Magma genesis in an ongoing rifting zone: The Tadjoura Gulf (Afar area)

0016-7037/93/$6.00

Gwchimica er Cosmochimica Acta Vol. 57. pp. 2291-2302 copyright 0 1993 Pe~mon Press Ltd.Printiin U.S.A.

+ .OO

Magma genesis in an ongoing rifting zone: The Tadjoura Gulf (Afar area) J . A . BARRAT,“* B. M. JAHN,’ S. FOURCADE,’ and J. L. JORON’ ‘G6osciences Rennes, Universit6 de Rennes 1, Campus de BeauIieu, 35042 Rennes Cedex, France tiroupe des Sciences de la Terre, Laboratoire Pierre Siie, C.E.N. Saclay, 9 1 t 90 Gif sur Yvette, France (Received April 22, 1992; accepted in revisedform November 10, 1992) Abstract-Basalts from the Tadjoura Gulf (Afar area, northeastern

Africa) have been analyzed for chemical and isotopic (Sr, Nd, 0) ~rn~~tions in order to understand processes of magma genesis in relation to the active rifting in this region. We use these data to characterize the mantle source of the basalts and to examine the potential role of lithospheric mantle in their genesis. The isotopic compositions of Sr, Nd, and 0 show considerable variation, with %r/%r = 0.703O.707, ENd= +8.1 to -0.5, and S”‘O = +4.6 to +8.4. Combined with major and trace-element data, these isotopic results suggest that the basalts were essentially produced by partial melting of the mantle sources comparable to those of southern Red Sea basalts namely, both depleted mantle and somewhat enriched plume sources are responsible for the generation of the magmas. The continent tith~phe~c mantle seems to have played only a very minor role in the petrogenesis. Prior to the opening ofthe Tadjoura Gulf and the Asal Rift, some lavas penetrated throu8h continental crust, resulting in variable degrees of crustal contamination. At present, the crust of the Tadjoura Gulf and the Asal Rift is dominantly basaltic. Some lavas of the “recent series” show very unusual chemical and isotopic characteristics: hii Ta/Th ratios, positive Eu anomalies, and low 6 ‘*O values. These features can be best explained by a ~n~nation process involving assimilation of hy~o~e~~y altered gabbros by some b&tic ma&as. INTRODUCTION

and CIVEXTA, 1984; BARRATet al., 1990,VIDAL et al., 1991). The purposes of this study are to better characterize the mantle reservoirs involved and to evaluate the roles of lithospheric mantle and crustal ~on~mination in the magma genesis.

THE ROLE OF LITHOSPHERIC MANTLE in the genesis of continental rift basalts is often ambiguous and has been debated vigorously in recent years. In fact, the possible intem~on between magmas and continental crust has complicated the interpretation of chemical and isotopic data, particularly when the true nature of the lithospheric mantle is not yet well understood. The combined use of incompatible tram elements and radiogenic and stable isotopes appears to be indis~n~ble in this context not only for the discussion of mantle soum: heterogeneity and partial melting effects, but also for a better understanding of the combined effects of fractional crystallisation and crustal contamination. In northeastern Africa and Arabia, enormous volumes of lava have erupted during the last 30 Ma. More than 350,000 km3 of lava has accumulated in the Yemen and Ethiopian plateau ( MOHR, 1983). The existence of a hot spot in the Afar and southern Red Sea region is now well recognized. The hot spot activities are intimately associated with the lithospheric opening and creation of the Red Sea and Aden Gulf ( SCHILING, 1973; SCHILLING et al., 1992; BARRAT et al., 1990; VIDAL et al., 1991). Basaltic rocks of the Gulf of Tadjoura area (Republic of Djibouti; Fig. 1) are particularly favorable for studying interactions between the lithosphere and the asthenosphere from the initial continental rifting to the formation of oceanic crust. Such interactions are reflected in the chemical and isotopic characteristics of the hvas erupted over time in various places ( BARBERI et al., 1980; JORON et al., 1980a,b, BETTON

DJIBOUTI MORE THAN 20 Ma OF VOLCANIC ERUPTIONS The opening of the Tadjoura Gulf has a complex history of tectonic and volcanic activities (Fig. 1). The Tadjoura Gulf is the western end of the Aden Gulf, and the active rifting is propagating into the Afi-ican ~ntinent (e.g., COURTILLOT et al., 1980). According to BARBERIet al. ( 1975) and RICHARD( 1979), the principai volcanic formations (in chronological order) are as follows.

1) The Adolti series, mainly altered basalts, with an unknown thickness as only the uppermost section is present. K-Ar ages of 2015 Ma have been obtained for these basalts ( BARBERI et al., 1975). These lavas are thought to be ~n~rn~~~~ with the eruptions of Yemen and Ethiopian trap basalts (MOHR, 1983; CHIESAet al., 1989; FERAUDet al., 1991). 2) The Mabla series ( 15-9 Ma; BARBERIet al., 1975), composed of thick flows of rhyolites and subordinate amounts of basalts ( GADALIAand VARET, 1983 ) . 3) The Dalha series, in which about 1000 m of basaltic flows erupted just before the initial opening of the Tadjoura Gulf (TREUiLand VARET, 1973; RICHARD,1979), which took place about 3.5 Ma ago. 4) The “initial series” of Tadjoura (partly below sea level), formed by a number of basaltic eruptions, are contemporaneous with the “stratoid series” of basal6 in central Afar. 5) The present volcanic activity is very important both on land and in submarine rifts, particularly in the Asal Rift, where the most recent eruption took place in 1978. SAMPLING

AND ANALYTICAL

PROCEDURES

For the present analyses, the fresh samples from the recent basaltic series were chosen to cover the entire compositional range as established by previous workers (STIELTJESet al., 1973; JORONet al.,

* Present address: University of Southampton, Dept. of Geology, Highfield, Southampton SO9 5NH, UK.

2291

2292

J. A. Barrat et al. standard = 0.7 1024 and La Jolla Nd standard = 0.5 11860. The precision for individual runs (2 standard errors) was generally better than 0.002%. The reproducibility was +0.004% during the course of the study. Because of the young ages of all the samples studied, their measured isotopic compositions are considered as initial values. Oxygen was extracted from rock powder samples by reaction with BrFs following the method ofCl_AYroN and MAYEDA ( 1963). Oxygen isotopic compositions wem measured using a VG SIRAmass spectrometer equipped with three collectors at the University of Rennes. The results are given in 6 notation relative to SMOW. The

12”N

6’rO of standard NBS-28 obtained during the period of study was +9.56 -+ 0.04, based on ten runs; and that of our internal standard CIRCE 93, a basaltic glass, was +5.66 If:0.02, based on thirteen runs. The last value is indistinguishable from that obtained by F. Pineau at La Jolla and in Paris (8 ‘*O = +5.68 & 0.14). AU extractions were duplicated, and the reproducibility (given for each sample in Table 3 ) was generally better than 0. I 6 unit. The uncertainties are assumed to be 0. I d unit.

1000 1500 Ta 3 iii,

RESULTS

AND DISCUSSION

Elemental Geochemistry

1 6

iP0 5 _ I

42’30’

I

43’E

I

(

-1

43’30’

FIG. 1. Simplified geological map of the Gulf of Tadjoura area (geological patterns: I = Adolei series; 2 = Mabla series; 3 = Dalha series; 4 = stratoid series; 5 = initial series; 6 = Asal Rift, 7 = Quaternary superficial formations) and sampling localities ofthe Cyuden expedition (zones l-6). The depth, Ta/Th ratio, and 6 ‘a0 variations with longitude are given in this figure. 1980a,b; BARRATet al., 1990). Additional samples from the Mabla and Dalha series were also selected for analyses. Sample descriptions and precise geographic localities of most of the samples can be found in STIELTJES( l973), RICHARD( 1979), GADALIA( 1980), and BARRATet al. ( 1990). Rock powder samples were prepared using only agate mortars in order to avoid any contamination of metals such as Co, Ta, and Nb. Multiple analytical methods have been employed. All new major elements and some trace elements (Nb, Sr, V, Y) were determined by the XRF method at University of Rennes using a Philips PW I404 spectrometer. Analytical precision for most major elements is 2% but may reach IO% for minor elements, such as MnO and P205. For trace elements, the uncertainties are generally better than 5%. For some samples, REEs and Sr were determined by the isotope dilution method. The uncertainties are 3% for La and Lu, 2% for other REEs, and 1% for Sr. The excellent reproducibility and internal consistency are shown by the results obtained with two separate dissolutions of the USGS standard rock BHVO I (Table 2 ) . Other traceelement abundances were determined by the INAA method at Centre d’Etudes NuclCaires, Saclay. The analytical procedures have been described by CHAYLAet al. ( 1973). The results are given relative to the values of CRPG standard GSN used at Saclay. The precision is better than 5%. Isotopic compositions of Sr and Nd were measured using a Fit&an MAT 262 mass spectrometer equipped with seven collectors at the University of Rennes. Ion intensities ofdifferent masses were collected in static mode. Isotopic ratios were normalized against 88Sr/86Sr = 8.375 and ltiNd/‘@Nd = 0.72 19 for Sr and Nd, respectively. Values obtained for standards during the time period of data acquisition are as follows: 14’Nd/ ‘“Nd = 0.5 I I834 +- 2 ( I a; n = 6) for La Jolla Nd standard, 87Sr/86Sr = 0.7 10223 ? 8 ( I cr; n = 14) for NBS SRM-987 Sr standard. The results are reported relative to NBS SRM-987 Sr

