Introduction

Large igneous provinces (LIPs) represent the most extensive volcanic events in Earth’s history, typically outpouring over 100,000 cubic kilometers of lava within a short duration of 1 to 5 million years1. The conventional model links LIP volcanism to massive carbon dioxide (CO2) emissions that drive global warming and environmental perturbations2,3. However, recent studies reveal that the relationship between LIP emplacement and atmospheric CO2 is more complex. In the Central Atlantic Magmatic Province, episodic eruptions produce immediate CO2 spikes, followed by declines attributed to enhanced weathering of freshly exposed volcanic rocks4. However, in the case of the Deccan Traps and Columbia River Basalt Group, CO2 emissions are observed to increase prior to the onset of their main flood basalt volcanism due to the pre-eruption intrusions5. Moreover, “cryptic” carbon emissions can persist for several million years after the main eruptive phase through sustained magmatic degassing and metamorphism of organic-rich sediments6. These varying patterns might reflect differences in magma volatile content, intrusive activity, crustal architecture, and the relative timing of competing processes, such as volcanic degassing and enhanced weathering.

The environmental impact of LIPs was particularly significant during the Permian Period, which experienced two distinct biotic crises. The end-Guadalupian event (~260 million years ago (Ma)) was a critical precursor that possibly set the stage for the devastating end-Permian mass extinction7,8. This event notably affected shallow-marine taxa such as fusulines, corals and brachiopods across the Guadalupian-Lopingian (G-L) boundary9,10,11,12,13. Geochronological data have established temporal synchronicity between the end-Guadalupian mass extinction and the emplacement of Emeishan LIP14,15,16,17,18. While volcanogenic CO2 has been hypothesized as the key driver of this biotic crisis19, the actual relationship between the Emeishan volcanism and atmospheric CO2 concentrations (pCO2) remains poorly constrained.

Here we present a high-resolution pCO2 record spanning the Mid to Late Permian using compound-specific carbon-isotope measurements of chlorophyll-derived biomarkers. By resolving pCO2 changes through different phases of the Emeishan LIP development, we evaluate whether this volcanic episode conforms to the conventional LIP–CO2 rise–extinction model, and how carbon cycle dynamics during this interval may have influenced the end-Guadalupian biotic crisis.

Results and Discussion

Carbon isotopic records of bulk sediment and phytane

High-resolution stable carbon isotope records of bulk sediments (δ13Ccarb, δ13Corg) and biomarker phytane (δ13Cphy) (Fig. 1) from exposed strata at Shangsi section (32°20′ N, 105°28′ E, Supplementary Fig. 1) document well-preserved marine sedimentation spanning from the Guadalupian (including the Roadian, Wordian, and Capitanian stages) to the Lopingian. One characteristic signature of G-L boundary sections is the negative carbon isotope excursion (CIE) recorded in carbonates across various regions (Supplementary Fig. 2), although localized depositional hiatuses may obscure this trend. Our δ13Ccarb profile shows a gradual negative shift from ~4‰ to a minimum of ~1‰ during the transition from the Guadalupian to the Lopingian (Fig. 1c), followed by a return to pre-CIE levels (average of 4‰) during the Wuchiapingian stage. However, the δ13Corg record exhibits a different pattern (Fig. 1d). At the G-L boundary, instead of a negative excursion, the values reach a transient maximum of ~ −23‰. This is followed by two distinct negative shifts throughout the Wuchiapingian stage, each interrupted by a temporary positive excursion.

Fig. 1: Guadalupian-Lopingian transitional interval chronology, stable isotope, and lipid biomarker stratigraphic records from Shangsi Section, China.
figure 1

a Chrono- and lithostratigraphy. b Geological and biological events from refs. 12,13,14. c Carbonate carbon isotopic compositions (δ13Ccarb, blue squares). d Bulk organic carbon isotopic compositions (δ13Corg) together with %TOC (blue circles, size of symbol proportional to %). e Phytane carbon isotopic compositions (δ13Cphy, blue circles). Horizontal error bars represent 1σ of the mean (n = 2). f εp (blue circles) calculated from δ13Cphy and δ13C of dissolved CO2 determined from (c). Shaded gray areas represent 68% (dark gray) and 95% (light gray) confidence intervals of εp calculation based on Monte Carlo simulations with 10,000 times (See Methods). Black dashed line represents the LOESS smoothing function (See Methods). Data in Supplementary Table 1. Fm Formation, W Wordian.

Full size image

Phytane is a diagenetic product derived from the phytol side chain of chlorophyll produced by photoautotrophic organisms, primarily algae and cyanobacteria20. The carbon isotopic composition of phytane preserves information about isotopic fractionation that occurred during photosynthetic carbon fixation by the primary producers that synthesized the original chlorophyll molecules. Consequently, the compound-specific δ13C values of sedimentary phytane can serve as an effective proxy for reconstructing geological pCO2 levels when appropriately calibrated21,22,23,24. At Shangsi, our δ13Cphy measurements show a positive shift from ~ −31‰ to −27‰ from the Capitanian into the Early Wuchiapingian, followed by a gradual decrease to a minimum of ~ −32‰ (Fig. 1e). When converting δ13Cphy values to the original biomass δ13C, we applied the established 3.3‰ enrichment factor documented by ref. 22. This correction accounts for the systematic isotopic fractionation that occurs during chlorophyll biosynthesis and subsequent diagenetic transformation to phytane.

To reconstruct atmospheric CO2 levels, we quantified the isotopic fractionation (εp) associated with carbon fixation during photosynthesis using our paired values of δ13Cphy and δ13Ccarb following established methods25 (See Methods). This parameter increases systematically with rising pCO2, though it can also be influenced by factors such as phytoplankton growth rate, cell geometry, and active carbon transport mechanisms26,27,28. In Fig. 1f, the calculated εp profile shows a negative shift from ~22‰ to a minimum of ~17‰ during the transition from the Capitanian to the onset of the Wuchiapingian, followed by a pronounced increase in εp. By the mid-Wuchiapingian stage, εp reaches a maximum value of ~24‰, and subsequently decreases to ~21‰.

