{"id":"0b75f1d1-67f7-4c14-8c20-9c8a42b0232b","arxiv_id":"2411.11953","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"JWST spectroscopy confirms SN 2023aeax as a normal Type Ia supernova at z=2.15, and its standardized distance is consistent with the standard cosmological model.","lead":"Astronomers used JWST to find and confirm a second Type Ia supernova beyond redshift 2, at z=2.15. Its measured distance agrees with the standard cosmological model, so the first two such objects do not yet prove that supernova brightness changes with cosmic time.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The no-evolution conclusion depends on extrapolating a low-z trained BayeSN SED model to rest-frame UV/blue colors and to the edge of its phase coverage; a mock-recovery test is needed before the 0.1σ agreement can be taken as evidence against intrinsic luminosity evolution.","rationale":"The reader's weakest-assumption analysis and mine converge on the same load-bearing point: the BayeSN SED model is being asked to work far outside the bulk of its low-z training data, in rest-frame UV/blue wavelength coverage, at the edge of its phase range, and with a very blue intrinsic color. This is a real correctness risk for the specific distance modulus of 46.10, not merely a philosophical extrapolation worry, because a modest SED mis-specification would shift the inferred mu by an amount comparable to the quoted uncertainty. That said, the paper is appropriately cautious: its main claim is that there is no definitive evidence for evolution, not that evolution is excluded, and it explicitly calls for a larger sample. The good-faith reading is that the object, classification, and photometry are credible and the analysis is transparent about its caveats. The CONDITIONAL verdict therefore remains appropriate: the result is a valuable data point, but its interpretation as a no-evolution constraint should be formally conditional on validating the BayeSN extrapolation through the concrete mock-recovery test described above. I do not see a basis for rejection, and the concern does not move the verdict.","tokens_in":17249,"tokens_out":6293,"duration_ms":67749,"concrete_test":"Refit the Table 1 photometry with the same BayeSN pipeline after omitting the final DDT epoch (MJD 60429) to test sensitivity to the +50 d training boundary. Then run mock-recovery simulations: draw 1000 low-z SNe Ia from the BayeSN training set with B-V near -0.3, place them at z=2.15, and sample them with the exact Table 1 epochs, filters, uncertainties, and upper limits. Fit each mock with the same code and compare the recovered distance modulus to the input value. If the mean recovery bias exceeds about 0.1 mag, or if omitting the last epoch shifts SN 2023aeax's mu by more than 0.15 mag, the 0.1sigma agreement is not robust evidence against intrinsic luminosity evolution.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central 0.1σ agreement rests on a distance modulus that Section 4.1 obtains by applying a phase-extended BayeSN model to a sparse three-epoch light curve (Table 1). The model is trained to rest-frame phases only through +50 d, with linear extrapolation in log-SED beyond that, and the final DDT epoch sits at roughly +50 rest-frame days, right at the training boundary. At z=2.15, the F115W and F150W bands sample rest-frame wavelengths around 0.37 and 0.48 microns, so much of the constraining photometry comes from the blue/UV edge of the model's calibrated range. The fitted B-V=-0.3 is also at the blue edge of the low-z cosmology sample (Figure 7), a region where the training set provides little leverage. If this extrapolation carries an unmodeled color- or phase-dependent bias of even ~0.15 mag, the reported statistical uncertainty on mu (±0.17 mag) understates the true error and the claimed consistency with LCDM becomes uninformative. Section 4.2's dismissal of selection bias is not the principal concern for a single-object distance; the load-bearing step is the SED model extrapolation, and the paper's own statement that this is the only model with the needed coverage underscores that the extrapolation is currently untested.