Geochemistry, Geophysics, Geosystems

Reconciling astrochronological and 40Ar/39Ar ages for the Matuyama-Brunhes boundary and late Matuyama Chron

Authors


Abstract

When five Matuyama-Brunhes (M-B) boundary records from the North Atlantic are placed on isotope age models, produced by correlation of the δ18O record directly or indirectly to an ice volume model, the M-B boundary lies consistently at the young end of marine isotope stage 19 with a mean age for the midpoint of the reversal of 773.1 ka (standard deviation = 0.4 kyr), ∼7 kyr younger than the presently accepted astrochronological age for this polarity reversal (780–781 ka). Two recently proposed revisions of the age of the 40Ar/39Ar Fish Canyon sanidine (FCs) standard to 28.201 ± 0.046 Ma and 28.305 ± 0.036 Ma would adjust 40Ar/39Ar ages applicable to the M-B boundary (and other reversals and excursions back to 1.2 Ma) to ages older than the new astrochronological ages by 8–24 kyr. The variables used to construct the ice volume models cannot account for the discrepancy. The FCs standard age that best fits the astrochronological ages is 27.93 Ma, which is within the uncertainty associated with the commonly used value of 28.02 (±0.16) Ma but younger than the recently proposed FCs ages. The EDC2 and EDC3 age models in the Dome C (Antarctic) ice core yield ages of 771.7 ka and 766.4 ka, respectively, for the 10Be flux peak that denotes the paleointensity minimum at the reversal boundary, implying that the EDC2 (rather than EDC3) age model is consistent with the observations from marine sediments, at least close to the M-B boundary.

1. Introduction

The Matuyama-Brunhes (M-B) boundary often constitutes a “golden spike” in the age calibration of sediment sequences. Knowing the age of this polarity chron (reversal) boundary is therefore important to a wide range of studies, although the duration, and possibly the age, of this and other polarity reversals is likely to be a function of site location at scales of a few millennia [Clement, 2004; Leonhardt and Fabian, 2007]. The currently popular age for the M-B polarity chron boundary (780 ka) comes from Shackleton et al. [1990], although the reversal age was adjusted to 781 ka in a more recent time scale [Lourens et al., 2004]. The publication of Shackleton et al. [1990] was revolutionary because it placed the polarity chron boundary 50 kyr older than the previously accepted age (730 ka) based on astrochronologies in marine sediments [Imbrie et al., 1984; Ruddiman et al., 1989] and K-Ar radioisotopic ages [see Mankinen and Dalrymple, 1979]. The new age implied a systematic (∼7%) error in K-Ar ages over the last few Myr, and that previous astrochronologies for the Brunhes chron had “lost” an obliquity cycle (or two precessional cycles). A similar astrochronological age for the M-B boundary (790 ka) was suggested some 8 years earlier [Johnson, 1982], however, it took the better constrained estimate of Shackleton et al. [1990] to achieve the necessary traction to overturn the younger (730 ka) age.

The 780 ka age for the M-B boundary was based on astronomical tuning of benthic and planktic δ18O records from Ocean Drilling Program (ODP) Site 677 in the eastern equatorial Pacific [Shackleton et al., 1990]. The isotope records from ODP Site 677 were matched to an ice volume model [Imbrie and Imbrie, 1980] that was based on the orbital (insolation) solutions of Berger and Loutre [1988]. The procedure was aided by a strongly modulated precession signal in the planktic isotope record (the modulation helping to check for “lost” precession cycles) and a strong obliquity signal in the benthic record. Unfortunately, magnetic stratigraphy could not be obtained at Site 677, and therefore the position of the M-B boundary at Site 677 was unknown. To place the M-B boundary in the Site 677 age model, Shackleton et al. [1990] transferred the M-B boundary from Deep Sea Drilling Project (DSDP) Site 607 to ODP Site 677 by correlation of the two isotope records (Figure 1). The position of the M-B boundary at Hole 607A is known within a few centimeters at 31.84 m depth [Ruddiman et al., 1989] and this level was transferred to the 30.40 m level at Site 677, indicating very similar mean sedimentation rates for the Brunhes Chron at the two sites (∼4 cm/kyr).

Figure 1.

Benthic oxygen isotope data from DSDP Site 607 (green [Ruddiman et al., 1989]), ODP Site 677 (red [Shackleton et al., 1990]), and the LR04 benthic oxygen isotope stack (blue [Lisiecki and Raymo, 2005]) placed on their independent age models. Marine isotope stages are numbered, and the traditional age (780 ka) of the Matuyama-Brunhes (M-B) boundary is indicated.

In the last few years, several high sedimentation rate records, with mean sedimentation rates in the 7–17 cm/kyr range, have become available which provide higher-resolution isotope-paleomagnetic correlations across the M-B boundary. In addition, the European Project for Ice Coring in Antarctica (EPICA) has now obtained deuterium measurements (δDice) at Dome C to 3260 m depth, beyond the M-B boundary [see Jouzel et al., 2007]. The current time scale at Dome C (EDC3) [Dreyfus et al., 2007; Parrenin et al., 2007] has evolved from that provided earlier (EDC2) by EPICA Community Members [2004]. EDC3 is based on an accumulation and ice flow model (as for EDC2), however, it is constrained by orbital tuning of air content and δ18O air records. The position of the M-B boundary can be estimated from 10Be flux measurements at Dome C [Raisbeck et al., 2006; Dreyfus et al., 2008].

