Multiple Sulfur Isotope Geochemistry during the Permian-Triassic Transition

Abstract

The end-Permian mass extinction was the largest biodiversity crisis in the Phanerozoic. Based on characteristic negative ∆³³S signals of sedimentary pyrite, previous multiple sulfur isotope studies suggested shoaling of anoxic/sulfidic deep-waters onto a shelf, leading to the shallow-marine extinction. However, the validity of this shoaling model has been controversial. I compiled previously-reported multiple sulfur isotope records during the Permian-Triassic transition interval, and examined a stratigraphic relationship between the extinction horizon, redox oscillation in the depositional settings, and the multiple sulfur isotope record in each studied section. The compilation shows that the negative ∆³³S signals do not correspond clearly to the extinction horizon or to the benthic anoxia/euxinia in the studied sections. The compilation also documents that the multiple sulfur isotope records during the Permian-Triassic transition are substantially variable, and that the negative ∆³³S signals were observed in various types of sediments including shallow-marine carbonates, carbonates/siltstones of relatively deep-water facies, and abyssal deep-sea cherts. Those observations allow me to infer that the negative ∆³³S signal is not a robust indicator of shoaling. Rather, this isotopic signal may reflect substantial sulfur isotope heterogeneity in the sediments controlled by local factors.

1. Introduction

The end-Permian mass extinction was the largest biodiversity decline in the Phanerozoic (e.g., [1]). Although the major extinction across the Permian-Triassic boundary (P-TB) (ca. 251.9 Ma) has been extensively studied since the 1980s (e.g., [2,3]), numerous studies revealed repeated biotic crises with protracted environmental deterioration during the Permian-Triassic transition interval (Figure 1a) (e.g., [4,5,6]), including the end-Guadalupian (Middle Permian) extinction (ca. 259.5 Ma) [7,8,9], and biotic upheavals in the Early Triassic (ca. 251.9 to 246.7 Ma) [10] with the extinction immediately before the Smithian-Spathian boundary (S-SB; ca. 250.6 Ma) in the Olenekian (Late Early Triassic) (e.g., [11,12]). Several geologic events around the Guadalupian-Lopingian (Late Permian) boundary (G-LB) have been proposed as the cause of the end-Guadalupian extinction, such as the eruption of the Emeishan flood basalts (e.g., [13,14]), a eustatic sea-level fall [7], a global cooling [15], and oceanic anoxia/euxinia [16,17,18]. As the cause of the P-TB extinction, a number of geologic phenomena have been proposed, including a bolide impact (e.g., [19]), volcanism of the Siberian Traps Large Igneous Province (STLIP) (e.g., [20,21]) concurrent with climate warming (e.g., [22]), anoxia/euxinia (e.g., [23,24]), hypercapnia [25], and oceanic acidification (e.g., [26]).

Previous studies also revealed biotic upheavals following the P-TB extinction in the Early Triassic (Figure 1a), which were presumably driven by sustained environmental stresses in this interval, such as recurrent STLIP volcanism (e.g., [27]), elevated pCO₂ (e.g., [28,29,30]) and pCH₄ (e.g., [31,32]), global warming [33], and ocean acidification [34] and anoxia (e.g., [10,35,36,37]). Especially, the late-Smithian extinction event was one of the severest biodiversity crises during the delayed recovery interval of the Early Triassic (Figure 1a) (e.g., [11]), and was characterized by a turnover in marine fauna including ammonoids, conodonts, and other invertebrates (e.g., [38,39]), and by a change in terrestrial ecosystems [11]. Previous studies suggested that the late-Smithian extinction generally coincided with several geologic events, such as the recurrent STLIP activity [40], large carbon cycle fluctuations [11,41], and extreme greenhouse conditions (late Smithian Thermal Maximum) [33,42] followed by a major cooling around the S-SB associated with ocean anoxia [12,43] and/or acidification [34]. Nevertheless, the ultimate trigger mechanisms of the “triple” extinction during the Permian-Triassic transition are not yet fully understood and are still hotly debated (e.g., [3,44]).

Multiple sulfur isotope geochemistry is useful for tracking the global sulfur cycle through Earth history (e.g., [45,46]). Several previous studies analyzed multiple sulfur isotopes in sedimentary records in the Permian-Triassic transition interval (Figure 1b), and tried to reconstruct the sedimentary sulfur cycle in association with bottom water redox conditions and benthos activity during the three extinction events. In particular, based on an anomalous sulfur isotopic signal observed in the analyzed sediments, initial studies proposed shoaling of anoxic/sulfidic deep-waters onto a shelf, leading to the shallow-marine extinction. However, the validity of this shoaling model has been questioned by later studies. In this article, I briefly review the multiple sulfur isotope system and its initial application to the Permian sedimentary records with the original shoaling models. Then, I review later multiple sulfur isotope studies in the Permian-Triassic transition interval, particularly focusing on a stratigraphic relationship between the extinction horizon, redox conditions in the depositional settings, and the multiple sulfur isotope record in each studied section. Finally, I compile all the previous multiple sulfur isotope records during the Permian-Triassic transition and examine the validity of the original shoaling models.

Figure 1. The Permian-Triassic transition interval. (a) “triple” extinction during the interval with the ages of the boundaries [47,48]; (b) paleo-geographic map [49] with sections of previous multiple sulfur isotope studies. The location of Gujo-Hachiman in mid-Panthalassa is not well constrained. E. Triassic: Early Triassic, M. Tri.: Middle Triassic, W.: Wordian, Wuchiaping.: Wuchiapingian, Chang.: Changhsingian, I.: Induan, Gries.: Griesbachian, D.: Dienerian, Smith.: Smithian.

Figure 1. The Permian-Triassic transition interval. (a) “triple” extinction during the interval with the ages of the boundaries [47,48]; (b) paleo-geographic map [49] with sections of previous multiple sulfur isotope studies. The location of Gujo-Hachiman in mid-Panthalassa is not well constrained. E. Triassic: Early Triassic, M. Tri.: Middle Triassic, W.: Wordian, Wuchiaping.: Wuchiapingian, Chang.: Changhsingian, I.: Induan, Gries.: Griesbachian, D.: Dienerian, Smith.: Smithian.

2. Multiple Sulfur Isotope Geochemistry and Its Initial Application to the Permian Sedimentary Records

2.1. Multiple Sulfur Isotope System

Sulfur has four stable isotopes (³²S, ³³S, ³⁴S, and ³⁶S), and the sulfur isotopic composition of sulfur-bearing material is described conventionally using delta notation:

δ^xS = ((^xS/³²S)_sample/(^xS/³²S)_reference − 1) × 1000 (x = 33, 34, or 36),

(1)

where (^xS/³²S)_sample and (^xS/³²S)_reference are the isotope ratios of sample and reference material, respectively. For the multiple sulfur isotope systematics, capital delta notation is generally used to describe a deviation from a mass-dependent fractionation (MDF) relationship, according to the following formulas [50]:

∆³³S = δ³³S − ((1 + δ³⁴S/1000)^0.515 − 1) × 1000,

(2)

and

∆³⁶S = δ³⁶S − ((1 + δ³⁴S/1000)^1.90 − 1) × 1000.

