The North Pacific Basin, the world’s deepest ocean basin, is rich and diverse in bathymetric features (Figure 1A): within its 22% share of the global ocean surface, the North Pacific encloses (by area) 66% of the world’s hadal zone (>6,000 m), 41% of the deep sea trenches, 39% of the seamounts, 53% of the guyots, and 29% of the submarine ridges (Harris et al., 2014). It also hosts three large igneous oceanic plateaus: the Shatsky Rise, the Hess Rise, and the Mid-Pacific Mountains. The 6,200 km long Emperor-Hawaiian seamount chain (ESC) stretches from the junction of the Kuril–Kamchatka and Aleutian Trenches at 55°N to the Island of Hawaii at 19°N (Kodama et al., 1978) separating the Northwest Pacific Basin from the Northeast Pacific Basin (Figure 1A). Of similar length is a topographic W-E divide at ∼20°N, composed of the Japanese Guyots, Wake Guyots, and Mid-Pacific Mountains. These structures form a sill that separates the two Northern Pacific Basins from the Central Pacific Basin. Our sediment cores fall into the central North Pacific sector from 30 to 50°N and from 60°E to 170°W (Figure 1B), encompassing both ridge flanks of the ESC and the adjacent Emperor and Chinook Troughs further east. The Emperor Trough consists of several bathymetric deeps extending in NNW orientation over a length of 1750 km (Erickson, 1970; Dickens and Owen, 1995). The Chinook Trough is bathymetrically expressed as a dominant deep in a series of discrete segments with large vertical relief and a maximum depth of 7,140 m (Rea, 1970; Rea and Dixon, 1983; Dickens and Owen, 1995).

North Pacific seamount bases connect in ridge- or platform-like structures that constrain and focus the lower abyssal bottom currents (Figure 1A). Lower Circumpolar Deep Water (LCDW) presently forms the homogeneous lowest deep water mass and ventilates the bulk of the deep North Pacific at water depths below 3,500 m (Mantyla and Reid, 1983; Johnson and Toole, 1993). LCDW is formed in the Southern Ocean by mixing of Antarctic Bottom Water (AABW), Pacific Deep Water (PDW) and North Atlantic Deep Water (NADW). In the North Pacific, LCDW is characterized by a potential temperature of <1.2°C and salinities ranging between 34.65 and 34.70 psu (Talley, 2003; Talley, 2013). On its way into the North Pacific, it ages and accumulates silica and nutrients, thereby loosing oxygen due to remineralization of organic carbon. Thus, the LCDW is relatively oxygen-poor with concentrations from 160–180 μmol O₂/kg in the modern North Pacific.

The 5,000 m deep Samoan Passage is the major bottleneck for bottom water transfer from the South to the North Pacific (Reid and Lonsdale, 1974; Roemmich et al., 1996). At present, ∼6.0 Sv of the LCDW supplied by the Southwest Pacific Deep Western Boundary Current (DWBC) traverse the Samoan Passage. This bottom current splits into a western branch, that flows through the east Mariana Basin, and an eastern branch, that travels through the Central Pacific Basin to partly enter the Northwest Pacific Basin through the 5,180 m deep Wake Island Passage (Rudnick, 1997; Kawabe et al., 2009). The separate LCDW branches finally reach the Northeast Pacific Basin either by a northern route along the Aleutian Trench, or through a gap in the Emperor Seamount chain at ∼39°N, or via a bypass of the Hawaiian Ridge at 20°N (Kawano et al., 2006; Komaki and Kawabe, 2009). Local upwelling and vertical mixing with overlying North Pacific Deep Water (NPDW) eventually recycles the dense LCDW waters trapped in the abyssal North Pacific basins (Talley and Joyce, 1992; Johnson et al., 2006).

