22. TRACE ELEMENT GEOCHEMISTRY AND PETROGENESIS OF BASALTS FROM DEEP SEADRILLING PROJECT SITES 556-559 AND 561-5641

J. Hertogen, Universiteit Leuven, Fysico-chemische geologieTh. Sachtleben and H.-U. Schmincke, Ruhruniversitàt Bochum, Mineralogisches Institut

andG. A. Jenner, Max-Planck-Institut für Chemie, Mainz2

ABSTRACT

Forty-three samples from DSDP Holes 556-559 and 561-564 were analyzed for rare earth elements (REE), Sc, Cr,Co, Hf, Ta, and Th by instrumental neutron activation analysis.

The recovered basalts range from those depleted in light REE (LREE) to those enriched in LREE. The two types ofbasalts occur together in Holes 558 and 561. The depleted basalts have remarkably constant La/Yb, La/Sm, and La/Tiratios and apparently derive from a large, homogeneous, mantle source underneath a segment (1200 km long) of theMid-Atlantic Ridge. The almost twofold variation in the concentrations of incompatible trace elements in the depletedbasalts is primarily due to different degrees of batch partial melting.

The variation of highly to moderately incompatible elements in the Leg 82 enriched basalts can be successfully ex-plained in terms of source mixing between depleted mantle sources and alkaline or nephelinitic magmas similar toAzores Islands magmas. However, the correlation of LREE enrichment with distance from the Azores Triple Junction istenuous at best, and the enriched alkaline component is probably not directly related to the Azores volcanism.

INTRODUCTION

The geochemical studies of mid-ocean ridge basalts(MORB) from different regions in the northern AtlanticOcean have revealed regionally marked differences inisotopic and trace element composition that are thoughtto be related to heterogeneities in the mantle source re-gions. One of these geochemically anomalous regions inthe northern Atlantic is the Azores Triple Junction. Var-ious hypotheses about the nature and origin of suchanomalies have been developed based on one of two as-sumptions: either a mantle plume, enriched with largeion lithophile (LIL) elements rising beneath the Azores(White and Schilling, 1978), or the presence of separatemagma reservoirs with different chemistries resulting froma complex melting history in the upper mantle (Wood,1979).

One objective of Leg 82 was to relate regional varia-tions of basalt geochemistry to geodynamic processes inan anomalous segment of the Mid-Atlantic Ridge, theAzores Triple Junction. Glomar Challenger drilled atnine sites southwest of the Azores (Fig. 1) during Leg82. The recovered basalts span a wide range in age andtrace element composition, varying from 14 to 37 Maand from samples enriched in LREE to samples depletedin LREE. In this paper we present rare earth element(REE) and other trace element (Sc, Cr, Co, Hf, Ta, Th)data from 43 selected samples. The trace elements have

Bougault, H., Cande, S. C , et al., Init. Repts. DSDP, 82: Washington (U.S. Govt.Printing Office).

2 Addresses: (Hertogen) Universiteit Leuven, Fysico-chemische geologie, B-3O3O Leuven,Belgium; (Sachtleben and Schmincke) Ruhruniversitàt Bochum, Mineralogisches Institut, D-4630 Bochum, West Germany; (Jenner, present address) Department of Earth Sciences, Me-morial University of Newfoundland, St. John's, Newfoundland A1B 3X5, Canada.

been determined by instrumental neutron activation anal-ysis (INAA). Two companion papers (Sachtleben et al.and Jenner et al., this volume) deal with the petrology,chemistry (major elements and Cr, Ni, Rb, Sr, Y, Zr,Nb, Ba), and Nd and Sr isotopic composition of thesame samples.

ANALYTICAL TECHNIQUES

Samples for this study were selected (1) to be representative of thechemical units (magma groups) as defined by the shipboard party andby onshore analyses (Sachtleben et al., this volume), and (2) to be asfresh as possible. These criteria were checked by major element analy-ses taking into account only samples with low H2O and/or CO2 when-ever possible. In addition, thin sections were examined for alteration.

Samples were coarsely crushed in an iron mortar and ground in anagate mill to 60 µm. Aliquots containing 800 mg of sample were irra-diated for 7 hours in a thermal neutron flux of about 2 × 1012 neu-trons cm"2 sec""1 in the Thetis reactor at Gent University, Belgium.The gamma-ray intensities were measured at Leuven University with alarge volume Ge(Li)-detector and a hyperpure Ge "Low Energy Pho-ton Detector" at intervals of 7 and 20-30 days after irradiation. Con-centrations were determined relative to a secondary "in-house" alkalibasalt standard, which has been repeatedly calibrated against USGSBCR-1 reference rock. For details of the analytical technique, see Her-togen and Gijbels (1971). Analytical precision for the different ele-ments can be judged from the errors (2 × standard error of the mean)reported for Sample 564-9-1, 1-4 cm (Table 1). The errors were calcu-lated from counting statistics and the observed dispersion of resultsfrom different countings and/or different gamma rays. The precisionis generally better than 2% for Sc, Cr, Co, Sm, and Eu; 6% for La,Tb, and Yb; and 12% for Ce and Lu. The errors for Nd, Ta, and Thrapidly increase with decreasing concentration and may be as high as15-20% for the most LIL-depleted samples.

