basaltic liquids and harzburgitic residues in the garrett

Ž .Earth and Planetary Science Letters 146 1997 243–258

Basaltic liquids and harzburgitic residues in the GarrettTransform: a case study at fast-spreading ridges

Yaoling Niu a,), Roger Hekinian b´a Department of Earth Sciences, The UniÕersity of Queensland, Brisbane, Queensland 4072, Australia

b IFREMER, Centre de Brest, DROrGM, B.P. 37, 29287 Plouzane, France´

Received 9 May 1996; revised 28 October 1996; accepted 30 October 1996

Abstract

X Ž .The peridotite–basalt association in the Garrett Transform, ;13828 S, East Pacific Rise EPR , provides a primeopportunity for examining mantle melting and melt extraction processes from both melts and residues produced in acommon environment beneath fast-spreading ridges. The peridotites are highly depleted, clinopyroxene-poor, harzburgites.Residual spinel, orthopyroxene and clinopyroxene in these harzburgites are extremely depleted in Al O , and plot at the2 3

Žmost depleted end of the abyssal peridotite array defined by samples from slow-spreading ridges including samples from.hotspot-influenced ridges , suggesting that these harzburgites are residues of very high extents of melting. The residual

Ž .peridotites from elsewhere at the EPR i.e., Hess Deep and the Terevaka Transform also are similarly depleted. Thissuggests that the extent of melting beneath the EPR is similar to, or even higher than, beneath ridges influenced by hotspotsŽ .e.g., Azores hotspot in the Atlantic Ocean and Bouvet hotspot in the Indian Ocean , and is significantly higher than F10%,a value that has been advocated to be the average extent of melting beneath global ocean ridges. Many of these harzburgitesamples, however, show whole-rock incompatible element abundances higher than expected. These same samples also havevarious amounts of excess olivine with forsterite contents as low as Fo . The total olivine modes correlate inversely with85

olivine forsterite contents, and positively with whole-rock incompatible element abundances. These correlations suggest thatthe excess olivine and incompatible element enrichment are both the result of melt–solid re-equilibration. The buoyant meltsthat ascend through previously depleted residues crystallize olivine at shallow levels as a result of cooling. Entrapment ofthese melts leads to whole-rock incompatible element enrichment. These observations contrast with the notion that meltsformed at depth experience little low pressure equilibration during ascent.

Keywords: harzburgite; mid-ocean ridge basalts; East Pacific Rise; mantle; melts; mid-ocean ridges

1. Introduction

Abyssal peridotites and mid-ocean ridge basaltsŽ .MORB are complementary products of mantlemelting and melt extraction processes that create the

) Corresponding author. Tel.: q61 7 3365 2372. Fax: q61 73365 1277. E-mail: [email protected]

w xocean crust. Studies of abyssal peridotites 1–4 andMORB showed that the extent of mantle melting ishigh beneath hotspot-influenced shallow ridges, andis low beneath deep ridges away from hotspots.These results have led to the recognition of globalcorrelation of MORB chemistry with ridge axial

w xdepth and crustal thickness 5 , and the notion thatmantle temperature variation exerts the primary con-trol on the extent of melting beneath global ocean

0012-821Xr97r$12.00 Copyright q 1997 Elsevier Science B.V. All rights reserved.Ž .PII S0012-821X 96 00218-X

( )Y. Niu, R. HekinianrEarth and Planetary Science Letters 146 1997 243–258´244

Ž . X Ž .Fig. 1. a Location of the Garrett Transform offsetting the EPR at ;13828 S. b Morphotectonic representation of the Garrett TransformŽ . w x Ž . Ž .showing the volcanic ridges Alpha, Beta, and Gamma and transform troughs 17,18 . c A detailed bathymetry 100 m intervals of the

Ž .Garrett Deep and northern flank of the East Median Ridge showing the extent of the outcrop of peridotite–basalt association shaded area .Ž .The thick lines are diving tracks e.g., GN 01, GN 03, etc. . GS 03 is the track of a deep-towed bottom camera run. The samples studied are

Ž . Ž .from dives GN 03, GN 04, GN 13, GN 14 and GN 15. The observed lithologies are: fresh B and altered b basalt flows, consolidated andŽ . Ž . Ž .metamorphosed breccia Mbr , metabasalt Mb , and serpentinized peridotites P .

w xridges 2,5–8 . This conclusion is, however, basedlargely on data from slow-spreading ridges in theAtlantic and Indian Oceans. At the fast-spreading

Ž .East Pacific Rise EPR , there is little correlationw xbetween MORB chemistry and ridge depth 8–11 ,

and EPR MORB show chemical systematics thatdiffer from those at slow-spreading ridges on both

w xregional and ridge segment scales 8–12 . Clearly, an

improved understanding of EPR MORB genesis re-quires petrological data on both residual peridotitesand associated basalts. However, there had been no

w xdetailed sampling of peridotites from the EPR 13,14w xuntil very recent submersible 15–20 and drilling

w x21 investigations at Hess Deep, and the Garrett andTerevaka transforms.

In this paper, we present geochemical data on

Notes to Table 1:a Ims impregnation; total volume of impregnation veinsrveinlets were estimated on board. Thin-section chips and analyzed portions ofharzburgite samples are fresh cuttings away from impregnation veinlets and hydration haloes.b Ž .The reported mineral modes vol% were obtained by point-counting at 1 mm intervals based on one or two thin sections for each sample,and are partially reconstructed. We assigned all textural serpentines, whether or not they have olivine psedomorphs, and associated

Ž . Ž . Ž .magnetite abundant to the serpentine Serp category; optically fresh unaltered olivine, either as individual grainsraggregates or asŽ .‘‘meshes’’ surrounded or veined by serpentine, to olivine Ol ; unaltered orthopyroxene, spatially associated bastite pseudomorphs and

Ž .peripheral talc to primary orthopyroxene Opx ; unaltered clinopyroxene and associated tremolitic amphiboles to primary clinopyroxeneŽ . Ž .Cpx ; spinel with or without magnetite rims to primary spinel Sp .

( )Y. Niu, R. HekinianrEarth and Planetary Science Letters 146 1997 243–258´ 245

Tab

le1

Gar

rett

tran

sfor

msa

mpl

ede

scri

ptio

nsa

bSa

mpl

eD

epth

Setti

ngSa

mpl

eSa

mpl

eIm

Mod

esof

sam

ples

anal

ysed

Cou

nts

size

desc

ript

ion

Ž.

Ž.

Ž.

Ž.

Ž.

mcm

On

boar

dvo

l%Se

rpO

lO

pxC

pxSp

tota

l

Serp

enti

nize

dha

rzbu

rgit

eG

N3-

943

24O

utcr

opon

25=

29=

13H

arzb

urgi

te;n

oim

preg

natio

n0

79.1

54.

4015

.00

nil

1.45

2000

asc

arp

Ž.

GN

14-1

5050

Tal

us20

=18

=18

Har

zbur

gite

;4ga

bbro

icve

inle

ts2

–5

mm

thic

k8

71.8

010

.16

15.2

40.

851.

9512

37Ž

.G

N14

-250

46T

alus

13=

15=

15H

arzb

urgi

te;1

gabb

roic

vein

let

-2

mm

thic

k10

83.8

98.

716.

620.

700.

0812

30G

N14

-549

08O

utcr

op12

=14

=12

Har

zbur

gite

;no

impr

egna

tion

049

.29

39.8

89.

470.

221.

1417

10Ž

.G

N14

-748

33T

alus

40=

40=

15H

arzb

urgi

te;1

gabb

roic

vein

let

;2

mm

thic

k1

73.9

27.

3117

.36

0.30

1.12

2007

GN

14-9

4660

Out

crop

16=

16=

5H

arzb

urgi

te;n

oim

preg

natio

n0

49.7

637

.18

11.6

90.

161.

