{"id":"aae675ff-4c04-4c80-a3cf-3784bcfca7a6","arxiv_id":"2411.16476","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"R-matrix atomic data support W III as the source of the 4.5 μm kilonova emission, yielding W III masses near 1.7e-4 solar masses for AT2017gfo and 9.4e-4 solar masses for AT2023vfi.","lead":"This paper uses new atomic collision data for tungsten, platinum and gold to predict the infrared lines these elements would emit in the late nebular phase of a neutron-star merger. The authors estimate the tungsten mass ejected in the kilonovae AT2017gfo and AT2023vfi from the observed 4.5 μm emission and compare it with nucleosynthesis models.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Ionisation balance and 4.5 μm band-fraction assumptions make the W III masses upper limits, so the comparison with total-W nucleosynthesis yields is not yet a closed test.","rationale":"The reader's weakest-assumption analysis correctly identifies the load-bearing premise: the 4.5 μm flux assigned to W III must be dominated by W III, and W III must be the dominant tungsten ionisation stage. My reading of the manuscript supports this. The collisional-radiative machinery in Sections 2–3 is internally consistent, the R-matrix data are published elsewhere, and the luminosity tables are useful; the paper also honestly maps the Te–ne degeneracy in Figure 6 and states the band-fraction limitation. However, none of that removes the dependence on external assumptions about the observed feature and the ionisation balance. The comparison with total W from merger simulations in Table 4 is the point where the argument becomes most fragile: the inferred quantity is a single ion-stage mass under a fixed band-fraction prior, while the theoretical quantity is a total elemental mass. The paper's own caveats ('Without a quantitative treatment of ionization fraction, an elemental mass is difficult to quantify'; 'We assume Wiii to be the dominant ionization state') are explicit admissions of the gap. This does not require a change of verdict: CONDITIONAL is already the appropriate classification. The proposed test—self-consistent nebular modelling with ionisation/recombination and competing species—would settle whether the mass estimates are point estimates or upper limits, and whether the nucleosynthesis comparison carries the weight the abstract gives it.","tokens_in":30878,"tokens_out":5252,"duration_ms":52666,"concrete_test":"Run a self-consistent nebular model including ionisation/recombination and both W III and Se III (plus other 4.5 μm candidates) for AT2017gfo and AT2023vfi, fitting the full 3–5.5 μm spectra. Report the best-fit W III mass and the W III/total-W fraction. If the fraction deviates from unity by more than a factor of two, or if the best-fit W III mass shifts by more than ~30% when Se III is included, the quoted masses should be treated as upper limits and the Table 4 comparison downgraded. A cheaper ancillary check: compute the predicted 6.05 μm W I+II luminosity for the quoted W III masses and compare with the deepest available 5–7 μm upper limits; absence of the predicted feature would directly weaken the ionisation-balance and blending assumptions.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The mass estimates in Section 3.2 (1.65e-4 M_sun for AT2017gfo, 9.4e-4 M_sun for AT2023vfi) use equation (3) to convert an adopted line luminosity into a W III mass. That conversion is sound only if (a) the adopted luminosity is actually emitted by W III rather than by Se III or other species, and (b) the inferred W III mass can be compared with the total W masses from Table 4. Condition (a) is not established: for AT2017gfo the paper adopts 5e37 erg/s as an upper-bound W III contribution to a 2e38 erg/s Spitzer band, without a quantitative Se III or continuum-subtraction model. Condition (b) fails unless W III is the dominant tungsten ionisation stage, but the collisional-radiative model deliberately omits ionisation and recombination, so it cannot determine the ionisation balance. The paper later asserts 'We assume Wiii to be the dominant ionization state' (Section 4) and elsewhere assumes 'reasonably similar fractions for the near neutral stages of W'. If W III is not dominant, the derived mass underestimates total W; if part of the 4.5 μm feature is Se III, it overestimates W III. The manuscript honestly flags the band-fraction ambiguity, but the ionisation-dominance assumption is treated as a prior rather than a derived quantity. The abstract's 'broad agreement' with nucleosynthesis therefore rests on an unquantified assumption, and the Table 4 comparison is not apples-to-apples.