Journal of Petrology Advance Access originally published online on October 11, 2007
Journal of Petrology 2008 49(4):633-664; doi:10.1093/petrology/egm051
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Global Correlations of Ocean Ridge Basalt Chemistry with Axial Depth: a New Perspective
1Department of Earth Sciences, Durham University, Durham DH1 3LE, UK
2Institute of Geography and Earth Studies, Aberystwyth University, Aberystwyth SY23 3DB, UK
RECEIVED FEBRUARY 25, 2007; ACCEPTED AUGUST 7, 2007
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONTHE GLOBAL MORB DATASET...CORRECTION FOR FRACTIONATION...GLOBAL RIDGE AXIAL DEPTH...DISCUSSIONSUMMARYSUPPLEMENTARY DATAREFERENCES |
|---|
The petrological parameters Na8 and Fe8, which are Na2O and FeO contents in mid-ocean ridge basalt (MORB) melts corrected for fractionation effects to MgO = 8 wt%, have been widely used as indicators of the extent and pressure of mantle melting beneath ocean ridges. We find that these parameters are unreliable. Fe8 is used to compute the mantle solidus depth (Po) and temperature (To), and it is the values and range of Fe8 that have led to the notion that mantle potential temperature variation of
TP = 250 K is required to explain the global ocean ridge systematics. This interpreted TP = 250 K range applies to ocean ridges away from ‘hotspots’. We find no convincing evidence that calculated values for Po, To, and TP using Fe8 have any significance. We correct for fractionation effect to Mg# = 0·72, which reveals mostly signals of mantle processes because melts with Mg# = 0·72 are in equilibrium with mantle olivine of Fo89·6 (vs evolved olivine of Fo88·1–79·6 in equilibrium with melts of Fe8). To reveal first-order MORB chemical systematics as a function of ridge axial depth, we average out possible effects of spreading rate variation, local-scale mantle source heterogeneity, melting region geometry variation, and dynamic topography on regional and segment scales by using actual sample depths, regardless of geographical location, within each of 22 ridge depth intervals of 250 m on a global scale. These depth-interval averages give Fe72 = 7·5–8·5, which would give TP = 41 K (vs 
250 K based on Fe8) beneath global ocean ridges. The lack of Fe72–Si72 and Si72–ridge depth correlations provides no evidence that MORB melts preserve pressure signatures as a function of ridge axial depth. We thus find no convincing evidence for TP > 50 K beneath global ocean ridges. The averages have also revealed significant correlations of MORB chemistry (e.g. Ti72, Al72, Fe72, Mg72, Ca72, Na72 and Ca72/Al72) with ridge axial depth. The chemistry–depth correlation points to an intrinsic link between the two. That is, the 5 km global ridge axial relief and MORB chemistry both result from a common cause: subsolidus mantle compositional variation (vs TP), which determines the mineralogy, lithology and density variations that (1) isostatically compensate the 5 km ocean ridge relief and (2) determine the first-order MORB compositional variation on a global scale. A progressively more enriched (or less depleted) fertile peridotite source (i.e. high Al2O3 and Na2O, and low CaO/Al2O3) beneath deep ridges ensures a greater amount of modal garnet (high Al2O3) and higher jadeite/diopside ratios in clinopyroxene (high Na2O and Al2O3, and lower CaO), making a denser mantle, and thus deeper ridges. The dense fertile mantle beneath deep ridges retards the rate and restricts the amplitude of the upwelling, reduces the rate and extent of decompression melting, gives way to conductive cooling to a deep level, forces melting to stop at such a deep level, leads to a short melting column, and thus produces less melt and probably a thin magmatic crust relative to the less dense (more refractory) fertile mantle beneath shallow ridges. Compositions of primitive MORB melts result from the combination of two different, but genetically related processes: (1) mantle source inheritance and (2) melting process enhancement. The subsolidus mantle compositional variation needed to explain MORB chemistry and ridge axial depth variation requires a deep isostatic compensation depth, probably in the transition zone. Therefore, although ocean ridges are of shallow origin, their working is largely controlled by deep processes as well as the effect of plate spreading rate variation at shallow levels. KEY WORDS: mid-ocean ridges; mantle melting; magma differentiation; petrogenesis; MORB chemistry variation; ridge depth variation; global correlations; mantle compositional variation; mantle source density variation; mantle potential temperature variation; isostatic compensation
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONTHE GLOBAL MORB DATASET...CORRECTION FOR FRACTIONATION...GLOBAL RIDGE AXIAL DEPTH...DISCUSSIONSUMMARYSUPPLEMENTARY DATAREFERENCES |
|---|
The global mid-ocean ridge system is the dynamic manifestation of mantle convection that governs plate tectonics. Our current understanding of ocean ridge magmatism, which creates the ocean crust, results from multidisciplinary efforts. Seafloor geological and geophysical observations have laid the foundation. Surface-ship dredging, submersibles, and deep-sea drilling have collected the magmatic products to infer magmatic processes. Theoretical considerations have established that ocean ridges are mostly passive features in the sense that the mantle beneath the ridge rises adiabatically and melts by decompression in response to plate separation. Experimental petrology has been particularly important in facilitating interpretations of mid-ocean ridge basalts (MORB) in terms of physical processes of mantle melting and magma evolution. Dave Green has been one of the pioneers (Green & Ringwood, 1963
, 1964, 1967, 1970; O’Hara & Yoder, 1963, 1967; O’Hara, 1965, 1967, 1968a, 1968b, 1970; Green, 1971, 1973; Wyllie, 1973; Green et al., 1979; Presnall et al., 1979; Jaques & Green, 1980; Stolper, 1980; Falloon & Green, 1987, 1988; Green & Falloon, 1998, 2005) in using peridotite melting experiments to study MORB petrogenesis.
The paper by Jaques & Green (1980) in particular marked a turning point for MORB petrogenesis research. It demonstrated for the first time how the bulk compositions of partial melts of mantle peridotite vary systematically with melting conditions and mantle composition, and thus laid the groundwork for the establishment of detailed petrogenetic grids and empirical models aimed at a quantitative understanding of ocean ridge melting and MORB genesis (e.g. Klein & Langmuir, 1987; McKenzie & Bickle, 1998); Niu & Batiza, 1991a; Kinzler & Grove, 1992; Langmuir et al., 1992; Kinzler, 1997; Niu, 1997. Figure 1, which summarizes the model results of Niu (1997), illustrates the concept of Jaques & Green (1980).
|
The paper by Klein & Langmuir (1987
) is a milestone contribution on MORB genesis. Following the idea of Dick & Fisher (1984) and Dick et al. (1984), Klein & Langmuir (1987) showed that location-averaged MORB chemistry correlates with ridge axial depth on a global scale. For example, Fe8 (i.e. FeO corrected for the effect of crystal fractionation to MgO = 8 wt%) increases whereas Na8 decreases as the ridge shallows (Fig. 2a and b). They interpreted such correlations as resulting from varying pressures and degrees of partial melting of a compositionally uniform mantle (Fig. 2c) caused by temperature variation in the subsolidus mantle of up to 250 K from beneath cold deep ridges to hot shallow ridges with a compensation depth of 150–200 km. In a more comprehensive treatment, Langmuir et al. (1992) re-examined the same problem further and confirmed the basic conclusions of Klein & Langmuir (1987).
|
These interpretations have been widely accepted as the way the mantle actually works despite numerous arguments for alternative working hypotheses (Brodholt & Batiza, 1989
; Natland, 1989; Albarède, 1992; Niu & Batiza, 1993, 1994; Shen & Forsyth, 1995; Niu, 1997, 2004; Niu & Hékinian, 1997a, 1997b; Castillo et al., 1998, 2000; Niu et al., 2001, 2002; Presnall et al., 2002; Presnall & Gudfinnsson, this issue).
In this paper we re-examine the Klein & Langmuir (1987) and Langmuir et al. (1992) interpretations, in particular the argument for 250 K mantle temperature variation and the suggestion for a compensation depth of 150–200 km required to explain global ridge axial depth variations. We offer an alternative approach for fractionation correction. We conclude that mantle temperature variation is likely to be inadequate and that a significant contribution from mantle compositional variation is required to explain global axial depth and MORB compositional variations and correlations on a global scale.
|
THE GLOBAL MORB DATASET AND NEW PERSPECTIVES |
|---|
|
TOPABSTRACTINTRODUCTIONTHE GLOBAL MORB DATASET...CORRECTION FOR FRACTIONATION...GLOBAL RIDGE AXIAL DEPTH...DISCUSSIONSUMMARYSUPPLEMENTARY DATAREFERENCES |
|---|
About 13 900 glass analyses are available in the global MORB database (PetDB, http://beta.www.petdb.org/). In our analyses, we have excluded samples (1) with SiO2 > 53 wt% (up to 71 wt%, basaltic andesite to rhyolite), (2) without water depth information, (3) from depths <400 m to minimize any ‘hotspot’ effect, and (4) with MgO <7 wt% to avoid unnecessary errors associated with fractionation correction for the more evolved samples because of the non-linear liquid lines of descent (LLDs) for most oxides (see Electronic Appendix A, available for downloading at http://www.petrology.oxfordjournals.org/). This leaves 9130 samples on a global scale to work with.
Figure 3 shows sample distribution with ridge axial depth. A logarithmic scale is used to magnify the details at both deep and shallow ridge depths. This log-normal sample distribution largely reflects the ridge axial depth variation on a global scale. Much of the global mid-ocean ridge lies between 2000 and 3000 m water depth in geographical regions readily accessible for sampling.
|
As expected, compositions of samples vary significantly along ridge segments and with respect to physical observables such as spreading rate (Niu & Batiza, 1993
; Niu & Hékinian, 1997a, 1997b) or ridge axial depth on all scales (Brodholt & Batiza, 1989; Niu & Batiza, 1993, 1994; Niu et al., 2001). In this case, averaging is often used to ‘average out’ outliers or reduce the scatter in pursuit of first-order systematics (e.g. Klein & Langmuir, 1987; Langmuir et al., 1992; Niu & Hékinian, 1997a; Rubin & Sinton, 2007). The way to average is not arbitrary, but follows some basic principles. As long as the analytical data are of good quality, there is no reason to exclude arbitrarily any ‘outliers’ because each analysis, whatever its composition may be, results from compound mantle processes. Assuming that the shallow-level fractional crystallization effect can be effectively removed (see below), the data still possess the signatures of fertile mantle compositional (lithological and mineralogical) variation on all scales, spreading-rate dependent variation, mean pressures of melting, and mantle temperature variation (Niu et al., 2001). The way to effectively deconvolve all these probable effects is to average the data with respect to a particular variable. For example, to examine the effect of spreading rate variation on MORB chemistry, we must average the MORB data with respect to spreading rate within chosen spreading rate windows (Niu & Hékinian, 1997a; Rubin & Sinton, 2007). To examine if MORB chemistry varies with ridge axial depth, we must average the MORB data with respect to ridge axial depth within chosen depth intervals. This is mathematically the same as seeking the partial derivatives of a quantity with respect to a particular variable, with all other variables held constant. There are two factors to consider when averaging MORB data with respect to ridge axial depth:
- The global ridge axial depth varies from near sea level at the Reykjanes Ridge immediately south of Iceland to as deep as 6000 m in portions of the Cayman Rise. What would be the ideal depth interval to average? We choose a 250 m depth interval because this value approximates the ridge depth variation over 500 km regional-scale lengths and because we will have >20 averaged data points to work with. This number of data points is large enough to be statistically significant without compromising first-order systematics, but is small enough to smooth out secondary effects while making the first-order global systematics prominent.
- Ridge axial depth can vary significantly on ridge segment scales, in particular along slow-spreading ridges such as the Mid-Atlantic Ridge (MAR). For example, within the 90 km long MAR ridge segment at 26°S (Niu & Batiza, 1994), the axis is 2500 m deep at the segment center, but deepens to 4000 m towards segment offsets.
How to do the averaging?
Klein & Langmuir (1987) chose to smooth the variation in ridge depth by using a derived single depth value for all samples from a particular ridge segment. Such smoothing of ridge depth is arbitrary, and such ‘pre-treatment’ of a physical observable potentially loses information. In contrast, we use the actual depths at which basalt samples were erupted in this study. Table 1 gives the averages for samples within each of the 22 depth intervals (Fig. 3). To minimize the Iceland ‘hotspot’ effect, we exclude samples in the 0–250 m depth interval. As there are no samples in the 250–400 m depth range with MgO >7 wt%, the 250–500 m depth interval actually includes only samples from the ridge axis at depths >400 m. We thus have 22 depth intervals of 250 m or 22 averaged data points to work with (see Fig. 3).
|
Chemistry–depth correlations: averages of fractionation-uncorrected raw data
Table 1 gives depth-interval averages of fractionation-uncorrected major (SiO2, TiO2, Al2O3, FeOt, MgO, CaO and Na2O) and minor (MnO, K2O and P2O5) element oxides. Figure 4 shows that TiO2, Al2O3, Na2O and Mg# increase whereas FeO, CaO and CaO/Al2O3 decrease with increasing ridge axial depth. The MgO–depth correlation is less significant partly because of the use of samples with MgO >7 wt% only. All these correlations are statistically significant at the >95% confidence level. SiO2 and the minor oxides MnO and P2O5 do not show systematic variation vs ridge axial depth. However, Fig. 5a shows, despite a higher value ‘outlier’ in the depth interval 2000–2250 m, that K2O does show a statistically significant decrease towards shallower ridges.
|
|
Except for the inverse CaO/Al2O3–depth correlation shown by Klein & Langmuir (1987
) and Brodholt & Batiza (1989) and the associations of high-Na2O MORB with deep ridges and low-Na2O MORB with shallow ridges identified by Bryan & Dick (1982), Dick & Fisher (1984), Dick et al. (1984) and Brodholt & Batiza (1989), Table 1 and Figures 4 and 5a represent the first demonstration that most major and minor element oxides (except SiO2, MnO and P2O5) in global MORB melts correlate with ridge axial depth, and such correlations are more prominent when depth-interval averages are plotted.
