{"id":"7aef5b86-937e-4d65-8f7e-bd367f181e0d","arxiv_id":"2411.09412","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"A modified 'red limit' in color-magnitude space selects some lensed type Ia supernovae, including on the falling edge of the light curve and at redshifts up to 3.","lead":"The authors propose a color-magnitude selection rule to quickly flag likely lensed type Ia supernovae in LSST survey data, and test it on simulated and observed supernovae. The rule works in principle on simulated data, but its real-world efficiency depends on survey conditions the paper does not yet model.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper's own validation data contradict the claim of negligible CC contamination: Sec. 4.4 reports <0.05% unlensed CC SNe passing the red limit, while Sec. 4.5 states the same cut selects a majority of observed SDSS CC SNe.","rationale":"I agree with the reader that the noiseless, in-sample simulations weaken the quantitative efficiency and rejection claims. However, I see a more specific, more load-bearing problem in the CC contamination claim. The paper's own validation data in Sec. 4.5 contradict the simulated negligible contamination reported in Sec. 4.4 and reiterated in the abstract. This is not only a realism or extrapolation concern; it is an internal discrepancy in the evidence presented. If real unlensed CC SNe pass the red limit at the majority rate the authors describe, then the central practical claim that the cut rejects most unlensed transients is unsupported, regardless of how clean the SNe Ia separation looks in simulations. I still do not think the paper should be rejected outright: the physical idea that lensed SNe Ia occupy a redder CM region is plausible and is partially supported by the three observed lensed systems on the falling edge, and the authors are transparent about limitations and about a companion pipeline study. The right status remains conditional: the method is promising, but the CC contamination rate and the LSST-transferable efficiency numbers need to be established on realistic, survey-matched data. Since the reader's verdict is already CONDITIONAL, my stress-test does not move the verdict; it sharpens the reason for it.","tokens_in":16468,"tokens_out":7411,"duration_ms":73931,"concrete_test":"K-correct the 88 observed SDSS-II CC SNe shown in Fig. 5 to the LSST r/i bands, add realistic photometric noise and a detection threshold, and apply Eq. 6 at the rising and falling epochs; report the exact fraction selected. If that fraction exceeds roughly 1% instead of the simulated <0.05%, the negligible-CC-contamination claim fails and the method cannot be used for LSST alert triage without a CC-specific rejection step.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the modified red limit (Eq. 6) rejects essentially all unlensed SNe while retaining roughly 44-67% of lensed SNe Ia. A direct contradiction appears in the CC-contamination estimate. Sec. 4.4 concludes from simulations that contamination by unlensed CC SNe is 'very low to negligible (<0.05%)', and the abstract repeats this claim. But Sec. 4.5, applying the same limit to the observed SDSS-II CC SNe used for validation, reports that 'the cut is, however, found to select a majority of the low redshift unlensed CC SNe sample.' Since CC SNe vastly outnumber lensed SNe Ia in an LSST alert stream, a majority selection of CC SNe would dominate the candidate list even if the SNe Ia rejection fractions are correct. The authors attribute the discrepancy to a small sample and defer a full study, but the tension is not resolved: either the simulated CC templates or luminosity functions are too faint or too blue, the red limit is not robust to real CC SEDs, or the bandpass mismatch makes the validation invalid for LSST. In any of these cases, the quantitative rejection claim in the abstract and Sec. 4.4 is not supported by the paper's own evidence. This is more load-bearing than the noiseless-simulation issue, because even with perfect photometry the proposed rule already fails to reject the observed CC sample.