pith. sign in

arxiv: 2605.15360 · v1 · pith:BYRDKTAWnew · submitted 2026-05-14 · 🌌 astro-ph.HE

Correlation Between X-Ray and Cosmic Neutrino Sources: From Obscured AGN to Blazars

Emma Kun , Imre Bartos , Claudio Ricci , Santiago del Palacio , Francis Halzen , Julia Becker Tjus , Peter L. Biermann , Anna Franckowiak This is my paper

Pith reviewed 2026-05-19 15:28 UTC · model grok-4.3

classification 🌌 astro-ph.HE
keywords neutrino sourcesblazarsactive galactic nucleihard X-ray emissionIceCubemultimessenger astronomyphotohadronic processes
0
0 comments X p. Extension
pith:BYRDKTAW Add to your LaTeX paper What is a Pith Number? 
\usepackage{pith}
\pithnumber{BYRDKTAW}

Prints a linked pith:BYRDKTAW badge after your title and writes the identifier into PDF metadata. Compiles on arXiv with no extra files. Learn more

The pith

Seven blazars are statistically consistent with the hard X-ray to neutrino luminosity relation calibrated on six AGN.

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

The paper checks whether seven NuSTAR-observed blazars with near-threshold IceCube excesses follow the same correlation between unabsorbed hard X-ray luminosity and high-energy neutrino luminosity previously reported for six active galactic nuclei. A Bayesian regression on the original six sources recovers a slope consistent with one and an intrinsic scatter of roughly 0.6 dex. All seven new objects pass a posterior-predictive consistency test under the assumption that the published neutrino best-fit values represent signal. A flux-space permutation test on the combined thirteen sources rejects random pairing at 3.23 sigma after controlling for distance bias.

Core claim

The seven new blazars lie within the posterior predictive distribution of the log L_nu versus log L_hX relation calibrated on the six AGN, and the joint sample shows a statistically significant rejection of random L_hX--L_nu pairing.

What carries the argument

Bayesian linear regression of log neutrino luminosity on log hard X-ray luminosity with errors in both coordinates and intrinsic scatter, followed by posterior predictive checks and a distance-controlled permutation test in flux space.

If this is right

  • Both the original AGN and the new blazars fall inside the photohadronic prediction band for neutrino production in compact photon-rich regions.
  • The relation implies that neutrino output scales roughly linearly with hard X-ray output across these populations.
  • Detection-level confirmation would require either more calibration sources or an X-ray-weighted stacking analysis of IceCube data.

Where Pith is reading between the lines

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

  • If the correlation is physical, X-ray catalogs could be used to prioritize future neutrino follow-up observations.
  • The scatter of 0.6 dex may trace differences in Doppler boosting or target photon density between blazars and obscured AGN.
  • Repeating the test on an independent sample selected by X-ray flux rather than neutrino significance would test selection bias.

Load-bearing premise

The published IceCube best-fit neutrino numbers reflect actual signals rather than background fluctuations.

What would settle it

A larger sample containing even one source with high hard X-ray luminosity but a neutrino flux far below the calibrated relation, or vice versa, would break the claimed consistency.

Figures

Figures reproduced from arXiv: 2605.15360 by Anna Franckowiak, Claudio Ricci, Emma Kun, Francis Halzen, Imre Bartos, Julia Becker Tjus, Peter L. Biermann, Santiago del Palacio.