The decreasing Mg# in MORB melts from deep ridges (Mg# 0·68) to shallow ridges (Mg# 0·55) (Fig. 4) is consistent with, and best explained by, advanced degrees of fractional crystallization towards shallow ridges (lowering Mg#). However, the systematics of other oxides (except Al2O3, MgO and, to a lesser extent, FeO in a qualitative sense) with ridge axial depth cannot be readily explained by the same process acting on a similar parental magma. Hence, part of the fractionation-uncorrected MORB compositional variation as a function of ridge axial depth must reflect parental magma compositional variation with ridge axial depth. We will address in subsequent sections why MORB parental melt composition varies with ridge axial depth and why parental melts delivered to shallow ridges crystallize (evolve) more than those delivered to deep ridges.
|
CORRECTION FOR FRACTIONATION EFFECTS |
|---|
|
TOPABSTRACTINTRODUCTIONTHE GLOBAL MORB DATASET...CORRECTION FOR FRACTIONATION...GLOBAL RIDGE AXIAL DEPTH...DISCUSSIONSUMMARYSUPPLEMENTARY DATAREFERENCES |
|---|
It is well established that MORB melts are not primary magmas, but are variably evolved derivatives, primarily through fractional crystallization as a result of cooling at shallow levels (e.g. O’Hara, 1968a
; Walker et al., 1979; Stolper, 1980; Langmuir, 1989; Sinton & Detrick, 1992; O’Hara & Herzberg, 2002). Therefore, to see the mantle signals that MORB melts might preserve, it is necessary to correct for the effect of crustal-level fractional crystallization.
The Klein & Langmuir (1987) method and interpretations
It is not possible to correct unambiguously the effect of fractional crystallization back to the status of primary magmas because the exact fractionation processes are unknown. Klein & Langmuir (1987, p. 8090) arbitrarily chose to correct to MgO = 8 wt%. The correction is made on Na2O–MgO and FeO–MgO variation diagrams by back-tracking along LLDs for samples with MgO <8 wt% and forward for samples with MgO >8 wt% to MgO = 8 wt%. Klein & Langmuir (1987) excluded samples with MgO <5 wt% to avoid possible correction errors. MORB melts from different ridges can have different LLDs, and different LLDs can exist in a given ridge segment. However, the LLDs often define parallel or sub-parallel trends on MgO variation diagrams. Hence, applying a single set of LLD slopes for Na2O–MgO and FeO–MgO to the global database is acceptable [see Klein & Langmuir (1987, Fig. 2) and also Langmuir et al. (1992)]. Na8 and Fe8 (Fig. 2) are such fractionation-corrected values for Na2O and FeO at MgO = 8·0 wt% in MORB melts by Klein & Langmuir (1987) and Langmuir et al. (1992). Na8 and Fe8 are now well-known petrological parameters widely used to indicate the extents and pressures of mantle melting required to generate MORB, and basalts from ocean islands and other tectonic settings, including continental flood basalts.
Figure 1 shows that, in primary melts, Na2O decreases with increasing degree of melting and FeO is higher in higher pressure melts (Jaques & Green, 1980). Consequently, Klein & Langmuir (1987) interpreted the correlations of MORB Na8 and Fe8 with ridge axial depth (Fig. 2) as resulting from varying extents and pressures of melting caused by temperature variation of up to 250 K in the subsolidus mantle from beneath cold deep ridges to hot shallow ridges. It should be noted that Klein & Langmuir (1987) did not use mantle potential temperature, TP, defined by McKenzie & Bickle (1988), but their ‘subsolidus mantle temperature variation’ can be conceptually referred to as temperature variation at the mantle solidus (TSolidus), so that both can be compared by the linear relationship TP = TSolidus – PSolidus(kbar) x 1·8°C/kbar, where 1·8°C/kbar is the assumed adiabat (McKenzie & Bickle, 1988). TSolidus (or To) and PSolidus (or Po) are explicitly the temperature and pressure at which the ascending mantle intersects the solidus.
The problem of the Fe8 correction of Klein & Langmuir (1987)
The hidden problem in the concept of fractionation correction of FeO to Fe8 is revealed by calculating Mg# = Mg8/(Mg8 + Fe8), where Mg8 is the cation proportion of Mg at MgO = 8 wt%, and Fe8 is the cation proportion of Fe at at MgO = 8 wt%. Figure 6a demonstrates that Klein & Langmuir's (1987) Fe8 values for the global MORB dataset correlate inversely with Mg#. This correlation suggests that the Klein & Langmuir (1987) fractionation-corrected data still largely reflect varying degrees of fractional crystallization from less evolved melts (e.g. Mg# 0·68) to highly evolved melts (e.g. Mg# 0·55), which does not relate in any straightforward way to depth of mantle melting and mantle potential temperature variation.
|
vs Fe8 (a) and ridge axial depth (b) to demonstrate the error in the concept of the fractionation-correction of Klein & Langmuir (1987
) (note: 10% total Fe is assumed to be Fe3+ in all Mg# calculations) using the atomic proportion of Fe8 and the atomic proportion of 8 wt% MgO. The significance of Fe8 is thus conceptually exactly the same as the significance of Klein & Langmuir's (1987
) inverse Fe8–depth correlation in Fig. 2b (reproduced in Fig. 6c for comparison) is conceptually identical to the positive Mg#–depth correlation (Fig. 6b). That is, Mg# decreases from 0·68 at deep ridges to 0·55 at shallow ridges. These melts can only be in equilibrium with highly evolved Fe-rich olivine of Fo79·6–88·1 (Fig. 6a and b), not with mantle olivine of Fo >89·5 (e.g. Roeder & Emslie, 1970). Therefore, by using Fe8 (total range 7 to 11) one examines the progressively more evolved melt, mostly at the crustal level, from deep ridges to shallow ridges, not the pressures of mantle melting.
Another hidden problem in the concept of the Fe8 correction is that using a constant MgO value (8 wt%) results in the loss of the intrinsically positive FeO–MgO correlation obvious in model primary melts (Fig. 1) (Niu & Batiza, 1991a; Niu, 1997; Herzberg & O’Hara, 1998, 2002) and melts of peridotite melting experiments (e.g. Jaques & Green, 1980; Falloon & Green, 1987, 1988; Falloon et al., 1988; Hirose & Kushiro, 1993; Baker & Stolper, 1994). It is inevitable that erupted MORB melts must inherit some of their major element compositional variability from their parental melts in the mantle, but the variable and elevated FeO values of 7 to 11 wt% in melts with Mg# = 0·55–0·68 cannot be interpreted as indicating pressures of melting and mantle temperature variation of 250 K.
Lecroart et al. (1997) attempted to correct for the effects of low-pressure fractionation using FeO/MgO ratios. They found that, in major element oxide ratio–ratio spaces using MgO as the denominator, many MORB suites converge to a common point with 8% MgO, and thus interpreted this convergence as supporting the premises of the Klein & Langmuir (1987) correction scheme. However, we show that Mg# varies significantly at MgO = 8 wt% (Fig. 6).
Correction for fractionation effects to Mg# = 0·72
Niu et al. (1999, 2002) corrected MORB compositions for fractionation effects to a constant Mg# value of 0·72, instead of MgO = 8 wt%. By correcting to Mg# = 0·72, we still do not recover the primary magmas whose compositions are not precisely known (see Fig. 1). By assuming that mantle melts passing through the Moho must be in equilibrium with olivine of >Fo89, then correction of MORB melts for fractionation to Mg# = 0·72 is justified because a melt with Mg# = 0·72 is in equilibrium with olivine of Fo89·6 (following Roeder & Emslie, 1970). Furthermore, the most primitive MORB melts sampled have Mg# 0·72 (not higher), which allows reliable fractionation correction up to this Mg# value (see below). C. H. Langmuir (personal communication, 2006) stated ‘it does not matter whether the data are corrected to a constant MgO or Mg#, the meaning is the same and the implication of the corrected results is the same’. It is explicit in both practice and concept that it does matter. Melts corrected for fractionation effect to Mg# = 0·72 are in equilibrium with mantle olivine of 
Fo89·6, thus reflecting more faithfully mantle signals or processes. In contrast, melts corrected to MgO = 8 wt% with Mg# = 0·55 0·68 are in equilibrium with more Fe-rich, non-mantle, olivine compositions of Fo79·6–88·1 and mainly record cooling-induced fractional crystallization at crustal levels (Fig. 6).
Our correction procedure to minimize the effects of low-pressure differentiation has been discussed previously (Niu et al., 1999, 2002). This is further described in Electronic Appendix A and has been applied to each of the 9130 samples used, and to every major and minor element oxide. As each oxide is corrected independently (other than Mg#), the sum of Si72, Ti72, Al72, Fe72, Mn72, Mg72, Ca72, Na72, K72 and P72, which is 100% within ‘analytical error’, suggests the effectiveness of the correction procedure.
Chemistry–depth correlations: Fe8 and Na8 are equivalent to fractionation-uncorrected raw data
Figure 7 compares our averages of fractionation-uncorrected data (Table 1 and Fig. 4) with Klein & Langmuir's (1987) fractionation-corrected data (i.e. Fe8, Na8 and Mg# calculated from Fe8 and MgO = 8%). The similarity is remarkable. There is no simpler way than Fig. 7 to demonstrate that Klein & Langmuir's (1987) apparently sophisticated fractionation correction procedure does not really correct for fractionation effects, but reproduces the raw data for samples with MgO >7 wt%. Thus, Fe8, which is effectively fractionation-uncorrected FeO, cannot be used to infer pressures and temperatures of mantle melting based on corrections such as in Figs 2 and 6.
|
Chemistry–depth correlations: averages of fractionation-corrected data
Table 2 gives depth-interval averaged data corrected for fractionation effects to Mg# = 0·72. The recalculated Mg# (from Fe72 and Mg72) is constant at 0·713 ± 0·002 (vs varying from 0·55 to
0·68 calculated from Klein & Langmuir's (1987) Fe8 and 8 wt% MgO), an internal test that proves that our fractionation correction procedure is effective (Electronic Appendix A and Fig. 8).
|
|
Figures 5b and 9 show that the systematics of Ti72, Al72, Fe72, Ca72, Na72, K72 and Ca72/Al72 with ridge axial depth seen in the fractionation-uncorrected raw data (Figs 4 and 5a) remain, although the abundance levels differ and the degrees of depth dependence vary (see Electronic Appendix B for graphic comparisons). Hence, erupted MORB melts inherit compositional systematics from their parental melts in the mantle. Figure 9a shows parallel Fe72–depth and Mg72–depth trends. That is, there is a clear positive Fe72–Mg72 correlation (Fe72 = 1·1256Mg72 – 3·331, R = 0·989) (derived from data in Table 2) that exists in model primary melts (Fig. 1) and melts in peridotite melting experiments, but is never seen in actual MORB melts, which are far too evolved.
|
Chemistry–depth correlations: more on Fe72 vs Fe8 and Na72 vs Na8
Given that the original Klein & Langmuir (1987
) concept of Fe8 and Na8 has become deeply rooted in the literature, it is necessary to discuss more about Fe8. Figure 10 compares our Fe72 (Fig. 10a) with Fe8 by Klein & Langmuir (1987) (Fig. 10b) plotted against ridge axial depth. Both parameters show a progressive increase with decreasing ridge axial depth, but the range of Fe8 (5· 0 Fe units) is five times greater than the range of Fe72 (1· 0 Fe unit). If the range of Fe8 corresponds to mantle temperature variation of 250 K beneath global ocean ridges as proposed by Klein & Langmuir (1987), then the small range in Fe72, from 7·5 to 8·5, would suggest only 50 K variation.
|
We can use Langmuir et al.'s (1992
) equation (29) (Po[solidus] = 6·11Fe8–34·5; To[solidus] = 1150 + 13Po) to calculate Po and To. For Klein & Langmuir's; (1987) Fe8 range of 6·5–11·5 (Fig. 10b), we obtain Po = 5·2–35·8 kbar and To = 1218–1615°C or TP = 1208–1551°C by using an adiabatic gradient of 1· 8°C/kbar (McKenzie & Bickle, 1988). It should be noted that a range of adiabat values (from 1· 0 to 1· 8°C/kbar) have been used by others (e.g. Langmuir et al., 1992; Turcotte & Phipps Morgan, 1992; Fang & Niu, 2003), but the exact value is not critical as we are interested in temperature variation ranges, although the TP value (converted from To) is greater for a smaller adiabat value. The Langmuir et al. (1992) data suggest that there is a huge mantle potential temperature variation of TP = 1551°C–1208°C = 342 K beneath global ocean ridges, which is actually greater than Klein & Langmuir's (1987) preferred TP = 250 K. More conservatively, if we use Fe8 = 7–11 (Fig. 10), we obtain Po = 8·3–32·7 kbar, To = 1258–1575°C (i.e. TP = 1243–1516°C) or TP = 274 K, which is still greater than, but similar to, Klein & Langmuir's (1987) preferred TP = 250 K. On the other hand, if we use Fe72 = 7·5–8·5 (Fig. 10a), we obtain Po = 11·3–17·4 kbar, To = 1297–1377°C (i.e. TP = 1277–1345°C) or TP = 68 K.
Two points should be noted from this exercise: (1) as we use samples from the deepest ridges such as the Cayman Rise (5500 m) and the shallowest ridges such as the Reykjanes Ridge (500 m), and if Fe72 indeed has pressure significance (see below), then the mantle potential temperature variation would be TP 70 K, rather than TP = 250 K (Klein & Langmuir, 1987; Langmuir et al., 1992) beneath global ocean ridges; (2) interestingly, the MORB mantle TP = 1277–1345°C based on Fe72 is similar to the value of 1280°C recommended by McKenzie & Bickle (1988) or the new estimate of 1315°C (D. McKenzie, personal communication, 2006) or 1350°C as suggested by Fang & Niu (2003).
