Introduction

Compared to more recent periods, interglacial atmospheric CO2 concentrations and Antarctic air temperatures were lower during the period between 800 and 430 thousand years ago (ka), commonly referred to as the “lukewarm interglacials”1,2. The shift to warmer interglacials at ~430 ka, known as the Mid-Brunhes Event (MBE)3 or Mid-Brunhes Transition (MBT)4, marks a key change in the Earth’s climate history.

The Southern Ocean (SO) is a critical region where carbon- and nutrient-rich deep waters are upwelling, causing enhanced surface ocean productivity and driving substantial carbon exchange between the deep ocean and the atmosphere5,6. In addition to upwelling, the SO is also a key region of Antarctic Bottom Water (AABW) formation (downwelling), which regulates ocean-atmosphere carbon exchange7. As such, the SO has been identified as an important region controlling atmospheric CO2 and climate dynamics across the MBE8,9,10,11,12. Mechanisms proposed to explain lower CO2 during the lukewarm interglacials include increased AABW production8,12, northward shifts of SO fronts13 and weakened SO upwelling11, all of which involve changes in basin-scale deep ocean circulation. However, reconstructions of SO circulation during the lukewarm interglacials remain scarce, in particular for regions south of the Antarctic Polar Front (APF) where key processes like AABW formation and upwelling take place.

Reconstructing past SO circulation is particularly challenging due to the lack of carbonate sediments from this region, which restricts the availability of reliable age data (based on 14C and δ18O) and limits the use of traditional paleo water mass tracers such as δ13C and Cd/Ca signatures in carbonate tests (shells) of fossil foraminifera. Additionally, other water mass tracers, such as seawater neodymium (Nd) and lead (Pb) isotopes, preserved in sediments are susceptible to alteration and contamination from high influxes of chemically immature glacial detritus from the Antarctic margin14,15.

In this study, we reconstruct SO circulation changes using radiogenic Pb isotope records from a hydrogenetic ferromanganese crust dredged on “Haxby Seamount”, one of the Marie Byrd seamounts on Antarctica’s Pacific continental margin (69 °8.16’S, 123 °12.76’W-69 °8.53’S, 123 °13.13’W; water depth: 1502–1734 m, Fig. 1)16. Hydrogenetic ferromanganese crusts precipitate slowly from seawater on hard substrates in areas with very low or absent sedimentation, making them reliable archives of past SO seawater compositions with minimal diagenetic alteration. Age control of ferromanganese crusts can be obtained using 230Th and 10Be dating methods. To address the challenges posed by their slow growth rates (a few millimetres per million years) and complex internal structures, we apply laser ablation coupled with multiple-collector inductively coupled plasma mass spectrometry (LA-MC-ICPMS)17,18 and two-dimensional (2D) mapping techniques19,20. This combined novel approach enables us to confidently resolve past Pb isotope variations on orbital timescales.

Fig. 1: Sample locations.
figure 1

a Meridional distributions of oxygen levels (colour shading) overlain by neutral density (γn) contours (white curves)81,82. Locations of ferromanganese crust Haxby and core PS2551 are shown by a red circle and a yellow diamond, respectively. Locations of ferromanganese nodules from the South Pacific (orange), Pacific sector of the SO (dark blue), and Drake Passage (light blue) are marked as squares in the inset map. b 208Pb/206Pb-206Pb/204Pb plot of preindustrial Pb isotope signals recovered from surfaces of ferromanganese nodules and crusts in the upper panel30,40,83. Southern Ocean nodules show that AABW in the Pacific sector is characterized by low 208Pb/206Pb (<2.0625) ratios, whereas South Pacific nodules show that Pacific waters, such as PDW, carry higher 208Pb/206Pb ratios (>2.0625). AABW Antarctic Bottom Water, PDW Pacific Deep Water, LCDW Lower Circumpolar Deep Water.

Full size image

Lead has three radiogenic isotopes, i.e., 206Pb, 207Pb and 208Pb produced from the decay of 238U, 235U and 232Th, respectively, and one primordial isotope 204Pb. Seawater Pb primarily originates from weathering of continental source materials supplied mainly via fluvial inputs21,22 and, in polar regions, via glacial meltwaters. Due to its relatively short residence time in the ocean (~50 to 200 years)23,24,25, Pb isotopes serve as sensitive tracers for continental inputs and basin-scale water mass mixing. In the SO, wind-driven upwelling transfers Circumpolar Deep Water (CDW), the main circum-Antarctic water mass, towards the surface south of the APF. In our study area, Lower CDW (LCDW) is mainly a mixture of AABW formed in the Ross Sea and North Atlantic Deep Water (NADW)7,26, whereas less dense Pacific Deep Water (PDW) is entrained in Upper CDW (UCDW)27 (Fig. 1a). At present, crust Haxby is located in the pathway of deep upwelling LCDW and thus ideally suited for monitoring deep water mixing in the high-latitude SO. Our data show marked changes in 208Pb/206Pb values across the MBE, which we attribute to shifts in the mixture between AABW and LCDW, highlighting the SO’s role in regulating the global carbon cycle.

