{"id":"a8298a9a-a4c4-48c8-8a9c-6a5024f88388","arxiv_id":"2411.12741","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"A membership probability analysis of Gaia and OGLE data finds 23 RR Lyrae candidates in 10 intermediate-age Magellanic Cloud clusters and infers lower-limit RR Lyrae production rates for young and intermediate-age populations.","lead":"The authors searched for RR Lyrae variable stars near intermediate-age star clusters in the Magellanic Clouds and found 23 candidate members in 10 clusters. If confirmed, these would be among the first direct evidence that RR Lyrae stars can form in populations only a few billion years old, not just ancient ones.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Intermediate-age DTD hinges on a control-field subtraction whose statistical significance is not demonstrated; Table 3 shows comparable cluster and control counts for several young/intermediate clusters.","rationale":"I read the paper in good faith and identify its central claim as the decontaminated DTD for young/intermediate-age clusters. The old-age (>8 Gyr) DTD is well-anchored by many clusters and large N_RRLs, and the paper's agreement with S21 and with Galactic globular cluster rates is independent support. The load-bearing weakness is not the membership model itself (which is standard and validated by the CMD locus), but the control-field subtraction that converts the raw DTD into the headline decontaminated values. The reader's weakest_assumption is exactly this control-field representativeness, so I agree. The paper itself acknowledges spatial background variations and that control-field counts sometimes exceed cluster expectations (Table 3), which means the subtraction can move the small-number signal by an order of magnitude. Since the paper presents the decontaminated values as the primary result and explicitly warns they are lower limits, the failure mode is over-subtraction: if the control fields overestimate contamination at the cluster centers (e.g. because the field RRL density is lower along the cluster sightline due to correlated star formation or crowding incompleteness), the true DTD could be higher, but if they underestimate it (e.g. because the cluster field's background is enhanced by the cluster's own extended population or by LMC structure), the young/intermediate DTD could be consistent with zero. The test I propose settles this by propagating the 8 control fields' distribution and comparing with a no-subtraction baseline. I recommend keeping CONDITIONAL because the paper is honest about radial-velocity confirmation being needed and about incompleteness; the concern strengthens the condition but does not change the verdict category. No ad hominem, no theatrics: the issue is statistical robustness, not integrity.","tokens_in":26075,"tokens_out":2322,"duration_ms":20758,"concrete_test":"Recompute the decontaminated DTD with the full posterior of N_cont: for each cluster, draw N_cont from the distribution of 8 control fields (or a Poisson-Gamma posterior), and marginalize in the hierarchical model of Section 4 rather than subtracting the median. Also rerun the inference with no control-field subtraction and with an alternative contamination prior based on the model's own background mixture term. If the 1-2 Gyr bin median changes by more than ~0.15 RRL/10^5 Msun or its 84th percentile includes 0, the claimed young/intermediate DTD is not robust to the control-field assumption.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim is the decontaminated DTD in the 1-2 and 2-8 Gyr bins (0.34 and 0.071 RRL/10^5 Msun). The decontamination proceeds by subtracting, for each cluster, the median control-field member count N_cont (Section 4.1, Table 3) from the cluster expectation <N_RRLs>. The implicit premise is that the control fields at 23 r_h provide an unbiased estimate of contamination inside the cluster field, i.e. that the field RRL population is statistically identical between cluster sightlines and off-cluster control fields in proper motion, photometric distance, and density. This premise is load-bearing and not demonstrated: (1) the LMC/SMC field RRL density and distance distribution vary spatially (the paper itself cites strong spatial variations in the background, Section 4.1), so an annulus at 23 r_h may not represent the field at the cluster center; (2) only 8 control fields per cluster are used, yet N_cont values are medians thereof, so the sampling uncertainty of N_cont is not propagated into the DTD posterior; (3) Table 3 shows the signal is