2112

37Ž

.G

N14

-14

4244

Out

crop

30=

20=

9H

arzb

urgi

te;3

gabb

roic

vein

lets

2–

10m

mth

ick

2074

.01

11.6

513

.07

0.24

1.03

1263

Ž.

and

1di

kele

t3

mm

thic

kG

N15

-246

47T

alus

,fla

nk15

=17

=17

Dun

ite;n

oim

preg

natio

n0

68.2

722

.97

7.41

nil

1.35

1258

ofth

eri

dge

GN

15-5

4688

Out

crop

,17

=10

=10

Har

zbur

gite

;no

impr

egna

tion

083

.49

0.34

14.5

0ni

l1.

6812

05fl

ank

ofth

eri

dge

GN

15-1

042

17O

utcr

op21

=12

=8

Har

zbur

gite

;no

impr

egna

tion

051

.87

22.3

724

.27

0.08

1.41

1212

Dia

base

GN

4-3

4602

Tal

us,a

ctiv

e25

=20

=15

Plag

iocl

ase–

oliv

ine

phyr

ic,

;eq

uala

mou

nts

plag

iocl

ase

and

clin

opyr

oxen

e,pl

ustr

ace

ofol

ivin

ean

dm

agne

tite

tect

onic

zone

GN

4-9

4013

Out

crop

28=

20=

20M

etam

orph

osed

,aph

yric

,;

equa

lam

ount

spl

agio

clas

ean

dcl

inop

yrox

ene

ingr

ound

mas

sG

N13

-937

64T

alus

40=

20=

8A

phyr

ic,e

qual

amou

nts

plag

iocl

ase

and

clin

opyr

oxen

epl

ustr

ace

oliv

ine

and

opaq

ueG

N15

-446

97T

alus

,fla

nk26

=16

=10

Aph

yric

,equ

alam

ount

spl

agio

clas

ean

dcl

inop

yrox

ene

plus

;3%

opaq

ueof

the

ridg

e

Bas

alt

GN

4-1

5062

Tal

us,a

ctiv

e25

=20

=20

Pillo

wfr

agm

ent;

aphy

ric

glas

ste

cton

iczo

neG

N4-

1137

20O

utcr

op,

20=

10=

10Pi

llow

frag

men

t;ap

hyri

cgl

ass

dike

sG

N13

-145

28O

utcr

op20

=15

=15

Shee

tflo

w;p

lagi

ocla

se–

oliv

ine

basa

ltic

glas

sG

N13

-636

57O

utcr

op25

=25

=12

Pillo

wfr

agm

ent;

aphy

ric

glas

sG

N13

-837

64T

alus

40=

20=

15Pi

llow

frag

men

t;ap

hyri

cno

n-gl

assy

basa

ltG

N15

-148

01T

alus

,28

=14

=12

Pillo

wfr

agm

ent;

oliv

ine–

phyr

icgl

ass

Cen

tral

Bas

in


peridotites and spatially associated basaltic samplesfrom the Garrett Transform. This peridotite–basaltassociation allows, for the first time, a close exami-nation of mantle melting and melt extraction pro-cesses from both melts and residues produced in acommon environment at the EPR. Also, this study isthe first attempt at whole-rock trace element system-atics of abyssal peridotites, and will be a usefuladdition to the geochemical data base for modelsconcerning MORB genesis and chemical geodynam-ics in general.

2. Geological background

The Garrett Transform is associated with theŽ . w xfastest 145 mmryr 22 strike-slip motion on the

Earth, and offsets the EPR at 13828XS by 130 kmw x Ž .23 Fig. 1a . This transform is one of the fewtransforms in which active volcanism is observedw x17–20,24 , and mantle peridotites are recoveredw x13,14,17–20 at the EPR. Previous studies of some

w xdredged samples from this transform 25,26 war-ranted an in situ observation by the submersible

w xNautile 17 . In the Garrett Transform, active volcan-ism is primarily taking place along three parallelridges oblique to the transform strike, but fresh lavasalso exist at deep locations near the peridotite out-

Ž .crops Fig. 1b . Mantle peridotites are exposed as an;10 km long sliver in the transform valley and

Žalong the northern flank of the median ridge Fig..1c . Gabbros occur as a dike complex within the

peridotite domain, and diabasic dikes of variablethickness are common. The peridotites are variouslyserpentinized, and are locally impregnated with

Žbasaltic veinsrveinlets of variable thickness -1.mm to a few centimetres .

3. Samples and analytical techniques

Field occurrence, hand specimen characteristics,and petrography of the samples studied are summa-rized in Table 1. These include 10 serpentinizedharzburgites; 4 fresh diabases; 1 fresh non-glassybasalt; and 5 fresh glassy basalts. The analyzedportions of harzburgites are fresh cuttings away fromimpregnation veinlets and seafloor weathering haloes.

Major element compositions of residual minerals inharzburgites and basaltic glasses were analyzed us-

w xing a Camebax SX 50 microprobe at IFREMER 18 .Whole-rock major elements for non-glass basalts,diabases and harzburgites, and Cr and Ni forharzburgites were analyzed by XRF at The Univer-

w xsity of Queensland following 27,28 . Trace elementsŽ .except for Cr and Ni in harzburgites were analyzedusing a FISONS PQ inductively coupled plasmamass spectrometer at The University of Queensland.The whole-rock analytical data and uncertainties forrepeated analyses of reference standards PCC-1 andBIR-1 for trace elements are given in Table 2. Wehave determined 37 trace elements for all the sam-ples but we report here rare-earth elements, Y, Zr,Nb, Hf, Ta, and Th, because all other elements areobserved to be mobile during serpentinization. Sam-ple preparation and analytical procedure can be ob-tained from the authors and are described in anEPSL Online Background dataset 1.

4. Data and interpretations

4.1. Basaltic liquids

Fig. 2 shows that the large range of incompatibleelement abundances in basaltic samples can be ex-plained by various extents of fractional crystalliza-

Žtion from a compositionally similar parent Fig. 2a–. Žd . This is better described by glass samples liquid

.compositions . The inflected trend on CaOrAl O2 3Ž .vs. MgO plot Fig. 2c defined by glass samples

suggests that crystallization follows the sequence ofolivine ™ olivine q plagioclase ™ olivine qplagioclase q clinopyroxene. The different abun-dance levels on chondrite-normalized rare-earth ele-

Ž . Ž .ment REE diagrams Fig. 2e–f are consistent withŽ .varying extents of fractional crystallization Fig. 2d .

The overall parallel REE patterns suggest that theirparental melts may have formed by similar extents ofmelting of a similarly depleted fertile mantle. Byassuming a fertile mantle that is slightly more de-

Žpleted than source for average MORB see caption

1 Žhttp:rrwww.elsevier.nlrlocaterepsl mirror site USA,.http:rrwww.elsevier.comrlocaterepsl


.of Fig. 2 , the primary melts parental to these basalticliquids can be explained by G20% melting by either

Ž . Ž .the batch Fig. 2g or fractional Fig. 2h meltingmodel. While the extent of melting calculated usingincompatible trace elements alone is subject to largeerrors, due to source compositional uncertainties, thiscalculated value is nevertheless comparable with 23

Ž"2% melting calculated from major elements Na8. w xf2.23"0.09% and Ca rAl f0.87"0.01 7 .8 8

4.2. Harzburgitic residues

4.2.1. Mineral modes and whole-rock major elementcompositions

Fig. 3a shows that, compared with abyssal peri-w xdotites from slow-spreading ridges 4,29,30 , Garrett

peridotites are extremely depleted harzburgites withŽ .clinopyroxene cpx being essentially absent. This

suggests that these harzburgites are residues of highextents of melting. Also, these harzburgites havevariable and high abundances of ‘‘total modal

Ž .olivine’’ olivineqserpentine . The apparent higholivine modes could be over-estimated because or-

Ž .thopyroxene opx is often subject to serpentinizationw xas well 31 , and volume expansion is often associ-

ated with serpentinization. However, petrographyshows that serpentinization is primarily after olivine

Ž .in these harzburgites see note to Table 1 , which isalso evident from the inverse correlation between

Ž .olivine and serpentine modes Fig. 3b . Note that thecorrelation in Fig. 3b is not due to data closurebecause the opx mode is not constant. The observa-tion that olivine pseudomorphs are preserved in mostsamples suggests that the effect of volume expansionon the estimated olivine mode can be neglected forthis purpose. Importantly, the significant inverse cor-relation of ‘‘total olivine’’ modes with olivine

Ž . Ž .forsterite Fo contents Fig. 3c demonstrates thepresence of excess olivine in these harzburgites priorto serpentinization. The excess olivine, by its vari-able and low Fo contents, is inconsistent with beinga residual phase, but is consistent with being crystal-lized from a basaltic melt. In terms of whole-rockmajor element compositions, the high FeO contents

Ž .in these harzburgites Table 2 are largely controlledby the presence of excess olivine of low Fo contentŽ .Fig. 3d . Therefore, these harzburgites are not sim-ple residues.