North Pacific bottom water is among the oldest water masses in the world (Stuiver et al., 1983). It accumulates high CO₂, thereby leading to dissolution of calcite. Sediments from below 4,700 m are practically carbonate-free (Lyle, 2003), but carry minor contents of siliceous microfossils like diatom frustules, radiolarian skeletons and sponge spicules, which possibly stabilize the sedimentary fabric and may thereby improve the stability and quality of its paleomagnetic signal.

The abyssal sediments in the Central North Pacific primarily consist of distal East Asian dust distributed across the North Pacific by westerly winds (Blank et al., 1985; Rea and Hovan, 1995; Snoeckx et al., 1995; Yamazaki and Ioka, 1997). Serno et al. (2014) observed that over 90% of the modern abyssal sediment in the studied area ranges in size from clay to fine silt (mode: ∼4 µm), typical of long-range eolian transport. Most common minerals are illite, chlorite, kaolinite, quartz, and smectite (Corliss and Hollister, 1979). The Subarctic Front and Mixed Water Region at ∼40–45°N dividing the cold subarctic Oyashio Current from the warm subtropical Kuroshio Current marks the approximate southern iceberg drift limit during past glacial intervals, north of which coarse (>250 µm) detrital particles, easily identified as ice-rafted debris (IRD), have been found in abyssal sediments (Kent et al., 1971; St. John and Krissek, 1999; McCarron et al., 2020). However, this source of sediment supply has to be considered only during glacial times, when the transport of ice-rafted debris may expand as far south as 38°N (McCarron et al., 2021).

Sporadic volcaniclastic injections also contribute as a siliciclastic sediment component, particularly along active margins like those in the northwest Pacific. The most relevant pyroclastic fraction in Neogene North Pacific sediments is tephra from the Japan-Kuril-Kamchatka and Izu–Bonin–Mariana Arcs (Straub and Schmincke, 1998). Tephra frequencies are highest in western North Pacific: Derkachev et al. (2020) counted 25 tephra layers within 215 kyr at the north-west terminus of the ESC downwind from the Kamchatka peninsula. Discernible tephra and pumice layers further south are far less frequent (Natland, 1993). Floating pumice formed by shallow submarine rhyolitic eruptions can however drift over large distances. This has also been observed in the North Pacific, but is not nearly as common as in the SW Pacific (Bryan et al., 2012). Fine-grained weathering materials from outcropping volcaniclastic aprons of the ESC may be dissipated by bottom current erosion and reach the adjacent ocean basins by nepheloid transport (Manville et al., 2009). Collapses of instable seamount flanks and slope deposits can generate extensive volcanoclastic plumes and turbidite lobes that may share some similarities with tephra layers.

The sediments investigated here were collected during two North Pacific expeditions of the old and new German research vessels R/V SONNE. The SO202-INOPEX cruise [Tomakomai (Japan)—Busan (South Korea)] in 2009 focused on the biogeochemistry and paleoceanography of the subarctic Pacific and Bering Sea, while the SO264-EMPEROR cruise in 2018 [Suva (Fiji)—Yokohama (Japan); Nürnberg, 2018] explored in particular the shallower North Pacific ESC, thereby addressing the North Pacific’s role in Plio-Pleistocene climate change and carbon cycle dynamics. As the paleo- and rock magnetic studies of Korff et al. (2016) showed cyclic Late Pleistocene glacial magnetite depletion at Northwest Pacific core SO202-39-3, we now investigate two Northeast Pacific abyssopelagic cores SO202-33-4 (Chinook Trough, 6,133 m) and SO202-35-1 (Emperor Trough, 5,507 m) to determine if such cyclic or episodic magnetite dissolution layers are present on either side of the ESC. We also selected three of the deepest SO264 cores (SO264-19-2, 22-2, and 56-2) from abyssal plain locations in proximity to the ESC at mid-latitudes of 39.5–47.7°N and water depths of 3,973–6,133 m (Table 1). All five core locations exhibited well-stratified, undisturbed seismic reflectors as indicated by sub-bottom profiling site surveys. The shipboard core lithological and logging data indicated largely homogenous muddy sedimentary sequences with strong and steady susceptibility signals. We shortly describe each core in the following and refer to the cruise reports (Gersonde, 2012; Nürnberg, 2018) for more detailed site and core descriptions.

