## Introduction

The mid-Cretaceous (Barremian–Turonian) is punctuated by repeated Oceanic Anoxic Events (OAEs), which represent intervals of global and episodic burial of organic-rich sediments on the seafloor under oxygen-deficient bottom waters. Several organic-rich sedimentary deposits are recorded in the uppermost Aptian to the Lower Albian, mainly in the Tethys and Atlantic Oceans1,2,3,4,5 (Fig. 1). In particular, four prominent black shale horizons (113/Jacob, Kilian, Urbino/Paquier, and Leenhardt equivalent levels) have been identified as the sedimentary expression of OAE1b4. OAE1b is characterized by (1) an exceptionally long duration (~ 3.8 Myr) with intermittent occurrences of oxygen-depleted bottom water conditions, (2) a global carbon-cycle perturbation, and (3) a major marine biotic turnover4. In particular, planktonic foraminifera experienced one of the most significant turnovers in their evolutionary history, where large, heavily calcified planktonic foraminifera of the Aptian were replaced by small, weakly calcified taxa characteristic of the Albian4,6,7,8.

Since 40Ar–39Ar ages of the Southern Kerguelen Plateau basalt (Fig. 1) (109.2–119.0 Ma)9 roughly correspond to the duration of OAE1b (~ 110.5–114.5 Ma), volcanic episodes associated with the break-up of Gondwana have been suggested as the trigger of OAE1b1,3,5. However, the issue remains debated because of the large chronological uncertainties of the 40Ar–39Ar ages of the basaltic rocks and the ages of the sedimentary sequences. Marine osmium (Os) isotopic records (187Os/188Os) reflect the balance between the continental Os flux (187Os/188Os ≈ 1.0–1.5) and mantle/hydrothermal and extraterrestrial Os fluxes (187Os/188Os ≈ 0.12–0.13) to the global ocean10. Thus, 187Os/188Os values of palaeo-seawater preserved in sedimentary rocks represent a robust proxy to constrain the timing of massive input of unradiogenic Os through submarine hydrothermal or volcanic eruption. Since the estimated residence time of Os in the modern ocean is 104–105 years, that is longer than the timescale of modern oceanic circulation10,11, seawater 187Os/188Os values are relatively constant throughout the global oceans. Here, we present Late Aptian to Early Albian palaeo-marine Os isotopic variations in the Poggio le Guaine (PLG) record (central Italy, deposited in the central to western Tethys) (Fig. 1 and Supplementary Fig. S1) and Deep Sea Drilling Program (DSDP) Site 463 (western Mid-Pacific mountains, central Pacific Ocean) (Fig. 1 and Supplementary Fig. S2) to constrain the timing of extensive volcanic episodes during OAE1b. For the Os isotopic analysis of organic-rich sedimentary rock samples of PLG section, we preferentially used borehole core12 samples drilled near the PLG section because initial Os isotopic information of these rocks are easily altered by weathering.

### Lithological description

#### PLG records.

The PLG section (Fig. 1, Supplementary Figs. S1, S3 and Supplementary Note 1.1) is one of the most continuous sedimentary successions encompassing OAE1b4. Since coarse silicate fragments are rare throughout this section (Fig. 2a–c), this sedimentary sequence was considered deposited in a pelagic environment. The pre-OAE1b interval mainly comprises reddish or olive-grey argillaceous limestone and marlstone enriched in heavily calcified large planktonic foraminifera (Figs. 2a, 3a, 4 and Supplementary Fig. S3a), whereas the OAE1b interval is characterized by a cyclic alternation of olive to greenish-grey marlstone/mudstone and black shale (Figs. 3a, 4 and Supplementary Fig. S3b,c) where the size and abundance of planktonic foraminifera decreased (Figs. 2b, 4). The 113/Jacob, Kilian, Urbino/Paquier, and Leenhardt equivalent levels are the regional sedimentary expression of OAE1b sub-events4 (Figs. 2c, 3 and Supplementary Figs. S3, S4). The 113/Jacob equivalent level is the first organic-rich sediment during OAE1b. A distinctive sequence of alternating reddish and black shales, called Monte Nerone interval, occurs between the Kilian and Urbino/Paquier equivalent levels (Fig. 3). Above 15.42 m stratigraphic level in the section (pink line in Figs. 3, 4), small, weakly calcified Albian planktonic foraminiferal taxa appear followed by the abrupt decrease in the number of large (> 250 μm), heavily calcified Aptian planktonic foraminifera4. The heavily calcified planktonic foraminifera characteristic of Aptian never appeared above the thick black shale horizon called Kilian level (16.97–17.37 m)4 (Fig. 4 and Supplementary Fig. S3b). This major planktonic foraminiferal species turnover is accompanied by a significant increase in the radiolarian abundance4 and a decrease in calcium carbonate (CaCO3) content (Fig. 4). The Kilian equivalent level is punctuated by a major negative carbon isotopic excursion (CIE) (Fig. 3) coupled with a marked decrease of CaCO3 content and the dominance of agglutinated foraminiferal taxa4 (Fig. 4).