Advances in 40Ar/39Ar geochronology including incremental heating of groundmass separates, spurred by the Shackleton et al. [1990] astronomical ages for the last several polarity reversals, led to the first direct dating of lava flows thought to record the M-B reversal [Baksi et al., 1992; Singer and Pringle, 1996]. Lavas recording reverse-transitional-normal magnetization directions in flow sequences on four volcanoes have been dated using the 40Ar/39Ar method. Using the widely adopted age of 28.02 ± 0.16 Ma for the Fish Canyon sanidine (hereafter referred to as FCs28.02) standard [Renne et al., 1998], these ages range from 798.4 to 775.6 ka [Singer et al., 2002; Brown et al., 2004; Coe et al., 2004; Singer et al., 2005]. According to Singer et al. [2005], only the lavas on Maui record the actual M-B reversal at 775.6 ± 1.9 ka, which was apparently preceded by a period of transitional and weakened field, as recorded by the other lava flows between about 798 and 792 ka. Although a paleointensity low does appear in sedimentary records at ∼792 ka, transitional magnetization directions of this age have not been recorded in sediments.

Recently Kuiper et al. [2008] proposed a calibration of the FCs standard that shifts its age to 28.201 ± 0.046 Ma, based on 40Ar/39Ar dating of tephra deposits in the Miocene Melilla basin for which astrochronologic age control is available. Subsequently, Renne et al. [2010] proposed an FCs standard age of 28.305 ± 0.036 Ma based on a statistical optimization that includes pairs of 238U-206Pb and 40Ar/39Ar data from selected samples. It is important to note that the Kuiper et al. [2008] and Renne et al. [2010] ages for FCs use total 40K decay constants of 5.463 × 10−10 a−1 and 5.5492 × 10−10 a−1, respectively, that differ from the value of 5.543 × 10−10 a−1 recommended by Steiger and Jaeger [1977]. If adopted, these recently proposed FCs standard ages, hereafter denoted FCs28.201 and FCs28.305, would shift the age of the lava flows on Maui, and hence the 40Ar/39Ar age of the M-B reversal, to 781 ka and 784 ka, respectively, both of which conflict with the astrochronologic estimates discussed here.

2. North Atlantic Sediment Cores

Coupled isotope-paleomagnetic records across the M-B boundary are available for numerous marine cores (see review by Tauxe et al. [1996]) although very few have sedimentation rates exceeding a few cm/kyr, and fewer still have coupled isotope-paleomagnetic records at high resolution across the boundary interval. Five sites that satisfy the criteria of high (>5 cm/kyr) sedimentation rates and high-resolution coupled isotope-paleomagnetic records are North Atlantic ODP Sites 980, 983, 984, 1063 and Integrated Ocean Drilling Program (IODP) Site U1308 (Figure 2).

Figure 2.

Map showing location of DSDP, ODP, and IODP drill sites mentioned in the text.

We compare the position of the M-B boundary relative to δ18O data at these five sites, to provide estimates for the age of the M-B boundary. The directional magnetic data across the M-B boundary can be represented as virtual geomagnetic polar (VGP) latitudes, derived from magnetization components determined in a particular demagnetization interval. Normalized remanence data provide paleointensity proxies. Benthic and planktic δ18O data are available over the M-B boundary interval for ODP Sites 983 and 984. Benthic δ18O data (only) are available at ODP Sites 980 and 1063 and IODP Site U1308. Details concerning the determination of magnetization components and paleointensity proxies, and the oxygen isotope analyses, over the M-B boundary interval are given by Channell and Kleiven [2000] for ODP Site 983, Channell et al. [2004] for ODP Site 984, Channell and Raymo [2003] and Raymo et al. [2004] for ODP Site 980, and Channell et al. [2008] and Hodell et al. [2008] for IODP Site U1308. For ODP Site 1063, new magnetic and benthic oxygen isotope data were acquired across the M-B boundary to augment the oxygen isotope data of Feretti et al. [2005] and the magnetic data of Shipboard Scientific Party [1998].

In Figures 3 and 4, we compare the oxygen isotope and magnetic records across the M-B boundary at ODP Sites 983 and 984. These records are compared with two alternative ice volume models [Imbrie and Imbrie, 1980]. One model is forced by midsummer insolation at 65°N (the traditional forcing function) based on the Laskar et al. [2004] orbital solution, and the other by integrated summer insolation at 65°N [see Huybers, 2006]. Correlation of isotope records to an Imbrie and Imbrie [1980] ice volume model is a widely used means of establishing the age of isotope records [e.g., Shackleton et al., 1990; Lisiecki and Raymo, 2005]. In Figures 5 and 6, we compare the oxygen isotope and magnetic records across the M-B boundary at ODP Site 980 and IODP Site U1308. Note that age models represented in Figures 36 have not been adjusted for this paper, are independent of one another, and correspond to the age models given in the original publications cited above.

Figure 3.

(top) ODP Site 983 planktic (blue) and benthic (red) oxygen isotope data compared with ice volume models based on midsummer (light blue) and integrated summer (light green) insolation forcing and LR04 from Lisiecki and Raymo [2005] (dashed black line). (bottom) ODP Site 983 virtual geomagnetic polar (VGP) latitude (red) and relative paleointensity proxy (blue). ODP Site 983 data from Channell and Kleiven [2000].

Figure 4.

(top) ODP Site 984 planktic (blue) and benthic (red) oxygen isotope data compared with ice volume models based on midsummer (light blue) and integrated summer (light green) insolation forcing and LR04 from Lisiecki and Raymo [2005] (dashed black line). (bottom) ODP Site 984 virtual geomagnetic polar (VGP) latitude (red) and relative paleointensity proxy (blue). ODP Site 984 data from Channell et al. [2004].

Figure 5.

(top) ODP Site 980 benthic (red) oxygen isotope data compared with ice volume models based on midsummer (light blue) and integrated summer (light green) insolation forcing and LR04 from Lisiecki and Raymo [2005] (dashed black line). (bottom) ODP Site 980 virtual geomagnetic polar (VGP) latitude (red) and relative paleointensity proxy (blue). ODP Site 980 data from Channell and Raymo [2003].

Figure 6.

(top) IODP Site U1308 benthic (red) oxygen isotope data compared with ice volume models based on midsummer (light blue) and integrated summer (light green) insolation forcing and LR04 from Lisiecki and Raymo [2005] (dashed black line). (bottom) IODP Site U1308 virtual geomagnetic polar (VGP) latitude (red) and relative paleointensity proxy (blue). IODP Site U1308 data from Channell et al. [2008] and Hodell et al. [2008].