The major element characteristics of basalts and the effects of fractional crystallisation (removal of olivine, plagioclase, clinopyroxene, and Fe-Ti oxides) have been described by previous workers ( STIELTJES et al., 1973; RICHARD, 1979; JORON et al., 1980a; GADALIA, 1980; AUDIN et al., 1990). They will not be repeated here in any detail. Some representative analyses are given in Table 1. According to the classification of YODER and TILLEY ( 1962), these basalts are mainly olivine tholeiites, with only a few samples containing normative quartz or nepheline (Table I ). Primitive basalts are rare, and most of them show differentiated features, often with iron enrichment. In terms of highly incompatible elements, the basalts show important variations in element ratios ( JORON et al., 1980a,b; BARRAT et al., 1990). Figure 2 illustrates the chemical diversity of these rocks, using some incompatible element plots. Moreover, the rocks have an impressive range of REE abundances, showing LREE-emiched to LREE-depleted N-MORB type patterns (Fig. 3). The N-MORB type basalts are now erupted in the east of the Tadjoura Gulf. There are other important distinctions between different volcanic series. For example, La/Ta ratios in the basalts older than the Dalha series are 8 to 20 (Fig. 2~); whereas in Dalha and younger basalts, the ratios are less than 12. Some recent basalts (e.g., 102-2, Fig. 2a) have very unusual elemental patterns, exhibiting enrichment in LREEs and Ta, Nb, but they are low in Th, U, and Rb relative to the primitive mantle or chondrites. Consequently, these rocks have unusually high Ta/Th ratios reaching a value of 3.5 (this ratio is about 1 for LREE-enriched oceanic basalts). The high Ta/Th ratio is not the only special feature of the Djibouti lavas. Basalts of the Asal Rift and of the Tadjoura “initial series” crystallised in low oxygen fugacity conditions (close to the Ni-NiO buffer; see RICHARD, 1979). Experimental studies have demonstrated that, in such conditions, plagioclase has a strong positive Eu anomaly (e.g., SUN et al., 1974; DRAKE and WEILL, 1975). In consequence, plagioclase fractionation would likely result in negative Eu anomaly in residual liquids. Despite the clear differentiated character shown in most LREE-enriched basalts (STIELTJES et al., 1973; RICHARD, 1979), no such negative Eu anomalies have been observed (Fig. 3). On the contrary, positive Eu anomalies are identified ( Eu / Eu * = I .OO- I .I6 )

3.3 14.0! 13.2: 0.11 4.1! 8.4' 3.5! 1.2' 0.51 0.5! 99.31 42.4:

23.77 10.93 0.96

0.53

29.04 17.86

27.98 18.36 0.43

Lailb LafTa TatTh

4.43 28.85 19.69 19.62 6.67 8.48

8.3 258 2.84 0.56 2.74 35 29.95 387 6.18 3.07 1.26 47 50 45.4 11.3 40.8 435

7.1! 30.04 18.8: 16.34 14.1;

22.47 10.42 1.17

3.0 ?12 1.53 0.32 1.79 23 18.65 432 4.05 2.31 0.63 29 158 50.8 69 42.7 406

2.84 24.79 24.39 23.01 2.21 12.11

23.67 10.53 1.18

4.2 228 1.93 0.33 2.27 28 23.91 378 5.25 2.74 1.01 34 135 48.5 52 43.8 463

3.0' 22.31 23.01 19.3! 14.9: 2.7(

811-234 816-250 46.98 46.2 2 3.15 3.47 14.28 13.3 4 15.13 15.3 7 0.21 0.1' Q 6.22 5.8, 2 10.97 9.9 1 2.93 2.6' 4 0.48 0.5 1 0.33 0.41 2 -0.28 1.41 0 100.40 QQ.O!9 48.94 48.0

Dalho baraltr

47.86 3.56 13.71 16.49 0.25 4.91 9.58 3.41 0.75 0.63 -0.32 100.85 40.97

tnts (in ppm) 31.5 18.1 492 583 3.53 4.44 0.68 0.76 1.89 1.87 28 34.70 33.40 493 7.37 7.16 3.01 3.21 1.15 1.24 45 38 2 36.5 35.2 2.8 20 31.1 29.4 320

5.67 28.85 20.19 12.93 16.14

Fe3+=0,15 Fe tot 3.82 0.7:

3.19 14.05 13.74 0.24 3.95 7.92 3.41 0.96 0.55 0.81 99.93 40.12

61

ace ele f% Sa Th U Ta N, La Sr Hf GJ Tb Y Cr CQ Ni SC V

HY 01

az Ne Or Ab Ail Di

;IPW noI

m9r

total

P205 L.O.I.

u20

E cp

Si02 Ti02 Al203 Fe203

I

1 86-214

19.63 9.46 1.53

4.4 188 1.10 0.32 1.68 21 15.90 491 3.94 2.28 0.81 27 307 51.0 163 35.5 353

2.95 20.73 26.98 19.37 7.57 10.87

0.50 0.35 0.65 99.49 58.01

GI-U 46.24 2.92 14.46 13.21 0.19 7.83 10.69 2.45

5.2 3.20 1.22 165 52.6 116 35.6

4.54 2.44 0.93 255 58.3 150 35.7

20.66 9.69 2.99

25.20

16.00

19.35 10.47 1.22

5.3 197 0.87 0.29 2.80

18.50

1.45 2.66 19.32 27.93 18.61

15.00 14.85 0.24 8.00 11.25 2.60 0.45 0.79 0.55 101.05 55.67

44.30 3.02

6.4 168 I.41 0.38 1.72

3.0: 22.9: 24.6! iQ.l( 5.0( 13.0'

TF727 t 46.0'7 2.8 1 7 14.0' 13.51 2 3 0.0: 7.8 1 10.2: 2 2.7 I 2 0.5: 0.41 D 7 0.9' 99.1:3 57.313

7047

~~(S~ELTJ~~~~.,

Initial series -r

TABLE l.Representative an~y~sofDji~u~an

19.75 10.35 0.99

128 45.5 54 37.7

5.3 2.90 1.20

23.70

13.8 355 2.31 0.62 2.29

15.05

0.15 4.14 28.50 20.05 20.48

47.10 2.98 13.70 15.09 0.25 6.30 9.80 3.40 0.70 0.58 1.25 101.15 49.32

s314

22.16 10.14 0.70

158 44.9 66 35.2

4.2 1.72 0.97

21.50

19.8 270 3.05 0.66 2.12

5.02 24.54 23.63 20.50 11.74 7.21

49.60 2.06 14.35 13.38 0.21 6.40 10.10 2.90 0.65 0.26 0.75 101.06 52.72

Rift 6883

Asal

4.2 2.24 0.85 353 58.0 248 31.5

3.2 1.65 0.72 280 47.3 97 44.3

17.53 9.20 2.03

14.90

14.30

19.66 11.92 1.48

6.9 275 0.80 0.17 1.62

2.31 20.7. 24.9. 19.81 8.2; 12.2:

13.60 14.44 0.21 8.80 10.50 2.45 0.40 0.42 0.30 100.97 58.69

3.65

S108 46.20

5.0 121 0.81 0.25 1.20

2.07 20.73 28.35 26.17 1.43 13.58

47.60 1.92 14.80 11.69 0.19 8.30 12.45 2.45 0.35 0.19 0.40 100.34 62.33

6890

3.50 10.00 0.78

180 47.7 139 37.9

0.95 0.69 0.40

1.40

0.18 0.28 0.14

0.53 17.43 32.78 25.04 9.94 8.20

0.09 0.05 2.47 101.04 67.79

15.50 9.78 0.13 8.83 12.90 2.06

0.&3

A303 46.45

8.26 10.74 0.89

3.5 30.4 0.56 0.37 0.50 10 5.37 111 2.01 1.17 0.85 29 212 47.8 119 40.3 353

1.06 19.72 28.66 21.97 20.92 1.2t

16.00 9.16 2.86

286 55.2 156 35.2 372

2.7 84 0.56 0.24 1.66 23 15.20 374 4.2 2.20 0.95 34

1.89 20.90 27.01 17.17 8.13 13.10

3.6 91.9 0.62 0.15 1.65 21 12.20 498 3.89 2.33 0.88 30 245 52.5 155 35.4 390 13.66 7.39 2.66 12.50 8.40 3.47 21.83 IO.97 2.44

2.01 20.9s 25.74 18.2: 8.9Z 14.01

I.2 65.1 0.36 0.14 1.25 19 10.50 464 3.74 2.11 0.64 30 218 56.6 17'2 36.1 430

1.42 20.39 27.71 18.14 7.57 12.79

124-1 45.4s 3.4; 13.81 14.9' 0.2( 8.5! 10.2: 2.41 0.3r 0.41 0.61 100.5; 57.2C

3.9 102 0.85 0.35 2.07 28 22.70 329 5.13 2.70 1.04 36 248 51.1 137 36.4 341

2.01 20.82 26.25 19.23 3.76 14.92

Tadioura Gulf, actlvc axir 1 to-4 116-5 114-S 102-2 49.93 45.48 45.47 45.25 2.73 3.58 1.35 3.08 14.77 14.38 14.53 14.31 15.52 14.27 12.32 15.20 0.21 0.21 0.25 0.20 7.78 7.87 7.60 7.90 11.10 10.42 11.37 10.22 2.47 2.46 2.41 2.33 0.34 0.24 0.18 0.32 0.42 0.51 0.28 0.14 -0.29 -0.62 -0.31 -0.42 99.54 99.19 99.39 99.84 54.17 55.96 59.61 54.78

1973;RICHARD, 1979;JoRoN~~&., 198Oa,b; BARRAT~~~~., 1990,andthiswork).