(3)

The δ³⁴S, ∆³³S, and ∆³⁶S values are reported in ‰ relative to the Vienna Cañon Diablo Troilite (V-CDT) standard.

Multiple sulfur isotope geochemistry has been a useful tool for understanding the sulfur cycle in the Archean [51], as Archean sedimentary records are characterized by their unique isotopic signals of mass-independent fractionation of sulfur (S-MIF) with substantially large ∆³³S (up to +15‰). Photochemical reactions of sulfur dioxide in the reducing atmosphere have been a leading candidate for the main S-MIF yielding mechanism for the Archean records ([52] and references therein). After the Great Oxidation Event at ca. 2.3 Ga, the ∆³³S values of Proterozoic and Phanerozoic sediments are substantially small (generally within 0 ± 0.2‰). [46,52]. However, in many cases, the ∆³³S values of the younger sediments deviate slightly but significantly from zero. These nonzero ∆³³S records in the Proterozoic and Phanerozoic, which are detectable by high-precision multiple sulfur isotope measurements using a fluorination technique, most likely reflected MDF processes in the oceanic biogeochemical cycles (e.g., [53]).

Sulfur isotope fractionation during a MDF process, such as microbial sulfate reduction (MSR), can be described generally by the ³⁴ε and ³³λ values [54]:

³⁴ε = (1 − (³⁴S/³²S)_product/(³⁴S/³²S)_reactant) × 1000,

(4)

and

³³λ = ln[(³³S/³²S)_product/(³³S/³²S)_reactant]/ln[(³⁴S/³²S)_product/(³⁴S/³²S)_reactant],

(5)

where (^xS/³²S)_product and (^xS/³²S)_reactant are the instantaneous sulfur isotope ratios of product and the remaining reactant, respectively (x = 33 or 34). The ³³λ value is approximately 0.515 during an equilibrium exchange reaction at temperature between 0 °C and 100 °C [50]. However, the ³³λ value is generally lower than 0.515 during kinetic MSR, and the ∆³³S value of instantaneously produced hydrogen sulfide (H₂S) is higher than the value of substrate sulfate [54]. Another consequence of the multiple sulfur isotope system is a nonlinear effect of sulfur mixing on ∆³³S [50]. When ³⁴S-depleted and ³⁴S-enriched sulfur are mixed, the ∆³³S value of mixed sulfur is generally lower than the values of initial sulfur pools. On a δ³⁴S-∆³³S cross plot, the mixed sulfur is along a convex downward curve between the ³⁴S-depleted and ³⁴S-enriched endmembers, according to the mixing ratio. In particular, a negative ∆³³S signal cannot be produced solely via a normal kinetic process (such as MSR) but suggests a mixing of ³⁴S-depleted and ³⁴S-enriched sulfur [50].

2.2. Original Shoaling Model for the P-TB Event

Shen et al. [55] first analyzed the multiple sulfur isotopes in the end-Permian sedimentary records. They reported the quadruple sulfur isotopic composition of sedimentary pyrite in P-TB shelf carbonates/shales at Meishan in South China, the Global Stratotype Section and Point (GSSP) for the P-TB (Figure 1b) [56]. The P-TB successions at Meishan have been extensively studied and previous studies revealed multiple lines of evidence for environmental changes around the P-TB, including the abrupt extinction [2,57], sea-level changes (e.g., [58,59]), a negative carbon isotopic excursion (e.g., [57,60]), petrological and geochemical evidence for anoxia/euxinia (e.g., [58,61,62]), oxygen and calcium isotope anomalies in conodont apatite [22,63], and anomalous mercury enrichment [64]. In the analyzed P-TB interval, bedded carbonates (Beds 22–24) are overlain by several cm-thick “event beds”, a gray-green ash clay bed (Bed 25) and a black organic-rich claystone bed (Bed 26). Overlying Bed 27 is composed of argillaceous wackestone and dolostone, and Bed 28 is a ~0.5-cm-thick gray-green ash clay, which is overlain by carbonate-dominant Bed 29 and organic rich mudstone (Bed 30). Beds 24e–28 are particularly regarded as the “maximum extinction interval” [2], with the major extinction at Beds 24e–25 and the second phase of extinction at Bed 28 [65]. Another biotic crisis occurred later in Beds 34–38 [66]. The P-TB is assigned within Bed 27 based on the First Appearance Datum (FAD) of conodont Hindeodus parvus, an index fossil for the base of the Triassic [56].

Shen Y. et al. [55] found negative ∆³³S values below the maximum extinction interval (Figure 2a). As described above, this anomalous sulfur isotopic signal is indicative of mixing of ³⁴S-depleted and ³⁴S-enriched sulfur [50]. Shen Y. et al. interpreted that the supposed ³⁴S-depleted sulfur source was H₂S produced via MSR in an open sediment with burrows under oxic conditions. Enhanced benthos activity may have supplied seawater sulfate from the overlying water column into the sediments via burrows. On the other hand, they suggested that the ³⁴S-enriched sulfur source was H₂S produced via quantitative MSR in an anoxic sediment with no burrows. A sulfate supply into the sediment may have been reduced under anoxic conditions with the absence of bioturbation, promoting quantitative MSR in the porewater within the closed sediment. Furthermore, they inferred that the supposed absence of bioturbation was caused by shoaling of anoxic deep-water onto the shallow shelf. According to their model, the bulk ∆³³S value of the sediment was positive under oxic conditions because ³⁴S-depleted H₂S produced within the open sediment was a predominant sulfur source for sedimentary pyrite. However, when the anoxic water shoaled, the shelf sediment shifted to a closed system due to the disappearance of bioturbation, and ³⁴S-enriched H₂S was newly produced in the sediment via quantitative MSR. ³⁴S-enriched H₂S precipitated in situ as pyrite within the sediment and was mixed with the formerly produced ³⁴S-depleted pyrite, resulting in the negative ∆³³S value of bulk sediment. Shen Y. et al. [55] therefore concluded that the anomalous negative ∆³³S values indicated shoaling of anoxic deep-water, leading to the shelf extinction.

2.3. Modified Shoaling Model on the G-LB Records

Zhang et al. [69] first analyzed the multiple sulfur isotopes in the G-LB sediments. They reported the quadruple sulfur isotopic composition of pyrite in G-LB carbonates at Penglaitan, the GSSP for the G-LB, and Tieqiao in Laibin, Guangxi, South China, and at the EF section in west Texas, U.S. (Figure 1b). The analyzed ~13-m-thick G-LB interval at Penglaitan accumulated in the Dian-Qian-Gui Basin, a southern extension of the Jiangnan Basin, in southern South China, and is composed of lower ~9-m-thick bioclastic limestones and upper ~4-m-thick cherty limestones (Figure 3) [70,71,72]. The lower limestones are the Laibin Limestone, the uppermost part of the ~250-m-thick Maokou Formation, and represent the lowstand systems tract accumulated in an upper slope setting [73]. Especially, the upper ~4-m-thick part of the Laibin Limestone is characterized by carbonate debris flow deposits with abundant crinoids. Pyrite framboid data suggest that the Laibin Limestone was generally deposited under oxic conditions [74], and this is supported by the occurrence of burrows in the limestones [71]. The contact between the Laibin Limestone and the overlying cherty limestones represents a major flooding surface. The overlying cherty limestones of deep-water facies are the basal part of the Lopingian Heshan Formation, accumulated in a deep-water basin [73]. Pyrite framboid data suggest their deposition under dysoxic conditions [75] with intermittent euxinia [74]. The G-LB is assigned in the uppermost part of the Laibin Limestone (the base of Bed 6k) based on the FAD of conodont Clarkina postbitteri postbitteri (Figure 3) [71].