Figure 1
Figure 1. Figure 1: Posterior-predictive consistency test. The blue line shows the plug-in posterior-predictive median of the K2024 relation calibrated on the six published sources only (shown as red filled circles), with shaded 68% (dark) and 95% (light) bands set by ±σint,med and ±1.96 σint,med around the median line. Open blue squares show the seven new blazars analyzed in this work. The dashed line marks LhX = Lν. All sev… view at source ↗
Figure 2
Figure 2. Figure 2: Null calibration of the posterior-predictive consistency test. Left: Distribution of χ 2 7 under a background-only null in which ˆns values are drawn from Poisson(λbg) truncated to ˆns > 0, for four background expectations λbg ∈ {0.3, 0.5, 1.0, 2.0}. The black dashed curve is the χ 2 7 distribution that would obtain if the test were correctly calibrated against the K2024 relation (median ∼ 7). The observed… view at source ↗
Figure 3
Figure 3. Figure 3: Distance-free log(LhX/Lν) ratio diagnostic. Red filled circles: six K2024 calibration sources; open blue squares: seven new blazars from this work. Vertical dashed lines show inverse-variance-weighted means with shaded 68% bands. The grey band indicates the photohadronic prediction range κ ∈ [0.1, 1]. Population-weighted κ values are 0.67 (calibration) and 0.43 (new sample); the two means differ by −0.19±0… view at source ↗
Figure 4
Figure 4. Figure 4: Flux-space permutation test as a finite-sample diagnostic. Left: Observed joint sample with the weighted forward regression and its ±σ w scat band. Middle: Null distribution from a luminosity-space permutation that breaks the d 2 L coupling in the null but preserves it in the observation, yielding an inflated nominal significance of 5.07 σ. Right: Null distribution from the flux-space permutation adopted h… view at source ↗
Figure 5
Figure 5. Figure 5: Monthly Fermi-LAT 0.1–100 GeV photon flux (top) and NuSTAR 15–55 keV energy flux (bottom) of 3C 454.3 as a function of time. Filled circles show Fermi-LAT detections (TS ≥ 4); downward triangles indicate upper limits (TS < 4). The dashed line marks the quiescent mean flux over 2017–2022. Orange shaded bands and dotted vertical lines indicate the epochs of NuSTAR observations. 14 http://www.astropy.org [PI… view at source ↗
Figure 6
Figure 6. Figure 6: Fermi-LAT and NuSTAR light curves of, from top to bottom: PKS 1441+25 (FSRQ, z = 0.939; the neutrino spectral index was fitted freely); MITG J201534+3710 (FSRQ, z = 0.858); and S3 0458−02 (FSRQ, z = 2.286). Symbols and lines as in [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Fermi-LAT and NuSTAR light curves of, from top to bottom: OJ 287 (BL Lac, z = 0.306); S2 0109+22 (IBL, z = 0.265; the neutrino spectral index was fitted freely); and Ton 599 (FSRQ, z = 0.725). Symbols and lines as in [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
read the original abstract

The origin of high-energy astrophysical neutrinos remains a key open question in multimessenger astrophysics. A correlation between unabsorbed hard X-ray emission and high-energy neutrino luminosity has been reported in a sample of six active galactic nuclei with the highest individual IceCube significances, linking neutrino production to compact, photon-rich environments near supermassive black holes. In this work we study whether the threshold-near IceCube excesses associated with seven NuSTAR-observed blazars are statistically consistent with that established relation. Calibrating the relation between the neutrino and hard X-ray luminosities as $\log L_\nu = \alpha + \beta \log L_\mathrm{hX} + \mathcal{N}(0, \sigma_{\rm int}^2)$ on the six published sources via a Bayesian regression with errors on both axes, the recovered slope is consistent with $\beta = 1$, and the intrinsic scatter is $\sim 0.6$\,dex. All seven new blazars are posterior-predictively consistent with this calibration ($\chi^2_7 = 1.58$, $p = 0.98$) under the working hypothesis that the published IceCube best-fit neutrino numbers $\hat{n}_s$ values reflect the signal. A null-injection test confirms that, given the present calibration sample size, the consistency test does not by itself adjudicate between signal and selected-background origins. A distance-free $L_\mathrm{hX}/L_\nu$ ratio diagnostic places both populations within the photohadronic prediction band, statistically indistinguishable. A flux-space permutation test on the 13-source joint sample, with construction-controlled $d_L^{\,2}$ distance bias, rejects random pairing $L_\mathrm{hX}$--$L_\nu$ with $p = 6.3 \times 10^{-4}$ ($3.23\,\sigma$). We interpret these results as a conditional consistency check; a detection-level statement requires either an enlarged calibration set or an X-ray-weighted IceCube stacking likelihood with internal data.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The manuscript calibrates a Bayesian linear regression relating hard X-ray luminosity L_hX to neutrino luminosity L_ν on a sample of six AGN with the highest individual IceCube significances, recovering a slope consistent with unity and intrinsic scatter ~0.6 dex. It then performs a posterior-predictive consistency check on seven additional NuSTAR-observed blazars, finding χ²_7 = 1.58 (p = 0.98) under the assumption that the published n̂_s values trace signal. A distance-controlled flux-space permutation test on the joint 13-source sample rejects random L_hX--L_ν pairing at p = 6.3 × 10^{-4} (3.23σ). The authors present these results as a conditional consistency check supporting a physical correlation, while acknowledging that an enlarged calibration set or X-ray-weighted stacking analysis would be needed for a detection-level claim.