4.2.2. Major element compositions of residual miner-als

a Ž wIt has been demonstrated that Cr sCrr Crqx.Al in spinel and Al O contents in opx and cpx of2 3

residual peridotites are sensitive indicators of thew xextent of melting 1,3,32,33 because Al is a moder-

ately incompatible element and its abundances inthese phases decrease with progressive melting. Fig.4 shows that these minerals in Garrett harzburgitesare extremely depleted in Al O , and plot at the2 3

most depleted end of the abyssal peridotite arrayŽdefined by samples from slow-spreading ridges i.e.,

the Mid-Atlantic Ridge and Southwest Indian Ocean.Ridge . Note that Garrett residual minerals are as

depleted as, or even more depleted than, peridotitesŽ .from hotspot-influenced ridges points 5–7 . Com-

parison with melting residues produced by peridotitew xmelting experiments 32 , the highly depleted resid-

ual mineral compositions and the point close to cpxŽ .exhaustion Fig. 3a suggests that these harzburgites

are residues of ;25% melting. This high extent ofmelting is comparable to that estimated from majorand REE abundances in the associated basaltic liq-

Ž .uids see above , but is significantly higher thanF10%, a value that has been advocated as theaverage extent of melting beneath global ocean ridgesw x34,35 .

4.2.3. Whole-rock minor and trace element systemat-ics

We have demonstrated that Garrett harzburgitesare not simple residues, but residues of very high

Ž .extents of melting Fig. 3aFig. 4 that have beenmodified with additional olivine with low Fo contentŽ .Fig. 3c . The low-Fo olivine can only be explainedby its crystallization from a cooling melt. The signa-ture of such a melt component is well preserved inthe whole-rock incompatible element compositionsŽ .Fig. 5 . The whole-rock incompatible element abun-dances correlate positively with ‘‘total olivine’’

Ž .modes Fig. 5a and inversely with olivine Fo con-Ž .tents Fig. 5b–d . These correlations, which are in-

consistent with melting, but consistent with refertil-ization of a melt, indicate that both excess olivineand elevated abundances of incompatible elements inthese harzburgites are the consequence of the samemelt–solid equilibration process. Buoyant melts thatascend through previously depleted melting residues


Tab

le2

Maj

oran

dtr

ace

elem

ent

anal

yses

ofha

rzbu

rgit

es,

diab

ases

and

basa

lts

from

the

Gar

rett

Tra

nsfo

rm

Sam

ple:

3-9

14-1

14-2

14-5

14-7

14-9

14-1

415

-215

-515

-10

PC

C-1

RS

D4-

34-

913

-813

-915

-44-

14-

1113

-113

-615

-1B

IR-1

RS

DŽ

.Ž

.H

arz

Har

zH

arz

Har

zH

arz

Har

zH

arz

Har

zH

arz

Har

z%

Dbs

Dbs

Bsl

tD

bsD

bsG

lass

Gla

ssG

lass

Gla

ssG

lass

%

Wei

ght

per

cen

tb

yX

RF

Wei

ght

per

cen

tb

ym

icro

pro

be

SiO

39.7

639

.54

43.8

540

.10

41.4

140

.64

39.9

038

.48

36.4

742

.43

51.0

050

.28

50.7

850

.83

49.0

750

.78

50.1

350

.58

50.4

750

.50

2

TiO

0.05

0.05

0.06

0.10

0.02

0.08

0.06

0.05

0.07

0.02

1.17

1.27

2.05

2.25

1.09

1.25

1.11

1.16

1.73

2.45

2

Al

O0.

600.

720.

750.

690.

580.

930.

690.

760.

450.

5314

.76

14.2

113

.48

12.6

915

.25

14.1

715

.40

15.5

914

.73

14.9

62

3

FeO

t8.

6410

.58

9.64

11.1

57.

9210

.36

10.1

512

.10

13.2

88.

309.

969.

8811

.76

12.6

38.

979.

819.

078.

9510

.15

10.0

7M

nO0.

060.

130.

160.

160.

110.

150.

130.

120.

130.

120.

180.

180.

200.

210.

190.

190.

190.

160.

140.

15M

gO39

.27

36.9

135

.37

36.3

039

.09

38.2

137

.25

38.0

538

.11

37.8

87.

668.

656.

876.

498.

528.

058.

958.

797.

576.

83C

aO0.

101.

243.

011.

230.

790.

780.

900.

200.

150.

212

.68

11.3

011

.31

11.5

112

.94

12.2

912

.24

12.3

411

.27

10.2

6N

aO

0.27

––

––

––

––

–2.

342.

482.

742.

701.

832.

232.

182.

372.

673.

712

KO

––

0.01

––

––

––

0.01

0.10

0.04

0.20

0.15

0.02

0.04

0.03

0.04

0.11

0.19

2

PO

0–

––

––

––

––

0.07

0.08

0.18

0.20

0.05

0.06

0.05

0.10

0.15

0.06

25

LO

I10

.87

10.0

46.

219.

809.

118.

1810

.86

9.53

10.5

410

.33

–0.

980.

09–

1.34

––

––

–T

otal

99.6

399

.20

99.0

699

.54

99.0

499

.33

99.9

499

.28

99.1

999

.82

99.9

199

.36

99.6

799

.66

99.2

610

0.24

99.3

510

0.09

98.9

999

.18

Mg

a89

.086

.186

.785

.389

.886

.886

.784

.983

.689

.160

.46

3.45

3.6

50.5

65.3

61.9

66.2

66.1

59.6

57.3

()

Par

tsp

erm

illi

onp

pm

by

XR

FC

r21

5421

6820

0324

3026

6440

3433

3823

2019

7519

62N

i18

4718

1616

3416

2622

1820

6821

1619

8217

7622

11

()

Par

tsp

erm

illi

onp

pm

by

ICP

–M

SS

c9.

6410

.511

.617

.89.

210

.39.

486.

917.

149.

298.

653.

936

.639

.839

.244

.837

.536

.426

.433

.336

.335

.943

.51.

6V

2428

3150

2332

2413

2421

23.3

20.

826

026

433

537

924

525

519

323

530

233

128

33.

5C

r–

––

––

––

––

––

–13

722

423

227

242

537

231

739

433

030

737

33.

1C

o10

711

311

111

710

911

511

014

013

311

412

73.

749

4946

4547

4840

4746

5155

7.1

Ni

––

––

––

––

––

––

4748

8966

102

6792

129

102

153

170

2.1

Ga

0.68

1.05

1.64

0.97

0.53

1.02

0.95

0.94

1.11

0.63

0.50

34.

116

16.2

17.8

18.1

15.7

15.7

15.4

15.8

17.5

18.9

15.4

2.3

Y0.

480.

991.

181.

850.

220.

881.

180.

681.

