Redox chemistry changes in the Panthalassic Ocean linked to the end-Permian mass extinction and delayed Early Triassic biotic recovery

Significance

To understand how most life on Earth went extinct 250 million years ago, we used multiple sulfur isotopes to investigate redox chemistry changes in the Panthalassic Ocean, comprising ∼85–90% of the contemporaneous global ocean. The S-isotopic anomalies from Canada and Japan provide evidence for the timing of the onset of euxinia and mixing of sulfidic and oxic waters. Our data suggest that shoaling of H₂S-rich waters may have driven the mass extinction and delayed the recovery of the marine ecosystem. This study illustrates how environmental changes could have had a devastating effect on Earth’s early biosphere, and may have present-day relevance because global warming and eutrophication are causing development of sulfidic zones on modern continental shelves, threatening indigenous marine life.

Abstract

The end-Permian mass extinction represents the most severe biotic crisis for the last 540 million years, and the marine ecosystem recovery from this extinction was protracted, spanning the entirety of the Early Triassic and possibly longer. Numerous studies from the low-latitude Paleotethys and high-latitude Boreal oceans have examined the possible link between ocean chemistry changes and the end-Permian mass extinction. However, redox chemistry changes in the Panthalassic Ocean, comprising ∼85–90% of the global ocean area, remain under debate. Here, we report multiple S-isotopic data of pyrite from Upper Permian–Lower Triassic deep-sea sediments of the Panthalassic Ocean, now present in outcrops of western Canada and Japan. We find a sulfur isotope signal of negative Δ³³S with either positive δ³⁴S or negative δ³⁴S that implies mixing of sulfide sulfur with different δ³⁴S before, during, and after the end-Permian mass extinction. The precise coincidence of the negative Δ³³S anomaly with the extinction horizon in western Canada suggests that shoaling of H₂S-rich waters may have driven the end-Permian mass extinction. Our data also imply episodic euxinia and oscillations between sulfidic and oxic conditions during the earliest Triassic, providing evidence of a causal link between incursion of sulfidic waters and the delayed recovery of the marine ecosystem.

The end-Permian mass extinction was the largest biotic catastrophe of the last 540 million years, resulting in the disappearance of >80% of marine species, and a full biotic recovery did not occur until 4–8 million years after the extinction event (1–6). Several lines of evidence from the low paleolatitude Paleotethys and high paleolatitude Boreal oceans, which accounted for ∼10–15% of the contemporaneous global ocean area, suggest that sulfidic (H₂S-rich) conditions may have developed widely during the end-Permian extinction (7–13). However, redox chemistry changes in the Panthalassic Ocean, comprising ∼85–90% of the global ocean area, remain controversial, with competing hypotheses proposing extensive deepwater anoxia (“superanoxic ocean”) or suboxic deep waters in combination with spatially constrained thermocline anoxia (14–18). Evidently, redox chemistry changes in the Panthalassic Ocean are central to an examination of the links between global-ocean conditions and the end-Permian extinction event as well as the subsequent delayed biotic recovery.

The preservation of Permian–Triassic boundary deep-sea sediments is limited because most oceanic crust of that age has been subducted, and the only surviving Panthalassic seafloor sediments are within accretionary terranes or marginal uplifts now located in western Canada, Japan, and New Zealand (19, 20). In this study, we report analyses of all four sulfur isotopes (³²S, ³³S, ³⁴S, and ³⁶S) for pyrite from the deep-sea sediments of western Canada and Japan. Our study samples from western Canada were collected at Opal Creek in southwestern Alberta (50.67542°N, 115.07992°W), which represents an outer shelf or upper slope setting in the eastern Panthalassic Ocean at a paleolatitude of ∼30°N (21, 22) (Fig. 1). A second set of study samples is from Gujo-Hachiman in central Japan (35.7355°N, 136.8489°E), which represents the abyssal plain of the equatorial Panthalassic Ocean (23) (Fig. 1).

Fig. 1.

Paleogeography for the end-Permian world (∼252 Ma) and the respective location of the Opal Creek and Gujo-Hachiman sections (18, 22).

The Opal Creek section consists of black shale and siltstone of the lower Phroso Siltstone Member and upper Vega Siltstone Member of the Sulfur Mountain Formation (Fig. 2). The lowermost Sulfur Mountain Formation contains abundant siliceous monaxon sponge spicules and triaxon hyalosponge spicules that disappear abruptly at 0.36 m above the base of the formation (20). The disappearance of sponge spicules coincided with a sudden influx of Late Permian conodonts belonging to the genus Clarkina, marking the onset of the end-Permian extinction event (20) (Fig. 2). Four conodont zones were defined at Opal Creek, with the first occurrence of Hindeodus parvus and Clarkina taylorae indicating the Permian–Triassic boundary (20, 24) (Fig. 2).

Fig. 2.

Permian-Triassic biostratigraphy and multiple S-isotopic data for the Opal Creek (Left) and Gujo-Hachiman (Right) sections (filled red circles indicate negative Δ³³S). The red line corresponds to the end-Permian mass extinction horizon (EPME). Conodont Zonation at Opal Creek is after refs. 20 and 24: (1) Mesogondolella sheni Zone; (2) Clarkina hauschkei–Clarkina meishanensis Zone; (3) H. parvus–C. taylorae Zone; (4) C. taylorae–Clarkina cf. carinata Zone.

The Gujo-Hachiman section consists of three units: Unit I, comprising green–gray ribbon-chert; Unit II, comprising green–gray to black siliceous claystone; and Unit III, comprising black shale with a few thin lenses of gray chert and layers of white–yellow siliceous claystone bearing uppermost Permian radiolarians (17, 18, 25) (Fig. 2). Radiolarian and conodont data suggest that the Permian–Triassic boundary is located near the base of Unit III (17, 18, 25, 26).

