Authigenic 10Be/9Be ratio signatures of the cosmogenic nuclide production linked to geomagnetic dipole moment variation since the Brunhes/Matuyama boundary.

Geomagnetic dipole moment variations associated with polarity reversals and excursions are expressed by large changes of the cosmogenic nuclide beryllium-10 (10Be) production rates. Authigenic 10Be/9Be ratios (proxy of atmospheric 10Be production) from oceanic cores therefore complete the classical information derived from relative paleointensity (RPI) records. This study presents new authigenic 10Be/9Be ratio results obtained from cores MD05-2920 and MD05-2930 collected in the west equatorial Pacific Ocean. Be ratios from cores MD05-2920, MD05-2930 and MD90-0961 have been stacked and averaged. Variations of the authigenic 10Be/9Be ratio are analyzed and compared with the geomagnetic dipole low series reported from global RPI stacks. The largest 10Be overproduction episodes are related to dipole field collapses (below a threshold of 2 × 1022 Am2) associated with the Brunhes/Matuyama reversal, the Laschamp (41 ka) excursion, and the Iceland Basin event (190 ka). Other significant 10Be production peaks are correlated to geomagnetic excursions reported in literature. The record was then calibrated by using absolute dipole moment values drawn from the Geomagia and Pint paleointensity value databases. The 10Be-derived geomagnetic dipole moment record, independent from sedimentary paleomagnetic data, covers the Brunhes-Matuyama transition and the whole Brunhes Chron. It provides new and complementary data on the amplitude and timing of millennial-scale geomagnetic dipole moment variations and particularly on dipole moment collapses triggering polarity instabilities.

Introduction

Knowledge of past geomagnetic dipole moment (GDM) variation is required to understand the past and present geodynamo regimes and anticipate future changes. Namely, the amplitudes and timing of these variations are the keys for understanding the underlying physical processes of dipole field instabilities [Hulot et al., 2010]. Paleomagnetic investigations of sediments and lavas provide information on past variations of the dipole field [e.g., Tauxe, 1993; Valet, 2003], and it is nowadays generally accepted that polarity reversals and excursions are associated with dipole field wanings [e.g., Laj and Channell, 2015; Valet and Fournier, 2016]. These studies yield continuous records of the relative paleointensity (RPI) that are stacked and averaged to produce global RPI records. The latter can be calibrated by using absolute paleointensities from lava‐flow records and transformed into virtual axial dipole moment (VADM) records. However, depositional and postdepositional remanent magnetization (DRM and pDRM) acquisition processes in sediments may introduce biases, hampering accurate geomagnetic interpretations [e.g., Tauxe et al., 2006; Roberts et al., 2013; Valet and Fournier, 2016]. Some discrepancies linked to such environmental biases are partially resolved by stacking and averaging individual records [e.g., Guyodo and Valet, 1996, 1999; Laj et al., 2000, 2004; Stoner et al., 2002; Thouveny et al., 2004; Yamazaki and Oda, 2002; Valet et al., 2005; Channell et al., 2009; Ziegler et al., 2011]. These regional and/or global RPI stacks document millennial‐ to million‐year‐scale dipole moment variations and are widely used for correlation and establishment of chronologies. However, smoothing effects inherent to stacking as well as potential distortions due to magnetization acquisition (smoothing and delays) cast doubt about time series analyses performed to extract frequencies of long‐term geomagnetic variations. For these reasons few teams of geophysicists and geochemists have initiated studies of cosmogenic isotopes over geological archives in order to provide complementary data sets on variations of the geomagnetic moment.

The cosmogenic nuclide beryllium‐10 (10Be) production in the stratosphere (~65%) and troposphere (~35%) is triggered by spallation reactions between highly energetic primary (mainly protons (H+) and alpha particles (4He2+)) and secondary (mainly neutrons) galactic cosmic ray particles and atmospheric nitrogen and oxygen atoms [Lal and Peters, 1967; Dunai and Lifton, 2014]. The penetration rate of the primary galactic cosmic ray particles into the atmosphere varies strongly with latitude as a function of the vertical cutoff rigidity, which quantifies the ability of charged particles to penetrate the geomagnetic field lines. Thus, increased 10Be production results from the deficient magnetospheric shielding due to weak GDM strength. Given a nearly constant galactic cosmic ray intensity assumption [Vogt et al., 1990], the factors modulating the flux of incoming primary cosmic ray particles, and then the production of cosmogenic nuclide 10Be, are directly related to solar activity and Earth magnetic field variability. Previous studies have demonstrated that—while solar activity inflects the production rate on short (decadal to centennial) timescales—10Be flux measured along ice and marine sediment sequences reflects large‐amplitude variations of the geomagnetic dipole moment at the millennial scale [e.g., Raisbeck et al., 1981; Yiou et al., 1985; Frank et al., 1997; Wagner et al., 2000; Carcaillet et al., 2004b; Muscheler et al., 2005; Ménabréaz et al., 2012, 2014]. After its production, the reactive 10Be2+ is quickly removed from the atmosphere, i.e., 1–3 years [Raisbeck et al., 1981; Beer et al., 1990; Baroni et al., 2011], and rapidly scavenged from the water column with typical residence time of ~100 years at ocean margins [Anderson et al., 1990]. Recent studies on marine and glacial records together with simulation using general circulation models provide conclusive arguments for a rapid global atmospheric mixing of 10Be before deposition [Ménabréaz et al., 2012; Heikkilä et al., 2009, 2013]. Altogether, the 10Be time response is significantly more rapid than GDM millennial‐scale fluctuations supporting the use of 10Be to estimate precise GDM timing.

The inverse relationship between the geomagnetic dipole moment and the atmospheric production rate of 10Be was first established by Elsasser et al. [1956] and progressively improved [Lal, 1988; Beer et al., 1990; Masarik and Beer, 1999, 2009; Kovaltsov and Usoskin, 2010]. It provides the approach, independent from paleomagnetism, to decipher past GDM variations from marine sedimentary sequences from which paleomagnetic information can also be extracted [e.g., Carcaillet et al., 2003, 2004b].

In this study, we present new authigenic 10Be/9Be ratio results obtained from cores MD05‐2920 (covering the time interval 20–386 ka) and MD05‐2930 (5–250 ka) collected in the west equatorial Pacific Ocean. These data are merged with a revised data set from cores MD05‐2930 (250–790 ka) [Ménabréaz et al., 2014] and MD90‐0961 (686–873 ka) [Valet et al., 2014]. The resulting stacked and averaged record provides the first nearly continuous sedimentary record of 10Be production rate covering the Brunhes Chron.

Materials and Methods

Core Description and Environmental Settings

Cores MD05‐2920 and MD05‐2930 were collected by using the giant piston corer CALYPSO during the MD‐148 PECTEN IMAGES XIII Cruise aboard the R/V Marion Dufresne in 2005 [Beaufort et al., 2005]. The core site MD05‐2920 (2.51°S, 144.32°E; 1848 m water depth) is located on the north coast of Papua New Guinea in the Bismarck Sea, at 100 km off the Sepik and Ramu river estuaries (Figure 1). The MD05‐2920 sequence is 36.7 m long and mainly composed of homogeneous grayish olive clays with dispersed bioclasts and foraminiferal tests. The two major components of the sediments are terrigenous clays and pelagic carbonates. The sedimentary elemental composition determined by X‐ray fluorescence analyses does not present any clear glacial‐interglacial variability, which suggests that glacial/interglacial alternation exerted a limited influence on the regional hydrological cycle that drives the main sedimentological changes in the region [Tachikawa et al., 2011, 2014]. The core site MD05‐2930 (10°32′S, 146°73′E; 1490 m water depth) is located in the Gulf of Papua (GOP) on a topographic high, protected from the influence of gravity flows from the Papua shelf (Figure 1). The 36.9 m long core consists mainly of homogenous gray‐green clayey and silty‐clayey muds rich in foraminifera. The core is composed of dominant siliciclastic fraction fed by major rivers that drain the southern Papua New Guinea Island and by less than 40% detrital carbonates coming mainly from the northern extension of the Great Barrier Reef [Beaufort et al., 2005; McFadden et al., 2006]. The GOP sedimentation has been largely influenced by sea level fluctuations during the last glacial‐interglacial cycles due to its proximity to the large and shallow continental margin between Australia and Papua New Guinea [Jorry et al., 2008]. We also used data from core MD90‐0961, collected during the SEYMAMA campaign of the R/V Marion Dufresne in 1990. The coring site is located in the Indian Ocean, east of the Maldives platform (05°03.71°N, 73°52.57°E; 2446 m water depth; Figure 1). The lithology of this 45.7 m long core is dominated by calcareous nanofossil ooze with abundant foraminifers and clayey muds [Valet et al., 2014].

Figure 1

Site locations. Location of the three studied cores (colored circles) along with authigenic 10Be/9Be ratio and 10Be flux from marine and glacial records discuss in the text. These references include Greenland Summit [Muscheler et al., 2005], EPICA Dome C (EDC) [Raisbeck et al., 2006; Cauquoin et al., 2015], ODP Sites 983 and 1063 [Knudsen et al., 2008; Christl et al., 2010], Portuguese margin [Carcaillet et al., 2004b; Ménabréaz et al., 2011], KR0515‐PC2 and KR0515‐PC4, and Dome Fuji [Horiuchi et al., 2016]. Results from core MD90‐0961 (green) [Valet et al., 2014] are included with our new results from cores MD05‐2920 (blue) (Bismarck Sea) and MD05‐2930 (red) (Gulf of Papua) to construct the authigenic 10Be/9Be ratio stack.

