A full sequence of the Matuyama–Brunhes geomagnetic reversal in the Chiba composite section, Central Japan

Geological records of the Matuyama–Brunhes (M–B) geomagnetic reversal facilitate the development of an age model for sedimentary and volcanic sequences and help decipher the dynamics of the Earth’s magnetic field. However, the structure of the geomagnetic field during the M–B geomagnetic reversal remains controversial due to its complex field behavior. In this study, we conducted paleo- and rock-magnetic analyses of samples from the Chiba composite section (CbCS), a continuous and expanded marine succession in Central Japan, to reconstruct the full sequence of the M–B geomagnetic reversal. We define an average stratigraphic position of the M–B boundary and estimate its age based on three sections in the CbCS and a neighboring drill core, TB-2. The average stratigraphic position of the M–B boundary in the CbCS is established at 1.1 ± 0.3 m above a widespread volcanic ash bed (the Byk-E tephra). Assuming a chronological error associated with orbital tuning of 5 kyr and stratigraphic uncertainty of 0.4 kyr, the M–B boundary in CbCS is at 772.9 ± 5.4 ka (1σ). The virtual geomagnetic pole, which is calculated from the paleomagnetic directions, shows several short fluctuations between 783 and 763 ka, with concomitant decreases in geomagnetic field intensity index. After termination of the field instabilities, the field intensity recovered and became higher than before the M–B boundary, with a stable normal polarity direction. The paleomagnetic records in the CbCS exhibit a field asymmetry between the axial dipole decay and field recovery, providing a full sequence of the M–B reversal, suggesting that the non-axial dipole field dominated several times during periods ca. 20 kyr long across the M–B boundary, due to depletion in the main axial dipole component. Our results provide probably the most detailed sedimentary record of the M–B geomagnetic reversal and offer valuable information to further understand the mechanism and dynamics of geomagnetic reversals.


Introduction
The Matuyama-Brunhes (M-B) geomagnetic reversal is the last polarity reversal and one of the most studied paleomagnetic events (e.g., Opdyke et al. 1966;Harrison 1974). The mid-point of the polarity switch interval (the directional transition from a reversed polarity to normal polarity) in the reversal (M-B boundary) is a critical stratigraphic marker for calibrating the age of sedimentary and volcanic sequences (e.g., Channell et al. 2010;Suganuma et al. 2015;Singer et al. 2019) and for defining the Lower-Middle Pleistocene boundary (Head et al. 2008;Head and Gibbard 2015;Suganuma et al. 2018). The timing of the M-B boundary has been consistently constrained to ca. 773 ka in recent studies based on the astronomical age of rapidly accumulated marine sediments (e.g., Channell et al. 2010;Valet et al. 2014Valet et al. , 2019Suganuma et al. 2018) and radiometric dating of volcanic ash and lava-flows (e.g., Suganuma et al. 2015;Singer et al. 2019).
Geomagnetic field structure during the M-B geomagnetic reversal has also been intensively studied (e.g., Hartl and Tauxe 1996;Brown et al. 2004;Singer et al. 2005Singer et al. , 2019Raisbeck et al. 2006;Sagnotti et al. 2014Sagnotti et al. , 2016Balbas et al. 2018;Macri et al. 2018). Paleomagnetic records from several marine sediments and lavas, and numerical model simulations, show that the polarity switch spans several thousand years depending on site location (e.g., Channell and Guydo 2004;Channell et al. 2010;Clement 2004;Singer et al. 2005;Amit et al. 2010;Olson et al. 2011;). On the other hand, a remarkably short transition has been identified in exposed lacustrine sediments in the Sulmona Basin of the Central Apennines, Italy (Sagnotti et al. 2014(Sagnotti et al. , 2016 and marine sediments of the Valle di Manche section from Calabria, Southern Italy (Macri et al. 2018). In these studies, the polarity switch at the M-B boundary, dated at ca. 786 ka, was inferred to have completed within a century (Sagnotti et al. 2014;Macri et al. 2018). Prior to the M-B boundary, a clear decay in geomagnetic field intensity has been identified as a "precursor" event in deep-sea sediments (e.g., Hartl and Tauxe 1996;Channell et al. 2009), lavas (e.g., Brown et al. 2004;Singer et al. 2005), and ice (Raisbeck et al. 2006;Dreyfus et al. 2008). This precursor field decay is generally dated to ca. 795 ka and appears to be of the same magnitude as that at the M-B boundary (e.g., Hartl and Tauxe 1996;Gotton et al. 2007). Recent paleomagnetic records from highly precisely 40 Ar/ 39 Ardated lavas have indicated the long-term processes of the M-B geomagnetic reversal (Balbas et al. 2018;Singer et al. 2019). Based on these data, the field decay is thought to have begun ca. 30 kyrs before the polarity switch at ca. 773 ka, and involved centennial-tomillennial-scale field instabilities and a depletion of the axial dipole field (Balbas et al. 2018;Simon et al. 2018a;Singer et al. 2019). However, volcanic sequences fail to capture the entire M-B geomagnetic reversal record due to the nature of sporadic volcanic eruptions, instead, recording near-instantaneous geomagnetic signals. Sedimentary sequences provide continuous records, but geomagnetic recordings in slowly accumulated sediments are affected by vertical and temporal offsets between the sediment/water interface and the lock-in depth of at least 10-15 cm, the zone where the paleomagnetic record is fixed (e.g., Channell and Guydo 2004;Liu et al. 2008;Suganuma et al. 2010;Ménabréaz et al. 2012;Simon et al. 2018b). This may generate a smoothing effect due to progressively changing sedimentation-fabric processes (e.g., Suganuma et al. 2011;Roberts et al. 2013;Valet and Fournier 2016). Moreover, the sediment-recording smoothing process is amplified during paleomagnetic measurements due to the averaging effect of the magnetometer response window, when U-channels are used for the measurements (e.g., Oda and Xuan 2014;Philippe et al. 2018). These complex behaviors and the lack of truly continuousinstantaneous recordings have resulted in an incomplete picture of centennial-to-millennial-scale field behavior during the M-B geomagnetic reversal.
A detailed paleomagnetic record of the M-B geomagnetic reversal with a centennial-scale resolution was reported from the 130 m thick Chiba composite section in Central Japan ( Fig. 1; Suganuma et al. 2015;Okada et al. 2017;Simon et al. 2019). On January 2020, the Chiba section, the main portion of the Chiba composite section, was ratified by the Executive Committee of the International Union of Geological Sciences as a Global Boundary Stratotype Section and Point (GSSP) for the Middle Pleistocene Subseries (Chibanian Stage/Age) (Fig. 1c). The M-B boundary, the primary marker of the GSSP, lies 1.1 m above the Byk-E tephra bed and is dated to 772.9 ka (Suganuma et al. , 2018Okada et al. 2017;Simon et al. 2019). However, a complete sequence across the M-B geomagnetic reversal was not then available because of a lack of data for the upper 36 m portion of the Chiba composite section.
We report new paleomagnetic and rock-magnetic measurements for the upper part of the Chiba composite section. We also combine the results with those of Suganuma et al. (2015), Okada et al. (2017), and Simon et al. (2019), to reconstruct a full sequence of the M-B geomagnetic reversal in the interval between 794 and 748 ka, which covers the entire interval of depletion of the geomagnetic field intensity, including the polarity switch and preceding transitional phase. Additionally, new samples identify the M-B boundary from a sub-section of the Chiba composite section, to confirm correlations between its sub-sections. For the first time using chronological and stratigraphic uncertainties between the Chiba composite section Okada et al. 2017; our results) and a neighboring sediment core (Hyodo et al. 2016), we constrain the average age and stratigraphic position of the M-B boundary at the GSSP for the Chibanian Stage/Age and Middle Pleistocene Subseries/Subepoch.

