{"id":"a8d703b4-3609-48af-9130-5ce1efce6c35","arxiv_id":"2507.07785","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A new calibrated atomic dataset for all singly and doubly ionized lanthanides provides 66,591 experimentally anchored transition wavelengths for kilonova modeling.","lead":"Astronomers need precise atomic data to read the light from kilonovae, the explosions that forge heavy elements. This paper computes and calibrates such data for rare earth elements in two ionization states, giving tens of thousands of transitions with accurately known wavelengths.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Calibrated-wavelength count rests on manual LS-label matches that are never validated against independent measured wavelengths; a wrong match silently produces a wrong Ritz wavelength for both-end calibrated transitions.","rationale":"The reader's weakest_assumption identifies manual LS matching as the reliability bottleneck; this stress-test agrees and sharpens it into a concrete, testable consequence: the 66,591 'experimentally calibrated' transitions are only trustworthy if the manual level matches are correct, and the paper provides no independent wavelength-based validation of those matches. The cited experimental line lists by Ferrara et al., Den Hartog et al., and Voith et al. are already used for gf comparisons, so a wavelength comparison is feasible without new observations. This concern does not by itself overturn the paper: the dataset is publicly released, the manual step is transparently acknowledged, and the gf benchmarks are genuine evidence. However, the abstract's headline claim is specifically about calibrated wavelengths enabling line identifications, so the missing holdout check is the decisive test. The reader's CONDITIONAL verdict already reflects the need for such validation, so the verdict should remain UNCHANGED. If the proposed check fails, the central claim should be hardened toward REJECT; if it passes, the remaining issue is largely the DREAM-anchored subset and the precision of the 'experimental' qualifier in the abstract.","tokens_in":61556,"tokens_out":8675,"duration_ms":110387,"concrete_test":"From the Zenodo release, select transitions whose lower and upper levels are both marked xmatch. For Pr II, Nd II, Sm II, Eu II, Dy II, Er II, match these to Ferrara et al. (2024) laboratory wavelengths by level pair; for Tm II use Den Hartog et al. (2024); for Gd II use Voith et al. (2025). Neither list was used for calibration. Compute delta = lambda_release - lambda_lab for each common line and report the RMS and the fraction of outliers beyond 0.5 A or 3 sigma. Then audit a random sample of 100 xmatch assignments per ion against literature Landé g-factors and independent GRASP/HFR eigenvector compositions, counting reassignments. A small outlier fraction and <5% reassignments would validate the calibration; large outliers or frequent reassignments would show the xmatch wavelength set is biased.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central deliverable is the 66,591 'experimentally calibrated' transitions (abstract; Table V). Section II C states the JJ2LSJ transformation 'is not bijective; multiple ASFs may share the same dominant LS label' and that 'all calibrations were performed manually.' The calibration pipeline is therefore only as reliable as its LS-purity-based level matches. If a theoretical level is matched to the wrong experimental level, it still receives the xmatch metadata flag, and the resulting 'calibrated' transition wavelength becomes a Ritz difference between two unrelated experimental levels; in dense lanthanide spectra this can be off by thousands of cm^-1, on the scale of the median corrections in Figures 1-2, while the transition looks fully anchored. The paper validates level-energy residuals and gf values (Figures 6-7), but never compares the calibrated wavelengths themselves against independent measured line positions. Since the abstract's feasibility claim concerns line identifications, the unvalidated manual matching is the load-bearing weak point. A secondary issue is that Section II C also admits calibration against theoretical DREAM levels, so 'experimental' may not apply to every member of the 66,591. No holdout validation is reported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript presents FAC-based relativistic configuration-interaction calculations of atomic structure and E1 transitions for all 28 singly and doubly ionized lanthanide ions (La II through Yb III). The authors optimize the fractional-mean-configuration potential with an SMBO algorithm (Eq. 3) and then manually calibrate computed levels to NIST