Chronostratigraphic Framework

The original chronologies of the MD05‐2920 and MD05‐2930 sedimentary sequences [Tachikawa et al., 2011 ; Ménabréaz et al., 2014] were obtained by radiocarbon dating and by correlating their oxygen isotope (δ18O) records with the LR04 reference benthic stack [Lisiecki and Raymo, 2005]. Benthic foraminifera, Cibicidoides wuellerstorfi and Uvigerina peregrina, were picked from the 250–355 mm size fraction, at 10 or 20 cm intervals in core MD05‐2920 [Tachikawa et al., 2011, 2014], and at 20 cm intervals in core MD05‐2930 [Ménabréaz et al., 2014]. The benthic foraminifera δ18O records display clear variations, which were tuned to glacial‐interglacial cycles from the LR04 [Tachikawa et al., 2014; Ménabréaz et al., 2014]. In this paper, we used the radiocarbon calibrations provided by Regoli et al. [2015] for core MD05‐2930, and we recalibrate 14C ages in core MD05‐2920 by using the Marine13 calibration curve [Reimer et al., 2013] (see Tables S1 and S2 in the supporting information). We improved the resolution of original chronologies by adding, and slightly shifting, several tie points from both cores to increase their correlation with the reference LR04 stack (Figure 2). The age uncertainties resulting from the alignments of δ18O records with LR04 is assumed at ±4 ka [Lisiecki and Raymo, 2005]. Moreover, we set the authigenic 10Be/9Be ratio peak corresponding to the well‐recognized Laschamp excursion to 41.3 ± 0.6 ka B.P. [Laj et al., 2014] in order to improve our time resolution within the marine oxygen isotope stage (MIS) 3 interval. That age corresponds to the Laschamp age from the GICC05 age model of Greenland ice cores, i.e., 41.2 ± 1.6 ka B.P. [Svensson et al., 2008] further supporting its reliability. This is the only firm tie point directly derived from the authigenic 10Be/9Be ratio records as we want to avoid any chronological application of the cosmogenic records in this study. Note that the comparison of records combining high‐resolution δ18O and authigenic 10Be/9Be ratio might be very useful to improve global correlations and evaluate regional offsets between benthic records, but this is beyond the scope of this paper. Core MD05‐2920 spans the 2–11 MIS interval (20 to 386 ka), while core MD05‐2930 extends to MIS 20 (5 to 790 ka). The MD90‐0961 δ18O record has been compared with oxygen isotopic stratigraphy from twin core MD90‐0963 [Bassinot et al., 1994] in order to derive a chronology that lasts from MIS 9 to MIS 22. The cosmogenic isotope record of core MD90‐0961 covers a stratigraphic interval spanning MIS 17 to MIS 22 [Valet et al., 2014]. According to their respective age models, the mean sedimentation rates in cores MD05‐2920, MD05‐2930, and MD90‐0961 are 9.3, 4.6, and 4.7 cm/ka, respectively (Figure 2a).

Figure 2

Chronostratigraphy. Oxygen isotope stratigraphy for cores MD05‐2920 (blue) and MD05‐2930 (red) derived from the δ18O recorded by benthic foraminifera Cibicidoides wuellerstorfi and Uvigerina peregrina. Marine oxygen isotope stages are labeled according to the LR04 ages [Lisiecki and Raymo, 2005]. (a) Depth‐age relationship and derived sedimentation rates of cores MD05‐2920 (blue) and MD05‐2930 (red). (b) δ18O records along the reference benthic stack LR04 of Lisiecki and Raymo [2005]. The records are placed on their own chronologies. The full‐colored triangles locate the radiocarbon ages and correlation tie‐points of both cores with the reference record LR04 (see Tables S1 and S2).

Authigenic Beryllium Normalization Procedures

The exchangeable‐10Be concentrations measured in the sediments, or authigenic 10Be (i.e., the fraction dissolved or chemically precipitated and adsorbed onto settling particles), do not only reflect atmospheric production at the time of their deposition but also depend on oceanic transport processes and detrital inputs from surrounding continents [Bourlès et al., 1989; Christl et al., 2010; Simon et al., 2016]. The normalization of authigenic 10Be concentration is necessary to correct for such sedimentary imprints and compare 10Be records with geomagnetic variability. The normalization of the authigenic 10Be cosmogenic nuclide by the authigenic stable 9Be isotope released by silicate rock weathering relies on the similar behavior of both isotopes once homogenized in seawater [Bourlès et al., 1989; Brown et al., 1992]. This method reliably corrects for ocean/continent secondary contributions and provides robust results clearly demonstrating an inverse relationship with the geomagnetic field [e.g., Henken‐Mellies et al., 1990; Robinson et al., 1995; Carcaillet et al., 2003, 2004a, 2004b; Thouveny et al., 2008; Knudsen et al., 2008; Ménabréaz et al., 2012, 2014; Valet et al., 2014; Horiuchi et al., 2016].

Authigenic 9Be and 10Be Extraction and Measurements

Samples of ~1 g (dry sediment) were collected from cores MD05‐2920 and MD05‐2930 with a 10 cm resolution representing a total of 300 and 378 samples and ~1 to ~2 ka nominal time resolution, respectively. These samples have been processed for the Be isotope analysis at the Centre Européen de Recherche et d'Enseignement de Géosciences de l'Environnement (CEREGE) National Cosmogenic Nuclides Laboratory (France) according to the chemical procedure established by Bourlès et al. [1989] and used by subsequent studies from the same group [e.g., Carcaillet et al., 2004b; Ménabréaz et al., 2012, 2014; Valet et al., 2014]. Authigenic 10Be and its stable isotope 9Be were extracted from each ~1 g dry samples by soaking them in 20 mL of 0.04 M hydroxylamine (NH2OH‐HCl) in a 25% acetic acid leaching solution at 95 ± 5°C for 7 h. A 2 mL aliquot of the resulting leaching solution was sampled for the measurement of the natural 9Be concentration. The remaining solution was spiked with 300 μL of a 9.8039 × 10−4 g.g−1 9Be carrier before Be purification by chromatography in order to accurately determine 10Be sample concentrations from the accelerator mass spectrometry (AMS) measurements of 10Be/9Be ratios (additional details on sample preparation and chemical procedures are given in Simon et al. [2016]). In addition to sample processing, several routine blanks and replicates were measured in order to evaluate the cleanliness and reproducibility during the chemical extraction.

The natural authigenic 9Be concentrations were measured by using a graphite‐furnace atomic absorption spectrophotometer (AAS) with a double‐beam correction (Thermo Scientific ICE 3400®). The standard‐addition method and the addition of a constant volume of Mg(NO3)2 solution were used to eliminate the matrix effects during the absorption and to allow measurements near the detection limit. Authigenic 9Be sample concentrations vary around 2.00 ± 0.38 × 10−7 g.g−1 (MD05‐2920) and 3.03 ± 0.43 × 10−7 g.g−1 (MD05‐2930). The associated uncertainties (2σ) based on the reproducibility of quadruplicated measurements and the least squares fitting between measured absorbance at each stages of the standard‐addition method varying around average values of 1.6% (MD05‐2920) and 1.9% (MD05‐2930). Similar 9Be concentration range and uncertainty have been measured in core MD90‐0961 from the Indian Ocean, i.e., 2.00 ± 0.32 × 10−7 g.g−1 and a mean uncertainty of 2.2% [Valet et al., 2014]. The natural authigenic 10Be concentration measurements were performed at the French AMS national facility Accélérateur pour les Sciences de la Terre, Environnement, Risques (ASTER) (CEREGE) [Arnold et al., 2010]. 10Be sample concentrations are calculated from the measured spiked 10Be/9Be ratios normalized to the National Institute of Standards and Technology 4325 Standard Reference Material (2.79 ± 0.03 × 10−11) [Nishiizumi et al., 2007] and are decay‐corrected by using the 10Be half‐life (T1/2) of 1.387 ± 0.012 Ma [Chmeleff et al., 2010; Korschinek et al., 2010]. The calculations of the radioactive decay of 10Be‐concentration results presented in Ménabréaz et al. [2014] and Valet et al. [2014] have been revised in this study. This is particularly important since we are dealing with multiple records and want to compare 10Be/9Be ratio variation amplitudes through time. Chemistry blank ratios range from 10−14 to 10−15, i.e., at least 3 orders of magnitude lower than the sample 10Be/9Be ratios. The authigenic 10Be/9Be ratios are calculated from 9Be‐AAS and 10Be‐AMS concentration measurements transformed in atoms [Simon et al., 2016]. Associated uncertainties (2σ) are derived from the propagation of both uncertainties and vary around 4.3%, 4.7%, and 5.6% in cores MD05‐2920, MD05‐2930, and MD90‐0961, respectively (see Tables S3–S5).

Results

All sample concentrations in atoms per gram and ratios are presented versus depth in Figures 3 and 4. The results and statistical averages are reported hereafter with a ±2 σ uncertainty.

Figure 3

Benthic δ18O record and authigenic beryllium isotope results from core MD05‐2920. (a) The oxygen isotope record from benthic foraminifera Cibicidoides wuellerstorfi and Uvigerina peregrina is expressed as δ18O versus VPDB (‰). (b) Authigenic 9Be concentration variations present a significant change at 1820 cm. The green dotted lines represent mean values computed over each interval (467 to 1820 cm and 1820 to 3610 cm). The white dots represent sample outliers from both intervals (data outside mean ± 1.96σ, p = 0.05) associated to sharp paleoenvironmental events (these levels are also represented by the white dots in Figures 3c and 3d). The red stars identify tephra layers recognized by magnetic susceptibility spikes. (c) The decay‐corrected authigenic 10Be concentrations reveal successive intervals of significant increase and three sharp minima. Three large 10Be lows correspond to 9Be minima associated with the tephra layers (dashed line). (d) Authigenic 10Be/9Be ratio with mean ± 1σ. The 10Be/9Be ratio increase intervals are highlighted by the vertical gray bars and numbered from II to X according to the labeling system proposed by Carcaillet et al. [2004a] and Ménabréaz et al. [2014].

