Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
ELSEVIER Earth and Planetary Science Letters 173 (1999) 7–23 www.elsevier.com/locate/epsl Depth versus age: new perspectives from the chemical compositions of ancient crust Eric Humler a,Ł , Charles Langmuir b , Valérie Daux c a Laboratoire de Pétrologie, Université Paris VII Denis Diderot, Case 110, 4 Place Jussieu, 75252 Paris Cedex 05, France b Lamont Doherty Geological Observatory, Palisades, NY, USA c Laboratoire de Géologie des Bassins Sédimentaires, Université P. and M. Curie, Case 116, 4 Place Jussieu, 75252 Paris Cedex 05, France Received 27 May 1999; accepted 6 September 1999 Abstract Petrological data provide a new approach to an evaluation of the depth–age problem for ancient seafloor. The correlations among basalt chemical composition, axial depth and mantle temperature at current ocean ridges allow the determination of initial depth and mantle temperature for any portion of ancient seafloor that was created at a spreading center, provided the chemical composition of the ancient crust is determined. It is then possible to calculate a petrologically constrained depth at any age, which can be compared to observed depths and depths from the classical half space models. We evaluate data from DSDP and ODP drill holes on crust older than 80 Ma, considering chemical composition, back-tracked depth and crustal thickness. The data are complex, and interpretation of their chemical composition requires consideration of alteration, absence of glass compositions, data quality, and the influence of off-axis volcanism and near-ridge hot spots. To check and expand the data set, we develop and use trace element proxies for major element compositions, since many trace element ratios are less influenced by alteration and by variable proportions of phenocrysts. The twenty drill holes for which reliable data can be obtained are well distributed around the globe, and include multiple sites on old crust in the Atlantic, Pacific and Indian ocean basins. Comparison of the chemical and crustal distributions between ancient and current N-MORB show that the oceanic crust older than 80 Ma has significantly lower Na8.0 , Zr=Y, Sm=YbN , and higher CaO=Al2 O3 , Fe8.0 and crustal thickness. Quantitative modeling of these results suggests that the mantle was hotter in this time period by about 50ºC, that the crust was several hundred meters shallower and 1–2 km thicker. These observations show that half to two thirds of the observed flattening relative to a half space model is due to the change in mantle temperature and crustal composition. Thus, only a few hundred meters of flattening by plate reheating by hot spots or by other mechanisms is required. These results are consistent with the existence of abundant oceanic plateaus even at fast-spreading rates in the Mesozoic, and with the apparent thickening of ocean crust with time. 1999 Elsevier Science B.V. All rights reserved. Keywords: oceanic crust; thermal history; lithosphere; vertical movements; Ocean Drilling Program; Deep Sea Drilling Project Ł Corresponding author. E-mail: [email protected] 0012-821X/99/$ – see front matter 1999 Elsevier Science B.V. All rights reserved. PII: S 0 0 1 2 - 8 2 1 X ( 9 9 ) 0 0 2 1 8 - 6 8 E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 1. Introduction Since the early developments of plate tectonics, it has been well understood that the depths of the ocean basins relate to the thickness of the basaltic crust and the temperature distribution within the underlying upper mantle [1,2]. The classic approach towards quantifying these relationships has been to assume that there is a ‘normal’ ocean crust with a thickness of 6–7 km, and then to model the two-dimensional problem of plate cooling with age. The first-order relationships between depth and the square root of age, as recently summarized by Smith and Sandwell [3] (Fig. 1), can be understood at young ages as a direct response to the cooling and thickening of the lithosphere, and are modeled successfully by a cooling half space [4,5]. However, crust older than 80 Ma is shallower than would be predicted from the half space model, and these shallow depths have been viewed classically from the perspective of perturbations to the cooling history of normal crust. Proposed perturbations include lithospheric re-heating by hot spots (e.g. [3,6]), secondary small-scale convection in the upper mantle [7], modification of the subsidence behavior by radiogenic heat generation and phase changes in the cooling lithosphere [8–10], and counterbalance of the effects of thermal contraction Fig. 1. Mean variation of the depth of the seafloor vs. its age deduced from satellite altimetry and ship depth soundings (reproduced from [3]). The dashed line shows the theoretical evolution of seafloor depth as a function of age according to the boundary layer model [4,5]. by plate scale flow [11,12]. The tacit assumption behind most of this work has been that zero-age depth for normal crust has been at steady state for the last 150 Ma and that by looking at ancient crust it is possible to observe what has happened to this normal crust through time. It has been recognized, of course, that the depth– age problem is complex in detail. Hayes [13] showed regional variations in depth vs. age, and Marty and Cazenave [14] and Calcagno and Cazenave [15] have noted that subsidence seemed to correlate with the original axial depth of the ridge, and be affected by the regional temperature of the underlying mantle. Kane and Hayes [16–18] noted large variations in subsidence rate and correlations between subsidence rate and inferred initial depth that were not consistent with the various geophysical models. Therefore as a broader perspective has emerged from consideration of the new global data sets, it has become clear that there remains much to be understood in detail from depth–age studies. A second approach to depth variations of the ocean floor has been to consider the orthogonal problem to depth vs. age: variations along the ridge axis, at ‘zero age’. Maps of the global ridge system show a substantial depth distribution, from above sea level at Iceland, to as much as 5000 m at some of the deepest ridges, such as in the Arctic, at the Australian Antarctic Discordance, and along portions of the Southwest Indian Ridge. These depth variations correlate with the chemical compositions of the basalts of the ocean crust [19,20] (fig. A in EPSL Online Background Dataset 1 ), and the thickness of the crust [21]. These relationships, the ‘global correlations’, can be modeled successfully as a response to variations in the potential temperature of the mantle [21–24], with some additional lithospheric cooling taking place at super-slow spreading ridges [22,25,26]. Keen et al. [27] investigated old crust compositions in the Atlantic in order to test Klein and Langmuir’s model [21]. They found that the chemical composition of ancient crust, back-tracked along a half space model to zero age, agreed with the depth–chemistry correlations of present-day ridges. 