Magmatic Processes at Superfast Spreading Mid-Ocean Ridges: Glass Compositional Variations Along the East Pacific Rise 13°–23°S

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 96, NO. 84, PAGES 6133-6155, APRIL 10, 1991

Magmatic Processes at Superfast Spreading Mid-Ocean Ridges' Glass Compositional Variations Along the East Pacific Rise 13ø-23øS

JOHN M. SINTON, SUZANNE M. SMAGLIK, AND JOHN J. MAHONEY

Department of Geology and Geophysics, University of Hawaii at Manoa, Honolulu

KEN C, MACDONALD

Department of Geological Sciences, University of California,. Santa Barbara

Major and minor element analyses of 496 natural volcanic glass samples from 141 locations along the superfast spreading (150 mm/yr) East Pacific Rise (EPR), 13ø-23øS, and near-ridge seamounts comprise 212 chemical groups. We interpret these groups to represent the average composition of individual lava flows or groups of dosely related flows. Groups slightly enriched in K20 (T-MORB) are distributed variably along the axis, in contrast to the Galapagos Spreading Center where T-MORB are extremely rare. This result is consistent with the interpretation that T-MORB magmas arise from low-melting temperature, K-rich heterogeneities in the subaxial EPR mantle. The Galapagos Spreading Center, which is migrating to the west in an absolute reference frame, is underlain by mantle previously processed and depleted in the T-MORB component during melting events giving rise to earlier EPR magmas. Excluding T-MORB, there are nearly monotonic, twofold increases in K/Ti and K/P of axial lavas from 23øS to 13øS. From 22øS to 17øS these gradients correlate with isotopic ratios, but north of 17øS there is a reversal of isotopic gradients, indicating (recent?) decoupling of the isotopic and minor element ratios in the subaxial mantle. A strong, southward increase in degree of differentiation for approximately 200 km north of the large offset at 20.7øS correlates with a gradient in bathymetry, consistent with previous interpretations that this offset is propagating to the south. Samples from recently abandoned ridges associated with this dueling propagator mainly carry the distinctive, evolved fractionation signatures of rift propagation, suggesting that propagating rift tips have been abandoned preferentially to failing rift tips. Glass compositional variations south of this offset are consistent with rift failure on the southern limb within 40 km of the offset, and possibly also south of 22øS; the latter region may be affected by deformation accompanying northward growth of the Easter Microplate. Near-ridge seamounts on the Pacific Plate between 18ø-19øS comprise two distinct populations: those aligned approximately parallel to the spreading direction are extremely variable in major element composition, but consistently enriched in Sr relative to nearby axial lavas; smaller seamounts aligned approximately parallel to the direction of absolute plate motion are uniformly depleted in minor elements and Sr relative to axial lavas. The degree of differentiation of axial lavas between 18ø-19øS can be related to the structural development of the rift axis and/or vigor of hydrothermal activity of individual segments. Glass compositional variations indicate that magmatic segmentation occurs on several different scales at the superfast spreading rate of this area. Primary magmatic segmentation mainly reflects mantle source variations, the boundaries of which correlate with the largest physical offsets in the rise axis between the Easter Microplate and Garrett Transform Zone. A secondary magmatic segmentation, defined by the along-axis continuity of similar parental magma compositions or liquid lines of descent, occurs with a length scale varying from 11 to 185 km, with an average of 69_+57 (lcr) km. The boundaries of these segments mainly occur at overlapping spreading centers. All first-, second- and third-order physical offsets correspond to secondary magmatic segment boundaries, but some secondary magmatic segment boundaries also occur at small, fourth-order ridge axis discontinuities. The secondary magmatic segments define the length scale of mantle melting variations, mainly variations in extent of melting, but not the scale of melt extraction processes that feed the axis. This scale must be smaller than that of the secondary magmatic segments and probably corresponds to the length scale of fourth-order physical discontinuities along axis. There is a good positive correlation of average secondary magmatic segment length with spreading rate for four well-sampled areas varying from 20 to 150 mm/yr. Secondary magmatic segments also become more variable in axial length with increasing spreading rate. The average lengths of secondary magmatic segments are smaller than those predicted by gravitational instability considerations at all spreading rates. Superposed on the axial magmatic segmentation are variations reflecting subaxial magmatic temperature, defined by extent of magmatic differentiation, which bears little systematic relation to physical or other kinds of magmatic segmentation. At 13ø-23øS, the length scale of this variation is 217+_60 (1o) km, approximately corresponding to the wavelength of "rolls" in the gravity field obsep/ed off-axis. Taken together, the various kinds and scales of magmatic variations obse•ed for this superfast spreading ridge suggest that regional temperature of the upwelling asthenosphere, magma supply to the axis, and crustal magmatic temperature reflect independent, regionally decoupled processes.

Copyright 1991 by the American Geophysical Union.

Paper number 90J802454. 0148-0227/91/90JB-02454505.00

INTRODUCTION

The magmatic processes involved in the transformation of upwelling asthenosphere into oceanic lithosphere at mid-ocean ridges include partial melting, melt segregation, and a variety of

6133

6134 SINTON ET AL.: SUPERFAST SPREADING MID=OCEAN RIDGES

processes presumed to occur in crustal magma chambers and/or conduits. Thermal considerations suggest that the geometrical nature of these processes will have spreading rate dependence. For example, Sleep and Rosendahl [1979] predicted that the size, longevity, and along-axis continuity of magma chambers should increase with magma supply, which in the absence of hotspot effects can be correlated generally with spreading rate. Whitehead et al. [1984] suggested that a first- order magmatic segmentation of the slow spreading Mid- Atlantic Ridge results from the diapiric rise of discrete gravitational instabilities in the asthenosphere occurring at a spacing of approximately 50 km along axis, and Schouten et al. [1985] predicted that the spacing between these gravitational instabilities should increase to about 85 km at 200 mm/yr spreading rate.

A variety of structural and petrologic studies indicate that mid-ocean ridges are segmented at a scale finer than that predicted by considerations of gravitational instability. MacdonaM et al. [1984, 1988a] showed that medium and fast spreading ridges are segmented structurally by a variety of irregularly spaced offsets that occur between major transform faults. The boundaries of these segments take the form of propagating rifts, overlapping spreading centers (OSCs), and saddle points, which denote bathymetric deeps along the axial depth profile. Langmuir et al. [1986] pointed out that even structurally minor deviations in axial linearity (devals), as well as the more pronounced OSCs and transform faults, can be important magmatic boundaries. Taken together, the work of Langmuir et al. [1986] and MacdonaM et al. [1988a] suggests a hierarchy in the segmentation of mid-ocean ridges ranging from first-order discontinuities, such as transform faults, to fourth- order discontinuities, such as devals (see MacdonaM et al. [1988a], for discussion). It is now widely assumed that ridge segments are the primary units of crustal accretion at mid- ocean ridges, and their spacing, longevity and migration along axis provide the fundamental controls on the architecture of oceanic lithosphere created at mid-ocean ridges.

Although these studies have greatly enhanced understanding of the nature of mid-ocean ridge systems, the effect of spreading rate on the geometry of the upwelling zones and of crustal magma reservoirs remains largely unknown. This partly reflects a lack of systematic data on the extremes of the spreading rate spectrum. The purpose of this paper is to report the preliminary results of a lava sampling program along and near the axis of the very fast spreading East Pacific Rise (EPR) from 13'-23' south latitude, in order to assess the variations in magmatic processes occurring at very high spreading rates. Although spreading rates along the world's mid-ocean ridge system vary from less than 10 mm/yr to almost 160 mm/yr [e.g., DeMets et al., 1990], the East Pacific Rise south of the Rivera Transform, which spreads at rates above about 80 mm/yr, has been widely described as "fast" spreading. In order to distinguish the present study area, we follow the terminology of Lonsdale [1977] and Francheteau and Ballard [1984] in referring to rates above 130-150 mm/yr as "superfast" or "ultrafast" spreading. The fastest currently active spreading ridge occurs south of the Easter Microplate near 31øS, with a spreading rate of about 159 mm/yr [Naar and Hey, 1989; DeMets et al., 1990]. The East Pacific Rise between the Easter Microplate and the Garrett Transform has a full spreading rate between 150-155 mm/yr, a value only about 3% slower than that at 3 I*S. Thus our study area constitutes a reasonable type area for superfast spreading mid-ocean ridge processes.

THE EAST PACIFIC RISE 13ø-23øS

The East Pacific Rise axis north of the Easter Microplate (Figure 1) has been mapped by Seabeam [Tighe et aL, 1988;

Lonsdale, 1989] and SeaMARC II [Macdonald and Cormier, 1991]. The region is free of any known hotspots. There is profound along-axis variation in axial depth [Macdonald et al., 1988b], with the shoalest and broadest portion occurring near 17ø-18øS, and a pronounced offset depression near 20.7øS. The southern boundary of the East Pacific Rise near 23øS is a left- stepping transform fault that marks the northern boundary of the Easter Microplate according to Herron [1972] and Naar and Hey [1986]. Although more recent work [Engeln et al., 1988; Naar and Hey, 1989] indicates that the northern boundary of the Easter Microplate is complex, the EPR bends to the west into a moderately well-defined transform zone at 23ø03'S. The EPR extends for almost 1100 km to the north

without disruption by a major transform fault until the Garrett Transform near 13ø22'S. A major right-stepping offset occurs near 20.7øS [Rea, 1978; 1981; Macdonald et al., 1988b]. Macdonald et aL [1988b] studied this offset with Seabeam and Deep Tow and showed that it represents the tip of a southward propagating rift. The offset is complex and contains evidence for periods of both northward and southward propagation (dueling propagation), although the long-term propagation is to the south. South of this offset the EPR axis is segmented by several, mainly right-stepping, OSCs and devals. To the north it is segmented by several left-stepping OSCs and devals, with the exception of a right-stepping deval near 15øS and the Garrett Transform, which is also right-stepping (Figure 1). There has been little off-axis mapping done in the region, but Lonsdale [1986] mapped several seamounts between 17 ø and 20øS. A few seamounts are known to occur near the 20.7øS offset [Rea, 1978], and Macdonald and Cormlet [1991] discovered several seamounts on the Pacific Plate between 18 ø and 19øS, many of which have produced recent eruptions as indicated by highly reflective lava flows evident on SeaMARC II side scan sonar

images.

SAMPLE DISTRIBUTION AND GLASS COMPOSITIONS

In November-December 1987 we conducted a sampling cruise to the East Pacific Rise, 13ø-23øS, (University of Hawaii RN Moana Wave Cruise MW8712). We occupied 107 dredge stations, of which 102 successfully recovered igneous rock samples. Of the 102 successful dredges, 85 were from the active EPR axis, 7 north of the Garrett Transform, and 78 between the Garrett and the Easter Microplate. Of the remaining dredges, nine were on abandoned ridges associated with the 20.7øS dueling propagator, six were on seamounts on the Pacific Plate between 18ø-19øS, and two were from the Garrett Transform Zone. In addition to the MW8712 dredges, we obtained 28 dredge samples from F/S Sonne "Geometep" Cruises 12 and 26, conducted by Preussag AG, and 11 samples recovered by dredging and Cyana submersible operations carried out by IFREMER [Renard et al., 1985] during the 1980 Searise cruise of N/O Jean Charcot and a 1984 Cyana submersible cruise with N/O Nadir as support ship. The average sample spacing is of the order of 10-15 km along axis, but in several places (e.g., 18ø-19øS and the vicinity of several major offsets) the sample spacing is much closer. In this paper we report the results of electron microprobe major and minor element analyses of natural glasses from axial, near-ridge seamount, and Garrett Transform samples.

