Sr, Nd and Pb Isotopic Variation Along the Pacific–Antarctic Risecrest, 53

Earth and Planetary Science Letters 154Ž. 1998 109±125

Sr, Nd and Pb isotopic variation along the Pacific±Antarctic risecrest, 53±578S: Implications for the composition and dynamics of the South Pacific upper mantle

Paterno R. Castillo a,), James H. Natland b, Yaoling Niu c, Peter F. Lonsdale a a Scripps Institution of Oceanography, UniÕersity of California, San Diego, 9500 Gilman DriÕe, La Jolla, CA 92093-0220, USA b Rosenstiel School of Marine and Atmospheric Science, UniÕersity of Miami, Miami, FL 33149, USA c Department of Earth Sciences, The UniÕersity of Queensland, Brisbane, Qld. 4072, Australia Received 22 October 1996; revised 16 September 1997; accepted 26 September 1997

Abstract

Sr, Nd and Pb isotope data for basalts from spreading axes and off-axis volcanoes near the Pacific±Antarctic risecrest, from Vacquier transform to just south of Udintsev transform, reveal an isotopically heterogeneous upper mantle. The isotopic composition of the mantle is represented by three end-members:Ž. 1 the `depleted' source of the bulk of Pacific normal-type mid-ocean ridge basaltsŽ.Ž. N-MORB ; 2 an `enriched' source that produces basalts of the Hollister Ridge; and Ž.3 a source, restricted to two adjacent sample locales, similar to that of Indian MORB. The distribution of these isotopic heterogeneities along the Pacific±Antarctic risecrest suggests two alternative hypotheses on the nature and dynamics of the south Pacific upper mantle. The whole area could be a single N-MORB mantle domain that shows a weak but continuous increase in 143Ndr 144Nd from northeast to southwest across more than 2000 km of sea floor. The gradient is unrelated to the Louisville hotspot because Louisville basalts have low 143Ndr 144Nd and the hotspot's influence along the ridge is spatially limited and near the high 143Ndr 144 Nd southwestern end of the gradient. The gradient appears consistent with a southwestward flow of the Pacific N-MORB-type mantle that has been proposed mainly on the basis of ridge morphology. That the N-MORB mantle domain is continuous across Heezen suggests that large-scale magmatic segmentation is not related to the largest structural offsets of the Pacific ridges. Alternatively, the higher 87Srr 86Sr, DNd and D8r4 of samples from southwest of the Heezen transform relative to those from the northeast could result from southwestward pumping of both plume and Indian Ocean-type mantle material by the Louisville hotspot. The Heezen transform forms a prominent tectonic and mantle domain boundary that prohibits the Louisville- and Indian Ocean-type mantle from flowing towards and contaminating the depleted Pacific-type source in the northeast. q 1998 Elsevier Science B.V.

Keywords: Sr-87rSr-86; Nd-144rNd-143; Pb-208rPb-204; isotope ratios; petrology; mid-ocean ridge basalts; Pacific-Antarctic Ridge; East Pacific Rise; South Pacific

1. Introduction

) Corresponding author. Fax: q1 619 534 0784. E-mail: Mid-ocean ridge basalts are produced by decom- [email protected] pression melting of mantle welling up beneath di-

Fig. 1. Along-axis bathymetric profile of the southern EPR and northern PAR in meters below sea level, plotted versus distance from the Vacquier transform fault. Note the overall shallowing of the ridge axis from the Vacquier transformŽ. northeast to south of the Udintsev transformŽ. southwest . Transform faults are indicated by solid lines, major non-transform ridge offsets by dashed lines, sampling stations by arrows and numbers. Inset shows location of the study areaŽ. shaded , bounded by the Vacquier transform in the northeast and the Udintsev transform in the southwest. verging plate boundaries. The composition of MORB the great petrologically unknown regions of the global and the morphologic and geophysical characteristics spreading ridge systemŽ. Fig. 1 . In this paper we of mid-ocean ridges are widely considered to be present Sr, Nd and Pb isotopic results on basaltic relatedŽ e.g.wx 1±5. . Exploration of these relation- glasses obtained during our sampling. Our objectives ships has up until now been concentrated on the are to constrain the composition of the MORB source more accessible portions of the global mid-ocean along the Pacific±Antarctic risecrest, and to relate ridge system in northern and tropical latitudes, par- these compositions to the scale and dynamics of the ticularly in the Pacific. Investigations of ridge seg- upper mantle in this region of the southern Pacific. ments at far southern latitudes are relatively fewŽ e.g. wx6±8. . Our Seabeam mapping and rock sampling of ad- 2. Geologic setting joining parts of the East Pacific RiseŽ. EPR and Pacific±Antarctic RidgeŽ. PAR in the far southern Our study areaŽ. Fig. 1 includes 950 km of Pa- Pacific during the austral summer of 1994wx 9±11 cific±Antarctic spreading axis subdivided by 6 represent the first systematic investigation of one of right-stepping transforms that have a combined length P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 111 of 1500 km. The two longest and most closely deeper than average. The 2035 m summit of the PAR spaced transforms, Heezen and Tharp, are often axial ridge at 558S is slightly shallower than any grouped together as the Eltanin fault system; this 850 EPR axial ridge, and in the eastern Pacific is topped km long shear zone is one of the longest breaks in only by the Cocos±Nazca spreading center in the the global system of spreading centers, and is the vicinity of the Galapagos hotspotwx 18 . The axis is traditional boundary between the EPR and PAR. shallow not because the axial ridge is unusually high, The crustal accretion rate along this central part of but because an axial ridge of average height rises the Pacific±Antarctic plate boundary, calculated from above a broad regional shoaling of the northern the plate rotation model of de Mets et al.wx 12 and the PAR. timescale of Cande and Kentwx 13 , decreases south- Most of our samples were collectedŽ by chain bag ward from 83 mmryr near Vacquier transformŽ. 538S dredges and wax-covered corers. from the floors of to 75 mmryr near Udintsev transformŽ. 56±578S. axial rift valleys, the crests of axial ridges, or the These are ìntermediate' spreading rates, comparable òvershot ridge' prolongations of axial ridges around to those on the northern part of the Pacific±Cocos transform-fault intersectionswx 8 . We also sampled boundarywx 14 and the central Indian±Antarctic young lavas erupted off-axis, at a chain of volcanoes boundarywx 15 . As at those other intermediate-rate and a volcanic ridge on the Pacific plate. The vol- spreading centers, there is much variation in the canic chainŽ. station 57 is in the ìnside corner' structural geomorphology of the risecrest. In the between the southernmost EPR axial ridge and faster-spreading northern half of our study area, most Heezen transform, and erupted along a fracture EPR spreading segments have axial ridges similar to opened by a small component of extension across the those on fast-spreading risecrests but there are short transformwx 8 . Stations 21 and 22 sampled the south- axial rift valleys immediately south of the Vacquier eastern end of the Hollister Ridge, a 1±3 km high, transformŽ. sampling station 64, Fig. 1 and within 450 km long volcanic ridge that extends obliquely the Raitt transformŽ. station 60wx 8 . The 55 km long down the western flank of the PAR, from within 10 mid-Eltanin spreading center, between the Heezen km of the 558S axial-ridge summit, across the Hollis- and Tharp transforms, also has an axial ridgewx 16 . ter fracture zone, onto ;15 Ma crust south of Tharp Along the northernmost PAR, we found, by multi- fracture zone. Hollister Ridge was discovered decades beam bathymetry during our sampling cruise, that ago with soundings of less than 250 m by U.S.N.S. the segment immediately north of the Udintsev trans- Eltanin wx19 but its extent, shape and strikeŽ. 3138 form has a 1 km deep axial rift valley, the one was defined more recently, by satellite altimetryŽ e.g. midway between the Hollister and Herron transforms wx20. . Altimetry also indicates that two or more lower at 558S has a 250 m high axial ridge, and most of the ridges parallel the main ridge, 30±60 km to the others have the low relief rifted morphology once southwest. We surveyed and sampled the southeast thought typical of intermediate spreading rateswx 17 . end of the Hollister Ridge, 10±75 km from the We assume, following interpretations of other inter- spreading axis, where it is a narrow line of coalesced mediate-rate spreading centersŽ e.g.wx 5,14,15. , that volcanos. In this respect, and in overall geomorphol- these structural variations result from inter-segment ogy and setting, Hollister Ridge seems similar to differences in the rate of magma supply, and that many other volcanic ridges that have been described axial ridges indicate the presence of long-lived crustal elsewhere on young, fast-moving Pacific lithosphere, magma chambers, whereas axial rift valleys indicate notably near the Pacific±Nazca boundary at 13±198S their absencewx 3 . Že.g.wx 21. . Whether these ridges grow in response to There is a range of almost 2 km in the depth regional extensional stresses in the Pacific platewx 22 below sea-level of the spreading axis in our study or to a secondary circulation in the underlying as- area, from just over 2000 m at the summit of a broad thenospherewx 23 , a common feature of all explana- hump in the long-profileŽ. Fig. 1 near 558S Ž station tory hypotheses is that the ridge azimuth is parallel 20 and 23. , to almost 4000 m in the rift valley north to the motion of the plate over the asthenosphere. of the Udintsev transformŽ. stations 14 and 15 . The For Hollister Ridge, this condition holds only if the short, isolated intra-Raitt and mid-Eltanin axes are underlying asthenosphere is moving southward in the 112 P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 hotspot reference frameŽ. see discussion below be- 3. Samples and methodology cause the ridge strikes about 178 oblique to Gripp and Gordon'swx 24 estimate of Pacific motion with Samples from the ridge axes, overshot ridges and respect to the hotspotswx 25 . Wessel and Kroenke wx 26 intratransform centers include both primitive and have suggested that this estimate of Pacific motion is moderately differentiated tholeiitic basalts, a ferroan- quite wrong, and that for the past 3 m.y. the Pacific desite and a dacite. Off-axis samples include tholei- plate in the vicinity of Hollister Ridge has been itic basalts and a few mildly alkalic basalts from movingŽ. relative to the hotspots not 2968 wx24 , but Hollister Ridgewx 10±12 . An important feature of almost due northŽ. 0038 . This hypothesized Pacific Pacific±Antarctic risecrest basalts is their overall motion, with a 708 change in direction at 3 Ma, similarity in their chemical composition with those places the summit of Hollister Ridge over the hotspot from the better sampled intermediate- to fast-spread- responsible for the Louisville seamount chain; more ing sections of the EPR north of ;308S, henceforth conventional estimates of plate motion locate this referred to as the `northerly EPR'. Consistent with hotspot ;350 km to the northeast, near where previous studiesŽ e.g.wx 29,30. , we have chosen to Lonsdalewx 27 surveyed and sampled a large Pleis- subdivide our basalt samples into geochemical groups tocene seamount with distinctively Louisville geo- using KrTi ratio. Basalts from the Pacific±Antarctic chemistrywx 28 . We reject Wessel and Kroenke's wx 26 risecrest, similar to those from the northerly EPR, `relocation' of the hotspot on tectonic grounds be- are dominated by depleted, low KrTi Ž.-0.14 N- cause the geography of the Pacific±Antarctic rise- MORB. There are a few transitional-typeŽ. T- MORB crest is inconsistent with their hypothetical Pacific Ž.KrTi)0.14 along the Pacific±Antarctic risecrest motionŽ there is no evidence of such a radical re- although these are fairly common in the northerly alignment of spreading axes and transforms at 3 Ma. , EPRŽ e.g.wx 1,29,30. . In the Pacific±Antarctic rise- and the geomorphology and pattern of Hollister ridge crest, T-MORB grade compositionally into enriched and its partners are quite unlike typical manifesta- Ž.KrTi)0.56 , mildly alkalic basalts from the south- tions of the LouisvilleŽ. or any other hotspot. A more eastern end of Hollister Ridge. Some near-ridge relevant question, which we will address below, is seamounts along the northerly EPR are also capped whether a Louisville hotspot located hundreds of with alkalic basaltsŽ e.g.wx 31. . Finally, a few compo- kilometers to the north beneath the EPR flankwx 27 sitionally distinct, highly primitive and extremely could increase the mantle temperature and magma depleted basalts were collected from the Raitt intra- supply of the northern PAR, and thereby explain its transform spreading center. These samples are chem- shallow depth and the prolific volcanism evidenced ically similar to those from the Garret and Siquieros by an axial ridge along the spreading center and high intratransform domainswx 32±34 . ridges along rise-flank fractures. Samples analyzed for Sr, Nd and Pb isotopic

