The studied Permian bedded chert consists of rhythmic successions in the form of chert and thin siliceous claystone beds. Previous sedimentological and geochemical studies have revealed that the chert beds are the SiO₂-diluted parts of siliceous claystone beds, as indicated by an increase in radiolarian tests (Hori et al., 1993; Takiguchi et al., 2006). The chert bed samples contain over 90% SiO₂, which results in the dilution of major oxides (TiO₂, MgO, CaO, Na₂O, K₂O, and P₂O₅) and trace elements by the large amount of biogenic silica. This SiO₂-dilution can enlarge the relative errors of compositional data, which affects the results of multivariate analysis (Soda and Onoue, 2019). Therefore, we focused on the geochemical compositions of the intercalated siliceous claystones.

We collected 114 claystone samples from the Iwaidani section for bulk-rock geochemical analysis. Each sample was crushed in an agate mortar to 5–10 mm in size and handpicked to avoid contamination from veins and intensely weathered material. These handpicked fragments were washed by ultrasonic cleaning in Milli-Q deionized water (>18 MΩ). After air-drying, the fragments were pulverized in an agate planetary mill for major and trace element analyses.

Major element (Si, Ti, Al, Fe, Mn, Mg, Ca, Na, K, and P) and Ni, Cu, and Zn contents were determined with an energy dispersive X-ray fluorescence (XRF) spectrometer (PANalytical Epsilon 3^XLE) with a Mo X-ray tube at Kyushu University, Fukuoka, Japan. Analyses were calibrated using 20 standard rock samples issued by the Geological Survey of Japan. Detection limits for trace elements were 3 ppm for Ni and 2 ppm for Cu and Zn. Reproducibility based on the replicate analysis of two standards issued by Geological Survey of Japan (JSd-1 and JSd-3) was better than ± 1% for Al, Mn, Na, Ti, Ca, K, Ni, Cu, and Zn, and ± 4% for P.

Samples were analyzed for trace and rare Earth element (REE) concentrations using inductively coupled plasma–mass spectroscopy (ICP–MS) with lithium metaborate/tetraborate fusion at Actlabs (package Code 4B2-STD), Ancaster, Canada. Ten international rock and mineral standards were also analyzed for quality control during trace and REE analyses. For details of the analytical techniques, see www.actlabs.com.

To extract paleoenvironmental changes from the compositional data, principal component analysis (PCA) was applied to data for the ten major oxides (SiO₂, Al₂O₃, MgO, CaO, K₂O, TiO₂, Na₂O, P₂O₅, Fe₂O₃, and MnO). We limited PCA to major elements, because some trace elements had large analytical uncertainties and/or were below the detection limit (Supplementary Table S1). To use the major element contents for PCA, the data were normalized using Ti contents and compared with those of post-Archean average Australian shale (PAAS) (Taylor and McLennan, 1985) to obtain enrichment factors, defined as follows: X EF   =   ( X sample / Ti sample ) / ( X PAAS / Ti PAAS ) where X and Ti are the weight contents of element X and Ti, respectively.

PCA is a multivariate statistical analysis that synthesizes numerous observational variables into several orthogonalized principal components (PCs) maximizing the respective variances. Before performing a Q-mode PCA on a correlation coefficient matrix using a singular value decomposition (Golub and Van Loan, 1989; Van Huffel and Vanderwalle, 1991; Albarède, 1995), the enrichment factors were converted into additive log ratios to map the simplex sample space onto the Euclidean real sample space for a constant sum (Aitchison, 1986). This is done because it transforms most of the compositional data into multivariate normal distributions (Aitchison and Shen, 1980; Reyment and Hirano, 1999), which is a prerequisite for PCA (e.g., Atkinson et al., 2004). Consequently, for elements that are less mobile during weathering and diagenesis, such as Ti, whose intermediate ionic potential leads to the formation of stable or metastable individual solids in the Earth surface environment (Stumm and Morgan, 1996; Langmuir, 1997; Railsback, 2007), PCA optimizes an otherwise cumbersome interpretation (Woronow and Love, 1990; Turner et al., 2003).

The major and trace element data obtained from the 114 siliceous claystone samples of the Iwaidani section are listed in Supplementary Table S1. In all of the samples, Al is positively correlated with Ti and high field strength (HFS) elements such as Zr, Hf, and Th (Figure 4). Since elements such as Al, Ti, Zr, Hf, and Th are not significantly mobilized during post-depositional processes, including diagenesis and weathering, these positive correlations indicate that most of these elements have a detrital origin; potassium (K) is also correlated with the elements in detrital materials (Figure 5). The elemental ratios between these elements in our samples are indistinguishable from those of PAAS (Figure 4). Their similarity to average shale composition (Taylor and McLennan, 1985) indicates that the detrital materials in the Iwaidani section samples have a terrigenous origin, which is consistent with the geochemistry of the Permian–Triassic siliceous rocks at the Sasayama section in Japan (Figure 1B; Kato et al., 2002). Conversely, no correlations were found between these terrigenous elements and elements such as P, Ba, Mn, V, U, and Mo (Figure 5), which are typically derived from biogenic and/or hydrogenous sources (Calvert and Pedersen, 2007).

To avoid the significant dilution effect by biogenic SiO₂, concentrations of the elements were normalized to Al concentrations and compared with those of the PAAS (Taylor and McLennan, 1985) to obtain enrichment factors. Figure 6 shows the enrichment factors of K, Ti, Zr, Hf, and Th. Values of K_EF and Th_EF (1.59 ± 0.10 and 1.02 ± 0.12, respectively; average ± 1SD) are fairly constant throughout the studied section and close to the value of 1, where 1 represents the sample having the exact same ratio as PAAS. In the Capitanian and Wuchiapingian, the values of Ti_EF, Zr_EF, and Hf_EF are similar to those of PAAS; in the late Changhsingian the enrichment factors for these elements show a rapid decrease.

