Feb 10, 1993 - clinopyroxene (aegyrine) and riebeckite amphibole. Most of these rocks are fresh, the alteration products being restricted to minor secondary ...
JOURNAL OF GEOPHYSICAL
RESEARCH, VOL. 98, NO. B2, PAGES 1819-1835, FEBRUARY
10, 1993
Mantle Sourcesand Magma-ContinentalCrustInteractionsDuring Early Red Sea-Gulf of Aden Rifting in SouthemYemen:ElementalandSr, Nd, Pb IsotopeEvidence GILLES CHAZOT AND HER VE BERTRAND LaboratoiredesSciencesde la Terre, EcoleNormaleSupdrieurede Lyonand UniversitdLyonL France
Large-scalemagmaticactivity,rangingfrom late Oligoceneto Quatemary,is associated with the Red SeaGulf of Aden titling throughoutthe Arabian passivemargin. The SouthemYemen area representsthe southernmost extremityof this magmaticrange,facing the Afar area, and providesa meansof studyingthe magmaticrecordsof early stagesof rifting (30-16 Ma) in a plume-related context.We investigatemajorand traceelements,andNd, Sr and Pb isotopesof a bimodalseriesof transitionalaffinity consistingof (1) thick elivine-basalt trapsoverlainby ignimbrificrhyolites,(2) basaltic,rhyolific,trachyticandperalkalinedykeswith a prevailingN120-140ø E orientation,and(3) gabbroic,syeniticand graniticplutons.Major and sometrace elementvariationsfrom basaltsto felsicrocksareconsistent withlow-pressure fractionalcrystallization. Mass balancecalculations usingmajorelementssuggestthe fractionationof clinopyroxene (6-9%), elivine (-6%), plagioclase(42-43%),magnetite(-12%), apatite(1-2%) + alkali feldspar(16%). However,LILE (largeion
lithophile element) enrichment andhighinitial87Sr/86Sr ratios (upto0.7074 and0.710inrhyolites and pantellerites, respectively) requirethefelsicrocksto be generated throughsignificantcrustalassimilation. Nd andSr isotopicratiosof the rhyolitictrapscanbe reproduced by bulk mixingbetweenmagmassimilarto the underlyingbasalticunit and the ArabianProterozoicbasement.On the otherhand,an assimilation-fractional crystallization processis requiredto accountfor theisotopicdiversityof the rhyoliticandperalkalinedykes. Rhyolitescanbe derivedfrom a basalticliquidby a moderatefractionationrate(F = 0.47) anda high crustal assimilation rate(R = 0.45), whereasthepantellerites requiremoresignificant fractionation rate(F = 0.07) and a very low assimilationrate (R = 0.05). Elementalandisotopicsignatures of the basaltsdo not supporta
significant crustal contribution intheir formation and204 their isototfic diversity (87Sr/86Sr from0.7034 to0.7051, 143 144 206 -
Nd/ Nd from 0.512676to 0.513045and Pb/ Pb from 17.96to 18.66)mainlyreflectsmantlesource heterogeneifies. Sr-Nd-Pbisotopicdataare consistent with a binarymixingbetweendepletedandenriched sourceregions.The depletedend-membercorresponds to an asthenospheric reservoirapproachingthat producingmid-oceanridge basaltsat Gulf of Aden/RedSea spreadingcenters.The enrichedreservoir, intermediatebetween enrichedmantle I and II end-members,is supposedto be located within old subcontinental lithosphere relatedto the Pan-Africanorogenicevents.UnlikethemodemvolcanicsfromAfar, no HIMU (high U/Pb ratio) signaturehasbeenrecognizedin our sampling.This rulesout any significant chemicalinfluenceof the Afar plume uponearly rift-relatedvolcanismin SouthernYemen and suggestsa continentalrift initiationof passivetype.[towever,onecansuspectthatthe Afar plumemay havesuppliedthe excessheatrequiredto produceso voluminoustrapsandto triggermeltingin the lithosphericmantle,making the distinctionbetweenpassiveandactiverifling moreambiguous.
INTRODUCTION
Seacoast,just in front of the Afar depression. It is thereforea
Crustal extension and transition from continental tooceanic rift veryfavorable areafora better understanding oftheigneous
eventslinked to the initiationof a rift in a R-R-R-typetriple in theRedSea-Gulfof Adenareais accompanied by widespread
magmatic activity throughout theArabian passive margin. Thejunction. In thispaper wepresent thefirstpetrological, search forchemical and isotopic evolution ofmagmas intime andearly geochemical and isotopic data obtained onthe lateOligoceneMiocenemagmatism, sincetheprecursory workof Moseley
spaceis a fruitfulwayto constrain themantleandcrustalmelting [1969]. Our main purposeis to assessthe nature of mantle processes involvedduringthedevelopment of therift. In thelast sources,the evolutionof magmas,andtheirinteractions with the years, numerous studies have been focused on the recent
volcanism, eitheralongtheaxialtrough[Schilling,1969;Altherr
continental crust involved in such a context.
et al., 1988, 1990; Eissen et al., 1989; Barrat et al., 1990;
Schillinget al., 1992], in the Afar areawith specialreferenceto theAfar plume[Schilling,1973;Barberi et al., 1980;Hart et al., 1989;Vidalet al., 1991]or in SaudiArabiaHarratsfAitherret al., 1990;CampandRoobol,1989; Campet al., 1991].On the other hand,geochemical worksrelevantto early volcanismare more fragmentary[Capaldi et al., 1986, 1987a;Petrini et al., 1988; Chiesaet al., 1989;HegnerandPallister,1989]. Oneof themost exciting areas for this purposeis the poorly known southern Yemen(Figure1) whichrepresents thesouthernmost extremityof a magmaticrange stretchingmore than 2000 km alongthe Red
GEOLOGICAL SEWrING
Southern Yemenbasement consists of complexassemblages of metasedimentary and mafic igneousrocksrestingupon older gneisses andintrudedby lategraniticandmaficplutonsfLareare, 1930; Greenwoodand Bleackley,1967]. Althoughpoorly documented, theseformationscan be reasonablyattributedto earlyProterozoic andPanAfricaneventsby comparison withdata acquiredfurthernorthin the SaudiArabianshieldfEngelet al., 1980;Staceyet al., 1980;Duyvermanet al., 1982].This basement is coveredlocallyby MesozoicandearlyCenozoicsedimentary rocks[JungwirthandAs Saruri, 1990].
