{"id":"3e317c10-50ff-43be-9f32-2a0988bce1af","arxiv_id":"2411.17358","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A homogeneous reanalysis of 26 rocky exoplanets shows a negative correlation between planet iron fraction and host star age, with denser planets around younger stars.","lead":"Rocky exoplanets around younger stars appear denser and more iron-rich than those around older stars. This suggests that planet composition depends on when in the Milky Way's history the system formed, not just on stellar abundance measurements.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"RV mass-measurement selection function, acknowledged in §4.4.2, is not quantified; a density-age bias in sample inclusion could manufacture the observed r=-0.62 correlation.","rationale":"The paper's central claim is a demographic correlation between rocky-planet iron fraction and host-star age. For that claim to hold, the sample must not be biased in the same direction. The paper itself identifies the most plausible bias in §4.4.2: younger, more active stars have noisier RV measurements, which can preferentially remove lower-density planets from the young-age subsample because lower density at a given radius implies lower mass and lower RV semi-amplitude. This is precisely the direction needed to create the observed negative correlation. The paper's rebuttal is qualitative: it checks that planet mass does not strongly correlate with age and argues that selection cannot explain the absence of high-density planets around old stars. Neither argument substitutes for a completeness model. Since the reader's weakest_assumption is the same RV selection effect and the reader already assigns CONDITIONAL, my stress test does not change the verdict. The requested forward-model test would settle whether the selection effect can explain the full slope; until then, the conditional verdict is appropriate.","tokens_in":24822,"tokens_out":5270,"duration_ms":64574,"concrete_test":"Build a forward model of the selection: draw a synthetic population of transiting rocky planets with an intrinsic mass-radius distribution and no density-age correlation; assign ages and activity jitter from the observed host-star sample (using an empirical R'HK-age relation); apply a transit SNR threshold and an RV detectability criterion K > n*(jitter+instrumental noise) with realistic observing strategies; apply the paper's radius-valley cut; recover masses/radii; infer %Fe exactly as in §4.1; run linmix. If the recovered α is within uncertainties of -8.0(+4.6,-5.0)%Fe/Gyr, the selection concern lands; if it is consistent with zero, the claim is robust to this bias. A complementary check is to repeat the analysis on a uniformly RV-observed sample (e.g., TESS follow-up programs with homogeneous K upper limits) and see whether the density-age correlation persists.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim rests on the assumption that the sample of 26 transiting rocky planets with RV masses is representative across stellar age. This is insecure: §4.4.2 concedes \"there may be a decreasing sensitivity to lower density planets towards younger ages\" because younger stars are more active, yet no completeness function, injection-recovery simulation, or selection-corrected analysis is provided. The mechanism is concrete: for a fixed transit-detected radius, planet mass and hence RV semi-amplitude K scale as K ∝ M_p^(2/3) P^(-1/3) M_*^(-2/3) (Eq. 7). Young active hosts add jitter, raising the minimum detectable K and preferentially excluding low-mass, low-density planets at small radii. A plot of planet mass versus age (§4.4.2, top-right of Fig. 12) is not a sufficient null test because density, not mass alone, is the selected quantity, and radii vary. Without quantifying this selection, the observed Pearson r=-0.62 between %Fe and age could be produced entirely by removing low-%Fe planets from the young-age end. The paper's argument that selection cannot explain the absence of high-density planets around old stars is helpful but does not remove the need to show the recovered slope survives a full forward model.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a homogeneous re-analysis of 26 transiting rocky exoplanets (R_p < 4 R_Earth, below the radius valley) and their host stars, using Gaia DR3 GSP-Spec parameters, parallaxes, and photometry as inputs to the BASTA stellar-modelling code. It recalculates planet radii and masses from the updated stellar parameters, infers iron mass fractions from the Zeng et al. (2019) interior grid, and reports a statistically significant negative correlation between inferred planet iron fraction and stellar age (Pearson r = -0.62, Spearman rho = -0.63, Bayesian linear-regression slope alpha = -8.0^{+4.6}_{-5.0} percent Fe per Gyr). The authors interpret the trend as evidence that rocky planets formed around younger, more iron-rich stars are denser and more iron-rich, linking exoplanet composition to Galactic chemical evolution.","tokens_in":25046,"tokens_out":11157,"duration_ms":99651,"significance":"If the trend is real, it is an important new demographic result: it would be the