The 22.7 m long piston core SO202-33-4 was recovered north of the Chinook Trough fracture zone from a hadal water depth of 6,133 m; it is the deepest core of expedition SO202 and also of our study. The sediment is described as diatom-bearing mud, and, below 15.52 m, as diatom ooze. The 19.29 m long gravity core SO202-35-1 was retrieved from an abyssal plain west of the Emperor Trough at 5,507 m water depth. Its lithology comprises homogeneous light brownish diatom-bearing clay, mottled in various extents. The presence of manganese nodules gives evidence of a very low sedimentation rate as these grow very slowly (Barnes and Dymond, 1967). The 14.37 m long gravity core SO264-19-2 was collected SW of Ninigi Seamount and NW of Nintoku Seamount at a water depth of 5,311 m. It is composed of light to dark gray silt and contains a few thin tephra and turbidite layers. A pumice fragment of 1.5 cm size was found at a core depth of 11.00 m. The 16.21 m long gravity core SO264-22-2 was recovered from a site near Soga Seamount north of Yomei Seamount at a water depth of 5,704 m. The sediment consists of light to medium gray and brownish silty ooze. A volcanic rock was found at 0.12 m depth and a tephra layer at 3.86–3.90 m depth; laminated layers were observed in core depths of 7.51–7.64, 8.21–8.25, 9.21–9.52 and 11.06–11.21 m, respectively. The 12.56 m long gravity core SO264-56-2 was taken on the base of Minnetonka Seamount at a bathyal to abyssal water depth of 3,973 m. The core’s uppermost 1.6 m consist of dark gray silt and pelagic ooze with intercalated sandy (turbidite) layers at 3.28–3.30, 3.84 m and tephra layers at 4.68–5.32 and 7.33–7.37 m. A sharp drop in radioactive Th²³⁰ concentration was discovered at ∼2.3 m depth in this core (Walter Geibert, pers. comm.) that implies a hiatus from 140 to 340 ka. This hiatus was not visually detected during initial core description.

After shipboard logging of whole-round cores with a GEOTEK Multi-Sensor Core Logger (MSCL) (Höfken et al., 2018), all core segments were split into halves. Magnetic volume susceptibility κ was measured at the University of Bremen on archive halves in 1 cm increments using a Bartington MS2 susceptometer with a high-resolution MS2F (Ø 15 mm) spot sensor mounted to an automated logger. Temperature drift effects were eliminated by performing a void measurement in air after each sediment measurement.

All working halves were continuously subsampled with non-magnetic 6.2 cm³ plastic cubes, obtaining a sample spacing of ∼2.3 cm; solely core SO202-35-1 had been subsampled at a lower resolution of 5 cm. Natural Remanent Magnetization (NRM) of these 3,122 cube samples was measured with an automated 2G Enterprises model 755R cryogenic magnetometer system with integrated alternating field (AF) demagnetization unit at the paleo- and rock magnetic laboratory at the University of Bremen (Mullender et al., 2016). NRM was AF demagnetized in 16 steps, using a 3-axis demagnetization mode with peak fields of 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 60, 70, 80, 90, and 100 mT. An Anhysterestic Remanent Magnetization [ARM, reflecting the submicron magnetite concentration (Egli and Lowrie, 2002)] was then imparted in a 100 mT peak AF and 50 µT DC bias field to all samples. The ARM of the three SO264 cores was demagnetized in the same 16 steps as NRM while the ARM of the two SO202 cores had been previously demagnetized in 11 steps with peak fields of 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 100 mT.