fragile: NGC 2121 has <N_RRLs>=3.5 with N_cont=4.0, NGC 1718 has 1.0 vs 1.3, NGC 1806 has 1.7 vs 1.7, and NGC 2213 has 3 vs 1.6 after Cuevas-Otahola's own Table 3. After subtracting N_cont, several clusters retain only a fraction of an RRL, so the aggregate 1-2 Gyr signal of 0.34 is substantially driven by a few clusters where cluster and control counts are within Poisson noise. Because the control fields are at a different radius and often different surrounding environment than the cluster line of sight, the subtraction could be over- or under-subtracting. The reader's weakest_assumption correctly identified this control-field representativeness as the load-bearing assumption.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper searches for RR Lyrae (RRL) stars in intermediate-age Magellanic Cloud clusters as a direct test of the existence of RRL populations younger than the canonical \\gtrsim 10 Gyr. The authors combine the Gaia DR3 Specific Object Study and OGLE-IV RRL catalogs with a catalog of LMC/SMC cluster parameters, and build a probabilistic mixture model that assigns membership probabilities on the basis of proper motions and photometric parallaxes derived from the Muraveva et al. (2018) G-band relation. They identify 23 RRL candidates with p>0.5 in 10 clusters with ages 1\\,---\\,8 Gyr, and infer the RRL delay time distribution (DTD) in three age bins with a hierarchical Bayesian model. After subtracting median control-field counts as a contamination correction, they report decontaminated DTD values of 0.34\\,RRL/10^5\\,M_\\odot in the 1\\,---\\,2 Gyr bin, 0.071\\,RRL/10^5\\,M_\\odot in the 2\\,---\\,8 Gyr bin, and 2.5\\,RRL/10^5\\,M_\\odot in the >8 Gyr bin. The old-age value agrees with the field DTD of Sarbadhicary et al. (2021), while the young and intermediate-age values are substantially lower, which the authors interpret as evidence for an order-of-magnitude lower RRL production rate at intermediate ages, possibly tied to binary evolution.","tokens_in":26486,"tokens_out":6793,"duration_ms":65252,"significance":"If the cluster associations are correct, this is the first direct evidence linking RRL stars to intermediate-age simple stellar populations, and it would provide a strong constraint on binary-evolution formation channels for RRLs. The paper is methodologically transparent: the membership model, the hierarchical DTD inference, and the control-field construction are described in enough detail to be reproduced from public catalogs, and the agreement of the old-age DTD with an independent field measurement by S21 is a nontrivial consistency check that supports the overall approach. The candidate list with light curves and membership probabilities is a useful resource for spectroscopic follow-up. The main weakness is that the young/intermediate-age DTD values rest on a small number of candidates after a control-field subtraction whose statistical uncertainty and systematic assumptions are not fully quantified; the quantitative rate measurements should be treated with caution until that treatment is strengthened.","major_comments":[{"comment":"The decontaminated DTD\\u2014the central result\\u2014is obtained by subtracting the median control-field count N_cont from each cluster\\u2019s expectation value <N_RRLs> and then feeding the difference into the hierarchical Gamma/Poisson likelihood of Eqs.\\ (7)\\u2013(8). However, N_cont is the median of only 8 control fields, is reported without an uncertainty, and for several clusters the Poisson sampling noise on N_cont is comparable to the signal itself (e.g., NGC\\ 2121 has <N_RRLs>=3.5 and N_cont=4.0; NGC\\ 1806 has 1.7 and 1.7; NGC\\ 1718 has 1 and 1.3). Treating N_cont as a known constant and not propagating its sampling distribution into the DTD posterior will understate the uncertainties on the 1\\u20132 and 2\\u20138 Gyr bins. I recommend including N_cont in the hierarchical model with its own prior (e.g., Gamma or Poisson) and marginalizing over it, or at minimum performing a sensitivity analysis in which N_cont is drawn from its sampling distribution and the DTD fit is repeated.","section":"Section 4.1, Table 3"},{"comment":"The control fields are placed at 23 r_h with all cluster parameters kept unchanged, which assumes that the field RRL population in proper motion, photometric distance, and density is statistically identical between the control sightlines and the cluster sightlines. The paper itself notes in this same section that the LMC/SMC background has strong spatial variations, yet no test of this