Shifts in magnetic mineral assemblages support ocean deoxygenation before the end-Permian mass extinction

The end-Permian mass extinction (EPME) eliminated up to ~81% of marine species over a geologically brief interval and is widely linked to widespread marine deoxygenation. Although numerous short-term redox reconstructions across the EPME exist (e.g., biomarkers, iron speciation, isotopic proxies), the longer-term redox evolution leading into the crisis remains unclear. This study tests the hypothesis that ocean deoxygenation began well before the EPME by integrating mineral magnetic analyses with cerium anomaly redox proxies through the Changhsingian into the Permian–Triassic boundary at the Meishan GSSP, aiming to establish a high-resolution, independently constrained redox history and its relation to volcanism, continental weathering, and ecosystem collapse and recovery.

Prior work implicates oceanic anoxia/dysoxia as a principal driver of the EPME based on biomarkers, Fe-speciation, and isotopic systems (U, Tl, Ni, Cr), but these proxies are calibrated to modern settings and often emphasize short-term events, leaving long-term trends uncertain. Magnetic minerals such as magnetite and hematite respond sensitively to redox state: hematite forms under oxidizing conditions and has high coercivity, whereas magnetite has high saturation magnetization and lower coercivity. Previous studies have linked magnetic mineralogy to redox in lacustrine and marine sediments (e.g., Paleolake Nonesuch, eastern Mediterranean), showing reductive dissolution can lower magnetic mineral abundance and alter grain size. At Meishan, extensive prior geochronology, geochemistry, cyclostratigraphy, biostratigraphy, and magnetostratigraphy provide a robust framework for correlation. However, a detailed magnetic mineralogical record tied to geochemical redox tracers across the late Permian–early Triassic had not been fully exploited to assess multi-Myr redox evolution.

• Site and sampling: 264 shallow-marine carbonate samples were collected from the Meishan section (sections D and C), spanning the topmost Lungtan Fm., Changxing Fm., and Yinkeng Fm. Average spacing 10–15 cm. Sample ages were modeled using Bchron based on weighted mean radioisotope ages and stratigraphic positions.
• Sample preparation: Chips (~5 mm) avoiding weathered surfaces and veins were powdered with an agate mortar.
• Rock magnetic measurements: Room-temperature hysteresis loops, IRM acquisition and DC demagnetization using a VSM (sensitivity 5×10⁻¹⁰ Am²). Parameters Ms, Mrs, Hc, Hcr obtained. IRM acquisition curves smoothed (Savitzky–Golay) and decomposed via coercivity spectrum analysis to quantify magnetic components. Magnetic susceptibility (χ) and high-T χ–T measured with an AGICO MFK1-FA. ARM imparted (0.05 mT DC over 100 mT AF) and measured with a 2G-760 magnetometer. SIRM imposed at 1 T; backfield at 0.3 T. S-ratio₀.₃T = 0.5×(SIRM₁T − IRM₀.₃T)/SIRM₁T used to track low- vs high-coercivity components. Thermal demagnetization and low-temperature experiments probed mineral identity (e.g., Verwey transition for magnetite at ~125 K).
• SEM/EDXS: Magnetic minerals concentrated from carbonates by weak acid buffering (acetic/sodium acetate, pH ~4) and observed via FE-SEM (Zeiss Gemini 450) with BSE imaging and EDXS at 15 kV, 2 nA, WD ~10 mm.
• Cerium anomalies: 41 bulk samples were dissolved (HF/HNO₃ protocols) and analyzed by ICP-MS (Agilent 7500). REE normalized to PAAS. Ce anomalies computed and screened. To minimize diagenetic or detrital contamination influences, samples were filtered using REE pattern criteria, Y/Ho vs trace-element correlations (Ti, Th, Cu, Sc), and Eu anomaly checks; 17 of 41 were excluded, leaving 24 that preserve primary REE–Ce signatures. Bulk Ce data were compared with published in-situ conodont Ce records.
• Stratigraphic division and proxies: Rock magnetic trends define four intervals (I–IV) with boundaries at ~252.8 Ma, ~251.9 Ma, and ~251.2 Ma. Magnetic grain size assessed via ARM/SIRM; concentration via χ, ARM, and Ms; coercivity and S-ratio track mineral assemblage. These were integrated with Ce-anomaly redox inference, iron speciation and published weathering proxies (Sr flux modeling) and volcanism timelines.