ARM normalized by its DC bias field is termed anhysteretic susceptibility κ_ARM. The κ_ARM/κ ratio is a diagnostic value of magnetite particle or crystal size (King et al., 1983). The low-field volume susceptibility κ used as normalizer was measured with a Bartington MS3 susceptometer and a MS2B discrete sample sensor. Each sample was measured three times with intermitting void measurements, after which the net values were averaged. The proxy ratio Fe/κ (Funk et al., 2004; Korff et al., 2016; Shin et al., 2019) compares total iron to magnetite content, i.e. paramagnetic (or total) Fe to ferrimagnetic Fe species concentration. The Fe/κ ratio usually carries a subtle primary signature, reflecting sediment source rock petrology and provenance, and a usually more prominent secondary imprint, indicating zones of magnetic mineral diagenesis (depletion, formation, reduction, oxidation).

Following the remanence and susceptibility measurements, we selected five representative samples of core SO264-19-2 (dotted lines in Figure 2) for First-Order Reversal Curves (FORC) analyses (Pike et al., 1999; Roberts et al., 2000). About 25 mg of dried sediments were stabilized with glue and measured with a Princeton Measurement Corporation (PMC) Micromag M2900 alternating gradient force magnetometer following the protocol of Egli (2013). To process and plot FORC distribution data we used the VARIFORC method of Egli et al. (2013).

Magnetic separates for electron microscopic observation were extracted from suspended sediment by magnetic techniques (Hounslow and Maher, 1996). Prior to dispersion, the susceptibility of each selected sample was measured to quantify magnetic extraction efficiency. About 3 cm³ sediments were dispersed in demineralized water by high-power ultrasonic agitation. Sodium polyphosphate was added as a peptizing agent to keep clay mineral particles in suspension. This suspension was recirculated in a closed extraction system, where magnetic particles attach to a “magnetic finger” (von Dobeneck et al., 1987). The magnetic extracts collected during 24 h operation were repeatedly washed with demineralized water to remove attached clays.

The magnetic particle fraction of >1 µm was investigated with a ZEISS DSM 940A scanning electron microscope (SEM) with energy-dispersive X-ray spectroscopy (EDX) analyzer at the Faculty of Geoscience, University of Bremen, operated at 15 kV. A drop of suspended magnetic extract was placed on the SEM stub, dried and coated with a carbon layer to prevent surface charging during SEM observation. For transmission electron microscopy (TEM) of the submicron fraction, we used a Zeiss EM 900N TEM at the Electron and Light Microscopy Service Unit, Faculty of Mathematics and Natural Sciences, University of Oldenburg, operated at 80 kV. For TEM, magnetic extracts were agitated in an ultrasonic water bath for 2 min to separate magnetic particles, then a drop of this suspension was placed on a carbon coated TEM copper grid and dried in air.

High-resolution records of the major elements Fe, Ca, Ba, Ti and Al were obtained by X-ray fluorescence scanning with an Avaatech Series 4 XRF-core scanner at the Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research (AWI) Bremerhaven Marine Geology Laboratory. Prior to scanning, the topmost sediment layer on the core archive half was removed to create a fresh, flat surface. SPEXCerti Prep Ultralene ^® film was placed on the sediment surface, all XRF-scanning was carried out with 1 cm sample resolution, a 10 mm slit size and a current of 0.15 mA. Three consecutive runs were performed with tube voltages of 10, 30, and 50 kV, and with acquisition times of 10, 30 and 50 s, respectively. Raw data was processed using Canberra Eurisys’ iterative least squares software (WIN AXIL) package. Element concentrations are reported as area counts.

The reliability of paleomagnetic records critically depends upon the magnetic, chemical and mechanical (orientation) stability of the NRM carriers. This is even more relevant for RPI-studies (Tauxe, 1993) than for reversal magnetostratigraphy. Analyzing the sediment’s magnetic properties prior to interpreting paleomagnetic results is therefore an established and essential first step in magnetostratigraphic studies. Here we present exemplary rock magnetic analyses for two cores from different settings, ESC core SO264-19-2 (0–2.01 Ma) and Chinook Trough core SO202-33-4 (0–3.45 Ma). High-resolution rock magnetic property logs first identify the trends, events, cycles, and diagenetic alterations of the sedimentary magnetic mineral assemblages. FORC analyses and electron microscopy of five representative samples then reveal the specific nature and likely origin of magnetic particles that are thought to carry reversal and RPI information over extended periods.