assumption is provided. Because several intermediate-age clusters have N_cont comparable to or larger than <N_RRLs>, the residual young/intermediate-age signal depends directly on this assumption. A concrete test would be to measure N_cont at multiple radii (e.g., 10, 23, and 40 r_h) and to compare with maps of the field RRL density; if the scatter is substantial, the DTD inference should include a systematic component for the contamination level.","section":"Section 4.1"},{"comment":"The paper states that the decontaminated results 'should be taken as lower limits' because incompleteness of the RRL catalogs in crowded cluster centers is not accounted for, yet the abstract and Table 4 present 0.34 and 0.071 RRL/10^5 M_\\odot as the inferred rates without this caveat. Since contamination and incompleteness act in opposite directions, the quoted numbers are not unbiased estimates of the cluster DTD; they are lower limits under the stated assumptions. The abstract should be reworded either to present these as lower limits or to provide a quantitative completeness correction, so that readers do not interpret the point estimates as the full answer.","section":"Abstract, Section 5.1, Conclusions"},{"comment":"The hierarchical DTD inference uses the expectation value <N_RRLs> = sum of membership probabilities as the observable N_obs in the Gamma-distribution likelihood of Eq. (8), discarding the posterior distribution P(N_RRLs|D) that the model in Section 3 explicitly computes for each cluster. For candidates with membership probability near 0.5, the difference between the expectation value and the actual count is not negligible, and the uncertainty in membership probabilities is not propagated into the DTD posterior. I suggest marginalizing over the full P(N_RRLs) using posterior samples in the hierarchical fit, at least for the clusters that drive the young/intermediate-age bins, rather than using the point estimate <N_RRLs>.","section":"Sections 3 and 4, Eqs. (7)-(8)"}],"minor_comments":[{"comment":"The text refers to 'the mean number of RRL members identified in control fields' but Table 3 lists the 'Median number of contaminants estimated in 8 control fields.' Please make the statistic used consistent.","section":"Section 4.1, Table 3"},{"comment":"The statement 'Star clusters with smaller symbol sizes were not considered for the analysis' is unclear, since Table 1 appears to include clusters with small symbols; please clarify which clusters were excluded and why.","section":"Figure 1 caption"},{"comment":"The sentence 'I21 and S21 thus reach a similar conclusion as S21 that intermediate-age RRL stars exist' is confusing; it should read that both studies reach a similar conclusion, or the wording should be revised.","section":"Introduction, paragraph 6"},{"comment":"The sentence 'all of these works predate Gaia DR2 and, not having any kinematic membership criteria, are likely to include back/foreground MC contaminant RRLs' is imprecise for the Kuehn et al. works; the kinematic criterion refers to the present analysis rather than to the earlier searches, so please clarify.","section":"Section 5.2"},{"comment":"The column header uses \\Delta/R_lim for the angular separation ratio, but the column description in the table notes does not define R_lim; a footnote stating that R_lim is R_t or 3 R_h as described in Section 3 would be helpful.","section":"Table 2"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a timely question and the candidate list is a valuable resource for follow-up, so I see clear value in this work. The main concern is whether the control-field subtraction can quantitatively support the quoted intermediate-age DTD values; this is fixable with a more careful propagation of control-field statistics and a sensitivity analysis. The agreement of the old-age DTD with S21 is a strong point in favor of the methodology. I recommend major revision, with the expectation that the authors can address the control-field uncertainty without changing the overall scope of the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this deserves a serious referee, and the CONDITIONAL verdict is the right one. The old-age DTD (>8 Gyr) agreeing with the S21 field measurement is a genuinely non-trivial consistency check: two independent methods, cluster versus field, produce the same RRL production rate. That gives me real confidence in the methodology.