Samples were selected for analysis based on phenocryst content, sample morphology (sheet versus pillow, etc.), and weathering characteristics in order to cover the age and lithologic variations in each dredge. Ultrasonically cleaned natural glass chips were analyzed using the University of Hawaii fully automated, wavelength-dispersive, three-spectrometer Cameca MBX microprobe. To date, a total of 496 individual

SINTON ET AL.: SUPERFA• SPREADING MID-OCEAN I•IDGES 6135

114øW 15øS

113øW 112øW 13øS

•4os } t N

},

__

IIOøW

13øS

14øS

15øS

M

16os 16øS I IøW

17os ,7os

K

}, 115øW

18oS ............................... _. • • 18os

19os 19øS t

20øS 20os •ONø ; COCOS '

, $•Q•\e.o$ i ....... PLATE

21os [ •':.- • øø •V•)I$c•'Fe••GOt'OU•,..,_ 63mm/yr F PA C/F/ :'4 e •'ø•'• •4

, ,

114øW 113øW

Fig. 1. Map of the East Pacific Rise axis based on Seabe. am [Tighe et at, 1988; Lonsdale, 1989] and SeaMARC II [Macdonald and Cormier, 1991]. Boundaries of secondary magmatic segments defined in this paper (capital letters) are shown by the double arrows. Regions where T-MORB (see text) were collected are shown by lowercase t. Spreading directions and rates shown in the inset are from DeMets et at[ [1990].

6136 SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES

sample analyses have been obtained, using a combination of natural glass (major elements) and mineral (minor elements) standards. A minimum of five replicate measurements on up to three glass chips from each rock sample were averaged for each analysis and normalized to glass standards VG-2 and A99. ZAF corrections were applied to all intensity data. All analyses used a alefocused beam (--20 #m), an accelerating voltage of 15 kV and a sample current of 12 hA. Count times for most elements were 10 seconds; P205 was analyzed separately using count times of 50 seconds.

Sample analyses with compositions identical within analytical precision were averaged into group means for individual station locations. Each group mean represents 1-8 individual glass analyses and is considered to represent the mean for a single lava flow or group of closely related flows. We presently recognize 212 glass groups, of which 86 representatives are presented in Table 1. Standard deviations for each oxide, based on analytical variation within one well-analyzed group, are also listed in Table 1. Complete data sets are available from the authors.

GLAss COMPOSITIONAL VARIATIONS

Normal and Transitional MORB

Most recovered samples are normal mid-ocean ridge basalts (N-MORB), although moderately high degrees of differentiation appear to be characteristic of this area. Analyzed samples span the compositional range from basalt to dacite; the most common lava type is ferrobasalt with about 11 wt % FeO* (total Fe as FeO). N-MORB data define coherent trends on most MgO variation diagrams, suggesting that crystal fractionation from similar parental magmas can explain most of their compositional variations. Variation in A1203 versus MgO for the main population of N-MORB indicates that plagioclase joins olivine as a fractionating phase below about 7.8 wt % MgO, and that clinopyroxene joins the fractionating assemblage below about 6.8 wt % MgO. Although not precisely defined by our data, variations in TiO2 and P205 indicate that magmas from this region begin to fractionate an oxide phase at about 4 wt % MgO and apatite at about 3 wt % MgO. A subset of N- MORB recovered between 20ø20'S and 19ø05'S have higher A1203 and lower CaO/A1203 for a given MgO relative to other N-MORBs (Figure 2). Tight clustering of most of the data on projections in the clinopyroxene-olivine-plagioclase-SiO2 tetrahedron (Figure 3) implies that most collected samples are multiply saturated with olivine and plagioclase, and many are saturated with olivine, plagioclase and clinopyroxene.

Although most of the samples are N-MORB and their differentiates, about 12% (26 of 212 recognized chemical groups) are slightly enriched in K20 relative to the N-MORB population at the same MgO (Figure 4). This enrichment is relatively mild compared to that in alkalic basalts and, following the terminology of Bass [1972], we refer to these lavas as transitional (T-MORB). In addition to enrichments in K20, many, but not all, T-MORB are also slightly enriched in A1203 and lower in CaO/A1203 relative to N-MORB at the same MgO (Figure 2). When compared to N-MORB from the same ridge segments, slight enrichments in Na20 and slight depletions in FeO* at the same MgO are evident. T-MORB have higher K•i and K/P ratios than associated N-MORBs (Figures 4 and 5), and preliminary trace element analyses by X- ray fluorescence indicate that T-MORBs also are enriched in Rb, St, Ba and Nb. These results suggest that the two lava types cannot be produced by variable degrees of melting of the same source, at least not at extents of melting greater than

about 5%. Rather, the formation of T-MORB appears likely to have involved melting of enriched heterogeneities in the subaxial mantle.

The distribution of T-MORB along axis is irregular (Figures 1 and 5). Several T-MORB localities correspond to ridge segment boundaries or minor deviations in ridge strike (Figure 1), although some others bear no apparent relationship between eruption location and structural or magmatic segmentation of the ridge. T-MORB recovered between 17 ø and 18øS correspond with the highest magma supply to the ridge as indicated by the depth and breadth of the axial region in this area.

Axial Variations

Figure 5 shows some variations in glass compositions for axial lavas between the Easter Microplate and the Garrett Transform plotted versus latitude. When only N-MORB are considered, it is apparent that there are roughly twofold, more or less monotonic, increases in K•i and K• from 230S to the Garrett Transform near 13øS. Between 17ø-22øS these

gradients correspond to an increase in 87Sr/86Sr and a concomitant decrease in 143Nd/144Nd previously documented by Macdougall and Lugmair [1986]. However, although the K•i and K• ratios of N-MORBs continue to increase to at

least 13øS, the isotopic trends reverse near 170S [Mahoney et al., 1989]. Thus the variations in isotopic and minor element ratios appear to be alecoupled north of 17øS. The K• variation, and to a lesser extent the variation in K•i, can be used to divide the ridge axis into discrete populations with broadly similar K• ratios, especially when the slight increase in K/P with decreasing MgO (Figure 4) is taken into account. The boundaries defined by the K• populations correspond to the 20.7øS offset, the 15øS right-stepping deval, and perhaps the 16øS OSC. The 20.7øS offset correlates with a profound bathymetric deep, the 16øS corresponds to the second biggest offset in the axis between the Easter Microplate and the Garrett Fracture Zone, whereas the 15øS deval has only a very minor bathymetric expression. It is interesting that a major geochemical change appears to occur at 15øS, which is the only right-stepping offset between 20.7øS and the Garrett, although the significance of this observation is unknown at present.

Along-axis MgO variations also suggest several major, long- wavelength populations, each with its own particular MgO systematics. Samples collected between the Easter Microplate and the 20.7øS offset define a broad cusp in MgO with differentiation decreasing (i.e., MgO increasing) toward both ends. The highest MgO N-MORB recovered along axis occur just north of the transform bounding the Easter Microplate. Between 20.7øS and 14•0'S the MgO contents of the total population of N-MORB define a broad dome centered around 17ø-19øS, where the ridge axis is broadest and shallowest. Thus in this nearly 700 km-long region, MgO crudely correlates with regional magma supply. Superimposed on this broad dome in MgO are three subregions, each defined by distinct gradients in average MgO of most N-MORB comprising them (shown by stippling on Figure 5). The steepest gradient in MgO in the study area occurs over 200 km of ridge axis to the north of the 20.7øS offset. The profound southward decrease in MgO north of 20.7øS is consistent with a net long-term southward propagation of the offset [Sinton et al., 1983; Rea, 1978; MacdonaM et al., 1988b]. There is also a southward decrease in average MgO from 17ø05'S to a population of low MgO samples occurring between 18ø22'S and 18•7'S. North of the sharp discontinuity in MgO at the small deval at 17ø05'S, MgO decreases toward a slight bend in the ridge at 14•0'S, where

SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES 6137

another sharp discontinuity in MgO occurs. North of 14ø30'S, MgO decreases into the Garrett Transform Zone.

Based on axial bathymetric survey data, Lonsdale [1989] suggested that offsets near 15.9øS and 19.6øS are propagating to the south, and that offsets near 18.4øS, 18.6øS and 19.0øS are propagating to the north. Of these, only 18.6øS represents a significant boundary in the MgO variation of Figure 5; that offset shows a gradient of increasing MgO south of the offset. Thus none of the propagators identified by Lonsdale [1989] has the distinctive decrease in MgO behind the tips, a characteristic of all known long-lived, well-developed propagating riffs [Sinton et al., 1983], unless the gradient to the north of the 18.6øS offset can be taken as evidence for southward propagation. Our chemical results suggest that none of the other offsets has had a significant history of propagation.

Except for the crudely similar gradients in MgO and bathymetry north of the 20.7øS offset, there is no good correlation between degree of magmatic differentiation, as indicated by glass MgO contents, and ridge axial depth for most of the EPR between 13 ø and 23øS. Furthermore, many of the discontinuities in MgO do not correlate with those defined by incompatible minor element or isotopic ratios. This result indicates that a variety of magmatic processes and source compositions contribute to magmatic segmentation of the axial region and that these processes can be regionally alecoupled.

Magmatic Segments

Along-axis variations shown in Figure 5 reflect regional gradients in mantle composition, magma supply and degree of differentiation. However, they do not independently define the along-axis continuity of coherent magmatic units or segments. One way to define magmatic segments is by the along-axis limits of samples with compositions consistent with evolution from similar parental magmas. Because the data considered in this paper are for natural glasses (= quenched liquids), an evaluation of oxide variations versus MgO (e.g., Figure 6) allows the determination of which samples are consistent with evolution along similar liquid lines of descent. We have noticed that spatially associated samples tend to lie along common linear arrays on MgO variation diagrams (liquid lines of descent), but that samples on either side of some ridge axis offsets lie on different, but subparallel trends. Samples from subparallel, but offset, liquid lines of descent probably arise from different parental magma compositions, and hence the locations along axis where changes in parental magma compositions occur can be used to delimit magmatic segments. For example, the data shown in Figure 6 suggest that the samples erupted between 15ø55'S and 16ø30'S (Segment L) evolved from parental magmas with slightly higher SiO2 and Na20, and possibly FeO*, but comparable CaO contents to those for the region between 16ø30'S and 17ø57'S (Segment K).

Because major element liquid lines of descent are relatively linear between 5.5 and about 8.5 wt % MgO, it is possible to calculate average parental magma compositions at some reference value of MgO by projecting the observed glass mean compositions back to this reference value, parallel to the average liquid paths for each oxide. Consistent with previous studies [e.g., Klein and Langrnuir, 1987], we have chosen 8.0 wt % MgO as the reference value. We determined the average slope of the liquid paths for each oxide by regressing lines through 133 axial group means with MgO values between 5.5 and 8.5 wt %. The average parental magma compositions at 8 wt % MgO for individual segments and equations for their calculation are presented in Table 2.

In many cases, more than one chemical type was recovered in a single dredge; the significance of this result is not entirely clear, but it probably reflects temporal variations in magma composition. Most of our dredge stations were relatively short (< 1 km bottom tracks) and precisely located using Seabeam and/or SeaMARC II charts. All stations that we consider to be

axial dredges are within 2 km of the axis defined by the bathymetry; most are within 1 km of the axis. Assuming continuous symmetric spreading at the approximately 150 mm/yr spreading rate for this area, all samples within 2 km of the axis were erupted within the last 27,000 years. Much of the variance determined for average parental magma compositions of individual segments (Table 2) results from variations occurring within individual sample locations. In the averages r•resented in Table 2, parental oxide compositions more than 20 from the mean were excluded, but only for segments composed of more than 10 chemical groups, and in no case were more than two values excluded for any oxide average. In the great majority of cases the average is within la of all sample values.

We have used this method to determine the locations along the axis where significant chemical boundaries exist. The criterion used is that the average parental magma compositions on either side of the boundary must be different in at least one oxide by an amount at least as great as the sum of the respective la values for that oxide. The average segment parental magmas presented in Table 2 all meet this criterion. Inspection of the data in Table 2 indicates that the average parental magmas to Segments K and L, for example, are significantly different with respect to SiO2 and Na20 at la, but due to the high variance of FeO* in Segment K, the apparent slightly higher average for segment L (Figure 6, Table 2) is not significant at the la level. This illustrates that our definition of magmatic segments is conservative, but its statistical basis should be relatively unbiased.