The drilling site of PLG-core is located 400 m northwest of the PLG Section12. The Upper Aptian–Lower Albian sedimentary and biostratigraphic record of PLG core is well correlated to that of PLG section. Therefore, the geochemical data of outcrop and borehole core can be confidently combined.

#### DSDP Site 463

The Lower Cretaceous record at DSDP Site 463 (Fig. 1 and Supplementary Fig. S2) mainly comprises varicoloured limestone and marlstone with minor chert layer (Fig. 2d–f, Supplementary Fig. S5, and Supplementary Note 1.2). The Upper Aptian to Lower Albian sedimentary sequence lacks organic-rich sediments13 and does not exhibit any marked lithological changes.

## Results

We measured carbonate carbon isotopic compositions (δ13Ccarb) of limestone and marlstone samples of PLG section and DSDP Site 463 for stratigraphic correlation (see Method 4) (Fig. 3, Supplementary Fig. S6 and Tables S1, S2, and Supplementary Note 2). The δ13Ccarb values of PLG section were 1.7–3.4‰ (Fig. 3a) and consistent with those of presented in ref4. Although the δ13Ccarb values of DSDP Site 463 (2.4–4.6‰; Fig. 3b) are higher than those ones at PLG section by ~ 1‰, their temporal variation is quite similar. The δ13Ccarb record at DSDP Site 463 shows a negative CIE at ~ 539 m below seafloor (mbsf), just above the Schackoina cepedai acme at 539.8 mbsf (Fig. 3b and Supplementary Note 3). Since the S. cepedai acme at the PLG section falls ~ 1.2 m below the 113/Jacob equivalent level (the first organic-rich horizon in the Tethys region during OAE1b) which is marked by a negative CIE4 (Fig. 3), we correlate the CIE in the Pacific Ocean with the 113/Jacob equivalent level (Fig. 3). A more distinct negative CIE (~ 1‰) occurs at 529.33 mbsf, just above the lowest occurrence of Microhedbergella renilaevis (529.56 mbsf). Since these features are consistent with those recorded across the Kilian equivalent level, which records the demise of the Aptian planktonic foraminifera in the PLG section, we correlate this CIE at DSDP Site 463 to that of the Kilian level (Fig. 3).

At the PLG section, the total organic carbon content (TOC) values of pre-OAE1b ranges from 0.01 to 0.04% and averages 0.03% (see Methods 5) (Supplementary Fig. S7 and Table S3). The TOC values of black shale horizons during OAE1b are higher than other horizons. In particular, TOC values across the Jacob and Urbino equivalent levels show extremely high values (up to 8%). Other black shale intervals (i.e., Kilian equivalent level, Monte Nerone interval, and Leenhardt equivalent level) reveal lower concentrations of TOC (up to 0.65%). TOC values of sedimentary rocks from the DSDP Site 463 are 0.04–0.12% and we could not find organic rich intervals (Supplementary Table S4).

