Timing is crucial to understanding the causes and consequences of events in Earth history. The calibration of geological time relies heavily on the accuracy of radioisotopic and astronomical dating. Uncertainties in the computations of Earth's orbital parameters and in radioisotopic dating have hampered the construction of a reliable astronomically calibrated time scale beyond 40 Ma. Attempts to construct a robust astronomically tuned time scale for the early Paleogene by integrating radioisotopic and astronomical dating are only partially consistent. Here, using the new La2010 and La2011 orbital solutions, we present the first accurate astronomically calibrated time scale for the early Paleogene (47–65 Ma) uniquely based on astronomical tuning and thus independent of the radioisotopic determination of the Fish Canyon standard. Comparison with geological data confirms the stability of the new La2011 solution back to ∼54 Ma. Subsequent anchoring of floating chronologies to the La2011 solution using the very long eccentricity nodes provides an absolute age of 55.530 ± 0.05 Ma for the onset of the Paleocene/Eocene Thermal Maximum (PETM), 54.850 ± 0.05 Ma for the early Eocene ash −17, and 65.250 ± 0.06 Ma for the K/Pg boundary. The new astrochronology presented here indicates that the intercalibration and synchronization of U/Pb and 40Ar/39Ar radioisotopic geochronology is much more challenging than previously thought.
 Temporal relations are the key to causality arguments in Earth's history [Renne et al., 1998; Min et al., 2000; Kuiper et al., 2008; Renne et al., 2010]. Our understanding of global change mechanisms and the dynamics of a rapidly changing climate system depends upon a precise knowledge of the dates as well as the rates of change in the geological past. An accurate timescale is the backbone for reconstructing Earth's history in order to identify mechanisms associated with past prominent global events like the Paleocene/Eocene Thermal Maximum (PETM) more than 55 myr ago. As a numerical dating method, astronomical tuning, the correlation of cyclic variations in the geological record to astronomical computations of the insolation quantities on Earth, has revolutionized the age calibration of the geological time scale [Shackleton et al., 1990; Lourens et al., 2004; Hilgen, 2010]. An astronomically tuned geological time scale is now available for the last 40 million years [Lourens et al., 2004; Pälike et al., 2006b]. Efforts to extend the astronomically calibrated time scale into the early Paleogene have been hampered by fundamental problems related to the early to-late Eocene “cyclostratigraphic gap,” an interval in the middle Paleogene that has not yet been completely covered by cyclostratigraphic studies in pelagic sediments [Pälike and Hilgen, 2008], by the uncertainties and limits of astronomical calculations [Laskar et al., 2004, 2011b], and by uncertainties in radioisotopic age constraints [Machlus et al., 2004; Westerhold and Röhl, 2009].
 Here we try to solve the dating dilemma by comparing the expression of the very long eccentricity-cycle minima in new orbital solutions for eccentricity (La2010 [Laskar et al., 2011a] and La2011 [Laskar et al., 2011b]) with geological data (Figure 1) that contain eccentricity-modulated precession cycles.
2. La2010 and La2011 Orbital Solutions for Eccentricity
 At present, the most advanced Earth orbital and rotational solution has been obtained by a direct numerical integration of the planet orbits and the precession of the Earth spin axis – the La2004 solution. This solution has been used for the astronomical calibration of the Neogene [Lourens et al., 2004] in the GTS2004 geological time scale [Gradstein et al., 2004]. The time of validity of this solution is estimated to be at ∼40 Ma. Extending the astronomically calibrated geological time scale into the Paleogene is hampered by the limited accuracy of available orbital solution in times older than 40 Ma [Laskar et al., 2004; Machlus et al., 2004] (Figure 2a). Nevertheless building an orbitally calibrated stratigraphic framework based on the identification of the stable long eccentricity cycle (405-kyr) is still possible [Laskar et al., 2004] for older periods. Although in the absence of long continuous records it results in floating time scales only.