The extinction horizons in the G-LB interval at Penglaitan are controversial (Figure 3). An ammonoid turnover occurs in the uppermost part of the Laibin Limestone immediately above the G-LB ([76]; cf., [77]). Kaiho et al. [78] suggested the abrupt extinction at the bottom of the Lopingian Heshan Formation, based on the lithofacies and carbon isotopic analyses. A turnover in conodont fauna from a Guadalupian genus Jinogondolella to a Lopingian Clarkina is recorded immediately below the G-LB [71]. The major extinction of Guadalupian large fusulinids occurred in the uppermost part of the Laibin Limestone (immediately below the G-LB) in the J. granti Zone in the latest Capitanian (Late Guadalupian) [71,79]. Huang et al. [80] reported the earlier demise of a metazoan reef system with the extinction of corals and alatoconchid bivalves at the top of Bed 4 in the J. granti Zone. Shen and Shi [79] reported the occurrence of brachiopods of Lopingian aspect in the lower part of the Laibin Limestone (Beds 2c and 3b), and suggested that the pre-Lopingian biotic crisis occurred in the J. xuanhanensis Zone in the late Capitanian. In addition, Wignall et al. [75] attributed the absence of large and complex fusulinaceans in the analyzed Laibin Limestone to their extinction significantly before the G-LB. Hence, although details are still unknown, the extinction may have occurred in three steps in the Laibin Limestone at Penglaitan [79,80]. It is uncertain whether this apparently stepwise extinction pattern at Penglaitan reflected the global extinction possibly occurred gradually in the Capitanian [81,82,83], or a local facies control. Arefifard and Payne [84] noted an effect of local facies control on the disappearance pattern of fusulinids in Capitanian successions on a global scale, and suggested the single and rapid extinction of large fusulinids in the latest Capitanian. On the other hand, Bond et al. [85,86] pointed out that the major extinction event occurred in the mid-Capitanian, significantly before the G-LB (cf., [87]). The timing and pattern of the end-Guadalupian (or “Capitanian”) extinction should be better constrained by further works.

Zhang et al. [69] reported that the δ³⁴S values of sedimentary pyrite ranged mostly from −20‰ to 0‰ through the analyzed interval (Figure 3), although the values dropped sharply at the G-LB to ca. −50‰. The negative ∆³³S values were observed throughout the interval.

Figure 3. Multiple sulfur isotope chemostratigraphy across the Guadalupian-Lopingian boundary (G-LB) at Penglaitan [69]. A number along the log is Bed number. The estimated three extinction intervals are after Shen and Shi [79] and Huang et al. [80].

Figure 3. Multiple sulfur isotope chemostratigraphy across the Guadalupian-Lopingian boundary (G-LB) at Penglaitan [69]. A number along the log is Bed number. The estimated three extinction intervals are after Shen and Shi [79] and Huang et al. [80].

The Tieqiao section is located ~10 km west of the Penglaitan section and has been regarded as a supplementary reference section of the GSSP for the G-LB [71]. The litho- and bio-stratigraphy of the analyzed ~17.5-m-thick G-LB interval at Tieqiao is similar to that at Penglaitan (Figure 4). The G-LB interval consists of the lower ~10.5-m-thick Laibin Limestone and upper ~7-m-thick cherty limestones of the Heshan Formation [70,71,72]. The Laibin Limestone is gray bioclastic limestones accumulated in an upper slope setting [73]. The similarity of lithofacies of the Laibin Limestone between Tieqiao and Penglaitan may suggest that the limestones were generally deposited under oxic conditions [74], although pyrite framboid data suggest that its uppermost part was deposited under dysoxic conditions [75]. The overlying cherty limestones of the Heshan Formation accumulated in a deep-water basin [73], under dysoxic/anoxic (or possibly euxinic) conditions [18,75].

At Tieqiao, the G-LB is assigned in the uppermost part of the Laibin Limestone based on the FAD of C. postbitteri (Figure 4) [71,72,88], although the extinction horizon in the analyzed G-LB interval has not been well constrained. Chen et al. [89] described a skeletal mound in the Laibin Limestone at Tieqiao, possibly correlated to a metazoan reef in the Laibin Limestone at Penglaitan [80], and reconstructed sea-level changes in the G-LB interval. Based on the results, they suggested that the potential extinction horizon was in the uppermost part of the Laibin Limestone, ~30 cm above the G-LB. In contrast, together with the observations at Penglaitan, Wignall et al. [75] proposed that the extinction of large fusulinaceans occurred by the J. granti Zone, significantly before the G-LB, as mentioned above. Zhang et al. [90] reported a sharp decline in diversity and abundance of smaller foraminifers at the top of the Laibin Limestone at Tieqiao. However, this apparent disappearance of smaller foraminifers immediately above the G-LB was largely due to a facies control according to the deepening in the earliest Lopingian, rather than the global extinction. This is because most genera of smaller foraminifers observed in the Guadalupian Maokou Formation reappear in the middle to upper part of the overlying Heshan Formation. They concluded that the magnitude of the extinction of smaller foraminifers across the G-LB was not significant [90]. Based on the updated conodont zonation at Tieqiao, Sun et al. [88] also suggested that conodonts did not suffer substantial losses during the Capitanian. As in the case of Penglaitan, the timing and pattern of the extinction in the G-LB interval at Tieqiao should be better constrained in the future.

Zhang et al. [69] reported that the δ³⁴S values of pyrite ranged mostly from –10‰ to +10‰ in the analyzed interval (Figure 4). However, the δ³⁴S values dropped sharply around the G-LB to ca. –50‰, and this negative δ³⁴S excursion was correlated with that at Penglaitan (Figure 3), and also with that of carbonate-associated sulfate around the G-LB at Tieqiao [18]. The negative ∆³³S values are observed through the interval.

At the EF section in west Texas (Figure 1b), the uppermost part of the Guadalupian Bell Canyon Formation is overlain by the Lopingian Castile Formation (Figure 5) [91]. The analyzed interval accumulated in the Apache Mountains along the southwestern flank of the Delaware Basin in western Pangea. The siliciclastic Bell Canyon Formation accumulated in a forereef slope setting with calcareous gravity-flow deposits and allochthonous reefal debris, which were derived from the coeval Capitan Limestone of reefal facies on a shallower shelf margin [92]. In contrast, the overlying evaporitic Castile Formation extensively accumulated in the Delaware Basin after a substantial shallowing around the G-LB, which was associated with a large eustatic regression [7,93,94].