Note to Table 1: All Sr and Nd isotope analyses were done at the Scripps Institution of Oceanography; all Pb isotope analyses were made at the San Diego State UniversityŽ. Dr. B. Hanan, analyst . " values are in-run precisions and refer to the last digits of the isotopic ratios. Multiple determinations of NBS 987Ž87 Srr 86 Srs0.71025. give analytical uncertainty of "0.000020 for87 Srr 86Sr; multiple determinations of the La Jolla Nd standardŽ143 Ndr 144 Nds0.511860. give analytical uncertainty of "0.000016 for143 Ndr 144 Nd; analytical uncertainties on Pb isotopic ratios, based on error propagation of the error in NBS 981 values and the analytical precision errors in the sample determination, are approximately "0.005 for 206 Pbr 204 Pb and 207 Pbr 204 Pb and "0.01 for 208 Pbr 204 Pb. Sr isotopic ratios were fractionation-corrected to 86 Srr 88 Srs0.1194; Nd isotopic ratios were measured in oxide forms and fractionation corrected to 146 NdOr 144 NdOs00.72225 Ž146Ndr 144 Nds0.7219. ; Pb isotopic ratios were corrected for mass fractionation based on average measured values for NBS 981 using the values of Todt et al.wx 54 . Sample preparation was done at Scripps. Total procedural blanks are -20 pg for Sr and -5 pg for Nd Ž;30 mg samples.Ž. and -0.5 ng for Pb 300±400 mg samples . Trace element data acquired through ICP±MS at Scripps using the procedure described in Janney and Castillowx 35 . Reproducibility for the method is generally better than 4% based on repeated analyses of international standards AGV-1 and BCR-1 and of an in-house standard of an EPR basalt. LarSmŽ. chon schondrite-normalizedwx 52 LarSm values.