 At present, there is a large international effort toward the construction of a complete astronomically calibrated geological time scale over the full Cenozoic era (GTSnext; http://www.gtsnext.eu). This will require the accomplishment of an orbital solution for the Earth motion back to 65 Ma. But this is not an easy task, as due to the chaotic motion of the system [Laskar, 1990, 1999], the initial uncertainty is multiplied by 10 every 10 myr. Extending the La2004 solution from 40 to 65 Ma is thus equivalent to an improvement of the full gravitational model of more than 2 orders of magnitude. In order to improve the long-term ephemerides, the group of the Institut de Mécanique Céleste et de Calcul des Éphémérides (IMCCE) at Paris Observatory developed a new high precision planetary ephemeris called INPOP [Fienga et al., 2008, 2009, 2011] (Intégration Numérique Planétaire de l'Observatoire de Paris). This precise model is adjusted to all available terrestrial and space mission tracking observations. An innovative feature of the INPOP ephemerides compared to the available equivalent solutions from JPL/NASA, is the capacity to be extended on long time, as much as 1 myr, which is then used as reference for the construction of the long-term planetary ephemerides La2010 [Laskar et al., 2011a] that provide a reliable evolution of the Earth orbit over 50 Ma. In fact, in order to evaluate the uncertainty of the long-term solution, several solutions were computed corresponding to various settings: La2010a, b, c were adjusted to the INPOP08 ephemeris [Fienga et al., 2009], while La2010d was adjusted to the previous model INPOP06 [Fienga et al., 2008].
 More recently, using an increased set of observations, the new high precision INPOP10a ephemeris was released [Fienga et al., 2011]. This allowed the construction of a new long-term solution La2011 [Laskar et al., 2011b] which benefited also from improvements in the numerical algorithm for the computation of the solution. As a result, the integration time in La2011 was decreased while the numerical accuracy was increased [Laskar et al., 2011b]. The solution La2011 was constructed together with clones of the solution (La2011m2, La2011p2, La2011m4) with changes in the initial position of the minor planets of respectively −15 m, +15 m, and −1.5 mm, much less than the present determination of their position which amounts to several hundreds of km. Moreover, the same study [Laskar et al., 2011b] demonstrated that the motion of Ceres and Vesta, is highly chaotic. As a consequence, a precise calculation of Earth's eccentricity beyond 60 Ma is not, and probably will never be, possible [Laskar et al., 2011b]. This makes direct tuning of geological data to the short eccentricity curve (100-kyr) impossible for interval older than 60 Ma and demonstrates that an astronomically tuned framework beyond 60 Ma can only be based on the identification of the stable long eccentricity cycle (405-kyr) [Laskar et al., 2004].
 On the opposite, before 50 Ma, the solution La2010 and even better, La2011 can be used for a fine tuning to the eccentricity curve. The interval 50 Ma to 60 Ma remains in question, as it is difficult at present to assert the validity of the orbital solutions La2010 and La2011 in this range at the short eccentricity level.
 In order to test the stability of the La2010 and La2011 nominal solutions with respect to the very long eccentricity modulation (∼2.4 myr), other solutions with slightly different settings have been computed and compared. The extraction of the amplitude modulation of the four different eccentricity solutions in La2010 suggests that this solution is not reliable beyond 52 Ma (Figure 2b). In the same way, the amplitude modulation of four different eccentricity solutions in the new La2011 shows that the solution cannot be trusted after 60 Ma (Figure 2c), but this upper bound does not provide information on how far these solutions can be trusted, as the different La2011 solutions are very small variations around the same solution. On the opposite, comparing the amplitude modulation between La2010a, La2010d and La2011 (Figure 2d) shows that the La2010d solution is similar to La2011 up to 54 Ma. This is a very strong indication that the La2010d and even better the La2011 solution is valid over 54 Ma. Indeed, La2010d is based on the INPOP06 ephemeris [Fienga et al., 2008], while La2010a, b, c are based on INPOP08 [Fienga et al., 2009]. The construction of INPOP10a [Fienga et al., 2011] has now demonstrated that the long-term behavior of INPOP08 [Fienga et al., 2009] is not as good as the one of INPOP06 [Fienga et al., 2008], which is closer to INPOP10a [Fienga et al., 2011] which is used as a reference for La2011. Comparison of the two independent solutions La2010d and La2011 is thus a good way to evaluate the time of validity of these solutions, which can be set to 54 Ma. At least the amplitude modulation of eccentricity is stable back to 54 Ma for the La2011 solution as can be concluded from the comparison between geological data and the La2010a, La2010d and La2011 solutions (see Discussion).
