Constraining reionization morphology and source properties with 21cm galaxy cross-correlation surveys

Anne Hutter; Caroline Heneka; Yannic Pietschke

REVIEW 2 major objections 5 minor 2 cited by

Constraining reionization morphology and source properties with 21cm galaxy cross-correlation surveys

T0 review · 2 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read The paper claims that 21cm-galaxy cross-correlation measurements can tightly constrain the escape fraction and star formation efficiency of reionization-era galaxies, parameters that 21cm auto-power measurements leave almost completely dege

desk verdict A genuinely useful in-mock survey-optimization study showing cross-power can break degeneracies auto-power cannot, but the source-property constraints are for power-law model coefficients, not physical f_esc. read the letter →

arxiv 2601.18627 v2 pith:F6HBRIXG submitted 2026-01-26 astro-ph.CO physics.data-an

Yannic Pietschke , Anne Hutter , Caroline Heneka This is my paper

classification astro-ph.COphysics.data-an

keywords Epochofreionization21cmcosmologycross-powerspectrumgalaxysurveysescapefractionstarformationefficiencysimulation-basedinference

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 tries to establish that the cross-power spectrum between 21cm neutral-hydrogen fluctuations and high-redshift galaxy positions is more than a foreground-robust confirmation of the 21cm signal—it is the key that unlocks the properties of the galaxies that reionized the universe. Using realistic mock observations, including 100 hours of low-frequency radio data with foreground masking and galaxy surveys of varying depth, area, and redshift precision, the authors train a simulation-based inference network to recover both the reionization timeline and, crucially, the parameters controlling ionizing photon escape and star formation. Their central quantitative claim is that cross-power alone recovers the four source parameters with R² > 0.92 and posterior volumes below 12% of the prior, while 21cm auto-power alone achieves R² < 0.47 and posterior volumes above 60%. If true, cross-correlation surveys become the observational route to measuring how efficiently early galaxies ionized the intergalactic medium.

What carries the argument

The central object is the 21cm–galaxy cross-power spectrum, the scale-dependent correlation between fluctuations in the 21cm brightness temperature and the galaxy overdensity field. Its characteristic anti-correlation at intermediate scales, zero-crossing at the typical ionized-bubble size, and positive correlation on small scales carry the source-bubble correlation that breaks degeneracies. The inference machinery is a conditional normalizing flow trained on thousands of forward-simulated, noise-realized power spectra, enabling likelihood-free posterior estimation without an explicit analytic likelihood.

What would settle it

Run the identical pipeline on forward simulations in which the escape fraction has a break or log-normal scatter around the power-law form; if the cross-power posteriors become biased or the claimed R² > 0.92 degrades, the central claim is simulation-specific. Alternatively, compare the inferred escape-fraction parameters with direct spectroscopically measured Lyman-continuum escape fractions at z ≈ 6–8; a systematic mismatch would falsify the recoverability claim.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the scale-dependent correlation between ionized bubbles and galaxies carries information about why and how efficiently galaxies ionized the universe—information the 21cm auto-power spectrum cannot supply. In simulation-based forecasts, cross-power spectra infer the normalization and mass slope of the escape fraction and star formation efficiency essentially without degeneracy, provided the galaxy survey has spectroscopic-quality redshifts and either reaches faint galaxies or is paired with aggressive foreground cleaning. The auto-power spectrum, by contrast, leaves these same parameters with more than 60% of the prior volume remaining, because

Load-bearing premise

The inference targets are the coefficients of power-law models for star formation efficiency and escape fraction versus halo mass, with halos treated as direct proxies for detectable galaxies; if real galaxies' detectability or ionizing output has a broken, scattered, or feedback-driven mass dependence, the recovered posteriors describe the toy model rather than physical escape fractions.

