{"id":"95b0f091-9ab0-4021-bdf1-adc3d7dcff94","arxiv_id":"2501.13286","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Improved HULLAC atomic data for singly ionized lanthanides lower the energy levels and raise kilonova opacities by factors of 3-10 per element and roughly 1.5 for a lanthanide-rich mixture, broadly matching ab initio GRASP results.","lead":"This paper recomputes atomic energy levels and transition data for singly ionized lanthanides (elements 59-70) using improved HULLAC calculations, and finds that kilonova opacities are up to 3-10 times higher per element, and about 1.5 times higher for a lanthanide-rich mixture, than the authors' previous Paper I values.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed 3–10x opacity increase is not robust: Table 1 shows 10–42% level-energy errors with no propagation to the opacity, and Gd II has unphysical level ordering.","rationale":"The central claim is that improved HULLAC data raise singly ionized lanthanide opacities by up to 3–10, and by about 1.5 for a mixture. The load-bearing condition is that the improved energy level distributions are accurate enough for the opacity change to be real and correctly sized. The paper's own validation shows residual errors (Delta_median = 10–25%, Gd II 42%) that are not propagated into the opacity. Because the expansion opacity depends on populations through exp(-E_l/kT) and on line positions through the wavelength grid, even a 20% error in a level a few eV above ground can alter the opacity by factors comparable to the claimed increase. The GRASP agreement is the best evidence that the new levels are better than Paper I, and for most elements this is an independent check; however, Pm II and Ho II are in-sample, and Tb/Dy/Ho still disagree with GRASP by about 3x in some bands. Thus the qualitative direction (higher opacity than Paper I) is well supported, but the quantitative factor 3–10 is not established without an uncertainty budget. A Monte Carlo perturbation of the levels by the reported errors directly tests this. The reader's weakest assumption (reference energy accuracy) is valid and is one source of the residual errors, but the concern is broader: it applies even if the references are perfect, because the HULLAC model and fitting procedure have intrinsic error. I therefore partially agree with the reader and recommend keeping the conditional verdict.","tokens_in":20688,"tokens_out":11192,"duration_ms":103515,"concrete_test":"Run a Monte Carlo sensitivity analysis: perturb each ion's HULLAC energy levels by random Gaussian shifts with sigma equal to the Table 1 median error (and 2x for Gd II), respecting parity and configuration ordering, and recompute the expansion opacity at T=5000 K and the ratio to Paper I over at least 1000 realizations. If the 68% interval of the ratio spans more than a factor of about 2, the claimed 3–10 factor is not supported by the stated atomic-data accuracy.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 2.1 reports median normalized energy errors of 10–25% (Table 1), with 42% for Gd II, where the optimized potential places 4f7 5d2 and 4f8 6s below the NIST ground state 4f7 5d 6s (Section 2.2). At T=5000 K (kT≈0.43 eV), the Boltzmann factor exp(-E_l/kT) is exponentially sensitive: a ~0.5 eV error in a 2 eV lower level changes its population by a factor ~3. The paper computes opacities (Section 3.1) from these levels but does not propagate the stated errors, nor does it test sensitivity to correlated shifts in energy distributions. The GRASP comparison is supportive in many cases, but for Pm II and Ho II it is in-sample (Section 2.1), and for Tb II, Dy II, Ho II factor ~3 discrepancies remain at 5000–10000 A (Section 4.2). Therefore the magnitude of the factor 3–10 increase is not quantitatively established; the direction may be right, but the headline number could change substantially under plausible systematic level errors.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents improved HULLAC atomic structure calculations for singly ionized lanthanides (Z=59–70), using optimized effective potentials and a restricted set of low-lying configurations. Compared with the authors' previous Paper I, the new calculations produce systematically lower energy level distributions, and the resulting LTE expansion opacities are higher by up to a factor of 3–10 for individual elements and by about a factor of 1.5 for a lanthanide-rich mixture (Ye=0.20). The opacities are compared with those from GRASP2K calculations (G19, R20, R21), and the paper identifies the transition arrays that dominate the opacity in different wavelength ranges. The atomic data are made publicly available.","tokens_in":21032,"tokens_out":3369,"duration_ms":29935,"significance":"If the reported opacity increase is robust, it would imply that kilonova models using Paper I data underestimated the singly ionized lanthanide opacity, affecting inferred r-process yields from observations like AT2017gfo. The paper's strengths include a systematic 12-element comparison, a clear identification of the configurations responsible for the opacity (transition-array analysis in Section 4.1), and a public data release. The GRASP comparisons provide a useful cross-check for most elements, and the observation that the mixture opacity changes by only ~1.5 despite larger per-element changes is an