Enrichment factors for Mn, Ni, Cu, Zn, Tl, V, U, and Mo have been widely used to characterize the redox conditions of marine environment (e.g., Calvert and Pedersen, 1993; Calvert and Pedersen, 2007; Algeo and Tribovillard, 2009; Takahashi et al., 2014; Algeo and Li, 2020), which can be generally classified as oxic, dysoxic, suboxic, anoxic, or euxinic state (presence of free H₂S) (Tyson and Pearson, 1991). Enrichment factors of Mn, V, U, Mo, Ni, Cu, and Zn in Figure 7 were calculated relative to PAAS (Taylor and McLennan, 1985), and Tl was calculated relative to upper continental crust (UCC; McLennan, 2001). In this study, the entire section exhibits a low enrichment factor for Mn (Mn_EF <1). Enrichment factors for V and U are generally close to the value of 1 during the Capitanian, but increase slightly across the GLT. Increases in Ni_EF, Cu_EF, Zn_EF, Tl_EF, and U_EF occur across the WCB. An increase in Mo_EF also occurs above the WCB. Vanadium, U, and Mo enrichment factors continued to increase during the latest Changhsingian. In particular, Mo_EF reached a peak of 5,000 (Mo content = 5,370 ppm).

Calcium, P, and Ba are widely used as proxies for the burial flux of biogenic material (Hollis et al., 2003; Takiguchi et al., 2006; Algeo et al., 2011b; Soda and Onoue, 2018; Soda and Onoue, 2019; Sato et al., 2020) based on their distribution and deposition in modern marine sediments (e.g., Schroeder et al., 1997). Figure 8 shows changes in the enrichment factors against the stratigraphic frame in the present study. The trends in Ca_EF and P_EF decreased abruptly in the late Capitanian, and these elements are positively correlated with each other (r = 0.62; Figure 9). There is an abrupt decrease in Ba_EF during the Wuchiapingian, and Ba_EF remains relatively low during the Changhsingian.

Figure 10 shows the rare Earth element (REE) patterns of the Iwaidani section normalized to PAAS (Taylor and McLennan, 1985). The REE patterns of the Guadalupian samples are moderately to highly enriched in middle REEs (MREE: Sm, Gd, Tb, and Dy), whereas most of the Lopingian samples yield rock/PAAS ratios of 1. The REE ratios for the Lower Triassic claystones are relatively flat and lower than 1. Since the enrichment of the MREEs exhibits stratigraphic changes, we calculate the relative MREE enrichments using PAAS-normalized MREE/MREE* ratios (Chen et al., 2015): MREE / MREE ∗ =   2   ×   ( avg .  MREE / avg . MREE PAAS ) ( avg .LREE / avg .LREE PAAS ) +   ( avg .HREE / avg .HREE PAAS ) , where LREE and HREE represent light REEs (La, Ce, Pr, and Nd) and heavy REEs (Er, Tm, Yb, and Lu), respectively.

The MREE/MREE* ratios throughout the studied section show a pronounced decline in MREE enrichment across the GLT (Figure 8). The MREE/MREE* ratios fluctuate between 0.9 and 1.5 during the Wuchiapingian and early Changshingian, but decreased in the late Changshingian.

Cerium has multiple valence states, which results in fractionation that can be quantified by the Ce anomaly (German and Elderfield, 1990; Holser, 1997): Ce / Ce ∗   =   2 Ce N / ( La N   +  Pr N ) ,

where the subscript “N” denotes PAAS-normalized values (e.g., La_N = La_sample/La_PAAS). If the Ce/Ce* value is lower than 1, Ce is plotted below the line connecting La and Pr, which is called as negative Ce anomaly (Figure 10). The Ce/Ce* ratio varies between 0.50 and 1.06, and fluctuates inversely with the MREE/MREE* ratios throughout the studied section (Figure 8). With the exception of a few of samples, there is no Ce anomaly (Ce/Ce* ∼1) when the REE patterns are relatively flat, but the Ce/Ce* values exhibit negative anomalies in the MREE-enriched samples (Figure 10).

Table 1 provides the component loadings and Figure 11 shows compositional biplots for the studied samples. We accepted PC1, PC2, and PC3 in this dataset from their eigenvalues (PC1 eigenvalue = 1.960, PC2 eigenvalue = 1.446, and PC3 eigenvalue = 1.179), which explain 81.3% of the total variance (PC1 proportion = 42.7%, PC2 proportion = 23.2%, and PC3 proportion = 15.4%). Other PCs (PC4 and PC5) did not show distinct geochemical components because of their low eigenvalues (<0.8). The loadings of PC1 show strong negative values (less than −0.900) for Si and Al, with subordinate negative values (less than −0.700) for Mg, Na, and K, and a moderate negative value for Ca (−0.493). PC2 shows peaks with high negative loadings (less than −0.500) for Fe, Mn and P. PC3 is characterized by a high negative loading for Ca, moderate negative loadings for Mg and P, and positive loadings for all other elements.

The stratigraphic trend of PC1 is relatively constant throughout the Capitanian and largely varies between 0 and 4 in the Wuchiapingian and most of the Changhsingian. However, the PC1 scores show an abrupt decrease to negative scores across the EPME and PTB (Figure 6). The PC2 scores are characterized by a gradual increase from negative to positive values across the GLT (Figure 7). The PC3 scores show a stratigraphic trend in which the Capitanian samples have negative scores, whereas the other Lopingian samples have near-zero or positive scores (Figure 8).