For some of the oxygen isotope records, the resolution of the data is sufficient to extract the precessional component from the ice volume models and from the δ18O records using a Gaussian band-pass filter centered on a frequency of 0.05 kyr−1 (period of 20 kyr) with a bandwidth of 0.02 kyr−1 (Figure 7). For Site U1308, the data gap in δ18O caused by a barren interval just prior to the M-B boundary (Figure 6) perturbs the output; however, for Sites 983 and 984 the results support the age models.

Figure 7.

Output of Gaussian band-pass filters centered on a period of 20 kyr applied to planktic (blue) and benthic (red) oxygen isotope records, compared with the output from the same filter applied to the ice volume models constructed from midsummer insolation (black) and integrated summer insolation (light green). There are insufficient benthic data at Site 984 and no planktic data at Site U1308.

In Figure 8, we compare the Site 1063 benthic oxygen isotope record of Feretti et al. [2005] based on a variety of species but principally Cibicidides wuellerstorfi or Nuttallides umbonifera with oxygen isotope data from Cibicidides wuellerstorfi and Oridorsalis umbonatus measured at the University of Florida on specimens taken from u channel samples. After completion of magnetic measurements, the u channel samples with 2 × 2 cm2 cross section were subsampled in 5 cm intervals. Specimens were picked from the >212 μm size fraction, and one to five individuals were used for δ18O analysis. The benthic δ18O data of Feretti et al. [2005] were derived from Holes 1063A, 1063B and 1063C in the M-B boundary interval, whereas the u channel samples were taken from Hole 1063D. The δ18O data of Feretti et al. [2005] was augmented using the u channel samples to test the shipboard hole-to-hole correlations built into the composite depth (mcd) scale. With Site 1063 data placed on a uniform age model, the alignments shown in Figure 8 indicate that the composite depth scale is consistent among holes in the M-B interval at Site 1063. Corrections for equilibrium of +0.64‰ and +0.2‰ were used for Cibicidides wuellerstorfi and Oridorsalis umbonatus, respectively.

Figure 8.

(a) ODP Site 983 planktic (blue) oxygen isotope data [Channell and Kleiven, 2000] compared with the IODP Site U1308 benthic (green) oxygen isotope data [Hodell et al., 2008] and the ODP Site 1063 benthic isotope data from Feretti et al. [2005] (red with joining line) and Site 1063 benthic data measured at the University of Florida on Cibicidides wuellerstorfi (orange circles) and Oridorsalis umbonatus (orange triangles) compared with ice volume models based on midsummer (light blue) and integrated summer (light green) insolation forcing. (b) ODP Site 983 (blue) and ODP Site 1063 (red) relative paleointensity proxies. (c) Virtual geomagnetic polar (VGP) latitudes for ODP Site 983 (blue) and ODP Site 1063 (red and orange) where red VGP symbols joined by line and maximum angular deviation (MAD) values (green) represent component magnetizations determined for a fixed 20–80 mT peak field demagnetization range. Orange VGP symbols and MAD values (ocher) represent component magnetizations determined individually (variable peak AF field range) at 1 cm intervals.

The natural remanent magnetization (NRM) components from Core 1063D-16H u channel samples were determined in the 20–80 mT peak alternating field (AF) demagnetization interval, and also by picking components individually from orthogonal projections of AF demagnetization data at 1 cm intervals (Figure 8c). Component declinations were uniformly rotated for the entire core to yield N/S declinations above/below the M-B boundary, respectively. Virtual geomagnetic polar (VGP) latitudes were calculated from the component declination and inclination data, and plotted with the maximum angular deviation (MAD) values [Kirschvink, 1980] associated with the NRM component directions (Figure 8c). Relative paleointensity (RPI) proxies for Site 1063 were determined from the slopes of the best fit lines of NRM versus isothermal remanence (IRM) and anhysteretic remanence (ARM) in the 20–60 mT peak field demagnetization interval (method of Channell et al. [2002]). The VGP latitudes and RPI data from Site 1063 are compared with the VGP latitudes and RPI data from ODP Site 983 (Figure 8b). Note that the ODP Site 983 data are plotted on their published age model [Channell and Kleiven, 2000], and the Site 1063 data are plotted on an isotopic age model that appears consistent with ODP Site 983 (Figure 8a).

3. Age and Duration of the M-B Boundary in Marine Sediments

Polarity reversal is a process in which the ∼180° directional change in the geomagnetic field vector may take a few thousand years and the duration of the directional change depends on site location, with increasing duration at higher latitudes [Clement, 2004]. Modeling of the M-B boundary has been interpreted to imply a likely millennial-scale M-B boundary age discrepancy between the Atlantic and Pacific site locations [Leonhardt and Fabian, 2007], however, a wider distribution of high-resolution M-B boundary records, beyond the Central and North Atlantic, is needed to substantiate this finding. The estimated duration of the M-B directional transition (the time for which the VGP latitudes are below the range associated with “normal” secular variation) is quite variable: 3.5 kyr (Site 980), 4.6 kyr (Site 983), 6.2 kyr (Site 984), 4.6 kyr (Site 1063) and 2.9 kyr (Site U1308) (Table 1). This variability can be attributed to sedimentation rate changes that are not fully accommodated by the age models, and/or differences in magnetic recording efficiency and timing (lock-in) of remanence acquisition.

