Pith. sign in

REVIEW 4 major objections 9 minor 112 references

Simulations calibrate cluster bias to deliver unbiased H0 at 1.5%

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · glm-5.2

2026-07-10 04:21 UTC pith:MOGHPROE

load-bearing objection Solid methods paper with a circularity problem it mostly acknowledges: the mock validation draws B from the same distribution used as the prior, so unbiasedness is guaranteed by construction. the 4 major comments →

arxiv 2607.08613 v1 pith:MOGHPROE submitted 2026-07-09 astro-ph.CO

The Three Hundred Project: validating H₀ inference from mock X-ray and millimetre analyses of galaxy clusters

F. De Luca , H. Bourdin , P. Mazzotta , E. Rasia , A. Kozmanyan , W. Cui , M. De Petris , D. de Andres
show 1 more author
G. Yepes
This is my paper
classification astro-ph.CO
keywords galaxy clustersHubble constantSunyaev-Zel'dovich effectX-ray observationscosmological parametersBayesian inferencehydrodynamical simulationsintracluster medium
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper uses hydrodynamical simulations of 972 galaxy cluster projections to characterize the 'cluster structure bias' B — the systematic discrepancy between temperatures reconstructed from X-ray and Sunyaev-Zel'dovich (SZ) observations that arises from the simplifying assumption that clusters are smooth, spherical objects. The authors show B follows a well-defined statistical distribution that depends on cluster morphology: relaxed clusters cluster near B≈1 with small scatter, while disturbed systems show larger values and wider scatter. They model this relationship both through a three-class scheme (relaxed, hybrid, disturbed) and through Gaussian Process regressions on a continuous morphological indicator. When these simulation-informed priors on B are fed into a Bayesian pipeline that jointly analyses X-ray and SZ temperature profiles to infer the Hubble constant from mock cluster samples, the method recovers H0 without bias, achieving roughly 4% precision with 100 clusters and 1.5% with 1000, with a systematic floor of 0.6–0.8 km/s/Mpc. The approach works without restricting analysis to the small subset of relaxed clusters, because the morphology-dependent prior on B corrects for the structural bias of each system.

Core claim

The structural bias that limits galaxy clusters as standard rulers for measuring H0 is not an irreducible nuisance but a quantifiable, morphology-dependent statistical quantity. When calibrated from hydrodynamical simulations and used as an informative prior in a Bayesian pipeline, joint X-ray and SZ temperature profiles yield unbiased H0 estimates whose precision scales with sample size down to a small systematic floor, rather than being limited by the need to select only relaxed clusters.

What carries the argument

The cluster structure bias B, defined as the ratio between X-ray and SZ reconstructed temperatures (B ≡ TX/T_SZ,X), quantifies how much the simplified smooth-spherical model of the intracluster medium distorts thermodynamic reconstructions. Its distribution is derived from The Three Hundred gadget-x simulations and modeled as a function of an X-ray morphological indicator MX via Gaussian Process regression, then used as a log-normal informative prior in a Bayesian inference of H0 from the cosmological factor C(z, ϑ) that relates apparent cluster sizes to angular diameter distances.

Load-bearing premise

The distribution of structural bias B, measured from a single hydrodynamical simulation at redshift zero, accurately represents the structural biases of real galaxy clusters across the redshift range 0.05–0.6.

What would settle it

If real cluster B values, measured from joint X-ray and SZ observations, deviate systematically from the simulation-predicted distribution in a way that depends on morphology differently than modeled, the informative priors would be miscalibrated and the Bayesian pipeline would produce biased H0 estimates rather than the unbiased results seen in mock tests.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 9 minor

Summary. This paper uses The Three Hundred hydrodynamical simulations (gadget-x run) to characterise the cluster-structure bias parameter B, which quantifies the discrepancy between ICM temperature profiles reconstructed from X-ray and SZ observations under simplified (spherically symmetric, smooth) modelling assumptions. The authors find that B depends on cluster dynamical and morphological state: relaxed clusters cluster near B=1 with small scatter, while disturbed systems show higher values and larger scatter. They model this dependence both via a trichotomous class-conditional model (3DS) and via Gaussian-process regressions (GPkernel, GPhetero) on a morphological indicator. These simulation-informed priors are then incorporated into a Bayesian cosmological pipeline that infers H0 from the ratio of SZ and X-ray temperature profiles. Mock tests yield unbiased H0 estimates with ~4% precision for 100 clusters and ~1.5% for 1000, with a systematic floor of 0.6-0.8 km/s/Mpc. The work extends the methodology of Kozmanyan et al. (2019) to a larger sample and adds morphology-dependent priors.