Our geochemical analysis of the upper Guadalupian to Lopingian bedded chert succession from the Iwaidani section records an enrichment in redox-sensitive elements, such as V, Ni, Cu, Zn, Tl, U, Mo. These redox-sensitive elements are divided into two categories on the basis of their behaviors under degrees of marine oxygenation (Calvert and Pedersen, 1993, Calvert and Pedersen, 2007). The first category includes those elements whose valence state varies with the redox environment. Vanadium, U, and Mo form highly soluble ions under oxygenated conditions, but under anoxic and euxinic conditions produce insoluble and particle-reactive materials in their lower valence state (Calvert and Pedersen, 1993; Helz et al., 1996; Zheng et al., 2000; Algeo and Maynard, 2004; Tribovillard et al., 2006; Algeo and Li, 2020). The second category includes elements that are primarily associated with sulfides or Mn-oxide phases. Elements such as Ni, Cu, Zn, and Tl are incorporated into precipitating sulfide phases or scavenged by Fe- and Mn-(oxyhydr)oxides (Calvert and Pedersen, 1993; Rehkämper and Nielsen, 2004; Tribovillard et al., 2006; Calvert and Pedersen, 2007). We use the changes in the enrichment factors, ratios, and contents of these elements as proxies for redox conditions through the studied section.

In the presence of dissolved oxygen, Mn forms insoluble Mn(III) or Mn(IV) hydroxides or oxides (e.g., MnO₂); these particulate forms are rapidly deposited (Calvert and Pedersen, 1993; Sholkovitz et al., 1994). Under suboxic conditions, Mn is reduced to Mn(II) and forms soluble cations (e.g., Mn²⁺ and MnCl⁺) (Algeo and Li, 2020). Since particulate Mn is deposited more readily under oxic conditions, while soluble Mn remains in the water column under suboxic conditions, a low Mn_EF (<1) in marine sediments suggest suboxic depositional conditions near the sediment surface. In the Iwaidani section, Mn_EF values are generally less than 1, where 1 represents the sample having the exact same ratio as PAAS (Figure 7). These Mn_EF values are lower than those of the Cisuralian red chert (Sugitani et al., 1991; 0.01–0.10% MnO, 5.2–29 Mn_EF), which lies stratigraphically below the studied section (Figure 2B; Sano, 1988a). The low Mn values in the Iwaidani section suggest that suboxic conditions has developed in the depositional environment already in the late Guadalupian, long before the PTB.

Vanadium and U enrichments in sediments are proxies for moderately to strongly reducing conditions in the depositional environment (Sadiq, 1988; Algeo and Maynard, 2004; Tribovillard et al., 2006). Vanadium and U reduction occurs under lower oxygen concentrations than Mn reduction (Algeo and Li, 2020). In oxic seawater, V is present as soluble V(V) in the quasi-conservative form of vanadate oxyanions (HVO₄ ^2– and H₂VO₄ ^2–). When conditions change from suboxic to weakly anoxic, V(V) converts to V(IV) and forms the vanadyl ion (VO²⁺), related hydroxyl species [VO(OH)^3–], and insoluble hydroxides [VO(OH)₂] (Breit and Wanty, 1991; Wanty and Goldhaber, 1992). Under H₂S-rich (euxinic) conditions, V(III) is further reduced and precipitates as relatively insoluble hydroxides [V(OH)₃] (Wanty and Goldhaber, 1992). Similar to vanadium, soluble U(VI) is reduced to insoluble U(IV) (e.g., UO₂) under anoxic conditions.

In the studied section, Capitanian U_EF and V_EF values are relatively constant and close to the value of 1 (Figure 7). Across the GLT, the V_EF and U_EF values increase slightly, and then U_EF continues to increase during the Changhsingian. The V_EF and U_EF values increase abruptly in the latest Changhsingian, with peak values at the EPME horizon (Figure 7). Similar enrichment trends to those observed here in V and U were observed in Th/U ratios, another common redox proxy, from end-Permain strata in Italy and Slovenia (Figure 7; Wignall and Twitchett, 1996). Consequently, V and U enrichments suggest that the development of suboxic to weakly anoxic conditions occurred across the GLT in this section, followed by anoxic–euxinic conditions during the Changhsingian.

Molybdenum is present in seawater as molybdate (MoO₄ ²⁻). Mo-removal from the water column requires the formation of particle-reactive thiomolybdate species (MoO_xS_4-x ²⁻), which occurs in the presence of aqueous H₂S (Helz et al., 1996; Erickson and Helz, 2000; Helz et al., 2011). Thus, Mo enrichment in sediments requires high H₂S levels in a euxinic environment (Algeo and Tribovillard, 2009; Tribovillard et al., 2012; Algeo and Li, 2020). In the studied section, Capitanian and Wuchiapingian samples have Mo contents below the detection limit (2 ppm). However, our data demonstrate that the Mo_EF begins to increase during the lower Changhsingian, which overlaps with the increases in U_EF (Figure 7). Subsequently, the Mo_EF values increase sharply during the latest Changhsingian, with a maximum value of ∼5,500. Covariations of the enrichment patterns of Mo and U (Algeo and Tribovillard, 2009; Tribovillard et al., 2012) (Figure 12) followed an evolution typical for open marine, hydrographically unrestricted settings (Algeo and Tribovillard, 2009; Tribovillard et al., 2012). Trends in U_EF and Mo_EF suggest a change from suboxic–anoxic to euxinic conditions in the deep-sea basin during the latest Changhsingian to Induan (Figure 12). The Mo_EF values in the thousands and Mo_EF/U_EF ratios in the hundreds are recorded immediately before the EPME (Figure 7), suggesting the development of sulphidic bottom water and scavenging of Mo by a Fe–Mn (oxyhydr)oxides particulate shuttle (Algeo and Tribovillard, 2009; Tribovillard et al., 2012). The occurrence of the particulate shuttle requires an oxic–anoxic redox boundary in the water column and rapid water replacement to maintain the predicted Fe–Mn redox behavior (Algeo and Tribovillard, 2009; Scholz et al., 2013). However, redox boundaries and vertical circulation in the pelagic Panthalassa remain unclear, and this needs to be addressed in future studies.