1 http://www.elsevier.nl/locate/epsl, mirror site: http://www.elsevier.com/locate/epsl E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 This work demonstrated that the chemical composition of ancient crust can be used to determine the likely depth of the ancient ridge axis. Keen et al., however, used only sites from the Atlantic, many of them influenced by hot spots, and did not explore the depth–age problem, which relates inherently to ‘normal crust’. Klein and Langmuir [21] noted that the chemical composition of ancient crust provided a tool with which to investigate depth variations with time. The new perspective that is possible with a petrological approach derives from the fact that the chemical composition of the crust gives an independent estimate of its initial depth. The composition also provides information about mantle temperature, although there is some controversy concerning the relative importance of mantle temperature and mantle composition [28]. Both original depth and mantle temperature are central to the evaluation of the depth–age problem, and neither of them has been able to be investigated from the classical geophysical approach. Petrology provides independent constraints on initial depth and mantle temperature, and thus allows us to address the possibility that the current shallow depths of ancient crust may result in part from shallower initial depths of ridges during the Mesozoic. While this would be an inherently speculative suggestion based solely on depth and age, it is a testable hypothesis using a petrological approach. In order to investigate a possible change in crustal composition through time and its implications for depth–age, we evaluate here the chemical composition, back-tracked depth and crustal thickness of DSDP and ODP drill holes on crust formed more than 80 Ma ago. Through this comparison, we can test the steady state assumption of most geophysical depth–age modeling, and determine whether there is evidence for changes through time in the distribution of ridge depths and mantle temperature beneath ridges. It is important to note that the appropriate comparison between the ancient sites and current ridges is the isochronous distributions of depth and chemistry. There is not a single depth of ocean ridges today, nor is there a single chemical composition (see fig. A in EPSL Online Background Dataset, see footnote 1). Therefore the critical question is not 9 whether one can find in the past some example that is an analogue for some ocean ridge today. Instead, the issue is whether there is a statistically significant difference in the distributions of depth and chemistry between the global system of ridges today and that in the Mesozoic. 2. Site and data selection The twenty years of ocean drilling by DSDP and ODP have provided a significant number of basement samples from older ocean crust. However, many of these samples may not be directly comparable to the zero-age data set of dredged glasses from active ridges, and hence cannot be used for a direct comparison with the data set from zero-age MORB. The sections below discuss the various complexities in dealing with data from drilled basalts, the screens we have applied to sites and to chemical data, and the steps we have taken to ensure the fidelity of the chemical signal. The drill sites considered along with documentation are more fully presented in the appendix available through EPSL Online Background Dataset, see footnote 1). 2.1. Geological setting Not every drill site on ancient crust that penetrates basement recovers material that was formed in place at an ocean ridge. Many sites are located on bathymetric highs that were created off-axis. Others penetrate basalts inferred to be off-axis sills on the basis of chilled margins and baked sediment contacts (e.g. sites 260, 261). Where these intrusions have been dated, they often have ages substantially younger than the spreading age inferred from magnetic anomalies. Some sites near continents sometimes drill transitional basement (e.g. sites 551 and 766A, where seismic velocity is intermediate between oceanic and continental crust). Some recovered materials, such as dolerites and pebbles (e.g. sites 170 and 316) also apparently do not represent materials recovered in place. These various sites are not applicable to the problem we wish to address. Other sites recover basement that was influenced by on-ridge or near-ridge hot spots. While these 10 E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 sites may be important for the overall distribution of depths for ancient ocean ridges, their chemistry is often anomalous and not subject to the clarity of interpretation that is possible for ocean ridges more distant from hot spots. Furthermore, it is obvious that such sites have shallow depths because of differences in initial conditions, and such sites have been intentionally excluded from geophysical discussions of the depth–age problem. We also have excluded samples that appear on a chemical basis to come from ridges influenced by hot spots, using the criterion of including samples only for which (La=Sm)N <1 or K2 O=TiO2 <0.1 if REE are not available (e.g. sites 170, 250A, 265, 258). These samples have been excluded both because they generally indicate a different tectonic setting than normal MORB, and also because such samples have different chemical systematics than normal MORB, as discussed further below. In summary, we have excluded sites that recovered off-axis materials, and sites that are on shallow bathymetric anomalies created either at the ridge or off-axis. We have also not used data from samples that appear to be exotic or not to have been created by zero age volcanism. 2.2. Alteration All old crust is altered, and most drilled material from ancient crust differs from the pristine magmatic glass that can be obtained at zero age. Despite the pervasive alteration, some old holes have small amounts of glass for which reliable major element compositions have been determined (e.g. holes 105, 417D, 418A–B, and 765D). For some drill sites, there are glass analyses, fresh rock analyses and altered rock analyses from the same hole. Comparison of these various analyses shows that even in old crust many whole rock analyses provide major element compositions that faithfully record the chemical signature of the basalts. Samples that are heavily altered, however, do not carry a reliable chemical signature. Altered samples generally have high water contents due to the presence of newly formed hydrous minerals formed during alteration, which is reflected in the ‘loss on ignition’ during analysis. Only samples with less than 2% loss on ignition have been considered. 2.3. Fractionation correction In order to remove the effects of fractionation, we have normalized the data to 8% MgO following the procedure of [21]. There have been some attempts to modify this correction procedure e.g. [29], and there are significant issues, particularly when the data are far removed from 8% MgO, or when one applies the same correction factor over a large compositional range. In the current case, our main aim is a comparison of drill hole data with the pre-existing MORB compilations [21,22]. To make this comparison accurately, the same correction factors need to be used. The results are not particularly sensitive to fractionation correction, because the mean MgO content of our data base for drill sites (7.3% MgO for all individual samples) is the same as the mean for the East Pacific Rise (EPR) [22], and the EPR is used for many of the comparisons below. In addition there is no relationship between the corrected chemical parameters and the MgO content of the data, as illustrated in fig. B in EPSL Online Background Dataset, see footnote 1). Because of the absence of glasses from many drill sites, it is also necessary to use whole rock data. The presence of phenocrysts in the whole rocks modifies their chemical compositions, which leads to uncertainty in the interpretation of the major elements. For example, 15% plagioclase in the whole rock can change Na8.0 from 2.6 to 2.2, and Fe8.0 from 9.5 to 6.4, while increasing Al2 O3 from 15.5 to 18.3 wt.%. Major element data for whole rock samples with high Al and low Fe and Ti contents have been excluded. 