2294

J.

A. Barrat et al.

The possibility of plagioclase accumulation as the cause of the positive Eu anomalies merits a further consideration. All analyzed basalts are aphyric to slightly porphyritic (except sample LGl (Fig. 3) from the Asal Rift, a tholeiite rich in bytownite megacrysts). Cumulative plagioclase would, in principle, suffice to explain the positive Eu anomalies. Several lines of evidence argue against such an hypothesis, and the positive Eu anomaly may be a genuine characteristic of certain basaltic liquids from Djibouti. The arguments are as follows.

1) In a diagram of Eu/Eu*

I

I

Rb Ba Th

I

U

I

I

K

I

Ta La

I

I

vs. A1203 (Fig. 4), the trends of plagioclase accumulation assuming different Eu concentrations ( 1.6 and 0.48 ppm) may explain some of the data points (some basalts from the Tadjoura Gulf and the Erta’Ale volcanic chain in northern Afar), but not the majority of basalts. 2) Some seafloor samples collected by the submersible during the Cyuden expedition ( 1984) consist of fresh glass devoid of any phenocrysts of plagioclase. These glass samples also show positive Eu anomalies (Fig. 5 ) . For two samples ( 102-2 and 124-2) , concentrations of REEs and Sr were determined for the glassy pillow margin and crystallized pillow center (Table 2). No significant difference has been discerned in terms of their REE characteristics.

-

In conclusion, for most samples, the Eu anomalies are not related to plagioclase accumulation. Furthermore, all lavas showing high Ta/Th ratios also possess positive Eu anomalies. However, the converse is not true: Two ferrobasalts from the Dalha series have positive Eu anomalies, but their Ta/Th ratios are “normal,” about 1. This point will be discussed later.

Hf Tb

FIG. 2. Primitive mantle normal&d element patterns for Djiboutian basalts (primitive mantle values from SUN and MCD~NOUGH, 1989): (a) Gulf of Tadjoura and Asal Rift basal& (b) Dalha basalts; (c) pre-Dalha basalts (JORON et al., 1980a,b; BARRATet al., 1990; and this work).

ITadioura Gulf 1

I

I

I

Lace

I

Nd

I

I

I

SmEuGd

I

I

Dy

Er

I

I

I

YbLu

1

i-1I Lace

I

I

I

I

Nd SmEuGd

I

I

Dy

Er

I

I

YbLu

FIG. 3. Chondrite-normalised REE distribution patterns for representative Djiboutian basalts ( BARRAT et al.. 1990: and this work). (Chondrite abundances from MASUDA et al., 1973, further divided by I .2.)

2295

Petrogenesis of basait in the Afar area of Africa 1.2

Eu Eu*

10

15

20 A1203

FIG. 4. Eu/Eu* vs. AlzO, plot. Ai24 abundances are expressed in wt%. Data from STIELTJE~ et al., 1973; RICHARD, 1979; BARRAT et al., 1990; acd this work. The plagioclase accumulation curves have been calculated with the following parameters:The basalt (L) contains 14.5% Al*O,, 4.63 ppm Sm, 1.6 ppm Eu, and 5.24 ppm Gd; the plagioclases Pll and P12both contain 29%A&O,, 0.23 ppm Sm, 0.26 ppm Gd, and 1.6 and 0.48 ppm Eu, respectively. The percent of

accumulative plagioclases is indicated on the curves.

Isotope Geochemistry (Sr, Nd, and 0) Like the incompatible elemental ratios, the isotopic compositions of Sr, Nd, and 0 of the Djibouti basalts also show an impressive variation (Table 3), as follows: 87Sr/86Sr from 0.704 to 0.707, $d from +8.1 to -0.5, and 6 IgO from +4.6 to +8.4%. The highest Sr and 0 isotopic ratios and the lowest 6Ndvalues are Only found among the oldest lavas. The lowest 6180 values are measured in young lavas. In the following, the discussion will be made in three parts: on the Miocene basalts older than the Dalha series, on the Daiha series, and on younger lavas from the Tadjoura Gulf and Asal Rift. The pre-Dalha basalts The pre-Daiha basalts show a large variation in “SrJg6Sr ratios from 0.7037 to 0.7067. Because of their similar geologic setting, it is reasonable to make a comparison with the “old” basalts from central Afar and the plateau basalts of Ethiopia and Yemen (Fig. 6). In these regions, lavas of high 87Sr/86Sr have also been identified recently ( CHIESAet al., 1989; HART et al., 1989). Such a feature of high 87Sr/86Sr ratios and low tNdvalues has been explained in two opposing manners: either ( 1) it is the genuine character of their sources, probably localized in the mantle lithosphere (HART et al., 1989; VIDAL et al., 199 1); or (2) it is the result of magma contamination by the continental crust during the emplacement ( CHIESA et al., 1989). The iavas of the pre-Dalha series are often altered. Among our set of samples, we have found only two fresh basalts with high “SrJg6Sr suitable for oxygen isotope determinations. The results show that their 6180 values are high, at about +8k (Table 3). However, because these are evolved basalts with low MgO, an oxygen isotopic shift by fractional crystallisation must also be taken into account. For all the known magmatic series which have evolved via closed-system fmctional cxystallisation, it has been observed that an increase of 0.2 to 0.4 6 units accompanies every 10% increase of SiO2, with a maximum of 1.5 6 units when the magma has evolved to rhyoiite (e.g., WWDHEAD et al., 1987). Hence, the SiOz

content of the basalts demonstrates that the high d IgO cannot be interpreted as due to a closed-system evolution. The d ‘*O values of pristine mantle-derived rocks are rarely outside the range +5- +6%0. Nevertheless, some 6 IgO values higher than +66 have been reported in rocks from the subcontinental lithospheric mantle (e.g., JAVOY, 1980). This type of mantle may possess a range of 6180 values between f4.5L and +7.5%0, although its average value (+6.2%0) is not dramatically different from that of “normal” oceanic mantle (see review in KYSER, 1990). Our limited oxygen data set does not rule out a contribution from lithospheric mantle to the genesis of the pre-Dalha basahs ( VIDAL et al., 199 1) , but a contribution of continental crust is strongly suggested. Other arguments favour a process of crustal contamination: ( 1) Mantle peridotite nodules from Arabia, Yemen, and Afar never have 87Sr/86Sr ratios higher than 0.7055 ( BARBERIet al., 1980; MENZIESand MURTHY, 1980; BETTONand CIVETTA, 1984; MENZIESand HAWKFSWORTII, 1987; HENJESKUNST et al., 1990); and (2) quartz xenocrysts are present in some of the oiivine basalts (e.g., sample IR6 ie; see GADALIA, 1980). Although the nature of the present crust (mafic or sialic) in Afar has long been debated ( BARBERIand VARET, 1973; BERKHEMER et al., 1975; MOHR, 1989), the crust of Djibouti was undoubtly of sialic nature when lavas of the pre-Dalha series were erupted. Hence, crustal contamination is not in contradiction with the available data, and the origin of the Ethiopian plateau basalts may be comparable to that of the Yemen lavas. Indeed, CHIESAet al. ( 1989) explain the geochemical variations (major and trace elements, Sr isotopic compositions) of the latter by an AFC assimilation-fractional crystallization (AFC) process. Unfortunately, in the present case, the extent of crustal contamination cannot be quantified, ( 1) because of the diversity of primary basaltic liquids and (2) because the precise composition of the basement is unknown. This precludes any firm determination on the nature of the mantle source involved. The basalts of the Dalha series Significant variations in Sr and Nd isotopic compositions, including the results published by VIDALet al. ( 199 1), are observed in the Dalha lavas: “Srjg6Sr = 0.70340 to 0.70485, and CNd= +6.3 to +2.2. The 6180 values we have obtained are around +5.5%0, but only three fresh samples with similar Sr isotopic compositions have been analysed (see Table 3 ). Further analyses are required to evaluate the real oxygen iso-

Tadjoura glasses with TaKhB2.5 and 61804 %

I

I

I

Lace

I

I

I

I

Nd SmEuGd

I

Dy

I

Er

III

YbLu

FIG. 5. Chondrite-normal&d REE distribution patterns for low d “0 Tadjoura glasses.

2296

J.