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"SN 2023aeax is a JWST-discovered, spectroscopically confirmed Type Ia supernova at z=2.15, with NIRCam photometry in three epochs and a NIRSpec prism spectrum. The paper classifies it as a normal SN Ia, fits the light curve with the BayeSN SED model (using a phase-extended version with linear extrapolation beyond +50 rest-frame days), and obtains a distance modulus of 46.10+0.17-0.18 mag, which is in ~0.1 sigma agreement with the LambdaCDM prediction of 46.12 mag. The authors conclude there is no definitive evidence for SN Ia luminosity evolution with redshift, while noting that the two-object sample (including SN 2023adsy) shows a ~1 sigma slope relative to LambdaCDM. The paper also re-fits SN 2023adsy with the same methodology for a direct comparison.","tokens_in":17574,"tokens_out":9176,"duration_ms":86649,"significance":"The discovery is timely and important: it is only the second spectroscopically confirmed, non-lensed SN Ia at z>2, and the first that would pass traditional low-redshift cosmology cuts on light-curve properties (though it sits at the blue edge of the color distribution). The data are presented in sufficient detail (Table 1) with a public MAST archive link, and the spectral classification is supported by multiple template comparisons. The paper is appropriately cautious in its conclusions and includes a useful consistency check by re-fitting SN 2023adsy with the same BayeSN model. However, the central distance measurement rests on an extrapolation of a low-redshift-trained SED model into uncalibrated phase and wavelength territory, and the reported uncertainty does not include the systematic risk of that extrapolation.","major_comments":[{"comment":"The phase-extended BayeSN model is the load-bearing element of the distance measurement, yet it is applied at or beyond the edge of its training domain: the model is trained through +50 rest-frame days and extrapolated linearly in log-SED beyond that, and the final DDT epoch (MJD 60429) lands at roughly +50 rest-frame days after the fitted tpk (MJD 60270.34; Table 4). At z=2.15, F115W and F150W sample rest-frame wavelengths near 0.37 and 0.48 microns respectively, in the blue/UV region that is sparsely calibrated by the low-z training set. A prior application to a lensed SN at z=1.78 (Pierel et al. 2024b) is not a substitute for a validation tailored to this object's phase, wavelength, and color coverage, especially since the fitted B-V=-0.3 is at the blue edge of the low-z sample (Figure 7). I request a mock-recovery test: generate simulated z=2.15 SNe with known distance moduli from the BayeSN model (or from an independent SED model) at the same phases/wavelengths, add noise appropriate to Table 1, and re-fit with the same code and priors. The recovered bias and scatter should be quoted as a systematic uncertainty. Without this, the claimed 0.1 sigma agreement is not robust to an extrapolation bias of order 0.15 mag, which would erase the agreement.","section":"Section 4.1"},{"comment":"The distance uncertainty (mu=46.10+0.14-0.12 before systematics) is derived from only three detections plus upper limits, with tpk, AV, theta, and mu fitted jointly. The posterior in Figure 6 shows AV pinned at the boundary (<0.04) and does not present covariances among the parameters. Because the object's color is at the extreme blue edge of the training distribution (B-V=-0.3), the intrinsic-scatter ('epsilon') model is unconstrained for this SN, and the posterior width may be driven by priors rather than by the data. I recommend reporting the full covariance matrix and performing a prior-sensitivity check (e.g., a wider theta prior, or a different AV prior that is not bounded so sharply at zero). The paper should also state how the upper limits are handled in the likelihood; this affects the reported uncertainty. As written, the statistical uncertainty likely understates the true uncertainty, making the 0.1 sigma agreement appear more significant than justified.","section":"Table 4 and Figure 6"}],"minor_comments":[{"comment":"The abstract contains an extra closing brace in \"The James Webb Space Telescope}\"; please correct.","section":"Abstract"},{"comment":"After the JHAT citation, the text has a stray \"2)\" that appears to be a footnote marker error.","section":"Section 2.1"},{"comment":"The name \"SN2023aeax\" and \"SN 2023aeax\" are used inconsistently; please standardize to the latter.","section":"Throughout"},{"comment":"The phrase \"an additional0.055z mag\" is missing a space; please write \"an additional 0.055z mag\" and clarify the lensing uncertainty formula.","section":"Section 4.2"},{"comment":"The sentence \"a larger sample needed\" should read \"a larger sample is needed.