(Data S1) Palynological analysis of IODP and ODP sediment cores around the Antarctic margin

The circum-Antarctic Southern Ocean is an important region for global marine food webs and carbon cycling because of sea-ice formation and its unique plankton ecosystem. However, the mechanisms underlying the installation of this distinct ecosystem and the geological timing of its development remain unknown. Here, we show, on the basis of fossil marine dinoflagellate cyst records, that a major restructuring of the Southern Ocean plankton ecosystem occurred abruptly and concomitant with the first major Antarctic glaciation in the earliest Oligocene (~33.6 million years ago). This turnover marks a regime shift in zooplankton-phytoplankton interactions and community structure, which indicates the appearance of eutrophic and seasonally productive environments on the Antarctic margin. We conclude that earliest Oligocene cooling, ice-sheet expansion, and subsequent sea-ice formation were important drivers of biotic evolution in the Southern Ocean.

Isotope composition of dolomite, calcite and pore-waters of ODP Leg 210 samples

Early diagenetic dolomite beds were sampled during the Ocean Drilling Programme (ODP) Leg 201 at four reoccupied ODP Leg 112 sites on the Peru continental margin (Sites 1227/684, 1228/680, 1229/681 and 1230/685) and analysed for petrography, mineralogy, d13C, d18O and 87Sr/86Sr values. The results are compared with the chemistry, and d13C and 87Sr/86Sr values of the associated porewater. Petrographic relationships indicate that dolomite forms as a primary precipitate in porous diatom ooze and siliciclastic sediment and is not replacing the small amounts of precursor carbonate. Dolomite precipitation often pre-dates the formation of framboidal pyrite. Most dolomite layers show 87Sr/86Sr-ratios similar to the composition of Quaternary seawater and do not indicate a contribution from the hypersaline brine, which is present at a greater burial depth. Also, the d13C values of the dolomite are not in equilibrium with the d13C values of the dissolved inorganic carbon in the associated modern porewater. Both petrography and 87Sr/86Sr ratios suggest a shallow depth of dolomite formation in the uppermost sediment (<30 m below the seafloor). A significant depletion in the dissolved Mg and Ca in the porewater constrains the present site of dolomite precipitation, which co-occurs with a sharp increase in alkalinity and microbial cell concentration at the sulphate-methane interface. It has been hypothesized that microbial 'hot-spots', such as the sulphate-methane interface, may act as focused sites of dolomite precipitation. Varying d13C values from -15 per mil to +15 per mil for the dolomite are consistent with precipitation at a dynamic sulphate-methane interface, where d13C of the dissolved inorganic carbon would likewise be variable. A dynamic deep biosphere with upward and downward migration of the sulphate-methane interface can be simulated using a simple numerical diffusion model for sulphate concentration in a sedimentary sequence with variable input of organic matter. Thus, the study of dolomite layers in ancient organic carbon-rich sedimentary sequences can provide a useful window into the palaeo-dynamics of the deep biosphere.

Nd isotopes of ODP Sites 146-893 and 167-1017 samples

Changes in the source of intermediate waters to the southern California margin may have caused variations in seafloor oxygen levels on stadial-interstadial time scales. We test this hypothesis using the Nd isotopic composition of benthic foraminifera and fossil fish debris from ODP Sites 893 and 1017 to track the composition of intermediate waters across interstadials 8-14 (~37-52 ka) during Marine Isotope Stage 3. The epsilon-Nd values of waters bathing the seafloor at Site 893 were typically ~-9 and those bathing Site 1017 were ~-7, both of which are significantly less radiogenic than waters that had originated in either the North Pacific or Southern Ocean (by the time such waters reached the southern California margin). Detrital silicate epsilon-Nd values of nearly -12 suggest that this offset toward lower epsilon-Nd values was likely caused by boundary scavenging that partially overprinted the water mass composition with local/regional fluvial Nd inputs. In spite of the evidence for boundary scavenging, the lack of systematic seawater Nd isotope changes on a stadial-interstadial basis suggests that the provenance of the intermediate waters did not change, and that the waters were derived from the Southern Ocean. Instead, changes in local/regional sea surface productivity may have caused the recorded changes in seafloor oxygenation.

Geochemistry and Mg-isotopes of carbonate sediments and porefluids of ODP Site 130-807

Deep-sea pore fluids are potential archives of ancient seawater chemistry. However, the primary signal recorded in pore fluids is often overprinted by diagenetic processes. Recent studies have suggested that depth profiles of Mg concentration in deep-sea carbonate pore fluids are best explained by a rapid rise in seawater Mg over the last 10-20 Myr. To explore this possibility we measured the Mg isotopic composition of pore fluids and carbonate sediments from Ocean Drilling Program (ODP) site 807. Whereas the concentration of Mg in the pore fluid declines with depth, the isotopic composition of Mg in the pore fluid increases from -0.78 per mil near the sediment-water interface to -0.15 per mil at 778 mbsf. The Mg isotopic composition of the sediment, with few important exceptions, does not change with depth and has an average d26Mg value of -4.72 per mil. We reproduce the observed changes in sediment and pore-fluid Mg isotope values using a numerical model that incorporates Mg, Ca and Sr cycling and satisfies existing pore-fluid Ca isotope and Sr data. Our model shows that the observed trends in magnesium concentrations and isotopes are best explained as a combination of two processes: a secular rise in the seawater Mg over the Neogene and the recrystallization of low-Mg biogenic carbonate to a higher-Mg diagenetic calcite. These results indicate that burial recrystallization will add Mg to pelagic carbonate sediments, leading to an overestimation of paleo-temperatures from measured Mg/Ca ratios. The Mg isotopic composition of foraminiferal calcite appears to be only slightly altered by recrystallization making it possible to reconstruct the Mg isotopic composition of seawater through time.