Figure 11 shows Klein & Langmuir's (1987) inverse Na8–Fe8 trend, which they call a global trend (Klein & Langmuir, 1989) and is interpreted as indicating mantle temperature control on pressures (depths) and extents of melting. That is, a hotter mantle rises and intersects the solidus at a greater depth (high Fe8), has a taller melting column, and thus melts more (lower Na8) than a cooler mantle. Langmuir et al. (1992) successfully modeled this ‘global trend’ by accumulated fractional melting as a result of increasing TP (increasing Po) from beneath deep ridges to beneath shallow ridges (see Langmuir et al., 1992, fig. 53). Figure 11 compares Klein & Langmuir's (1987) Na8–Fe8 with Na72–Fe72, for which both 9130 data points and depth interval averages (Figs 3, 7 and 9) are plotted. Explicitly, by using Fe8, Klein & Langmuir (1987) overestimated the ranges of melting pressures towards higher values. It should be noted that a significant inverse Na72–Fe72 trend is well defined by the depth-interval averaged data (Fig. 11b). The Na8–Fe8 ‘global trend’ (Fig. 11a) was modeled by accumulated fractional melting by Langmuir et al. (1992); however, the Na72–Fe72 trend (Fig. 11b) cannot be modeled by the same process as this would require changes in all possible parameters (e.g. P, T, mode and style of melting, source mineralogies and relevant partition coefficients). The relevant question is if the small Fe72 variation (1· 0 Fe unit) defined by depth-interval averages on a global scale (Figs 9–11
) has any significance at all in terms of pressures of melting. If so, how much might the pressure difference be, and if it is pressure-independent, what could cause the small Fe72 variation, the inverse Na72–Fe72 correlation, and all the oxide–depth correlations prominent in both fractionation-uncorrected and -corrected data (Figs 4, 5, 7, 9 and 10).
|
The small Fe72 variation might have no pressure significance
Figures 1 and 12a show explicitly that SiO2 in primary melts is as sensitive as FeO to the pressure of peridotite melting; FeO increases whereas SiO2 decreases with increasing melting pressure in the range of 5–30 kbar. Hence, if fractionation-corrected MORB melts preserve the chemical signature of melting pressures, a significant inverse Si72–Fe72 trend should exist.
|
However, this is not the case as shown in Fig. 12b; neither individual data points of the 9130 samples nor averages define any significant trend. Figure 12c shows that the data span a large Si72 range at a given ridge axial depth, but do not vary in any systematic way as a function of ridge axial depth as illustrated by the averages. It is likely that MORB melts do preserve signals of melting pressures, and that the large Fe72 (Figs 10 and 11) and Si72 variations (Fig. 12) displayed by individual data points could partly result from varying depths of melting. However, without first-order Si72 decrease, but with only a small Fe72 increase, towards shallow ridges, it is unconvincing that mantle melting depth (Po, thus To) increases with decreasing ridge axial depth (Fig. 12c). This means that the small Fe72 variation (1· 0 Fe unit) with ridge axial depth (Fig. 10a) might have a different origin (see below).
|
GLOBAL RIDGE AXIAL DEPTH VARIATION—THE FUNDAMENTAL CONSTRAINT |
|---|
|
TOPABSTRACTINTRODUCTIONTHE GLOBAL MORB DATASET...CORRECTION FOR FRACTIONATION...GLOBAL RIDGE AXIAL DEPTH...DISCUSSIONSUMMARYSUPPLEMENTARY DATAREFERENCES |
|---|
Mantle potential temperature variation (
TP), plate spreading rate variation (RSR) and mantle source compositional variation (XMS) are the three major variables that determine the extent of mantle melting and MORB compositions (e.g. Niu et al., 2001; Niu, 2004). Of these, RSR can be directly observed, but both TP and XMS must be inferred. The global ocean ridge axial depth variation (DRA) is not a governing variable, but a physical consequence of mantle processes. The lack of apparent correlation between RSR and DRA (Niu & Hékinian, 1997a; Sinton & Rubin, 2007) suggests that DRA is independent of RSR. Therefore, DRA = f(TP, XMS) should be true. This reasoning is logical because DRA must be caused by sub-ridge material density variation, and the latter must result from either TP or XMS or perhaps both. Our task is thus to evaluate the possible contributions of TP and XMS to DRA by carrying out some simple tests, recognizing the significant correlations of DRA with global MORB compositional variations (XMORB; Figs 4, 5, 7, 9 and 10). It is noteworthy that the DRA in question (i.e. Figs 4, 5, 7, 9 and 10) contains no contribution from ‘dynamic topography’ because the latter (1) is of local or regional origin, (2) should be removed by the averaging with respect to sample depth regardless of geographical location (Fig. 3), and (3) should not correlate with XMORB on a global scale.
Significance of mantle potential temperature variation
The Klein & Langmuir (1987) and Langmuir et al. (1992) interpretation and current views
We have demonstrated above that it is invalid to use Fe8 to infer melting pressures or initial depths of melting (Figs 6, 7, 8, 10 and 11). Hence, the Klein & Langmuir (1987) [and Langmuir et al. (1992)] interpretation of a global mantle potential temperature variation of TP = 250 K beneath mid-ocean ridges needs revising. The association of low Na8 with shallower ridge depths could be consistent with a higher extent of melting (Fig. 7a), and thus higher TP. We must emphasize, however, that Fe8 correlates with Na8 and ridge axial depth (Figs 2 and 7), and this is in fact the essence of the Klein & Langmuir (1987) and Langmuir et al. (1992) theory [see Klein & Langmuir, 1987, Fig. 12; Langmuir et al., 1992, fig. 53 and equation (29)].
The thickness of melt or magmatic crust (vs crust interpreted from seismic data) produced is predicted to vary with TP (Klein & Langmuir, 1987; McKenzie & Bickle, 1988; Langmuir et al., 1992) provided that the mantle source composition is uniform, that the effect of spreading rate variation is small, and that melt extraction from the mantle is complete (Niu, 1997; Niu et al., 2001). However, determination of crustal thickness is not a straightforward observation, but an interpretation based on seismic velocity data (White et al., 1992, 2001). In addition, whether the interpreted thickness of the seismic crust represents magmatic crust or not remains debatable (Hess, 1962; Christensen, 1972; Forsyth, 1992; Detrick et al., 1994; Cannat, 1996; Niu, 1997; Niu & Hékinian, 1997a, 1997b; White et al., 2001). We consider that the thickness of seismic crust could be used as a reference, but prefer to use straightforward observations such as ridge axial depth variation as a constraint in evaluating models of mantle processes.
McKenzie & Bickle (1988) originally argued for TP = 1280°C for normal sub-ridge mantle; TP = 1315°C is now considered more reasonable (D. McKenzie, personal communication, 2006). ‘Because the thickness of the oceanic crust varies very little, most of the upper mantle apart from hot rising jets (plumes) must have approximately the same temperature’ (McKenzie & Bickle, 1988, p. 641), implying that TP 
0 (or negligible) beneath global ocean ridges. Shen & Forsyth (1995) argued for TP 
60 K and that the TP = 250 K of Klein & Langmuir (1987) and Langmuir et al. (1992) was an artifact from mantle heterogeneity. Presnall et al. (2002) argued for TP 20 K (TP = 1240 ± 10°C) and that the TP = 250 K of Klein & Langmuir (1987) and Langmuir et al. (1992) resulted from the effects of source heterogeneity, such as CO2 content that affects FeO in the melt and thus Fe8. Herzberg et al. (2007) argued for TP = 1400 – 1280°C = 120 K based on petrological modeling. If we use Fe72 from equation (29) of Langmuir et al. (1992), we would obtain TP <70 K (68 K; see above). The range of estimated TP = 0 K, 20 K, 60 K, 70 K, 120 K, 250 K is clearly rather large. All the above researchers have their own reasons why they prefer a particular value or values, but these are not always clear. McKenzie & Bickle (1988) argued for uniform TP (or TP 0) because of (1) ‘the absence of any pressure effect’ (p. 642) in MORB melts, and (2) the uniform thickness of seismic crust beneath normal ocean ridges (excluding Iceland). Klein & Langmuir (1987) and Langmuir et al. (1992) insisted on TP = 250 K because of (1) the large variation in the thickness of seismic crust, including both Iceland and the Cayman Rise, and (2) the need to explain the large range of Fe8.
One may argue that Klein & Langmuir (1987) and Langmuir et al. (1992) are consistent with other models for ‘normal ocean ridges’ with TP < 100 K if Iceland is excluded. However, this is not the case. Even if samples from shallow ridges (< 2000 m deep) are excluded, Fe8 still has a large range from 7 to 10·5 (see Fig. 6), which would mean TP = 240 K (i.e. To = 1258–1482°C and TP = 1243–1482°C). So, TP = 250 K by Klein & Langmuir (1987) and Langmuir et al. (1992) refers to normal ocean ridges unaffected by hotspots or mantle plumes (e.g. Iceland) on a global scale. Adopting the same approach using Fe72 and samples from ridges >2000 m deep gives Fe72 = 7·5–8·1, and TP = 41 K.
All these simple exercises tell us that the large TP = 250 K variation proposed by Klein & Langmuir (1987) and Langmuir et al. (1992) cannot be because the dataset includes hotspots (e.g. Iceland) and hotspot-influenced ridges (e.g. Reykjanes Ridge), but because of using Fe8, which led them to reach erroneous conclusions (Figs 6, 7, 8, 10 and 11). We still do not know precisely what TP is beneath the global ocean ridge system, but (1) if we take into account some robust aspects of the arguments by other workers (e.g. McKenzie & Bickle, 1988; Shen & Forsyth, 1995; Presnall et al., 2002; Herzberg et al., 2007), and (2) if we assume TP = 68 K (based on Fe72; see above) using a dataset that considers all ridges including the Reykjanes Ridge, then we can say with confidence that TP is likely to be <<100 K. As this latter statement assumes that Fe72 has pressure significance that is actually unsupported by the data (i.e. lack of inverse Si72–Fe72 correlation and lack of Si72 systematics with ridge axial depth; Fig. 12), there is no convincing evidence to argue for TP >50 K beneath global ocean ridges.
The rapidity of thermal diffusion also does not favor the development of large lateral thermal gradients in the shallow mantle that could persist in time and space. More importantly, the seismic low-velocity zone (LVZ), which corresponds to a low-viscosity zone (lvz) at about 100 to 250 km depth in the mantle continuous beneath global ocean ridges (Phipps Morgan et al., 1995; Ekström & Dziewonski, 1998), is arguably the most dynamic region in the upper mantle in terms of vigor in convection and scales of lateral flow (Phipps Morgan et al., 1995; Niu et al., 1999; Niu & Hékinian, 2004; Niu, 2005). Hence, the LVZ processes would diminish, not create or enhance, temperature variation, if any, beneath ocean ridges.
We foresee that a more precise and definitive estimate of TP beneath ocean ridges will be possible when we seriously consider mantle compositional variations on local, regional and global scales (e.g. Natland, 1989; Albarède, 1992; Langmuir et al., 1992; Niu & Batiza, 1994, 1997; Niu et al., 1996, 1999, 2001, 2002; Niu, 1997, 2004; Castillo et al., 1998, 2000). We emphasize the effect of major element compositional variation, not minor incompatible elements such as K2O (Shen & Forsyth, 1995) or CO2 (Presnall et al., 2002) because it is not obvious how K2O and CO2 variations could explain the 5000 m ocean ridge axial depth variation. Comparison of MORB melts with hotspot melts can help to evaluate MORB mantle TP and TP (e.g. McKenzie & Bickle, 1988; Herzberg & O’Hara, 2002). However, Green et al. (2001) and Green & Falloon (2005) have argued that TP = 1430°C for both MORB mantle and hotspot mantle such as that beneath Hawaii. These workers emphasize that it is not TP variation, but mantle compositional heterogeneity as a byproduct of plate tectonics that is responsible for the diverse range of intraplate hotspot and MORB magmas.
So, what is the actual contribution of TP to the range of axial depths of the global ridge system (DRA = 5 km)? We do not know, but it is likely to be very small because we have shown above that TP is likely to be 50 K beneath ocean ridges. Additionally, there are some fairly robust arguments that can be made on the basis of some straightforward physics.
Contribution of TP to DRA: a simple isostasy test
In this section, we examine whether any possible TP can explain the axial ridge depth difference DRA = 5 km between two end-member ridges (i.e. deep ridge, DWater 5500 m at the Cayman Rise; shallow ridge, DWater 500 m at the Reykjanes Ridge), which encompass the 5 km ridge axial depth variation on a global scale (Figs 3–10). Of the two isostasy hypotheses (i.e. ‘Airy isostasy’ and ‘Pratt isostasy’), ‘Pratt isostasy’ is relevant to ocean ridges because there is no ‘mountain root’ involved. As the effect of dynamic topography, by the very nature of our averaging process, is removed in the chemistry–depth relationships (Figs 4, 5, 7, 9 and 10), the straightforward physics is that ridge axial depth variation must be determined by sub-ridge material density variation whether the latter is caused by temperature variation or compositional variation or both.
Figure 13 shows that whether the ridge is shallow or deep, the mass between the two columns above the common compensation depth (DC) must be the same as described by the equation
|
|
|
|
DRA = 5 km variation between the two end-member ridges, we need to consider the sub-ridge density contributions of the crust, the residual mantle (after MORB melt extraction) and the fertile mantle as described in the caption to Fig. 13 and detailed in Electronic Appendix C. Consideration of (1) crustal thickness variation (e.g. from 3 km beneath deep ridges such as the Cayman Rise to 9 km beneath shallow ridges such as the Reykjanes Ridge) and (2) residual mantle density variation as a result of melt extraction (less depleted and less buoyant residues beneath deep ridges and more depleted and more buoyant residues beneath shallow ridges) (Fig. 13) is inadequate to explain the DRA = 5 km. We thus evaluate the additional effect of reducing the density 
S by increasing mantle temperature beneath the shallow ridge relative to the deep ridge. Figure 14a shows density reduction (%) with increasing temperature by using a volume thermal expansion coefficient of 
= 3 x 10–5/K. The solid curves in Fig. 14b and c, which already consider the effect of crustal and residual mantle density variation (see Electronic Appendix C for details), show that to explain the DRA = 5 km, how DC changes in response to density reduction (%) beneath the shallow ridge relative to beneath the deep ridge as a result of temperature variation.
|
|
We consider two scenarios following the conclusions of Klein & Langmuir (1987
), assuming: (1) TP = 250 K and (2) DC = 150–200 km.
Scenario 1. For TP = 250 K (labeled [1] in Fig. 14a and c), which has an effect in density reduction of 0·75% (labeled [2] and [3] in Fig. 14a and b), DC = 380 km (labeled [4] in Fig. 14b and c). That is, the 380 km long mantle column beneath the shallow ridge must be 250 K hotter than the 375 km long (= 380 km – 5 km water column) mantle column beneath the deep ridge. This temperature variation requires continuous (vs discrete) TP increase from beneath deep ridges to beneath shallow ridges by 250 K in the upper 380 km of the mantle to explain the systematically correlated variations of MORB chemistry with ridge axial depth on a global scale (Figs 4, 5, 7, 9 and 10). It is physically not straightforward how TP = 250 K could persist in time and space in the upper 380 km of the mantle beneath global ocean ridges, which is dynamically the most vigorous portion of the mantle.