We conducted Re-Os analysis of limestone, marlstone, mudstone, and black shale samples collected from the PLG record (outcrop and core) and limestone and marlstone samples of the DSDP Site 463 (see Method 6) (Fig. 3, Supplementary Figs. S8, S9 and Tables S5, S6 and Supplementary Note 4). Considering the positive correlations between TOC and Re, Os concentrations throughout the PLG section, these elements are derived from the hydrogenous fraction associated with organic matters (Supplementary Fig. S10)14,15. Since there are no clear relationships between other elements (e.g. Fe and Mn), the contribution of Os and Re derived from other fractions (such as ferromanganese oxides)16 is considered minor (Supplementary Fig. S11 and Table S7). Most of the PLG core samples were treated with aqua regia digestion. However, for some samples of PLG section, we applied both CrO3-H2SO4 and inverse aqua regia to check whether our method extract hydrogenous fraction or not. On the basis of these results, we could not find any significant differences in 187Os/188Osi (see Method 6) (Supplementary Figs. S12, S13 and Supplementary Note 4), which supports our approach to successfully extract the hydrogenous information.

### Osmium isotopic records across the Aptian–Albian boundary

In the pre-OAE1b interval, 187Os/188Osi values (i.e., initial 187Os/188Os values, corrected for the radioactive decay of 187Re to 187Os, see methods) range from 0.52 to 0.57 (average 0.54) in the PLG section and from 0.49 to 0.51 (average 0.50) at DSDP Site 463 (Figs. 3, 4, 5). These values are slightly lower than the pre-OAE1a values of the Lower Aptian (~ 0.7)17 and pre-OAE2 values of the Upper Cenomanian (~ 0.7)18. Since several seawater temperature proxies (belemnite oxygen isotopic ratios, the organic palaeothermometer TEX86, and the existence of glendonites at higher latitudes) suggest a cooler climate during the Late Aptian than during the Early Aptian and Late Cenomanian19,20,21, these low 187Os/188Osi values may reflect weak continental weathering under cool climatic conditions.

The 187Os/188Osi values began to fluctuate immediately after the onset of OAE1b (i.e. the deposition of the Jacob level) in the PLG section. The first sharp negative spike of 187Os/188Osi values appears few-cm below the major planktonic foraminiferal turnover (15.48 m; pink line in Fig. 3) at the PLG section. Similar Os isotopic spike can be also recognized at the DSDP Site 463 (531.08 mbsf: Fig. 3a). The 187Os/188Osi values again sharply decline in the lower part of the Kilian level in the PLG section, then rapidly increase to 0.73 (Fig. 3a). Although the pattern of 187Os/188Osi variations is quite similar between the two sites, the amplitude of 187Os/188Osi variations at the Kilian equivalent level at DSDP Site 463 (0.54–0.62) is much smaller than that in the PLG Section (0.40–0.73) (Fig. 3b). Considering the long residence time of Os in the present ocean (10–100 kyr)10,11, the significant difference in the range of 187Os/188Osi values between these two localities is a conundrum. Borehole core samples from both sites are fresh, and the influence of weathering seems insignificant. We consider active bioturbation or a small hiatus around the Kilian equivalent horizon at DSDP Site 463 to be the most probable explanation of the decreased 187Os/188Osi variations there. Indeed, the amplitude of the negative δ13Ccarb spike at the Kilian equivalent level of DSDP Site 463 (~ 1‰) is smaller than the one at PLG section (> 2‰), which also implies the mixing of sediments or existence of minor hiatus during Kilian event. The sedimentary expression of the Kilian level at the PLG section is a 40 cm-thick black shale with minor bioturbation4 whereas that of DSDP Site 463 is bioturbated lime/marlstone thinner than 20 cm (Fig. 2e and Supplementary Fig. S5). Therefore, we consider the Os isotopic variations in the PLG section to be less disturbed and to better reflect the initial Os isotopic signature of seawater than those at the DSDP Site 463.