Table 1. Age and Duration of the M-B Polarity Transition
SiteOnset Age (ka)End Age (ka)Duration (kyr)Reversal Age (ka)
ODP 980774.8771.33.5773.1
ODP 983775.0770.44.6772.7
ODP 984776.5770.36.2773.4
ODP 1063775.0770.44.6772.7
IODP U1308775.0772.02.9773.5
Mean reversal age775.3770.94.4773.1
SD −1σ (kyr)0.70.71.30.4
Dome Ca (EDC2)773.6769.83.8771.7
Dome Ca (EDC3)768.2764.73.5766.4

The directional records across the M-B boundary at ODP Sites 983 and 984 are among the highest-resolution sedimentary records of a polarity reversal. The directional records from both sites show characteristic VGP clusters in the South Atlantic and NE Asia [Channell and Lehman, 1997; Channell et al., 2004]. The ODP Site 984 record shows a prereversal directional “excursion” at ∼781 ka that is not present at other site(s) (Figure 4). The correlation of paleointensity records between ODP Site 983 and 984, using independent age models at the two sites, supports the contention that this apparent prereversal excursion predates the onset of the reversal not only at Site 984, but also at ODP Site 983.

At all five sites (Figures 36 and 8), the M-B directional change, as indicated by VGP latitude, occurs in the transition from marine isotope stage (MIS) 19 to MIS 18. The onset of the polarity transition certainly postdates the δ18O minimum that represents the zenith of MIS 19 (MIS 19.3). Although the published age models for the five sites yield a range of duration estimates for the M-B polarity transition, the age of the midpoint of the polarity transition (labeled “reversal age” in Table 1) has a mean value of 773.1 ka and a standard deviation of 0.4 kyr.

For the majority of marine cores that have yielded records across the M-B boundary, the younger age promoted here (773 ka) is more compatible with the data than the more traditional age (780 ka). Two of the higher-resolution coupled isotope-paleomagnetic records are those from Core MD900963 from the Maldives, Indian Ocean [Bassinot et al., 1994] and from the ODP Site 769 from the Sulu Sea [Oda et al., 2000]. For Core MD900963, the mean Brunhes sedimentation rate is 4 cm/kyr. At this site, the M-B boundary appears to lie in the MIS 19/MIS 18 transition. The planktic δ18O record was correlated to an ice volume model to yield a M-B boundary age of 775 ± 10 ka [Bassinot et al., 1994]. At ODP Site 769, the mean Brunhes sedimentation rate is 8 cm/kyr. The relatively high sedimentation rate is offset by the low resolution of the δ18O record (one sample/20 cm). The age of the M-B boundary was determined by correlation of the planktic δ18O to that of Core MD900963 yielding an M-B boundary age of 778.9 ka [Oda et al., 2000]. Reestimation of the M-B boundary age at these two sites [Tauxe et al., 1996], by correlation of the δ18O records to an ice volume model, yielded M-B boundary ages of 776.7 ka (Core MD900963) and 776.8 ka (ODP Site 769). These age estimates exceed the M-B boundary ages estimated here (Table 1). Note, however, here that the oxygen isotope data in Core MD900963 [Bassinot et al., 1994] and at ODP Site 769 [Linsley and von Breymann, 1991] are compromised by low sedimentation rates (Core MD900963) and low δ18O sampling frequency (ODP Site 769).

4. EPICA Dome C

The deep-sea benthic δ18O stack (LR04 [Lisiecki and Raymo, 2005]) and the δDice, at Dome C [Jouzel et al., 2007] are in general, although not precise, agreement back to ∼800 ka (Figure 9). As the age models for the marine and ice core records are independent, this indicates a close linkage between global ice volume and Antarctic air temperatures. The close consistency of the δDice record with the CO2 record [Lüthi et al., 2008] shows that carbon dioxide concentration is also strongly correlated with global ice volume as recorded by δ18O (LR04) and Antarctic temperatures as recorded by δDice. Similar CO2 oscillations in MIS 3 in the Taylor Dome ice core [Indermühle et al., 2000] coincide with Antarctic Isotope Maximum (AIM) warming events implying that the MIS 19–18 transition has characteristics in common with MIS 3 [Lüthi et al., 2008]. Pol et al. [2010] support this conclusion by labeling the δDice oscillations in the MIS 18–19 transition at Dome C, from younger to older, as AIM A, AIM B and AIM C.

Figure 9.

(a) ODP Site 983 planktic oxygen isotope record (blue) placed on the published age model [Channell and Kleiven, 2000] compared with the LR04 benthic isotope stack (red) [Lisiecki and Raymo, 2005] and with ice volume models based on peak summer (black) and integrated summer (light green) insolation forcing. (b) Dome C δDice on the EDC2 age model (red) and on the EDC3 age model (black) [Jouzel et al., 2007], with AIM events labeled after Pol et al. [2010]. (c) Virtual geomagnetic polar (VGP) latitudes for Site 983 (blue) and Site U1308 (red) denoting the M-B boundary. (d) Median 10Be flux [Raisbeck et al., 2006] placed on the EDC2 age model (red) and EDC3 age model (black). (e) Relative paleointensity proxy for IODP Site U1308 (red) [Channell et al., 2008] correlated to the PISO paleointensity stack [Channell et al., 2009] calibrated to virtual axial dipole moment (VADM).

The interhemispheric phasing of millennial-scale climate change during MIS 3 has been inferred by both methane synchronization of Greenland and Antarctic ice cores [EPICA Community Members, 2006] and by comparing the timing of planktic and benthic δ18O changes in cores from the Portuguese margin that apparently record both Antarctic and Greenland signals [Shackleton et al., 2000, 2004]. For reasons not well understood [Skinner et al., 2003], the benthic δ18O signal on Portuguese Margin, and elsewhere in the North Atlantic [Hodell et al., 2010], resembles Antarctic temperature variations whereas planktic δ18O records from the same locations mimic stadial-interstadial oscillations in Greenland. By comparing the timing of planktic and benthic δ18O changes, Shackleton et al. [2000, 2004] confirmed the phasing deduced by methane synchronization of Greenland and Antarctic ice cores [Blunier et al., 1998; Blunier and Brook, 2001; EPICA Community Members, 2006]. The Antarctic warm events occur during the longest, coldest stadials in Greenland, and are followed by abrupt warming in Greenland as Antarctica begins to cool. This pattern has been referred to as the “bipolar seesaw” [Broecker, 1998] and is explained by changes in heat transport related to thermohaline circulation referred to as Atlantic Meridional Overturning Circulation (AMOC).