important, non-obvious result. However, the central quantitative claim (factor 3–10) is not yet supported with uncertainty estimates or sensitivity tests, and some of the validation is in-sample.","major_comments":[{"comment":"Table 1 reports median normalized level-energy errors of 10–25% (42% for Gd II), but the opacity calculations in §3.1 do not propagate these errors or test sensitivity to systematic shifts in the level distributions. Since the expansion opacity depends on the Boltzmann factor exp(-E_l/kT), a ~0.5 eV error in a 2 eV lower level changes its population by a factor ~3 at T=5000 K (kT≈0.43 eV), so the quoted factor 3–10 increase relative to Paper I is not quantitatively established. Please provide an uncertainty estimate, e.g., by recomputing opacities with level energies shifted by the reported errors (or by correlated shifts of entire configurations).","section":"§2.2, §3.1"},{"comment":"For Gd II, the optimized potential places the lowest levels of 4f7 5d2 and 4f8 6s below the NIST ground state 4f7 5d 6s, and the median error is 42% (Table 1). This unphysical level ordering changes the partition function and the populations of the lower levels used in Eq. (6)–(7); the opacity increase for Gd II may therefore be partly an artifact of the potential choice. Please quantify how the Gd II opacity changes when the ground configuration is enforced, or restrict the reported factor to elements for which the level ordering is physical.","section":"§2.2, Table 1"},{"comment":"For Pm II and Ho II, the GRASP energies are used both as the tuning target for the effective potential and as the validation of the resulting opacities, so the agreement between HULLAC and GRASP for these ions is not an independent check. Moreover, for Tb II, Dy II, and Ho II, the HULLAC and GRASP opacities still differ by a factor ~3 at 5000–10000 Å (Figures 5 and 10). This limits the strength of the claim that the new opacities are benchmarked by the ab initio calculations; please state explicitly which comparisons are independent and discuss the residual discrepancies.","section":"§2.1, §4.2"},{"comment":"The restricted RCI configuration set (4f^q(6s,5d,6p) plus 4f^{q-1}(5d^2,5d6s,6s^2,6s6p,5d6p)) is assumed to be sufficient for opacity-relevant bound-bound transitions, but the paper does not test convergence with respect to adding further correlation configurations (e.g., 4f^{q-2} 5d^2 6s or 4f^{q-1} 5d 6p^2). Since the analysis in §4.1 identifies transition arrays among these configurations as controlling the opacity, a missing array could change the factor 3–10 for individual elements. A convergence test for one or two representative ions (e.g., Sm II and Ho II) would support the completeness claim.","section":"§2.1, §4.1"}],"minor_comments":[{"comment":"The text and Figure 6 caption contain 'Plank mean opacity'; this should be 'Planck mean opacity' (the same typo appears in Appendix A captions).","section":"§3.2, Figure 6"},{"comment":"The abbreviations G19, R20, R21 are defined in the introduction, but the reference list uses full author-year citations; please make the connection explicit at first use and ensure the reference list includes all three papers (Gaigalas et al. 2019; Radžiūtė et al. 2020, 2021).","section":"§1, references"},{"comment":"The sentence 'the black line in each calculation represent the characteristic features' has a subject-verb agreement error; also, 'we here analyze' is awkward.","section":"§4.1, Figure 7"},{"comment":"The caption states that the first and second rows are HULLAC results of the present calculation and Paper I, respectively; please clarify whether the first row is always the present work, as the row ordering appears different for some ions (e.g., Yb II) where the strategies are identical.","section":"Table 2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of MNRAS and addresses an important need in kilonova opacity modeling. The main issue is that the headline factor 3–10 is presented without uncertainty propagation or sensitivity tests, and the GRASP validation is partly in-sample for Pm II and Ho II. I would support publication after the authors add an uncertainty/sensitivity analysis and clarify the independence of the benchmarks."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe main thing to know: this paper gives the kilonova community a noticeably better set of HULLAC atomic data for singly ionized lanthanides, and it identifies which transition arrays actually control the opacity. The headline claim — opacities up to a factor of 3–10 higher than Paper I for individual elements — is plausible and probably in the right direction, but the exact magnitude is not yet pinned down. The factor ~1.5 for a lanthanide-rich mixture is on firmer ground.\n\nWhat's genuinely new: the authors expanded the configuration set (adding 4f^{q-1}5d^2, 5d6p, etc.), tuned the effective potential per ion against NIST and GRASP energies, and show that this lowers the excited-state energy distributions relative to Paper I. The comparison with the ab initio GRASP opacities for twelve ions is a real validation step, and for most elements the new HULLAC opacities land much closer to GRASP than Paper I did. The line-array analysis (e.g., for Sm II and Ho II) is a useful physical explanation for where the opacity comes from and why Paper I had a dip at 3000–4000 Å. The data are made available in a public database, which is commendable.