boundaries on ranges are inclusive, while upper boundaries a…
Figure 6
Figure 6. Figure 6: Performance when trained on simulated clusters. The model is tested on simulated clusters, real SDSS clus￾ters, and a bright subset of these SDSS clusters satisfying rBCG ≤ 16.5. Displayed on top is a cumulative distribution function (CDF) of the Euclidean error across…
Figure 8
Figure 8. Figure 8: A comparison of R2 x, R2 y, and AT for different networks tested on the real and simulated cluster datasets. The threshold distance for AT is set to be 25 kpc. Labels to the left of the arrow represent the dataset used for train￾ing. They are abbreviated S for simulate…
Figure 9
Figure 9. Figure 9: The blue curve represents the accuracy of the neural network used in this work for each simulation snap￾shot. The black curve represents the accuracy achieved by selecting the brightest galaxy in the r-band within 200 kpc of the cluster center, while the gray curve rep…
Figure 10
Figure 10. Figure 10: The proportion of predictions within a given distance to the ground truth at various redshifts. The model trained on a range of 0.15 ≤ z ≤ 0.25 is compared to a model trained over the entire redshift range of the catalog. Both are tested on real SDSS clusters. Solid l…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 5 canonical work pages

  1. [1]

    2011, ApJS, 193, 29, doi: 10.1088/0067-0049/193/2/29

    Aihara, H., Allende Prieto, C., An, D., et al. 2011, ApJS, 193, 29, doi: 10.1088/0067-0049/193/2/29

    doi:10.1088/0067-0049/193/2/29 2011

  2. [2]

    2017, MNRAS, 470, 2617, doi: 10.1093/mnras/stx721 2 https://www.the300-project.org

    Alam, S., Ata, M., Bailey, S., et al. 2017, MNRAS, 470, 2617, doi: 10.1093/mnras/stx721 2 https://www.the300-project.org

    doi:10.1093/mnras/stx721 2017

  3. [3]

    1996, A&AS, 117, 393, doi: 10.1051/aas:1996164

    Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393, doi: 10.1051/aas:1996164

    doi:10.1051/aas:1996164 1996

  4. [4]

    E., Stalder, B., de Haan, T., et al

    Bleem, L. E., Stalder, B., de Haan, T., et al. 2015, The Astrophysical Journal Supplement Series, 216, 27, doi: 10.1088/0067-0049/216/2/27

    doi:10.1088/0067-0049/216/2/27 2015

  5. [5]

    H., Teimoorinia, H., et al

    Bottrell, C., Hani, M. H., Teimoorinia, H., et al. 2019, MNRAS, 490, 5390, doi: 10.1093/mnras/stz2934 13

    doi:10.1093/mnras/stz2934 2019

  6. [6]

    L., Bah´ e, Y

    Brough, S., Ahad, S. L., Bah´ e, Y. M., et al. 2024, MNRAS, 528, 771, doi: 10.1093/mnras/stad3810

    doi:10.1093/mnras/stad3810 2024

  7. [7]

    2003, MNRAS, 344, 1000, doi: 10.1046/j.1365-8711.2003.06897.x

    Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000, doi: 10.1046/j.1365-8711.2003.06897.x

    work page arXiv 2003

  8. [8]

    2003, PASP, 115, 763, doi: 10.1086/376392

    Chabrier, G. 2003, PASP, 115, 763, doi: 10.1086/376392

    doi:10.1086/376392 2003

Show all 33 references
  1. [9]

    2021, A&A, 649, A42, doi: 10.1051/0004-6361/202040245

    Chu, A., Durret, F., & M´ arquez, I. 2021, A&A, 649, A42, doi: 10.1051/0004-6361/202040245

    2021 doi

  2. [10]

    2022, Astronomy & Astrophysics, 666, A54, doi: 10.1051/0004-6361/202243504

    Chu, A., Sarron, F., Durret, F., & M´ arquez, I. 2022, Astronomy & Astrophysics, 666, A54, doi: 10.1051/0004-6361/202243504

    2022 doi

  3. [11]

    2016, Monthly Notices of the Royal Astronomical Society, 456, 2566, doi: 10.1093/mnras/stv2839

    Cui, W., Power, C., Biffi, V., et al. 2016, Monthly Notices of the Royal Astronomical Society, 456, 2566, doi: 10.1093/mnras/stv2839

    2016 doi

  4. [12]

    2018, MNRAS, 480, 2898, doi: 10.1093/mnras/sty2111

    Cui, W., Knebe, A., Yepes, G., et al. 2018, MNRAS, 480, 2898, doi: 10.1093/mnras/sty2111

    2018 doi

  5. [13]

    2022, MNRAS, 514, 977, doi: 10.1093/mnras/stac1402 De Lucia, G., & Blaizot, J

    Cui, W., Dave, R., Knebe, A., et al. 2022, MNRAS, 514, 977, doi: 10.1093/mnras/stac1402 De Lucia, G., & Blaizot, J. 2007, Monthly Notices of the Royal Astronomical Society, 375, 2–14, doi: 10.1111/j.1365-2966.2006.11287.x De Propris, R., West, M. J., Andrade-Santos, F., et al....