RESULTS

The trace element data obtained by INAA are pre-sented in Table 1. Chemical unit assignments listed inTable 1 are taken from Sachtleben et al. (this volume).

449

J. HERTOGEN, TH. SACHTLEBEN, H.-U. SCHMINCKE, G. A. JENNER

Table 1. Trace element concentrations (ppm) in Leg 82 basalts.

Hole-Core-Section(interval in cm)

556-2-1, 83-85556-2-5, 5-7556-3-2, 19-21556-6-5, 68-70556-7-1, 85-87556-8-2, 27-29556-12-3, 78-80556-16-2, 131-133557-1-1, 105-108558-27-3, 69-72558-27-3, 116-118558-28-3, 118-120558-29-1, 63-65558-29-4, 83-85558-31-1, 79-82558-32-1, 123-125558-33-2, 71-73558-34-1, 30-32558-35-3, 110-113558-35-4, 106-108558-38-1, 34-36558-38-2, 61-63558-40-1, 92-94558-40-2, 7-9559-1-1, 25-29559-1-2, 52-56559-8-1, 127-129561-1-1, 33-38561-2-1, 143-146561-2-2, 138-141561-3-2, 47-53562-1-1, 110-112562-4-1, 91-93562-6-3, 120-123562-6-4, 1-5562-8-2, 6-9563-23-1, 30-34563-25-1, 19-22563-25-2, 21-24564-1-2, 21-25564-1-2, 38-40564-6-4, 71-74564-9-1, 1-4

Chemicalunit

IIIIIIIIIIIIIIIIV

IIIIIIIIIIIIIIII

IVAIVAIVBIVB

VIVAVIVIIIIIIIIIII:

iiil

IIiIIIIII

Sc

38.836.038.837.638.740.838.440.633.932.332.338.535.233.934.835.635.735.537.336.335.237.136.536.039.037.939.034.843.641.439.139.639.141.036.438.544.741.143.943.742.439.940.0

±0.5a

Cr

3062432612753642813944733

478488402342321335337370366368362446342462500271260265391239232232228228244247254334321257333259227219± 5

Co

51.340.344.847.144.643.642.842.948.948.150.140.244.240.343.547.737.944.150.242.849.744.050.241.545.442.547.450.547.866.542.944.344.448.542.443.544.645.252.046.3

50.742.5

±0.8

La

1.82.62.82.41.51.451.72.3

19.45.75.92.79.89.49.7

10.08.98.76.86.64.87.85.0

13.59.28.59.4

12.32.72.62.42.62.73.12.52.42.01.853.61.93.33.84.3

±0.2

Ce

6.19.29.68.35.04.85.97.9

47.114.414.28.4

22.621.323.022.920.220.516.615.811.818.511.827.620.519.721.525.9

9.08.37.89.09.3

10.06.87.86.45.7

11.05.6

10.310.611.2

±0.7

Nd

5.98.89.48.15.25.05.86.9

28.69.39.07.3

13.413.213.213.811.612.210.89.97.1

11.56.7

17.810.811.812.313.58.98.37.39.38.19.56.77.75.95.39.55.69.98.79.5

±0.8

Sm

2.283.233.443.022.091.882.113.017.132.722.692.493.373.283.363.423.073.032.892.792.133.052.183.693.453.363.563.43.122.952.773.283.293.412.552.702.202.063.302.143.133.133.28

±0.06

Eu

0.891.181.281.120.880.790.871.182.421.021.010.941.191.15

.21

.22

.12

.09

.051.010.811.100.83

.33

.26

.20

.28

.2

.221.131.061.251.231.310.981.030.910.831.260.891.201.171.20

±0.03

Gd

3.2

4.74.03.12.7

7.83.53.43.7

3.7

3.63.73.33.4

2.63.62.8

4.53.74.34.03.94.84.55.03.7

3.0

3.1

4.24.3

±0.4

Tb

0.610.850.940.800.570.510.570.801.290.630.610.650.720.640.690.680.640.600.650.600.510.650.500.620.770.760.810.690.880.790.760.890.870.930.700.730.650.570.880.610.840.810.84

±0.03

Yb

2.683.533.743.252.442.202.353.223.52.342.252.632.452.352.502.452.152.182.482.352.022.532.061.853.002.933.062.513.723.483.293.613.583.692.793.102.782.513.742.643.403.403.53

±0.15

Lu

0.450.630.690.570.420.390.420.590.620.420.410.470.420.410.470.430.430.380.470.410.360.450.390.310.500.580.590.460.690.620.580.640.660.700.500.560.540.430.700.480.610.590.64

±0.07

Hf

1.562.352.552.14

.42

.29

.45'.131.96.84.83.70

2.192.102.172.241.971.941.891.771.322.071.402.182.422.372.522.162.112.001.852.302.332.471.721.821.511.412.341.472.252.252.25

±0.08

l a

0.090.130.140.110.070.070.070.131.950.530.530.150.960.900.970.980.860.860.620.590.460.710.471.320.960.910.991.290.170.160.140.140.170.210.160.160.120.110.250.11

0.280.32

±0.02

Th

0.150.150.200.220.20.15

<0.150.41.810.630.650.311.121.001.081.110.970.940.650.690.520.830.551.451.090.941.071.520.340.260.19

<0.150.200.250.20.190.170.240.290.210.60.350.38

±0.07

a 2 times standard error of the mean.