Major element data acquired through electron probe at the University of Queensland. TiO22 and Na O fractionation-corrected values are for major element contents corrected for fractionation to 8 wt% MgO using the method described in Niu and Batizawx 2 . P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 113 88 Sm Ti Na r basalts Ti La r ) Ž.Ž. Ti Ba r K " r Pb " r Pb " r Pb " r Nd " Ž. Ž. Ž. Ž. Ž. r 87 143 206 207 208 86 144 204 204 204 Hollister Ridge South West Sr Nd Pb Pb Pb 100 chon Ž.Ž. r Normal-MORB Sample Lat. Long. Sr Table 1 Sr, Nd and Pb isotopic ratios, incompatible element ratios and fractionation-corrected major element composition of representative south Pacific D1-2D7-7-1D14-1 55.90 56.02D16-7A 56.11 144.91D17-2A 143.87 56.68 0.702502D18-1 56.27 0.702536 140.17 139.67 19GC32 14 0.702442 56.25 139.31 0.702433D41-2 18 0.513113 53.92 0.702423 0.513158 18D42-1A 138.88 9 55.33 14 8D53-1B 55.10 0.513173 0.702678 134.57 0.513199D56-1 55.76 11 127.76 10 0.513177 0.702787 8 127.61D58-1 0.702352 9 14 55.67 122.89 18.203 0.702336D60-1A 0.513076 16 17.981 55.15 0.702487 16D60-2 54.43 0.001 9 0.513152 121.44 18.297 17 0.001D61-1 0.513174 15.448 7 121.14 0.702512 0.513165 54.43 119.35 0.004D62-1A 15.445 7 18.732 0.702524 11 0.001 0.513103 7 54.45 0.702945D63-1A 15.484 54.01 119.35 17.684 0.001 16 0.002 37.618 10 22D64-2 53.67 0.513109 118.42 0.003 0.702954 37.363 18.039 15.501 0.008 118.14 0.002D64-3B 0.513090 9 18.444 0.702452 37.815 20 0.512991 53.13 117.97 15.460 0.003 0.702441 0.001 0.002Transitional-MORB 0.059 53.13 10 21 9 0.004 0.008 0.702540 0.067 14D31-1B 38.184 15.433 0.007 0.512987 0.09 117.70 15.493 117.70 0.054 14 0.08D23-1 0.513090 54.27 37.216 6 0.003 0.002 0.702415 0.513111 18.497 0.702418 0.05D23-3E 0.003 0.48 9 0.080 37.363 23 0.016 0.513065 8 54.88 134.89 0.49 13 0.002D59-1 37.761 54.88 18.508 0.16 0.120 9 0.005 0.702852 0.45Seamounts 15.560 0.513090 1.33 137.67 0.008 0.001 0.513106 54.31 0.40 137.67 0.076 20 1.27 8 2.56 0.002 0.702824 0.119 0.79 18.882 15.542 7 0.703508 0.07 1.33 2.87 120.77 37.838 24 0.513163 0.15 0.65 0.001 0.002 14 2.58 18.805 0.702417 12 1.65 0.004 0.56 37.764 15.538 0.513060 12 0.003 0.78 0.512978 2.63 1.17 0.052 9 17.852 0.002 0.001 15.506 7 2.46 0.117 0.03 1.33 0.513122 0.108 0.054 38.194 0.002 1.46 0.003 0.18 2.87 9 0.14 0.03 15.441 18.768 0.003 2.69 0.35 38.075 0.110 0.73 0.001 0.002 0.71 0.007 0.33 0.27 1.53 37.303 15.536 0.057 1.28 2.97 0.003 1.17 0.001 0.10 1.52 0.89 2.60 0.074 0.154 2.78 38.377 2.92 0.62 0.08 0.38 0.003 1.47 0.210 2.73 0.101 0.58 0.75 1.60 0.53 0.112 0.13 2.35 0.15 1.36 1.21 1.37 0.75 2.87 2.63 0.85 0.090 0.088 1.67 0.08 1.32 0.15 2.84 1.48 2.75 2.77 0.71 0.057 0.74 0.10 1.36 1.56 2.73 0.62 2.74 0.142 0.31 1.79 0.175 2.38 0.89 0.19 0.95 1.25 2.65 1.26 2.74 D21-2BD21-6E 54.59D22-1A 54.59 138.53D57-3 54.84 138.53 0.702624 55.56 137.88 0.703486 14 0.703518 22 122.17 0.513113 22 0.702536 0.512922 8 14 0.512974 8 9 0.513067 18.889 10 18.782 0.002 18.498 0.010 15.539 15.544 0.005 0.002 15.518 0.008 38.665 38.397 0.004 0.004 37.749 0.021 0.574 0.273 0.009 1.14 0.54 0.162 2.34 0.24 1.40 0.88 1.99 1.62 3.55 2.84 1.46 2.75 0.061 0.08 0.55 1.70 2.89 114 P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 ratios were first ground into 1±2 cm sized fragments are presented in Table 1. Analytical procedures for in a stainless steel mortar and pestle, and then soni- isotopic analysis are listed as notes in Table 1; cally cleaned in distilled water. After drying, glass analytical procedures for major and trace element fragments free of phenocrysts and alterations were analysis are presented elsewhereŽwx 35 ; manuscript in hand picked under a binocular microscope. The se- prep.. . lected glass fragments were sonically cleaned again in distilled water and then acetone. Finally, glass 4. Analytical results samples for isotopic analysis were leached with warm, ;2.0 N HCl for about 15 min prior to The Pacific±Antarctic lavas are isotopically het- dissolution. Isotopic analyses and some trace element erogeneous, but their Sr, Nd and Pb isotopic ratios data of selected glass samples from the study area display overall correspondence and their

Fig. 2.Ž. A87 Srr 86Sr versus143 Ndr 144 Nd andŽ. B206 Pbr 204 Pb versus207 Pbr 204 Pb and208 Pbr 204 Pb ratios for the Pacific±Antarctic basalts listed in Table 1. Typical analytical uncertainties are about the size or smaller than the symbols used. Fields for MORB from the Indian Ocean ridgesŽ. IOR , East Pacific Rise Ž. EPR and Louisville Seamount Chain Ž. LSC are shown for reference. Sources of data are fromwx 6,7,28,37±43 and references therein. Data for samples D31-1BŽ.Ž. T-MORB , GC32 N-MORB and N-MORB from the Raitt intratransform spreading center are labelled separately for clarity. Off-axis samples represent samples from the linear volcanic chain in the inside corner of Heezen and from the southeastern end of Hollister Ridge. P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 115

206 Pbr 204 Pb ratios correlate well with 207 Pbr 204 Pb and208 Pbr 204 PbŽ. Figs. 2 and 3 . Pacific±Antarctic N-MORB are similar to typical ridge basalts elsewhere, being depleted in incompatible trace elements and having variable but generally low 87Srr 86 Sr and high 143 Ndr 144 Nd ratios, except for those from the Raitt intratransform spreading centerŽ. Fig. 2A . The intratransform basalts have relatively higher 87Srr 86 Sr and lower 143 Ndr 144 Nd than the other N-MORB, despite their extreme depletion in incompatible elements; they plot at the high 87Srr 86 Sr and low 143 Ndr 144 Nd end of the MORB field for the northerly EPR, approaching the composition of the Louisville Seamount Chain lavas. Moreover, samples GC32 and D31-1B, which are N-MORB and T- MORB, respectively, from a short PAR segment bounded by the Hollister and Tharp transforms, have slightly higher 87Srr 86 Sr for given 143 Ndr 144 Nd than the other Pacific±Antarctic N-MORB and the EPR MORB. The Sr and Nd isotopic ratios of Pa- Fig. 3.206 Pbr 204 Pb versusŽ. A87 Srr 86Sr andŽ. B143 Ndr 144 Nd cific±Antarctic T-MORB are variable, ranging from ratios for Pacific±Antarctic basalts. Symbols and sources of data low 87Srr 86 Sr and high 143 Ndr 144 Nd, similar to the as in Fig. 2. majority of the N-MORB, to high 87Srr 86 Sr and low 143 Ndr 144 Nd, close to those of the Louisville EPR field, but within the field for MORB from the Seamount Chain lavas. The sample from the volcanic Indian Ocean ridges. Raitt intratransform and off-axis chain in the inside corner of Heezen transform has basalt data also plot outside the EPR field, and inside low 87Srr 86 Sr but high 143 Ndr 144 Nd similar to the the field for MORB from the Indian Ocean ridges. N-MORB, but samples from Hollister Ridge have Sr Hollister Ridge basalts do not plot inside the and Nd isotopic values that overlap with those of the Louisville Seamount Chain field because they have Louisville Seamount Chain lavas. The Pb isotope low 206 Pbr 204 Pb for given 87Srr 86Sr and ratiosŽ. Fig. 2B of all Pacific±Antarctic ridge basalts 143 Ndr 144 Nd relative to the Louisville Seamount overlap and generally fall within the EPR MORB Chain lavas. They are isotopically similar to the field in the 206 Pbr 204 Pb versus 208 Pbr 204 Pb dia- lavas from seamounts in the central and northern gram. However, the intratransform N-MORB and portions of the Hollister RidgeŽwx 36 ; L. Dosso, pers. samples D31-1B and GC32 have relatively higher commun., 1997. . Thus, lavas from Hollister Ridge 207 Pbr 204 Pb for given 206 Pbr 204 Pb than the rest of may not have entirely originated from the Louisville the ridge basaltsŽ. Fig. 2B . Samples from the Hollis- mantle plume, but from a mixture of Louisville and ter Ridge have among the highest 206 Pbr 204 Pb and depleted EPR mantle materials. plot close to, although do not overlap, values of the Except for the Raitt intratransform basalts and Louisville Seamount Chain lavas. samples GC32 and D31-1B, the 87Srr 86Sr, Additional details of the isotopic characteristics of 143 Ndr 144 Nd and 207 Pbr 204 Pb of Pacific±Antarctic the Pacific±Antarctic lavas are shown by the plots of ridge and off-axis basalts show good correlations 87Srr 86 Sr and 143 Ndr 144 Nd versus 206 Pbr 204 Pb with highlyrmoderately incompatible element ratios, Ž.Fig. 3A,B . The majority of Pacific±Antarctic N- from the most depleted abyssal tholeiites to the MORB data indeed plot inside the EPR MORB field; Hollister alkalic basaltŽ correlation coefficients, r, are data for the volcanic chain in the inside corner of )0.80; Fig. 4A±F. . Correlations of206 Pbr 204 Pb Heezen transform also plot within the EPR MORB and 208 Pbr 204 with the majority of highlyrmod- field. Samples GC32 and D31-1B plot outside the erately incompatible element ratios for the Pacific± 116 P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125