The analyzed ~11-m-thick interval at EF is composed mainly of bedded carbonates and is subdivided into five sedimentary units, Units A to E, in ascending order (Figure 5) [91,92,95]. The ~3.3-m-thick basal Unit A is composed of skeletal carbonate mudstone to wackestone. The overlying ~1.7-m-thick Unit B (“the first siltstone unit” in Wardlaw and Nestell [91]) is composed of thinly interbedded skeletal carbonate and calcareous siltstone. The ~3.7-m-thick Unit C consists of bedded wackestone to carbonate mudstone. The ~1.2-m-thick Unit D (“the second siltstone unit” in Wardlaw and Nestell [91]) is composed of interbedded skeletal carbonate and calcareous siltstone. The ~1.0-m-thick uppermost Unit E consists of the lower ~0.3-m-thick skeletal limestones (Bed E1 and E2), and the upper ~0.6-m-thick dolomitic and gypsiferous limestones (Bed E3–E5). The E2–E3 boundary in the Unit E is the boundary between the Bell Canyon and Castile formations [91,92]: the lower E1 and E2 are the uppermost part of the Bell Canyon Formation, whereas the upper E3–E5 are the basal part of the Castile Formation. Lambert et al. [96] reported the occurrence of conodont C. hongshuiensis in Bed E1 (Figure 5), which characterizes the latest Capitanian [97], although the G-LB has not been precisely assigned at EF.

Wu et al. [98] reported high-precision U-Pb geochronology of zircons from ash beds of the Belle Canyon Formation in the Guadalupe Mountains along the northwestern flank of the Delaware Basin. Based on the results, they suggested that the end-Guadalupian extinction occurred within the last 1 million years of the Capitanian and in the earliest Lopingian. Smith et al. [99] conducted multiple geochemical analyses of shallow- and deep-water carbonates from the Guadalupe and Apache Mountains. They proposed that the restricted geometry of the Delaware Basin stimulated drastic changes in water chemistry during a sea-level fall in the Capitanian, such as elevated salinity, water-column stratification, and bottom water anoxia, leading to the severe and physiologically selective extinction in the isolated basin. Shen et al. [94] correlated the entire Guadalupian series in the Guadalupe Mountains to that in South China. In the framework of composite bio- and chemo-stratigraphy and geochronology, they concluded that the fossil records of the uppermost Capitanian rocks in the Guadalupe Mountains were truncated by the abrupt lithofacies change from the marine successions to the Castile evaporites. At the EF section in the Apache Mountains, the analyzed interval of the uppermost part of the Bell Canyon Formation accumulated in the early to latest Capitanian on the basis of conodont zonation [91,96]. In particular, the latest Capitanian strata are exceptionally preserved in its uppermost part (Figure 5). Nevertheless, the timing and pattern of the extinction in the analyzed interval have been poorly constrained.

Zhang et al. [69] reported that the δ³⁴S values increased from ca. –30‰ to 0‰ upward (Figure 5), and then decreased to ca. –35‰ in the uppermost part of the analyzed interval. In particular, the δ³⁴S decrease in the latest Capitanian could be correlated with the negative δ³⁴S excursion around the G-LB at Penglaitan and Tieqiao (Figure 3 and Figure 4). The ∆³³S values are generally negative in the siltstone-bearing units B and D.

Figure 5. Multiple sulfur isotope chemostratigraphy across the G-LB at EF [69]. Alphabet along the log is Unit name. The latest Capitanian interval in the uppermost part of the Bell Canyon Formation is characterized by the occurrence of conodont Clarkina hongshuiensis [96]. C.: Castile Fm.

Figure 5. Multiple sulfur isotope chemostratigraphy across the G-LB at EF [69]. Alphabet along the log is Unit name. The latest Capitanian interval in the uppermost part of the Bell Canyon Formation is characterized by the occurrence of conodont Clarkina hongshuiensis [96]. C.: Castile Fm.

Zhang G.J. et al. [69] detected the negative ∆³³S values in the analyzed intervals at Penglaitan, Tieqiao, and EF (Figure 6). For explaining the negative ∆³³S records, they slightly modified the shoaling model in Shen Y. et al. [55]. In the original shoaling model by Shen Y. et al., the negative ∆³³S signals indicate a mixing of ³⁴S-depleted and ³⁴S-enriched sulfur, and ³⁴S-depleted H₂S was produced via MSR in an open sediment with burrows under oxic conditions. In their revised model, Zhang G.J. et al. [69] interpreted that ³⁴S-depleted H₂S could also have been produced via MSR in the anoxic water column, which resulted in the emergence of a sulfidic (not anoxic) deep-water mass. ³⁴S-depleted H₂S in the deep-water mass was then supplied to the shelf sediment via shoaling, and was mixed with ³⁴S-enriched H₂S produced via quantitative MSR in the shelf sediment according to the shutdown of bioturbation. Consequently, the bulk ∆³³S value of the sediment became negative. According to the revised shoaling model, Zhang G.J. et al. [69] concluded that shoaling of sulfidic deep-waters may have contributed to the end-Guadalupian extinction on a global scale.

3. Later Multiple Sulfur Isotope Studies in the Permian-Triassic Transition Interval

Following Shen Y. et al. [55] and Zhang G.J. et al. [69], several studies reported the multiple sulfur isotope records during the Permian-Triassic transition, particularly focusing on the three extinction events (Figure 1a).

3.1. G-LB

To test the shoaling model in Zhang G.J. et al. [69], Saitoh et al. [100] analyzed the multiple sulfur isotopes in sedimentary pyrite from the G-LB interval at Chaotian, South China (Figure 1b), focusing on the benthic redox conditions in the depositional settings. At Chaotian along the northwestern edge of South China, carbonates and terrigenous clastics of relatively deep-water facies accumulated on a slope/basin setting during the Permian to earliest Triassic [16]. The G-LB interval is composed of ~25-m-thick bioclastic Maokou limestone (Limestone Unit), ~11-m-thick thinly bedded black mudstone/chert (Mudstone Unit), a ~3-m-thick felsic Wangpo tuff bed, and ~20-m-thick bioclastic Wujiaping limestone, in ascending order (Figure 7a). Based on the litho- and bio-facies, the Maokou and Wujiaping limestones accumulated on a shallow oxic shelf with common burrows, whereas the Mudstone Unit accumulated on a relatively deep disphotic slope/basin under sulfidic conditions [16,101]. The Limestone Unit and the ~8-m-thick lower part of the overlying Mudstone Unit accumulated in the early Capitanian based on the occurrence of J. postserrata and J. shannoni [16,102,103]. The ~3-m-thick upper part of the Mudstone Unit possibly accumulated in the middle Capitanian although no conodont occurs in this part. The upper Capitanian rocks are possibly missing at Chaotian due to an abrupt shallowing around the G-LB [16,102]. The extinction horizon is tentatively assigned at the top of the Mudstone Unit. The G-LB is assigned at the bottom of the Wujiaping limestone based on the abundant occurrence of small fusulines, Codonofusiella and Reichelina.

