{"id":"bf76e52c-7ab8-4d8b-9379-1571d789609e","arxiv_id":"2505.24559","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"About 31-33% of red supergiants in M31 and M33 show ultraviolet excess from hot companions, giving new HST-based binary fractions.","lead":"Using Hubble Space Telescope images, the authors counted red supergiant stars in the Andromeda (M31) and Triangulum (M33) galaxies and identified likely binary systems by looking for extra ultraviolet light. They find that about a third of these giant stars have a hot companion star, matching predictions from binary evolution models.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Binary fraction rests on an unvalidated UV-excess criterion; Eq. (10) covers only log g, omitting [M/H], Teff systematics and chromospheric UV, so the false-positive rate of Eq. (9) is unmeasured and the headline may be biased upward.","rationale":"The paper's strongest contributions are the HST-resolved RSG samples and the M33 measurement. The extinction-sensitivity test in Section 4.4 shows only a few-percent swing in the binary fraction, and the M31 fraction agrees closely with Neugent (2021), so the measurement has real support at its current level. However, the identification step is not independently validated: no control sample, no chromospheric correction, and no [M/H] error term enter the binary criterion. That is the same weak point the reader identified, and it is load-bearing because every headline number passes through Eq. (9). The concern is real but not demonstrated to be fatal; a relabeling plus a control-sample false-positive estimate would address it. I therefore keep the reader's CONDITIONAL verdict and report UNCHANGED.","tokens_in":16501,"tokens_out":7884,"duration_ms":105485,"concrete_test":"Use the RSGs in the PHAT/PHATTER footprints that Neugent (2021) classified as single from spectroscopy, pass them through the exact pipeline of Sections 3.1 and 3.2 (SED fit on F814W/F110W/F160W, then Eq. 9 on F275W/F336W), and compute the false-positive rate. The paper already cross-matched 33 objects and found 11 disagreements; the decisive untested direction is whether spectroscopically single RSGs are flagged as binaries. If even roughly 10 percent of confirmed single RSGs show the UV excess, the headline fraction cannot be attributed to companions without an empirical correction.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The entire binary measurement is produced by Eq. (9): a 3-sigma UV excess over the L97 single-RSG model is read as a companion. The model error in Eq. (10) only spans log g = 0 to 0.5 (0.6 for Teff < 3500 K). It does not include (i) the fixed [M/H] = +0.3 (M31) / +0.1 (M33), although F275W lies on the steep Wien tail of a ~3700 K star and plausible +/-0.2 dex metallicity variations can change the blanketed UV flux by more than the 3-sigma threshold; (ii) Teff systematics from the adopted extinction, which Section 4.4 shows affects the fitted Teff distribution; or (iii) possible chromospheric or wind UV emission from the RSG itself, dismissed only by the statement that RSGs emit very weakly in the UV band. No control sample of spectroscopically confirmed single RSGs is run through Eq. (9), so the false-positive rate is not measured. Paper I labeled the same observable a lower limit; this paper labels it the binary fraction. The reported 33.4% and 30.9% are therefore best read as UV-excess fractions with an unknown bias.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents new RSG samples in M31 and M33 constructed from HST PHAT/PHATTER photometry, derives Teff, radius, and luminosity for each star by fitting L97 single-star SEDs to optical and NIR bands, and identifies binary candidates by requiring a 3-sigma UV excess in F275W/F336W relative to the single-star model prediction. The reported binary fractions are 33.4% +/- 0.9% (828/2481) for M31 and 30.9% +/- 0.8% (966/3129) for M33, with 31.6% +/- 1.9% and 34.7% +/- 1.8% for the log L/L_sun > 4.0 subsamples. The paper also estimates total RSG populations of about 6,563 (M31) and 7,572 (M33), roughly 19% and 148% larger than ground-based counts by Ren et al. (2021), and claims good agreement with BPASS binary evolution predictions.","tokens_in":16781,"tokens_out":7662,"duration_ms":86345,"significance":"If the binary fractions are correct, this is a valuable empirical result: it is one of the largest homogeneous samples of RSG binaries in external galaxies, and the HST-based photometry provides a substantial improvement in resolution and completeness over ground-based surveys. The paper has clear strengths: strict photometric quality cuts, a multi-step foreground removal that combines color-color diagrams with Gaia astrometry, cross-checks against the Ren et al. (2021) catalog and Neugent (2021) spectroscopic classifications, machine-readable output tables, and a systematic extinction test in Section 4.4. The central measurement, however, currently rests on a UV-excess criterion whose false-positive rate is unquantified and whose relation to the total binary fraction is not established; the BPASS comparison is also partly constructed by the adopted selection filters. These issues are load-bearing for the headline numbers, so