Estimates of atmospheric pCO2

New geochronological data provide a precise temporal framework of the eruption sequence within the Emeishan volcanic province14,15,16,17. The main eruptive phase of the Emeishan flood basalt volcanism occurred from 260.55 Ma to 259.1 Ma (refs. 15,17), with the earliest documented eruption at ca. 263.5 Ma (ref. 14). The waning stage of Emeishan LIP volcanism extended until ca. 257.4 Ma in the early Wuchiapingian16.

Our εp-based estimates of pCO2 (Fig. 2e) show that CO2 concentrations remain stable at approximately 700 parts per million by volume (ppm) throughout the Roadian and Wordian stages. Coincident with the onset of the Capitanian stage, pCO2 values begin to decrease at ~263.5 Ma, reaching an estimated minimum of 369 ppm, with a running mean average of ~550 ppm across the G-L boundary. This interval of decreasing pCO2, from ~263.5 to 259.1 Ma, temporally corresponds with the early and main phases of the Emeishan flood basalt volcanism (Fig. 2a, e). Counterintuitively, the minimum pCO2 level coincides with the termination of the main phase of flood basalt eruptions, after which pCO2 values increase despite reduced eruption volumes. This surprising relationship is further highlighted by the correspondence between lowest pCO2 values and peak mercury (Hg) anomalies19 (Fig. 2b), demonstrating that the most intense volcanic activity during the main eruption phase paradoxically resulted in atmospheric CO2 drawdown rather than the expected increase. In the Early Wuchiapingian, pCO2 value reaches an estimated maximum of 1000 ppm, with a running mean average of ~900 ppm. This interval of elevated CO2 level coincides with explosive silicic eruptions rather than the earlier flood basalt volcanism16,29. Following the termination of the Emeishan volcanism, the pCO2 values decrease to approximately 600 ppm.

Fig. 2: Estimates of atmospheric CO2 and major geological events associated with the Emeishan LIP spanning from the Guadalupian to the Lopingian.
figure 2

a Evolutionary stages and high-precision chronology of the Emeishan LIP. Data from refs. 14,15,16,17. b Mercury record (Hg/TOC) for the timing and strength of volcanism19. c, d The degree of continental weathering based on (c) chemical index of alteration (CIA) and d seawater lithium isotope (δ7Lisw) records38,39,40. e pCO2 estimates based on εp. Shaded gray areas represent 68% (dark gray) and 95% (light gray) confidence intervals of pCO2 estimation based on Monte Carlo simulations (See Methods). f Carbonate carbon isotopic compositions (δ13Ccarb) from this study. g Biological events from refs. 14,57,104. Solid and dashed lines in (c–f) represent the LOESS smoothing function.

Full size image

Our observed pCO2 trend from the Emeishan LIP reveals several unexpected findings in carbon cycle behavior that fundamentally challenge conventional understanding of LIP-climate interactions. First, while the standard model predicts atmospheric CO2 increases during peak flood basalt volcanism4, the Emeishan LIP exhibits precisely the opposite trend: a sustained CO2 decrease from ~700 to ~350 ppm during its most intense volcanic phase. This decoupling between volcanic activity and atmospheric CO2 is further highlighted by the correspondence between minimum pCO2 values and peak Hg anomalies, demonstrating that the most vigorous eruption phase paradoxically coincided with maximum carbon drawdown rather than addition. Second, our data reveal that the negative CIE at the G-L boundary occurred alongside CO2 drawdown rather than increase, contrasting with the typical coupling between negative CIEs and CO2 spikes observed in numerous other environmental perturbations throughout Earth history23,30,31. Third, the subsequent CO2 increase during the later silicic eruptive phase—despite its lower eruption volume—further demonstrates that eruption intensity does not directly correlate with atmospheric CO2 response in this case, reversing the expected pattern of volcanogenic carbon forcing. These multiple discrepancies from established patterns collectively indicate that a distinct and previously unrecognized mechanism must have dominated the carbon cycle during the Emeishan event, effectively counterbalancing and even overwhelming the expected volcanic CO2 contributions.

Mantle plume-driven uplift and carbonate weathering as a carbon sink

The key to understanding the paradoxical CO2 pattern during the Emeishan LIP lies in the unique tectonic processes and geomorphological changes that preceded the main volcanic eruptions. Unlike many other LIPs, the Emeishan event was characterized by extensive pre-eruptive crustal doming and uplift that fundamentally altered regional weathering dynamics32,33,34. Biostratigraphic and sedimentological studies of 67 sections within the Middle Permian Maokou Formation, beneath the Emeishan basalts in South China, provide compelling evidence of this uplift. These studies document systematic thinning of the Maokou strata toward the center of the province and widespread development of unconformities that become increasingly prominent approaching the center of the domal structure (Supplementary Fig. 3). The observed pattern of stratigraphic thinning outlines a subcircular uplifted area with a radius of approximately 800 km that experienced uplift of up to 1000 m32 (Fig. 3). This uplift, estimated to have lasted approximately 3 million years, is consistent with theoretical models of lithospheric response to mantle plume impingement35,36,37.

Fig. 3: Time series and evolutionary stages of Emeishan LIP emplacement, and the carbon cycle spanning from the Guadalupian to the Lopingian.
figure 3

a Stable carbonate platform prior to the mantle plume formation. b Kilometer-scale crustal uplift and erosion due to the Emeishan mantle plume during the Stage I, with the orange dashed area representing the weathered Maokou Formation, leading to the decrease of CO2. c The main eruption of the Emeishan flood basalts during the Stage II, with CO2-poor basaltic magmatism and low CO2 levels. d Emeishan silicic eruptions and associated greenhouse gas emissions (e.g., CO2) during the Stage III. The widths of the curved arrows with isotopic signals in (b, c) represent the magnitude of carbon flux. Note that the thicknesses of Earth’s internal layers in this figure are not to scale. This figure is intended as a conceptual illustration and does not reproduce the full geometry or complexity of the LIP.