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs steady-state collisional-radiative (CR) models for the first three ionization stages of W, Pt and Au using published Dirac R-matrix effective collision strengths (Smyth et al. 2018; Dunleavy et al. 2022; McCann et al. 2024; Bromley et al. 2023; McCann et al. 2022) and applies them to late-time kilonova conditions (Te = 0.15–0.35 eV, ne = 10^5–10^7 cm^-3). It presents photon emissivity coefficients, line luminosities for the ten strongest lines of each ion (Tables 1–3), and synthetic spectra (Figures 2–5). The W III fine-structure lines at 4432.23 and 4535.16 nm are identified as the strongest 4.5 μm features. Imposing the observed 4.5 μm luminosities, the authors invert equation (3) to obtain W III masses of 1.65e-4 M_sun for AT2017gfo and 9.4e-4 M_sun for AT2023vfi. These are compared with total W masses from neutron-star merger nucleosynthesis models (4e-5 to 2e-4 M_sun, Table 4); the body reports broad consistency mainly for AT2017gfo, with the AT2023vfi estimate exceeding models by up to an order of magnitude. The paper also converts the inferred W fraction into lanthanide/actinide and third-peak mass fractions using model correlations, and uses merger models to predict W line-profile shapes that may distinguish short- from long-lived remnants.","tokens_in":31054,"tokens_out":22659,"duration_ms":183628,"significance":"The atomic-physics core is solid: the mass estimates follow from equation (3) with no parameter fitted to the target observations, and the full effective-collision-strength tables in Appendix A are a valuable community resource. The paper is also honest about its input assumptions—the limiting role of the Spitzer band fraction is stated in Section 3.2, the Te–ne–mass degeneracy is mapped openly in Figure 6, and Section 4 notes that the comparison with models could be an overestimate or an upper limit. If the W III identification holds, this is a genuinely new third-peak r-process diagnostic for kilonovae and a potentially powerful proxy for neutron-rich ejecta, with falsifiable line-profile predictions (Section 5). The significance is currently capped by the two load-bearing assumptions the review process flagged: the fraction of the 4.5 μm band assigned to W III is taken as an upper bound without a quantitative Se III model, and the W III-to-total-W ionization correction is assumed rather than derived. Both are acknowledged in the text but not bounded, so the headline 'broad agreement' claim runs ahead of what is demonstrated.","major_comments":[{"comment":"The inferred masses are W III ion masses, but the comparison in Table 4 and the mass-fraction statements (0.33%, 1.6%, and the lanthanide/third-peak conversions of Figure 10) are phrased in terms of total elemental W. The CR model deliberately omits ionization and recombination (Section 2), so it cannot determine whether W III is the dominant tungsten stage; the paper instead states the assumption directly ('We assume Wiii to be the dominant ionization state', Section 4; 'Assuming reasonably similar fractions for the near neutral stages of W', Section 3.2). Because the total-W comparison and the X(W) ≈ 0.3% central value are linearly sensitive to this ionization-correction factor, the agreement claim is conditional on an unquantified quantity. I recommend either adding a quantitative ionization-balance estimate (e.g., Saha at the adopted Te/ne with published ionization potentials) or explicitly reframing the Table 4 comparison as a bounded W III-to-total-W statement rather than a point comparison.","section":"Sections 2, 3.2, and 4"},{"comment":"For AT2017gfo the input luminosity is 5e37 erg/s, quoted from Hotokezaka et al. (2022) as the maximum W III contribution to a 2e38 erg/s Spitzer 4.5 μm band, and the mass estimate scales linearly with this number. The paper