Results and discussion

Acquisition of orbital-resolution Pb isotopes time series

Previous methods18 for acquiring orbital-scale seawater isotope records from ferromanganese crusts and nodules have been limited by issues such as age profile and proxy data measured along different laser tracks, as well as unidentifiable post-diagenetic effects, both of which introduce significant uncertainty into the time series records. In this study, we overcome these challenges by performing sequential 2D scans of elemental and isotopic compositions on the same area of the crust (detailed in Methods). This allows us to precisely extract elemental ratios, age data, and Pb isotope compositions from each pixel (Figs. 2 and 3). Furthermore, post-diagenetic material can be clearly identified and excluded, ensuring consistent, high-resolution Pb isotope records across different analyzed areas.

Fig. 2: Mn/Fe composition and age of the studied ferromanganese crust.
figure 2

a Mn/Fe ratios with age contours marking glacial intervals (ages in ka). The glacial age contours always correspond to relatively low Mn/Fe ratios, while relatively high Mn/Fe ratios reflect interglacials. b and c show 2D/3D age images for the selected areas indicated in a.

Full size image
Fig. 3: Elemental and isotope compositions of a selected area of crust Haxby.
figure 3

a Backscatter Electron (BSE) image analysed by electron microprobe. The area marked by the red rectangle was used for Pb and Tl elemental as well as Pb isotope analyses; b 2D and 3D age images; c Mn/Fe ratios; d Pb/Tl ratios; and e208Pb/206Pb ratios. The rectangles outlined in green and blue in b–d mark the laminated sections (Sec-1 and Sec-2) selected for the extraction of the high-resolution time series. The close consistency between Mn/Fe and Pb/Tl ratios indicate minimal spatial shifts between individual scan sections.

Full size image

First, a scan of Mn/Fe ratios was conducted over a large area, which revealed two main growth patterns. The first is represented by a prominent laminated structure with alternating high and low Mn/Fe bands (Fig. 2a). The Mn/Fe ratios in this structure range from 1 to 2 and reflect a typical hydrogenetic growth pattern (Mn/Fe ratios of 0.5-2)28. The second pattern comprises discrete patches with Mn/Fe ratios below 0.5, which disrupt the laminations and suggest post-diagenetic process.

The age information based on the 230Thex method (detailed in Methods) revealed average growth rates of each line scan in Figs. 2b and 2c ranging from 1.0 to 1.5 mm/Myr. These are within the typical range for hydrogenetic growth29 and are consistent with the 1.7 mm/Myr growth rate derived from 10Be/9Be dating30. While ferromanganese crusts are generally assumed to grow at nearly constant rates over glacial-interglacial cycles, our 230Thex results reveal episodic growth patterns alternating between extremely slow growth (0.3-0.6 mm/Ma) and fast growth with the fastest rate exceeding 10 mm/Myr (Supplementary Fig. 6). These fast growth episodes, which all occurred during glacial periods at average ages of ~22 ka, ~61 ka, ~189 ka, ~277 ka, ~359 ka, and ~439 ka, are associated with reduced flow velocity31, nutrient enrichment32 and oxygen depletion33 in LCDW.

The rapid growth episodes fall into two distinct categories. The first and more common type retains the laminated Mn/Fe structure (Mn/Fe = 1-2) typical of hydrogenetic growth, and these intervals were used to extract the seawater Pb isotope records. While such fast growth rates exceed the common hydrogenetic range (<6 mm/Myr)29, our results show that they can occur under specific environmental conditions without compromising their hydrogenetic origin. The second type, for example, observed at ~61 ka, displays Mn/Fe ratios below 0.5 and Pb/Tl ratios exceeding 15 (Fig. 3d). This indicates either mixed hydrothermal/hydrogenetic growth, characterized by lower Tl concentrations28, and/or extensive post-depositional dissolution of Tl-bound Mn oxides under reductive glacial LCDW conditions33. Given these complexities, this section was excluded from the Pb isotope time series. Although fast-growing layers cover relatively large areas in the 2D scans, they represent only brief time intervals and therefore exert only limited influence on the overall Pb isotope record.