Compatible Trace Elements (Sc, Cr, Co)

Sc, Cr, and Co concentrations are within the rangestypical of MORB from the North Atlantic (Bougault etal., 1979; Wood et al., 1979).

Sc and (to a lesser extent) Co contents are rather con-stant, which is in keeping with the scarcity of highlyfractionated basalts in the Leg 82 holes. Neither system-atic hole-to-hole differences nor clear-cut correlationswith magma groups have been observed for Sc and Co.

Cr contents correlate with MgO values; for more de-tails, see Sachtleben et al. (this volume), who report Crdata for a larger number of samples.

Incompatible Trace Elements (REE, Hf, Ta, Th)

Ta and Th strongly correlate with La, whereas Hfcorrelates with Sm. The following discussion is, there-fore, largely restricted to the REE data.

Hole 556. The eight analyzed samples represent fourmagma groups and are all "normal" LREE-depleted ba-salts (N-MORBs) (Fig. 1). Absolute REE abundancesspan a rather large range, but the relative patterns aremarkedly similar (i.e., virtually constant La/Yb ratios;see Fig. 8 for details). Magma Groups II and V have

consistently higher TiO2 contents (Fig. 9) and lower Mg-numbers (58 + 2 versus approximately 67 ± 2; Sachtle-ben et al., this volume) than Groups I and III. The samedichotomy shows up in the absolute REE abundances(Fig. 1).

Hole 557. Basalts from this site are unique amongLeg 82 samples; they are highly fractionated (Mg-valueof 44) and rich in Fe and Ti (15% FeOtot, 3.6% TiO2)(Sachtleben et al., this volume). The single sample ana-lyzed for trace elements (557-1-1, 105-108 cm) is strong-ly enriched in LREE (Fig. 1), Hf, Ta, and Th.

Basalts of similar major and trace element composi-tions are known from Hole 407 in the Reykjanes Ridgeat 63°N (Wood et al., 1979). However, the strongly LREE-enriched basalts from Holes 410 and 410A (Mid-Atlan-tic Ridge, 45 °N) have a distinctly different major ele-ment chemistry; most are much less fractionated (Mg-values up to 66), and all have lower TiO2 (1.4-1.5%)and FeOtot (7-9%) contents (Wood et al., 1979). TheHole 557 basalts also differ from the Fe-rich and Ti-richbasalts from the Galapagos spreading center, which aretypically depleted in LREE (Clague et al., 1981).

Hole 558. The 15 analyzed samples represent 6 mag-ma groups, and all but one (558-28-3, 118-120 cm; Group

450

TRACE ELEMENT GEOCHEMISTRY, PETROGENESIS OF BASALT

I I I I ILa Ce Nd Sm Eu Gd Tb Yb Lu

Figure 1. Rare earth element (REE) patterns of basalts normalized tochondrites from Holes 556 (8 samples), 557 (1 sample), and 558 (15samples). Samples from different magma groups (chemical units inRoman numerals) generally have different REE contents. The sam-ples from Hole 558 show the greatest diversity of REE patterns,ranging from LREE-enriched to LREE-depleted. The hatched fieldfor Hole 558 covers the closely spaced patterns of 11 E-MORBsamples from Groups I, III, and IV. The irregular REE distribu-tion of Sample 558-40-2, 7-9 cm could be due to alteration.

II) are LREE-enriched basalts (Fig. 1). For the LREE-enriched basalts, a correlation between magma groupand LREE enrichment is observed: V < I < IVA <IVB < III. Group VI is apparently a special case, be-cause the two samples show grossly different REE pat-terns (558-40-1, 92-94 cm and 558-40-2, 7-9 cm). TheGroup VI samples belong lithologically to a breccia unit,but do not seem to form a single chemical unit (Sachtle-ben et al., this volume). The irregular REE pattern of558-40-2, 7-9 cm could be due to alteration; the CO2content (0.94%) of this sample is among the highestmeasured in our sample suite.

The LREE-depleted basalt (558-28-3, 118-120 cm) al-so has the high La/Ta and Hf/Ta ratios typical ofN-MORB (Fig. 5). However, this sample is unusualamong the Leg 82 N-MORBs, because its 143Nd/144Ndisotopic ratio is virtually indistinguishable from that of

the E-MORBs (enriched MORB) from Hole 558 (Jenneret al., this volume).

Hole 559. The basalts from this hole form a singlemagma group. The three analyzed samples are LREEenriched (Fig. 2) and show only limited variations oftrace element abundances.