Significance. The paper addresses a real and recognised problem in cluster-based cosmology: structural biases from asphericity, clumping, and dynamical state affect the interpretation of joint X-ray/SZ observations. The approach of using simulation-informed priors on B, conditioned on observable morphological indicators, is a sensible strategy to move beyond restricting analyses to small relaxed subsamples. The simulation-based calibration (Sect. 4.4.3) and the explicit decomposition of variance into statistical and systematic terms (Fig. 7) are commendable. The GP regression framework for modelling heteroscedastic scatter in B as a function of morphology is a useful methodological contribution. The precision forecasts, if they survive scrutiny regarding their transferability to real data, would be competitive for an H0 probe.

major comments (4)
  1. Sect. 4.4, Eq. (26): The mock validation is structurally circular. The mock eta_T catalogues are generated by drawing B from the simulated The300 gadget-x distribution, and the same distribution (or its 3DS/GP model) is used as the informative prior in the Bayesian pipeline. Unbiasedness is therefore guaranteed by construction under this setup. The simulation-based calibration in Sect. 4.4.3 is also internal: parameters are drawn from the same priors, data are simulated, and posteriors are checked for coverage. This tests the pipeline's self-consistency, not whether the gadget-x B distribution matches reality. The paper acknowledges this risk in Sect. 5.1 ('a residual dependence of the inferred B distribution on the adopted simulation model cannot be excluded'), but the headline precision claims (4% at 100 clusters, 1.5% at 1000, systematic floor 0.6-0.8 km/s/Mpc) are all derived under a
  2. Sect. 5.1: The paper notes that gizmo-simba (another The300 run) shows different ICM properties (higher temperatures, lower central densities) compared to gadget-x. Since B is derived from relative mismatches between thermodynamic reconstructions, a concrete test of robustness would be to recompute B from the gizmo-simba run and compare the distributions. The300 provides exactly this framework (same haloes, different hydrodynamical implementations). Without such a test, the reader cannot assess how sensitive the B distribution and the resulting H0 forecasts are to subgrid physics. This is load-bearing because the entire prior calibration rests on the fidelity of the gadget-x B distribution. At minimum, the authors should discuss whether the gizmo-simba differences are expected to affect B specifically (not just the absolute profiles), and ideally provide a quantitative comparison.
  3. Sect. 4.4.2, Fig. 7: The systematic floor sigma_s = 0.6-0.8 km/s/Mpc is derived by fitting sigma^2(N) = sigma_0^2/N + sigma_s^2 to the dispersion of H0 MAP estimates across mock realisations. However, since the mock B values are drawn from the same distribution used as priors, the residual scatter at large N reflects only the irreducible scatter within the simulated B distribution itself, not any mismatch between simulation and reality. The systematic floor is therefore a lower bound that would only apply if the simulated B distribution perfectly described real clusters. The paper should clarify this explicitly: the 0.6-0.8 km/s/Mpc value is the systematic limit under the assumption that the gadget-x B distribution is correct, not an absolute systematic floor for real observations.
  4. Sect. 3.1, Eq. (11): The generalisation introducing b_n to account for observational systematics (X-ray temperature calibration, missing relativistic corrections) is mentioned but never tested. In the mock analysis, b_n = 1 by construction. Since calibration uncertainties in X-ray temperature measurements are known to be non-negligible (Schellenberger et al. 2015; Migkas et al. 2024, both cited), and Wan et al. (2021) included an empirical prior on this bias, the absence of any test of how b_n affects H0 recovery is a gap. Even a simple sensitivity test (e.g., injecting a constant b_n offset and checking whether the pipeline recovers a biased H0) would strengthen the paper's claim of robustness for future observational applications.
minor comments (9)
  1. Sect. 2.2: The relaxation indicator chi (Eq. 4) uses x_{0,i} = 0.1 for both f_s and Delta_r. The sensitivity to this choice is not discussed. A brief comment on whether the classification is robust to the normalisation would help.
  2. Sect. 3.1: The integration length of 20 Mpc is adopted, with alternatives tested in Appendix A. The KS test p-values are reported, but it would help to state explicitly whether the B values for individual clusters change significantly between integration lengths, not just whether the overall distributions are consistent.
  3. Fig. 2 inset: The comparison with the K19 distribution is shown but the two samples differ in cosmology and redshift. A brief note on whether the K19 distribution was rescaled or adjusted for the different cosmology would improve clarity.
  4. Sect. 4.3: The GPkernel model maps M_X onto a normalised variable via the CDF of the normal distribution. The choice of this specific transformation is motivated but somewhat ad hoc. A brief comparison with alternative transformations, or at least a statement that the choice was validated, would be useful.
  5. Table 1: The fitted parameters for the single log-normal model (mu_fit = 1.08e-2, sigma_fit = 13.29e-2) seem inconsistent with the median B = 1.01. For a log-normal distribution, the median should be exp(mu). Please check whether the reported mu_fit is in log-space or linear-space, and clarify.
  6. Sect. 4.4.1: The statement that 'Omega_m and Y posteriors are essentially prior-dominated' is important but buried. Given that the method is primarily an H0 probe, this limitation should be stated more prominently, perhaps in the abstract or conclusions.
  7. Appendix D, Eq. (D.1): The derivation assumes xi << 1. The typical metallicity values used should be stated for context.
  8. The manuscript uses 'The300' and 'The Three Hundred' interchangeably. Consistency would improve readability.
  9. Sect. 5.1: The discussion of redshift evolution of B is qualitative. Even a simple check using available snapshots at z > 0 (The300 has multiple redshifts) would substantially strengthen the claim that evolution is negligible.