and DREAM reference data (§II C). The headline deliverable is a public dataset of 146,856 levels, 28,690,443 E1 transitions, and 66,591 transitions whose upper and lower levels are both flagged as calibrated, which the abstract claims carry 'experimentally calibrated wavelength information.' The paper benchmarks level energies, oscillator strengths against experimental and semi-empirical data (§IV A), and LTE expansion and Planck-mean opacities against HULLAC, GRASP, and HFR datasets (§IV B). The authors report roughly 0.5 dex scatter for strong transitions (log(gf) > -1), agreement of Planck-mean opacities with HULLAC V2 and GRASP to within about 20%, and factors-of-several-to-ten differences with newer HFR opacities, which they attribute to level-density differences at intermediate excitation energies.","tokens_in":61810,"tokens_out":9339,"duration_ms":105434,"significance":"The dataset addresses a recognized bottleneck in kilonova spectroscopy: the scarcity of wavelength-calibrated atomic data for lanthanides. If the calibration pipeline is trustworthy, the 66,591 anchored transitions would materially advance line-identification efforts for AT2017gfo-like spectra, and the level/transition tables plus the Zenodo data release provide a reproducible resource for radiative-transfer modeling. Strengths of the paper include systematic coverage of 28 ions, explicit benchmarks against several independent codes (HULLAC V2, GRASP, HFR) and against recent laboratory gf measurements, a transparent statement of the non-bijective JJ2LSJ labeling problem, and machine-readable public data. The oscillator-strength and opacity benchmarks are genuinely independent and generally supportive of the calculations. The main risk is concentrated in the unvalidated manual level-matching step that underpins the calibrated-wavelength count; this is fixable within the manuscript's scope because the authors already cite the measured line lists needed for a direct wavelength validation.","major_comments":[{"comment":"The paper's headline deliverable, the 66,591 transitions claimed to have 'experimentally calibrated wavelength information' (abstract; Table V), is not validated against independent measured wavelengths. The calibration pipeline in §II C depends entirely on manual LS-label matching between FAC levels and NIST/DREAM levels, with the JJ2LSJ transformation explicitly acknowledged to be non-bijective and with all calibrations performed by hand. The reported benchmarks do not test the correctness of these matches: Figures 1-2 show residuals for the very levels used in the calibration, and Figures 6-9 compare oscillator strengths, not line positions, for transitions whose levels were both identified. Because a wrong match still receives the 'xmatch' flag and produces a Ritz wavelength that can be off by thousands of cm-1, the abstract's feasibility claim for line identifications is not yet supported. I request a direct validation of the calibrated wavelengths against measured line positions, for example the NIST line lists and the Ferrara et al. (2024), Den Hartog et al. (2024), and Voith et al. (2025) line lists already used in §IV A, or a hold-out test in which a subset of known levels is excluded from calibration and used only for verification.","section":"II C, IV A, Table V"},{"comment":"The 'experimentally calibrated' label overstates the anchoring for a substantial, unquantified fraction of the 66,591 transitions. Section II C states that calibration aligns computed levels with 'experimental energy levels from the NIST ASD and, where applicable, theoretical data from DREAM'; DREAM is a semi-empirical HFR database, not a set of measurements. Table V shows DREAM reference levels for many ions (e.g., Ce II, Pr II, Tm II), so some of the 66,591 transitions are anchored on both ends to theoretical levels. The abstract and §III B should either qualify the claim (for example, 'experimentally or semi-empirically anchored') or report the breakdown of NIST-anchored versus DREAM-anchored transitions in Table V. This is not a purely semantic point: a user relying on the calibrated wavelengths for line identifications needs to know which transitions carry which anchoring status.","section":"II C, Table V"},{"comment":"The level-energy 'agreement' reported in Figures 1-2 and discussed throughout §III is partly by construction and should be framed as fit residuals rather than predictive validation. The FMC occupation weights are optimized against experimental reference levels via the loss function in Eq. (3), and the calibration shifts of §II C are applied to the same experimental levels that are then used as benchmarks. The genuinely