Samples were prepared by crushing ∼0.2 g of bulk sediment in an agate mortar. The crushed samples were further disaggregated using 10% hydrogen peroxide to remove organic matter. After 48 h of reaction, the samples were passed through 2 mm sieves and dried in an oven. After drying, sodium pyrophosphate was added to prevent the particles from flocculating. The Microtrac Bluewave equipment was used to measure the particles by laser diffraction in water, with their size being calculated according to the modified Mie theory for non-spherical particles (Chýlek et al., 1976). To establish the basic statistical parameters of particle size, as well as particle size distribution, the GRADISTAT v. 4.0 program, developed by Blott and Pye (2001), was used to establish the basic statistical parameters of particle size, as well as particle size distribution.

Rock magnetic analyses were conducted on 21 selected samples distributed along the studied interval to estimate downcore variations in the composition, concentration, and grain size of magnetic minerals. These analyses included volume normalized low-field magnetic susceptibility (κ), hysteresis loops, with maximum fields of 0.5 T, isothermal remanent magnetization (IRM) imparted in progressively (100 acquisition steps) to saturation (SIRM), or to the maximum attainable direct field of 1.8 T. IRM curves were subdivided into discrete coercivity components, as in Robertson and France (1994), with cumulative log-Gaussian (CLG) functions, using the spreadsheet developed by Kruiver et al. (2001) and Heslop et al. (2002).

In addition, we investigated the temperature dependence of magnetic susceptibility using a MFK1-FA equipped with CS4 furnace (AGICO) Kappabridge, to discriminate ferromagnetic mineralogy. Samples were measured up to a maximum temperature of 700°C in an argon atmosphere, in order to minimize thermochemical alteration.

In order to remove secondary NRM components, an Alternating Field (AF) demagnetization technique was employed, with stepwise demagnetization at peak AF fields of 2.5, 5, 10, 15, 20, 30, 40, 50, 60, 70, 80, 90 and 100 mT. Vector component diagrams (Zijderveld, 1967) were used to display stepwise demagnetization data. Characteristic remanent magnetization (ChRM) directions were determined using principal component analysis (PCA) with linear best fits calculated from three or more demagnetization steps using the PuffinPlot paleomagnetic analysis application (Lurcock and Wilson, 2012; Lurcock and Florindo, 2019).

The lithology of the studied interval is predominantly clay. The average grain size is 11.16 phi (0.435 μm), while the mode is 11.17 phi (0.432 μm), with a standard deviation of 0.33. The coarsest average grain size (9.87 phi; 1.065 μm) is found at 158.09 mbsf. The finest average grain size (12.41 phi; 0.181 μm) was observed at 162.11 mbsf. Several samples have multiple peaks in the grain size distribution (Figure 2).

A drop-stone, 2 cm in diameter, was found at 319.74 mbsf, and additional ice rafted debris (IRD) particles greater than 5 mm in diameter were identified at 258.5 and 247.5 mbsf. Frakes (1975) also reported the presence of a coarse fraction in Cores 19 and 18 (180.5 and 161.5 mbsf, respectively) interpreted to reflect ice rafting events.

We observed a lower proportion of clay and a higher proportion of sand in the interval between 180 and 130 mbsf (Cores 19–14, respectively), where the variability of the average grain size is greater and ranges between 10 and 12 phi (0.98–0.24 μm). Between 180 and 160 mbsf (Cores 19–18, respectively) the average grain size is finer between 11 and 12 phi (0.48–0.24 μm) and coarsens between 160 and 130 mbsf (Cores 17–14) to values ranging between 9.7 and 11.7 phi (1.20–0.30 μm). The average grain size is finer in the lower interval of the core (between ∼300 and ∼180 mbsf; Cores 34–20) with values ranging between 11 and 11.5 phi (0.48–0.34 μm), with a slight coarsening in average grain size in the interval ∼320–∼304 mbsf (Cores 34–33). Abrupt changes in grain size at ∼262, ∼240 and ∼224 mbsf suggest the possibility of minor unconformities in the sedimentary succession.

Magnetic susceptibility (κ) values range widely between 1.8 × 10⁻⁶ and 92.9 × 10⁻⁶ SI with a mean value of 10.9 × 10⁻⁶ SI (Figure 2). Below the unconformity at ∼180 mbsf, the highest κ values occur in samples between 280 and 210 mbsf and are associated with IRD intervals with coarser sediments and larger clasts. Above 180 mbsf, κ gradually increases reaching peak values at 140 mbsf.