Figure 4

Benthic δ18O record and authigenic beryllium isotope results from core MD05‐2930. (a) The oxygen isotope record from benthic foraminifera Cibicidoides wuellerstorfi and Uvigerina peregrina is expressed as δ18O versus VPDB (‰). (b) Authigenic 9Be concentration variations present increase intervals corresponding to transition from interglacial to glacial periods. The white dots represent sample outliers (data outside mean ± 1.96σ, p = 0.05) associated to sharp paleoenvironmental events (these levels are also represented by the white dots in Figures 3c and 3d. The red stars identify tephra layers recognized by magnetic susceptibility spikes. Large 9Be concentrations highlighted by vertical yellow bar are found at 2180–2280 cm depth corresponding to the transition between MIS 12 and 13 periods. (c) The decay‐corrected authigenic 10Be concentrations reveal successive intervals of significant increase and two sharp minima. The 10Be‐peaks are not related to major 9Be changes, while minimum 10Be‐concentrations are found at depth embedded within sharp environmental imprint (white dots). (d) Authigenic 10Be/9Be ratio with mean ± 1σ.

Authigenic 9Be Concentration

The authigenic 9Be concentrations vary from 0.58 to 2.11 × 1016 at.g−1 in core MD05‐2920 with an average value and standard deviation of 1.34 ± 0.25 × 1016 at.g−1 (Figure 3b). Higher values ranging from 1.14 to 3.00 × 1016 at.g−1 with an average value and standard deviation of 2.03 ± 0.29 × 1016 at.g−1 are observed throughout core MD05‐2930 (Figure 4b). The highest 9Be concentration range is found within the core presenting the lowest sedimentation rates (MD05‐2930), contrary to what had been observed on Portuguese margin sediments [Carcaillet et al., 2004b]. This is likely explained by different Be scavenging affinity with particle type: i.e., higher carbonate content in core MD05‐2920 compared to siliciclastic sediments of core MD05‐2930 [Chase et al., 2002]. This interpretation is supported by similar 9Be concentration ranges measured in carbonate‐rich MD90‐0961 [Valet et al., 2014] and MD05‐2920 cores. Four 9Be‐concentration lows (Figure 3b) in core MD05‐2920 coincide with layers enriched in tephra material characterized by high magnetic susceptibilities. Extreme 9Be‐concentration values in both cores, outside from the 95% normal distribution, are represented by the white dots in Figures 3b and 4b. The long‐term average 9Be concentration of core MD05‐2920 presents a major change at mid‐MIS 6 stage (near 165 ka). Average values vary significantly around (i) 1.19 ± 0.17 at.g−1 beneath 1829 cm and (ii) 1.56 ± 0.18 at.g−1 between 1820 and 0 cm (Figure 3b). Such a significant shift is not observed neither in the δ18O signals nor in the XRF data [Tachikawa et al., 2011, 2014], precluding a simple climatic interpretation. A slight increasing trend of Fe percentage accompanied with a decreasing trend of the CaCO3% suggests a progressive intensification of the rivers discharges, and dissolved 9Be delivery into the Bismarck Sea, highlighting the role of past hydrological changes on the sediment composition of core MD05‐2920 [Tachikawa et al., 2011]. Five 9Be‐concentration peaks, statistically distinct from the long‐term average of each interval, are observed during glacial periods (MIS 4, 5.d, 6, 7.d, and 10) likely supporting sharp paleoenvironmental changes at these levels. In contrast with core MD05‐2920, the authigenic 9Be concentrations of core MD05‐2930 do not show any clear trend but exhibit some sharp 9Be‐concentration increases at transition intervals between interglacial and glacial periods (Figure 4b). This timing corresponds to periods characterized by sea level drops, oceanic circulation, and water‐column property (e.g., thermocline depth) changes that contribute to modify drastically the sedimentation realm within the GOP [Jorry et al., 2008]. While 9Be‐concentration peaks are generally rather small, a large interval of high 9Be concentrations is found between 2180 and 2280 cm, at the transition between MIS 12 and 13. Strong correlation with terrigenous proxies such as Ti or Fe elemental concentrations supports a close association of 9Be signature with terrigenous signals (Tachikawa, personal communication) highlighting the influence of sea level changes on 9Be input in the GOP. During sea level lowstands (glacials), the large and shallow continental margin between Australia and Papua New Guinea was subaerially exposed, the main rivers extended to the present shelf edge and reef systems karstified. This led to an increased dilution of the carbonate fluxes by the siliciclastic fluxes explaining the relatively high 9Be concentration of these layers.

Authigenic 10Be Concentration

Authigenic 10Be (decay‐corrected) concentrations vary from 2.84 to 15.39 × 108 at.g−1 in core MD05‐2920 with an average value of 8.6 ± 2.0 × 108 at.g−1. Smaller variations are observed in core MD05‐2930 with authigenic 10Be (decay‐corrected) concentrations ranging from 2.08 to 12.00 × 108 at.g−1 and an average value of 6.6 ± 1.4 × 108 at.g−1. The main 10Be‐concentration features of both cores are the presence of large peaks and some sharp minima (Figures 3c and 4c). The 10Be intervals that correspond to extreme 9Be‐concentration values, outside from the 95% normal distribution, are represented by the white dots in Figures 3c and 4c. In core MD05‐2920, the decay‐corrected authigenic 10Be‐concentrations denote several important increase intervals and three sharp minima. These 10Be lows correspond to major 9Be changes associated with tephra layers at 1387, 2059–2069, and 3089–3119 cm. The 10Be concentrations yield a slightly increasing trend through the MD05‐2920 core (r 2 = 0.2) similar to its 9Be concentration variation (Figure 3b). This observation is coherent with regional hydrological cycle variations and the progressive intensification of rivers discharge that likely support higher scavenging rates of dissolve Be isotopes from the water column. Core MD05‐2930 does not display any trend in the 10Be‐concentration signature but exhibits seven large peaks and three lows. Two of these low 10Be‐concentration intervals, i.e., 1290 and 1720 cm, correspond to major 9Be lows associated with magnetic susceptibility spikes related to tephra layers. The interval at 2280 cm that presents the lowest 10Be‐concentration value is also characterized by a significant 9Be‐concentration low associated with a major environmental change immediately prior to the 2180–2280 cm interval (highlighted by a yellow bar in Figure 4). Detailed comparison of authigenic Be concentration signals with environmental proxies along the MD05‐2920 and MD05‐2930 cores is beyond the scope of this paper; however, for few specific layers the homogeneous mixing of both dissolved 10Be and 9Be isotopes in the water column has been questioned, stressing the need for cautious geomagnetic interpretations of the authigenic 10Be/9Be ratios in these intervals.

Authigenic 10Be/9Be Ratio

The authigenic 10Be/9Be ratio varies from 3.68 to 11.35 × 10−8 and 1.40 to 5.90 × 10−8 in cores MD05‐2920 and MD05‐2930, respectively. The long‐term average value of core MD05‐2920 is 6.50 ± 1.34 × 10−8, 2 times higher than the long‐term mean value from core MD05‐2930, 3.26 ± 0.72 × 10−8. This difference is mainly supported by higher 9Be‐concentration variations (about 75% of the variance) and, to a lower degree, by 10Be‐concentration changes. The authigenic 10Be/9Be ratio from core MD90‐0961 varies from 2.26 to 8.27 × 10−8 and presents a long‐term average of 3.93 ± 1.35 × 10−8 [Valet et al., 2014]. Interestingly, the ~2:1 ratio between the mean 10Be/9Be ratios is similar to the sedimentation rate ratios between those cores supporting a direct link between siliciclastic sedimentation rates and scavenging efficiency. Three of the four intervals from core MD05‐2920 that present large 9Be‐concentration lows associated with tephra layers (red star in Figure 3b) yield significantly distinct 10Be/9Be ratios (Figure 3d). In core MD05‐2930, the interval 2180–2280 cm associated with large 9Be‐ and 10Be‐concentration variations clearly shows a substantial 10Be/9Be ratios inconsistency (Figure 4d). These intervals associated with environmental artefacts represent less than 3% of the total samples from both cores (5 samples over 300 and 10 samples over 378 in cores MD05‐2920 and MD05‐2930, respectively). Other samples presenting large 9Be concentration changes likely associated with environmental imprints (white dots in Figures 3d and 4d) yield coherent 10Be/9Be ratio with their neighbor samples. Furthermore, no similarities nor correlations are observed between the authigenic 10Be/9Be ratio series and δ18O records or their 9Be normalizer (r 2 < 0.1), supporting the 9Be normalization method, even within the intervals characterized by rapid environmental changes.

Despite some amplitude differences, the two authigenic 10Be/9Be ratio records exhibit similar series of peaks (Figures 3d and 4d). Intervals characterized by 10Be/9Be ratio increases are numbered from I to XXII according to the labeling system proposed by Carcaillet et al. [2004a] and used by Ménabréaz et al. [2014]. Eventually, these labels are accompanied by the names of documented geomagnetic excursions for comparison purpose (Table 1). When compared to the long‐term average Be ratio values, all peaks labeled along the MD05‐2920 and MD05‐2930 represent a multiplication by 1.2 to 1.8. When compared with the framing Be ratio, they represent multiplication factors up to 2.1 (Figures 3d and 4d). In the two cores MD05‐2920 and MD05‐2930, the averages of the highest 10Be/9Be values (above the mean + 1σ) are 8.7 × 10−8 and 4.5 × 10−8, respectively, and are about 2 times higher than the averages of the lowest values (below mean‐1σ), 4. 6 × 10−8 and 2.4 × 10−8, respectively. In core MD90‐0961 where the measured series covers mainly the interval containing the polarity transition, this ratio is 2.7.