The two sharp declines of 187Os/188Osi values to ~ 0.4 (Figs. 3 and 4) can be explained either by (1) rapid declines in continental weathering, (2) large meteorite impacts, or (3) increases in hydrothermal activity. Here, we apply a simple box model calculation to evaluate these possibilities (see Method 7). The first one requires an abrupt 35% decrease in radiogenic continental Os within a few hundred thousand years (Supplementary Table S8), probably through rapid cooling events. However, no such abrupt cooling of the climate has yet been reported. Instead, intensive warming was reported at the Kilian equivalent level in the Atlantic Ocean19,21, which, on the contrary, would have accelerated chemical weathering. Decreased continental weathering therefore seems unlikely as the cause of the decreased 187Os/188Osi values. Although a meteorite impact could explain the sharp decline of 187Os/188Osi values, neither a large meteorite crater nor tektites have been reported in correspondence of this stratigraphic interval. Therefore, extensive volcanic events on the Southern Kerguelen Plateau9, which could have released a large amount of unradiogenic Os through hydrothermal activities, seems the most reasonable explanation. Our box-model calculation shows Kerguelen volcanic episodes at that time promoted a 60–90% increase in the unradiogenic Os flux through hydrothermal activities. The 187Os/188Osi values abruptly increase from 0.40 to 0.73 in the upper part of the Kilian level in the PLG section (Fig. 3). Based on our box-model calculation, this abrupt increase in 187Os/188Osi values can be explained by either a 41% decrease in the hydrothermal Os flux associated with the oceanic crustal production or a 67% increase in the continental weathering rate. Considering the intensive warming during the Kilian event19,22, a rapid increase in the radiogenic Os flux through continental weathering appears plausible. Based on the characteristics of Kerguelen lava flows, a large part of the volcanic plateau was emplaced subaerially, and the submarine eruption was short lived23. We interpret this positive 187Os/188Osi anomaly as a later eruptive phase when active CO2 degassing via subaerial eruptions enhanced continental weathering and weakened the hydrothermal input of unradiogenic Os. Similar positive 187Os/188Osi anomalies have not been reported yet for other Cretaceous OAEs (OAE1a17 and OAE218). This might be ascribed to the entirely submarine volcanic eruptions of the Ontong Java Plateau and Caribbean Plateau, which triggered OAE1a and OAE2, respectively, and continuously supplied a large amount of unradiogenic Os through hydrothermal activities throughout the eruptions.

The 187Os/188Osi values above the Kilian equivalent level (average 0.63 both in the PLG section and at the DSDP Site 463) are higher than the pre-OAE1b background values (0.54 in the PLG section and 0.50 at the DSDP Site 463) (Fig. 3). Planktonic foraminifera oxygen isotopic records in the Atlantic Ocean (Ocean Drilling Program Site 1,049) indicate a rise in seawater temperature after the Aptian–Albian transition relative to the pre-OAE1b interval24. Furthermore, the disappearance of glendonite from Arctic sediments21 supports a warming climate after the Aptian–Albian boundary, which could have sustained high continental weathering rates and high marine 187Os/188Osi values. The radiogenic strontium isotopic ratio (87Sr/86Sr) of seawater also continuously declined from the Early to Late Aptian, with an incipient increase above the Aptian–Albian boundary25. Such trends suggest a larger contribution of continental weathering during OAE1b (Early Albian) than during the pre-OAE1b interval (Late Aptian). After the initial sharp declines of 187Os/188Osi values to ~ 0.4, 187Os/188Osi values fluctuated throughout OAE1b from 0.53 to 0.68 in the PLG section and from 0.58 to 0.71 at DSDP Site 463 (Fig. 3). Considering the protracted volcanic eruptions of the Southern Kerguelen Plateau9, extended minor eruptions after the main volcanic pulse might have caused fluctuations of the continental weathering rate and/or the hydrothermal Os influx, eventually contributing to the prolonged perturbation of the Os cycle (Fig. 3).

### Major planktonic foraminiferal turnover

The two volcanic Os isotopic signals correspond to the major planktonic foraminiferal turnover interval characterized by the extinction of large, heavily ornamented calcified planktonic foraminifera and the speciation of smaller thin, weakly calcified species that lack ornamentation in the PLG Section4,8. These volcanic horizons are accompanied by abrupt decreases in CaCO3 content and the poor preservation of calcareous microfossil tests, which means the dissolution of CaCO3 on the seafloor. Moreover, the relative abundance of agglutinated benthic foraminifera, which is resilient to ocean acidification, increased at the same time in the Tethys region4 (Fig. 4). These features suggest a shallowing of the carbonate compensation depth (CCD)26,27,28 at these volcanic horizons. A massive input of volcanic CO2 could have prompted ocean acidification, potentially leading to dwarfism of planktonic foraminifera and shallowing of the CCD. Positive relationships among pH, test thickness, and growth rates of foraminifera have been documented in laboratory experiments and geological observations29,30,31. Therefore, the drastic turnover and reduction in the size of planktonic foraminifera may reflect a decline of pH caused by the massive release of volcanic CO2 (Fig. 4). Indeed, since acidified condition prevents the growth of ornamentation and spines of foraminifera28,31,32, the lack of ornamentation in smaller Albian species4,8 might reflect adaptation to the acidified oceanic condition. No major extinction of calcareous nannoplankton has been reported during OAE1b. Marine organisms with extremely small calcified tests might have been less influenced by ocean acidification33,34, probably, because of their more efficient proton pumping systems in the carbon pools inside their cells relative to larger calcareous shell-forming organisms35.