Subsequent laser scans for Tl and Pb isotopes were solely conducted in the section shown in Fig. 2b, in which the youngest surface is preserved. Age data and Pb isotope compositions from two separate laminated sections (sec-1 and sec-2 in Fig. 3) were processed separately, providing two independent Pb isotope time series. Consistent 206Pb/204Pb and 208Pb/206Pb records from these two areas (Supplementary Fig. 10) validate its reproducibility.

Controls of Pb isotope variations

The 208Pb/206Pb data show pronounced cyclic variations over the past 800 kyr, with generally lower 208Pb/206Pb during warm times than observed during cold periods (Fig. 4d). A key observation is that pre-MBE interglacials had systematically higher 208Pb/206Pb ratios than post-MBE interglacials. While variations in terrestrial Pb sources may have influenced these records18, the evidence presented below strongly supports the conclusion that water mass mixing34 has been the dominant control on Pb isotope signatures of the crust Haxby record.

Fig. 4: Compilation of paleoclimatic records during the past 800 ka.
figure 4

a Stack of sea level reconstructions84; b Ba/Fe ratio at ODP Site 109411; c Babio at site PS255142; d208Pb/206Pb time series in sec-1, note that this record is shown inverted. The black line represents 5-point runing mean values; e sea salt Na (ssNa) flux, a proxy for sea ice extent85; f atmospheric CO22; g Antarctic air temperature anomaly1. The white and yellow vertical shadings denote glacial and interglacial periods, respectively, based on the standard marine isotope stage (MIS) classification from the LR04 benthic δ18O stack86. Interglacial MISs are numbered at the top. The box with the grey dotted outline illustrates the time interval with lukewarm pre-MBE interglacial periods recorded in crust Haxby. MBE Mid-Brunhes Event.

Full size image

Modern seawater survey and model simulations show that the distribution of Pb isotope compositions in the SO reflects water mass advection and mixing35. Similarly, spatial patterns of Pb isotope compositions recovered from Pleistocene sediments and surfaces of ferromanganese crusts/nodules in the Pacific sector of the SO also clearly show a closer alignment with water mass distributions than with shifts in terrestrial Pb sourcing30. In support of this, the Pb isotope time series in crust Haxby after 220 ka closely covaried with the authigenic Pb isotope record of sediment cores SO213-59-2 and SO213-60-1 located in the sub-Antarctic Pacific further north (Supplementary Fig. 10), which has been interpreted in terms of variable water masses contributions to LCDW over the past 220 ka36. This is also in line with records of other water mass tracers, such as radiogenic Nd isotopes and δ13C, from the same cores36.

Given crust Haxby’s proximity to the Antarctic continent, it is important to assess whether variable detrital Pb inputs may have influenced its isotopic compositions. While benthic fluxes are known to overprint seawater Pb and Nd isotope signatures in authigenic phases of some Antarctic continental margin sediments15,37, such processes are minimized in crusts Haxby, which grew on hard substrates on a seamount, where sedimentation rates are known to be extremely low36. Moreover, it had been previously shown that at present boundary processes on the West Antarctic continental margin do not significantly change the Nd isotopic composition of bottom waters there38. Had detrital input from West Antarctica dominated the Pb isotope signals of crust Haxby, one would, for example, expect 208Pb/206Pb signatures of 2.044–2.056, which had been measured on near-coastal sediments from the Amundsen Sea embayment directly south of our sampling location39. These values are significantly lower than those recorded in crust Haxby (Fig. 4d), which displays more Pacific-sourced signals, consistent with upwelling of Pacific-derived deep waters in this region30. In contrast, sediment core SO213-60-1, located ~20° further north, actually recorded a more pronounced contribution of Antarctic Pb signatures, consistent with its greater depth and stronger influence by AABW. Furthermore, input from variable Antarctic sources has led to distinct Pb isotope distributions along the Antarctic margin. Moving from the western to the eastern Pacific sector of the SO and into Drake Passage, 208Pb/206Pb ratios systematically increase while 206Pb/204Pb ratios decrease in the surfaces of ferromanganese nodules/crusts (Figs. 1 and 5)30,40. In our record (Fig. 5), the interglacial Pb isotope signatures before and after the MBE show a clear shift in 208Pb/206Pb, while the 206Pb/204Pb ratios essentially remained unchanged. This pattern is inconsistent with changes in inputs from Antarctic Pb sources. The absent shift in 206Pb/204Pb further rules out a primary control by “incongruent weathering”, which would have produced and released a more radiogenic, higher 206Pb/204Pb, signal in the weathering solutions than that of the source rocks as a consequence of preferential chemical weathering or radiation damage of fresh labile accessory silicate minerals, during the early phases of interglacial ice sheet retreat18,41. Nearby sediment cores such as PS2551 also show no significant change in clay mineralogy across the MBE, despite strong glacial-interglacial variability, further suggesting that the provenance of terrestrial material remained the same42.

