Introduction

The end-Permian mass extinction (EPME) was marked by the loss of up to 81% of marine species in a short time interval of 61 ± 48 kyr and is considered the most severe biocrisis in Earth history1,2,3. Unraveling the mechanisms for this biocrisis is crucial for understanding the relationships between environmental changes and biological macroevolution. Numerous studies suggest that widespread marine anoxia likely drove the marine mass extinction4,5,6,7. A shift toward more dysoxic conditions at the EPME has been implicated in most global oceanic settings based on biomarkers8, iron speciation9,10, and geochemical and isotopic proxies4,6,7,11. Depending on their empirical calibration to modern marine sediments, these redox proxies provide understanding of significant insights into short-term redox history across the EPME. However, the long-term redox conditions remain undetermined.

Magnetite (Fe2+[Fe3+]2O4) and hematite ([Fe3+]2O3) are commonly occurring magnetic minerals in sediments and rocks12. Magnetite has high saturation magnetization (Ms) that extends up to 92 Am2/kg and low coercivity (Hc, ~10’s of mT) at room temperature13. Hematite forms under oxidizing conditions and has a much lower Ms (0.4 Am2/kg) than magnetite. Meanwhile, its Hc can be zero for superparamagnetic particles and up to 1000 s of mT for high coercivity pigments13. Previous studies indicated that magnetic mineral variations of terrestrial sediments (e.g., loess, red beds) are contributed to paleoclimatic patterns relating to rainfall and temperature14,15. Besides, a recent analysis of the 1.1 billion-year-old Paleolake Nonesuch sediments in North America has shown that redox processes are related to magnetite and hematite variations16. Likewise, the magnetic properties of eastern Mediterranean sediments provide evidence for a causal relationship between fluctuating redox conditions and the relative magnetic mineral assemblage17. Reductive dissolution may result in low magnetic mineral abundances and decreased particle size, leading to weakened magnetization and altered magnetic mineral assemblages. From these investigations, it is apparent that in some situation, accurately identifying the magnetic mineralogy of sediments may provide a useful tracer of paleo-oceanic redox conditions.

The renowned Meishan section (31°4.644′N, 119°42.510′E) was situated in the eastern Paleo-Tethys during the end-Permian to early Triassic18 (Fig. 1a). It has been well studied for geochronology1,19, multiple geochemical proxies8,20,21, cyclostratigraphy22, high-resolution biostratigraphy23, and magnetostratigraphy24, which provide a robust framework for correlation to other marine sections. In this study, we integrate mineral magnetism and microscopy to examine magnetic minerals combined with geochemical proxies through the Changhsingian and across the Permian-Triassic boundary (PTB) at the Meishan section to assess long-term redox conditions, associated with end-Permian deterioration and ecosystem collapse of a shallow marine environment and its early Triassic recovery.

Fig. 1: Locality map and end-Permian stratigraphy of the study section.
figure 1

a Paleogeographic location of the Meishan section at the end-Permian (ca. 252 Ma)modified from ref.69. b Lithological column for the Meishan section and sampling range (green line). LIP, large igneous province. PTB, Permian-Triassic boundary. GSSP, Global Boundary Stratotype Section and Point. Wu., Wuchiapingian. Lun., Lungtan Formation. Fm., Formation.

Full size image

Results

Magnetic mineral components

Hysteresis and isothermal remanent magnetization (IRM) results indicate the presence of mainly low- and high-coercivity magnetic minerals (Supplementary Fig. S1). Wasp-waisted loops indicate the presence of at least two magnetic phases (Supplementary Fig. S1c)25. Wide-open hysteresis loops likely indicate the presence of stable single-domain hematite (e.g., Supplementary Fig. S1i). Coercivity spectrum analysis of IRM acquisition curves26,27 also reveals the presence of one or two components in limestone samples (Fig. 2): Component 1 with B1/2 of 426–501 mT and a dispersion parameter (DP) of 0.28–0.33; Component 2 with B1/2 of 31.6–55.0 mT and a DP of 0.35–0.60 (Supplementary Table S1). In addition, the relative proportion of each component changes synchronously with the S-ratio0.3T that calculated based on 0.5 × (SIRM1T - IRM-0.3T)/SIRM1T28 (see the details in the Samples and Methods section). S-ratio0.3T values of > 0.97 indicate that Component 2 is the primary magnetic carrier, which has a granular morphology (Fig. 2d-f). Lower S-ratio values indicate an increasing proportion of Component 1 that consists of tabular Fe-oxide aggregates (Fig. 2a). Although large particles are often easily observed, experimentally determined magnetization and coercivity estimates reveal two distinct magnetic phases within the selected samples (Fig. 2 and Supplementary Fig. S1). The χ–T signal is weak in limestone samples, which prevents clear magnetic mineral identification.