2.4. Data quality The chemical compositions of most zero-age MORB are analyses of glasses by electron microprobe in well known laboratories for which interlaboratory correction factors are often available. Where possible, we have applied interlaboratory corrections to the major element data [22]. Many of the data from drill sites, however, were collected by the initial investigators who were at sea over 20 years ago, using techniques that are of inadequate precision (e.g. emission spectrograph, ‘rapid method’ major elements; e.g. holes 303–313 and 581), or from labora- E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 tories for which there is no record of interlaboratory calibration and quality. In addition, much of the data is on whole rocks with no petrographic descriptions and sometimes no indication of the loss on ignition. In some cases, these older chemical analyses simply do not make sense in terms of what we know now about chemical systematics. A small number of modern analyses that pass the screen for loss on ignition also have chemical compositions that are not consistent with igneous systematics (e.g. 5% Na2 O and low TiO2 ). Therefore we have used our judgment in some cases to exclude data where there are a small number of analyses that are questionable. These data are noted in the appendix available through EPSL Online Background Dataset, see footnote 1). Sites for which there are modern data of known reliability have been included, along with several sites for which there are only older data. These various criteria lead to our selection of sites and data for investigation of the depth–age problem. Table 1 lists the twenty sites with reliable chemical data from appropriate geological settings. Fig. C in EPSL Online Background Dataset (see footnote 1) presents a map of the distribution of sites we have been able to consider. The sites are well distributed around the globe, and include multiple sites on old crust in the Atlantic, Pacific and Indian ocean basins. 11 1). Therefore where data for Zr=Y or Sm=Yb exist, we can use these ratios to check major element data, and to calculate Na8.0 values from the trace element proxies, based on the correlations. Because the correlation is tighter, we use the (Sm=Yb)N ratio to calculate the Na8.0 proxy, which is labelled as Na8.0 Ł . The equation is: ln Na8:0 Ł D 0:976.Sm=Yb/N 0:133 (1) Ł Na8.0 has an absolute uncertainty from the width of the Sm=Yb correlation of š0.15%. The trace element proxies are particularly important because they are insensitive to alteration and to fractionation correction. Zr, Y and the rare earth elements such as Sm and Yb have been shown by many studies to be unchanged by most alteration processes, e.g. [30–32], and ratios among these elements are particularly robust. These elements are also relatively incompatible, and hence are not contained in significant amounts in mineral phases pertinent to MORB. This is illustrated in fig. E in EPSL Online Background Dataset (see footnote 1), which shows the invariance of the Zr=Y and Sm=Yb ratios across a range of MgO contents for suites of samples. Because of these characteristics, these ratios are not sensitive to fractionation correction, nor to mineral accumulation in whole rock samples. 2.5. Trace element proxies for the major elements The major element data are important because they can be linked quantitatively to mantle temperature, axial depth and crustal thickness. Because of problems with alteration and whole rock analyses for the major elements, it is useful to find other parameters that can be used as proxies for the major elements. This permits a larger number of sites to be included, and provides an independent test of the validity of the major element results. The use of trace element proxies requires a clear definition of the relationships between the trace elements and the major elements. Several investigators [21,25] have shown that the (Sm=Yb)N ratios of MORB correlate with the Na8.0 contents for basalts that are not enriched in incompatible elements. In addition, available data show that the Zr=Y ratios of MORB correlate with the (Sm=Yb)N ratios (fig. D in EPSL Online Background Dataset, see footnote 3. Are drilled basement samples directly comparable to dredged samples from ocean ridges? Dredged samples from ocean ridges are mostly from active volcanoes still in the midst of their eruptive phase, while drill sites recover the top few meters of basement that represent the final products of the entire crust formation process. The last lavas erupted could be of rather different composition than the mean lava from the active ridge. Differences between mean on-axis eruptions and final off-axis eruptions could also vary as a function of spreading rate. Therefore it is necessary to evaluate whether drilled sample compositions are comparable to zeroage dredges. To address this question, we consider two regions, one from the slow-spreading Mid-Atlantic Ridge (MAR) and the other from the fast-spreading East 12 Table 1 DSDP=ODP sites older than 80 Ma with reliable chemical data (references are given in appendix on EPSL Online Background Dataset) Age Water depth Normal sites 17=164 17=166 20=197 26=257 32=304 32=303 32=307 51-53=417D-418A=B 80=550B 86=581 91=595 129=801C 110 132 151 135 133 130 148 118 97 116 155 167 5485 4950 6143 5278 5640 5609 5696 5502 4420 5467 5596 5685 Hot spot margins 11=100 11=105 27=259 27=261 123=765D 158 156 110 152 139 Ontong–Java plateau 17=169 149 61-89=462-462A 148 129=802A 115 Na8.0 Fe8.0 CaO=Al2 O3 Na8.0 Ł Zr=Y N Ce=SmN N Sm=YbN N 190 2.41 2.43 2.51 2.24 2.23 2.26 2.35 2.37 2.96C-0.28 2.82 š 0.06 3.07 š 0.15 2.38 2.58 š 0.06 2.79 š 0.08 2.36 š 0.00 2.50 š 0.01 2.69 š 0.09 2 8 3 11 13 6 6 112 12 0.855 0.699 š 0.010 0.684 š 0.005 0.655 š 0.018 0.653 š 0.007 0.668 š 0.011 0.591 š 0.031 0.547 š 0.024 1 4 2 2 6 3 6 93 1.038 1.048 š 0.030 1.080 š 0.031 0.961 š 0.045 0.958 š 0.016 0.970 š 0.069 1.010 š 0.046 1.019 š 0.024 1 4 2 7 6 3 3 94 1 30 2.17 2.54 2.44 2.66 š 0.10 2.80 š 0.15 15 30 0.691 š 0.080 0.766 š 0.035 0.686 š 0.046 4 3 9 0.931 š 0.030 1.092 š 0.058 1.051 š 0.026 4 3 9 5 15 1 2 10 0.821 š 0.070 0.850 š 0.052 0.774 0.872 š 0.011 0.868 š 0.019 5 15 1 2 10 4 31 18 1.089 š 0.030 1.034 š 0.047 1.046 š 0.043 4 30 16 N 2.56 š 0.06 9.82 š 0.54 0.77 š 0.02 5 2.30 š 0.20 10.29 š 0.41 0.83 š 0.02 6 2.27 š 0.02 10.64 š 0.04 0.823 š 0.003 2.38 2.33 š 0.08 10.35 9.83 š 0.24 0.81 0.84 š 0.02 5325 5245 4712 5667 5724 2.27 š 0.10 2.11 2.13 1.98 š 0.16 10.14 š 0.09 8.95 9.18 9.26 š 0.46 0.785 0.775 0.819 š 0.013 5415 5179 5969 2.00 š 0.19 2.03 š 0.05 2.06 š 0.07 9.70 š 0.22 10.72 š 0.12 10.90 š 0.28 0.89 š 0.01 0.824 š 0.004 0.82 š 0.01 1.79 š 0.61 2.33 š 0.29 2 16 1 22 1.95 2.01 1.86 2.05 2.04 2.38 š 0.06 2.32 š 0.04 2 44 0.785 š 0.178 0.766 š 0.070 0.790 0.646 š 0.052 0.611 š 0.044 8 85 33 2.53 2.40 2.43 2.27 š 0.09 2.34 š 0.03 2.62 š 0.03 8 127 33 0.780 š 0.020 0.862 š 0.043 0.936 š 0.050 9 Age (Ma), water depth (m), Na8.0 , Fe8.0 (weight%) and CaO=Al2 O3 with LOI <2%; Na8.0 Ł are calculated from Sm=YbN using Na8.0 –Sm=YbN correlation for N-MORB (Eq. 1). Ce=SmN and Sm=YbN are chondrite-normalized ratios. N D number of data. E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 Leg=Hole E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 Pacific Rise (EPR). The MAR south of the Kane fracture zone, known as the MARK area, is one of the best-studied portions of the ridge system, with abundant samples collected both along the neo-volcanic ridge, and from the floors of the rift valley and the rift valley walls (e.g. [33,34]). These data can be compared with the extensive drilling that has taken place both east and west of the ridge axis in this region, at sites 395 and 396, where crustal ages are about 10 Ma [35–37]. Zero-age and older basalts have similar compositions, and there is also no significant change in the chemistry of the basement with depth of drilling penetration. Similar results are obtained for the EPR (table A and fig. F on EPSL Online Background Dataset, see footnote 1, data from [22,38–40]). Therefore at both slow and fast spreading drill sites often provide chemical compositions comparable to dredges at zero age. Of course, off-axis eruptions or small seamounts may create discrepancies in some cases, particularly when basement penetration is minimal. 