Using this procedure, we have identified 17 magmatic segments between the Easter Microplate and 13øS, with 15 segments occurring south of the Garrett Transform (Figure 1). These segments include from 1 to 18 sample locations and are defined by 2 to 19 chemical groups. Virtually all major structural boundaries (OSCs, propagating rift tip, Garrett Transform) correspond to magmatic segment boundaries, but several magmatic boundaries (6 of 15 south of the Garrett) correspond to devals with only very minor structural definition. Furthermore, there are many fourth-order structural boundaries, some of which are clearly evident in the bathymetric data, that do not emerge from our analysis as significant boundaries with respect to the coherence of magmatic liquid lines of descent. The processes giving rise to these segments are discussed in a later section.

KEY FEnTUr•.s OFTHE EAST PACIFIC RISE, 13-23øS

The 20. 7•S Propagating Rift

The most obvious disturbance in the geochemical signal along the entire ridge corresponds to the major bathymetric depression associated with the 20.7øS propagating rift. Boundaries in glass K/P and K/Ti coincide with this offset, and a gradient in degree of differentiation characterizing the northern rise axis for almost 200 km culminates just north of the offset (Figure 5). Macdonald et al. [1988b] studied this offset with Seabeam and Deep Tow and showed that its fine scale structure is complex for two primary reasons. First, it is a "dueling propagator" produced by successive short-term episodes of northward and southward propagation at rates of 160-230 mm/yr, with long-term net migration to the south at

6138 S•NTON ET AL.: SUPERFAST SPREADING M•D-OCEAN PdDGES

SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES 6139

6140 S•NTON ET AL.: SUPERFAST SPREADING MID=OCEAN R•DGES

19

18

• 17

16 15

< 13 12

0.9

0.7

0.6

0.• • , I • I • I , I • I •

4 5 6 7 8 9 10

Mg0 wt. z Fig. 2. A1203 and CaO/AI203 variations versus MgO for glass data in this study. Most N-MORBs plot within the field enclosed by the solid curve. The field enclosed by a dashed line includes selected samples from propagating rift segment G. T-MORBs are shown as lowercase letters denoting the segment from which they were recovered. The glass group for dredge 103 from the Garrett Transform is shown by a triangle. Data for southern Hump seamounts (dash double dot) and northern Hump seamount samples (dotted field) are also shown. Pluses on the upper left of each diagram show 2or variations in analytical uncertainty.

about 20 mm/yr. Second, it bears many characteristics of OSCs as well as some of those of classic propagating rifts.

We sampled the northern and southern limbs of the axis in this region, as well as several associated off-axis ridges that MacdonaM et al. [1988b] considered to be abandoned by short- term propagation events as the propagating rifts curved either

0.1

KaO wt. •. n g -3- 2a' ! g k

4 5 6 7 8 9 10

MgO wt. z Fig. 4. K20, K/Fi and KIP versus MgO for data of this study. Symbols are the same as those in Figure 2.

to the inside or outside of previous curving ridge tips. Lavas from propagating rift segments north of the offset are significantly more differentiated than those from the failing or doomed southern rift segments (Figures 2, 7, and 8). The fields enclosing data from the southern rift segments also generally extend to higher Na20 and SiO2 for a given MgO, relative to those from the propagating rift (Figure 7, Table 2). Maximum differentiation on the presently active axis occurs near the southern tip of the northern (propagating) axis (Figure 8). This situation contrasts with that for the Galapagos 95øW

CPX PLG

x

OLV $1L

Fig. 3. Projections of the glass data in the CMAS tetrahedron. End-members are those of Walker et aL [1979] except that molecular FeO = FeO+MnO - (Fe203+TiO2) and CaO = CaO - 3.3 P205. Most N-MORB plot in the field enclosed by a solid curve. Those lying outside of this field are shown by an open circle. High A1203 propagating rift samples from Figure 2 plot within the short-dashed field on the projection from plagioclase (left-hand diagram). Other sample symbols are T-MORBs (plus), southern Hump seamounts (cross) and Garrett dredge 103 (triangle). The long-dashed curves show the multiple saturation surfaces (cotectics) inferred from the data. The data enclosed in the dotted field are interpreted to represent magmas formed by mixing of end-members lying in the directions of the short arrows. See text for discussion.

SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN R•DGES 6141

2600

2800

3000

8.0

0 6.0

4.0

3.0

2.0

0.2

/ '•."•;....•!;:•ii: .7- o ::.• •..::::• . ;::•:::: • •....::::•..:./•

o I o .... •:• i i I

I I I I i I I I I

I4cr

o o o o

o o

o

o o o o o o ø o o

'

I2o'

24øS

Fig. 5. Along-axis variations in axial depth [from Macdonald et al., 1988b], and glass MgO, K/Ti and K• for axial samples from this study. T-MORB samples are shown as open symbols. Only samples with less than 52 wt % SiO2 are shown on the K/Ti and K• plots. Stippled fields approximately 2•r (K• and K/Ti) or 40' (MgO) wide (Table 1) are drawn through the majority of the N-MORB data to show general trends. These fields encompass almost the entire spread of N-MORB data for K/P and KfFi; the observed range of data for MgO (dashed curve) is locally greater than

propagator where maximum differentiation occurs about 15 km behind the propagator tip [Christie and Sinton, 1981]. However, at the Galapagos propagator, true seafloor spreading begins about 15 km behind the tip defined by the intersection of pseudofaults [Hey et al., 1989]; the variation in differentiation behind the spreading center tip of the Galapagos 95øW propagator shows a similar relationship to that at 20.7•.

Samples collected from the 20.7• propagating rift segment G define two distinct populations with respect to their variation in AI203 and C, aO/AI203 relative to MgO. Those within 40 km of the propagating rift tip lie along similar trends to those of the doomed and failing rifts, which in turn are similar to most N- MORB from the area (Figure 2). However, those erupted 40 to 180 km behind the rift tip mainly have higher AI203 and lower CaO/AI203 for a given MgO compared to other N- MORB. Furthermore, the high AI propagating rift lavas project to a field displaced toward olivine in the Cpx-olivine- SiO2 projection (Figure 3), suggesting that they became saturated with clinopyroxene earlier in their evolutionary history, relative to other N-MORB of the area. If so, these magmas may have fractionated at greater depths than either those within 40 km of the propagating rift tip or other magmas of the EPR in this region. High-AI propagating rift samples are almost all aphyric and so their early crystallization histories are not recorded in their mineralogy.

The compositions of two sample groups in dredge 33 from abandoned ridge 2 of Macdonald et al. [1988b] fall well off the differentiation trends defined by all other samples from the region (Figures 3 and 7). One of these groups (33B) is essentially aphyric but a sample from Group 33C contains abundant rounded grains of clinopyroxene and complexly zoned plagioclase grains with once partially melted cores. The groundmass of a Group 33C lava contains fasciculate clinopyroxene in a texture reminiscent of those produced from superheated liquids [cf. Walker et al., 1979]. These dredge 33 compositions are best explained as mixed lavas and appear to represent an unusual kind of mid-ocean ridge magma mixing in which the end member compositions must have been highly evolved silicic magma and relatively undifferentiated basalt. The best model solution requires about 20% of our most differentiated composition (P12-33) to be mixed with 80% of a basalt with about 6.8 wt % MgO (Table 3). Christie and Sinton [1986] argued that similar mixed magmas requiring at least one highly evolved end-member from the Galapagos 95øW propagator probably represent single-event mixing of isolated magma batches rather than mixing in long-lived magma chambers. Such single-event mixing with extreme end-member compositions may be a unique characteristic of propagating rifts. As pointed out by Walker et al. [1979], magma mixing involving one highly evolved, low temperature end-member can

6142 SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN R•DGES

52

51

50

12

10

12

2.8

2.6

2.4

16030 ' OSC

SiOz

L LL L Fi L L L .... ..... i i I -'IK i i

2 LL L

FeO* ß .... I .... I .... I ..... • • • ,

K

CaO /K •K

L L L

.... I .... I .... I , • • , I • , ,

L ß

L L K

KKK • L L K LI•KK K Na20

6 6.5 7 7.5 8 8.5

MgO wt.. z Fig. 6. Compositional variations for lavas from the segment to the north (L) and south (K) of the 16ø30 ' OSC. The hatchured and open rectangles represent the calculated average "parental" oxide values (_+ lc• high) at 8 wt % MgO for the samples from segments K and L, respectively. As discussed in the text, the calculated (oxide)8.0 for one sample (Group 81A) is more than 2o' from the mean for FeO* and Na20 (solid triangle) and therefore excluded from the calculation of the average values for those oxides. The sample is included in the means for SiO2 and CaO (open triangles). According to the statistical definition of segments used in this study, the parental magmas, and hence the liquid lines of descent, for segments K and L are significantly different (at lc 0 in SiO2 and Na20, but not in FeO* and CaO.

produce superheated magmas, a relation consistent with the groundmass textures present in some samples from Group 33C.

Lavas recovered from abandoned ridges associated with the 20.7øS offset are quite variable in composition but mainly consist of highly differentiated ferrobasalts. Relatively undifferentiated samples, similar to those of the failing rift, are rare in the abandoned ridge collections. The most evolved lavas recovered from the area are andesites and dacites from

the recently abandoned ridges 2 and 4 of Macdonald et al. [1988b] (Figure 7). Thus most of the abandoned ridges appear to carry the distinctive, evolved fractionation signatures of rift propagation [Sinton et al., 1983] and generally lack the undifferentiated character of failing rifts. Significantly, this result implies that propagating rift tips in this area tend to be abandoned preferentially to failing rift tips; this appears true even for abandoned ridges with eastward-concave curvature, indicating that they probably were abandoned from the southern (doomed) limb. The explanation is that abandoned ridge tips are created by self-decapitation as the active ridge cuts inside of the soon-to-be abandoned ridge tip [Macdonald

et aL, 1987]. Thus even a long-term "failing ridge", in the parlance of Hey et al. [1989], briefly can propagate during the process leading to ridge abandonment. Because the 20.7• offset has a history of dueling propagation, previously propagating tips occur on both the northern and southern limbs (see Figure 14 of Macdonald et aL [1988b]). The presence of fractionated signatures typical of propagating rifts on most of the abandoned ridge samples suggests that ridge abandonment in this area is favored by successive propagation events along the northern and southern limbs of the offset.

Failing Rifts of the East Pacific Rise, 20. 7- 23•S

As the ridge north of the 20.7øS offset propagates to the south, the southern ridge will fail. In the terminology of Hey et al. [1989] this is the doomed rift; that part of the doomed rift that has already begun to fail is the failing rift. In the Galapagos 95øW propagating/failing rift system the petrological characteristics of rift failure include restriction to relatively magnesian lava compositions, for the most part unaffected by magma mixing. Batiza and Vanko [1984] recovered a wide range of lithologies from the Mathematicians failed ridge; their collections mainly consist of relatively undifferentiated rock types, including some alkalic rocks. The steady northward decrease in degree of magmatic differentiation beginning about 40 km south of the 20.7øS offset (Figures 5 and 8) may indicate that this part of the southern ridge is beginning to fail as the 20.7øS offset propagates to the south. The tendency for the southern rift segments to have higher Na20, P205 and SiO2, and possibly slightly lower FeO* at a given MgO relative to the propagating rift (Figure 7) is similar to compositional relationships between propagating and failing rift magmas of the Galapagos area. Yonover [1989] interpreted these results to indicate that Galapagos failing rift magmas form by lower degrees of shallower melting than propagating rift magmas. The 20.7øS failing and propagating rift magmas appear to bear similar compositional relationships to those of the Galapagos area.