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a color-magnitude (CM) selection criterion for identifying strongly lensed type Ia supernovae (SNe Ia) in LSST-era surveys. The authors simulate LSST-like photometry of lensed and unlensed SNe Ia using SNCosmo and lensing observables from a realistic lens population, and show that a subset of lensed SNe Ia occupy a redder region of the CM plane than unlensed SNe Ia on both rising and falling phases of the light curve and out to z=3. They define a modified 'red limit' (Eqs. 6 and 7), report selection efficiencies of roughly 44-67% of lensed SNe Ia at 91-99.5% rejection of unlensed SNe Ia, find negligible simulated contamination from unlensed core-collapse SNe, and validate the criterion against archival samples of SNe Ia, CC SNe, SLSNe, and three known lensed SNe. The central claim is that the modified red limit is a rapid, effective pre-filter for lensed SNe Ia candidates in LSST alerts.","tokens_in":16683,"tokens_out":2777,"duration_ms":29199,"significance":"If the claimed separation and rejection rates hold under realistic survey conditions, this would be a practically useful, inexpensive filter for a regime (lensed SNe Ia) that is scientifically valuable but extremely rare. The paper's strengths include the use of a realistic lens population from the HSC catalog, the extension of CM selection to the falling edge and to z=3 via z-y colors, and the independent check against the known lensed systems PS1-10afx, iPTF16geu, and SN Zwicky, which do lie above the proposed limit. The pipeline is built from public tools (SNCosmo, Glafic, SNData) and the methodology is reproducible in principle. However, the quantitative efficiency claims are weakened by in-sample fitting of the selection curve and by an unresolved internal contradiction between the simulated CC contamination estimate and the observed CC validation sample; these issues must be addressed before the headline numbers can be taken at face value.","major_comments":[{"comment":"There is a direct internal contradiction in the core-collapse contamination estimate. Section 4.4 concludes from simulations that contamination by unlensed CC SNe is 'very low to negligible (<0.05%)', and the abstract repeats this. Section 4.5, applying the same limit to the observed SDSS-II CC SNe, states that 'the cut is, however, found to select a majority of the low redshift unlensed CC SNe sample.' Since CC SNe vastly outnumber lensed SNe Ia in an LSST alert stream, a majority selection of observed CC SNe would dominate the candidate list even if the simulated CC rejection fractions are correct. The authors attribute the discrepancy to a small sample and defer a full study, but the tension is not resolved: either the simulated CC templates or luminosity functions are too faint or too blue, the red limit is not robust to real CC SEDs, or the bandpass mismatch makes the validation invalid for LSST. In any of these cases, the quantitative rejection claim in the abstract and Sec. 4.4 is not supported by the paper's own evidence. This must be resolved by either re-fitting the CC simulation ingredients, quantifying the expected number of CC SNe passing the cut relative to lensed SNe Ia using realistic number densities, or explicitly retracting the '<0.05%' claim.","section":"4.4-4.5 and Abstract"},{"comment":"The selection efficiencies (≈44%/67% for Set 1 low-z, ≈46%/45% for high-z) are computed on the same simulated distributions that were used to fit the modified red limit. The curve in Eq. (6) is described as selected to 'better suit our Set 1 distribution,' and the reported completeness figures are therefore in-sample calibration, not independent predictions. The paper should either perform out-of-sample validation (e.g., fitting on one simulation draw and testing on an independent draw, or a cross-validation split) or clearly state that the reported numbers are in-sample efficiencies that may overstate performance when applied to real data.","section":"4.1, Eqs. (6)-(7)"},{"comment":"The simulations contain no photometric noise, detection limits, PSF size, cadence, or microlensing, as the authors acknowledge in Sec. 2.3 and Sec. 5. These are not minor omissions for a method whose purpose is to filter LSST alerts: real detections will be biased toward brighter, noisier, and epoch-restricted light curves, and photometric scatter will smear objects across the sharp red limit in Eqs. (6)-(7). The paper already notes a companion study that injects SNe into HSC data with an LSST-like cadence, which is appropriate, but as it stands the quantitative LSST applicability of Sec. 4.1 is not established. I recommend that the paper either present the idealized efficiencies as upper limits with an explicit caveat, or include a simple noise-injection experiment (e.g., adding