Table 1

Authigenic 10Be/9Be Increase Periods in the MD05‐2920, MD05‐2930, and MD90‐0961 Stack Compare to Corresponding Geomagnetic Dipole Lows (GDL) From References

Stack 10Be/9Be Ratio Peaks						Categorye 2 of 10Be Overproduction, Deviation > 1σ				Category 1 of 10Be Overproduction, Deviation > 2σ
Labelsa	Age (ka)	Enhancement Factorb	Standardized Valuesc	10Be‐Derived VADM (1022 Am2)	% of the Average Fieldd	Time Interval (ka)	Duration (ka)	10Be‐Derived VADM (1022 Am2)	% of the Average Field	Time Interval (ka)	Duration (ka)	10Be‐Derived VADM (1022 Am2)	% of the Average Field	GDL Labelf	Names potentially associated to excursionsg	Marine Oxygen Isotope Stages
I*	27	1.23	1.09	3.81	51.57	27–28	1	3.84	51.9						Rockall/Mono Lake (3α)	2/3
II*	41	1.60	2.89	1.34	18.09	39–44	5	2.63	35.6	40–42	2	1.36	18.4	a	Laschamp (3β)	3
III*	62	1.28	1.33	3.33	45.07	61–65	4	3.69	49.8					a'	Norwegian Greenland Sea (4α)	4
IVa*	98	1.22	1.05	3.91	52.91	98	<1	3.91	52.91					b1	Post‐Blake/Fram Strait I (5α)	5.3
IVb*	119	1.24	1.12	3.74	50.57	118–120	2	3.84	52.0					b2	Blake (5β)	5.5
V*	190	1.66	3.16	1.21	16.36	187–194	7	2.44	33.0	188–191	3	1.60	21.6	c	Iceland‐Basin (7α)	6/7
VI*	217	1.32	1.53	2.96	40.04	212–219	7	3.30	44.7					d	Pringle‐Falls ( (7β)	7
VII	258	1.43	2.06	2.14	28.91	256–263	7	3.06	41.4	258	<1	2.14	28.9		Calabrian Ridge 0/Fram Strait (8α)	8
VIII	286	1.33	1.57	2.90	39.18	284–288	4	3.39	45.8					g	Portuguese margin	8
IX	309	1.14	0.70	4.77	64.47		n.s.h							h	Calabrian Ridge 1 (9α)	9
X	378	1.11	0.52	5.22	70.53	n.s.								i1	Laguna del Sello/Levantine/Bermuda (11β)	10/11
XI	406 (412)	1.26 (1.21)	1.23 (1.01)	3.51 (3.99)	47.5 (53.9)	404–406 (412)	2 (+1)	3.73	50.4					i2		11
XII	(438) 447	1.11 (1.17)	(0.52) 0.82	5.20 (4.41)	70.2 (59.6)	n.s.								j1 & j2	Emperor	12
XIII	491	0.98	−0.07	6.95	93.92	n.s.								k	(13α)	13
XIV	523	1.13	0.64	4.88	66.02	n.s.								l1	Big Lost/Calabrian Ridge 2/West Eifel 4 and 5/(14α)	13
XV	538	1.43	2.02	2.15	29.02	534–540	6	2.50	33.8	538	<1	2.15	29.0	l2		13/14
XVI	550	1.09	0.43	5.43	73.37	n.s.								m		14
XVII	586	1.34	1.61	2.78	37.58	585–586	1	3.38	45.7					n	Calabrian Ridge 3 / Los Tilos / La Palma (15β)	15
XVIII	615	1.2	0.93	4.16	56.28	n.s.								o	La Palma (15β)/West Eifel 2	15/16
XIX	700	1.1	0.44	5.26	71.15	n.s.								p, q	Stage 17 (17α)/Delta	17
XX	715	0.81	−0.55	10.00	135.26			n.s.						r	West Eifel 1	18
XXI	772	1.84	3.16	1.24	16.77	765–777	12	1.46	19.8	767–775	8	1.23	16.7	s	Brunhes/Matuyama transition (B/M)	19
XXII*	839	1.36	1.05	2.61	35.33	839–840	1	2.64	35.7						Kamikatsura	21

Adapted following the nomenclature proposed by Ménabréaz et al. [2014]. Stared roman numbers (*) are from this study.

The long‐term average of the normalized stack.

The distance to the mean, expressed in units of the standard deviation, of the arithmetic stack constructed from each standardized records ((x‐mean)/σ) and sampled at 1 ka interval.

The long‐term average geomagnetic dipole moment covering the 5–850 ka period derived from the authigenic 10Be/9Be ratios stack = 7.4 ± 2.7 × 1022 Am2.

The clustering is based on standard deviation values from the standardized stack.

According to a nomenclature proposed by Thouveny et al. [2008] and used in Ménabréaz et al. [2014].

Names are used here for clarity only and are those cited in Langereis et al. [1997], Channell et al. [2006, 2016], Lund et al. [2001, 2006], Laj and Channell [2015]; Singer et al. [2002, 2005, 2008a, 2008b, 2014], and Thouveny et al. [2004, 2008].

n.s. indicates nonsignificant (<1 standard deviation) 10Be/9Be ratio enhancement.

Authigenic 10Be/9Be Ratio Stack

All authigenic 10Be/9Be ratio records are presented on their respective time scales and discussed in the chronological context (Figures 5, 6, 7). Figure 5a presents the standardized 10Be/9Be ratio records of the MD05‐2920, MD05‐2930, and MD90‐0961 cores. The three records exhibit authigenic 10Be/9Be ratio enhancements interpreted as the results of geomagnetic dipole lows (GDLs) which occurred over the covered time span: the last 850 ka. This interpretation can be assessed by comparison with the RPI stacks SINT‐2000 [Valet et al., 2005] and PISO‐1500 [Channell et al., 2009] (Figures 5d and 5e). In these RPI reference curves GDL intervals have been labeled from a to u (Figures 5 and 6), following Thouveny et al. [2008]. The overall agreement between the overlapping parts of the three records and the correspondence of 10Be/9Be ratio peaks with GDLs confirms that the main signal responds to the 10Be overproduction under global geomagnetic modulation (Figures 5 and 6d). An averaged 10Be production record was constructed by stacking the three authigenic 10Be/9Be ratio records by using two distinct methods presented in Figures 5b–5d. The first method uses series normalized to their own long‐term mean (Figure 5b). The second method uses records standardized to their own mean and standard deviation value (Figure 5c). Both methods yield similar results. Each normalized and standardized records have been linearly re‐sampled at 1 ka interval, corresponding roughly to the average sampling resolution of core MD05‐2920, prior to stacking. The averaging was performed by computing arithmetical means along each individual 10Be/9Be ratio record. The uncertainties were calculated by error propagation where cores overlapped and correspond to analytical uncertainty of the MD05‐2930 core within the 387–685 ka interval (Figure 5). These errors likely represent minimal uncertainty. Further, 10Be/9Be ratio and δ18O results from additional records will contribute to increase the global and temporal resolutions of our regional stack, strengthening its interpretation.

Figure 5

Authigenic 10Be/9Be ratio records and stacks from cores MD05‐2920, MD05‐2930, and MD90‐0961 compared with two published VADM stacks. All records are plotted on their respective time scales. (a) Standardized authigenic 10Be/9Be ratios for cores MD05‐2920 (blue), MD05‐2930 (red), and MD90‐0961 (green). (b) Normalized and (c) Standardized 10Be/9Be ratio partial stacks computed from the three core records. VADM variations reconstructed from the (d) SINT‐2000 [Valet et al., 2005] and the (e) PISO‐1500 [Channell et al., 2009] RPI stacks. The 10Be/9Be ratio increase intervals are highlighted by the vertical gray bars and labeled by roman numerals. Atmospheric 10Be‐production increases are synchronous with geomagnetic dipole lows (GDLs) that are labeled by letters from a to u (according to Thouveny et al. [2008]) on the SINT‐2000 and PISO‐1500 records. The vertical yellow bar highlights the disturb interval exclude for the geomagnetic interpretation.

Figure 6

Authigenic 10Be/9Be ratio results over the 0–400 ka interval compared with results of previous studies (see Figure 1 for core locations). (a) Stack of KR0515‐PC2 and KR0515‐PC4 and Dome Fuji 10Be flux (red dots) across the Iceland Basin excursion (IBE) [Horiuchi et al., 2016]. Normalized 10Be deposition rates (blue line) [Frank et al., 1997]. (b) Authigenic 10Be/9Be ratio records from Portuguese margin cores [Carcaillet et al., 2004a, 2004b]. (c) 10Be flux (230Thxs‐normalized) from ODP Sites 1063 and 983 [Knudsen et al., 2008; Christl et al., 2010]. (d) Authigenic 10Be/9Be ratio results from core MD05‐2920 (blue) and MD05‐2930 (red) (this study).

Figure 7

Snapshots of authigenic 10Be/9Be ratio results versus ice core records for the (a) Laschamp excursion (LE), (b) Iceland Basin event (IBE), (c) 270–355 ka interval, and (d) the Matuyama‐Brunhes Boundary (MBB). We used the standardized records of both MD05‐2920, MD05‐2930, and MD90‐0961 cores plotted on their respective time scales. EDC 10Be fluxes [Cauquoin et al., 2015; Raisbeck et al., 2006, 2007] are plotted on the AICC2012 ice age [Bazin et al., 2013]. GISP2/GRIP 10Be flux [Muscheler et al., 2005] is plotted on the GICC05 time scale [Svensson et al., 2008]. Dome Fuji 10Be‐flux [Horiuchi et al., 2016] is plotted on the DFO‐2006 ice age time scale [Kawamura et al., 2007].

The standardized stack provides a straightforward means for clustering the 10Be overproduction episodes (Figure 5c). The stack exhibits major 10Be production peaks (defined by 10Be/9Be ratios higher than the long‐term average +2σ) at intervals II, V, VII, XV, and XXI. Significant increases are defined by 10Be/9Be ratios higher than the long‐term average +1σ. Other minor increases, characterized by 10Be/9Be ratios lower than long‐term average +1σ, are related to GDLs in the SINT‐2000 and/or PISO‐1500 records. A comprehensive interpretation of these three categories of 10Be/9Be ratio increase intervals, and their relationships with reported geomagnetic excursions, is proposed in the following section (Table 1).