A. Barrat et al.

TABLE 2. REE and Sr abundanc~(in ppm) determinedby isotopedilution (glforglasshction)

La

Ce

_!!L

460.6 359.9 544.2

27.76 15.54 48.41

68.86 38.24 118.1

39.82 23.36 71.59

109.4 435.3 426.6 492.7 486.8 486.9

1.66 10.06 10.21 12.12 17.75 16.34

4.90 27.72 27.66 32.54 43.93 42.40

329.2 244.1

13.58 19.26

Sr hIal Rnl 7047 6870 6881

Sm

7

EulEu’

Gd

-!L

Yb

a/Yb)r

5.38 16.46

3.06 1.88 5.38

8.88 5.31 16.64

7.62 4.65 14.89

3.90 2.72 7.62

3.44 2.31 6.29

5.33 4.44 5.08

1.05 1.09 1.00

4.33 20.39 20.41 23.49 30.14 30.07

1.58 5.52 5.45 6.02 7.46 7.37

0.63 2.03 2.04 2.26 2.69 2.67

2.30 5.83 5.84 6.30 7.43 7.56

2.96 5.66 5.64 5.95 6.80 6.74

1.88 2.98 2.98 3.05 3.51 3.50

1.77 2.47 2.50 2.63 2.85 2.80

0.28 0.37 0.38 0.38 0.h2

0.63 2.69 2.70 3.16 4.11 3.85

1.02 1 11 1.12 1.13 1.12 1.10

32.30 47.04

20.41 30.23

5.32 7.71

1.81 2.63

5.40 8.41

5.15 8.47

2.83 4.83

2.40 4.30

0.37 0.65

3.74 2.96

1.04 1.01

29.57 18.02 22.41

66.67 42.47 51.45

38.56 24.74 30.11

8.72 5.82 6.93

2.96 2.13 2.45

0.78 5.81 6.89

8.08 5.26 6.15

4.27 2.75 3.22

3.58 2.23 2.63

0.55 0.33 0.40

5.45 5.34 5.63

1.04 1.13 1.10

33.76

75.44

41.37

9.08

2.99

8.81

7.66

3.97

3.28

0.48

6.80

1.03

4.86 5.12 5.25

0.92 1.05 1.05

9.08

AL

radjwra GL A3D3 102-Z 102-291 124-191 124-Z 124-291

nitial Series TF715 TF746A Dalha series 66-214 811.234 816-250

Fabla serie! 836.167

-

Standards BUS-1 BHVO-1 BHVO-1

380.3 379.5

25.00 14.80 14.94

53.80 37.48 37.68

28.82 24.21 24.26

6.59 6.00 5.96

1.97 2.05 2.03

topic diversity of these lavas. Whatever the case, the distribution of the data points in the +,d vs. *‘Sr/%r (Fig. 6) suggests that the eruption of the Dalha series is a “turning point” in the Djiboutian magmagenesis. Clearly, the involvement of the component characterized by high 87Sr/S6Sr ratios and low tN,, values decreases during this period, and basalts similar to some of the recent volcanics are produced. The recent lavas,fkom Tadjoura Gulfand Asal Rifi

Using hygromagmaphile trace elements and Nd isotopic arguments, BARRAT et al. ( 1990) identified three reservoirs which are involved in the generation of recent lavas in Djibouti: ( I ) a depleted MORB mantle ( DMM ), (2 ) a hot spot type reservoir (Ramad-enriched component, or REC), and (3) a probable lithospheric reservoir characterized by high Ta/Th ratio (Tadjoura-enriched component, or TEC). Our new results, covering a wider spectrum of Djibouti basalts (Table 3 ), allow us to construct an improved petrogenetic model. Large variations in Sr and Nd isotopic compositions are observed in the basalts: 87Sr/86Sr = 0.7030 to 0.7046, and tNd = +8.1 to +4.5. The distribution of data points (Fig, 6) clearly indicates involvement of multiple components. The existence of depleted basalt (depleted in LREEs, relatively high cNdof +8.1, and low 87Sr/86Sr of about 0.703, as represented by sample A3D3) suggests the participation of the DMM component. On the other hand, basalts enriched in LREEs have isotopic compositions falling in the field defined by the basalts of the southern Red Sea and Erta’Ale in northern Afar. This indicates that the hot spot type REC component has played a significant role in the magma genesis. Two important points need to be addressed here.

1) A basalt from the Asal Rift (S314; see Table 3 and Fig. 6) has peculiar isotopic characteristics comparable with those of peridotite nodules from Assab ( BARBERIet al., 1980; BETTON and CIVETTA, 1984), which represent the cnly known samples from the lithospheric mantle of the

6.69 6.04 6.00

6.40 5.18 519

3.65 2.49 2.49

3.40 1.91 1.88

0.52 0.28 0.28

Afar region (Assab is situated about 200 km to the north of the Tadjoura Gulf). Provided that part of the mantle beneath the Asal Rift is also character&d by this composition, a lithospheric mantle source may be involved in the genesis of this particular sample. Its oxygen isotopic composition is not in contradiction with this interpretation (&I80 = 5.6%). 2) Figure 7 shows that rocks with high Ta/Th ratios do not show any obvious distinction in Sr and Nd isotopic compositions with respect to the rest of samples. By contrast, 6 I80 and Ta/Th ratios show a negative correlation. The sensitivity of oxygen isotopic composition due to alteration and hydrothermal processes is well recognized. Therefore, we have analyzed only a few exceptionally fresh samples devoid of any visible alteration, including fresh glasses. Despite this effort, the young basalts from the Tadjoura Gulf, Asal Rift, and ferrobasalts from the Dalha series show a range of oxygen isotopic compositions from +4.6 to

10 E Nd

r

t

8 6 4 t-central

2

Afar

0 -2

I

0.703

I

0.704

0.705

0.706

I

8% r/*% r FIG. 6. tNd vs. *‘Sr/%r plot for Djiboutian basalts (data from BETTON and CIVETTA, 1984; H~~~etal., 1989; EISSEN etal., 1989; VIDAL et al., 1991; BARRAT et al.. 1990; unpubl. results; and this work). 14’Nd/144Nd= 0.5 12638 corresponds to tNd = 0 today.

2291

Petrogenesis of basalt in the Afar area of Africa

TABLE 3. FezOl (%), trace-element (ppm) contents, and isotopic composition for Djiboutian lavas and Ethiopian alkali basalts (TB: transitional basalt: PI-TB: transitional basalt rich in plagioclase; Haw: hawaiite; Qz T: quartz tholeiite). Data from JORON et al. (1980a,b); GADALIA (l&O); BE'ITONand CIVETTA(1984); BARRATet al. ( 1990); and this work.

~

Ta w

rock

l-h

La INAA

Tb w

_aiTa

ran-h

iu/Eu’