independent accuracy evidence comes from the gf comparisons against experiments (§IV A) and the opacity comparisons against other codes (§IV B), and that evidence is supportive; however, the manuscript should state this distinction explicitly and, ideally, report residuals on levels lying above the calibration range or on a hold-out subset of levels. In addition, because 'all calibrations were performed manually,' the Zenodo release should include the mapping between each matched FAC level and the specific NIST or DREAM energy level; without that mapping, the 66,591-transition claim is not auditable by other groups.","section":"II B, II C, Figures 1-2"}],"minor_comments":[{"comment":"The captions of Figures 1 and 2 both read 'doubly ionized lanthanide ions,' but the ion labels and the text of §II C indicate that Figure 1 shows singly ionized species and Figure 2 doubly ionized species; please correct the Figure 1 caption.","section":"Figures 1 and 2 captions"},{"comment":"The sentence 'These constraints ensure that the optimization explores physically meaningful configurations while allowing sufficient flexibility to improve the accuracy of the calculations.' appears twice verbatim; please delete the duplicate.","section":"II B"},{"comment":"The text in the gadolinium subsection states that the FAC calculation 'recover[s] a ground state configuration of 4f7 5d1 6s1, which differs from the experimental ground state 4f7 5d2,' whereas Table IV lists the FAC ground state as 4f7 5d2 and the NIST ground state as 4f7 5d1 6s1; these statements contradict each other and must be reconciled.","section":"III A 8 and Table IV"},{"comment":"The transition table excerpt contains formatting artifacts in which wavelength and log(gf) values run together (e.g., '7068.15-1.0213' and '4437.09.-1.4071'); the machine-readable version may be unaffected, but the printed excerpt should be cleaned up.","section":"Table VII excerpt"},{"comment":"The continuation tables of Table I are headed 'TABLE II. *' and 'TABLE III. *'; the stray asterisk placeholder should be removed.","section":"Tables II and III headers"},{"comment":"The caption of Figure 36 says the black horizontal lines show NIST experimental data, but the text of §III A 10 states that Dy III calibration uses level information from Spector et al. (1997) because the NIST ASD only contains the ground state for Dy III; the caption should be updated to reflect the actual calibration source.","section":"Figure 36 caption"}],"recommendation":"major_revision","confidential_remarks":"This is a data-resource paper in the style of ApJS or A&A. My recommendation of major revision rests on a single load-bearing gap: the absence of any independent validation of the calibrated wavelengths that constitute the paper's central deliverable. The underlying FAC calculations and the independent gf and opacity benchmarks are credible, so the gap appears fixable within the scope of a revision. Please also ensure the authors correct the internal inconsistencies (Figure 1/2 captions and the Gd II ground-state contradiction between Table IV and the text) in the same round."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nRead the Flörs et al. lanthanide atomic data paper. Bottom line: it's a serious data production effort, and the dataset is worth having, but the paper's central claim about line-identification readiness rests on a calibration step that is not independently validated.\n\nWhat's new: previous work by this group developed the FAC potential optimization and calibration method for Nd II/III and U III. Here they push it through all 28 singly and doubly ionized lanthanides, bring in the new Nd III measurements from Ding et al., and release a large dataset on Zenodo: 146k levels, 28.7M E1 transitions, and 66,591 transitions flagged as having both upper and lower levels matched to experimental energies. That breadth is the contribution. The opacity comparisons to HULLAC V2, GRASP, and HFR are genuinely useful, and the gf comparisons for strong lines look credible. The metadata on calibration status per transition is thoughtful.\n\nThe soft spot is exactly where the stress-test says: the calibration is done by manually matching theoretical levels to experimental ones using LS labels. The paper itself notes the JJ2LSJ transformation is not bijective and that all calibrations were performed manually. A wrong match produces a Ritz wavelength that looks fully anchored but is actually a difference between two unrelated experimental levels. In dense lanthanide spectra, that can be off by thousands of cm^-1, on the scale of the median corrections shown in Figures 1-2. The paper validates level-energy residuals and gf values, but never compares the calibrated wavelengths themselves against independent measured line positions. No holdout validation. So the 66,591 number is only as good as the manual matching, and we have no direct test of it.