4. Comparison of the distributions of chemical composition for ancient and current MORB A first-order comparison between ancient and current ridges can be made by comparing the chemical compositions on the Na8.0 –Fe8.0 diagram. Fig. 2 shows that most of the old drill site data occupy the depleted half of the field for current MORB. Some of the data plot outside the normal MORB field; these will be discussed further below. From this figure it is clear that the ancient data have a different distribution and mean than the recent data. One can ask whether there might be some systematic bias that could account for these different distributions. For example, much of the current database for MORB comes from slow-spreading ridges, where essentially all of the high Na8.0 , high (Sm=Yb)N samples have been recovered. Since most of the drill site data are from fast-spreading crust, that could lead to an inappropriate comparison. But the mean composition from the East Pacific Rise is very similar to the mean composition from slow-spreading ridges (fig. G on EPSL Online Background Dataset, see footnote 1) and Table 2). The mean EPR composition, indicated by the large gray dot in the figures, lies 13 Fig. 2. Na8.0 vs. Fe8.0 for basalts far from hot spots (light gray field), at hot spot margins (open squares) and ODP=DSDP basalts older than 80 Ma (triangles). Most of the drill sites data occupy the depleted half of the field for current MORBs. All the ODP Site samples have La=SmN <1. The different triangular symbols refer to different ODP=DSDP basalt types as discussed further below (see also Figs. 4 and 5). The large open circle represents the mean for zero-age MORB. outside the data field for the ancient crust. Therefore the contrast between ancient and recent ocean crust is independent of spreading rate. Rigorous statistical comparison is complicated by several factors. Regional averages of surface samples from current ocean ridges are a different sample of the seafloor than a drill hole through old crust. The drill hole provides multiple flow units from a single geographical location and provides a short-term temporal average at a single spot. The regional averages from current ridges provide multiple locations at a single time. If one treats these averages as statistically equivalent, then one can make histograms and carry out statistical tests. Histograms of the distributions of the major element parameters for zero-age and >80 Ma old basalts are shown in Fig. 3 (and fig. G on EPSL Online Background Dataset, see footnote 1), all for normal MORB with (La=Sm)N less than 1 (or K2 O=TiO2 less than 0.1 when REE are not available). For trace elements, a similar statistical test would require a global compilation and evaluation of trace element data from MORB, which is a major undertaking that is beyond the scope of the present study. Instead, we make use of the fact that the mean and standard deviation of individual samples from the EPR are similar to the mean and standard deviation 14 E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 Table 2 Comparison of ODP sites >80 Ma with present ocean ridges Mean σ 2σ 2.20 2.24 9.98 0.82 2.54 0.97 0.18 0.17 0.63 0.03 0.30 0.10 0.10 0.08 0.36 0.02 0.14 0.05 12 20 12 11 18 19 Current MORB (regional averages) Na8.0 2.56 0.40 Fe8.0 9.61 0.96 CaO=Al2 O3 0.76 0.06 0.08 0.19 0.013 100 100 84 EPR basalts (regional averages) Na8.0 2.66 0.15 Fe8.0 9.93 0.37 CaO=Al2 O3 0.80 0.02 0.06 0.14 0.01 28 28 25 EPR basalts (individual samples) Na8.0 2.61 Fe8.0 9.87 CaO=Al2 O3 0.79 Zr=Y 3.15 (Sm=Yb)N 1.11 0.23 0.59 0.04 0.43 0.10 0.03 0.12 0.002 0.11 0.02 295 295 295 62 70 Crustal thickness ODP >80 Ma <15 Ma 0.97 0.77 0.33 0.31 34 24 ODP >80 Ma Na8.0 Na8.0 Ł Fe8.0 CaO=Al2 O3 Zr=Y (Sm=Yb)N 7.43 6.62 N Mean Na8.0 , Fe8.0 , CaO=Al2 O3 , Zr=Y and Sm=YbN for zero-age MORBs and EPR basalts, and ODP=DSDP basalts older than 80 Ma. Mean zero-age values are from [22] and values for ODP=DSDP from Table 1 (references in the appendix on EPSL Online Background Dataset, see footnote 1). 2σ is two standard deviations of the mean. The old crust compositions are clearly distinct from average MORB today and from the EPR. of the global data set of regional averages, as documented in Table 2, and shown in fig. G in EPSL Online Background Dataset (see footnote 1). Since we have a large data set for individual EPR samples, a substantive comparison can be made. Furthermore, most of the drill sites under investigation were created at fast-spreading ridges, so the comparison is apt in this sense as well. Histograms comparing (Sm=Yb)N and Zr=Y ratios in individual EPR samples and the ODP data are shown in Fig. 3. The shifts in the trace element ratios are consistent with one another and with the major elements, indicating that the ODP samples occupy the depleted half of the current MORB data set. The consistency among all param- eters indicates that the results are not dependent on data quality, alteration, whole rock data, nature of the statistical sample, etc. The different distribution of the data is confirmed by all parameters. Table 2 presents the average data for the various sample sets. Compared to current MORB and EPR compositions, the basalts older than 80 Ma have lower Na8.0 , Zr=Y and (Sm=Yb)N and higher CaO=Al2 O3 and possibly higher Fe8.0 (see fig. H in EPSL Online Background Dataset, see footnote 1). 5. Discussion The available data suggest that in comparison to current ridges, the ancient seafloor was made up of a much greater proportion of low Na8.0 , (Sm=Yb)N , and Zr=Y and high CaO=Al2 O3 basalts. Based on the Na8.0 –depth correlation (fig. Aa on EPSL Online Background Dataset, see footnote 1), the data suggest that more of the ridge was shallower, and the underlying mantle hotter, in the period from 80 to 170 Ma. Despite the clarity of the statistical tests, there are uncertainties to this conclusion. First, the number of sites is limited, and the data may not be representative of the mean old crust. Second, while the drill sites come from the Atlantic, Pacific and Indian Oceans, there are large regions of the seafloor that are not well sampled, and therefore some geographical bias is possible. While the question remains as to what extent the low Na basalts were a global phenomenon in that time period, there is no doubt that the sampled sites did reflect very different mean conditions of formation than current ocean ridges. Such an observation has been made on the basis of isotopic compositions as well. Janney and Castillo [41] point out that drilled Mesozoic MORB younger than 125 Ma seem to have been derived from sources that differ in their radiogenic isotope composition. At first glance it would seem straightforward to convert the petrological results into direct implications for depth–age systematics by making use of the Na8.0 –depth correlation for ocean ridges. However, the various chemical characteristics of the ODP samples are quite similar to samples that are currently found near hot spots, and this complicates the interpretation. E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 15 Fig. 3. Comparison of chemical distributions of Na8.0 (a), CaO=Al2 O3 (b) Sm=YbN (c) and Zr=Y (d) between current MORBs (a, b) or EPR (c, d) and ancient basalts. Current MORBs are represented by white bars, EPR by gray bars and drill sites by black bars. Old basalts have lower Na8.0 , Zr=Y and Sm=Yb N and higher CaO=Al2 O3 compared to current MORB and EPR basalts. Each population has been normalized to 100%. Old basalts are distinguishable from global MORB and EPR at better than 1% significance for Na 8.0 , CaO=Al2 O3 , Zr=Y and Sm=YbN . Old ODP=DSDP basalts are slightly Fe8.0 -richer than current MORB but this difference is not statistically significant (see fiig. H on EPSL Online Background Dataset (footnote 1) and Table 2). 