The large negative δ13Ccarb spike at the Kilian equivalent level implies a more intensive volcanic input of 13C-depleted CO2 than during the first volcanic pulse at the beginning of the biotic turnover (Fig. 5). The resulting intensive climatic warming and subsequent massive input of nutrients into the ocean through continental weathering could have triggered increased primary production, leading to oceanic anoxia in the Tethyan and Atlantic Oceans (Fig. 5).

No evidence of an O2-depleted environment has yet been reported for the first volcanic pulse. Furthermore, O2-depleted conditions during the Kilian event (i.e., during the latest stage of the biotic turnover) were limited to the Tethyan and the Atlantic Oceans. These facts might imply that oceanic anoxia was not the direct trigger of the major marine biota turnover during OAE1b and possibly, the ocean acidification induced by volcanic events could have contributed to the biotic crisis. This conclusion suggests that oceanic anoxia and the extinction of marine biota are different phenomena that should be separately discussed.

## Summary

Here, we presented continuous marine Os isotopic variation across the Aptian–Albian boundary at Tethyan pelagic section (PLG section) and Pacific sediments (DSDP Site 463). We found two sharp negative shifts around the Aptian–Albian boundary, which suggest submarine volcanic eruptions. Since these intervals corresponds to the major planktonic foraminiferal species turnover and neither of them correspond to global oceanic anoxia, we concluded this turnover was likely caused by ocean acidification triggered by two major volcanic events.

## Methods

### Micropalaeontological studies

All samples were analysed for planktonic foraminifera. On average, 8 g of rock were processed for each sample. Owing to the hard lithology, all samples were mechanically disaggregated into small fragments (3–8 mm) and treated following the cold acetolysis technique of ref36 by sieving through a 63 μm mesh and drying at 50 °C. The cold acetolysis method enables the extraction of easily identifiable foraminifera. This technique offers accurate taxonomic determination and detailed analysis of planktonic foraminiferal assemblages, allowing a more precise placement of several bioevents and zonal boundaries.

Moderately to well-preserved planktonic foraminifera were present in almost all samples. Planktonic foraminifera from the washed residues of each size fraction (63–100, 100–125, 125–150, 150–180, 180–250, 250–355, 355–500, and > 500 μm were separately studied under a stereomicroscope. Taxonomic concepts for genera and species identification in this study mainly follow ref6 and the online Mikrotax database (https://mikrotax.org/pforams/index.php?dir=pf_mesozoic), where the type materials for most species are illustrated by recent scanning electron microscopy images37. The biostratigraphic framework used in this study is after ref6,38, and references therein.

### Measurement of CaCO3 content

The sedimentary rock samples of PLG section were crushed and ground into fine powder in an agate mortar. Powdered samples were dissolved in 10% vol. HCl and CaCO3 content were obtained by measuring CO2 volume using a Dietrich–Frühling calcimeter. Detailed procedure is the same of ref4..

### Sample trimming

The surfaces of the collected sediment samples were trimmed with a spatula and weathered parts were removed. The samples were then crushed into coarse fragments, from which fresh fragments were collected. These samples were washed in deionized water in an ultrasonic bath and dried in an oven at 60 °C for at least 12 h. The dried samples were crushed and ground to a fine powder using an agate mill. These powdered samples were used for stable carbon isotopic analyses, total organic carbon measurements to determine the host phases of Os and Re, and finally for Re-Os isotopic analyses.