Fig. 4.Ž.Ž. A ± F87 Srr 86Sr,143 Ndr 144 Nd and207 Pbr 204 Pb versus BarTi and chondrite-normalized LarSm ratios for Pacific±Antarctic basalts. Chondrite normalizing values are fromwx 55 . Correlation coefficientsŽ. r values for the regression lines are calculated for all the basalts except for the Raitt intratransform, D31-1B and GC32 basalts. Plots of 87Srr 86 Sr and 143 Ndr 144 Nd versus other highlyrmoderately incompatible element ratios such as RbrTi, KrTi, KrP and SmrNd also yield similarly high correlation coefficient valuesŽ. r)0.80 . P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 117

143 r 144 87 r 86 r Fig. 5. Variations ofŽ. A Nd Nd;BNa;CTi;DŽ.88 Ž. Ž. Sr Sr;Ž. E DNd; and Ž. F D8 4 of MORB vs. distance from Vacquier 143 r 144 transform along the Pacific±Antarctic risecrest. Data are interpreted as either a weak overall increase in Nd Nd and decreases in Na8 and Ti8 of N-MORB from the northeastŽ. Vacquier transform toward the southwest Ž.Ž.Ž. Udintsev transform A ± C or a systematically higher 87Srr 86Sr, DNd, and D8r4 of the northern PARŽ.Ž. southwest of Heezen transform relative to the southern EPR northeast of Heezen 143 r 144 Ž.Ž.D ± F . The approximate location of the intersection of the PAR with Hollister Ridge, which has low Nd Nd and high Na88 , Ti , 87Srr 86 Sr, DNd, and D8r4, is shown for reference. Off-axis basalt data are not shown for clarity. 118 P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125

Antarctic basalts also generally exist but are not as Pacific±Antarctic N-MORB also exist. In contrast to good as those with 87Srr 86 Sr, 143 Ndr 144 Nd and 143 Ndr 144 Nd values, which represent individual 207 r 204 - 206 r 204 Pb PbŽ. r 0.65 , except that Pb Pb is samples, the Na88 and Ti values represent averages well correlated with RbrSrŽ. rs0.78, not shown . of compositionally identicalŽ. within analytical error Although the good correlations are mainly controlled samples from a given sampling station. These geo- by the few extreme values of T-MORB and enriched graphical patterns of the Ti88 and Na signify an Hollister Ridge basalts, similar though more limited orderly spatial variation in the bulk major element trends are generally shown by the N-MORB alone. composition of the upper mantle beneath the Pa- Thus, our data suggest that the Sr, Nd and Pb cific±Antarctic risecrest. isotopic and incompatible element ratios are gener- Another way of looking at the data is to consider ally coupled in the majority of Pacific±Antarctic a stepped rather than a gradational transition in basalts. This observation contrasts with that in the isotopes along these ridges, with the step correspond- superfast spreading southern EPR at 13±238S where ing to the long offset across the two transform faults the Nd, Sr and Pb isotopic ratios of MORB are of the Eltanin system. The two geographical groups totally decoupled from trace element ratioswx 37,38 . are shown by plots of 87Srr 86 Sr, DNd and D8r4 Sr, Nd and Pb isotopic and incompatible element values versus along axis distanceŽ. Fig. 5D±F . DNd ratios are decoupled only in a few samples such as represents deviations of 143 Ndr 144 Nd from the man- samples D31-1B and GC32 and the Raitt intratrans- tle plane and D8r4 represents departures of form basalts. The intratransform basalts have some- 208 Pbr 204 Pb from the northern hemisphere reference what higher 87Srr 86 Sr and lower 143 Ndr 144 Nd than linewx 39,40 . The peak in all of these values is they might, given their extremely low concentrations centered on the geographic intersection of the Hollis- of highly incompatible elements. An Indian MORB- ter Ridge with the PAR, suggesting that the Louisville type mantle, proportionally higher in 87Srr 86 Sr and hotspot influences the isotopic signature of a fairly BarTi than other N-MORBŽ. Table 1 and Fig. 4 , is broad region of the mantle underlying the Pacific± represented by samples GC32 and D31-1B from the Antarctic risecrest. two adjacent sampling stations, 31 and 32, in the PAR segment between the Hollister and Tharp trans- formsŽ. Fig. 1 . 5. Discussion Finally, N-MORB northeast of the Udintsev trans- The variable isotopic and highlyrmoderately in- form have higher 143 Ndr 144 Nd than those from the compatible element ratios of Pacific±Antarctic rise- EPR southwest of the Vacquier transform. The dif- crest and off-axis basalts indicate that the upper ference reflects a weak, gradual increase in mantle there, like that beneath the northerly EPR and 143 Ndr 144 Nd to the southwest, from the Vacquier to northeast Pacific ridges, is compositionally heteroge- Udintsev transformŽ. Fig. 5A . Note that the Raitt neousŽ e.g.wx 37,38,41±43. . The overall pattern of intratransform basalts appear to control the low end these heterogeneities among Pacific±Antarctic N- of the gradient but, even if these are excluded from MORBŽ. Fig. 5A±F leads us to propose two alterna- the N-MORB group, the rest of the samples still tive hypotheses about dynamic processes in the man- show a general increase in the range of 143 Ndr 144 Nd tle beneath the Pacific±Antarctic risecrest. to the southwest. Sr isotopic ratios show only a slight 87 r 86 overall decrease in the range of Sr Sr toward 5.1. A southwestward flow of the South Pacific upper the same directionŽ. Fig. 5D but a clear systematic mantle decrease in 87Srr 86 Sr to the southwest is observed along the PAR immediately west of our study area, This hypothesis is based on the premise that, 8 X 8 8 143 r 144 from the Udintsev transform to 63 27 S, 166 04 W although the gradients in Nd Nd and Na8 and wx 36 . Slight along-axis decreases from northeast to Ti8 values among N-MORB in our study area are southwest in Na22 O and TiO contents corrected for weak, these nevertheless represent a coherent geo- fractionation to 8 wt% MgOŽ Na88 and Ti , respec- chemical structure of the N-MORB mantle source in tively; Fig. 5B,C. of the different sample groups of the South Pacific. The large geochemical anomaly at P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 119 the PAR±Hollister Ridge intersection caused by the ridge-axis morphological features, the fact that there influence of the Louisville plume is neither related is a systematic decrease in 87Srr 86 Sr along the PAR X X to, nor responsible for, the geochemical gradient from the Udintsev transform to 63827 S, 166804 W among N-MORB. First, Hollister Ridge lavasŽ this wx36 , which is part of the proposed area of mantle study;wx 36. and Louisville seamount chain lavaswx 28 flow, indicates that there may indeed be a relation- have low 143 Ndr 144 Nd and are enriched in incom- ship between morphologic expression of the spread- patible elements; these are unlike the high ing axis and the composition of the upper mantle. 143 Ndr 144 Nd and incompatible element depleted N- Thus, we envision that pulses or waves of high MORB just north of UdintsevŽ. Fig. 4A±C . Second, 143 Ndr 144 Nd Pacific N-MORB mantle come from the occurrence along the Pacific±Antarctic risecrest northeast of the Vacquier transform and flow toward of plume-influenced T-MORB is merely as co-inci- the AAD. dental as the occurrence of transitional-MORB and Under this hypothesis, an Indian MORB-type enriched-MORB on some ridge axes near other non- mantle, like the Louisville mantle plume, is simply hotspot seamount provincesŽ e.g.wx 31,44. . Hollister entrained into a large southwestward flow of de- Ridge, itself, thus is not necessarily related tectoni- pleted Pacific-type asthenosphere. However, the ex- cally to the Louisville Seamount Chain, as summa- act origin of the Indian MORB-type source is un- rized earlier from the perspective of its location and known. This could have been simply swept upward orientation, although the isotopic similarity between by the Louisville mantle plume, but the presence of the Hollister and Louisville lavas suggests that the Indian MORB-like lavas along the Chile Rise farther ridge and seamounts share a common source. The to the eastwx 49 indicates that it is not a unique question then is: what is the cause of the southwest- feature of the Louisville hotspot. Perhaps, designa- ward gradation in composition among N-MORB? tion of `Indian MORB-type' mantle implies no ge- Several morphologic and geochemical studies of netic link with the Indian Ocean; therefore, similar southern Pacific ridgeswx 7,45±48 have proposed that mantle at remote locations elsewhere reflects a pro- the Pacific upper mantle is flowing southwestward. cess that may be or have been dominant in the Indian We suggest that this could be responsible for the Ocean, but is not restricted to it. For example, the observed geochemical gradient in our region. Based Indian Ocean-like mantle material could have origi- on the composition of MORB from the PAR and nated all the way from the south Central Pacific, Southeast Indian Ridge on the east and west sides of where the large concentration of hotspots suggests a the Australian±Antarctic DiscordanceŽ. AAD , re- large-mantle convection, sweeping up lower mantle spectively, Pyle et al.Žwx 7 ; see also w 45 x. have pro- materialswx 40,50 . Alternatively, the process usually posed that Pacific asthenosphere has been flowing invoked for the Indian Ocean mantle is that of southwestward during the last 30 m.y., displacing the involvement or retention of regions, or even streaks, Indian Ocean upper mantle between Australia and of subcontinental mantle left from the dispersal of Antarctica. Such flow may have resulted from Gondwanan continents and continental fragments in shrinkage of the Pacific region as the Gondwanan the widening ocean basinsŽ e.g.wx 6. . The Eltanin plates grewwx 46 . More recent interpretations suggest fault system can be traced toward Antarctica to the that evidence for the southwestward flow of the southeast and toward the general area of the Chatham Pacific asthenosphere also occurs along the PAR Rise, which is a continental fragment, to the north- between the AAD and Udintsev transformwx 47,48 . west. These continental boundaries allow us to spec- This includes:Ž. 1 the occurrence of abrupt changes ulate that small fragments of old subcontinental man- in ridge morphology and subsidence of the ridge, tle may be dispersed as far away as the modern from rift valley with small subsidence to axial high Pacific±Antarctic risecrest in the Eltanin region. In with normal subsidence, across Fracture Zone XII at either process, the production of Indian Ocean-type ;648S, 1718Wwx 47 ; andŽ. 2 the presence of a mantle has little to do with the Indian Ocean. Still boundary between rough and faulted seafloor and another possibility is that these lavas could result smooth seafloor in the segment between the Udint- from interaction of mantle melt with hydrothermally sev transform and 1578Wwx 48 . Although these are altered oceanic crust, rich in hydrothermally mobile 120 P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 components such as 87Sr and Ba, although the char- material consequently displaces and eventually joins acteristically low 206 Pbr 204 Pb of these lavas make surrounding masses of depleted MORB astheno- this possibility unlikely. We thus see that Indian sphere flowing to the southwest. Thus the isotopic MORB-type mantle in this region could have come composition of Hollister Ridge connects the from sources other than the Louisville plume. The Louisville plume with the Pacific±Antarctic N- distinctive Pb and Sr isotopic signature and high Ba MORB sourceŽ. Fig. 6 . In addition to giving Hollis- content of the basalts suggest that their mantle source ter Ridge its distinctive isotopic signature, the is different from those of the Hollister lavas and Louisville plume also enhances production of crust Pacific±Antarctic N-MORBŽ. Figs. 2±4 and Figs. 6 . and greater total volume of melting along both Hol- As mentioned earlier, this asthenospheric flow lister Ridge and its intersection with the PAR be- may also be responsible for the azimuth of Hollister cause the plume mainly consists of enriched, low Ridge and the other ridges parallel to it, which could temperature melting peridotites. be the vector sum of the northwestward and south- The asthenospheric flow proposed in this hypoth- westward motions of the Pacific plate and astheno- esis is not restricted to this region because it also sphere, respectively. The Hollister family of ridges evidently occurs, albeit in a more subtle form, along are a plate-tectonic feature of yet unknown origin the superfast spreading segment of the EPR at 13± Žcf.wx 22,23,25. . Hollister Ridge is the largest and 238S. There, highlyrmoderately incompatible ele- most prominent among these ridges because it is the ment ratios consistently decrease from north to south northernmost and longest, so that it was able to tap along the entire segment but Sr, Nd and Pb isotopic X most of the Louisville plume material entrained in ratios define a broad, smooth peak between 15848 S X the southwest-flowing asthenosphere. Louisville and 20852 S, with a maximum enrichment at 17± X plume material has been ascending near the 17830 Swx 37,38 . Mahoney et al. wx 38 proposed that Pacific±Antarctic risecrest and Eltanin fault system the material with plume-like isotopic composition for more than 40 million yearswx 27,28 and this that causes this isotopic hump may have originated