The M-B boundary at ODP Site 983 lies within the oldest of the three planktic δ18O minima (Figure 3), at an age that would coincide with the oldest of the δDice oscillations, on the EDC3 or EDC2 age model, at Dome C (Figure 9). Due to the shielding effect of the geomagnetic field on cosmic ray flux, geomagnetic field intensity can be estimated from cosmogenic isotope flux in ice cores [see Muscheler et al., 2005], and the position of the M-B boundary in the Dome C ice core can be estimated from the flux of 10Be [Raisbeck et al., 2006; Dreyfus et al., 2008]. The median flux 10Be at Dome C shows two broad peaks, separated by about 20 kyr at ∼770 and ∼790 ka, that appear to correlate with two prominent lows in the paleointensity (RPI) records, particularly in the Site U1308 RPI record (Figure 9). The two lows in paleointensity are less well defined at ODP Sites 980/983/984, although the initial dip in paleointensity at all four sites occurs coincidentally at ∼792 ka (Figures 35). If we associate the maximum in 10Be flux with the M-B boundary at Dome C, the apparent age of the M-B boundary age in the marine cores is more consistent with the EDC2 age model than the EDC3 age model that yield M-B boundary ages of 772 and 766 ka, respectively (Table 1).

At Site 983, the planktic δ18O minima are phase shifted relative to weakly manifest δ18O oscillations in the benthic record (Figure 10). The decreases in benthic δ18O occur on the MIS 19–18 transition appear to occur at times of maximum planktic δ18O (i.e., cold events). This is similar to MIS 3 when the Antarctic warm events coincide with the Greenland stadials [EPICA Community Members, 2006]. Oscillations of similar age are observed in the Site U1308 benthic record (Figure 6) and the Site 1063 benthic record (Figure 8), although the age models are insufficiently precise to resolve synchronicity or phase shifts. Following Shackleton et al. [2000, 2004], the benthic δ18O signal from the particular locations in the North Atlantic tracks Antarctic temperature variations. Assuming a similar relationship for Site 983, we synchronize the δDice at Dome C to the benthic δ18O at Site 983 (Figure 10). The adjustment of the δDice record from the EDC2/EDC3 age models (Figure 9) results in a close synchronization of increase and decrease in 10Be flux with the onset and end of the directional transition (as indicated by VGP latitude) at the M-B boundary (Figure 10), providing tacit support for the synchronization of the benthic δ18O at Site 983 and the δDice at Dome C, and hence for the “bipolar seesaw.”

Figure 10.

(a) ODP Site 983 planktic oxygen isotope record (blue) placed on the published age model [Channell and Kleiven, 2000] compared with ice volume models based on midsummer (black) and integrated summer (light green) insolation forcing. (b) Dome C δDice (red) from Jouzel et al. [2007] shifted in age to correlate with the ODP Site 983 benthic oxygen isotope record (black). Purple vertical lines emphasize the correlation of Site 983 benthic δ18O (and δDice) minima to Site 983 plantkic δ18O maxima (analogous to observations in MIS 3 in the North Atlantic, see text). (c) The median 10Be flux [Raisbeck et al., 2006] placed on the resulting age model for Dome C brings the peak in 10Be flux at 771 ka in coincidence with (d) the change in VGP latitude denoting the Matuyama-Brunhes (M-B) boundary at Site 983. Light gray shading indicates the M-B boundary polarity transition interval at ODP Site 983 and its implied correlation to the Dome C ice core record adjusted in age to synchronize Site 983 benthic δ18O and δDice.

5. The 40Ar/39Ar Ages

Due to the sporadic nature of volcanic eruption and the brief (few thousand years) duration of polarity reversal, the capture of magnetization directions recording polarity transitions is highly fortuitous. Arguably, the best constrained 40Ar/39Ar age for the M-B boundary is from the Haleakala caldera on Maui [Coe et al., 2004]. Here, 24 flows over about 36 m of section record transitional magnetization directions characterized by large-scale R-N-R swings in VGP latitude. Six of these flows yielded 40Ar/39Ar ages of 773.0 ± 3.0 ka to 785.1 ± 8.0 ka (not in stratigraphic order) that gave an inverse variance weighted mean of 775.6 ± 1.9 ka [Coe et al., 2004; Singer et al., 2005], relative to FCs28.02. The paleomagnetic record at Haleakala (Maui) is complicated by a hiatus in the volcanic activity of more than 220 ka. The 24 transitional flows that record the M-B reversal lie directly atop a sequence of 16 flows, also transitionally magnetized, that span about 30 m of section where six 40Ar/39Ar determinations give a mean age of 900.3 ± 4.7 ka. These ∼900 ka excursional magnetization directions are associated with the Kamikatsura excursion [Coe et al., 2004]. This excursion has not been convincingly recorded in sediments and derives its name from transitional magnetizations in the Kamikatsura Tuff of SW Japan [Maenaka, 1983; Takatsugi and Hyodo, 1995], and has been identified in a single transitionally magnetized lava flow on Tahiti that is 40Ar/39Ar dated relative to FCs28.02 at 904 ± 11 ka [Singer et al., 1999].

