{"id":"897b63c7-b827-46b3-b514-22dc9480302a","arxiv_id":"2501.09181","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"The delayed MeV emission after the SGR 1806-20 giant flare matches the predicted radioactive decay signal of about 10^-6 solar masses of newly synthesized r-process nuclei.","lead":"A previously unexplained pulse of gamma-rays seen after a 2004 magnetar flare may be the radioactive glow of freshly forged heavy elements. If correct, magnetar giant flares become the second confirmed factory of r-process elements in the universe, with implications for galactic chemistry and cosmic rays.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Spectral evidence is not yet demonstrated: Fig. 3 compares RHESSI data to intrinsic (unattenuated) synthetic spectra at t ≈ 1000 s, when the ejecta are still optically thick, so the claimed spectral match is not a valid test of the r-process interpretation.","rationale":"The reader's weakest assumption is the giant-flare baryon ejection mechanism and the associated ejecta mass, velocity, and Ye. That is indeed a major external dependency. My stress-test instead targets an internal modeling inconsistency in the paper's spectral evidence: the only spectral dataset (RHESSI, t ≈ 400–900 s) is compared to synthetic spectra that explicitly omit extinction, even though the same paper's light-curve calculation uses per-layer attenuation and places t_peak near 10^3 s for the fiducial model. This matters because the central claim is 'direct observational evidence' for r-process synthesis, and a key part of that evidence is the spectrum. If the properly attenuated spectrum does not reproduce the observed MeV continuum, the argument loses one of its three matching observables (light curve, fluence, spectrum). The proposed test is a pure recomputation with already-described ingredients, not a request for new data. I do not think this elevates the concern beyond the reader's CONDITIONAL verdict, because the paper's qualitative spectral comparison was already identified as a weakness, and the fluence/decay-rate agreement is independent of this specific issue. Hence I recommend keeping the reader's verdict unchanged while adding this specific technical condition: the authors should show the attenuated spectra or explicitly restrict the comparison to t > 3000 s where extinction is negligible.","tokens_in":21325,"tokens_out":7771,"duration_ms":91315,"concrete_test":"Recompute the synthetic spectra at t = 1000, 3000, and 12000 s using the same per-layer attenuation and thermalization treatment already used for the light curve (Hotokezaka et al. 2016; Eq. 2 with κγ = 0.1 cm^2 g^-1), summing over velocity layers, and compare the resulting emergent spectrum to the RHESSI count spectrum with its instrument response and background model from Boggs et al. (2007). If the attenuated spectrum falls below the reported flux or has a significantly different slope in the 0.1–2.5 MeV band (≳2σ), the spectral identification is not supported. If it matches within uncertainties, the present concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In §2.2, the synthetic spectra in Fig. 3 are explicitly computed 'accounting for Doppler broadening due to the ejecta expansion but excluding extinction effects.' However, the RHESSI spectrum used for comparison (Boggs et al. 2007) was measured at t ≈ 400–900 s, i.e. just before or at the gamma-ray photosphere time t_peak ≈ 10^3 s (Eq. 4). At t ≈ 1000 s, the bulk of the ejecta is still optically thick (τγ ≳ 1 for velocity layers near and below the photosphere), so the escaping spectrum is not the intrinsic line forest shown in Fig. 3. The light-curve model correctly applies per-layer attenuation via the Hotokezaka et al. (2016) treatment, but the spectral comparison does not use the same emergent spectrum. Since the only spectral information for the delayed component comes from exactly this optically thick interval, the agreement with the kBT ≈ 1.9 MeV thermal-bremsstrahlung fit is not yet established. Attenuation would preferentially suppress low-velocity inner layers, which contribute narrow line components, and Compton scattering would down-scatter MeV photons; either effect could shift or dim the predicted spectrum relative to the RHESSI data. A correct spectral comparison could still agree, but the paper currently shows only a qualitative match between an intrinsic spectrum and an observed spectrum, while the one quantitative spectral measurement is taken from the epoch where the model's own transport treatment says extinction matters most.