Full size image

The critical environmental consequence of this uplift was the exposure of vast expanses of previously submerged marine carbonates to subaerial weathering conditions. The development of extensive paleoweathering crusts, paleosols, and karst topography at the unconformity between the Maokou Formation and overlying Emeishan basalts provides direct evidence for intensified weathering during this period. This enhanced weathering is further corroborated by geochemical proxies, including a substantial decline in seawater lithium isotopes (δ7Lisw)38 (Fig. 2d) and elevated Chemical Index of Alteration (CIA) values39,40 (Fig. 2c), both reliable indicators of increased continental weathering intensity.

In addition, the distinctive CO2 record of the Emeishan LIP is closely linked to its broader Late Paleozoic weatherability context. Marcilly et al. (ref. 41) reconstruct high-weatherability zones (HWZ) from climatically sensitive lithologies and highlight their importance for the evolution of the Late Paleozoic Ice Age (LPIA). During the Permian, HWZ preferentially shifted toward the China blocks and Siberia rather than the arid central Pangean interior41, placing South China within a warm, high-runoff belt conducive to strong CO2 uptake. In such low-latitude settings, tectonic uplift can shift weathering from supply-limited to kinetically controlled regimes and thereby enhance CO2 consumption42. Within this framework, plume-related doming above Emeishan would have amplified weathering of the rapidly exposed Yangtze carbonate platform, producing a pronounced CO2 drawdown. Therefore, our Emeishan record reflects a focused, characteristic, plume-amplified weathering pulse operating within global LPIA-scale changes in weatherability.

Carbonate weathering is typically not considered a major driver of long-term atmospheric CO2 changes because it operates on relatively short timescales with a balanced carbon budget—CO2 consumed during weathering is largely released during subsequent carbonate precipitation in the oceans (e.g., GEOCARB III43). Under steady-state conditions, the net effect on atmospheric CO2 is minimal. However, the Emeishan scenario represents a departure from this steady-state scenario. The rapid and extensive uplift of carbonate platforms created a transient but powerful carbon sink by dramatically increasing the weathering flux over approximately 3 million years—a timeframe long enough to constantly impact pCO2 levels but short enough to prevent full compensation through the carbonate burial feedback system.

Notably, the Late Permian predates the Mid‑Mesozoic proliferation of pelagic calcifiers and the establishment of a well-buffered deep‑sea carbonate sink. In such a ‘Neritan’ ocean state, carbonate burial was dominated by platforms and shelves, and the negative feedback associated with deep-sea carbonate compensation was inherently weaker and slower44. Consequently, a large, transient alkalinity–DIC injection from the rapid exposure and weathering of the Maokou carbonates would not have been effectively buffered by pelagic dissolution or precipitation. This inefficiency likely produced a stronger atmospheric pCO2 response than in the modern ocean, sustaining pCO2 drawdown on Myr timescales and amplifying the climatic consequences of enhanced carbonate weathering during the crustal doming of the Emeishan LIP. This Permian carbon-cycle architecture strengthens the plausibility of our interpretation that tectonic uplift may have dominated the atmospheric signal across the Emeishan LIP interval.

Quantitatively, this uplifted carbonate province had a substantial weathering potential. Assuming the uplifted area extended across a circular region with an 800 km radius and experienced average erosion of 150 m thickness32, we estimate ~754,000 Gt of limestone was subjected to weathering, based on an average limestone density of 2.5 × 1012 kg/km3. This erosion would have mobilized ~90,480 Gt of inorganic carbon and 9500 Gt of organic carbon (see Methods), resulting in a net consumption potential of 80,980 Gt C from the atmosphere. This value substantially exceeds the estimated contemporary atmospheric CO2 reservoir of ~1480 Gt C, based on the initial concentration of 700 ppm during the Roadian and Wordian stages (see Methods). Even if marine carbonate compensation and buffering mechanisms limited the atmospheric impact to only 10% of the weathering flux (~8098 Gt C), this still exceeds the entire atmospheric carbon reservoir, demonstrating the enormous potential of this mechanism to drive significant CO2 drawdown.

This large-scale weathering of carbonate platforms represents a powerful carbon sink that could readily overwhelm concurrent volcanic CO2 emissions, particularly given the distinctive geochemical composition of the Emeishan basalts. Analyses of the volatile content in Emeishan volcanic rocks provide critical context for understanding this carbon cycle dynamic. Tang et al. (ref. 45) conducted comprehensive geochemical and isotopic analyses of volatiles in Emeishan picrites and basalts, using both bulk-rock measurements and studies of melt inclusions trapped in olivine phenocrysts. These investigations revealed that Emeishan basalts contain remarkably low CO2 concentrations, averaging only ~135 ppm. This is dramatically lower than typical ocean island basalt values of 6500–28,000 ppm and much less than other continental flood basalts associated with major extinction events. Note that the value of 135 ppm is lower than typical values observed in modern systems and may have been affected by loss during preservation. Therefore, this value should be interpreted with caution as a direct quantitative measure of magmatic degassing. Nevertheless, it provides strong comparative evidence that the Emeishan basaltic magmatism was notably low in CO2 and that degassing was not primarily magmatic.

The relatively CO2-poor nature of the Emeishan basalts stands in stark contrast to other LIPs like the Siberian Traps, which have been associated with massive carbon emissions during the end-Permian mass extinction. Isotopic analyses of helium and carbon in the Emeishan basalts indicate they were derived from a relatively degassed mantle source, from which much of their volatile content had been depleted prior to the eruption. Beyond the magmatic volatile budget, thermogenic CO2 generated where sills intrude organic-rich sedimentary basins can be both volumetrically significant and rapid46,47,48,49,50. In the Siberian Traps, high‑precision geochronology documents a transition from early extrusive lavas/pyroclastics to widespread sill emplacement into the volatile‑rich Tunguska Basin51. The emplacement of these spatially extensive sills led to enormous thermogenic degassing46,47,48,49. An independent pCO2 reconstruction23 shows a coeval CO2 rise with the onset of this sill-dominated intrusive phase in the Siberian Traps, supporting this interpretation. In contrast, the Emeishan LIP is characterized by layered mafic intrusions concentrated in the inner-zone Panxi region52,53,54 (see Methods for discussion of intrusive activity). Comparable province-wide sill intrusion into organic-rich sedimentary sequences analogous to those in the Tunguska Basin has not been demonstrated in the Emeishan LIP. The predominance of carbonate host rocks and associated contact aureoles around Emeishan intrusions suggests that metamorphic carbon liberated by decarbonation of carbonate country rocks was likely the main degassing source55. However, the limited precision of existing age data for Emeishan intrusions hinders precise constraints on the relative contribution and timing of thermogenic CO2 release. Nevertheless, because intrusions are confined to the inner zone, their impact on atmospheric CO2 was likely limited. Therefore, the combination of enhanced weathering capacity and relatively CO2-poor magmatism created conditions whereby carbon sequestration would dominate over volcanic degassing, even during periods of intense eruptive activity.