3. Geological Data
 The geological data we use are iron (Fe) intensity data based on X-ray fluorescence (XRF) core-scanning measurements of marine sediments drilled at Ocean Drilling Program (ODP) Sites 1258 [Westerhold and Röhl, 2009] (Leg 207; Demerara Rise) and 1262 [Westerhold et al., 2007, 2008] (Leg 208; Walvis Ridge), located in the Atlantic Ocean (Figure 1). Both records have a robust cyclostratigraphic framework based on the stable 405-kyr cycle from 47 to 62 Ma and exhibit well expressed very long eccentricity cycle minima [Lourens et al., 2005; Westerhold et al., 2007, 2008; Westerhold and Röhl, 2009]. Most important, within this rigid stratigraphic framework, the two data sets can only be moved in 405-kyr steps.
3.1. Additional Bulk δ13C and XRF Core Scanning Data
 Stable isotope measurements on 218 powdered (freeze-dried) bulk sediment samples from ODP Site 1258 were performed on a Finnigan MAT 251 mass spectrometer equipped with an automated carbonate preparation line at MARUM - University Bremen. The carbonate was reacted with orthophosphoric acid at 75°C. Analytical precision based on replicate analyses of in-house standard (Solnhofer Limestone) averages 0.05‰ (1σ) for δ13C and 0.07‰ (1 σ) for δ18O. All data are reported against VPDB after calibration of the in-house standard with NBS-19 and available in Table S1 in theauxiliary material.
 Detailed comparison of the Fe intensity and bulk δ13C data of ODP Sites 1258 and 1262 (Figure 3) reveal a doubling of sediment due to a fault at 126.14 rmcd in Core 1258B-14R. About 1.22 m of the sedimentary succession has been doubled by the fault. A similar feature which might indicate another doubling of sediment in the section was not observed. Consequently the cyclostratigraphy for ODP Site 1258 [Westerhold and Röhl, 2009] was revised. The 100-kyr filter (seeFigure 3) indicates that one short eccentricity cycle less is present between H1 and I1 as proposed before. The revision results in a one step shift in the short eccentricity cycle numbers from E10021 to E10086 (e.g., E10021 shifts to E10020). It is important to note that the above correction by one short eccentricity cycle does not affect the number of 405-kyr cycles previously identified at Site 1258 [Westerhold and Röhl, 2009].
4. Astronomical Calibration of the Paleocene and Early Eocene
 The well expressed very long eccentricity cycle in the XRF core scanning Fe data from ODP Site 1258 and 1262 provide the unique opportunity (1) to test the La2010 and La2011 solutions in the stable part from 47 to 54 Ma and (2) to anchor the floating stratigraphic framework for the Paleocene and early Eocene to obtain accurate absolute ages. The basic age model for this comparison is the consistent recognition of the number of stable 405-kyr cycles in the geological data [Lourens et al., 2005; Westerhold and Röhl, 2006; Westerhold et al., 2007, 2008; Hilgen et al., 2010] from 47 to 60 Ma. This cyclostratigraphic framework can be shifted by 405-kyr increments in time.
4.1. Testing the Stability of the La2010 and La2011 Solutions
 Before we can start the test we need to know the phase relationship between δ13C, Fe intensity data and eccentricity. Bulk δ13C data from ODP Site 1258 and 1262 show that minima in δ13C correspond with peaks in Fe (i.e., carbonate dissolution), both of which appear to be in phase with maxima in eccentricity [Lourens et al., 2005; Zachos et al., 2010]. A very similar relation has been documented for the Oligocene and Miocene [Zachos et al., 2001; Pälike et al., 2004, 2006a, 2006b; Holbourn et al., 2007] as well as the Paleocene and late Cretaceous [Herbert, 1997; Zachos et al., 2010; Westerhold et al., 2011]. At ODP Site 1258 and 1262 the sedimentary cyclicity is dominated by precession modulated by the short and long eccentricity cycles. The astronomical phase relationship is well documented [Lourens et al., 2005] suggesting that the short and long eccentricity cycle are expressed in amplitude variations in the data. This way, large amplitude variations in Fe intensity data reflect large amplitude variations in orbital eccentricity and vice versa. Hence, minima in the very long eccentricity cycle will be expressed by very low amplitude modulations in the Fe intensity data and can be used as a distinctive fingerprint [Lourens et al., 2005; Westerhold et al., 2007].