Interpretation of the Authigenic 10Be/9Be Ratio Variation

This new record is characterized by the compilation of multiple records over the 20–386 ka and the 688–790 ka intervals. The results from the 280–790 ka interval obtained from core MD05‐2930 are those from Ménabréaz et al. [2014], after an appropriate correction of the radioactive decay and slight modification of the chronology (see section 2.2). The long‐term average has been computed over the whole time interval (i.e., the last 850 ka interval that includes the Brunhes Chron and the terminal part of the Matuyama Chron). The 10Be overproduction episodes are described following the three categories defined from the standardized stack (Figure 5c). Category 1 represents standardized values above 2σ. Category 2 includes 10Be overproduction episodes comprise between 1σ and 2σ. Category 3 represents some minor increase intervals below 1σ (see Table 1 for summary). The sampling resolution together with the stacking process prevents any interpretation with higher resolution than 1 ka.

Category 1 10Be Overproduction Episodes

Interval II documents a large 10Be/9Be ratio increase representing a 10Be overproduction episode associated with the Laschamp excursion. The 1.8‐ and 1.7‐fold 10Be/9Be ratio increase in core MD05‐2920 and MD05‐2930, respectively, are very similar with the Laschamp GDL in the global paleointensity stack (GLOPIS) [Laj et al., 2004], SINT‐2000, and PISO‐1500 records (Figure 5). This major 10Be/9Be ratio increase (values above 2σ) spans a 2 ka duration interval within the interpolated stack record and appears shorter within individual records (~1 ka). It is globally coherent with 10Be overproduction signal from other marine sequences [Carcaillet et al., 2004b; Christl et al., 2010] (Figures 6b and 6c) and is contemporaneous and similar to a 10Be‐flux doubling in ice core records from Antarctica (European Project for Ice Coring in Antarctica (EPICA) Dome C) [Raisbeck et al., 2007] and Greenland (Greenland Ice Core Project/Greenland Ice Sheet Project 2 (GRIP/GISP2)) [Muscheler et al., 2005] (Figure 7a).

Interval V shows a significant 10Be overproduction episode lasting 3 ka at the MIS 6/7 boundary (Figure 5). The 1.7‐ and 1.6‐fold 10Be/9Be ratio increase in cores MD05‐2920 and MD05‐2930, respectively, is centered at 190 ka. This corresponds to GDL associated with the Iceland Basin excursion (IBE) [Channell et al., 1997; Channell, 1999, 2014; Laj et al., 2006]. The directional swings and RPI low reported for this excursion and GDL yield a midpoint age of 190 ka at eight North Atlantic sites [Channell, 1999, 2014], on the Portuguese margin [Thouveny et al., 2004], at the South Atlantic Ocean Drilling Program (ODP) Site 1089 [Stoner et al., 2003], and at western equatorial Pacific Ocean sites [Yamazaki and Yoka, 1994]. The IBE GDL also lies in the MIS 6/7 transition in the Lake Baikal records [Oda et al., 2002; Demory et al., 2005]. The IBE corresponds to the GDL c in the SINT‐2000 and PISO‐1500 records (Figures 5d and 5e). It is synchronous with 10Be overproduction signal reported from several marine records worldwide [Frank et al., 1997; Carcaillet et al., 2004b; Knudsen et al., 2008; Christl et al., 2010; Horiuchi et al., 2016] (Figure 6) and is expressed by a doubling of the 10Be flux recorded in the Antarctica Dome Fuji ice core [Horiuchi et al., 2016] (Figure 7b).

Interval VII shows a large 10Be production enhancement within the 256–263 ka interval. The peak centered at ∼258 ka presents a 1.5‐ and ‐1.3‐fold 10Be/9Be ratio increase in cores MD05‐2920 and MD05‐2930, respectively. It is not directly related to a major GDL in the RPI stacks (Figures 5d and 5e) but may correspond to few events such as the Calabrian Ridge 0 excursion reported from the Ionian Sea (Mediterranean Sea) by Langereis et al. [1997], the 8α excursion described from Integrated Ocean Drilling Program Leg 172 Sites 1060–1063 by Lund et al. [2001], and the Fram strait excursion [Nowaczyk and Frederichs, 1999].

Interval XV yields a major (1.4‐fold increase) 10Be production episode centered at 538 ka associated with GDL l in SINT‐2000 and l2 in PISO‐1500. It is the major event of a multiple structure composed of three distinct 10Be overproduction peaks: XIV (522–529 ka), XV (534–540 ka), and XVI (550 ka), respectively, related to GDL l and m in SINT‐2000, and to GDL l1, l2, and m in PISO‐1500. These peak intervals can be tentatively correlated with several identified excursions (Table 1). Interval XV is at the age of prominent RPI lows and excursional directions at Sites 983 and 984 [Channell et al., 2004] and at ODP Site 1062 (14α) [Lund et al., 2001]. The so‐called Big Lost excursion was first described in lava flows from Idaho dated at 565 ± 14 ka [Champion et al., 1988] and re‐dated at 558 ± 20 ka [Lanphere, 2000]; anomalous directions and weak paleointensities in lava flows of the Eifel dated by 40Ar/39Ar at 555 ± 4 ka supported its global occurrence [Singer et al., 2008a].

Interval XXI displays the largest authigenic 10Be/9Be ratio peak (1.8‐fold increase compared to the long‐term average) within the 767–775 ka period (values > +2σ). This 10Be production enhancement recorded in core MD05‐2930 is in phase with the midpoint 10Be production enhancement recorded in core MD90‐0961 at 772 ka. It is related to the dipole field collapse linked to the Brunhes/Matuyama (B/M) geomagnetic polarity reversal (labeled GDL s in Figure 5) [e.g., Valet and Fournier, 2016]. It coincides with the minimum VADM in PISO‐1500 at ~770–775 ka and reasonably agrees with the minimum VADM in SINT‐2000 at ~777 ka (Figures 5d and 5e). Its amplitude agrees with the near doubling of the 10Be production rate of the B/M reversal reported from previous 10Be/9Be ratio or 10Be‐flux studies of marine sediments [Raisbeck et al., 1985; Carcaillet et al., 2003; Suganuma et al., 2010]. It also agrees with the 10Be production rate increase reported from the EPICA Dome C (EDC) ice core [Raisbeck et al., 2006]. Figures 7c and 7d present the EDC 10Be flux recalculated from raw decay‐corrected 10Be concentrations [Cauquoin et al., 2015; Raisbeck et al., 2006] according to the AICC2012 ice age [Bazin et al., 2013] and filtered by using the locally weighted least squares error (Lowess) method. The 10Be overproduction episode recorded in core MD05‐2930 is enclosed within the EDC 10Be‐flux peak spanning the 766–777 ka interval, while it presents a broader signature in the MD90‐0961 core. No 10Be overproduction episode precedes the main signature of the reversal in contradiction with the frequent reporting of a marked RPI low interpreted as a precursor of the B/M reversal (labeled s′ in Figure 5e). This appears also in contradiction with the presence of a 10Be flux peak in the EDC record [Raisbeck et al., 2006] (Figure 7d). However, the anticorrelation between the 10Be‐flux and δD in the EDC record [Raisbeck et al., 2006, Figure 1] might suggest that (i) an environmental component might still be prevailing in the 10Be‐flux and/or (ii) that it corresponds partially to an artifact due to an underestimation of the EDC accumulation rates from δD [Cauquoin et al., 2015]. Such an anticorrelation is expected for 10Be concentrations because of the inverse relationship between the 10Be dry deposition and snow accumulation rates on the East Antarctic Plateau, but should be removed in the 10Be‐flux term. Moreover, very noisy raw EDC 10Be concentrations characterized by numerous sharp spikes [Raisbeck et al., 2006], together with new geochemical results [Tison et al., 2015], point out that the 10Be peak associated with the “precursor” (Figure 7d) might also partially result from potential 10Be migration within the veins network between large ice crystals, and during fusion recrystallization processes in the extreme bottom layers of ice core [Landais et al., 2004]. These elements cast doubt on the precise geomagnetic interpretation of a precursor at this level in the EDC record. The absence of any strong RPI collapse corresponding to the precursor in the new HINAPIS‐1500 stack from North‐Atlantic high‐sedimentation rate sedimentary cores [Xuan et al., 2016] also questions its strong signature within the PISO‐1500 stack [Channell et al., 2009]. Before any conclusion, further studies are required on other sediment series and for other reversals.

Category 2 10Be Overproduction Episodes

Interval I is only expressed in core MD05‐2930 by a 1.5‐fold 10Be/9Be ratio increase centered at 27.5 ka (Figures 5a and 6d). It corresponds with a minor GDL in the PISO‐1500 record (Figure 5e). Excursions were documented at 25.5–27 ka in Arctic Ocean sediments [Nowaczyk and Knies, 2000] and North Atlantic sediments [Channell et al., 2016]. Considering the age uncertainty, the 10Be overproduction interval (values higher than mean + 1σ; Figure 5c) spanning the 24–32 ka interval could be correlated to the Mono Lake excursion at ∼32–34 ka [Liddicoat and Coe, 1979; Wagner et al., 2000; Laj and Channell, 2015]. This signature, if confirmed, would provide the first cosmogenic evidence obtained from marine sediments of the GDL linked to the Mono Lake excursion, otherwise recognized in the 36Cl signal in the GRIP ice core [Muscheler et al., 2005].