Fig. 5: 208Pb/206Pb vs 206Pb/204Pb ratios.
figure 5

Pb isotope data of crust Haxby are shown by circles. The purple and blue circles mark 5-point running mean Pb isotope values during post- and pre-MBE interglacials, respectively. The red circle indicates the bulk composition analysed by the conventional approach30. Coloured squares denote Pb isotope data from discrete surface scrapings of ferromanganese nodules40,83 shown in Fig. 1. These Pb isotope data from Southern Ocean and South Pacific nodules are used to reflect pre-industrial Pb isotope signals of southern- and north Pacific-sourced deep waters, respectively. MBE Mid-Brunhes Event.

Full size image

By contrast, Antarctic-sourced deep waters, such as AABW, exhibit a much lower 208Pb/206Pb signal than Pacific-sourced deep water, e.g., PDW, while these two water masses show a similar range of 206Pb/204Pb ratios (Fig. 5). Therefore, the variability in the 208Pb/206Pb record of crust Haxby is best explained by shifts in the relative contributions of AABW and PDW to LCDW. We conclude that the observed changes in 208Pb/206Pb are primarily driven by water mass mixing between these deep water sources.

Implications for SO circulation changes

Our record from crust Haxby documents higher glacial than interglacial 208Pb/206Pb ratios of LCDW. This suggests increased proportions of PDW relative to AABW in LCDW during glacials. More radiogenic Nd isotope signatures43,44,45 and increased dissolved inorganic carbon (DIC) concentrations46,47 in LCDW during glacial periods have provided evidence that PDW replaced shoaled NADW as a dominant component of LCDW6,47. Moreover, a larger gradient in Nd isotopic compositions between AABW and LCDW during the last glacial period43 suggests reduced vertical mixing, likely due to a more stratified deep SO and weakened upwelling. This change would have led to less AABW being incorporated into LCDW. In addition, both modern data and paleorecords show that upper water column Pb isotope signals can be efficiently transferred to the deep ocean via sinking particles without changing water mass mixing35,48,49,50. However, biological productivity, represented by the Babio record from nearby site PS2551 (Fig. 4c)42 and Babio, biogenic opal and diatom concentrations at site PS58/25451,52, was lower during glacials than during interglacials, suggesting that enhanced particle sinking was unlikely responsible for the higher 208Pb/206Pb ratios of glacial LCDW. Therefore, we attribute the elevated 208Pb/206Pb ratios in glacial LCDW to increased PDW subduction and a reduced upward supply of AABW.

Likewise, the higher 208Pb/206Pb ratios in LCDW during the lukewarm interglacials than during more recent interglacials suggest reduced mixing of AABW relative to PDW (Fig. 6). Unlike the shoaled and likely more sluggish Atlantic Meridional Overturning Circulation (AMOC) during recent glacial periods53,54, evidence from Nd isotopes10 and sediment grain size55 records in the Atlantic Ocean suggests that AMOC geometry and strength were similar during lukewarm and recent interglacials. Compilation of global benthic foraminifera δ13C data covering the last 800 ka shows little change, by inference, AMOC strength between pre- and post-MBE interglacials56, which is consistent with a benthic foraminifera δ13C record from one of the Marie Byrd Seamounts57. Furthermore, comparable Nd isotope compositions of LCDW during interglacials before and after the MBE indicate little increase in the mixing of PDW into LCDW during the lukewarm interglacials58,59,60,61. Taken together, the elevated 208Pb/206Pb ratios in LCDW primarily resulted from reduced admixture of AABW into LCDW, indicative of a more stratified deep SO, during lukewarm interglacials.

Fig. 6: Schematic illustration of Southern Ocean circulation changes before and after the Mid-Brunhes Event (MBE).
figure 6

During pre-MBE interglacials, the deep Southern Ocean was more strongly stratified than in more recent interglacials, with reduced mixing of Antarctic Bottom Water (AABW), characterized by low 208Pb/206Pb, into the Lower Circumpolar Deep Water (LCDW). This enhanced stratification, together with weaker upwelling and expanded sea-ice cover around Antarctica, would have limited CO2 exchange between the deep ocean and the atmosphere. PDW Pacific Deep Water.