Editorial extensions

If this is right

Cross-power measurements reduce posterior volumes for global reionization properties by 20–30% over auto-power alone, confirming genuinely complementary information.
Spectroscopic-quality redshifts are a hard requirement; photometric-quality redshifts render cross-correlation measurements uninformative regardless of survey area or depth.
Tight source-property constraints are achievable through either deep galaxy surveys reaching roughly 10^10 solar-mass halos under moderate foregrounds, or brighter galaxy surveys reaching 10^11 solar-mass halos combined with aggressive 21cm foreground cleaning.
The escape fraction parameters and star formation efficiency parameters—normalizations and power-law slopes—are recoverable from cross-power with R² > 0.92 and posterior volumes below 12%, while remaining degenerate in auto-power.
The mutual information gain from cross-power grows toward higher redshift and peaks at scales corresponding to ionized-bubble sizes, identifying where survey effort yields the most scientific return.

Reading between the lines

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

If true, the result reframes survey design priorities: investment in spectroscopic redshift capability buys more science return than enlarging area, because photometric surveys add almost nothing to the cross-correlation measurement.
A natural extension would be to apply the same inference pipeline to higher-order summaries, such as the bispectrum or Minkowski functionals, to test whether the cross-power advantage persists when morphology is summarized beyond the power spectrum.
The finding that adding auto-power to cross-power actually dilutes source-property constraints at fixed network capacity suggests future multi-probe analyses should weight or architect observables explicitly rather than simply concatenating them.
The assumptions of a perfect halo-galaxy correspondence and pure power-law mass scaling are the most likely places where real data could break the claim; a scatter, break, or feedback-driven turnover in the mass dependence would likely bias the recovered coefficients.

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.

Desk Editor's Note

read the letter

Quick take: this is the first SBI treatment of 21cm-galaxy cross-power with a systematic survey scan, and the central qualitative result—that cross-power breaks degeneracies among ionizing source parameters that 21cm auto-power cannot—comes through convincingly. The 27-configuration study gives concrete advice: spectroscopic redshifts are essential, photometric z is useless, and deep galaxy surveys can compensate for small fields. That advice is well-supported within the model and worth having.

The paper is technically sound. EoRFlow and the uncertainty code are public; the authors evaluate on held-out test sets with bootstrap error bars; Appendix A honestly reports that 1D power spectra outperform 2D for cross-power, which is not what you'd naively guess. The MI decomposition is a reasonable addition.

Now the soft spots. The headline claim is stated too absolutely. 'Fundamentally inaccessible' overreaches: only 1D auto-power was tested, and the whole exercise lives inside the 21cmFAST power-law family. Real f_esc(M_h) almost certainly has scatter, a break, or feedback-driven turnover; if the true relation is outside the training family, the nominal 12% posterior volume does not cover the physical escape fraction. The halo-as-galaxy proxy in Sec. 2 compounds this—no in-mock test can catch that bias. The paper does flag the proxy, but the abstract and conclusions don't carry the caveat. The abstract also says PV~19% for the M_h,min=1e11 optimistic-foreground case, while Appendix C says 11–15%; one of those numbers is wrong and it must be fixed. The mutual-information gains are computed from the same trained flows used for inference, so they measure information accessible within this simulation family, not an external measurement; the paper almost says this but not quite. Finally, the trained models are held back until acceptance; for a survey-strategy scan, releasing them with the preprint would make the results easier to check.

Bottom line: the in-mock claims hold up and the qualitative hierarchy is probably robust. The paper is worth a serious referee—send it to review. I'd cite it for the survey-optimization conclusions and the per-bin MI map, less for the absolute source constraints. Also bring to reading group.

Referee Report

2 major / 5 minor

Summary. The paper extends the EoRFlow simulation-based inference framework to 21cm-galaxy cross-power spectra. Using 21cmFASTv4 mocks at z=6–8 with power-law f_esc(M_h) and f_*(M_h), SKA-Low noise, and galaxy survey effects, the authors train conditional normalizing flows to infer x_HI(z) and ⟨1+δ_HI⟩(z), then study survey parameter dependence, mutual information gains, and constraints on four source parameters. They report that cross-power alone tightly constrains source parameters (R²>0.92, PV<12% at M_h,min=1e10 M⊙) while auto-power is uninformative (R²<0.47, PV>60%), and that optimistic foreground removal enables similar constraints at M_h,min=1e11 M⊙. Appendices compare 1D vs 2D power spectra and test the optimistic foreground scenario.