Interval III displays a large 10Be production enhancement between 61 and 65 ka, in phase with GDL a′ from the SINT‐2000 and PISO‐1500 records (Figure 5). The core MD05‐2930 is characterized by higher (1.5‐fold increase) and long‐standing 10Be/9Be ratio interval compared to the 1.4‐fold increase 10Be/9Be ratio from core MD05‐2920 (Figure 5a). At this age the Norwegian Greenland Sea excursion has been recorded by low intensity and large direction swing in several studies [e.g., Bleil and Gard, 1989; Nowaczyk and Frederichs, 1999; Laj et al., 2004; Simon et al., 2012; Nowaczyk et al., 1994, 2013]. This cosmogenic overproduction episode was previously documented by Frank et al. [1997], Carcaillet et al. [2004b], and Christl et al. [2010] (Figures 6b and 6c).

Between 95 and 125 ka, intervals IVa and IVb document two successive 10Be/9Be ratio increase episodes (1.2‐fold increase) associated with a durable GDL, subdivided into two subunits (GDLs b1 and b2) in the SINT‐2000 and PISO‐1500 records (Figures 5d and 5e). Low RPI events within this time frame are referred as the post‐Blake and Blake events, respectively. Their directional and intensity signatures have been largely documented in many sedimentary records, lava flows, and in a speleothem since the first discovery of the Blake event observed on the Blake Outer Ridge sediment sequences [e.g., Smith and Foster, 1969; Denham, 1976; Creer et al., 1980; Tucholka et al., 1987; Tric et al., 1991; Thouveny et al., 1990, 2004; Bourne et al., 2012; Osete et al., 2012; Singer et al., 2014]. This long‐lasting GDL is also identified by successive large 10Be production enhancements in three studies: the global 10Be‐flux stack of Frank et al. [1997] (Figure 6a), Portuguese margin [Carcaillet et al., 2004b] (Figure 6b), and ODP Site 983 [Christl et al., 2010] (Figure 6c). The 10Be/9Be ratios of cores MD05‐2920 and MD05‐2930 present some discrepancies. Interval IVa dated at 98 ka presents a small‐amplitude difference (1.3‐fold versus 1.2‐fold increase). Interval IVb presents significant amplitude differences and timing offsets between both cores. In core MD05‐2920, peak IVb is centered at 124 ka and consists in a 1.2‐fold increase, while in core MD05‐2930 peak IVb points at 119 ka and consists in a 1.4‐fold increase. This likely result in an underestimation of the 10Be overproduction episodes associated with the post‐Blake and Blake events.

Interval VI documents a significant (1.3‐fold) 10Be production signal spanning the 212–217 ka interval (peak at 217 ka). It corresponds to the GDL d in the PISO‐1500 record (Figure 5e) that appears as a minor feature in SINT‐2000 (Figure 5d). The Pringle Falls excursion has been signaled and dated in the same time interval in sedimentary sequences at Pringle Falls (Oregon) at 218 ± 10 ka [Herrero‐Bervera et al., 1994] and at ODP Site 919 (North Atlantic Ocean) between 205 and 225 ka [Channell, 2006], in volcanic sequences of Albuquerque Volcanoes (New Mexico) at 211 ± 13 ka [Singer et al., 2008b], and possibly confused with the Mamaku excursion recorded in an ignimbrite of New Zealand dated at 227 ± 8 ka [McWilliams, 2001]. This GDL was also documented by 10Be/9Be ratio increases associated with RPI low in the Portuguese margin core MD95‐2040 [Carcaillet et al., 2004b] (Figure 6b) and by high 10Be‐flux at ODP Site 1063 [Christl et al., 2010] (Figure 6c).

Interval VIII yields a significant 10Be overproduction episode centered at 286 ka (1.5‐ and 1.3‐fold increases in cores MD05‐2920 and MD05‐2930, respectively) associated with the GDL g in both SINT‐2000 and PISO‐1500 (Figures 5d and 5e). This GDL at ∼290 ka was recorded as a deep RPI low accompanied by excursional directions in two cores MD95‐2039 and MD95‐2040 of the Portuguese margin [Thouveny et al., 2004], thus identifying the Portuguese Margin excursion, strongly supported by a large 10Be enhancement [Carcaillet et al., 2004b] (Figure 6b). Several 10Be‐flux peaks, with a maximum 1.4‐fold increase, are also reported from the EDC record during this time interval [Cauquoin et al., 2015] (Figure 7c).

Interval XI, between 394 and 415 ka, documents two 10Be overproduction episodes (1.3‐ and 1.2‐fold increases, respectively) centered at 406 and 412 ka and associated with the double configuration of GDL i2 documented in PISO‐1500 (Figure 5e). This double structure is coherent with repeated RPI minima at 402 and 413 ka in equatorial Pacific sediments [Valet and Meynadier, 1993]. Although GDL i consists of a unique low in SINT‐2000, the multiple structures of intervals X and XI agree with identification of successive excursions within that time frame: Levantine [Langereis et al., 1997] and 11α/Bermuda [Lund et al., 2001; Channell et al., 2012].

Interval XVII at 582–590 ka reports a 1.3‐fold 10Be production enhancement centered at ∼586 ka emerging from a massive structure (Figure 5c) associated with GDL n at 590–600 ka in the SINT‐2000 and PISO‐1500 records. Excursional directions and RPI lows corresponding to this period were described from ODP sites 983 and 984 [Channell et al., 2004] and ODP Leg 172 [Lund et al., 2001]. Other evidences of excursion were reported from lava flows at La Palma (Canary) Island, dated at 602 ± 24 ka [Quidelleur et al., 1999] and at 580 ± 8 ka [Singer et al., 2002], and from lava flows of the Eifel (Germany) and from Tahiti, dated at 578 ± 8 and 579 ± 6 ka, respectively [Singer et al., 2008a].

Interval XXII corresponds to a 10Be overproduction episode (1.4‐fold increase) covering a long‐lasting interval in core MD90‐0961, peaking at 839 ka. This episode is associated with GDL u in the SINT‐2000 and PISO‐1500 records and might be related to a geomagnetic instability dated at 822.2 ± 8.7 ka at La Palma (Canary Islands) [Singer et al., 2002] or to the Kamikastura excursion recorded in Icelandic lavas and dated at 862 ± 51 ka [Camps et al., 2011].

Category 3 10Be Production Episodes

This third category includes minor 10Be production episodes that emerge slightly over the long‐term average 10Be/9Be ratio (<1σ). Some of those episodes have been labeled and briefly described if recorded in both cores and related to well‐defined GDL in SINT‐2000 and PISO‐1500.

Interval IX documents several minor 10Be overproduction episodes corresponding to GDL h characterized by a double GDL minimum in PISO‐1500 (Figure 5e). Core MD05‐2920 presents a similar structure with two distinct peaks (1.3‐ and 1.1‐fold increases) centered at ∼310 and ∼327 ka, respectively. The 10Be/9Be ratio signature of core MD05‐2930 presents a rather massive interval of enhanced 10Be production (1.1‐fold increase) with smaller peaks centered at 312, 319, and 327 ka. Two 10Be‐flux peaks (1.3‐fold increase) have also been identified during that time interval (307 ka, 320–330 ka) in the EDC record [Cauquoin et al., 2015] (Figure 7c).

Interval X yields several increasing 10Be‐production intervals (∼1.2‐fold increase) within the 367–383 ka period in cores MD05‐2920 and MD05‐2930 and might be associated with GDL i1 from the PISO‐1500 record.

The timing of intervals IX and X is consistent with excursions reported within the 310–330 ka and 360–420 ka intervals at several marine sediment sites [Langereis et al., 1997; Lund et al., 2001; Channell et al., 2012].

Interval XII displays two 10Be production enhancements (1.1‐ and 1.2‐fold increases) at 438 and 447 ka, respectively. These twin peaks might be related to the double structure of GDL j in PISO‐1500 (j1 at 430–445 ka and j2 at 455–460 ka), instead of a single GDL j, as reported in SINT‐2000.

Interval XIII centered at 490–500 ka is barely significant with only a maximum 1.0‐fold 10Be/9Be ratio increase, but it correlates with GDL k, a wide GDL of small amplitude in SINT‐2000 and a narrow GDL of large amplitude in PISO‐1500.

Intervals XIV and XVI yield two 10Be production increase episodes centered at 523 and 550 ka, respectively. These episodes are associated with the category 1 10Be overproduction interval XV (see above).

Interval XVIII documents a 1.2‐fold 10Be production increase centered at 615 ka and associated with GDL o in the PISO‐1500 record. Evidences of an excursion corresponding to this period arise from anomalous directions in the Lishi loess sequence [Liu et al., 1988] and from La Palma (15β)/West Eifel 2 [Lund et al., 2006; Singer et al., 2008a].

Interval XIX reports a small 10Be overproduction episode between 690 and 704 ka with a maximum peak at 700 ka (1.2‐fold increase). This large interval is the “tip of an iceberg” that mirrors the large GDLs p and q structure present in both SINT‐2000 and PISO‐1500 records (Figures 5d and 5e). The interval XIX can also be related to a double inclination anomaly recorded in sediments deposited during MIS 17 in Osaka Bay [Biswas et al., 1999].

The interval XX at 715–730 ka is hardly significant, documenting a 0.9‐fold 10Be/9Be ratio increases centered at 715 and 729 ka in cores MD05‐2930 and MD90‐0961, respectively. The chronological offset of this interval between the two records likely explains some over smoothing at this level where GDL r is recorded at 730 ka (Figure 5).

Two minor 10Be/9Be ratio increase episodes deserve additional analyses before labeling: they include two peaks coherent with the Albuquerque and Mamaku excursions centered at 153 ka and 240 ka [Peate et al., 1996; Shane et al., 1994], respectively. The Albuquerque event is expressed solely in core MD05‐2930 and is not put forth by a clear GDL in the SINT‐2000 and PISO‐1500 records. The Mamaku event corresponds to a 10Be/9Be ratio increase (1.2‐fold), but its age deviates in both cores (232 and 240 ka in core MD05‐2930 and MD05‐2920, respectively). A slight alignment of both peaks by 4 ka (Figure 5a) would be coherent with a 10Be production enhancement at ~236 ka in core MD95‐2040 from the Portuguese Margin (Figure 6b) [Carcaillet et al., 2004b] corresponding to GDL e from the PISO‐1500 record (Figure 5e). This would also agree with the age of 230 ± 12 ka obtained on excursional lava flows at Mamaku (New Zealand) [Shane et al., 1994].