900.

120.

079

7.9

24.0

24.9

40.1

43.3

23.0

22.6

18.3

22.6

34.2

45.6

13.5

2.4

Zr

0.39

0.96

1.88

0.80

0.12

1.37

2.27

0.57

1.64

0.03

0.19

26.

548

.356

.411

712

542

.549

.136

.046

.398

.113

113

.12.

0

()

()

Par

tsp

erb

illi

onp

pb

by

ICP

–M

SP

arts

per

mil

lion

pp

mb

yIC

P–

MS

Nb

118.

816

128.

214

127.

668

6.1

4217

.50.

921.

123.

133.

300.

860.

930.

570.

772.

442.

050.

541.

2L

a25

3772

7313

4540

538

5937

327.

61.

411.

813.

994.

141.

371.

420.

981.

243.

183.

610.

583.

4C

e29

106

169

218

1686

124

9327

811

057

3.3

5.02

6.33

13.0

113

.59

5.03

5.01

3.55

4.63

10.6

112

.89

1.84

1.7

Pr

4.7

2130

382.

819

2716

7017

814

.71.

001.

232.

402.

521.

020.

980.

720.

961.

972.

480.

381.

8N

d33

125

180

225

1511

816

677

465

7128

14.1

5.55

6.56

12.1

812

.98

5.64

5.36

4.10

5.39

10.1

113

.07

2.24

1.7

Sm

1753

7811

56.

455

7629

203

137

26.7

2.23

2.47

4.24

4.54

2.24

2.05

1.64

2.15

3.54

4.68

1.02

1.5

Eu

4026

4155

3.5

2328

8.3

1330

249

.50.

880.

951.

471.

570.

890.

820.

660.

871.

251.

550.

491.

7G

d31

9013

621

013

9312

655

287

148

34.1

3.22

3.51

5.72

6.27

3.17

2.94

2.39

3.09

4.81

6.39

1.67

3.2

Tb

7.8

1926

412.

819

2412

512.

32

28.8

0.59

0.63

1.02

1.11

0.57

0.55

0.44

0.57

0.85

1.15

0.32

2.6

Dy

6814

519

532

627

140

188

9634

416

127.

84.

204.

417.

127.

734.

033.

823.

123.

965.

958.

012.

382.

2H

o18

3444

718.

732

4025

724.

23

21.0

0.88

0.92

1.48

1.61

0.84

0.81

0.66

0.83

1.23

1.67

0.52

2.4

Er

6111

214

521

731

107

132

8422

115

1311

.32.

632.

714.

404.

742.

462.

401.

952.

423.

664.

911.

542.

5T

m10

1923

335.

417

2214

352.

63

20.5

0.40

0.41

0.66

0.72

0.37

0.36

0.30

0.37

0.56

0.75

0.22

2.1

Yb

8113

915

821

944

122

148

9723

026

227.

32.

522.

594.

164.

512.

312.

301.

862.

313.

514.

691.

492.

8L

u15

2427

358.

620

2517

395.

36

9.6

0.38

0.39

0.63

0.68

0.34

0.34

0.28

0.34

0.53

0.70

0.22

2.1

Hf

1630

6545

3.7

4979

1944

1.3

817

.11.

491.

703.

303.

561.

401.

431.

111.

472.

733.

700.

531.

8T

a6.

82.

22.

81.

73.

23.

22.

12.

67.

12.

111

7.0

0.07

90.

097

0.24

70.

265

0.07

20.

077

0.05

90.

065

0.19

40.

177

0.04

91.

1T

h1.

31.

82.

51.

61.

41.

63.

81.

68.

11.

810

9.8

0.04

50.

057

0.18

90.

200

0.04

00.

044

0.03

00.

037

0.14

80.

111

0.03

06.

6


crystallize olivine at shallow levels as a result ofcooling. Entrapment of these melts raises the abun-dances of incompatible elements in the whole-rockharzburgites.

Fig. 6 shows chondrite-normalized REE abun-dances in the harzburgites. For comparison, model

Ž .melting residues by batch Fig. 6a and fractionalŽ .Fig. 6b melting also are shown. Obviously, boththe abundance levels and patterns are inconsistentwith their being simple melting residues. The abun-dance levels are variable and higher than expectedgiven the similarly depleted residual mineral compo-sitions in these harzburgites. The slopes for lightREEs are too shallow for melting residues. Also,batch melting cannot in any way explain samples of

Ž .low REE abundances i.e., GN14-7 and GN15-10 .Some samples show no obvious Eu anomalies,whereas others show either negative or positive Euanomalies. Note that GN15-2 has anomalously highLa, which, together with smaller La anomalies inseveral other samples, is interpreted as La beingmobile during serpentinization. Any process that in-volves a melt would have affected Ce and Pr, at thevery least, as well as La. Although the effects ofplagioclase may be invoked to explain the Euanomalies, the absence of plagioclase in the harzbur-gites analyzed does not support this interpretation.Given that Sr lacks any systematics in these rocksŽ .not shown , it is possible that Eu, like Sr, may alsobe quite mobile during serpentinization. In any case,the overall patterns and variable abundance levels

Ž .are consistent with refertilization of a melt Fig. 5c .

4.2.4. EÕaluation of the extent of refertilizationTo evaluate precisely the extent of refertilization

is not straightforward. However, the relative extentof refertilization can be estimated using REEs byadding a melt with composition similar to GN4-11 to

residues of 25% fractional melting. Such a melt isŽ .chosen because: 1 all the Garrett basaltic liquids

Ž . Ž .have similar REE patterns Fig. 2e ; 2 GN4-11 hasŽonly olivine on the liquidus see Fig. 2c; olivine is

the only phase observed to have crystallized from the. Ž . arefertilizing melt ; and 3 GN4-11 has an Mg of

Ž 3q w 3q 2qx;63.8–66.2 assuming Fe r Fe qFe f0–.0.1 in the melt , which would be in equilibrium with

olivine of Fo , comparable to the observed85.4 – 86.7

excess olivine. A residue of 25% fractional meltingŽ .is indicated by the near exhaustion of cpx Fig. 3a

and highly depleted residual mineral compositionsŽ .Fig. 4 , and batch melting cannot explain the low

Ž .REE abundance samples Fig. 6 . The results areŽ .shown in Fig. 7a–c for representative light Ce ,

Ž . Ž .intermediate Tb , and heavy Yb REEs. The amountŽ .of GN4-11 f required to match the data varies

from 0.1% to ;12%. Note that the f values derivedfrom REEs also explain the abundances of minorŽ . Ž .TiO and major FeO element variations, and2

Ž .olivine Fo contents in the harzburgites Fig. 7d–f .

( )4.2.5. Issues on high field strength elements HFSEsFig. 8a–c plots primitive mantle normalized in-

compatible element abundances in the harzburgites.The patterns display several anomalous features con-cerning HFSEs. Not shown, Garrett basaltic liquidsexhibit rather smooth patterns. Th is more depletedthan Nb in the basaltic liquids, yet the opposite isobserved in the harzburgites. Nb is more depletedthan Ta in both liquids and harzburgites. A positiveTi anomaly exists in all but GN15-5 samples. Whilepositive Zr–Hf anomalies are present in someharzburgites, the large negative anomalies in othersare striking. Since the refertilization process dis-cussed above affects Ti as well as the ‘‘adjacent’’

Ž .REEs Fig. 7 , positive Ti anomalies would havealready existed prior to refertilization, which seems

Note to Table 2:Harzsserpentinized harzburgite; Dbssdiabase; Bsltsnon-glassy basalt; Glasssbasaltic glass; FeOts total Fe expressed as FeO;

a w 2qx 2q 2qLOIs loss on ignition; Mg sMgr MgqFe , assuming total FeO as Fe for harzburgites, and 90% total FeO as Fe for diabasesand basalts. All the analyses are whole-rock compositions except for basaltic glasses. Major element data for glasses were analyzed at

w x w xIFREMER using a Camebax SX 50 18 . Whole-rock XRF analyses were done at the University of Queensland following 27,28 . Except forCr and Ni in harzburgites, which were analyzed by XRF, all the trace elements were analyzed at the University of Queensland using a