Circularity Check

2 steps flagged

Validation is structurally circular: mock data are drawn from the same B distribution used as informative priors, so unbiasedness is guaranteed by construction.

specific steps
  1. fitted input called prediction [Section 4.4, Eq. (26), and Table 2]
    "ln ηmock_T = ln[C(z; H0^ref, Ω_m^ref, Y^ref)] + ln B + N(0, 0.14), with an additional log-normal dispersion (of the same order) to simulate real observations consistent with the observational findings of De Luca et al. (2026). With these mock catalogues, we then tested our Bayesian analysis with the PyMC package, adopting the prior definitions summarised in Table 2"

    The mock ηT catalogues are generated by drawing B from the The300 gadget-x simulated sample (Eq. 26). The same simulated B distribution—fitted into the 3DS or GP models (Sect. 4.1, Table 1)—is then used as the informative prior in the Bayesian pipeline (Table 2). The 'unbiased H0' result (Sect. 4.4.2, Fig. 6) is thus guaranteed by construction: the data-generating distribution for B matches the prior. The simulation-based calibration (Sect. 4.4.3) is also internal: parameters are drawn from the same priors, data simulated, and posteriors checked for coverage. This tests MCMC self-consistency, not whether the gadget-x B distribution matches reality. The precision claims (4% at 100 clusters, 1.5% at 1000, systematic floor 0.6–0.8 km/s/Mpc) are all derived under this circular setup. The paper

  2. fitted input called prediction [Section 4.4.3, simulation-based calibration]
    "instead of adopting a fixed cosmology and a random selection of B values as detailed in Sect. 4.4, we generated 200 realisations of ηsim_T by sampling parameters according to the priors listed in Table 2: Θ_sim = (ϑ_sim, B_sim) ~ p(Θ). With these parameters, we then simulated new data as ηsim_T ~ p(ηT|Θ_sim) and drew new posteriors."

    The SBC test draws B_sim from the same prior p(B) derived from the simulation, simulates data, and checks whether posteriors recover the input. This is a standard SBC self-consistency check—it verifies the MCMC and marginalization are correctly implemented, not that the prior matches reality. The paper acknowledges this limitation (Sect. 5.1: 'a residual dependence of the inferred B distribution on the adopted simulation model cannot be excluded'), but the headline precision claims are all derived under this circular setup.

full rationale

The paper's central validation claim—that simulation-informed B priors yield unbiased H0—rests on a circular test: mock data are generated from the same B distribution (The300 gadget-x) that is used as the informative prior in the Bayesian pipeline. The simulation-based calibration in Sect. 4.4.3 is also internal. This is a genuine structural circularity in the validation, not merely self-citation. However, the paper is transparent about this limitation (Sect. 5.1), the B distribution itself is derived from first-principles hydrodynamical simulations (not fitted to the target result), and the methodology is designed for future application to real data. The circularity is in the validation setup, not in the B derivation itself. Score 4 reflects that the central claim has independent content (the B distribution is a simulation output, not a fit to H0) but the unbiasedness demonstration is guaranteed by construction.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 0 invented entities