\n\nAlso, the FMC weights and calibration shifts are fit to the same experimental levels used as benchmarks, so the reported level-energy agreement is partly by construction. That's not disqualifying—the gf and opacity comparisons are independent—but it lowers the evidential value of the level residuals.\n\nMinor issue: the generation code is not released, only the data. For a paper whose purpose is to provide a reusable resource, that's a small miss.\n\nWho should read this: anyone doing kilonova radiative transfer with lanthanides, and atomic physicists interested in calibration practice. It deserves a serious referee. I'd send it to review, but ask the authors to add a validation of the calibrated wavelengths, e.g., withholding a subset of measured lines from the calibration and checking the predicted positions, or at least reporting a confusion estimate for the manual matches. Without that, the headline claim is stronger than the evidence.\n\nRecommendation: accept for review with the expectation of major revision or an explicit caveat.","headline":"A useful lanthanide dataset with a real but addressable calibration-validation gap; the opacity benchmarks are solid, but the 66k calibrated wavelengths need a direct test before line-identification claims carry weight.","tokens_in":62346,"tokens_out":2648,"would_cite":true,"duration_ms":28620,"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":"This paper builds a wavelength-calibrated lanthanide atomic dataset—66,591 transitions pinned to experiment out of 28.7 million—so kilonova spectra can be matched line by line.","keywords":["lanthanides","kilonova","atomic opacities","energy levels","oscillator strengths","energy level calibration","radiative transfer","r-process nucleosynthesis"],"falsifier":"Take a random sample of the 66,591 calibrated transitions for ions with independent high-resolution laboratory data, such as Fourier-transform spectra recorded after this dataset was assembled, and check whether each predicted wavelength coincides with an observed line within the paper's stated calibration uncertainty. A mismatch rate well above the few-percent level for transitions flagged as having both levels matched to experiment would show the manual level-assignment step is biased, whereas near-complete agreement would confirm the method and the opacities derived from it.","tokens_in":2134,"feed_emoji":"⚛️","tokens_out":3946,"duration_ms":125407,"temperature":0.7,"pith_summary":"After a neutron-star merger, the ejected heavy elements imprint their absorption lines on the optical and near-infrared light we call a kilonova, and the lanthanide elements (atomic numbers 57–70) are the main source of that blanketing. This paper tries to give radiative-transfer models atomic data accurate enough for line-by-line spectral matching: it computes the energy levels and electric-dipole transitions of all 28 singly and doubly ionized lanthanides with a relativistic configuration-interaction code, then calibrates low-lying levels against measured laboratory energies. The payoff claimed is 146,856 energy levels and 28,690,443 E1 transitions, of which 66,591 have experimentally anchored wavelengths—enough, the authors argue, to make lanthanide line identifications in kilonova spectra feasible. They also report that strong lines agree with experiment and with the best prior calculations, and that the calibration barely changes the derived opacities, which matters because opacity governs the brightness and color of the transient.","feed_headline":"Kilonova line identifications get 66,591 calibrated wavelengths","feed_subtitle":"Complete level and transition data for all 28 singly and doubly ionized lanthanides, anchored to measured spectra.","key_machinery":"The load-bearing machinery is a two-stage pipeline. First, the Flexible Atomic Code diagonalizes the Dirac–Coulomb Hamiltonian for a carefully chosen set of configurations, with the local central potential optimized by a Bayesian sequential-model optimization (SMBO) that minimizes a Boltzmann-weighted root-mean-square deviation between computed and reference energies—this weighting is what makes low-lying levels, the ones thermally populated in kilonova ejecta, come out right. Second, a manual calibration step converts the code's $jj$-coupled levels into LS labels via a transformation routine, matches them to experimental energies within parity and total-angular-momentum groups, and