 An astronomically tuned framework beyond 50 Ma can only be based on the identification of the stable long eccentricity cycle (405-kyr) [Laskar et al., 2004] and not on the direct tuning to the short eccentricity cycle. Therefore, we establish a modified minimal tuning age model for Site 1262 similar to Site 1258. This minimal tuning assigns a tie point every 405-kyr based on the recognition and position of the long eccentricity cycle as given in the relevant publications [Westerhold et al., 2007, 2008; Westerhold and Röhl, 2009]. We follow this approach to label the long eccentricity cycles in each epoch from the base to the top. The minimal tuning tie points are given in Table S3. We also extracted the amplitude modulation (AM) of the short eccentricity signal (95-kyr) of the orbital solutions and Fe intensity data of Sites 1258 and 1262 using the freeware ENVELOPE [Schulz et al., 1999].
 We start the comparison with plotting the data and the extracted 95-kyr amplitude modulation of the data on the stable cyclostratigraphic framework using the modified option 1 age model [Westerhold et al., 2007], with the onset of the PETM at 55.53 Ma. Subsequently we shift the records by 405-kyr steps to age model option 2 and option 3 moving the onset of the PETM to 55.93 Ma and 56.33 Ma, respectively (Figure 4). The best match of the 95-kyr AM between the orbital solutions and the Fe intensity data is obtained (Figure 5) choosing tuning option 1 for Sites 1258 and 1262 [Westerhold et al., 2007, 2008; Westerhold and Röhl, 2009]. Especially the eccentricity cycle minima in the La2010a, La2010d and La2011 solutions are consistent with the geological data applying tuning option 1 (Figure 5). Linear correlation coefficients between orbital solutions and geological data also reveal that age model option 2 and option 3 show a significant mismatch to the solutions between 47 and 54 Ma (Figures 4 and 5 and Table 1). The comparison to Site 1262 does not show a clear picture. This empathizes that none of the orbital models are correct beyond 52 to 54 Ma. To evaluate the match between Site 1262 data and orbital solutions for the likely stable part of the orbital solutions we calculated the correlation coefficient from 52 to 55 Ma (Table 1). The evaluation reveals the best match is obtained between Site 1262 and all three orbital solutions using tuning option 1.
Table 1. Linear Correlation Coefficient of the 95-kyr Amplitude Modulation Between Orbital Solutions and Fe Intensity Data of ODP Sites 1258 and 1262
Comparison of Entire Dataset Dominated by Precession Cycles
Comparison of 1262 Dataset From 52 to 55 Ma
 Very long eccentricity minima in the Fe data are present at around 48.0 Ma, 49.8 Ma, 51.9 Ma and 54.1 Ma, which are consistent with the La2010d and La2011 solutions. These findings suggest that geologic data confirm the stability of the La2011 solution as far back as 54 Ma. The very long eccentricity minima at 51.9 Ma and 54.1 Ma are close to the onset of instability in the La2011 solution. But geological data indicate that the minima are indeed at the correct position. Any match beyond 54 Ma cannot be evaluated because of the high uncertainty of the eccentricity solutions older than 54 Ma. Moreover, the detailed comparison shows that geological data are better represented by La2010d and La2011 than by La2010a (Figure 5). This is in good agreement with the otherwise understanding that the long-term behavior of La2010d is more reliable than the one of La2010a,b,c suggesting that only La2010d and La2011 should be used beyond 47 Ma.
 Because of the growing uncertainty in the expression and shape of short eccentricity cycles in the very long eccentricity minima after 50 Ma we refuse to compare or even try to tune geological data to the short eccentricity cycles of the La2010 and La2011 solution. Due to this uncertainty a tuning of the early Eocene record of Site 1262 to the short eccentricity pattern is not possible. Unfortunately, for the same reason a comparison cannot resolve if there are 19 or 20 short eccentricity cycles between the onset of the PETM and the Elmo event [Lourens et al., 2005; Westerhold et al., 2007]. To do so an orbital solution is required that has a very small uncertainty in the short eccentricity cycle amplitude pattern in the very long eccentricity node at 54 Ma.