The paper introduces no new physical entities or postulated particles. The B parameter is a phenomenological bias factor defined in K19; the GP models and 3DS classification are methodological tools, not new physics. All free parameters are fitted to simulation data or assumed for mock generation. The main circularity concern is that the B priors and the mock data used to validate them come from the same simulation.

free parameters (5)
  • σ₀ (statistical scaling) = 26.6–26.9 km/s/Mpc
    Fitted to the variance-vs-sample-size relation σ²(N) = σ₀²/N + σₛ² in Sect. 4.4.2.
  • σₛ (systematic floor) = 0.6–0.8 km/s/Mpc
    Fitted as the asymptotic floor of the variance scaling relation.
  • B distribution parameters (μ, σ per class) = See Table 1
    Log-normal parameters fitted to the simulated B values for full, relaxed, hybrid, and disturbed subsamples.
  • GP hyperparameters (amplitude a, length scale ℓ) = Not explicitly reported
    Optimised during GP regression for both GPkernel and GPhetero models; informative priors placed on them but posterior values not tabulated.
  • Per-cluster ηT uncertainty = 14%
    Assumed in mock catalogue generation (Eq. 26) to mimic observational uncertainties; not derived from a realistic instrument model.
axioms (5)
  • domain assumption The cluster structure bias B is separable from the cosmological bias C in the discrepancy factor ηT = bn·C·B (Eq. 11).
    Invoked in Sect. 3 (Eq. 11); assumes cluster morphology and cosmological distance effects are uncorrelated, which may not hold if morphology evolves with redshift in a cosmology-dependent way.
  • domain assumption The gadget-x simulation of The300 accurately reproduces the ICM thermodynamic structure of real galaxy clusters.
    Invoked throughout; the paper cites supporting evidence (Bartalucci et al. 2023; Rossetti et al. 2024; Lovisari et al. 2024) but acknowledges in Sect. 5.1 that residual dependence on the simulation model cannot be excluded.
  • domain assumption B does not significantly evolve with redshift for z < 0.6.
    Invoked implicitly by using the z=0 snapshot for all mock tests; discussed in Sect. 5.1 where the authors note that current observations show no clear evolution but this is not definitively established.
  • domain assumption Three orthogonal projections of the same halo provide statistically independent B measurements.
    Invoked in Sect. 3.1 to increase sample size to 972; tested via Monte Carlo resampling in Appendix B, which shows no significant bias, but independence is not formally proven.
  • domain assumption Log-normal distributions adequately describe the B distribution for each dynamical class.
    Invoked in Sect. 4.1; selected via BIC comparison among scipy distributions, but the choice is empirical rather than theoretically motivated.

pith-pipeline@v1.1.0-glm · 31543 in / 3186 out tokens · 516490 ms · 2026-07-10T04:21:46.202214+00:00 · methodology

0 comments
read the original abstract

Measurements of thermodynamical quantities in galaxy clusters are differently affected by simplified modelling of radially averaged observables in the X-ray and millimetre bands. This includes assumptions about the cosmological model and the morphology of the cluster intracluster medium (ICM). Within a large sample of clusters extracted from The Three Hundred hydrodynamical simulations, we assess the systematic differences expected from the morphological assumptions between ICM temperatures as inferred from X-ray spectroscopy or joint X-ray and millimetre imaging. We find that these differences show a well-defined statistical behaviour that correlates with the cluster dynamical and morphological indicators. We then investigate how joint inferences of cluster temperature profiles, a priori informed by this statistical behaviour, allow us to constrain cosmological parameters inferred from the apparent cluster sizes. Assuming a flat $\Lambda$CDM model and priors on $\Omega_\mathrm{m}$ and the helium abundance, this method provides us with unbiased estimates of the Hubble constant, $H_0$, characterised with a precision of about $4\%$ and $1.5\%$ for samples of 100 and 1000 clusters, respectively, and ultimately limited by systematic uncertainties of about $0.6$--$0.8\, {\rm km\, s^{-1} Mpc^{-1}}$. This work highlights the potential of joint X-ray and millimetre observations of galaxy cluster samples to place tight constraints on $H_0$.

Figures

Figures reproduced from arXiv: 2607.08613 by A. Kozmanyan, D. de Andres, E. Rasia, F. De Luca, G. Yepes, H. Bourdin, M. De Petris, P. Mazzotta, W. Cui.

