Rock-magnetic characteristics
The results of the thermomagnetic and low-temperature magnetic analyses of the two selected specimens are shown in Fig. 3. The thermomagnetic experiments, performed in both air and a vacuum, demonstrate that the specimens have a single Curie/Néel temperature at approximately 580 °C (Fig. 3a). The thermomagnetic curve for specimen KJ100 shows no abrupt changes throughout the heating and cooling processes, but specimen Rhr44 exhibits a small plateau at approximately 450 °C during the heating process. The normalized induced magnetization (J/J0) during cooling in air is slightly smaller than that during heating for both specimens. In contrast, J/J0 during cooling in a vacuum is slightly larger for KJ100 but substantially larger than that during heating for Rhr44 (Fig. 3a). The results of KJ100 indicate that the specimen contains (titano) magnetite with a Curie temperature close to that of magnetite at 585 °C (e.g., Hunt et al. 1995). The cooling curves are irreversible because of the degradation of iron oxyhydroxides within the specimens into ferrimagnetic minerals due to the heating or minor thermal alteration of titanomagnetite (Bowles et al. 2013).
Conversely, the results of Rhr44, in which the specimen was heated in air, exhibit a small plateau at approximately 450 °C, indicating the creation of a ferromagnetic mineral phase. The thermomagnetic curve of the cooling process in air has a lower magnetization than the equivalent heating curve, indicating that the ferromagnetic or ferrimagnetic minerals created at approximately 450 °C were further altered by continued heating and oxidized to form a higher Curie/Néel temperature mineral, such as hematite. In contrast, the specimen heated in a vacuum has a single Curie/Néel temperature at 580 °C, the same as that heated in air, but almost no increase in magnetization at approximately 450 °C. Moreover, unlike during the heating process, the curve exhibits a drastic increase in magnetization throughout the cooling process. The thermomagnetic behavior seen in Rhr44 was also observed in the Kazusa Group (Okada et al. 2017) and has been reported as one of the typical behaviors of greigite-bearing samples (Roberts et al. 2011).
The low-temperature remanence curves (Fig. 3b) indicate the presence of a Verwey transition, where the magnetite transforms from a monoclinic to cubic spinel structure between 110 and 120 K (Verwey 1939; Özdemir et al. 1993). The curves, which indicate a declining ratio of magnetization per unit temperature, are characterized by a rapid remanence decline from 10 to 50 K, with a broad Verwey transition from 100 to 120 K. Özdemir et al. (1993) reported that the Verwey transition tends to shift to lower temperatures with a broader transition temperature range due to maghemitization (oxidation) on the surface of magnetite. The rapid decrease in remanence between 10 and 50 K is considered to be due to the influence of superparamagnetic particles (Özdemir et al. 1993; Özdemir and Dunlop 2010). The Day diagram of selected specimens is shown in Fig. 4. The boundaries of the magnetic domain state regions for the Day diagram follow the work of Dunlop (2002). All specimens fall into the vortex state (PSD) region (Roberts et al. 2017), which can also be interpreted as a mixture of particles in the SD and MD states, within a relatively small area where the data are distributed. Although, as Roberts et al. (2018) recently discussed, the Day diagram is fundamentally ambiguous for domain state diagnosis; therefore, our diagnosis described above should be treated as just a reference for the domain states.
To evaluate the variation in rock-magnetic properties, the low-field magnetic susceptibility (kLF) and the ARM susceptibility (kARM), as well as the ratio of both parameters throughout the stratigraphic sequence, are shown in Fig. 5a, b. As the ARM experiment was only conducted for the upper part above 36 m, i.e., where the hybrid method was applied, the kARM and the kARM/kLF ratio are not plotted below 36 m. Although several spikes are observed in the kARM/kLF ratio, which likely corresponds to tephra and/or sandy layers, these data indicate that the rock-magnetic characteristics are relatively homogeneous for this interval (Fig. 5b). The kLF and kARM curves, which exhibit variations well within an order of magnitude, suggest that the specimens from the KJ route are suitable for estimating RPIs (Tauxe 1993).