Geological and stratigraphic settings
The Kazusa Group is the Pleistocene age marine succession exposed in the Boso Peninsula, Central Japan ( Fig.  1a, b), which has been investigated in a number of studies (e.g., Kanehara et al. 1949;Kawai 1951;Shinada et al. 1951;Mitsunashi 1954;Mitsunashi et al. 1959;Nakajima 1978). The group is characterized by excellent and Haneda et al. Progress in Earth and Planetary Science (2020)  continuous exposure and exceptionally high sedimentation rates (ca. 2 m/kyr) . The presence of many volcanic ash beds in the Boso Peninsula enables the reconstruction of a detailed stratigraphy (reviewed by Satoguchi and Nagahashi 2012;Fig. 1c). In addition, well-preserved marine microfossils (calcareous nannofossils, benthic and planktonic foraminifera, radiolaria, and diatoms) from the Kazusa Group have also been used to reconstruct biostratigraphies and a foraminiferal oxygen isotope stratigraphy (Aoki 1968;Oda 1977;Sato et al. 1988;Okada and Niitsuma 1989;Pickering et al. 1999;Cherepanova et al. 2002;Suganuma et al. 2018;Fig. 1c). These features provide a continuous, high-resolution chronostratigraphic framework. The Chiba composite section in the middle part of the Kokumoto Formation, Kazusa Group, comprises the Yanagawa, Urajiro, Yoro River (including the Chiba section, the GSSP of the Chibanian Stage/Age), Yoro-Tabuchi, and Kokusabata sub-sections (Figs. 2 and 3; Suganuma et al. 2018). A summary of the stratigraphic levels and locality of the Chiba composite section is shown in Table 1. The lithofacies of the middle part of the Chiba composite section is dominated by bioturbated silty beds with minor sandy beds, showing no evidence of episodic deposition and hiatus (Nishida et al. 2016), although thicker sandy beds are observable in the lowermost and upper parts (Fig. 3). Numerous tephra beds (Ku1, Ku2A, Ku2B, Tas, Tap, and Byk) occur in the Chiba composite section and demonstrate good correlation between sub-sections ( Fig. 3; Suganuma et al. 2018). The Byk-E tephra bed, originating from the Older Ontake volcano in Central Japan (Takeshita et al. 2016), was ratified as the base of the Chibanian Stage/Age.

