Joint JWST-DECam Lensing Reveals That the Bullet Cluster Is a Minor Merger

Boseong Young Cho; Hyungjin Joo; Kim HyeongHan; M. James Jee; Sangjun Cha

REVIEW 3 major objections 5 minor 3 references

Joint JWST-DECam Lensing Reveals That the Bullet Cluster Is a Minor Merger

T0 review · 3 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read JWST-DECam lensing shows the Bullet Cluster is a 10:1 minor merger, not a major one.

desk verdict First robust three-halo mass decomposition of the Bullet Cluster, giving a ~10:1 ratio, but the result flips to ~2:1 if you impose a mass-concentration relation, and the profile tests don't quite close that gap. read the letter →

arxiv 2512.03150 v2 pith:ESIWBJFF submitted 2025-12-02 astro-ph.GA

Boseong Young Cho , M. James Jee , Hyungjin Joo , Sangjun Cha , Kim HyeongHan This is my paper · ORCID

classification astro-ph.GA

keywords galaxyclustersweaklensingstrongBulletClustermergersdarkmattervirialmassNFWprofile

topics Dark Matter

open problems Dark Matter

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 claims the Bullet Cluster's collision was a minor merger: the main cluster is roughly ten times more massive than the infalling subcluster. Combining JWST's dense background-galaxy sample with DECam's wide-field shear measurements, the authors resolve three separate halos and derive virial masses of 15.11 and 1.49 times 10^14 solar masses for main and subcluster, a mass ratio of 10.14. Previous estimates ranged from about 2:1 to 100:1, while the hydrodynamical simulations that reproduce the bullet's shock need mass ratios near 5:1 to 10:1. If correct, this resolves a two-decade tension and supplies revised initial conditions for dark-matter and merger studies of this benchmark system.

What carries the argument

Three-halo Navarro-Frenk-White (NFW) profile fitting with freely varying concentrations, anchored by projected masses from strong lensing within 150 kpc of each brightest cluster galaxy and constrained by wide-field DECam weak-lensing shear out to roughly 7 Mpc. The strong-lensing masses act as boundary conditions that break the mass-concentration degeneracy and mitigate merger-induced model bias, while the wide-field coverage removes the need for uncertain extrapolation to the virial radius.

What would settle it

Measure the subcluster's virial mass independently of lensing, for example through deep spectroscopy of its member galaxies (expected velocity dispersion of roughly 550-650 km/s for a 1.5e14 solar-mass halo, versus 800-950 km/s for a 3.8e14 halo) or through spatially resolved X-ray temperature and density mapping; a true subcluster mass near 4e14 solar masses would falsify the minor-merger claim.

Watch

Extended reading notes

Core claim

The paper reports the first virial masses for all three components of the Bullet Cluster, derived from a joint weak-plus-strong lensing analysis. The main cluster splits into two comparable halos, Main-SE and Main-NW, with M200c = 6.09 and 4.74 times 10^14 solar masses, and the subcluster has M200c = 1.49 times 10^14 solar masses. Summing the two main halos gives M200c = 15.11 times 10^14 solar masses, so the main-to-subcluster mass ratio is 10.14, classifying the Bullet Cluster as a minor merger. This reconciles the previously conflicting observational ratios with the initial conditions required by simulations that reproduce the bow shock and dark-matter/gas offsets.

Load-bearing premise

The minor-merger conclusion rests on letting each halo's concentration float freely rather than imposing the standard mass-concentration relation; the paper's own robustness test shows that imposing that relation flips the inferred mass ratio from ~10:1 to ~2:1.

Editorial extensions

If this is right

Resolves the long-standing tension between observational mass ratios (~2:1 to ~100:1) and the ~5:1 to 10:1 initial conditions that simulations need to reproduce the bullet's shock and dark-matter/gas offset.
Shows the main cluster is itself bimodal, with two comparable-mass halos separated by 170 kpc, implying the main cluster experienced a prior merger before the bullet collision.
Demonstrates that earlier strong-lensing-only mass ratios were driven largely by the arbitrary choice of truncation radius, not by the data.
Provides updated halo masses and concentrations for future hydrodynamical modeling of the Bullet Cluster, replacing the simplified two-halo initial conditions.
Establishes that fitting merging clusters with free halo concentrations is decisively preferred over imposing a standard mass-concentration relation (log Bayes factor 15.86 in favor of free concentrations).