\"","section":"Section 5"},{"comment":"The caption \"The Bayesn light curve model parameters\" has inconsistent capitalization; use \"BayeSN\" to match the rest of the text.","section":"Table 4"},{"comment":"The phrase \"χ2 per degree of freedom (ν)\" is redundant; use \"reduced χ2\" for clarity.","section":"Section 3"},{"comment":"The one-sentence dismissal of selection bias addresses detection brightness but not the search selection function; for the population-level interpretation, please expand this discussion even if only qualitatively.","section":"Section 4.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is a valuable observational contribution, and the requested mock-recovery test is feasible within the scope of a revision. I do not see the extrapolation issue as grounds for rejection, but the current version overstates the significance of the distance agreement because the quoted uncertainty excludes the dominant systematic risk. I would accept a revised version that includes the validation or explicitly widens the uncertainty."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a careful, honest measurement of a genuinely new object—the first spectroscopically confirmed normal-color SN Ia at z>2—but the central 0.1σ agreement with ΛCDM should not be read as strong evidence against luminosity evolution. The distance rests on applying BayeSN, a low-z-trained model, at the edge of its phase coverage and beyond its blue color calibration, so the quoted uncertainty is likely too small.\n\nWhat the paper does well: the photometry is careful, combining three JWST programs with proper alignment and PSF-fitting; the classification is convincing, with NGSF template matching strongly favoring a normal SN Ia; and the authors re-fit SN 2023adsy with the same model to put both objects on a consistent footing. They are also transparent about the model's limitations, including the phase extrapolation and the unmodeled selection function. The data are public, which is good practice.\n\nThe soft spot is the load-bearing one. BayeSN in this version is trained to +50 rest-frame days with linear extrapolation in log-SED beyond that, and the last photometric epoch lands right around +50 days. At z=2.15, F115W and F150W sample rest-frame ~0.37 and 0.48 microns, and the fitted B-V=-0.3 is at the blue edge of the cosmology sample (Figure 7). If the model carries a color- or phase-dependent bias of even 0.15 mag—plausible given the training coverage—the distance shifts by a comparable amount to the reported ±0.17 mag uncertainty. That makes the 0.1σ agreement with ΛCDM effectively uninformative about evolution, not a meaningful constraint. The authors acknowledge this in part, but they still present the agreement as a headline result.\n\nThe selection function concern is real but secondary for a single object; a bias correction is not obviously needed when the SN is well above the detection limit, but the population-level claim will require it. The combined ~1σ slope with SN 2023adsy is stated properly as a hint, not a result.\n\nWho is this for: anyone working on high-z SNe Ia or on SED-model extrapolation. It's a solid data paper, not a paradigm changer. The measured distance is a data point that will be used in future samples, but the evolution question is not settled by this object.\n\nRecommendation: send to peer review, but the referee should push for a mock-recovery test of the BayeSN extrapolation at z~2, or at least an explicit quantification of the systematic from the phase/color extrapolation. With that, it's publishable as a measurement paper.","headline":"A useful new z>2 SN Ia point, but the 0.1σ agreement with ΛCDM is softer than it looks once the BayeSN phase/color extrapolation is accounted for.","tokens_in":18305,"tokens_out":2801,"would_cite":true,"duration_ms":27010,"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":"The paper reports that SN 2023aeax, a spectroscopically confirmed Type Ia supernova at $z=2.15$, has a BayeSN-standardized distance modulus of $\\mu=46.10^{+0.17}_{-0.18}$ mag, matching the $\\Lambda$CDM prediction of $46.12$ mag at the…","keywords":["Type Ia supernova","high redshift","JWST","luminosity distance","dark energy","SN Ia standardization","BayeSN","SN 2023aeax"],"falsifier":"Fit SN 2023aeax's photometry with an independent standardization model such as SALT3-NIR, or simulate $z\\simeq2.15$ Type Ia light curves from low-redshift BayeSN training SEDs with known distance moduli and run them through the same phase-extended pipeline; if the recovered distances are biased by more than the quoted $0.1\\text{--}0.2$ mag uncertainty, or if the two models