Saitoh et al. [100] showed that the Mudstone Unit was characterized by consistently low δ³⁴S (mostly < −30‰) and high ∆³³S (up to +0.10‰) values (Figure 7). Those sulfur isotopic signals probably reflected the enhanced water-mass sulfate reduction in the anoxic water column, which resulted in the emergence of a sulfidic deep-water mass on the relatively deep slope/basin. This interpretation is consistent with the petrological and geochemical observations of the Mudstone Unit (e.g., the abundant occurrence of small-sized pyrite framboids). Saitoh et al. [100] suggested that the sulfidic deep-water mass may have contributed to the extinction on the shallow shelf via shoaling, and this scenario was apparently consistent with the shoaling model in Zhang G.J. et al. [69]. However, a shelf record during the extinction event is missing at Chaotian due to a local subsidence of the sedimentary basin and to a G-LB unconformity [16,102]. It was therefore difficult to test the Zhang G.J. et al.’s shoaling model by examining the redox and ∆³³S record of the shelf sediment during the extinction at Chaotian.

Interestingly, Saitoh et al. [100] found negative ∆³³S signals in the Maokou and Wujiaping limestones (Figure 7). These bioclastic limestones accumulated on the oxic shelf with common burrows [16,101]. According to the shoaling model in Zhang G.J. et al. [69], the negative ∆³³S value of the sediment was a result of mixing of ³⁴S-depleted and ³⁴S-enriched sulfide in a sediment via shoaling of sulfidic deep-water. Thus, this anomalous isotopic signal should have been limited to euxinic sediments. The commonly observed negative ∆³³S signals in the oxic limestones at Chaotian are apparently inconsistent with the previous shoaling model. To explain the isotope records, Saitoh et al. [100] modified the sulfur mixing model in Zhang G.J. et al. [69]. They interpreted that the negative ∆³³S signals in the shelf limestones were owing to local sulfur isotope heterogeneity in the sediments. According to their revised sulfur mixing model, both of the ³⁴S-depleted and ³⁴S-enriched sulfides were produced in the oxic sediments. ³⁴S-depleted sulfide was produced via MSR within the burrows, which were connected to the sulfate-enriched water column and were a local open system in the sediments. On the other hand, ³⁴S-enriched sulfide was produced via quantitative MSR in a local closed system within the sediments. The bulk ∆³³S value of the sediments became negative by their mixing. As described in Section 2.2, Shen Y. et al. [55] assumed that an oxic sediment was a fully open system with respect to sulfate due to enhanced benthos activity. However, the observed negative ∆³³S signals in the shelf limestones at Chaotian suggested that ³⁴S-depleted and ³⁴S-enriched sulfides coexisted in the local oxic sediments, even when the sediments were well bioturbated. The Chaotian data also suggested that the sulfur isotopic composition of pyrite was spatially heterogenous in the sediments at the hand-specimen scale.

3.2. P-TB

Zhang et al. [67] analyzed the multiple sulfur isotopic composition of sedimentary pyrite across the P-TB at Opal Creek in British Columbia, western Canada, and at Gujo-Hachiman in central Japan (Figure 1b). At Opal Creek, the >20-m-thick Phroso Siltstone Member of the Sulphur Mountain Formation is composed mainly of black to dark gray shale and siltstone, accumulated on an outer shelf or slope in the Lopingian to Early Triassic (Figure 8a) [104]. In the siliciclastic succession, the main extinction horizon ~0.4 m above the basal unconformity is characterized by a turnover in conodont fauna and the disappearance of bioturbation. The P-TB is assigned at the horizon ~1 m above the extinction horizon in the shale, based on the FAD of H. parvus (Figure 8a) [105]. Schoepfer et al. [104] suggested that the basal ~4-m-thick shale/siltstone, including the extinction horizon and P-TB, were deposited under sulfidic conditions, whereas the overlying >15-m-thick siltstone-dominant part accumulated under oxic to suboxic conditions, on the basis of pyrite framboid and multiple geochemical data.

Zhang et al. [67] found negative Δ³³S values in the basal siltstone immediately below the extinction horizon (Figure 8a), and attributed them to shoaling of sulfidic deep-water following the Shen Y. et al.’s model [55]. In contrast to the negative ∆³³S values of the basal siltstone, the value of the overlying shale across the P-TB is consistently positive, up to +0.06‰, with relatively low δ³⁴S values (mostly < −20‰). Zhang et al. [67] interpreted that those sulfur isotopic signals showed an increased influx of syngenetic pyrite with positive Δ³³S and negative δ³⁴S into the sediments under sustained euxinic conditions. Their interpretation is consistent, at least in part, with the Schoepfer et al. [104]’s observation that the basal ~4-m-thick shales/siltstones accumulated under sulfidic conditions.

The negative Δ³³S values reappear in the lower part of the overlying siltstone (Figure 8a). Zhang et al. [67] inferred episodic shoaling of sulfidic water and oscillations between sulfidic and oxic conditions in the depositional setting during the deposition of the part. This scenario is apparently inconsistent with the previous suggestion that the siltstone was deposited predominantly under oxic to suboxic conditions [104]. The shoaling inflow of sulfidic deep-water was substantially reduced and the sulfidic deep-water may have shoaled intermittently during the siltstone deposition (cf., [106]). The substantial decline in inflow of sulfidic water with ³³S-enriched H₂S may have been responsible for the observed Δ³³S decrease, from the positive values of the P-TB shale to the negative values of the overlying siltstone. The supposed decline in shoaling inflow of sulfidic deep-water is consistent with the cessation of coastal upwelling during the siltstone deposition [104]. The ∆³³S values increase to be positive in the upper part of the siltstone, though its cause is uncertain. Zhang et al. [67] concluded that the multiple sulfur isotope record at Opal Creek was generally consistent with the shoaling scenario in Shen Y. et al. [55] at Meishan, and implied that episodic shoaling of sulfidic deep-water occurred on a global scale at the end-Permian.

At Gujo-Hachiman in central Japan, bedded cherts and siliceous claystones occur as an allochthonous block in a Jurassic accretionary complex in the Mino-Tanba belt. The Gujo-Hachiman succession accumulated originally in an abyssal setting on the deep-ocean floor in mid-Panthalassa in the Guadalupian to possibly earliest Triassic (Figure 1b) (e.g., [107,108]). The analyzed ~4-m-thick P-TB interval at Gujo-Hachiman consists of ~3.2-m-thick gray chert (unit I), ~0.2-m-thick gray siliceous claystone (unit II), and ~0.7-m-thick black siliceous claystone (unit III), in ascending order (Figure 9a) [107,109]. The lower cherts (unit I) and the overlying gray siliceous claystones (unit II) are dated to Changhsingian (Late Lopingian) based on the occurrence of radiolarians such as Neoalbaillella optima and Albaillella triangularis [109,110,111]. The Changhsingian age of the unit I cherts is supported by conodont zonation [111]. The unit II is overlain by the black siliceous claystones. The accumulation of black siliceous claystone around the P-TB is commonly observed in mid-Panthalassan deep-sea sequences in Japan, and has been thought to be a result of the demise of radiolarians and the decline in biogenic silica production during the global extinction [5,112,113]. Based on regional correlations, the extinction horizon at Gujo-Hachiman could be placed at the base of the black siliceous claystones (unit III) (Figure 9a).