\n\nWhat is new: the membership probability catalog for RRL around 40 Magellanic clusters, the 23 candidates in 10 intermediate-age clusters, and the first cluster-based DTD for RRL. The paper is transparent about the limits: no radial velocities yet, incompleteness in cluster centers not quantified, decontaminated values labeled lower limits. I also appreciate that they present the light curves and period-amplitude diagram for the 23 candidates; that makes the list immediately usable for follow-up.\n\nThe soft spot is exactly where you put it. Decontamination subtracts the median N_cont from the cluster expectation, and Table 3 shows several young/intermediate clusters where the control-field count is comparable to or above the cluster expectation: NGC 2121 has N_cont=4.0 vs <N_RRLs>=3.5, NGC 1806 1.7 vs 1.7, NGC 1718 1.3 vs 1.0, NGC 2213 1.6 vs 3.0. Since the control fields are at 23 r_h and the background varies spatially—the paper says so itself—the subtracted counts are not obviously unbiased estimates of the contamination inside the cluster field. The median of eight control fields also has sampling noise that never enters the DTD posterior. So the 1–2 and 2–8 Gyr rates of 0.34 and 0.071 RRL/10^5 Msun are plausible lower limits, but their statistical significance is not established. The central claim of the paper, however, does not collapse: the 23 candidates themselves, with p>0.5 and inside 3 r_h, stand independently of the subtraction. Three were already found by Salinas et al. and previously assigned to the field; that is a nice external point.\n\nFor a referee, I would want the N_cont sampling distribution propagated into the DTD, and a discussion of what could break the representativeness assumption. I would also like a comparison of control-field and cluster-field RRL properties, not just counts, to justify the subtraction. The literature coverage is solid, with proper credit to S21, I21, Walker 1989, and the binary evolution models.\n\nFor who this is for: anyone working on RR Lyrae formation channels, binary evolution, or Magellanic Cloud star clusters. It deserves a serious referee and, with the requested revisions, is a solid contribution. I'd bring it to reading group.","headline":"Careful, honest search for intermediate-age RR Lyrae; old-age DTD is a strong result, but the young/intermediate signal leans on a control-field subtraction of unproven representativeness.","tokens_in":27070,"tokens_out":5539,"would_cite":true,"duration_ms":45796,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"23 RR Lyrae stars are probable members of 10 intermediate-age (1–8 Gyr) Magellanic Cloud clusters, and the inferred delay time distribution puts RR Lyrae production in young populations about an order of magnitude below the old-population…","keywords":["RR Lyrae stars","intermediate-age stellar populations","Magellanic Clouds","star clusters","delay time distribution","Gaia DR3","OGLE IV","binary stellar evolution"],"falsifier":"Measure radial velocities of the 23 candidates at about 1 km/s precision: true cluster members must match each cluster's systemic velocity within its few-km/s dispersion, and a systematic offset would overturn the membership claim. A companion calculation would recompute the DTD with clusters whose control-field contaminant counts rival their member counts — NGC 1806 (1.7 members, 1.7 contaminants) in the young bin, and NGC 2121 (3.5 members, 4.0 contaminants) in the intermediate bin — removed; if the young or intermediate signal disappears, the decontamination subtraction is carrying the result.","tokens_in":25904,"feed_emoji":"⭐","tokens_out":22205,"duration_ms":169620,"temperature":0.7,"pith_summary":"RR Lyrae stars are the canonical tracers of the oldest stellar populations, but this paper argues they can also form in populations only 1–8 Gyr old. Because no individual RR Lyrae can be dated directly, the authors searched for them inside intermediate-age clusters of the Magellanic Clouds, whose ages are reliably known, and identified 23 RR Lyrae as probable members of 10 such clusters using Gaia DR3 proper motions and photometric distances in a probabilistic membership model. From the ensemble they infer a delay time distribution — the number of RR Lyrae produced per unit stellar mass as a function of time since star formation — of $0.34^{+0.17}_{-0.12}$, $0.071^{+0.073}_{-0.041}$, and $2.5^{+0.4}_{-0.3}$ RR Lyrae per $10^5\\,M_\\odot$ for populations aged 1–2, 2–8, and >8 Gyr, after subtracting contamination estimated from off-cluster control fields. If the memberships survive radial-velocity confirmation, this would be the first direct proof that intermediate-age RR Lyrae exist, supporting binary mass