IRM acquisition curves obtained for the samples were separated into four coercivity components (Figure 3). Both samples (at 218.92 and 285.41 mbsf) show a component with very low coercivity (mean 8 mT) and dispersion parameter of 0.30, which corresponds to coarse detrital magnetite with different coercivity or activation of superparamagnetic (SP) particles dispersed in the material (Heslop et al., 2004; Abrajevitch and Kodama, 2011; Savian et al., 2014; Savian et al., 2016). The two middle-coercivity components show values consistent with single domain (SD) magnetite (mean coercivity of 23 mT, dispersion parameter 0.23 in samples 218.92 mbsf and 45 mT, dispersion parameter 0.19 in sample 285.41 mbsf, respectively). Those components are assumed to represent low-coercivity SD magnetite and high-coercivity SD magnetite, probably of biogenic origin (Egli, 2004; Jovane et al., 2012; Yamazaki , 2012; Heslop et al., 2004; Rodelli et al., 2019). The component characterized by mean coercivities of 100 and 280 mT (dispersion parameter = 0.15 and 20, respectively), are interpreted as hematite (Kruiver et al., 2001; Abrajevitch et al., 2009; Rodelli et al., 2019). The overall contribution to the IRM in sample 218.92 mbsf from the SD magnetite is 72, and 9% is linked instead to detrital magnetite. Hematite only represents 19% of the total contribution. IRM acquisition curves for a sample at 285.41 mbsf, typical of samples with lower κ, was fitted with only four components (Figure 3E): the low coercivity SD magnetite component is missing, and the hard SD magnetite is less abundant, corresponding to only 64% of the magnetic signal, and the detrital magnetite is more abundant (18%). Hematite is slightly less abundant (18%).

A κ-T curve for sample at 218.92 mbsf (Figure 3C) shows increasing values up to about 330°C, which can be interpreted as maghemite (Rodelli et al., 2018; Rodelli et al., 2019), possibly related to partial oxidation of magnetite. This is followed by a peak at about 520°C and a major inflection at 579°C, calculated with the two-tangent method of Grommé et al. (1969), and consistent with chemically pure magnetite (Thompson and Oldfield, 1986). A κ-T curve for sample at 285.41 mbsf (Figure 3F) has a sharp drop at 565°C consistent with magnetite probably mixed with titano-magnetite (Ozima and Larson, 1970). The cooling curve for sample at 285.41 mbsf has higher overall κ than the heating curve, which indicates production of new magnetic phases during heating while sample at 218.92 mbsf maintain similar values.

NRM intensities (Figure 2A) for the DSDP 274 succession range widely between 9.36 × 10⁻² and 6.58 × 10⁻⁵ A/m about a mean of 4.72 × 10⁻³ A/m. At the base of the sequence (320–310 mbsf), values are higher than for the rest of the sequence below the unconformity, then increase again above 180 mbsf. Glaciomarine sediments at the base are coarser and contain larger clasts that can increase intensities and lead to non-reliable paleomagnetic behavior (e.g., unstable remanence, low inclination or multi-component magnetizations). For these reasons, samples at the base were excluded from magnetostratigraphic interpretations.

Some samples exhibit a low coercivity horizontal component of remanence (Figure 4) interpreted to be a drilling induced radial overprint related to strong magnetic fields in the cutting-shoe and/or core barrel (e.g., Acton et al., 2002; Florindo et al., 2003). In most cases, this is removed by AF demagnetization in a peak field of 20 mT, but in some cases the horizontal overprint is harder and persists to higher coercivities (>80 mT; Figure 4D) and prevents the isolation of a ChRM. In a few samples, characterized by the lowest NRM intensities, this overprint could not be removed by AF demagnetization. We cannot rule out that drying may have further hardened the radial overprint.

The ChRM is identified in most samples, regardless of polarity, between AFs of 20 and 70 mT (Figure 4). Approximately half of the ChRMs are directed to the origin of vector component plots (Figures 4A,C) and half follow a trajectory toward a high coercivity component not removed by peak AFs of 100 mT (Figures 4F,G). The high coercivity component of samples containing a small amount of hematite is assumed to be a secondary chemical remanent magnetism and unanchored ChRM components are reported.

Of the total of 287 samples, 65.51% (188) were used to determine the magnetic polarity zonation, with discrete magnetozones defined as an interval where at least two consecutive samples show polarities different from adjacent intervals. The other 99 samples were discarded according to the following criteria. Samples were discarded from the final analysis if they met the following conditions: 1) samples with a radial overprint (Figure 4D); 2) single samples showing a ChRM with an inclination of opposite polarity to that of the previous and following samples; 3) samples with anomalous low inclination (shallower than ±25°) were also not considered (Figure 4G); 4) samples containing IRD which do not demagnetize due to the presence of high-coercivity minerals and 5) samples with overprints (drilling overprint (Fuller et al., 1998) or core-splitting overprint (Wilson et al., 2000; Florindo et al., 2001) that could not be removed by low field AF demagnetization.