Figure 1
Figure 1. Figure 1: Overall distributions (grey histograms) of M500 (left panel), the spectroscopic-like temperatures (centre), and the dynamical indicator χ (right). Relaxed, hybrid, and disturbed subsamples are shown with red, green, and blue colours, respectively. Currently, The300 collects the results produced with gadget￾music (Sembolini et al. 2012), gadget-x (Rasia et al. 2015), and gizmo-simba (Cui et al. 2022) suites… view at source ↗
Figure 2
Figure 2. Figure 2: Distributions of B for the full sample (grey histogram) and the relaxed (red), hybrid (green), and disturbed (blue) subsamples, together with their best fitting log-normal models (solid curves). The black line shows the combined 3DS class-based model, obtained from the three log-normal components fitted to the dynamical subsamples. A compar￾ison between the K19 distribution and our findings is shown in the… view at source ↗
Figure 3
Figure 3. Figure 3: Correlations between B and the cluster temperature, mass, and the dynamical (χ) and morphological (MSZ, for SZ maps) indicators, respectively from upper to lower panels. Relaxed clusters are shown as red circles, while green triangles and blue squares are used for hybrid and disturbed systems. The χ axis is reversed to match the relaxation ordering of MSZ. MX ≲ −2.5), which should be closer to the spherica… view at source ↗
Figure 4
Figure 4. Figure 4: Results of the GP regressions with 1σ and 2σ dispersions for the GPkernel (muted teal) and GPhetero (muted amber) models. The trichoto￾mous classification is shown with red circles for relaxed, green triangles for hybrid, and blue squares for disturbed systems, respectively [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Fractional differences in the H0 MAP estimates with respect to the reference value adopted in the test for the 3DS (black), GPhetero (blue) and GPkernel (red) models of the informative priors, respectively. of the standard deviations always close to unity. In a flat ΛCDM framework, ηT depends mainly on three parameters: H0, Ωm, and Y, as described by Eq. (12). However, only one observable is fitted and thi… view at source ↗
Figure 8
Figure 8. Figure 8: Empirical cumulative distribution of the normalised ranks for the 3DS (black dashed line), GPkernel (red), and GPhetero (blue) models. The solid black line and the grey envelopes show the expectation for a uniform distribution and the confidence intervals at 68% and 95% around it. The coverage of the posteriors at 68% and 95% are also shown in the figure, for the three models. atic terms are small, but non… view at source ↗

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

112 extracted references · 112 canonical work pages · 1 internal anchor

  1. [1]

    2025, Phys

    Abdul Karim, M., Aguilar, J., Ahlen, S., et al. 2025, Phys. Rev. D, 112, 083515

    work page 2025

  2. [2]

    2023, PeerJ Comput

    Abril-Pla, O., Andreani, V ., Carroll, C., et al. 2023, PeerJ Comput. Sci., 9

    work page 2023

  3. [3]

    W., Evrard, A

    Allen, S. W., Evrard, A. E., & Mantz, A. B. 2011, ARA&A, 49, 409

    work page 2011

  4. [4]

    Angulo, R. E. & Hahn, O. 2022, Living Rev. Comput. Astrophys., 8, 1

    work page 2022

  5. [5]

    2020, A&A, 634, A113

    Ansarifard, S., Rasia, E., Biffi, V ., et al. 2020, A&A, 634, A113

    work page 2020

  6. [6]

    W., Piffaretti, R., et al

    Arnaud, M., Pratt, G. W., Piffaretti, R., et al. 2010, A&A, 517, A92 Astropy Collaboration. 2013, A&A, 558, A33 Astropy Collaboration. 2018, AJ, 156, 123 Astropy Collaboration. 2022, ApJ, 935, 167

    work page 2010

  7. [7]

    W., Démoclès, J., & Lovisari, L

    Bartalucci, I., Arnaud, M., Pratt, G. W., Démoclès, J., & Lovisari, L. 2019, A&A, 628, A86

    work page 2019

  8. [8]

    2023, A&A, 674, A179

    Bartalucci, I., Molendi, S., Rasia, E., et al. 2023, A&A, 674, A179

    work page 2023

  9. [9]

    R., Pfrommer, C., & Sievers, J

    Battaglia, N., Bond, J. R., Pfrommer, C., & Sievers, J. L. 2012, ApJ, 758, 74

    work page 2012

  10. [10]

    M., Murante, G., Arth, A., et al

    Beck, A. M., Murante, G., Arth, A., et al. 2015, MNRAS, 455, 2110

    work page 2015

  11. [11]