candidly notes that the band-fraction assignment is 'the limiting factor,' but in Section 4 and the abstract the resulting 1.65e-4 M_sun is treated as a point mass whose proximity to the upper end of the model range (4e-5 to 2e-4 M_sun, Table 4) is described as 'particularly good' agreement. Without a quantitative model for the Se III or continuum fraction of the band, the AT2017gfo comparison should be framed as an upper-envelope consistency test rather than a two-sided match, or else a quantitative bound on the Se III contribution should be supplied.","section":"Section 3.2"},{"comment":"The quoted masses assume single (Te, ne) points (3500 K, 1e6 cm^-3 for AT2017gfo; 3000 K, 3e5 cm^-3 for AT2023vfi), while Figure 6 shows the required mass varying by close to three orders of magnitude across the explored grid—for example ~2.1e-1 M_sun at 500 K versus ~5.6e-4 M_sun at 0.86 eV for L = 1e38 erg/s at fixed density. The paper deserves credit for displaying this degeneracy, but the abstract and conclusions do not carry it: the 'broad agreement' statement is made without attaching any systematic range to the headline masses. The central claims should propagate the Figure 6 range explicitly, and the agreement with nucleosynthesis should be reported against the allowed grid rather than at single points.","section":"Section 3.2 and Figure 6"},{"comment":"The abstract's 'broad agreement with the inferred ion masses of W' is stronger than what the body itself supports: Section 4 states that for AT2023vfi the inferred 9.4e-4 M_sun exceeds the simulation predictions by 'up to one order of magnitude,' and for AT2017gfo most models give W masses a factor of 2–3 below the estimate, with the spread of models merely encompassing it. The conclusions (Section 6) repeat 'broad consistency' while also reporting the factor-2–3 offset. The abstract should carry the same qualification as the body: good agreement for AT2017gfo at the upper end of the model range, with a genuine tension for AT2023vfi that the paper itself discusses.","section":"Abstract and Section 4"}],"minor_comments":[{"comment":"As printed, equation (3) multiplies by Σ_i N_i, but with N_j defined relative to the ground level (Section 2) the renormalization to total ion population requires division by Σ_i N_i (or equivalently sum-normalized populations with Σ_i N_i = 1). The numerical entries of Table 1 are consistent with the division form, so the printed equation should be corrected or clarified.","section":"Equation (3)"},{"comment":"The text assigns the W I lines at 6041.42 nm and 6646.86 nm to 5D1–5D3 and 5D3–5D4, while Table 1 lists them as 5D1–5D2 and 5D2–5D3; the text should match the table.","section":"Section 3.1"},{"comment":"The 'ground configuration' of W III is written as (5d4 6s2), but the W III ground configuration is 5d4; the 6s2 belongs to the W I ground configuration. The transitions discussed are the 5d4 fine-structure lines, so the configuration label is a typo.","section":"Section 3.2"},{"comment":"The statement 'neutron-rich material should roughly amount to about 40 times (14 times) the inferred W mass for HFB21 (DZ31), i.e. to 6.6e-3 M_sun (2.3e-3 M_sun) for AT2017gfo and to 2.8e-2 M_sun (9.8e-3 M_sun) for AT2023vfi' is internally inconsistent for AT2023vfi: 40 × 9.4e-4 = 3.76e-2 and 14 × 9.4e-4 = 1.32e-2, neither matching the printed values; please reconcile the numbers.","section":"Section 4"},{"comment":"The converted elemental ranges differ between the two sections: Section 4 gives 1.8% ≲ X_La+X_Ac ≲ 5.5% and 1.6% ≲ X(3rd peak) ≲ 9.9%, while Section 6 gives 0.85%–11% and 0.73%–19%. The factor-of-two uncertainty propagation needs to be defined once and applied consistently.","section":"Sections 4 and 6"},{"comment":"Minor presentation items: 'colradpypackage' (Section 2) and 'spectropscopic' (Section 5) need fixing; the references for Kasliwal et al. (2019) and Kasliwal et al. (2022) carry identical volume/page data (510, L7) and should be checked; and the introduction's outline omits Section 5.