Mean ChRM inclinations were computed separately for normal and reverse polarity populations using a maximum likelihood method (Arason and Levi, 2010) as implemented in the PuffinPlot application (Lurcock and Florindo, 2019). The normal polarity intervals have a mean inclination of −55.01° (N = 56, α₉₅ = 6.67°, θ₆₃ = 27.03°), the reverse polarity intervals have a mean of 68.3° (N = 132, α95 = 4.45°, θ₆₃ = 27.8°) (Figure 5). Some samples have lower inclination than the expected Geocentric Axial Dipole (GAD) inclination (±64.5° at 20 Ma and ±63.9° at 40 Ma; Torsvik et al., 2012) for this period and for this site latitude. A post depositional tilt of ∼5° the strata is implied by the difference in normal and reversed polarity inclinations.

We recognize eight normal and eight reversed magnetozones (Figure 6). The base of magnetozone N1 is not identified due to an interval of undefined polarity at the base of the core with most samples characterized by radial overprints. The upper part of magnetozone N1 is well defined by four normal polarity samples. Samples 313.21 (normal polarity) and 311.4 (reversed polarity) mbsf define the boundary of magnetozone N1 with R1. Between 312.3 and 302.38 mbsf magnetozone R1 is clearly identified, although it includes several samples with noisy data and others that do not demagnetize. Magnetozone N2 is well defined by three normal polarity samples between 302.82 (top of magnetozone R1) and 297.72 mbsf, which is the base of magnetozone R2. Between 298.4 and 280.09 mbsf, magnetozone R2 includes a long series of samples with reversed polarity with only occasional samples that do not demagnetize. The boundary between magnetozones N3 and R3 lies within an 8-m-interval where polarity could not be determined due to samples in which the radial overprints do not demagnetize. The interval between ca. 270 and 241.08 mbsf includes two intervals of well-defined reversed polarity: magnetozones R3 and R4, with a long interval magnetozone N4 where polarity is unclear with several samples not demagnetizing and others with persistent overprints. Magnetozone N5 is defined at its base between 241.08 and 240.49 mbsf and at its top between samples 236.37 and 235 mbsf. Magnetozones R5, N5 and N4 contain several samples that do not fully demagnetize. The boundary between magnetozones R5 and N6 is not well defined due to an interval including some samples that do not demagnetize and are noisy and have radial overprints. Magnetozone N6 is represented by four samples but also contains some samples that do not demagnetize. The boundary between magnetozones N6 and R6 is at 219.83 mbsf; at the midpoint between samples of opposite polarity at 220.74 and 218.92 mbsf. Magnetozone N7 includes only two samples and is separated from magnetozone R6 by an interval of questionable polarity conditions containing several noisy samples that do not demagnetize and have radial overprints. The interval of the core above 210 mbsf includes a well-defined N-R-N-R pattern of polarity, with the boundary between magnetozones N7 and R7 at 206.70 mbsf between samples 206.94 and 206.47 mbsf. Magnetozone R7 includes several noisy samples that do not demagnetize and have radial overprints. The boundary between magnetozones R7 and N8 is at 175.87 mbsf midway between a reversed sample at 177.53 mbsf and a normal sample at 174.22 mbsf. The boundary between magnetozone N8 and R8 is identified at 160.65 mbsf between samples at 162.11 and 159.20 mbsf.

Although most of the magnetozone boundaries were identified, some of the boundaries, particularly in the lower interval of the core, could not be located precisely because of intervals with uncertain polarity due to strong radial overprints, samples that do not demagnetize and others which were truncated by sedimentary unconformities identified by abrupt grain size changes. The boundaries between magnetozones N6 and R6, R4 and N5, and within N4 occur within sedimentary unconformities.

No further biostratigraphic information has been reported for the 316–123 mbsf interval (Cores 34–14) since the original biostratigraphic first and last occurrence information provided by the Shipboard Report (Table 1; Hayes et al., 1975). Here, we update those occurrences accounting for more recent age and occurrence data from other Southern Ocean drill holes in order to correlate our magnetic polarity stratigraphy with the Geomagnetic Polarity Time Scale (GPTS) (Figure 7).