Fig. 2: Mineralogy analysis results.
figure 2

Bulk coercivity spectrum, S-ratio0.3T and backscatter electron (BSE) micrographs of selected samples from the Meishan section, China. S-ratio0.3T = 0.5 × (SIRM1T - IRM-0.3T)/SIRM1T28. Samples shown in Fig. 2a-c are from Interval I, whereas those displayed in Fig. 2c-f belong to the Interval II-IV, respectively. These sample positions have been marked in Fig. 3f using red dot symbols.

Full size image

From rock magnetic and SEM analyses, magnetic mineral assemblages in the Meishan section are dominated by high-coercivity component 1 (e.g., hematite, goethite) and low-coercivity component 2 (e.g., magnetite, maghemite). However, thermal demagnetization results for limestone samples indicate no noticeable remanent magnetization decrease at 80–150 °C24, which excludes high-coercivity goethite and suggests the presence of hematite12,29,30. Usually, low-coercivity maghemite, which is cubic with inverse spinel crystal structures31, coexists with magnetite in sediments. Nevertheless, magnetite is identified from low-temperature magnetic experiments through an apparent Verwey transition at 125 K for limestone samples24, which is the main low-coercivity magnetic mineral (component 2). Thus, combining the rock magnetic and SEM results, low-coercivity magnetite and high-coercivity hematite are interpreted to be the main magnetic carriers in the Meishan section.

Magnetic Property Variations

Following magnetic parameter trends (Supplementary Table S2), the section can be divided into four intervals (Intervals I - IV in ascending order) with boundaries at ca. 252.8 Ma, ca. 251.9 Ma and ca. 251.2 Ma, respectively (Fig. 3). Anhysteretic remanent magnetization (ARM), Ms and Hcr variations suggest that the ferrimagnetic particle concentration fluctuates considerably in Interval I (ca. 253.8 – 252.8 Ma) (Fig. 3b-d). However, χ values (Fig. 3a) are nearly constant (-0.46 × 10-8 - 1.20 × 10-8 m3/kg), probably due to dominant diamagnetic components (e.g., carbonate minerals). Moreover, the ARM to saturation isothermal remanent magnetization (ARM/SIRM1T) ratio oscillates with an increasing trend in this interval, which reflects a magnetic grain size decrease (Fig. 3e). Additionally, S-ratio0.3T (0.26 - 1.00) fluctuations are associated with synchronous Hcr (17.6 – 535.0 mT) and concentration-dependent magnetic parameter changes (Fig. 3c-f). Based on coercivity spectrum analysis (Fig. 2a-c), these variations are interpreted to indicate hematite and magnetite content changes in Interval I.

Fig. 3: Rock magnetic parameters and geochemical proxies for the Meishan section.
figure 3

a–f Rock magnetic parameters. Red dots in (f) indicate the samples selected for coercivity spectrum analysis in Fig. 2. g Cerium anomalies (ΩCe), including bulk samples (blue dots, this study) and conodont samples (blue line)32. h Ratio of riverine to mantle Sr flux (FR/FM) to indicate weathering35. Conodont zones are from Yuan et al. 23. The magnetic polarity sequence is from M. Zhang et al. 24. χ, magnetic susceptibility. ARM, anhysteretic remanent magnetization. Ms, saturation magnetization. Hcr, coercivity of remanence. SIRM1T, saturation isothermal remanent magnetization. S-ratio0.3T = 0.5 × (SIRM1T - IRM-0.3T)/SIRM1T28. Blue horizontal dashed line is the Permian-Triassic boundary (PTB); the red horizontal dashed line is the onset of the end-Permian mass extinction (EPME).

