{"id":"86de6d0c-a557-4e28-a586-3d8803708d59","arxiv_id":"2501.12525","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A simulation forecasts that a small-telescope array can detect several unresolved lensed supernovae per year and, with follow-up, constrain H0 to about 2.7 percent.","lead":"The paper simulates how often the 7DT telescope array, twenty 50-cm telescopes in Chile, could spot unresolved gravitationally lensed supernovae by watching known lens candidates, predicting about seven Type Ia and more than nine core-collapse events per year in its deepest r-band mode. The authors also forecast that follow-up of the detected systems could measure the Hubble constant to about 2.7 percent precision.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Source-redshift prescription in §3.1 is internally inconsistent: truncated normal lower bound at 1 excludes zs < 0.7, yet the rates rely on zs < 0.7.","rationale":"The reader correctly identified the source-redshift distribution in Section 3.1 as the weakest assumption, but their concern was that the distribution's shape is ad hoc and unvalidated. My stress test sharpens this into a concrete internal inconsistency: the stated truncated normal with lower bound at 1 cannot produce any of the zs < 0.7 sources on which all detection rates and the H0 forecast depend. This is more directly load-bearing than the general uncalibrated-shape concern, because it raises the possibility that the simulation as described is impossible and the text does not match the code. A single computational check—rerunning with the stated prescription—would settle whether this is a typo in the manuscript or a fundamental flaw in the simulation. If the contradiction is resolved by clarifying that the 'truncated normal' applies to a ratio or by correcting the bound, the conditional verdict can stand; if the code also produces zero events, the paper requires major revision. The reader's other concern about the 10% time-delay distance uncertainty is secondary: even if that assumption is optimistic, it only affects the secondary H0 precision claim, whereas the source-redshift issue undermines the primary detection-rate results. For these reasons I do not change the reader's conditional verdict, but I would add a specific condition that the authors must clarify and validate the source-redshift sampling procedure.","tokens_in":21439,"tokens_out":10575,"duration_ms":96802,"concrete_test":"Obtain the simulation code (the Sheu et al. 2023 pipeline or the authors' scripts) and run it with the exact source-redshift prescription stated in Section 3.1: multiply the lens redshift distribution by a truncated normal N(2,0.5) with lower bound 1, then retain only systems with zs < 0.7, and recompute the annual detection rates. If the resulting rates are zero, the text is internally inconsistent and the central rates in Table 2 need re-derivation from a corrected, explicitly described redshift distribution; if nonzero, the implementation differs from the text and the discrepancy must be resolved before the predictions can be trusted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.1 states that the source redshift distribution is obtained by multiplying the lens redshift distribution by a truncated normal N(2,0.5) with lower bound at 1, and then says 'We limit our selection to systems where the glSNe redshift is below 0.7'. Taken literally, this yields zero systems: a truncated normal with lower bound 1 has P(zs < 0.7) = 0, and the lens catalog redshifts are mostly 0.1–0.6 (Figure 3). Yet Figures 3, 5, and 6 show simulated zs below 0.7 and nonzero detection rates. This is not merely an uncalibrated assumption; it makes the simulation irreproducible as written. Either the text misdescribes the actual code (e.g., 'multiplying' may mean zs = zl × x with x drawn from the truncated normal, or the bound is misstated), or the code does something different from the paper. Under the first reading the predicted rates, and therefore the H0 forecast that depends on detecting 7 Ia and 7 CC systems per year, are unsupported; under the second reading the description must be corrected and the rates re-validated against the actual code.