    2012, MNRAS, 420, 3545

    Biffi, V ., Dolag, K., Böhringer, H., & Lemson, G. 2012, MNRAS, 420, 3545

    work page 2012

  12. [12]

    1999, Phys

    Birkinshaw, M. 1999, Phys. Rep., 310, 97

    work page 1999

  13. [13]

    E., et al

    Bocquet, S., Grandis, S., Bleem, L. E., et al. 2024, Phys. Rev. D, 110, 083510 Böhringer, H., Dolag, K., & Chon, G. 2012, A&A, 539, A120 Böhringer, H. & Werner, N. 2010, A&A Rev., 18, 127

    work page 2024

  14. [14]

    K., LaRoque, S

    Bonamente, M., Joy, M. K., LaRoque, S. J., et al. 2006, ApJ, 647, 25

    work page 2006

  15. [15]

    2017, ApJ, 843, 72

    Bourdin, H., Mazzotta, P., Kozmanyan, A., Jones, C., & Vikhlinin, A. 2017, ApJ, 843, 72

    work page 2017

  16. [16]

    Buote, D. A. & Tsai, J. C. 1995, ApJ, 452, 522

    work page 1995

  17. [17]

    G., Ettori, S., Lovisari, L., et al

    Campitiello, M. G., Ettori, S., Lovisari, L., et al. 2022, A&A, 665, A117

    work page 2022

  18. [18]

    E., Holder, G

    Carlstrom, J. E., Holder, G. P., & Reese, E. D. 2002, ARA&A, 40, 643

    work page 2002

  19. [19]

    1979, A&A, 75, 322

    Cavaliere, A., Danese, L., & de Zotti, G. 1979, A&A, 75, 322

    work page 1979

  20. [20]

    2025, A&A, 699, A141 CHEX-MATE Collaboration

    Chappuis, L., Eckert, D., Sereno, M., et al. 2025, A&A, 699, A141 CHEX-MATE Collaboration. 2021, A&A, 650, A104

    work page 2025

  21. [21]

    2018, MNRAS, 477, 139

    Cialone, G., De Petris, M., Sembolini, F., et al. 2018, MNRAS, 477, 139

    work page 2018

  22. [22]

    R., Gelman, A., & Rubin, D

    Cook, S. R., Gelman, A., & Rubin, D. B. 2006, J. Comput. Graph. Stat., 15, 675

    work page 2006

  23. [23]

    Cowie, L. L. & Perrenod, S. C. 1978, ApJ, 219, 354

    work page 1978

  24. [24]

    M., Evrard, A

    Crone, M. M., Evrard, A. E., & Richstone, D. O. 1996, ApJ, 467, 489

    work page 1996

  25. [25]

    Croton, D. J. 2013, PASA, 30, e052

    work page 2013

  26. [26]

    2022, MNRAS, 514, 977

    Cui, W., Dave, R., Knebe, A., et al. 2022, MNRAS, 514, 977

    work page 2022

  27. [27]

    2018, MNRAS, 480, 2898

    Cui, W., Knebe, A., Yepes, G., et al. 2018, MNRAS, 480, 2898

    work page 2018

  28. [28]

    2017, MNRAS, 464, 2502 De Luca, F., Bourdin, H., Mazzotta, P., et al

    Cui, W., Power, C., Borgani, S., et al. 2017, MNRAS, 464, 2502 De Luca, F., Bourdin, H., Mazzotta, P., et al. 2026, A&A, 707, A32 De Luca, F., De Petris, M., Yepes, G., et al. 2021, MNRAS, 504, 5383 Di Valentino, E., Said, J. L., Riess, A., et al. 2025, Phys. Dark Universe, 49, 101965

    work page 2017

  29. [29]

    2015, MNRAS, 447, 2198

    Eckert, D., Roncarelli, M., Ettori, S., et al. 2015, MNRAS, 447, 2198

    work page 2015

  30. [30]

    2013, Space Sci

    Ettori, S., Donnarumma, A., Pointecouteau, E., et al. 2013, Space Sci. Rev., 177, 119

    work page 2013

  31. [31]

    2020, Astron

    Ettori, S., Ghirardini, V ., & Eckert, D. 2020, Astron. Nachr, 341, 210

    work page 2020

  32. [32]

    2014, ApJ, 783, L10

    Gaspari, M., Brighenti, F., Temi, P., & Ettori, S. 2014, ApJ, 783, L10

    work page 2014

  33. [33]