Significance. If the central claim holds, 21cm-galaxy cross-correlations become a qualitatively new probe of reionization sources: they would break degeneracies that 21cm auto-power alone leaves unresolved, while also providing foreground-robust morphology constraints. The paper has genuine strengths: public code for the cross-power uncertainty model and EoRFlow, held-out test sets, bootstrap error bars, and an honest appendix (Appendix A) showing that 1D spherical averaging outperforms 2D for the cross-power. The main caveat is that the headline source-property result is an in-mock demonstration for a specific power-law parameterization and a halo-mass threshold model; no out-of-family test is possible with the current setup. That limits the strength of the physical claim but does not invalidate the in-mock method.

major comments (2)

[§4.4, Fig. 6; Eqs. (2)-(3)] The headline source-property constraints are in-mock results for the two-parameter power-law model of Eqs. (2)-(3). Training and test sets are drawn from the same 21cmFASTv4 family, so R²>0.92 and PV<12% demonstrate that the flow can invert this specific forward model, not that physical f_esc(M_h) and f_*(M_h) are constrained. If the real mass dependence contains scatter, a break, or feedback-driven turnover, the posteriors are projections onto toy-model coefficients and nominal credible intervals need not cover truth. The phrase 'fundamentally inaccessible to 21cm auto-power alone' (§4.4, Conclusions) is therefore too strong and should be reframed as 'within the 21cmFASTv4 power-law family', or supported by robustness tests with alternative parameterizations and stochastic scatter.
[§2 (halo proxy)] The mock galaxy overdensity is constructed by a one-to-one halo selection at a fixed M_h,min threshold. There is no luminosity scatter, duty cycle, or selection function; the statement that line luminosity scales with halo mass is not implemented quantitatively. The cross-power amplitude and scale dependence, and therefore the inferred f_esc/f_* values, depend directly on this selection. A realistic galaxy sample with scatter could shift the cross-power and bias the inference. Since coverage checks use only test sets from the same halo-selection model, no in-mock test detects this. This limitation should be stated in Section 5 and, ideally, tested with occupancy-variation mocks.

minor comments (5)

[Abstract vs Appendix C] The abstract states PV~19% for the optimistic-foreground M_h,min=1e11 case, but Appendix C reports R²>0.81 and PV between 11% and 15%, and Section 4.4 quotes PV<15%. This numerical inconsistency should be corrected.
[Table 1] Table 1 lists 'log10 fesc,10 U[0.005,0.5]' and 'log10 f∗ U[0.005,0.5]', but the text and Fig. 6 axes suggest linear ranges for f_esc,10 and f_*,10. Please clarify whether these parameters are sampled in log or linear space and use f_*,10 consistently.
[§4.1 and §4.4] R² and 'posterior volume' are not defined in the text. PV is computed as the standard deviation of posterior samples normalized by the prior standard deviation, which is a marginal-width ratio, not a multi-dimensional volume; the name is misleading. R² should state whether it uses posterior means or another point estimator.
[§4.4] The cross+auto model degrades source constraints relative to cross alone (R² 0.89–0.92, PV 19–27%). The capacity-dilution explanation is plausible, but this is a caveat to the general statement that combining measurements improves constraints; the domain of the 20–30% PV-reduction claim in §4.1 should be made explicit.
[§4.3] The mutual information estimates use the trained flow rather than the true posterior. The statement that 78% of bins show >100% fractional gain should be phrased as a property of the trained model, not an exact information decomposition of the data.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the inference targets are simulation inputs, not fitted outputs; the derivation chain is a forward-model inversion forecast.