Significance. If the correlation is physical, the result would link high-energy neutrino production to compact, photon-rich environments traced by unabsorbed hard X-rays, providing support for photohadronic models in AGN. The Bayesian regression with errors in both coordinates, the explicit null-injection test for the consistency check, and the construction-controlled permutation test are methodological strengths that increase the credibility of the analysis relative to simpler correlation tests. The small calibration sample (n=6) and the explicit demonstration that the consistency test cannot yet distinguish signal from selected background are appropriately highlighted as limitations.

major comments (1)
  1. [Permutation test] Permutation test (flux-space analysis on the 13-source joint sample): the reported p = 6.3 × 10^{-4} is presented as independent evidence against random pairing. However, no equivalent null-injection test is described under the background-only hypothesis for the n̂_s values, unlike the explicit null test performed for the posterior-predictive consistency check. Because source selection is based on IceCube significance thresholds and the calibration sample is small, it remains possible that the low p-value arises from residual selection effects or distance correlations rather than a physical L_hX--L_ν relation; a background-only permutation test is required to establish that the reported significance is robust.
minor comments (2)
  1. [Abstract and Methods] The abstract and main text should explicitly state the number of free parameters in the Bayesian regression (intercept, slope, and intrinsic scatter) when reporting the recovered values.
  2. [Results] Figure captions for the L_hX/L_ν diagnostic plot should include the exact photohadronic prediction band boundaries used for the statistical comparison.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful and constructive review. We address the major comment below and agree that the suggested addition will strengthen the robustness discussion.

read point-by-point responses

  1. Referee: [Permutation test] Permutation test (flux-space analysis on the 13-source joint sample): the reported p = 6.3 × 10^{-4} is presented as independent evidence against random pairing. However, no equivalent null-injection test is described under the background-only hypothesis for the n̂_s values, unlike the explicit null test performed for the posterior-predictive consistency check. Because source selection is based on IceCube significance thresholds and the calibration sample is small, it remains possible that the low p-value arises from residual selection effects or distance correlations rather than a physical L_hX--L_ν relation; a background-only permutation test is required to establish that the reported significance is robust.

    Authors: We appreciate the referee highlighting this important point. The flux-space permutation test evaluates the null of random L_hX--L_ν pairing on the observed values while explicitly controlling for d_L² distance bias to remove the most obvious selection-induced correlation. The n̂_s values (and thus L_ν) are fixed at their published values, which already embed the IceCube significance-based selection. Nevertheless, we agree that an explicit background-only injection test for the permutation statistic itself would provide a more complete demonstration that the observed p-value is not an artifact of how the joint sample was assembled. We will therefore add this test in the revised manuscript: mock n̂_s will be drawn from the background-only distribution consistent with IceCube point-source sensitivity at the source declinations, the corresponding L_ν computed, and the full distance-controlled permutation procedure repeated to obtain the distribution of p-values under the background-only hypothesis. The results will be reported in Section 4 (or a new appendix) with the same transparency used for the existing null-injection test on the consistency check. This directly addresses the concern without altering the main conclusions. revision: yes

Circularity Check

0 steps flagged

No circularity: calibration and permutation test remain independent

full rationale

The derivation calibrates a luminosity relation via Bayesian regression on six sources, then applies posterior-predictive consistency to seven new blazars and a distance-controlled flux-space permutation test to the joint sample. Neither step reduces to its inputs by construction: the consistency check evaluates new data against the fitted posterior without forcing agreement, while the permutation test directly assesses whether the observed L_hX–L_ν pairing is improbable under random reassignment. The paper explicitly flags the working hypothesis on n̂_s and reports a null-injection test for the consistency step, confirming the analysis does not equate fitted parameters with predictions. No self-citations, uniqueness theorems, or ansatzes are invoked as load-bearing premises. The statistical tests supply external validation independent of the calibration values themselves.