Copyright1993by the AmericanGeophysical Union. Papernumber9ZIB02314. 0148-0227/93/92JB-02314505.00
In late Oligocene-early Miocenetimes,a first widespread magmaticcycleoccurred. This is mainlyexpressed aslavaflows formingtheYemenplateau(theso-calledTrapSeries),but alsoas dykeswarmsandplutons. Theseformations werestudied mostly 1819
1820
CfIAZOTANDBERTRAND: EARLYREDSEA-GU• OFADENRIFTING
"i::.!-
Radfan RS
•
A
u
o
I
Zabargad •• 25 Km I
I
AFRICA
Trip series
Samplingarea
O, ß 20pkrn
Fig. 1. Schematic mapof theSouth-East extremityof theYementrapSeriesandsamplingareas,from Moseley[1969],modified. Theinsetshowsthedistribution of theYemenTrapSeries,theHarratsof SaudiArabiaandthethickgabbroicdykesalongtheRed Seamargin.
in NorthernYemen,wheretwo phasesof bimodaltrapbuilding
(upto 2000m thick)havebeenproposed, between 30 and26 Ma andbetween23 and 19 Ma, respectively[Capaldi et al., 1986; Chiesaet al., 1989]. However,Menzieset al. [1990] suggestthat
apparentperiodof quiescence, a secondmagmaticcyclehasbeen identified locally and attributed to upper Miocene to early Pliocenetimes(10-5 Ma) in bothNorthernYemen[Capaldiet al., 1987c; Manetti et al., 1991] and SouthernYemen (the so-called
the initiation of the traps could date back to 43.5 Ma. Aden Line [Gassand Mallick, 1968; Cox et al., 1969, 1970, 1977; Contemporaneously, dykeswarms[Capaldiet al.., 1987a,23-20 Mallick et al., 1990]). Plio-Quaternaryvolcaniceventscomplete Ma] andgraniticplumns[Capaldiet al., 1987b,26-20Ma] were the magmatichistoryof this area[Cox et al., 1977; Civettaet al., emplaced. In Southern Yemen,the onlypreviousstudyof this 1980;Huchonet al., 1991]. We investigatedseveral sectorspreviously describedby magmaticcycle is that of Moseley[1969], bearingon the part of the Yemen Trap southernmost endingof the YemenTrap Seriesand associated Moseley[1969], in the southernmost dyke swarms. These events have recently been dated bythe40Ar- Series(Figure1). The Oligo-Miocenemagmatismin thisregionis by thickolivine-basalttrapsoverlainby ignimbritic 39Atmethod between 29 and16 Ma, showing thatmostof the characterized
dykesareyounger thanthetrapsin thisarea[Fdraudet al., 1991; rhyolites,dippinggently 20ø towardsthe SW. In the Dhala area, ¾. Zumboet al., manuscriptin preparation,1992]. Preliminary the exposedthicknesscould approximate3000 m, accordingto Moseley [1969]. In the Alanad and Radfan Mountains areas
geochemical dataare givenby Chazotet al. [1991]. After an
CItAZOT AND BERTRAND:EARLYRED SEA-GUI.• OFADEN RIFIIN6
numerousdykes were emplaced through the basementand
sedimentarycover. They have a prevailingN120-N140ø E orientation but someof them are trendingN-S or N70ø E. They consistof basalts,trachytes,rhyolites and peralkalinerocks (comendites and pantellerites).In the Dhala area,someplutons are formedby gabbrosandsyenitesand intrudebothbasement
1821
clinopyroxene (aegyrine)andriebeckiteamphibole.Most of these rocksare fresh,the alterationproductsbeingrestrictedto minor secondary K-feldspars. Plutonic Rocks
The plutonsampledto the Northof Dhalaconsists of gabbros andsyenites. Gabbrosare locallylayered,andevolveto a finegrainedfaciesneartherim. They containcumulative plagioclase PETROGRAPIIY (An59_62) andintercumulus augite(En35_40, Wo39_42, Fs10_18), Each magmaticgroup (i.e., traps,dyke swarmsor plutons) magnesianbiotite and euhedral apatite. Opaque oxides are
andtraps.
(9% to 19% TiO2). Syenitescontain displays a strong bimodal distribution characterizedby the ilmeniteandtitanomagnetite associationbetweenbasaltsor gabbrosand rhyolites,syenitesor alkali feldspar, plagioclase,small amountsof quartz, green peralkalinerocks.
hornblende and minor biotite which is often associated with
Bas'alts
chlorite,calciteandepidote.
apatiteandzircon.Hydrothermalalterationproducessecondary In bothtrapsanddykes,basaltshavemicroliticto doleriticor ophitic textures. The microlitic basalts contain euhedral to subhedral phenocrysts of plagioclase, clinopyroxene and sometimesolivine, especially in the traps. The groundmassis dominatedby plagioclase,clinopyroxene,opaque oxides and,
GEOCItEMISTRY
AnalyticalMethods
Thirty threerepresentative sampleswere selectedfor major, trace and rare earth element (REE) determinations(Table 1).
occasionally, olivine.The olivinecomposition rangesfrom Fo77_ Analyseswereperformedby inductively coupledplasma-emission 86for phenocrysts to Fo53_86 for olivinein thegroundmass. Many of the phenocrysts have undergone iddingsitisation. Clinopyroxenephenocrystsare diopsideand augite (En35_50, TABLE 1. Summaryof GeologicalFeatures of theAnalyzedSouthern Yemen Samples Wo40_50,Fs5_15)without clear chemical zoning, except for increaseof TiO 2 at therims (from 0.5-1.5%in thecoreup to 2% Sample Site RockType T Orientation in the rims). The groundmass clinopyroxenesare very similar to . YS3 A pantell erite D N 165 the phenocrysts,althoughslightly Fe-enriched.The phenocrysts YS4 A pantelledte D N165
andthemicrolites ofplagioclase arelabrador (An52.70). Someof them display Na-enrichmentfrom core to rim (from An63 to
An52 ). Oreminerals aretitanomagnetites containing 20-25%of TiO2, andsomeiron sulfides.Somebasaltscontainsmallcoarse-
grainedxenoliths consisting of chromiferous augitc(En44_54 ' Wo30_36, Fs9_16 ) andenstatite (En80-85, Fs10-15), surrounded by coronitic
reactions
rims.
Most
of these basalts have suffered
hydrothermal alteration, which is generally more strongly developedwithin the dykes,producingcalcite,chlorite,zeolites, prehniteandclay minerals.
YS5
A
comendite
D
YS7
A
rhyolite
D
N 130/140
NO
YS8
A
pantelledte
D
N125
YS12
M
basalt
F
YS14 YS15
R R
basalt basalt
D D
N100 N75
YS16
R
trachyte
D
N120
YS 17
R
basalt
D
N60
YS 18
R
basak
S
YS19
R
trachyte
D
YS20
R
basalt
F
YS21 YS22
R R
basalt basalt
D D
YS23
R
basalt
S
YS24 YS26 YS27
R R R
basalt basalt basalt
D D D
greatvariabilityin composition, rangingfromoligoclase-andesine YS29 (An26_37) to sanidine(Or39-48)or pure Na or K end-member YS31 (Ab97_100 or Or98).Opaqueoxidesconsistof magnetite,ilmenite YS32 (40% to 65% TiO2) or fitanomagnetite (10% to 27% TiO2). Some YS33
D D
gabbro syenite
P P
D
basalt
F
D
basalt
F
YS34
D
basalt
F
YS37
D
basalt
F
YS38
D
rhyolite
F
YS39
D
basalt
F
YS41 YS42
A A
pantellerite pantellerite
D D
N 120 N120
YS43 YS44
A A
basalt comendite
D D
N120 N 140
YS45
A
basalt
S
YS46
A
RhyolitesandTrachytes Rhyolites and trachytes are found both as dykes and ignimbrificunits overlyingthe trapsin the Dh',daarea.They are slightlyporphyriticor aphyric.In the ignimbrites,phenocrysts are euhedralquartzandscarcesanidine(Or36_47),while in thedykes,
theyareanorthoclase or albite(Ab75_98). Microlites displaya
samplescontain calcite vesiclesand chlorite and clay minerals developedafter ferromagnesian minerals. Comendites and Pantellerites
Theseperalkalinerocks,onlyfoundasdykes,aredefinedusing both chemical (see next section) and petrographic criteria. Referringto the latter criteria,they differ from rhyolitesby their texture,mainly granophyric, andtheirmineralogy:quartzwhichis very abundantbothasphenocrysts andin the groundmass coexists with alkali feldsparswhichrangefrom anorthose(Ab70) to pure K-end-member (Or98_100) in bothphenocrysts and granophyric intergrowths. The peralkalineferromagnesian mineralsare Na-
pantellerite
D
N130 N 120 N95 NO N120 N120
N150
Site:A, Alanad; M, Musaymir; R, RadfanMountains; D, Dhala(Figure 1). ColuInnT: D, dyke;F, flow; S, sill;P, pluton.