Ni, Cu, Zn, and Tl are incorporated into precipitating sulfide phases or scavenged by Fe- and Mn-(oxyhydr)oxides (Calvert and Pedersen, 1993; Koschinsky and Hein, 2003; Algeo and Maynard, 2004; Rehkämper and Nielsen, 2004; Tribovillard et al., 2006; Calvert and Pedersen, 2007). Since the entire section contains low Mn-enrichment factors (Mn_EF <1), except for two high-Mn_EF peaks (Mn_EF >6) in the Changhsingian (∼13.5–14.5 m), enrichment patterns for Ni, Cu, Zn, and Tl may be related to authigenic sulfides (e.g., pyrite) in a reducing environment (Huerta-Diaz and Morse, 1992; Morse and Luther, 1999; Tribovillard et al., 2006; Belzile and Chen, 2017). In the studied section, Ni_EF, Cu_EF, Zn_EF, and Tl_EF begin to increase in the lower Changhsingian and remain high until the uppermost Changhsingian (Figure 7). This supports increasingly anoxic and/or euxinic bottom water conditions toward the latest Changhsingian, as suggested by the V_EF, U_EF, and Mo_EF values. Relatively high Tl_EF values are recorded in the stratigraphic interval from ∼13.5 to 14.5 m, which appears to be associated with Mn enrichment (Koschinsky and Hein, 2003).

In addition to the redox-sensitive elements mentioned above, the Ce anomaly is also a common tracer for redox conditions in pelagic sediments (Kato et al., 2002; Takahashi et al., 2014; Fujisaki et al., 2019). Our data reveal that low Ce/Ce* values are associated with increased MREE (Figure 13), and there is no Ce anomaly when the REE patterns are relatively flat in the studied section (Figures 8, 10). Since the MREE-enriched patterns are probably due to the presence of fossilized biogenic apatite (e.g., conodonts) that recrystallized during diagenesis (Reynard et al., 1999; Chen et al., 2015; Zhang et al., 2016), the stratigraphic trends of Ce/Ce* values in the Iwaidani section may reflect the change in biogenic apatite content during diagenesis, rather than recording the redox conditions.

The PC1 scores of the present study represent the compositional variations in terrigenous detrital materials (aluminosilicates), because the loadings show strong negative values for Si, Al, Mg, Ca, Na, and K (Table 1). Prominent negative loadings for both Si and Al and the strong positive correlation coefficient between Si_EF and Al_EF (r = 0.89; Figure 9) indicate PC1 is related to the presence aluminosilicates, such as clay minerals. The same directions in the relatively strong loadings for alkali and alkaline Earth elements (Mg, Ca, Na, and K) with those of Si and Al suggest that these elements have been substituted for Si–Al units and layers or Al octahedral sheets in clay minerals (e.g., Brigatti et al., 2006). Thus, these elemental behaviors in the PC1 axis indicate the intensity of chemical weathering.

To verify the PC1 interpretation, we employed the chemical index of alteration (CIA; Nesbitt and Young, 1982) and the weathering index of Parker (WIP; Parker, 1970). These indices are widely used for reconstructing paleoclimate conditions by assessing the degree of hinterland weathering, which is controlled mainly by climatic factors (Nesbitt and Young, 1982; Nesbitt et al., 1996; Ohta and Arai, 2007; Sheldon and Tabor, 2009; Yang et al., 2016). Values of CIA indicate the extent of decomposition of feldspar minerals, which are the most abundant mineral group in the upper continental crust (Taylor and McLennan, 1985). The CIA is defined as Al₂O₃/(Al₂O₃ + CaO* + Na₂O + K₂O) × 100 (molar content). CaO in silicate phases (CaO*) is estimated by subtracting CaO in phosphate phases, predicted from P₂O₅ contents (McLennan, 1993; Price and Velbel, 2003). In samples with low CaO and high P₂O₅ contents, this calculation may yield negative values. In this case, values are set to 0. The WIP is defined as 100 × (2Na₂O/0.35 + MgO/0.9 + 2K₂O/0.25 + CaO*/0.7) (molar content). We used WIP to assess the weathering intensity of silicate rocks based on the proportions of alkali and alkaline Earth elements in the weathered products (Parker, 1970). The smaller WIP values denote stronger chemical weathering, which is opposite to the CIA values.

Figure 6 shows stratigraphic variations in CIA, WIP, and PC1. The stratigraphic profile of the PC1 scores correlates well with the WIP profile, supporting the interpretation that PC1 reflects the intensity of hinterland weathering. The general trend of the CIA index is also similar to that of PC1. The PC1 scores decrease with increasing CIA and decreasing WIP values across the EPME, suggesting intensified chemical weathering in hinterland areas during the latest Changhsingian to earliest Induan.

The PC2 scores explain the accumulation of Fe (oxyhydr)oxides in the section, because of the strong negative loadings (< −0.800) of Fe and P (Table 1). Negative P loading suggests adsorption or co-precipitation of phosphate on sinking Fe-oxyhydroxide particles (e.g., FeOOH–PO₄ particles; Yao and Millero, 1996; Dellwig et al., 2010). In addition, relatively strong loading of Mn may also result from adsorption and/or co-precipitation of Mn as a particulate in the Mn–Fe–P-shuttle (MnOx–FeOOH–PO₄; Dellwig et al., 2010). Such Fe- and P-rich particles are reductively dissolved under anoxic–euxinic conditions, and release Fe²⁺ and phosphate to the water column (Shaffer, 1986; Dellwig et al., 2010). The correlation coefficient between Fe_EF and P_EF under the anoxic–euxinic conditions during the Changhsingian and Induan (r = 0.30) is lower than that of the Capitanian and Wuchiapingian (r = 0.62). The more negative PC2 scores (Figure 7) suggest a greater accumulation of Fe- and P-rich particles in a more oxygen-rich environment under the suboxic conditions of the Capitanian and early Wuchiapingian, but this interpretation is not valid under the anoxic–euxinic conditions after the late Wuchiapingian.