applies corrections that range from a direct replacement (when the dominant LS component exceeds 0.7) through a scaled partial correction (0.5–0.7) to a symmetry-group average (below 0.5). The output carries per-level and per-transition metadata stating whether both, one, or neither endpoint was matched to experiment, which is what lets a modeler trust the 66,591 calibrated wavelengths and treat the rest as statistically placed lines.","core_discovery":"On its own terms, the paper's central claim is that the bottleneck for identifying heavy elements in kilonova spectra is not a lack of lines but a lack of lines with trustworthy wavelengths, and the paper removes that bottleneck for the lanthanides. Using the Flexible Atomic Code with an optimized local central potential, the authors computed relativistic configuration-interaction structures for all 28 singly and doubly ionized lanthanide ions (La II/III through Yb II/III), yielding 146,856 bound energy levels below the ionization threshold and 28,690,443 E1 transitions between them, with the correct ground-state configuration recovered for every ion except Gd II. Low-lying levels were then matched by hand, within parity and total-angular-momentum groups, to experimental levels from standard databases, shifting or partially correcting each theoretical level so that 66,591 transitions have both endpoints anchored to measurement; the average calibration corrections are about 2950 $\\mathrm{cm}^{-1}$ for singly and 2571 $\\mathrm{cm}^{-1}$ for doubly ionized species. The authors further claim that strong transitions ($\\log(gf) > -1$) agree with experimental and semi-empirical values to a scatter of roughly 0.5 dex, while theoretical weak lines are systematically weaker than measured ones; that the resulting LTE opacities agree with the updated HULLAC and GRASP datasets to within about 20 percent but disagree with one recent HFR dataset whose near-ground level density is judged inconsistent with experiment; and that calibrating the energies leaves the opacities almost unchanged, because the bulk of the opacity comes from the many uncalibrated lines that fill in between the measured ones.","pith_inferences":["If the manual matching bottleneck could be automated, the same calibration pipeline would transfer to actinides and to higher ionization stages, where experimental anchors are sparser and the payoff for kilonova and supernova modeling would be comparable.","The systematic weakness of theoretical gf values for faint lines, which the authors note appears across independent codes and experiments, implies that light-curve models built purely on strong-line data may understate line blanketing; testing this would mean recomputing synthetic spectra with the weak-line population boosted to match experiment.","Since calibration moves opacities little but moves wavelengths a lot, the dataset's decisive test is spectral: synthetic spectra computed with calibrated lines should match narrow observed absorption features in AT2017gfo markedly better than the same model with uncalibrated lines.","The authors' statement that Pm III alone lacks calibration anchors suggests promethium will remain the hardest lanthanide to identify; a targeted laboratory measurement campaign on Pm III would be the most direct way to extend the calibrated list."],"forward_implications":["Radiative-transfer models fitted to AT2017gfo can now be run with wavelength-calibrated lanthanide line lists, and any synthetic feature that survives this test becomes a candidate real line rather than a numerical artifact.","The 66,591 calibrated lines constitute a concrete search list of optical and near-infrared wavelengths where lanthanide absorption should appear, giving observers a direct target for confirming or excluding third-peak r-process elements.","Because strong lines are reliable to about 0.5 dex while theoretical weak lines are systematically too weak, opacity and ejecta-mass inferences from synthetic spectra should be re-checked for sensitivity to the weak-line population.","The comparison discriminates between existing opacity datasets: calibrated FAC opacities sit within about 20% of the updated HULLAC and GRASP results, while the HFR dataset's factor-of-several higher opacities for some ions are explained by level densities that conflict with measured level counts.","Per-transition calibration flags let future spectral synthesis report, for every predicted absorption feature, whether its wavelengths are experimentally anchored or purely theoretical."],"supporting_citations":[{"why":"The relativistic configuration-interaction code that produces all energy levels and E1 transition data in