Exact placement of the end-Permian extinction horizon in the two study sections allows us to use multiple sulfur isotopes to explore temporal changes of redox chemistry in the Panthalassic Ocean and their potential link to the end-Permian mass extinction as well as the following protracted marine biotic recovery.

The multiple sulfur isotope data of pyrite from the Opal Creek and Gujo-Hachiman sections are presented using conventional delta notation: δ³ⁱS = 1,000 × ((³ⁱS/³²S)_sample/(³ⁱS/³²S)_reference ‒ 1), where 3i = 33, 34, or 36. Capital delta notation is also defined to describe relationships involving the least abundant isotopes: Δ³³S = δ³³S − 1,000 × ((1 + δ³⁴S/1,000)^0.515 – 1), Δ³⁶S = δ³⁶S − 1,000 × ((1 + δ³⁴S/1,000)^1.90 – 1). The capital delta notation is defined with exponents of 0.515 and 1.90 to approximate the deviation from single-step low-temperature equilibrium exchange reactions (27, 28). Delta and capital delta values are given in units of per mille (‰). Figs. 2 and 3 present δ³⁴S and Δ³³S data of pyrite from the study sections (full analytical data of the samples and reference materials available as Tables S1–S3).

Fig. 3.

Cross-plot of δ³⁴S vs. Δ³³S for pyrite from the Opal Creek and Gujo-Hachiman sections. The colored nonlinear line represents a mixing line generated by the model (blue: Opal Creek, green: Gujo-Hachiman). The model field for sulfate reduction (black outlined field) and sulfate reduction combined with disproportionation (black dotted-line field) represent the field of all possible fractionations predicted by a model of sulfate reducers, and is larger than in ref. 12 because that field was calculated for a more limited range of fractionations produced by experiments with sulfate reducers. Whereas the expanded range of the model field may capture solutions that are non–steady-state, it provides a more conservative set of criteria to identify a mixing scenario than the field plotted by ref. 12.

Table S1.