Full size image

The S-ratio0.3T increases sharply at the end of Interval I (ca. 252.8 Ma) and then remains largely invariable with an average value of 0.98 to the top of the section; Hcr also does not vary much (Fig. 3d, f). These parameters, coupled with IRM decomposition results and SEM observations (Fig. 2d-f), indicate that magnetite is the dominant magnetic carrier from Interval II to Interval IV. Furthermore, the ARM/SIRM1T ratio fluctuates with a decreasing trend, which indicates a magnetic mineral grain size increase in Interval II (Fig. 3e). In addition, χ, ARM and Ms values rise gradually from the bottom of Interval III (ca. 251.9 Ma) and then decline in the middle-upper part of Interval III and return to stable values in Interval IV (ca. 251.2 Ma) (Fig. 3a-c). These oscillations indicate that the magnetite concentration first increases and then decreases in Interval III.

Cerium anomaly (ΩCe)

ΩCe values of bulk sample in the Meishan section range from −0.37 to 0.00 (Supplementary Table S3). They are less than -0.1 in Intervals I-II, indicating oxic-dysoxic conditions. ΩCe values fall into the anoxic range (> -0.1) in Intervals III-IV. The ΩCe trend of bulk samples is consistent with published conodont ΩCe data32 (Fig. 3g).

Discussion

Enhancement of chemical weathering

Magnetite in marine sediments generally originates from detrital inputs, bacterial activity (extracellular and intracellular), and authigenesis during diagenesis33. The different degrees of roundness and abrasion of the studied particles (Fig. 2) exclude an authigenic origin and indicate that the magnetite had a detrital origin. Also, paleomagnetic investigations confirm that magnetite at the Meishan section was deposited during limestone deposition24. Thus, magnetite concentration variations correspond to changes in continental silicate weathering and erosion. The lowest part of Interval III is dominated by mudstone. It is reasonable to expect that siliciclastic rocks would contain more abundant magnetite particles that were mostly reworked and transported. Rock magnetic parameters indicate that detrital magnetite concentrations started to increase at the Interval II-III boundary in the Meishan section (Fig. 3a-c), coinciding with a lithofacies change to high clay contents34. Meanwhile, the increasing of detrital magnetite content is consistent with major element Al variations at the Meishan10. Therefore, these magnetic minerals are related to lithologic types and are controlled by continental weathering. Notably, published lithium isotope data from the Meishan section reveal that the rate of chemical weathering rapidly increased ≥ 0.3 Ma before the main extinction event21. 87Sr/86Sr modeling for Meishan supports the interpretation that gradually enhanced weathering started from the bottom of Interval II, and at the early Triassic continental weathering rates were intensified by a factor of >1.9 compared to those of the Late Permian (Fig. 3h)35. These slight differences in the onset of enhanced weathering may have been caused by different sampling densities, which need further investigation. Additionally, our data suggest that chemical weathering reached a maximum at ca. 251.5 Ma (Fig. 3a-c), synchronous with the waning of volcanic activity 36,37 and with a positive zinc isotope (δ66Zn) shift20. These processes probably delayed oceanic ecosystem recovery in the early Triassic after the mass extinction.

Latest permian qceanic anoxia and implications

Experimentally determined magnetization and coercivity estimates for samples indicated that magnetite and hematite were the main magnetic minerals before ca. 252.8 Ma (Interval I) and that magnetite then became dominant. The observed shifts in magnetic mineralogy were attributed to redox conditions or provenance dynamics. Notably, previous studies indicated that the Meishan section has not undergone later tectonic disturbances38. In addition, biostratigraphic and chemostratigraphic correlations corroborated the continuity of sedimentary deposition in the Meishan strata23,37, denoting a notable absence of sedimentary hiatuses. These lines of evidence collectively suggest the stability in provenance within the Meishan section. Thus, the shifts in magnetic mineral assemblage are interpreted as being caused by the redox fluctuation, as indicated by the cerium-anomaly (ΩCe = CeN/(LaN0.5 × PrN0.5)-1) geochemical proxy.