1822
CtIAZOT AND BERTRAND: EARLY RED SEA-GULF OF ADEN RIYIZNG
spectroscopy at Centre de Recherches P6trographiqueset G6ochimiques (CRPG), Nancy. Sometraceelements(Y, Sr, Rb, Zr, Nb, Sc, Co, V, Cr, Ba) were alsomeasuredat the Universityof Lyonby X ray fluorescence spectroscopy. All isotope measurementswere performed at Unit6 de Recherches Associ6e(URA) 10, ClermontFerrand.For Pb, Sr and Nd isotopeanalyses, 600 mg of gravelwereleachedwith 2N HC1 and decomposed in HF with a few dropsof HNO3 in a closed teflonbeaker.Pb was elutedwith 6N HC1on an anionexchange columnand loadedon a single Re filamentwith silica gel and H3PO4. Isotopicratiosweremeasured on a CAMECA TSN 206 massspectrometer. Maximum uncertaintyis 0.018, 0.02 and0.07
50.69% to 60.79% SiO2) betweenferrobasaltsand trachytes. Nevertheless, the variationsdefinea consistentevolution:MgO decreases whenTh increases (frombasaltsto pantellerites), while
without blank correction. Sr and REE were eluted with 2.5N HC1
However, the relative scatterof the data set, for a givenTh value,
SiO2 increases throughout the sequence. K20 increases from basalts to felsic rocks and then follows a flat evolution towards
the most differentiatedpantellerites,whereasNa20 displaysa
muchmorescattered distribution. A1203,P205.TiO2 and,to a lesserextent,CaO and FeO, increasefrom basaltsto ferrobasalts,
and then decreasestrongly towards trachytes,rhyolites and pantellerites. Theseobservedmajor elementvariationscould correspondto thoseexpectedfrom fractionalcrystallization processes.The trends are qualitatively consistentwith the on 206pb/204pb,207pb/204pband208pb/204pb ratios, predominance of olivineandclinopyroxene fracfionation at the respectively. Measured ratios are correctedrelative to the beginning of thedifferentiation. Decrease of CaO,P205andTiO2 National Bureauof Standards(NBS) 981 standardfor which the registers crystallization of plagioclase, apatite and followingvalueswere obtained:206pb/204pb = 16.908, titanomagnetite,respectively.The end of the sequenceis 207pb/204pb= 15.44911,208pb/204pb= 36.53936.Total characterizedby the fractionationof alkali feldsparin the proceduralblanks were 0.4 ng. Pb isotoperatios are reported pantellerites (constant or decreasing contents of K20 andNa20). on a cationexchangecolumn.Nd was subsequently purified by inversed-phase chromatography with Di-(2-ethylhexyl) phosphoricacid (HDEHP) on teflonpowder.Sr andNd isotopic ratiosweremeasuredon a MicromassVG54E massspectrometer using singleTa and triple Ta-Re filaments,respectively.Mass
in somediagramscannotbe explainedby sucha process.A part
of thisscatter isprobably caused by alteration (forexample Na20 variations)but otherprocesses may havebeensuperimposed on crystalfractionation.
fractionation ofNdwascorrected bynormalizing to146Nd/144Nd = 0.7219. Values obtained for the NBS 987 and La Jolla standards
Trace Elements
were 87Sr/86Sr= 0.710201and 143Nd/144Nd = 0.511833,
The SouthernYemenrocksexhibita widerangeof trace element contents (Table2). Selected elements areplotted against correction relative to these values. Maximum errors are 0.002% Th in Figure 3. The behavior of the compatible elements confrrms 87 on Sr/86Sr ratiosand0.003%on---•,•a/ratios. x•_ age of olivineand clinopyroxene (early correction wasmade forPbandNdisotopic ratios. The87Sr/86Srthe early fractionation decrease of Ni, Cr, Co)followed byplagioclase (decrease of St). ratioswerecorrected usingthe39Ar-40Aragewhenknown Some basalts appear tobeslightly cumulative (Cr> 700ppm,Ni [Fdraudet al., 1991] or assumingan ageof 20 Ma. The •qd were > 300 ppm). When plotted against Th, the incompatible elements calculatedaccordingto DePaolo and Wasserburg[1976], using describe an overall positive trend from basalts to pantellerites but 143Nd/144Nd = 0.512638 forpresent-day chondrite uniform in most diagrams, the points fail to follow a single straight line reservoir(CHUR). respectively.Nd and Sr isotope ratios are reported without
passingthroughthe origin. This confirmsobservations made
Major Elements
previously frommajorelements andrequiresthe petrogenetic study to take parameterssuch as crustalcontamination or
Major elementcontentsarereportedin Table 2. All thebasalts, from both traps and dykes, are olivine and hypersthene normafive,exceptfor four of themwhichare slightlynepheline normafive.Felsicrocksaresubalkaline rhyolitesandtrachytes and
sourceheterogeneityinto account,in additionto crystal fractionation. The rareearthelementpatterns of representative rocksnormalized to chondrites areshownin figures4 and5. All
into comendites andpantellerites, according to MacDonaldand Bailey's[1973] andMacDonald's[1974] classification basedon
whole,thebasaltsfrom thedykeshavemoreLREE-enriched and variablepatternsthan thosefrom the traps.The felsic rock (rhyolites topantellerites) patterns areparalleltoeachotherandto thoseof somebasaltsfrom the dykes(Figure4). Moreover,a negativeEu anomalyappearsin the peralkalinelavas. This testifiesthe importanceof crystalfractionationin the evolution frommaficto felsicmagmas.The spiderdiagrams of thesame samples arereportedin Figure6. Sr andTi negativeanomalies observedin the felsiclavasfrom basaltsto rhyolitesandfrom rhyolites to pantelleritesare related to plagioclaseand titanomagnetite fractionation,while the strongnegativeBa anomalydisplayedby the peralkalinerocksis relatedto alkali feldspar fractionation attheendof thesequence. A crucialfeature of thesepatterns is thenegative Nb anomaly whichappears in the rhyolificpatterns, andwhichis vanishing in theperalkaline rocks. If controlledby fractionalcrystallization, thisbehaviorwould requirethefractionation of a mineralthatselectively removes Nb
thebasalts arelight-REE-enriched but theydisplaya variable La/Ybratiorangingfrom6 to 20 (Figure5). On the peralkaline rocks(high(Na20 + K20) / A1203ratio)subdivided enrichment, normafivequartzandfemicmineralcontents.
The amountof H20 variesgreatlyin thebasalts, rangingfrom 0.62 % to 10.52 %, accordingto the secondary phasesobserved. The rhyolites,trachytesandsyeniteare somewhat altered(LOI
(lossof ignition)rangingfrom2.17 % to 5.34 %), whereasthe peralkaline rocksaremostlyunaltered, LOI neverexceeding 1.64 %.