Reading between the lines

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

If the 10:1 classification holds, dark-matter self-interaction cross-section constraints calibrated with ~2:1 initial conditions will need to be re-evaluated, since those simulations assumed a much more equal-mass encounter.
The ~50-degree tilt between the main cluster's internal merger axis and the primary bullet axis is a testable prediction: deeper radio observations should reveal a second relic oriented perpendicular to that NW-SE axis.
The same JWST-plus-wide-field-plus-strong-lensing anchoring approach could be applied to other merging clusters; if similar 10:1 systems turn out to be common, mass-function and scaling-relation studies of disturbed clusters may require revision.
The subcluster's unusually high inferred concentration (c ~ 9.7) hints at strong tidal stripping, which would make its NFW-based virial mass more sensitive to modeling assumptions than the quoted uncertainties suggest.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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

This is a serious, data-rich paper. The JWST source density and the DECam wide-field coverage let the authors do something nobody has done before: separate the two main-cluster halos and the subcluster and assign each a virial mass. The fiducial result is M200c = 15.1e14 and 1.49e14 Msun for main and sub, a 10.1:1 ratio, consistent with the minor-merger initial conditions that hydro simulations of this system need. The WL-only configuration gives the same ratio, which is reassuring, and the paper's systematics tests are unusually thorough: PSF diagnostics, LSS error budgets, two- vs three-halo Bayes factors, Einasto and truncated profiles, HST comparison, and a demonstration that only the JWST data can separate the halos. That is real work and real progress.

The soft spot is the concentration assumption. The authors leave three NFW concentrations free, and the main halos come out very low (c ~ 1.7 and 2.5). If instead you impose the Diemer & Joyce M-c relation, the main-cluster mass drops to 7.4e14, the subcluster rises to 3.8e14, and the ratio becomes ~1.95:1. The data strongly prefer the free-c model by a Bayes factor of e^15.9, and the authors correctly note that merging clusters deviate from average relations. But the Bayes factor compares two NFW-based models; it does not address the possibility that the outer density profiles of merging halos are not NFW at all. The paper's TNFW test uses the Oguri & Hamana truncation prescription, which they themselves argue is unsuitable for rapid mergers, citing Walker et al. 2025. A smaller truncation radius, as expected from high accretion rates, would steepen the outer profiles, lower the main-cluster mass, raise the subcluster mass, and shrink the ratio. That scenario is left untested. The low concentrations and the Main-SE concentration hitting the prior boundary in the Einasto test suggest the inner data are not pinning down the profile shape.

The SL anchoring comes from the authors' own Cha et al. 2025 reconstruction, which is a self-citation. I don't think that's disqualifying here because the WL-only three-halo result is statistically identical and the anchoring regions are reasonably chosen, but independent SL constraints would strengthen the case.

This deserves a serious referee. The central claim is plausible and the analysis is far ahead of prior attempts, but the decisive 10:1 classification rests on a modeling choice that the paper's own alternative assumptions flip. A good referee should push for a test with an explicitly smaller truncation radius or some other physically motivated outer-profile model, and ideally for the release of the catalogs and code. I'd bring it to a reading group and would consider citing it after the profile question is addressed.