Ž .Plasma Quad Inductively coupled mass spectrometer ICP–MS . PCC-1saverage of 6 repeated analyses of 2 digestions. BIR-1saverageof 18 repeated analyses of 12 digestions. %RSD is the relative standard deviations about the averages for PCC-1 and BIR-1. Samplepreparation and analytical procedures can be obtained from the authors and are described in an EPSL Online Background dataset.


to be a common feature in highly depleted peri-w xdotites 36 . However, the anomalies associated with

Zr–Hf and Nb–Ta need attention.Zr–Hf anomalies: As Zr and Hf do not show

anomalies in basalts, these anomalies in residualharzburgites are thus unexpected. Rampone et al.w x37 interpreted the negative anomalies of Zr and Srin an N-MORB type ophiolite from Internal Lig-

Ž . Ž . Ž . Ž .Fig. 2. a – d MgO variation diagrams of representative major, minor, and trace elements for Garrett basaltic rocks. e and fw x Ž . Ž . Ž .Chondrite–normalized 42 REE abundances for these basaltic rocks. g and h Comparisons of model melts by batch melting g and

Ž . Ž .fractional melting h with a hypothetical primary melt parental to Garrett basaltic liquids thick shaded line . The primary melt isŽapproximated by adding 12% olivine to GN4-11 to be in equilibrium with mantle olivine Fo i.e., the primary melt would have90

Fe – Mg w x. Ž .Mgaf72, assuming olivinermelt K s0.3 56 . The preferred fertile mantle source thick solid line is the average of R717, thedw xleast depleted, and R238, the moderately depleted, Ronda spinel lherzolites 57 ; this choice was made because Garrett basaltic liquids arew xslightly more depleted than average N-type MORB, for which R717 is ideal 57 . This fertile mantle has subsolidus modes of 66% olivine

Ž . Ž . Ž . Ž .Ol , 24% orthopyroxene Opx , 9% clinopyroxene Cpx , and 2% spinel Spl , but we have assumed high pressure modes of 55% Ol, 30%w x w xOpx, 13% Cpx, and 2% Spl following 58 . The melting modes are taken from the polybaric melting reaction of 52 : 0.464 Cpxq0.681

Opxq0.048 Spl´0.193 Olq1 Melt. Mineralrliquid partition coefficients are given in Appendix A.


uride, North Italy, as reflecting ‘‘changes in therelative magnitude of Zr, Sr, and REE partitioncoefficients, depending on specific melting condi-tions’’. As Sr in Garrett harzburgites is quite mobile,whether Sr exhibits similar features prior to serpen-

Ž .tinization is unknown. Since both positive weakŽ .and negative strong Zr–Hf anomalies are present

Ž .Fig. 8a–c , variation in melting condition may notbe important. Both positive and negative Zr anoma-lies have been observed in melt inclusions in high-Mg

w xolivine phenocrysts in basalts 38,39 , and were in-terpreted as representing unmixed ‘‘true’’ instanta-neous melt fractions. If so, this implies that ‘‘micro-scale’’ heterogeneities exist with respect to Zr andHf in the mantle. The presence of a Zr–Hf richphase in the fertile mantle is possible, but it isunlikely for it to survive in residues after )20%melting. As Garrett harzburgites have undergone melt

refertilization, it is possible that such a Zr–Hf phaseŽ .ZrO –baddeleyite is a more likely candidate may2

have co-precipitated with the excess olivine. If so,Ž .two possibilities exist: 1 uneven distribution of this

phase in the harzburgites would lead to the observedŽ .Zr–Hf anomalies; 2 positive Zr–Hf anomalies may

w xbe expected in harzburgites 36 , but large negativeZr–Hf anomalies may be due to incomplete diges-tion of this phase during sample preparation. Acareful evaluation of these scenarios are warranted.

Nb–Ta fractionation, and the peculiar Nb–Threlationship: it has been widely accepted that Nb andTa are geochemically identical, and that the NbrTaratio is constant, and is close to chondritic valuesŽ . w x16–18 in all oceanic basalts 40–42 . Very recent

w xstudies 43,44 have shown that NbrTa ratios inMORB are not constant, but vary significantly, from18 in highly enriched lavas to as low as 9 in ex-

Ž . Ž . Ž .Fig. 3. a Orthopyroxene opx and clinopyroxene cpx modes are plotted against olivine modes to compare Garrett harzburgites withw xabyssal peridotites from slow-spreading ridges discussed by 4,29,30 . For Garrett harzburgites, the ‘‘total olivine’’ plotted is the sum of

Ž . Ž . Ž .olivine and serpentine see Table 1 modes. b Olivine modes correlate inversely with serpentine modes in Garrett harzburgites. c OlivineŽ . Ž .forsterite Fo contents inversely correlate with ‘‘total olivine’’ in Garrett harzburgites. d Relative to whole-rock analyses of abyssal

w x w xperidotites from the Southwest Indian Ocean Ridge 59 and to reconstructed whole-rock abyssal peridotite compositions 52 using thew x Žpublished modes and mineral analyses of Dick 4,29,30 , whole-rock Garrett harzburgites show low MgOrFeO and SiO rFeO ratios high2

.FeO; see Table 2 , which are consistent with the presence of low-Fo olivine. The diamonds symbols are olivine compositions plotted forŽ .reference. Note that Garrett whole-rock compositions show MgO deficiency relative to SiO Table 2 , perhaps due to MgO loss during2

serpentinization.


tremely depleted ones, indicating that Nb is moreincompatible than Ta during melting. If so, onewould expect even lower NbrTa ratios in residualharzburgites. This is indeed the case. Fig. 8f showsthat the NbrTa ratio increases with increasing extent

Fig. 4. In comparison, residual minerals in Garrett harzburgitesplot at the most depleted end of the data array defined by those in

w xabyssal peridotites from slow-spreading ridges 2,4,29,30,60,61 .These harzburgites are as depleted as, or even more depleted than,

Ž .peridotites from hotspot-affected ridges e.g., points 5–7 . TheGarrett mineral data are from samples with olivine Fo)89,

w xincluding new analyses of samples previously studied 25,26 .These indicate that, prior to olivine addition, Garrett harzburgites

Ž .were residues of very high extents ;25% of melting. Thenumbered symbols are regional averages: 1sAtlantis II faultzone; 2sVulcan fault zone; 3s Islas Orcadas fault zone; 4s

Ž .Bullard fault zone; 5sBouvet fault zone near Bouvet hotspot inŽ w x.the Indian Ocean; 6s158N axial valley near 148N hotspot 62 ;

Ž .7s438N fault zone in the Atlantic near Azores hotspot .