Scenario 2. For DC = 200 km (or 150 km) (labeled [5] and [6] in Fig. 14), it requires a density reduction of 1·78% (or 2·88% if DC = 150 km). This requires the 200 km long (or 150 km if DC = 150 km) mantle column beneath the shallow ridge be 594 K (or 961 K if DC = 150 km) hotter than the 195 km (or 145 km if DC = 150 km) mantle column beneath the deep ridge. This is physically improbable. Such temperature difference (i.e. 594 K or 961 K) is equivalent to or greater than the negative thermal buoyancy required for oceanic lithosphere subduction.
The above two scenarios are different expressions of the same theory of Klein & Langmuir (1987) and Langmuir et al. (1992). If either scenario fails, both scenarios fail. We thus cannot avoid the conclusion that TP = 250 K is inconsistent with DC = 150–200 km. Consideration of sub-ridge density variation as a result of (1) crustal thickness variation and residual mantle density variation as well as TP = 250 K cannot explain the DRA = 5 km unless DC >380 km (vs 150–200 km) (Fig. 14). More realistically, if we consider TP < 100 K (i.e. < 0·30% density reduction), then DC must be >847 km, which makes it more difficult to explain the observations. The situation would be even more difficult if TP < 50 K (i.e. < 0·15% density reduction), because a DC of >1625 km would be required. The straightforward conclusion is that mantle potential temperature variation is incapable of explaining the DRA = 5 km because of the rather small volume thermal expansion coefficient of = 3 x 10–5/K.
The DRA = 5 km could be explained with DC = 150–200 km if the crust at the Reykjanes Ridge (not Iceland) were >26 km thick. However, a crustal thickness of 9 km (Klein & Langmuir, 1987; Langmuir et al., 1992), not 26 km, is inferred from the available seismic data. Furthermore, a systematic crustal thickness increase from deep ridges to shallow ridges [i.e. crustal thickness (km) = –4·6 x ridge depth (km) + 28·3] would be required to explain MORB chemistry (Figs 4, 5, 7, 9 and 10), which has not been observed.
Significance of mantle compositional variation
The fundamental concept
The significant correlations of ridge axial depth with fractionation-uncorrected (Figs 4 and 5a) and fractionation-corrected (Figs 5b and 9) MORB major element data are striking. These correlations suggest that both DRA and XMORB are consequences of a common process. We argue above that DRA = f(TP, XMS) should be true, and we have also demonstrated that TP is inadequate to explain DRA = 5 km on a global scale. It follows that fertile mantle compositional variation, XMS, is the best candidate to cause DRA = 5 km as well as MORB chemistry. As it is the major elements that determine the phase equilibria, mineralogy, lithology, and physical properties of the rock (trace elements and isotopes are often, but not always, decoupled from major elements; see Niu et al., 1996, 1999, 2001, 2002), we emphasize the importance of major element compositional variation. The global variation in mantle potential temperature, TP, has been inferred from ridge axial depth (assumed thermal buoyancy and thickness of the buoyant crust) and MORB chemistry based on parameters such as Fe8 and Na8 (Klein & Langmuir, 1987; Langmuir et al., 1992). In fact, XMS can be inferred from the same observations—ridge axial depth (buoyancy as a result of mantle compositional variation) and MORB composition, which is, to a large extent, inherited from its mantle source although magma generation processes tend to modify the source signals.
MORB chemistry allows us to infer both magma generation–differentiation processes and sources. The common perception of ‘processes vs sources’ (or vice versa) is somewhat misleading, tends to focus on one or the other, and, as a result, conceptually neglects the fundamental controls of mantle compositional variation on melt generation ‘processes’. It is generally accepted that TP determines the initial depth and extent of melting, melt production and melt composition, and is reluctantly accepted that RSR determines the depth of melting cessation, thus the extent of melting, melt production and composition, but the role of XMS as driving force for magma generation processes has been largely neglected. The correlations of MORB chemistry with ridge axial depth (Figs 4, 5, 7, 9 and 10) could be explained by TP and XMS, yet petrologists often choose to favor TP, assuming that XMS = 0 (i.e. there is a uniform mantle source composition for MORB).
Mantle compositional variation, XMS, can (1) cause density variation in the mantle, which in turn (2) controls mantle convection and (3) compensates surface topography, and controls (4) rate of mantle upwelling and decompression melting, (4) extent of melting, (5) melt compositions (both inherited and process-enhanced), and (6) paths of melt evolution. In fact, XMS is much more efficient than TP in causing mantle density variations (e.g. O’Hara, 1973; Niu & Batiza, 1991b, 1991c; Niu et al., 2003). Given the small volume thermal expansion coefficient of 3 x 10–5/K, a 1· 0% density difference as a result of compositional variation requires a T of 333 K to compensate. We can see from Fig. 14 that if TP <100 K, beneath global mid-ocean ridges, we are dealing with less than a 0·3% density difference, which requires DC = 847 km to account for the DRA = 5 km (phase changes are neglected for conceptual clarity). A 0·3% density variation in the upper 847 km of the mantle as a result of compositional variation is physically likely (see below), but it is physically difficult to maintain TP = 100 K in the upper 847 km of the mantle beneath global ocean ridges in time and space, a model that also requires TP to increase continuously from beneath deep ridges to beneath shallow ridges. Again, it is physically possible that an 0·6% density variation may exist beneath ocean ridges as a result of compositional variation above DC = 458 km, but it is physically difficult to maintain TP = 200 K in the upper 458 km of the mantle beneath global ocean ridges.
Physical consequences of MORB mantle compositional variation
We interpret the significant correlations of ridge axial depth with MORB chemistry on a global scale (Figs 4, 5, 7, 9 and 10) as the result of melting a progressively less depleted fertile mantle peridotite source from beneath shallow ridges to beneath deep ridges. Figure 15a and b is reproduced from Fig. 9 with only the data regression lines plotted to illustrate the concept. The increase in Al72, Na72 and Ti72 (also K72 in Fig. 5b) and decrease in Ca72 and Ca72/Al72 in MORB melts with increasing ridge axial depth is consistent with a progressively more enriched (or less depleted) peridotite source (i.e. high Al2O3, Na2O, TiO2 and K2O, and low CaO/Al2O3). Higher Al2O3 ensures a greater amount of modal garnet, and higher Na2O and Al2O3 and lower CaO ensure high jadeite/diopside ratios in the clinopyroxene of the source peridotite below deep ridges. The decrease of Fe72 and Mg72 in MORB melts with increasing ridge axial depth is also consistent with a more enriched (or less depleted) source peridotite with a lower modal proportion of olivine beneath deep ridges. All these observations are consistent with the fertile MORB mantle source peridotite being not only more enriched (or less depleted) but also denser beneath deep ridges, which explains why they are deep.
|
0 beneath global ocean ridges. If the mantle beneath shallow ridges (e.g. the Reykjanes Ridge) is indeed hotter (e.g. Figure 15c shows schematically that ridges are mostly passive features, and that the dense asthenospheric mantle beneath deep ridges should rise more slowly (or ‘reluctantly’) in response to plate separation. The reduced upwelling rate beneath the deep ridge (1) allows conductive cooling to penetrate to a greater depth, (2) makes the cold thermal boundary layer (CTBL) thicker, (3) causes melting cessation at a deeper level, (4) results in a shorter decompression melting column (Po – Pf), (5) leads to a lower extent of melting, and (6) to possibly thinner magmatic crust with less depleted (or more enriched) major element compositions at deep ridges than at shallow ridges. Such a physical consequence is conceptually the same as for slow-spreading ridges (see Niu & Hékinian, 1997a
, Fig. 4). If TP is the same (i.e. TP 0 K) beneath deep and shallow ridges, the solidus would be at a shallower depth beneath the shallow ridge because of the more refractory fertile source (the thick solid line). If TP is a few tens of degrees hotter beneath shallow ridges, then the solidus would be deeper, perhaps similar to the depth of the solidus beneath the deep ridge (see below and also Electronic Appendix D for detailed discussion).
Possible systematics of mantle compositional variation beneath global ocean ridges
Fertile mantle composition and its variability cannot be directly observed, but can be modeled (e.g. McDonough & Sun, 1995; see Niu, 1997, for a review), and can also be inferred from the compositions of mantle xenoliths from various tectonic settings, orogenic massif peridotites, abyssal peridotites, ophiolites, etc. Griffin et al. (1999) carried out a comprehensive survey of the composition of the sub-continental lithosphere mantle. Figure 16 shows their 52 average bulk-rock peridotite compositions on MgO variation diagrams. For comparison, the primitive mantle composition of McDonough & Sun (1995) and the model melting residues (0–25% melt extraction) of Niu (1997) are also plotted. As expected, the data are scattered because of the complex histories of these mantle rocks including depletion, refertilization (metasomatism), etc. Nevertheless, the data define SiO2–MgO, TiO2–MgO, Al2O3–MgO, FeOt–MgO, CaO–MgO and Na2O–MgO trends that are all roughly consistent with varying extents of melt depletion. Although most sampled peridotites are expected to be some sort of mantle melting residues, some of the samples are much less depleted and plot towards the primitive mantle composition. Given the known process of plate tectonic recycling (oceanic lithosphere subduction, subcontinental lithosphere ‘delamination’, etc.), it is reasonable to imagine that these data may reflect or even represent the range of upper mantle compositions beneath ocean ridges, at least the less depleted ones with MgO <41 wt%.
|
For our purpose, we derive a set of bulk-rock compositional data with varying MgO contents from 37 to 48·5% along the regression lines (the thick grey lines). These derived bulk-rock compositions are used to calculate the modal mineralogy of lower-pressure spinel peridotites and higher-pressure garnet peridotites.
Density of bulk-rocks and constituent minerals as a function of melt depletion under spinel peridotite stability conditions
Figure 17a shows the calculated modes of olivine, opx and cpx (see Niu, 1997, appendix C) as a function of bulk-rock MgO; these resemble the modal systematics of melting residues after varying extents of melt depletion (Niu, 1997). Figure 17b shows calculated mineral density as a function of melt depletion (i.e. increasing MgO) using DENSCAL (Niu & Batiza, 1991b, 1991c). As the isobaric volume thermal expansion coefficients and isothermal compressibilities are very similar for all the silicate minerals of interest, we calculated mineral densities at low pressure and room temperature conditions. We compare density variation as a function of compositional variation (depletion) at the same conditions. Olivine density decreases with increasing extents of melt extraction, which is proportional to bulk-rock MgO (Niu, 1997, Fig. 8), and reaches a density reduction (increasing Fo/Fa ratio) of 1·2% from bulk-rock MgO = 37 to 47· 5% (Fig. 17c). Opx density declines by 0·88%, and cpx density declines by 1· 6%. However, the bulk-rock density reduction is only 0· 66%. This is because olivine has the greatest density throughout (Fig. 17b), and olivine modes increase as the bulk-rock is progressively depleted (Fig. 17a). It is thus conceptually important that although melting will make the residue compositionally more depleted and physically more buoyant, the overall density reduction is small in the spinel peridotite stability field (vs that in the garnet peridotite stability field; see below). As mantle melting residues beneath ocean ridges are located mostly in the spinel peridotite stability field, the density variation as a result of varying extents of melt extraction is small, and thus has limited contribution to ridge axial depth variation (much less than considered in Figs 13 and 14).
|
Density of bulk-rock and constituent minerals as a function of melt depletion under garnet peridotite stability conditions
The density of garnet peridotites with the same bulk-rock compositions as those in Figs 16 and 17 should decline rapidly with increasing extent of depletion (increasing MgO) because Al2O3, and thus modal garnet, are depleted accordingly. The average bulk-rock compositions of garnet peridotites from representative xenolith suites allowed Griffin et al. (1999
) to establish the modal systematics of garnet peridotite as a function of bulk-rock Al2O3 content. Figure 18 shows these modal systematics, which can be used to calculate mineral modes in garnet peridotites. The results are given in Fig. 19a. With increasing bulk-rock MgO (or melt depletion) (see Niu, 1997, Fig. 8), both olivine and opx modes increase whereas garnet and cpx modes decrease, which is consistent with an incongruent melting relationship in the garnet peridotite stability field (O’Hara & Yoder, 1967; Herzberg, 1992): a Cpx + b Ol + c Gnt = 1· 0 Melt + d Opx.
|
|
Figure 19b compares the density differences between garnet peridotite and spinel peridotite of the same bulk-composition (at a given MgO) and how the densities change with progressive depletion (increasing MgO). Figure 19c compares the density reduction (%) of spinel peridotite and garnet peridotite with progressive depletion (increasing MgO). The greater density of garnet peridotite than spinel peridotite of the same bulk-rock composition (Fig. 19b) is known to be determined by the dense garnet, and the greater density reduction of garnet peridotite than spinel peridotite with depletion again results from the decreasing garnet mode because of decreasing Al2O3 in the bulk-rock. This suggests that melting and melt extraction in the garnet (vs spinel) peridotite stability field can create spatially small-scale, but large-amplitude buoyancy variation in the mantle.
It should be noted that the density reduction of garnet peridotites from bulk-rock MgO = 37 to 42% can be as much as 1·2%, which is huge, and is equivalent to the effect of a temperature variation of 400 K. Let us assume the fertile garnet peridotite is less depleted with 38% MgO in the deep ridge mantle and is more depleted with 41% MgO (possibly the most depleted fertile source) in the shallow ridge mantle. There would be 0·675% density difference (equivalent to an 225 K temperature variation). This would require DC = 415 km to account for DRA = 5 km (see Fig. 14b and c). If we assume 40 wt% MgO in the shallow ridge mantle, there would be a 0·46% density difference (153 K equivalent), which would require DC = 576 km to account for DRA = 5 km.