Isotopic response and carbon cycle implications

The large-scale carbonate weathering triggered by pre-eruptive uplift not only affected atmospheric CO2 levels but also left a distinctive signature in the carbon isotope record. To evaluate the isotopic implications of this mechanism, we applied a mass balance approach to quantify the expected δ13C shifts resulting from enhanced carbonate weathering. The weathering of the Maokou Formation would have introduced a substantial amount of carbon with distinctive isotopic signatures into the global carbon cycle. Using a mass balance calculation, we estimated the integrated δ13C value of the weathering input. Assuming average δ13C values of 0‰ for the weathered carbonate component (typical of marine carbonates) and −25‰ for the organic carbon component (with an average TOC of 1.26% derived from Maokou Formation samples), the composite weathering input would have a δ13C value of approximately −2.38‰ (see Methods). This represents a significantly 13C-depleted flux compared to the background marine carbonate values of approximately +1‰ recorded before the excursion.

Under a hypothetical scenario without oceanic buffering or carbonate compensation, the input of 99,980 Gt C with δ13C value of −2.38‰ into the ocean-atmosphere system would result in a new ocean δ13C value of −1.39‰. This would produce a negative excursion in marine carbonate δ13C of about 2.39‰ (from +1‰ to −1.39‰) over the 3-million-year period of enhanced weathering. In a more realistic scenario incorporating carbonate compensation, where for every mole of CO2 added from weathering, approximately 0.8 moles of carbonate sediment dissolves, the expected shift would be moderated to about 1.28‰, with a new equilibrium δ13C value of −0.28‰.

The magnitude of these calculated isotopic shifts aligns remarkably well with our observed CIE at the G-L boundary, where δ13C values decline from ~4‰ to ~1‰ (Fig. 1c). This coherence between our theoretical calculations and the actual isotopic record provides independent validation of the weathering mechanism we propose. Importantly, the timing of this negative excursion coincides precisely with the interval of decreasing atmospheric CO2, supporting our interpretation that both signals reflect the same underlying process of enhanced weathering.

This isotopic pattern represents another distinctive feature of the Emeishan LIP compared to other major volcanic episodes. Typically, negative CIEs associated with LIPs are interpreted as evidence for massive injections of isotopically light carbon into the atmosphere, often accompanied by warming and ocean acidification. In the Emeishan case, however, the negative CIE occurs alongside CO2 drawdown rather than an increase, demonstrating that different carbon-cycle mechanisms were involved. This pattern further emphasizes the unique nature of the Emeishan event and underscores the importance of considering the full geological context—including pre-eruptive processes—when interpreting carbon-cycle perturbations associated with LIPs.

Our findings challenge the conventional model linking LIPs directly to CO2 increases and global warming. The Emeishan case demonstrates that pre-eruptive mantle-lithosphere interactions can drive carbon-cycle perturbations millions of years before the first lava flows, potentially triggering atmospheric responses opposite to traditional expectations. More broadly, the climatic expression of LIP volcanism is shaped by a combination of factors, including emplacement style (extrusive vs. intrusive), mantle volatile loading, lithospheric structure, paleogeography, and the prevailing ocean–carbon state, etc. The Emeishan LIP developed in a tropical, high-weatherability setting, where crustal uplift exposed kilometer-thick Yangtze platform carbonates to intense erosion and weathering under a Late Paleozoic ocean characterized by weak pelagic buffering, while the associated basaltic magmatism was notably CO2-poor. Together, these factors explain why the Emeishan LIP is distinct among continental flood basalts in exhibiting pCO2 decline during the main basaltic phase. This fundamentally reframes how we understand deep Earth-surface environment connections, suggesting that a LIP’s environmental impact follows more complex pathways than previously recognized, potentially explaining why some LIPs are linked to mass extinctions, whereas others are not. The distinctive CO2 drawdown pattern we observe emphasizes the importance of examining each LIP within its complete geological context rather than applying a universal template. Future research should adopt an integrated Earth System approach that examines the full sequence of mantle-to-atmosphere interactions across all stages of LIP development—crucial for accurately interpreting cause-and-effect relationships between volcanism and environmental change throughout Earth’s history.

Methods

Materials

Samples were collected from the Shangsi section (32°20′ N, 105°28′ E, Supplementary Fig. 1), Shangsi Village, Guangyuan City, Sichuan Province, China. The Shangsi section, a prominent candidate for the Global Stratotype Section and Point (GSSP) of the Permian-Triassic boundary, is distinguished by its well-exposed marine successions and extensive research foundation. Based on paleogeographic reconstruction, Shangsi was located in the eastern region of the Paleo-Tethys during the Middle to Late Permian. All samples were collected from outcrops where the surface layers had been completely removed, exposing fresh layers to avoid weathered or otherwise contaminated rocks. Stratigraphically, the Shangsi section includes the Maokou and Wuchiaping formations, which correspond to the Guadalupian and Lopingian series, respectively. The G-L boundary is placed at the base boundary of the Wuchiaping Formation56. Fossil records indicate that the end-Guadalupian extinction was a multi-phase biotic crisis characterized by a gradual decline in biodiversity and episodic turnovers beginning in the mid-Capitanian. It affected multiple taxonomic groups such as fusulines, corals, ammonoids, brachiopods, conodonts and bivalves, leading to widespread ecosystem collapse9,10,11,12,13.