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes that the previously unexplained delayed MeV emission observed after the 2004 giant flare of SGR 1806-20 (Mereghetti et al. 2005; Frederiks et al. 2007; Boggs et al. 2007) is the gamma-ray signature of ~10^-6 solar masses of freshly synthesized r-process nuclei ejected from the neutron star crust during the flare. The authors construct a power-law velocity-distribution ejecta model (Eq. 1), derive analytic scalings for the gamma-ray photosphere time (Eq. 4), the radioactive heating rate (Eq. 5), the decay luminosity (Eq. 6), and the fluence (Eq. 8), and augment these with a multizone SkyNet nucleosynthesis calculation and approximate gamma-ray transport following Hotokezaka et al. (2016). They report a new analysis of INTEGRAL SPI-ACS data (Appendix A) yielding a post-peak decay index delta = 1.2 +/- 0.1 that matches the r-process heating index, and they compare their fiducial model (Mej ~ 1.2e-6 solar masses, vbar = 0.15c, Ye = 0.40) with the observed light curve, total fluence ~ 9e-4 erg/cm^2, and the RHESSI spectrum parameterized as kBT ~ 1.9 MeV bremsstrahlung. On this basis they conclude that magnetar giant flares are a second confirmed r-process site and discuss implications for Galactic chemical evolution, r-process cosmic rays, and future MeV observatories such as COSI, together with a falsifiable 'nova brevis' optical counterpart prediction.","tokens_in":2557,"tokens_out":2735,"duration_ms":207484,"significance":"If the identification is correct, this is a landmark result: magnetar giant flares would become the second directly confirmed r-process site, with short delay relative to star formation, an estimated 1-10% contribution to the Galactic r-process inventory, and a novel source of r-process cosmic rays. The paper ships concrete assets: transparent analytic scalings (Eqs. 4-8) that expose the parameter dependencies; a genuinely new, model-independent archival measurement (the SPI-ACS decay index, Appendix A); detailed line-level spectra built from the JINA REACLIB and ENSDF nuclear data via SkyNet; and several falsifiable predictions (nova brevis with m_AB ~ 4 at 15 minutes; COSI detectability of line bumps from the next Galactic giant flare; late hard-X-ray emission for NuSTAR-like instruments). The authors are commendably explicit about the principal assumption (Sec. 2.1: the C24 ejection mechanism 'remains under active investigation') and about the difficulty of excluding non-thermal alternatives (Sec. 3.4).","major_comments":[{"comment":"The spectral comparison in Fig. 3 is not, as presented, a valid test of the r-process identification. The synthetic spectra are computed 'accounting for Doppler broadening due to the ejecta expansion but excluding extinction effects' (Fig. 3 caption; Sec. 2.2), whereas the RHESSI measurement used for comparison (Boggs et al. 2007) was accumulated over t = 400-900 s (Appendix A), i.e., during the rising phase at or before the model's gamma-ray photosphere time t_peak ~ 10^3 s (Eq. 4). At that epoch the bulk of the ejecta (which for beta = 6 is concentrated in low-velocity layers) still has tau_gamma > 1, so the escaping radiation is not the intrinsic total line forest shown in the figure: attenuation preferentially suppresses the narrow components from the inner layers, the visible fast outer layers have a different nucleosynthesis history than the bulk, and inelastic Compton scattering down-scatters MeV photons, the last effect being conceded in Sec. 2.2 but not applied to the plotted spectra. Because the r-process specificity of the claim (as opposed to generic radioactive decay of neutron-rich ejecta) rests on the predicted spectral character (Table 1), the authors should compare the RHESSI data with emergent spectra computed using the same per-layer attenuation employed for the light curve (Hotokezaka et al. 2016), or show quantitatively that attenuation leaves the qualitative match intact. As it stands, the abstract's claim that the spectrum 'matches' is not supported by the evidence shown.","section":"Sec. 2.2, Fig. 3"},{"comment":"The inferred r-process mass and the Galactic-yield estimates built upon it are calibration outcomes, not independent predictions, and the paper should say so. Since the fluence scales as