the paper requires revision before the claims are fully supported.","major_comments":[{"comment":"The model error term F_err_mod only includes the variation of log g from 0 to 0.5 (or 0.6 for Teff < 3500 K), while [M/H] is fixed a priori (+0.3 for M31, +0.1 for M33) and the Teff error is said to be smaller than the 50 K grid spacing without propagating systematic uncertainties from extinction and reddening. Because F275W and F336W lie on the steep Wien tail of the RSG spectrum, a plausible +/-0.2 dex change in [M/H] can alter the predicted UV flux by an amount comparable to or larger than the 3-sigma threshold in Eq. (9). The paper should either extend Eq. (10) to cover these systematics or validate Eq. (9) on a control sample of spectroscopically confirmed single RSGs, for instance the Neugent (2021) catalog; without one of these, the false-positive rate of the binary identification is unmeasured and the headline fractions may be biased.","section":"Section 3.2, Eq. (10)"},{"comment":"Paper I (Dai et al. 2025) explicitly reported lower limits on the RSG binary fraction in the LMC and SMC using the same UV-excess method, because the method detects only binaries with hot, UV-luminous companions. The present paper drops the 'lower limit' qualifier and presents the values as 'the binary fraction.' Unless the authors demonstrate that the UV-excess selection is complete for all binary configurations that contribute to the RSG population, the numbers in the abstract should be restated as lower limits or as 'binary fractions of RSGs with detectable hot companions,' and the comparison with BPASS should be framed accordingly.","section":"Sections 3.2, 5, and abstract"},{"comment":"The BPASS 'prediction' is filtered by the observed Teff range (Eqs. 12-13) and then restricted to companions with L >= 60.7 L_sun (a B9V star), which is a hand-chosen threshold. This selection makes the claimed 'good agreement' with the observed 31.6%/34.7% fractions partly a consequence of the filter rather than an independent test. Please justify the luminosity threshold, report the fraction of BPASS binaries that pass it, and show how the predicted fraction changes when the threshold is varied by, say, +/-0.5 dex in luminosity.","section":"Section 4.5"},{"comment":"The M31 binary fraction changes from 32.7% to 36.9% across the three extinction scenarios, a spread of about 4 percentage points, which is larger than the quoted statistical uncertainty of +/-0.9%. This systematic should be propagated into the final uncertainty or at least discussed in the abstract; as written, the +/-0.9% implies a precision that the method does not achieve.","section":"Section 4.4"},{"comment":"The total RSG population estimates (6,563 in M31, 7,572 in M33) are derived by scaling the HST sample using the Ren et al. (2021) coverage fractions of 39.8% and 43.5%. Those fractions were computed from the ground-based RSG sample; because the HST sample is substantially deeper (148% more RSGs in M33), the spatial distribution of the newly resolved faint RSGs may not follow the same pattern, which would bias the total estimates. Please recompute the coverage fractions using the HST sample or test the sensitivity of the totals to the adopted fraction.","section":"Section 4.2"}],"minor_comments":[{"comment":"There is a missing conjunction or period between 'M31' and '3,294 RSGs' in the sentence describing the sample sizes.","section":"Abstract"},{"comment":"The sample sizes are inconsistent: the text states 584 RSGs in M31 and 735 in M33, but the fractions are computed with denominators 585 and 733, respectively. Please correct the numbers.","section":"Section 4.3"},{"comment":"The caption says the comparison is 'in M31,' but the text and the surrounding analysis describe both M31 and M33; the caption should refer to both panels.","section":"Figure 9"},{"comment":"The reduced chi-square definition uses weights w(lambda_j) that depend on both the model and the observed flux; this is not the standard definition of chi-square and its statistical interpretation should be clarified.","section":"Eq. (5)"},{"comment":"The RSG branch boundaries are described as visually determined; a sensitivity test (for example, shifting the boundaries by 0.05 mag) would help establish how robust the RSG sample is to the chosen borders.","section":"Eqs. (1)-(3)"}],"recommendation":"major_revision","confidential_remarks":"The paper contains a strong dataset and a sensible analysis chain, but the authors have not resolved the tension with their own Paper I, which called the same UV-excess observable a lower limit. The current manuscript's abstract and summary present the values as the binary fraction without qualification, and the BPASS agreement is partly manufactured by the companion luminosity and Teff selection windows. I recommend major revision: the authors should reframe the claims as lower limits or restricted binary fractions, add a control sample or a fuller model-error budget, and treat the extinction systematics as part of the uncertainty. Once these points are addressed, the paper would be a solid contribution to the field."