Chronology of the Chiba composite section
The chronological framework for dating the Chiba composite section is provided by correlation between the benthic oxygen isotope stratigraphy (Suganuma et al. 2018) and a sea-level proxy curve obtained from Ocean Drilling Project (ODP) Site 1123, offshore New Zealand (Elderfield et al. 2012), assisted by a U-Pb radiometric age of the Byk-E tephra . The resulting age model reveals that the Chiba composite section corresponds to an age range from the late part of marine isotope stage (MIS) 20 (801.1 ka) to the early part of MIS 18 (747.6 ka) with the sedimentation rate varying between 0.44 and 3.87 m/kyr (Suganuma et al. 2018;Haneda et al. 2020). In the Chiba section, the M-B boundary is 1.1 m above the Byk-E tephra bed in an interval with a sedimentation rate of 89 cm/kyr Suganuma et al. 2018). The Byk-E tephra has a Th/U-corrected U-Pb zircon age of 772.7 ± 7.2 ka  and an astronomical age of 774.1 ka, which forms the basal age of the Chibanian Stage (Suganuma et al. 2018).

Previous paleomagnetic studies in the Kokumoto Formation
The M-B boundary was identified in the middle part of the Kokumoto Formation of the Kazusa Group in the central Boso Peninsula by Nakagawa et al. (1969) and Niitsuma (1971Niitsuma ( , 1976 (Fig. 1c). These pioneering studies identified the M-B boundary approximately 2 m below the Byk-E tephra bed using alternating field (AF) demagnetization, which removes secondary remanences by exposing a specimen to an alternating magnetic field (Niitsuma 1971;Okada and Niitsuma 1989;Tsunakawa et al. 1995Tsunakawa et al. , 1999Aida 1997).
Three paleomagnetic studies for the Chiba composite section have provided detailed records of geomagnetic field behavior across the M-B boundary Okada et al. 2017;Simon et al. 2019). Suganuma et al. (2015) performed thermal demagnetization, which removes secondary remanence by heating and cooling specimens in a zero magnetic field. This was using specimens from the Chiba and Yanagawa sections between − 6.55 and 5.75 m (indicated by blue horizontal bars in Fig.  3; Additional file 1: Table S1). The results of the thermal demagnetization indicate the M-B boundary at 0.8 m above the Byk-E tephra bed at the Yanagawa section-not below the tephra bed as reported in earlier studies (Fig. 3). This difference to previous studies was interpreted as due to secondary remanence in the specimens, resulting from high coercivity magnetic minerals, that were not removed by AF demagnetization in the early studies. However, relative paleointensity could not be reconstructed, because minerals in the specimens underwent heat-alteration during the thermal demagnetization process, producing new magnetic phases . Okada et al. (2017) conducted demagnetization using a hybrid method of thermal demagnetization at 300°C followed by progressive AF demagnetization for specimens from the Chiba and Yoro-Tabuchi sections between − 14.6 and 9.7 m (indicated by black horizontal bars in Fig. 3; Additional file 2: Table S2). This hybrid demagnetization method removes the secondary (See figure on previous page.) Fig. 1 Location map of the study sites and stratigraphy of the Kazusa Group. a Map of the Japanese archipelago. b Distribution of marine successions on the Boso Peninsula and the location of the Chiba composite section. The black square indicates the area of the geological map in Fig. 2. c Stratigraphy of the Kazusa Group after Kazaoka et al. (2015): (i) Niitsuma (1976), (ii) Okada and Niitsuma (1989), (iii) Pickering et al. (1999), and (iv) Tsuji et al. (2005)  remanence, which cannot be removed by AF demagnetization alone, importantly with minimum heat-alteration, therefore allowing reconstruction of relative paleointensity (such a procedure is widely used for similar reasons in many other sedimentbased studies). Based on this hybrid approach, the M-B boundary was observed, with a depleted geomagnetic field intensity at 1.1 m above the Byk-E tephra bed in the Chiba section (Fig. 3). The redefined stratigraphic position of the M-B boundary is supported by paleomagnetic data from the TB-2 sediment core drilled 190 m northeast of the Chiba section (Hyodo et al. 2016;Figs. 3 and 4). However, the recovery phase of the geomagnetic field (after the  Table S2) and beryllium isotope measurements of samples from the Chiba and Yoro-Tabuchi sections up to 37.5 m (indicated by pink squares in Fig. 3). Based on their measurements, the paleomagnetic direction and relative paleointensity records were obtained, along with a beryllium isotope record ( 10 Be, 9 Be, and 10 Be/ 9 Be ratio), as independent proxies of the geomagnetic field intensity. These authors suggested that the M-B boundary involved geomagnetic dipole moment collapse and observed a sequence of rapid directional oscillations between the M-B boundary and the later field recovery phase, in which a stable dipole is reestablished on the opposite direction. The directional oscillations were identified as instability clusters (IC1 = 771.9-768.5 ka, IC2 = 768.5-765.2 ka, and IC3 = 763.5-762.5 ka) by Simon et al. (2019).
These previous paleomagnetic studies of the Chiba composite section have reconstructed a centennialresolution paleomagnetic record from the stratigraphic interval between -14.6 and 37.5 m from the Byk-E tephra bed. Exceptionally high sedimentation rates of the Chiba composite section are thought to limit the lock-in depth and smoothing effect of the geomagnetic field recording process (e.g., Suganuma et al. 2011;Valet and Fournier 2016). Therefore, the sedimentary succession documents short-duration instabilities of the geomagnetic field that are almost simultaneous with deposition. However, paleoand rock-magnetic records from the strata above 37.5 m have not yet been obtained. In addition, although the M-B boundary is located 1.1 m above the Byk-E tephra bed in the Chiba section, this boundary has not been examined in the Yoro-Tabuchi section, 70 m northeast of the Chiba section (Figs. 2c and 4).