F_gamma proportional to Mr (Eq. 8), the statement that the data reveal 'the synthesis of ~10^-6 M_sun of r-process elements' (Abstract) is essentially the measured fluence re-expressed once the r-process interpretation, the heating-rate normalization (Eqs. 5-6), the distance (8.7 +/- 1.5 kpc), and the integration window out to 10 t_peak are assumed; the peak time constrains only the combination (Mej/f_Omega)/vbar^2 (Eq. 4). Independent support for the fiducial Mej ~ 1.2e-6 M_sun is weak: the radio-afterglow modeling quoted in Sec. 1 spans 10^24.5-10^27 g, and Appendix C shows that Mej = 10^-8 M_sun would fail by orders of magnitude while remaining consistent with the radio constraints. The strongest non-circular evidence is the post-peak decay index delta = 1.2 +/- 0.1 measured from archival SPI-ACS data (Appendix A), which is independent of the model normalization and agrees with the r-process heating index alpha ~ 1.2; I recommend that the Abstract and Sec. 4 present this as the primary evidence and label Mej, vbar, f_Omega, Ye, and Mr as fitted parameters. In addition, the fluence normalization itself is derived by scaling the RHESSI fluence by SPI-ACS count ratios with acknowledged comparable statistical and systematic errors (Appendix A); these uncertainties, together with the distance error, should be propagated into the 1-10% Galactic-r-process contribution quoted in Sec. 3.1.","section":"Sec. 2.1, Eqs. (4)-(8); Sec. 3.1; Appendix A"},{"comment":"The robustness claim in the Summary, that the r-process interpretation 'is likely to remain robust' to improved understanding of the ejection mechanism because a mass greater than about 10^-7 to 10^-6 M_sun must be excavated from neutron-rich depths, is not established by the material in the paper. The mass requirement is itself derived from the observed fluence under the r-process interpretation (Eqs. 6-8), so it cannot independently vouch for that interpretation. Moreover, the nucleosynthesis outcome is not guaranteed by crustal composition alone: the r-process in these ejecta operates via alpha-rich freeze-out at Ye = 0.40, which requires specific entropy and expansion conditions (Sec. 1; Patel et al. 2025), and Appendix C shows that the synthesized mass fraction and the late-time decay slope both vary with Ye. Since Sec. 2.1 states that the ejection process 'remains under active investigation', the abstract's 'confirmed r-process sites' is premature. The correct claim at present is that the delayed MeV emission is consistent with r-process decay for a plausible but non-unique set of ejecta parameters, with the temporal decay index providing the strongest quantitative support; confirmation would come from the paper's own falsifiable predictions (COSI line bumps; nova brevis at m_AB ~ 4).","section":"Sec. 4 (Summary) and Sec. 2.1"}],"minor_comments":[{"comment":"The arXiv abstract block cites 'Patel et al. (in prep.)' while the manuscript text and references cite Patel et al. (2025) (arXiv:2501.17253); make the citation consistent and verify that the companion paper is publicly available.","section":"Abstract vs. Sec. 1"},{"comment":"There are typos in the stated integration ranges: 't0 + 120000 s' should read 't0 + 12000 s' in two places, and 'tot0 + 14000 s' is missing a space.","section":"Appendix A and Appendix B"},{"comment":"The reference list contains two separate entries for Boggs et al. (2007) that are the same paper; merge them into one.","section":"References"},{"comment":"The text states that the delayed component decays below the background level at t greater than about 3000-8000 s, whereas Fig. 2 and Appendix A use data out to 12000 s; harmonize the quoted decay-end times.","section":"Sec. 1 and Fig. 2"},{"comment":"Please report the numerical heating-rate normalization actually used in the analytic estimates; the text states that qdot_r is 'normalized based on our nucleosynthesis calculations' (Sec. 2.1) but does not give the resulting constant, so Eqs. (6)-(8) are not fully reproducible.","section":"Eqs. (5)-(6)"},{"comment":"The new decay index delta = 1.2 +/- 0.1 differs from the value delta = 0.85 quoted in Mereghetti et al. (2005), with the difference attributed to the fit window; showing the residuals of the power-law fit and the sensitivity of delta to the chosen start time (675 s) and background model would strengthen the comparison with alpha ~ 1.2.","section":"Appendix A, Fig. 