The southern boundary of the field area is a transform fault, which according to Herron [1972] and Naar and Hey [1986] marks the northern boundary of the Easter Microplate. There is a southward increase in axial depth for about 90 km north of this transform, becoming especially pronounced over the last 30 km (Figure 5). However, unlike many other regions with well- defined axial bathymetric gradients (e.g., north of the 20.7øS offset, south of the Garrett Transform (Figure 5), south of Tamayo [Bender et al., 1984], or north of the Clipperton Fracture Zone [Thompson et al., 1985, 1989; Langmuir et al., 1986]), where lavas tend to become increasingly differentiated as the axis deepens, lavas north of the 23ø03'S transform become less differentiated as axial depth increases approaching the transform intersection.

Recent work on the northern boundary of the Easter Microplate indicates that there is a broad wedge of diffuse seismicity [Engeln et al., 1988] and disturbed structure [Naar and Hey, 1989] east of the East Pacific Rise between about 22øS and the Easter Microplate. According to Naar and Hey [1989], spreading rates on the EPR south of 21ø59'S are slower than those predicted by a simple Pacific-Nazca pole of opening, suggesting that this portion of the EPR may be affected by a broad region to the east extending the northern "boundary" of the Easter Microplate. Our geochemical data show that the gradient of increasing MgO along with axial depth north of the Easter Microplate also begins at about 22øS. Thus there appears to be a coincidence near 22øS between the boundary of the region of diffuse seismicity east of the axis, the southern

SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES 6143

6144 SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES

20.7øS PR 3

SiO z 2

65 2 1 6O

2 0

55 '•.•

4 • "••• -• • 0.4 50 4

45 ' ' ' 0.2

Na20 2 o

4 2

4 1•] 1,5 3

"- •"i' 10 2 • • • 5

4 %TiO 2 ß 2 •2 ,

• I

4

2

2 2 i i 1

• FeO*

2 4 6 8 2 4 6 8

M•O wt. •, M•O wt. •, Fig. 7. Variation in SiO2, Na20, TiO2, P205 and FeO* versus MgO for samples from the 20.7øS offset region; data for T- MORBs from this region are not plotted. Data for samples from the present axis of the northern (propagating) rift segments lie in fields enclosed by solid curves; those from the southern (doomed and failing) rift segments fall in the fields enclosed by dashed curves. Samples from abandoned ridges are plotted as numbers corresponding to the abandoned ridge designations of Macdonald et al. [1988b], or as open squares for ridges farther away from the axis than those considered by MacdonaM et al. [1988b]. The compositions of the samples from abandoned ridge 2 which we consider to be best explained by magma mixing between dacite and ferrobasalt end-members (arrows show the directions of end-member compositions) are designated by an asterisk.

limit of "normal" spreading rates and the beginning of the gradient of increasing MgO with increasing axial depth. The correlation between increasing axial depth and decreasing extent of fractionation is similar to that for failing rifts of the

FALLING RIFT

ß [] ß

I I I I

EPR 20.7øS PROPAGATING RIFT

ß ß

i ß I ii ß

ß ß

o ß

o

o ß

o o

i i i

o

o

o

o

0 I I I I I I I I I I I I I

120 80 40 0 40 80 120 160

Distance (km) Fig. 8. FeO*/MgO and K/P versus along-axis distance from the 20.7% propagating rift tip. The position of the propagating rift tip is shown by the vertical dashed line. T-MORB are shown as open symbols.

Galapagos 95øW and EPR 20.7øS regions. Thus one interpretation of the along-axis variation in degree of differentiation and magma supply (as indicated by axial depth) south of 22øS is that this portion of the EPR axis is beginning to fail as the northern boundary of the Easter Microplate extends to the north.

The Hump Region, 18 ø - 19%

Macdonald and Cormier [1991] mapped an area of about 14,000 km 2 centered on the EPR at 18-19øS, known as "The Hump", using SeaMARC II (Figure 9). The axis in this region consists of three segments separated by left-stepping OSCs with offsets of about 1-3 km. The middle (18ø37'-18ø22'S) segment is characterized by an axial graben about 1 km wide and about 50 m deep. This segment was previously sampled during Preussag Geometep Cruise 26 by Backer et al. [1985]. The axial segments to the north and south do not have an axial graben and are instead characterized by axial domes or rectangular cross sections. We sampled all three axial segments by dredging and have also analyzed samples from the Preussag program. In addition, we ran bottom camera surveys over all three axial regions in order to assess variations in volcanic style and hydrothermal activity in these three regions with contrasting structural development.

Affal variations. The 18ø37'S and 18ø22'S offsets are small

OSCs that mark magmatic segment boundaries (Table 2). The 25-km-long segment between these offsets (Segment I) is among the shortest defined by our data for the EPR axis in the study region. This rifted segment contrasts markedly in having erupted lavas extending to substantially more differentiated compositions than the unrifted segments on either side (Figures 5 and 10). This result suggests that there may be a relation

SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES 6145

TABLE 3. Magma Mixing Solutions for Group 33B

44B P 12-33 80A:20B 33B A

SiO2 50.78 66.93 54.10 54.00 0.04 TiO2 1.88 1.06 1.72 1.72 0.01 33203 14.70 13.80 14.24 14.54 0.15 FeO* 11.61 7.12 10.51 10.74 0.23 MnO 0.17 0.13 0.22 0.16 0.06

MgO 6.86 1.15 5.63 5.75 0.12 CaO 11.03 4.20 10.02 9.70 0.32

Na20 2.72 4.79 3.21 3.13 0.08 K20 0.09 0.68 0.17 0.21 0.03 P205 0.16 0.12

Sum of squares of residuals - 0.204.

between degree or range of differentiation of erupted lavas and stage of tectonic development of the spreading center axis in this area. Our photographic data indicate that the central, rifted segment is floored by recent eruptions, but the presence of an axial graben implies that eruptions have not kept pace with spreading. The presence of relatively differentiated lavas in this segment requires that the eruptions have tapped subaxial

magma that has not received volumetrically significant, recent replenishments from the mantle. In contrast, where the eruptions have kept pace with spreading such that an axial graben has not formed or has been recently flooded (e.g., the axial segments to the north and south), the erupted magmas are uniformly less differentiated.

Relation to hydrothermal activity. Camera traverses approximately 5-7 km long were run oblique to the rise axis in each of the three axial segments of the Hump region. The camera results contain information relating to volcanic style, eruption frequency, and hydrothermal activity in each of the three segments. A variety of lava forms and evidence for hydrothermal activity were present in each segment. However, sediment cover was generally more extensive in the middle segment, suggesting that the eruption frequency was somewhat less there. In the southern segment, evidence for recent hydrothermal activity was extremely rare and young lava forms (Plate la) dominate the bottom photographs. In the northern segment, yellow hydrothermal deposits (nontronite?) were locally distributed along cracks and interflow interstices. Small, apparently extinct, hydrothermal chimneys were photographed in only one place in the northern segment (Plate l d). These results suggest that this segment may have experienced recent

114ø00'W 113ø30 '

t ,65

62

113o00 '

o - )0'

S

o

18 ø - 30'

19 ø 00'

Fig. 9. Bathymetric map of the Hump Area (after Macdonald and Cormier [1991]). The stippled regions less than 2800 m effectively show the rise axis and many of the off-axis seamounts. Numbered MW8712 dredge tracks are shown as bold lines; other samples (open circles) are from Preussag cruises and one sample recovered during MW8712 bottom camera run 11 (sample location not well known).

6146 SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES

HUMP Axis and Seamounts lO

200

150

100

50

9

8

J

7 J

6

J

o

17.9

d

ß ß H ß J ß

J ß I ß ß j ß

JJ i•[ HH

• H

T ß

, m I , • I , m I , ,

18.2 18.5 18.8 19.1

Degrees South Latitude

Fig. 10. Variations in MgO (wt %) and Sr (ppm) for samples from the Hump axial segments (H, I, J) and seamounts (triangles). MgO values were determined on glass by electron microprobe. Semiquantitative Sr analyses were performed at sea by energy-dispersive XRF. Shore- based mass spectrometric isotope dilution and wavelength-dispersive X-ray fluorescence analyses for Sr on samples analyzed at sea indicate that our shipboard Sr analyses were too high by an average of 33 ppm (range = 7 o 48 ppm). Data from the ridge axis samples plotted here all have been reduced by 33 ppm from the original shipboard analyses; all seamount data plotted here are isotope dilution mass spectrometric analyses (D. G. Waggoner, unpublished). Relative to the axial samples, all southern seamount samples (inverted triangles) are enriched in Sr and the northern seamounts (upright triangles) are depleted in Sr.

hydrothermal activity and/or that hydrothermal outflow in this region may be widespread, but poorly organized. A variety of biological communities, dominated by anemones and crinoids, were associated with the hydrothermal deposits.

In contrast, the camera run on the middle segment contained the best evidence for localized hydrothermal activity, in the form of yellow and green hydrothermal deposits and a number of extinct and active chimneys, including one active "black smoker" (Plate lc). The chimneys were best developed near the axial graben-bounding fault scarps. Taken together, these data suggest that present hydrothermal activity is relatively weak in the southern segment, more recently active in the northern segment, and presently very vigorous and well organized in the middle segment. This order corresponds with one for decreasing average MgO of the three segments (7.9, 7.6, and 6.9, respectively, for the southern (H), northern (J) and middle (I) segments) (Figure 10). Perhaps hydrothermal cooling of the subaxial magma system is the critical parameter affecting magmatic differentiation in the Hump area.

Near-axis seamounts. MacrohaM and Cormier [1991] mapped a total of 13 seamounts in 7000 km 2, all occurring on the west flank of the Hump area (Figure 9). We sampled six of these seamounts that contained lava flow fields with highly reflective side scan sonar characteristics, indicative of young volcanic activity. Four of these dredges were on seamounts from the southern part of the Hump area, 18035 ' to 18ø55'S, which define a lineament approximately parallel to the spreading direction. Two dredges were on seamounts in the northern part of the Hump area that are part of a chain

oriented northwest, approximately parallel to the direction of absolute plate motion (hotspot reference frame); this trend intersects the EPR axis near 18ø27'S.

Lavas recovered from the southern (relative motion) seamounts mainly consist of a variety of mildly alkalic compositions ranging from weakly nepheline-normative (up to 1.22 wt % calculated with 0.1 FeO* as Fe203) to weakly hypersthene-normative (up to 1.57 wt %), broadly similar to some lavas recovered from near-ridge seamounts on the east flank of the East Pacific Rise near 21øN [Batiza, 1989]. These samples plot on the SiO2-deficient side of the cpx-olivine join in the cpx-olivine-SiO2 projection (Figure 3). Only one axial group with similar normafive characteristics (mainly attributable to relatively low SiO2) was recovered. This group (51A) comes from the EPR axis at 19ø00'S, just north of the 19ø05'S offset. Among the southern seamount samples, two glass groups are K20-enriched T-MORB. In addition to alkalic lavas, some strongly hypersthene-normative basalts similar to nearby axial lavas were recovered in dredge 55, although the seamount samples have higher A120 3 for their MgO contents, relative to the nearby axial lavas (Figure 2). Samples recovered from the northern (absolute motion) seamounts are tholeiites with similar amounts of normafive hypersthene (13 to 21.3 wt %) to the axial lavas (mainly in the range 12.3 to 19.4 wt %). However, these seamount lavas are distinct from the axial lavas in being relatively depleted in A120 3 (Figure 2) and K20 contents, and in having lower K• and K/Ti ratios (Figure 4).