Fig. 6.Ž. A Triangular plot of relative abundance of207 Pb, 208 Pb and 206 Pb in Pacific±Antarctic basalts. The compositional ranges of207 Pb, 208 Pb and 206 Pb in the whole triangle are 15.00±16.00, 37.00±40.00 and 16.50±20, respectively. Also plotted in the triangle is the field for Indian MORB Pb isotopes for referenceŽ fromwx 6,41 and references therein.Ž. . B Enlarged view of the shaded portion in the triangle. The 3 lines with arrows represent the three isotopic groups mentioned in the text: 1sPacific N-MORB array, 2sLouisville plume and depleted Pacific±Antarctic MORB mixing trend, and 3sdepleted Pacific±Antarctic MORB and Indian MORB, specifically the low 206 Pb and high 207 Pb type, mixing trend. P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 121 long ago from a distant hotspot which now lies northerly location toward PAR segments south of the embedded in the MORB mantle matrix. One of the Eltanin fault system. possible sources of the material is the Marquesas The simplest case, of course, would be if the hotspot in the south Central Pacific, separated from Louisville hotspot is located along Hollister Ridge the East Pacific risecrest with lithosphere riddled wx26 . We have ruled out this possibility earlier but, with subparallel volcanic ridges, in many ways simi- even if this is the situation, the stepped increases in lar to the Hollister ridges on the north flank of the thicknesses of the lithosphere across the Eltanin fault PAR. Thus, in both the Pacific±Antarctic South system block the lateral asthenospheric flow at the Pacific and southern Equatorial Pacific, the Pacific base of the lithosphere to the east and north. The plate moves over an asthenosphere that appears to result is accumulation of upwelling mantle beneath, move in different directions; in the former, the as- and a responsive southward flow along, the nearest thenosphere is moving toward the southwest and in adjacent ridge axis. The most immediate effect on the latter it seems to be moving toward the east. The ridge elevation is along the nearest adjacent segment. origin and exact direction of movement of the as- Thus, no matter where the hotspot is, some south- thenosphere are beyond the scope of our study but ward flow of shallow asthenosphere along or adja- may be related to the yet unknown component of cent to the ridge axis, and pushed by the Louisville plate tectonic motion Ð that of the internal structure plume, is predicted. Besides the shallow, inflated of mantle convection, which is decoupled from ob- ridges, such flow might also produce deviatoric servable lithospheric motions. stresses on the base of the thin lithosphere, which would account for the obliquity of Hollister Ridge to any plausible hotspot trace, as well as to the axis of 5.2. Pacific±Antarctic Ridge±LouisÕille hotspot in- the Pacific±Antarctic risecrest. As in the previous teraction hypothesis, diverted flow also provides a mechanism explaining the plume-like geochemical signature of Our second hypothesis is based on the geochemi- basalts from Hollister Ridge. Whatever Hollister cal anomaly centered on the intersection of the PAR Ridge is structurally, with its unusual orientation, with Hollister Ridge. Here the PAR is also shallow- and in relationship to the hotspot, it appears to be a est, it has an inflated axial cross section, and it has a particularly close encroachment of the plume-in- correspondingly pronounced positive axial gravity duced high-pressure region on the adjacent spreading anomalywx 20 . These isotopic and geographic pertur- centers, which accentuates both subaxial mantle flow bations are commonly observed along the mid-ocean and ridge topography. Neither elevation nor inflation ridge when a hotspot is close to the risecrestŽ e.g. of risecrest topography require that the Louisville wx18,43,51. . In our model, lateral mantle flow at the plume impinge directly on the spreading center but base of the lithosphere will be guided southward by they are a result of its proximity. Hence, the Hollis- the step-wise thinner lithosphere beneath the Eltanin ter Ridge, and the other ridges concentric to it on the system. Because of the right-lateral offsets on the Pacific ridge flank, are consequences of volcanism fracture zones, the lithosphere is thinner to the south along oblique rents in the sea floor caused by south- of the Louisville hotspotŽ. Fig. 1 ; this is the direction westward lateral pumping of asthenosphere along the of pressure release for lateral flow in the shallow ridge axis, prompted by proximity to the Louisville asthenosphere. Another possibility is that Hollister plume. Ridge is a natural consequence of a spreading center The pumping of the Louisville plume material migrating toward a hotspot. Smallwx 25 proposed that, southwest to the PAR is responsible for the step-wise in such a case, plume material is redistributed and increase in 87Srr 86 Sr, DNd and D8r4 values in the migrates updip along the base of the lithosphere, northern PARŽ. Fig. 4C±E . The Hollister Ridge away from the cool end of the spreading center alkalic basalt falls within, or close to, the plotted toward the hotter and thinner center of the segment. field for the Louisville Seamount Chain, the enriched In this scenario, Hollister Ridge is the trace of potential mixing end-member in all permutations of Louisville plume material migrating from its colder, the Sr, Nd and Pb isotope diagramsŽ. Figs. 2 and 3 . 122 P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125