Paleomagnetic core samples
We collected 25 mm-diameter drill core samples from the upper part of the Yoro River section and the lowermost and middle parts of the Yoro-Tabuchi section using a portable drill (Figs. 3,4,and 5). For the Yoro River section (above 37.5 m), the drill cores were collected from 64 levels with a ca. 50 cm spacing (the same interval as used for deriving the foraminiferal oxygen isotope stratigraphy of Suganuma et al. 2018 andHaneda et al. 2020). The locations of the collected samples are indicated by red horizontal bars in Fig. 3 and red letters in Fig. 5. In the lowermost (0 m to 3.8 m) and middle (11.6 m to 13.4 m) part of the Yoro-Tabuchi section, drill cores were collected from 34 levels and 9 levels with a ca. 10 to 20 cm spacing, as indicated by red horizontal bars in Fig. 3 and red letters in Fig. 4. These samples were oriented with a magnetic compass. Each core was cut into 2 cm-long specimens for subsequent paleomagnetic and rockmagnetic measurements at Ibaraki University and National Institute of Polar Research (NIPR).

Paleomagnetic measurements
The natural remanent magnetization (NRM) for each specimen was measured using a three-axis cryogenic magnetometer SRM-760R (2G Enterprises, USA) in a magnetically shielded room at the NIPR. To extract the primary magnetic component, we removed the secondary remanence from the NRMs using a hybrid of thermal demagnetization to 300°C and AF demagnetization ). The thermal demagnetization was conducted using a TDS-1 (Natsuhara Giken, Japan) at the NIPR. The progressive AF demagnetization was performed in 5 to 10 mT increments between 5 and 20 mT before the thermal demagnetization and between 20 and 80 mT after the thermal demagnetization, using static threeaxis AF demagnetization coils in an SRM-760R magnetometer at the NIPR.
We identified the characteristic remanent magnetization (ChRM) at peak AF fields between 30 to 50 mT after the thermal demagnetization at 300°C Simon et al. 2019) and calculated paleomagnetic directions by principal component analysis (Kirschvink 1980) using these step points with anchoring to the origin. The virtual geomagnetic pole (VGP) was calculated using the ChRM directions with

Rock-magnetic measurements
To determine rock-magnetic properties and calculate the relative paleointensity indices, we measured low-field magnetic susceptibility (k LF ), laboratory-induced magnetizations of anhysteretic remanent magnetization (ARM), and isothermal remanent magnetization (IRM). Measurements of the k LF were performed on specimens using a KLY-3S Kappabridge susceptibility meter (AGICO, Czech Republic) at Ibaraki University, before any paleomagnetic or rock-magnetic measurements were made. For specimens whose NRM was demagnetized by the hybrid method, ARM was acquired in a 30 μT DC field with an 80 mT alternating field using the SRM-760R magnetometer at the NIPR. In each case, after ARM acquisition, we repeated the hybrid demagnetization (thermal demagnetization at 300°C and AF demagnetization) at 10 mT increments between 20 and 80 mT. ARM susceptibility (k ARM ) was calculated from the initial ARM intensity. k ARM is mainly preferentially acquired by small magnetic grains with weak magnetostatic interaction. On the other hand, k LF mainly reflects the total amount of magnetic minerals in a specimen, irrespective of the magnetic grain size. Therefore, the k ARM /k LF ratio is a conventional proxy for the grain size of magnetic particles in a specimen; an increase in the k ARM /k LF ratio indicates the relative contribution of finer magnetic particles and vice versa (e.g., Yamazaki 2008;Yamazaki and Ikehara 2012;Shin et al. 2019).
After the ARM experiments, we exposed selected specimens to saturated IRM (SIRM) in a 1.5 T field using a MMPM-9 pulse magnetizer (Magnetic Measurements Ltd., UK). The SIRM was measured using a JR-6A magnetometer (AGICO, Czech Republic) and demagnetized using a TD-48 thermal demagnetizer (ASC Scientific, USA) and AF demagnetizer DEM-8601C (Natsuhara-Giken, Japan) at Ibaraki University. Results from the hybrid demagnetization method used on the ARM, along with the SIRM, were used to calculate relative paleointensity indices as outlined below.
To determine S-ratios, IRM acquisition was performed on unprocessed specimens. S-ratios indicate the relative abundance of antiferromagnetic minerals (such magnetically hard hematite and goethite) in a mixture with ferrimagnetic minerals (such as magnetically soft magnetite and maghemite); values close to unity indicate that the remanence is dominated by magnetically soft ferrimagnets and vice versa (Bloemendal et al. 1992;Evans and Heller 2003). We imparted SIRM on selected specimens from the Yoro River and Yoro-Tabuchi sections above 9.7 m using a 1.5 T field using the MMPM-9 pulse magnetizer (Magnetic Measurements Ltd., UK). The IRM was then measured using the JR-6A magnetometer (AGICO, Czech Republic) at Ibaraki University. IRMs at 0.1 T and 0.3 T fields were acquired in the opposite direction of the SIRM. The S-ratio -0.1T and Sratio -0.3T were calculated as according to Bloemendal et al. (1992).