The ΩCe value, one of the most useful geochemical proxies available for gaining insights into local redox conditions39, has been used for the Meishan section (Fig. 4g) to infer water-column anoxia throughout the late-Permian to early-Triassic32. The potential diagenetic alternation is a matter of significant concern across all carbonate samples. Luckily, lines of evidence in accordance with previous studies indicate that rare earth element (REE) in carbonate rocks exhibit relative immobility during fluid-rock interaction. Additionally, they can uphold a seawater signature even after dolomitization, which predominant diagenetic process influencing carbonate rocks40,41,42.To limit the potential impact of post-depositional alternation and non-carbonate phase contamination on ΩCe, we applied filtering criteria to ΩCe data based on individual REE patterns and the correlations between of Y/Ho ratio and trace elements (e.g., Ti, Th, Cu, Sc) associated with diagenetic alternation43,44. Twelve samples with high Ti and Th contents or low Y/Ho ratio and five samples showed negative Eu anomalies indicate the influence of detrital contaminations or hydrothermal fluid-rock interactions43,44,45 and were ruled out. Out of forty-one samples, twenty-four samples preserved the primary REE pattern and Ce anomalies (Supplementary discussion and Figs. S3–4). Both our new and published ΩCe data are generally inferred the Intervals I-II fall in an equivocal zone that could have been deposited under an oxygenated or dysoxic conditions32. The magnetic mineral assemblage changes from hematite mixed with magnetite to magnetite across the Interval I-II boundary (ca. 252.8 Ma) (Fig. 3), which indicates an oxic-dysoxic transition. Sediments in Interval I were oxic, which allowed preservation of detrital hematite and magnetite that is consistent with the magnetic observations. Also, the magnetic mineral in Interval II was dominated by magnetite, which indicated a non-sulfidic dysoxic condition33. Multiple S-isotopic evidence (δ34S and Δ33S) also reflects that a lower degrees of sulphate reduction only episodically occurred at the top of Interval II (Beds 22-28)5. 87Sr/86Sr modeling for the Meishan section implies that gradually enhanced weathering started from Interval II (Fig. 3h), which led to the eutrophication and then resulted in dysoxic condition. From Interval III to Interval IV, the ΩCe values are higher than -0.132, indicating an anoxic depositional environment (Fig. 3g). The anoxic condition is also evidenced by iron speciation10 and biomarkers8 in the Meishan section. The published iron speciation indicate nonsulfidic ferruginous conditions in the shallow water column based on the combination of FeT > 0.5%, FeHR/FeT > 0.38 and Fepy/FeHR < 0.710. Ferruginous dysoxia prevailed but sulfide availability was restricted and was insufficient to sulfidize available reactive irons in Interval III-IV10,33.

Fig. 4: Compilation of redox proxy data and geological events across the end-Permian mass extinction (EPME).
figure 4

a Redox proxies, including magnetic minerals (this study), iron speciation10, sulfur isotopes5, thallium isotopes7 and nickel isotopes46. b Volcanic activity: arc volcanism37,54, coal fly ash55, plume melting56 and Siberian Traps36. c Species diversification rate48. d Carbon isotopes (δ13Ccarb) from bulk carbonates8.