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this paper delivers a real, reusable product: HST-resolution RSG catalogs for M31 and M33 in machine-readable form, with roughly double the M33 RSG count of ground-based work and 19% more in M31. Second, the M33 binary measurement is genuinely new, but the 'binary fraction' label outruns the method. The paper measures the fraction of RSGs with a 3-sigma UV excess over a single-star model, and Paper I called that same quantity a lower limit.\n\nWhat it does well. The sample construction is careful: three color-color diagrams plus Gaia for foreground removal, a documented TRGB boundary, and an explicit AGB cut. Section 4.4 is a real check—three internal-extinction scenarios shift the M31 fraction by only about 4 percentage points. In M31 the paper lands at 33.4% against Neugent's 33.5%, a completely different method (spectra plus machine learning), and that convergence carries weight. The 11 objects where the paper disagrees with Neugent are examined with HST images, and the aperture-contamination explanation is plausible. The BPASS comparison is transparent about its selection cuts.\n\nSoft spots, in proportion. The main one is the label, and the stress-test note hits it: Eq. (10) only covers log g uncertainty, while [M/H] is fixed at +0.3/+0.1 and F275W sits on the steep Wien tail for a ~3700 K star. Plausible metallicity scatter can move the predicted UV flux by more than the 3-sigma threshold. Chromospheric UV from the RSG itself is dismissed in a sentence, and no control sample of confirmed single RSGs is run through Eq. (9). The false-positive rate is unmeasured, and the quoted uncertainties do not include this systematic. What partly rescues the claim is the independent agreement with Neugent in M31 and with the LMC/SMC values from Paper I; a large bias would make those coincidences unlikely. But the honest reading of 33.4% and 30.9% is 'fraction with detectable UV-bright companions,' and a referee should push for relabeling or calibration. Secondary: the BPASS agreement is a filtered consistency check (observed Teff window plus a hand-chosen 60.7 Lsun companion threshold), and the total population numbers are simple scalings by survey coverage—fine as ballpark, not more.\n\nWho this is for: RSG evolution, binary population synthesis, resolved stellar populations in M31/M33. It deserves a serious referee, with the expectation of a conditional verdict. I would send it out, and I would cite the catalogs and the M33 measurement either way.","headline":"Worth sending to review: solid HST-resolution RSG catalogs and a genuinely new M33 measurement, but 'binary fraction' overstates what the UV-excess method demonstrates.","tokens_in":17368,"tokens_out":6828,"would_cite":true,"duration_ms":76045,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper measures the binary fraction of red supergiants in M31 and M33 using HST photometry, finding about one third have hot companions, and shows HST resolves far more RSGs than ground-based surveys.","keywords":["red supergiants","binary fraction","ultraviolet excess","M31","M33","Hubble Space Telescope","spectral energy distribution","binary population synthesis"],"falsifier":"Measure radial velocities for a sample of RSGs classified as single in this work; if a large fraction show orbital motion with hot companions, the UV-excess method misses binaries, whereas if single-star RSGs show intrinsic UV variability or UV emission, the method overcounts. Alternatively, spatially resolve the UV emission of binary candidates with integral-field spectroscopy to confirm a genuine hot companion rather than scattered light.","tokens_in":16184,"feed_emoji":"🌟","tokens_out":6305,"duration_ms":64529,"temperature":0.7,"pith_summary":"The paper uses Hubble Space Telescope photometry from the PHAT and PHATTER surveys to build cleaner, more complete samples of red supergiants in M31 and M33, then measures their binary fraction by looking for ultraviolet excess over a single-star model. It finds binary fractions of 33.4% ± 0.9% in M31 and 30.9% ± 0.8% in M33, with high-luminosity subsets at 31.6% ± 1.9% and 34.7% ± 1.8%. It also finds that HST resolves roughly 19% (M31) and 148% (M33) more red supergiants than previous ground-based catalogs, and that the measured binary fractions agree with the BPASS binary population synthesis model. If correct, about one in three red supergiants in these galaxies has a detectable hot companion, which matters for predicting supernova progenitors and the fate of massive stars.","feed_headline":"One in three red supergiants in M31 and M33 has a hot companion","feed_subtitle":"HST-resolved samples show roughly one in three red supergiants have UV excess, matching binary evolution models.","key_machinery":"The identifying mechanism is the ultraviolet excess test. A star is declared a binary when its observed F275W or F336W flux exceeds the flux predicted by a single-red-supergiant model spectrum by more than 3$\\sigma$ of the observational error plus 3$\\sigma$ of the model uncertainty, where the model uncertainty accounts for surface gravity up to $\\log g = 0.5$. Because red supergiants emit very weakly in the UV, any such excess is attributed to a hot main-sequence companion. Stellar parameters are obtained from SED fitting to the redder bands using the adopted