Relative paleointensity
Proxies of magnetic grain concentration, k LF , ARM, or IRM, are usually employed to normalize NRM when calculating the relative paleointensity index (e.g., Kent and Opdyke 1977;Tauxe 1993;Tauxe and Shackleton 1994;Guyodo and Valet 1999;Yamazaki 1999;Valet 2003;Yamazaki and Kanamatsu 2007;Suganuma et al. 2008). Previous studies of the Chiba composite section calculated the relative paleointensity by normalizing the vector difference of the NRM between 30 mT and 50 mT (after thermal demagnetization at 300°C) with the ARM intensity removed between the same demagnetization steps after thermal demagnetization at 300°C Simon et al. 2019). However, the pseudo-Thellier approach of Tauxe et al. (1995) or other related method has been widely employed to deduce the paleointensity. In the approach of Tauxe et al. (1995), the slope of the NRM intensity left at a certain demagnetization step normalized against ARM or IRM intensity gain at the same imparted field step was used as the paleointensity proxy. In practice, the slope of the NRM intensity left at selected demagnetization steps against the ARM or IRM left at the same demagnetization steps has been more widely used (e.g., Channell et al. 2002;Valet et al. 2014). Moreover, normalization by ARM may over-compensate magnetic concentration changes because ARM can be sensitive to magnetostatic interaction among grains (Yamazaki and Kanamatsu 2007). As a result, the efficiency of ARM acquisition in specimens can decrease with an increase in magnetic concentration.
Therefore, we normalized the NRM against the ARM (or IRM) as NRM 30-50 /ARM 30-50 (or NRM 30-50 /IRM 30-50 ) from NRM lost versus ARM (or IRM) lost between the 30 and 50 mT demagnetization steps (after thermal demagnetization at 300°C). We employed these ratios as relative paleointensity indices. k LF was not used as a normalizer in this study, because it may be less effective due to possible influence from diamagnetic and paramagnetic minerals and large ferrimagnetic grains that do not carry NRM (Shin et al. 2019). The Pearson linear correlation coefficient (R-value ARM and R-value IRM ), associated with the slope of the NRM intensity left at certain demagnetization steps against the ARM or IRM left at the same demagnetization steps, was calculated to monitor the goodness of the slope fit (Channell et al. 2002;Shin et al. 2019). A comparison of the proxies used for relative paleointensity between this study and the previous studies Simon et al. 2019) is in Figure S1.

Remanent magnetization
Representative orthogonal vector diagrams (Zijderveld 1967) of the NRMs are shown in Fig. 6. Some of the data for intermediate and reversed polarities before and after the M-B boundary are taken from Okada et al. (2017) and Simon et al. (2019) (Fig. 6c, d, h, and i). The magnetization from the Chiba composite section consists of one or more components (Fig. 6). The component before thermal demagnetization at 300°C generally is the secondary component. After thermal demagnetization at 300°C, several specimens showed a ChRM between 30 and 50 mT, which decays toward the origin (Fig. 6b-i). Okada et al. (2017) revealed that some samples from the Chiba composite section became unstable to thermal demagnetization at temperatures above 300°C, and the following AF demagnetization between 30 and 50 mT appeared to reflect-well the ChRM decay towards the origin. Because the samples have a single Curie/Néel temperature at 580°C Okada et al. 2017), the component between 30 and 50 mT is thought to be carried by magnetite. Based on the result of Suganuma et al. (2015) and Okada et al. (2017), also  ). Therefore, we calculated paleomagnetic directions of the ChRM between the 30 and 50 mT steps, following the thermal demagnetization at 300°C.