4.2. Paleocene and Early Eocene Astronomical Time Scale to La2010 and La2011
 Anchoring the geological data by minimal tuning age model option 1 results in the finding that an age estimate of ∼56.0 Ma for the PETM [Kuiper et al., 2008; Hilgen et al., 2010] has to be rejected. The first-order age calibration to the very long eccentricity cycle minima in the La2011 solution suggests that tuning option 1 is correct for Sites 1262 [Westerhold et al., 2007, 2008] and 1258 [Westerhold and Röhl, 2009]. Calibration to the very long eccentricity nodes proposes an absolute age of 55.53 ± 0.05 Ma for the onset of the PETM (Figure 6 and Table 2). Additionally, we obtain an astronomically tuned age of 54.85 Ma ± 0.05 Ma for ash −17, a well-dated prominent ash layer in the Fur Formation of Denmark [Storey et al., 2007]. This ash layer has been tied to ODP Site 1262 by a detailed cyclostratigraphy procedure [Westerhold et al., 2009]. Independent 40Ar/39Ar dating of ash −17 [Storey et al., 2007] resulted in an age of 55.473 ± 0.120 Ma, calculated to an FC standard age of 28.201 Ma [Kuiper et al., 2008] (Table 2). The comparison of radioisotopic to astronomical ages reveals a substantial discrepancy of 473 to 773 kyr.
5. Consequences for the Fish Canyon Radioisotopic Dating Standard
 Radioisotopic 40Ar/39Ar dating depends on the accuracy of the FC standard monitor age, the decay constant, and the precision of the measurement [Renne et al., 1998]. Using FC ages of 28.02 Ma [Renne et al., 1998], 28.201 Ma [Kuiper et al., 2008] or 28.305 Ma [Renne et al., 2010] results in significantly older ages for ash −17 and the PETM compared to the astronomically calibrated age (Figure 7 and Table 2). Although detailed biostratigraphy and cyclostratigraphy patterns indicate that the cyclostratigraphic position of ash −17 is correct [Westerhold et al., 2009], a lower number of cycles between ash −17 and the PETM has been proposed [Hilgen et al., 2010]. If the position of ash −17 with respect to the onset of the PETM is displaced by an order of 120–250 kyr [Hilgen et al., 2010] (55.02–55.15 Ma), then the age of the PETM is still within the estimate using 28.02 Ma [Renne et al., 1998] for the FC, but clearly younger than the estimates using 28.201 [Kuiper et al., 2008] or 28.305 [Renne et al., 2010] Ma (Table 2). Assuming that our astronomical age of ash −17 is correct, we recalculate an age of 27.89 Ma for the FC. This is within the error range of 28.02 ± 0.28 Ma [Renne et al., 1998] and surprisingly close to the recent independent estimate of 27.93 Ma based on detailed and revised astronomical dating of the Pleistocene Matuyama-Brunhes boundary [Channell et al., 2010].
 The astronomically tuned absolute age of the onset of the PETM is independent of the relative distance to ash −17 and its radioisotopic age. Recently, a high-precision U/Pb single-crystal zircon age for an ash layer within the PETM from Spitsbergen provided an estimate of 55.829 ± 0.10 Ma [Charles et al., 2011]. This is ∼300 kyr older than the astronomically calibrated age and supports the findings that U/Pb dates are systematically older than 40Ar/39Ar dates [Min et al., 2000; Schoene and Bowring, 2006; Schoene et al., 2006]. It should be noted that the U/Pb age of 55.785 ± 0.086 Ma for the lower bentonite in the Spitsbergen PETM is the weighted mean of the five youngest analyses. If the youngest single zircon age of 55.71 ± 0.14 Ma [Charles et al., 2011] (55.57–55.85) approximates the eruption age, then the onset of the PETM is 55.61–55.89 Ma. This is close to but still 30–360 kyr older than the astronomical age, and might be related to residence time issues of zircon crystals [Min et al., 2000; Simon et al., 2008].