full rationale

The paper's analysis is a simulation-based inference forecast, not a derivation that re-imports its conclusions. The forward model (21cmFASTv4) generates both the summary statistics (P21, P21,g) and the labels (x_HI, <1+delta_HI>, and the four power-law source parameters). EoRFlow is trained on one subset and evaluated on a disjoint held-out test set, with R^2 and posterior-volume metrics computed against known simulation inputs. Thus the 'predictions' are tests of invertibility of the forward model, not quantities fitted to the same data they are claimed to predict. The parameterization f_*(M_h) and f_esc(M_h) (Eqs. 2-3) is an input ansatz; the paper does not claim to derive it. The inferred source constraints are therefore conditional on this parametric family and on the halo-as-galaxy proxy, a model-dependence/correctness limitation rather than circularity. The only self-references (EoRFlow Pietschke et al. 2025; cross-power framework Hutter & Heneka 2025) point to public code and are not used as an unverified uniqueness theorem. One minor point: the abstract's PV~19% for the M_h,min=1e11 optimistic-foreground case differs from Appendix C's 11-15%; this is an internal inconsistency to correct, not a circular step. Overall, no step in the claimed chain equates a prediction to an input by construction.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The paper introduces no fitted-to-data constants; its free choices are the hand-chosen prior widths, the power-law source model, the halo-galaxy proxy, and the foreground model. Because every posterior-volume figure is normalized by the prior standard deviation, the headline '10%', '12%', and '19%' values are prior-relative, not absolute measurement errors. The optimistic-foreground appendix is the most fragile choice: it postulates recovery of all wedge modes to the primary-beam scale, and the M_h,min=1e11 source-property results depend entirely on it. No new physical entities are introduced; EoRFlow is software, and the four astrophysical parameters are prior-drawn inference targets.

free parameters (4)

log10 f_esc,10 prior = U[0.005, 0.5]
Uniform prior width chosen by hand (Table 1); sets the escape-fraction normalization and forms the denominator of the posterior-volume metric.
alpha_esc prior = U[-0.8, 0.5]
Hand-chosen prior on escape-fraction mass dependence; the reported source-property constraints are relative to this width.
log10 f_*,10 prior = U[0.005, 0.5]
Hand-chosen prior on star-formation-efficiency normalization (Table 1).
alpha_* prior = U[-0.3, 0.9]
Hand-chosen prior on SFE mass dependence; the headline PV<12% source-property result is normalized by this prior.

assumptions (6)

domain assumption Saturated spin temperature T_S >> T_CMB at z=6-8, so deltaT_b = T_0 x_HI (1+delta) (Eq. 1)
Assumed in Section 2 based on HERA constraints; if X-ray heating is incomplete at z~8 the signal carries spin-temperature structure the inference ignores.
domain assumption Power-law parameterization of f_*(M_h) and f_esc(M_h) (Eqs. 2-3)
The inferred 'source properties' are coefficients of these power laws; real departures from power-law form break the physical interpretation of the posteriors.
domain assumption Halo mass is a direct proxy for galaxy detectability in emission-line surveys (Section 2)
Galaxy selection reduces to a mass threshold; line-luminosity scatter, dust, duty cycle, and AGN contamination are not included in the mock galaxy field.
domain assumption Foreground wedge model: moderate = horizon + 0.1 Mpc^-1 buffer; optimistic = all modes down to primary-beam scale (App. C)
The optimistic scenario is load-bearing for source-property constraints at M_h,min=1e11 M_sun; if wedge modes cannot be cleaned to the beam scale, that pathway collapses.
ad hoc to paper Prior truncation: simulations with x_HI>0.7 or x_HI<0.2 across all z=6-8 are discarded (Section 2)
Filters the simulation suite to a chosen reionization-timing window, shaping the distribution of morphologies seen in training and test data.
domain assumption Fixed Planck-2018 cosmology (Omega_b=0.049, Omega_m=0.31, h=0.68)
Cosmology is an input, not marginalized; the T_0=27 mK normalization of Eq. (4) depends on it.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Constraining reionization morphology and source properties with 21cm galaxy cross-correlation surveys." pith.science (2026). https://pith.science/paper/F6HBRIXG