The 775.6 ka M-B boundary age from Maui becomes 781 ± 2 ka if calculated relative to FCs28.201 [Kuiper et al., 2008] and 783.5 ± 2 ka relative to FCs28.305 [Renne et al., 2010] (Figure 11). Kuiper et al. [2008, Table S3] recalculated 16 published 40Ar/39Ar ages applicable to the M-B boundary thereby shifting M-B boundary ages from the 772–800 ka range to the 780–813 ka range. It is appropriate, when comparing 40Ar/39Ar ages to independent astrochronologies, to expand uncertainty estimates to include, in addition to analytical error, intrinsic uncertainties associated with the age of the FCs standard and 40K decay constant. Accordingly, the recalculated Maui M-B age and accompanying uncertainty using FCs28.201 [Kuiper et al., 2008] becomes 781 ± 4 ka. Note that the values given by Kuiper et al. [2008, Table S3] do not include the full uncertainties. The supposed M-B boundary ages from Chile, Tahiti and La Palma (summarized by Singer et al. [2005]) yield average ages of 791.7, 794.8, and 798.4 ka, respectively, relative to FCs28.02. Recalculation of these ages to FCs28.201 yields ages of 797, 800 and 804 ka, and to FCs28.305 yields ages of 800, 803 and 807 ka (Table 2 and Figure 11). Even before recalculation, the M-B boundary age estimates from Chile, Tahiti and La Palma are over 15 kyr older than the estimate from Maui, leading Singer et al. [2005] to link the age estimates from Chile, Tahiti and La Palma with transitional magnetization directions associated with a paleointensity low that precedes the M-B boundary, as seen at Site U1308 and Site 980 (Figures 5 and 6) and in the 10Be flux at Dome C (Figure 9). This prereversal paleointensity low has not (as yet) been associated with transitional magnetization directions in sedimentary records.

Figure 11.

The 40Ar/39Ar ages from transitionally magnetized lavas on four volcanoes computed using the FCs28.305 calibration (green), FCs28.201 calibration (blue), and “optimal” FCs27.93 estimate (red). Astrochronological age and uncertainty and EDC2 and EDC3 ages for the Matuyama-Brunhes (M-B) boundary (gray vertical bars) are from Table 1. All uncertainties shown at the two sigma level of precision. Note that for the Maui lavas at the M-B boundary, propagating the full uncertainty of the FCs28.201 or the FCs28.305 calibrations (including errors in the standard age and 40K decay constant) would result in an uncertainty of ±4 ka, which would double the width of the mean age boxes in blue and green; yet this remains insufficient to bring the 40Ar/39Ar ages into overlap with the astrochronological estimate.

Table 2. Age Estimates for Reversals and Excursionsa
 Event and Location
M-B, MauiPreM-B, ChilePreM-B, TahitiPreM-B, La PalmaS. RosaTop JarBase JarPunaCobb
Referenceb111123334
Ar-Ar age in reference (ka)775.6791.7794.8798.49361001106911221193
FCs28.3057848008038079461011108011331205
FCs28.2017817978008049421008107611291201
FCs27.93773789792796933998106611181189
Astro (ka)773794794794932987106811151190
EDC2772797797797     
EDC3766794794794     

For reversals and excursions between the M-B boundary and 1.2 Ma, the FCs28.201 calibrations yield a consistent discrepancy of 1%–2% (8–21 kyr), reaching over 2% (12–24 kyr) for FCs28.305, relative to ages derived from astronomically calibrated marine sediments (Table 2 and Figure 12).

Figure 12.

Astronomical estimates and ice core estimates (using EDC 2 and EDC 3 age models) for the ages of geomagnetic reversals and excursions in the 700–1250 ka interval compared with 40Ar/39Ar ages calculated relative to FCs27.93 (the optimal FCs for consistency with astrochronology), FCs 28.201, and FCs28.305. Data from Table 2.

6. Astrochronologies

Astrochronological ages are determined by matching cycles in sedimentary sequences to astronomic solutions for orbital cycles, or to calculations of solar insolation for a particular latitude and time of year. Bedding rhythms, physical property data and geochemical parameters, almost any parameter that appears cyclic in sedimentary sequences, have been matched to astronomic solutions to construct astrochronologies. For eccentricity, the orbital solution is considered to be precise at least over the last 40 Myr, however, the solution for precession and obliquity is less accurate due to uncertainties related to tidal dissipation in the Earth-Moon system. These uncertainties can be gauged by computing the difference between the solution for present-day values of tidal dissipation and half the present-day value of tidal dissipation, and these differences increase to ∼1 cycle for both precession and obliquity at ∼15 Ma [Lourens et al., 2004]. For the last few million years, uncertainties in astronomical solutions that use modern values of dynamical ellipticity and tidal dissipation, are probably insignificant, however, another uncertainty comes to the fore for young sequences, namely, the time lag between orbital forcing and response, that may exceed several kyr, and is therefore particularly important for young (Quaternary) ages.

Benthic oxygen isotope records, that are the basis for age models described here, are generally considered to represent a measure of global ice volume, and for this reason, can be used as a global stratigraphic tool, although benthic δ18O record is influenced by changes in benthic temperature and water chemistry [e.g., Skinner et al., 2003; Sosdian and Rosenthal, 2009]. Following Imbrie and Imbrie [1980], the lag between insolation forcing and response for ice volume is traditionally accommodated by construction of an ice volume model of the form:

equation image

where the ice volume (y) depends on the insolation forcing (x), a mean time constant (Tm) and a nonlinearity function (b) subtracted during ice growth and added during ice decay. The calculation of insolation requires selection of latitude and time of year, and, according to classic Milankovitch theory, midsummer (21 June) at 65°N is generally selected. Huybers [2006] has argued that integrated summer insolation (at 65°N) is a more appropriate forcing function as it incorporates the duration of the summer season. The benthic isotope stack of Lisiecki and Raymo [2005], referred to as LR04, was calibrated using an ice volume model constructed from midsummer insolation at 65°N using the orbital solution of Laskar et al. [1993] with b and Tm values of 0.6 and 15 kyr, respectively, for the last 1.5 Myr. In LR04, the time constant of 15 kyr was selected to maximize agreement with independent age estimates for the last 135 kyr. For the 5.3–3.0 Ma interval of LR04, values of b and Tm of 0.3 and 5 kyr, respectively, were used with linear increases to values of 0.6 and 15 kyr during the 3.0–1.5 Ma interval.