INAA

INAA

INAA

k

0.85 0.86 0.89 0.85

1.00 0.99 1.02 0.96

iI180

87Sr 86Sr

143Nd 144Nd

5,51f0,05 5,56K),Ol 5,92kO,o2 5,47*0,05

0.70315 0.70312 0.70319 0.70313

3.513028 D.512977 0.513029 D.512970

0.70309

0.513052

0.70363 0.70359 0.70359 0.70375 0.70353 0.70354

0.512927 0.512957 0.512894 0.512888 0.512912

iulf of

type wn

106-191 110-4 DRl-3gi

'-MORB --MORB --MORB

12.51 12.27 12.23 12.47

0.66 0.66 0.56 0.67

0.56 0.57 0.50 0.57

5.8 5.6 5.4 6.0

0.66 0.69 0.65 0.69

8.77 8.12 8.26

8.70

10.34 9.82 10.74 8.95

dredge A; A3D3

S-MORE

9.19

0.18

0.14

1.4

0.4

3.50

10.00

0.78

1.02

zone2 114-1 114-5 116-5 116-4 102-291

TB TB TB TB TB

13.80 14.16 14.80 15.05

0.96 0.85 0.58 0.74

2.10 2.07 1.66 2.19

23.0 22.7 15.2 21.3

1.09 1.04 0.96 1.16

21.10 21.83 16.00 18.36

10.95 10.97 9.16 9.73

2.19 2.44 2.86 2.96

1.03 1.02 1.06 1.05 1.12

5,451tO,O2 4,Q9*0,04 5,39M,02 4,Q8*0,01 4,70?00,06

102-2

TB

15.10

0.36

1.25

10.5

0.84

12.50

8.40

3.47

1.11

4,97*0,02

zone4 124-191 124-291 124-3 124-5

TB TB TB TB

14.33 14.89 14.64 15.01

0.62 0.66 0.63 0.62

1.65 2.06 2.04 2.10

12.2 15.8 15.8 16.5

0.88 1.05 1.05 1.09

13.86 15.05 15.05 15.14

7.39 7.67 7.75 7.86

2.66 3.12 3.24 3.39

1.13 1.10 1.10 1.11

4,75iOo,03 4.59+0,04 4,83+_0,06 4.98*0,08

0.70353 0.70354 0.70353 0.70356

tnitials~1 TF746A TF727 GH3 TF717 TF715

'S TB TB TB TB TB

20.36 15.48 13.24 14.73 13.93

1.69 1.41 1.10 0.86 0.51

1.84 1.72 1.68 1.38 1.12

20.2 18.0 15.9 14.8 12.8

1.29 0.93 0.81 0.89 0.79

15.67 19.35 19.63 16.63 16.20

10.98 10.47 9.46 10.72 11.43

1.09 1.22 1.53 1.60 2.19

1.01

5.72rtO.04 5,31*0.07 5,82fO,Ol 5,39+0,02 5,17+0,01

0.70378 0.70390 0.70368 0.70382 0.70404

TB PI-TB TB TB

12.81 8.65 13.02 15.09

0.83 0.46 3.05 2.31

1.65 0.73 2.12 2.29

16.9 6.4 21.5 23.7

0.85 0.47 0.97 1.2

19.88 13.62 22.16 19.75

10.24 8.77 10.14 10.34

1.99 0.98 0.69 0.99

1.10 1.16

Haw TB TB TB TB TB TB TB TB

15.26 14.70 11.16 12.93 14.53 14.44 12.22 16.01 15.05

4.20 1.33 0.81 0.68 1.13 0.80 0.77 0.88 '0.87

4.28 1.60 1.20 1.15 1.84 1.62 1.87 2.60 2.60

50.0 16.8 14.3 9.6 19.8 14.9 16.0 26.0 25.2

2.34 1.06 0.72 0.74 1 0.85 0.83 1.18 1.22

21.36 15.85 19.86 13.00 19.80 17.53 19.28 22.03 20.66

11.68 10.50 11.92 8.36 10.20 9.20 8.56 10.00 9.69

1.01 1.20 1.48 1.69 1.72 2.02 2.43 2.95 2.99

1.00

TB TB TB TB

16.23 14.85 15.02 15.57

2.84 1.44 1.53 1.93

2.74 1.83 1.79 2.27

30.0 18.6 18.7 23.9

1.26 0.84 0.83 1.01

23.77 22.18 22.47 23.67

10.93 10.18 10.42 10.53

0.96 1.27 1.17 1.18

1.04

12.25

1.90 2.85 6.28 3.34 0.91 2.44 3.06

1.67 1.01 3.16 2.42 0.95 2.30 1.55

22.2 15.6 26.2 46.2 12.9 27.6 29.5

1.13 0.74 1.01 1.4 0.67 1.19 0.96

19.65 21.08 25.94 33.00 19.25 23.19 30.80

13.29 15.60 8.29 19.09 13.58 12.00 19.02

0.88 0.35 0.50 0.72 1.04 0.94 0.51

13.48 12.83

4.44 3.53

1.89 1.87

34.7 33.4

1.24 1.15

27.98 29.04

18.26 17.86

0.44 0.53

3.60

40.80

2.81

70.0

0.54

129.63 24.91

0.07

Ta ~CWe axi zone 1 '-MORB 103-l gl

Ass1 Rift 120-l LGl 6883 s314 6881 6942 6890 LG2 6897 S108 6870 7043 7047 Mha ser 86-214 811-232 811-234 816-250

Standard GSN

1.04

5.98fl.02 5,62%X05

1.09 1.05

6,22+0,01 558~0.01 5,17fo,lO 5,38&0,04 5,22kO,O4 5,65&0,07 5,31kO,o3 5,23+0,03 5,34+0.03

0.70357 0.70365 0.70382 0.70456 0.70455 0.70374 0.70359 0.70391 0.70361 0.70372 0.70375 0.70399 0.70375 0.70386

0.512924

0.51287; 0.512W 0.51288;

1.51286,:

0.5129Oi

6

Jre-Dalhc olcanic TB IR57b IFi61e TB TA? IR67f TA IR69d TB IR73b TB IR75d QzT? B35-157 836-167 115-10

1.10

0.51287C 0.51288Z 0.51290:

C&T QZT

1.13 1.10

1.03

5,64iO,lO 5,42iO,O5 5,32*0,07

0.70386 0.70390 0.70390 0.70408

0.70376 0.70613 0.70386 0.70429 0.70372 0.70406 0.70573 0.70573 8,37ztO,O8 0.70664 7,98*0,10 0.70571

0.512846

0.512611 0.51261E

2298

J. A. Barrat et al. r

87Sr

86sr 0.704

6 6180

2 Ta/Th FIG. 7. “Sr/“Sr, cNdra’*0 vs. Ta/Th plots for the recent Djiboutian basal& the Erta’Ale chain ( E’A) , and the South of the Red Sea ( RS ) See Fig. 4 for the symbols. (Data from BARBER]et al., 1980; JORON et al., 1980a,b; BE’II-ONand CIVETTA,1984; EISSENet al., 1989; BARRATet al., 1990; unpubl. results; and this work.)

+6.2k, a range which exceeds that recorded in oceanic basalts. The 6”O value of MORBs is +5.7 -t 0.3%0 ( MUEHLENBACHSand CLAYTON, 1972; RNEAU et al., 1976; ITO et al., 1987), and OIBs may possess a less restricted range of isotopic compositions (al80 = +.5%0 to ca +6.5Ym; see review in KYSER, 1990). The oxygen isotopic ratios of most young basalts analyzed are “normal,” compared to those reference values; but some samples, located in zones 2 and 3 of the active rift in the Tadjoura Gulf, display anomalously low 6 “0 values. Moreover, such low 6 “0 values are detected in fresh, optically isotropic glasses. Hence, these lavas appear to have been erupted as low IgO magmas. Several mechanisms have been invoked to explain the origin of low 6 IgO lavas (e.g., MUEHLENBACHSet al., 1974; FRIEDMAN et al., 1974; HILDRETH et al., 1984; TAYLOR, 1987). They include ( 1) direct interactions between magmas and fluids, (2) partial melting of a low 6 IgO source, and (3) contamination of an isotopically normal magma by a crustal component with low 6l8O. Among them, the first hypothesis may be excluded as it has been argued that fluid-magma interactions cannot result in low 6 “0 magmas (TAYLOR and SHEPPARD, 1986). Thus, it appears that a special component depleted in IgO is involved in the genesis of Djibouti basalts. AS shown earlier, a clear negative correlation exists between oxygen isotopic compositions and Ta/Th ratios (Fig. 7). All basalts with low 6180 also have high Ta/Th ratios and positive Eu anomalies (see Elemental Geochemistry section, Table

3, and Fig. 5). Therefore, the low IgO feature is also a characteristic of the special component TEC. The question now is: Is the TEC a mantle component, or is it only a crustal contaminant? Is the TEC a mantle component? The ratio of Ta/Th in melts would be the same as the source because of the extremely low distribution coefficients of both Ta and Th (co.0 1). Accordingly, the correlation between Ta/Th and 6 IgO (Fig. 7) could be interpreted as resulting from various contributions from two mantle reservoirs: a “normal” mantle with 6 IgO = 5.5-6.0%0, and Ta/Th = 1; and an “abnormal” mantle component (TEC) with a low SIB0 value (~4.6460), and Ta/Th 23.5. However, the nature of this abnormal component is not clear. An increase of Ta/Th ratio may be marginally achieved in a source which has previously experienced melt extraction, through preferential partitioning of Th into the liquid (residual source with high Ta/Th ratio), or through metasomatic processes which enriched Ta relative to Th ( O’REILLY and GRI~N, 1988; PACESand BELL, 1989; BARRAT et al., 1990). However, neither low 6 IgO values nor positive Eu anomalies are readily understandable in the light of such processes. The geographic variations in Ta/Th ratios ( BARRATet al., 1990) and in S “0 of recent lavas (Fig. 1) allow us to localize the zones of TEC intervention (low 6180; high Ta/Th) to the region in the west ofthe Tadjoura Gulf and the Asal Rift. In contrast, in the east of the Tadjoura Gulf (Zone 1 and site A3), the lavas have similar compositions to those from the southern Red Sea region. Their 6180 values are near those of MORB, and none of them shows high Ta/Th ratios. It appears from the bathymetric profile that the TEC intervention is limited to the emerged zones or to water depths less than 1300 m. This strongly suggests that the TEC is localized in the Djibouti lithosphere (see next section). Although they are rare, mantle rocks depleted in 180( 6 IgO < +5%) are known to exist. For example, eclogite nodules in kimberlites have been found to have 6 IgO values between about +2 to +80/, (e.g., GARLICK et al., 1971; ONCLEY et al., 1987; NEAL et al., 1990). Most of the eclogite nodules are generally interpreted as relics of ancient hydrothermally altered oceanic crust which has been subducted into the mantle. One of the most compelling arguments in favour of this interpretation (e.g., MACGREGOR and MANTON, 1986) is the similarity of oxygen isotopic composition with that of ophiolite sequences. Since partial melting of mantle rocks does not significantly fractionate oxygen isotopic ratios, it has been argued that low “0 basaltic magmas could be derived by melting of mantle sources containing subducted oceanic crust ( CARTWRIGHT and VALLEY, 199 1). Recycling of oceanic lithosphere through subduction processes has been used to explain some special geochemical and isotopic characteristics of certain oceanic basalts (e.g., HOFMANN and WHITE, 1982). To our knowledge, however, no oceanic basalts with enriched LREEs have shown any Ta / Th ratios as high as 3.5, and their oxygen isotopic composition is higher. Moreover, as previously mentioned, the restricted geographic location of the TEC argues against the possibility that it could be associated with the enriched plume-type component. Thus, the localization of the TEC within the local lithosphere is highly probable. The peculiar chemical and isotopic