Axiom & Free-Parameter Ledger

3 free parameters · 2 axioms · 0 invented entities

The central claim depends on a fitted linear relation whose parameters are determined from the six-source sample and on the assumption that IceCube n̂_s values trace real neutrino signals. The permutation test reduces reliance on the fit but the small sample size remains a limiting factor.

free parameters (3)
  • intercept α
    Fitted parameter in the Bayesian regression log L_ν = α + β log L_hX
  • slope β
    Fitted parameter recovered from Bayesian regression on the six AGN; reported as consistent with 1
  • intrinsic scatter σ_int
    Fitted scatter term reported as ~0.6 dex from the calibration sample
axioms (2)
  • domain assumption Published IceCube best-fit n̂_s values reflect the neutrino signal rather than background
    Explicitly invoked as the working hypothesis for the posterior-predictive consistency test on the seven blazars
  • domain assumption The six AGN and seven blazars are drawn from populations that share the same underlying L_hX–L_ν relation
    Required to interpret the consistency and joint permutation results as evidence for a common physical mechanism

pith-pipeline@v0.9.0 · 5943 in / 1734 out tokens · 74004 ms · 2026-05-19T15:28:16.379688+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages · 3 internal anchors

  1. [1]

    G., Abbasi, R., Abdou, Y., et al

    Aartsen, M. G., Abbasi, R., Abdou, Y., et al. 2013, PhRvL, 111, 021103, doi: 10.1103/PhysRevLett.111.021103

    work page doi:10.1103/physrevlett.111.021103 2013

  2. [2]

    G., Ackermann, M., Adams, J., et al

    Aartsen, M. G., Ackermann, M., Adams, J., et al. 2020, PhRvL, 124, 051103, doi: 10.1103/PhysRevLett.124.051103

    work page doi:10.1103/physrevlett.124.051103 2020

  3. [3]

    2022, Science, 378, 538, doi: 10.1126/science.abg3395

    Abbasi, R., Ackermann, M., Adams, J., et al. 2022, Science, 378, 538, doi: 10.1126/science.abg3395

    work page doi:10.1126/science.abg3395 2022

  4. [4]

    2024, arXiv e-prints, arXiv:2406.07601, doi: 10.48550/arXiv.2406.07601

    Abbasi, R., Ackermann, M., Adams, J., et al. 2024, arXiv e-prints, arXiv:2406.07601, doi: 10.48550/arXiv.2406.07601

    work page doi:10.48550/arxiv.2406.07601 2024

  5. [5]

    2023, ApJS, 265, 31, doi: 10.3847/1538-4365/acbb6a

    Abdollahi, S., Ajello, M., Baldini, L., et al. 2023, ApJS, 265, 31, doi: 10.3847/1538-4365/acbb6a

    work page doi:10.3847/1538-4365/acbb6a 2023

  6. [6]

    B., et al

    Ackermann, M., Ajello, M., Atwood, W. B., et al. 2015, ApJ, 810, 14, doi: 10.1088/0004-637X/810/1/14

    work page doi:10.1088/0004-637x/810/1/14 2015

  7. [7]

    L., Ansoldi, S., Antonelli, L

    Ahnen, M. L., Ansoldi, S., Antonelli, L. A., et al. 2015, ApJL, 815, L23, doi: 10.1088/2041-8205/815/2/L23

    work page doi:10.1088/2041-8205/815/2/l23 2015

  8. [8]

    L., Biermann, P

    Allen, M. L., Biermann, P. L., Chieffi, A., et al. 2024, Astroparticle Physics, 161, 102976, doi: 10.1016/j.astropartphys.2024.102976

    work page doi:10.1016/j.astropartphys.2024.102976 2024

  9. [9]