Multiple S-isotopic data and standard deviation (1σ) of Opal Creek section

Sample	Depth, m	δ33S, ‰	δ34S, ‰	δ36S, ‰	Δ33S, ‰	Δ36S, ‰	1σ(δ33S), ‰	1σ(δ34S), ‰	1σ(Δ33S), ‰
OCS+1880	43.02	−13.42	−26.04	−49.93	0.075	−1.04	0.01	0.01	0.005
OCS+1825	42.47	−11.61	−22.58	−43.44	0.084	−0.98	0.00	0.01	0.005
OCS+1745	41.67	−13.77	−26.66	−50.96	0.052	−0.92	0.01	0.01	0.005
OCS+1360	37.82	−10.44	−20.30	−39.13	0.067	−0.91	0.02	0.03	0.003
OCS+1350	37.72	−2.97	−5.86	−11.70	0.055	−0.60	0.01	0.00	0.010
OCS+1285	37.07	−12.72	−24.67	−47.13	0.058	−0.78	0.01	0.01	0.010
OCS+1070	34.92	−7.96	−15.50	−29.79	0.048	−0.55	0.01	0.01	0.008
OCS+1005	34.27	−5.20	−10.14	−19.61	0.030	−0.44	0.00	0.00	0.004
OCS+845	32.67	−7.73	−15.04	−28.90	0.045	−0.51	0.01	0.00	0.005
OCS+670	30.92	−8.40	−16.28	−31.25	0.018	−0.54	0.01	0.01	0.017
OCS+615	30.37	−5.17	−10.04	−19.28	0.012	−0.29	0.01	0.01	0.009
OCS+565	29.87	−7.55	−14.71	−28.23	0.051	−0.46	0.02	0.01	0.016
OCS+530	29.52	−5.31	−10.30	−19.78	0.013	−0.30	0.00	0.01	0.005
OCS+485	29.07	−7.00	−13.56	−26.01	0.012	−0.40	0.01	0.01	0.007
OCS+440	28.62	−5.38	−10.42	−20.18	0.005	−0.48	0.01	0.03	0.009
OCS+395	28.17	−7.73	−15.01	−28.84	0.031	−0.51	0.02	0.01	0.012
OCS+205	26.27	−5.82	−11.31	−21.83	0.015	−0.45	0.01	0.01	0.010
OCS-240	21.82	−7.95	−15.46	−29.74	0.045	−0.56	0.01	0.01	0.006
OCS-325	20.97	−12.36	−23.95	−45.96	0.043	−0.95	0.01	0.01	0.008
OCS-390	20.32	−10.03	−19.42	−37.41	0.023	−0.83	0.01	0.01	0.016
OCS-580	18.42	−7.22	−14.00	−27.09	0.016	−0.65	0.01	0.01	0.007
OCS-760	16.62	−7.06	−13.72	−26.61	0.030	−0.70	0.01	0.00	0.005
OCS-835	15.87	−12.25	−23.73	−45.50	0.037	−0.89	0.01	0.00	0.010
OCS-945	14.77	−12.76	−24.69	−47.15	0.030	−0.77	0.02	0.03	0.007
OCS-1010	14.12	−10.77	−20.73	−39.77	−0.034	−0.74	0.02	0.03	0.003
OCS-1055	13.67	−8.57	−16.63	−31.94	0.032	−0.57	0.02	0.03	0.006
OCS-1085	13.37	−8.99	−17.35	−33.40	−0.014	−0.69	0.02	0.01	0.009
OCS-1125	12.97	−0.34	−0.66	−1.71	0.003	−0.46	0.01	0.01	0.013
OCS-1135	12.87	−0.47	−0.94	−2.19	0.017	−0.40	0.00	0.01	0.003
OCS-1175	12.47	−7.77	−15.06	−29.00	0.011	−0.58	0.01	0.00	0.010
OCS-1205	12.17	−4.93	−9.56	−18.50	0.003	−0.42	0.01	0.00	0.010
OCS-1235	11.87	−4.65	−8.99	−17.44	−0.013	−0.43	0.00	0.00	0.003
OCS-1275	11.47	−5.02	−9.74	−18.92	0.011	−0.49	0.01	0.01	0.004
OCS-1330	10.92	−4.96	−9.63	−18.66	0.011	−0.43	0.01	0.01	0.006
OCS-1355	10.67	−4.34	−8.43	−16.21	0.008	−0.26	0.00	0.01	0.006
OCS-1395	10.27	−5.84	−11.30	−21.79	−0.003	−0.43	0.01	0.01	0.007
OCS-1435	9.87	−5.29	−10.24	−19.75	−0.001	−0.39	0.01	0.01	0.002
OCS-1495	9.27	−3.61	−7.00	−13.23	0.005	0.03	0.01	0.01	0.010
OCS-1550	8.72	−4.34	−8.34	−15.98	−0.037	−0.19	0.00	0.00	0.004
OCS-1585	8.37	−3.50	−6.76	−12.97	−0.011	−0.17	0.00	0.00	0.001
OCS-1595	8.27	−6.75	−13.07	−25.07	−0.003	−0.38	0.01	0.00	0.010
OCS-1655	7.67	−7.62	−14.77	−28.39	0.014	−0.52	0.00	0.00	0.005
OCS-1680	7.42	−7.21	−13.94	−26.91	−0.008	−0.59	0.00	0.00	0.002
OCS-1715	7.07	−5.85	−11.32	−21.86	−0.001	−0.46	0.01	0.00	0.010
OCS-1765	6.57	−1.66	−3.26	−6.54	0.021	−0.36	0.01	0.01	0.006
OCS-1795	6.27	−5.12	−10.01	−19.55	0.047	−0.61	0.01	0.00	0.012
OCS-1890	5.32	−12.10	−23.46	−44.96	0.053	−0.86	0.02	0.00	0.018
OCS-1940(P)	5.12	−0.24	−0.59	−2.07	0.060	−0.95	0.01	0.00	0.005
OCS-1955	5.06	−10.76	−20.85	−39.89	0.036	−0.65	0.01	0.00	0.003
OCS-2015	4.45	−11.97	−23.21	−44.57	0.053	−0.94	0.01	0.00	0.005
OCS-2100	3.58	−11.16	−21.65	−41.85	0.048	−1.12	0.01	0.00	0.004
OCS-2160	2.96	−9.55	−18.49	−35.47	0.020	−0.63	0.01	0.00	0.006
OCS-2265	1.89	−8.25	−15.98	−30.96	0.013	−0.82	0.01	0.00	0.009
OCS-2305	1.48	−10.35	−20.06	−38.58	0.027	−0.81	0.00	0.00	0.005
OCS-2390	0.61	−15.09	−29.14	−55.24	0.029	−0.60	0.02	0.01	0.018
Chang+28	0.36	−11.96	−23.06	−44.02	−0.017	−0.66	0.00	0.00	0.003
Chang+12	0.22	−12.92	−24.97	−47.44	0.021	−0.53	0.01	0.00	0.005
Chang+2	0.14	−5.47	−10.56	−20.14	−0.015	−0.18	0.01	0.00	0.003
Chang+0	0.12	−5.28	−10.17	−19.65	−0.028	−0.41	0.01	0.01	0.010

Table S3.

Multiple S-isotopic data and standard deviation (1σ) of International Atomic Energy Agency (IAEA) reference materials

Sample	δ33S, ‰	δ34S, ‰	δ36S, ‰	Δ33S, ‰	Δ36S, ‰
IAEA-S1	−0.17	−0.50	−1.78	0.087	−0.82
	−0.13	−0.42	−1.43	0.086	−0.63
	−0.14	−0.44	−1.58	0.088	−0.74
	−0.17	−0.50	−1.70	0.089	−0.74
	−0.13	−0.43	−1.68	0.093	−0.86
Average	−0.15	−0.46	−1.64	0.088	−0.76
1σ	0.02	0.04	0.13	0.003	0.09
IAEA-S2	11.34	22.06	42.26	0.041	−0.07
	11.37	22.12	42.19	0.040	−0.25
	11.31	22.00	41.92	0.042	−0.29
	11.42	22.22	42.43	0.036	−0.21
	11.44	22.26	42.56	0.035	−0.15
Average	11.38	22.13	42.27	0.039	−0.19
1σ	0.05	0.11	0.25	0.003	0.09
IAEA-S3	−16.85	−32.61	−62.14	0.078	−1.10
	−16.89	−32.68	−62.16	0.080	−0.97
	−16.91	−32.73	−62.32	0.078	−1.05
	−16.91	−32.72	−62.25	0.077	−1.00
	−16.84	−32.59	−61.95	0.080	−0.95
Average	−16.88	−32.67	−62.16	0.079	−1.01
1σ	0.04	0.07	0.14	0.001	0.06