Chronology

The age control for the G-L interval is derived from the latest high-precision geochronology. The tie-points are as follows: 273.01 ± 0.14 Ma for the base of the Roadian57, 266.9 ± 0.4 Ma for the base of the Wordian58, 264.28 ± 0.16 Ma for the base of the Capitanian58, 259.51 ± 0.21 Ma for the base of the Wuchiapingian59 and 254.14 ± 0.07 Ma for the base of the Changhsingian60.

Intrusive activity during the Emeishan LIP emplacement

The Emeishan intrusive system is expressed by a cluster of layered mafic-ultramafic complexes in the inner zone, typically surrounding the Panxi area (Panzhihua-Xichang, Supplementary Fig. 3a), including Panzhihua, Baima, Hongge, Taihe, Binggu and Xinjie18,52,53,54,61. Geophysical imaging along a seismic profile crossing the Emeishan LIP shows that crustal thickening and magmatic modification are most pronounced in the inner zone, where a thick mafic underplating layer is present, whereas the intermediate and outer zones lack comparable underplating features, implying that magmatic intrusion in these zones was relatively limited62. To date, only a single field study has documented composite, zoned diabase sills occurring in the outer zone, at Luodian in southern Guizhou63. Overall, the prevailing view on the distribution of intrusions in the Emeishan LIP is that they are mainly concentrated in the inner zone.

The timing of the Emeishan LIP intrusions broadly overlaps the main basaltic eruptions and extends from the earliest magmatic records to the waning stage (Supplementary Fig. 3d; refs. 14,15,16,17,18,52,53,54,61,63,64). Although the current consensus is that Emeishan LIP extrusive and intrusive activities were essentially coeval, the analytical quality and precision of available age determinations for the intrusions are insufficient to establish a precise intrusive chronology comparable to that of the Siberian Traps.

The country rocks intruded by the Emeishan bodies are predominantly carbonates. In the inner zone, intrusions were emplaced largely into Neoproterozoic (Sinian) dolostones that are nearly pure (containing very little clay or quartz), interbedded with siliceous limestones, marlstones and minor shales of the Dengying Formation65. In the outer zone, the Luodian sills intrude Early-Middle Permian Sidazhai Formation carbonates63. The development of contact aureoles within these carbonate successions suggests metamorphic decarbonation and localized carbonate partial melting of the country rocks during emplacement65,66.

Sample cleaning and preparation

All rock samples were cleaned and prepared following procedures described in ref. 67. Specifically, the rock samples were sonicated in three solvents (methanol, dichloromethane, and n-hexane) to remove any organic contamination. After cleaning, the samples were crushed into fine powder using a ring and puck mill and then stored in pre-combusted glass jars.

Bulk isotopic analysis

Bulk carbon isotopic values of carbonate (δ13Ccarb) were analyzed using a Thermo Scientific GasBench coupled with a Delta V Plus isotope ratio mass spectrometer (IRMS), with Chinese national standards, GBW 04416 and GBW 04417 as references, at the Institute of Geology and Geophysics, Chinese Academy of Sciences. The carbon isotopic ratios are reported using standard δ notation (δ13C in ‰) relative to Vienna Pee Dee Belemnite (VPDB). Standard precision is better than 0.05‰ (1σ).

Rock powders were completely decarbonated by reacting with dilute hydrochloric acid and subsequently rinsed with Barnstead Nanopure water until the pH was neutral. Bulk carbon isotopic values of organic matter (δ13Corg) were analyzed using a Thermo Scientific Flash elemental analyzer (EA) coupled with a Delta V Plus IRMS, with in-house laboratory standards along with reference standards IAEA-600 and USGS 40, at the Institute of Geology and Geophysics, Chinese Academy of Sciences. All samples were measured in triplicate. Standard precision is better than 0.1‰ (1σ).

Lipid extraction and biomarker analysis

Samples were extracted using a Thermo Scientific™ Dionex™ ASE™ 350 Accelerated Solvent Extractor in 9:1 dichloromethane/methanol mixture at 100 °C to obtain total lipid extracts (TLEs). To maximize biomarker yield, more than 150 g of powdered rock was used per sample for extraction, and three ASE extraction cycles were performed until the extract color became lighter, ensuring sufficient biomarker abundance for subsequent isotopic analyses. Biomarker analysis was conducted at the Stable Isotope Geoscience Facility, Texas A&M University and the Institute of Urban Environment, Chinese Academy of Sciences. Saturated hydrocarbons were eluted through SiO2-gel chromatography with 100% hexane. To remove n-alkanes, these fractions were transferred to cyclohexane and treated with activated 5Ǻ molecular sieves at 85 °C for 24 h. Before isotopic analysis, we verified phytane identity and purity by GC-MS and required baseline-resolved phytane peaks with good peak shape. Any chromatogram showing peak overlap (coelution) was excluded. Representative GC-MS chromatograms illustrating phytane identification and baseline separation are shown in Supplementary Fig. 7. Compound-specific isotopic compositions of phytane (δ13Cphy) were analyzed using a Thermo Trace Ultra gas chromatograph coupled with a MAT 253 IRMS, equipped with a HB-5 column (50 m × 0.2 mm × 0.33 μm) and a programmable temperature vaporization injector with helium as the carrier gas. The oven program was set at 60 °C upon sample injection, held for 1.4 min, then ramped up uniformly at 10 °C min−1 to 180 °C, 4 °C min−1 to 260 °C, and 20 °C min−1 to 320 °C, where it was held for 10 min. An external standard of n-alkanes with known δ13C values (A7, Arndt Schimmelmann, Indiana University) was used to monitor the instrument’s performance. The carbon isotopic ratios are reported relative to Vienna Pee Dee Belemnite. Hexadecane (C16) or eicosane (C20) with known δ13C values was used as the internal standard. All samples were measured in duplicates and corrected for drift and linearity. Standard precision is better than 0.3‰.