Rock-magnetic properties and paleomagnetic directions in the Yoro River section
The rock magnetic properties in the upper part of the Yoro River section (indicated by red and orange circles in Fig. 7; Additional file 2: Table S2) are relatively homogenous as are those from lower in the Chiba composite section. Proxies of magnetic grain concentration (k LF , k ARM ) show simultaneous and minor fluctuations by a factor of < 10 throughout the Chiba composite section (Fig. 7k). The k ARM /k LF ratio-a proxy of magnetic grain size-varies between 1.5 and 4.4 throughout the section (Fig. 7l). The S-ratio -0.3T , which represents the relative abundance of antiferromagnetic minerals in a mixture with ferrimagnetic minerals, fluctuates moderately throughout the Chiba composite section except in the Yoro River section above 67 m (Fig. 7 m). The average S-ratio -0.3T of 0.96 indicates that the magnetic grains throughout the Chiba composite section mostly consist of ferrimagnetic minerals (Frank and Nowaczyk 2008).
While the S-ratio -0.1T fluctuates between 0.66 and 0.90 in the Chiba and lower Yoro-Tabuchi sections , the fluctuation amplitude is suppressed in the upper parts of these sections (Fig. 7m). The periodic decrease of the S-ratio -0.1T in the Yoro-Tabuchi and Chiba sections is thought to be not attributed to the increase in the proportion of hematite or goethite, but mainly to the increase (decrease) in the amount of fine (coarse) size magnetic grain, because of its cooccurrence with changes in k ARM /k LF (Yamazaki et al. 2003;Suganuma et al. 2008;Okada et al. 2017). This interpretation is supported by the high average value of the S-ratio -0.3T (0.96), indicating the dominance of ferrimagnetic minerals throughout the Chiba composite section (Frank and Nowaczyk 2008; Fig. 7m). Additionally, the amplitude of the k ARM /k LF changes is limited to a factor of < 3. These data indicate that the magnetic properties throughout the Chiba composite section are generally uniform, although minor grain size changes occur in the lower interval. Therefore, the rockmagnetic data of the Chiba composite section fulfill the established criteria for the construction of a robust relative paleointensity record with respect to magnetic grain concentration, magnetic grain size, and magnetic mineralogy (Tauxe 1993). However, the S-ratio -0.3T and Sratio -0.1T in the Yoro River section above the 67 m level decrease to 0.86 and 0.46, respectively (Fig. 7m). These decreases are accompanied by decreases in the relative paleointensity indices, linear correlation coefficients (Rvalue ARM and R-value IRM ), NRM intensity, and proxies of magnetic concentration (Fig. 7e-h, j-k, and m). No distinct lithological change was observed in this stratigraphic interval in the field. As the interval above 67 m corresponds to just after the full glacial interval of MIS 18, in the oxygen isotope stratigraphy (Fig. 7b; Haneda et al. 2020), the change in the magnetic properties may have originated from local environmental change (e.g., supplies of terrestrial and/or organic matter to the basin, hydrographic change in the bottom water, sea level change, oceanic circulation changes, or oxygenation). The indices above the 67 m level may not represent realistic variation in geomagnetic relative intensity. Therefore, we used an S-ratio -0.3T of less than 0.9 as a cut-off limit for useful relative paleointensity indices, so the indices above the 67 m level are not used in the following discussions.
The VGP latitude profile above the 38 m level (indicated by red open circles in Fig. 7d) exhibits stable normal polarity with three intermediate polarities at 42.4, 54.0, and 69.2 m (indicated by gray arrows in Fig. 7d). The NRM 30-50 /ARM 30-50 ratio above the 38 m level is generally higher than the overall average of 0.08 and widely fluctuates between 0.04 and 0.29 in this interval (Fig. 7e). In comparison, the VGP latitude and rockmagnetic properties remain relatively stable.
The NRM 30-50 /IRM 30-50 profile was also obtained from selected specimens obtained from the Chiba composite section (Fig. 7g). This ratio shows similar variations to the NRM 30-50 /ARM 30-50 ratio with the minimum value (at 1.9 m) co-occurring with the M-B boundary, a slight increase between 6 and 13 m, and a recovery phase to full intensity at 15-25 m. The high linear correlation coefficient between NRM 30-50 /ARM 30-50 and NRM 30-50 / IRM 30-50 of 0.94 indicates no systematic differences between the relative paleointensity indices obtained from the Chiba composite section.
We calculated the correlations between the relative paleointensity indices and the normalizers to test for independence. The linear correlation coefficients for ARM and NRM 30-50 /ARM 30-50 , and SIRM and NRM 30-50 / IRM 30-50 are 0.44 and 0.52, respectively (Additional file 4: Figure S2a-b). These weak positive correlations imply that the relative paleointensity indices could be partially contaminated by the normalizers. On the other hand, the linear correlation coefficients for authigenic 10 Be/ 9 Be versus NRM 30-50 /ARM 30-50 and NRM 30-50 /IRM 30-50 are − 0.72 and − 0.70, respectively (Additional file 4: Figure  S2c-d). Because the cosmogenic nuclide 10 Be production is mainly modulated by variations in magnetospheric shielding on a millennial time scale, 10 Be concentration records from marine sediments constitute an independent proxy of the geomagnetic dipole moment, providing that secondary depositional processes have been removed through a normalization procedure (e.g., Simon et al. 2016). The similarity of NRM 30-50 /ARM 30-50 and NRM 30-50 /IRM 30-50 ratios with authigenic 10 Be/ 9 Be demonstrates that the relative paleointensity from the Chiba composite section is reliable proxies for the geomagnetic field intensity variation. Therefore, we used the highresolution NRM 30-50 /ARM 30-50 ratio as the best relative paleointensity index.