The magnetostratigraphic-RPI dating approach generally follows four steps: 1) establishment of an initial age model based on geomagnetic reversals, 2) selection of the optimal RPI normalization method, 3) refinement of the reversal age model by correlating the RPI logs to an astronomically tuned RPI reference record, and 4) corroboration of the RPI-based age model by orbital tuning of a cyclic proxy record, if available.

As shown by early rock magnetic studies of the abyssal North Pacific (Vali et al., 1987; Yamazaki and Ioka, 1997), magnetofossils are omnipresent and abundant in pelagic sediments, and, by virtue of their SD domain state and magnetic stability, efficient NRM carriers. The benthic production of biogenic magnetite appears to be rather uniform, quite in contrast with the very dynamic deposition of eruptive or redeposited volcaniclastic materials. Magmatic titanomagnetite particles, ranging from coarse phenocrysts to fine silicate-embedded magnetite inclusions, may not be efficient PDRM carriers and bear a risk of creating bias in RPI normalization by their undefined magnetic properties and fluctuating presence. While biogenic and volcanogenic magnetic mineral accumulation is not subject to long-term trends, the eolian influx of fine dust particles from Inner Asia increases significantly through the Plio-Pleistocene and progressively controls the rock and paleomagnetic characteristics of abyssal North Pacific sediments, as Yamazaki and Ioka (1997) first showed. The increase of κ relative to Fe signalizes a shift from more paramagnetic to more ferrimagnetic minerals (Figure 2), which may imply a growing impact of physical (cold, arid, magnetite-preserving) Late Pleistocene weathering regimes relative to a more chemical (warm, humid, magnetite-oxidizing) pedogenesis in the Early Pleistocene. Similar changes are known from Chinese loess records (e.g., Heslop et al., 2000; Zhang et al., 2020a, 2020b).

To analyze the impact of magnetic grain size and composition variations and sediment lithology on the RPI records, we first compared κ_ARM/κ distribution of all cores (Figure 10A) and then RPI vs. κ_ARM/κ scatter plots for each individual core (Figures 10B–F), distinguishing volcaniclastic and diagenetic layers as well as normal and reverse sections. As Figure 10A shows, the range of magnetic grain sizes varies largely between cores. The easternmost core SO202-33-4 displays the widest range of magnetic grain sizes from fine to coarse, which actually represents a temporal trend (Figure 2, Figure 10B) rather than sediment heterogeneity, resulting from increasing detrital magnetite (eolian dust) input since the Mid-Pleistocene. The cross plots in Figures 10B–F reveal remarkable differences in terms of magnetic grain-size variation and its impact on RPI. The low and rather constant κ_ARM/κ ratios of cores SO264-56-2 (∼5–10) and SO264-22-2 (∼10–25) reflect greater proximity to the ESC resulting in higher volcaniclastic input from nearby seamount flanks. This proximal input is evidenced by the high frequency and greater thickness of volcaniclastic layers and the microscopically observed coarseness of magnetic particles (Figures 4A–F). Given that the ECS seamounts have been volcanically inactive for over 40 Myr, such phases of volcaniclastic deposition and massive erosion (both cores have a large hiatus within the Brunhes Chron) could be contouritic or tubiditic, driven either by deep-water currents along the ESC (Figure 1) or by downslope turbidity currents triggered by local seamount slope failures. As the impact seemed stronger during the Brunhes Chron, we prefer the first explanation that provides a paleoceanographic link to a possible Mid- to Late Pleistocene intensification of bottom current intensity. Neither contouritic nor turbiditic sedimentation is desirable for RPI records as sediment rates and grain sizes could vary at short scales, adding randomness to the RPI pattern.