Hole 561. The two magma groups have distinctly dif-ferent trace element (Fig. 2) and Nd isotopic (Jenner etal., this volume) characteristics. The occurrence of bothE-MORBs and N-MORBs in a single hole underscoresthe small scale of mantle heterogeneities or the localizednature of mixing processes.

Hole 562. The five analyzed samples are all LREEdepleted (Fig. 3). Basalts from magma Group I have con-sistently higher REE abundances than those from mag-ma Group II.

Hole 563. The two magma groups are LREE depletedbut are clearly separated by absolute REE abundances,which are much lower in Group I than in Group II (Fig. 3).Note the similarities between Groups I and II fromHoles 563 and 564.

Hole 564. The sample from magma Group I (564-1-2,21-25 cm) is a typical, depleted N-MORB (Fig. 3). TheGroup II basalts from Holes 563 and 564, though LREEdepleted, have a more transitional character; the La/Ybratios are higher (Fig. 8) and the Hf/Ta and La/Ta ratios(Fig. 5) are lower than in depleted basalts from the otherLeg 82 holes. They are also clearly separated from the

La Ce Nd Sm Eu Gd Tb Yb Lu

Figure 2. Rare earth element patterns of basalts normalized to chon-drites from Holes 559 and 561. The Hole 559 basalts form a single,very homogeneous, magma group. Enriched (Group I) and depleted(Group II) basalts occur together in Hole 561.

451

J. HERTOGEN, TH. SACHTLEBEN, H.-U. SCHMINCKE, G. A. JENNER

20 -

10 —

-\ r-

Hole 564

La Ce NdSm Eu Gd Tb

Yb Lu

Figure 3. Rare earth element (REE) patterns of basalts normalized tochondrites from Holes 562, 563, and 564. Note the similarity ofREE patterns of Group I and II samples from Holes 563 and 564.The Group II basalts from these two holes have somewhat higherLa/Sm and La/Yb (see also Figs. 6 and 8) ratios than the LREE-depleted MORBs from the other Leg 82 holes.

common depleted basalts in the TiO2 versus La varia-tion diagram (Fig. 9).

DISCUSSION

Trace Element Variations in the Leg 82 N-MORBs

For the present discussion N-MORBs are defined asthose samples that have a La/Yb ratio (normalized tochondrites) less than 0.5 and a La/Ta ratio greater than15 (Figs. 5 and 8). The transitional Group II basaltsfrom Holes 563 and 564 (Fig. 3) will be discussed to-gether with the E-MORBs.

The Leg 82 N-MORBs have remarkably constant La/Yb, La/Sm, and La/Ti ratios, in spite of the almosttwofold variation of absolute REE abundances (Figs. 6,8, 9). This large variation of REE and other incompati-ble elements may be due to either substantial (40-50%)crystal fractionation or different degrees of partial melt-ing. To determine which of these two processes is themost important or the most plausible, we will consider

the variation of some critical major and trace elementsin the Hole 556 N-MORBs that encompass the wholerange of chemical variation in the Leg 82 N-MORBs(see also Sachtleben et al., this volume). Data for somekey elements have been summarized in Table 2 to facili-tate the discussion.

The chemical variation within the Hole 556 sequencecannot be explained by high-pressure clinopyroxene frac-tionation (above 10 kbar; Bender et al., 1978; Fujii andBougault, 1983). This would lead to a noticeable in-crease in the La/Yb and La/Ti ratios and a significantdecrease of Sc content with increasing La content. Thesetrends are not observed (Figs. 8 and 9; Table 2). Low-pressure fractionation of olivine alone can also be dis-missed because it would rapidly deplete the magmas inMgO (Fig. 4) and Ni. The depletion in MgO and Nicould be largely offset by simultaneous plagioclase crys-tallization. However, extensive plagioclase fractionationis definitely ruled out by the absence of large negativeEu-anomalies and by the increase of the Al2O3/CaO ra-tio in the Hole 556 N-MORBs (Table 2). Fractionationof plagioclase with Al2O3/CaO around 1.8—as observedin the cores of phenocrysts—would produce an oppositeAl2O3/CaO variation trend. The Ca/Al fractionation byplagioclase could in turn be offset by cocrystallizationof clinopyroxene. Least-squares mass balance calcula-tions, using high-pressure (7-10 kbar) olivine, clinopy-roxene, and plagioclase compositions from Fujii andBougault (1983), suggest that Sample 556-3-2, 19-21 cmcould be related to Sample 556-8-2, 27-29 cm by 21%clinopyroxene plus 12% plagioclase fractionation. Thisresult conflicts with the constancy of Sc concentrationsand La/Yb and La/Ti ratios. Moreover, clinopyroxeneplus plagioclase without olivine does not appear to be aliquidus phase assemblage within the 1 atm to 15 kbarpressure range (Bender et al., 1978; Fujii and Bougault,1983). Therefore, if the major and trace element varia-tion in the Leg 82 N-MORBs are to be explained bycrystal fractionation, complex mechanisms have to beinvoked, such as the fractionation of a continuously re-plenished magma chamber (O'Hara, 1977). It is, how-ever, beyond the scope of this paper to test the plausibil-ity of this class of models.