 Recalculations of various radioisotopic 40Ar/39Ar ages for the K/Pg boundary, applying the revised estimate of 27.89 Ma for the FC, provide ages between 65.08 and 65.26 Ma (Table 2), which are consistent with the 65.28 Ma estimate from tuning option 1 of the Paleocene time scale [Westerhold et al., 2008]. The duration of the Paleocene Epoch based on the astronomically calibrated age for the PETM and the recalculated radioisotopic ages for the K/Pg boundary is between 9.55 and 9.75 myr, suggesting that the entire Paleocene contains 24 [Westerhold et al., 2008] 405-kyr eccentricity cycles, not 25 [Hilgen et al., 2010]. Retuning of the Paleocene time scale to the La2011 405-kyr cycle results in an age of 65.25 ± 0.06 Ma for the K/Pg boundary.
 In this context, our results are highly controversial to those obtained from the Mediterranean sections. Absolutely crucial, however, for the approach of Kuiper et al.  and Rivera et al. is an absolutely perfect analysis of cyclicity expressed in the road-cut outcrops at the Sorbas and the Faneromeni section. Land sections are often more difficult to interpret than the multiple hole and transect approach of ODP sediment cores. It is essential to note that a small error in the tuning of 20 kyr around 6.5–7.0 Ma will increase by an order of magnitude to a mismatch of more than 200 kyr at 50 to 60 Ma. With respect to our new result for the absolute age of ash −17, the PETM and the K/Pg boundary this possibly indicates a likely tuning error for the Faneromeni and Messâdit section of 40–60 kyr or 2–3 precession cycles. This error is much larger than any error introduced by possible local phase relation changes to precession and/or obliquity [Laepple and Lohmann, 2009]. Tuning the sections 60 kyr too old will lead to 0.8% older 40Ar/39Ar ages than expected if using the 28.02 Ma (or 27.9 Ma argued for here) age for the Fish Canyon sanidine.
6.1. Tuning of the Sorbas, Faneromeni and Messâdit Sections
 The tuning of the Sorbas and Faneromeni sections is thought to be robust because the observed sapropel pattern exactly matches precession/obliquity interference patterns in the interval from C3An.1n to C3Bn [Krijgsman et al., 1999] and the overall cyclostratigraphic correlation to the long and short eccentricity cycles is well established [Hilgen et al., 1995]. Nevertheless, looking into detail reveals some discrepancies that might indicate that the tuning of these sections is not absolutely correct down to the individual precession cycle as required for refining the FC.
 The tuning of the Sorbas section [Krijgsman et al., 1999] was based on the age of magnetochron C3An.1n of the previously established ‘tuning’ of the early Messinian [Hilgen et al., 1995]. But this age was not obtained by direct orbital tuning due to the absence of tunable sections in the Mediterranean from 6.7–5.3 Ma [Hilgen, 1991]. Instead it was established by linear interpolation of seafloor spreading rates between the top of C3n.4n and the top of C3An.2n [Hilgen et al., 1995] (see Table 3). The age of the top of C3n.4n is well established and consistent with seafloor spreading rate changes [Wilson, 1993; Cande and Kent, 1995; Lourens et al., 1996; Lourens et al., 2004]. The age of 6.677 Ma for the top of C3An.2n in Hilgen et al.  is based on tuning of the Faneromeni section and correlation between the Kastelli (Crete), Metochia (Gavdos), and Giblescemi (Sicily) sections. Tuned ages for magnetochrons older than C4n.1n (7.528 Ma) have been confirmed [Hüsing et al., 2007, 2009], but ages for C4An.1n to C3An.1n are uncertain because sedimentary succession cannot clearly be related to orbital cycles in the Faneromeni, Kastelli and Metochia sections [Hilgen et al., 1995]. Therefore, the age of the top of C3An.1n used for the initial tuning of the Sorbas section could be questioned. In fact, the age of C3An.1n (o) in Hilgen et al.  is 58 kyr older than CK95 [Cande and Kent, 1992; 1995] and leads to 139 kyr older age after tuning of the Sorbas section [Krijgsman et al., 1999] (Table 3). The combination of the tuned magnetostratigraphy of the Sorbas and the Abad composite (Sorbas and Nijar basin) [Sierro et al., 2001] used in the ATNTS/GPTS2004 [Lourens et al., 2004] even show a difference of up to 205 kyr with respect to CK95 at C3Bn (y), a discrepancy of