A larger range of the magnetogranulometric κ_ARM/κ values is observed (Figures 10B–D) for the older complete cores SO202-33-4 (10–70), SO202-35-1 (15–50) and SO264-19-2 (10–50 without diagenesis), which are similar to the values of Yamazaki and Ioka (1997). We interpret these large magnetic grain size variations as increased admixing of coarser detrital (titano-)magnetite to a constant benthic production of biogenic magnetite (Franke et al., 2007). Older sections of the RPI record with their finer (biogenic) magnetic particle sizes have consistently lower values. The RPI vs. κ_ARM/κ biplots of these three cores data follow hyperbolic trends (e.g., red line in Figure 10B), i.e., RPI decreases downwards, together with the fining of magnetic grain size (Figure 2, Figure 6, Figure 10B). We tentatively use this inverse relation to rescale the SO202-33-4 RPI record (dotted line of SO202-33-4 in Figure 11A) by adjusting the NRM/ARM level by multiplication with κ_ARM/κ. This is not regarded as a general method, but visibly enhances the patterns in the lower part of the RPI record.

Such age-dependent decrease of RPI amplitudes is not seen in the EPAPIS-3 Ma stack (Figure 8). Yamazaki and Oda (2005) related long-term RPI trends rather to changes in accumulation rates, which are not found in our records (Figure 9). For South China Sea RPI records, Ouyang et al. (2014) reported biogenic particles to be 2–4 times more efficient PDRM carriers than detrital components. Chang et al. (2016) argued that small particles could better overcome the hydrodynamic forces than larger particles and would therefore be more reliable paleomagnetic recorders. Chen et al. (2016) also found that the biogenic SD component of east Equatorial Pacific sediment carried 2–4 times higher NRM/ARM than a detrital component consisting mainly of SD sized magnetic inclusions in silicate crystals. However, different PDRM acquisition efficiencies of biogenic and detrital components seemed not to have important consequences as both carrier types recorded the same signal with overall coherent patterns.

The higher RPI efficiencies of biogenic SD magnetite suggested by previous studies contrast our findings for which various explanations are discussed in the following: 1) viscous remagnetization, 2) different magnetic carriers efficiencies, 3) reductive dissolution, or 4) oxidation of fine magnetite particles.

Viscous normal overprinting of reverse sections is observed in Zijderveld plots (e.g., Figure 7B), but could not have been the leading cause for the lower NRM/ARM in older sections, since the normal Gauss record section of SO202-33-4 has even lower RPI values than the reverse Matuyama section above (Figure 8) and the soft 0–20 mT AF demagnetization track sections were omitted from RPI.

A possible mineralogical explanation could lie in differences of the detrital magnetic carriers in the studied areas. Magnetite nanoparticle inclusions in silicates (Chen et al., 2016) should indeed be less suitable carriers in terms of PDRM theory, while similarly large isolated detrital magnetite particles in the vortex state (former PSD) size spectrum could be more efficient remanence carriers than twisted chains or interacting clusters of magnetofossils, in particular, if their NRM would in parts be a Depositional Remanent Magnetization (DRM) (Valet et al., 2017). At the studied central North Pacific mid-latitude sites, the influx of long-range wind-transported ultra-fine continental dust is bound to carry only very fine weathering-resistant magnetite and maghemite particles. Likewise fine, proximally sourced magmatic titanomaghemite particles from weathering rocky outcrops at seamount flanks, dispersed and sorted by bottom or turbidity currents, would also be accumulated as individual particles and should likewise make good DRM or PDRM carriers. Figures 10B,D,E, show that the data points of isolated volcaniclastic layers often have lower κ_ARM/κ values and a wider RPI range. They can record magnetic paleointensity, but with greatly varying efficiency. The two large RPI peaks near the bottom of core SO202-33-4 actually have their origin in such volcaniclastic layers.