Different degrees of batch partial melting of a lherzo-lite mantle source readily account for the twofold varia-tion of incompatible elements and for the constant La/Yb and La/Ti ratios of the Leg 82 N-MORBs. A cal-culated partial melting path has been superimposed onFigure 8 to illustrate this point. The melting model as-sumed a depleted source consisting of 65% olivine (OL),25% orthopyroxene (OPX), 10% clinopyroxene (CPX)with 0.26 ppm La (0.76 × chondrites), and 0.4 ppmYb (2 × chondrites). The melting proportions wereOL0OPXo.2CPX0.8 from 0 to 12.5% melting andOL02OPX08 above 12.5% melting. Crystal/liquid par-tition coefficients for OL, OPX, and CPX were 0.001,0.001, and 0.03 for La and 0.001, 0.05, and 0.3 for Yb.The outcome of calculations for a garnet-bearing man-tle source is virtually identical provided that the garnetis exhausted from the source at 8 to 10% melting. Thesystematic decrease of the La/Sr ratio with decreasing

452

TRACE ELEMENT GEOCHEMISTRY, PETROGENESIS OF BASALT

Table 2. Selected data for Hole 556 samples.

Measuredvalue

MgO (%)Mg-numberAl2O3 /CaONi (ppm)Sc (ppm)La (ppm)Sr (ppm)(La/Sr) × 104

8-227-29

8.0968.8

1.1113640.8

1.4598

148

7-185-87

6.7464.6

1.0312838.7

1.5106142

Core-Section (interval in cm)

12-378-80

7.9967.2

1.2311238.4

1.795

179

2-183-85

7.8067.2

1.2312138.8

1.8114158

16-2131-133

6.5756.4

1.375540.6

2.3120192

6-568-70

6.9461.6

1.3410937.62.4

103233

2-55-7

6.1856.4

1.347936.02.6

108241

3-219-21

6.1057.6

1.377238.82.8

108260

Note: From this work and Sachtleben et al. (unpublished data).

8

MgO (%)

Figure 4. Ytterbium (YbN) content normalized to chondrites versusMgO content for Leg 82 basalts. In order to produce the twofoldincrease of heavy rare earth element (REE) abundances in theN-MORBs (e.g., from Hole 556), low-pressure olivine (OL) + pla-gioclase (PL) fractionation should be dominated by plagioclase, asshown by the two fractionation vectors. Because extensive plagio-clase fractionation is ruled out by the lack of large negative Eu-anomalies and the variation of the Al/Ca ratios, the large varia-tion of REE abundances appears to be the result of different de-grees of partial melting (see also Fig. 8).

La content (Table 2) and the small negative Eu-anoma-lies in some N-MORBs (Figs. 1-3) could imply that pla-gioclase was a residual source mineral throughout mostof the melting range. This would be consistent with themodel proposed by Fujii and Bougault (1983).

However, all Leg 82 N-MORBs have MgO contents andMg-numbers that are too low to qualify the N-MORBsas primary or near-primary mantle melts (Rhodes andDungan, 1979; Frey et al., 1980). So, even if the twofoldvariation of incompatible trace elements was primarilyestablished by partial melting, the parental magmas ofall N-MORBs most likely went through a stage of oliv-ine fractionation before eruption (Stolper and Walker,1980). Figure 4 suggests that the N-MORBs could haveevolved from a parallel sequence of high-Mg magmas(> 10% MgO) by 5 to 15% olivine fractionation.

Although the constant La/Yb and La/Ti ratios of theLeg 82 N-MORBs require a very homogeneous mantlesource, the substantial variation of the La/Ta, Hf/Ta,and La/Hf ratios (Fig. 5) nevertheless suggests somesource heterogeneity or mixing. Indeed, it is generallyassumed that basalts inherit the Hf/Ta and La/Ta ratiosof their source region, because these incompatible ele-ments are not substantially fractionated from each otherby partial melting and fractional crystallization (Bou-gault et al., 1979; Wood et al., 1979; Langmuir et al.,1978).

The linear array of the Leg 82 N-MORBs in the Hf/Ta versus La/Ta diagram (Fig. 5) could be interpreted asa mixing line between N-MORB magmas (or sources)and E-MORB magmas or alkaline magmas similar toAzores alkali basalts. Mixing relations are further ex-plored in Figure 7. The mixing curve was calculated formixing between Sample 556-8-2, 27-29 cm and the alka-li basalt F33 from Faial Island, Azores Island (White etal., 1979). This particular Faial sample (42.2 ppm La,4.2 ppm Ta) has been chosen as being representative ofrelatively primitive (Mg-number > 65), enriched mantlemagmas. The La/Ta ratios of the N-MORBs (La/Ta >15) are very poorly correlated with La content, which isconsistent with the tenet that La/Ta and La/Hf ratiosare not systematically fractionated by moderate degreesof melting or crystallization. Figure 7 offers one expla-nation for the La/Ta variations in the Leg 82 N-MORBs,i.e., mixing of N-MORB magmas with less than 2% ofan enriched alkaline component. This small amount ofmixing will have a minor effect on absolute LREE con-tents but will not significantly alter La/Yb, La/Ti, andLa/Sm ratios (Figs. 6, 8, 9). Actually, the E-MORB mag-mas also qualify as enriched end-members, but the re-