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper simulates the detection of unresolved strongly lensed supernovae with the 7DT small-telescope array, using 5807 strong lens candidates from DECaLS/DESI Legacy Surveys. It predicts annual detection rates for Type Ia and several core-collapse subtypes under different 7DT observing strategies, and uses a Gaussian-process, model-independent method with Pantheon SNe Ia to forecast H0 precision from follow-up of the detected systems. The headline results are 7.46 SNe Ia yr^-1 and 9.49 CC SNe yr^-1 (sum of subtypes) in the r-band deep target mode at 22.04 mag, and 2.53 SNe Ia yr^-1 in the WTS medium-band program, leading to a forecasted 2.7% H0 precision (70.03 ± 1.9 km/s/Mpc) from 7 Ia + 7 CC systems.","tokens_in":21692,"tokens_out":4533,"duration_ms":41791,"significance":"If the rates and H0 precision hold, this would demonstrate that a modest-aperture, wide-field telescope array can contribute to glSNe discovery and time-delay cosmography, complementing LSST and Roman. The paper includes a concrete simulation pipeline, explicit SN-rate formulas imported from the literature, and a reproducible GP-based analysis framework. The main value is in showing the feasibility of unresolved glSNe detection with small telescopes; however, the quantitative forecasts are currently undermined by inconsistencies in the source-redshift prescription and an unsupported time-delay distance uncertainty assumption.","major_comments":[{"comment":"The source-redshift prescription as written is internally inconsistent: a truncated normal N(2,0.5) with lower bound 1 has zero probability below zs=0.7, yet the text immediately applies a selection zs < 0.7, and Figures 3, 5, and 6 show simulated source redshifts below 0.7. Taken literally, the simulation would produce no systems at all, while the paper reports nonzero rates (e.g., 7.46 SNe Ia yr^-1 and the 7 Ia + 7 CC sample used in Section 4). Please correct the description (e.g., if the implementation is zs = zl × x with x drawn from a truncated normal, state that explicitly) or re-run the simulation with the stated distribution and re-derive the rates. As written, the central detection-rate and H0-precision forecasts are not reproducible.","section":"Section 3.1"},{"comment":"The H0 forecast assumes sigma_DDelta_t = 0.1 D for every mock system, including unresolved systems whose simulated time delays have a median of 6.83 days (Section 3.3). The paper itself notes that short delays are disadvantaged for precise time-delay measurements. A 10% time-delay distance uncertainty for a few-day delay in an unresolved, blended system is optimistic and is the main driver of the 2.7% precision in Table 2. Please justify this assumption with a realistic error budget or present the H0 precision as a function of sigma_DDelta_t; otherwise the forecast is a self-consistency check of the assumed noise rather than a prediction of the 7DT+follow-up program.","section":"Section 4, Eq. (4)"},{"comment":"The headline detection rates are quoted as 'maximum expected' under ideal conditions, but the simulation uses the full 5807-candidate sample without a lens-likelihood threshold (Figure 2 shows many candidates with likelihood near 0.1–0.3), assumes 100% observing efficiency and 7-hour nights with no weather or moon-phase losses, and does not model the lens-galaxy subtraction or image blending in detail. Please provide a fiducial scenario (e.g., applying a likelihood cut, adopting a weather fraction, or including a detection-efficiency factor) alongside the ideal-case maxima, so that the annual rates quoted in the abstract are not read as realistic expectations.","section":"Sections 3.1 and 3.4"}],"minor_comments":[{"comment":"Change 'microlensig' to 'microlensing'.","section":"Section 3.3"},{"comment":"Clarify the relation between the 5807-candidate parent sample and the ~1225 systems within the 7DT footprint; the text says 'around 1225 lens systems and candidates' but does not state the declination or likelihood cuts used.","section":"Section 3.1"},{"comment":"Unify the WTS depth: the abstract quotes 20.61 mag for the m6000 filter, while Table 1 lists 20.60 for m600; also the filter is referred to as both m6000 and m600.","section":"Abstract and Table 1"},{"comment":"The caption states zs <= 0.6, while Section 3.4.1 says systems with 0.6 < zs < 0.7 are not detectable; make the redshift ranges consistent across text and figures.","section":"Figure 5 caption"},{"comment":"Equation (3) uses H0DDelta_t but does not define DDelta_t before first use; add a sentence defining the time-delay distance.","section":"Section 4"},{"comment":"Fix rendering artifacts in author names (e.g., 'Sagu´ es', 'Ca˜ nameras') and the spacing in '2 .7%' in the abstract.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of A&A and the topic is timely given 7DT operations and LSST survey start. The main issue is the internal inconsistency in §3.1, which the authors should be able to resolve by clarifying the code; I did not find evidence of intentional misrepresentation. I would support a major revision requiring a corrected source-redshift description, a sensitivity analysis of the H0 forecast to the adopted time-delay distance uncertainty, and a non-ideal detection-rate scenario."