Full size image

Several processes may have contributed to reduced SO deep water mixing during the lukewarm interglacials11,12,13. One potential mechanism is the reduced dynamic range of SO overturning during these periods, as suggested by diminished productivity inferred from Ba/Fe ratios (Fig. 4b) from the Atlantic sector of the SO11. Similar reductions in productivity, indicated by Babio data, have also been observed at site PS2551 (Fig. 4c)42. However, productivity records across the SO as a whole and even within its individual ocean basins are not consistent. While some studies report elevated productivity during specific interglacials before the MBE42,51, another study suggests little decline in productivity during lukewarm interglacials57. These discrepancies may reflect spatial variability in sea-ice cover or the positioning of SO fronts, yet the overall trend suggests that productivity during lukewarm interglacials was generally lower than during more recent interglacials. Another potentially important process is an increase in AABW density. Enhanced sea ice formation (Fig. 4e) and lower Antarctic air temperatures (Fig. 4g) during lukewarm interglacials could have facilitated AABW production and the formation of denser AABW62,63. This hypothesis is supported by the observed steeper latitudinal δ13C gradient between northern- and southern-sourced waters in the equatorial Atlantic Ocean8, as well as model simulations12. The resulting increased density contrast within the deep SO would have strengthened ocean stratification between mid- and deep-waters, limiting mixing between AABW and LCDW.

The stronger stratification of the deep SO during the lukewarm interglacials coupled with reduced upwelling and extended sea ice cover around Antarctica, would have diminished the exchange of carbon between the deep ocean and the atmosphere (Fig. 6). A box model study64 suggests that the combination of these SO processes (i.e., increased sea ice coverage and enhanced deep ocean stratification) under presence of a strong NADW export10 lowered atmospheric CO2 levels by as much as 36 ppm. This finding helps to explain the observed CO2 concentration difference of 30-40 ppm between lukewarm and post-MBE interglacials2,65. If deep ocean stratification diminishes66 and sea ice production decreases67 in response to anthropogenic warming in the future, our results may suggest a long-term future weakening of the SO’s capacity to sequester carbon in the ocean interior.

Methods

Sample preparation

The surface of crust Haxby was initially coated with epoxy resin to preserve its structure before being cut into slabs suitable for laser ablation analysis. After each cut, the sample was re-embedded in epoxy resin to ensure continued protection of its natural structure. Ultimately, a slab approximately 5 cm long, 2 cm wide, and 0.3 cm thick was sectioned using a microtome saw. The exposed surface for LA-MC-ICP-MS analysis was then polished using Bühler “Micropolish” (Bühler GmbH, Duesseldorf, Germany). Following polishing, the sample surface was thoroughly cleaned with Milli-Q water to remove any potential contaminants.

Laser ablation analyses

Two-dimensional (2D) elemental and isotopic ratio maps were generated from selected areas of the sample. Prior to laser ablation, backscattered electron images were acquired using an electron microprobe. A target area of 2000 μm x 2775 μm (width x length) for LA-MC-ICP-MS analyses is outlined by a red rectangle in Supplementary Fig. 1. These analyses were performed using a 193 nm Analyte Excite Excimer Laser Ablation System connected to a Thermo Scientific Neptune Plus MC-ICP-MS at GEOMAR Kiel. Three distinct measurements were conducted in sequence, each with specific instrumental settings:

  1. 1).

    Mn and Fe isotope analyses (Supplementary Fig. 2): Semi-quantitative overview images (10 µm spot size and 10 × 8 µm step size) were obtained to assess the overall Mn and Fe concentration variability and Mn/Fe ratios in the sample. This step was used to visualize the sample’s growth structure and guide the selection of areas for more detailed analyses.

  2. 2).

    Th and U isotope analyses (Supplementary Figs. 3 and 4): Using a 25 µm spot size and 25×20 µm step size, measurements were taken to create age images in two selected areas: area-I (0-800 µm width, 0-1900 µm length, Fig. 2b) and area-II (1000-2000 µm width, 0-900 µm length, Fig. 2c).

  3. 3).

    Pb and Tl isotope analyses (Supplementary Fig. 5): Using a 25 µm spot size and 25 × 20 µm step size, measurements were conducted to map Pb and Tl concentrations and generate 206Pb/204Pb and 208Pb/206Pb ratio images.