magnitude errors and a detection threshold) to show how the efficiencies degrade.","section":"2.3, 4.1, 5"},{"comment":"The observed validation uses photometry from survey-specific r/i bands (DES, ESSENCE, JLA, ZTF, SDSS) that differ from the LSST bandpasses for which the red limit was defined. The paper acknowledges this in Sec. 4.5 but does not quantify the effect. As a result, the 'works well on observed data' claim is qualitative and cannot be used to independently confirm the simulated rejection percentages. The authors should either apply bandpass corrections (or approximate transformations) to place the observed data on the LSST system, or explicitly restrict the validation claim to a demonstration of qualitative separation without quoting effective rejection rates.","section":"3 and 4.5"}],"minor_comments":[{"comment":"Sec. 3 contains a typo: 'Trasient' should be 'Transient' in 'Zwicky Trasient Facility'.","section":"3"},{"comment":"The units for the magnitude and color in Eqs. (6) and (7) are not stated; since the paper defines AB magnitudes, it would be helpful to write 'magnitudes in AB' explicitly near the equations. Also, the thresholds 21.02 and 23.63 should be described as apparent i- and y-band magnitudes, respectively.","section":"4.1"},{"comment":"The text says the criterion 'selects the known lensed SNe Ia systems on the falling edges of respective light curves,' but Fig. 5 shows both rising and falling edge panels. Please clarify whether the known lensed systems are selected on both edges or only the falling edge, and whether the rising-edge panels are also consistent with the textual claim.","section":"4.5 / Fig. 5"},{"comment":"The phrase 'Mane et al., 2025, in prep.' is not a citable reference; if the companion study is not yet public, it should be referred to as 'companion study in preparation' in the text and omitted from the reference list, or the arXiv number should be provided if posted.","section":"2.3 / 5"},{"comment":"The data availability statement says 'All simulated data are available from the corresponding author upon request.' For reproducibility, it would be better to place the simulation scripts, the HSC galaxy catalog subset, and the generated light curves in a public repository (e.g., Zenodo or GitHub), since the paper aims to provide a practical filter for LSST.","section":"Data Availability"}],"recommendation":"major_revision","confidential_remarks":"The main reason for major revision is not the idealized simulations per se, but the unresolved internal contradiction between Sec. 4.4 and Sec. 4.5 on CC contamination. The abstract's 'negligible' CC contamination claim is contradicted by the paper's own observed validation, and the authors' proposed explanation (small sample) does not resolve which of the two results should be trusted. This is a central claim of the paper (the filter must reject unlensed transients), so it cannot be left as a deferred 'future study.' If the authors can demonstrate with realistic number densities that the observed CC events that pass the cut are a negligible fraction of the LSST alert stream, or can re-tune the CC simulation to reproduce the observed sample, the paper would be suitable for publication after those changes."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper is a useful incremental step, but the headline efficiency numbers are not yet trustworthy. The red limit is fitted to the same simulated data used to measure completeness, and the paper's own observed CC SNe sample contradicts its claim of negligible contamination.\n\nThe genuinely new bits: modified red limit curves for LSST rizy bands (Eqs. 6, 7), extension to the falling edge of light curves, and a z-y selection for high-redshift (z > 1.64) SNe. The authors also test against a handful of known lensed systems, and those do sit above the curve — that is real independent support, albeit tiny.\n\nNow the soft spots. First, Sec. 4.4 reports simulated unlensed CC SNe contamination below 0.05%, while Sec. 4.5 states the same cut selects a majority of the observed SDSS CC SNe. The authors attribute this to small sample size, but the tension is not resolved. Since CC SNe vastly outnumber lensed SNe Ia in any real alert stream, this is load-bearing: the rejection claim in the abstract is not supported by the paper's own evidence.