Petrogenesis of basalt in the Afar area of Africa characteristics of this component are hardly attributable to the subcontinental lithospheric mantle. Some eclogite nodules (Type I; see MACGREGOR and MANTON, 1986) have clinopyroxenes with LREE enrichment and occasionally show positive Eu anomalies. Pyroxenites from the Beni Bousera Massif also show positive Eu anomalies (PEARSON et al., I99 1) . Partial melting of these kinds of rocks could, in principle, generate magmas with the observed geochemical characteristics. However, all of these rocks have 6 “0 values greater than 5%0 ( JAVOY, 1980; MACGREGOR and MANTON, 1986; PEARSONet al., 199 1). Other eclogite nodules (Type II) possess the appropriate low “0 values but, to our knowledge, not the other characteristics of the TEC (positive Eu anomalies; high Ta/Th ratio). Moreover, their isotopic composition of Sr and Nd (e.g., NEAL et al., 1990) is extremely different from thQse of recent Tadjoura basalts (Table 3). We therefore conclude that the TEC probably does not reside in the lithospheric or asthenospheric mantle. Is the TEC localized in the Djibouti crust?We now consider the possibility of magma contamination by a crustal component with low 6’*0. The fact that all lavas with high Ta/ Th ratios are differentiated iron-rich basalts (Tables 1 and 3) suggests that they are more susceptible to crustal contamination due to their longer residence time within magma chambers. The nature of the crust in Djibouti and the Afar region has always been controversial (BERKHEMER et al., 1975; BARBERI and VARET, 1975; MOHR, 1989). Existing hypotheses suggest that the crust is either a stretched thin continental crust injected by basaltic dykes, or alternatively oceanic crust. Seismic refraction studies suggest the presence of a 5-l 1 km thick oceanic crust (RUEGG, 1974). In view of the great quantity and the differentiated character of lavas erupted prior to the opening of the Tadjoura Gulf (greater than one km thickness of lavas for the Dalha series only), the crust here must be predominantly of basic composition, containing a large quantity of gabbroic cumulates. We suggest that such rock types, located in the Djibouti-Afar crust, may represent the TEC, based on the following line of reasoning. Although such gabbroic cumulates are not found as enclaves in the recent lavas, their geochemical characteristics may be estimated from the cogenetic basalts using observations from major layered complexes. Except for some recent lavas from the east of the Tadjoura Gulf, all the basalts (old and young) are enriched in LREEs. Using published distribution coefficients, we conclude that their gabbroic cumulates would also be enriched in LREEs and have variable positive Eu anomalies depending on their plagioclase contents. The Ta/Th ratios are more difficult to estimate. Because of the low Th Kd ( ~0.0 1) for all the major mineral phases, the cumulates will be depleted in Th. However, Ta and Nb contents may be controlled by opaque phases (GREEN and PEARSON,1987 ) . Consequently, the gabbroic cumulates could have high and variable Ta / Th ratios. Hydrothermal activity is very important in the Afar region, including the Asal Rift and the Tadjoura Gulf basin (BOSCH et al., 1977; CHOUKROUNEet al., 1988). At present, the hydrothermal fluids in Asal are dominantly of marine origin ( SANJUANet al., 1990). The effects of hydrothermal alteration in the upper 1550 m of crust have been studied using drill cores ( FOUILLACet al., 1989 ). The intense fracturing caused

2299

by extensional tectonics and the resulting high permeability of these rocks have greatly favoured the circulation of hydrothermal fluids. The results show that whole-rock 6 “0 values are about + 12960in the first hundred meters and are reduced to +5%0 in the core section, at a depth of about 1550 m, where the temperature reaches 260°C. Studies on ophiolite complexes and oceanic crust drill cores have demonstrated that the upper parts are enriched in 180; whereas the lower parts, altered in temperature conditions as high as 35O”C, are often depleted in “0, resulting in low 8 “0 values of +3 to +4%0, even though such an isotopic structure is not systematic (see MUEHLENBACHS,1986, for a review). The existence of a magma chamber beneath Asal at about 5 km of depth is generally accepted (T. Souriot, pers. commun.), and we postulate that gabbroic cumulates of low 6180 may be generated in this region as the result of intense hydrothermal alteration. The important Plio-Pleistocene lacustrine and fluvial deposits of the region ( GASSE et al., 1980) suggest that the climate of Djibouti has not always been arid. Prior to the opening of the Tadjoura Gulf, the hydrothermal fluids could have had a meteoric origin. The Afar precipitations at present have a 6180 value of -3%0 (GONFIANTANI et al., 1973), but those in the Pliocene-Pleistocene time might have had even lower values if the climate was less arid. If so, the fluids could produce depletions of ‘*O in the rocks much more efficiently than marine fluids during water-rock interactions. This is illustrated by the Jabal at Tirf Complex in southern Saudi Arabia. This gabbroic complex of about 20 Ma has been interpreted as a relic of oceanic crust formed in the initial stage of Red Sea opening (COLEMAN et al., 1979; FERAUD et al., 199 1). Most analyzed samples show “0 depletion, and some have 6 “0 values as low as -2%0 (TAYLOR, 1980). It is possible to estimate the Sr and Nd isotopic compositions of the crustal gabbros. Before interactions with hydrothermal fluids, the isotopic compositions of the lavas and associated cumulates were identical. The effects of the hydrothermal alteration are qualitatively predictible. As the Nd content of hydrothermal fluids is negligible compared to concentrations in the gabbros, no important effects on the isotopic compositions are produced, whatever the values of the water/rock ratios. The Sr content of hydrothermal fluids is, however, not always insignificant (~8 ppm or more for marine hydrothermal fluids). A shift to more radiogenic “Sr/ ?Sr values is possible (especially for marine hydrothermal fluids), but its extent is strongly dependent on the water/rock ratios and the Sr contents of the gabbros. Ophiolite studies have shown that the lower oceanic crust (which has experienced “0 depletion) is not significantly contaminated by marine Sr due to the low water/rock ratios (e.g., the Samail ophiolite; see GREGORY and TAYLOR, 1981; MCCULLOCH et al., 198 1). In addition, the high Sr contents of the LREEenriched lavas imply that the Sr abundances of the Djiboutian cumulates are higher than in ophiolitic gabbros. Consequently, their Sr isotopic compositions are less sensitive to hydrothermal alteration. We conclude that ( 1) low 6 ‘*O crustal gabbroic cumulates and the parent lavas of the low 6 “‘0 basalts probably have similar Sr and Nd isotopic compositions, and that (2) the basalts contaminated by this kind of material will not show perturbed Sr and Nd isotopic compositions. The fact that the basalts with high and low Ta/Th

2300

J. A. Barrat et al.

ratios cannot be distinguished by their tNdand 87Sr/86Sr (Fig. 7) strongly favours this hypothesis. Contamination of basaltic magmas by these kinds of gabbroic cumulates could have been responsible for the unusual chemical and isotopic characteristics of the Djibouti lavas. We emphasize that in the gabbroic cumulates, the Ta/Th and Eu/Eu* ratios are independent variables because they are not related to the same phases ( Fe-Ti oxides and plagioclase, respectively). The presence of ferrobasalts (samples B 11 and B 16; see Tables 2 and 3 ) with positive Eu anomalies but normal Ta/Th and 6 ‘*O values can be explained by the diversity of contaminants (= gabbroic cumulates more or less affected by hydrothermalism) . Quantitative modelling of this process is difficult because its mechanism and the parameters needed for the calculation are poorly known or underconstrained. Estimations can be made assuming an AFC model ( DEPAOLO, 198 1) with the following assumptions: The uncontaminated basalt, the gabbroic contaminant, and the contaminated basalt are assumed to have a 6’*0 of +5.5%0, OL, and +4.7%0 respectively; is assumed to be -0.3%0. It is worth noting here a, umulate_liq that the calculated F value ( = mass of contaminated magma/ initial mass of magma) is a function of the ratio of assimilated contaminant to fractionated phases (Y). Assuming r = 0.2, we obtain F = 0.4; for r = 0.4, F = 0.75. Contamination of basaltic magmas by a young, hydrothermally altered mafic crust is our preferred model to explain the peculiar characteristics of some Djiboutian basalts. The exact mechanism of such contamination is difficult to constrain, but it is worth noting that a similar process is known elsewhere. In Iceland, too, contamination by a low “0 hydrothermally altered basic crust produces low ‘*O magmas (e.g., MUEHLENBACHSet al., 1974; SIGMARSSONet al., 1992; and references therein). In this area, the process often occurs in evolved lavas (Fe-Ti-rich basalts, quartz-tholeiites), and the crust is basic and thick. Finally, both areas are a site of extensional tectonics and are favourable for the production of low ‘*O hydrothermally altered basement and possibly, through a two-stage process, of low “0 magmas. CONCLUSIONS The involvement of two mantle components, DMM and REC, in the generation of magmas in northeastern Africa is well established and has been confirmed by several studies (SCHILLING, 1973; HART et al., 1989; BARRAT et al., 1990; VIDALet al., 199 1; SCHILLINGet al., 1992). In addition, traceelement and radiogenic isotopic data have led many authors to postulate a significant participation of lithospheric source in the magma genesis ( JORON et al., 1980b; BETTON and CIVETTA, 1984; HART et al., 1989; BARRAT et al., 1990; VIDALet al., 1991; SCHILLING et al., 1992). The application of the oxygen isotopic tracer technique has added a further constraint to the nature of the participating lithosphere. A few basalts erupted prior to the opening of the Tadjoura Gulf show both high *‘Sr/%Sr and 6 ‘*O values. The presence of quartz xenocrysts in the lavas ( GADALIA, 1980) suggests contamination by the continental crust. The generation of recent lavas is more complicated. The diversity of mantle sources alone cannot satisfactorily explain the chemical and isotopic characteristics of the lavas. Assim-

ilation of hydrothermally altered gabbroic cumulates (component TEC) by basaltic magmas is believed to be responsible for the origin of low 6 I80 and high Ta/Th liquids and positive Eu anomalies. The role of lithospheric mantle in the magma genesis seems to be minor. Of the fifty samples analyzed, only one (Sample No. S314) shows chemical and isotopic compositions that may be inherited from the lithospheric mantle. Besides a minor contribution of the lithospheric mantle, the source of erupted lavas appears to be dominated by the two types of components: DMM and REC. Our data do not require additional mantle components but are compatible with the involvement of two types of crust related to two stages of rift development: sialic crust in the early stages of rifting, then mafic crust when the rift becomes more mature. Acknowledgmen&--J.A.B.