4"}],"recommendation":"major_revision","confidential_remarks":"To the editor: this is an attractive and potentially important candidate identification, and the authors are unusually transparent about assumptions and alternatives (Sec. 2.1, Sec. 3.4, Appendix A). My major_revision recommendation is driven by (i) the spectral-match pillar of the abstract, which is not demonstrated at the epoch of the only spectral measurement because the plotted spectra omit extinction at t ~ t_peak, and (ii) the gap between the 'direct evidence / confirmed r-process site' headline and the fit-calibrated nature of the inferred mass. Both are fixable without new observations: recompute emergent spectra with the paper's own transport treatment, and reframe the claims with the decay index as the primary evidence. I would also ask that the editor confirm availability of the companion papers (Cehula et al. 2024; Patel et al. 2025, arXiv:2501.17253), on which the ejection and nucleosynthesis pillars rest, since the abstract still cites 'in prep.' for one of them. The single-event, parameter-calibrated nature of the identification is a scope consideration: this is more a discovery claim than a settled measurement, and the journal's standard for 'direct evidence' wording should be applied accordingly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What's actually new here: no one before connected the unexplained delayed MeV tail of the SGR 1806-20 giant flare to r-process decay emission from freshly ejected crust. That is a real and interesting step, and the paper does a lot of things well. It assembles the archival INTEGRAL, Konus-WIND, and RHESSI data, refits the decay index, checks the solar-origin hypothesis with GOES data, and generates synthetic spectra from actual nucleosynthesis network calculations. The decay index match (1.2 ± 0.1 versus the expected ~1.2) is a genuine, non-circular point in the model's favor. The paper also honestly flags that the baryon ejection mechanism remains under active investigation (Section 2.1), which is the right level of candor.\n\nThe soft spots are real but not fatal. The title and abstract claim \"direct observational evidence,\" which is too strong. The r-process mass is essentially read off the fluence through Eq. (8), and several parameters (Mej, vbar, beta, f_Omega, Ye) are adjusted to match the same light curve and fluence that set the normalization. That is a consistency check, not an independent measurement. The decay index is the main external test, and it is encouraging but not definitive.\n\nThe stress-test concern about the spectrum lands. Figure 3 compares the RHESSI measurement (t ≈ 400–900 s) to intrinsic, unattenuated synthetic spectra at t = 1000 s and later, when the ejecta are still optically thick. The light-curve model does apply per-layer attenuation, but the spectral comparison does not use the emergent spectrum. Attenuation and Compton down-scattering would preferentially remove and soften the narrow lines from slow inner layers, so the claimed spectral agreement is not yet a demonstrated test. The paper's passing mention of Compton softening does not fix that.\n\nWho is this for? Anyone working on r-process sites, magnetar giant flares, or MeV astronomy. The paper deserves a serious referee: the hypothesis is testable, the archival data are public, and the model is coherent enough to be engaged with even while the strong claim is dialed back.\n\nRecommendation: send to peer review, but require (1) softening of \"direct evidence\" to something like \"strong evidence\" or \"consistent with,\" and (2) a spectral comparison using the attenuated emergent spectrum, or at minimum a quantitative estimate of extinction/Compton effects on the predicted lines. Shipping the model inputs and code would also help future checks. I'd cite this as a candidate site with caveats, and I'd happily discuss it in our reading group.","headline":"A genuinely new and plausible identification of the 2004 SGR 1806-20 delayed MeV component as r-process decay emission, but the \"direct evidence\" claim outstrips what the light-curve fit and an unattenuated spectral comparison can support.","tokens_in":22242,"tokens_out":1607,"would_cite":true,"duration_ms":18188,"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 delayed MeV glow after the SGR 1806-20 giant flare is the radioactive signature of freshly made r-process nuclei, making magnetar giant flares a confirmed