    Arnaud, K. A. 1996, in Astronomical Society of the Pacific Conference Series, Vol. 101, Astronomical Data Analysis Software and Systems V, ed. G. H. Jacoby & J. Barnes, 17 Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, A33, doi: 10.1051/0004-6361/201322068 Astropy Collaboration, Price-Whelan, A. M., Sip˝ ocz, B. M., et a...

    work page doi:10.1051/0004-6361/201322068 1996

  10. [10]

    J., Cowie, L

    Barger, A. J., Cowie, L. L., Mushotzky, R. F., & Richards, E. A. 2001, AJ, 121, 662, doi: 10.1086/318742

    work page doi:10.1086/318742 2001

  11. [11]

    C., Blandford, R

    Begelman, M. C., Blandford, R. D., & Rees, M. J. 1984, Reviews of Modern Physics, 56, 255, doi: 10.1103/RevModPhys.56.255

    work page doi:10.1103/revmodphys.56.255 1984

  12. [12]

    2022, ApJ, 939, 43, doi: 10.3847/1538-4357/ac9588

    Eichmann, B., Oikonomou, F., Salvatore, S., Dettmar, R.-J., & Becker Tjus, J. 2022, ApJ, 939, 43, doi: 10.3847/1538-4357/ac9588

    work page doi:10.3847/1538-4357/ac9588 2022

  13. [13]

    , keywords =

    Haardt, F., & Maraschi, L. 1991, ApJL, 380, L51, doi: 10.1086/186171

    work page doi:10.1086/186171 1991

  14. [14]

    2022, International Journal of Modern Physics D, 31, 2230003, doi: 10.1142/S0218271822300038

    Halzen, F. 2022, International Journal of Modern Physics D, 31, 2230003, doi: 10.1142/S0218271822300038

    work page doi:10.1142/s0218271822300038 2022

  15. [15]

    , keywords =

    Hewett, P. C., & Wild, V. 2010, MNRAS, 405, 2302, doi: 10.1111/j.1365-2966.2010.16648.x HI4PI Collaboration, Ben Bekhti, N., Fl¨ oer, L., et al. 2016, A&A, 594, A116, doi: 10.1051/0004-6361/201629178

    work page doi:10.1111/j.1365-2966.2010.16648.x 2010

  16. [16]

    Data analysis recipes: Fitting a model to data

    Hogg, D. W., Bovy, J., & Lang, D. 2010, arXiv e-prints, arXiv:1008.4686, doi: 10.48550/arXiv.1008.4686 IceCube Collaboration. 2022, Dataset, doi: 10.21234/03fq-rh11

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1008.4686 2010

  17. [17]

    2020, ApJL, 891, L33, doi: 10.3847/2041-8213/ab7661

    Inoue, Y., Khangulyan, D., & Doi, A. 2020, ApJL, 891, L33, doi: 10.3847/2041-8213/ab7661

    work page doi:10.3847/2041-8213/ab7661 2020

  18. [18]

    2022, ApJ, 938, 87, doi: 10.3847/1538-4357/ac8794

    Kawamuro, T., Ricci, C., Imanishi, M., et al. 2022, ApJ, 938, 87, doi: 10.3847/1538-4357/ac8794

    work page doi:10.3847/1538-4357/ac8794 2022

  19. [19]

    I., Kovalev, Y

    Kellermann, K. I., Kovalev, Y. Y., Lister, M. L., et al. 2007, Ap&SS, 311, 231, doi: 10.1007/s10509-007-9622-5

    work page doi:10.1007/s10509-007-9622-5 2007

  20. [20]

    Kelly, B. C. 2007, ApJ, 665, 1489, doi: 10.1086/519947

    work page doi:10.1086/519947 2007

  21. [21]

    J., Ricci, C., Trakhtenbrot, B., et al

    Koss, M. J., Ricci, C., Trakhtenbrot, B., et al. 2022, ApJS, 261, 2, doi: 10.3847/1538-4365/ac6c05

    work page doi:10.3847/1538-4365/ac6c05 2022

  22. [22]

    Y., Pushkarev, A

    Kovalev, Y. Y., Pushkarev, A. B., G´ omez, J. L., et al. 2025, A&A, 700, L12, doi: 10.1051/0004-6361/202555400

    work page doi:10.1051/0004-6361/202555400 2025

  23. [23]