Remanent magnetization
Typical orthogonal vector diagrams (Zijderveld 1967) for the NRMs obtained from pAFD, pThD, and the hybrid method are shown in Fig. 6. The results of the pAFD (Fig. 6a, d) indicate that the NRMs consist of two magnetic components. The low-coercivity (LC) components were removed at a peak field of 15–20 mT and the remaining high-coercivity (HC) remanences demagnetized almost linearly, but slightly curvilinearly toward the origin (grey arrows in Fig. 6a, d). The demagnetization paths for the pThD (Fig. 6b, e) also indicate the existence of two components, among which the low-temperature (LT) components are demagnetized by 250 °C. The remaining high-temperature (HT) component of KJ85 (Fig. 6b) is almost linearly demagnetized toward the origin but slightly digresses at 600 °C, whereas that of KJ82 (Fig. 6f) demagnetized with a curvilinear path (green arrows in Fig. 6b, e), indicating that the HT component in KJ82 consists of a mixture of two different components. One of those components in HT had the same direction with LT but might have had blocking temperatures overlapped with the other component in HT. Moreover, the demagnetization path digresses substantially from 450 °C (red arrow in Fig. 6e), probably affected by magnetic minerals newly created during the ThD process. Although the secondary components seem to be removed more from the pThD paths than from the pAFD paths, the HT component in pThD may not be a primary component (Fig. 6e). In contrast, the hybrid results (Fig. 6c, f) for both specimens indicate that the secondary components are effectively removed by 15–20 mT of AFD after 250 °C of ThD with no curvilinear demagnetization paths (orange arrow in Fig. 6f). Here, we assume that both the LC component and HC component in pAFD consist of a low unblocking temperature component and a high unblocking temperature component, respectively. We can express the low and the high unblocking temperature component in the LC as LTLC and HTLC, and those in the HC as LTHC and HTHC, respectively. As above, we assume that both the LT and HT components in pThD consist of a low-coercivity and a high-coercivity component, respectively. Here, too, we can express the low- and the high-coercivity component in the LT as LTLC and LTHC, and those in the HT as HTLC and HTHC, respectively. In the hybrid results, the component below 250 °C corresponds to the LT in pThD, and the remaining two components from 250 °C to 15 mT (blue arrow in Fig. 6f) and higher than 15 mT (orange arrow in Fig. 6f) correspond to HTLC and HTHC, respectively. The comparison between the hybrid and the pThD results (Fig. 6f, e) clearly shows that the HT component (green arrow in Fig. 6e) consists of the sum of HTLC and HTHC (dashed blue and orange arrows in Fig. 6e), indicating that the HT component includes the HTLC component as a secondary magnetization. The HTLC component is not able to be separated by the pThD method alone since the spectrum of unblocking temperature of HTLC is comparable with that of the HTHC component, which probably corresponds to the primary magnetization.
In the reversals test, if the ChRMs indicating reversed polarity are distributed symmetrically with those indicating normal polarity, the ChRMs can be treated as a primary paleomagnetic record without any secondary component. For the reversals test, we used all the ChRMs, excluding the intervals of the reversal boundaries and the lower normal polarity zone (40–47 m), where the RPIs indicate much lower than the average value. In Fig. 7a, the α95 circle of reversed polarity ChRMs does not overlap with the α95 of normal polarity, indicating that the secondary components are not removed from the pAFD results. The same results are observed for pThD in Fig. 7b. This situation, whereby both pAFD and pThD failed to remove the secondary components, was also observed in the lower Chikura Group (Okada et al. 2012). In contrast, the hybrid result (Fig. 7c) shows that the α95 of reversed polarity includes the average of the normal polarity, and vice versa, which indicates that the ChRMs deduced by the hybrid method can be treated as the primary paleomagnetic record with a level of significance of 5% or less. This result indicates, for the first time, that the hybrid method is statistically confirmed to be able to successfully remove the secondary components since Okada et al. (2017), which did not conduct any field tests, including the reversals test. This result also indicates that the interpretation that the HTHC corresponds to the primary magnetization is proven.
According to the results of the hybrid method (Fig. 7c), we calculated the average declination value and the associated α95 interval of the hybrid ChRMs as 8.6 ± 3.3°, indicating that a clockwise tectonic rotation of ~ 10° occurred in the KJ route region, which is similar to the results of a previous study (Kotake et al. 1995). We subtracted this value from the declinations to remove the rotation effect for subsequently determining the VGPs. All the ChRM directions, MADs, VGP latitudes, and RPIs, which were used following the discussion of geomagnetic field variations, were deduced by the hybrid method, which was conducted for all horizons above 36 m (Fig. 5c–g). However, we also plotted the ChRM directions, MADs, and VGP latitudes deduced by the pThD below 36 m as reference data (Fig. 5c–f). The ChRM inclinations are in close agreement with the expected values for the geocentric axial dipole (dashed lines in Fig. 5d). The MADs generally remain below 5°, while they exhibit much higher values in some intervals, mainly associated with geomagnetic reversals. We avoided using ChRMs with MADs greater than 15° (Fig. 5c) in the subsequent discussion.
Magnetostratigraphy
The VGP latitudes plotted in Fig. 5f indicate that two normal polarity zones are identified in the stratigraphic intervals above 142 m and between 38.6 and 44.6 m, respectively. An intermediate polarity zone identified from 12 to 15 m will be discussed in a future study after applying the hybrid method. Kotake et al. (1995) reported that the Olduvai subchronozone was identified in the middle part of the Hata Formation, and a marker tephra bed named the KO tephra was intercalated in the Olduvai interval (Fig. 1e). The KO tephra is described as HF tephra in Kawakami and Shishikura (2006). They reported that the HF tephra was identified on a horizon slightly above the interval of this study. On the other hand, Okada et al. (2012) reported that the Gauss/Matuyama boundary was identified in the Minamiasai Formation, which is just below the Hata Formation. These observations confirm that the upper and lower normal polarity zones identified in this study correspond to the Olduvai and Réunion subchronozones, respectively. The average sedimentation rates between the lower Réunion boundary at 39.3 m and upper Réunion boundary at 44.6 m, and between the lower Olduvai boundary at 142.0 m and the upper Réunion boundary, are calculated as 25 cm/ky and 57 cm/ky respectively, using boundary ages of 2137 ka and 2116 ka for the lower and upper Réunion, respectively (Channell et al. 2016), and 1945 ka for the lower Olduvai (Ogg 2012). Since these sedimentation rates are higher than those in any deep-sea core (e.g., Channell et al. 2002, 2016; Valet et al. 2014), this sedimentary sequence has the potential to provide a paleomagnetic record that will play an important role in finding unknown field behaviors during the geomagnetic reversals.