Together, Louisville and Hollister form a linear data dynamically related to the Louisville plume in this array that trends to the low 206 Pbr 204 Pb end of the region. The plume influence on the ridge, albeit Pacific±Antarctic N-MORB arrayŽ. Fig. 6 . All points subtle in geochemical terms, is nevertheless quite are a strong indication that the trends result from broad geographically. The plume itself may be geo- simple two-component mixing, with one end-mem- chemically zonedŽ cf.wx 52. , or it may have punched ber being represented by the most depleted abyssal a deep layer of Indian MORB-type mantle upward tholeiite in the region, the other by the alkalic basalt into a depth range where it could begin to melt. from Louisville hotspot. The mixing may occur at crustal levels, between magma batches, or by blend- 5.3. Large-scale magmatic segmentation ing of melt strains near the mantle sourceŽ e.g.wx 42. : it does not matter which. That the mantle itself is In the northerly EPR, MORB erupted along con- sharply partitioned into strongly depleted and more tiguous sections of the risecrest that have coherent enriched zones on a very small scale is indicated by isotopic and trace element compositions, such as the occurrence of samples near both extremes even those along the superfast spreading segments of the in single dredge haulsŽ cf.wx 32. . The enriched com- EPR at 13±238Swx 37,38,52 , have been proposed to ponents within the mantle are dominated by the one define `primary magmatic segments'wx 38,44 . Each end-member represented by Hollister Ridge alkalic segment is thought to contain magmas derived from basalt. The proportion of the more enriched compo- a common, compositionally distinct mantle domain nent is strongest where the ridge is shallowest, but it wx53 . Our two hypotheses for the composition and is clearly present to some extent everywhere. The dynamics of the upper mantle beneath the Pacific± Indian MORB-type mantle, in stations 31 and 32 Antarctic risecrest have different implications on the Ž.Fig. 1 , marks the first appearance of the high size of the mantle domain in the south Pacific. In the 87Srr 86Sr, DNd and D8r4Pb valuesŽ. Fig. 4E±G ridge±hotspot interaction hypothesis, the mantle south of Heezen. Whatever its origin, Indian Ocean- source of the Pacific±Antarctic MORB is divided like mantle was entrained along with the ascending into a northeastern Pacific-type mantle source and a Louisville plume mantle, and introduced to the vicin- southwestern, heterogeneous, Louisville plume-in- ity of the Pacific±Antarctic risecrest at the same fluenced mantle source. The boundary of the two time. mantle domains is the Heezen transform of the To summarize, the ridge±hotspot interaction hy- Eltanin fault system, and is thus consistent with the pothesis claims that there are two general mantle proposal that these large mantle domains are bounded types along the Pacific±Antarctic risecrest. One type by the largest ridge tectonic offsetsŽ e.g.wx 4,44. . is represented by depleted abyssal tholeiites compa- If there is a southwestward flow of the south rable to those found elsewhere along the EPR, the Pacific asthenosphere, on the other hand, then the 143 r 144 other by an array of related compositions ranging gradient in Nd Nd ratios, Na88 and Ti of from comparatively depleted tholeiites, with similari- N-MORBŽ. Fig. 4A±C indicates that the mantle ties to Indian Ocean type-MORB, and enriched from the Vacquier to Udintsev transforms belongs to basalts having in the extreme a Louisville plume-like a single mantle domain distributed over at least 1800 signature. All of these may be derived from mantle km beneath the Pacific±Antarctic risecrest. The ob- materials added to this region, whether in dispersed served continuity of mantle source composition or concentrated form, by advection associated with across the Eltanin fault system also indicates that the plume. The strongly depleted mantle northeast of primary mantle domains are not necessarily bounded Heezen, which might be termed Pacific-type mantle, by the largest ridge tectonic offsets. Primary mantle is fairly strongly influenced south of Heezen by both domains reflect longŽ at least several hundred million types of source peridotite along the risecrest for a years. time-integrated histories of these mantle distance of at least 300 km. This, and the shallow sources to come up with detectable differences in depths to the risecrest, are the immediate conse- radiogenic isotope ratios. On the other hand, the size quences of proximity of the Louisville hotspot to the of large ridge tectonic offsets such as transform PAR. In addition, the Indian MORB-type mantle is faults are lithospheric tectonic features and grow P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 123 independently of the composition of the underlying represented by alkalic basalts from the southeastern mantle. These tectonic features grow in response to end of the Hollister Ridge. The third isotopic end- changing spreading rate and direction of the separat- member is similar to that of Indian MORB-type ing ridge axes, which in the case of the Eltanin fault mantle. The appearance of these two end-members in system was only during the last 75 MaŽ e.g.wx 16. . isolated samples along or close to the ridge axis Hence, it is not required that the boundaries of suggests that their mantle sources are effectively primary mantle domains coincide with the largest isolated from the rest of the Pacific±Antarctic rise- ridge tectonic offsets. Of course, neither the mantle crest upper mantle, and thus their resultant lavas nor spreading ridges are stationary, at least with were able to survive the typical melting and mixing respect to one another, such that, in certain cases, phenomena underneath the ridge axis. boundaries of mantle geochemical zones and litho- The overall pattern of distribution of these hetero- spheric tectonic boundaries unavoidably coincide. geneities along the Pacific±Antarctic risecrest leads Hence, it is not surprising to observe that some of us to propose two hypotheses to explain the nature the boundaries of the primary magmatic segments in and dynamics of the upper mantle in the south the northerly EPR coincide with large-offset trans- Pacific. One claims that the depleted N-MORB form faultsŽ e.g.wx 44. . source forms a generally coherent reservoir, which displays a geochemical gradient with 143 Ndr 144 Nd increasing and Na and Ti decreasing from north- 6. Summary and conclusions 88 east to southwest over 1800 km of ridges and inter- The variable isotopic and highlyrmoderately in- vening transform faults. This primary mantle domain compatible element ratios of Pacific±Antarctic rise- is apparently unaffected by the presence of the crest and off-axis basalts indicate that the upper large-offset transforms of the Eltanin system. The mantle there, like that beneath the northerly EPR and origin of the systematic northeast±southwest compo- northeast Pacific ridges, is compositionally heteroge- sitional gradient of the N-MORB mantle source be- neous. We discern three isotopic end-members in the neath the ridges is not directly related to mixing south Pacific upper mantle:Ž. 1 the `depleted' type of between the depleted MORB sources and enriched material which is the source of Pacific N-MORB;Ž. 2 Louisville plume; the gradient is only locally inter- an `enriched' source similar to that which produces rupted by the influence of the Louisville hotspot and basalts of the nearby Hollister Ridge; andŽ. 3 a the presence of Indian Ocean-type mantle. Instead, source, restricted to two adjacent sample locations, the geochemical gradient may be related to a pro- similar to that of Indian MORB. As is generally true posal that the Pacific N-MORB mantle is flowing in the northerly Pacific, the source of the incompati- southwestward in response to the shrinking size of ble-element-depleted Pacific N-MORB is the upper the Pacific basin, but the relationship is highly specu- mantle, which itself is variable in composition. The lative at this time and needs further investigation. linear isotopic and trace element ratio arrays of The other hypothesis claims that there are two N-MORB are a strong indication that the trends primary mantle domains beneath the Pacific± result from simple two-component mixing, with one Antarctic risecrest. Northeast of the Heezen trans- end-member being depleted, the other by variably form, of the Eltanin fault system, which is the most geochemically depleted to enriched sources which prominent and stable physical segment boundary in may occur as blobs, veinlets, or streaks within the the Pacific, the mantle is similar to that elsewhere depleted mantle domain. Mixing may occur in the along the northern Pacific ridges, being mainly the mantle source prior to melting between solid materi- depleted source of Pacific N-MORB. Southwest of als, near the mantle source of the melts derived from the Heezen transform, the Pacific N-MORB mantle these materials, or at crustal levels between magma source is contaminated by Louisville mantle plume batches or melt strains. The second end-member is material, which also includes entrained Indian that of an enriched source, the most extreme compo- MORB-type mantle. The Louisville plume is pumped sition of which in this region is represented by lavas to the PAR through Hollister Ridge, represented by from the Louisville hotspot. In our study area, it is mildly alkalic lavas along its southern end. 124 P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125