453

J. HERTOGEN, TH. SACHTLEBEN, H.-U. SCHMINCKE, G. A. JENNER

25

La/Ta

Figure 5. Hf/Ta versus La/Ta for Leg 82 basalts. The linear data arraysuggests that the variation is due to mixing involving an enrichedalkaline end-member similar to Azores alkali basalt magmas (Whiteet al., 1979). See Figure 7 for details of the mixing model.

quired amount of enriched component would then befive to ten times larger. The discussion in the next sec-tion will show that interpretation of the La/Ta varia-tions in terms of magma mixing is not unique. The vari-able La/Ta ratios might equally well reflect source mix-ing (less than 0.1% enriched component) or local sourceheterogeneities that are too small to affect the ratios ofother incompatible elements.

Mixing Relations in the Leg 82 E-MORBs

Two models have been proposed to explain the occur-rence of E-MORBs along the Mid-Atlantic Ridge be-tween 30 and 40°N. The first one invokes binary mixingbetween "normal" depleted MORB (or their source re-gions) and Azores Mantle Plume magmas enriched inincompatible trace elements and radiogenic Sr, Nd, andPb isotopes (Schilling, 1975; White and Schilling, 1978).The second model postulates the existence of regionalchemical and isotopic heterogeneities in the suboceanmantle, whose origins might be unrelated to the Azoresvolcanism (O'Nions et al., 1978; Wood et al., 1979;Davies, 1981). Because of its conceptual simplicity, theAzores Mantle Plume mixing hypothesis can be testedquite rigorously. To do so we will investigate whether ornot the trace element variations in the E-MORBs areconsistent with magma or source mixing and whether or

not the Azores magma are plausible enriched end-mem-bers.

Before proceeding with the discussion of mixing rela-tions, we would like to call attention to two chemical as-pects of the Leg 82 E-MORBs. Firstly, the MgO (Fig. 4),Cr (Table 1), and Ni (Sachtleben et al,. this volume) con-tents are generally higher than in the Leg 82 N-MORBs.This precludes crystal fractionation being of any impor-tance in establishing the trace element characteristics ofthe Leg 82 E-MORBs. (The fractionated basalts fromHole 557 are an exception). Secondly, there is no corre-lation between LREE enrichment and HREE abundancein the Leg 82 E-MORBs. Crossing REE patterns, as ob-served in the Hole 558 E-MORBs, are the hallmark ofthe "dynamic melting" model proposed by Langmuir etal. (1977) for basalts from the FAMOUS area. However,no indication for dynamic melting could be found in theREE systematics of the Leg 82 N-MORBs. Becausethere is no obvious reason why dynamic melting shouldbe restricted to particular areas along a ridge, or to en-riched sources for that matter, we cannot accept dy-namic melting as the preferred model.

Two mixing models will be explored: (1) mixing be-tween N-MORB magmas and alkaline magmas; and (2)mixing of depleted N-MORB mantle sources with alka-line magmas. For the sake of clarity, the alkali basaltF33 from Faial Island, Azores, (White et al., 1979) hasbeen chosen to represent alkaline magmas for both mod-els.

We will first consider the mixing relations in the chon-drite-normalized La/Yb versus La/Sm diagram (Fig. 6).With the exception of the basalts from Holes 557 and561, the Leg 82 E-MORBs fit reasonably well to themixing curve between N-MORB magmas and the alkalibasalt F33. The magma mixing model predicts 8 to 4O°7oalkaline component for the Hole 558 E-MORBs, 13 to14% for the Hole 559 E-MORBs, and less than 3% forthe transitional MORBs from Holes 562, 563, and 564.If we consider the magma mixing curve, crystal frac-tionation after magma mixing does not appear to be aplausible explanation for the deviation of the E-MORBsfrom Holes 557 and 561, because crystallization of mod-erate amounts of olivine, plagioclase, and clinopyrox-ene does not significantly change the La/Sm and La/Ybratios. So, if the REE geochemistry of these samples isdue to mixing, the enriched end-members must havebeen different. The alignment of the Leg 82 E-MORBsalong a mixing line can also be explained in terms ofsource mixing. The heavy arrows in Figure 6 indicate thecalculated La/Sm and La/Yb ratios of magmas gener-ated by 20% partial melting of N-MORB sources "con-taminated" with 1.5, 3, and 5% F33 alkaline compo-nent. The following assumptions were made for Sm: aconcentration of 0.32 ppm (1.5 × chondrites) in theN-MORB source and partition coefficients of 0.001,0.01, and 0.15 for olivine, orthopyroxene, and clinopy-roxene. The other input parameters of the melting modelwere the same as used in the preceding section for mod-eling the La/Yb variation in the N-MORBs. Results ofthis source mixing/partial melting model will be dis-cussed more in detail in subsequent paragraphs.