\n\nWhere I'd push back: the paper reports 10–25% median level-energy errors, and 42% for Gd II, and then does not propagate these to the opacity. At T=5000 K a 0.5 eV error in a low-lying level changes its Boltzmann population by a factor of ~3, so the factor 3–10 for single elements is not quantitatively established. The GRASP comparison is partly in-sample: for Pm II and Ho II the same GRASP energies were used to optimize the potential (Section 2.1) and then used for validation. For Tb II, Dy II, and Ho II, the HULLAC and GRASP opacities still differ by ~3 in the 5000–10000 Å range. The Gd II case is also a bit awkward: the optimized potential puts 4f^7 5d^2 and 4f^8 6s below the NIST ground state, which the authors report honestly but which is a red flag for that one ion. These caveats don't overturn the direction of the change — the GRASP comparison for the other ions supports it — but they do mean the headline factor is a range, not a precise number.\n\nI think this deserves a serious referee. The dataset is a real improvement, the analysis of transition arrays is valuable, and the limitations are transparent. The major revision should ask for uncertainty estimates (e.g., opacity sensitivity to shifts in the level energies) and for an independent benchmark for at least one of Pm/Ho. For the mixture opacity, the factor ~1.5 is likely robust, and the paper's own statement that the impact on light curves is limited is an honest caveat.\n\nFor peer review: send it out. It's a solid, useful paper with a clearly stated limitation that the authors are open about. A careful reviewer could push on the error propagation, but this is not a desk-reject.","headline":"A genuinely improved lanthanide opacity dataset, with a likely-but-not-quantitatively-pinned-down factor 3-10 increase; the mixture factor ~1.5 is the more robust result.","tokens_in":21486,"tokens_out":2883,"would_cite":true,"duration_ms":31095,"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":"Improved atomic calculations for singly ionized lanthanides lower their predicted energy levels and raise kilonova opacities by up to a factor of 3–10 for individual ions and about 1.5 for a lanthanide-rich mixture.","keywords":["kilonova","opacity","lanthanides","atomic data","neutron star mergers","r-process","energy levels","expansion opacity"],"falsifier":"A laboratory measurement of the cumulative number of energy levels below 6 eV for a singly ionized lanthanide such as Sm II or Ho II would settle whether the new level distributions are right, because if the measured level density matches Paper I rather than the new calculations, the opacity increase would disappear.","tokens_in":20535,"feed_emoji":"💥","tokens_out":9331,"duration_ms":82491,"temperature":0.7,"pith_summary":"This paper claims that the opacity of singly ionized lanthanides in kilonova ejecta has been underestimated, because the earlier calculations placed excited energy levels too high. Using an improved tuning strategy for the atomic effective potential and a larger set of configurations, the authors obtain systematically lower level distributions for $Z=59$–$70$ and find that the resulting expansion opacities at $T=5000$ K are higher than those from the previous study by up to a factor of 3–10, with about a factor of 1.5 for a lanthanide-rich mixture. They also show that the new opacities broadly agree with independent ab initio GRASP2K calculations, and they trace the opacity structure to specific transition arrays. If correct, kilonova models using the previous data understate how effectively lanthanides absorb and reprocess radiation in neutron-star merger ejecta.","feed_headline":"Better atomic data raise kilonova opacity up to tenfold","feed_subtitle":"Tuned energy levels for singly ionized lanthanides push the opacity of neutron-star merger ejecta above earlier values.","key_machinery":"The load-bearing mechanism is the optimized effective central-field potential inside the HULLAC code, a relativistic configuration-interaction atomic-structure calculation. The potential is parameterized by orbital occupation numbers and mean radii, and the parameters are varied by Nelder–Mead minimization until the first-order energies of the ground and low-lying states are minimized, giving potentials tuned to reproduce the lowest level of each configuration. Along with the configuration set $4f^q(6s,5d,6p)$ and $4f^{q-1}(5d^2,5d\\,6s,6s^2,6s\\,6p,5d\\,6p)$ for $q=3$–$14$, this lowers the entire level distribution relative to Paper I. The opacity side uses the standard expansion-opacity formula with Sobolev optical depths, LTE level populations, and Saha ionization, evaluated at $T=5000$ K, $\\rho=10^{-13}$ g cm$^{-3}$, and $t=1$ day.","core_discovery":"The paper's central discovery is that the energy-level distributions of singly ionized lanthanides are systematically lower than the authors' previous Paper I results, and that this shift is what raises the opacity. For each ion from Pr II to Yb II, the new HULLAC calculations include configurations such as $4f^{q-1}5d^2$, $4f^{q-1}5d\\,6p$, and $4f^{q-1}6s\\,6p$, and the effective potential is optimized against the lowest levels of each configuration in the standard atomic database and, for Pm II and Ho II, against the GRASP benchmark data. The median deviations from reference levels