is very grateful for the tutorage of J. Cornichet, J. Mace, F. Martineau, M. Le Coz, and M. Lemoine in chemical separation procedures, mass spectrometry, and XRF analyses. We thank P. Choukroune and B. Auvray for providing the Cyaden samples, and N. Bonhommet and A. Gadalia for the Dalha and Mabla samples. 1. W. Croudace, P. K. Kepezhinskas, R. W. Nesbitt, S. Roberts, and R. N. Taylor are gratefully acknowledged for improving the text. We thank D. W. Mittlefehldt and the formal reviewers A. R. Basu, R. Kay, and M. Menzies for their constructive comments. Financial support program (ATP-PIROCEAN, 1986-1987, DBT 199 1) is highly acknowledged. This paper is contribution 5 and 8 of the INSU-DBT program. Editoriul handling: D. W. Mittlefehldt

REFERENCES AUDIN J.,VELLUTINIP.J.,COULON C.,PIGUET P., and VINVENT J. ( 1990) The 1928-1929 eruption of Kammourta volcano-Evidence of tectono-magmatic activity in the Manda Inakir Rift and comparison with the Asal Rift, Afar depression, Republic of Djibuti. Bull. Volcanol. 52, 55 l-56 I. BARBER]F. and VARET J. (1975) Nature of the Afar crust: A discussion. In Afar Depression QfEthiopia (ed. A. PILGERand A.

ROESLER),pp. 375-378. Schweizerbart, Stuttgart. BARBERIF., FERRARAG., SANTACR~CE R., and VARETJ. (1975) Structural evolution of the Afar triple junction. In Afar Depression of Ethiopia (ed. A. PILGERand A. ROESLER).Schweizerbart, Stuttgart. BARBERIF., CIVETTA L., and VARETJ. ( 1980) Sr isotopic composition of Afar volcanics and its implication for mantle evolution. Earth Planet. Sci. Lett. 50, 247-259.

BARRATJ. A., JAHN B. M., JORONJ. L., AUVRAYB., and HAMDI H. ( 1990) Mantle heterogeneity in northeastern Africa: Evidence from Nd isotopic compositions and hygromagmaphile element geochemistry of basaltic rocks from the Gulf of Tadjoura and southern Red Sea regions. Earth Planet. Sci. Lett. 101,233-241. BERKHEMERH., BARERB., BARTELSENH., BEHLEA., BURHARDT H., GEBRANDEH., MAKRISJ., MENZELH., MILLERH., and VEES R. ( 1975) Deep seismic soundings in the Afar region and on the highland of Ethiopia. In Afar Depression ofEthiopia (ed. A. PILGER and A. ROESLER), pp. 89-107. Schweizerbart, Stuttgart. BETTONP. J. and CIVETTAL. (1984) Strontium and neodymium isotopic evidence for the heterogeneous nature and development of the mantle beneath Afar (Ethiopia). Earth Planet. Sci. Lett. 71, 59-70.

BOSCHB., DESCHAMP~J., LELEU M., LOP~UKHINEM., MARCEA., and VILBERTB. ( 1977) Thegeothennal zone of Lake Assal (FTAI): Geochemical and experimenta studies. Geothermics 5, 165-175. CARTWRIGHTI. and VALLEYJ. W. ( 1991) Low-‘*0 Scourie dike magmas from the Lewisian Complex, northwestern Scotland. Geology 19,578-58 I. CHAYLA B., JAFFREZICH., and JORON J. L. (1973) Analyse per activation dans les neutrons tpithermiques: Application a la ditermination d’elements en traces dans les roches. C. R. Acad. Sci. Paris 277, 273-275.

Petrogenesis of basalt in the Afar area of Africa CHIESAS., CIVE~A L., DE FINO M., LA VOLPE L., and ORSI G.

( 1989) The Yemen trap series: Genesis and evolution of a continental flood basalt province. J. Volcanol. Geotherm. Res. 36,331350. CHOUKROUNEP., FRANCHETEAUJ., AUVRAYB., AUZENDEJ. M., BRUN J. P., SICHLERB., ARTHAUD F., and LUPINEJ. C. ( 1988) Tectonics of an incipient oceanic rift (the western extension of the Aden rift within the Gulf of Tadjoura, Republic of Djibouti). Mar. Geophys. Res. 9, 147-163. CLAYTON R. N. and MAYEDA T. K. ( 1963) The use of bromine pentafluoride in the extraction of oxygen from oxides and silicates for isotopic analysis. Geochim. Cosmochim. Acta 27,43-52. COLEMANR. G., HEDGE C. E., FLECKR. J., GHENT E. D., and DONATO M. M. ( 1979) Miocene Tihama Asir complex and its bearing on the opening of the Red Sea. In Evolution and Mineralization of the Arabian-Nubian Shield (ed. A. M. AL SHANTI), Vol. 1, pp. 173-186. Pergamon Press. COURTILLOTV., GALDEANOA., and LE MOUEL J. L. ( 1980) Propagation of an accreting plate boundary: A discussion of new magnetic data in ihe Gulf of Tadjoura and Southern Afar. Earth Planet. Sci. Lett. 47, 144-160. DEPAOLOD. J. ( 198 I ) Trace element and isotopic effects of combined wall rock assimilation and fractional crystallization. Earth Planet. Sci. Lett. 53, 189-202. DRAKE M. J. and WEILL D. F. (1975) Partition of Sr, Ba, Ca, Y, Eu2+, Eu3+, and other REE between plagioclase feldspar and magmatic liquid: An experimental study. Geochim. Cosmochim. Acta 39,689-7 12. EISSENJ. L., JUTEAU T., JORON J. L., DUPRB B., HUMLER E., and AL’MUKHAMEWV A. I. ( 1989) Petrology and geochemistry of the basalts from the axial rift of the Red Sea at 18”North. J. Petrol. 30,79 l-839. WRAUD G., ZUMBO V., and SEBAI A. (1991) “Ar/39Ar age and duration of tholeiitic magmatism related to the early opening of the Red Sea rift. Geophys. Res. Lett. 18, 195-198. FOUILLACA. M., FOUILLACC., CFSBRONF., PILLARDF., and LEGENDRE0. ( 1989) Water-rock interaction between basalt and high salinity fluids in the Asal Rift, Republic of Djibouti. Chem. Geol. 76,27 l-289. FRIEDMANI., LIPMANP., OBRADOVICHJ. D., GLEASONJ. D., and CHRISTIANSENR. L. ( 1974) Meteoric water in magmas. Science 184, 1069-1072. GADALIAA. ( 1980) Les rhyolites du stade initial de I’ouverture dun rift: Exemple des rhyolites mio&tes de I’Afar. Thesis, Univ. Orsay. GADALIAA. and VARET J. ( 1983) Les rhyolites mioctnes de 1’Est de I’Afar. Bull. Sot. Geol. France 7, 139-153. GARLICKG. D., MACGREGOR1. D., and VOGELD. E. ( 197 1) Oxygen isotope ratios in eclogites from kimberlites. Science 172, 1025 1027. GASSE F., RICHARD O., ROBBE D., ROGNON P., and WILLIAMS M. A. J. ( 1980) Evolution tectonique et climatique de 1’Afarcentral d’aspr& les sediments plio-pleisto&nes. Bull. Sot. Geol. France 7, 987-1001. GONFIANTANIR., B~RSI S., FERRARA G., and PANICHIC. (1973) Isotopic composition of waters from the Danakil depression (Ethiouia). Earth Planet. Sci. Lett. 18. 13-2 I. GREEN T. H. and PEARSONN. J. ( 1987) An experimental study of Nb and Ta partitioning between Ti-rich minerals and silicate liquids at high pressure and temperature. Geochim. Cosmochim. Acta 51, 55-62. GREGORY R. T. and TAYLOR H. P., JR. ( 198 1) An oxygen isotope profile in a section of cretaceous oceanic crust, Samail ophiolite, Oman: Evidence for 6’*0 buffering of the oceans by deep (>5 Km) seawater-hydrothermal circulation at mid-ocean ridges. J. Geophys. Res. 86,2737-2755. HART W. K., WOLDEGABRIELG., WALTERR. C., and MERTZMAN S. A. ( 1989) Basaltic volcanism in Ethiopia: Constraints on continental rifting and mantle interactions. J. Geophys. Res. 94,77317748. HENJES-KUNSTF., ALTHERRF., and BAUMANNA. ( 1990) Evolution and composition of the lithospheric mantle underneath the western Arabian peninsula: Constraints from Sr-Nd &type systematics of mantle xenoliths. Contrib. Mineral. Petrol. 105,460-472. HILDRETHW., CHRISTIANSENR. L., and O’NEIL J. R. ( 1984). Cat-