Cyclic magnetite dissolution by glacial oxygen depletion and carbon trapping as described by Korff et al. (2016) and Shin et al. (2018) for the abyssal NW Pacific can reduce RPI, but was only detected in core SO264-19-2, which features one broad and two fine magnetite dissolution layers (Figure 3, Figure 8). In the RPI vs. κ_ARM/κ biplot (Figure 10D), these samples form an isolated cluster of strongly reduced RPI and very low (2–8) κ_ARM/κ values. This underlines the finding by Korff et al. (2016) that magnetite-depletion entails severe PDRM loss and dramatically deteriorates RPI patterns. The phenomenon of glacial benthic anoxia and related magnetite depletion seems to be rare at the ESC and absent further east, at least in the latitudinal range covered by the here studied sediment cores. From the SO264 magnetic susceptibility logs (Nürnberg, 2018), we expect that this situation should be different in the subarctic North Pacific (north of 50°N).

From our viewpoint, the most convincing explanation for long term NRM loss in abyssal sediments can be found in early paleomagnetic work in the North and South Pacific by Opdyke and Foster (1970), Kent and Lowrie (1974), Prince et al. (1980) and Yamazaki and Katsura (1990). These authors believed that the unstable magnetic behavior of the Pliocene abyssal sediments of the cores was related to the presence of highly oxidized fine-grained magnetite, which we would now consider as biogenic. Their reasoning for increasing magnetic instability with oxidation (i.e. maghemitization) state is a shift of the superparamagnetic (SP) threshold of stable SD particles from 30 nm for magnetite (Dunlop, 1973) up to 120 nm for stoichiometric maghemite (Readman and O’Reilly, 1972). As biogenic magnetite particles mostly fall into that size range and become fully oxidized when being exposed to oxic bottom and pore water conditions over long periods (Chang et al., 2013), the NRM and thus RPI loss may be best explained by oxidation-related unblocking of fine biogenic magnetite particles. A marked loss of ARM during the Gauss Chron (Figure 6D) without decrease in κ/Fe evidences magnetite oxidation rather than magnetite reduction. Possibly this destabilization only concerns the numerous small octahedral magnetofossils, while less abundant larger, e.g., prismatic magnetofossils (Figure 4) may still preserve a stable SD state. Yamazaki and Katsura (1990) proposed, that the “unstable-to-stable transition” after late Pliocene resulted from less oxidizing benthic conditions in the Pleistocene.

We next compare our new abyssal RPI records to previously published RPI target stacks and records and also briefly test cyclostratigraphic alternatives using existing paleoclimate proxy records of our cores (Figure 11A, Figure 11B). In order to keep this graphic compilation manageable in size, we spliced the incomplete RPI records of SO264-19-2 and -22-2 together and only show the RPI records of SO202-35-1 and -33-4 in full length. The depicted RPI target records are EPAPIS-3Ma, SINT-2000, PISO-1500 and ODP Leg 138 (Figure 11, for references see Table 1). From these published RPI records, only the EPAPIS-3Ma ages have been modified to the GPTS2020 timescale (from dotted to solid lines). It appears that the other RPI stacks are largely synchronous with GPTS 2020, probably due to their δ¹⁸O tuned chronologies.

In the Mid- to Late Pleistocene Brunhes Chron, the RPI patterns of our cores agree generally well with all target records. Some major intensity minima linked to Brunhes excursions at ∼41 ka (Laschamp) and ∼188 ka (Iceland Basin; Channell et al., 2020) are prominent in our RPI records. Characteristic peaks at ∼350–400 ka and ∼600–700 ka also exist in all records. The early Brunhes “West Eifel Suite” (530–730 ka) of 5–6 geomagnetic excursions (Böhnel et al., 1987; Singer, 2014; Channel et al., 2020) is at least partly reflected in large and well recognizable RPI amplitudes.