\n\nSecond, the red limit (Eq. 6) is defined on the same simulated Set 1 sample used to quote the 44–67% completeness and 91–99% rejection. So those numbers are in-sample fits, not predictions. The noiseless simulations without detection limits, cadence, PSF, or microlensing reinforce that the quantitative claims should be treated as optimistic. No code or data is released, which also limits reproducibility.\n\nWhat the paper does well: the physical argument is coherent and consistent with earlier work (Quimby 2014, Arendse 2024), and the authors are explicit about the idealized setup and defer realistic testing to a companion paper. That is honest, but it means this manuscript is a step toward a practical tool, not a demonstration of one.\n\nBottom line: worth a serious referee, because the extension is useful and the observed lensed SNe check is encouraging. But the referee should insist the CC contamination contradiction be addressed, and the efficiency claims be re-derived on held-out simulated data with realistic noise. The paper is for people working on LSST transient triage and lensed SN searches; I'd send it to review, expecting major revision.","headline":"Useful incremental step, but the headline efficiencies are in-sample and the CC contamination claim is contradicted by the paper's own observed sample.","tokens_in":17389,"tokens_out":2734,"would_cite":true,"duration_ms":25224,"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":"The paper claims that a modified 'red limit' in color-magnitude space can pick out strongly lensed type Ia supernovae from unlensed ones across redshifts up to z=3.","keywords":["gravitational lensing","type Ia supernovae","color-magnitude selection","LSST","core-collapse supernovae","supernova identification","cosmological probes"],"falsifier":"Run the proposed red limits on the first season of real LSST difference-imaging detections with spectroscopic classifications, and compare the fraction of confirmed unlensed SNe Ia that fall above the limit to the simulated 91–99% rejection; if real photometric noise, detection thresholds, or PSF-blended images push ordinary SNe Ia above the line more often, the reported efficiencies will not reproduce.","tokens_in":16110,"feed_emoji":"🔭","tokens_out":6583,"duration_ms":56830,"temperature":0.7,"pith_summary":"Strongly lensed type Ia supernovae are rare but cosmologically valuable because their time delays and standardizable brightness can pin down the Hubble constant. This paper tries to establish that a simple color-magnitude cut, a 'red limit', can flag them quickly in the flood of LSST-style transient alerts. Simulating LSST-like photometry in $rizy$ bands, the authors find that some lensed SNe Ia sit redward of unlensed SNe Ia of the same apparent magnitude, on both the rising and falling phases of the light curve, out to $z=3$. They propose two explicit selection lines and report that the cut recovers a substantial fraction of simulated lensed SNe Ia while rejecting 91–99% of unlensed SNe Ia. If this holds in real data, the method gives a fast triage step before expensive follow-up or full lens modeling.","feed_headline":"New color limit flags lensed supernovae at up to 67% efficiency","feed_subtitle":"A brightness-color boundary in LSST alerts can flag lensed SNe Ia while rejecting ~99% of ordinary ones","key_machinery":"The central object is the observed color-magnitude diagram (CMD) built from SALT2 light-curve simulations and lensing observables; the load-bearing device is the proposed red limit, a straight line in that plane. The light curves are evaluated at fixed observer-frame epochs, three days before and seven days after the $i$-band peak, for both unresolved total flux and resolved individual images. The red limit does the selection: points above the line are candidates for lensed SNe Ia, and the analysis maps how the lensed and unlensed populations fall on either side under varying redshift, phase, and supernova type.","core_discovery":"The paper's central claim is that a subset of strongly lensed SNe Ia occupies a region of the observed color-magnitude diagram that unlensed SNe Ia do not, and that a straight-line boundary captures this separation. Lensing magnification makes the supernova appear brighter, while the preferentially higher source redshifts shift its spectral energy distribution redward, pushing candidates above the boundary. The proposed modified red limit is $m_r-m_i > 0.52 m_i - 10.96$ when $m_i > 21.02$ for sources with $z<1.64$, and $m_z-m_y > 0.59 m_y - 13.67$ when $m_y > 23.63$ for $1.64