@misc{pith2026260118627,
  author       = {Pith},
  title        = {Pith review of: Constraining reionization morphology and source properties with 21cm galaxy cross-correlation surveys},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F6HBRIXG}},
  note         = {Machine review of arXiv:2601.18627}
}

abstract

Cross-correlations between 21cm observations and galaxy surveys provide a powerful probe of reionization by providing robustness against foreground contamination while linking ionization morphology to galaxies. We quantified the constraining power of 21cm galaxy cross-power spectra for inferring the neutral hydrogen fraction, $x_\mathrm{HI}(z),$ and mean overdensity, $\langle 1+\delta_\mathrm{HI} \rangle(z)$, exploring dependence on the field of view; redshift precision, $\sigma_z$; and minimum halo mass, $M_\mathrm{h,min}$. We employed our simulation-based inference framework EoRFlow for likelihood-free parameter estimation. Mock observations include thermal noise for 100h of SKA-Low with foreground avoidance and realistic galaxy-survey effects. For a fiducial survey ($\mathrm{FOV}=100\,\mathrm{deg}^2$, $\sigma_z=0.001$, $M_\mathrm{h,min}=10^{11}\mathrm{M}_\odot$), cross-power spectra yield unbiased constraints with posterior volumes (PVs) of $\sim$10% relative to priors. Cross-power measurements reduce the PV by 20-30% versus 21cm auto-power alone. With foreground avoidance, spectroscopic redshift precision is essential; photometric redshifts render cross-correlations uninformative. Notably, cross-power spectra constrain ionizing source properties, the escape fraction $f_\mathrm{esc,}$ and the star formation efficiency $f_*$, which remain degenerate in auto-power (PV >60%). Tight constraints require either deep surveys detecting faint galaxies ($M_\mathrm{h,min} \sim 10^{10}\mathrm{M}_\odot$) with moderate foregrounds (PV~11%) or conservative mass limits with optimistic foreground removal (PV~19%). 21cm galaxy cross-correlations enhance morphology constraints beyond auto-power while enabling previously inaccessible source property constraints. Realizing full potential requires precise redshifts and either faint galaxy detection limits or improved 21cm foreground cleaning.

Figures

Figures reproduced from arXiv: 2601.18627 by the authors.

**Figure 1.** 21cm-Galaxy cross-power spectrum at z = 7.6 (xHI = 0.75) for the fiducial survey configuration explored in the inference analysis in Section 4.1 (FOV = 100deg2 , σz = 0.001, Mh,min = 1011M⊙). The blue line shows the physical signal, the shaded blue region indicates the 1σ uncertainty, and orange points represent the mock observation including instrumental noise. Three distinct regimes characterize the cross-power s… view at source ↗

**Figure 2.** Marginalized posteriors for the neutral fraction [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Informativeness measured by the normalized posterior volume on the neutral fraction [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Posterior volume (averaged over neutral fraction, density, redshifts, and test observations) as a function of galaxy survey [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Fractional mutual information gain from including 21cm [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

**Figure 6.** Figure 6: Inference of astrophysical source parameters from 21cm auto- and 21cm-Galaxy cross-power spectra for a randomly chosen [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Nearest Neighbour-Based Statistics for 21cm-Galaxy Cross-Correlations in the Epoch of Reionization
astro-ph.CO 2026-02 conditional novelty 7.0 of 10

kNN CDF statistics detect 21cm-galaxy cross-correlations more effectively than two-point methods and distinguish reionization models at fixed ionized fraction even with noise and foregrounds.
A hidden reionization prior biases cosmological inference
astro-ph.CO 2026-06 unverdicted novelty 5.0 of 10

No monotonic reionization history fits Planck τ, patchy kSZ from SPT/ACT, and Lyα endpoint; an early ionization phase at z≳12 relaxes ∑mν<0.39 eV and shifts σ8 via As-τ degeneracy.