5.1. Distinguishing samples from hot spot margins Ocean ridges that are influenced by hot spots often show contrasting systematics to the global correlations for normal ridges [21,22]. This is apparent in figs. A in EPSL Online Background Dataset (see footnote 1) and Fig. 2. Ridges near hot spots, indicated by the open squares in the figures, can have low Na8.0 for their depth, and form an offset field on the Na8.0 –Fe8.0 diagram. The screen of (La=Sm)N <1 does not exclude all samples that are from the margins of hot spots, which can still be influenced by the hot spot signature. Careful consideration of all these samples is im- portant because there is often the opposite relationship between depth and chemistry around certain hot spots, such as the Azores and the Galapagos [21,22]. This leads to the presence of samples from hot spot margins that are relatively deep but have low Na8.0 and thin crust, which contrasts with the global systematics for normal ridges. This is illustrated most strikingly at hole 504B, at the margins of the Galapagos hot spot. This site is very depleted in Na8.0 and (Sm=Yb)N . Despite its low Na8.0 , the crust is thin and deep. Collins et al. [42] report a crustal thickness of only 5 km for this region. Hence the crust here seems to be forming from material that has been previously depleted by the upwelling associated with the hot 16 E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 spot, leading to less crustal formation and very depleted compositions. Samples from such regions do not have as thick crust, nor were they formed at as shallow depths, as would be inferred from the Na8.0 . Therefore this effect needs to be taken into account for a more careful evaluation of the drill site data. How can such samples be distinguished on the basis of their chemical compositions? In addition to the slightly displaced field on the Na8.0 –Fe8.0 diagram, such samples are distinguishable using rare earth element ratios. Fig. 4 compares REE data from hot spot margins with normal ridges. The hot spot margins have higher (Ce=Sm)N ratios at the same (Sm=Yb)N ratio, and are distinguished from normal MORB across the entire compositional spectrum. This diagram, along with the major element systematics (Fig. 2), provides a means to distinguish samples that are influenced by nearby hot spots even though they are light REE depleted and would pass our chemical screen for hot spot samples. On both these diagrams, several of the drill sites clearly seem to be from hot spot margins (sites 100, 105, 259, 261 and 765D, crossed triangles). These sites are likely to have thinner crust and have been deeper at zero age than indicated by their Na8.0 and (Sm=Yb)N . Therefore they are not suitable for discussion of the depth–age problem. 5.2. Distinguishing samples erupted on old crust Most of the drill site data have chemical systematics that are consistent with the distribution of normal MORB (Figs. 2–4). A few sites, however, deviate from these systematics. In particular, the four sites that are nearest to the Ontong–Java plateau (holes 169, 462=462A and 802A, open triangles) do not plot on the Na8.0 –(Sm=Yb)N correlation (Fig. 5), and despite their high values of (Ce=Sm)N relative to (Sm=Yb)N (Fig. 4b) the data plot mostly within the normal MORB field on the Na8.0 –Fe8.0 (Fig. 2) diagram, in contrast to other samples with slightly elevated (Ce=Sm)N ratio. The origin of basalts from these sites has been debated, because the magnetic anomalies suggest that the crust is Jurassic, while the age of the samples is Cretaceous. On this basis, the recovered basalts were originally considered to be off-axis eruptions rather than basement samples. Castillo et al. [43] and Janney and Castillo Fig. 4. Sm=YbN vs. Ce=SmN (a) for zero-age basalts far from hot spot (black symbols) and at hot spot margins (white symbols). See [22] for references. All the samples have La=SmN <1. (b) Sm=YbN vs. Ce=SmN for zero-age basalts (far from hot spot: light gray area; at hot spot margin: darker gray area) and ODP=DSDP basalts older than 80 Ma (triangles). Most of the drill sites are consistent with the distribution of normal MORB (gray triangles; sites 164, 166, 197, 257, 303, 304, 307, 417=418, 581, 595 and 801C). Some of the drill sites plot in the MORB field at the margins of hot spots (crossed triangles, sites 100, 105, 259, 261, 765D). Sites close to the Ontong–Java plateau have been represented with white upside down triangles (holes 169, 462=462A and 802A). The gray circle represents the mean value of EPR samples having La=SmN <1. [44], however, have argued that there may have been a Cretaceous rifting event within the Jurassic crust, which led to the formation of these samples at spreading centers. The major and trace element data, while not conclusive in this regard, are suggestive of an off-axis origin, because of the combination of relatively high (Sm=Yb)N , Fe8.0 and (Ce=Sm)N . Eruption off-axis, beneath thicker lithosphere, would cause higher (Sm=Yb)N and Fe8.0 due to formation at a higher mean pressure. Because of the different E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 17 obtain 2.4% Na8.0 , the mean value for ancient crust, the intersection of the solidus occurs at 25 kbar, with a mantle potential temperature 50ºC hotter, forming 8.6 km of crust. These results suggest a mean temperature difference of about 50ºC, and a difference in crustal thickness of 1–2 km. The petrological approach can be applied quantitatively to individual drill sites as follows: the half space equation has the form: D D B.age/1=2 C D0 Fig. 5. Sm=YbN vs. Na8.0 for basalts far from hot spots (light gray area) and ODP=DSDP basalts older than 80 Ma (triangles). The symbols are defined in Fig. 4. The samples used have La=SmN <1. The large gray circle represents the mean for the East Pacific Rise. Sites 169, 462=462A and 802A (open triangles pointing down) deviate substantially from the normal systematics of ocean ridges and other ODP sites. chemical systematics, and the geological uncertainties with respect to these samples, these sites also are suspect for depth–age evaluation. 5.3. A petrological approach to depth vs. age The remainder of the sites, designated as ‘normal’ sites in Table 1 and by gray triangles in the diagrams, are good candidates for a petrological approach to the depth–age problem. These sites alone are also different from current MORB at better than 1% statistical significance. A first-order qualitative interpretation of the data can be based on the empirical Na8.0 –depth correlation for ridges today (see fig. Aa on EPSL Online Background Dataset, see footnote 1). The mean Na8.0 (or Na8.0 Ł ) of 2.37 for the remaining ancient drill sites suggests a zero-age depth of 2250 m, about 350 m shallower than average ridge crest today. A more quantitative approach makes use of the Klein and Langmuir model [21] for the depth– chemistry systematics at zero age as a response to mantle temperature. Then we can calculate the mantle temperature, initial depth and crustal thickness that correspond to any value of Na8.0 . Using mean values, current ocean ridges are formed by a mantle potential temperature of 1335ºC, with an intersection of the solidus for major melting occurring at 20 kbar, and creating 7 km of crust with Na8.0 of 2.66%. To (2) where the depth, D, is related to the initial depth D0 and the slope, B. B has a linear temperature dependence (e.g. [5,45,46]). In fact, of course both B and D0 are dependent on mantle temperature, because hotter mantle melts more, creating thicker crust and shallower initial depths. The Klein and Langmuir model relates both the mantle temperature and initial depth to Na8.0 . Regression of the empirical Na8.0 –depth correlation (fig. Aa on EPSL Online Background Dataset, see footnote 1) can be used to determine D0 : D0 D 2076 ð Na8:0 2690 (3) The relationship between Na8.0 and mantle temperature [21], along with the half space formulation for B [14] can then be used to determine B as a function of Na8.0 : BD 56:2 ð Na8:0 C 464 (4) Both D0 and B change with temperature of the mantle, because the initial depth varies, and the hotter ridges subside faster, as is obvious from