10Be‐Derived Virtual Dipole Moments

Since atmospheric cosmogenic 10Be production rates are modulated at global scale by the geomagnetic dipole moment (GDM) (Figure 5), empirical and theoretical arguments [e.g., Lal and Peters, 1967; Masarik and Beer, 2009] can be used to calibrate the 10Be/9Be ratio by using absolute virtual (or virtual axial) dipole moment values (i.e., VDM or VADM) in order to reconstruct a 10Be‐derived geomagnetic dipole moment record (Figures 8 and 9). Absolute GDM values were extracted from absolute paleointensity databases: Geomagia50.v3 (http://geomagia.gfz‐potsdam.de, December 2015) [Brown et al., 2015a, 2015b] and PINT2015‐05 (http://earth.liv.ac.uk/pint/, December 2015) [Biggin et al., 2009]. Reliability criteria were used for selection. They include (i) quality of methods and number of samples used to produce the paleointensity (PI) values, (ii) the presence of suitable paleomagnetic direction information, and (iii) a reasonable error bar on the age. Only PI data provided with standard deviation and age errors lower than 10% were retained: a total of 487 absolute VDM or VADM data (∼30% of the total data set) are available for our calibration procedure (Figure 10d). This selection allows excluding extreme values that arise from distinct measurement procedures and/or that does not provide sufficient confidence (these values are represented by open circle in Figure 10d). After this first data filtering, we have associated clusters of high, intermediate, and low GDM values with clusters of low, medium, and high values of the 10Be/9Be ratio stack within four distinct time intervals following the procedure used by Ménabréaz et al. [2012, 2014] in order to account for uncertainties inherent to volcanic paleomagnetic data and to their nonuniform distribution through time. For each time interval, average values are calculated to discriminate the data in three clusters defined as data lower than mean ± 1σ, data comprised between mean − 1σ and mean + 1σ, and data higher than mean ± 1σ limits (see Table S6).

Figure 8

Calibration of the normalized authigenic 10Be/9Be ratio stack using absolute values of the virtual axial dipole moment (VADM) extracted from the Geomagia50.v3 [Brown et al., 2015a, 2015b] and PINT2015‐05 [Biggin et al., 2009] databases. Average VADM and authigenic 10Be/9Be ratio values are computed for different clusters (see text and Table S6 for details). Associated error bars correspond to the standard deviation of the values used in each cluster. The polynomial fit (purple line) between these calibration points is used to compute the 10Be‐derived GDM presented in Figures 9 and 10. The dotted red line represents the theoretical relationship between the relative strength of the magnetic field and 10Be cosmogenic nuclide production rate according to model from Masarik and Beer [1999, 2009] and Wagner et al. [2000].

Figure 9

10Be‐derived GDM reconstruction of the 5–850 ka interval. From 0 to 10 ka, the red line is the global field model CALS10k.1b [Korte et al., 2011]. The 10Be/9Be ratio increase intervals are highlighted by vertical grey bands and numbered from I to XXII, adapted from the labeling of Carcaillet et al. [2004a] and Ménabréaz et al. [2014]. Geomagnetic dipole lows (GDLs) are labeled a to u. VADM rates of changes (x 1022 Am2 ka−1) are shown. Field threshold percentages are calculated versus the time‐averaged GDM for the past 850 ka (7.4 ± 2.7 × 1022 Am2). The Laschamp (II), Iceland Basin (V), and MBB (XXI) present field value percentages below 20%, while the other major events—VII and XV—are below a 30% threshold. Secondary 10Be‐production episodes are found below ~50% of the average field value (~3.7 × 1022 Am2). The yellow band underlines the disturb interval excluded for geomagnetic interpretation.

Figure 10

10Be‐derived GDM for the 5–850 ka interval (black) compared to (a) the PISO‐1500 VADM record [Channell et al., 2009], (b) the SINT‐2000 VADM record [Valet et al., 2005], (c) the southeast Pacific deep‐sea floor magnetization record [Gee et al., 2000], and (d) the absolute intensity from the Geomagia50.v3 [Brown et al., 2015a, 2015b] and PINT2015‐05 database [Biggin et al., 2009]. The full dots were used for the 10Be‐GDM calibration; the empty circles were excluded (see text for details). All records are plotted on their respective time scales.

For the first time interval, data are restricted to the 5 to 50 ka including the Laschamp excursion and are compared with data extracted from the Geomagia50.v3 database. This interval has an average absolute GDM value of 4.97 ± 2.43 × 1022 Am2 and an average 10Be/9Be ratio value of 1.09 ± 0.18 × 10−8. The second time interval—i.e., 50–750 ka—excludes the Laschamp excursion and the Brunhes‐Matuyama reversal periods. This interval has an average absolute GDM of 7.43 ± 2.87 × 1022 Am2 and an average 10Be/9Be ratio value of 0.96 ± 0.17 × 10−8. The third time interval, 750–850 ka, includes the Brunhes‐Matuyama reversal and has an average absolute GDM of 6.50 ± 3.42 × 1022 Am2 and an average 10Be/9Be ratio value of 1.05 ± 0.30 × 10−8. Finally, the whole time interval 5–850 ka was considered with an average absolute GDM of 6.90 ± 2.97 × 1022 Am2 and an average authigenic 10Be/9Be ratio value of 0.98 ± 0.20 × 10−8.

All clusters computed for all time intervals are used to define an empirical calibration curve by polynomial regression through all calibration points (Figure 8). As raised by Ménabréaz et al. [2014], the calculation of arithmetic means inside the clusters introduces slight errors given the nonlinear relationship between geomagnetic field variation and 10Be production rate. These errors are included in the 2σ uncertainty associated with the 10Be‐derived GDM record presented in Figures 9 and 10 (Table S7). The used polynomial regression is in good agreement with the theoretical relationship established by Masarik and Beer [1999, 2009] between the magnetic field strength and 10Be cosmogenic nuclide production rate (Figure 8).

Discussion

10Be‐Derived GDM Variations

Dipole moment values of the 10Be‐derived GDM record range from 1.15 to 13.90 × 1022 Am2 around an average value and standard deviation of 7.4 ± 2.7 × 1022 Am2 (Figure 9 and Table S7). The mean value of the standard error associated with the stack is 0.73 Am2. The average 10Be‐derived GDM value is remarkably consistent with the long‐time average values of 7.3 ± 1.8 × 1022 Am2 and 7.5 ± 2.6 × 1022 Am2 computed for the last 850 ka from PISO‐1500 and SINT‐2000, respectively [Channell et al., 2009; Valet et al., 2005] (Figures 10a and 10b). These values are not significantly different from the long‐term VADM or VDM averages computed from the PADM2M (6.2 ± 1.2 × 1022 Am2) [Ziegler et al., 2011] and from the absolute PI data set used for calibration (6.9 ± 3.0 × 1022 Am2) (Figure 10d). The average VADM value is also relatively similar with the present dipole moment value (7.78 × 1022 Am2). In Figure 9, the 10Be‐derived GDM record is completed by a 2 ka overlap with the global field model CALS10k.1b [Korte et al., 2011]. The lack of correlation of 10Be‐derived GDM record with archeomagnetic VADM reconstruction between 5 and 8 ka could be attributed to disturbed sediments at the MD05‐2930 core top resulting from coring processes (similar weakly constrain recent field behaviors are often observed in RPI data from piston cores) together with low global coverage and, especially, large disparity of the absolute intensity data from that period (Figure 10d). Our new 10Be‐derived GDM record presents strong compatibilities with published VADM stacks, despite slight age offsets (i.e., at 330–350 ka, 550–570 ka, and 710–720 ka) (Figures 10a and 10b) that may result from various causes (e.g., age model uncertainties and pDRM lock‐in effects in RPI curves). Moreover, this new record agrees with the southeast Pacific seafloor magnetization record obtained from deep‐towed magnetometry [Gee et al., 2000], thus confirming their common dipolar geomagnetic moment origin (Figure 10c).

The compatibility of the 10Be‐derived GDM with paleomagnetic reference curves based on several distinct approaches and calibration procedures supports a posteriori the reliability of principles and methods of the reconstruction of geomagnetic dipole moment variations by using cosmogenic nuclide production. Therefore, this record provides an independent way to evaluate critical GDM values, change rates, and thresholds linked to large‐amplitude geomagnetic field changes such as excursions and reversals. The chronology of the record based on the astronomically calibrated LR04 stack yields an updated time series of GDL linked to Brunhes excursions and B/M reversal, which provides a helpful template for discussing specifically each geomagnetic events by using complementary techniques such as paleomagnetic measurements and 40Ar/39Ar dating on lava flows.

Geomagnetic Dipole Low (GDL) Frequency

Our results confirm that GDLs, correlated or not with reported excursions, are intrinsic features of the geodynamo behavior, reflecting its instability, and triggering mechanisms that may also generate polarity reversals [Laj and Channell, 2015]. The future analyses of frequency, timing, amplitude, and rates of geomagnetic instabilities seen from the cosmogenic perspective will provide critical constraints, complementary of paleomagnetic reconstructions, to feed and assess the validity of numerical and analogic dynamo models [Busse, 1978; Vanyo et al., 1995; Olson et al., 2010]. One striking features observed in Figure 9 is the apparent variation of GDL density across the Brunhes Chron. From 8 to 300 ka, nine major GDLs are associated with significant 10Be overproduction episodes, while between 301 and 775 ka only three of those significant GDLs are observed. This suggests an increase in GDL frequency associated with recognized excursions between the early Brunhes and the late Brunhes phase. The average GDM value of the 8–300 ka interval (6.4 ± 2.3 × 1022 Am2) is lower than that of the 301–775 ka interval (8.1 ± 2.6 × 1022 Am2), which is coherent with a reciprocal relationship between the number of GDL, their length, and the average value of the dipole moment. A similar GDM distinction is found when excluding the major GDLs from the calculation: 7.0 ± 1.9 × 1022 and 8.4 ± 2.3 × 1022 Am2 for the 8–300 ka and 301–775 ka intervals, respectively. These calculations give an identical ∼20% lower average dipole moment for the 8–300 ka interval compared to the 301–775 ka first half of the Brunhes chron (Figure 9). This implies that (i) a fairly weak GDM triggers, or is triggered by, frequent and/or durable instabilities (GDL and excursions), and on the contrary (ii) a strong GDM triggers, or is triggered by, a low frequency of instabilities (few GDL and excursions).