Referee Report

3 major / 5 minor

Summary. This paper presents a joint weak+strong lensing analysis of the Bullet Cluster using JWST/NIRCam and DECam data. The authors model the system as three NFW halos (Main-SE, Main-NW, Subcluster) with free concentrations and no imposed mass-concentration relation, anchored by strong-lensing projected masses from their earlier Cha et al. (2025) reconstruction. They report M200c = 15.11^{+2.48}_{-2.10} × 10^14 Msun for the main cluster and 1.49^{+0.32}_{-0.25} × 10^14 Msun for the subcluster, yielding a mass ratio 10.14^{+3.22}_{-2.47}, which they claim definitively classifies the Bullet Cluster as a minor merger. The paper includes extensive robustness tests: WL-only vs WL+SL, two-halo vs three-halo Bayes factors, alternative density profiles, LSS error estimation, HST comparison, and tests of imposing the Diemer & Joyce M-c relation. The central claim is that this resolves the long-standing discrepancy between observed mass ratios (~2:1 to ~100:1) and simulation-based requirements (~5:1 to 10:1).

Significance. If the result holds, it would be a significant step: using JWST's high source density to resolve three halos and wide-field DECam to constrain the virial scale would provide the first observational determination of the Bullet Cluster's mass ratio from lensing alone, with implications for dark-matter self-interaction constraints and merger modeling. The paper's strengths are the unusually thorough systematic checks: PSF diagnostics, LSS error accounting, model-comparison Bayes factors, and cross-checks between JWST and HST. The WL-only and WL+SL results are statistically consistent (Section 5.2.6), mitigating concerns about circularity in the SL anchoring. However, the minor-merger classification is not robust to the choice of concentration prior: imposing the standard M-c relation flips the ratio to ~1.95 (Section 5.2.1, Table 3). The paper argues the data decisively favor free concentrations, but this comparison does not resolve the underlying profile-extrapolation uncertainty. The claim of a 'definitive' classification is therefore overstated in its current form.

major comments (3)

[§5.2.1, Table 3] The mass ratio is 10.14 when concentrations are free (3cS) but 1.95 when the Diemer & Joyce (2019) M-c relation is imposed (3MS), with the main-cluster mass dropping from 15.11 to 7.36 ×10^14 Msun and the subcluster rising from 1.49 to 3.79 ×10^14 Msun. The paper's 'decisive' Bayes factor (ln BF = 15.86) compares two NFW models differing only in the concentration prior; it does not quantify the physical prior uncertainty. Because merging clusters are known to deviate from average relations, the uniform prior is motivated, but the conclusion that the Bullet Cluster is a minor merger rests entirely on this choice. The authors should present the ratio as conditional on the adopted prior and provide a quantitative assessment (e.g., a prior-weighted mixture or a simulation-based prior for merging systems) rather than asserting a definitive classification.
[§5.2.3] The TNFW test uses the Oguri & Hamana (2011) truncation radius, which the authors themselves state is unsuitable for rapidly merging halos, citing Walker et al. (2025) that such systems have much smaller truncation radii. A smaller truncation radius would steepen the outer density profile and reduce the extrapolated M200c of the main halos, potentially lowering the mass ratio. The paper does not test this physically motivated alternative. The fact that Main-SE's concentration hits the lower prior boundary (c=1) in the Einasto and TNFW tests indicates the inner profile shape is weakly constrained, so the fiducial masses—dominated by outer NFW extrapolation—are not robust to this omission.
[§4.2.4, Table 2] The main-cluster halos are inferred with very low concentrations (c_SE ≈ 1.7, c_NW ≈ 2.5), which drive the large M200c values. The paper attributes this to merging, but no independent validation is provided, and the concentration priors (uniform 1–20) are wide. The mass-ratio conclusion is therefore sensitive to the prior range on c; the authors should demonstrate stability under narrower or physically motivated priors.

minor comments (5)

[§5.2.1] The M-c relation is imposed as a delta function without scatter. The authors should note that including intrinsic scatter would weaken the Bayes-factor comparison and partially bridge the two ratios.
[§5.2.2] The statement that the two-halo model 2cS 'failed to converge' is abrupt; clarify whether this is a posterior-sampling failure or a physical impossibility (e.g., the SL-anchored masses cannot be jointly reproduced).
[§3.4.2] The description of the redshift assignment for DECam sources is clear, but the effective source redshift z_eff = 0.560 and the β correction would benefit from a brief justification of the DESY3GOLD GOODS-S control field choice versus other control fields.
[Figure 9] The caption states 'definitively establishes' the minor-merger scenario; given the major-comment concerns, this wording should be softened to reflect the model dependence.
[General] There are a few typographical issues (e.g., 'corresponding author' line formatting, a duplicated reference entry for Finner et al. 2023a/2023b). These do not affect the science.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular-by-construction derivation; central ratio is independently reproduced by the WL-only configuration, and the self-citations are not load-bearing.