We emphasize the importance of garnet because the MORB source is largely in the garnet (vs spinel) peridotite stability field under subsolidus conditions, and because the compensation depth must be several hundreds of kilometers deep (see Figs 13 and 14) beneath ocean ridges where garnet is a stable phase. It is also conceptually important to note that density differences of 0·3–0·6% (equivalent to a temperature effect of 100–200 K) in the MORB mantle as a result of compositional variability (varying degrees of depletion; e.g. Fig. 16) is possible. This would require DC = 847–458 km to account for the DRA = 5 km. Because of the extremely slow rates of chemical diffusion, density differences caused by compositional differences can be ‘permanent’ (long-lived) and can persist in space and time in that depth range. However, it is physically unlikely to be able to maintain 100–200 K temperature difference in space and time beneath global ocean ridges to that depth range because of much faster thermal (vs chemical) diffusion, and effective thermal homogenization by large-scale lateral asthenospheric flow beneath the oceanic lithosphere (see above).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONTHE GLOBAL MORB DATASET...CORRECTION FOR FRACTIONATION...GLOBAL RIDGE AXIAL DEPTH...DISCUSSIONSUMMARYSUPPLEMENTARY DATAREFERENCES |
|---|
What controls the global ocean ridge axial depth variation?
We propose that plate tectonics dictates where ocean ridges begin and how they evolve, but that mantle density variation as a result of compositional (vs temperature) variation determines the axial depth of ridges. That is, mantle compositional variation results in mantle density variation that isostatically compensates the global ocean ridge axial relief of
5 km. Figures 15, 18 and 19 show explicitly that less depleted mantle peridotites with higher Al2O3 content favor the formation of a greater amount of garnet, making the bulk-rock denser beneath deep ridges than beneath shallow ridges. We elaborate the argument further below.
The ridge axial depth varies from 6000 m at the deep end of the spectrum (e.g. the Cayman Rise) to 200 m at the shallow end (e.g. the Reykjanes Ridge) on a global scale. On local and regional scales, the ridge axial depth varies from <400 m along fast-spreading ridges (e.g. the East Pacific Rise, EPR) to as much as 1500 m within some segments of slow-spreading ridges (e.g. the Mid-Atlantic Ridge, MAR). In general, ridge axial depth is a function of sub-ridge material density, which is in turn a function of temperature and mantle composition beneath the ridge. However, on local and regional scales, the ridge axial depth variation may also be caused by differential mantle upwelling rate or lithospheric flexure; that is, a component of ‘dynamic topography’ that is not in isostatic equilibrium. As our data averaging is done with respect to sample depths (i.e. 250 m depth intervals) regardless of geographical location (Fig. 5), the ‘dynamic topography’ effect has been averaged out. Thus, the global ridge axial depth variation (DRA) with its correlations with MORB chemistry (Tables 1 and 2; Figs 4, 5, 7, 9 and 10) is in isostatic equilibrium; that is, it reflects the effect of sub-ridge material density variation. We have demonstrated above that consideration of possible crustal thickness variation and residual mantle density variation, as well as TP, cannot explain the DRA = 5 km on a global scale. It follows that DRA = f(XMS) is largely true. One may argue that deep ridges such as the Cayman Rise, Southwest Indian Ridge and Arctic ridges are deep and spread slowly, so spreading rate may exert some control on ridge axial depth. This is not the case because (1) there is no apparent correlation between ridge axial depth and spreading rate on a global scale (Niu & Hekinian, 1997a; Rubin & Sinton, 2007), (2) shallow ridges such as the Reykjanes Ridge spread slowly (< 25 mm/year), and (3) deep ridges such as the Australian Antarctic Discordance (AAD; > 4 km) spread at an intermediate (75 mm/year; not slow) rate.
Why is the AAD deep?
The AAD along the Southeast Indian Ridge (SEIR) is a unique feature of the global ocean ridge system that needs mention. The ADD is a complex 400 km long segment of the SEIR with axial depth >4000 m, about 1000–1500 m deeper than the adjacent ‘normal ridges’. The AAD has been interpreted as a cold ridge segment (e.g. Forsyth et al., 1987; Ritzwoller et al., 2003) or even the site of downwelling where the Pacific and Indian asthenospheres meet because of ancient cold subducted oceanic lithosphere (crust + lithospheric mantle) that is sinking beneath the AAD (Gurnis et al., 1998). Whether the mantle beneath the AAD is cold or not is unknown, but we do know that the AAD is deep, and must be underlain by dense material. If the mantle is cold and the ‘cold anomaly’ is currently maintained in the uppermost <150 km of the mantle (Ritzwoller et al., 2003), then the ‘anomaly’ beneath the AAD would have to be unrealistically cold (theoretically >900 K colder). If, on the other hand, the ‘cold anomaly’ extends down to the 660 km discontinuity (Gurnis et al., 1998), then the mean temperature beneath the AAD would be 150 K colder than that beneath the Reykjanes Ridge (in the context of our model in Figs 13 and 14). Otherwise, a fertile mantle compositional anomaly would better explain the deep topographic anomaly. As the subducted lithosphere has finite dimension, and because the AAD is a local or regional anomaly, the cold-slab model with an 150 K thermal anomaly extending down to the 660 km discontinuity (vs <150 km) is physically plausible.
Because ‘the SEIR is the only ridge to override an old subduction zone’, and ‘this overriding is probably the reason why the AAD is a unique feature of the global system of ridges’ (Gurnis et al., 1998, p. 1503), the cold-slab model for the AAD, if correct, does not apply to deep ridges elsewhere because of the much longer wavelengths (thousands of kilometers) of ridge axial depth variation than the 400 km of the AAD.
Why is Iceland shallow?
Iceland is another unique feature of the global mid-ocean ridge system with its voluminous melt production and anomalous ridge axial depth above sea level. These features have resulted in the classification of Iceland as a ‘hotspot’. This ‘hotspot’ has been widely accepted as the surface manifestation of a hot and deeply rooted thermal mantle plume originating from great depths—perhaps from the core–mantle boundary or from the base of the upper mantle. The ‘enriched’ incompatible element and radiogenic isotope characteristics of Iceland rocks have led to the wide acceptance that mantle plumes are derived from enriched fertile mantle peridotite sources at depths significantly deeper than the more depleted MORB mantle source. However, the enrichment or fertility of the plume source with its high Fe/Mg (and high Al2O3) makes the plume source ‘negatively’ buoyant relative to MORB mantle, creating physical difficulties for the rise of the Iceland plume (O’Hara, 1973). An excess mantle potential temperature of >300 K relative to the ambient MORB mantle may explain the Iceland hotspot by (1) possible thermal buoyancy, (2) buoyant crust and (3) buoyant residual mantle as a result of a large extent of melting caused by the hot mantle source. However, Foulger et al. (2005) have contested the mantle plume model for Iceland, suggesting that it is inconsistent with a thermal mantle plume because (1) there is no evidence for a hot mantle source (probably less than a few tens of degrees Kelvin hotter than the MORB mantle), and (2) it is a shallow feature, certainly confined to the upper mantle. These researchers offered an alternative to explain the melting anomaly by invoking the presence of abundant eclogite (or mantle peridotite fertilized by ‘resorbed eclogite’) in the Iceland mantle source left behind by the subducted oceanic crust of the Caledonian suture. They suggested that such a source can produce several times more melt than normal mantle peridotite without the need for high temperatures.
The plate separation rate at Iceland is slow, 25 mm/year (full rate). Therefore, it is not possible to explain the large melt volumes of Iceland by a low extent of melting of a large amount of source material per unit time, as the residual mantle cannot be carried away fast enough [even recognizing the along-axis flow (Niu & Hékinian, 2004)]. Hence, a large extent of melting for a given parcel of fertile mantle is necessary. The presence of abundant eclogite with a lower solidus temperature than the ‘host’ peridotite would certainly help. However, there is a physical problem with this model. The density of the bulk-rock fertile peridotite mantle (assumed to be 3·35 g/cm3 for simplicity) will be increased by addition of the dense eclogite component (assumed to be 3·50 g/cm3). The fertile mantle density will increase (%) by 0·05 x eclogite (vol.%). That is, 3% eclogite addition increases the bulk-rock density by 0·15% (equivalent to decreasing the temperature by 45 K), and 6% eclogite addition increases the bulk-rock density by 0·3% (equivalent to decreasing the temperature by 90 K). If 10% eclogite is needed to produce more melt to make the thick (>20 km) Iceland crust (vs 6 km ‘normal ocean crust’), then the Iceland fertile mantle would be 0· 5% denser (or 150 K cooler) than the MORB source mantle. The implication of such a simple calculation is that to add the dense eclogitic material into the Iceland fertile mantle would make Iceland the deepest, not the shallowest ridge (see the discussion on the AAD above). If the mantle potential temperature beneath Iceland is the same as that of the normal ocean ridge mantle, then Iceland must be underlain by compositionally depleted (low Fe/Mg, Al2O3) and physically buoyant mantle material (like subcontinental lithospheric material?).
If the Reykjanes Ridge is indeed affected by the Iceland ‘plume’, then we can infer from Fig. 15 that it is physically plausible that the Iceland fertile mantle is more depleted in terms of major elements (also see possible depletion in K2O from Fig. 5). We thus have the Iceland paradox. Our model for the Reykjanes Ridge can help with the Iceland scenario, but cannot explain the overall Iceland phenomena. Perhaps the Iceland fertile mantle is rich in volatiles (i.e. wet melting of a depleted peridotite source), or maybe it is indeed hot (i.e. hot melting of depleted peridotite source). We do not have an answer to the Iceland paradox, but we propose the need for a more serious, comprehensive and unbiased examination of the Iceland problem.
What controls the global MORB compositional variation?
The systematic and correlated variations of MORB chemistry with ridge axial depth (Figs 4, 5, 7 and 9) suggest that they are both consequences of a common cause in the mantle. We maintain that this is the major element compositional variation in the mantle beneath global ocean ridges. The observed MORB compositional variation (Figs 9 and 15), after fractionation correction to Mg# = 0·72, results from the combination of two different, but genetically related, processes: (1) mantle source inheritance; (2) melting process enhancement.
The systematic variations of Al72, Na72, Ti72, K72, Ca72, Ca72/Al72, Fe72 and Mg72 in MORB melts with increasing ridge axial depth (Figs 5b, 9 and 15) are consistent with a progressively more enriched (or less depleted) peridotite source from beneath shallow ridges to beneath deep ridges. Figure 15c shows schematically that fertile mantle compositional (mineralogical and lithological) variation leads to progressively denser fertile mantle from beneath shallow ridges to beneath deep ridges. In response to plate separation, the denser mantle rises slowly, resulting in a reduced melting rate, deepened conductive geotherm, shortened melting column, and thus lowered extents of melting from beneath shallow ridges to beneath deep ridges. That is, deep ridge MORB melts have the signatures of lower extents of melting than shallow ridge MORB melts (Figs 5b, 9 and 15). For example, among all the major and minor elements, the small decrease of Fe72 and Mg72 from shallow ridges to deep ridges is also consistent with decreasing extent of melting (Fig. 1).
It may sound counterintuitive that the less depleted (or enriched) fertile mantle beneath deep ridges melts less than the more depleted and refractory mantle beneath shallow ridges. However, the explanation is straightforward if we consider the two fundamental factors that control melting: (1) adiabatic upwelling; (2) enthalpy of melting. Without adiabatic upwelling, there would be no decompression melting, and no melt would be produced regardless of how fertile the peridotite source was. The retarded upwelling rate beneath deep ridges (greater density) leads to retarded melt production by decompression, and hence a lower melting rate than beneath shallow ridges. Our model for the Reykjanes Ridge can help with the Furthermore, the restricted amplitude of upwelling (greater density)beneath deep ridges results in a short decompression melting interval (Po – Pf) (Fig. 15). Thus shallower ridges are characterized by overall higher extents of melting relative to the mantle beneath deeper ridges, producing more melt per unit time. The observation that the most depleted abyssal peridotites still have cpx as a residual phase (Dick & Natland, 1996; Niu & Hékinian, 1997a, 1997b) indicates that ocean ridge melting does not eliminate cpx. That is, partial melting from a few per cent to high extents before cpx elimination, which mostly occurs in the olivine–opx–cpx–spinel four-phase field, does not require excess heat as long as adiabatic upwelling occurs. This reasoning is corroborated by an experimental study of ‘second-stage melting’ (Duncan & Green, 1987), and by the refractory MORB melts from the Garret Transform formed as a result of intra-transform extension-induced decompression melting of previous melting residues beneath the EPR axis (Hékinian et al., 1995; Niu & Hékinian, 1997b; Wendt et al., 1999).
What controls the initial depth (Po) of melting?
It is conceptually correct that a hotter parcel of mantle that rises adiabatically will intersect the solidus and begin to melt deeper than a cooler parcel of mantle. On the other hand, as the solidus is a material property, its location in P–T space is composition dependent. As we argue that the fertile MORB source is less refractory beneath deep ridges than beneath shallow ridges, and if TP is similar beneath all ridges (i.e. TP 50 K), then the initial depth of melting would be slightly shallower (not deeper) beneath shallow ridges as indicated in Fig. 15c. If TP beneath shallow ridges (e.g. the Reykjanes Ridge) were indeed 40 K hotter than TP beneath deep ridges, this would place the onset of melting at a slightly deeper level indicated by the thick dashed line on the left of Fig. 15c. The existing data indicate that the onset of melting increases (1) by 0·89 kbar per 10 K TP increase (McKenzie & Bickle, 1988; Langmuir et al., 1992), (2) by 0·29 kbar per 1·0 unit Mg# decrease (e.g. from 91 to 90) in the peridotite source, and (3) by 0·76 kbar per 0·05 wt% total alkali (Na2O + K2O) increase (Herzberg et al., 2000; Hirschmann, 2000; Niu et al., 2001). By assuming that the shallow ridge mantle is 40 K hotter, and more depleted (2 Mg# units higher, 0·05 wt% total alkalis less) than the deep ridge mantle, then the ascending mantle beneath the shallow ridge would intersect the solidus 2 kbar (+ 3·5 – 0·76 – 0·58 = 2·16 kbar) deeper than the mantle beneath the deep ridge, which is well within the uncertainties. This suggests that the net effect of source composition and small TP contributions on the initial depth of melting is likely to be very small, and the depth of initial melting beneath the global ocean ridge system must be very similar. The extent of melting, which is proportional to Po – Pf, is thus largely controlled by the final depth of melting (Niu & Hékinian, 1997a).
What controls the final depth (Pf) of melting?