first homogeneous, sample-level evidence that rocky exoplanet composition depends on host-star age, connecting exoplanet interiors to Galactic chemical evolution and with consequences for the interpretation of planet formation and habitability. The paper's strengths are the careful homogenisation of stellar parameters, the multi-method age validation in Appendix A2 (asteroseismology, gyrochronology, kinematics, and chemical abundances), the use of established external interior and radius-valley models, and the explicit discussion of observational biases. The correlation is measured rather than derived, so circularity is not a concern. However, the sample is small (26 planets), the selection-function caveat in Section 4.4.2 is not quantified, and the statistical evidence is not reported in full; the result is therefore promising but not yet established at the strength the paper claims.","major_comments":[{"comment":"The central claim requires the sample to be representative in planet density across stellar age, but the acknowledged 'decreasing sensitivity to lower density planets towards younger ages' is not quantified. The mass-versus-age panel in Fig. 12 is not a sufficient null test: because the radial-velocity semi-amplitude scales as K ∝ M_p^(2/3) P^(-1/3) M_*^(-2/3) (Eq. 7), a selection in detectable K at fixed transit radius translates into a selection in density, not in mass alone. The authors should provide an RV completeness model (e.g., injection-recovery into the actual K uncertainties and activity levels) or a forward model that injects the proposed age-density relation and demonstrates that the observed r = -0.62 is recovered. Without this, the trend could be produced entirely by removing low-%Fe planets from the young-age end of the sample.","section":"Section 4.4.2, Fig. 12"},{"comment":"The assignment of %Fe = 0 to every planet whose best-fit mass and radius fall above the pure-rock track is a censoring step whose effect on the regression is not reported. The authors should state how many of the 26 planets receive this boundary value, show the distribution of the pre-truncation interpolated values, and rerun the linmix and ODR fits treating those measurements as upper limits (e.g., with a censored regression or a model that allows negative inferred Fe). This is needed to demonstrate that the slope and correlation are not artifacts of piling points up at the %Fe = 0 boundary.","section":"Section 4.1, Fig. 5"},{"comment":"The statement that the correlation is 'highly significant' is not backed by a reported p-value or a permutation/bootstrap test. With N = 26 and typical age uncertainties of several Gyr, the linmix posterior alone is not a substitute for a robustness analysis. The authors should report p-values for the Pearson and Spearman coefficients, perform leave-one-out or bootstrap resampling, and show that the slope is not driven by the cluster of high-%Fe points at young ages in the upper-left of Fig. 5.","section":"Section 4.2"}],"minor_comments":[{"comment":"The text near Fig. 3 says 'we plot the stellar density as a function of age'; this should read 'planet density'.","section":"Section 4.1"},{"comment":"The weighting term x_Θ in the marginalized posterior is not defined; a brief definition would make the Bayesian computation reproducible.","section":"Section 2.3, Eq. (4)"},{"comment":"The causal interpretation in terms of Galactic chemical evolution is plausible but is not directly tested (e.g., no Mg/Si or alpha-abundance measurements are used for the planet hosts); consider softening the causal language in the abstract and conclusions.","section":"Section 5 and Abstract"},{"comment":"The paper states that all data are in Tables 1-3, but it does not provide the posterior draws for the inferred %Fe values or the BASTA age posteriors; making these available as machine-readable files would improve reproducibility.","section":"Data availability"}],"recommendation":"major_revision","confidential_remarks":"The main obstacle is the unquantified RV selection effect in Section 4.4.2; if the authors cannot supply a completeness or forward model, the paper should be reframed as reporting a candidate trend with a clearly stated caveat rather than a confirmed relation. The manuscript is otherwise within the journal's scope, and I saw no evidence of prior publication or novelty-disclosure problems."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a read: the first homogeneous look at rocky exoplanet composition versus host star age, and the stellar ages are genuinely carefully derived. The raw density-age trend is visible before any interior modeling, which is a good sign; the inferred %Fe correlation just puts a number on it. The age validation is solid—asteroseismology, gyrochronology, kinematics, chemical abundances, and the LEGACY sample all behave. Credit where due: the authors name the main threat themselves and give a physically sensible galactic chemical evolution interpretation without overclaiming.