The black siliceous claystones (unit III) yield no radiolarians or conodonts, except for the occurrence of radiolarians in siliceous claystone lenses/layers in the lower part, i.e., Albaillella triangularis and Neoalbaillella sp. 10 cm above the unit base, and Neoalbaillella sp. 30 cm above the base [109]. The occurrence of Changhsingian radiolarians above the estimated extinction horizon has been problematic [107], though their absence in the matrix of the black claystones suggests that it was due to reworking of the fossils into the younger sediments [114] or their survival in the aftermath of the global mass extinction [115]. The assignment of the P-TB at Gujo-Hachiman is also uncertain due to the poor occurrence of index fossils [107,116]. At Ubara in the Kyoto Prefecture, located ~160 km southwest of Gujo-Hachiman, a P-TB interval of similar lithological units (i.e., bedded gray cherts, gray siliceous claystones, and black siliceous claystones, in ascending order) is exposed as another deep-sea sediment block in the Mino–Tanba Belt, and H. parvus has been identified 10 cm above the base of the black siliceous claystones [117]. Based on this observation, Algeo et al. [107] placed the P-TB at Gujo-Hachiman tentatively near the base of the black siliceous claystones (unit III) (Figure 9a).

The deep-ocean redox conditions during the accumulation of the analyzed P-TB interval at Gujo-Hachiman have been controversial. Based on the composite stratigraphy of the abyssal sediments in Japan and in British Columbia, Canada, Isozaki [5,118] proposed prolonged deep-ocean anoxia in Panthalassa during the Permian-Triassic transition interval (superanoxia). Although the long duration of superanoxia has been challenged by later studies (e.g., [119]), the deep-ocean anoxia/euxinia during the P-TB extinction event has been supported by petrological and geochemical observations on the abyssal sediments in Japan and New Zealand (e.g., [119,120,121]). Nevertheless, based on petrological and multiple geochemical analyses of the present Gujo-Hachiman succession, Algeo et al. [107] proposed that euxinic water masses developed within the oxygen-minimum zone while the bottom waters remained mostly suboxic through the P-TB interval.

Zhang et al. [67] analyzed pyrite in the Gujo-Hachiman sediments and found the negative ∆³³S values in the uppermost part of the gray cherts (unit I) and the overlying gray siliceous claystones (unit II) (Figure 9a). As the Gujo-Hachiman succession accumulated in the deep abyssal setting, shoaling of deep-water cannot be assumed for explaining the negative Δ³³S values. Zhang et al. interpreted that the negative ∆³³S signals recorded oscillations between sulfidic and oxic conditions in the deep ocean, and suggested that the supposed oscillations in deep oceanic redox in mid-Panthalassa in the Changhsingian were consistent with shoaling of sulfidic deep-water at Meishan in eastern Paleotethys and at Opal Creek in western Pangea around the P-TB (Figure 1b).

To re-examine the previous shoaling model, Saitoh et al. [122] recently analyzed the multiple sulfur isotopes in sedimentary pyrite across the P-TB at Chaotian (Figure 1b). The ~40-m-thick P-TB interval consists of ~11-m-thick bioclastic Wujiaping limestone, ~25-m-thick Dalong Formation of deep-water facies, a characteristic ~1.4-m-thick marl unit, and ~3.5-m-thick micritic Feixianguan limestone, in ascending order (Figure 10a). The major extinction horizon is at the bottom of the marl unit, and the P-TB is assigned at the bottom of the Feixianguan limestone on the basis of the FAD of H. parvus [123]. Based on the litho- and bio-facies, the Wujiaping limestone accumulated on a shallow oxic shelf, whereas the overlying thinly-bedded black siliceous/calcareous mudstones in the lower to middle part of the Dalong Formation were deposited on a relatively deep disphotic slope/basin under sulfidic conditions [124]. The depositional setting shifted to a shallower oxic slope in the upper part of the Dalong Formation. In the aftermath of the extinction, the marl unit and overlying Feixianguan limestone accumulated most likely on the slope under sulfidic conditions [122].

Saitoh et al. [122] showed that δ³⁴S and Δ³³S values are substantially low (mostly <−35‰) and high (up to +0.150‰), respectively, in the lower to middle part of the Dalong Formation (Figure 10). These isotopic signals indicated enhanced water-mass sulfate reduction and the emergence of a sulfidic deep-water mass on the deep slope/basin. The ∆³³S record of the analyzed P-TB interval at Chaotian was apparently inconsistent with the previous shoaling model in Shen Y. et al. [55] and Zhang et al. [67], for the following two reasons. Firstly, the negative ∆³³S values were observed in the oxic Wujiaping limestones with common burrows (Figure 10a). According to the previous shoaling model, the negative ∆³³S value should have been restricted to anoxic/sulfidic sediments, as mentioned in Section 3.1. Saitoh et al. [122] explained the negative ∆³³S data by the mixing of ³⁴S-depleted and ³⁴S-enriched sulfides produced in the oxic limestones, as in the case of the Maokou and Wujiaping limestones in the G-LB interval (Figure 7) [100]. Secondly, the ∆³³S values remained consistently positive across the extinction horizon, regardless of the sudden disappearance of bioturbation with the emergence of euxinia.

To better constrain the relationship between the sulfur isotope record and benthic redox conditions, Saitoh et al. [122] further examined a correlation between ∆³³S value and ichnofabric index of the P-TB sediments at Chaotian (Figure 11). According to the previous shoaling model, the negative ∆³³S value should have been restricted to anoxic/sulfidic sediments. Hence, if the shoaling model was correct, a positively correlated variation between the ∆³³S value and ichnofabric index would have been observed. However, the Chaotian data do not show such a variation. Saitoh et al. [122] expanded the previous shoaling model to resolve the apparent discrepancy between the sulfur isotope record at Chaotian and the model. When the deep-water was sulfidic and its shoaling rate was relatively high, a substantial amount of ³³S-enriched H₂S was supplied to a shelf sediment via shoaling, and the bulk ∆³³S value of the sediment became positive. This scenario was similar to that for the positive ∆³³S record of the euxinic P-TB shale at Opal Creek (Figure 8a) [67]. Saitoh et al. [122] further inferred that the observed variation in Δ³³S around the P-TB on a global scale reflected a substantial variation in H₂S concentration and/or in upwelling rate of shoaled deep-waters.