The AFM (Alkali-FeO-MgO)diagramshowsa Fe-enriched trend, from basalts to ferrobasalts. This trend, as well as the
normafivecomposition, alkali andTiO2 (2 %) contentsof the basaltsand the peralkalinity of the felsic rocks have been documented in manytransitionalmagmaticseries[Bass,1972' Treuil andVaret, 1973;Barberi et al., 1975].They havebeenalso
emphasized in thevolcanicseriesof NorthYemen[Capaldiet al., 1987c; Chiesa et al., 1989].
Major elementvariationsrelativeto Th, whichis usedas an
[Dupuy andDostal,1984]whichisunlikelyandnotsupported by
indexof differentiation,are shownin figure 2. The bimodalityof
petrographic observations. On the otherhand,a negativeNb
our samplingis underlinedby the gap of silicacontent(from anomaly mayresultfroma contamination process involving the
.
.
II o
16
18
14
16
-eel
14
12
12 10
10
8
8 /",
6
6
A /,,,
4
4-
[]
I
I
I
10
15
2o
2
Th
J
•
A
25
25
P205"
0.8
A
TI02 ß ß
0.6
ß
ß
ß
ß
0,4 []
0.2
A
I
I
I
lO
16
20
0.0 o
18 16
o
A
Th
•• 25
Th
20
AI205
16
_....
14 lo
&, 12
6
10
4
..
6
5
2
•
I
I
I
10
15
20
A
o
25
A
25
o
K20
SI02
A
A
.
4
-
3
-
2
-
0
0
I
I
I
I
6
10
15
20
Th I
I
Th
25
25
No20
[] []
ß
_
ß
[] ß
Fig. 2
[]
A
.-•% ß
0
!
I
I
I
5
10
15
20
Th
25
Fig.2.Majorelement oxide (wt%)variation diagrams withTh(ppm) used asindex ofdifferentiation. Solidcircles: basalts; squares: rhyolites andtrachytes; trianglescomendites andpantellerites.
1826
CttAZOT AND BERTRAND:F.ARLY RED SEA-GUI• OF ADEN RIFTING
1400
1200 $r
Or
ß
1200
go0
ß
1000 800
600 eß•ß
--o
600 ß
400
300
g
200
0
ß n•_,._ I
•
1200
t /x 10
o
4, ./,,. __..•__.•• Th 15
2o
0
25
t 0
5
• z• 10
z• 15
20
25
Rb
Zr
go0 []
z• zx 600
300
i
i
i
!
5
10
15
20
Th 25
0
5
i
!
i
10
15
20
25
20
25
35
160
120 21 z•
8O
[]
14 4O
7
0 5
10
15
20
25
I 0
5
10
15
Fig.3. Variation of selected compatible andincompatible traceelements (ppm)relative toTh.Samesymbols asin Figure2.
uppercontinental crust[DupuyandDostal,1984].Whateverare the processes implicatedin the genesisof the rhyolites,the disappearance of the Nb anomaly betweenrhyolites and pantellerites excludes thepossibility for thesecond toderivefrom theformersolelyby a fractionalcrystallization process, despite theirparallelREE patterns.
suspectedto have been somewhatshifted away from the mantle arraytowardsa higherSr ratio by secondaryalterationprocesses.
However, the87Sr/86Sr ratios of themostaltered basalts (YS17, YS24,YS43andYS26)arewellcorrelated to the143Nd/144Nd ratios (assumedto be insensitiveto alteration), so that these samplesremain near the mantle array. So, althoughwe do not
deny thedisturbin,g, roleofalteration, weargue thatithasaweak
influence on the 87Sr/86Sr signature of the Yemen rocks. The felsicrocksaremorescattered. The syeniteYS31 hasthesameNd assomeof the basalts,therhyolites Twentyrockscoveringthecompositional rangefrombasaltsto andSr isotopiccompositions rhyolitesandpantellerites havebeenanalyzedfor Sr, Nd andPb have higher Sr ratios and the lowest Nd values, and the isotopes(Table3). Nd andSr ratiosshowimportantvariations: peralkalinerocks show Nd values similar to thoseof the least Isotopes
initial87Sr/86Sr ratios range from0.7034to0.7051in thebasalts radiogenicbasaltsbut they are shifted towards the highest and from 0.70406 to 0.7101 in the felsic volcanics. •.Nd values 87Sr/86Sr ratios. Thislargespread is strongly suggestive of rangefrom+0.7 to +7.9 in thebasaltsandfrom-2•1to +3.8 in the variable crustalinfluencesin the genesisof the felsic rocks.The spana relatively limitedrangeof 206pb/204pb ratios, felsicvolcanics.In Figure7, mostof the basaltsandthe gabbro samples and arestretching alongthemantlearraybetweenthemid-oceanridge from17.92to 18.83(Figure8). Their207pb/204pb basalts(MORB) field and the enrichedmantlesourcesEMI and
208pb/204pb ratios range from15.50to15.72andfrom38.16to
39.24, respectively.They define an elongateddomainslightly 87Sr/86Sr ratiowithinthebasalts, because someof theinhave abovethe MORB field and the NorthernHemisphereReference sufferedextensivealteration.For example,the rock YS15 is Line (NHRL) as defined by Hart [1984]. This domain is
EMII. Specialattentionmust be paid to the variationsof the
CHAZOT AND BERTRAND: EARLY RED SEA-GULF OF ADEN RL•'T•O
•oooRock/Chondrite
1827
•)00 Rock/Chondr•te
-''-__
-- Basalts
•"'•f•',-
''h
I '-
Basalts .
-_
,•'
La
•
Pr
•
Sm
Eu
•
Tb
Dy
Er
•
Ba
•
Th
•
La
•
Sr
•
Sm
•
•
T•O2
Y
.
Dy
Er
Fig. 4. Chondrite-normalized REE patternsfor selectedbasalts,rhyolites- Fig. 6. Spiderdiagrams of representative rocksfrom thedykes.Chondrite trachytesandperalkalinerocksfrom the dykesfrom SouthernYetnen. normalizingvaluesfrom Thompsonet al. [1982]. The arrowsindicatethe Normaliz2ng valuesfromThompson et al. [1982]. samples usedfor thepetrol•enetic modelling.
microprobe(Table4). The resultsare shownin Table 5. A good
characteristic ofa weak D,u•al signature. Thefelsic rocks cover fit is obtainedbetweencalculatedand observedmajor element nearly thesame range of206pb/204pb asthebasalts, but(exceptcompositions with fractionationof clinopyroxene (6-9%), olivine the syenite YS31) they tend to have significantly higher
(-6%), plagioclase(42-43%), magnetite(~12%) and apatite(12%) (plus16% of alkali feldsparwhenpantelleriteis concerned). From these results (F values = 0.11 and 0.32), the Rayleigh PETROGENESIS equationwas appliedto testfurtherthe model with selectedtrace elements,usingdistributioncoefficientsfrom the literature(Table The strong bimodality displayed by the Southern Yemen 6). We obtaineda good fit betweenobservedand calculated Tertiary igneousrocks, and the chemicaland isotopicdiversity compositionsfor the pantellerite YS3 (Figure 9) but the prevailingbetweenandwithin marie andfelsicgroupsoutlinethe calculation failedtoreproduce therhyoliteYS7 pattern(Figure9). complexity of their genesis at mantle levels (source This modellingindicatesthat crystalfractionationof basaltic heterogeneities, melting rates) as well as at crustallevels (lowmelts can producethe major and trace element compositions pressurefractionation,continentalcrustinteractions). observedin the pantelleriticmagmas,whereasit encounters more difficulties in the case of rhyolitic magmas.One can therefore FractionalCrystallization suspectthesesilicic magmasto be generatedthroughdifferent Major and trace elementvariations(e.g., Figures2 and 4) processes. This assumptionis reinforcedby the large spreadof suggestlow-pressurefractionationto be partly responsiblefor isotopiccompositions(Figure 7) which suggeststhat crustal somemafic-felsic trends,even thoughit is obviousfrom an contaminationplayed a significantrole, in additionto crystal isotopicpoint of view (Figure 7) that this processcannotact fractionation. alone.To testquantitativelythe validity of this assumption, we usedthe leastsquaresmethodof Wrightand Doherty[1970] on Crustal Contaminationand Generationof major elements.Among the bestparent-daughter candidates(on SilicicMagmas the basisof their parallelREE patterns,Figure6), we computed The precedingobservations indicatethatcrustalassimilation is the fractionationfrom the basaltYS24 towardsthe rhyolite YS7 a likely process during the genesis of the felsic rocks. However, andthepantelleriteYS3, usingmineralcompositions measured by thereis a major difficulty in assessing the interactionsbetween mantle-derived magrnas andcontinental crustin SouthernYemen, because of the poor knowledge of the nature and isotopic lOO
207pb/204pb and208pb/204pb foragiven 206pb/204pb.