• Magnetic mineral assemblage shift at ~252.8 Ma: Interval I (ca. 253.8–252.8 Ma) contains mixed high-coercivity hematite and low-coercivity magnetite; from Interval II onward, magnetite dominates, evidenced by S-ratio₀.₃T rising sharply to near-constant high values (avg ~0.98), stable Hcr, and IRM decomposition/SEM showing granular magnetite.
• Timing of deoxygenation: The hematite-to-magnetite dominance transition at ~252.8 Ma, together with Ce-anomaly trajectories, indicates an oxic/dysoxic shift and the onset of deoxygenation ~0.9 Myr before the EPME.
• Redox state from Ce anomalies: Bulk-sample Ce anomalies range from −0.37 to 0.00. Intervals I–II show values < −0.1 (oxic–dysoxic), whereas Intervals III–IV exceed −0.1, indicating anoxia. This agrees with published conodont Ce data and iron speciation (FeHR/Fet > 0.38, FePY/FeHR < 0.7; ferruginous, nonsulfidic conditions) for Meishan.
• Grain size and concentration trends: In Interval I, ARM/SIRM increases (grain size decrease) with variable concentration; in Interval II, ARM/SIRM trends downward (grain size increase). In Interval III, χ, ARM, and Ms increase then decrease, indicating a transient rise and fall of detrital magnetite concentration.
• Enhanced weathering: Detrital magnetite abundance (and mudstone/clay content) increases from the II–III boundary, consistent with Al variations and Sr-isotope-based modeling that indicates Early Triassic continental weathering rates intensified by >1.9× relative to Late Permian. Weathering appears to peak at ~251.5 Ma, coincident with waning Siberian Trap and arc volcanism and a positive δ66Zn shift.
• Early Triassic ferruginous dysoxia: Post-EPME strata dominated by magnetite reflect ferruginous dysoxic (nonsulfidic) water-column conditions in the earliest Triassic.
• Global context: Independent redox proxies elsewhere (e.g., Tl and Ni isotopes) also suggest deoxygenation began 0.7–1.0 Myr before the EPME. Biodiversity records at Meishan and broader regions show early declines near the I–II boundary, consistent with protracted environmental deterioration preceding the mass extinction.

The integration of magnetic mineralogy and Ce anomalies demonstrates that deoxygenation at Meishan initiated roughly 0.9 Myr before the EPME and progressed from oxic conditions (hematite plus magnetite preserved) to dysoxia (magnetite-dominant, nonsulfidic) and then to ferruginous anoxia in the earliest Triassic. The magnetic transition is unlikely to reflect provenance shifts given the lack of tectonic disturbance, continuous deposition, and biostratigraphic/chemostratigraphic continuity. Instead, it records redox evolution corroborated by Ce anomalies, iron speciation, and independent geochemical constraints. The temporal alignment between the onset of deoxygenation and the initiation of Siberian plume activity (~252.8 Ma) and prior arc volcanism suggests volcanic emissions (CO₂, SO₂, thermogenic gases) drove climate warming, oxygen loss, and enhanced chemical weathering. Elevated weathering increased nutrient fluxes (Fe, PO₄), promoting eutrophication and sustaining dysoxia/anoxia. The observed peak in detrital magnetite and weathering proxies around 251.5 Ma coincides with waning volcanism, after which magnetite concentrations decreased toward pre-extinction levels. These findings imply that prolonged deoxygenation primed ecosystems for collapse, contributed to the severity of the EPME, and hindered early Triassic recovery through sustained ferruginous conditions and elevated sediment/nutrient fluxes.

Magnetic mineral assemblages and cerium anomalies from the Meishan GSSP reveal a protracted redox deterioration beginning ~0.9 Myr before the end-Permian mass extinction. A shift from hematite–magnetite mixtures to magnetite dominance at ~252.8 Ma marks an oxic-to-dysoxic transition, with ferruginous anoxia prevailing into the earliest Triassic. Enhanced continental weathering, closely tied to volcanic activity, increased detrital magnetite flux and nutrient delivery, reinforcing deoxygenation and delaying biotic recovery. These results highlight magnetic mineralogy—as integrated with Ce anomalies and other proxies—as a powerful tracer of long-term ocean redox evolution. Potential future work includes: expanding high-resolution, multi-proxy magnetic–geochemical records across diverse paleobasins; refining age models to better synchronize redox and volcanic events; quantifying contributions of authigenic versus detrital magnetic phases; and coupling data with Earth-system models to test mechanistic links among volcanism, weathering, nutrient cycling, and redox states.

• Potential diagenetic alteration of carbonate REE–Ce signatures necessitated stringent screening; despite filtering, residual diagenetic effects cannot be entirely excluded.
• The χ–T signal in limestones was weak, limiting direct high-temperature mineral identification in some intervals.
• Onset timings for enhanced weathering differ slightly among proxies (e.g., Sr modeling vs magnetic data), potentially due to sampling-density differences and proxy sensitivities.
• Correlation of the precise onset of deoxygenation between Meishan and other sections is limited by geochronologic uncertainties and local environmental variability.
• While provenance shifts are argued against based on stratigraphic and tectonic evidence, unrecognized subtle changes in sediment sources cannot be completely ruled out.