Before the end-Permian extinction, δ³⁴S compositions at Opal Creek vary from –24.97‰ to –10.17‰ with Δ³³S from –0.028‰ to +0.021‰ (Fig. 2). In the same interval at Gujo-Hachiman, δ³⁴S ranges from –40.18‰ to +13.27‰ with Δ³³S from –0.099‰ to +0.095‰ (Fig. 2). The end-Permian extinction horizon at Opal Creek (sample Chang+28) exhibits δ³⁴S of –23.06‰ and Δ³³S of –0.017‰ (Fig. 2). The same horizon at Gujo-Hachiman (sample ITJ-13) displays δ³⁴S of –25.24‰ and Δ³³S of +0.075‰ (Fig. 2). After the end-Permian extinction, δ³⁴S at Opal Creek varies from –29.14‰ to –0.59‰ with Δ³³S from –0.037‰ to +0.084‰ (Fig. 2). For the same interval, the Gujo-Hachiman section shows δ³⁴S values from –39.21‰ to –26.95‰ with Δ³³S of +0.061‰ to +0.143‰ (Fig. 2).

We use a previously described box model (29–32) to evaluate the significance of variation in our multiple sulfur isotopes for redox chemistry changes in the Panthalassic Ocean. In this model, covariation of δ³⁴S and Δ³³S and their positions in Quadrants I–IV provide insight regarding the sulfate reduction, sulfur oxidation, and sulfur disproportionation processes (Fig. 3) (calculations available in Sulfur-Cycle Model; Figs. S1 and S2). The model adopts an isotopic composition for seawater of δ³⁴S = +19.2‰ and Δ³³S = +0.022‰, as estimated from multiple S-isotopic analysis of contemporaneous marine carbonate-associated sulfate (33).

Fig. S1.

Network structure for metabolic pathways of dissimilatory sulfate reduction modified after refs. 30 and 32. Sulfate is reduced to sulfide by sulfate-reducing bacteria with multiple enzyme-catalyzed processes. The sulfur reservoirs include external and internal sulfate reservoirs, an APS (adenosine-5′-phosphosulfate) reservoir, a sulfite reservoir, a trithionate reservoir, a thiosulfate reservoir, and internal and external sulfide reservoirs. The network involves transfer pathways of sulfur: SoSi, SiSo, SiA, ASi, AS, SA, STTh, TThS, TThTS, TSTTh, TThS-D, TSS, TSHi, HiTS, HiHo, HoHi, and their associated isotopic fractionation factors (e.g., ^jα_SiSo, ^jα_SA, ^jα_AS, etc.).

Fig. S2.

Sulfur cycle with concurrent operation of MSR and sulfur disproportionation processes. Sulfate can be reduced to sulfide through the MSR process; part of the sulfide can then be reoxidized into sulfur intermediate valence products (e.g., S⁰, SO₃^2-). The reoxidation products can be transformed to sulfate and sulfide through the disproportionation process.

At Opal Creek before the end-Permian mass extinction, pyrite shows negative δ³⁴S in combination with either positive Δ³³S (Quadrant II) or negative Δ³³S (Quadrant III) (Figs. 2 and 3). At the extinction horizon, pyrite exhibits negative Δ³³S with negative δ³⁴S (Quadrant III) (Figs. 2 and 3). After the extinction, pyrite shows positive Δ³³S with negative δ³⁴S (Quadrant II) (Figs. 2 and 3). However, negative Δ³³S with negative δ³⁴S values (Quadrant III) are observed for an interval between ∼7 m and ∼14 m (Figs. 2 and 3). Higher in the Sulfur Mountain Formation, all multiple S-isotopic data show positive Δ³³S with negative δ³⁴S (Quadrant II) (Figs. 2 and 3).

At Gujo-Hachiman before the extinction event, the lower to middle part of Unit I exhibits positive Δ³³S with positive δ³⁴S (Quadrant I) or negative δ³⁴S (Quadrant II) (Figs. 2 and 3). The upper part of Unit I and lower Unit II are characterized by negative Δ³³S (Quadrants III and IV) except for two samples with positive Δ³³S (Quadrant II) (Figs. 2 and 3). The extinction horizon exhibits positive Δ³³S with negative δ³⁴S (Quadrant II), as do all samples from higher in the section (Figs. 2 and 3).

The paleoenvironmental significance of multiple S-isotopic data from the Opal Creek and Gujo-Hachiman sections can be inferred based on their positions in Quadrants I–IV (Fig. 3). Data in Quadrant II (positive Δ³³S with negative δ³⁴S) may be due to fractionation relationships produced by microbial sulfate reduction, disproportionation of sulfur intermediate compounds, and the conservation of mass in the sulfur cycle (30). Quadrant II samples situated within the sulfate reduction field (black outlined field) imply sulfate reduction in an open system in which disproportionation could occur, but is not required (Fig. 3). However, other samples in Quadrant II (positive Δ³³S with negative δ³⁴S) as well as two samples in Quadrant I (positive Δ³³S with positive δ³⁴S) are situated outside the sulfate reduction field but within the sulfur disproportionation field (black dotted-line field), suggesting that they may have been produced by a combination of sulfate reduction and sulfur disproportionation (Fig. 3). It is important to note that the inclusion of sulfur disproportionation in our model is not to imply that this process must occur in any of these environments, but simply because its operation will expand the field of steady-state solutions and reduce the possibility of false identification of mixing scenarios.