Ž .Fig. 5. a ‘‘Total olivine’’ modes correlate positively with incom-Ž .patible element e.g., Ga and Ti abundances in Garrett harzbur-

Ž . Ž . Ž .gites. b – d Olivine forsterite Fo contents in these harzburgitesŽcorrelate inversely with incompatible element e.g., TiO , Sm and2

.Nb abundances. The values in parentheses are confidence levelsof the correlation coefficients. Note that no olivine is analyzed

Žfrom sample GN15-2 due to almost complete serpentinization see.Table 1 .

of refertilization, indicating that the NbrTa ratio intruly depleted residues unaffected by melt refertiliza-tion would be as low as 2. Given that Nb and Ta are


w x Ž .Fig. 6. Chondrite 42 normalized rare earth element abundances of Garrett harzburgites. For comparison, calculated depletion curves by aŽ .batch and b fractional melting models are also plotted. The model parameters are as for Fig. 2g–h. Note that the abundance levels and

patterns are inconsistent with their being simple melting residues by either model, given their overall similarly depleted in residual mineralŽ .compositions Fig. 4 , but are consistent with melt refertilization. Also note that the anomalous La in GN3-9, 14-7, and 15-2 is interpreted as

resulting from relative mobility of La during serpentinization.

chemically quite similar, it is difficult to explain thevariable NbrTa ratios by chemical considerationalone. Because Nb and Ta have very different atomic

93 181 Žmasses, Nb vs. Ta the superscripts are atomic.masses , we propose that diffusion may control their

apparent incompatibility. Heavier elements woulddiffuse slower than lighter ones, and would be evenslower when the abundance level is extremely lowŽ .in residues due to reduced compositional gradient

across the mineral grains undergoing melting. Thisexplains the curved NbrTa vs. Nb and Th trendsŽ .Fig. 8d,e with large NbrTa changes in more de-

w x w x 90 – 92pleted rocks 43 . The observation 43 that Zr ismore incompatible than 177 – 180 Hf, and 85,87Rb ismore incompatible than 133Cs in highly depletedMORB can be explained likewise. The positive Th

Ž .anomaly relative to Nb in the harzburgites mayŽ232 93also be explained likewise Th vs. Nb; i.e., low


Ž . Ž . Žw x . Ž . Ž .Fig. 7. a – c Comparisons between the data and calculated abundances CN , chondrite-normalized of light Ce , intermediate Tb , andŽ . Ž . Ž .heavy Yb REE abundances. The calculation is done by adding various amounts f of a basaltic melt MsGN4-11 to a residue

Ž . Ž . Ž .Rs25% fractional melting by simultaneously solving the equation fs C yC r C yC for all, but La and Eu, REEs. Note thatData R M RŽ .the fits are less good for light REEs given their more variable patterns for some samples Fig. 6 , but are excellent for intermediate and

Ž . Ž .heavy REEs. d – f Plots of whole-rock TiO and FeO, and olivine forsterite contents with the calculated f to show that these significant2

correlations are consistent with melt refertilization.

.NbrTh ratios despite their different chemical prop-erties.

5. Implications

5.1. What causes the high extents of melting beneaththe EPR: a hot mantle or fast spreading rate?

We have shown that Garrett harzburgites are ex-tremely depleted, as depleted as, or even more de-pleted than, abyssal peridotites from hotspot-in-fluenced ridged in the Atlantic and Indian oceansŽ .Fig. 4 . In addition, recently drilled samples from

w xHess Deep 21 , and our unpublished data from bothw xHess Deep and the Terevaka Transform 19 also are

similarly depleted. One may interpret these highlydepleted harzburgites as resulting from a depletedfertile mantle, but this requires that the observationsat the three locations be fortuitous. Importantly,MORB from the broad EPR region are not highlydepleted. Therefore, the highly depleted harzburgitesfrom these locations are consistent with very high

Ž .extents of melting ;25% . If the extent of meltingrelates in a simple way to mantle temperature varia-

w xtion 5,8 , this would suggest a hotter EPR mantle.However, the EPR in this broad region has normaltopography and morphology thermally unaffected by


Ž . Ž . w xFig. 8. a – c Primitive mantle 42 normalized incompatible element abundances in Garrett harzburgites to show some anomalous featuresŽ .about Ti, Zr, Hf, Nb, Ta and Th relative to REEs. d Plot of NbrTa vs. Nb of all the lithologies analyzed from the Garrett Transform to

w x Ž .show that the curved trend is consistent with the observation that Nb is more incompatible than Ta 43,44 . e Plot of NbrTa vs. Th toŽ . Ž .show that NbrTa ratio in a rock increases with increasing incompatible element e.g., Th abundances. f The significant correlation

Ž .between NbrTa ratio and the calculated refertilization parameter Fig. 7 attests that the true melting residues would have NbrTa-2, andthe variably higher NbrTa ratios result from the refertilization process discussed in the text.

any known hotspots. Hence, factors other than man-tle temperature variation may play a role. We pro-pose that plate spreading rate variation may be asimportant as mantle temperature variation. The highextents of melting beneath the EPR may result froma fast spreading rate. Given the fact that mantleupwelling beneath ridges results from plate separa-tion, fast separation leads to fast mantle upwellingw x45–48 . Fast upwelling allows the adiabat to extend

Žto shallower levels i.e., against the conductive cool-.ing to the seafloor , thus decompression melting

continues up to a shallower level, and more meltforms from a given parcel of mantle.

That oceanic crustal thickness is generally con-stant, and is independent of plate spreading ratew x49,50 , has been the primary constraint leading tothe idea that the extent of melting beneath oceanridges is independent of spreading rate. However, thecrustal thickness is entirely inferred from seismicdata. The fact that thin crust and peridotite outcropshave been observed within axial zones at slow-spreading ridges but not at fast-spreading ridges


suggests that crustal thickness derived from seismicw xdata must be used with caution 51 . The significant

Ž .implications are: 1 the notion that MORB representF10% melting beneath global ocean ridges needsrevision with the available peridotite data from the

Ž .EPR; and 2 models on MORB genesis that neglectspreading rate variation need re-evaluation.

5.2. Do MORB melts experience low pressure equili-bration during ascent?

Recent studies have led to the idea that meltsformed at depth must be extracted rapidly withoutexperiencing low-pressure equilibration during as-

w xcent 8 . The presence of excess olivine and incom-patible element refertilization seen in Garrettharzburgites attest the significant melt–solid equili-bration during melt ascent. As excess olivine isobserved on petrographic scales in abyssal peri-

w xdotites from slow-spreading ridges as well 52 ,porous flows may be as important as channel flowsduring melt migration. Therefore, melt–solid equili-bration during melt ascent at shallow levels is un-avoidable. Caution is thus necessary when inferringmelting pressures from basalt chemistry, as has been

w xpointed out in recent studies 53–55 .

Acknowledgements

YN acknowledges support from The University ofQueensland and from the Australian Research Coun-cil. RH acknowledges support from Ministere desAffairs Etrangeres in France as well as the Depart-ment of Marine Geosciences at IFREMER. Both YNand RH are grateful for the Bilateral Science andTechnology Program of Australia and the Ministerede l’Education de l’enseignement Superieur et de laRecherche of France, without which this work wouldnot have been completed. We thank Robyn Frank-land and Kim Baublys for helping with samplepreparations, Peter Colls for thin sections, and FrankAudsley for XRF analyses at the University ofQueensland, and Marcel Bohn for assistance with themicroprobe analysis at IFREMER. We are gratefulfor the constructive comments of Jon Snow and two

[ ]anonymous reviewers. FA

Appendix A. Rare earth element mineralrrrrr liquidpartition coefficients used in this study

Kd Clino- Ortho- Spinel Olivinevalues pyroxene pyroxene

La 0.0560 0.0053 0.0003 0.0067Ce 0.0920 0.0090 0.0005 0.0060Pr 0.1610 0.0127 0.0007 0.0060Nd 0.3030 0.0163 0.0008 0.0059Sm 0.4450 0.0200 0.0009 0.0067Eu 0.5005 0.0275 0.0010 0.0074Gd 0.5560 0.0350 0.0011 0.0095Tb 0.5700 0.0480 0.0013 0.0115Dy 0.5820 0.0610 0.0018 0.0154Ho 0.5825 0.0740 0.0023 0.0193Er 0.5830 0.0838 0.0030 0.0256Tm 0.5625 0.0935 0.0038 0.0374Yb 0.5420 0.1033 0.0045 0.0491Lu 0.5060 0.1130 0.0050 0.0600

w xData from 63–67 .