In addition, we found a very characteristic tephra bed, in which cummingtonite is exclusively predominant among the heavy minerals, at the 43.55 m horizon that is 1 m below the upper Réunion boundary (see S. Figure 1). This tephra bed could be useful as a local marker tephra bed to indicate a horizon of the upper Réunion boundary.
Variations of VGP and RPI during the reversals
The VGP paths, latitudes, and RPIs between 36 and 47 m, which cover the entire Réunion subchronozone, are shown in Fig. 8a–c, respectively. In the same manner, those for the Réunion subchron from ODP Site 981 (Channell et al. 2003) are shown in Fig. 8d–f as a corresponding record. The VGPs of this study passed across the equator within a similar longitudinal band over Africa during both the upper and lower Réunion reversals, which is also observed in the Site 981 record, although the longitude is different (Fig. 8d). Moreover, the VGPs of this study settled in a VGP cluster area around China, termed “A” in this study (Fig. 8a) and is identified between 43.0 and 43.5 m immediately below the upper boundary (Fig. 8b). A similar cluster area with the same timing, but on a different longitudinal band from this study, is also observed in the Site 981 record (marked “A” in Fig. 8d, e). These observations suggest that these two records might reflect the nondipole feature of the geodynamo during the Réunion subchron. The VGPs in this study settled in a VGP cluster area around Argentina during the upper boundary reversal, termed “B” in this study (Fig. 8a, b). However, this is not apparent in Site 981 (Fig. 8d). In Fig. 8c, although the RPI values indicate minima at both the upper and lower Réunion boundaries, the RPI values for the entire Réunion interval are generally lower than the average of the entire interval of this study (Fig. 5g). This observation suggests that the dipole moment during the Réunion subchron never recovered to the average level after the first reversal ended, which was also observed in the RPI record from deep-sea sediments of the North Atlantic Ocean (Channell et al. 2016; Channell et al. 2020), and authigenic 10Be/9Be record from a deep-sea core of the western Pacific Ocean (Simon et al. 2018).
In the same manner in Fig. 8, we show our record from the stratigraphic interval between 138 and 145 m covering the lower Olduvai reversal in Fig. 9a–c, and the Site 984 record (Channell et al. 2002, 2013a, 2013b) in Fig. 9d–f. In the lower Olduvai polarity transition, the VGP did not move across the equator within a narrow longitudinal band, unlike in the Réunion subchron, but instead settled in several VGP cluster areas. The VGP can be observed in three areas (Fig. 9a): (A) the southern Indian Ocean, (B) North America, and (C) the southern South Pacific Ocean off South America. The VGP moved rapidly between these clusters. The RPI started to decline at 139.4 m immediately before the onset of the paleomagnetic directional change, where the VGP settled in cluster A in the southern hemisphere (Fig. 9b, c). Immediately following this, the RPI declined rapidly, reaching a minimum at 140.1 m, which corresponds to the VGP settling in cluster B in the northern hemisphere. Subsequently, the RPI gradually recovered to a peak at 144.2 m. During this recovery, the VGP settled in cluster C in the southern hemisphere at approximately 141 m, before moving back again to cluster B in the northern hemisphere.
Similar features to these, particularly for the asymmetric RPI profile consisting of a rapid decline and slow recovery and the locations of the VGP cluster areas, are well documented in the Site 984 record (Channell et al. 2002, 2013a, 2013b) shown in Fig. 9d–f. The cluster areas A, B, and C are observed in almost the same locations in both records. Similar VGP cluster areas have also been identified in lava records (e.g., Hoffman 1991, 1992) as well as other sedimentary records (e.g., Channell et al. 2002; Kusu et al. 2016). These cluster positions have been discussed to be related to the vertical flux patches in the present non-axial dipole (NAD) field (e.g., Channell et al. 2003, 2004; Hoffman et al. 2008). Kusu et al. (2016) reported that the VGP cluster areas in the upper Olduvai boundary were observed in similar locations to downward flux patches of the NAD, which occur in North America and the southern South Pacific, where the vertical component is less than − 15 μT according to the average geomagnetic field for the past 400 years (Constable 2007). The locations of these flux patches of the NAD correlate relatively well with our record, especially for cluster B in the Réunion subchron and clusters A and B in the lower Olduvai reversal, which suggests that the VGPs during geomagnetic reversals associated with low RPIs were strongly affected by the NAD field.