Klein & Langmuir (1987) and McKenzie & Bickle (1988) assumed that decompression melting continues up to the Moho beneath ocean ridges. Niu & Batiza (1991a), however, showed that mantle melting beneath ocean ridges stops at levels significantly deeper than the Moho because MORB chemistry is inconsistent with melting to the Moho and with basic magma crystallization immediately above the Moho. Shen & Forsyth (1995) emphasized the importance of conductive cooling in determining the final depth of melting, but did not specify what could cause varying depth of melting cessation. Detailed studies of abyssal peridotites and MORB chemistry suggest that the final depth of melting (Pf) increases with decreasing spreading rate (Niu, 1997, 2004; Niu & Hékinian, 1997a, 1997b; Niu et al., 1997). All these results, together with the S-wave seismic velocity data of Zhang & Tanimoto (1993), are consistent with melting cessation at depths >30 km beneath very slow-spreading ridges, but significantly less than 30 km beneath fast-spreading ridges (Niu, 1997; Niu & Hékinian, 1997a). The argument of spreading-rate control on the depth of melting cessation is based on the concept that ridges are mostly passive features and the rate of adiabatic upwelling beneath ocean ridges is proportional to the rate of plate separation. This concept applies here.
The final depth of melting (Pf) is the result of a balance between conductive cooling to the seafloor and adiabatic upwelling of hot mantle (Fig. 15c). As the temperature on the seafloor is fixed, 0–3°C, the boundary condition for the conductive cooling is fixed. Hence, Pf is controlled by the rate of asthenospheric upwelling, which is known to depend on spreading rate (e.g. Reid & Jackson, 1981; Niu, 1997; Niu & Hékinian, 1997a) and also on mantle density variation as we argue here. The fertile mantle with greater density beneath deep ridges retards the upwelling, and allows conductive cooling to penetrate to a great depth, forcing melting to stop at a deeper level than beneath shallow ridges (Fig. 15c).
Why are MORB melts more evolved at shallow ridges?
Figures 4, 6 and 7 show that MORB melts erupted at shallower ridges are progressively more evolved with Mg# decreasing from 0·68 at deep ridges to 0·55 at shallow ridges. The explanation is straightforward. As discussed above, the rate of decompression melting, the extent of melting and the overall melt production increase from beneath deep ridges to beneath shallow ridges. It follows that there is progressively more frequent melt supply to the crust from beneath deep ridges to beneath shallow ridges. As a result, a progressively longer-lived magma chamber is better developed beneath shallower ridges. The key is that the crust at shallower ridges will have a higher time-averaged mean temperature (warmer) than the crust at deeper ridges. Such a crustal-level thermal state allows melts supplied to shallower ridges to experience more advanced extents of differentiation or evolution (resulting in low Mg#). In contrast, infrequent melt supply at deep ridges leads to crustal cooling, and the melts would experience limited extents of differentiation before freezing (high Mg#), or may even freeze in the mantle before reaching the crust. The latter case [i.e. the lack of erupted lavas at some deep (e.g. the Arctic) and slow-spreading (e.g. the 15°N MAR) ridges] has led to the view of ‘amagmatic’ spreading in the recent literature (see Dick et al., 2003; Michael et al., 2003).
The above argument that the extent of MORB melt differentiation is a result of mantle melt production and melt supply frequency is consistent with similar arguments on MORB petrogenesis as a function of spreading rate (Sinton & Detrick, 1992; Niu & Hékinian, 1997a, 1997b; Rubin & Sinton, 2007).
Why do we invoke garnet-rich peridotites rather than eclogites or garnet pyroxenites beneath deep ridges?
Our inferred chemical and mineralogical compositions of the fertile MORB mantle beneath deep ridges are qualitatively consistent with the presence of progressively more abundant recycled eclogites or garnet pyroxenites, which have been argued to be important in the source regions of oceanic basalts (e.g. Hirschmann & Stolper, 1996; Pertermann & Hirschmann, 2003). However, the lack of pyroxenites (converted from eclogites and garnet pyroxenites) in mantle melting residues (abyssal peridotites) sampled from the seafloor (e.g. Dick & Fisher, 1984; Dick et al., 1984; Dick, 1989; Niu & Hékinian, 1997b) suggests that eclogites and garnet pyroxenites may not be abundant in MORB source regions (also see Niu et al., 2002; Niu & O’Hara, 2003). It is possible that pyroxenites may be melted out, which explains the absence of such lithologies in abyssal peridotites. However, pyroxenites are not observed in mantle samples collected from Arctic ridges, where the extent of melting diminishes (Michael et al., 2003). All these observations suggest that although eclogites and garnet pyroxenites must exist in the mantle, as inferred from massif peridotites and mantle xenoliths, they must be volumetrically insignificant (vs garnet-rich peridotites) to explain the observed global correlations of ridge axial depth and MORB chemistry. Recycled eclogites (+ lithospheric mantle) may indeed exist beneath the AAD (see above), but the AAD is a unique feature of the global system of ocean ridges.
Why is there a stronger garnet signature in MORB melts from deep ridges than from shallow ridges?
MORB major element compositions and melting relationships recorded in abyssal peridotites (Niu, 1997) are consistent with MORB being generated largely in the spinel peridotite stability field, but incipient melts formed in the garnet peridotite stability field are expected because of volatile-induced melting at greater depths. High Sm/Yb ratios in some MORB melts are interpreted as indicating the presence of garnet as a residual phase in the source region. 230Th excesses in MORB melts are also interpreted as reflecting the presence of garnet in the melting region (Beattie, 1993). This interpretation is further supported by Hf isotope systematics (Salters & Hart, 1989). Salters (1996) reported an interesting observation based on Hf–Nd isotopes that MORB melts from deep ridges (e.g. Cayman Rise and AAD) have stronger garnet signatures than MORB melts from shallow ridges (e.g. Kolbeinsey Ridge and Reykjanes Ridge). To be consistent with Fe8-based interpretations, Salters (1996) proposed that the onset of melting below shallow ridges starts deeper (high Fe8) than for deep ridges (low Fe8), and that shallow ridges are characterized by a columnar-shaped melting region with a narrow base in the garnet stability field, whereas melts from deep ridges sample a triangular-shaped melting regime with a wide base where garnet is a stable phase. Although this interpretation is interesting, it is not necessary to invoke peculiarly different geometries of the melting regions beneath shallow and deep ridges. A straightforward interpretation is that the intensity of the garnet signature in MORB melts is inversely related to the extent of dilution; it is diluted less in melts produced by low extents of melting beneath deep ridges and is diluted more in melts produced by high extents of melting beneath shallow ridges (Niu et al., 1999).
Are ocean ridges of shallow or deep origin?
Recognizing that ocean ridges are mostly passive features in response to plate separation, we accept that ocean ridges are of shallow origin. The ridge process-affected part of the mantle may be limited to the upper 150–200 km of the asthenosphere (e.g. Ekström & Dziewonski, 1998) although the actual affected depth may vary, depending on the viscosity structure with depth and plate separation rate. The plate separation rate is important because it is proportional to the ridge suction force that drives the material flow towards ridges to form the crust and much of the lithospheric mantle (Niu & Hékinian, 2004). If mantle viscosity increases exponentially with depth, then the material needs at the ridge must be supplied by lateral asthenospheric (low-velocity zone) flow towards ridges. This leads to the inevitable decoupling of the ridgeward asthenospheric flow and the superjacent plate movement away from the ridge (Niu et al., 1999; Niu & Hékinian, 2004). The roughly 150–200 km depth that may be affected by passive ridge melting processes is perhaps the very limit that Klein & Langmuir (1987) placed on the DC = 150–200 km, although this depth range is inconsistent with a TP = 250 K as proposed in their model.
We have demonstrated above that mantle potential temperature variation, if significant at all (probably TP 50 K) (plus the density contributions of varying crustal thickness and residual mantle; see Figs 13 and 14), is inadequate to explain the global ridge axial relief of DRA = 5 km, and fertile mantle compositional variation is required. The latter is predicted to be characterized by progressively more enriched (or less depleted) fertile peridotite with higher Al2O3 (and co-variations of other major element oxides seen in Fig. 16) and thus more dense peridotite from beneath shallow ridges to beneath deep ridges. This explains not only the ridge axial depth variation, but also its correlations with MORB chemistry on a global scale (Figs 4, 5, 7, 9 and 10).
Using the peridotite data compiled by Griffin et al. (1999) (Fig. 16) as a reference, it is clear that the density reduction (Fig. 19) from less depleted peridotite (say MgO = 38 wt%) to the most depleted (yet still fertile) peridotite must be in the range of 0·3–0·6%, with 0·6% being the maximum required to produce the observed MORB compositional variation. This requires a compensation depth of 847 km for 0·3% density variation and 458 km for 0·6% density reduction without considering the effects of phase transitions (i.e. the 410 km and 660 km discontinuities). Consideration of the effects of phase transitions shallows the compensation depth, but the latter remains likely to be within the 410–660 km transition zone, if not in the lower mantle. It follows that whereas ocean ridges form or originate as a consequence of plate tectonics, and are thus mostly passive features, the actual controls on the surface expression of ridges (first-order ridge axial depth variation on a global scale), rate of mantle upwelling, extent of melting, melt production, and melt composition are likely to be deep.
Therefore, ocean ridges are of shallow origin, but their working is largely controlled by deep processes as well as the effect of plate spreading rate variation at shallow levels. An intriguing question is if such small compositional variations (large enough to cause significant density differences; see both Fig. 15 and Fig. 19) have any effect on the behavior of the 410 km (if not the 660 km) discontinuity. If so, is seismology sensitive enough to detect this behavior along the global ocean ridge system as a function of ridge axial depth variation, recognizing the fact that the ridge axial depth variation correlates with MORB chemistry? We consider this to be a worthwhile topic for future study towards a better understanding of how the Earth works.
|
SUMMARY |
|---|
|
TOPABSTRACTINTRODUCTIONTHE GLOBAL MORB DATASET...CORRECTION FOR FRACTIONATION...GLOBAL RIDGE AXIAL DEPTH...DISCUSSIONSUMMARYSUPPLEMENTARY DATAREFERENCES |
|---|
- The petrological parameters Na8 and Fe8 of Klein & Langmuir (1987) have been widely used as indicators of the extents and pressures of mantle melting beneath ocean ridges, and have also been used to elucidate the petrogenesis of basalts from other tectonic settings. However, these parameters are unreliable.
- Fe8 is the essence of the Klein & Langmuir (1987) and Langmuir et al. (1992) theory, and is used to compute the mantle solidus depth (Po) and temperature (To). It is the values and range of Fe8 that allowed Klein & Langmuir (1987) to argue that mantle temperature variation of up to 250 K is required to explain the global ocean ridge systematics. This interpreted 250 K range actually applies to normal ocean ridges unaffected by ‘hotspots’. We found no convincing evidence that calculated values for Po, To, To, Po and TP on the basis of Fe8 have any significance.
- MORB melts, when corrected for fractionation effects to a constant Mg# = 0·72, reflect mostly signals of mantle processes because melts with Mg# = 0·72 are in equilibrium with mantle olivine of Fo89·6 (vs evolved olivine of Fo88·1–79·6 in equilibrium with melts of Fe8). The Fe72 values of the global depth-interval averages range from 7·5 to 8·5, which, using the model of Klein & Langmuir (1987) and Langmuir et al. (1992), would give TP = 41 K (vs 250 K based on Fe8) beneath global ocean ridges unaffected by ‘hotspots’. The lack of Fe72–Si72 and Si72–ridge depth correlations provides no evidence that MORB melts preserve pressure signatures as a function of ridge axial depth. There is thus no convincing evidence for TP >50 K beneath global ocean ridges unaffected by ‘hotspots’.
- The significance of ridge axial depth correlations with MORB chemistry is better revealed by averaging out possible effects of other processes such as spreading rate variation, local-scale mantle source heterogeneity, melting region geometry variation, and dynamic topography on regional and segment scales using actual sample depths regardless of geographical locations within each of the 22 ridge depth intervals of 250 m on a global scale.
- Significant correlations exist between ridge axial depth and fractionation-uncorrected raw data for major and minor element oxides except for SiO2, MnO and P2O5. The systematic variations of TiO2, Al2O3, CaO, Na2O, K2O and CaO/Al2O3 with ridge axial depth are largely inherited from the parental magmas. The systematic FeO increase, MgO decrease and the associated Mg# decrease towards shallow ridges are consistent with progressively advanced degrees of melt differentiation towards shallow ridges. High extents of melting and frequent melt supply from mantle to the crust at shallow ridges favor development of a warm crust and long-lived magma chambers (as in the case of the fast-spreading EPR), resulting in advanced degrees of differentiation (low Mg#), whereas low extents of melting and infrequent melt supply from mantle to the crust at deep ridges lead to a cold crust and prohibit the development of long-lived magma chambers, such that the melt may freeze in the mantle or experience limited degrees of differentiation before solidification in the crust.
- Significant correlations between ridge axial depth and fractionation-corrected MORB chemistry (Ti72, Al72, Fe72, Mg72, Ca72, Na72, K72 and Ca72/Al72) exist. The fractionation correction has recovered the expected positive FeO–MgO correlations in model primary melts and melts produced in peridotite melting experiments. The MORB compositional systematics at Mg# = 0·72 result from the combination of two different, but genetically related processes—mantle source inheritance and melting process enhancement.
- The correlations between MORB chemistry and ridge axial depth suggest that they are both consequences of a common mantle process. The subsolidus mantle major element compositional (mineralogical, lithological, etc.) variation, not mantle temperature variation, results in mantle density variation that isostatically compensates the global ridge axial relief of DRA = 5 km as well as correlations with MORB chemistry. That is, a progressively more enriched (or less depleted) fertile source (i.e. high Al2O3 and Na2O, and low CaO/Al2O3) beneath deep ridges ensures a greater amount of modal garnet (high Al2O3) and high jadeite/diopside ratios in clinopyroxene (high Na2O and Al2O3, and lower CaO), making a denser mantle, and thus deeper ridges. The small decrease in Fe72 and Mg72 in MORB melts with increasing ridge axial depth is consistent with the more enriched (or less depleted) source with lower modal olivine beneath deep ridges. Hence, MORB chemistry is a manifestation of subsolidus mantle composition.