Reference graph

Works this paper leans on

55 extracted references · 7 linked inside Pith · cited by 2 Pith papers

[1]

Abdurashidova, T. H. C. Z., Adams, T., Aguirre, J. E., et al. 2023, ApJ, 945, 124

2023
[2]

E., Alexander, P., et al

Abdurashidova, Z., Aguirre, J. E., Alexander, P., et al. 2022, ApJ, 924, 51

2022
[3]

E., Alexander, P., et al

Abdurashidova, Z., Aguirre, J. E., Alexander, P., et al. 2022, ApJ, 925, 221

2022
[4]

2024, MNRAS, 534, L30

Acharya, A., Mertens, F., Ciardi, B., et al. 2024, MNRAS, 534, L30

2024
[5]

Aghanim, N. et al. 2020, Astron. Astrophys., 641, A6, [Erratum: As- tron.Astrophys. 652, C4 (2021)]

2020
[6]

2019, MNRAS, 488, 4440

Alsing, J., Charnock, T., Feeney, S., & Wandelt, B. 2019, MNRAS, 488, 4440

2019
[7]

2024, in 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, V olume 2 (ASPLOS ’24) (ACM)

Ansel, J., Yang, E., He, H., et al. 2024, in 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, V olume 2 (ASPLOS ’24) (ACM)

2024
[8]

2019, ApJ, 874, 133

Beane, A., Villaescusa-Navarro, F., & Lidz, A. 2019, ApJ, 874, 133

2019

Show all 55 references

[9]

R., Mertens, F

Ceccotti, E., Offringa, A. R., Mertens, F. G., et al. 2025, First upper limits on the 21-cm signal power spectrum of neutral hydrogen atz=9.16 from the LOFAR 3C196 field

2025
[10]

K., Bianco, M., et al

Cerardi, N., Giri, S. K., Bianco, M., et al. 2025, arXiv e-prints, arXiv:2511.11568

2025 arXiv
[11]

K., Witte, S

Cole, A., Miller, B. K., Witte, S. J., et al. 2022, J. Cosmology Astropart. Phys., 2022, 004

2022
[12]

Cooper, N., Norregaard, C., Meriot, R., & Pritchard, J. R. 2025, arXiv e-prints, arXiv:2508.16329

2025 arXiv
[13]

Cover, T. M. & Thomas, J. A. 2006, Elements of Information Theory, 2nd edn. (Hoboken, NJ: Wiley-Interscience)

2006
[14]

E., Mesinger, A., & Murray, S

Davies, J. E., Mesinger, A., & Murray, S. G. 2025, A&A, 701, A236

2025
[15]

R., Parsons, A

DeBoer, D. R., Parsons, A. R., Aguirre, J. E., et al. 2017, Publications of the Astronomical Society of the Pacific, 129, 045001

2017
[16]

2019, Neural Spline Flows

Durkan, C., Bekasov, A., Murray, I., & Papamakarios, G. 2019, Neural Spline Flows

2019
[17]

Furlanetto, S. R. & Lidz, A. 2007, ApJ, 660, 1030

2007
[18]

R., Oh, S

Furlanetto, S. R., Oh, S. P., & Briggs, F. H. 2006, Phys. Rep., 433, 181

2006
[19]

E., & Mesinger, A

Gagnon-Hartman, S., Davies, J. E., & Mesinger, A. 2025, A&A, 699, A131

2025
[20]

2022, arXiv e-prints, arXiv:2206.14908

Greene, J., Bezanson, R., Ouchi, M., Silverman, J., & the PFS Galaxy Evolution Working Group. 2022, arXiv e-prints, arXiv:2206.14908

2022 arXiv
[21]

& Cooray, A

Heneka, C. & Cooray, A. 2021, MNRAS, 506, 1573

2021
[22]

2017, ApJ, 848, 52

Heneka, C., Cooray, A., & Feng, C. 2017, ApJ, 848, 52

2017
[23]

& Mesinger, A

Heneka, C. & Mesinger, A. 2020, MNRAS, 496, 581 HERA Collaboration, Abdurashidova, Z., Adams, T., et al. 2023, ApJ, 945, 124

2020
[24]