S-isotopic data in Quadrants III and IV cannot be attributed simply to either sulfate reduction and/or sulfur disproportionation (12, 29, 34). Although a few negative Δ³³S values within the disproportionation field might be due to a combination of sulfate reduction and sulfur disproportionation (Fig. 3), this interpretation is not strongly supported by observations from biological experiments of sulfate reduction and/or sulfur disproportionation. Biological culture experiments have shown that, under sulfate-unlimited conditions, sulfide produced by sulfate reduction is more enriched in ³³S compared with reactant sulfate (35–37). Under sulfate-limited conditions, the sulfide generated by sulfate reduction should have similar δ³⁴S and Δ³³S to the reactant sulfate if the latter is completely reduced. Also, culture experiments of sulfur disproportionation have shown that products of sulfide and sulfate are enriched in ³³S relative to the reactant elemental sulfur (38). In this case, the negative Δ³³S projected by the model may be due to the specified boundary conditions used in the model, e.g., the proportion between sulfide that is reoxidized and disproportionated and sulfide produced by sulfate reduction. Previous studies have shown that mixing of sulfide having strongly negative δ³⁴S and positive Δ³³S with sulfide having positive δ³⁴S and positive Δ³³S (the latter being similar to seawater sulfate) can produce sulfide with negative Δ³³S values (12, 29, 34). This negative Δ³³S mixing signature is thus interpreted to account for the negative Δ³³S compositions observed in the Opal Creek and Gujo-Hachiman sections.

Negative Δ³³S compositions have been linked to shoaling of sulfidic waters and resultant mixing of sulfidic and oxic waters in the Permian oceans before or coincident with the mass extinction (12, 29). In this interpretation, shoaling of sulfidic waters changed the sulfur cycle within the sediment, yielding a mixture of pyrite already formed in an open-system environment with pyrite formed in a nearly closed-system environment caused by a sharp reduction of bioturbation as a result of the demise of benthic fauna by sulfide poisoning.

Two endmembers were used to generate the mixing lines in Fig. 3. One endmember is the isotopic composition of latest Permian–Early Triassic seawater sulfate and the other is the most negative δ³⁴S value observed at Opal Creek and Gujo-Hachiman, respectively (Fig. 3). The ranges of the mixing lines encompass all negative Δ³³S data from the two study sections (Fig. 3).

Before the end-Permian extinction, a combination of negative δ³⁴S and Δ³³S values indicates shoaling of sulfidic waters and transition to and from sulfidic condition at Opal Creek (Fig. 2). A similar pattern in the correlative interval at Gujo-Hachiman also provides evidence for the timing of the onset of euxinia and oscillations between sulfidic and oxic conditions (Fig. 2). These results are consistent with the multiple S-isotopic data from the Paleotethys Ocean that suggest shoaling of sulfidic waters and resultant mixing of sulfidic and oxic waters before the end-Permian mass extinction event (12) (Fig. 4). Our multiple S-isotopic data from Opal Creek and Gujo-Hachiman thus imply that shoaling of sulfidic waters and oscillations between sulfidic and oxic conditions may be of global significance before the end-Permian extinction.

Fig. 4.

The negative Δ³³S data from Opal Creek and Gujo-Hachiman (this study) and from Meishan (12) are aligned using the extinction horizon as a datum. The thickness of each conodont zone is not to scale.

The negative Δ³³S with negative δ³⁴S at Opal Creek coincided exactly with the extinction horizon, linking sulfidic conditions to the extinction event (Figs. 2 and 4). This observation suggests that shoaling of sulfidic waters and oscillations between sulfidic and oxic conditions may have driven the end-Permian mass extinction. The lack of negative Δ³³S during the extinction at Gujo-Hachiman (Figs. 2 and 4) might reflect preservation biases, which could be tested by future multiple S-isotopic analyses of pyrite from the extinction horizon in other sections.

Above the end-Permian mass extinction horizon, positive Δ³³S with negative δ³⁴S values are present in the Opal Creek and Gujo-Hachiman sections (Fig. 2). At Opal Creek, abundant small framboidal pyrites and trace metal data provide evidence that the lower 4 m of the Sulfur Mountain Formation were deposited under sulfidic conditions (22). The absence of negative Δ³³S values in this interval and Unit III at Gujo-Hachiman suggests that accumulation of abundant syngenetic pyrite having positive Δ³³S compositions overwhelmed the signal of negative Δ³³S, consistent with sustained euxinic conditions. The reappearance of negative Δ³³S above this interval at Opal Creek (Figs. 2 and 4) provides evidence of an episodic incursion of sulfidic waters and oscillations between sulfidic and oxic conditions during the Early Triassic. Conditions of this type may have contributed to the delayed marine biotic recovery, as reflected in the disappearance of benthic fauna and bioturbation in the Phroso Siltstone Member of the Sulfur Mountain Formation (20).

Our multiple S-isotopic study provides insights into temporal changes of redox chemistry in the latest Permian–Early Triassic Panthalassic Ocean. Our data suggest that episodic shoaling of sulfidic waters, oscillations between sulfidic and oxic conditions, and related environmental deterioration led up to the end-Permian extinction. The negative Δ³³S anomaly at the extinction horizon at Opal Creek provides critical evidence that shoaling of sulfidic waters and oscillations between sulfidic and oxic conditions may have been the main killing agents during the mass extinction. Moreover, the episodic incursion of sulfidic waters and mixing of sulfidic and oxic waters indicated by our multiple S-isotopic data may have played an important role in the protracted marine biotic recovery from the end-Permian mass extinction. Future isotopic studies carried out in the framework of biostratigraphy and sedimentary facies worldwide can test our model and improve our understanding of the relationships between redox chemistry changes and the mass extinction as well as biotic recovery.