full rationale

The central mass ratio M_Main/M_Sub = 10.14 is not equal by construction to any input. It is obtained from MCMC fits of three NFW halos to unbinned galaxy ellipticities (Eq. 8) plus projected-mass anchoring. The anchoring masses come from the authors' own Cha et al. (2025) reconstruction and are imposed as hard constraints ('We discard any proposed MCMC sample whose projected masses are inconsistent with the anchoring values'), which is a self-citation. However, this input is not load-bearing for the minor-merger claim: the WL-only 3cL configuration gives M_Main = 15.68e14 Msun, M_Sub = 1.54e14 Msun, ratio = 10.21 (Table 2), statistically identical to the fiducial 3cS result. The fragile modeling choice is the free NFW concentration with a wide uniform prior and no M-c relation; this is supported by a data-driven Bayes factor, ln(BF(3cS)(3MS)) = 15.86 (Section 5.2.1), rather than by the self-cited Lee et al. (2023)/Finner et al. (2025) simulation claims alone. The reported alternative with an imposed Diemer & Joyce (2019) relation (ratio ~1.95) is a model-dependence/fragility issue, not a circular step. Similarly, the TNFW test explicitly notes Walker et al. (2025) makes the standard truncation formula 'unsuitable' for rapid mergers, leaving a smaller-truncation model untested; this is an acknowledged robustness limitation, not a reduction of the prediction to the input. No fitted parameter is renamed as a prediction, and no uniqueness theorem is imported from the authors' prior work. Score 2 reflects only the minor non-load-bearing self-citations.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The paper adds no new physical entities. The central inference rests on standard lensing assumptions plus two modeling choices: the NFW profile and the decision to leave concentration free. The free-concentration choice is the most consequential, because an alternative M-c relation produces a qualitatively different mass ratio. All support for the three-halo decomposition comes from the JWST mass map and the authors' companion paper.

free parameters (6)

M200c, Main-SE = 6.09e14 Msun
NFW mass of the Main-SE halo, fitted to WL+SL data (Table 2).
c200c, Main-SE = 1.68
NFW concentration of the Main-SE halo, fitted with uniform prior (1-20).
M200c, Main-NW = 4.74e14 Msun
NFW mass of the Main-NW halo, fitted to WL+SL data (Table 2).
c200c, Main-NW = 2.50
NFW concentration of the Main-NW halo, fitted with uniform prior.
M200c, Subcluster = 1.49e14 Msun
NFW mass of the subcluster halo, fitted to WL+SL data (Table 2).
c200c, Subcluster = 9.72
NFW concentration of the subcluster halo, fitted with uniform prior.

assumptions (7)

domain assumption Each component is modeled with an NFW density profile.
Section 4.2.1: 'Each halo is modeled with an NFW profile without imposing an M-c relation.' This is a standard but not physically necessary choice for cluster halos.
domain assumption There are exactly three halos, centered at the three BCG positions.
Section 4.2.1: based on the JWST WL+SL mass map, which shows three mass peaks. The paper tests a two-halo model and finds it strongly disfavored, but the three-halo choice is a model assumption.
domain assumption The mass-concentration relation is not valid for these merging halos, so concentration is left free.
Section 5.2.1: when the Diemer & Joyce (2019) M-c relation is imposed, the mass ratio becomes ~1.95:1. The paper argues that merging clusters deviate from the relation, but this is a modeled assumption rather than an independent measurement.
domain assumption Flat Lambda-CDM cosmology with H0=70, Omega_M=0.3, Omega_Lambda=0.7.
Standard cosmological framework used for distance and density calculations throughout the paper.
domain assumption The DESY3GOLD control field redshift distribution is representative of the DECam background sources.
Section 3.4.2: statistical redshift assignment for DECam sources assumes the GOODS-S control field galaxy population is appropriate after magnitude-binning weights.
domain assumption The SL-projected masses from Cha et al. (2025) used as anchors are correct.
Section 4.2.1: MCMC samples are discarded unless the model reproduces the projected masses within the SL-anchored regions. The WL-only model gives consistent results, partially mitigating this dependency.
domain assumption The line-of-sight separations of the three halos are negligible when integrating the 3D density grid.
Section 4.2.4: 'assuming that differences in their line-of-sight positions are negligible' when computing the combined main-cluster mass.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Joint JWST-DECam Lensing Reveals That the Bullet Cluster Is a Minor Merger." pith.science (2026). https://pith.science/paper/ESIWBJFF