Paleomagnetic directions of the lowermost part of the Yoro-Tabuchi section
Paleomagnetic directions of the lowermost part of the Yoro-Tabuchi section were obtained from 16 levels from specimens in which the ChRM component MAD was ≤ 15°(Additional file 5: Table S3). The polarity switch from a reversed polarity (VGP latitude < − 45°) to a normal polarity (VGP latitude > 45°) occurs in the stratigraphic interval between 1.1 and 2.1 m. The M-B boundary, defined as a middle of the polarity switch interval, is located 1.6 m above the Byk-E tephra bed in the Yoro-Tabuchi section ( Fig. 8; Table  1). After the polarity switch is completed, the VGP latitude largely rebounds across the Equator. The stratigraphic position of the M-B boundary and subsequent directional instability are consistent with previous paleomagnetic data from the Yanagawa section , the Chiba section , and the TB-2 sediment core (Hyodo et al. 2016). b Benthic oxygen isotope stratigraphy . c Maximum angular deviation (MAD) with the 15°cut-off limit indicated, for the principal component analysis. d VGP latitude calculated from the results of the hybrid demagnetization method. e-h NRM 30-50 /ARM 30-50 and NRM 30-50 /IRM 30-50 as relative paleointensity indices calculated following the pseudo-Thellier approach of Tauxe et al. (1995), and linear correlation coefficients (R-value ARM and R-value IRM ) of the indices. i Authigenic 10 Be/ 9 Be  as an independent proxy of the field intensity. j Natural remanent magnetization intensity. k Low-field (k LF ) and ARM magnetic susceptibility (k ARM ) as proxies of magnetic grain concentration. l k ARM /k LF ratio as a proxy of magnetic grain size. m S-ratio at − 0.3 T and − 0.1 T as a proxy of the relative abundance of antiferromagnetic to ferrimagnetic minerals. New data points obtained from the Yoro-Tabuchi section are shown as blue and light blue open circles and lines, and new data points from the Yoro River section are shown as red and orange open circles and lines. Data points from previous studies Simon et al. 2019) are shown as black and gray closed circles and lines. The assignment of marine isotope stages is after Haneda et al. (2020). Instability clusters (IC1-IC3) are after Simon et al. (2019) Haneda et al. Progress in Earth andPlanetary Science (2020)  The NRM 30-50 /ARM 30-50 ratio as a relative paleointensity index was obtained from the 16 specimens in which the ChRM component MAD was ≤ 15° (Fig. 7e-f; Additional file 5: Table S3). The depletion of the relative paleointensity occurs 0.2 m above the Byk-E tephra bed, although the values are already lower than the average for the whole Chiba composite section (Fig. 7e). The interval of minimum relative paleointensity is observed from 3.0 m above the Byk-E tephra bed and is followed by subsequent directional instabilities, a phenomenon also observed in the Chiba section .

Comparison of the M-B boundary in the Chiba composite section
In this study, the M-B boundary is defined as a midpoint of the polarity switching interval from a reversed polarity (VGP latitude < − 45°) to a normal polarity (VGP latitude > 45°). An alternative definition is the stratigraphic position at which the VGP latitude = 0. However, using this definition, the M-B boundary is identified using only two intermediate polarity data points, in the case of the Chiba composite section. Therefore, we adopt the former definition of the M-B boundary, which is more robust.
The M-B boundary has been identified in the Yanagawa , Chiba , and Yoro-Tabuchi (this study) sections of the Chiba composite section, and in the TB-2 sediment core (Hyodo et al. 2016) (Fig. 8; Table 1). The depth scale of the TB-2 sediment core was converted into the equivalent thickness scale of the Chiba composite section, using 7 tie points based on lithology and tephrostratigraphy (indicated by red solid lines in Fig. 3). The average polarity switch occurs in an interval between 0.6 ± 0.3 and 1.6 ± 0.4 m (1σ) above the Byk-E tephra bed ( Table 2). The average midpoint of the polarity switch in the three sections and sediment core is 1.1 ± 0.3 m (1σ) above the Byk-E tephra bed (Table 2). Based on the age model of Suganuma et al. (2018), the M-B boundary is dated at 772.9 ± 0.4 ka (1σ) and the polarity switch is completed within 1.1 ± 0.4 kyrs (1σ), although the field intensity minimum and directional instability follow it ( Fig. 8; Simon et al. 2019). Assuming a potentially chronological uncertainty of 5 ka in orbital tuning methods Valet et al. 2019), the average age of the M-B boundary in the Chiba composite section is estimated as 772.9 ± 5.4 ka (1σ). This is consistent with the inferred ages obtained from sedimentary archives (e.g., Channell et al. 2010;Valet et al. 2019).