References

w x1 H.J.B. Dick and R.L. Fisher, Mineralogic studies of theresidues of mantle melting: Abyssal and alpine type peri-dotites, in: The Mantle and Crustal–mantle Relationships —Mineralogical, Petrological, and Geodynamic Processes, 3rdInt. Kimberlite Conf., II, J. Kompobst, ed., pp. 295–308,New York, NY, 1984.

w x2 H.J.B. Dick, R.L. Fisher and W.B. Bryan, Mineralogicalvariability of the uppermost mantle along mid-ocean ridges,Earth Planet. Sci. Lett. 69, 88–106, 1984.

w x3 P.J. Michael and E. Bonatti, Peridotite composition from theNorth Atlantic: Regional and tectonic variations and implica-tions for partial melting, Earth Planet. Sci. Lett. 73, 91–104,1985.

w x4 K.T.M. Johnson, H.J.B. Dick and N. Shimizu, Melting in theoceanic upper mantle: an ion microprobe study of diopside inabyssal peridotites, J. Geophys. Res. 95, 2661–2678, 1990.

w x5 E.M. Klein and C.H. Langmuir, Global correlations of oceanridge basalt chemistry with axial depth and crustal thickness,J. Geophys. Res. 92, 8089–8115, 1987.

w x6 D. McKenzie and M.J. Bickle, The volume and compositionof melt generated by extension of the lithosphere, J. Petrol.29, 625–679, 1988.

w x7 Y. Niu and R. Batiza, An empirical method for calculatingmelt compositions produced beneath mid-ocean ridges: appli-

Ž .cation for axis and off-axis seamounts melting, J. Geophys.Res. 96, 21,753–21,777, 1991.


w x8 C.H. Langmuir, E.M. Klein and T. Plank, Petrological sys-tematics of mid-ocean ridge basalts: Constraints on meltgeneration beneath ocean ridges, in: Mantle Flow and MeltGeneration at Mid-ocean Ridges, J. Phipps Morgan, D.K.Blackman and J.M. Sinton, eds., AGU Geophys. Monogr.71, 183–280, 1992.

w x9 J.P. Brodholt and R. Batiza, Global systematics of unavengedmid-ocean ridge basalt compositions: Comments on ‘‘Globalcorrelations of ocean ridge basalt chemistry with axial depthand crustal thickness by E.M. Klein and C.H. Langmuir’’, J.Geophys. Res. 94, 4231–4239, 1989.

w x10 E.M. Klein and C.H. Langmuir, Local versus global variationin ocean ridge basaltic composition: A reply, J. Geophys.Res. 94, 4241–4252, 1989.

w x11 R. Batiza, R. Hekinian, D. Bideau and J. Francheteau, Chem-´Ž .istry of deep 3500–5600 m Pacific MORB — Why is the

Pacific anomalous?, Geophys. Res. Lett. 22, 3067–3070,1995.

w x12 Y. Niu and R. Batiza, Chemical variation trends at fast andslow spreading ridges, J. Geophys. Res. 98, 7887–7902,1993.

w x13 Yu.P. Neprochnov and G.L. Kashintsev, Composition of themain crustal layers of the East Pacific Rise, Oceanology 239,199–201, 1978.

w x14 H. Backer, J. Lange and V. Marchig, Hydrothermal activityänd sulfide formation in axial valleys of the East Pacific Risecrest between 188S and 228S, Earth Planet. Sci. Lett. 72,9–22, 1985.

w x15 J. Francheteau, R. Armijo, J.L. Cheminee, R. Hekinian, P.´ ´Lonsdale and N. Blum, 1 Ma East Pacific Rise oceanic crust

Žand upper mantle exposed by rifting in Hess deep equatorial.Pacific Ocean , Earth Planet. Sci. Lett. 101, 282–295, 1990.

w x16 R. Hekinian, D. Bideau, J. Franchteau, J.J. Cheminee, R.Ármijo, P. Lonsdale and N. Blum, Petrology of the EastPacific Rise crust and upper mantle exposed in Hess DeepŽ .eastern equatorial Pacific , J. Geophys. Res. 98, 8069–8094,1993.

w x17 R. Hekinian, D. Bideau, M. Cannat, J. Franchteau and R.´Hebert, Volcanic activity and crust–mantle exposure in theúltrafast Garrett transform fault near 13828XS in the Pacific,Earth Planet. Sci. Lett. 108, 259–273, 1992.

w x18 R. Hekinian, D. Bideau, R. Herbert and Y. Niu, Magmatism´ ´Ž X .in the Garrett transform East Pacific Rise near 13827 S , J.

Geophys. Res. 100, 10,163–10,185, 1995.w x19 M. Constantin, R. Hekinian, D. Ackermand and P. Stoffers,´

Mafic and ultramafic intrusions into upper mantle peridotitesŽfrom fast spreading centers of the Easter microplate South-

.east Pacific , in: Mantle and Lower Crust Exposed in OceanicRidges and Ophiolites, R.L.M. Visser and A. Nicolas, eds.,pp. 71–120, Kluwer, Dordrecht, 1995.

w x20 R. Hekinian, R. Armijo, J.P. Cogne, M. Constantin, J.´ ´Francheteau, J. Girardeau, R.N. Hey, D.F. Naar and R.Searle, Petrology of the Easter microplate region in the southPacific, J. Geophys. Res., submitted, 1996.

w x21 H.J.B. Dick and J.H. Natland, Late stage melt evolution andtransport in the shallow mantle beneath the East Pacific Rise,Proc. ODP Sci. Results 147, 103–134, 1996.

w x22 C. DeMets, R.G. Gordon, D.F. Argus and S. Stein, Currentplate motions, Geophys. J. Int. 101, 425–478, 1990.

w x23 P. Lonsdale, Segmentation of the Pacific–Nazca spreadingcenter, 18N–208S, J. Geophys. Res. 94, 12,179–12,225, 1989.

w x24 R. Batiza, B.R. Rosendahl and R.L. Fisher, Evolution ofocean crust 3, Petrology and geochemistry of basalts fromthe East Pacific Rise and Siqueiros transform fault, J. Geo-phys. Res. 82, 265–276, 1977.

w x25 R. Hebert, D. Bideau and R. Hekinian, Ultramafic and mafic´ ´rocks from the Garrett Transform fault near 13830X N on theEast Pacific Rise: igneous petrology, Earth Planet. Sci. Lett.65, 107–125, 1983.

w x26 M. Cannat, D. Bideau and R. Hebert, Plastic deformation and´magmatic impregnation in serpentinized ultramafic rocks

Ž .from the Garrett transform fault East Pacific Rise , EarthPlanet. Sci. Lett. 101, 216–232, 1990.

w x27 K. Norrish and J.T. Hutton, J.T. An accurate X-ray spectro-graphic method for the analysis of a wide range of geologicalsamples, Geochim. Cosmochim. Acta 33, 431–453, 1968.

w x28 Y. Tao, P.A. Pella and R.M. Rousseau, NBSGSC — aFORTRAN program for quantitative X-ray fluorescenceanalysis, NBS Techn. Note 1213, 1985.

w x29 H.J.B. Dick, Abyssal peridotites, very slow spreading ridgesand ocean ridge magmatism, in: Magmatism in the OceanBasins, A.D. Saunders and M.J. Norry, eds., Geol. Soc.London Spec. Publ. 42, 71–105, 1989.

w x30 K.T.M. Johnson and H.J.B. Dick, Open system melting andthe temporal and spatial variation of peridotite and basaltcompositions at the Atlantis II F.Z., J. Geophys. Res. 97,9219–9241, 1992.

w x31 D.R. Janecky and W.E. Seyfried, Hydrothermal serpentiniza-tion of peridotite within the oceanic crust: Experimentalinvestigations of mineralogy and major element chemistry,Geochim. Cosmochim. Acta 50, 1357–1378, 1986.

w x32 A.L. Jaques and D.H. Green, Anhydrous melting of peri-dotite at 0–15 kb pressure and the genesis of tholeiiticbasalts, Contrib. Mineral. Petrol. 73, 287–310, 1980.