Full size image

The observed long-term magnetic mineral shifts together with ΩCe values suggest that the onset of deoxygenation probably occurred before the EPME by ~0.9 Myr at the Meishan section (Fig. 4a), which is consistent with the published global geochemical redox proxies and biodiversity changes. For example, thallium isotope (ε205Tl) profiles for the Gujo-Hachiman and Ubara sections range from -4.0 to -2.0, suggesting expansion of seafloor dysoxia by at least 1.0 Myr before the EPME7. Additionally, the nickel isotope data (δ60Ni) from Buchanan Lake indicate that deoxygenation occurred before the EPME by ~0.7 Myr46 (Fig. 4a). The different onset age estimates for oceanic deoxygenation may be due to poor geochronologic constraints in these sections or to environmental difference at different locations. As a result, it is difficult to correlate precisely between the Meishan and other sections. However, despite the differences, these proxies were all potentially indicative of oceanic deoxygenation before the EPME. Also, the published paleontological data from the Meishan section indicate a species diversity decrease near the Interval I-II boundary47. Moreover, a similar biodiversification declined across South China and the northern peri-Gondwanan region suggested environmental deterioration ~1.2 Myr before the EPME48 (Fig. 4c). Carbon isotopes (δ13Ccarb) also show a subtle excursion8 (Fig. 4d). Such deoxygenation events occurred over the prolonged end-Guadalupian biocrisis resulting in a gradual diversity decrease and selective extinction49. Interestingly, recent data from the terrestrial realm also indicated the extirpation of late Permian plant communities occurred some hundreds of thousands of years before the end-Permian50,51,52,53.

Previous geochronological investigations have constrained that extensive arc volcanism occurred before 253 ± 0.5 Ma37,54 and that Siberian plume melting launched at ca. 252.8 Ma55,56 (Fig. 4b). Oceanic deoxygenation is temporally consistent with this volcanism (Fig. 4). The synchronization indicates that early volcanic activity, which released volcanic volatiles and thermogenic gases56, potentially contributed to the onset of ocean deoxygenation. Following major eruption sequences36,37,57, the ocean became ferruginous anoxic.

While the existence of anoxic conditions before EPME is disputed, a combination of the published proxies provides conceptual insights in ocean redox evolution with synchronous volcanism during the late Permian Changhsingian (Fig. 5). Based on the presumed pathways, initial development of rising magma and plume melting at around 252.8 Ma probably contributed to the onset of oceanic deoxygenation54,56 and led to upwelling of deep dysoxic water and allowed eukaryotic algae to proliferate5,58, which developed a dysoxia condition in shallow marine waters. Massive volcanism also outgassed large amounts of CO2 and SO2 aerosols into the atmosphere59, contributing to global hyperwarming and accompanying oceanic O2 loss60. Moreover, this process caused rapid enhancement of chemical weathering from hot acid rain on freshly erupted lava. Enhanced chemical weathering brought soluble ions (e.g., Fe2+, PO43-, Ca2+ and Mg2+) that increased nutrient fluxes to the ocean, leading to eutrophication61. Together, these processes led to widespread ocean dysoxia, which resulted in the EPME and delayed recovery of oceanic ecosystems.

Fig. 5: A Conceptual model for the end-Permian mass extinction (EPME) linked to environmental and climatic perturbations.
figure 5

The initiation of oceanic deoxygenation was triggered by the ascent of magma and mantle plume melting. Subsequent episodes of mass volcanism eruptions (e.g., Siberian Traps) and their relevant environmental impacts (e.g., global hyper-warming, oceanic O2 escape, eutrophication) contributed to widespread oceanic deoxygenation, ultimately resulting in the EPME.

Full size image

Conclusion

Long-term magnetic mineral variations in the Meishan GSSP section may reveal a complex redox history during the EPME. Magnetic mineral assemblage variations record an onset of oceanic deoxygenation ~0.9 Myr before the EPME. The initial development of dysoxic water masses was detrimental to some life forms, which turned to widespread end-Permian anoxia, resulting in catastrophic extinction. This premature environmental deterioration may be considered a prelude to the impending mass extinction. Moreover, magnetite concentrations reveal that chemical weathering increase gradually over the EPME interval and decreased with the waning of volcanic activity. These records indicate that oceanic anoxia coupled with climatic changes played a crucial role in mass mortality and post-extinction recovery.

Samples and methods

Samples

Two hundred sixty-four shallow-marine limestone samples were collected from the Meishan section, including 186 samples from the topmost Lungtan Formation to the Changxing Formation in Meishan section D and 78 samples from the Yinkeng Formation in Meishan section C (Fig. 1b). The average sampling interval was 10–15 cm, and all sample ages were computed using the Bchron software package62,63 based on weighted mean radioisotope age estimates and their relative stratigraphic positions from the Meishan section1,19. Samples were broken into pieces (diameter ~5 mm) by hammer and were selected to avoid weathered surfaces and carbonate veins. Finally, the pieces were ground into powder using an agate mortar and pestle for magnetic measurements.