Previous work68 has shown that plasma conditions strongly influence elemental fractionation and ionization efficiency during laser ablation. The Normalized Argon Index (NAI), introduced by Ref. 68, serves as a proxy for plasma temperature and ionization efficiency. To ensure consistent and optimal plasma conditions across different sessions, we carefully adjusted laser parameters, such as spot size, fluence, and scan speed, and MC-ICP-MS settings (see Supplementary Table 1) to achieve high and stable NAI values (≥1) for all measurement sessions. This ensured minimal mass-dependent fractionation and enhanced ionization efficiency for the elements analyzed.

Mn and Fe images

As Mn and Fe are the primary constituents of ferromanganese crusts, Mn and Fe measurements provide direct insights into the sample’s growth patterns. The purpose of these overview scans was to evaluate the internal elemental variability and distribution of Mn and Fe within the selected area. No external standards were used for calibration during these scans. Signals for 55Mn, 56Fe, and 57Fe were simultaneously collected on Faraday cups (Supplementary Table 2).

While 56Fe is the most abundant stable isotope of iron (91.75%), its signal overlaps with the strong signal from the 40Ar16O cluster in the gas blank. In contrast, 57Fe, though free from major argon-derived interference, has a lower signal intensity due to its low abundance (2.11%). The interferences for all isotopes were corrected based on gas blank data collected prior to each line ablation. Despite these different interferences in the gas blank during analysis, the consistent intensity patterns between 56Fe and 57Fe (Supplementary Fig. 2a and S2b) suggest both provide reliable information towards the internal Fe distribution. On the other hand, Mn measurements are more straightforward since 55Mn, the only stable isotope of Mn, is free from major argon interferences. We normalized the 55Mn intensity to the combined 56Fe and 57Fe ion intensities to indicate the Mn/Fe variability (Supplementary Fig. 2d) and derive the Mn/Fe ratio (Fig. 2a) based on their isotope abundances. No smoothing of data points was applied in the Mn/Fe image.

Th and U images

Following the Mn and Fe overview scans, Th and U isotope images were generated. For correction, line measurements of the NIST-SRM610, USGS NOD-A-1, and USGS NOD-P-1 standards were taken after each set of sample lines. The USGS NOD-A-1 and USGS NOD-P-1 standards, which consist of ferromanganese nodule powders, were pressed into pellets without chemical binders.

Two scan profiles were carried out to investigate the use of the 230Thex/232Th and 234Uex/238U as chronology tools. The first scan collected 230Th, 232Th and 238U signals with 232Th and 238U on Faraday cups and 230Th on an ion counter. The second scan measured 232Th and 238U on Faraday cups and 234U on the ion counter (Supplementary Table 2). Two areas (I in Supplementary Fig. 3 and II in Supplementary Fig. 4), covering nearly the entire Mn/Fe image surface, were scanned twice to capture paired Th and U isotope measurements.

The raw data were background-corrected based on gas blank measurements collected before each line ablation. To enhance image quality, data smoothing was applied once. For each individual pixel, the value was averaged with a weighted combination of its neighbouring data points: the once-weighted values from the four diagonal neighbours, the three-times weighted values from the two nearest horizontal and two nearest vertical neighbours, and the seven-times weighted value from the central pixel.

As shown Supplementary Figs. 3a and 4a, 230Th signals decrease with increasing depth in the sample profiles and eventually reach secular equilibrium. However, there is no decreasing trend in 234U signals with increasing depth in the sample profiles. Early studies showed that 234Uex/238U profiles generally give higher growth rates than those derived from 230Thex/232Th data profiles69,70 because of the higher diffusivity of U in the ferromanganese crusts71,72. Our results also suggest that ferromanganese crusts are not a closed system for U, ruling out 234Uex/238U as a reliable chronology tool for dating this crust. In addition, 230Th concentrations in the topmost surface part in area-I are systematically higher than in area II (Supplementary Figs. 3a and 4a), indicating the topmost surface materials in area-II have been lost during recovery of the sample. The following laser ablation analyses thus only focused on the area-I.

Pb and Tl images

Prior to measurements of Pb and Tl isotopes, the sample surface was pre-ablated by triple laser pulses with 3 J cm−2 laser power density in order to eliminate potential surface contamination. Image data for Pb and Tl have been run with the line measurements of NIST-SRM610, USGS NOD-A-1 and USGS NOD-P-1 standards before and after every set of sample lines to monitor drift control and to obtain a reliable normalization of instrumental mass fractionation. The signals of 202Hg, 203Tl, 204Pb, 205Tl, 206Pb, 207Pb, 208Pb were collected by Faraday cups (Supplementary Table 2). The 202Hg signal was specifically monitored to correct for isobaric interference of 204Hg on 204Pb. Because the 202Hg/204Hg ratio can vary between measurement sessions, we applied an optimized 202Hg/204Hg value based on our recent calibration method30, rather than using a fixed literature value. After the blank correction for the gas blank data collected before each line ablation, the signal intensity images of 208Pb and 205Tl are shown in Supplementary Fig. 5a and 5b. Due to the similarity of Pb and Tl isotope variations, we only show the intensity images of the most abundant isotopes, 208Pb and 205Tl. The intensity ratio of 208Pb/205Tl is used to indicate Pb/Tl variations, and the Pb/Tl ratios are derived based on their isotope abundances, as shown in Fig. 2. To improve the image quality via data smoothing, individual pixel intensities were averaged twice with neighbouring data points for both Pb and Tl.