454

TRACE ELEMENT GEOCHEMISTRY, PETROGENESIS OF BASALT

12 -

1.5(La/Sm)N

Figure 6. La/Yb versus La/Sm for Leg 82 basalts; both ratios are nor-malized to chondrites. The N-MORBs fall within a very narrowfield (hatched square). The position of the E-MORBs (except forHoles 557 and 561) is consistent with either magma or source mix-ing. The mixing curve is calculated for magma mixing betweenSample 556-8-2, 27-29 cm and alkali basalt F33 from Faial Island,Azores (White et al., 1979); numbers along the mixing curve referto percentage of F33 component. The three heavy arrows indicatethe calculated position of magmas derived by 20% partial melting(PM) of N-MORB sources mixed with 1.5,3, and 5% of F33 com-ponent.

Plots involving ratios of incompatible and slightlycompatible elements, such as Figures 5 and 6, do notdistinguish between magma or source mixing for theLeg 82 E-MORBs. It might, therefore, be instructive toconsider mixing plots involving absolute concentrations,such as Figures 7-9.

The percentages of magma mixing indicated by theLa versus La/Ta plot (Fig. 7) are not inconsistent withthe amounts predicted by Figure 6, if allowance is madefor minor variations in the composition of the depletedand enriched end-members and for changes in the Lacontents by moderate degrees of crystal fractionation af-ter magma mixing. However, when the E-MORBs fromthe FAMOUS area (Bougault et al., 1979) and Hole 332B(Bougault, 1980) are considered, it appears that the chem-istry of the E-MORBs has been controlled by sourcemixing rather than magma mixing. Indeed, the majorityof the Hole 332B and FAMOUS MORBs plot well offmagma mixing curves between common N-MORB andalkaline magmas. The near-vertical array of the Hole

La/Ta

Figure 7. La versus La/Ta for Leg 82 basalts. As the La/Ta ratio is notchanged by fractional crystallization (FC) or partial melting (PM),the variation of the La/Ta ratio must be ascribed to mixing orsource heterogeneity. End-members of the mixing curve are thesame as in Figure 6. The moderately enriched MORBs from Hole332B (Bougault, 1980) and the FAMOUS area (Bougault et al.,1979) fall well off magma mixing curves between Azores alkalinemagmas and N-MORBs. This suggests that the trace element chem-istry of the E-MORBs is controlled by source mixing rather thanby magma mixing.

332B, FAMOUS, and Leg 82 E-MORBs in Figure 7 ismore readily explained by different degrees of partialmelting of N-MORB sources contaminated with > 1.5%alkaline magma.

To further explore mixing relations, magma mixingcurves have been superimposed on the Yb versus La/Yband TiO2 versus La plots (Figs. 8 and 9). It is immedi-ately apparent that the REE systematics and La/Ti vari-ation of the Leg 82 E-MORBs are not consistent withmagma mixing involving a unique depleted end-mem-ber. The whole array of N-MORB magmas must be con-sidered as depleted end-members, and the magma mix-ing curves shown in Figures 8 and 9 are examples from awhole family of mixing curves converging towards thecomposition of basalt F33. Here again, source mixingoffers an explanation that is at least as viable. The dashed,near-vertical lines in Figures 8 and 9 are batch partialmelting paths calculated for depleted N-MORB sourcescontaminated with 1.5% (line A) and 3% (line B) F33alkaline component. The numbers along the meltingcurves refer to the percentage of batch melting. (The melt-

455

J. HERTOGEN, TH. SACHTLEBEN, H.-U. SCHMINCKE, G. A. JENNER

(La/Yb)N

Figure 8. Yb versus La/Yb ratio normalized to chondrites for Leg 82basalts. The twofold variation of rare earth element (REE) contentin the Leg 82 N-MORBs (La/Yb < 0.5) is primarily due to differ-ent degrees (8 to 18%) of partial melting (PM). The dashed curvepassing through the N-MORBs is a calculated batch partial melt-ing path; see text for input parameters of the melting model. TheREE systematics of the Leg 82 E-MORBs can be interpreted interms of magma or source mixing. The full lines are two examplesfrom a whole family of magma mixing curves between N-MORBmagmas and Azores alkali basalt F33; numbers along mixingcurves refer to percentage of F33 component. The dashed linesmarked A and B are calculated partial melting paths for N-MORBmantle sources mixed with 1.5% (line A) and 3% (line B) F33magma; numbers refer to percentage of partial melting.

ing paths from Figure 9 are based on the reasonable as-sumptions that the La/Ti ratio of the N-MORB sourcewas identical to the La/Ti ratio of the N-MORBs andthat Ti behaved as an incompatible element after exhaus-tion of clinopyroxene from the mantle sources.) Ouranalysis of the data shows that the geochemistry of thetransitional and enriched Leg 82 MORBs is consistentwith >0.5 to 5°Io admixture of alkaline component.