Terrestrial organic matter input. To assess the contribution of terrigenous organic matter in the Shangsi samples, we used n‑alkane and isoprenoid indices (Supplementary Table 2). The abundances of these compounds were determined using an Agilent 8890 gas chromatograph equipped with a flame ionization detector (GC-FID) and a DB-5 capillary column (30 m × 0.25 mm × 0.25 μm). The temperature program started at 70 °C, then increased at a rate of 10 °C/min to 210 °C, followed by 3 °C/min to 300 °C, where it was held for 30 min. Based on peak areas, we calculated (i) the terrigenous-aquatic ratio (TAR; TAR = [C27 + C29 + C31]/[C15 + C17 + C19]), which contrasts long‑chain (C27–C31) higher-plant wax alkanes with short‑chain (C15–C19) algal/bacterial alkanes68, (ii) the carbon preference index (CPI; CPI = [C25 + C27 + C29 + C31]/[C24 + C26 + C28 + C30]) of long‑chain n‑alkanes as an indicator of odd‑over‑even predominance typical of fresh higher‑plant waxes69,70, and (iii) pristane/phytane (Pr/Ph), which generally reflects redox and depositional setting71. At Shangsi, 33 out of 39 samples have TAR values below 1, with an average of 0.58 for these samples. The six samples with TAR values greater than 1 are not concentrated at any particular stratigraphic level but occur randomly throughout the section. Long‑chain CPI values cluster around 1 (0.67–0.98; average = 0.89 ± 0.07, n = 39), and Pr/Ph ranges from 0.23–1.03 (average = 0.59 ± 0.14; n = 38), with no values exceeding 1.5. Together, these indices indicate that the organic matter in the Shangsi samples is primarily derived from aquatic/marine sources, with a relatively minor higher-plant contribution.

pCO2 calculations

Phytane-based pCO2 reconstructions rely on the measurements of carbon isotopic composition from both phytane (δ13Cphy) and associated carbonates (δ13Ccarb). The carbon isotope fractionations (εp) during phytoplankton photosynthesis are utilized to reconstruct seawater and atmospheric CO2 concentrations throughout the Phanerozoic era. εp values are derived from the δ13C values of dissolved CO2 (δd) and photosynthetic biomass (δp) (ref. 25). δp is calculated from the isotopic difference between phytol and photoautotrophic biomass. Prior culture studies estimate this difference to be 3.3 ± 1.3‰ (ref. 22). Meanwhile, δd is calculated from δ13Ccarb, using the temperature-dependent relationship between the δ13C values of dissolved CO2 and solid carbonate72,73. The coeval seawater temperatures were derived from a compilation of high-resolution conodont apatite δ18O records57,74,75. Given the varying sampling resolutions, a LOESS smoothing function was applied to derive continuous temperature data across the timeline. The 1σ (68%) and 2σ (95%) uncertainties in εp calculations were determined through Monte Carlo simulations that accounted for various sources of uncertainty, including analytical uncertainty in δ13Cphy, a standard deviation (SD) of 1.3‰ in the isotopic offset between phytane and biomass, and a 4 °C SD in seawater temperature estimation.

In pCO2 reconstructions, it is assumed that dissolved CO2 is primarily transported to the site of carbon fixation in photosynthetic cells by diffusion76. Carbon concentrating mechanisms (CCMs) were unlikely to be fully active during the relatively high CO2 levels of the Permian and Triassic periods22, as this active carbon uptake by CCMs requires substantial energy, a characteristic observed in many modern algae species77. The concentrations of aqueous CO2 [CO2(aq)] were calculated using the equation: εp = εf − b/[CO2(aq)], where εf represents the net kinetic isotope fractionation of cellular enzymatic processes, b represents the physiological factors78,79,80. In this study, εf was assigned a value of 28‰. The b-value was calculated using the equation: b = 82.05 × [PO43−] + 86.28, where the [PO43−] represents a phosphate concentration of 0.2 μM as a representative level. Further discussion on εf and b-value are discussed below. Finally, atmospheric pCO2 concentrations were calculated from CO2(aq) using Henry’s law under the assumption of air-sea equilibrium81. The 1σ (68%) and 2σ (95%) uncertainties in pCO2 calculations were determined through Monte Carlo simulations that accounted for various sources of uncertainty, including analytical uncertainty in δ13Cphy, a SD of 1.3‰ in the isotopic offset between phytane and biomass, a 4 °C SD in seawater temperature estimation, and uncertainty in [PO43−], which ranged from 0.17 to 0.27 μM.

Uncertainties in pCO2 estimates

The calculation of pCO2 involves various assumptions, including the dependence of term b on [PO43-] and the value of enzymatic isotope fractionation (εf), along with potential uncertainties in seawater temperature that can impact pCO2 estimates. These factors are discussed below.

Estimates of b

In the process of photosynthetic carbon fixation, the parameter b synthesizes the effects of growth rate, cellular morphology, and membrane permeability on carbon isotopic discrimination. In this paper, we adopt the estimates for b utilized in previous CO2 reconstructions for the Permian-Triassic boundary23. The refined equation for b as a function of [PO43-] is derived from sediment data and is expressed as b = 82.05 × [PO43−] + 86.28 (ref. 82). This formula yields b values that enhance the accuracy of CO2 reconstructions, providing better alignment with sedimentary records. In this equation, we used a representative [PO43−] concentration of 0.2 μM at a depth of 20 m, representing the average value within the modern observed range of 0.17 to 0.27 μM from the ocean surface to a depth of 50 meters. Due to the absence of data on the temporal variability of b, we maintained it as a constant value throughout the study interval. Regardless of the uncertainty of b, its impact is confined to the absolute values of pCO2, as sensitivity analysis demonstrates that pCO2 changes by less than 6% when the maximum or minimum b values are applied (see Supplementary Fig. 4). Consequently, these variations do not alter the overall trend of pCO2 calculations over time.

Estimates of εf

Enzymatic isotope fractionation (εf) is primarily driven by the activity of the enzyme Rubisco and/or contributions from β-carboxylases. Experimental studies on the growth of eukaryotic algae and cyanobacteria have determined that εf values typically are within the range of 25‰ to 28‰ (refs. 27,79,83,84). In this study, we adopt the εf value of 28‰, as β-carboxylation is typically inhibited under higher CO2 levels, leading to larger εf value85. Supplementary Fig. 5 shows the theoretical relationship among εp, phosphate concentration, and seawater temperature, based on the εf value of 28‰.