r-process site.","keywords":["r-process nucleosynthesis","magnetar giant flares","SGR 1806-20","delayed MeV emission","gamma-ray line emission","radioactive decay","galactic chemical evolution","neutron star crust ejection"],"falsifier":"Observe the next Galactic giant flare with a MeV spectrometer such as COSI: the model predicts Doppler-broadened line bumps from identified isotopes (for example $^{89}$Rb lines near 658, 1032, and 1248 keV and $^{92}$Sr near 1384 keV) appearing around $10^3$–$10^4$ seconds and decaying as $t^{-1.2}$. If those features are absent while the continuum matches, the r-process interpretation is ruled out. A targeted reanalysis of archival SGR 1900+14 1998 flare data, looking for a delayed MeV component without pulsations, would provide a second, independent test.","tokens_in":21116,"feed_emoji":"🌟","tokens_out":8902,"duration_ms":81332,"temperature":0.7,"pith_summary":"This paper argues that a previously unexplained hard gamma-ray glow seen minutes after the December 2004 giant flare of the magnetar SGR 1806-20 was the radioactive afterglow of heavy elements made during the flare. The authors show that the observed rise, decay, fluence, and MeV-peaked spectrum match predictions for gamma-ray line emission from roughly $10^{-6}\\,M_\\odot$ of freshly synthesized r-process ejecta. If correct, magnetar giant flares become the second directly confirmed astrophysical site of r-process nucleosynthesis, after neutron star mergers, and a source that can enrich the earliest metal-poor stars. The case matters because it would turn a 20-year-old orphan signal into one of the few direct observations of heavy-element synthesis in action.","feed_headline":"A 2004 magnetar flare forged heavy elements, new analysis shows","feed_subtitle":"The unexplained MeV glow after SGR 1806-20 matches radioactive decay of freshly made r-process nuclei.","key_machinery":"The mechanism is the radioactive decay chain of freshly synthesized r-process nuclei. Each $\\beta$-decaying species emits a set of nuclear gamma-ray lines, and the high ejecta velocities $v \\gtrsim 0.1c$ Doppler-broaden the lines into a quasi-continuous spectrum peaking near 1 MeV, so the discrete atomic carriers are hidden but their collective decay power is not. The gamma-ray luminosity follows $\\dot{Q}_\\gamma = M_r \\epsilon_\\gamma \\dot{q}_r(t)$ with $\\dot{q}_r(t) \\approx 5\\times10^{12}\\,(t/10^3\\,\\mathrm{s})^{-1.2}\\,\\mathrm{erg\\,s^{-1}\\,g^{-1}}$, and the ejecta become transparent on a timescale $t_{\\rm peak}\\sim10^3\\,\\mathrm{s}$ set by the opacity, mass, and velocity. The r-process itself operates through the $\\alpha$-rich freeze-out mechanism, which allows neutron capture to build heavy nuclei even at the moderate electron fraction $Y_e\\approx0.4$ expected in ejected neutron-star crust.","core_discovery":"The central claim is that the delayed MeV emission component detected by INTEGRAL, Konus-WIND, and RHESSI in the aftermath of the SGR 1806-20 giant flare—rising to a broad peak near 600–800 seconds and then decaying roughly as $t^{-1.2}$—is gamma-ray line emission from freshly synthesized radioactive r-process nuclei expelled during the flare. Using an ejecta model with mass $\\sim 1.2\\times10^{-6}\\,M_\\odot$, r-process mass $\\sim 7\\times10^{-7}\\,M_\\odot$, minimum velocity $0.15c$, covering fraction $3/4$, and electron fraction $Y_e = 0.40$, the paper reproduces the light curve, the total fluence of about $9\\times10^{-4}\\,\\mathrm{erg\\,cm^{-2}}$, and the hard spectrum peaking near 1 MeV. It concludes that this is direct observational evidence for the synthesis of $\\sim 10^{-6}\\,M_\\odot$ of r-process elements and that magnetar giant flares are confirmed r-process sites.","pith_inferences":["If the interpretation is right, the 1998 giant flare from SGR 1900+14 should have produced a similar delayed MeV component; a modern reanalysis of archival Konus-WIND data might recover it despite the high solar background at that time.","Resolving individual lines in a future flare would test the predicted abundance pattern (first-peak nuclei around $A\\sim90$ dominating the MeV output); absence of the predicted bumps would falsify the r-process identification even if the continuum matched.","The 1–10% Galactic contribution is a lower bound: if less energetic flares eject comparable r-process mass, magnetars could be a substantially larger share of the Galactic r-process