- The extent of mantle melting is controlled by the initial (Po) and final (Pf) depths of melting, i.e. Po – Pf. Po is expected to increase with increasing TP and mantle fertility. The lack of convincing evidence for huge TP variation and the small effect of MORB mantle compositional variation on Po suggest limited Po variation beneath global ocean ridges. Hence, the extent of melting depends largely on Pf, which is controlled by the rate of asthenospheric upwelling, and increases with decreasing plate separation rate. We emphasize here that the rate and amplitude of the mantle upwelling also depends on mantle density variation. The fertile mantle with greater density beneath deep ridges retards the rate and restricts the amplitude of the upwelling, leads to reduced melting rate by decompression, gives ways to conductive cooling to a great depth, forces melting to stop at a great depth with a short melting interval (Po – Pf), and thus produces less melt, and probably thinner magmatic crust in comparison with the less dense fertile mantle beneath shallow ridges.
- The variability of fertile mantle composition is limited by whether the source rock is fertile enough to produce melts with expected compositions by reasonable extents of melting. If the fertile mantle beneath deep ridges, for example, has MgO = 38 wt%, then the density reduction for the more refractory peridotites must be in the range of 0·3–0·6% to produce the observed MORB compositional variation. It would require compensation depths of 847 km for 0·3% density variation and 458 km for 0·6% density variation. Therefore, whereas ocean ridges are of shallow origin, their working is largely controlled by deep processes as well as the effect of plate spreading rate variation at shallow levels. With phase changes considered, the compensation depths would be within the 410–660 km transition zone (if not deeper). Do such small compositional variations have observable effects on the behavior of the 410 km and 660 km seismic discontinuities?
- The anomalously deep AAD may be the prime candidate for a cold sub-ridge mantle. Elsewhere, it is subsolidus mantle compositional (vs temperature) variation that determines the mineralogy, lithology and density variations that control the 5 km global ocean ridge relief and its correlation with the first-order MORB compositional variation. Current mantle plume debate on Iceland exposes an ‘Iceland Paradox’, which awaits a resolution.
|
SUPPLEMENTARY DATA |
|---|
|
TOPABSTRACTINTRODUCTIONTHE GLOBAL MORB DATASET...CORRECTION FOR FRACTIONATION...GLOBAL RIDGE AXIAL DEPTH...DISCUSSIONSUMMARYSUPPLEMENTARY DATAREFERENCES |
|---|
Supplementary data for this paper are available at Journal of Petrology online.
|
ACKNOWLEDGEMENTS |
|---|
It is our great pleasure to contribute to the special volume in honor of David H. Green for his numerous significant scientific contributions to our field in understanding the petrogenesis of mafic and ultramafic rocks on Earth, in particular the petrogenesis of mid-ocean ridge basalts in the context of global tectonics and chemical geodynamics. This work is partially supported by the Chinese 111 Project (No. B07011). Y.N. is particularly grateful to Dave for his encouragement over the years. M.J.O. has had a longstanding engagement with fertile mantle density. The principal idea in this paper has been developed over the past 16 years and discussions with many colleagues have been beneficial: Wolfgang Bach, Rodey Batiza, Pat Castillo, Henry Dick, Tony Ewart, Trevor Falloon, Godfrey Fitton, Fred Frey, Dave Green, Roger Hékinian, Claude Herzberg, Al Hofmann, Jill Karsten, Charlie Langmuir, John Maclennan, Dan McKenzie, Jim Natland, Mike Norry, Marcel Regelous, Andy Saunders, Jean-Guy Schilling, John Sinton, Shen-su Sun, Fred Tilmann, Bob White, Peter Wyllie and Youxue Zhang. A pleasant debate with Charlie Langmuir in a Melbourne restaurant (during the 2006 Goldschmidt Conference) was useful in clarifying some issues. A challenging yet friendly discussion with Dan McKenzie at a Bullard laboratory seminar was extremely useful. Journal reviewers Pat Castillo, John Sinton and Francis Albarède are thanked for their very constructive comments on an early version of the manuscript. The great editorial effort by Marjorie Wilson has improved the paper significantly, for which, and for her exceptional patience, we are grateful.
*Corresponding author. Telephone: +44-19-1334-2311. Fax: +44-19-1334-2301. E-mail: Yaoling.Niu{at}durham.ac.uk
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONTHE GLOBAL MORB DATASET...CORRECTION FOR FRACTIONATION...GLOBAL RIDGE AXIAL DEPTH...DISCUSSIONSUMMARYSUPPLEMENTARY DATAREFERENCES |
|---|
Albarède F. How deep do common basaltic magmas form and differentiate? Journal of Geophysical Research (1992) 97:10997–11009.[CrossRef]
Baker MB, Stolper EM. Determining the composition of high-pressure mantle melts using diamond aggregates. Geochimica et Cosmochimica Acta (1994) 58:2811–2827.[CrossRef][Web of Science]
Batiza R, Niu Y. Petrology and magma chamber processes at the East Pacific Rise 9°30'N. Journal of Geophysical Research (1992) 97:6779–6797.
Beattie P. Uranium-thorium disequilibria and partitioning on melting of garnet peridotite. Nature (1993) 363:63–65.[CrossRef]
Brodholt JP, Batiza R. Global systematics of unaveraged mid-ocean ridge basalt compositions: Comments on ‘Global correlations of ocean ridge basalt chemistry with axial depth and crustal thickness by Klein, E. M. and Langmuir, C. H.’ Journal of Geophysical Research (1989) 94:4231–4239.
Bryan WB, Dick HJB. Contrasted abyssal basalt liquidus trends: evidence for mantle major element heterogeneity. Earth and Planetary Science Letters (1982) 58:15–26.[CrossRef][Web of Science]
Cannat M. How thick is the magmatic crust at slow-spreading oceanic ridges? Journal of Geophysical Research (1996) 101:2847–2857.[CrossRef]
Castillo PR, Natland JH, Niu Y, Lonsdale P. Sr, Nd, and Pb isotopic variation along the Pacific ridges from 53 to 56°S: Implications for mantle and crustal dynamic processes. Earth and Planetary Science Letters (1998) 154:109–125.[CrossRef][Web of Science]
Castillo PR, Klein E, Bender J, Langmuir C, Shirey S, Batiza R, White W. Petrology and Sr, Nd, and Pb isotope geochemistry of mid-ocean ridge basalts from 11°45'N to 15°N segments of the East Pacific Rise. Geochemistry, Geophysics, Geosystems (2000) 1. paper number 1999GC000024.
Christensen NI. Abundance of serpentinites in oceanic crust. Journal of Geology (1972) 80:709–719.[Web of Science]
Detrick R, Collins J, Stenphen R, Swift S. In situ evidence for the nature of the seismic layer 2/3 boundary in oceanic crust. Nature (1994) 370:288–290.[CrossRef]
Dick HJB. Abyssal peridotites, very slow spreading ridges and ocean ridge magmatism. In: Saunders, A. D. & Norry, M. J. (eds). Magmatism in the Ocean Basins. Geological Society, London, Special Publications (1989) 42:71–105.
Dick HJB, Fisher RL, Bryan WB. Mineralogical variability of the uppermost mantle along mid-ocean ridges. Earth and Planetary Science Letters (1984a) 69:88–106.[CrossRef][Web of Science]
Dick HJB, Fisher RL. Mineralogic studies of the residues of mantle melting: Abyssal and alpine-type peridotites. In: The Mantle and Crustal–Mantle Relationships—Mineralogical, Petrological, and Geodynamic Processes of the Third International Kimberlite Conference—Kornprobst J, ed. (1984b) Vol. 2, Amsterdam : Elsevier, pp. 295–308.
Dick HJB, Natland JH. Late-stage melt evolution and transport in the shallow mantle beneath The East Pacific Rise. In: Proceedings of the Ocean Drilling Program, Scientific Results, 147—Mével C, Gills KM, Allan JF, eds. (1996) College Station, TX: College Station. 103–134.
Dick HJB, Lin J, Schouten H. An ultraslow-spreading class of ocean ridge. Nature (2003) 426:405–412.[CrossRef]
Duncan RA, Green DH. The genesis of refractory melts in the formation of oceanic crust. Contributions to Mineralogy and Petrology (1987) 96:326–342.[CrossRef][Web of Science]
Ekström G, Dziewonski AM. The unique anisotropy of the Pacific upper mantle. Nature (1998) 394:168–172.[CrossRef]
Falloon TJ, Green DH. Anhydrous partial melting of MORB pyrolite and other peridotite compositions at 10 k bar: Implications for the origin of MORB glasses. Mineralogy and Petrology (1987) 37:181–219.[CrossRef][Web of Science]
Falloon TJ, Green DH. Anhydrous partial melting of peridotite from 8 to 35 kb and the petrogenesis of MORB. Journal of Petrology, Special Lithosphere Issue (1988) 379–414.
Falloon TJ, Green DH, Hatton CJ, Harris KL. Anhydrous partial melting of a fertile and depleted peridotite from 2 to 30 kb and application to basalt petrogenesis. Journal of Petrology (1988) 29:1257–1282.
Fang N, Niu Y. Late Paleozoic ultramafic lavas in Yunnan, Southwest China and their geodynamic significance. Journal of Petrology (2003) 44:141–158.
Forsyth DW. Geophysical constraints on mantle flow and melt generation beneath mid-ocean ridges. Mantle Flow and Melt Generation at Mid-ocean Ridges American Geophysical Union Monograph—Phipps Morgan J, Blackman DK, Sinton JM, eds. (1992) 71:1–66.
Forsyth DW, Ehrenbard RL, Chapin S. Anomalous upper mantle beneath the Australian–Antarctic Discordance. Earth and Planetary Science Letters (1987) 84:471–478.[CrossRef][Web of Science]
Foulger GR, Natland JH, Anderson DL. Genesis of the Iceland melt anomaly by plate tectonic processes. Plates, Plumes, and Paradigms. Geological Society of America, Special Papers—Foulger GR, Natland JH, Presnall DC, Anderson DL, eds. (2005) 388:595–625.[CrossRef]
Green DH. Compositions of basaltic magmas as indicators of conditions origin: Application to oceanic volcanism. Philosophical Transaction of the Royal Society of London (1971) A268:707–725.[CrossRef]
Green DH. Experimental studies on a modal upper mantle composition at high pressure under water-saturated and water-undersaturated conditions. Earth and Planetary Science Letters (1973) 19:37–53.[CrossRef][Web of Science]
Green DH, Falloon TJ. Pyrolite: A Ringwood concept and its current expression. In: The Earth's Mantle, Composition, Structure, and Evolution—Jackson I, ed. (1998) Cambridge: Cambridge University Press. 311–380.
Foulger GR, Natland JH, Presnall DC, Anderson DL. Primary magmas at mid-ocean ridges, ‘hotspots,’ and other intraplate settings: Constraints on mantle potential temperature. Plates, Plumes, and Paradigms. Geological Society of America, Special Papers (2005) 388:217–248.[CrossRef]
Green DH, Ringwood AE. Mineral assemblages in a model mantle composition. Journal of Geophysical Research (1963) 68:937–945.
Green DH, Ringwood AE. Fractionation of basalts at high pressures. Nature (1964) 201:1276–1279.[CrossRef]
Green DH, Ringwood AE. The genesis of basaltic magmas. Contributions to Mineralogy and Petrology (1967) 15:103–190.[CrossRef]
Green DH, Ringwood AE. Mineralogy or peridotite compositions under upper mantle conditions. Physics of the Earth and Planetary Interiors (1970) 3:359–371.[CrossRef]
Green DH, Hibberson WO, Jaques AL. Petrogenesis of mid-ocean ridge basalts. In: The Earth: Its Origin, Structure and Evolution—McElhinny MW, ed. (1979) London: Academic Press. 265–290.
Green DH, Falloon TJ, Eggins SM, Yaxley GM. Primary magmas and mantle temperatures. European Journal of Mineralogy (2001) 13:437–451.
Griffin WL, O' Reilley SY, Ryan CG. The composition and origin of subcontinental lithosphere. Geochemical Society Special Publication—Fei Y, Bertka CM, Mysen BO, eds. (1999) 6:13–46. Mantle Petrology: Field Observations and High-Pressure Experimentation: A Tribute to Francis R. (Joe) Boyd.
Gurnis M, Muller RD, Moresi L. Cretaceous vertical motion of Australia and the Australian–Antarctic Discordance. Science (1998) 279:1499–1504.
Hékinian R, Bideau D, Herbért R, Y. Niu. Magmatic processes at upper mantle–crustal boundary zone: Garrett transform (EPR South). Journal of Geophysical Research (1995) 100:10163–10185.[CrossRef]
Herzberg C. Depth and degree of melting of komatiites. Journal of Geophysical Research (1992) 97:4521–4540.
Herzberg C, O’Hara MJ. Phase equilibrium constraints on the origin of basalts, picrites, and komatiites. Earth-Science Reviews (1998) 44:39–79.
Herzberg C, O’Hara MJ. Plume-associated ultramafic magmas of Phanerozoic age. Journal of Petrology (2002) 43:1857–1883.
Herzberg C, Raterron P, Zhang J. New experimental observations on the anhydrous solidus for peridotite KLN-1. Geochemistry, Geophysics, Geosystems (2000) 1. paper number 2000GC00089.
Herzberg C, Asimow PD, Arndt N, Niu Y, Albarède F, Lesher CM, Fitton JF, Cheadle MJ, Saunders AD. Temperatures in ambient mantle and plumes: Constraints from basalts, picrites and komatiites. Geochemistry, Geophysics, Geosystems (2007) (in press).
Hess HH. History of the ocean basins. In. In: Petrologic Studies: A Volume to Honour A.F. Buddington (1962) Denver, CO: Geological Society of America. 599–620.
Hirose K, Kushiro I. Partial melting of dry peridotites at high pressures: Determination of compositions of melts segregated from peridotite using aggregates of diamonds. Earth and Planetary Science Letters (1993) 114:477–489.[CrossRef][Web of Science]
Hirschmann MM. Mantle solidus: Experimental constraints and the effects of peridotite composition. Geochemistry, Geophysics, Geosystems (2000) 1. paper number 2000GC000070.