the equations and pointed out as well on an empirical basis by Marty and Cazenave [14]. These equations allow the calculation of half space crustal subsidence curves for crust of any known age and composition of Na8.0 . Note that these curves differ from the classical geophysical model, because both slope and intercept are constrained by composition, which gives an inferred initial depth and temperature. The crustal composition gives an independent estimate of D0 and B, which differs for every drill site, while the classical geophysical approach assumes constant values of D0 and B. The petrologically constrained half space model depth can then be compared with the actual depth corrected for sediment loading [47]. This provides 18 E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 an independent estimate of the magnitude of the ‘flattening’ with respect to the half space model. We can also turn the calculation around, and use the observed depth to calculate what the Na8.0 of the crust would be if the half space model were to apply. This requires an iterative calculation, which converges to better than 1% after five cycles. Note that these calculations are independent for each drill site, and therefore are relevant to the depth–age problem whether the drilled sites are ‘representative’ of ancient crust or not. Results for these calculations are presented in Table 3. For the old drill sites (except those affected by hot spot margins effects), the mean discrepancy in depth is 300 m, which is about a third as much calculated from the classical geophysical approach using the constant values of D0 D 2600 m and B D 365 m=Ma1=2 [48]. A direct comparison of the classical calculation with the petrologically constrained calculation is shown in Fig. 6 and Table 3. Some sites show no apparent flattening at all relative to the petrologically constrained half space (sites 164, 417=418, 303 and 304). The mean deviations from the half space model are reduced by about two thirds — that is, two thirds of the depth–age flattening is accounted for by temporal change in mantle temperature and initial ridge depth. 6. Seismic evidence for thicker old crust? The chemical results can be tested by comparing the mean crustal thickness that would be predicted from the melting calculations with observed crustal thicknesses on old crust. The petrological calculations suggest a mean crustal thickness of 9:4 š 1:3 km in the past as compared to the 6–7 km at current normal ocean ridges. This is qualitatively consistent with the long-term observation of the apparent ‘thickening’ of oceanic crust with age [49,50]. The chemical results suggest that there is not thickening with time, but rather that old crust was thicker at the time of its formation. In order to test the distribution of the data quantitatively, we have compiled from the recent literature seismic refraction data that are not influenced by anomalous regions such as fracture zones and hot spots [6,51]. The data have been divided into two groups based on the age of the oceanic crust. A first group of 24 data corresponding to young oceanic crust (from 0 to 15 Ma) has been extracted from the White et al. compilation [51]. The average crustal thickness for this set of data is 6:6 š 0:8 km. A second group of 34 sites older than 80 Ma have been compiled from White et al. [51] and Nagihara et al. [6]. The average ‘old crust’ is thicker (7:4 š 1 km) in comparison to ‘young crust’. The crustal thickness difference between both groups is about 1 km. Applying statistical tests to the data shows that the crustal thickness difference between the two groups are significant at better than 1% probability. While the thicker crust in the past is consistent with hotter mantle beneath ridges >80 Ma, the increase in thickness is not as large as predicted from the petrological calculations (see Table 3). Although the crustal thickness measurements and compositional data are not linked data sets, this discrepancy is probably real. There are several possible explanations. First, it is possible that some of the remaining sites are still influenced by hot spot margin effects. Note in Table 3 the abnormally high values of the calculated crustal thickness for the ‘hot spot margin’ sites, as a result of their anomalously low Na8.0 . A second possible explanation comes from the recent results on crustal thickness from the southern EPR collected during the MELT experiment. Canales et al. [52] report crustal thicknesses for the normal fastspreading EPR of only 4.8–5.7 km at 17º150 S and 5.4–6.3 km at 15º550 S. The petrological calculations assume that all the melt from the melting regime is segregated to form the crust. These seismic results suggest that at fast spreading rates, some of the melt may not be segregated from the mantle. In this case, the appropriate comparison for many of the ancient drill sites may be about 5.5 km, 2 km thinner than the average measurement on old fast-spreading crust. A 2 km increase would be more consistent with the petrological calculations. Furthermore, it is possible that the deeper and broader melting regimes associated with hotter mantle may not have as efficient melt extraction at fast spreading rates, which would also lead to lower crustal thicknesses than our calculations, which assume complete melt extraction. These questions clearly require careful seismic ex- Table 3 DSDP=ODP sites older than 80 Ma with normal and perturbed systematics (hot spot margin) Hole Age Unloaded Model depth Residual Model depth Unloaded Model crustal Actual crustal Actual Na8.0 Calculated Na8.0 Calculated depth from Na8.0 depth 1=2 Sp minus 1=2 Sp thickness thickness from depth minus actual 5313 5579 5977 6224 5813 380 744 69 552 259 6428 6903 7100 7159 7188 1495 580 1054 1487 1634 Mean value Normal chemical systematics 164 110 5678 581 116 5726 417D 418A=B 118 5737 303 130 5817 166 132 5173 304 133 5872 257 135 5469 307 148 5919 197 151 6343 595 155 5649 801C 167 6043 For normal mean Systematics st.dev. For all sites 12.4 14.6 12.1 10.2 15.1 8.44 2.11 1.98 2.13 2.27 1.95 1.86 2.49 2.17 1.87 1.77 0.25 0.29 0.04 0.40 0.18 2.41 2.17 2.27 2.26 2.56 2.3 2.24 2.35 2.51 2.38 2.33 2.38 2.34 2.32 2.24 1.78 2.25 1.95 2.12 2.38 1.86 2.02 0.03 0.17 0.05 0.02 0.78 0.05 0.29 0.23 0.13 0.52 0.31 2.34 0.12 2.13 0.21 0.20 0.26 12.9 5759 5499 5677 5844 6303 5945 5875 6227 6489 6362 6451 81 227 60 27 1130 73 406 308 146 713 408 273 384 211 6428 6531 6565 6762 6794 6809 6841 7040 7085 7144 7317 750 805 828 945 1621 937 1372 1121 742 1495 1274 8.5 11.5 10.2 10.3 7.1 9.8 10.6 9.2 7.5 8.9 9.4 1081 315 1134 9.4 1.3 7.28 8.6 6.7 6.8 Unloaded depth (m) is calculated using the procedure of LeDouaran and Parsons [47]. Model depth from Na8.0 is the depth (in m) calculated from a thermal dependent age–depth equation (Eqs. 2–4) using Na8.0 or Na8.0 Ł . Residual depth (m) is the difference between the depth calculated from Na8.0 (or Na8.0 Ł ) and the observed unloaded depth. Model depth 1=2 Sp (half space model) is the reconstructed depth (in m) based on a constant slope and intercept in the age–depth equation (see [48]). Unloaded minus 1=2 Sp is the difference between the observed unloaded depth and the depth calculated with the 1=2 Sp model. Model crustal thickness is the thickness of the crust (in km) produced during a passive upwelling of the mantle, calculated from the Na8.0 (or Na8.0 Ł ) using: Crustal thick. D 171 ð 10 0:54ðNa8:0 . Actual crustal thickness (km) is measured by seismic refraction (references are given in the appendix on EPSL Online Background Dataset (see footnote 1). Calculated Na8.0 from depth is inferred from the depth back-tracked along a half space model to zero and compared to zero-age basalts using Eq. 3. Calculated minus actual is the difference between the calculated and the observed Na8.0 . E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 Perturbed chemical systematics 259 110 4933 765D 139 6323 261 152 6046 105 156 5672 100 158 5554 19 20 E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 Fig. 6. Model depth vs. observed unloaded depth for drill sites older than 80 Ma with normal systematics. Open and closed symbols show model depths calculated in two ways. The open symbols are the classical half space model. Closed symbols are petrologically constrained model depths calculated using the equations given in the text. The petrologically constrained depths, which reflect variations in mantle temperature, are much closer to the observations. Petrologically constrained depths were calculated using Na8.0 (sites 166, 304, 417=418, 595, 801C) or Na8.0 Ł (sites 164, 197, 257, 303, 307, 581). Thin lines show plus or minus 10% variations around the 1:1 thicker line. periments carried out in coordination with drilling on old crust, and a more comprehensive understanding of melting and melt segregation as a function of spreading rate and mantle temperature. 