@misc{pith2026251203150,
  author       = {Pith},
  title        = {Pith review of: Joint JWST-DECam Lensing Reveals That the Bullet Cluster Is a Minor Merger},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ESIWBJFF}},
  note         = {Machine review of arXiv:2512.03150}
}

abstract

We present the first robust virial masses of the Bullet Cluster's three individual components from a joint weak+strong lensing analysis combining JWST/NIRCam and DECam observations. Despite its status as the benchmark system for dark matter and merger studies, inferred mass ratios for the Bullet Cluster have spanned a wide range from $\sim$2:1 to $\gtrsim$10:1 over more than two decades. We revisit this tension through three key advances: (1) JWST's exceptional data quality enables us to resolve three distinct halos, (2) DECam's wide-field coverage beyond its virial radius eliminates the need for extrapolation, and (3) high-fidelity strong-lensing priors mitigate weak-lensing model bias. We obtain $M_{200c} = 15.11^{+2.48}_{-2.10} \times 10^{14}M_{\odot}$ for the main cluster and $1.49^{+0.32}_{-0.25} \times 10^{14}M_{\odot}$ for the subcluster, yielding a mass ratio of $10.14^{+3.22}_{-2.47}$, definitively classifying the Bullet Cluster as a minor merger. This result reconciles the long-standing tension in the mass ratio and provides updated initial parameters for future modeling of this iconic system.

Figures

Figures reproduced from arXiv: 2512.03150 by the authors.

**Figure 1.** Wide-field view of the Bullet Cluster region. The background shows our DECam g+r+i color composite, covering the 1◦×1 ◦ (∼16 Mpc×16 Mpc at z = 0.296) region centered on the target. The red dashed rectangle marks the JWST/NIRCam F200W coverage (∼6 ′ × 2.5 ′ ). The green circle marks the total system’s virial radius (R Total 200c = 2.27 Mpc; §4.2.4), while the blue circle indicates the boundary of the DECam WL analysi… view at source ↗

**Figure 2.** PSF correction quality for JWST F200W (left) and DECam i-band (right). Blue points show observed stellar ellipticities while red points show the ellipticity residuals after PSF correction (i.e., observed minus modeled ellipticity components). Residuals centered at (0,0) with small scatter demonstrate accurate and precise PSF modeling. The lower-left inset in each panel displays PSF size residuals. 10 0 10 1 Separat… view at source ↗

**Figure 3.** PSF model diagnostics for DECam i-band. Left: D1 and D2 statistics (Rowe 2010) showing auto-correlation of PSF model residuals (D1; red circles) and cross-correlation between observed stellar ellipticities and residuals (D2; black crosses). The D1 (D2) statistic remains below 10−6 (10−5 ) across all angular scales, demonstrating exceptional PSF modeling accuracy. Right: Star-galaxy correlation functions following Je… view at source ↗

Figures from the paper (8 more)

**Figure 5.** Figure 5: Spectroscopic and photometric redshift comparison for JWST sources and galaxy number density validation for DECam. Top: Spectroscopic versus photometric redshifts for 103 JWST sources with spectroscopic data compiled from Fo¨ex et al. (2017), Puccetti et al. (2020),…