disagree by that amount, the reported $0.1\\sigma$ agreement with $\\Lambda$CDM would be an artifact of model extrapolation rather than evidence against luminosity evolution.","tokens_in":17094,"feed_emoji":"🔭","tokens_out":12656,"duration_ms":102775,"temperature":0.7,"pith_summary":"This paper reports the discovery and JWST follow-up of SN 2023aeax, a Type Ia supernova at $z=2.15$, only the second spectroscopically confirmed, non-lensed SN Ia in the dark-matter-dominated universe at $z>2$. Applying the BayeSN standardization model to a NIRCam multi-band light curve, the authors measure a distance modulus of $\\mu=46.10^{+0.17}_{-0.18}$ mag, which is in $0.1\\sigma$ agreement with the flat $\\Lambda$CDM prediction of $\\mu=46.12$ mag at that redshift. The supernova has a normal decline rate ($\\Delta m_{15}(B)\\sim1.25$) but an unusually blue peak rest-frame color ($B-V\\sim-0.3$), placing it at the edge of traditional low-redshift cosmology cuts. The paper concludes that this object provides no definitive evidence for evolution in Type Ia standardization with redshift, while noting that the two known $z>2$ SNe Ia together yield a $\\sim1\\sigma$ distance slope relative to $\\Lambda$CDM that needs a larger sample to interpret. A sympathetic reader would care because systematic evolution of SN Ia luminosities could masquerade as dark energy, and $z>2$ objects are the cleanest test of that systematics.","feed_headline":"Type Ia supernova at z=2.15 agrees with dark-energy cosmology at 0.1σ","feed_subtitle":"JWST confirms a second z>2 Type Ia supernova; its distance agrees with Lambda-CDM at 0.1 sigma.","key_machinery":"The load-bearing machinery is the phase-extended BayeSN SED model, a Bayesian model trained on low-redshift Type Ia supernovae that jointly fits the light-curve shape parameter $\\theta$, rest-frame host-galaxy dust extinction $A_V$, intrinsic scatter, and the distance modulus $\\mu$ directly from multi-band photometry. The variant used here, trained on the sample of Ward et al. (2023), covers rest-frame phases to about 50 days after peak and extends beyond that by linear extrapolation in log-SED space, which is necessary because SN 2023aeax's observations sample late phases. Supporting machinery includes difference imaging with HOTPANTS, PSF-fitting photometry with space_phot, the NGSF spectral classification code, and the EAZY host SED fit. The NIRSpec spectrum establishes the classification and redshift ($z=2.15\\pm0.01$), and the BayeSN fit then converts the photometry into the standardized distance that is compared with $\\Lambda$CDM.","core_discovery":"On the paper's own terms, the central discovery is that SN 2023aeax is a normal Type Ia supernova at $z=2.15\\pm0.01$ whose standardized luminosity distance agrees with $\\Lambda$CDM: fitting the NIRCam photometry with the phase-extended BayeSN model yields $\\mu=46.10^{+0.17}_{-0.18}$ mag against a predicted $46.12$ mag, a $0.1\\sigma$ difference. The host-subtracted NIRSpec spectrum is best matched by normal low-redshift SN Ia templates, with Si II and Ca II features present and core-collapse templates strongly disfavored. The measured light-curve shape parameter corresponds to $\\Delta m_{15}(B)\\sim1.25$, within the normal population, while the peak rest-frame color $B-V\\sim-0.3$ is bluer than almost all low-redshift cosmological SNe Ia. Re-fitting the other known $z>2$ SN Ia, SN 2023adsy, with the same BayeSN model gives $47.14^{+0.21}_{-0.24}$ mag, and combining both objects with the lower-redshift sample yields a $\\sim1\\sigma$ distance slope relative to $\\Lambda$CDM, consistent with current cosmological constraints and not significant enough to claim luminosity evolution.","pith_inferences":["The opposite extreme colors of the two known $z>2$ SNe Ia (blue for SN 2023aeax, red for SN 2023adsy) hint that the high-redshift population may have a broader intrinsic color distribution; if confirmed, current low-redshift color cuts could exclude a meaningful fraction of high-redshift SNe Ia and bias distance measurements.","The paper's ad hoc treatment of the host-galaxy mass step -- applying half the measured low-redshift step as a systematic error -- may underestimate distance uncertainties if the mass step evolves with redshift; a dedicated high-redshift host-mass calibration would be a natural next step.","The BayeSN phase extrapolation beyond 50 rest-frame days is not directly tested at $z\\sim2$; a rest-frame UV-optical spectrum of a $z>2$ SN Ia at late phases, or a simulation campaign