There is a lag between the original EPAPIS-3Ma record and the PISO-1500 and SINT-2000, which may be due to a delayed response of the carbonate system to changes in continental ice volume and sea level resulting in carbonate concentration variations lagging benthic δ¹⁸O (Hodell et al., 2001). This is one of the reasons, why we adjusted EPAPIS-3Ma to GPTS2020 in the Early Pleistocene Matuyama Chron. Our linear approach for this may have resulted in some over-corrections as the ∼20 kyr offset of a large RPI peak at ∼880–940 ka relative to other reference RPI records suggests. The Jaramillo Subchron RPI pattern lacks some detail, but still matches the target record very well. Interestingly, the previously observed long “CM single RPI intensity minimum” can also be recognized in the PISO-1500, SINT-2000, and ODP leg 138 records, where single, very broad and low RPI minima instead of two sharp framing minima are observed. The stretched minima agree well with the length of the revised CM duration. The few cosmogenic records also agree that geomagnetic intensity low persists for ca. 20 kyr (Carcaillet et al., 2004; Simon et al., 2018). For the Olduvai Subchron, our RPI patterns are not as detailed as in EPAPIS-3Ma; SO264-22-2 is actually more similar to SINT-2000. For the beginning of the Matuyama Chron, we are left just with the Pacific RPI targets where some uncertainties in the pattern matching are obvious.

The Late Pliocene Gauss Chron was best covered by core SO202-33-4, but lower RPI amplitudes make the match with EPAPIS-3Ma more challenging. The κ_ARM/κ adjusted RPI record of SO202-33-4 (dotted line in Figure 11A) shows more prominent variations, which are quite consistent with EPAPIS-3Ma. As the reliability of this RPI correction method has not been elaborated and validated, we leave these two options for comparison, but not for correlation.

In Figure 11B, we present available high-resolution paleoclimate proxy records (S_-0.3T and Ba/Al) for our cores, scaled to their RPI age models. North Pacific glacial-interglacial dust flux cycles are often well reflected in relative hematite variations depicted by the rock magnetic S_-0.3T ratio (e.g., Yamazaki and Ioka, 1997). With their larger distance from Asia, the S-ratio records of our cores vary just minimally between 0.95 and 0.98. These small changes may also mirror other processes than dust flux, e.g., diagenesis (Korff et al., 2016; Shin et al., 2018). Any attempt to correlate one of our rock magnetic records to a climate stack like LR04 is risky and definitely far less promising than RPI matching, at least for most parts of these rock magnetic records. On the other hand, RPI matching also avoids the circular problem of using climatic constrains to discuss climatological changes.

In the North Pacific, Ba/Ti or Ba/Al records are indicative of changes in paleo-export production changes and deep water ventilation. These records have been used as correlation targets where lower values in these ratios reflect glacial periods (Jaccard et al., 2010; Serno et al., 2014). Korff et al. (2016) correlated the SO202-39-3 Ba/Ti record to the astronomically tuned Ba/Al record of ODP 882 to solve the diagenesis-related RPI uncertainties in the upper core section. Shin et al. (2018) compared the Ba/Ti and Ba/Al records of various Shatsky Rise cores and found out that their minima in MIS 10, 8 and 6 were synchronous. Among all various inspected element ratios of our cores, Ba/Al records also had the most characteristic, in some parts even recognizably cyclic features (Figure 11). Without further detailed cyclostratigraphic analysis, common cyclic patterns are recognizable in some parts of the Ba/Al records of the herein studied cores, e.g., from ∼1,050 to 1,650 ka. However, in other sections, e.g., from ∼500 to 1,000 ka and ∼1,700 to 3,500 ka, the Ba/Al records have little in common, although certain cyclic and singular features do exist, e.g., the 2.73 Ma break (Bailey et al., 2011) in core SO202-33-4. It is also worth mentioning that we observe typical relative age offsets in the order of ∼20–40 kyr, which may be related to the complex regional Ba/Al signal formation and preservation processes (Serno et al., 2014) or to the RPI age model itself. This situation hinders the construction of a high-resolution cyclostratigraphic network using Ba/Al records, and thus prevents to perform an independent test of our RPI-based age models as initially intended.