    2023, A&A, 679, A46, doi: 10.1051/0004-6361/202346710

    Kun, E., Bartos, I., Becker Tjus, J., et al. 2023, A&A, 679, A46, doi: 10.1051/0004-6361/202346710

    work page doi:10.1051/0004-6361/202346710 2023

  24. [24]

    2021, ApJL, 911, L18, doi: 10.3847/2041-8213/abf1ec

    Kun, E., Bartos, I., Becker Tjus, J., et al. 2021, ApJL, 911, L18, doi: 10.3847/2041-8213/abf1ec

    work page doi:10.3847/2041-8213/abf1ec 2021

  25. [25]

    B., et al

    Kun, E., Bartos, I., Tjus, J. B., et al. 2024, PhRvD, 110, 123014, doi: 10.1103/PhysRevD.110.123014

    work page doi:10.1103/physrevd.110.123014 2024

  26. [26]

    On the Apparent Correlation between X-ray and Neutrino Luminosities of Active Galactic Nuclei

    Luo, J.-J., Lu, M.-X., & Liang, Y.-F. 2026, On the Apparent Correlation between X-ray and Neutrino Luminosities of Active Galactic Nuclei, https://arxiv.org/abs/2605.13588

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  27. [27]

    S., French, H

    Miller, J. S., French, H. B., & Hawley, S. A. 1978, in BL Lac Objects, ed. A. M. Wolfe, 176–187

    work page 1978

  28. [28]

    S., & M´ esz´ aros, P

    Murase, K., Kimura, S. S., & M´ esz´ aros, P. 2020, PhRvL, 125, 011101, doi: 10.1103/PhysRevLett.125.011101

    work page doi:10.1103/physrevlett.125.011101 2020

  29. [29]

    Neronov, A., Savchenko, D., & Semikoz, D. V. 2024, PhRvL, 132, 101002, doi: 10.1103/PhysRevLett.132.101002

    work page doi:10.1103/physrevlett.132.101002 2024

  30. [30]

    B., et al

    Oh, K., Koss, M., Markwardt, C. B., et al. 2018, ApJS, 235, 4, doi: 10.3847/1538-4365/aaa7fd

    work page doi:10.3847/1538-4365/aaa7fd 2018

  31. [31]

    M., Assef, R

    Padovani, P., Alexander, D. M., Assef, R. J., et al. 2017, A&A Rv, 25, 2, doi: 10.1007/s00159-017-0102-9

    work page doi:10.1007/s00159-017-0102-9 2017

  32. [32]

    2016, MNRAS, 458, 2836, doi: 10.1093/mnras/stw472

    Paiano, S., Landoni, M., Falomo, R., Scarpa, R., & Treves, A. 2016, MNRAS, 458, 2836, doi: 10.1093/mnras/stw472

    work page doi:10.1093/mnras/stw472 2016

  33. [33]

    V., Burenin, R

    Plavin, A. V., Burenin, R. A., Kovalev, Y. Y., et al. 2024, JCAP, 2024, 133, doi: 10.1088/1475-7516/2024/05/133

    work page doi:10.1088/1475-7516/2024/05/133 2024

  34. [34]

    2023, ApJL, 952, L28, doi: 10.3847/2041-8213/acda27

    Ricci, C., Chang, C.-S., Kawamuro, T., et al. 2023, ApJL, 952, L28, doi: 10.3847/2041-8213/acda27

    work page doi:10.3847/2041-8213/acda27 2023

  35. [35]

    2024, MNRAS, 528, 5990, doi: 10.1093/mnras/stae273

    Israyelyan, D. 2024, MNRAS, 528, 5990, doi: 10.1093/mnras/stae273

    work page doi:10.1093/mnras/stae273 2024

  36. [36]

    2000, ApJ, 542, 914, doi: 10.1086/317016

    Wilms, J., Allen, A., & McCray, R. 2000, ApJ, 542, 914, doi: 10.1086/317016

    work page internal anchor Pith review doi:10.1086/317016 2000