TABLE4. Compositions of theMineralsUsedin theLeastSquares Calculations(SeeTable 5)
K20
lO
CaO
1
,,
La
'
Ce
i
Pr
i
Nd
i
)
Sm
Eu
C.d Tb
Oy
Er
Yb
Fig. 5. Chondrite-normalized REE patternsfor the basalticdykes compared withthetraps(shaded area).Samenormalizing valuesasin Figure4.
Cpx
Oliv
Plag
0.02
0.02
0.23
20.41
TiO2
1.03
FeO
6.69
0.32
0.03 14.83
12.55
K-Feld Apa 6.41 0.71
0.01 53.36
Mag 0.02 0.25
0.10
0.43
0.04
24.70
0.60
0.67
0.72
70.32
0.00
Na20
0.32
0.05
4.16
6.36
0.00
SiO2
50.48
40.18
52.60
65.35
0.40
1.30
3.32
0.05
29.35
19.89
0.01
2.05
MgO
15.29
43.97
0.15
0.02
0.23
0.67
P205
0.00
0.00
0.00
0.00
45.00
0.00
A1203
1828
CHAZOT^ND BERTRAND:EARLYRED SEA-GULFOFADENRIFTING
noneof theNd-Sr isotopiccompositions of thefelsicdykescanbe
reproduced by mixingcurvesbetweencontemporaneous basaltic dykes and the crustalcomponents. This suggeststhat the ignimbritictrapscouldderivefromcrustalmeltinginducedby abundant injectionof basicmagmas followedby magmamixing. 4YS24.\ Magmasupplymayhavebeenmorelimitedduringdykeinjection, precluding significant crustalmelting. YS3 2 Consequently,we have tested an assimilation-fractional crystallization model(AFC model)[DePaolo,1981],onthesame 0 • Arabian shield: parent-daughter compositions asakeadydiscussed in theprevious section(i.e., basaltYS24, rhyoliteYS7, pantelleriteYS3), using the granodioriteZ103 as contaminant. The resultsof this •1 •r,--•rI modelling, applied to Nd-Sr isotopes, are plotted in Figure10 and 0.702 0.704 0.706 0.708 0.710 0.712 pointout thepossibleinvolvement of thisAFC process andthe Fig.7. endversus S?Sr/g6Sri diagram fortheSouthern Yemen rocks. contrasting history of rhyolitic and peralkaline magmas. The Samesymbolsas in Figure2. Data sources: EMI and EMII manfie liquidYS24through a componc-nts andMORBfieldfromZindlerand11art[1986];manfiearray rhyoliteYS7 canderivefromthebasaltic from Faure[1986];ArabianShieldfrom Staceyand Hedge[1984] and moderate crystalfractionationrate(fractionof liquidremaining HegnerandPallister[1989]. F = 0.47)anda highcrustalcontamination rate(R = assimilation rate/crystallization rate = 0.45), whereasthe genesisof the
cr•YS7 II JYS38 •- grou•roup IIZ103
-2
-4
iEM'• -•
EMIl87 Sr '86Sr I'œNd--'45'4 .
compositionof the basement.Nevertheless, referringto the likely southwardsextensionof the Pan-African and reworked early Proterozoic
structures identified
in the Saudi Arabian
shield
pantellerite YS3 requiresan important fractionation rate (F = 0.07) anda very low contamination rate(R = 0.05).The different values of contaminationrates precludeperalkalinemagmasto
derivefromrhyoliticliquids.Onecannotethecontrasting isotopic
[Staceyet al., 1980;Duyvermanet al., 1982;Staceyand Stoeser, evolutionof Sr and Nd ratios from basaltto pantellerite:as the 1983; Stacey and Hedge, 1984; Hegner and Pallister, 1989; 87Sr/86Sr ratiostrongly increases from0.70396 to 0.71014, the Kr6neret al., 1989],onecanreasonably considertheseformations 143Nd/144Nd ratioonlyslightlydecreases from0.512834 to as potential contaminantsfor the Southern Yemen Tertiary 0.512789. This is related to the strongdecreaseof Sr content magmas.Accordingto the abovereferences,thesebasement duringfractionation processes, causedby extensive removalof formationswereschematically dividedintotwo groups(Figures7 o,,,• •
r,......
I ....
;o,o ,,r
;.....
;•'• oceanic
....
' accreted
during the Pan-African events and is presumablydevoid of significantamountsof oldercontinental components [Staceyand Stoeser,1983;HegnerandPallister,1989].Thisgroupdominates in the western arc œerranesbut has also been identified in the
easternpart of the shield[Duyvermanet al., 1982]. GroupII, mainly recordedin the easternterranes,is a compositecrust involving ancient (early Proterozoicto Archean) continental material,possiblyreworkedduringPan-Africantimes[Delevaux et al., 1967; Staceyet al., 1980; Staceyand Hedge, 1984]. The
rocksfrom this groupexhibithigher207pb/204pb and 208pb/204pb ratios andmuchlessradiogenic Ndratios thanthose of groupI. The very wide isotopicrange displayedby the two groups makes it difficult to use their composition to constrain unequivocallythe crustal influence.Nevertheless,the Pb-Pb constraints(Figure 8) precludethe derivationof felsic from basalticmagmasthroughcontamination by ajuvenilePan-African crust, because, on the whole, the felsic magmas are more
radiogenic in207pb and208pb thanthebasaltic ones, which are themselves at leastas radiogenicas the Pan-Africancrust.The compositionsof the felsic rocks are more consistentwith a contaminationby basementrocks involving an old continental
158 2O-,pb/204 Pb Arabian shield
•57 -
/ (•ou•
/
cr•
-•
Z103•N
Arabianshield
/ (groupl)._
15.6 / YS39,• . 15.4
I
,'
..-'
......•
17.5
4o 39.5
18
206pb204 b
•
•
18.5
19
/ P
19.5
I
o8Pb/ 204Pb Arabian
39
(group II) YS38
38.5
component (much more radiogenic in207pb and208pb foragiven 206pb/204pb ratio). Among them, weused thegranodiorite (Z103) 38
shield
[]a/
•
,,,a ©x,j..... .x9,•/./ ." ,.' • .,'
•e.•.-"..,,..•.,.-'"'
YS39.•
....''''