Acknowledgements transform faults, EOS Trans. Am. Geophys. Union 75Ž. 1992 742. wx Field and laboratory work was supported by Na- 12 C. DeMets, R.G. Gordon, D.F. Argus, S. Stein, Current plate motions, Geophys. J. Int. 101Ž. 1990 425±478. tional Science Foundation grants OCE93-00275 wx13 S.C. Cande, D.V. Kent, A new geomagnetic polarity time Ž.Castillo , OCE94-11771 Ž. Natland , OCE93-00352 scale for the Late Cretaceous and Cenozoic, J. Geophys. Res. Ž.ŽC.H. Langmuir for Niu , and OCE92-17707 Lons- 97Ž. 1992 13917±13951. wx dale. . We thank B. Hanan for assistance in Pb 14 C.M. Weiland, K.C. Macdonald, Geophysical study of the 8 8 isotope analysis, J. Hawkins for critical review of an East Pacific Rise 15 N±17 N: An unusually robust segment, J. Geophys. Res. 101Ž. 1996 20257±20273. early version of the manuscript, and J. Mahoney, M. wx15 SEIR Scientic Team, J.R. Cochran, J.-C. Sempere, The Perfit and an anonymous reviewer for helpful re- Southeast Indian Ridge between 888E and 1188E: Gravity views. []FA anomalies and crustal accretion at intermediate spreading rates, J. Geophys. Res. 102Ž. 1997 15463±15487. wx16 P. Lonsdale, Tectonic and magmatic ridges in the Eltanin fault system, South Pacific, Mar. Geophys. Res. 8Ž. 1986 References 203±242. wx17 K.C. Macdonald, Mid-Ocean ridges: Fine-scale tectonic, vol- wx1 C.H. Langmuir, J.F. Bender, R. Batiza, Petrologic and tec- canic, and hydrothermal processes within the plate boundary X tonic segmentation of the East Pacific Rise between 5830 ± zone, Annu. Rev. Earth Planet. Sci. 10Ž. 1982 155±190. X 14830 N, Nature 222Ž. 1986 422±429. wx18 J.-G. Schilling, R.H. Kingsley, J.D. Devine, Galapagos hot wx2 Y. Niu, R. Batiza, An empirical method for calculating melt spot-spreading center system 1. Spatial Petrological and geo- compositions produced beneath mid-ocean ridges: applica- chemical variationsŽ. 838W±1018W , J. Geophys. Res. 87 tion for axis and off-axisŽ. seamounts melting, J. Geophys. Ž.1982 5593±5610. Res. 96Ž. 1991 21753±21777. wx19 J.A. Heirtzler, D.E. Hayes, E.M. Herron, W.C. Pittman, wx3 J.M. Sinton, R.S. Detrick, Mid-ocean ridge magma cham- Preliminary report of volume 20 U.S.N.S. Eltanin Cruises bers, J. Geophys. Res. 97Ž. 1992 197±216. 16-12, Lamont-Doherty Survey of the World Oceans, Tech. wx4 K.C. Macdonald, P. Fox, L. Perram, M. Eisen, R. Haymon, Rep. 3, 1969, 122 pp. S. Miller, S. Carbotte, M. Cormier, A. Shor, A new view of wx20 W. Smith, D.T. Sandwell, Measured and estimated seafloor the mid-ocean ridge from the behavior of ridge axis disconti- topographyŽ. version 4.2 , World Data Center A for Marine nuities, Nature 335Ž. 1988 217±225. Geology and Geophysics research publication RP-1, poster, wx5 P. Lonsdale, Segmentation and disruption of the East Pacific 1997. Rise in the mouth of the Gulf of California, Mar. Geophys. wx21 D.S. Scheirer, K.C. Macdonald, D.W. Forsyth, S.P. Miller, Res. 17Ž. 1995 323±359. D.J. Wright, M.-H. Cormier, C.H. Weiland, A map series of wx6 J.J. Mahoney, A.P. le Roex, Z. Peng, R.L. Fisher, J.H. the southern East Pacific Rise and its flanks, 158Sto198S, Natland, Southwestern limits of Indian Ocean ridge mantle Mar. Geophys. Res. 18Ž. 1996 1±12. and the origin of low 206 Pbr 204 Pb mid-ocean ridge basalt: wx22 D.T. Sandwell, E.L. Winterer, J. Mammerickx, R.A. Duncan, isotope systematics of the central Southwest Indian Ridge M.A. Lyych, D.A. Levitt, C.L. Johnson, Evidence for diffuse Ž.178±508E , J. Geophys. Res. 97 Ž. 1992 19771±19790. extension of the Pacific plate from Pukapuka ridges and wx7 D.G. Pyle, D.M. Christie, J.J. Mahoney, R.A. Duncan, Geo- cross-grain gravity lineations, J. Geophys. Res. 100Ž. 1995 chemistry and geochronology of ancient southeast Indian and 15087±15099. southwest Pacific seafloor, J. Geophys. Res. 100Ž. 1995 wx23 W.F. Haxby, J.K. Weissel, Evidence for small-scale conven- 22261±22282. tion from Seasat altimeter data, J. Geophys. Res. 91Ž. 1986 wx8 P. Lonsdale, Geography of the southernŽ. Pacific±Antarctic 3507±3520. East Pacific Rise and origin of its structural segmentation, J. wx24 A.E. Gripp, R.G. Gordon, Current plate velocities relative to Geophys. Res., 1994. the hotspots incorporating the NUVEL-1 plate motion model, wx9 P.R. Castillo, J.H. Natland, Y. Niu, Petrology and Sr, Nd and Geophys. Res. Lett. 17Ž. 1990 1109±1112. Pb isotope geochemistry of MORB from the fast-spreading wx25 C. Small, Observations of ridge±hotspot interactions in the ridges in the South Pacific, EOS Trans. Am. Geophys. Union Southern Ocean, J. Geophys. Res. 10Ž. 1995 17931±17946. 76Ž. 1995 F530. wx26 P. Wessel, L. Kroenke, A geometric technique for relocating wx10 P.R. Castillo, Y. Niu, J.H. Natland, S.H. Bloomer, P.F. hotspots and refining absolute plate motions, Nature 387 Lonsdale, Trace element and isotopic investigation of the Ž.1997 365±369. Pacific±Antarctic East Pacific Rise basalts; preliminary re- wx27 P. Lonsdale, Geography and history of the Louisville hotspot sults, EOS Trans. Am. Geophys. Union 75Ž. 1994 742. chain in the southwest Pacific, J. Geophys. Res. 93Ž. 1988 wx11 J.H. Natland, Y. Niu, P.R. Castillo, P.F. Lonsdale, S.H. 3078±3104. Bloomer, Petrologic exploration of the Pacific±Antarctic East wx28 Q. Cheng, K.-H. Park, J.D. Macdougall, A. Zindler, G.W. Pacific Rise in the vicinity of the Udintsev and Eltanin Lugmair, H. Staudigel, J. Hawkins, P. Lonsdale, Isotopic P.R. Castillo et al.rEarth and Planetary Science Letters 154() 1998 109±125 125