The PC3 axis explains the calcium phosphate accumulation, because the loadings show relatively strong negative values for both P and Ca in the compositional biplot (Figure 11). The enrichment factors for Ca and P show a relatively strong positive correlation (r = 0.62; Figure 9) that is consistent with the occurrences of conodont fossils in the studied section (Sano et al., 2012a). Indeed, the stratigraphic trends in P_EF and Ca_EF are largely similar to those in PC3 (Figure 8). Thus, these characteristics of PC3 are related to biogenic apatite accumulation, as reported from multivariate statistical analyses of a Middle Triassic bedded chert succession in the Mino Belt, Japan (Soda and Onoue, 2019). The more negative direction of the PC3 scores indicate the presence of biogenic apatite, because the loadings of P and Ca have the same directions (negative). Therefore, negative PC3 scores in the Capitanian (Figure 8) imply higher accumulations of biogenic apatite, whereas the near-zero PC3 scores in the Wuchiapingian to Induan samples suggest minor contributions of biogenic apatite to the sediments. Ca_EF and P_EF are relatively high in the Capitanian and clearly decrease across the GLT, which supports this interpretation of PC3. Indeed, the REE patterns in the Capitanian samples display a moderate to high enrichment in MREE (Figure 8), which is probably due to the presence of fossil biogenic apatite that recrystallized during diagenesis (Reynard et al., 1999; Chen et al., 2015). Thus, the relative depletion of MREEs in the Wuchiapingian samples is probably due to the paucity of biogenic apatite, which is consistent with the interpretation that the decrease in PC3 was associated with decreased contents of biogenic apatite across the GLT.

The detailed redox history of the Iwaidani section, from the Capitanian to Induan, is summarized in Figure 14. The entire section shows a low enrichment factor for Mn (Mn_EF <1), suggesting that reducing conditions has developed in the depositional environment already in the Capitanian. The enrichment patterns of the other redox-sensitive elements show a stepwise development of oxygen-depleted conditions in the studied section, referred to here as stages 1, 2, and 3 (in ascending order) (Figure 14).

Stage 1 consists of a slight enrichment of V and U (∼1.5–2.0 times that of the Capitanian), suggesting the initiation and development of reducing conditions across the GLT (Figure 14). The stratigraphic variations in PC2 scores indicate that the accumulation of Fe- and P-rich particles decreased across the GLT, suggesting the depositional environment changed from suboxic to weakly anoxic conditions, meaning that these particles could have been reductively dissolved. Subsequently, increases in U_EF, Mo_EF, Ni_EF, Cu_EF, Zn_EF, and Tl_EF during stage 2 indicate the development of anoxic conditions, which began at the WCB transition. Stage 3 (latest Changhsingian) is characterized by marked enrichments in Mo contents. The Mo_EF values and Mo_EF/U_EF ratios (Figure 7) demonstrate that H₂S-rich (euxinic) conditions may have developed in the latest Changhsingian, concurrent with the EPME. Since the average sedimentation rate for the Changhsingian bedded chert of the Hashikadani Formation is ∼0.3 cm·kyr⁻¹ (Onoue et al., 2019), the anomalous increases in Mo_EF corresponding to the 60 cm stratigraphic interval at the end of the Permian indicate that the euxinic conditions in stage 3 began ∼200 kyr before the EPME (Figure 14). The increase in V_EF and U_EF does not coincide with the increase in Mo_EF, and their values increase significantly at the EPME horizon (Figure 7). Since Mo is adsorbed onto Fe–Mn particulates, whereas U and V are not, this pattern is consistent with the development of a particulate shuttle (Algeo and Tribovillard, 2009) immediately before the EPME. The appearance of a particulate shuttle at that time has also been reported from the Akkamori section in Northeast Japan (Takahashi et al., 2014). The development of the particulate shuttle in the pelagic Panthalassa may have depended on development of redox boundaries and vertical circulation in the early phase of stage 3, and this needs to be addressed in future studies.

Our redox-sensitive element data and PCA are consistent with previous interpretations of the gradual expansion of reducing conditions starting around the GLT (Isozaki, 1997; Kato et al., 2002). Isozaki, 1994 and Isozaki, 1997 coined the term “deep-sea anoxia” to describe this unique long-term oceanographic phenomenon in the deep sea of the mid-Panthalassa. In this study, we have newly identified three distinct ocean states (stages 1–3) during the Lopingian. Our results support the hypothesis of deep-sea anoxia, but contradict recent interpretations of oxic-dominant conditions in deep-sea Panthalassa during the Lopingian, as evidenced by positive Ce anomalies and limited increases in U and Mo from the PTB section at Gujo-Hachiman (Algeo et al., 2010; Algeo et al., 2011b; Fujisaki et al., 2019). The differences in redox conditions between the Iwaidani and Gujo-Hachiman sections are discussed below, in Section Anoxic vs. Oxic Condition in Deep-Sea Panthalassa.