Our method applied for the normalisation of instrumental mass fractionation of Pb isotope is based on the assumption that the fractionation factors during measurements of samples and the selected standards are identical when the plasma conditions of MC-ICP-MS are consistent for all measurements68. The plasma states of analyses for samples and standards were tuned to NAI values of about 168. These NIST-SRM610, USGS NOD-A-1 and USGS NOD-P-1 standards cover a wide 206Pb/204Pb and 208Pb/206Pb range with different matrix compositions. The combination of these three different standards thus provides the best approach to maximize accuracy and precision of the measurements covering similar Pb isotope signal intensities and considering matrix composition variations. The Pb isotope fractionation factor was obtained from the optimal values of these three standards using the linear regression normalization method73. The details of this method are described in a recent paper30. The final images for 208Pb/206Pb and 208Pb/206Pb are shown in Supplementary Fig. 5c and Fig. 3e. Six repeat analyses of these the reference materials (three analyses before and three analyses after analytical sessions) yielded the following ratios and uncertainties: 206Pb/204Pb = 17.052 ± 0.003 and 208Pb/206Pb = 2.1695 ± 0.0005 for NIST SRM 610, 206Pb/204Pb = 18.935 ± 0.013 and 208Pb/206Pb = 2.0555 ± 0.0023 for USGS NOD-A-1, and 206Pb/204Pb = 18.689 ± 0.019 and 208Pb/206Pb = 2.0681 ± 0.0010 for USGS NOD-P-1 (all uncertainties are provided as 2 SD (n = 2640)). The good agreement between the double spike solution data74 and the LA-MC-ICPMS data show that our measurements are reliable. NIST SRM 610, which has Pb and Hg concentrations similar to those of the Haxby crust, was used as the principal standard to assess measurement precision30.

Age model

The U/Th images reflect the temporally variable Th concentrations in the crust. The 230Th data are normalized to 232Th to obtain the variability of the 230Th flux. As shown in Supplementary Figs. 3 and 4, 234U and 238U covary, and 234U/238U is well diffused in the scanned area. The excess 230Th (230Thex), unsupported by 234U decay, of each data point is thus calculated as:

$$\left[{\scriptstyle{{230}}\atop}{{{\rm{Th}}}_{{{\rm{ex}}}}}/{\scriptstyle{{232}}\atop}{{{\rm{Th}}}}\right]=\left[{\scriptstyle{{230}}\atop}{{{\rm{Th}}}}/{\scriptstyle{{232}}\atop}{{{\rm{Th}}}}\right]-{{{\rm{\beta }}}}\times \left[{\scriptstyle{{238}}\atop}{{{\rm{U}}}}/{\scriptstyle{{232}}\atop}{{{\rm{Th}}}}\right]$$
(1)

The β is the combined factor for the yield of ion counter, instrumental Th isotope fractionation, 238U/232Th fractionation and 234U/238U ratio. The [230Thex/232Th] value is expected to reach secular equilibrium after about seven half-lives of 230Th. A value of 1500 for β was derived as it leads to [230Thex/232Th] values in the old layers approaching zero (i.e., secular equilibrium of 230Th/234U). Because no trend of 234U concentration decrease compared to 238U concentrations was observed and the influence of 234U decay is overall small, its variability as a function of time has not been considered.

The age of each data point can be calculated:

$${{{\rm{Age}}}}={{{\mathrm{ln}}}}\left({\left[{\scriptstyle{{230}}\atop}{{{{{\rm{Th}}}}}_{{{{\rm{ex}}}}}}/{\scriptstyle{{232}}\atop}{{{\rm{Th}}}}\right]}_{0}/\left[{\scriptstyle{{230}}\atop}{{{{\rm{Th}}}}_{{{\rm{ex}}}}}/{\scriptstyle{{232}}\atop}{{{\rm{Th}}}}\right]\right)/{{{\rm{In}}}}\left(2\right)\times {{{{\rm{T}}}}}_{1/2}$$
(2)

The half-life (T1/2) of 230Th is 75690 yrs. The initial Th isotope ratio [230Thex/232Th]0 of 35000 (counts/V) is obtained from the highest measured values in the scanned surface area. The age distribution images are shown in Fig. 2 of the main manuscript, and the corresponding variations in incremental growth rate are presented in Supplementary Fig. 6.