\n\nThe soft spot is the one they name in §4.4.2: there may be decreasing sensitivity to lower-density planets toward younger ages due to stellar activity, yet there is no completeness model, injection-recovery simulation, or selection-corrected fit. The stress-test mechanism is concrete: K scales as Mp^(2/3) P^(-1/3) M*^(-2/3), so young active hosts preferentially exclude low-mass, low-density planets. The top-right panel of Fig. 12 is not a sufficient null test because density, not mass alone, is the selected quantity and radii vary. The authors' counterpoint—selection cannot explain the absence of high-density planets around old stars—has force, but it does not remove the need to show the recovered slope survives a forward model. If low-%Fe planets are missing from the young-age end, the observed r = -0.62 could be manufactured.\n\nThe small sample (26 planets) and the model-dependent %Fe assignment via Zeng et al. (2019) are secondary; the trend in raw density mitigates the model dependence, and the sample size just calls for caution, not dismissal.\n\nThis deserves a serious referee. The claim is new, plausible, and falsifiable with more data and a proper selection model. A referee should push for the quantified completeness analysis rather than reject on the basis of the acknowledged bias.","headline":"Plausible new age-composition correlation, but the unquantified RV selection function is the gap that needs closing before this is solid.","tokens_in":25630,"tokens_out":1949,"would_cite":true,"duration_ms":18753,"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":"Denser, more iron-rich rocky planets are found around younger stars.","keywords":["rocky exoplanets","exoplanet composition","stellar age","galactic chemical evolution","iron mass fraction","radius valley","homogeneous stellar characterisation","mass-radius relation"],"falsifier":"An injection-recovery test that adds synthetic low-density planets to the radial-velocity data of young, active stars and measures how detection completeness varies with stellar age would settle the matter: if completeness declines steeply toward younger ages and the trend disappears after correction, the central claim is an artifact. A larger sample with asteroseismic ages that shows a flat density–age relation would also refute it.","tokens_in":24636,"feed_emoji":"🪐","tokens_out":12381,"duration_ms":103635,"temperature":0.7,"pith_summary":"The paper argues that the interior composition of rocky exoplanets is tied to the age of their host stars. After homogenising stellar parameters for a sample of 26 small planets below the radius valley, it finds that denser planets with higher inferred iron content orbit younger stars, while older stars host less dense, less iron-rich rocky planets. Quantitatively, the iron mass fraction and stellar age show a Pearson correlation of $-0.62$ and a Bayesian regression slope of about $-8$ percent iron per gigayear. The authors interpret the trend as a fingerprint of galactic chemical evolution: the material available for planet formation has grown more iron-rich over time, so planets forming today may differ from those that formed billions of years ago, including Earth.","feed_headline":"Denser rocky planets orbit younger stars","feed_subtitle":"A homogeneous survey of 26 super-Earths finds denser, more iron-rich planets around younger stars.","key_machinery":"The argument is carried by a homogeneous re-characterisation of host stars: a grid-based Bayesian stellar modelling code takes space-based astrometry, photometry, and spectroscopy as inputs and returns stellar masses, radii, and ages with relative precision of roughly 2 Gyr. Planet radii and masses are recomputed from transit depths and radial-velocity semi-amplitudes using these stellar values, and only planets below the radius valley, the observed gap near 1.8 Earth radii that separates rocky super-Earths from sub-Neptunes, are kept. Each planet's iron mass fraction is obtained by linearly interpolating its mass-radius position on a grid of rocky interior models. The statistical quantity that carries the discovery is the negative correlation between inferred iron fraction and age, with a Pearson coefficient of $-0.62$ and a Bayesian regression slope of $\\alpha=-8.0^{+4.6}_{-5.0}$ percent iron per gigayear.","core_discovery":"The central claim is that rocky planet composition correlates with host-star age: planets that are denser and inferred to be richer in iron orbit younger stars. The paper reports correlation coefficients of $-0.62$ (Pearson) and $-0.63$ (Spearman), and a Bayesian linear-regression slope of $\\alpha=-8.0^{+4.6}_{-5.0}$ percent iron per gigayear, with the trend confirmed by an orthogonal distance regression. It interprets this as the first observational link between rocky exoplanet composition and stellar age, driven by galactic chemical evolution: younger stars are more iron-rich and less $\\alpha$-enhanced, and the material from which their planets formed follows the same enrichment. The paper argues the trend is not simply an effect of stellar mass or metallicity, because the correlation with age is stronger and the age-metallicity relation is flat until about 10 Gyr.","pith_inferences":["If the trend survives a completeness correction, the average density of rocky