Furthermore, the present study demonstrates that the changes from suboxic to euxinic conditions through the Lopingian reach a maximum immediately before the EPME. Takahashi et al. (2014) provided evidence of such euxinic conditions in the deep sea of the central Panthalassa Ocean immediately before and during the EPME, as evidenced by extremely high concentrations of U and Mo (enrichment factors of ∼6 and ∼7,600, respectively) in the uppermost Changhsingian black claystone from the Akkamori section in Northeast Japan (Figure 1B). Their values are consistent with those from the black claystone bed (∼6 U_EF, ∼5,500 Mo_EF) at the top of the middle unit in the Iwaidani section. These studies suggest that the “euxinic maximum” (Takahashi et al., 2014) at the end of the Permian is most likely a phenomenon that occurred on the deep seafloor of the Panthalassa Ocean (Figure 14). Subsequent abrupt decreases in U_EF and Mo_EF are recorded by the Induan black claystones in the Iwaidani section. Similar decreases in U_EF and Mo_EF after the EPME have also been reported for the pelagic deep-sea PTB section at Akkamori. The concurrent drawdown of U_EF and Mo_EF in both sections might reflect a drawdown of these trace elements in the global oceanic inventory (Takahashi et al., 2014).

Oxygen-depleted conditions during the latest Changhsingian have been proposed from other pelagic deep-sea PTB sections in Japan, such as the Waidani, Gujo-Hachiman, Akkamori, Ubara, and Tenjinmaru sections (Figure 1B; Kajiwara et al., 1994; Algeo et al., 2010; Wignall et al., 2010; Algeo et al., 2011b; Sano et al., 2012b; Takahashi et al., 2014; Kaiho et al., 2016; Zhang et al., 2017; Onoue et al., 2019). Of these, the Waidani, Akkamori, Ubara, and Tenjinmaru sections have limited stratigraphic records from the late Changhsingian to Early Triassic, whereas the Gujo-Hachiman section preserves a longer record from the Capitanian to the latest Changhsingian (Algeo et al., 2010; Zhang et al., 2017; Fujisaki et al., 2019). Therefore, here we discuss the redox conditions in the Lopingian by comparing the Iwaidani and Gujo-Hachiman sections.

The Gujo-Hachiman section (Kuwahara, 1997; Kuwahara et al., 1998) in the Mino Belt crops out ∼18 km east of the Iwaidani section (Figure 2B). Several large slabs (∼2–5 km wide and ≥15 km long) of ocean-island basalts and carbonate rocks are distributed between the Iwaidani and Gujo-Hachiman sections, and the rocks of both sections originated on a mid-oceanic seamount in the Panthalassa Ocean (Sano, 1988a, Sano, 1988b; Sano, 1989; Jones et al., 1993; Ichiyama et al., 2008). Therefore, both PTB sections are interpreted as pelagic facies that extend from the deep-ocean floor to the lower flank of a mid-Panthalassic seamount. This is evidenced by thin intercalations of dolomite rocks in the lower Guadalupian bedded cherts at Gujo-Hachiman (Kuwahara et al., 1998), which are likely reworked deposits from a shallow-water carbonate buildup, as previously suggested by Sano (1988a). Biostratigraphic analysis of radiolarians and conodonts revealed that the Gujo-Hachiman section contains the GLT and WCB (Kuwahara et al., 1998; Yao et al., 2001; Nishikane et al., 2011; Nishikane et al., 2014).

Recent studies of redox-sensitive elements (e.g., U, Mo, and Ce) and the quantitative analysis of framboidal pyrite from the Gujo-Hachiman section suggest oxic-dominant conditions in the deep-sea during the Lopingian and a rapid expansion of the sulphidic OMZ at shallower depths immediately before the EPME (Algeo et al., 2010; Wignall et al., 2010; Fujisaki et al., 2019). Based on positive Ce anomalies and low Mo and U enrichment factors, Fujisaki et al. (2019) suggested that the Gujo-Hachiman section records well-oxygenated conditions of deep-sea Panthalassa during the Guadalupian and most of the Lopingian, and episodic occurrences of suboxic conditions during the middle–late Lopingian.

These studies contradict our interpretation of the stepwise development of suboxic–euxinic conditions during the Lopingian (Figure 14), even though the Iwaidani and Gujo-Hachiman sections are located near each other and thus their depositional environments should have been similar. Since Mo-U co-variance trends for the Iwaidani section reflect open-marine, hydrographically-unrestricted depositional settings (Figure 12), it is unlikely that only the sediments in the Iwaidani section were deposited under restricted basin with stagnant conditions around a mid-oceanic seamount.

We suggest that the cherts and siliceous claystones from the Gujo-Hachiman section are intensely altered by the intrusions of nearby Cretaceous rhyolites and felsites (Figure 2B) and may not record the redox conditions in the Guadalupian and Lopingian. In previous studies, red hematite nodules have been described throughout the Gujo-Hachiman section, which has been referred to as negative evidence for anoxic conditions (Wignall et al., 2010). However, outcrop observations indicate that the hematite nodules with irregular shape and margins occur in association with the development of white and red veins and vugs of possible hydrothermal origin (Supplementary Figure S1). The development of these nodules and veins is characteristic of the Permian chert beds at Gujo-Hachiman and is not found in other Permian sections of the Funabuseyama area (Sano, 1988a; Sano, 2018). It is possible that the nodules and veins characteristic of the Gujo-Hachiman section were formed by the intrusions of the Cretaceous Okumino rhyolites distributed around the section (Figure 2B). Although their origin needs further investigation, the development of such veins is likely to significantly alter the chemical compositions of the siliceous claystone samples from the Gujo-Hachiman section.