Preliminary analyses for Sr show that the seamounts also define relatively incompatible-element enriched (southern seamounts) and depleted (northern seamounts) populations relative to the axial lavas (Figure 10). On Sr-MgO variation diagrams (not shown) it is evident that the Sr enrichment of the southern seamounts, relative to the axial lavas, persists throughout the mildly alkalic, strongly tholeiitic and T-MORB types. Sr enrichment appears to characterize the source regions for the southern seamounts, irrespective of their melting history. The result that seamounts approximating a relative motion trend are more enriched, and those from along an absolute motion trend are more depleted than associated axial lavas is consistent with findings for other Pacific near-ridge seamount provinces [Batiza and Vanko, 1984; Fornari et al., 1988; Batiza, 1989], although the processes responsible for this result are presently unknown. The southern seamounts (enriched) generally have higher Al203 and significantly lower CaO/A1203 than axial N-MORB (Figure 2), probably indicating that they formed by lower extents of melting than axial lavas. The northern seamounts are among the highest CaO/A1203 and lowest A120 3 of any samples from the area, possibly indicating that they formed from unusually high degrees of melting. Alternatively, they could have arisen from mantle already depleted by melting events giving rise to nearby axial lavas. We are continuing the study of seamounts and axial lavas from this region using a variety of isotopic and elemental abundances, the results of which will be reported in a separate publication.

The Garrett Transform Zone

The Garrett Transform Zone has been mapped using Seabeam [Fox and Gallo, 1989; Lonsdale, 1989] and SeaMARC II [MacrohaM and Cormier, 1991]. It contains three small pull-apart basins that are oriented slightly oblique, in a NE-SW sense, to the spreading direction. High reflectivity on the SeaMARC II side scan images indicates that the westernmost basin might be floored by young lavas. We sampled the westernmost and easternmost basins in order to

SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN R•DGES 6147

6148 SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES

determine the compositions of lavas possibly erupted in such unusual settings. Samples from the easternmost basin (dredge 111, Table 1) are quite weathered and have chemical compositions consistent with eruption from the nearby rise axis north of Garrett (segment Q). However, lavas collected from the westernmost basin (dredge 103, Table 1) have fresh glassy rinds and many unique chemical characteristics compared to either nearby lavas or any other axial samples in the study area. They are the least differentiated samples collected in this study (Figures 2-4) and appear to have formed from a magma enriched in Fe, depleted in SiO2 and CaO, and with some of the lowest CaO/AI203 (Figure 2) and K/Ti (Figure 4) values of any axial N-MORB from the region. We suggest that the distinctive lavas from dredge 103 were probably erupted in the transform zone. Their undifferentiated nature argues against the presence of a magma chamber beneath the pull-apart basin. Their high FeO and low SiO2 suggest relatively high pressure melting [Langrnuir and Hanson, 1980; Klein and Langrnuir, 1987]. The low CaO and CaO/AI203 (Figure 2) are consistent with either low degrees of melting or early (high pressure) fractionation of clinopyroxene, whereas the low K/Ti probably indicates derivation from a depleted source. These compositional characteristics do not permit us to distinguish whether these magmas formed by simple decompression melting in response to the transform pull-aparts, or alternatively, if they reflect small melt anomalies beneath the transform as proposed by Fornari et al. [1989] for Siqueiros Transform lavas.

DISCUSSION

Transitional MORB of the East Pacific Rise

T-MORB are variably distributed along axial segments of the study area with 26 of the chemical groups defined by the major

and minor element data here classified as T-MORB. Of the 15

magmatic segments occurring south of the Garrett Transform, 7 have at least one T-MORB group, although T-MORB are particularly abundant in segments G, K and N (Figures 4 and 5). The nature of the element enrichments and distinctive minor element ratios of T-MORB indicate that these samples probably represent melting of mantle containing discrete, K- enriched heterogeneities, rather than peculiarities in the melting of homogeneous mantle. Lava samples with similar enriched characteristics have been reported from the EPR 6 ø- 14øN [Langmuir et al., 1986], and a compilation of the reported analyses for the whole EPR between 23• and 14øN (Figure 11) shows that the eruption of T-MORB is a common characteristic of most of the EPR. T-MORB of the type described here are extremely rare on the Galapagos Spreading Center (GSC), however, suggesting that the mantle source for T-MORB is absent within the melting region beneath the GSC.

In the hotspot frame of reference, the East Pacific Rise is migrating WNW at about 19 mm/yr [Hey, 1977]. Therefore the shallow mantle now available for melting beneath the GSC already may have undergone melting events giving rise to earlier EPR magmas. The lack of T-MORB along the GSC is consistent with this interpretation if the source for T-MORB has a low melting temperature. Autio and Rhodes [1983] suggested that the abundance of highly depleted lavas from Hole 504B in crust formed on the Costa Rica Rift might be related to the "recycling of previously melted mantle (from the East Pacific Rise)." Our results and the summary shown in Figure 11 support this possibility.

Implications for Magma Chamber Processes beneath Mid-Ocean Ridges

The collection of multiple lava types in single dredges, and the variable distribution of disparate magma types along axis,

0.5

o

-25

s

EPR

-15 -5 5

Degrees Latitude

15

0.5

o

lO5 8o

GSC

lOO 95 90 85

Degrees W. Longitude

Fig. 11. Variations in K/Ti along the East Pacific Rise (EPR) and Oalapagos Spreading Center (OSC). Dashed curves represent the approximate upper bounds of all the N-MORB data; samples plotting above these lines are T-MORBs. T- MORBs characterized by moderately high K/Ti are very abundant along the EPR. Although high IGTi provinces are locally present on the GSC, particularly in the vicinity of the Oalapagos hotspot near 91øW, T-MORBs are extremely rare along the GSC. We interpret this result to indicate that the mantle source for T-MORBs has a low melting temperature and that the source for GSC magmas has been depleted in this component by previous melting events giving rise to EPR magmas. Data sources are Byer/y et aL [1976], Christie and Sinton [1981, 1986], Clague et aL [1981], Fisk et aL [1982], Fornari et aL [1983], Hekinian [1971], Langmuir [1988], Melson et aL [1976], Natland and Melson [1980], Perfit and Forttad [1983], Puchelt and Emmermann [1983], Scheidegger and Corliss [1981], Thompson et aL [1989], this study and unpublished University of Hawaii data (1985-1991).

SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES 6149

has important implications for magmatic plumbing beneath the EPR. The data seem to indicate that discrete magma types can be generated from different sources beneath the axis, and make their way to the surface, often in close proximity to one another, without becoming completely mixed in subaxial magma chambers. The uniformity in axial depth and other structural characteristics extending over several tens of kilometers for some segments, as well as the generally aphyric, yet moderately evolved, nature of most samples in our collections, suggests that subaxial melt-dominated magma reservoirs are probably quite extensive beneath this part of the EPR. However, the distribution of T-MORB and other, more subtle, compositional variations indicate that such chambers must be relatively inhomogeneous over time scales of approximately 27,000 years or less. This heterogeneity can be accommodated either by extremely limited convective mixing along axis, or by relatively frequent replenishments of new magma within the time resolution afforded by our sampling program. Although we cannot preclude the latter explanation, a variety of theoretical and experimental considerations [e.g., Marsh, 1989; Jaupan and Brandeis, 1986; Oldenburg et al., 1989] indicate that mixing by convection in magma chambers is likely to be limited in an along-axis sense. Even where magma chambers might be more or less continuous along strike, the chambers should be divided into mixing cells. Many T-MORBs were recovered from the vicinity of offsets, and of the four dredges (12, 50, 81 and P26- 171) clearly sited within an offset zone, all include at least one T-MORB group. We interpret this result to indicate that mixing is least likely to homogenize subaxial magma near offsets, where magma chambers are most likely to be disrupted by thermal and structural complexities in the subaxial plumbing systems. Notably, our data suggest that magma chambers are incompletely mixed even at the high magma supply rates expected to be associated with superfast spreading. At slower spreading, where the rates of eruption and supply are correspondingly less, the along-axis heterogeneity should be even more profound.

Magnatic Segmentation at Superfast Spreading

The East Pacific Rise between 13 ø and 23øS is structurally segmented by transform faults, a large, dueling, propagating rift, and at least 50 small OSCs and devals. Some sort of structural disruption in the axis can be found in the bathymetric data every 15 to 30 km along strike, similar to the East Pacific Rise at slower spreading [e.g., Langrnuir et al., 1986; Macdonald et al., 1988a]. If there is any difference between the structural segmentation at superfast spreading and that at slower rates, it is that the ridge is mainly broken by many small offset boundaries, with medium and large offsets being rather rare.

Langrnuir et al. [1986] proposed that the various scales of ridge axis offsets may reflect different scales of melt supply resulting from variations in the depth and spatial scale of melt generation from an upwelling mantle. Through an analysis of the off-axis structural expressions of ridge axis offsets, MacdonaM et al. [1988a] documented the character and time scales of the various scales of ridge axis discontinuities. The combined result of these studies is a hierarchy of physical segmentation of ridge axes. According to Macdonald et al. [1988a], first-order discontinuities at both fast and slow spreading centers are transform faults. They persist for 107 years or more, and define rigid offsets of the ridge axis exceeding 20-50 km. They segment the ridge on a length scale of 100-1000 km, commonly 200-500 km. On fast spreading centers, second-order discontinuities partition the ridge on length scales of 50-300 km. They are nonrigid and are typically

OSCs that offset the ridge 2-20 km and persist of the order of 106 years. A third-order segmentation, on a length scale of 30- 100 km, is defined by small offset OSCs (0.5-2 km) whose longevity is only of the order of 104-105 years. On an even shorter length scale of 10-50 km are fourth-order discontinuities which last less than 104 years. They are manifested as devals, which are small bends or changes in strike of the axis, small (<0.5 km) nonoverlapping offsets of the axial summit graben, or saddle points, which are small depth anomalies ( < 50 m).

There are several different ways in which the EPR 13ø-23øS can be partitioned based on chemical data, which suggest several different scales of magmatic segmentation (Figure 12). In order, to avoid confusion with the four orders of physical segmentation of Macdonald et al. [1988a], we refer to the various scales of magmatic segmentation as primary, secondary and tertiary. Within the region bounded by the Easter Microplate and the Garrett Transform Fault, the isotopic data [Mahoney et al., 1989] and minor element ratios suggest a primary segmentation with boundaries at the large 20.7øS offset and near the 16øS OSC. Geochemically, this segmentation reflects boundaries in mantle source composition. The lengths of these three segments range from 261 to 548 km.

Secondary magmatic segments defined by similar parental magma compositions vary in axial length from 11 to 185 km with an average of 69 + 57 (17) km. All identified OSCs correspond to secondary magmatic segment boundaries, but some of the magmatic segment boundaries occur at fourth- order physical discontinuities. Thus secondary magmatic segment boundaries mainly occur at second- and third-order physical discontinuities, with some occurring at fourth-order physical discontinuities. Because these segments are defined by contrasting magmatic lineages, their boundaries mark differences that most likely arise by variations in the melting processes beneath the axis. In this interpretation, their lengths define the scales over which mantle melting variations are not obliterated by crustal magma chamber processes.

Although, in our interpretation, the secondary magmatic units defined here denote the length scales of mantle melting variations, they do not necessarily define the length scale of mantle melt extraction, i.e., the spacing of melt segregation events that feed the axis. Our presently available geochemical data do not permit the resolution of different injection centers having similar major element magma compositions, only the scales over which the magma inputs vary in composition. The spacing of injection centers along axis could be less than the lengths of secondary magmatic segments, but not greater. However, it is clear from our data that there exists a tertiary magmatic organization at length scales smaller than the segments defined by similar parental magmas. For example, many of the secondary segments include short sections of ridge axis containing T-MORB, and some secondary magmatic segments show gradients in isotopic or minor element ratios. Furthermore, it is evident that a fourth-order physical segmentation of the axis occurs at a length scale averaging only 15-25 km, an interval only 1-3 times our sample spacing. Depending on how hard one squints at the SeaMARC II side scan records, about a third to a half of the tectonically identified fourth-order discontinuities correspond to secondary magmatic segments defined by similar melting histories. The other fourth-order discontinuities may simply mark the terminations of major fissure eruptions along the rise and therefore bear no relation to underlying magmatic segmentation [Macdonald et al., 1988a]. However, our results indicate that the chemical variability of most of these segments is not consistent with single eruptive events.