budget than the fiducial estimate."],"forward_implications":["Magnetar giant flares join neutron star mergers as directly confirmed r-process sites, contributing at least about 1–10% of the Galactic r-process inventory.","Because magnetars form promptly with star formation, their flares can enrich very metal-poor stars and help explain the early iron/r-process correlation in the Galactic halo.","The same ejecta, when shocked by the surrounding medium, can accelerate r-process nuclei to cosmic-ray energies, making magnetars potentially major sources of heavy cosmic rays.","The predicted Doppler-broadened decay lines give a concrete spectral target for COSI and next-generation MeV telescopes during the next Galactic giant flare.","The model predicts a kilonova-like optical/UV transient, a 'nova brevis' peaking at about magnitude 4 roughly 15 minutes after the flare, from the same ejecta."],"supporting_citations":[{"why":"Hydrodynamical model showing that a giant flare can shock-heat and eject neutron-star crust at $v \\gtrsim 0.1c$ with conditions promising for an r-process.","marker":"C24"},{"why":"SkyNet nuclear reaction network calculations confirming the r-process in giant-flare ejecta and providing the abundances and heating rates used to predict gamma-ray emission.","marker":"Patel et al. (2025)"},{"why":"INTEGRAL SPI-ACS discovery of the delayed MeV emission component starting near 400 seconds after the flare.","marker":"Mereghetti et al. (2005)"},{"why":"Konus-WIND detection of the delayed component and the late-time fluence measurement in the 80–750 keV band.","marker":"Frederiks et al. (2007)"},{"why":"RHESSI spectrum of the late MeV component, modeled as thermal bremsstrahlung with $k_BT\\approx1.9$ MeV, used to estimate total fluence and rule out pulsations.","marker":"Boggs et al. (2007)"},{"why":"Provides the radioactive heating-rate formalism and power-law decay index used to model r-process decay emission.","marker":"Metzger et al. (2010)"},{"why":"Supplies the gamma-ray thermalization/attenuation treatment and Doppler-broadening prescription used to compute light curves and spectra.","marker":"Hotokezaka et al. (2016)"},{"why":"Quantifies the gamma-ray opacity at ~0.1–1 MeV and the fraction of beta-decay energy released as gamma-rays.","marker":"Barnes et al. (2016)"},{"why":"Determines the distance to SGR 1806-20 used to convert predicted gamma-ray energy into observed fluence.","marker":"Bibby et al. (2008)"}],"fun_headline_variants":["Magnetar flare's MeV glow reveals heavy element forge","Delayed MeV flash ties magnetar flare to r-process origin","SGR 1806-20 afterglow confirms heavy element synthesis","Giant flare emission cracked: r-process nuclei detected"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The identification rests on the assumption that the giant flare really ejected about a millionth of a solar mass of neutron-star crust at speeds near a tenth of the speed of light with an electron fraction near 0.4; the ejection mechanism is complex, not directly observed, and still under active investigation.","fun_headline_variants_meta":{"raw":{"variants":["Magnetar flare's MeV glow reveals heavy element forge","Delayed MeV flash ties magnetar flare to r-process origin","SGR 1806-20 afterglow confirms heavy element synthesis","Giant flare emission cracked: r-process nuclei detected"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000248,"raw_usage":{"total_tokens":1635,"prompt_tokens":1119,"completion_tokens":516,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":735,"completion_tokens_details":{"reasoning_tokens":446}},"tokens_in":735,"tokens_out":516,"duration_ms":5038,"temperature":1.0,"reasoning_tokens":446,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:09:44.855708+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Observe the next Galactic giant flare with a MeV spectrometer such as COSI: the model predicts Doppler-broadened line bumps from identified isotopes (for example $^{89}$Rb lines near 658, 1032, and 1248 keV and $^{92}$Sr near 1384 keV) appearing around $10^3$–$10^4$ seconds and decaying as $t^{-1.2}$. If those features are absent while the continuum matches, the r-process interpretation is ruled out. A targeted reanalysis of archival SGR 1900+14 1998 flare data, looking for a delayed MeV component without pulsations, would provide a second, independent test.","supporting_citations":[],"review_version":1}