Hirschmann MM, Stolper EM. A possible role for garnet pyroxenite in the origin of the ‘garnet signature’ in MORB. Contributions to Mineralogy and Petrology (1996) 124:185–208.[CrossRef][Web of Science]
Jaques AL, Green DH. Anhydrous melting of peridotite at 0–15 kb pressure and the genesis of tholeiitic basalts. Contributions to Mineralogy and Petrology (1980) 73:287–310.[CrossRef][Web of Science]
Kinzler RJ. Melting of mantle peridotite at pressures approaching the spinel to garnet transition: Application to mid-ocean ridge basalt petrogenesis. Journal of Geophysical Research (1997) 102:853–874.[CrossRef]
Kinzler RJ, Grove TL. Primary magmas of mid-ocean ridge basalts, 2, applications. Journal of Geophysical Research (1992) 97:6907–6926.
Klein EM, Langmuir CH. Global correlations of ocean ridge basalt chemistry with axial depth and crustal thickness. Journal of Geophysical Research (1987) 92:8089–8115.
Klein EM, Langmuir CH. Local versus global variation in ocean ridge basaltic composition: A reply. Journal of Geophysical Research (1989) 94:4241–4252.
Kushiro I. Partial melting of a fertile mantle peridotite at high pressures: An experimental study using aggregates of diamond. Geophysical Monograph, American Geophysical Union (1996) 95:109–122.
Langmuir CH. Geochemical consequences of in situ crystallization. Nature (1989) 340:199–205.[CrossRef]
Langmuir CH, Klein EM, Plank T. Petrological systematics of mid-ocean ridge basalts: Constraints on melt generation beneath ocean ridges. Mantle Flow and Melt Generation at Mid-ocean Ridges. American Geophysical Union Monograph—Phipps Morgan J, Blackman DK, Sinton JM, eds. (1992) 71:183–280.
Lecroart P, Albarède F, Cazenave A. Correlations of mid-ocean ridge basalt chemistry with the geoid. Earth and Planetary Science Letters (1997) 153:37–55.[CrossRef][Web of Science]
McDonough WF, Sun, S.-s. The composition of the Earth. Chemical Geology (1995) 120:223–253.[CrossRef][Web of Science]
McKenzie D, Bickle MJ. The volume and composition of melt generated by extension of the lithosphere. Journal of Petrology (1988) 29:625–679.
Michael PJ, Langmuir CH, Dick HJB, Snow JE, Goldstein SL, Graham DW, Lehnert K, Kurras G, Jokat W, Muhe R, Edmonds HN. Magmatic and amagmatic seafloor generation at the ultraslow-spreading Gakkel ridge, Arctic Ocean. Nature (2003) 423:956–961.[CrossRef]
Natland JH. Partial melting of a lithologically heterogeneous mantle: Inferences from crystallisation histories of magnesian abyssal tholeiites from the Siqueiros Fracture Zone. Magmatism in the Ocean Basins—Saunders AD, Norry MJ, eds. (1989) 42. Geological Society, London: Special Publications. 41–70.
Niu Y. Mantle melting and melt extraction processes beneath ocean ridges: Evidence from abyssal peridotites. Journal of Petrology (1997) 38:1047–1074.[CrossRef][Web of Science]
Niu Y. Bulk-rock major and trace element compositions of abyssal peridotites: Implications for mantle melting, melt extraction and post-melting processes beneath ocean ridges. Journal of Petrology (2004) 45:2423–2458.
Niu Y. Generation and evolution of basaltic magmas: Some basic concepts and a hypothesis for the origin of the Mesozoic–Cenozoic volcanism in eastern China. Geological Journal of China Universities (2005) 11:9–46.
Niu Y, Batiza R. An empirical method for calculating melt compositions produced beneath mid-ocean ridges: application for axis and off-axis (seamounts) melting. Journal of Geophysical Research (1991a) 96:21753–21777.[CrossRef]
Niu Y, Batiza R. In-situ densities of silicate melts and minerals as a function of temperature, pressure, and composition. Journal of Geology (1991b) 99:767–775.[Web of Science]
Niu Y, Batiza R. DENSCAL: A program for calculating the densities of silicate melts and minerals as a function of temperature, pressure, and composition in magma generation environment. Computers and Geosciences (1991c) 17:679–687.[CrossRef]
Niu Y, Batiza R. Chemical variation trends at fast and slow spreading ridges. Journal of Geophysical Research (1993) 98:7887–7902.
Niu Y, Batiza R. Magmatic processes at the Mid-Atlantic Ridge 26° S. Journal of Geophysical Research (1994) 99:19719–19740.[CrossRef]
Niu Y, Batiza R. Trace element evidence from seamounts for recycled oceanic crust in the eastern Pacific mantle. Earth and Planetary Science Letters (1997) 148:471–483.[CrossRef][Web of Science]
Niu Y, Hékinian R. Spreading rate dependence of the extent of mantle melting beneath ocean ridges. Nature (1997a) 385:326–329.[CrossRef]
Niu Y, Hékinian R. Basaltic liquids and harzburgitic residues in the Garrett transform: A case study at fast-spreading ridges. Earth and Planetary Science Letters (1997b) 146:243–258.[CrossRef][Web of Science]
Niu Y, Hékinian R. Ridge suction drives plume–ridge interactions. In: Oceanic Hotspots—Hékinian R, Stoffers P, Cheminée J.-L, eds. (2004) New York: Springer. 285–307.
Niu Y, O’Hara MJ. The origin of ocean island basalts (OIB): A new perspective from petrology, geochemistry and mineral physics considerations. Journal of Geophysical Research (2003) 108. doi:10.1029/2002JB002048.
Niu Y, Waggoner DG, Sinton JM, Mahoney JJ. Mantle source heterogeneity and melting processes beneath seafloor spreading centers: The East Pacific Rise, 18°–19°S. Journal of Geophysical Research (1996) 101:27711–27733.[CrossRef]
Niu Y, Langmuir CH, Kinzler RJ. The origin of abyssal peridotites: A new perspective. Earth and Planetary Science Letters (1997) 152:251–265.[CrossRef][Web of Science]
Niu Y, Collerson KD, Batiza R, Wendt JI, Regelous M. The origin of E-Type MORB at ridges far from mantle plumes: The East Pacific Rise at 11°20'N. Journal of Geophysical Research (1999) 104:7067–7087.[CrossRef]
Niu Y, Bideau D, Hékinian R, Batiza R. Mantle compositional control on the extent of melting, crust production, gravity anomaly and ridge morphology: a case study at the Mid-Atlantic Ridge 33–35°N. Earth and Planetary Science Letters (2001) 186:383–399.[CrossRef][Web of Science]
Niu Y, Regelous, M. Wendt JI, Batiza R, O’Hara MJ. Geochemistry of near-EPR seamounts: Importance of source vs process and the origin of enriched mantle component. Earth and Planetary Science Letters (2002) 199:329–348.
Niu Y, O’Hara MJ, Pearce JA. Initiation of subduction zones as a consequence of lateral compositional buoyancy contrast within the lithosphere: A petrologic perspective. Journal of Petrology (2003) 44:851–866.
O’Hara MJ. Primary magmas and the origin of basalts. Scottish Journal of Geology (1965) 1:19–40.
O’Hara MJ. Crystal–liquid equilibria and the origins of ultramafic nodules in igneous rocks. In: Ultramafic and Related Rocks—Wyllie PJ, ed. (1967) New York: John Wiley. 346–349.
O’Hara MJ. Are ocean floor basalts primary magmas? Nature (1968a) 220:683–686.[CrossRef]
O’Hara MJ. The bearing of phase equilibria studies in synthetic and natural systems on the observation of volcanic products. Physics of the Earth and Planetary Interiors (1968b) 3:236–245.[CrossRef]
O’Hara MJ. Upper mantle composition inferred from laboratory experiments and observation of volcanic products. Physics of the Earth and Planetary Interiors (1970) 3:236–245.[CrossRef]
O’Hara MJ. Non-primary magmas and dubious mantle plume beneath Iceland. Nature (1973) 243:507–508.[CrossRef]
O’Hara MJ. Geochemical evolution during fractional crystallization of a periodically refilled magma chamber. Nature (1977) 266:503–507.[CrossRef]
O’Hara MJ. Importance of the ‘shape’ of the melting regime during partial melting of the mantle. Nature (1985) 314:58–62.[CrossRef]
O’Hara MJ. Trace element geochemical effects of integrated melt extraction and ‘shaped’ melting regime. Journal of Petrology (1995) 36:1111–1132.
O’Hara MJ, Herzberg C. Interpretation of trace element and isotope features of basalts: Relevance of field relations, petrology, major element data, phase equilibria, and magma chamber modeling in basalt petrogenesis. Geochimica et Cosmochimica Acta (2002) 66:2167–2191.[CrossRef][Web of Science]
O’Hara MJ, Mathews RE. Geochemical evolution in an advancing, periodically replenished, periodically tapped, continuously fractionating magma chamber. Journal of the Geological Society, London (1981) 138:237–277.
O’Hara MJ, Yoder HS. Partial melting of the mantle. Carnegie Institution of Washington Yearbook (1963) 62:66–71.
O’Hara MJ, Yoder HS. Formation and fractionation of basic magmas at high pressures. Scottish Journal of Geology (1967) 3:67–117.
O’Reilly SY, Griffin WL. Mantle metasomatism beneath western Victoria, Australia: I, Metasomatic processes in Cr-diopside lherzolites. Geochimica et Cosmochimica Acta (1988) 52:433–447.[CrossRef][Web of Science]
Pertermann M, Hirschmann MM. Partial melting experiments on a MORB-like pyroxenite between 2 and 3 GPa: Constraints on the presence of pyroxenite in basalt source regions from solidus location and melting rate. Journal of Geophysical Research (2003) 108. doi:10.1029/2000JB000118.
Phipps Morgan J, Morgan WJ, Zhang Y-S, Smith WHF. Observational hints for a plume-fed, suboceanic asthenosphere and its role in mantle convection. Journal of Geophysical Research (1995) 100:12753–12767.[CrossRef]
Presnall DC, Gudfinnsson GH. Origin of the Oceanic Lithosphere. Journal of Petrology (2007) 49:613–630.
Presnall DC, Dixon JR, O’Donnell TH, Dixon SA. Generation of mid-ocean ridge tholeiites. Journal of Petrology (1979) 20:3–35.
Presnall DC, Gudfinnsson GH, Walter MJ. Generation of mid-ocean ridge basalts at pressures from 1 to 7 GPa. Geochimica et Cosmochimica Acta (2002) 66:2073–2090.[CrossRef][Web of Science]
Regelous M, Niu Y, Wendt JI, Batiza R, Greig A, Collerson KD. An 800 ka record of the geochemistry of magmatism on the East Pacific Rise at 10°30'N: Insights into magma chamber processes beneath a fast-spreading ocean ridge. Earth and Planetary Science Letters (1999) 168:45–63.[CrossRef][Web of Science]
Reid I, Jackson HR. Oceanic spreading rate and crustal thickness. Marine Geophysical Research (1981) 5:165–172.
Ritzwoller MH, Shapiro NM, Leahy GM. The Australian–Antarctic mantle anomaly as the principal cause of the Australian–Antarctic Discordance. Journal of Geophysical Research (2003) 108. doi:10.1029/2003JB002522.
Roeder PL, Emslie RF. Olivine–liquid equilibrium. Contributions to Mineralogy and Petrology (1970) 29:275–289.[CrossRef][Web of Science]
Rubin KH, Sinton JM. Inferences on mid-ocean ridge thermal and magmatic structure from MORB compositions. Earth and Planetary Science Letters (2007) 260:257–276.[CrossRef][Web of Science]
Salters VJM. The generation of mid-ocean ridge basalts from the Hf and Nd isotope perspective. Earth and Planetary Science Letters (1996) 141:109–123.[CrossRef][Web of Science]
Salters VJM, Hart SR. The hafnium paradox and the role of garnet in the source of mid-ocean ridge basalts. Nature (1989) 342:420–422.[CrossRef]
Shen Y, Forsyth DW. Geochemical constraints on initial and final depth of melting beneath mid-ocean ridges. Journal of Geophysical Research (1995) 100:2211–2237.[CrossRef]
Sinton JM, Detrick RS. Mid-ocean ridge magma chambers. Journal of Geophysical Research (1992) 97:197–216.
Stolper EA. phase diagram for mid-ocean ridge basalts: Preliminary results and implications for petrogenesis. Contributions to Mineralogy and Petrology (1980) 74:13–27.[CrossRef][Web of Science]
Saunders AD, Norry MJ. Chemical and isotopic systematics in ocean basalt: Implication for mantle composition and processes. Magmatism in the Ocean Basins (1989) 42. Geological Society, London: Special Publications. 313–345.
Turcotte DL, Phipps Morgan J. Magma migration and mantle flow beneath a mid-ocean ridge. In: Phipps Morgan, J. Blackman, D. K. & Sinton, J. M. (eds) Mantle Flows and Melt Generation at Mid-ocean Ridges. American Geophysical Union Monograph (1992) 71:155–182.
Walker D, Shibata T, DeLong SE. Abyssal tholeiites from the Oceanographer Fracture Zone, II, Phase equilibria and mixing. Contributions to Mineralogy and Petrology (1979) 70:111–125.[CrossRef][Web of Science]
Wendt JI, Regelous M, Niu Y, Hékinian R, Collerson KD. Geochemistry of lavas from the Garrett Transform Fault: insights into mantle heterogeneity beneath the eastern Pacific. Earth and Planetary Science Letters (1999) 173:271–284.[CrossRef][Web of Science]
White RS, McKenzie DP, O’Nions RK. Oceanic crustal thickness from seismic measurements and rare earth elements inversions. Journal of Geophysical Research (1992) 97:19683–19715.[CrossRef]
White RS, Minshull TA, Bickle MJ, Robinson CJ. Melt generation at very slow-spreading oceanic ridges: Constraints from geochemical and geophysical data. Journal of Petrology (2001) 42:1171–1196.
Wyllie PJ. Experimental petrology and global tectonics: a preview. Experimental Petrology and Global Tectonics. Tectonophysics (1973) 17:189–209.
Zhang Y-S, Tanimoto T. High resolution global upper mantle structure and plate tectonics. Journal of Geophysical Research (1993) 98:9793–9823.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  L. Beccaluva, G. Bianchini, C. Natali, and F. Siena Continental Flood Basalts and Mantle Plumes: a Case Study of the Northern Ethiopian Plateau J. Petrology, May 11, 2009; (2009) egp024v1. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||