6150 SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES

AVE.

87Sr / 86Sr 0 70248 +.00005 I I

L K J I

L K H l G 070262 + 00006

I I I I t 16 18 20

DEGREES SOUTH LATITUDE

0 70247 + 00004

I I 22

Fig. 12. Along-axis magmatic segmentation and sample distribution along the East Pacific Rise 13ø-23øS. The nature of the primary and secondary magmatic segmentation is discussed in the text. Note that most secondary magmatic segment boundaries occur at the more pronounced bathymetric depressions in the axial depth profile.

Two fourth-order physical discontinuities mark abrupt changes in the MgO variations evident in Figure 5. An evaluation of the chemistry of lavas on either side of the 17ø05'S and 14ø30'S devals suggests that samples from opposing limbs of the devals can be related to similar parental magmas by variable extents of crystal fractionation. In other words, lava variations on either side of these boundaries define liquid lines of descent similar within analytical precision and segment variance. However, both of these devals mark significant boundaries in extent of fractionation. Lavas from north of the 14•30'S deval have uniform MgO contents of 7.6 + 0.2 wt %, values significantly higher than those to the south, which are much more variable (Figure 5). The lack of any significant change in depth or width of the axis across these boundaries suggests that the entire ridge is nearly equilibrated with respect to magmatic buoyancy forces (magmastatic equilibrium}. We therefore suggest that the 14•30'S and 17ø05'S devals represent mixing boundaries within continuous magma chambers.

Taken together, the length scales of magmatic segmentation defined by the geochemical results suggest a good, but not perfect, correspondence with physical discontinuities along the EPR 13ø-23øS (Figure 12). Primary boundaries in mantle source geochemistry generally coincide with the larger physical offsets, whereas, secondary magmatic boundaries mainly correspond to second- and third-order physical discontinuities. Many of these secondary segments correspond to well-defined bathymetric domes 50-100 km in axial length (Figure 12). Fourth-order physical discontinuities that do not mark secondary magmatic segment boundaries may correspond to the length scales of magma injection centers and/or mixing cells along axis.

Mantle Melting Variations AlongAxis

The nature of variation of the parental magma compositions feeding the secondary magmatic segments has implications for the processes controlling their composition. Parental Na20 and Al203 are strongly correlated (Figure 13}, suggesting that variable extents of partial melting can explain much of the observed variation [Jaques and Green, 1979; Klein and Langrnuir, 1987; Sinton and Fryer, 1987]. The only segment showing deviation from this correlation is the propagating rift segment G where parental Al203 is higher than the overall

trend. Although segment G samples lying in the dashed field of Figure 2 were excluded from the calculation of parental Al203 contents for this segment, its average parental Al203 is still higher than for the overall trend versus (Na20)8.0. We suspect that the persistently high (Al203)8.0 of segment G lavas is more likely to mainly reflect anomalous (high-pressure) fractionation histories involving clinopyroxene rather than alecoupling of Na and Al during melting.

The total variation of (Na20)8.0 of our data is from 2.26- 2.70 wt % (Table 2); this range encompasses most of that observed for all Pacific spreading centers [Klein and Langmuir, 1987]. This variation is only about 25% of the global variation discussed by Klein and Langmuir [1987], which they attribute to variations in extent of melting from -8-20 %. A comparison to the global data set suggests that the range in variation in extent of melting observed for magmatic segments of the EPR 13 ø- 23øS is only about 3 %, ranging from approximately 12-15 % melting.

There is a poor, inverse correlation between parental FeO* and Na20, broadly similar to the slope determined for the global data set of MORB by Klein and Langmuir [1987]. The lack of correlation between K20 and Na20, contrary to expectations from variable partial melting of a mantle with uniform Na20 and K20 contents, indicates that the K20 of parental magmas for this portion of the East Pacific Rise probably also reflects variations in the mantle source composition. This conclusion is supported by the latitudinal variation in (K20)8.0 shown in Figure 14, which mimics the variation in K•i (Figure 5).

Overall there is a poor correlation between average parental magma compositions for secondary segments and axial depth, although we note that the region between about 18øS and the Garrett, which has relatively uniform axial depth (Figures 5 and 12), shows little variation in any parental magma compositional parameter except K20 (Figure 14). South of the large offset at 20.7øS there are general trends of southward increasing (FeO*)8.0 and decreasing (Na20)8.0, (Al203)8.0 (Figure 14) and (SIO2)8.0 (not plotted) to the boundary of the Easter Microplate near 23øS. These relations are consistent with a general trend of increasing extents of deeper melting [Klein and Langmuir, 1987] south of 20.7øS, even though this effect is apparently not reflected in the axial depth. The two segments north of the Garrett have slightly lower average (FeO*)8.0 and higher (K20)8.0 than any of the segments south of it.

SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES 6151

16.0

15.5

15.0

14.5

14.0

• 10

• 9

0.12

,•, 0.09

N 0.06

'•' 0.03

QA I,,MO

G

EB E4R H CIF

L

QA •0 (• N C IF

Q N R kMo U

A CE B j C I F

0 • I • I

2.2 2.4 2.6 2.8

(Na20)•.o Fig. 13. Variation in selected calculated "parental" oxide values at 8 wt % MgO (Table 2) versus (Na20)8.0 for secondary magmatic segments shown in Figure 1. The good correlation between (A1203)8.0 and (Na20)8.0 suggests that variations in the extent of partial melting play an important role in controlling the parental magma compositions of secondary magmatic segments. The crude anticorrelation between (FeO)8.0 and (Na20)8.0 is approximately parallel to that for the global data set considered by Klein and Langmuir [1987]. Variations in (K20)8.0 and (Na20)8.0 are decoupled; Samples from south of 20.7øS (A-F) and north of Garrett (Q, R) have the lowest and highest (K20)8.0 values, respectively.

Spreading Rate Variations on Magmatic Seg•nentation

The physical segmentation of slow spreading ridges is harder to define than at faster spreading rates, mainly because the axis is much more difficult to identify precisely. Second-order discontinuities are small-offset (<20 km), nonrigid transform faults, often just bends in the rift valley [Grindlay et al., 1990]. Third-order and fourth-order discontinuities are probably intervolcano gaps between chains of volcanoes in the inner floor of the rift valley, while fourth-order discontinuities may be intravolcano gaps within a given chain. Hence it is difficult to compare the length scale of physical segmentation between fast and slow spreading centers, because of the limitations in our ability to define and detect high-order discontinuities at slow spreading [Macdonald et aL, 1988a]. Furthermore, it is harder to remove temporal magmatic variability in dredge collections from slow spreading ridges. Nevertheless, the procedure used to delimit the secondary magmatic segments defined in this paper can be applied to other ridge systems that have been sampled adequately. For example, we have determined the along axis continuity of common liquid lines of descent using data from the EPR 8.5ø-14øN [Langrnuir, 1988], the GSC near 95øW and the FAMOUS area (Figure 15). It is clear from this analysis that there is evidence to suggest that secondary magmatic segment lengths increase with spreading rate. It is also apparent that secondary magmatic segments become more variable in length with faster spreading.

Although the increase in magmatic segmentation length with spreading rate proposed here is qualitatively consistent with predictions by Schouten et al. [1985], the secondary magmatic segments are much shorter than predicted by those authors at all spreading rates. When the nature of the secondary magmatic segments defined in this paper is considered, a positive correlation between magmatic segment length and spreading rate might be expected. Because the secondary magmatic segments represent the length scales over which extents of melting are more or less uniform, these lengths apparently bear some systematic relationship to mantle temperature variations at the depths of melt segregation, or to the geometry of the asthenospheric upwelling zone beneath the axis. The width of the upwelling zone at depth can be inferred from the relationship of plate thickening with age determined by Parker and Oldenburg [1973]. At the same depth this width simply increases proportionally to spreading rate. The apparently nonlinear relationship determined for the four areas plotted in Figure 15 suggests that magmatic segment length may not be simply proportional to spreading rate. However, it is possible that the true relationship is inadequately described by the presently available data. In order to define the segment lengths in the way we have in this paper requires relatively closely spaced sampling over several segment lengths. The relationship shown in Figure 15 suggests that this sampling interval is less than 1 km at slow spreading to 5-15 km at faster spreading. As sampling studies from other areas emerge that meet these criteria the relationship may be modified, although

3.0

2.8

2.6

2.4

2.2

2.0

15.5

15.0

14.5

14.0

10

0.09

0.06

0.03

' (Na•O)•.o .

jIH F% E B

A

i , I ,

(AI20:])a.o

ML

, I I

jIH

I I I

M R

(FeO*)e.o

N o M

K JH c E I FD C

, , I I

L K H I S

J (Kz0)8.0

• , i I

12 1,.5 18 21 24

Degrees S. Latitude Fig. 14. Variation in selected calculated "parental" oxide values at 8 wt % MgO for the secondary magmatic segments of Figure 1 plotted against latitude. Segments south of 20.7øS (A-F) and north of Garrett (Q, R) are plotted as bold symbols. The decrease in (K20)8.0 with latitude mimics the variation in K/Ti (Figure 5) and can be attributed to a northward increase in K20 of the N-MORB mantle source.

6152 SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES

120

90

60

30

o

50

1• 30 EPR •-' o 13o_23os

o / • 01 I I I I I / •'

30 90 150 / SPREADING RATE • •,,•" '

(mm/yr)

EPR 8.5ø-14øNJ• • •

GSC •- 1 FAMOUS

i i i

o 5o lOO 15o

FULL SPREADING RATE (mm/yr)

Fig. 15. Average secondary magmatic segment lengths for four spreading center regions plotted against spreading rate. The data for the EPR 13ø-23øS arc from this paper; data for segments of the EPR 8.5ø-14øN arc from the East Pacific Rise Synthesis [Langmuir, 1988]. Data for the Galapagos Spreading Center near 95øW are from Christie and Sinton [1981, 1986 and unpublished, 1985-1991]; and from Bryan [1979] for the FAMOUS area. The variation in average secondary magmatic seemerit lengths for the four regions can be fit by y = 0.01X 1'7T (R 2 = 0.993), where y = average segment length in km and X = full spreading rate (mm/yr). The inset shows that the deviations from the means (1or) for the four regions also show a positive correlation with spreading rate.

we would be surprised if some sort of positive correlation between segment length and spreading rate does not survive additional data. If a linear relationship emerges from further sampling, then it can be argued that average secondary magmatic segmentation lengths vary directly with the cross section size of the zone of upwelling asthenosphere beneath the axial region.

The data for the four areas shown in Figure 15 are very well fit by a power law relationship (y = X 1.77 (R 2 = 0.993). Should such a relationship survive additional data from other areas then it must be concluded that factors in addition to the simple (two-dimensional) geometry of asthenospheric upwelling contribute to the length scales over which mantle melting is more or less uniform. Such factors might include irregular temperature variations within the melting zone, second-order variations in the along-axis geometry of the melting zone or variations in the segregation depths of melts related to different thermal geometries at different spreading rates.

Thermal and Magma Supply Variations Beneath the EPR 13ø-23øS

Superposed on the magmatic and physical segmentation of the EPR 13ø-23øS is the along-axis variation in MgO contents (Figure 5); similar variation is seen for all indices of degree of magmatic differentiation or magmatic temperatures. Five regions with relatively smooth variation in MgO contents are apparent. Most of the boundaries of these regions correspond to major physical or secondary magmatic segment boundaries, but two occur within secondary magmatic segments. Where degree of magmatic differentiation correlates with axial depth, it can be argued that magmatic temperature is primarily controlled by magma supply [Christie and Sinton, 1981]. However, the general lack of an inverse correlation between axial depth and MgO for much of the present study area

implies that crustal magmatic temperatures can be affected by factors in addition to magma supply. The possible relationship to hydrothermal cooling discussed for the Hump region illustrates one additional complicating factor.