Planktonic foraminifera occur exclusively within Core 21 (190–199.5 mbsf) Introduction (92–88 cm) and Material and Methods (73–69 cm), including Globigerina (Subbotina) angiporoides (LO 29.18–30.28 Ma, FO 40.40–41.89 Ma; Olsson et al., 2006; Wade et al., 2018) and Globoturborotalita labiacrassata Jenkins (LO 21.12–22.44 Ma, FO 30.28–32.10 Ma; Spezzaferri et al., 2018). There is some indication that the assemblage may be reworked, but a penecontemporaneous redeposition was suggested by Kaneps (1975, Table 1).

Calcareous nannofossils are present only in isolated horizons and burrows in the interval covered by Cores 28–21 (266–190 mbsf), but only in Core 21 (199.5–190 mbsf), Introduction, 93 cm and Material and Methods, 70 cm, was the assemblage sufficient for an age determination (Chiasmolithus altus assemblage) (Burns, 1975; Bukry, 1975, Table 1). Fioroni et al. (2012), in a Southern Ocean calibration, attributed the Chiasmolithus altus biozone to the time interval from chrons C12r to C8n, which was later confirmed by Kulhanek et al. (2019). Dictyococcites scrippsae (middle Eocene to Upper Oligocene; Bukry and Percival, 1971), Coccolithus pelagicus (Danian to present; Wallich, 1877) and Reticulofenestra species (Ypresian to present, Fioroni et al., 2015) were also found in the same samples.

Diatoms are generally abundant with preservation ranging from fair/poor (Cores 19–1) to excellent (Cores 34–20). Down through Core 19 (180.5–171 mbsf), diatoms are of Miocene age (Denticula antarctica Zone) and below this, through Core 34 (323–313.5 mbsf), of Oligocene age (Pyxilla prolongata; See Figures 8A,B of McCollum, 1975, Table 1).

Ciesielski (1975) identified silicoflagellates in the interval between Cores 19–13, from 179 to 114 mbsf, with this interval containing an assemblage of low diversity and abundance, poor preservation, and broken specimens. Distephanus speculum and D. boliviensis are identified throughout this interval providing poor age-diagnostic forms (Guiry and Guiry, 2011). However, the interval comprising Cores 34 to 20 (321.5–181 mbsf) displays a uniform assemblage of silicoflagellates with excellent preservation and high abundance. Mesocena apiculata, Dictyocha deflandrei s.l, Distephanus cf. D. boliviensis, Distephanus crux, Corbisema apiculata, and Naviculopsis trispinosa (See Tables 6C, D of Ciesielski, 1975) define this interval as belonging to the Dictyocha deflandrei Zone, which is subdivided into two subzones: 1) Dictyocha deflandrei Subzone (Cores 30–20) characterized by the presence of Dictyocha frenguellii and 2) Mesocena apiculata subzone (Cores 34–30) (Guiry and Guiry, 2011, Table 1).

Chen (1975) originally reported radiolarian occurrences at Site 274. Many of the species he found or described are however junior synonyms (see Table 1 for their correct names, along with a reinterpretation of his reported occurrences of bioevents at Site 274). A very typical modern to Pliocene sequence of radiolarian events is observed in cores 11–1, followed by several late to middle Miocene LADs occurring in Core 12 (Antarctissa deflandrei, 5.87 Ma, Cycladophora spongothorax, 8.62 Ma, Actinomma golownini, 10.87 Ma, Cycladophora golli, 14.92 Ma). In the interval between Cores 19–15 (180.5–133 mbsf), radiolarian associations are dominated by Cyrtocapsella isopera indicating a middle to early Miocene age (see Table 9A, Chen 1975; Weaver, 1983). Associations are completely different in the interval between cores 34 and 20, where abundant radiolarians are present, such as Amphisphaera sp., Calocyclas sp., Cenosphaera sp., Eucyrtidium sp. This interval is assigned an age of Chattian or older, based on Lithomelissa challengerae, Lithomelissa robusta, Lithomelissa sphaerocephalis, Lithomelissa tricornis, Prunopyle frakesi and Prunopyle hayesi (see Table 9B, Chen, 1975; Table 1).