**Figure 6.** Figure 6: WL+SL mass reconstruction of the Bullet Cluster from JWST/NIRCam data using the MARS algorithm (Cha & Jee 2022). We reproduced the mass contours in Cha et al. (2025), ranging from κ = 0.15 to 1.35 in steps of 0.15. Green dashed circles indicate the SL-anchored regions …

**Figure 7.** Figure 7: Wide-field mass reconstruction of the Bullet Cluster from DECam data, covering 1◦ × 1 ◦ . Left: The background grayscale represents the galaxy number density map derived from our red-sequence catalog ( [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

**Figure 8.** Figure 8: MCMC posterior distributions for our fiducial case (3cS; this notation from [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]

**Figure 9.** Figure 9: Compilation of Bullet Cluster virial mass (M200c) measurements spanning more than two decades. Measurement methods are color-coded: equilibrium-based methods (purple: velocity dispersion, X-ray, SZ), WL (blue), SL (green), our joint WL+SL analysis (red), and hydrodynam…

**Figure 10.** Figure 10: Tangential shear profile comparison for three-halo and two-halo NFW models. The three panels show measurements centered on Main-SE, Main-NW, and Sub BCGs, respectively, with tangential shear ⟨g+⟩ (upper panels) and cross shear ⟨g×⟩ (lower panels). Black points and cro…

**Figure 11.** Figure 11: Limitations of HST WL. Left: HST-only mass reconstruction using FIATMAP with S/N contours (purple, levels from 3 to 7 in steps of 1), achieving peak S/N of 7.90 (main cluster) and 4.63 (subcluster). JWST WL S/N contours are overlaid in yellow (levels from 3 to 17 in s…

**Figure 12.** Figure 12: Same as [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

3 extracted references

[1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter doi edition editor eprint howpublished institution journal key month number organization pages publisher school series title misctitle type volume year version url label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts ...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION format.url url empty "" new.block "" url * "" * if FUNCTION format.eprint eprint empty "" archivePrefix empty "" archivePrefix "arXiv" = new.block " " eprint * " " * new.block " " eprint * " " * if if if FUNCTION format.doi doi empty "" " " doi * " " * if FUNCTION format.pid doi empty eprint empty ur...
[3]

b^y3 y+ `֬Vķ N o5 I` S A5F N W;!j֪f

thebibliography [1] 20pt to REFERENCES 6pt =0pt -12pt 10pt plus 3pt =0pt =0pt =1pt plus 1pt =0pt =0pt -12pt =13pt plus 1pt =20pt =13pt plus 1pt \@M =10000 =-1.0em =0pt =0pt 0pt =0pt =1.0em @enumiv\@empty 10000 10000 `\.\@m \@noitemerr \@latex@warning Empty `thebibliography' environment \@ifnextchar \@reference \@latexerr Missing key on reference command E...

arXiv 2021

Pith tools

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

[1] [1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter doi edition editor eprint howpublished institution journal key month number organization pages publisher school series title misctitle type volume year version url label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts ...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION format.url url empty "" new.block "" url * "" * if FUNCTION format.eprint eprint empty "" archivePrefix empty "" archivePrefix "arXiv" = new.block " " eprint * " " * new.block " " eprint * " " * if if if FUNCTION format.doi doi empty "" " " doi * " " * if FUNCTION format.pid doi empty eprint empty ur...

[3] [3]

b^y3 y+ `֬Vķ N o5 I` S A5F N W;!j֪f

thebibliography [1] 20pt to REFERENCES 6pt =0pt -12pt 10pt plus 3pt =0pt =0pt =1pt plus 1pt =0pt =0pt -12pt =13pt plus 1pt =20pt =13pt plus 1pt \@M =10000 =-1.0em =0pt =0pt 0pt =0pt =1.0em @enumiv\@empty 10000 10000 `\.\@m \@noitemerr \@latex@warning Empty `thebibliography' environment \@ifnextchar \@reference \@latexerr Missing key on reference command E...

arXiv 2021