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a useful but soft forecast. The paper adapts an established lensed-SN simulation pipeline (Sheu et al. 2023) to the 7DT array, with updated time-delay and magnification statistics from Bag et al. (2024) and Sagúes Carracedo et al. (2024), and produces the first 7DT-specific detection rates: ~7 Type Ia and ~9 core-collapse lensed SNe per year in r-band at 22.04 mag under ideal conditions, plus a 2.7% H0-precision projection from a mock follow-up sample. That is a legitimate, useful extension: small-telescope arrays monitoring known lens fields could genuinely open a new discovery channel for unresolved lensed SNe.\n\nThe paper is honest about its idealizations—microlensing is ignored, observing conditions are ideal, and the CC templates are crude. It also gives credit where due. The best part is the setup: a concrete observational program (1125 tiles, depth 22.04, 7–14 day cadence) derived from control-time analysis. That is actionable.\n\nThe soft spots are real but not fatal. The source-redshift construction in §3.1 is ad hoc: it looks like zs = zl × X with X drawn from a truncated normal N(2,0.5) lower-bounded at 1, then zs < 0.7 is selected. The text is ambiguous—it could be read as multiplying distributions, which would be contradictory—but Figure 3's ratio plot (zl/zs < 1) supports the multiplicative interpretation. Either way, the recipe is not calibrated against any observed lensed SN redshift distribution, and the predicted yield is highly sensitive to the shape of the low-zs tail because 7DT sees only zs < 0.7. A different plausible assumption could shift the annual rates by a factor of several. The lens sample is also unfiltered: all 5807 candidates are used regardless of lens probability (Figure 2 shows many low-likelihood entries), so the rates are likely upper bounds. The H0 forecast assumes 10% time-delay distance uncertainties for systems with median delays of a few days; that seems optimistic given the short-delay disadvantage the paper itself notes, and no systematic budget for lens modeling or microlensing is included. So 2.7% H0 precision should be read as a best-case forecast, not a realistic expectation.\n\nWho is this for? Survey strategists and people thinking about small-telescope time-domain programs; also as a reference for lensed-SN rate estimates. It deserves a serious referee—the methodology is imported and the conclusions are conditional, but the specific numbers are useful and the assumptions are transparent enough to be tested. I'd recommend sending it to review, with a request that the authors clarify the source-redshift construction, validate it against existing glSNe or lensing cross-section expectations, and soften the H0 claim to reflect the idealized uncertainty model.","headline":"A useful, honest forecast of how a small-telescope array could find unresolved lensed SNe; the order-of-magnitude rates are plausible, but the specific numbers and the H0 forecast rest on assumptions a referee should probe.","tokens_in":22345,"tokens_out":4977,"would_cite":true,"duration_ms":47499,"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 authors forecast that the 7-Dimensional Telescope, an array of twenty 50-cm telescopes, can detect about 7 Type Ia and 9 core-collapse lensed supernovae per year in known lens fields, and that follow-up of a subset could measure the…","keywords":["gravitationally lensed supernovae","time-delay cosmography","Hubble constant","small-aperture telescope array","transient detection rates","Gaussian process regression","strong lens candidates","unresolved lensed images"],"falsifier":"Count the number of the 5,807 candidate lens fields that contain a bright background source at $z_s \\lesssim 0.7$ using existing deep imaging; the predicted annual yields are nearly proportional to that count, so a much smaller population would collapse the forecast. A direct empirical check is the first year of the proposed 7DT $r$-band monitoring: observing far fewer than roughly 17 unresolved lensed supernovae would indicate that the source-redshift tail, the magnification distribution, or the unresolved fraction is more optimistic than reality.","tokens_in":21195,"feed_emoji":"🔭","tokens_out":10850,"duration_ms":97902,"temperature":0.7,"pith_summary":"This paper argues that a small-telescope array can serve as a discovery engine for gravitationally lensed supernovae even though it cannot resolve the separate lensed images. Simulating synthetic light curves for 5,807 strong-lens systems and candidates, the authors forecast that the 7-Dimensional Telescope, twenty 50-cm telescopes with a combined 25 deg$^2$ field of view, will detect at most about 7.46 Type Ia and 9.49 core-collapse lensed supernovae per year by monitoring roughly 1,125 target