I .•_•.Z103 •.'"'./""-••
%%'"'Arab,an shield
studiedin detail by Staceyand Hedge [1984] in an attemptto quantifythe participationof sucha crustin the genesisof the felsic rocks.The Nd-Sr isotopiccompositionof the ignimbritic .--' Pb / Pb upper trap (YS38) can be satisfactorilyfitted by bulk mixing 37 17 17.5 18 18.5 19 19.5 [Volliner,1976] betweentheunderlyingbasalticunit (YS39) and the granodioriticbasement,in the ratio of 4 to 1, respectively. However, the required contaminant must have higher Fig. 8. Pb isotopic compositionsof the SouthemYemen samples compared withdatafromtheArabianShield[Stacey et al., 1980;Stacey ..
37.5 / •
•
.-"•--..MORB 206204
206pb/204pb, 207pb/204pb and208pb/204pb ratiosthanZ103
andStoeser,1983;Staceyand11edge,1984;HegnerandPallister,1989].
sample,i.e., somewherehalf way betweenthe two fields of old componentbearingcrustreportedin Figure8. On the otherhand,
MORB field from Wilson [1989]. NItRL from tlart [1984]. Same symbolsasin Figure2.
CIg•ZOTANDBERTP•ND:EARLY]•EDSEA-GULF OFADEN]•O
dd
d
o
o
o
o
o
d
o
o
o
o
d
o
o
o
1829
o
o
o
1830
CIIAZOT AND BERTRAND:EARLY RED SEA-GULF OF ADEN RIYI'ING
TABLE 5. Major ElementLeastSquares Calculations RelatingYS24 to YS3 andYS24 to YS7 YS24=0.11YS3+0.09Cpx+0.06Oliv+0.43Plag+0.16 K-Feld+0.12Mag+0.01Apa Y$24 Observed.%
Y$24 Calcu][atcd.%
K20
1.57
1.66
CaO
8.36
8.29
TiO2
3.04
3.20
FeO
11.56
11.50
Na20 SiO2 A1203 MgO P205
4.07 49.12 17.25
3.41 49.16 17.30
4.21
4.20
0.60
0.69
YS24=0.32YS7+0.06Cpx+0.06O1iv+0.42Pl;g+0.12Mag+0.02Apa K20 1.57 1.21 CaO
8.36
8.24
TiO2
3.04
3.12
FeO
11.56
11.53
Na20 SiO2 A1203
4.07 49.12 17.25
3.43 49.19
MgO
4.21
4.16
P205
0.60
0.75
17.27
andER2 = 0.6fortheSecond. F_,R 2 = 0.49fortheFirstCalculation
TABLE6. Distribution Coefficients UsedintheTraceElement Modelling Relating (Top)YS24toYS3and (Bottom)YS24 to YS7
Oliv.
Cpx
Plag.
Oxides
Apa.
K-Feld
Ba
0.03
0.04
0.56
0.40
0.08
3.60
Rb
0.04
0.04
0.13
0.47
0.04
0.30
Th
0.03
0.06
0.05
0.90
0.01
0.09
Nb
0.01
0.80
0.06
2.00
0.10
0.01
La
0.03
0.10
0.20
0.39
6.00
0.13
Ce
0.02
0.12
0.14
0.34
8.00
0.09
Sr
0.02
0.16
2.50
0.68
0.06
7.00
Sm
0.009
0.25
0.07
0.30
30.00
0.07
Zr
0.07
0.50
0.13
0.40
0.50
0.27
Eu
0.008
0.30
0.32
0.30
27.00
1.13
Ti
0.025
0.35
0.04
13.25
0.10
0.05
Y
0.06
0.50
0.20
0.40
8.00
Yb
0.05
0.30
0.03
0.40
21.00
Oliv.
Cpx
Plag.
Oxides
Apa.
Ba
0.005
0.00
0.33
0.40
0.08
Rb
0.04
0.04
0.13
0.47
0.04
Th
0.03
0.06
0.05
0.90
0.01
Nb
0.01
0.30
0.01
1.00
0.10 6.00
La
0.03
0.08
0.10
0.39
Ce
0.02
0.10
0.09
0.34
8.00
Sr
0.02
0.16
2.70
0.68
0.06
Sm
0.009
0.25
0.07
0.30
30.00
Zr
0.07
0.50
0.13
3.60
0.50
Eu
0.008
0.30
0.32
0.30
27.00
Ti
0.02
0.30
0.04
12.00
0.10
Y
0.06
0.50
0.20
0.40
8.00
Yb
0.05
0.30
0.03
0.40
21.00
0.10 0.012
CHAZOTANDBERTRAND:EARLYRED SEA-OUI• OFADENRIFTING
1831
ROCK/CHONDRITE 1000
calculated pattern YS3 .f
:.-.'t "
.Sr/
,_-.-:-:• •"-" '-''"::='--=-•' •'- •Z• ....
.-' t-.-"• /
/ v
'""
30
d
-
.........•..•..•$•calculated pattern YS7 .
.•...• ..,..-...•.
-
I- - - 7-'",-•
.,•..:-• ;:. ,t ... •.......••-"... o--,'/•'
I,
.. ,, .... ' ':'"'• ..,,•
\', '-J...' //;
:'.
,--__ t,..... '-.
.;;
,,•-'•z---,..! .... ';•. ",4.'•-..
--',
///: .•,o,,,•YS,•.-•x ,-,.'.-11 •
Pantellerite YS3
'. •::... ...,• f'x... -.. '
/:.
15
%-...-::./
10
•,,..' []
5
',/
oo o Si02
Be
Fb
Th
Nb
La
Ce
S•
Sm
Zr
Eu
Ti
Y
Yb
45
Fig.9. Observed andcalculated chondrite-normalized patterns relatbgthe basaltYS24 to therhyoliteYS7 andto thepantellerite YS3. Distribution coefficients used for the calculations are shown in Table 6. Chondrite
50
55
60
65
70
75
80
Fig. 11. Sr/Nd ratio variationsversusSiO2 content(wt %) for the Southern Yemenrocks.Samesymbolsasin Figure2.
normalizing valuesfromThompson et al. [ 1982].
plagioclase(seeFigures6 and 9). Meanwhile,Nd, which hasan incompatiblebehavior,is concentrated in the residual liquid, so that the Sr/Nd ratio evolves from 33.9 in the basalts (26.6 for
crust[LeemahandHawkesworth, 1986;Carlsonandttart, 1987], all the analyzedbasaltsare within the field of commonmantle
meltsdefined bf Hartetal. [1989],andthereisnocorrelation between 87Sr/8øSr andK/Pratios. Thisstrongly indicates thatthe 87Sr/86Sr ratiovariations result fromsource heterogeneity rather
YS24usedin themodel)to 0.15 in theperalkaline rocks(0.21for whichis in goodagreement with YS3) (Figure11).Consequendy, evenif thecontamination rateis thanfromcrustalassimilation, the alignment of these basalts along the mantle array in the verylow, theAFC processwill inducemuchhighervariationson 143Nd/144Nd versus 87Sr/86Sr plot (Figure 7). Unlike the basalts, the Sr thanon the Nd isotopicratios,providedthat significant theperalkalinerocksclearlyexperienced crustalcontamination, as mounts of plagioclaseareremoved.
theydisplay a strong concurrent increase in both87Sr/86Sr and
K/Pratios.Ontheotherhand,considering thisdiagram, onecould
MantleSourcesof theBasalts:Implications for Rifling Processes The previouspart of this studyhasdocumented the influence of crustal contamination
in the formation of the differentiated
rocks.It is thereforeimportantto assess whetheror not thebasalts alsohaveundergone crustalcontamination beforespeculating on theirmantlesources. For thispurposewe havecombinedevidence fromtraceelementratioswiththatfromradiogenic isotoperatios.