7. Mantle temperature in space and time Data from ocean ridges today suggest that there are temperature variations in the upper mantle of as much as 250ºC from the hottest to the coldest ridges [21–24]. Within this large amplitude of thermal variation within the earth, there is clearly room for significant temporal changes in the ambient temperature of the upper mantle. The petrological results from the old drill sites suggest that the mean temperature of the upper mantle beneath ridges at >80 Ma may have been significantly hotter (by about 50ºC) than at present. This is consistent with the large number of oceanic plateaus being created in this time period [53]. In the Pacific today, no giant plateaus are being created, and the EPR is remarkably constant in depth. At greater than 80 Ma, it appears that the Pacific was in a very different state of activity, driven by hot mantle material [53]. This material would have created the oceanic plateaus, and generally contributed to higher mean temperatures in the upper mantle. Other sites we have investigated are adjacent to passive margins, where melting effects associated with continental break-up may also have influenced crustal generation. Are we mostly seeing a ‘hot spot effect’ in this analysis? The data we have considered have been selected in order to avoid a hot spot effect, first by eliminating all holes on shallow crust with (La=Sm)N greater than 1, and then by eliminating those sites which appear to have come from hot spot margins. This leads to the elimination of a significant proportion of the data. This in itself provides additional evidence for a particularly warm mantle during this period, since hot spots and hot spot influence are so difficult to avoid. That is, hot spots and continental break-up may have led to rather different mean conditions at ocean ridges for crust older than 80 Ma in comparison to conditions today. This led to shallow crust, which has influenced our modern interpretation of depth–age systematics. Of course, the hot spot influence we are discussing is very different from the plate reheating by hot spots which has long been suggested to contribute to the flattening of the depth–age curve. This latter influence occurs when crust passes over a hot spot during its traverse across the seafloor. In contrast, the results reported here are indicative of hot mantle activity at the time of crust formation. Finally, we note that since ridges are passive features, large temperature differences along ridges imply large differences in the mantle beneath the plates as well. This has other important implications for the interpretation of the drill site data. Today the East Pacific Rise lies over mantle of normal temperature. If it were displaced 2000 km to the west, it would be in line with the mid-Pacific hot spots, and presumably overlie hotter mantle. Was the mean mantle temperature hotter in the past? Or were ridges simply over the parts of the mantle that had higher temperatures? Therefore an apparent ‘change with age’ needs to be considered in both space and time. It is not clear from the data if the ambient temperature of the upper mantle was actually hotter in this time period, or whether ridges were simply sampling hotter regions of the mantle. Certainly a E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 decrease of 50ºC over 100 Ma cannot be normal secular cooling of the upper mantle. For the depth–age problem, the four-dimensional change in mantle temperature also needs to be considered. The older portions of plates traveling across the globe are influenced by the temperature of the mantle over which they pass. Therefore variations in subsidence will result not only from temporal variations in temperature at zero age, but also will result from temperature variations in the underlying mantle across which the plate spreads with age. This could lead to varying subsidence rates, and subsidence that is not regular on a global basis, but instead can flatten and steepen at different ages in response to temperature variations in the asthenosphere. 8. Conclusions Data from drill sites older than 80 Ma that have penetrated the basement of the seafloor reveal that the global distribution of basalt compositions at that time may have been very different from the distribution at present. Basalts had chemical systematics similar to those at present ridges, but the mean compositions had lower Na8.0 , (Sm=Yb)N , and Zr=Y and higher CaO=Al2 O3 and possibly Fe8.0 . These systematics suggest that the mantle was hotter in this time period by about 50ºC, that the depth of the ridge was several hundred meters shallower, and the crust 1–2 km thicker. The possibility of a hotter mantle has important implications for a wide range of associated problems, including sea level change and carbonate compensation (through effects on basement depth), and possibly even arc volcanism. It also accounts easily for the apparent ‘thickening of ocean crust with time’. The crustal composition provides an independent estimate of cooling and subsidence for the seafloor. These estimates show that about one half to two thirds of the observed flattening relative to a boundary layer model is due to the change in mantle temperature and crustal composition. A few hundred meters of flattening is all that is required by plate reheating by hot spots or by other mechanisms. Indeed, temperature variations in the upper mantle — beneath old plates as well as ridges — are likely 21 to lead to a wide variety of depth–age relationships, such as are appearing in the more comprehensive evaluations that are possible using the new satellite bathymetry [3,54]. Finally, we note that the database from which these conclusions are drawn is very limited. Not only are there few sites with ample basement penetration, but the data for many samples that do exist is incomplete. A fuller evaluation will require further off-axis drilling combined with seismic experiments and detailed geochemical studies. Acknowledgements We are grateful to C. Jaupart, J. Ludden and E. Klein for fruitful discussions, and for reviews by R. Batiza, P. Castillo and P. LeCroart. This study has been supported by INSU ‘Intérieur de la Terre’ and ‘CNRS Fédération de Recherche No. 32’ and by the US National Science Foundation. [AC] Appendix A Appendix available on EPSL Online Background Dataset 2 . References [1] M.G. Langseth, X. LePichon, M. Ewing, Crustal structure of mid-ocean ridges, 5: Heat flow through the Atlantic Ocean and convection currents, J. Geophys. Res. 71 (1966) 5321–5355. [2] D.P. McKenzie, J.G. Sclater, Heat flow in the eastern Pacific and sea-floor spreading, Bull. Volcanol. 