evidence for a hotspot origin of the Louisville Seamount wx42 Y. Niu, D.G. Waggoner, J.M. Sinton, J.J. Mahoney, Mantle Chain, in: B. Keating, P. Fryer, R. Batiza, G. BoehlertŽ. Eds. , source heterogeneity and melting processes in the Hump Seamounts, Islands and Atolls, Am. Geophys. Union Monogr. region: East Pacific Rise 188±198S, J. Geophys. Res. 101 43Ž. 1987 283±296. Ž.1996 27711±27733. wx29 G. Thompson, W.B. Bryan, R. Ballard, K. Hamuro, W.G. wx43 B. Hamelin, B. Dupre, C. Allegre, Lead±strontium isotopic Melson, Axial processes along a segment of the East Pacific variations along the East Pacific Rise and the Mid-Atlantic Rise, 108±128N, Nature 318Ž. 1985 429±433. Ridge: a comparative study, Earth Planet. Sci. Lett. 67 wx30 M.R. Perfit, D.J. Fornari, M.C. Smith, J.F. Bender, C.H. Ž.1984 340±350. Langmuir, R.M. Haymon, Small-scale spatial and temporal wx44 J.M. Sinton, S.M. Smaglik, J.J. Mahoney, Magmatic pro- variations in mid-ocean ridge crest magmatic processes, Ge- cesses at superfast spreading mid-ocean ridges: glass compo- ology 22Ž. 1994 375±379. sitional variations along the East Pacific Rise 13±238S, J. wx31 R. Batiza, D. Vanko, Petrology of young Pacific seamounts, Geophys. Res. 96Ž. 1991 6133±6155. J. Geophys. Res. 89Ž. 1984 11235±11260. wx45 R. Lanyon, A.J. Crawford, S.M. Eggins, Westward migration wx32 J.H. Natland, Partial melting of a lithologically heteroge- of Pacific Ocean upper mantle into the Southern Ocean neous mantle: inferences from crystallization histories of region between Australia and Antarctica, Geology 23Ž. 1995 magnesian abyssal tholeiites from the Siquieros Fracture 511±514. Zone, in: A.D. Saunders, M.J. NorryŽ. Eds. , Magmatism in wx46 W. Alvarez, Geologic evidence for the plate driving mecha- the Ocean Basins, Geol. Soc. London Spec. Publ. 42Ž. 1989 nism: The continental undertow hypothesis and the Aus- 41±70. tralian±Antarctic Discordance, Tectonics 9Ž. 1990 1213± wx33 R. Hekinian, D. Bideau, R. Hebert, Y. Niu, Magmatism in 1220. X the Garret transform faultŽ East Pacific Rise near 13827 S. , J. wx47 K.M. Marks, J.M. Stock, Variations in ridge morphology and Geophys. Res. 100Ž. 1995 10163±10185. depth±age relationships on the Pacific±Antarctic Ridge, J. wx34 M.R. Perfit, D.J. Fornari, W.I. Ridley, P.D. Kirk, J. Casey, Geophys. Res. 99Ž. 1994 531±541. K.A. Kastens, J.R. Reynolds, M. Edwards, D. Desonie, R. wx48 M. Sahabi, L. Geli, J.-L. Olivet, L. Gilg-Capar, G. Roult, H. Shuster, S. Paradis, Recent volcanism in the Siquieros trans- Ondreas, P. Beuzart, D. Aslanian, Morphological reorganiza- form fault: picritic basalts and implications for MORB magma tion within the Pacific±Antarctic Discordance, Earth Planet. genesis, Earth Planet. Sci. Lett. 14Ž. 1996 91±108. Sci. Lett. 137Ž. 1996 157±173. wx35 P.E. Janney, P.R. Castillo, Geochemistry of Mesozoic Pacific wx49 E.M. Klein, J. Karsten, Ocean ridge basalts with convergent MORB: Constraints on melt generation and the evolution of margin geochemical affinities from the southern Chile Ridge, the Pacific upper mantle, J. Geophys. Res. 102Ž. 1997 5207± Nature 374Ž. 1995 52±57. 5229. wx50 P.R. Castillo, The Dupal anomaly as a trace of the upwelling wx36 I. Vlastelic, L. Dosso, J. Etoubleau, J.-L. Joron, H. Bougault, lower mantle, Nature 336Ž. 1988 667±670. L. Geli, PACANTARCTIC Sci. Party, New geochemical wx51 B.B. Hanan, J.-G. Schilling, Easter microplate evolution: Pb data from the Pacific±Antarctic Ridge and the Hollister isotope evidence, J. Geophys. Res. 94Ž. 1989 7432±7448. Ridge, Terra Nova 9 Abstr. Suppl. a1Ž. 1997 515. wx52 R.A. Duncan, M.A. Richards, Hotspots, mantle plumes, flood wx37 J.J. Mahoney, J.M. Sinton, M.D. Kurz, J.D. Macdougall, K.J. basalts, and true polar wander, Rev. Geophys. 29Ž. 1991 Spencer, G.W. Lugmair, Isotope and trace element character- 31±50. istics of a super-fast spreading ridge: East Pacific Rise, wx53 R. Batiza, Magmatic segmentation of mid-ocean ridges: a 13±238S, Earth Planet Sci. Lett. 121Ž. 1994 173±193. review, in: C.J. Macleod, P. Taylor, C.L. WalkerŽ. Ed. , wx38 W. Bach, E. Hegner, J. Erzinger, M. Satir, Chemical and Tectonic, Magmatic, Hydrothermal and Biological Segmenta- isotopic variations along the superfast spreading East Pacific tion of Mid-Ocean Ridges, Geol. Soc. London Spec. Publ. Rise from 6 to 308S, Contrib. Mineral. Petrol. 116Ž. 1994 118Ž. 1996 103±130. 365±380. wx54 W. Todt, R.A. Cliff, A. Hanser, A.W. Hofmann,202 Pbq205 wx39 S.R. Hart, D.C. Gerlach, W.M. White, A possible new Pb double spike for lead isotopic analysis, Terra Cognita 4 Sr±Nd±Pb mantle array and consequences for mantle mix- Ž.1984 209. ing, Geochim. Cosmochim. Acta 50Ž. 1986 1551±1557. wx55 N.M. Evensen, P.J. Hamilton, R.K. O'Nions, Rare earth wx40 S.R. Hart, Heterogeneous mantle domains, genesis and mix- element abundances in chondritic meteorites, Geochim. Cos- ing chronologies, Earth Planet. Sci. Lett. 90Ž. 1988 273±296. mochim. Acta 42Ž. 1987 145±152. wx41 W.M. White, A. Hofmann, H. Puchelt, Isotope geochemistry of Pacific mid-ocean ridge basalt, J. Geophys. Res. 92Ž. 1987 4881±4893.