A recent modern‑ocean calibration of the phytol εp-[CO2(aq)] relationship, based on a Deming regression, reports εf ≈ 23.9‰ and b ≈ 82.4 (ref. 24). This study refines and strengthens the phytol/phytane‑based approach to pCO2 reconstruction, thereby improving its reliability and expanding its applicability for future studies. In our study, the b value is parameterized from the phosphate relationship, which yields b ≈ 102. This value is substantially lower than the traditional constant of b = 170 and broadly consistent with the low‑b range in modern calibrations24. As for εf, while the modern calibration provides an important reference framework, several considerations suggest that its εf value may not be directly applicable to our study. First, the calibration domain of modern datasets is limited at the high‑[CO2(aq)] end. The open ocean rarely attains very high [CO2(aq)], and near-shore waters, where [CO2(aq)] can be higher, are typically excluded to minimize terrestrial inputs. This restriction reduces the opportunity to observe maximum fractionation, thereby leading to a lower fitted εf (ref. 24). Second, carbon‑concentrating mechanisms (CCMs) and active HCO3- uptake (β ‑carboxylation) are pervasive in modern phytoplankton and tend to depress the apparent fractionation. During deep-time intervals, these pathways were likely down‑regulated under elevated pCO2, implying a higher effective εf. In addition, because of the asymptotic nature of the εp-pCO2 relationship, adopting a low εf places more of our data near the εp ≈ εf limit, where small uncertainties in εp can propagate into disproportionately large pCO2 (ref. 24). For these mechanistic and methodological reasons, we adopt εf = 28‰ in our pCO2 reconstructions.

Correlation of records of δ13Ccarb

The G-L CIE is characterized by a pronounced decrease in δ13C values near the boundary57. Our δ13Ccarb profile exhibits trends and patterns that are consistent with those of other G-L records observed in various paleo-continental regions (Supplementary Fig. 2). This consistency suggests that the δ13Ccarb record from Shangsi reflects the typical marine δ13CDIC signal, which was subsequently used to calculate the isotopic composition of dissolved CO2.

LOESS

The running average was generated using the locally estimated scatterplot smoothing (LOESS) function in SigmaPlot V14 software. For this calculation, a first-order polynomial was applied as the smoothing degree, and the sampling frequency was set to 0.4 to define the proportion of data points included in the smoothing process.

Estimating the impact of Maokou Formation weathering on the carbon cycle

Estimating the volume and mass of eroded Maokou limestone

The uplifted area of the Maokou Formation is approximately 2,010,619 km2, assuming a circular area with a radius of 800 km (ref. 32). The estimated average erosion thickness is 150 m (ref. 32), resulting in an eroded volume of 301,593 km3 (calculated as 2,010,619 km2 × 0.15 km). With an average limestone density of 2.5 × 1012 kg/km3, the total mass of the eroded limestone is calculated to be 7.54 × 1017 kg or 754,000 Gt.

Total organic carbon content in Maokou Formation samples

To evaluate the total organic carbon (TOC) content of the Maokou Formation, we compiled previously published TOC data from different sections. Reported average TOC values are 1.64% at Chaotian86, 1.96% at Shangsi87, 7.58% at Xibeixiang88 and 0.19% at Tieqiao89. As the Xibeixiang data are limited to the uppermost Maokou Formation, they are excluded from subsequent analyses. The remaining three sections exhibit relatively consistent values, with a weighted average of 1.26% for the Maokou Formation. Note that the Shangsi value is taken from an independent study87 rather than our own measurements; however, our TOC data yield a similar value (2.13%), supporting the reliability of the published result.

Limestones generally contain less organic carbon than fine-grained siliciclastic rocks (e.g., shales) deposited in the same settings. The TOC content in most marine limestones is relatively low, typically below 0.5%. Modern carbonate sediments can accumulate organic matter in amounts comparable to shales, but much of the organic carbon is lost during burial through oxidation or decay, resulting in only a small residual TOC. However, certain limestone units do display elevated TOC under some specific conditions, such as anoxic bottom waters, high primary productivity and rapid burial, which facilitate the formation of organic-rich limestones.

Widespread anoxic conditions during the deposition of the Maokou Formation have been documented across various regions, including mid- to high-latitude areas of Panthalassa and the Tethys Ocean86,90,91,92,93,94. These oxygen-deficient conditions were not restricted to deep marine basins but also extended into intermediate to shallow water depths40,86,90,91,92,95. For instance, sulfur (δ34S) isotope records coupled with pyrite iron speciation analyses from the shelf carbonates at Chaotian reveal that the expansion of the oxygen minimum zone (OMZ) during the Capitanian drove the development of euxinic conditions86. Similarly, multiple sulfur isotopes (δ34S and Δ33S) in pyrites from the Penglaitan and Tieqiao sections further support the widespread shoaling of sulfidic waters90. Furthermore, negative shifts in δ238U values of marine limestones at Xikou also suggest intensified marine anoxia, with an estimated ~10% anoxic seafloor coverage lasting >3 Myr (ref. 96). In addition, remarkably high nitrogen isotope values (δ15N > 6‰, with peaks ≥ 10‰) in Maokou Formation limestones have been reported across several regions, exceeding most Phanerozoic δ15N records86,91,97,98,99. In the modern deep ocean, nitrate δ15N values typically range from 4‰ to 6‰. However, strongly denitrifying OMZs (e.g., the eastern tropical Pacific Ocean and the Arabian Sea) often exhibit exceptionally high δ15N values of 15–20‰ (refs. 100,101). These regions with such high δ15N values are characterized by high-productivity and low-oxygen levels, where intense water-column denitrification occurs. Therefore, these findings suggest that the Maokou Formation was likely deposited under oxygen-depleted conditions, which may have facilitated the accumulation of relatively high TOC in carbonate sediments.