Sulfur-Cycle Model

The sulfur-cycle model used in this study incorporates three processes: microbial sulfate reduction (MSR), sulfur disproportionation (SD), and mixing process. The results of the MSR calculations (based on refs. 29–32) correspond to the field in Fig. 3 in which disproportionation could occur but is not required. The results of the SD calculations correspond to the field in Fig. 3 in which disproportionation is required in addition to MSR. The calculations of the mixing process were presented in ref. 29, table DR5.

MSR Process.

The details of the steady-state microbial sulfate reduction process were given in refs. 30 and 32 and are outlined briefly below. A series of branch terms (fi) is defined to illustrate the division of sulfur flux (φ_i) within the metabolic network (Fig. S1), and can be generally expressed as

$$f_1=\frac{{\phi }_1}{{\displaystyle \sum _{i=\mathrm{1..}n}{\phi }_i}},f_2=\frac{{\phi }_2}{{\displaystyle \sum _{i=\mathrm{2..}n}{\phi }_i}},f_{n-1}=\frac{{\phi }_{n-1}}{{\displaystyle \sum _{i=n-1,n}{\phi }_i}},$$ [S1]

where φ_i represents the flux of sulfur transferred along pathway i and all f values range between 0 and 1.

The mass conservation equations for different reservoirs in the network are written as

$${}^{j}R_{\text{TSHi}}=f_{\text{HiTS}}{}^{j}R_{\text{HiTS}}+\left(1-f_{\text{HiTS}}\right){}^{j}R_{\text{in/out}},$$ [S2]

$$\begin{array}{c}{}^{j}R_{\text{TThTS}}=f_{\text{TSTTh}}\frac{{}^j\alpha _{\text{TSTTh}}}{{}^j\alpha _{\text{TSHi}}}{}^{j}R_{\text{TSHi}}+\left(1-f_{\text{TSTTh}}\right)\\ \times \left[f_{\text{TSS}}\frac{{}^j\alpha _{\text{TSS}}}{{}^j\alpha _{\text{TSHi}}}{}^{j}R_{\text{TSHi}}+\left(1-f_{\text{TSS}}\right){}^{j}R_{\text{in/out}}\right],\end{array}$$ [S3]

$$\begin{array}{c}{}^{j}R_{\text{STTh}}=f_{\text{TThS}}\frac{{}^j\alpha _{\text{TThS}}}{{}^j\alpha _{\text{TThTS}}}{}^{j}R_{\text{TThTS}}+\left(1-f_{\text{TThS}}\right)\\ \times [f_{\text{TThS}-\text{D}}\frac{{}^j\alpha _{\text{TThS}-\text{D}}}{{}^j\alpha _{\text{TThTS}}}{}^{j}R_{\text{TThTS}}+\left(1-f_{\text{TThS}-\text{D}}\right)\\ \times \left[f_{\text{TSS}}\frac{{}^j\alpha _{\text{TSS}}}{{}^j\alpha _{\text{TSHi}}}{}^{j}R_{\text{TSHi}}+\left(1-f_{\text{TSS}}\right){}^{j}R_{\text{in/out}}\right]],\end{array}$$ [S4]

where ^jR = (^jS/³²S) and ^jα_pathway refers to the isotopic fractionation factors (^jS/³²S)_pathway/(^jS/³²S)_reservoir (j = 33, 34, 36). Then the fractionation factors between sulfur reservoirs within the subunits and the sulfur flux out of each subunit can be solved:

$${}^j\alpha _{\text{TS}-\text{out}}=\frac{f_{\text{HiTS}}\left({}^j\alpha _{\text{HiTS}}-1\right)+1}{{}^j\alpha _{\text{TSHi}}},$$ [S5]

$${}^j\alpha _{\text{TTh}-\text{out}}=\frac{f_{\text{TSTTh}}{}^j\alpha _{\text{TSTTh}}{}^j\alpha _{\text{TS}-\text{out}}+\left(1-f_{\text{TSTTh}}\right)\left[f_{\text{TSS}}{}^j\alpha _{\text{TSS}}{}^j\alpha _{\text{TS}-\text{out}}+\left(1-f_{\text{TSS}}\right)\right]}{{}^j\alpha _{\text{TThTS}}},$$ [S6]

$$\begin{array}{c}{}^j\alpha _{\text{S}-\text{out}}=\frac{f_{\text{TThS}}{}^j\alpha _{\text{TThS}}{}^j\alpha _{\text{TTh}-\text{out}}}{{}^j\alpha _{\text{STTh}}}+\frac{\left(1-f_{\text{TThS}}\right)f_{\text{TThS}-\text{D}}{}^j\alpha _{\text{TThS}-\text{D}}{}^j\alpha _{\text{TTh}-\text{out}}}{{}^j\alpha _{\text{STTh}}}\\ +\frac{\left(1-f_{\text{TThS}}\right)\left[\left(1-f_{\text{TThS}-\text{D}}\right)\left[f_{\text{TSS}}{}^j\alpha _{\text{TSS}}{}^j\alpha _{\text{TS}-\text{out}}+\left(1-f_{\text{TSS}}\right)\right]\right]}{{}^j\alpha _{\text{STTh}}}.\end{array}$$ [S7]