transfer as their formation channel and lowering the age floor usually attached to this star class.","feed_headline":"RR Lyrae stars found in clusters as young as 2 Gyr","feed_subtitle":"If the memberships hold, RR Lyrae form in young populations via binary evolution — overturning the 10-Gyr age floor.","key_machinery":"The argument runs on three pieces of machinery. First, a two-component mixture model assigns each RR Lyrae a membership probability: the cluster component is a Gaussian in photometric parallax and proper motion centred on the cluster's values, and the background component is a Gaussian mixture fitted to an annular control region around the cluster; because RR Lyrae photometric distances from the $M_G$–[Fe/H] relation carry errors below 6%, the distance term separates cluster members from the Magellanic field far more cleanly than Gaia's own parallaxes. Second, a hierarchical Bayesian model converts per-cluster expected member counts into the delay time distribution: the expected number of RR Lyrae in a cluster is $\\lambda = M \\times \\mathrm{DTD}$, modelled with a Gamma likelihood, a generalisation of the Poisson distribution suited to the real-valued expected counts, that marginalises over the uncertainty in each cluster's initial mass under uniform priors on $\\log \\mathrm{DTD}$ and $\\log M$. Third, a control-field contamination test reruns the identical membership model on eight off-cluster fields centred at 23 half-light radii, and the median number of false-positive members found there is subtracted from each cluster's expected count before the DTD is inferred.","core_discovery":"RR Lyrae stars are not exclusively old: 23 RR Lyrae with membership probability $p>0.5$ are probable members of 10 Magellanic Cloud clusters with ages between 1 and 8 Gyr — three in the Small Magellanic Cloud (NGC 339, NGC 361, NGC 419) and seven in the Large Magellanic Cloud (Hodge 14, NGC 1718, NGC 1806, NGC 1846, NGC 2121, NGC 2153, NGC 2213). Modelling each cluster's expected RR Lyrae count against its initial mass, the authors infer the delay time distribution of RR Lyrae production and, after subtracting control-field contamination, obtain $0.34^{+0.17}_{-0.12}$ RR Lyrae per $10^5\\,M_\\odot$ at 1–2 Gyr, $0.071^{+0.073}_{-0.041}$ at 2–8 Gyr, and $2.5^{+0.4}_{-0.3}$ at >8 Gyr. The old-population rate agrees with the delay time distribution previously inferred from the LMC field population, while the young and intermediate rates sit roughly an order of magnitude lower than the field estimates; the paper presents them as lower limits because the RR Lyrae catalogues are incomplete in crowded cluster centres. Together the three bins assert that RR Lyrae do form in populations far younger than the canonical 10 Gyr, at rates so low that only clusters of order $10^5\\,M_\\odot$ or more are expected to host even one — which is why intermediate-age RR Lyrae have remained undetected until now.","pith_inferences":["A decisive check could be made with data already in hand: if the 23 candidates differ systematically from old-population RR Lyrae of the same galaxies in their period–amplitude distribution, that would independently support a distinct formation route before any spectroscopy is obtained.","The decontaminated values are lower limits by the authors' own account; future Gaia releases that recover RR Lyrae in crowded cluster cores could push the 1–2 Gyr and 2–8 Gyr rates upward, possibly closing the gap with the field DTD and removing the need for an environmental explanation.","If binary evolution is the channel, the DTD's shape — higher at 1–2 Gyr than at 2–8 Gyr — is a quantitative prediction for binary population synthesis codes, which so far report no DTD for their binary-evolution RR Lyrae.","The same membership machinery transfers to Andromeda's massive clusters once deep RR Lyrae catalogues exist there, provided the cluster ages come from turn-off photometry rather than integrated light, which is known to misjudge ages."],"forward_implications":["The 23 candidates — especially the three in NGC 1846, which an earlier variable-star search had assumed to be field stars — are concrete targets for radial-velocity and metallicity follow-up that could deliver the first direct confirmation of intermediate-age RR Lyrae.","A $10^5\\,M_\\odot$ cluster of age 1–8 Gyr is expected to contain fewer than one RR Lyrae, which explains why the much less massive intermediate-age clusters of the Milky Way have never been found to host any.","The old-population DTD of $2.5$ RR Lyrae per $10^5\\,M_\\odot$, equivalent to about one RR Lyrae per $10^4\\,M_\\odot$ at the present day, matches the rate observed in Galactic globular clusters and validates the inference method.","The cluster DTD sits below the field DTD at young and intermediate ages, implying either that the field estimate was inflated by Milky Way foreground contamination or that dense cluster environments suppress the binaries that produce young RR Lyrae.","The inferred present-time frequencies predict at most 0.2 RR Lyrae in clusters younger than 8 Gyr with present masses below $10^4\\,M_\\odot$, consistent with the absence of RR Lyrae in all known Milky Way intermediate-age clusters."],"supporting_citations":[{"why":"Supplies the field-population delay time distribution and the intermediate-age rate expectation of about one RR Lyrae per hundred thousand solar masses against which the cluster DTD is measured.","marker":"S21"},{"why":"The Gaia DR3 Specific Object Study catalogue is one of the two RR Lyrae data sources searched around the clusters.","marker":"Clementini et al. 2023"},{"why":"The OGLE-IV catalogue provides the deep RR Lyrae survey coverage of the Magellanic Clouds, the other data source.","marker":"Soszyński et al. 2016"},{"why":"The absolute-magnitude–metallicity relation used to convert G-band photometry into the precise photometric distances that separate members from field stars.","marker":"Muraveva et al. 2018"},{"why":"The probabilistic mixture-model formalism on which the membership probability inference is built.","marker":"Foreman-Mackey 2014"},{"why":"The earlier targeted search that found no RR Lyrae in SMC intermediate-age clusters and concluded RR Lyrae do not occur younger than about 11 Gyr; this paper's positive result reverses that baseline.","marker":"Walker 1989"},{"why":"The initial-mass-function parameters used to convert present-day cluster masses into initial masses, which enter the DTD denominator.","marker":"Baumgardt et al. 2023"},{"why":"Binary stellar evolution models that produce RR Lyrae at intermediate ages; the formation channel the positive young DTD would support.","marker":"Bobrick et al. 2022"}],"fun_headline_variants":["Young RR Lyrae: 23 stars in 1-8 Gyr Magellanic clusters","RR Lyrae not just old: 23 in intermediate-age clusters","Intermediate-age RR Lyrae: 23 in Magellanic clusters","23 RR Lyrae in 1-8 Gyr clusters challenge 10-Gyr age rule"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole result hinges on the assumption that the eight control fields placed at 23 half-light radii from each cluster centre contain the same background population of RR Lyrae stars — in density, distance, and proper motion — as the line of sight to the cluster itself, so that subtracting their false-member counts removes contamination without removing real members.","fun_headline_variants_meta":{"raw":{"variants":["Young RR Lyrae: 23 stars in 1-8 Gyr Magellanic clusters","RR Lyrae not just old: 23 in intermediate-age clusters","Intermediate-age RR Lyrae: 23 in Magellanic clusters","23 RR Lyrae in 1-8 Gyr clusters challenge 10-Gyr age rule"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001503,"raw_usage":{"total_tokens":6192,"prompt_tokens":1269,"completion_tokens":4923,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":885,"completion_tokens_details":{"reasoning_tokens":4836}},"tokens_in":885,"tokens_out":4923,"duration_ms":32332,"temperature":1.0,"reasoning_tokens":4836,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T17:12:14.597196+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure radial velocities of the 23 candidates at about 1 km/s precision: true cluster members must match each cluster's systemic velocity within its few-km/s dispersion, and a systematic offset would overturn the membership claim. A companion calculation would recompute the DTD with clusters whose control-field contaminant counts rival their member counts — NGC 1806 (1.7 members, 1.7 contaminants) in the young bin, and NGC 2121 (3.5 members, 4.0 contaminants) in the intermediate bin — removed; if the young or intermediate signal disappears, the decontamination subtraction is carrying the result.","supporting_citations":[{"cited_title":"R., 1989, @doi [ ] 10.1086/132470 , https://ui.adsabs.harvard.edu/abs/1989PASP..101..570W 101, 570","cited_arxiv_id":null,"evidence_quote":"The earlier targeted search that found no RR Lyrae in SMC intermediate-age clusters and concluded RR Lyrae do not occur younger than about 11 Gyr; this paper's positive result reverses that baseline."}],"review_version":1}