    2022

  6. [14]

    H., Gettings, D

    Gonzalez, A. H., Gettings, D. P., Brodwin, M., et al. 2019, The Astrophysical Journal Supplement Series, 240, 33, doi: 10.3847/1538-4365/aafad2

    2019 doi

  7. [15]

    2019, MNRAS, 483, 3545, doi: 10.1093/mnras/sty3203

    Gozaliasl, G., Finoguenov, A., Tanaka, M., et al. 2019, MNRAS, 483, 3545, doi: 10.1093/mnras/sty3203

    2019 doi

  8. [16]

    2015, Deep Residual Learning for Image Recognition

    He, K., Zhang, X., Ren, S., & Sun, J. 2015, Deep Residual Learning for Image Recognition. https://arxiv.org/abs/1512.03385 Ivezi´ c,ˇZ., Kahn, S. M., Tyson, J. A., et al. 2019, ApJ, 873, 111, doi: 10.3847/1538-4357/ab042c

    2015 arXiv

  9. [17]

    P., & Ba, J

    Kingma, D. P., & Ba, J. 2017, Adam: A Method for Stochastic Optimization. https://arxiv.org/abs/1412.6980

    2017 arXiv

  10. [18]

    2023, ApJS, 267, 41, doi: 10.3847/1538-4365/ace052

    Kluge, M., & Bender, R. 2023, ApJS, 267, 41, doi: 10.3847/1538-4365/ace052

    2023 doi

  11. [19]

    2011, arXiv e-prints, arXiv:1110.3193, doi: 10.48550/arXiv.1110.3193

    Laureijs, R., Amiaux, J., Arduini, S., et al. 2011, arXiv e-prints, arXiv:1110.3193, doi: 10.48550/arXiv.1110.3193

    2011 doi arXiv

  12. [20]

    2022, A&A, 661, A2, doi: 10.1051/0004-6361/202141120

    Liu, A., Bulbul, E., Ghirardini, V., et al. 2022, A&A, 661, A2, doi: 10.1051/0004-6361/202141120

    2022 doi

  13. [21]

    L., & Gonzalez, A

    Mancone, C. L., & Gonzalez, A. H. 2012, Publications of the Astronomical Society of the Pacific, 124, 606–615, doi: 10.1086/666502

    2012 doi

  14. [22]

    Muzzin, A., Wilson, G., Yee, H. K. C., et al. 2009, The Astrophysical Journal, 698, 1934, doi: 10.1088/0004-637X/698/2/1934

    2009 doi

  15. [23]

    2010, in , 807–814 Planck Collaboration, Ade, P

    Nair, V., & Hinton, G. 2010, in , 807–814 Planck Collaboration, Ade, P. A. R., Aghanim, N., et al. 2016, A&A, 594, A27, doi: 10.1051/0004-6361/201525823

    2010 doi

  16. [24]

    2024, The Open Journal of Astrophysics, 7, 65, doi: 10.33232/001c.122309

    Roche, C., McDonald, M., Borrow, J., et al. 2024, The Open Journal of Astrophysics, 7, 65, doi: 10.33232/001c.122309

    2024 doi

  17. [25]

    S., Rozo, E., Busha, M

    Rykoff, E. S., Rozo, E., Busha, M. T., et al. 2014, The Astrophysical Journal, 785, 104, doi: 10.1088/0004-637x/785/2/104

    2014 doi

  18. [26]

    2023, Astronomy & Astrophysics, 671, A57, doi: 10.1051/0004-6361/202245138

    Seppi, R., Comparat, J., Nandra, K., et al. 2023, Astronomy & Astrophysics, 671, A57, doi: 10.1051/0004-6361/202245138

    2023 doi

  19. [27]

    C., Ryczanowski, D., Bianconi, M., et al

    Smith, J. C., Ryczanowski, D., Bianconi, M., et al. 2023, Research Notes of the American Astronomical Society, 7, 51, doi: 10.3847/2515-5172/acc333

    2023 doi

  20. [28]

    2022, AJ, 163, 146, doi: 10.3847/1538-3881/ac5030

    Somboonpanyakul, T., McDonald, M., Noble, A., et al. 2022, AJ, 163, 146, doi: 10.3847/1538-3881/ac5030

    2022 doi

  21. [29]

    2020, Journal of Open Source Software, 5, 2505, doi: 10.21105/joss.02505 van den Bosch, F

    Tejero-Cantero, A., Boelts, J., Deistler, M., et al. 2020, Journal of Open Source Software, 5, 2505, doi: 10.21105/joss.02505 van den Bosch, F. C., Weinmann, S. M., Yang, X., et al. 2005, MNRAS, 361, 1203, doi: 10.1111/j.1365-2966.2005.09260.x

    2020

  22. [30]

    R., Forman, W., et al

    Vikhlinin, A., McNamara, B. R., Forman, W., et al. 1998, ApJ, 502, 558, doi: 10.1086/305951

    1998 doi

  23. [31]

    L., Han, J

    Wen, Z. L., Han, J. L., & Liu, F. S. 2012, The Astrophysical Journal Supplement Series, 199, 34, doi: 10.1088/0067-0049/199/2/34

    2012 doi

  24. [32]

    Wilson, G., Muzzin, A., Yee, H. K. C., et al. 2009, ApJ, 698, 1943, doi: 10.1088/0004-637X/698/2/1943

    2009 doi

  25. [33]

    G., Adelman, J., Anderson, John E., J., et al

    York, D. G., Adelman, J., Anderson, John E., J., et al. 2000, AJ, 120, 1579, doi: 10.1086/301513

    2000 doi

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.