Hutter, A., Dayal, P., Müller, V ., & Trott, C. M. 2017, ApJ, 836, 176

2017
[25]

& Heneka, C

Hutter, A. & Heneka, C. 2025, arXiv e-prints, arXiv:2509.15906

2025
[26]

2018, MNRAS, 479, 2754 La Plante, P., Mirocha, J., Gorce, A., Lidz, A., & Parsons, A

Kubota, K., Yoshiura, S., Takahashi, K., et al. 2018, MNRAS, 479, 2754 La Plante, P., Mirocha, J., Gorce, A., Lidz, A., & Parsons, A. 2023, ApJ, 944, 59

2018
[27]

Lipman, Y ., Chen, R. T. Q., Ben-Hamu, H., Nickel, M., & Le, M. 2022, arXiv e-prints, arXiv:2210.02747

2022 arXiv
[28]

& Hutter, F

Loshchilov, I. & Hutter, F. 2019, Decoupled Weight Decay Regularization

2019
[29]

I., Brinchmann, J., et al

Mainieri, V ., Anderson, R. I., Brinchmann, J., et al. 2024, arXiv e-prints, arXiv:2403.05398

2024 arXiv
[30]

2020, The Messenger, 180, 24

Maiolino, R., Cirasuolo, M., Afonso, J., et al. 2020, The Messenger, 180, 24

2020
[31]

G., Mevius, M., Koopmans, L

Mertens, F. G., Mevius, M., Koopmans, L. V . E., et al. 2020, Monthly Notices of the Royal Astronomical Society, 493, 1662

2020
[32]

2011, Monthly No- tices of the Royal Astronomical Society, 411, 955, _eprint: https://academic.oup.com/mnras/article-pdf/411/2/955/4099991/mnras0411- 0955.pdf

Mesinger, A., Furlanetto, S., & Cen, R. 2011, Monthly No- tices of the Royal Astronomical Society, 411, 955, _eprint: https://academic.oup.com/mnras/article-pdf/411/2/955/4099991/mnras0411- 0955.pdf

2011
[33]

2024, MNRAS, 530, 3183

Moriwaki, K., Beane, A., & Lidz, A. 2024, MNRAS, 530, 3183

2024
[34]

G., Greig, B., Mesinger, A., et al

Murray, S. G., Greig, B., Mesinger, A., et al. 2020, Journal of Open Source Soft- ware, 5, 2582

2020
[35]

2025, SciPost Phys., 18, 155

Ore, A., Heneka, C., & Plehn, T. 2025, SciPost Phys., 18, 155

2025
[36]

Park, J., Kim, H.-S., Wyithe, J. S. B., & Lacey, C. G. 2014, MNRAS, 438, 2474

2014
[37]

2019, Monthly Notices of the Royal Astronomical Society, 484, 933–949

Park, J., Mesinger, A., Greig, B., & Gillet, N. 2019, Monthly Notices of the Royal Astronomical Society, 484, 933–949

2019
[38]

R., Backer, D

Parsons, A. R., Backer, D. C., Foster, G. S., et al. 2010, The Astronomical Jour- nal, 139, 1468

2010
[39]

Pietschke, Y ., Heneka, C., Schlenker, T., Ore, A., & Schosser, B. 2025, J. Cos- mology Astropart. Phys., 2025, 039

2025
[40]

C., Liu, A., Dillon, J

Pober, J. C., Liu, A., Dillon, J. S., et al. 2014, ApJ, 782, 66

2014
[41]

C., Parsons, A

Pober, J. C., Parsons, A. R., DeBoer, D. R., et al. 2013, The Astronomical Jour- nal, 145, 65

2013
[42]

D., Weniger, C., Acedo, E

Saxena, A., Meerburg, P. D., Weniger, C., Acedo, E. d. L., & Handley, W. 2024, RAS Techniques and Instruments, 3, 724

2024
[43]

2025, SciPost Phys

Schosser, B., Heneka, C., & Plehn, T. 2025, SciPost Phys. Core, 8, 037

2025
[44]