model grid.","core_discovery":"The central claim is that the binary fractions of red supergiants in M31 and M33 are about one third, obtained by applying a UV-excess SED-fitting method to HST-resolved samples. The paper reports 828 binary candidates among 2,481 usable RSGs in M31 and 966 among 3,129 in M33, yielding 33.4% ± 0.9% and 30.9% ± 0.8%. For RSGs with $\\log L/L_{\\odot} > 4.0$, the fractions are 31.6% ± 1.9% and 34.7% ± 1.8%. These observed fractions agree with BPASS predictions once binaries with low-luminosity main-sequence companions that would not yet be observable are removed from the model outputs.","pith_inferences":["The UV-excess method only detects companions hot enough to produce F275W or F336W flux; binaries with cool, low-mass companions are likely missed, so the true RSG binary fraction may be higher than the measured 31-33%.","If the agreement with BPASS holds, the same method could be applied to other nearby galaxies with HST UV coverage to map the binary fraction as a function of environment without requiring spectroscopy.","The M33 sample covers only the inner region, so the reported fraction may not represent the whole galaxy; combining with outer-region data could test whether the inner/outer gradient suggested by spectroscopic work persists."],"forward_implications":["About one in three red supergiants in M31 and M33 has a detectable hot companion, reshaping expectations for supernova progenitors that come from binary channels.","HST-resolved samples contain substantially more RSGs than ground-based surveys (19% more in M31, 148% more in M33), implying previous RSG population counts in nearby spirals are incomplete.","The observed binary fractions match BPASS predictions when low-luminosity companions are excluded, supporting current binary evolution models for post-main-sequence massive stars.","The derived RSG physical parameters ($T_{\\rm eff}$, $R$, and $L$) provide a sizable catalog for future studies of stellar evolution in M31 and M33.","The lack of a metallicity trend in binary fractions across M31, M33, the LMC, and the SMC suggests the binary fraction is roughly independent of galaxy metallicity."],"supporting_citations":[{"why":"Supplies the synthetic stellar spectra used for SED fitting and for predicting the single-star UV flux.","marker":"Lejeune et al. 1997"},{"why":"Provides the ground-based RSG catalog used for comparison, for scaling to total galaxy populations, and for the foreground-star removal method.","marker":"Ren et al. 2021"},{"why":"Gives previous RSG binary fractions in M31 and M33 that this work compares against and partly reconciles via HST resolution.","marker":"Neugent 2021"},{"why":"Describes the BPASS binary population synthesis model from which the predicted binary fractions are extracted.","marker":"Stanway & Eldridge 2018"},{"why":"Provides the PHATTER survey catalog of M33 used as the HST photometric dataset.","marker":"Williams et al. 2021"},{"why":"Provides the PHAT V3 catalog of M31 used as the HST photometric dataset.","marker":"Williams et al. 2023"},{"why":"Paper I, establishing the SED-fitting UV-excess method that this work applies to M31 and M33.","marker":"Dai et al. 2025"}],"fun_headline_variants":["1/3 of red supergiants in M31 and M33 are binaries","Hubble finds one in three red supergiants in M31 and M33 has a companion","HST: Third of red supergiants in M31 and M33 have binary companions","HST data: 33% of RSGs in M31 and 31% in M33 have UV excess"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method assumes red supergiants have negligible intrinsic ultraviolet emission, so that any UV excess must come from a hot companion; if RSG chromospheres or an underestimated model UV flux produce the excess, the binary fractions are overestimated.","fun_headline_variants_meta":{"raw":{"variants":["1/3 of red supergiants in M31 and M33 are binaries","Hubble finds one in three red supergiants in M31 and M33 has a companion","HST: Third of red supergiants in M31 and M33 have binary companions","HST data: 33% of RSGs in M31 and 31% in M33 have UV excess"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001147,"raw_usage":{"total_tokens":4834,"prompt_tokens":1100,"completion_tokens":3734,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":716,"completion_tokens_details":{"reasoning_tokens":3636}},"tokens_in":716,"tokens_out":3734,"duration_ms":29542,"temperature":1.0,"reasoning_tokens":3636,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:19:47.670231+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure radial velocities for a sample of RSGs classified as single in this work; if a large fraction show orbital motion with hot companions, the UV-excess method misses binaries, whereas if single-star RSGs show intrinsic UV variability or UV emission, the method overcounts. Alternatively, spatially resolve the UV emission of binary candidates with integral-field spectroscopy to confirm a genuine hot companion rather than scattered light.","supporting_citations":[{"cited_title":"2025, , 539, 1220, 10.1093/mnras/staf560","cited_arxiv_id":null,"evidence_quote":"Paper I, establishing the SED-fitting UV-excess method that this work applies to M31 and M33."}],"review_version":1}