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope and the atomic-data basis is genuine; my main concern is message discipline. The body of the paper is admirably candid about the band-fraction and ionization-balance limitations, but the abstract and conclusions outrun those caveats, and the two load-bearing assumptions are stated without quantitative bounds. These are fixable in revision by reframing, propagating the Figure 6 range, and adding a simple ionization-balance estimate; I do not see grounds for rejection. Also worth checking editorially: the Kasliwal 2019/2022 duplicate reference entry."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper does something genuinely useful: it takes the new R-matrix collision data for W, Pt, and Au and produces luminosity tables for nine ion stages under kilonova nebular conditions. The tabulated effective collision strengths in the appendix are a reproducible contribution on their own. The revised W III mass for AT2017gfo (1.65e-4 M_sun) and the first estimate for AT2023vfi (9.4e-4 M_sun) are the headline results, and the correlation-based conversion from W mass to lanthanide and third-peak fractions is a clever way to extract more from a single measurement.\n\nThe paper is also refreshingly honest about its own limits. It explicitly says it omits ionization and recombination, it maps the Te-ne degeneracy in Figure 6 rather than hiding it, and it flags the 'what fraction of the Spitzer band' problem as the limiting factor. The body of the paper is appropriately cautious about AT2023vfi, noting the estimate overshoots the models by up to an order of magnitude and that the discrepancy could be the approach or the event. That is the right posture.\n\nThe soft spots are real but they are exactly the ones the paper names. The mass inference requires that the adopted 5e37 erg/s is genuinely W III emission and not Se III or a continuum artifact, and that W III is the dominant tungsten ionization stage. Neither is derived; the first is an assumption from Hotokezaka et al., the second is stated as 'we assume Wiii to be the dominant ionization state.' If either fails, the mass becomes an upper limit on W III, and the comparison with total W yields from merger models in Table 4 is not apples-to-apples. The abstract's 'broad agreement' is a bit stronger than the body supports, since the AT2023vfi point is up to an order of magnitude off. That is a wording issue, not a fatal flaw.\n\nWho gets value from this? Atomic physicists building heavy-element collisional-radiative models and anyone interpreting the 4.5 μm kilonova feature. The data are reproducible and the method is transparent. It deserves a serious referee; the referee should push for a quantitative treatment of the band fraction and ionization balance, but the paper is a solid contribution that belongs in the literature.","headline":"A solid, honest atomic-data paper whose headline W III mass estimates rest on two clearly stated assumptions, so the masses are conditionally useful upper limits rather than closed measurements.","tokens_in":647,"tokens_out":1173,"would_cite":true,"duration_ms":28223,"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":"Doubly ionised tungsten (W III) is identified as the source of the 4.5-micron emission in two kilonovae, and its measured luminosity is used to infer how much tungsten the explosions ejected.","keywords":["kilonova","r-process nucleosynthesis","tungsten","collisional-radiative modelling","R-matrix atomic data","AT2017gfo","AT2023vfi","mid-infrared spectroscopy"],"falsifier":"Look for the predicted 6.05 $\\mu$m W I/W II blend in the same late-time spectra: the model predicts it should be comparable in luminosity to the 4.5 $\\mu$m feature when W I or W II masses are similar to W III, so its absence would contradict the W III identification; alternatively, a high-resolution mid-infrared spectrum that fails to show the two W III components at their predicted ~6:1 intensity ratio would falsify the claim.","tokens_in":30478,"feed_emoji":"⚛️","tokens_out":6240,"duration_ms":50989,"temperature":0.7,"pith_summary":"The paper argues that the unexplained 4.5-micron emission seen in two kilonovae, AT2017gfo and AT2023vfi, is produced by doubly ionised tungsten (W III), and uses this identification to measure how much tungsten the explosions ejected. It builds collisional-radiative models from new R-matrix electron-impact excitation data for the first three ionisation stages of