3.3. S-SB

Thomazo et al. [125] analyzed the multiple sulfur isotopic composition of sedimentary pyrite across the S-SB at Mineral Mountains in Utah, USA (Figure 1b). In the Early Triassic, the Mineral Mountains section was located in the southern Sonoma Foreland Basin on the western Pangea margin, in which carbonates and marls of the Thaynes Group accumulated [126]. The analyzed ~120-m-thick S-SB interval at Mineral Mountains is subdivided into the unit A, B, SSB, and C, in ascending order (Figure 12a). The ~20-m-thick unit A is composed mainly of microbial/bioclastic carbonates accumulated in inter- to sub-tidal settings on an inner ramp. The overlying ~50-m-thick unit B is composed mainly of mudstones with storm-induced deposits accumulated in an upper offshore setting on a mid to outer ramp. The overlying ~10-m-thick unit SSB is composed of marls accumulated in a lower offshore setting on a mud outer ramp. The uppermost ~40-m-thick unit C shows an upward shallowing trend, from mudstones accumulated on an outer ramp to microbial carbonates on an outer to mid ramp. The analyzed S-SB interval therefore recorded a transgression–regression cycle, with the maximum flooding event during the deposition of the unit SSB. Although the S-SB is not well assigned, regional litho- and bio-stratigraphic correlations indicated that the unit SSB corresponds to the Smithian-Spathian transition interval [126,127]. In spite of the relative sea-level changes, the S-SB interval was deposited consistently under oxic to dysoxic conditions [128]. The common occurrence of trace fossils (except for the unit SSB and the lower part of unit C) and multiple geochemical data indicated the absence of water-column anoxia through the analyzed interval.

The δ³⁴S and ∆³³S values of the carbonates in the unit A were relatively high and low, respectively (Figure 12). In particular, some negative ∆³³S values were recorded in the carbonates with burrows. Thomazo et al. [125] explained the sulfur isotope data of the unit A by a mixing of ³⁴S-depleted and ³⁴S-enriched sulfides both produced in the carbonates via MSR. They particularly suggested that benthos activity supplied seawater sulfate into the sediments via burrows and promoted MSR to produce ³⁴S-depleted H₂S. Their sulfur mixing model was similar to that for the negative ∆³³S data of the bioclastic limestones in the G-LB and P-TB intervals at Chaotian [100,122]. The overlying unit B was characterized by the negative δ³⁴S and positive ∆³³S values, suggesting MSR under sulfate-enriched conditions. Thomazo et al. suggested that ³⁴S-depleted sulfides were produced in the sediments, an open system with respect to sulfate due to the common burrows, as in the case of unit A. In contrast, in the overlying unit SSB and C, the δ³⁴S values were mostly positive though the ∆³³S values remained high. Thomazo et al. emphasized that the systematic increase in δ³⁴S from the unit B to the overlying units SSB and C (with the consistently positive Δ³³S) corresponded to the disappearance of trace fossils. They interpreted that the analyzed sediments shifted from an open system to a close system by the shutdown of bioturbation associated with the late Smithian extinction. They further inferred that ³³S-enriched accumulative H₂S was produced in the closed sediments via partial reduction of porewater sulfate, according to Rayleigh distillation (Figure 12b). The accumulative H₂S was mixed with ³⁴S-depleted sulfides produced earlier in the open sediments, resulting in the positive δ³⁴S and ∆³³S values of bulk sediments in the units SSB and C. Their work therefore illustrated a clear correlation between bioturbation and the sedimentary sulfur isotope record during the Smithian-Spathian transition.

4. Negative ∆³³S Signal as an Indicator of Shoaling?

Shen Y. et al. [55] and Zhang G.J. et al. [67,69] regarded the negative ∆³³S signature in the sediments as an indicator of shoaling of anoxic/sulfidic deep-water. However, later multiple sulfur isotope studies were apparently inconsistent, at least in part, with the previous shoaling model. For instance, the negative ∆³³S signals were observed in oxic shelf carbonates with common burrows in the G-LB and P-TB intervals at Chaotian [100,122] and in the S-SB interval at Mineral Mountains [125], with the absence of evidence for anoxia. On the other hand, the ∆³³S value of sediment accumulated during shoaling of sulfidic deep-water may not have been necessarily negative. This is because the bulk ∆³³S value of euxinic sediment was strongly controlled by a supply of H₂S (with positive ∆³³S) to the sediment via shoaling, as observed in the P-TB euxinic shale at Opal Creek [67] and in the earliest Triassic Feixianguan carbonates at Chaotian [122]. Even when the sediments shifted from an open system to a closed system by the shutdown of bioturbation according to the previous model, the bulk ∆³³S value of the sediments could remain positive. As Thomazo et al. [125] demonstrated at Mineral Mountains, when accumulative H₂S was produced via partial reduction of porewater sulfate in the closed sediments, its mixing with ³⁴S-depleted sulfides (produced earlier in the open sediments) resulted in the consistently positive ∆³³S values of the bulk sediments (Figure 12b). Furthermore, a positively correlated variation between the ∆³³S value and ichnofabric index of the sediments, which would have been expected according to the previous shoaling model, is not observed in the P-TB interval at Chaotian [122].

Additionally, a stratigraphic relationship between the extinction horizon, redox oscillation in the depositional settings, and the negative ∆³³S signals in the studied sections does not support the previous shoaling model strongly, although Shen Y. et al. [55] and Zhang G.J. et al. [67,69] emphasized their stratigraphic correspondence. The negative ∆³³S values are observed through the G-LB interval at Penglaitan (Figure 3), and that is apparently consistent with the supposed stepwise extinction in the interval. However, the Laibin Limestone was generally deposited under oxic conditions [74], and it is inconsistent with shoaling of anoxic/sulfidic deep-waters. In the G-LB interval at Tieqiao and at EF, the extinction horizon has been poorly constrained (Figure 4 and Figure 5). At Chaotian, the middle to upper Capitanian shelf record is missing (Figure 7a) [100]. In these three sections, it is difficult to ascertain a clear stratigraphic correspondence between the extinction event and negative ∆³³S signals. In the maximum extinction interval across the P-TB at Meishan (Figure 2a), the ∆³³S values are generally positive. The ∆³³S values are also consistently positive across the extinction horizon in the P-TB interval at Chaotian (Figure 10a). At Opal Creek, the negative ∆³³S signals are observed at the extinction horizon and that is apparently consistent with the shoaling model (Figure 8a). As the Gujo-Hachiman P-TB sediments accumulated on the deep-ocean floor in mid-Panthalassa, Zhang et al. [67] interpreted that the negative ∆³³S values of the abyssal sediments reflected oscillations between sulfidic and oxic conditions in the deep ocean (Figure 9a). However, as discussed in Section 3.2, the redox conditions in the Panthalassan bottom waters during the accumulation of the analyzed Gujo-Hachiman sediments are controversial. The ∆³³S values are positive in the SSB unit at Mineral Mountains although there are few analyzed points in the unit (Figure 12) [125].

The present compilation shows that the multiple sulfur isotope records during the Permian-Triassic transition are substantially variable (Figure 13), and that the negative ∆³³S signals are observed in various types of sediments including shallow-marine carbonates, carbonates/siltstones of relatively deep-water facies, and abyssal deep-sea cherts/ siliceous claystones. As Shen Y. et al. [55] and Zhang G.J. et al. [67,69] suggested, the negative ∆³³S values of the sediments are indicative of mixing of ³⁴S-depleted and ³⁴S-enriched sulfur [50]. Nonetheless, the negative ∆³³S records observed in various depositional environments strongly suggest that the supposed sulfur mixing is not unambiguously associated with a shoaling scenario. Under the circumstances, the negative ∆³³S signal may not be a robust indicator of shoaling of anoxic/sulfidic deep-waters. Although the shoaling model in Shen Y. et al. [55] and Zhang G.J. et al. [69] is fascinating, the sedimentary sulfur cycle during the Permian-Triassic transition interval may have been more complicated than originally thought.