The average length scale of the five regions apparent on Figure 5 is 217 +__ 60 (lcr) km. This value approximates those of the undulations in gravity observed off axis in the vicinity of our study area, which Haxby and Weissel [1986] suggested may result from shallow-level convection in the upper mantle that has been sheared into rolls by the motion of the overlying Pacific Plate. Winterer and Sandwell [1987] showed that the gravity highs correspond to "cross-grain" topographic ridges on the Pacific Plate, and Macdonald [1988] proposed an analogy between these highs and mantle upwelling associated with OSCs. If the gravity "rolls" are present at the rise axis, then a relation between short-wavelength upwelling of asthenosphere in a three-dimensional framework of mantle convection and

magma supply to the rise axis might be expected. Macdonald et al. [1988b] noted that a large positive gravity anomaly approaches the EPR axis between 18 ø and 19øS, more or less corresponding to the broad maxima in axial bathymetry and MgO, despite the complexities in the Hump region discussed earlier. Although the exact nature of any relationships between the off-axis gravity signal, asthenospheric upwelling and mantle convection, and variations in subaxial thermal structure and magma supply remains to be elucidated, the general correspondence between the length scale and location of axial thermal variations with the extrapolation of the off-axis gravity pattern to the axis suggests that these features may be related. Haxby and Weissel [1986] suggested that the gravity rolls may reflect secondary undulations present in the asthenophere that is being tapped passively beneath the East Pacific Rise in this region.

One result of our study concerns the relationships between the various kinds of thermal segmentation affecting East Pacific Rise magmas in this region and the structural segmentation of the rise axis. Secondary magmatic segments, which are mainly controlled by varying degrees of melting of upwelling asthenosphere bear a good, but not perfect, relationship to structural boundaries, implying a perhaps surprising relationship between average asthenospheric temperature and shallow structure of the accreting axis. However, the results from this area cannot be extended to the general case because we noted in our analysis of the secondary magmatic segmentation of the EPR 8.5ø-14øN that there is little change in parental magma composition on either side of the Clipperton Transform or the 11ø45'N OSC. The case of the Clipperton is particularly instructive because, in contrast to the uniform MgO and axial depth south of the transform, there is a strong gradient in axial depth and magmatic differentiation to the north [Langrnuir et al., 1986]. The chemical evidence suggests similar degrees of melting providing magmas to the EPR north and south of Clipperton, but with substantially different magma supply. Magma supply and extent of melting can be alecoupled if the flux of asthenosphere into the melt zone is independent of asthenospheric temperature (which, together with the length of the melting column, determines degree of melting). Applying this reasoning to the Clipperton region requires that although the extent of melting is similar on either side of the transform, the magma supply is less to the north, and therefore the amount of mantle undergoing melting (asthenospheric flux) must be correspondingly less. This result raises the interesting possibility that some transform faults may overlie shear zones in the asthenosphere separating regions flowing with different velocities.

SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES 6153

The observation that regions of the EPR 13ø-23øS with smoothly varying indices of magmatic temperature have a similar spatial scale to that of gravity rolls off axis, possibly implying a relationship between asthenospheric structure and crustal magmatic temperature, remains enigmatic. The common link to these features may be some aspect of magma supply, for which a general correspondence can be observed for the region between about 16øS and the propagating rift tip at 20.7øS. However, as noted previously, much of the local variation in degree of differentiation appears to be independent of magma supply, as indicated by axial depth or width. Taken together the data indicate that, although certain spatial and length scale correlations can be made, there are many cases where extent of melting, magma supply and degree of magmatic differentiation appear to vary independently. Taken together, the various kinds and scales of magmatic variations observed for this superfast spreading ridge suggest that temperature of the upwelling asthenosphere, magma supply to the axis, and crustal magmatic temperature reflect independent, regionally alecoupled processes.

CONCLUSIONS

1. The average composition of lavas erupted at the superfast spreading EPR 13ø-23øS is moderately differentiated ferrobasalt with about 11 wt % FeO*. T-MORB lavas, with slight enrichments in K20 and some other incompatible elements, are variably erupted along axis with a preference for eruption near ridge axis offsets; they represent melting of mantle containing discrete heterogeneities in the subaxial region. There is a more or less monotonic, twofold increase in K/Ti of erupted N-MORB from 23øS to 13øS, which can be related to a northward increase in this ratio in the mantle

source for N-MORB.

2. The large offset near 20.7øS carries the distinctive fractionation signature of long-lived propagating rifts, consistent with long-term, net propagation of this offset to the south. The generally differentiated nature of samples from abandoned ridges associated with this offset suggests that ridge abandonment in this area is favored by successive propagation events on single ridges along the northern and southern limbs of the offset.

3. Off-axis seamounts on the Pacific Plate between 18 ø and

19øS are of two types. Small seamounts generally aligned parallel to the absolute motion of the Pacific Plate are composed of MORB-type basalts highly depleted in incompatible elements. Larger seamounts, generally aligned parallel to relative motion directions are widely variable in composition but consistently enriched in incompatible elements relative to the nearby ridge lavas.

4. The highly fractionated nature of lavas recovered from a short axial segment with a well-developed axial summit graben between 18B7'S and 18ø22'S suggests that there may be a relationship between subaxial magmatic temperature and structural style of the axial region, time since replenishment of the subaxial magma reservoir, and/or vigor of hydrothermal activity.

$. The EPR 13ø-23øS is magmatically segmented on various scales. A primary segmentation corresponding to the major structural offsets in the axis can be related to regions of the mantle with distinct compositions. Secondary magmatic segmentation is defined by the length scales of samples with similar liquid lines of descent, which mainly can be related to variations in extent of melting beneath the axis. There is a good, but not perfect, relationship between secondary

magmatic segmentation and the physical segmentation of the axis with all first-, second- and third-order physical segment boundaries corresponding to secondary magmatic segment boundaries. A tertiary magmatic organization is indicated, but not well defined by our sampling program. Some fourth-order physical discontinuities apparently represent the limits of mixing cells in subaxial magma reservoirs. The length scale of magma inputs to the crust must be smaller than the secondary magmatic segment scale and may correspond to the scale of tertiary magmatic organization.

6. There is a positive correlation between average secondary magmatic segment length and spreading rate for four areas ranging from 20 mm/yr to 150 mm/yr, suggesting a correspondence between the length scale over which mantle melting variations are uniform and the geometry of the upwelling asthenosphere beneath the axis.

7. Superposed on the chemical magmatic segmentation are regions with smoothly varying indices of magmatic differentiation. These regions generally correspond to the positions and length scale of undulations in gravity observed off axis. The apparent alecoupling between the secondary magmatic segments (controlled by variations in extent of melting), axial depth, and the regions defined by smoothly varying magmatic temperature gradients suggests that average asthenospheric temperature in the melting region, magma supply to the crust, and crustal magmatic temperature can reflect independent, regionally alecoupled processes.

Acknowledgments. We are indebted to the Captain and crew of RN Moana Wave Cruise MW8712, whose collective competence and professional operations at sea made a succ•:ssful field program both possible and pleasurable. We would especially like to thank Dave Gravatt for tireless help with the dredging operations, Will Hervig for help with the camera system, Maile Sakamoto for artistic expertise, and the rest of the S & M team, Grant Gribble, John Hoffmann, Dan Johnson, Patty Lee, Lori Liu, Jill Wessel, Ruth Multhaup, JoAnn Sinton, Steve Spengler, Frank Trusdell, Bruce Tsutsui and Gall Yamada for exceptional diligence in describing the rock collections' and preparing over 550 rock powders for geochemical analysis at sea. The generosity of Roger Hekinian (IFREMER) and Harald Bllcker (Preussag), who donated samples from the area, is gratefully acknowledged. We thank Jeff Fox for permission to use unpublished Seabeam charts of much of the ridge and Garrett Fracture Zone, Bill Bryan for his program used to calculate CMAS end-members, Marie-Helene Cormier for help in producing the bathymetric map of the Hump area, Guy Waggoner for preliminary isotope dilution mass spectrometric analyses of Sr on Hump seamounts, Tina Mueller and Warren Ulmer for help with sample preparation, Adam Weiner for some microprobe analyses and Brooks Bays and Nancy Hulbirt for drafting. Discussions with Rodey Batiza, John Bender, Bill Bryan, Rhett Butler, Charlie Langmuir, Anton leRoex and Brian Taylor contributed to our thinking on the problems of magmatic segmentation. Reviews by Jamie Allan and John Bender led to substantial improvements in the manuscript. This research was supported by National Science Foundation grants OCE86-08805 and OCE86-09706. This is University of Hawaii SOEST contribution 1980.

REFERENCES

Autio, L. IC, and J. M. Rhodes, Costa Rica Rift Zone basalts: Geochemical and experimental data from a possible example of multistage melting, in Initial Report DSDP, 69, 729-745, 1983.

6154 SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES

Bficker, H., J. Lange, and V. Marchig, Hydrothermal activity and sulphide formation in axial valleys of the East Pacific Rise crest between 18 and 22øS, Earth Planet. Sci. Lett. 72, 9-22, 1985.

Bass, M. N., Occurrence of transitional abyssal basalt, Lithos 5, 57-67, 1972.

Batiza, R., Seamounts and seamount chains of the eastern Pacific, in Decade of North American Geology, The Eastern Pacific Ocean and Hawaii, edited by E. L. Winterer, D. M. Hussong, and R. W. Decker, Geological Society of America, Boulder, Colo., 1989.

Batiza, R., and D. Vanko, Petrology of young Pacific seamounts, J. Geophys. Res. 89, 11,235-11,260, 1984.

Bender, J. F., C. H. Langmuir, and G. N. Hanson, Petrogenesis of basalt glasses from the Tamayo region, East Pacific Rise, J. Petrol., 25, 213-254, 1984.

Bryan, W. B., Regional variation and petrogenesis of basaltic glasses from the FAMOUS area, Mid-Atlantic Ridge, J. Petrot, 20, 293- 429, 1979.

Byerly, G. R., W. G. Melson and P. R. Vogt, Rhyodacites, andesites, ferrobasalts and ocean tholeiites from the Galapagos spreading center, Earth Planet. Scœ Lett., 30, 216-221, 1976.

Christie, D. M., and J. M. Sinton, Evolution of abyssal lavas along propagating segments of the Galapagos spreading center, Earth Planet. Scœ Lett., 56, 321-335, 1981.

Christie, D. M., and J. M. Sinton, Mantle processes in the generation of MORB from the 95øW Galapagos propagating rift area, Contrib. Mineral PetroL, 94, 274-288, 1986.

Clague, D. A., F. A. Frey, G. Thompson, and S. Rindge, Minor and trace element geochemistry of volcanic rocks dredged from the Galapagos spreading center: Role of fractional crystallization and mantle heterogeneity, J. Geophys. Res., 86, 9469-9482, 1981.

DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein, Current plate motions, Geophys. J. Int., 101,425-478, 1990.

Engeln, J. F., S. Stein, J. Werner, and R. G. Gordon, Microplate and shear zone models for oceanic spreading center reorganizations, J. Geophys. Res., 93, 2839-2856, 1988.

Fisk, M. R., A. E. Bence and J-G. Schilling, Major element chemistry of Galapagos rift zone magmas and their phenocrysts, Earth Planet. ScL Lett., 61,171-189, 1982.

Fornari, D. J., M. R. Perfit, A. Malahoff and R. Embley, Geochemical studies of abyssal lavas recovered by DSRV ALVIN from eastern Galapagos rift, Inca transform and Ecuador rift: 1. major element variations in natural glasses and spatial distribution of lavas, J. Geophys. Res., 88, 10,519-10,529, 1983.