    2019, ApJ, 884, 169

    Gaspari, M., Eckert, D., Ettori, S., et al. 2019, ApJ, 884, 169

    work page 2019

  34. [34]

    & Rubin, D

    Gelman, A. & Rubin, D. B. 1992, Stat. Sci., 7, 457

    work page 1992

  35. [35]

    2024, A&A, 689, A298

    Ghirardini, V ., Bulbul, E., Artis, E., et al. 2024, A&A, 689, A298

    work page 2024

  36. [36]

    2019, A&A, 621, A41

    Ghirardini, V ., Eckert, D., Ettori, S., et al. 2019, A&A, 621, A41

    work page 2019

  37. [37]

    2022, MNRAS, 518, 4238

    Gianfagna, G., Rasia, E., Cui, W., et al. 2022, MNRAS, 518, 4238

    work page 2022

  38. [38]

    J., Klein, M., & Dolag, K

    Grandis, S., Bocquet, S., Mohr, J. J., Klein, M., & Dolag, K. 2021, MNRAS, 507, 5671

    work page 2021

  39. [39]

    2024, MNRAS, 532, 1031

    Haggar, R., De Luca, F., De Petris, M., et al. 2024, MNRAS, 532, 1031

    work page 2024

  40. [40]

    E., Pearce, F

    Haggar, R., Gray, M. E., Pearce, F. R., et al. 2020, MNRAS, 492, 6074

    work page 2020

  41. [41]

    R., Millman, K

    Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357–362

    work page 2020

  42. [42]

    Hastings, W. K. 1970, Biometrika, 57, 97

    work page 1970

  43. [43]

    A., Barnes, D

    Henson, M. A., Barnes, D. J., Kay, S. T., McCarthy, I. G., & Schaye, J. 2017, MNRAS, 465, 3361

    work page 2017

  44. [44]

    Hoffman, M. D. & Gelman, A. 2014, J. Mach. Learn. Res., 15, 1593

    work page 2014

  45. [45]

    Hunter, J. D. 2007, Comput. Sci. Eng., 9, 90

    work page 2007

  46. [46]

    1986, MNRAS, 222, 323

    Kaiser, N. 1986, MNRAS, 222, 323

    work page 1986

  47. [47]

    2008, ApJ, 674, 11

    Kawahara, H., Kitayama, T., Sasaki, S., & Suto, Y . 2008, ApJ, 674, 11

    work page 2008

  48. [48]

    2024, A&A, 686, A97

    Kim, J., Sayers, J., Sereno, M., et al. 2024, A&A, 686, A97

    work page 2024

  49. [49]

    2016, MNRAS, 457, 4340

    Klypin, A., Yepes, G., Gottlöber, S., Prada, F., & Heß, S. 2016, MNRAS, 457, 4340

    work page 2016

  50. [50]

    Knollmann, S. R. & Knebe, A. 2009, ApJS, 182, 608

    work page 2009

  51. [51]

    2019, A&A, 621, A34

    Kozmanyan, A., Bourdin, H., Mazzotta, P., Rasia, E., & Sereno, M. 2019, A&A, 621, A34

    work page 2019

  52. [52]

    Kravtsov, A. V . & Borgani, S. 2012, ARA&A, 50, 353

    work page 2012

  53. [53]

    V ., Vikhlinin, A., & Nagai, D

    Kravtsov, A. V ., Vikhlinin, A., & Nagai, D. 2006, ApJ, 650, 128

    work page 2006

  54. [54]

    Kumar, R., Carroll, C., Hartikainen, A., & Martin, O. 2019, J. Open Source Softw., 4, 1143

    work page 2019

  55. [55]

    T., Hearin, A

    Lau, E. T., Hearin, A. P., Nagai, D., & Cappelluti, N. 2020, MNRAS, 500, 1029

    work page 2020

  56. [56]

    T., Nagai, D., Avestruz, C., Nelson, K., & Vikhlinin, A

    Lau, E. T., Nagai, D., Avestruz, C., Nelson, K., & Vikhlinin, A. 2015, ApJ, 806, 68

    work page 2015

  57. [57]

    2023, MNRAS, 523, 1228

    Li, Q., Cui, W., Yang, X., et al. 2023, MNRAS, 523, 1228

    work page 2023

  58. [58]