tungsten, platinum and gold, then computes optically thin line luminosities at kilonova nebular conditions. For tungsten, enforcing the observed 4.5-micron luminosity yields W III masses of about $1.65 \\times 10^{-4}\\,M_\\odot$ for AT2017gfo and $9.4 \\times 10^{-4}\\,M_\\odot$ for AT2023vfi. These estimates are broadly consistent with tungsten yields from neutron-star merger nucleosynthesis models, especially for AT2017gfo, and the paper shows that a measured tungsten mass can be converted, via model correlations, into abundance constraints for lanthanides, actinides and third r-process peak elements. The result matters because it would make tungsten a practical observational tracer of very neutron-rich ejecta in kilonovae.","feed_headline":"Tungsten W III identified in kilonova 4.5-micron glow","feed_subtitle":"New atomic data give W III masses for AT2017gfo and AT2023vfi, tying the infrared line to neutron-rich r-process ejecta.","key_machinery":"The engine is a steady-state collisional-radiative model solved with the colradpy package, driven by R-matrix effective collision strengths and Einstein A-coefficients for W I-III, Pt I-III and Au I-III. Line emission is expressed through photon emissivity coefficients (PEC, the upper-level population times the A-coefficient divided by electron density), and the luminosity formula $L = (hc/\\lambda)\\, n_e\\, \\mathrm{PEC}\\, (M_{\\mathrm{ion}}/m_{\\mathrm{ion}}) / \\sum_i N_i$ lets the authors invert an observed line luminosity into an ion mass once temperature and density are fixed. The two W III lines at 4432.23 and 4535.16 nm are the strongest computed features at kilonova nebular conditions, which is what carries the identification of the 4.5 $\\mu$m feature.","core_discovery":"On its own terms, the paper claims that the 4.5 $\\mu$m excess in AT2017gfo and AT2023vfi can be attributed to the two strongest fine-structure transitions of W III, at 4432.23 nm and 4535.16 nm within the $5d^4\\,^5D$ ground term, and that inverting the observed luminosity with a full collisional-radiative model gives W III masses of about $1.65 \\times 10^{-4}\\,M_\\odot$ for AT2017gfo (adopting $5 \\times 10^{37}\\,\\mathrm{erg\\,s^{-1}}$ from the Spitzer band) and $9.4 \\times 10^{-4}\\,M_\\odot$ for AT2023vfi (adopting $1.0 \\times 10^{38}\\,\\mathrm{erg\\,s^{-1}}$). With assumed total ejecta masses this corresponds to roughly 0.33% and 1.6% of the ejecta as W III, and up to about 1.0% and 4.7% as tungsten if neighbouring ion stages contribute similarly. Compared with hydrodynamical merger models and nuclear network yields, the AT2017gfo estimate sits near the top of the predicted range $4 \\times 10^{-5}$ to $2 \\times 10^{-4}\\,M_\\odot$, while the AT2023vfi estimate exceeds the models by up to an order of magnitude; the paper interprets this as either an observational overestimate from blending or evidence that AT2023vfi ejected more neutron-rich material, possibly from a neutron star-black hole merger.","pith_inferences":["If the W III identification survives, the 4.5 $\\mu$m band becomes a direct probe of the mass of ejecta with electron fraction below about 0.2, effectively a calorimeter for the most neutron-rich component of a kilonova; this extends the paper's correlation argument.","A clean test is to look for the predicted 6.05 $\\mu$m W I/W II blend at comparable luminosity to the 4.5 $\\mu$m feature in the same objects: detection would strengthen the identification, while non-detection would challenge the assumed ionisation balance.","The paper's mass estimates scale almost inversely with the fraction of the observed band assigned to W III; future JWST/MIRI observations at higher spectral resolution could separate the two W III components from any Se III contribution and break that degeneracy.","Because R-matrix datasets now exist for W, Pt, and Au, the same forward-modelling pipeline could be applied to other r-process elements with available collision data, turning kilonova spectra into a systematic abundance inventory."],"forward_implications":["The 4.5 $\\mu$m feature in AT2017gfo and AT2023vfi can be modelled as blended W III 4432.23 nm and 4535.16 nm emission, with combined luminosity $5 \\times 10^{37}\\,\\mathrm{erg\\,s^{-1}}$ (AT2017gfo) or $1.0 \\times 10^{38}\\,\\mathrm{erg\\,s^{-1}}$ (AT2023vfi).","The implied W III masses are $1.65 \\times 10^{-4}\\,M_\\odot$ (AT2017gfo) and $9.4 \\times 10^{-4}\\,M_\\odot$ (AT2023vfi), corresponding to roughly 0.33% and 1.6% of each ejecta, and possible total W fractions up to about 1.0% and 4.7%.","These masses are broadly consistent with theoretical W yields from neutron-star merger models for AT2017gfo; for AT2023vfi the estimate exceeds the models by up to an order of magnitude, which may indicate more neutron-rich ejecta or an observational overestimate.","A measured W mass fraction can be converted into constraints on lanthanide plus actinide and third-peak r-process element mass fractions (1.8% to 5.5% and 1.6% to 9.9% for AT2017gfo using the adopted correlations), connecting the line to r-process yield and opacity.","The predicted W III line profile shape, computed from merger-model velocity distributions, could distinguish short-lived versus long-lived neutron-star remnants because low-velocity W is depleted in long-lived remnant models."],"supporting_citations":[{"why":"Supplies the W III R-matrix effective collision strengths and A-coefficients that drive the mass inversion.","marker":"McCann et al. (2024)"},{"why":"Provides the W I R-matrix atomic data used in the collisional-radiative models.","marker":"Smyth et al. (2018)"},{"why":"Provides the W II R-matrix atomic data used in the collisional-radiative models.","marker":"Dunleavy et al. (2022)"},{"why":"Introduced the W III identification for the 4.5 $\\mu$m feature and the adopted AT2017gfo band luminosity.","marker":"Hotokezaka et al. (2022)"},{"why":"Gives the measured 4.5 $\\mu$m luminosity for AT2023vfi and the line width used in the synthetic spectra.","marker":"Gillanders & Smartt (2024)"},{"why":"Supplies the JWST spectrum of AT2023vfi and the adopted temperature, density, and total ejecta mass.","marker":"Levan et al. (2024)"},{"why":"Provides the hydrodynamical neutron-star merger models and nucleosynthesis yields used for comparison.","marker":"Just et al. (2023)"},{"why":"Adds short-lived remnant merger models that bracket the predicted W mass range.","marker":"Sneppen et al. (2024)"},{"why":"Reports the Spitzer 4.5 $\\mu$m observation of AT2017gfo that anchors the AT2017gfo luminosity.","marker":"Kasliwal et al. (2022)"}],"fun_headline_variants":["W III lines explain kilonova infrared glow","Kilonova 4.5-micron glow traced to tungsten ions","W III mass inferred from kilonova IR excess","Tungsten lines pin down kilonova glow origin"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 4.5 $\\mu$m feature is assumed to be dominated by W III emission rather than Se III or other species, and W III is taken as the dominant tungsten ionisation stage; the adopted electron temperature and density enter the mass estimate directly and change it by up to three orders of magnitude across the explored grid.","fun_headline_variants_meta":{"raw":{"variants":["W III lines explain kilonova infrared glow","Kilonova 4.5-micron glow traced to tungsten ions","W III mass inferred from kilonova IR excess","Tungsten lines pin down kilonova glow origin"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000741,"raw_usage":{"total_tokens":3444,"prompt_tokens":1217,"completion_tokens":2227,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":833,"completion_tokens_details":{"reasoning_tokens":2159}},"tokens_in":833,"tokens_out":2227,"duration_ms":14372,"temperature":1.0,"reasoning_tokens":2159,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:05:37.929631+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Look for the predicted 6.05 $\\mu$m W I/W II blend in the same late-time spectra: the model predicts it should be comparable in luminosity to the 4.5 $\\mu$m feature when W I or W II masses are similar to W III, so its absence would contradict the W III identification; alternatively, a high-resolution mid-infrared spectrum that fails to show the two W III components at their predicted ~6:1 intensity ratio would falsify the claim.","supporting_citations":[{"cited_title":"arXiv:2411.03427","cited_arxiv_id":null,"evidence_quote":"Adds short-lived remnant merger models that bracket the predicted W mass range."}],"review_version":1}