Fornari, D. J., M. R. Perfit, J. F. Allan, R. Batiza, R. Haymon, A. Barone, W. B. F. Ryan, T. Smith, T. Sirekin and M. A. Luckman, Geochemical and structural studies of the Lamont seamounts:

seamounts as indicators of mantle processes, Earth Planet. Scœ Lett., 89, 63-83, 1988.

Fornari, D. J., D. G. Gallo, M. H. Edwards, J. A. Madsen, M. R. Perfit, and A. N. Shor, Structure and topography of the Siqueiros Transform Fault system: Evidence for the development of intra- transform spreading centers, Mar. Geophys. Res., 11,263-299, 1989.

Fox, P. J., and D. Gallo, Garrett Fracture Zone in Decade of North American Geology, The Eastern Pacific Ocean and Hawaii, edited by E. L. Winterer, D. M. Hussong, and R. W. Decker, pp. 111-124, Geological Society of America, Boulder, Colo., 1989.

Francheteau, J., and R. D. Ballard, The East Pacific Rise near 21øN, 13øN and 20øS: Inferences for along-strike variability of axial processes of the mid-ocean ridge, Earth Planet. Scœ Lctt., 64, 93-116, 1984.

Grindlay, N. R., P. J. Fox and K. C. Macdonald, Second- order ridge axis discontinuities in the south Atlantic: Morphology, structure, evolution and significance, Mar. Geophys. Res., in press, 1990.

Haxby, W. F., and J. K. Weissel, Evidence of small-scale convection from Seasat altimeter data, J. Gcophys. Res., 91, 3507-3520, 1986.

Hekinian, R., Chemical and mineralogical differences between abyssal hill basalts and ridge tholeiites in the eastern Pacific Ocean, Mar. Geol., 11, 77-91, 1971.

Herron, E. M., Two small crustal plates in the South Pacific near Easter Island, Nature Phys. ScL, 240, 35-37, 1972.

Hey, R. N., Tectonic evolution of the Cocos-Nazca spreading center, Bult Geol. Soc. Am., 88, 1404-1420, 1977.

Hey, R. N., J. M. Sinton, and F. K. Duennebier, Propagating rifts and spreading centers, in Decade of Nonh American Geology, The Eastern Pacific Ocean and Hawaii, edited by E. L. Winterer, D. M. Hussong, and R. W. Decker, pp. 161-176, Geological Society of America, Boulder, Colo., 1989.

Jaques, A. L., and D. H. Green, Anhydrous melting of peridotite at 0- 15 kb pressure and the genesis of tholeiite basalts, Contrib. Minerat Petrol., 73, 287-310, 1979.

Jaupart, C., and G. Brandeis, The stagnant bottom layer of convecting magma chambers, Earth Planet. Sci. Lett., 80, 183-199, 1986.

Klein, E. M., and C. H. Langmuir, Global correlations of ocean ridge basalt chemistry with axial depth and crustal thickness, J. Geophys. Res., 92, 8089-8115, 1987.

Langmuir, C. H., Petrology database, Vol. 2 and Vol. 3, in East Pacific Rise Data Synthesis Final Report, edited by S. Tigre, Joint Oceanographic Institutions, Inc. (publ.), Washington, D.C., 1988.

Langmuir, C. H., and G. N. Hanson, An evaluation of major element heterogeneity in the mantle sources of basalts, Philos. Trans. R: Soc. London, $er. A, 297, 383-407, 1980.

Langmuir, C. H., J. F. Bender, and R. Batiza, Petrological segmentation of the East Pacific Rise, 5ø30'-14ø30'N, Nature, 322, 422-429, 1986.

Lonsdale, P., Structural geomorphology of a fast-spreading rise crest: the East Pacific Rise near 3025 'S, Mar. Geoœhys. Res., 3, 251-293, 1977.

Lonsdale, P., Site surveys in the south Pacific with a Seabeam swath- mapping system, in Initial Report DSDP 92, edited by M. Leinen, D. K. Rea, et al., pp. 229-236, U.S. Government Printing Office, Washington, D.C., 1986.

Lonsdale, P., Segmentation of the Pacific-Nazca spreading center, IøN-20øS, J. Geophys. Res., 94, 12,197-12,225, 1989.

Macdonald, K. C., Cracks in the Pacific Plate and mantle convection, Nature, 331,395, 1988.

Macdonald, K. C., J.-C. Sempere, and P. J. Fox, East Pacific Rise from Siqueiros to Orozco Fracture Zones: Along strike continuity of axial neovolcanic zone and structure and evolution of overlapping spreading centers, J. Geophys. Res., 89, 6049-6069, 1984.

Macdonald, K. C., J.-C. Sempere, P. J. Fox and R. Tyce, Tectonic evolution of ridge-axis discontinuities by the meeting, linking, or self-decapitation of neighboring ridge segments, Geology, 15, 993- 997, 1987.

Macdonald, K. C., P. J. Fox, L. J. Perram, M. F. Eisen, R. M. Haymon, S. P. Miller, S. M. Carbotte, M.-H. Cormier, and A. N. Shor, A new view of the mid-ocean ridge from the behaviour of ridge-axis discontinuities, Nature, 335, 217-225, 1988a.

Macdonald, K. C., R. M. Haymon, S. P. Miller, J.-C. Sempere, and P. J. Fox, Deep-Tow and Sea Beam studies of dueling propagating ridges on the East Pacific Rise near 20ø40'S, J. Geophys. Res., 93, 2875-2898, 1988b.

Macdonald, K. C., and M.-H. Cormier, The East Pacific Rise 3ø-23øS, Mar. Geophys. Res., in press, 1991.

Macdougall, J. D., and G. W. Lugmair, Sr and Nd isotopes in basalts from the East Pacific Rise: Significance for mantle heterogeneity, Earth Planet. Sci. Lett., 77, 273-284, 1986.

Mahoney, J. J., J. M. Sinton, K. J. Spencer, and S. M. Smaglik, Isotopic characteristics of a superfast spreading ridge: EPR 13ø-23øS (abstract), Eos Trans. AGU, 70, 1317, 1989.

Marsh, B. D., Magma chambers, Ann. Rev. Earth Planet. Sci., 17, 439- 474, 1989.

Melson, W. G., T. L. Vallier, T. L. Wright, G. Byerly, and J. Nelen, Chemical diversity of abyssal volcanic glasses erupted along Pacific, Atlantic and Indian Ocean seafloor spreading centers, in The Geophysics of the Pacific Ocean Basin and Its Margin, Geophys. Monogr. $er., vol. 19, edited by G. H. Sutton., et al., pp. 351-367, AGU, Washington D.C., 1976.

Naar, D. F., and R. N. Hey, Fast rift propagation along the East Pacific Rise near Easter Island, J. Geophys. Res., 91, 3425-3438, 1986.

Naar, D. F., and R. N. Hey, Recent Pacific-Nazca plate motions, in The Evolution of Mid-Ocean Ridges, Geophys. Monogr. Ser., vol. 57, IUGG vol. 8, edited by J. M. Sinton, pp. 9-30, AGU, Washington, D. C., 1989.

Natland, J. H., and W. G. Melson, Compositions of basaltic glasses from the East Pacific Rise and Siqueiros Fracture Zone, near 9øN, Initial Report DSDP, 54, 705-723, 1980.

Oldenburg, C. M., F. J. Spera, D. A. Yuen, and G. Sewell, Dynamic mixing in magma bodies: Theory, simulations and implications, J. Geophys. Res., 94, 9215-9236, 1989.

Parker, R. L., and D. W. Oldenburg, Thermal model of ocean ridges, Nature Phys. Scœ, 242, 137-139, 1973.

Perfit, M. R., and D. J. Fornari, Geochemical studies of abyssal lavas recovered by DSRV ALVIN from eastern Galapagos rift, Inca

SINTON ET AL.: SUPERFAST SPREADING MID-OCEAN RIDGES 6155

transform and Ecuador rift; 2, Phase chemistry and crystallization history, J. Geophys. Res., 88, 10,530-10,550, 1983. Puchelt, H., and R. Emmermann, Petrogenetic implications of

tholeiitic glasses from the East Pacific Rise and the Galapagos spreading center, Chern. GeoL, 38, 39-56, 1983.

Rea, D. K., Asymmetric sea-floor spreading and a nontransform axis offset: The East Pacific Rise 20øS survey area, Bull GeoL Soc. Am., 89, 836-844, 1978.

Rea, D. K., Tectonics of the Nazca-Pacific divergent plate boundary, in Nazca Plate: Crustal Formation and elndean Convergence, edited by L. D. Kulm et al., pp. 27-62, GeoL Soc. elm. Mem-, vol. 152, Geological Society of America, Boulder, Colo., 1981.

Renard, V., R. Hekinian, J. Francheteau, R. D. Ballard, and H. Biicker, Submersible observations at the axis of the ultra-fast spreading East Pacific Rise (17ø.30 ' to 21ø30'S), Earth Planet. Sci. Lett., 75, 339-353, 1985.

Scheidegger, K. F., and J. B. Corliss, Petrogenesis and secondary alteration of upper layer 2 basalts of the Nazca Plate, p. 77-108 in Nazca Plate: Crustal Formation and elndean Convergence, GeoL Soc. elm. Mern., vol. 152, edited by L. D. Kulm et al., Geological Society of America, Boulder, Colo., 1981.

Schouten, H., K. D. Klitgord, and J. A. Whitehead, Segmentation of mid-ocean ridges, Nature, 317, 225-229, 1985.

Sinton, J. M., and P. Fryer, Mariana Trough lavas from 18øN: Implications for the origin of back arc basin basalts, J. Geophys. Res., 92, 12,782-12,802, 1987.

Sinton, J. M., D. S. Wilson, D. M. Christie, R. N. Hey, and J. R. Delaney, Petrologic consequences of rift propagation on oceanic spreading ridges, Earth Planet. Sci. Lett., 62, 193-207, 1983.

Sleep, N.H., and B. R. Rosendahl, Topography and tectonics of mid- ocean ridge axes, J. Geophys. Res., 84, 3831-.3839, 1979.

Thompson, G., W. B. Bryan, R. D. Ballard, K. Hamuro, and W. G. Melson, Axial processes along a segment of the East Pacific Rise, 10-12øN, Nature, 318, 429-433, 1985.

Thompson, G., W. B. Bryan and S. E. Humphris, Axial volcanism on the East Pacific Rise, 10-12øN, in Magmatism in the Ocean Basins, edited by A. D. Saunders and M. J. Norry, GeoL Soc. London Spec. PubL, 42, pp. 181-200, 1989.

Tighe, S. A., R. S. Detrick, P. J. Fox, and R. C. Tyce, East Pacific Rise Data Synthesis, Section VI, VII, in East Pacific Rise Data Synthesis and Final Report, edited by S. A. Tighe, Joint Oceanographic Institutions, Inc. (publ.), Washington, D.C., 1988.

Walker, D., T. Shibata, and S. E. DeLong, Abyssal tholeiites from the Oceanographer Fracture Zone, II, Phase equilibria and mixing, Contrib. Mineral PetroL, 70, 111-125, 1979.

Whitehead, J. A., H. J. B. Dick, and H. Schouten, A mechanism for magmatic accretion under spreading centers, Nature, 312, 146-148, 1984.

Winterer, E. L., and D. T. Sandwell, Evidence from en-echelon cross- grain ridges for tensional cracks in the Pacific Plate, Nature, 329, 5.34-537, 1987.

Yonover, R. N., Petrological Effects of Rift Failure at the Galapagos Spreading Center near 95øW, PhD dissertation, 193 pp., University of Hawaii, Honolulu, Hawaii, 1989.

K. C. Macdonald, Department of Geological Sciences, University of California, Santa Barbara, CA 93106.

J. J. Mahoney, J. M. Sinton, S. M. Smaglik, Department of Geology and Geophysics, University of Hawaii at Manoa, Honolulu, HI 96822.

(Received January 3, 1990; revised November 5, 1990;

accepted November 5, 1990.)