On the basis of radiolarian assemblages and their comparison with foraminiferal and calcareous nannofossil assemblages, Chen (1975) hypothesized a hiatus between cores 20 (190.0–180.5 mbsf) and 19 (180.5–171.0 mbsf). This interpretation seems to be supported by his data, with the Miocene and younger portion of the sequence being well constrained based on radiolarian marker species which are well calibrated, and an older section starting with Core 20 (sample 20R-3, 40–50 cm) which is substantially different, older and much more uniform in its radiolarian assemblage (Table 1). The LADs recorded in Core 20 include Larcopyle frakesi (22.08 Ma), Lithomelissa sphaerocephalis (23.35–24.73 Ma), Lithomelissa challengerae (?28.28–29.40 Ma), and Calocyclas (?) semipolita (30.48-?33.06 Ma) which, while clearly identifying this core as substantially older than Core 19, make it difficult to assign it a well-constrained age.

Correlation of the paleomagnetic polarity zonation with the GPTS (Gradstein et al., 2012) provides an age interpretation between the upper Eocene and the Middle Miocene (from Chron C15n to C5Br) (Figure 7).

Biostratigraphic data identify a major unconformity at ∼180 mbsf (Hayes et al., 1975) separating Early/Late Oligocene strata below from Early/Middle Miocene strata above. More recent constraints on age data (Table 1) restrict the interval of core beneath the unconformity to the late Eocene–Early Oligocene and above the unconformity to the late Early–Middle Miocene (Figure 7).

Beneath the unconformity, magnetozones N1 to R7 are correlated to chrons C15–C9 as follows (Figure 7): Between 315 and 281 mbsf the normal magnetozones N1 and N2 are correlated with chrons C15n and C13n, respectively, and reversed magnetozones R2 and R2 with chrons C13r and C12r, respectively. This correlation implies a sedimentation rate of ∼1 cm/kyr for the late Eocene/Early Oligocene interval of the core with the Eocene/Oligocene boundary at ∼304 mbsf between Core 33 and Core 32. Two correlations are considered between 280 and 180 mbsf (magnetozones N3–R7; Figure 7). The younger correlates magnetozones N3. N5, R5, N6, R6, N7 and R7 with chrons C12n, C11n.2n, C11n.1r, C10n.2n, C10n.1r, C10n.1n and C9r (model D), respectively and implies that the last appearances of radiolarians Cyclampterium milowi and Calocyclas semipolita are almost one million year younger than their currently identified last occurrences. The older correlation links magnetozones N5, R5, N6, R6, N7 and R7 with chrons C12n, C11r, C11n.2n, C11n.1r, C11n.1n and C10r (model E) and has a smaller lag and consequently more consistent with the radiolarian late occurrence data but leaves magnetozone N3 with no corresponding chron in the GPTS. While we expect to see magnetozones truncated and potentially missing from the DSDP Site 274 record, magnetozone N3 is represented by several samples and multiple normal polarity chrons, and we would not expect to see an additional magnetic polarity interval not identified in the GPTS. Regardless, we observe that the sediment accumulation rate in the mid Oligocene interval of the core (280–180 mbsf) is more than four times that of the lower Eocene-Oligocene boundary interval, with 25% of the time interval accounted for by distributed unconformities as identified by marked grain size changes (Figures 2, 7).

Above the unconformity at 180 mbsf, the interval between 180 and 123 mbsf (cores 19–14), has few biostratigraphical constraints (Hayes et al., 1975, Table 1), and we identify three possible magnetostratigraphic age model reconstructions: where R8 N8 and upper part of R7 correspond to (A) chrons C5Br, C5Cn and C5Cr (B) chrons C5Cr, C5Dn and C5Dr. and (C) chrons C5Dr., C5En and C5Er, respectively (Figure 7). However, sediment accumulation rates are quite high for all reconstructions (between 2 and 5 cm/kyr), similar to the sediment accumulation rates observed beneath the unconformity.

Our new age model implies that the unconformity at 180 mbsf accounts for between 9 Myr at its shortest and 11 Myr at its longest, which is significantly shorter than the >15 Myr identified from biostratigraphy only in the original age depth model of Hayes et al. (1975).

According to our correlation and calculation of the sedimentary accumulation rate, the studied section encompasses ∼20 Myr. For the interval 180–123 mbsf (cores 19–14), we produced the following estimates of mean sedimentation rates for the two most likely age reconstructions: (model A) 2.0 cm/kyr (C5B-C5C), (model B) 5.1 cm/kyr (C5C-C5D) and (model C) 3.3 cm/kyr (C5D-C5E). The interval between 320 and 180 mbsf (cores 34–20) has a mean sediment accumulation rate of 1.5 cm/kyr (model D) and 1.4 cm/kyr (model E) with a mean of 0.8 cm/kyr between chrons C12n and C15n. The highest sedimentation rate is observed during C5Cn (following model B) with 5.1 cm/kyr and lowest sediment accumulation rate is observed during Chron C12n 0.7 cm/kyr.