Schosser, B., Heneka, C., & Schäfer, B. M. 2025, arXiv e-prints, arXiv:2508.10094

2025 arXiv
[45]

2016, MNRAS, 459, 2741 The HERA Collaboration, Abdurashidova, Z., Adams, T., et al

Sobacchi, E., Mesinger, A., & Greig, B. 2016, MNRAS, 459, 2741 The HERA Collaboration, Abdurashidova, Z., Adams, T., et al. 2025, arXiv e- prints, arXiv:2511.21289

2016
[46]

J., Goeke, R., Bowman, J

Tingay, S. J., Goeke, R., Bowman, J. D., et al. 2013, PASA, 30, e007

2013
[47]

2023, arXiv e-prints, arXiv:2302.00482

Tong, A., Fatras, K., Malkin, N., et al. 2023, arXiv e-prints, arXiv:2302.00482

2023 arXiv
[48]

M., Jordan, C

Trott, C. M., Jordan, C. H., Midgley, S., et al. 2020, Monthly Notices of the Royal Astronomical Society, 493, 4711 van Haarlem, M. P., Wise, M. W., Gunst, A. W., et al. 2013, A&A, 556, A2

2020
[49]

2022, ApJS, 259, 61

Villaescusa-Navarro, F., Genel, S., Anglés-Alcázar, D., et al. 2022, ApJS, 259, 61

2022
[50]

2020, MNRAS, 492, 4952

Vrbanec, D., Ciardi, B., Jeli´c, V ., et al. 2020, MNRAS, 492, 4952

2020
[51]

2016, MNRAS, 457, 666

Vrbanec, D., Ciardi, B., Jeli´c, V ., et al. 2016, MNRAS, 457, 666

2016
[52]

2022, ApJ, 928, 1

Wang, Y ., Zhai, Z., Alavi, A., et al. 2022, ApJ, 928, 1

2022
[53]

H., Kulkarni, G., & Haehnelt, M

Weinberger, L. H., Kulkarni, G., & Haehnelt, M. G. 2020, MNRAS, 494, 703

2020
[54]

Yoshiura, S., Line, J. L. B., Kubota, K., Hasegawa, K., & Takahashi, K. 2018, MNRAS, 479, 2767

2018
[55]

optimistic foreground model

Yoshiura, S., Pindor, B., Line, J. L. B., et al. 2021, Monthly Notices of the Royal Astronomical Society, 505, 4775 Article number, page 12 of 15 Pietschke et al.: Constraining Reionization Morphology and Source Properties with 21cm-Galaxy Cross-Correlation Surveys Appendix A:...

2021

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.

[1] [1]

Abdurashidova, T. H. C. Z., Adams, T., Aguirre, J. E., et al. 2023, ApJ, 945, 124

2023

[2] [2]

E., Alexander, P., et al

Abdurashidova, Z., Aguirre, J. E., Alexander, P., et al. 2022, ApJ, 924, 51

2022

[3] [3]

E., Alexander, P., et al

Abdurashidova, Z., Aguirre, J. E., Alexander, P., et al. 2022, ApJ, 925, 221

2022

[4] [4]

2024, MNRAS, 534, L30

Acharya, A., Mertens, F., Ciardi, B., et al. 2024, MNRAS, 534, L30

2024

[5] [5]

Aghanim, N. et al. 2020, Astron. Astrophys., 641, A6, [Erratum: As- tron.Astrophys. 652, C4 (2021)]

2020

[6] [6]

2019, MNRAS, 488, 4440

Alsing, J., Charnock, T., Feeney, S., & Wandelt, B. 2019, MNRAS, 488, 4440

2019

[7] [7]

2024, in 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, V olume 2 (ASPLOS ’24) (ACM)

Ansel, J., Yang, E., He, H., et al. 2024, in 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, V olume 2 (ASPLOS ’24) (ACM)

2024

[8] [8]

2019, ApJ, 874, 133

Beane, A., Villaescusa-Navarro, F., & Lidz, A. 2019, ApJ, 874, 133

2019