planets in a survey volume could be used as an independent probe of the local star-formation and chemical-enrichment history.","A direct test of the interpretation would be to compare the compositions of debris accreted onto old and young white dwarfs: the cooling age of the white dwarf should correlate with the iron content of the accreted rocky material.","Extending the analysis to M dwarfs and pre-main-sequence stars with gyrochronological ages would show whether the relation continues below about 2 Gyr or flattens.","The paper itself notes a possible selection effect: young, active stars may hide low-density planets in radial-velocity mass measurements, and an injection-recovery study quantifying that completeness is the most direct way to separate the physical trend from an observational one."],"forward_implications":["Rocky exoplanet composition becomes a clock: denser, more iron-rich planets form in more recent epochs of galactic chemical evolution.","Planets forming today in the solar neighbourhood may have larger iron cores, higher surface gravity, and different internal heat and magnetic-dynamo behaviour than Earth, which formed several gigayears ago.","Old stars, above about 8 Gyr, may be unable to produce rocky planets with more than roughly 50 percent iron, narrowing the range of possible compositions around ancient systems.","The absence of a strong planet-composition–metallicity correlation is consistent with the trend being driven by alpha-element abundances and age rather than by [Fe/H] alone.","Future samples with precise asteroseismic ages, such as those expected from PLATO, should recover the same age–density relation and refine its slope."],"supporting_citations":[{"why":"It provides the Bayesian stellar modelling code used to derive homogeneous stellar ages from the adopted input parameters.","marker":"Aguirre Børsen-Koch et al. (2022)"},{"why":"It defines the radius-valley location as a function of orbital period and stellar mass, which selects the rocky planet sample.","marker":"Ho & Van Eylen (2023)"},{"why":"It supplies the interior-model grid from which each planet's iron mass fraction is interpolated.","marker":"Zeng et al. (2019)"},{"why":"It provides the Bayesian linear regression used to quantify the age–iron fraction slope and correlation coefficient.","marker":"Kelly (2007)"},{"why":"It justifies the effective-temperature limit for reliable model ages and offers demographic context for exoplanet trends with age.","marker":"Berger et al. (2020)"},{"why":"It supplies asteroseismic ages used to validate the homogeneous age determinations.","marker":"Silva Aguirre et al. (2015)"},{"why":"It provides the radial-velocity semi-amplitudes for most of the sample, from which planet masses are recomputed.","marker":"Bonomo et al. (2023)"},{"why":"It connects stellar iron-mass-fraction to planet iron fraction and density, and its age trend independently validates the stellar ages.","marker":"Adibekyan et al. (2021)"},{"why":"It provides the observed relation between alpha-element abundances and stellar age that grounds the galactic chemical evolution interpretation.","marker":"Ciucă et al. (2021)"},{"why":"It supplies the GSP-Spec effective temperatures, metallicities, and calibrations used as inputs to the stellar modelling.","marker":"Recio-Blanco et al. (2023)"}],"fun_headline_variants":["Rocky planet density tied to host star age","Younger stars host denser rocky planets","Iron-rich rocky worlds orbit young stars","Stellar age predicts rocky planet composition","Galactic chemical evolution shapes rocky planet makeup"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the observed density–age trend is not created by an age-dependent selection effect in the radial-velocity mass measurements; the paper acknowledges that younger, more active stars may be less sensitive to low-density planets, but provides no completeness model to rule this out.","fun_headline_variants_meta":{"raw":{"variants":["Rocky planet density tied to host star age","Younger stars host denser rocky planets","Iron-rich rocky worlds orbit young stars","Stellar age predicts rocky planet composition","Galactic chemical evolution shapes rocky planet makeup"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001255,"raw_usage":{"total_tokens":5147,"prompt_tokens":952,"completion_tokens":4195,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":568,"completion_tokens_details":{"reasoning_tokens":4130}},"tokens_in":568,"tokens_out":4195,"duration_ms":30108,"temperature":1.0,"reasoning_tokens":4130,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:12:17.514798+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An injection-recovery test that adds synthetic low-density planets to the radial-velocity data of young, active stars and measures how detection completeness varies with stellar age would settle the matter: if completeness declines steeply toward younger ages and the trend disappears after correction, the central claim is an artifact. A larger sample with asteroseismic ages that shows a flat density–age relation would also refute it.","supporting_citations":[],"review_version":1}