Termination IX (the MIS 19–20 boundary) is the prominent marker in δ18O records just prior to the M-B boundary. In Figure 13, we test the sensitivity of the age of Termination IX in the ice volume models to changes in b, Tm and the choice of latitude for insolation calculation. Two types of insolation data are used for the test: midsummer insolation calculated using the Laskar et al. [2004] orbital solutions, and the integrated summer insolation estimated using the Huybers [2006] approach. For the integrated summer insolation, a threshold of 275 W/m2 is used for the calculation at 65°N and 85°N, and a threshold of 425 W/m2 is used for the calculation at 35°N. For each ice volume model, the age of Termination IX is determined as the age when the rate of ice volume decrease reaches a maximum within the 810–785 ka interval. Ice volume models calculated using integrated summer insolation yield Termination IX ages about 1–2 kyr older than those estimated using the midsummer insolation data (Figure 13). The value of nonlinearity (b) in the 0.3–0.6 range does not make a difference of more than 0.5 kyr in the age of Termination IX. The most important variable in the ice volume model is the choice of Tm where values ranging from 5 to 15 kyr change the age of the Termination by about 2 kyr (Figure 13). Note little (<0.5 kyr) change in the age of the Termination as Tm increases beyond 15 kyr. For ice volume models calculated using midsummer insolation, low (<0.5 kyr) dependence of Termination age on latitude is observed. Choice of latitude changes the age of Termination IX by about 1–2 kyr if integrated summer insolation is used.

Figure 13.

Age of Termination IX (MIS 19/20 boundary) in ice volume models as a function of b (nonlinearity), Tm (mean time constant), and (left) latitude of midsummer insolation and (right) latitude of integrated summer insolation. For integrated summer insolation, an insolation threshold of 275 W/m2 was used for latitudes other than 35°N, for which 425 W/m2 was used.

7. Conclusions

The five North Atlantic marine sites discussed here carry the highest-resolution coupled isotope-paleomagnetic records across the M-B boundary. These records have higher resolution (higher sedimentation rates) than records used in earlier analyses of the M-B boundary [e.g., Tauxe et al., 1996; Leonhardt and Fabian, 2007; Suganuma et al., 2010]. The recent records, when placed on their independent age models, yield closely consistent estimates for the age of the M-B boundary with a mean age of 773.1 ka, and standard deviation of 0.4 kyr (Table 1). This M-B boundary age is close to the age of the 10Be peak in EPICA Dome C ice core when placed on the EDC2 age model (772 ka), whereas the EDC3 age model yields an even younger age (766 ka). The M-B boundary coincides with the oldest of three millennial-scale fluctuations in planktic and benthic δ18O in the North Atlantic (Figure 3), and with similar oscillations in δDice at Dome C (Figure 9). At Site 983, the planktic and benthic δ18O oscillations appear out of phase and may be interpreted, following similar observations on Portuguese Margin in MIS 3 [Shackleton et al., 2000, 2004], as out-of-phase Greenland and Antarctic temperature signals manifest in planktic and benthic δ18O, respectively. If Site 983 benthic δ18O is synchronized with δDice oscillations at Dome C by shifting the δDice record, consistent with the Portuguese Margin MIS 3 analog, the ice core 10Be flux peak coincides with the marine M-B boundary, thereby supporting the revised age for the M-B boundary.

In three Pacific cores with mean sedimentation rates in the 0.66–1.19 cm/kyr range, the M-B boundary from paleomagnetic measurements is offset down-core by ∼15 cm from sedimentary 10Be flux maxima associated with M-B boundary paleointensity minima [Suganuma et al., 2010]. From isotopic tracers, the bioturbated surface layer in pelagic sediments has thickness around 10 cm, varying in the 3–30 cm range, and is largely independent of sedimentation rate [Boudreau, 1994, 1998]. The efficiency of mixing in the bioturbated layer [see Guinasso and Schink, 1975] is often sufficiently high that remanence acquisition (lock-in) must occur below this layer [see Channell and Guyodo, 2004]. Although a 15 cm lock-in depth corresponds to ∼15 kyr remanence delay in the Pacific cores studied by Suganuma et al. [2010], a similar lock-in depth in the sediments studied here would correspond to a delay in remanence acquisition of ∼1 kyr, assuming bioturbation depth and hence lock-in depth are independent of sedimentation rate.

The M-B reversal occurs at the young end of MIS19, at the onset of the transition to MIS 18. Termination IX in LR04 lies at ∼788–789 ka because Lisiecki and Raymo [2005] calibrated their δ18O stack using an ice volume model forced by midsummer insolation at 65°N with time constant (Tm) equal to 15 kyr, and nonlinearity (b) of 0.6 (Figure 13). For midsummer insolation forcing, values of Tm of 5 kyr would increase the age of the Termination to ∼790 ka but the flexibility in age of the Termination is less than 2 kyr for reasonable values of Tm and b (Figure 13). For integrated summer insolation forcing, for Tm = 15, the age of Termination IX (at 790 ka) is older by ∼2 kyr than for the midsummer insolation calculation using the same variables. In summary, flexibility in ice volume models can account for variations in the age of Termination IX of ∼3 kyr.

The observed discrepancy between astrochronological and 40Ar/39Ar age estimates for the M-B boundary and late Matuyama reversals and excursions (Table 2 and Figure 12), when the FCs28.201 or the FCs28.305 standard age is used, cannot be accounted for by changing variables associated with the ice volume models that provide the astrochronological calibration of benthic δ18O records. The age of the FCs standard that best fits the astrochronological ages is 27.93 Ma, within the error range of the widely adopted value of 28.02 ± 0.16 Ma [Renne et al., 1998].