fields in the $r$-band at a depth of 22.04 mag. The physical mechanism is that unresolved systems add their lensed images together, so the short-delay systems that small telescopes can see are the brighter ones. With dedicated follow-up of a subset of the detected systems, the paper forecasts a model-independent measurement of the Hubble constant to 2.7% precision, $H_0 = 70.03 \\pm 1.9$ km/s/Mpc. If these rates hold, a modest ground-based facility becomes a viable feeder for one of the few independent cosmological probes of the expansion rate.","feed_headline":"17 lensed supernovae a year forecast for a small-telescope fleet","feed_subtitle":"A fleet that cannot resolve lensed images could still tie the Hubble constant to 2.7 percent.","key_machinery":"The detection forecast is carried by synthetic light curves: SNCosmo with the SALT3 model for Type Ia and six template families for core-collapse subtypes, with image multiplicities, log-normal magnifications, and exponential time delays (mean 6.83 days) tuned to the unresolved, short-delay population inferred from ZTF and LSST studies, and with a system counted as detected when the combined (unresolved) light curve rises above the survey depth for a control time long enough to be caught. The cosmological forecast is carried by the time-delay distance relation $H_0 D_{\\Delta t} = (1+z_l)\\, H_0 D_l\\, H_0 D_s / H_0 D_{ls}$, evaluated with distances from Gaussian-process reconstructions of the unanchored luminosity distance $H_0 D_L(z)$ built from the Pantheon sample, and then compared against mock time-delay distances drawn from a flat $\\Lambda$CDM fiducial model with 10% noise; $H_0$ is obtained by marginalizing over the Gaussian-process realizations.","core_discovery":"The central claim is that unresolved, blended gravitationally lensed supernovae are detectable at useful rates with a small-aperture multi-telescope system, and that the systems thereby found are sufficient for a model-independent constraint on the Hubble constant. Under ideal conditions, monitoring the strong-lens and candidate sample in the $r$-band at a depth of 22.04 mag yields maximum annual rates of 7.46 Type Ia and 9.49 core-collapse detections, split among the core-collapse subtypes (2.49 Ic, 0.80 IIb, 0.52 IIL, 0.78 IIn, 3.75 IIP, 1.15 Ib); a shallower medium-band program at 20.61 mag yields 2.53 Type Ia events per year. Taking the detected systems as the input to a model-independent Gaussian-process analysis that anchors the Pantheon supernova distances to measured time-delay distances, the forecast returns $H_0 = 70.03 \\pm 1.9$ km/s/Mpc, i.e., 2.7% precision, for a combined sample of seven Type Ia and seven core-collapse systems.","pith_inferences":["The yield is calibratable before the survey runs: a count of $z_s \\lesssim 0.7$ background sources behind the candidate fields, from existing deep multi-band imaging, would pin down the dominant uncertainty in the rate forecast.","The 2.7% $H_0$ precision should be read as an ideal-case bound rather than a guaranteed outcome, since the simulated systems have short delays and the paper itself flags these as disadvantaged for precise time-delay measurement; including microlensing, which the paper sets aside, would tend to add scatter on exactly these short-delay systems.","The same control-time machinery transfers to other small-telescope arrays and to larger future lens catalogs, so the reusable result is the method for turning an unresolved-transient survey into a lensed-supernova yield estimate, not just the 7DT numbers."],"forward_implications":["Under ideal conditions, the 7DT target program in the $r$-band at depth 22.04 mag should detect about 7.46 Type Ia and 9.49 core-collapse lensed supernovae per year within the ~1,125 fields covering 5,807 strong-lens systems and candidates.","The 7DS wide-field medium-band program at the m6000 filter depth of 20.61 mag is forecast to detect about 2.53 Type Ia lensed supernovae per year.","Following up seven Type Ia and seven core-collapse detections with larger telescopes, the model-independent Gaussian-process method yields $H_0 = 70.03 \\pm 1.9$ km/s/Mpc, a 2.7% precision measurement.","Unresolved glSNe, having median time delays of a few days, are the bright, blended channel that small telescopes can catch; the trade-off is that short delays worsen time-delay-measurement precision for $H_0$.","A complementary strategy, proposed but not simulated, places 7DT monitoring in LSST's low-cadence rolling-season fields to add early detection and medium-band classification where LSST coverage is sparse."],"supporting_citations":[{"why":"Supplies the synthetic glSNe light-curve and rate methodology, including the source-redshift simulation and star-formation-history treatment, on