Millennial-scale VGP deviation and its distribution
The time evolution of the VGP latitude and the VGP locations in the Chiba composite section are shown in Fig. 8 Comparison of the M-B boundary observed from the Chiba composite section and the TB-2 sediment core. VGP latitudes based on thermal demagnetization in Suganuma et al. (2015) are shown as blue cross marks. VGP latitudes based on the hybrid demagnetization method are shown as black closed circles and lines ) and red open circles and lines (this study). The TB-2 sediment core was drilled 190 m northeast of the Chiba section, and VGPs obtained from the core as reported by Hyodo et al. (2016) are shown as blue closed circles and lines (AFD) and red closed circles and lines (ThD). Lithology of the TB-2 sediment core was from Hyodo et al. (2016Hyodo et al. ( , 2017. Blue, black, and red horizontal bars indicate the sampling level of Suganuma et al. (2015), Okada et al. (2017), and this study, respectively. Black and red dashed lines show the mid-thickness of polarity switch indicating the M-B boundary and the correlation of the volcanic ash Figs. 9, 10, and 11 alongside the relative paleointensity index (NRM 30-50 /ARM 30-50 ) and Be ratio-derived dipole moment (BeDM) profiles. Each temporal profile was calculated based on the latest age model of the Chiba composite section by Suganuma et al. (2018). The longest section record shown in Fig. 9 is integrated with data from the Chiba composite section and new results from the Yoro River section above 37.5 m, which were obtained using the hybrid demagnetization method. The new record from the lowermost part of the Yoro-Tabuchi section spans between 774.1 and 769.9 ka, which corresponds to the period from the polarity switch to the early phase of the directional instability (Fig. 10). The results of the progressive thermal demagnetization from the Chiba and Yanagawa sections show the VGP path and the VGP latitude profile between 787.1 and 767.7 ka ( Fig. 11; Suganuma et al. 2015). The VGP rebounds, which are large deviations from the north pole, after the M-B boundary are directional instability clusters (IC1, IC2, and IC3 in Figs. 9, 10, and 11;Simon et al. 2019).
The VGP latitude profiles show millennial-scale fluctuations with intermittent depletions in relative paleointensity during periods where the BeDM was lower than its modern level ( Fig. 9; Simon et al. 2019). The VGPs in the Chiba composite section prior to the M-B boundary intermittently deviate at ca. 783, 780, and 775 ka and mainly plot in the southeastern Pacific (Figs. 9 and 11). The VGPs then mainly shift from the Southern Hemisphere to Northern Hemisphere at ca. 773 ka without the longitudinal preference of the VGP path suggested by Clement (1991) and Laj et al. (1991). The VGPs of the directional instability clusters are mainly located in North America and the northeastern Pacific and are scattered across Eurasia and the western Pacific (Fig. 9, 10, and 11). After the termination of IC3 at ca. 763 ka, the VGPs is mostly stable north of 70°N.
Millennial-scale geomagnetic field instability has been identified as a characteristic feature of the geomagnetic reversal process (e.g., Valet et al. 2012;Balbas et al. 2018;Singer et al. 2019). Valet et al. (2012) proposed the presence of three successive phases in both reversednormal and normal-reversed transitions during the last 180 Ma. Based on ten volcanic records, they suggested phasing of (i) a precursor event, (ii) a 180°polarity switch, and (iii) a rebound. Based on accurate 40 Ar/ 39 Ar dating of lava flows, Singer et al. (2019) have argued that the M-B geomagnetic reversal began in association with a large VGP deviation and field depletion at ca. 795 ka, a process recording the dominance of a non-axial dipole field, due to the collapse of the axial dipole. The interval of VGP fluctuation and depletion of the relative paleointensity at 783 ka in the Chiba composite section are consistent with the geomagnetic field instability recorded in lava flows in Chile and Guadeloupe at ca. 784 ka (Fig. 12;Singer et al. 2019). The VGP at ODP sites 983 and 984 in the North Atlantic shows repeatedly large fluctuation, which begins from ca. 780 and 783 ka, respectively ( Fig. 12b; Channell and Guydo 2004). These VGP fluctuations observed in North Atlantic cores may correspond to the VGP fluctuations in the Chiba composite section seen from 783 ka (Fig. 12). Simon et al. (2019) reported the succession of large VGP deviations and low BeDM between 773.9 and 768.5 ka in the Chiba composite section corresponding to the interval between the polarity switch and IC1. The large VGP deviations just after the M-B boundary agree with the polarity switch and subsequent rebound phases of the successive reversal as modeled by Valet et al. (2012)  ). The directional instability associated with the geomagnetic reversal terminated ca. 3 kyr prior to the field recovery at ca. 760 ka. After the field recovery, following the termination of IC3, the VGP profile is generally stable, showing only occasional deviations. Because these deviations consist of single data points and are not accompanied by large decreases in the relative paleointensity, it is reasonable to consider them as noise. The relative paleointensity after the field recovery is higher than that before the field decay, although the rock magnetic properties are relatively stable (Figs. 7 and 12). This field asymmetry across the M-B boundary is consistent with the virtual axial dipole moment series from the relative paleointensity global stacks ( Fig. 12c; Valet et al. 2005;Channell et al. 2009). The paleomagnetic record of the Chiba composite section shows the full sequence of the M-B geomagnetic reversal between 783 and 763 ka, indicating the dominant contribution of the non-axial dipole field, during the depletion of the main axial dipole, at ca. 20 kyrs interval across the M-B boundary at ca. 773 ka.

Conclusions
Paleomagnetic and rock magnetic study of the Chiba composite section located in the Boso Peninsula, Central Japan, allows reconstruction of the full sequence of the M-B geomagnetic reversal in the interval between 794 and 748 ka. This covers the entire interval of depletion of the geomagnetic field intensity, including the polarity switch and preceding transitional phase. We have defined an average stratigraphic position and age of the M-B boundary between three (sub) sections in the Chiba composite section and a neighboring drilled core based on newly obtained and previously reported data.
The VGP latitude and relative paleointensity results show a clearly defined polarity switch, with a geomagnetic field minimum, 1.6 m above the Byk-E tephra bed in the Yoro-Tabuchi section. This is consistent with the previously reported paleomagnetic data from the Chiba composite section. The average stratigraphic position of the M-B boundary between the Chiba, Yoro-Tabuchi, and Yanagawa sections, and the TB-2 sediment core, lies 1.1 ± 0.3 m (1σ) above the Byk-E tephra bed. Accounting for uncertainty due to a 5 kyr chronological error with the orbital tuning and a stratigraphic deviation of 0.4 kyr, the age of the M-B boundary in the Chiba composite section is estimated at 772.9 ± 5.4 ka (1σ). The results from the upper part of the Yoro River section exhibit homogeneous rock-magnetic properties and