The Kuiper et al. [2008] FCs28.201 calibration of this standard is based on 40Ar/39Ar dating of sanidine phenocrysts extracted from tephra layers in the Messinian Melilla Basin (Morocco). The 40Ar/39Ar ages were correlated, using six biostratigraphic events, to astrochronologies (based on sapropels and bedding rhythms) in the Sorbas Basin of Spain [Sierro et al., 2001; van Assen et al., 2006]. Sedimentary cycles in the Melilla sections [Kuiper et al., 2008, p. 500] “lack the expression of characteristic details.…common in Mediterranean sapropel sequences.” The absence of an interpretable cyclostratigraphy in the Melilla sequences from which the 40Ar/39Ar ages were derived means that the astrochronological calibration of the Melilla tephras relies heavily on the biostratigraphic correlations between the Melilla and Sorbas basins [Sierro et al., 2001; van Assen et al., 2006]. Note that no usable magnetic stratigraphy was resolved in the Mellila sections [van Assen et al., 2006], and hence the correlation of 40Ar/39Ar ages to the magnetic stratigraphy (C3Ar-C3r) in the Sorbas Basin [Sierro et al., 2001] also relies on the correlation of these six biostratigraphic events.

Part of the attraction of the new FCs28.201 age, and even more so the FCs28.305 calibration, is that it [Kuiper et al., 2008, p. 502] “eliminates the documented offset of the conventionally calibrated 40Ar/39Ar and U/Pb dating systems in many volcanic rocks” which has become a problem in geochronology as the analytical precision has improved. However, there are exceptions in the observation of offsets, for example, the normally magnetized Bishop Tuff in California lies just above the M-B boundary in sediment sections across the western USA and laser fusion of sanidine from this unit gives a 40Ar/39Ar age of 774 ± 3 ka for the FCs28.02 [Sarna-Wojcicki et al., 2000]. Sedimentation rate estimates from the various sections led Sarna-Wojcicki et al. [2000] to conclude that the M-B reversal occurred 15 kyr before the eruption of the Bishop Tuff, at 789 ka. If the 774 ka age of the Bishop Tuff is recalculated to 779 ka using FCs28.201, the age of the M-B boundary becomes 794 ka, although Kuiper et al. [2008] incorrectly report (their Table S3) an age of 791 ka. An age of 794 ka for the M-B boundary would imply that it predates Termination IX, clearly unacceptable. Isotope dilution thermal ionization mass spectrometry of zircons in the Bishop Tuff [Crowley et al., 2007] yields a 206Pb/238U age that is indistinguishable from the 40Ar/39Ar age of Sarna-Wojcicki et al. [2000] using FCs28.02. Whereas the 206Pb/238U age of the Bishop Tuff requires a large correction for Th/U fractionation, and is thus not without controversy, Crowley et al. [2007] were able to use a Th/U ratio of 2.81 measured in quartz-hosted melt inclusions as a reasonable constraint. An exceptionally high melt Th/U ratio of 6.0 would be required to shift the 206Pb/238U age to 779 ka, consistent with the sanidine age of Sarna-Wojcicki et al. [2000] using FCs28.201, and an even more extreme melt Th/U ratio would be required to push the 206Pb/238U age to a value consistent with FCs28.305. Kuiper et al. [2008] noted that their new Fish Canyon standard age pushes 40Ar/39Ar ages for the M-B boundary older than astrochronological estimates, and suggested that the M-B boundary (and presumably other reversal/excursion transitions) are diachronous between the marine sediments and the terrestrial lava sequences from which the 40Ar/39Ar ages were derived. Based on the systematic age discrepancy in astrochronological and 40Ar/39Ar ages using FCs28.201 and FCs28.305 (Table 2 and Figure 12), there appears to be a fundamental problem with either the 40Ar/39Ar and/or the U-Pb dating systems at this level of resolution.

40Ar/39Ar ages using FCs28.02 are clearly more compatible with Quaternary astrochronological age estimates than ages using FCs28.201 or FCs28.305. In spite of its more controversial 206Pb/238U age, the Bishop Tuff 40Ar/39Ar age most consistent with the M-B boundary astrochronological age is compatible with FCs28.02, but not with either FCs28.201 or FCs28.305. The problem with FCs28.201 may stem from the astronomical age model for the Melilla Basin used by Kuiper et al. [2008], possibly through facies-related diachroneity of the biostratigraphic events that link the Sorbas Basin (Spain), where the astrochronologies originate, to the Melilla Basin (Morocco), where the 40Ar/39Ar ages originate.

Does the problem lie in the dating of the lava flows correlated to the marine astrochronology? The large numbers of 40Ar/39Ar incremental heating experiments on groundmass separated from the transitionally magnetized lavas from Maui [Coe et al., 2004] and lavas recording the reversals at the top and base of the Jaramillo subchron and the Punaruu excursion [Singer et al., 1999] reveal no evidence that argon loss may have lowered the ages. Moreover, the 40Ar/39Ar ages for the Santa Rosa excursion and Cobb Mountain Subchron (Table 2) are based on laser fusion analyses of sanidine, a mineral widely regarded as resistant to argon loss. We find it difficult to imagine a process whereby the ages of both lava groundmass and sanidine crystals in this large set of lava flows could be reduced systematically such that they match each of the astrochronologic estimates.

Our findings illustrate that the standardization of the 40Ar/39Ar method remains incompletely resolved. Although a FCs age of 27.93 Ma is consistent with the Quaternary astrochronological ages documented here, and is close to the age of 28.02 Ma advocated by Renne et al. [1998], Smith et al. [2010] find that FCs28.201 [from Kuiper et al., 2008] provides consistency between 40Ar/39Ar and U-Pb ages for ash beds in the Eocene Green River Formation, indicating that Cenozoic discrepancies between 40Ar/39Ar and U-Pb ages are apparently not resolved by choice of any one FCs age, but may be attributed to other issues such as inheritance in U-Pb ages and decay constant uncertainty.

Acknowledgments

We thank Paul Renne, Joel Baker, Matt Heizler, and an anonymous reviewer for journal reviews and Kyle Min and David Sawyer for comments on an early version of the manuscript. Patricia Feretti provided stable isotope data from the M-B boundary interval at ODP Site 1063. Research was supported by NSF through grants OCE-0850413, OCE-1014506, and EAR-0959108.

Ancillary