Finally, the sulfur isotopic fractionation associated with the entire metabolic network is obtained:

$${}^j\alpha _{{\text{SO}}_4-\text{sulfide}}=\frac{f_{\text{SiSo}}\left({}^j\alpha _{\text{SiSo}}\times \left[\frac{f_{\text{SA}}\left[\left({}^j\alpha _{\text{SA}}\times {}^j\alpha _{\text{S}-\text{out}}\right)-1\right]+1}{{}^j\alpha _{\text{AS}}}\right]-1\right)+1}{{}^j\alpha _{\text{SoSi}}},$$ [S8]

and

$$\delta {}^{34}\text{S}_{{\text{SO}}_4-\text{sulfide}}=\mathrm{1,000}\left({}^{34}\alpha _{{\text{SO}}_4-\text{sulfide}}-1\right),$$ [S9]

$$\mathrm{\Delta }{}^{33}\text{S}_{{\text{SO}}_4-\text{sulfide}}=\mathrm{1,000}\left[{}^{33}\alpha _{{\text{SO}}_4-\text{sulfide}}-{\left({}^{34}\alpha _{{\text{SO}}_4-\text{sulfide}}\right)}^{0.515}\right].$$ [S10]

Consequently, when the sulfur isotopic compositions of ancient seawater sulfate (δ³⁴S_SO4, Δ³³S_SO4) are given, the sulfur isotopic compositions of sulfide produced by the MSR pathway can be calculated:

$$\delta {}^{34}\text{S}_{\text{sulfide}}=\mathrm{1,000}\left[\left(\frac{\delta {}^{34}\text{S}_{{\text{SO}}_4}}{\mathrm{1,000}}+1\right)\times 1/{}^{34}\alpha _{{\text{SO}}_4-\text{sulfide}}-1\right],$$ [S11]

$$\begin{array}{c}\mathrm{\Delta }{}^{33}\text{S}_{\text{sulfide}}=\mathrm{1,000}[1/{}^{33}\alpha _{{\text{SO}}_4-\text{sulfide}}\times \left[\frac{\mathrm{\Delta }{}^{33}\text{S}_{{\text{SO}}_4}}{\mathrm{1,000}}+{\left(\frac{\delta {}^{34}\text{S}_{{\text{SO}}_4}}{\mathrm{1,000}}+1\right)}^{0.515}\right]\\ -{\left[\left(\frac{\delta {}^{34}\text{S}_{{\text{SO}}_4}}{\mathrm{1,000}}+1\right)\times 1/{}^{34}\alpha _{{\text{SO}}_4-\text{sulfide}}\right]}^{0.515}].\end{array}$$ [S12]

Sulfur Disproportionation Combined with MSR.

The sulfur isotope mass balance for sulfide is (Fig. S2)

$$\frac{d}{dt}\left(m_{\text{sulfide}}·{}^{j}R_{\text{sulfide}}\right)=m2·{}^j\alpha _{\text{MSR}}·{}^{j}R_{\text{sulfate}}-m3·{}^j\alpha _{\text{reox}-\text{Disp}}·{}^{j}R_{\text{sulfide}}-m1·{}^j\alpha _{m1}·{}^{j}R_{\text{sulfide}},$$ [S13]

where m represents the masses of reservoirs (sulfide) or fluxes (e.g., MSR, reox-Disp) at time t, and ^jα_MSR has been solved in Eq. S8. The fractionation factors of disproportionation (^jα_reox-Disp) through the reoxidative (reox) branch are modified from ref. 31, table S2 and ref. 38, table 2.

Table S2.

Multiple S-isotopic data and SD (1σ) of Gujo-Hachiman section

Sample	Depth, cm	δ33S, ‰	δ34S, ‰	δ36S, ‰	Δ33S, ‰	Δ36S, ‰	1σ(δ33S), ‰	1σ(δ34S), ‰	1σ(Δ33S), ‰
ITJ-1	58	−16.94	−32.85	−62.81	0.114	−1.33	0.03	0.05	0.004
ITJ-2	54	−17.05	−33.08	−63.38	0.126	−1.46	0.03	0.04	0.013
ITJ-3	44	−14.36	−27.93	−53.72	0.119	−1.33	0.02	0.03	0.004
ITJ-4	41	−13.83	−26.95	−52.05	0.143	−1.46	0.01	0.03	0.010
ITJ-5	36	−13.91	−27.09	−52.33	0.130	−1.49	0.03	0.04	0.005
ITJ-6	32	−19.83	−38.26	−72.80	0.061	−1.36	0.02	0.02	0.017
ITJ-7	28	−20.30	−39.21	−74.51	0.091	−1.32	0.03	0.04	0.008
ITJ-8	24	−18.10	−35.07	−67.03	0.116	−1.45	0.01	0.02	0.006
ITJ-9	19	−19.15	−36.99	−70.37	0.077	−1.26	0.01	0.02	0.003
ITJ-10	14	−16.80	−32.56	−62.23	0.106	−1.27	0.01	0.01	0.001
ITJ-11	8	−15.89	−30.76	−58.77	0.070	−1.13	0.01	0.01	0.005
ITJ-12	5	−19.44	−37.52	−71.42	0.064	−1.33	0.02	0.01	0.012
ITJ-13	0	−13.00	−25.24	−48.75	0.075	−1.34	0.01	0.01	0.008
ITJ-14	−10	−14.63	−28.24	−53.63	0.009	−0.66	0.02	0.01	0.008
ITJ-15	−12	−4.62	−8.76	−16.24	−0.099	0.35	0.01	0.02	0.004
ITJ-16	−15	−3.15	−6.01	−11.41	−0.048	−0.03	0.01	0.01	0.007
ITJ-17	−19	−4.42	−8.48	−15.97	−0.046	0.08	0.02	0.01	0.014
ITJ-18	−22	−19.88	−38.19	−72.12	−0.033	−0.82	0.02	0.02	0.007
ITJ-19	−27	−19.05	−36.63	−69.43	−0.014	−0.98	0.01	0.01	0.000
ITJ-20	−39	6.61	12.91	24.77	−0.015	0.09	0.01	0.01	0.013
ITJ-21	−47	−15.99	−30.87	−58.61	0.027	−0.77	0.02	0.04	0.010
ITJ-22	−52	−7.07	−13.67	−26.08	−0.002	−0.26	0.01	0.03	0.002
ITJ-25	−61	−20.88	−40.18	−75.74	0.021	−0.77	0.01	0.00	0.008
ITJ-26	−67	6.07	11.80	22.34	0.009	−0.19	0.01	0.01	0.005
ITJ-27	−77	−13.35	−25.95	−49.76	0.095	−1.04	0.01	0.00	0.003
ITJ-28	−89	−18.77	−36.15	−68.76	0.007	−1.20	0.03	0.05	0.003
ITJ-47	−298	−0.87	−1.80	−4.07	0.059	−0.66	0.01	0.01	0.007
ITJ-49	−316	6.81	13.27	25.27	0.004	−0.09	0.01	0.01	0.010