injecting low-redshift SEDs into high-redshift light curves, could confirm or refute the extrapolation's neutrality.","If the $\\sim1\\sigma$ slope persists as JWST collects more objects, it will become a target for distinguishing mild luminosity evolution from new physics, but the current two-object slope cannot do that."],"forward_implications":["The $0.1\\sigma$ agreement between SN 2023aeax and $\\Lambda$CDM means this object alone does not support intrinsic luminosity evolution at $z>2$.","The two spectroscopically confirmed $z>2$ SNe Ia, with opposite extreme colors, produce a $\\sim1\\sigma$ distance slope relative to $\\Lambda$CDM, so a larger sample is required to decide whether the slope is real.","JWST is expected to add roughly ten more spectroscopically confirmed $z>2$ SNe Ia in the next two years, turning the current two-object sample into a population test.","SN 2023aeax passes low-redshift cosmology cuts, so future surveys can include objects like it rather than treating high-redshift SNe Ia as pathological."],"supporting_citations":[{"why":"Supplies the BayeSN SED model that standardizes the light curve and directly infers the distance modulus.","marker":"Mandel et al. 2022"},{"why":"Defines the training set and wavelength coverage of the BayeSN variant used in this analysis.","marker":"Ward et al. 2023"},{"why":"Extends BayeSN to the late phases needed to fit SN 2023aeax's observed light curve.","marker":"Grayling et al. 2024"},{"why":"Establishes the methodology for JWST z>2 SN Ia distance measurements and provides the comparison SN 2023adsy.","marker":"Pierel et al. 2024d"},{"why":"Reports the first spectroscopically confirmed z>2 SN Ia, SN 2023adsy, whose distance is re-fit here with BayeSN.","marker":"DeCoursey et al. 2024"},{"why":"Provides the low-redshift cosmological sample, the host-galaxy mass-step correction, and the comparison cosmology constraints.","marker":"Brout et al. 2022"},{"why":"Describes the COSMOS-Web survey whose NIRCam imaging led to the discovery of SN 2023aeax.","marker":"Casey et al. 2023"},{"why":"Supplies the PRIMER template images used for difference imaging and photometric calibration.","marker":"Dunlop et al. 2021"},{"why":"Provides the PANORAMIC serendipitous first epoch of SN 2023aeax.","marker":"Williams et al. 2021"},{"why":"Supplies the NGSF spectral classification code that identifies SN 2023aeax as a normal Type Ia.","marker":"Goldwasser et al. 2022"}],"fun_headline_variants":["JWST finds no luminosity evolution in z>2 Type Ia supernovae","SN 2023aeax: distance matches ΛCDM at 0.1σ","Second z>2 SN Ia shows no evolution beyond ΛCDM","Normal SN Ia at z=2.15 agrees with dark energy","z=2.15 SN Ia offsets no evolution in luminosity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the BayeSN SED model, trained on low-redshift Type Ia supernovae, can be extrapolated to $z=2.15$ and to phases beyond its 50-day training range without introducing a redshift-dependent bias in the inferred distance modulus.","fun_headline_variants_meta":{"raw":{"variants":["JWST finds no luminosity evolution in z>2 Type Ia supernovae","SN 2023aeax: distance matches ΛCDM at 0.1σ","Second z>2 SN Ia shows no evolution beyond ΛCDM","Normal SN Ia at z=2.15 agrees with dark energy","z=2.15 SN Ia offsets no evolution in luminosity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000788,"raw_usage":{"total_tokens":3595,"prompt_tokens":1185,"completion_tokens":2410,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":801,"completion_tokens_details":{"reasoning_tokens":2313}},"tokens_in":801,"tokens_out":2410,"duration_ms":16851,"temperature":1.0,"reasoning_tokens":2313,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T18:04:08.265872+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fit SN 2023aeax's photometry with an independent standardization model such as SALT3-NIR, or simulate $z\\simeq2.15$ Type Ia light curves from low-redshift BayeSN training SEDs with known distance moduli and run them through the same phase-extended pipeline; if the recovered distances are biased by more than the quoted $0.1\\text{--}0.2$ mag uncertainty, or if the two models disagree by that amount, the reported $0.1\\sigma$ agreement with $\\Lambda$CDM would be an artifact of model extrapolation rather than evidence against luminosity evolution.","supporting_citations":[{"cited_title":"C., Oesch, P., Barrufet, L., et al","cited_arxiv_id":null,"evidence_quote":"Provides the PANORAMIC serendipitous first epoch of SN 2023aeax."}],"review_version":1}