w x33 H.J.B. Dick and T.B. Bullen, Chromian spinel as a petroge-netic indicator in abyssal and alpine-type peridotites andspatially associated lavas, Contrib. Mineral. Petrol. 86, 54–76,1984.

w x34 E.M. Klein, T. Plank and C.H. Langmuir, Constraints onmodels for mantle melting beneath ocean ridges, RIDGEEvents 2, 11–12, 1991.

w x35 D.W. Forsyth, Crustal thickness and the average depth anddegree of melting in fractional melting models of passiveflow beneath mid-ocean ridges, J. Geophys. Res. 98,16,073–16,079, 1993.

w x36 W.F. McDonough, H.-G. Stosch and N.G. Ware, Distributionof titanium and the rare earth elements between peridotiticminerals, Contrib. Mineral. Petrol. 110, 321–328, 1992.

w x37 E. Rampone, A.W. Hofmann, G.B. Piccardo, R. Vannucci, P.Bottazzi and L. Ottolini, Trace element and isotope geochem-istry of depleted peridotites from an N-MORB type ophioliteŽ .Internal Liguride, N. Italy , Contrib. Mineral. Petrol. 123,61–76, 1996.

w x38 A.V. Sobolev, and N. Shmizu, Ultra-depleted primary melt


included in an olivine from the Mid-Atlantic Ridge, Nature363, 151–154, 1993.

w x39 A.V. Sobolev, A.A. Gurenko and N. Shmizu, Ultra-depletedmelts from Iceland: data from melt inclusion studies, Min-eral. Mag. 58A, 860–861, 1994.

w x40 H. Bougault, J.-L. Joron and M. Treuil, Alteration, fractionalcrystallization, partial melting, mantle properties from traceelements in basalts recovered in the North Atlantic, in: DeepDrilling Results in the Atlantic Ocean: Ocean Crust, M.Talwani, C.G. Harrison and D.E. Hayes, eds., pp. 352–367,1979.

w x41 A.W. Hofmann, Chemical differentiation of the Earth: therelationship between mantle, continental crust, and oceaniccrust, Earth Planet. Sci. Lett. 90, 297–314, 1988.

w x42 S.-S. Sun and W.F. McDonough, Chemical and isotopicsystematics of ocean basalt: Implications for mantle composi-tion and processes, in: Magmatism of the Ocean Basins, A.D.Saunders and M.J. Norry, eds., Geol. Soc. London Spec.Publ. 42, 323–345, 1989.

w x43 Y. Niu and R. Batiza, Extreme mantle source heterogeneitiesbeneath the northern East Pacific Rise — trace elementevidence from near ridge seamounts, EOS Trans. AGU 76,F595, 1995.

w x44 T. Plank and W.M. White, Nb and Ta in Arc and mid-oceanridge basalts, EOS Trans. AGU 76, F655, 1995.

w x45 Y. Bottinga and C.J. Allegre, partial melting under spreading`ridges, Philos. Trans. R. Soc. London A 288, 501–525, 1978.

w x46 I. Reid and H.R. Jackson, Oceanic spreading rate and crustalthickness, Mar. Geophys. Res. 5 165–172, 1981.

w x47 D.W. Forsyth, Geophysical constraints on mantle flow andmelt generation beneath mid-ocean ridges, in: Mantle Flowand Melt Generation at Mid-ocean Ridges, J.P. Morgan, D.K.Blackman and J.M. Sinton, eds., AGU Geophys. Monogr.71, 1–66, 1992.

w x48 D.L. Turcotte and J. Phipps Morgan, Magma migration andmantle flow beneath a mid-ocean ridge, in: Mantle Flow andMelt Generations Beneath Mid-Ocean Ridges, J. PhippsMorgan, D.K. Blackman and J.M. Sinton, eds., AGU Geo-phys. Monogr. 71, 155–182, 1993.

w x49 Y.J. Chen, Oceanic crustal thickness versus spreading rate,Geophys. Res. Lett. 19, 753–756, 1992.

w x50 J.W. Bown and R.S. White, Variation with spreading rate ofoceanic crustal thickness and geochemistry, Earth Planet. Sci.Lett. 121, 435–449, 1994.

w x51 M. Cannat, How thick is the magmatic crust at slow spread-ing ridges?, Geophys. Res. 101, 2847–2857, 1996.

w x52 Y. Niu, Mantle melting processes beneath slow-spreadingocean ridges: evidence from abyssal peridotites, J. Petrol.,submitted, 1996.

w x53 J.H. Natland, Partial melting of a lithologically heteroge-neous mantle: Inferences from crystallisation histories of

magnesian abyssal tholeiites from the Siqueiros FractureZone, in: Magmatism in Ocean Basins, A.D. Saunders andM.J. Norry, eds., Geol. Soc. London Spec. Publ. 42, 41–70,1989.

w x54 F. Albarede, How deep do common basaltic magmas form`and differentiate?, J. Geophys. Res. 97, 10,997–11,009, 1992.

w x55 Y. Shen and D.W. Forsyth, Geochemical constraints oninitial and final depth of melting beneath mid-ocean ridges, J.Geophys. Res. 100, 2211–2237, 1995.

w x56 P.L. Roeder and R.F. Emslie, Olivine–liquid equilibrium,Contrib. Mineral. Petrol. 29, 275–289, 1970.

w x57 F.A. Frey, C.J. Suen and H. Stockman, The Ronda hightemperature peridotite: Geochemistry and petrogenesis,Geochim. Cosmochim. Acta 49, 2469–2491, 1985.

w x58 M.B. Baker and E.M. Stolper, Determining the compositionof high-pressure mantle melts using diamond aggregates,Geochim. Cosmochim. Acta 58, 2811–2827, 1994.

w x59 J.E. Snow and H.J.B. Dick, Pervasive magnesium loss bymarine weathering of peridotite, Geochim. Cosmochim. Acta59, 4219–4235, 1995.

w x60 T. Shibata and G. Thompson, Peridotites from the mid-Atlantic ridge at 438N and their petrogenetic relation toabyssal tholeiites, Contrib. Mineral. Petrol. 93, 144–159,1986.

w x61 M. Cannat, D. Bideau and H. Bougault, Serpentinized peri-dotites and gabbros in the Mid-Atlantic Ridge axial valley at15837X N and 16852X N, Earth Planet. Sci. Lett. 109, 87–106,1992.

w x62 H. Bougault, L. Dmitriev, J.-G. Schilling, A. Sobolev, J.L.Joron and H.D. Needham, Mantle heterogeneity from traceelements, MAR triple junction near 148N, Earth Planet. Sci.Lett. 88, 27–36, 1988.

w x63 A.J. Irving, A review of experimental studies of crystal liquidtrace element partitioning, Geochim. Cosmochim. Acta 42,743–770, 1978.

w x64 T.H. Green, S.H. Sie, C.G. Ryan and D.R. Cousens, Protonmicroprobe determined partitioning of Nb, Ta, Zr, Sr, and Ybetween garnet, clinopyroxene and basaltic magma at highpressure and temperature, Chem. Geol. 74, 201–216, 1989.

w x65 G.A. McKay, Crystalrliquid partitioning in basaltic systems:Extreme fractionation of REE in olivine, Geochim. Cos-mochim. Acta 50, 69–79, 1986.

w x66 W.E. Graham and R.L. Nielson, The partitioning of Sc, Y,and the rare earth elements between high-Ca pyroxene andnatural mafic to intermediate lavas at one atmosphere,Geochim. Cosmochim. Acta 56, 2387—2404, 1992.

w x67 H. Fujimaki, M. Tatsumoto and K. Aoki, Partition coeffi-cients of Hf, Zr, and REE betwen phenocrysts and ground-mass, Proc. Lunar Planet. Sci. Conf. 14, part 2, J. Geophys.Res. 89, suppl. B662–B672, 1984.

basaltic liquids and harzburgitic residues in the garrett

Documents