Estimating the carbon content of eroded material

Inorganic carbon (carbonate): assuming a carbon content of 12% in CaCO3, the mass of inorganic carbon is estimated at 90,480 Gt C (calculated as 754,000 Gt × 0.12). Organic carbon: the average total organic carbon content is 1.26 wt%, resulting in an estimated mass of organic carbon of 9,500.4 Gt C (calculated as 754,000 Gt × 0.0126).

Impact on the carbon cycle

Carbonate weathering involves the chemical reaction:

$${{{{\rm{CaCO}}}}}_{3}+{{{{\rm{CO}}}}}_{2}+{{{{\rm{H}}}}}_{2}{{{\rm{O}}}}\to {{{{\rm{Ca}}}}}^{2+}+{{2{{{\rm{HCO}}}}}_{3}}^{-}$$
(1)

and consumes 90,480 Gt of atmospheric CO2. Conversely, organic carbon weathering occurs through the reaction:

$${{{{\rm{CH}}}}}_{2}{{{\rm{O}}}}+{{{{\rm{O}}}}}_{2}\to {{{{\rm{CO}}}}}_{2}+{{{{\rm{H}}}}}_{2}{{{\rm{O}}}}$$
(2)

and releases 9,500.4 Gt C into the atmosphere. The net effect on the carbon cycle yields a net consumption of 80,979.6 Gt C. Over a period of 3 million years, this results in a rate of change of approximately 0.027 Gt C per year.

Impact on the marine carbonate δ13C

Assuming the δ13C of weathered and original marine carbonate is approximately 0‰, and that of organic carbon is −25‰, the mass balance calculation for δ13C of the total input gives δ13Ctotal as (90,480 × 0‰ + 9500.4 × −25‰) / 99,980.4 = −2.375‰. Without buffering effects, the initial ocean carbon stands at 38,000 Gt C with a δ13C of +1‰. After including the input from weathering, which totals 99,980.4 Gt C at −2.375‰, the new δ13C calculates to (38,000 × 1‰ + 99,980.4 × −2.375‰) / 137,980.4 = −1.39‰. This suggests a potential negative shift in marine carbonate δ13C of approximately 2.39‰, from +1‰ to −1.39‰, over a period of 3 million years.

Considering buffering, the carbonate compensation assumes that for every mole of CO2 added from weathering, about 0.8 moles of carbonate sediment dissolve, totaling 99,980.4 Gt C × 0.8 = 79,984.32 Gt C. Assuming δ13C of carbonate sediments similar to the initial ocean DIC at +1‰, a new mass balance calculation for the total carbon, which now amounts to 38,000 + 99,980.4 + 79,984.32 = 217,964.72 Gt C, yields a new δ13C of (38,000 × 1‰ + 99,980.4 × −2.375‰ + 79,984.32 × 1‰) / 217,964.72 = −0.28‰. Thus, the buffered change in δ13C equates to −0.28‰ from the initial +1‰, resulting in a change of −1.28‰.

Impact on the atmospheric CO2

The initial Permian atmospheric CO2 concentration was set to be 700 ppm. To convert this value into gigatons of carbon (Gt C), we assume the present-day atmospheric mass of 5.1 × 1018 kg and a molecular weight ratio of CO2 to air of 1.52. Using these values, 700 ppm is equivalent to 700 × 5.1 × 1018 × 1.52 / (106 × 106) = 5426.4 Gt CO2. Converting this to carbon, we yield 5426.4 Gt CO2 × (12/44) = 1479.9 Gt C. Without buffering effects, the net consumption of 80,979.6 Gt C would far exceed the initial atmospheric CO2 content of 1479.9 Gt C, indicating the potential for complete depletion of atmospheric CO2. However, considering buffering mechanisms such as carbonate compensation, we conservatively estimate that only 10% of the net change would affect the atmosphere. This results in 80,979.6 Gt C × 0.1 = 8097.96 Gt C net removal from the atmosphere, which still exceeds the initial atmospheric CO2 content. A more realistic scenario might involve atmospheric CO2 decreasing by approximately half, leading to a 1479.9 Gt C × 0.5 = 739.95 Gt C decrease. This would reduce the new atmospheric CO2 level to 1479.9—739.95 = 739.95 Gt C. Converting this back to ppm, we find that 739.95 Gt C is equivalent to 2713.15 Gt CO2. Using the same conversion factors, we estimate the new atmospheric CO2 concentration to be approximately 350 ppm.

Sensitivity analysis of net CO2 consumption

To assess robustness, we quantified how the carbonate-weathering sink scales with the geometry of the uplifted dome and the organic-carbon content by exploring dome radius of 600–900 km, average erosion thickness of 50–300 m, and bulk TOC values of 0.5–2 wt%. For all combinations, the net carbon sink ranges from 1.41 × 104 Gt C at (600 km, 50 m, 2%) to 2.19 × 105 Gt C at (900 km, 300 m, 0.5%), and is therefore ≫ 103 Gt C under all tested conditions (Supplementary Fig. 6). Assuming a 2–4 Myr weathering interval, mean removal rates span 0.003–0.11 Gt C yr-1. Note that our estimate of the carbonate-weathering volume and rate represents a simplified, first-order approximation intended to illustrate the likely range of values rather than capture the full complexity of the system.

In summary, the weathering of the Maokou Formation over a 3-million-year period likely had significant impacts on the global carbon cycle. The weathering process could have caused a substantial negative shift in marine carbonate δ13C, potentially up to −2.39‰ without buffering effects. When considering buffering, the shift is estimated to be around −1.28‰, which remains significant and detectable in the geological record. This could have had important implications for marine ecosystems and carbon cycling. Additionally, a significant drawdown of atmospheric CO2 is predicted, with levels potentially decreasing from 700 ppm to approximately 350 ppm, even after accounting for buffering effects. This substantial decrease in CO2 would likely have had significant climatic implications in the lead-up to the Emeishan volcanism event.

The actual magnitude of these effects, however, would depend on several factors, including ocean chemistry, biological feedbacks, and the precise operation of the carbon cycle in the Permian period. Nonetheless, the estimated changes align with emerging geological evidence, supporting the hypothesis that pre-LIP mantle-crustal processes significantly impacted the carbon cycle and global environment during this period.