which the simulation is built.","marker":"Sheu et al. (2023)"},{"why":"Provides the strong-lens and candidate catalog in the DESI Legacy Imaging Surveys footprint that defines the target list for 7DT monitoring.","marker":"Huang et al. (2020, 2021)"},{"why":"Extends the catalog of strong-lens candidates, adding the systems whose spatial distribution sets the ~1,125 target fields.","marker":"Storfer et al. (2024)"},{"why":"Provides the model-independent Gaussian-process method for anchoring SNe Ia distances with time-delay distances to estimate $H_0$.","marker":"Liao et al. (2019, 2020)"},{"why":"Supplies the time-delay-distance likelihood and posterior-marginalization procedure adopted for the $H_0$ forecast.","marker":"Li et al. (2024)"},{"why":"Supplies the Pantheon SNe Ia relative-distance dataset used for the Gaussian-process reconstructions of $H_0 D_L(z)$.","marker":"Scolnic et al. (2018)"},{"why":"The original time-delay cosmography formalism that yields the $H_0 D_{\\Delta t}$ relation used in the forecast.","marker":"Refsdal (1964)"},{"why":"Establishes that unresolved glSNe have shorter time delays and brighter summed light curves, motivating the small-telescope detection channel and the delay distribution used here.","marker":"Bag et al. (2024)"},{"why":"Provides the ZTF-based distributions of angular separation, magnification, and time delay that replace the earlier broad distributions in the simulation.","marker":"Sagués Carracedo et al. (2024)"},{"why":"Supplies the core-collapse and Type Ia supernova rate normalizations and the delay-time distribution used in the annual-rate estimates.","marker":"Shu et al. (2018)"}],"fun_headline_variants":["Small scopes forecast 17 lensed supernovae a year","Lensed supernovae spotted unresolved: 17 yearly from small scopes","A fleet of 50-cm telescopes could detect 17 lensed SNe per year","2.7% Hubble constant precision from unresolved lensed SNe","Unresolved lensed supernovae yield 17 yearly detections and H0 at 2.7%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire detection-rate forecast rests on the assumed redshift distribution of the supernova sources behind the lens candidates, specifically the low-redshift tail below $z_s = 0.7$ that the 7DT can actually see; the paper imposes this tail by multiplying the lens redshift distribution with a truncated normal distribution rather than measuring it from the lens catalog or from observed lensed supernovae.","fun_headline_variants_meta":{"raw":{"variants":["Small scopes forecast 17 lensed supernovae a year","Lensed supernovae spotted unresolved: 17 yearly from small scopes","A fleet of 50-cm telescopes could detect 17 lensed SNe per year","2.7% Hubble constant precision from unresolved lensed SNe","Unresolved lensed supernovae yield 17 yearly detections and H0 at 2.7%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000337,"raw_usage":{"total_tokens":1992,"prompt_tokens":1197,"completion_tokens":795,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":813,"completion_tokens_details":{"reasoning_tokens":687}},"tokens_in":813,"tokens_out":795,"duration_ms":7761,"temperature":1.0,"reasoning_tokens":687,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:06:56.869392+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Count the number of the 5,807 candidate lens fields that contain a bright background source at $z_s \\lesssim 0.7$ using existing deep imaging; the predicted annual yields are nearly proportional to that count, so a much smaller population would collapse the forecast. A direct empirical check is the first year of the proposed 7DT $r$-band monitoring: observing far fewer than roughly 17 unresolved lensed supernovae would indicate that the source-redshift tail, the magnification distribution, or the unresolved fraction is more optimistic than reality.","supporting_citations":[{"cited_title":"2023, ApJ, 952, 10","cited_arxiv_id":null,"evidence_quote":"Supplies the synthetic glSNe light-curve and rate methodology, including the source-redshift simulation and star-formation-history treatment, on which the simulation is built."},{"cited_title":"2020, ApJ, 894, 78","cited_arxiv_id":null,"evidence_quote":"Provides the strong-lens and candidate catalog in the DESI Legacy Imaging Surveys footprint that defines the target list for 7DT monitoring."},{"cited_title":"E., & Linder, E","cited_arxiv_id":null,"evidence_quote":"Provides the model-independent Gaussian-process method for anchoring SNe Ia distances with time-delay distances to estimate $H_0$."},{"cited_title":"S., Mao, S., et al","cited_arxiv_id":null,"evidence_quote":"Supplies the core-collapse and Type Ia supernova rate normalizations and the delay-time distribution used in the annual-rate estimates."}],"review_version":1}