Download Depth versus age: new perspectives from the chemical compositions

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.