33 (1969) 101–118. [3] W.H.F. Smith, D.T. Sandwell, Global sea-floor topography from satellite altimetry and ship depth soundings, Science 277 (1997) 1956–1962. [4] D.L. Turcotte, E.R. Oxburgh, Finite amplitude convection cells and continental drift, J. Fluid Mech. 28 (1967) 29–42. [5] E.E. Davis, C.R.B. Lister, Fundamentals of ridge crest topography, Earth Planet. Sci. Lett. 21 (1974) 405–413. [6] S. Nagihara, C.R.B. Lister, J.G. Sclater, Reheating of old oceanic lithosphere: deductions from observations, Earth Planet. Sci. Lett. 139 (1–2) (1996) 91–104. [7] B. Parsons, D.P. McKenzie, Mantle convection and the 2 http://www.elsevier.nl/locate/epsl, mirror site: http://www.elsevier.com/locate/epsl 22 [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 thermal structure of the Plates, J. Geophys. Res. 83 (1978) 4485–4495. D.W. Forsyth, The evolution of the upper mantle beneath mid-ocean ridges, Tectonophysics 38 (1977) 89–118. G.T. Jarvis, W.R. Peltier, Mantle convection as a boundary layer phenomenon, Geophys. J.R. Astron. Soc. 68 (1982) 389–472. B.J. Wood, D.A. Yuen, The role of lithospheric phase transitions on seafloor flattening at old ages, Earth Planet. Sci. Lett. 66 (1983) 303–314. Y. Ricard, C. Froidevaux, L. Fleitout, Global plate motions and the geoid: a physical model, Geophys. J. 93 (1988) 477–484. J.P. Morgan, W.H. S Smith, Flattening of the sea-floor depth age curve as a response to asthenospheric flow, Nature 359 (1992) 524–527. D.E. Hayes, Age–depth relationships and depth anomalies in the South-east Indian ocean and South Atlantic Ocean, J. Geophys. Res. 93 (1988) 2937–2954. J.C. Marty, A. Cazenave, Regional variations in subsidence rate of oceanic plates; a global analysis, Earth Planet. Sci. Lett. 94 (1989) 301–315. P. Calcagno, A. Cazenave, Subsidence of the seafloor in the Atlantic and Pacific oceans: regional and large scale variations, Earth Planet. Sci. Lett. 126 (1994) 473–492. K.A. Kane, D.E. Hayes, Tectonic corridors in the South Atlantic: evidence for long-lived mid-ocean ridge segmentation, J. Geophys. Res. 97 (1992) 17317–17330. D.E. Hayes, K.A. Kane, Long-lived mid-ocean ridge segmentation of the Pacific–Antarctic ridge and the Southeast Indian Ridge, J. Geophys. Res. 99 (1994) 19679–19692. K.A. Kane, D.E. Hayes, Long-lived mid-ocean ridge segmentation: constraints and models, J. Geophys. Res. 99 (1994) 19693–19706. S.B. Sherman, J.L. Karsten, E.M. Klein, Petrogenesis of axial lavas from the southern Chile Ridge: major element constraints, J. Geophys. Res. 102 (1997) 14963–14990. J.G. Schilling, C. Ruppel, A.N. Davis, B. McCully, S.A. Tighe, R.H. Kingsley, J. Lin, Thermal structure of the mantle beneath the equatorial Mid-Atlantic Ridge: inferences from the spatial variation of dredged basalt compositions, J. Geophys. Res. 100 (1995) 10057–10079. E.M. Klein, C.H. Langmuir, Global correlations of ocean ridge basalt chemistry with axial depth and crustal thickness, J. Geophys. Res. 92 (1987) 8089–8115. C.H. Langmuir, E.M. Klein, T. Planck, Petrological Systematics of Mid-ocean Ridge Basalts: Constraints on Melt Generation beneath Ocean Ridges, AGU, Washington, 1992. E. Humler, J.L. Thirot, J.P. Montagner, Global correlations of mid-ocean ridge basalt chemistry with seismic tomographic images, Nature 364 (1993) 225–228. C.H. Langmuir, Deep thoughts on the mantle, Nature 364 (1993) 191–192. Y. Shen, D.W. Forsyth, Geochemical constraints on initial and final depths of melting beneath mid-ocean ridges, J. Geophys. Res. 100 (2) (1995) 2211–2237. J.W. Bown, R.S. White, Variation with spreading rate of [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] oceanic crustal thickness and geochemistry, Earth Planet. Sci. Lett. 121 (1994) 435–449. M.J. Keen, E.M. Klein, W.G. Melson, Ocean-ridge basalt compositions correlated with paleobathymetry, Nature 345 (1990) 423–426. Y. Niu, R. Hekinian, Spreading rate dependence of the extent of mantle melting beneath ocean ridges, Nature 385 (1997) 326–329. P. Lecroart, F. Albarède, A. Cazenave, Correlation of MidOcean Ridge Basalt chemistry with the geoid, Earth Planet. Sci. Lett. 153 (1997) 37–55. P. Bienvenu, H. Bougault, J.L. Joron, M. Treuil, L. Dmitriev, MORB alteration: rare-earth element=non-rareearth hygromagmaphile element fractionation, Chem. Geol. 82 (1990) 1–14. A. Spivack, H. Staudigel, Low alteration of the upper crust and alkalinity budget of seawater, Chem. Geol. 115 (1994) 239–247. K.P. Jochum, S.P. Verma, Extreme enrichment of Sb, Tl and other trace elements in altered MORB, Chem. Geol. 130 (1996) 289–299. W.B. Bryan, G. Thompson, J.N. Ludden, Compositional variation in normal MORB from 22º to 25ºN: Mid-Atlantic Ridge and Kane fracture zone, J.Geophys. Res. 86 (1981) 11815–11836. J.R. Reynolds, C.H. Langmuir, Petrological systematics of the Mid-Atlantic ridge south of Kane: implications for models of ocean crust formation, J. Geophys. Res. 102 (B7) (1997) 14915–14946. M.A. Dungan, J.M. Rhodes, P.E. Long, D. Blanchard, J.C. Brannon, K.V. Rodges, The petrology and geochemistry of basalts from site 396, Leg 45–46 of the Deep Sea Drilling Project, Proc. ODP 46 (1978) 89–113. M.A. Dungan, P.E. Long and J.M. Rhodes, The petrography, mineral chemistry and one atmosphere phase relations of basalts from site 395–leg 45 DSDP, Init. Rep. DSDP 46 (1978). J.M. Rhodes, D.P. Blanchard, M. A Dungan, K.V. Rodges, J.C. Brannon, Chemistry of Leg 45 basalts, Proc. ODP 46 (1978) 447–459. R.K. Srivastava, R. Emmermann, H. Puchelt, Petrology and geochemistry of basalts from Deep Sea Drilling Project Leg 54, Proc. ODP 54 (1980) 671–693. J.H. Natland, W.G. Melson, Compositions of basaltic glasses from the East Pacific rise and the Siqueiros fracture zone near 9ºN, Proc. ODP 54 (1980) 705–723. S.E. Humphris, R.N. Thompson, I.L. Gibson, G.F. Marriner, Comparison of geochemistry of basalts from the East Pacific Rise and Siqueiros fracture zone, Deep Sea Drilling Project 54 (1980) 635–649. P.E. Janney, P. Castillo, Geochemistry of Mesozoic Pacific mid-ocean ridge basalts: constraints on melt generation and the evolution of the Pacific upper mantle, J. Geophys. Res. 102 (1997) 5207–5229. J.A. Collins, G.M. Purdy, T.M. Brocher, Seismic velocity structure at deep sea drilling project site 504B, Panama E. Humler et al. / Earth and Planetary Science Letters 173 (1999) 7–23 [43] [44] [45] [46] [47] [48] basin: evidence for thin oceanic crust, J. Geophys. Res. 94 (1989) 9283–9302. P. Castillo, R. Batiza, R.J. Stern, Petrology and geochemistry of Nauru basin igneous complex: large-volume, offridge eruptions of MORB-like basalt during the Cretaceous, Proc. ODP 89 (1986) 555–576. P.E. Janney, P.R. Castillo, Basalts from the Central Pacific Basin: evidence for the origin of Cretaceous igneous complexes in the Jurassic western Pacific, J. Geopys. Res. 101 (B2) (1996) 2875–2893. B. Parsons, J.G. Sclater, An analysis of the variation of ocean floor bathymetry and heat, J. Geophys. Res. 82 (1977) 803–827. D.L. Turcotte and G. Schubert, D.L., Geodynamics. Applications of Continuum Physics to Geological Problems. Wiley, New York, NY, 1982. S. LeDouaran, B. Parsons, A note on the correction of ocean floor depths for sediment loading, J. Geophys. Res. 87 (1982) 4715–4722. C.A. Stein, S. Stein, A model for the global variation in oceanic depth and heat flow with lithospheric age, Nature 23 359 (1992) 123–129. [49] X. LePichon, R.E. Houtz, C.L. Drake, E. Nafe, Crustal structure of the mid-ocean ridges, 1: Seismic refraction measurements, J. Geophys. Res. 87 (1965) 8417–8425. [50] J.P. Goslin, P. Beuzart, J. Francheteau, X. LePichon, Thickening of the oceanic layer in the Pacific Ocean, Mar. Geophys. Res. 1 (1972) 418–427. [51] R.S. White, D. McKenzie, R.K. O’Nions, Oceanic crustal thickness from seismic measurements and rare earth element inversions, J. Geophys. Res. 97 (B13) (1992) 19683– 19715. [52] J.P. Canales, R.S. Detrick, S. Bazin, A.J. Harding, J.A. Orcutt, Off axis crustal thickness across and along the East Pacific Rise within the melt area, Science 280 (1998) 1218– 1221. [53] R.L. Larson, Latest pulse of Earth: evidence for mid-Cretaceous superplume, Geology 19 (1991) 547–550. [54] W.H.L. Smith, D.T. Sandwell, Bathymetric prediction from bands altimetry and sparse shipboard bathymetry, J. Geophys. Res. 99 (1994) 21803–21824.