2301

astrophic isotopic modification of rhyolitic magma at times of caldera subsidence, Yellowstone Plateau volcanic fields. J. Geophys. Res. 89,8339-8369. HOFMANNA. W. and WHITE W. M. ( 1982) Mantle plumes from ancient oceanic crust. Ibid. 57, 421-436. ITO E., WHITE W. M., and G~PEL C. ( 1987) The 0, Sr, Nd, and Pb isotope geochemistry of MORB. Chem. Geol. 62, I57- 176. JAVOY M. (1980) ‘sO/16O and D/H ratios in hiah temoerature neridotites.‘Coll&e Intern. CNR’S No. 272; Ass-wiations Majiques Ultramafiques duns les Orogenes, 279-287. BORON J. L., TREUILM., JAFFREZIC H., and VILLEMANTB. ( 1980a) Etude geochimique des elements en traces darts less&s de roches volcaniques du rift d’Asal. Identification et analyse des processus d’accretion. Bull. Sot. Geol. France 7,85 l-86 I. JORONJ. L., TREUILM., JAFFREZIC H., VILLEMANTV., and RICHARD 0. ( 198Ob) G6ochimie des elements en traces du volcanisme de I’Afar et de la m&structure Mer Rouge-Afar-golfe dAden: Implications p&og&tiques et geodynamiques. Bull. Sot. Geol. France 7,945-957. KYSER T. K. ( 1990) Stable isotopes in the continental lithospheric mantle. In Continental Mantle (ed. M. A. MENZIES), pp. 127156; Oxford Monogr. Geol. Geophys. No. 16. MASUDAA., NAKAMURAN., and TANAKAT. ( 1973) Fine structure of mutually normalized ram earth patterns of chondrites. Geochim. Cosmochim. Acta 49,205 I-2067. MCCULLOCHM. T., GREGORYR. T., WASSERBURGG. J., and TAYLOR H. P., JR. ( 1981) Sm-Nd, Rb-Sr, and ‘8O/‘6O isotopic systematics in an oceanic crustal section: Evidence from the Samail ophiolite. J. Geophys. Res. 86, 2721-2735. MACGREGORI. D. and MANTONW. I. ( 1986) Roberts Victor cclogites: Ancient oceanic crust. J. Geophys. Res. 91, 14,063-14,079. MENZIES M. A. and HAWKESWORTHC. J. (1987) Upper mantle processes and composition. In MantleXenoliths (ed. P. H. NIXON), pp. 725-738. J. Wiley and Sons. MENZIESM. A. and MURTHY V. R. ( 1980) Nd and Sr isotope geochemistry of hydrous mantle nodules and their host alkali basalts: Implications for local heterogeneities in metasomatically veined mantle. Earth Planet. Sci. Lett. 46, 323-334. MOHR P. ( 1983) Ethiopian flood basalt province. Nature 303,577584. MOHR P. ( 1989) Nature of the crust under Afar: New igneous, not thinned continental. Tectonophysics 167, I- I I. MUEHLENBACHSK. ( 1986) Alteration of the oceanic crust and the history of seawater. In Stable Isotopes in High Temperature Geological Processes (ed. J. W. VALLEY, H. P. TAYLOR, JR., and J. R. O’NEIL); Rev. Mineral. 16, pp. 425-444. MUEHLENBACHSK. and CLAYTON R. N. ( 1972) Oxygen isotope studies of fresh and weathered submarine basalts. Canadian J. Earth Sci. 9, I72- 184. MUEHLENBACHSK., ANDERSONA. T., and SIGVALDASONG. E. ( 1974) Low “0 basahs from Iceland. Geochim. Cosmochim. Acta 38,577-588. NEAL C. R., TAYLORL. A., DAVIDSONJ. P., HOLDENP., HALLIDAY A. N., NIXON P. H., PACESJ. B., CLAYTONR. N., and MAYEDA T. K. ( 1990) Eclogites with oceanic and crustal mantle signatures from the Bellbank kimberlite, South Africa, Part 2: Sr, Nd, and 0 geochemistry. Earth Planet. Sci. Lett. 99, 362-379. ONGLEYJ. S., BASUA. R., and KYSERT. K. ( 1987) Oxygen isotopes in coexisting garnets, clinopyroxenes, and phlogopites of Roberts Victor eclogites: Implications for pctrogenesis and mantle metasomatism. Earth Planet. Sci. Left. 99, 362-319. O’REILLY S. Y. and GRIFFIN W. L. (1988) Mantle metasomatism beneath western Victoria, Australia: 1. Metasomatic processes in Sr diopside Iherzolites. Geochim. Cosmochim. Acta 52,433-447. PACESJ. 8. and BELLK. ( 1989) Nondepleted sub-continental mantle beneath the Superior Province of the Canadian Shield: Nd-Sr isotopic and trace-element evidence from Midcontinent Rift basalts. Geochim. Cosmochim. Acta 53,2023-2035. PEARSOND. G., DAVIESG. R., NIXON P. H., GREENWOODP. B., and MA~TEYD. P. ( 199 I ) Oxygen isotopic evidence for the origin of pyroxenites in the Beni Bousera peridotite massif, North Morocco: Derivation from subducted oceanic lithosphere. Earth Planet. Sci. Lett. 102, 289-301. PINEAUF., JAVOYM., HAWKINSJ. W., and CRAIGH. ( 1976) Oxygen

2302

J. A. Barrat et al.

isotope variations in marginal basins and ocean-ridge basalts. Earth Planet. Sci. Lett. ts, 299-307. RICHARD 0. ( 1979) Etude de la transition dorsale oceanique-rift tmerg6: Le Golfe de Tadjoura (Rkpublique de Djibouti): Approche &logique, g&chtonologique, et p&ologique. Thesis Univ. Orsay. RUEGG J. C. ( 1974) Structure profonde de la croilte tertestre au moyen de grands profils sismiques: Application a l%tudes des zones de distension (sud-est Afar et Massif Central fmncais). Thesis, Univ. Paris VI. SANJUANB., MICHARDG., and MICHARD A. ( 1990) Origine des substances dissoutes dans les eaux des sources thermales et des foraees de la r&on Asal-Ghoubbet ( R&ubliaue de Diibouti ) . J. “ Vol~anol.Geotierm. Res. 43, 333-352._ SCH.ILLINGJ. G. ( 1973) Afar mantle plume: Rare earth evidence. Nature 242,2-5. SCHILLINGJ. G., KINGSLEYR. H., HANAN B. B., and MCCULLY B. L. ( 1992) Nd-Sr-Pb isotopic variations along the Gulf of Aden: Evidence for mantle plume-continental lithosphere interaction. J. Geophys. Res. 97, 10,927- 10,966. SIG~~~ARSSO~.' Q..COND~MINESM., and FOURCADES. ( 1992) Mantle and crustal contribution in the genesis of recent basalts from olfrift zones in Iceland: Constraints from Th, Sr, and 0 isotopes. Earth Planet. Sci. Left. 110,149-162. STIELTJESL. ( 1973) L’axe tectono-volcanique d’Asal. Thesis, Univ. orsay. STIELTJESL., JORONJ. L., TREUIL M., and VARETJ. ( 1973) Le rift d’Asa1, segment de dorsale emerge: Discussion petrologique et geochimique. Bull. Sot. Geol. France 7,85 l-862. SUN S. S. and MCDONOUGH W. F. ( 1989) Chemical and isotopic systematics of oceanic basal& Implications for mantle composition and processes. In Magmatism in the Ocean Basins (ed. A. D. SAUNDERSand M. J. NORRY ); GSA Spec. Publ. 42,3 13-345.

SUN C. O., WILLIAMSR. J., and SUN S. S. ( 1974) Distribution coefficients of Eu and Sr for plagioclasoliquid and clinopyroxene liquid equilibria in oceanic ridge basalt: An experimental study. Geochim. Cosmochim. Acta 38, I4 1S- 1433. TAYLORH. P., JR. ( 1980) Stable isotope studies of spreading centers and their bearing on the origin of granophyres and plagiogranites. Colloque Intern. CNRS No. 272; Associations Majques Ultramafiques dartsles Orogtkes, pp. 149- 165. TAYLORH. P., JR. ( 1987) Comparison of hydrothermal systems in layered gabbros and granites, and the origin of low- “0 magmas. In Magmatic Processes, Physicochemical Principles (ed. B. 0. MYSEN);GSA Spec. Publ. 1,337-357. TAYLOR H. P.. JR.. and SHEPPARDS. M. F. ( 1986) Inneous rocks: 1. Processes of isotopic fractionation and isotope’ s$tematics. In Stable Isotopes in High- Temperature Geological Processes (ed. J. W. VALLEY,H. P. TAYLOR,JR., and J. R. O’NEIL); Rev. Mineral. 16, pp. 227-272. TREUIL M. and VARET J. (1973) Crit&es petrologiques, geochimiques, et structuraux de la genese et de la differentiation des magmas basaltiques: Exemple de 1’Afar. Thesis, Univ. Orl&ms and orsay. VIDAL PH., DENIEL C., VELLUTINIP. J., PIGUET P., COULON C., VINCENTJ., and AUDIN J. ( 199 1) Changes of mantle sources in the course of rift evolution: The Afar Case. Geophys. Res. Lett. 18,1913-1916. WOODHEADJ. D., HARMONR. S., and FRASERD. G. ( 1987) 0, S, Sr, and Pb isotope variations in volcanic rocks from the Northern Mariana Islands: Implications for crustal recycling in intraoceanic arcs. Earth Planet. Sci. Lett. 83, 39-52. YODER H. S. and TILLEY C. E. ( 1962) Origin of basaltic magmas: An experimental study of natural and synthetic rock systems. J. Petrol. 3, 342-532.