5. Future Perspectives

Several topics remain to be examined in future to better constrain the sedimentary sulfur cycle in association with bottom water redox during the Permian-Triassic transition.

5.1. Correlation between the Benthic Redox Conditions and ∆³³S Records

A stratigraphic relationship between the benthic redox conditions, benthos activity, and the multiple sulfur isotopic composition of sedimentary pyrite should be examined in more detail. So far, Saitoh et al. [122] is the only study that examined a stratigraphic correlation between the ∆³³S value and ichnofabric index of the sediments. Clarkson et al. [129] reported iron speciation data of shelf to basin sediments in several sections along the Arabian Margin, and suggested that anoxic ferruginous (non-sulfidic) conditions were prevalent in Neotethys during the Lopingian to Early Triassic. Those Neotethyan “ferruginous” sediments would be an interesting target for a future multiple sulfur isotope study because those sediments may have recorded pristine isotopic signatures of the sedimentary sulfur cycle for the following reasons. When the sediments were overlain by the ferruginous water column, the oxidation of H₂S produced via sedimentary MSR may have been suppressed under oxygen-depleted conditions. The escape of H₂S from the sediments to the overlying water column may also have been prevented. Instead, H₂S would have precipitated as iron sulfide at or below the sediment–water interface, although Clarkson et al. [129] did not specify that the bottom waters were enriched in ferrous ion on the analyzed sediments. On the other hand, the addition of sulfide from the overlying ferruginous water column to the sediments (e.g., liberated H₂S and sinking pyrite framboids) may have been substantially limited. Furthermore, the sediments would not have been disturbed by bioturbation under oxygen-depleted conditions. Hence, the Neotethyan “ferruginous” sediments would be a good recorder of the intrinsic sedimentary sulfur cycle during the Lopingian to Early Triassic.

5.2. SIMS Sulfur Isotope Record

Multiple sulfur isotope measurements by SIMS are useful for identifying a microscopic isotopic variation in sedimentary sulfide (e.g., [130,131]), although SIMS sulfur isotope records in the Permian-Triassic transition interval have not yet been reported. The reproducibility for ∆³³S in SIMS analyses (generally 2σ > 0.1‰; cf., [132]) is too large to detect the ∆³³S variation in the Permian-Triassic sediments, while that for δ³⁴S (generally 2σ < 0.5‰) is small enough to identify local δ³⁴S heterogeneity in the sediments. In particular, the previous bulk studies interpreted that the obtained negative ∆³³S records were a result of mixing of ³⁴S-depleted and ³⁴S-enriched sulfur in the analyzed sediments (e.g., [55,69,100]). In other words, those studies assumed the coexistence of ³⁴S-depleted and ³⁴S-enriched sulfur in the analyzed rock samples at the hand-specimen scale. To verify the sulfur mixing model in the bulk studies, it will be important to examine the supposed substantial δ³⁴S variation in the sediments during the Permian-Triassic transition by SIMS.

It is worth noting that some pyrite grains occur within a burrow in shelf limestones in the Permian-Triassic transition (see Figure 3a in Saitoh et al. [122], for example). According to the sulfur mixing model in Saitoh et al. [100], these pyrites were produced in situ within the burrows, a local open system connected to the sulfate-enriched water column, and thus their δ³⁴S values are substantially low. A systematic difference in δ³⁴S value between sulfides within the burrows and in the matrix in the same rock samples would be identified by future SIMS analyses. However, Harazim et al. [133] reported anomalously ³⁴S-enriched pyrites in burrows in Early Ordovician delta sediments in Newfoundland, Canada. The δ³⁴S values of pyrites within the burrows in the mudstones analyzed by SIMS were substantially scattered, ranging from +16‰ to +58‰ with the average value of ~+40‰, but somewhat higher than the bulk δ³⁴S values of the sediments (+13‰ to +42‰) and the estimated δ³⁴S value of contemporary seawater sulfate (between +30‰ and +40‰). These data indicated that some pyrites precipitated in a closed system during a Rayleigh-type distillation of the residual sulfate pool, and that the burrows were, at least in part, a closed system with respect to sulfate. Their results were therefore apparently inconsistent with the prediction by Saitoh et al. [100].

By using nanoSIMS, Wang et al. [134] recently reported a clear correlation between δ³⁴S value and morphology of pyrite in the Ediacaran strata in South China. In the analyzed shales/carbonates, syndepositional pyrite framboids were characterized by substantially low δ³⁴S values (up to −40‰), possibly produced in the anoxic water column/sediments. In contrast, diagenetic euhedral to subhedral pyrites were characterized by high δ³⁴S values (up to +40‰), produced in the closed sediments according to Rayleigh distillation. The bulk δ³⁴S values of the analyzed sediments were intermediate between the values of pyrite framboids and of euhedral to subhedral pyrites, showing that the bulk values were determined by a mixing of the two distinct types of pyrite. Although the relationship between the morphology and sulfur isotopic composition of pyrite could have been complicated by authigenic overgrowth on syndepositional sulfides (cf., [135]), their results showed the coexistence of ³⁴S-depleted and ³⁴S-enriched sulfides in the same sediments. It is also worth noting that the analyzed sediments accumulated in the Ediacaran and thus bore no burrows. Their results were apparently consistent with the shoaling model in Zhang G.J. et al. [67,69], in which the coexistence of ³⁴S-depleted and ³⁴S-enriched sulfides was associated with the absence of bioturbation. At any rate, a microscopic isotopic variation in sedimentary sulfide during the Permian-Triassic transition should be examined by SIMS analyses in future.

5.3. Negative ∆³³S Record during the Phanerozoic

The present compilation shows that the negative ∆³³S signals are observed in various depositional environments from a shallow shelf to an abyssal plain (Figure 13). Besides the Permian-Triassic transition interval, the negative ∆³³S signatures of sedimentary pyrite were recently found in the Ediacaran-Cambrian transition [136], in the Ordovician [137], and in the Triassic-Jurassic transition [138]. Although further works are required, the negative ∆³³S records may prevail more widely than previously thought in the Phanerozoic. If the negative ∆³³S signatures would be observed commonly in the Phanerozoic sediments, they should be attributed to a common sedimentary/diagenetic process, not a peculiar one like shoaling of toxic deep-water. Recent studies have extensively emphasized heterogeneity in the sulfur isotopic composition of pyrite in a sediment and in a single sedimentary system, largely controlled by local/diagenetic factors (e.g., [133,134,139,140]). Those observations allow me to infer that the sedimentary negative ∆³³S records during the Permian-Triassic transition interval (and possibly in the entire Phanerozoic) reflect local sulfur isotope heterogeneity in the sediments, rather than shoaling of anoxic/sulfidic deep-waters on a global scale. This idea is testable by examining how the negative ∆³³S signals prevail in the Phanerozoic sedimentary records in future.