In general, the correlations of Al, K and Rb contents attributed to alluminosilicates (mainly clay minerals) are strong in the bedded charts of the Mino Belt (Kato et al., 2002; Soda and Onoue, 2018, Soda and Onoue, 2019; Sato et al., 2020), and high correlations of these elements are also found in the Iwaidani section (Figures 5, 9). This implies that Al, K, and Rb in alluminosilicate minerals in the Permian bedded cherts were not significantly mobilized during post-depositional alteration processes. In contrast, there is no clear correlation between Al, K, and Rb in the Gujo-Hachiman section (Supplementary Figure S2). Compared with Iwaidani, the Gujo-Hachiman section shows a depletion of K and Rb relative to Al, but instead contains an excess of Si and Fe. This may indicate that the leaching of alkaline elements (K and Rb) and precipitation of Si and Fe occurred due to hydrothermal alteration (Pirajno, 2009). Chemical weathering indices (CIA and WIP) are sensitive to changes in bulk chemistry that is also associated with hydrothermal alteration (Price and Velbel, 2003; Pirajno, 2009). In contrast to the Iwaidani section, the Gujo-Hachiman samples are characterized by highly variable CIA and WIP values (Supplementary Figure S3) with highly to completely weathered nature (e.g., CIA >80, WIP <20; Duzgoren-Aydin et al., 2002; Ohta and Arai, 2007), suggesting that the samples are subjected to strong chemical alteration, probably due to hydrothermal alteration and accentuated weathering. If this interpretation is correct, elemental redox proxies, such as V and U, may have been affected by the chemical alterations, including oxidation leaching of these elements during hydrothermal alteration. Further geochemical and mineralogical studies are needed to constrain these interpretations.

The much higher U_EF and Mo_EF values, indicative of euxinia, are also observed in the latest Changhsingian at the Iwaidani section but not in the Gujo-Hachiman section. Algeo et al. (2011b) suggested that the bottom water of the Gujo-Hachiman section was suboxic, as evidenced by the limited increases in V, U, and Mo across the chert–black shale facies transition. This may reflect the hydrothermal alterations (as discussed above), but the lack of a stratigraphic record corresponding to the euxinic condition may be a more realistic explanation of this difference (Takahashi et al., 2014). As shown in Supplementary Figure S4, the rocks in the uppermost Changhsingian are strongly deformed compared with the lower stratigraphic intervals of the section. The Gujo-Hachiman section lacks the Hg/TOC anomalies and negative Δ³³S anomalies found at the uppermost Changhsingian in other PTB sections in Japan (Zhang et al., 2017; Shen et al., 2019), which can be explained by the absence of the stratigraphic interval corresponding to the euxinic conditions, due to minor shear faults at Gujo-Hachiman (Supplementary Figure S4). This explanation was also proposed by Takahashi et al. (2014) and Kaiho et al. (2016) to explain the different redox conditions between the Gujo-Hachiman and Akkamori sections during the latest Changhsingian.

Based on quantitative analyses of the framboidal pyrite size distribution, Algeo et al. (2010) and Wignall et al. (2010) suggested that a shallow-water sulphidic OMZ rapidly expanded immediately before the EPME. If this were the case, evidence of a shallow water OMZ should be found in the Iwaidani section. However, our interpretation of redox conditions based on our trace element data does not provide evidence on the position of chemoclines in the water column. Future geochemical research, such as iron speciation, total sulfur and stable sulfur isotopic composition, and the size of pyrite grains, will help confirm or refute this hypothesis (Wignall et al., 2010; Zhang et al., 2017; Takahashi et al., 2019; Maruoka and Isozaki, 2020).

Previous studies have suggested that the release of volcanic and contact metamorphism-related gases (CO₂ and CH₄) from the Siberian Traps would have enhanced global warming just before the PTB, which is supported by oxygen-isotope studies of marine biogenic calcite and apatite (Kearsey et al., 2009; Sun et al., 2012; Chen et al., 2016). The resulting temperature increase and the collapse of vegetation on land may have resulted in intensified chemical weathering via acceleration of the hydrological cycle (Sephton et al., 2005; Algeo and Twitchett, 2010; Sun et al., 2018), which would have caused increased nutrient discharge (e.g., nitrates and phosphates) to the global ocean and a resulting increase in primary productivity, leading to global oceanic anoxic conditions (Wignall, 2007; Algeo et al., 2011a; Zhang et al., 2018b; Schobben et al., 2020).

During stage 3, the siliceous rocks of the Iwaidani section record evidence for enhanced chemical weathering accompanied by a change from anoxic to euxinic conditions (Figure 14). Our study shows negative PC1 scores with increasing CIA and decreasing WIP values across the PTB, implying intensified chemical weathering in hinterland regions during the latest Changhsingian to earliest Induan. Given that the average sedimentation rate of the Changhsingian bedded chert is ∼0.3 cm·kyr⁻¹ (Onoue et al., 2019), enhanced continental chemical weathering occurred ca. 200 kyr before the EPME. These results support the interpretation of an acceleration of the hydrologic cycle and increased weathering rates immediately before the PTB, as evidenced by geochemical, biomarker, and isotopic studies (Sephton et al., 2005; Sheldon, 2006; Algeo and Twitchett, 2010; Kaiho et al., 2016; Sun et al., 2018; Cao et al., 2019).