In the87Sr/86Sr versus K/Pdiagram (Figure12a),whichis expectedto be a goodindicatorof the influenceof the continental
0.5130
143Nd/144Nd
thesebasaltsto be inheritedfrom heterogeneities withintheir sourceregions.
basalt YS24 0 0
pAa•&J;ellerite YS3 In thefollowingdiscussion,we referto thefour"end-member" 15)
•3:•_000 0
0.5128
0000 0
0.5126
o
0
mantlecomponentssummarizedby Zindler and Hart [1986] andHart [1988](DMM = depletedMORB mantle,EMI andEMII = enrichedmantleI and1I, HIMU = highU/Pb ratio),andto a fifthmorehypothetical one,PREMA = prevalent mantle[Zindler
A
0
ø•o o
rhyolite YS7 /
0.5124
wonderwhetherthe rhyolitessufferedcrustalcontamination or representdifferentiatesof more radiogenicbasaltsthan those studiedpresently.In fact, the previousdiscussions indicatethat bothprocesses probablyoccurred.Similarly,the use of trace elementssuchas Th versusNb (normalizedto Zr to reducethe effectsof fractionalcrystallization, Figure12b) showsthatthe basaltsstretchalongthe mantlearray from depletedtowards enrichedend-members, withoutevidenceof significantcrustal influence.Thereforealthoughcrustalcontamination processes cannotbe completelydiscarded,our data do not supportthe hypothesisthat they might have significanteffectson basalt chemistry.Consequently, we considerthe isotopicvariationsof
oooo
o o
andHart, 1986].Theisotopic compositions of thesecomponents andof theYemenbasaltsareplottedin Figures13, 14 and15,
0.5122
whichcombine87Sr/86Sr ' 143Nd/144Nd ' 207pb/204pb with
0.5120
continental
crust
206pb/204pb ratios. Inthe143Nd/144Nd versus 87Sr/86Sr (Figure
0.711
nearbythe N-MORBs of the Gulf of Aden and Red Sea[Betton and Civetta, 1984; Eissen et al., 1989; Altherr etak, 1990;
13),Southern Yemenbasaltsspreadalongthemantlearrayfrom
0.5118 0.703
0.705
0.707
0.709
0.713
87Sr/86Sr
Schillinget al., 1992]towardstheenriched mantlecomponents EMII and EMI. This trend is consistent with a pseudo-binary Fig. 10. AFC modellingfromthebasaltYS24to therhyoliteYS7 andthe depletedandenrichedsourceregionsandgives pantelleriteYS3. The asteriskis the Proterozoic granodiorite Z103 from mixingbetween SaudiArabia[StaceyandHedge,1984]usedascrustalcontaminant. '11ae evidence that the sources tappedduringthe earlyriftingwere circlesareresults of calculations obtained fordecreasing F (withstepof differentfrom the present-day MORB asthenospheric sources 0.05)FortherhyoliteYS7, partitioncoefficients are1.5and0.4 for Sr and beneath Red Sea and Gulf of Aden troughs. Nd, respectively, andR = 0.45 (seetext). For the pantelleriteYS3, The inferredsources canbe moreaccurately investigated by pan/t/oncoefficients are2.5and0.11forSrandNd,respectively, andR = 0.05. Other rhyolites(squares)and comendites and pantellerites
examining the87Sr/86Sr versus 206pb/204pb diagram (Figure 14)
(triangles)are alsoshown.
which better discriminatesthe four principal end-member.
1832
CHAZOTAND BERTRAND:EARLYRED SEA-GULFOFADEN •G
0.712
3.5136
143Nd/144Nd
8 7Sr/ 8 6 Sr 3.5134 071
....:..,:::.
anfie melts fi•.ld
0.708
MORB Adenand Red Sea
3.5132
.PREMA
ß• ..r•_•u:i Arabia
0.5130
0.706
0.5128
,.._ 0.704
KIP 0.702
10
0
20
30
40
'
'
'
50
60
70
(-
-•- ......
-....... ::....:
Ethiopi/ ..........':-:":' ......................
0.5126
0.5124 80
0.5122
0.5120
lO
Th*100/Zr
0.702
' 87Sr/86SI 0.703
0.704
0.705
0.706
0.707
0.708
0.709
Fig.13.The143Nd/144Nd versus 87Sr/86Sri diagram fortheSouthern Yemenbasalts.Forcomparison thefieldsof contemporaneous basaltic rocks(>15 Ma) from Ethiopia[ttart et al., 1989],Afar [Vidal et al., 1991],A1 Lith area,SaudiArabia[tlegnerand Pallister, 1989],Saudi Arabia(H. BertrandandG. Chazot,unpublished data,1991)andrecent basaltsfrom Red Sea and Gulf of Adentroughs[Eissenet al., 1989; Schillinget al., 1992]are shown.Mantlecomponents arefromZindler and !tart [1986].
peninsula.These componentshave probablybeenincorporated during the Pan-Africanorogeniceventsor even earlier [Betton and Civetta, 1984; Hegner and Pallister, 1989; Altherr et al., 1990; Henjes-Kunstet al., 1990; Schilling et al., 1992]. The resultingisotopicheterogeneity of the Pan-Africanlithosphere is thereforenaturaland may explain the relative scatteringof the data set towardsenrichedcomponents.The involvementof these two contrastingreservoirs,depletedasthenosphere and enriched lithosphere, is alsosupported by Pb-Pbrelationships (Figure15),
ß
ß
which tend to favour the involvement of the EMIl rather than EMI
component in thelithosphericsource. Concerning thedepletedreservoir,an alternativeinterpretation arisesif we considerthe PREMA compositionasa distinctmantle componentrather than a mixture of the other end-member
N b' 1 00/Zr
O.1
.....
1
•
10
1OO
components, as discussed by Zindlerand Ilart [1986].It canbe notedthat in all isotopicdiagrams,the Yemenbasaltdataset is preciselyrootedon the PREMA field from which it seemsto diverge towardsenrichedend-members,none of the samples fallingbetweenPREMA andDMM poles(Figures13, 14, 15). This could argue for the implication of a PREMA-type asthenospheric reservoir(depletedoceanisland basalts(OIB)
Fig.12.(a)The87Sr/S6Sri versus K/PratiofortheSouthern Yemen rocks.Manfiemeltsfield fromHart et al. [1989].(b) Th*100/Zrversus Nb*100/Zrdiagramfor theSouthern Yemenbasalts. MantleArrayfrom Beccaluva etal. [1984].N.MORBandE.MORBarerepresentative values
fromSunandMcDonough [1989].b.C.areuppercrustvalues fromTaylor andMcLennan [1985]andWeaverandTarhey[ 1984].Samesymbols as
0.707
0.706
in Figure2.