As there is no distinct sulfur isotope fractionation of the pathway for transfer of sulfide from cell to external pool (m1 in Fig. S2),

$${}^j\alpha _{m1}=1.$$ [S14]

Assuming that the size and isotope composition of the sulfide pool are at steady state, the derivatives after the time of sulfide production (Eq. S13) become zero:

$$m2\times {}^j\alpha _{\text{MSR}}\times {}^{j}R_{\text{sulfate}}=m3\times {}^j\alpha _{\text{reox}-\text{Disp}}\times {}^{j}R_{\text{sulfide}}+m1\times {}^{j}R_{\text{sulfide}}=\left(m3\times {}^j\alpha _{\text{reox}-\text{Disp}}+m1\right)\times {}^{j}R_{\text{sulfide}}.$$ [S15]

As

$$m2=m1+m3,$$ [S16]

$$f_{\text{reox}-\text{Disp}}=\frac{m3}{m2}.$$ [S17]

Rearranging Eq. S15,

$$\frac{{}^{j}R_{\text{sulfide}}}{{}^{j}R_{\text{sulfate}}}=\frac{m2\times {}^j\alpha _{\text{MSR}}}{m3\times {}^j\alpha _{\text{reox}-\text{Disp}}+m1}=\frac{{}^j\alpha _{\text{MSR}}}{f_{\text{reox}-\text{Disp}}\times {}^j\alpha _{\text{reox}-\text{Disp}}+\left(1-f_{\text{reox}-\text{Disp}}\right)}.$$ [S18]

Therefore, the total fractionation factor of disproportionation is

$${}^j\alpha _{\text{Disp}-\text{total}}=\frac{{}^j\alpha _{\text{MSR}}}{f_{\text{reox}-\text{Disp}}\times {}^j\alpha _{\text{reox}-\text{Disp}}+\left(1-f_{\text{reox}-\text{Disp}}\right)}.$$ [S19]

Similar to the MSR process (Eqs. S11 and S12), the S-isotopic composition of pyrite generated via disproportionation can be calculated as (Fig. S2)

$$\delta {}^{34}S_{\text{sulfide}}=\mathrm{1,000}\left[\left(\frac{\delta {}^{34}S_{{\text{SO}}_4}}{\mathrm{1,000}}+1\right)\times {}^{34}\alpha _{\text{Disp}-\text{total}}-1\right],$$ [S20]

$$\begin{array}{c}{\mathrm{\Delta }}^{\mathrm{33}}{S}_{\text{sulfide}}=\mathrm{1,000}{[}^{33}{\alpha }_{\text{Disp}-\text{total}}\times \\ \left[\frac{{\mathrm{\Delta }}^{\mathrm{33}}{S}_{{\text{SO}}_4}}{\mathrm{1,000}}+{\left(\frac{{\delta }^{34}{S}_{{\text{SO}}_4}}{\mathrm{1,000}}+1\right)}^{0.515}\right]-{\left[\left(\frac{{\delta }^{34}{S}_{{\text{SO}}_4}}{\mathrm{1,000}}+1\right){\times }^{34}{\alpha }_{\text{Disp}-\text{total}}\right]}^{0.515}].\end{array}$$ [S21]

Methods

Pyrite sulfur was extracted and trapped as Ag₂S (39). During this procedure, the product H₂S was carried by nitrogen gas through a condenser and a bubbler filled with milli-Q water, and collected as silver sulfide by reacting with silver nitrate solution.

For multiple S-isotopic analysis, Ag₂S was converted to SF₆ quantitatively by a fluorination reaction in a Ni reaction vessel with a 10-fold excess of F₂ at 250 °C for 8 h (12). After the reaction, SF₆ was purified cryogenically by distillation in a liquid N₂–ethanol slurry at −110 °C, and chromatographically on a 120 molecular sieve 5 Å/Hasep Q column with a thermal conductivity detector (TCD). The He carrier flow was set at 20 mL/min. The SF₆ peak was registered on a TCD and then isolated by freezing into a liquid-nitrogen-cooled trap. The isotopic abundance of the purified SF₆ was analyzed on a Finnigan MAT 253 at m/e⁻ values of 127, 128, 129, and 131 (³²SF₅⁺, ³³SF₅⁺, ³⁴SF₅⁺, ³⁶SF₅⁺).

One-sigma uncertainties on mass-dependent reference materials are better than ±0.2‰, ±0.01‰, and ±0.2‰ in δ³⁴S, Δ³³S, and Δ³⁶S, respectively. Uncertainties on the measurements reported here are estimated to be better than ±0.2‰, ±0.01‰, and ±0.2‰ for δ³⁴S, Δ³³S, and Δ³⁶S, respectively.