drop to 10–25% (42% for the special Gd II case), much smaller than the 20–100% differences in Paper I, and the cumulative level counts below 6 eV rise substantially. Since opacity under local thermodynamic equilibrium is driven by the Boltzmann population of excited states, the lower levels translate into more bound-bound transitions and higher expansion opacities. The opacities from the new data are higher than Paper I by up to a factor of about 10 at wavelengths below roughly 5000 Å for eight of the twelve ions, and for the $Y_e=0.20$ lanthanide-rich mixture the Planck mean opacity is higher by a factor of about 1.5–1.6 at $T=4000$–$5000$ K; in both comparisons the new results move toward, and mostly agree with, the ab initio GRASP opacities.","pith_inferences":["Editorial inference: if the systematic lowering seen here extends to neighboring ionization stages, the total opacity of lanthanide-rich ejecta could rise by more than the factor of 1.5 found for singly ionized states alone, because the missing ionization states already contribute at higher temperatures.","Editorial inference: an independent laboratory measurement of the low-lying levels of one benchmark ion, such as Sm II or Ho II below 6 eV, would directly test whether the tuned potentials produce the true level distribution rather than merely reproducing the calculations used for tuning.","Editorial inference: for Pm II and Ho II, the same GRASP data are used both to set and to validate the potentials, so the good agreement for those ions is partly circular; an independent experimental or theoretical check would remove that degeneracy."],"forward_implications":["For individual ions such as Pm II, Sm II, Eu II, Gd II, Tb II, Dy II, Ho II, and Er II, the improved data raise expansion opacities at wavelengths below about 5000 Å by factors up to roughly 10 relative to Paper I.","For a lanthanide-rich ejecta composition ($Y_e=0.20$), the Planck mean opacity of singly ionized lanthanides at $T=4000$–$5000$ K rises by a factor of about 1.5–1.6 over Paper I.","The new opacities largely match the independent ab initio GRASP opacities, supporting the conclusion that Paper I underestimated the singly ionized lanthanide opacity rather than the benchmark being wrong.","The transition-array decomposition shows that accurate energy levels for specific configurations, especially $4f^{q-1}5d^2$ and $4f^{q-1}5d\\,6p$, are required to avoid order-of-magnitude opacity errors in the UV and blue.","Because only singly ionized lanthanides were recomputed, the full effect on kilonova light curves will remain uncertain until similar benchmarks are done for the other ionization states that dominate at other temperatures."],"supporting_citations":[{"why":"Provides the previous HULLAC opacities and level distributions that define the baseline; the new results are measured against it as Paper I.","marker":"Tanaka et al. 2020"},{"why":"Supplies the HULLAC code with which all new atomic-structure calculations are performed.","marker":"Bar-Shalom et al. 2001"},{"why":"Supplies the reference energy levels from the standard atomic database used to tune and validate the effective potentials.","marker":"Kramida et al. 2018"},{"why":"Supplies benchmark ab initio GRASP data for Nd II used to validate the level distributions and opacities.","marker":"Gaigalas et al. 2019"},{"why":"Supplies benchmark ab initio data for Pr and Pm–Gd; for Pm II these data also serve as the tuning reference.","marker":"Radžiūtė et al. 2020"},{"why":"Supplies benchmark ab initio data for Tb–Yb; for Ho II these data also serve as the tuning reference.","marker":"Radžiūtė et al. 2021"},{"why":"Provides the $Y_e=0.20$ r-process trajectory whose lanthanide abundance pattern defines the mixture-opacity comparison.","marker":"Wanajo et al. 2014"}],"fun_headline_variants":["Lower lanthanide levels boost kilonova opacity up to 10x","New atomic data raise kilonova opacity up to tenfold","Improved lanthanide calculations make kilonova ejecta more opaque"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole factor-of-3–10 result assumes that the reference energy levels used to tune the potentials are accurate; if the database levels or the benchmark calculations are biased, the lowered level distributions and the opacity increase inherit that bias.","fun_headline_variants_meta":{"raw":{"variants":["Lower lanthanide levels boost kilonova opacity up to 10x","New atomic data raise kilonova opacity up to tenfold","Improved lanthanide calculations make kilonova ejecta more opaque"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000844,"raw_usage":{"total_tokens":3761,"prompt_tokens":1118,"completion_tokens":2643,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":734,"completion_tokens_details":{"reasoning_tokens":2581}},"tokens_in":734,"tokens_out":2643,"duration_ms":20367,"temperature":1.0,"reasoning_tokens":2581,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T16:18:02.208340+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A laboratory measurement of the cumulative number of energy levels below 6 eV for a singly ionized lanthanide such as Sm II or Ho II would settle whether the new level distributions are right, because if the measured level density matches Paper I rather than the new calculations, the opacity increase would disappear.","supporting_citations":[],"review_version":1}