Our results show that the initial 230Thex/232Th ratio was slightly variable. For instance, the outer, stratigraphically “younger” layer on the age plateaus of 189 ka and 277 ka shows slightly older “apparent” ages (Supplementary Fig. 7). This seemingly paradoxical observation, in our opinion, has to be taken as an indication, that certain assumptions in the 230Thex dating method have come to a limit. One of those assumptions is a constant initial 230Thex/232Th ratio, applicable for the entire crust. However, both the 230Thex and the 232Th concentration can change through time (e.g., due to changing lateral transport by different water masses). This finding is supported by the variable initial 230Thex/232Th ratios found in deep-sea corals from nearby locations75. The changes of initial 230Thex/232Th ratios are likely related to water mass changes, because the declined initial 230Thex/232Th ratios (indicated by reversed ages) generally coincide with low 208Pb/206Pb ratios (Supplementary Fig. 7b). The influence of the initial 230Thex/232Th ratio variability on calculated ages is less prominent during the slow-growing episodes, so only the ages during these periods are used to date the crust (as shown in Supplementary Fig. 7). Due to the significant age uncertainty caused by variable initial 230Thex/232Th ratios, our age model is not reliable for resolving high temporal resolution variations, such as sub-orbital or millennial scale changes. As a result, we focused our analysis on changes occurring on glacial-interglacial timescales.

The dating method based on the 230Thex approach enables us to construct an age model extending back to approximately 450 ka (5-6 half-lives of 230Th). For deeper layers, ages are inferred from Pb/Tl variations. In hydrogenetic ferromanganese crusts, Tl is mainly bonded to vernadite (δ-MnO2), the most abundant mineral phase, whereas Pb has a high affinity with the amorphous FeO-OH phase76. Therefore, the Pb/Tl ratio shows a negative correlation with Mn/Fe, which is supported by the high consistency between Pb/Tl and Mn/Fe ratios in crust Haxby (Fig. 3 and Supplementary Fig. 7a). The data after 450 ka clearly show systematic variations between high Pb/Tl (low Mn/Fe) during glacial times and low Pb/Tl (high Mn/Fe) during interglacial times (Supplementary Figs. 8 and 9). These variations may be related to changes in oxygen concentration in LCDW33,77,78,79. Consequently, we derived the ages between ~800 and 450 ka by tuning the Pb/Tl ratios to the Antarctic temperature record (Supplementary Figs. 8 and 9). Tie points were primarily placed at intervals of declining Pb/Tl values, which correspond to glacial periods in the Antarctic temperature record.

Data reproducibility for Pb isotope time series

The time series of Mn/Fe, Pb/Tl, 206Pb/204Pb and 208Pb/206Pb from two different laminated sections (sec-1 and sec-2 in Fig. 3) are presented in Supplementary Fig. 7 to assess the data reproducibility. Due to the influence of epoxy interference in the very surface part (topmost 25 µm as shown in the Supplementary Fig. 7b), the 208Pb/206Pb in this part cannot be well normalised and is thus excluded in the time series record. The Pb isotope data from porous areas (intensities of 208Pb < 0.4 V) are also excluded.

Manganese concentrations in ferromanganese crusts are sensitive to changes in redox conditions and serve as a redox proxy80. Although a general qualitative relationship exists between redox elements in crust Haxby and deep-ocean redox conditions, the interpretation of specific elemental ratios, such as Mn/Fe needs to be considered with caution. As illustrated in Supplementary Fig. 10b, the Mn/Fe time series records are inconsistent both along different lines in the same section and between different sections.

In contrast, the time series of 206Pb/204Pb and 208Pb/206Pb are consistent along the different analysis sections (Supplementary Fig. 10d and 10e). The overall 206Pb/204Pb and 208Pb/206Pb data extracted via laser ablation align well with the bulk Pb isotope compositions measured by conventional methods30. Moreover, the laser ablation 206Pb/204Pb and 208Pb/206Pb data from ~220 to 70 ka covary with authigenic Pb isotope data from the sediment cores recovered at sites SO213-59-2 and SO213-60-1 further north in the sub-Antarctic Pacific sector36, further validating the reliability of the Pb isotope time series obtained by the LA-MC-ICPMS technique.