the paper.","marker":"[52]"},{"why":"The Bayesian sequential-model optimization procedure used to tune the local central potential, which the paper credits for the improved uncalibrated level energies.","marker":"[61]"},{"why":"The authors' earlier neodymium and uranium study that introduced the level-calibration scheme which this paper generalizes to all 28 lanthanide ions.","marker":"[42]"},{"why":"The transformation that converts the code's jj-coupled levels into LS labels, which underpins every theoretical-to-experimental level match in the calibration.","marker":"[66]"},{"why":"The standard experimental database of measured atomic energy levels used as the calibration reference for most ions.","marker":"[64]"},{"why":"The semi-empirical dataset used as a supplementary calibration reference and as the main comparison for theoretical oscillator strengths.","marker":"[65]"},{"why":"Recent high-resolution laboratory oscillator-strength measurements used to benchmark the computed strong-line values for six singly ionized lanthanides.","marker":"[67]"},{"why":"The updated HULLAC opacity dataset that the paper finds agrees with its calibrated FAC opacities to within about 20 percent.","marker":"[48]"},{"why":"The recent HFR opacity dataset whose strongly elevated low-lying level densities produce the factor-of-several opacity disagreement the paper analyzes.","marker":"[49]"}],"fun_headline_variants":["66,591 calibrated lanthanide wavelengths unlock kilonova spectra","Kilonova spectroscopy gets 66k calibrated lanthanide lines","All 28 lanthanide ions: 28.7M transitions, 66k calibrated","Calibrated atomic data for every lanthanide ion in kilonovae","New lanthanide data pin down kilonova line identifications"],"cache_read_input_tokens":64512,"weakest_assumption_plain":"The load-bearing premise is that the manual matching of computed levels to measured ones—which the authors note is ambiguous because multiple computed levels share the same dominant LS label—is essentially correct; a modest rate of mismatch would silently corrupt the 66,591 calibrated wavelengths and everything built on them.","fun_headline_variants_meta":{"raw":{"variants":["66,591 calibrated lanthanide wavelengths unlock kilonova spectra","Kilonova spectroscopy gets 66k calibrated lanthanide lines","All 28 lanthanide ions: 28.7M transitions, 66k calibrated","Calibrated atomic data for every lanthanide ion in kilonovae","New lanthanide data pin down kilonova line identifications"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000662,"raw_usage":{"total_tokens":3158,"prompt_tokens":1210,"completion_tokens":1948,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":826,"completion_tokens_details":{"reasoning_tokens":1843}},"tokens_in":826,"tokens_out":1948,"duration_ms":14532,"temperature":1.0,"reasoning_tokens":1843,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T18:33:17.584111+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a random sample of the 66,591 calibrated transitions for ions with independent high-resolution laboratory data, such as Fourier-transform spectra recorded after this dataset was assembled, and check whether each predicted wavelength coincides with an observed line within the paper's stated calibration uncertainty. A mismatch rate well above the few-percent level for transitions flagged as having both levels matched to experiment would show the manual level-assignment step is biased, whereas near-complete agreement would confirm the method and the opacities derived from it.","supporting_citations":[{"cited_title":"Composition Effects on Kilonova Spectra and Light Curves: I","cited_arxiv_id":"1904.13298","evidence_quote":"The relativistic configuration-interaction code that produces all energy levels and E1 transition data in the paper."},{"cited_title":"Visible spectra of W8+ in an electron-beam ion trap","cited_arxiv_id":"2101.11193","evidence_quote":"The Bayesian sequential-model optimization procedure used to tune the local central potential, which the paper credits for the improved uncalibrated level energies."},{"cited_title":"Constraints on the presence of platinum and gold in the spectra of the kilonova AT2017gfo","cited_arxiv_id":"2101.08271","evidence_quote":"The authors' earlier neodymium and uranium study that introduced the level-calibration scheme which this paper generalizes to all 28 lanthanide ions."},{"cited_title":"The Spectrum and Energy Levels of the Low-lying Configurations of Nd III","cited_arxiv_id":"2307.09282","evidence_quote":"The semi-empirical dataset used as a supplementary calibration reference and as the main comparison for theoretical oscillator strengths."}],"review_version":1}