This observation favors the hypothesis that the geodynamo becomes periodically unstable when the dipole field is weak [e.g., Singer et al., 2008b; Olson et al., 2010] and tends to reject the hypothesis of pure stochastic unstabilities of the geodynamo [Zhang and Gubbins, 2000].

GDL Thresholds for 10Be Overproduction Episodes

As 10Be production at millennial scale is solely associated with the geomagnetic dipole moment modulation, only GDM drops can support the overproduction episodes. A critical value for triggering excursions or reversals was assessed by Channell et al. [2009] at ∼2.5 × 1022 Am2, representing about 30% of the present dipole moment value. All category 1 10Be overproduction episodes, including the fourth largest excursions (i.e., Laschamp, Iceland Basin, Calabrian Ridge 0, and Big Lost) and the B/M reversal, result from such events and provide such low calibrated 10Be‐derived GDM values (Figure 9). The B/M reversal and the Laschamp excursion present the weakest GDM values of 1.2 and 1.4 × 1022 Am2, respectively. This is equivalent to less than 20% of the mean average GDM over the period (Figure 9). The Iceland Basin excursion yields a GDM value of 1.6 × 1022 Am2 (∼22%), while two other events—Calabrian Ridge 0 and Big Lost—present GDM values of 2.1 × 1022 Am2 (29% of the mean GDM). The somehow lower 10Be/9Be ratio peaks of these last two events that hardly exceed the +2σ limit of category 1 events supposed that they might be related to the upper limit of the category 2 10Be overproduction episodes rather than being associated with the others three GDL maxima. All secondary Brunhes excursions of category 2 10Be overproduction episodes correspond with GDM values below 3.3–3.9 × 1022 Am2 (45 to 52%) of the mean GDM.

Such observations may help to understand the GDM thresholds necessary to trigger geomagnetic excursions and/or reversals [e.g., Laj and Channell, 2015; Valet and Fournier, 2016]. A threshold at 50% of the present GDM is evidenced since each 10Be‐derived GDM values lower than this limit are associated with paleomagnetic excursions, i.e., a directional instability largely overpassing the limits of secular variation. The B/M reversal, the Laschamp excursion, and Iceland Basin excursion, often considered as transient full reversals, (and marked as such in the ocean floor anomaly record (Figure 10c)) occurred below a GDM of 1.6 × 1022 Am2, i.e., ∼20% of the present GDM value. Therefore, the threshold of 50% of the actual GDM would suffice to trigger category 2 excursions, while the critical threshold of 20 to 30% seems to be required to trigger category 1 excursions and/or reversals.

GDL Durations and Decay/Recovery Rates

In order to compare the field features of the last reversal versus the largest Brunhes excursions, defined by category 1 10Be overproduction episodes, we derived the GDM decay/recovery rates from our record. These rates represent the best relative estimates considering the resolution and uncertainty associated with our record. Although they can be discussed due to forthcoming improvements in their accuracy and precision, the strong compatibilities of our record with RPI stacks support the proposed approach. Furthermore, the occurrence of systematic uncertainty changes between specific events, or between the decay or recovery phases, is hardly possible, which makes their comparison relevant. We calculated these rates for two types of GDL durations: 5 and 10 ka to test possible disparities. Computed average decay rates for the 5 and 10 ka intervals are 0.9 ± 0.5 and 0.6 ± 0.2 × 1022 Am2 ka−1, while their mean average recovery rates are 0.8 ± 0.3 and 0.5 ± 0.2 × 1022 Am2 ka−1, respectively. These coherent results suggest an increase of the decay/recovery rates in the 5 ka before and 5 ka after the major event peak (Figure 9). Maximum rates concern the Laschamp excursion with 1.5 × 1022 Am2 ka−1 for the decay and 0.9 × 1022 Am2 ka−1 for the recovery. Similar decay rates have been calculated for the Laschamp excursion from ice core records [Wagner et al., 2000; Muscheler et al., 2005], from the high‐resolution GLOPIS record [Laj et al., 2004], and from the 10Be derived GDM global stack by Ménabréaz et al. [2012]. Lower decay and recovery rates are computed for the B/M transition, associated with the 8 ka duration of the low‐intensity interval (Table 1). Change rates computed on longer term, over ±30 ka before and after the B/M transition between the GDM minimum and surrounding GDM maxima, are 0.36 × 1022 Am2 ka−1 for the decay and 0.42 × 1022 Am2 ka−1 for the recovery. The same computation for the Laschamp excursion gives 1.1 and 0.9 × 1022 Am2 ka−1 for the 41–50 and 35–41 ka intervals. The decay and recovery rates for all events present a small asymmetric feature with consistently higher decay rates of about 25–40% compared to recovery rates, although important variation between the events is revealed by large standard deviations. By contrast, the change rates of the B/M transition are lower for the decay than for the recovery for all calculation methods. Independently, from the time window selected for calculating the decay/recovery rates, the results of this study suggest lower change rates for the B/M transition compared to major excursions/events. This observation suggests that geodynamo behaviors are different for excursions/events and reversal. If confirmed our observation would support the Gubbins [1999] hypothesis that excursions reflect relatively rapid and short‐lived reversal of the field generated by flow of liquid in the outer core, while reversal involve longer‐term deeper processes of geomagnetic field generation and interactions with the inner core induced field. According to that model, excursions could not stand longer than 3 ka, while reversals would span longer time intervals. Our study shows that all major GDL associated with the Laschamp, Iceland Basin, Calabrian Ridge 0, and Big Lost events (i.e., II, V, VII, and XV) lasted less than 3 ka, while the episode associated with the B/M transition lasted 8 ka (Table 1). The case of category 2 GDLs is less clear: they are characterized by a large range of duration (1 to 7 ka), likely related to a weaker definition of 10Be overproduction episode limits on the stack record. Further investigations on this record associated with paleomagnetic measurements, as well as other higher‐resolution 10Be cosmogenic analyses of new records globally distributed, are currently underway to refine field behavior systematics.

Finally, it is interesting to compare the minimum and maximum decay rates computed from our long‐term GDM variation study during the Brunhes Chron (0.4 and 1.5 × 1022 Am2 ka−1) with the decay rate computed from archeomagnetic results that shows an acceleration for the last millennia (3 × 1022 Am2 ka−1) [Genevey et al., 2003; Gallet et al., 2014].

Conclusions

The authigenic 10Be/9Be ratio records obtained from low‐latitude cores MD05‐2920, MD05‐2930, and MD90‐0961 reveal the variation of the atmospheric 10Be global production rate between 5 and 850 ka. Correlation of the δ18O records with the standard LR04 stack enables the setting of a robust chronology, allowing comparison with paleomagnetic stacked records of the virtual axial dipole moment obtained from sediments and lavas, as well as with the southeast equatorial Pacific seafloor magnetic anomalies series. These comparisons confirm that significant 10Be enhancements are chronologically related with the global occurrence of geomagnetic dipole lows. A first category of 10Be overproduction episodes corresponds to major geomagnetic dipole lows associated with well‐known paleomagnetic excursions (Laschamp, Iceland‐Basin, Calabrian Ridge 0, and Big Lost) and with the Brunhes‐Matuyama transition. Weaker 10Be overproduction episodes are associated with minor geomagnetic dipole lows associated with less constrained paleomagnetic excursions. The calibration of the 10Be/9Be ratio stacked record using virtual dipole moment values extracted from absolute paleointensities drawn from the PINT and Geomagia databases is supported by the nonlinear relationship empirically established between 10Be production and virtual axial dipole moment values, and its compatibility with the production models. The 10Be‐derived geomagnetic dipole moment record covering the Brunhes Chron and the Brunhes‐Matuyama transition is in general agreement with sedimentary paleomagnetic stacks (namely with the PISO‐1500 stack). It allows to assess characteristic critical states of the dipole moment that triggered either excursion types 1 and 2, or reversals. A critical threshold of ∼4 × 1022 Am2 (50% of the current virtual dipole moment) triggers category 2 excursions, while dipole moment values below 2 × 1022 Am2 trigger category 1 excursions, or reversals. All category 1 10Be overproduction episodes are associated with geomagnetic dipole lows linked to geomagnetic excursions that have durations below 3 ka, while the episode associated with the Brunhes‐Matuyama transition spans a longer (8 ka) time interval. A possible influence of the average dipole moment value on the occurrence of geomagnetic dipole lows is suggested by different mean dipole moment values computed for different time intervals: the 8–300 ka interval with a mean of 7.0 ± 1.9 × 1022 Am2 is characterized by more frequent dipole lows than the 301–775 ka interval with a mean of 8.4 ± 2.3 × 1022 Am2. This study provides new insight on the long‐term geodynamo variation with robust information on amplitudes and timing of geomagnetic dipole moment. The pacing, duration, and change rates characterizing geomagnetic dipole lows reported in this study provide new constraints for geodynamo models.

Supporting Information S1

Click here for additional data file.

Tables S1 to S7

Click here for additional data file.

Data Set S1

Click here for additional data file.