    2020, MNRAS, 495, 2930 Article number, page 11 of 18 A&A proofs: manuscript no

    Li, Q., Cui, W., Yang, X., et al. 2020, MNRAS, 495, 2930 Article number, page 11 of 18 A&A proofs: manuscript no. main

    work page 2020

  59. [59]

    2024, A&A, 682, A45

    Lovisari, L., Ettori, S., Rasia, E., et al. 2024, A&A, 682, A45

    work page 2024

  60. [60]

    V ., Dutton, A

    Maccio, A. V ., Dutton, A. A., Van Den Bosch, F. C., et al. 2007, MNRAS, 378, 55

    work page 2007

  61. [61]

    H., Clowe, D., et al

    Markevitch, M., Gonzalez, A. H., Clowe, D., et al. 2004, ApJ, 606, 819

    work page 2004

  62. [62]

    E., & Mohr, J

    Mathiesen, B., Evrard, A. E., & Mohr, J. J. 1999, ApJ, 520, L21

    work page 1999

  63. [63]

    2004, MNRAS, 354, 10

    Mazzotta, P., Rasia, E., Moscardini, L., & Tormen, G. 2004, MNRAS, 354, 10

    work page 2004

  64. [64]

    W., Bayliss, M., et al

    McDonald, M., Allen, S. W., Bayliss, M., et al. 2017, ApJ, 843, 28

    work page 2017

  65. [65]

    2010, in Proceedings of the 9th Python in Science Conference, ed

    McKinney, W. 2010, in Proceedings of the 9th Python in Science Conference, ed. S. van der Walt & J. Millman, 56 – 61

    work page 2010

  66. [66]

    2010, A&A, 514, A93

    Meneghetti, M., Rasia, E., Merten, J., et al. 2010, A&A, 514, A93

    work page 2010

  67. [67]

    W., Rosenbluth, M

    Metropolis, N., Rosenbluth, A. W., Rosenbluth, M. N., Teller, A. H., & Teller, E. 1953, J. Chem. Phys., 21, 1087

    work page 1953

  68. [68]

    2024, A&A, 688, A107

    Migkas, K., Kox, D., Schellenberger, G., et al. 2024, A&A, 688, A107

    work page 2024

  69. [69]

    2026, in Encyclopedia of Astrophysics (First Edition), first edition edn., ed

    Miyatake, H. 2026, in Encyclopedia of Astrophysics (First Edition), first edition edn., ed. I. Mandel (Oxford: Elsevier), 410–430

    work page 2026

  70. [70]

    J., Fabricant, D

    Mohr, J. J., Fabricant, D. G., & Geller, M. J. 1993, ApJ, 413, 492

    work page 1993

  71. [71]

    V ., & Vikhlinin, A

    Nagai, D., Kravtsov, A. V ., & Vikhlinin, A. 2007, ApJ, 668, 1

    work page 2007

  72. [72]

    Neal, R. M. 2003, Ann. Stat., 31, 705

    work page 2003

  73. [73]

    F., Gao, L., Bett, P., et al

    Neto, A. F., Gao, L., Bett, P., et al. 2007, MNRAS, 381, 1450

    work page 2007

  74. [74]

    2010, A&A, 523, A22

    Nevalainen, J., David, L., & Guainazzi, M. 2010, A&A, 523, A22

    work page 2010

  75. [75]

    A., et al

    Nurgaliev, D., McDonald, M., Benson, B. A., et al. 2017, ApJ, 841, 5 Pandas development team. 2020, pandas-dev/pandas: Pandas

    work page 2017

  76. [76]

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, J. Mach. Learn. Res., 12, 2825 Planck Collaboration VI. 2020, A&A, 641, A6 Planck Collaboration XIII. 2016, A&A, 594, A13 Planck Collaboration XXIV. 2016, A&A, 594, A24

    work page 2011

  77. [77]

    Planelles, S., Schleicher, D. R. G., & Bykov, A. M. 2015, Space Sci. Rev., 188, 93

    work page 2015

  78. [78]

    W., Arnaud, M., Biviano, A., et al

    Pratt, G. W., Arnaud, M., Biviano, A., et al. 2019, Space Sci. Rev., 215, 25

    work page 2019

  79. [79]

    2015, ApJ, 813, L17

    Rasia, E., Borgani, S., Murante, G., et al. 2015, ApJ, 813, L17

    work page 2015

  80. [80]

    2013, Astron

    Rasia, E., Meneghetti, M., & Ettori, S. 2013, Astron. Rev., 8, 40

    work page 2013

Showing first 80 references.