The shift to a warmer climate during the transition between anoxic and euxinic depositional environments is also supported by the decrease in the enrichment factors of Ti, Zr, and Hf (Figure 6), which are primarily concentrated in heavy mineral assemblages such as ilmenite, rutile, titanomagnetite, and zircon (Schmitz, 1987; Shimmield and Mowbray, 1991; Calvert and Pedersen, 2007). Because the distribution of heavy minerals in deep-sea sediments is affected by changes in the grain size of wind-blown dust, the enrichments of these elements indicate increases in the intensity of aeolian transport and wind strength (Boyle, 1983; Beveridge, 1994; Calvert and Pedersen, 2007). Previous studies on past atmospheric transport of terrestrial materials (Rea et al., 1985; Rea, 1994; Anderson et al., 2006) suggest that decreases in Ti_EF, Zr_EF, and Hf_EF may reflect a transition to a more humid climate, and more specifically to humid conditions within the dust source regions and/or precipitation along the pathways of dust transport. Consequently, our results suggest that the climate changed to relatively warm and humid conditions during the ca. 200 kyrs prior to the EPME. The causes of this climatic change are debated, and hypotheses include an extraterrestrial influence on the influx of interplanetary dust particles (Onoue et al., 2019; Takahata et al., 2019); however, large-scale eruption of the Siberian Traps is commonly favored to explain the global warming and rapid enhancement of continental weathering (Wignall, 2007; Algeo et al., 2011a; Schobben et al., 2020). Two-thirds of the total lava volume of the Siberian Traps was deposited ca. 300 kyr before the EPME (Burgess and Bowring, 2015; Burgess et al., 2017). Grasby et al. (2011) suggested that a release of coal fly-ash generated by magma-coal thermal metamorphism in the Siberian Traps, which was dispersed globally during the 500–750 kyr before the EPME. Therefore, it seems reasonable to assume that intensified continental weathering indicated by our PC1 and weathering proxy data was related to initial emplacement of the Siberian Traps before the EPME.

In theory, enhanced chemical weathering via acceleration of the hydrological cycle would cause increased nutrient discharge (e.g., nitrates and phosphates) to the global ocean, with a consequent increase in biological productivity (Jones and Jenkyns, 2001; Jenkyns, 2010; Pogge von Strandmann et al., 2013). Elevated values of total organic carbon (TOC) and paleoproductivity proxy data (e.g., Ba and P) in PTB black claystones in Japan have been cited as evidence of increased primary productivity across the PTB in the Panthalassa Ocean (Suzuki et al., 1998; Algeo et al., 2011b), though nitrogen and carbon isotope data along the northwestern margin of Pangea (Grasby et al., 2016) and the eastern margin of the Panthalassa (Schoepfer et al., 2013) suggest nutrient-limited conditions and reduced primary productivity in the earliest Triassic. Barium is commonly used as a tracer of paleoproductivity (Zachos et al., 1989; Dymond and Collier, 1996; McManus et al., 1998; Algeo et al., 2011b), since barite precipitation occurs in decaying particulate organic matter while it sinks to the seafloor (Dehairs et al., 1980; Bishop, 1988; Dymond and Collier, 1996), which is supported by enhanced Ba_EF values below high-productivity areas (Nürnberg et al., 1997). Our data show that Ba_EF decreased significantly during the Wuchiapingian and then remained consistently low until the Induan (Figure 8). Thus, the Iwaidani section demonstrates no apparent link between increased chemical weathering and paleoproductivity during the latest Changhsingian. However, Ba cannot be used as a proxy for paleoproductivity under anoxic to euxinic conditions because 1) low sulfate concentrations in seawater, which are inferred for the late Permian–Early Triassic (Bottrell and Newton, 2006; Luo et al., 2010), reduce the solubility product of barite in seawater; and 2) anoxic conditions in sediment pore water cause the reductive dissolution of sedimented barite (Algeo et al., 2011b). Since the stratigraphic profile of Ba_EF mimics that of Th/U ratios and is inversely correlated with U_EF (Figures 7, 8), Ba_EF might not record variations in paleoproductivity, but instead reflect the dissolution of barite in the anoxic–euxinic environments during the Changhsingian and Induan. Consequently, the variation in paleoproductivity in relation to increased chemical weathering could not be investigated in the Iwaidani section, and the causal connection between chemical weathering and the development of euxinic conditions during the latest Changhsingian remains unclear. We need further investigation, such as carbon and nitrogen isotope analyses, to evaluate this connection.

Like the PTB, climate change induced by large-scale volcanism has also been proposed for the Capitanian (e.g., Wignall et al., 2009; Bond et al., 2010). The Emeishan Traps in South China were identified as a possible cause of the global warming during the late Capitanian (Chen et al., 2011); thus, enhanced chemical weathering during this period is expected (Zhang et al., 2019). However, the PC1 scores in the Iwaidani section did not change during the Capitanian and earliest Wuchiapingian (Figure 14). Chen et al. (2011), Chen et al. (2013) used oxygen isotopes from conodonts and found an increase in the paleotemperature during the late Capitanian, which reflects the end-Guadalupian global regression (Haq and Schutter, 2008) and shallowing of conodont habitats. This interpretation appears to be more appropriate in explaining the increase in seawater temperature in South China during the late Capitanian.

Our data reveal that suboxic to weakly anoxic conditions developed ca. 8 Myr prior to the PTB. There is no simultaneous increase in chemical weathering recorded by the negative PC1 scores during the prolonged suboxic–anoxic conditions in the Lopingian, except near the PTB (Figure 14). The cause of the onset of deep-sea anoxia remains uncertain, but our new data largely exclude the possibility that the event was induced by abrupt warming and subsequent acceleration of the hydrological cycle, triggered by degassing related to large-scale volcanic activity. The weakening of ocean circulation with the end of the Gondwana glaciation and Kamura cooling event (Isozaki et al., 2007a; Isozaki et al., 2007b) might be one possible explanation for a relatively stagnant ocean (Kajiwara et al., 1994; Isozaki, 1997) that led to the development of deep-sea anoxia during the Lopingian. Across the GLT, a decrease in the PC3 scores was associated with decreased concentration of biogenic apatite, which is sourced primarily from conodonts and pelagic vertebrates such as fish. Regions of upwelling around seamounts exhibit increased nutrient concentrations in surface waters, which result in the proliferation of fish in pelagic realms (Ohta et al., 2020). The decreasing PC3 trend in this study may therefore reflect the weakening transport of deep nutrient-rich cold water to the surface around the seamount by oceanic upwelling, due to more stagnant ocean circulation. However, a causal link between a reduced nutrient supply and conodont productivity has not been found. Further geological and geochemical research is needed to verify the exact nature of the deep-sea anoxia.