almost 3%. Differences in the absolute ages for magnetochron C3r to C4n.2n (Table 3) can lead to some significant changes in seafloor spreading rates (Figure 8). We have calculated apparent changes in seafloor spreading rates using the distances of the South Atlantic spreading center [Cande and Kent, 1992] as done for the GPTS 2004 [Ogg and Smith, 2004]. The comparison shows minor to major differences to CK95 during chrons C3r to C4n.2n. Moderate changes in spreading rate are consistent with the ATNTS in the Pliocene [Wilson, 1993], but unknown thereafter. By far the highest peak in South Atlantic seafloor spreading rates during the entire Cenozoic occurs at chron C3Bn using the ATNTS/GPTS2004 [Lourens et al., 2004; Ogg and Smith, 2004], around 7 Ma. This might point to potential errors in the astronomically calibrated time scale for C3An.1n to C4n.1n derived from the Sorbas basin sections. It is difficult to identify directly where the error in tuning might be because the correlation between the Sorbas, the Melillia basin and Cretian sections is complex. Hodell et al.  established a correlation between the Mediterranean, ODP Site 982 (North Atlantic) and the Salé Briqueterie section (NW Marocco [Hodell et al., 1994]), but rely for an initial age model on the astronomical ages for foraminiferal events derived from Hilgen et al. . Thus, an independent astronomically tuned interval spanning C3r to C4n.1n from outside the Mediterranean with high quality magnetostratigraphy would be needed for final clarification.
Table 3. Differences in Geomagnetic Polarity Time Scale Calibrations
 Geological data from deep-sea sediments confirm the exact positions of very long eccentricity minima in the new La2011 orbital solution from 47 to 54 Ma. For the first time, this allows direct correlation between geological data and a stable astronomical solution in the early Eocene independent from radioisotopic dating. First-order calibration of the geological data to the very long eccentricity minima in La2011 suggests that the astronomically calibrated absolute ages for ash −17, the PETM and the K/Pg boundary are in conflict with radioisotopically dated40Ar/39Ar ages based on recently recalibrated Fish Canyon (FC) standard ages [Kuiper et al., 2008; Renne et al., 2010; Rivera et al., 2011]. In contrast to a recent confirmation [Rivera et al., 2011], but in line with the study of Channell et al. , our results point to a 0.8% younger age for the FC of ∼27.9 Ma than the Kuiper et al.  estimate.
 Although the analytical precision of radioisotopic dating has improved tremendously, the discrepancy between the astronomical ages presented here and radioisotopic ages is profound. Our results point to intrinsic problems with the decay constants and/or standard ages used in 40Ar/39Ar dating, and possible biasing of the U-Pb clock in zircon crystals due to the memory of pre-eruption crystallization of magma that forms large volcanic ashfall deposits, or both. If so, the synchronization of40Ar/39Ar and U/Pb radioisotopic clocks as proposed by Kuiper et al.  is problematic.
 Our results indicate that the tuning of the late Miocene Mediterranean sections (Sorbas, Faneromeni, Messâdit, etc.), which contain the ash layers used for the intercalibration of the FC [Kuiper et al., 2008; Rivera et al., 2011], could be in error by 40–60 kyr. Independent astronomically tuned sections from outside the Mediterranean with high-quality magnetostratigraphy are required to solve the time scale controversy.
 We thank Monika Segl and her team for stable isotope analyses. We are indebted to H. Pfletschinger and V. Lukies (MARUM) for assisting in XRF core scanning. This research used samples and data provided by the Integrated Ocean Drilling Program (IODP). IODP is sponsored by the U.S. National Science Foundation (NSF) and participating countries under the management of Joint Oceanographic Institutions (JOI), Inc. Financial support for this research was provided by the Deutsche Forschungsgemeinschaft (DFG), ANR-ASTCM, PNP-CNRS, and CS from Paris Observatory. We thank three anonymous reviewers who provided thoughtful and thorough reviews improving the paper. The data reported in this paper are tabulated in theauxiliary material and archived at the Pangaea database (doi.pangaea.de/10.1594/PANGAEA.783354).