Rock magnetic measurements

Room-temperature hysteresis loops, isothermal remanent magnetization (IRM) acquisition curves and DC demagnetizations of IRM were measured using a Princeton Measurements Corporation vibrating sample magnetometer (VSM3900, sensitivity: 5 × 10-10 Am2). Hysteresis parameters, such as Ms, saturation remanent magnetization (Mr), Hc and the coercivity of remanence (Hcr), were obtained. IRM acquisition curves were smoothed using the Savitzky-Golay algorithm before coercivity spectrum analysis26,27. Magnetic susceptibility (χ) and high-temperature magnetic susceptibility (χ–T) were measured using an MFK1-FA Kappabridge system (AGICO Ltd., Brno). An anhysteretic remanent magnetization (ARM) was imparted to samples in a 0.05 mT DC field superimposed on a peak alternating field of 100 mT and measured with a 2G-Enterprises Model 760 cryogenic magnetometer (sensitivity: 5 × 10-11 Am2). A saturation IRM (SIRM) was imparted in a 1 T applied field using an MMPM10 pulse magnetizer, followed by the application of a 0.3 T backfield magnetization. The corresponding remanence was measured, termed SIRM1T and IRM-0.3T, respectively. The S-ratio0.3T was calculated as 0.5 × (SIRM1T - IRM-0.3T)/SIRM1T28.

Scanning electron microscope analysis

Selected fresh samples were ground into powder using an agate mortar and pestle and were then dissolved into a mildly acidic buffer solution (pH~4) for 2-3 days to beneficiate the trace concentration of magnetic minerals in limestone samples. The buffer solution was made of 2 M acetic acid (CH3COOH) and 1 M sodium acetate (NaCH3COOH) at a 4:1 ratio64. Then, residual extracts were prepared for scanning electron microscope (SEM) observations. Backscattered electron (BSE) images and SEM-based energy-dispersive X-ray spectroscopy (EDXS) micro-analyses were obtained for eight samples with a Zeiss (Germany) field emission SEM (Gemini 450) at a 15 kV accelerating voltage and 2 nA current with a working distance of ~10 mm.

Cerium anomaly measurements

Forty-one bulk samples through the entire Meishan section were selected for cerium anomaly analysis following the protocol described in Jenner et al.65. All samples were powdered into <75 μm using an agate TEMA mill after removal of the weathered surfaces and secondary veins. Subsequently, 50 mg of each powdered sample underwent dissolution in sub-boiling distilled HF, followed by evaporation to incipient dryness at 150 °C to eliminate the silicate. The resulting residuals were further dissolved in a mixture of sub-boiling distilled HF and HNO3 in a high-pressure Teflon bomb, allowing for a thorough digestion process lasting 96 h at 190 °C. After evaporation of the HF-HNO3 mixture, the sample was solubilized in HNO3, evaporated to dryness, taken up in 2-3 ml of HNO3, transferred to a 50 ml centrifuge tube and diluted to a final volume of 50 ml. Meanwhile, 1 ml 500 ng/g Rh was added to each sample as an internal standard for correcting matrix effects and instrument drift for measurements. The analyses were performed with an Agilent 7500 inductively coupled plasma mass spectrometer (ICP-MS) (USA) at the Beijing ZKKY GeoAnalysis Laboratory Co., Ltd. (Beijing, China). Analytical errors were better than ± 5% for most elements based on USGS standards and two Chinese national standards (GBW07107, GBW07108).

The rare earth elements (REE) were normalized to Post Archean Australian Shale (PAAS)66,67. Cerium anomalies (ΩCe) were calculated using the formula CeN/(LaN0.5 × PrN0.5)-168 and subsequently integrated with previously published in-situ ΩCe of conodont at the Meishan section32.