87Sr/86 Sr '-:.;;(,,.,,, • / '"--• .,.'""'.e, i e'5. •;.......... AfarArabia I '", .Saudi
0.705
components. Thedatasetisrooted justabove theleftboundary of theMORB field andstretches towardsenriched components intermediatebetween EMI and EMII. The lower end-memberof
0.704
this trendcan be attributedto the depletedasthenospheric
PREMA
related to recycling of continental crustmaterial and/or0.703
MORB) reservoir. The upper one (hybrid EMI-EMII) is usually
metasomatism
within
the mantle
and is inferred
RB Aden and Red Sea
to be located
withinan old subcontinental lithosphere [Zindlerandttart, 1986;
Hart,1988]. Thisstatement issupported byseveral studies which
0.702
16
"''
17
18
I
I
19
20
2?6 Pb/204pb 21
22
have documentedthe existenceof enrichedcomponents(EMI
Fig.14.The87Sr/86Sri versus 206pb/204pb diagram forSouthern Yemen
and/or EMIl) in the lithosphericmantle underneaththe Arabian
basalts.Data sourcesas in Figure 13.
CHAZOTANDBERTRAND:EARLYRED SEA-GULFOFADEN RIFHNG
1833
evolution.ConcerningSouthernYemen,no HIMU componenthas
159
inoursampling inwhich thehighest 206pb/204pb •..........beenidentified
2ø7pb/aø4Pb
and 208pb/204pbratios do not exceed18.66 and 38.84
15.8
respectively, andnoneof the trendspointtowardstheHIMU endmember (Figures 13, 14, 15). Another argument can be put forward againstthe participationof the HIMU component:the
basalts ofSouthern Yemen display aslow206pb/204pb values as .•..'-_•.-'"'-m
'
•MORB
Aden and Red Sea
15.5
.;....•• . ..?PREMA
thoseobservedin SaudiArabia or in the recentAdenridge,eastof 50øE,that is beyondthe presumedinfluenceof the plume [White and McKenzie, 1989; Schilling et al., 1992]. These resulsare consistent with the conclusions of Huchon et al. [1991]
establishingthat the Afar plume was not the primary causefor rifting. 2o6pb/2O4pb Arguing from theseobservations,our resultssupporta rift 15.3 ' initiationmodelof "passivetype"(meaningthat it shouldnot be 16 17 18 19 20 21 directly triggeredby the mantle plume upwelling), which was Fig.15.The207pb/204pb versus 206pb/204pb diagram forSouthemprobablyguidedby preexistingstructuresin responseto tensional Yemenbasalts.Datasources asin Figure13. forces,such as advocatedby Berhe [1986], Dixon et al. [1987, 1989], Bohannon et al. [1989], and McGuire and Bohannon [1989]. Following Almonds [1986a, b] ideas, we propose a reservoir?),differentfrom the MORB reservoir,suchasproposed decoupling between the incipient rifting involving both by Hart et al. [ 1989]in theEthiopian rift andWestCentralAfar. lithospheric and asthenosphericmelting (our data) and a However, the existence of a distinct PREMA end-member subsequent stagecharacterized by themeltingof the HIMU-type componentremainshypothetical.It is not requiredby the plumeof Afar [Barrat et al., 1990;Vidal et al., 1991;Schillinget
15.4
worldwide compilation of tlart [1988],andSchillinget al. [1992] al., 1992]. rejectthisproposal arguingthattheisotopictrendsobtained in the Alternatively, the combined melting of N-MORB type overallAfro-Arabianregiondonotconverge on a singleuniversal asthenosphere and continentallithosphereis consistentwith the but on severaldifferent PREMA-type sources.So, the evidence model of a westward propagatingoceanic rift through Afroof a PREMA reservoirunderneathYemenis still an openquestion Arabiancontinentasproposedby Courtillot[1980] andCourtillot andneedsfurtherinvestigation. et al. [1980]. In that case the rifting processshouldbe active Forcomparison, we showed in Figures13, 14and15 thefields insofaras it would be inducedby the penetrationof the Indian of earlyrift-relatedbasaltsfromneighboring localitieswhichare ridge andits relatedthermalanomaly,evenif the directinfluence contemporaneous with thosequotedin thispaper,i.e., morethan of theAfar plumeat thattimeremainsquestionable. 15 Ma old: Afar [Vidal et al., 1991], Ethiopianrift [tlart et al., On the otherhand,thenonrecognition of the isotopicsignature 1989],SaudiArabiacoastalplain[HegnerandPallister,1989;H. (HIMU-like) of the plumedoesnot necessarilymeanthat it was BertrandandG. Chazot,unpublished data,1991]. As pointedout inactive at that time. One can wonder whether it may have by theseauthors, all thesedataindicatethatmeltingof enriched supplied(at leastnearits centralaxis) the excessheatrequiredto subcontinental lithosphericsourcesis likely to be a constant triggermeltingwithinthe continentalmanfielithosphere, provided featureduringtheinitialstagesof rifting.Nevertheless, we point thatthismantlewedgehaspreviouslybeenhydratedduringPanoutthatthebasalts withthehighest 87Sr/86Sr ratios areamong the African events [Engel et al., 1980; Harris and Gass, 1981; mostalteredso thattheirSr sourceisotopicratiosmay be slightly Stoeserand Camp, 1985; Gallagherand Hawkesworth,1992]. If overestimated, although,in ouropinion,thegeneraltrendsarenot this assumptionis correct,however,it can hardly accountfor the questionable. In Ethiopiaand SaudiArabia,the trendspartly meltingof lithosphericmantleunderneathSaudiArabia as well, overlaptheYemenfieldandarealsoconsistent witha mixingof which is locatedmuch farther from the hot plume axis. It may enriched lithospheric anddepleted asthenospheric sources [Hartet explainwhy the SaudiArabia datasetis shiftedmore towardsthe al., 1989; Hegner and Pallister, 1989]. In thesethree areas, DMM component(Figure 14). This early plume hypothesis asthenospheric meltingcertainlypredominates overlithospheric appearscompatiblewith the chronologicalconstraints: the oldest
melting.For example,in A1 Lith area,Schillinget al. [1992] 40Ar-39Ar ages (29Ma)obtained onSouthern Yemen volcanics calculations attribute5-15%meltingto thelithosphere against 75- [Fdraudet al., 1991; V. Zumbo et al., manuscriptin preparation, 80% to the asthenospheric mantle.Similarestimatesarisefrom 1992] arebroadlycontemporaneous with the initial impingement SouthernYemen,usingthe samemixing model.The Afar case of the starting plume head against the lithosphere,which is appears to be somewhat differentbecausetheisotopicevolution assumedto have occurred around 30 Ma ago [White and defined by theAdoleibasalts doesnotestablish theinfluence of McKenzie,1989].The relative timingof therift-relatedtectonic asthenospheric depletedmantle,whereastwo of thesebasalts eventsand igneousactivity, a key point with regard to the exhibita HIMU signature,suggesting a binarymixtureof mechanism of rifting,is still to be improved. lithosphere andHIMU-type mantle[Vidal et al., 1991]. However, comprehensive isotopicstudiesrelative to the onsetof rifting in theseregionsarestill scarceandneedto be developed. Vidal et al. [1991] andSchillinget al. [1992] call attentionto the key point of the contributionof a HIMU componentto the composition of therecentbasaltsin theAfar area.They interpret this contributionas the resultof the increasinginfluenceof the uprisingAfar plume [Schilling,1973] in the courseof the rift
CONCLUSIONS
1. Oligo-Miocenetraps,dykesandplutonsare associated with theearlyRed Sea-Gulfof Adenrifting in SouthernYemen.They displaya bimodalsuite,rangingfrombasaltsandferrobasalts to trachytes-rhyolites andcomendites-pantellerites. 2. Major and trace element distributionsand mass-balance calculations indicate that the evolution from basalts to felsic rocks