### Stable carbon and oxygen isotopic compositions of carbonate

Carbonate stable carbon (δ13Ccarb) and oxygen isotopic compositions (δ18Ocarb) were measured in 39 samples from the PLG section and 68 samples from DSDP Site 463 using a GV IsoPrime instrument at the Japan Agency for Marine-Earth Science and Technology (JAMSTEC). The isotopic compositions are expressed in delta notation as per-mil variations relative to Vienna Peedee Belemnite (VPDB). Analytical errors (1σ) on δ13Ccarb and δ18Ocarb were estimated to be within 0.1‰ and 0.2‰, respectively on the basis of repeated measurements of in-house standard materials. The isotopic analytical method is detailed by ref39

### Stable carbon isotopic compositions of organic matter and total organic carbon contents (TOC)

We determined total organic carbon contents (TOC) and organic stable carbon isotopic compositions (δ13Corg) for sedimentary rock samples from the PLG section and 11 sedimentary rock samples from the PLG core. Powdered and weighed samples were decalcified with 2 M HCl, rinsed with purified water using a MilliQ Water purification system. After evaporation to dryness, these samples were weighted to calculate the change in weight during decalcification. These samples were wrapped in a Sn cup for measurements. Carbon contents and δ13Corg were measured at JAMSTEC via on-line system of isotope-ratio mass spectrometry (DeltaPlus XP; Finnigan, Waltham, MA, USA) coupled to a Flash EA 1,112 Automatic Elemental Analyzer through a ConFlo III interface40 Analytical errors (1σ) on δ13Corg is better than 0.37‰ based on repeated measurements of the in-house standard (L-Tyrosine).

### Rhenium and osmium analysis

We used CrO3-H2SO4 digestion41 and inverse aqua regia digestion methods to extract Re and Os from the samples. Samples for Re-Os analyses varied from 0.1 to 1 g depending on the expected concentration of Os in the samples. In fact, organic-rich black shale samples generally show high Os and Re concentration and marl/limestone samples with low TOC shows lower Os and Re concentrations. After spiking using 190Os- and 185Re-rich solutions, each sample was sealed in a Carius tube42 with 4 mL of CrO3-H2SO4 (0.2 g CrO3 per 1 mL of 4 N H2SO4) or inverse aqua regia (mixture of 1 mL of 30 wt% HCl and 3 mL of 68 wt% HNO3) digestion solutions. The sample solutions were heated at 240 °C for 48 h. Through this process, Os and Re were completely extracted from the samples as Os(VIII)O4 and Re(VII)O4, respectively, and isotopic equilibrium between the spike and sample was achieved. The supernatant (leachate) was separated from the residue by centrifugation. After removing the residue, Os was separated from the leachate with 3 mL of carbon tetrachloride (CCl4) in three successive extractions. The volatile Os(VIII)O4 was reduced to non-volatile Os(IV)Br62– by adding 3 mL of 9 N HBr. Extracted Os was purified by micro-distillation43. Re was separated from the leachate thorough following two steps: (1) 2 mL of Bio-Rad AG1-X8 anion exchange resin (100–200 mesh) and (2) 0.3 mL of Bio-Rad AG1-X8 anion exchange resin (100–200 mesh).

Os abundances and isotopic compositions were determined by negative thermal ionization mass spectrometry (Thermal Electron TRITON) (ref44 and references therein) at JAMSTEC (Japan) and Re abundances and isotopic compositions by quadrupole inductively coupled plasma mass spectrometry (iCapQ) at JAMSTEC using in-house Re standards. All data were corrected for procedural blanks, whose respective averages were 0.57 pg Os and 5.4 pg Re. The average of 187Os/188Os of procedural blank is 0.13. Instrumental reproducibility (standard error) was monitored based on replicate analyses of the in-house standard for 187Os/188Os = 0.106838 ± 0.000015 (2σ)45.

The 187Re decays to 187Os in sediments with a decay constant of 1.666 × 10–11 yr–l46. Therefore, the initial 187Os/188Os of sediments (187Os/188Osi) was calculated as:

$${}^{187}Os/{}{}^{188}{Os}_{i}={}{}^{187}Os/{}{}^{188}{Os}_{m}-[exp\left\{\lambda \times age\left(yr\right)\right\}-1] \times {}{}^{187}Re/{}{}^{188}{Os}_{m}$$
(1)

Here, λ is the decay constant of 1.666 × 10–11 year–l and the subscript ‘m’ indicates measured values. For the sedimentary rock samples from the PLG section and PLG core, ages used for the correction were 119 Ma from 0 to 6 m in the section, 117 Ma from 6 to 9 m, 115 Ma from 9 to 12 m, 114 Ma from 12 to 15 m, 113 Ma from 15 to 20 m, 112 Ma from 20 to 23 m, and 111 Ma from 23 to 26 m. In the DSDP Site 463 core, ages used for the correction were 114 Ma for DSDP Site 463, cores 62–61 and 113 Ma for DSDP Site 463, cores 61–58 (Supplementary Tables S5, S6). These ages were determined based on refs4,47.

### Calculation of Os fluxes using a simple box model

We calculated the changes in the Os flux through hydrothermal activities and continental weathering using a zero-dimensional box model based on ref17. This model assumes the ocean to be a unique Os reservoir, and its Os content and isotopic composition to reflect the balance between continental input, hydrothermal input related to oceanic crustal production and volcanic eruptions at the Southern Kerguelen Plateau, extraterrestrial input, and a sedimentary sink. These relationships are described as:

$$\frac{d{M}_{ocean}}{dt}={F}_{cont}+{F}_{hydr}+{F}_{cosm}+{F}_{Kerg}-{F}_{sed}$$
(2)
$$\frac{d({M}_{ocean}{R}_{ocean})}{dt}={F}_{cont}{R}_{cont}+{F}_{hydr}{R}_{hydr}+{F}_{cosm}{R}_{cosm}+{F}_{Kerg}{R}_{Kerg}-{F}_{sed}{R}_{sed}$$
(3)

where M, F, and R indicate the amount, flux, and isotopic ratio (187Os/188Os) of Os, and the subscripts ‘ocean’, ‘cont’, ‘hydr’, ‘cosm’, ‘Kerg’, and ‘sed’ represent the oceanic reservoir, continental input, hydrothermal input, extraterrestrial input, input from the Southern Kerguelen Plateau, and sedimentary output, respectively. Since the isotopic fractionation of Os is negligible between sediments and seawater, we assumed that Rsed coincides with Rocean. Then, the above equations can be combined as:

$$\frac{d{R}_{ocean}}{dt}=\frac{[{F}_{cont}\left({R}_{cont}-{R}_{ocean}\right)+{F}_{hydr}\left({R}_{hydr}-{R}_{ocean}\right)+{F}_{cosm}({R}_{cosm}-{R}_{ocean})]}{{M}_{ocean}}$$
(4)

Here, we used the present-day values of Fcont = 295 t/kyr, Rcont = 1.5410,48, Rhydr = 0.12649 Fcosm = 17.6 t/kyr, and Rcosm = 0.12610,49 for the steady background conditions below the Jacob level. The steady background oceanic condition is Rocean = 0.54, and we set Fhydr to 532.7 t/kyr to match this value because several studies have suggested that hydrothermal activities associated with the production of oceanic crust in the Cretaceous were more active than today (e.g., ref50). For the output parameters, we assumed that Rsed varies proportionally to Rocean, and set the coefficient of proportionality at 0.056 following ref17. Since there are no Os isotopic data available for Southern Kerguelen Plateau basalts, we used the isotopic composition of Kerguelen Archipelago lavas (187Os/188Os = 0.16)51,52 for RKerg. At steady state, FKerg is considered to be 0 t/kyr. Only Os isotopic values from the PLG section were used for Rocean because the PLG section is more continuous and spans a longer time interval than DSDP Site 463. Detailed parameter values throughout the cores are listed in Supplementary Table S8.

### Major elemental composition of bulk sedimentary rock samples

The powdered samples and lithium tetraborate flux (Li2B4O7, MERCK) were heated at 110 °C for more than 12 h. Dried powdered samples were weighed and heated at 950 °C for 7 h. After measurement of their weight to calculate loss of ignition, about 0.4 g samples and exactly ten-times larger amount of Li2B4O7 were mixed and heated to make glass beads. They were analyzed by ZSX Primus II XRF spectrometer (Rigaku) installed in Atmosphere and Ocean Research Institute (AORI), the University of Tokyo, Japan to determine major elemental composition.