Whereas the trace element characteristics of the Leg82 E-MORBs can be used to promote both the magmaand source mixing models, there is some evidence to de-finitely favor the source mixing model:

1. The Nd isotopic data are not easily reconciled withmagma mixing, even if allowance is made for variabilityin the isotopic and trace element composition of theend-members (Jenner et al., this volume).

La (ppm)

Figure 9. TiO2 versus La for Leg 82 basalts. Note the constant La/Tiratio of the N-MORBs. The full lines marked F33 are magma mix-ing curves, the dashed lines marked A and B are partial meltingpaths; see Figure 8 caption for further explanations.

2. Magma mixing does not appear to be a viable modelfor the moderately enriched FAMOUS and Hole 332BMORBs (Fig. 7), and there is no obvious reason to in-voke special circumstances for either the Leg 82 or theFAMOUS and Hole 332B basalts.

3. The LREE-depleted basalt from Hole 558 (Sam-ple 558-28-3, 118-120 cm) might provide an importantclue. Its 143Nd/144Nd ratio is within the range of E-MORBs (Jenner et al., this volume), although source ormagma mixing calculations point to a much smaller ad-mixture of enriched component. This sample apparentlyderives from an enriched source that lost a small frac-tion of melt before the main melting event. In contrast,it is not so easy to imagine a simple process that can de-plete a melt in incompatible trace elements.

We are still left with the question of whether the sourcemixing process is directly related to the Azores MantlePlume magmatism. The Azores magmas indeed appearto be suitable enriched end-members, but any alkalineor nephelinitic melt generated in the subocean mantle isprobably quite similar to the Azores magmas with re-spect to trace element geochemistry. It seems reasonableto attribute the occurrence of E-MORBs in Hole 557 toAzores plume input, but the correlation of light REEenrichment with distance from the Azores Triple Junc-tion is tenuous at best for the other Leg 82 holes. This

456

TRACE ELEMENT GEOCHEMISTRY, PETROGENESIS OF BASALT

observation favors alternative models, such as episod-ic source mixing with "fracture zone-derived" enrichedmagmas (Flower, 1981).

Finally, there is the problem of recent mixing versuslong-lived mantle heterogeneities (0'Nions et al., 1978;Wood et al., 1979; Davies, 1981). Although trace ele-ment concentrations do not provide information on thetiming of the enrichment event, the data for the Leg82 basalts may indirectly have a bearing on this ques-tion. Indeed, N-MORBs with amazingly constant La/Sm (Fig. 6), La/Yb (Fig. 8), and La/Ti (Fig. 9) ratioshave been derived from a large, homogeneous mantlesource over a distance of 1200 km along the Mid-At-lantic Ridge during a period of approximately 20 Ma(37-15 Ma). These data seem to be at variance with therecent emphasis on small-scale mantle heterogeneity inthe area under study.

CONCLUSIONSThe almost twofold variation of absolute concen-

trations of incompatible trace elements in the Leg 82N-MORBs is primarily due to different degrees (8 to18%) of batch partial melting. The Leg 82 N-MORBshave MgO and Ni contents that are too low to qualifythe MORBs as primary mantle melts; they most likelyevolved from a sequence of more primitive magmas by 5to 15% olivine fractionation before eruption. The re-markably constant La/Yb, La/Sm, and La/Ti ratios ofthe Leg 82 N-MORBs apparently require a large, homo-geneous mantle source underneath a segment (1200 kmlong) of the Mid-Atlantic Ridge.

The variation of highly to moderately incompatibleelements (REE, HF, Ta, Ti) in the Leg 82 E-MORBsis consistent with magma or source mixing between N-MORB magmas or sources and an enriched componentsimilar to Azores alkali basalts. However, the combinedtrace element and Nd isotopic data (Jenner et al., thisvolume) strongly favor source mixing (>0.5 to 5% en-riched component). The association of depleted MORBswith transitional and enriched MORBs in Holes 561,563, and 564 underscores the episodic and localized na-ture of mixing processes.

The absence of a regular spatial distribution of Leg82 E-MORBs as a function of distance from the AzoresTriple Junction casts serious doubts on the validity ofAzores Mantle Plume mixing models advocated by Schil-ling (1975) and White and Schilling (1978). The enrichedalkaline magmas involved in the source mixing are prob-ably not directly related to the Azores magmatism.

ACKNOWLEDGMENTS

This paper has greatly benefited from constructive reviews by J. W.Delano and A. J. Irving. We thank G. Van Den Eynde for technical as-sistance with the analyses. This work was supported in part by the Bel-gian Inter-University Institute for Nuclear Sciences (I.I.K.W.) and theDeutsche Forschungsgemeinschaft (Grants Schm 250-28, 29).

REFERENCES

Bender, J. F., Hodges, F. N., and Bence, A. E., 1978. Petrogenesis ofbasalts from the project FAMOUS area: experimental study from 0to 15 kbars. Earth Planet. Sci. Lett., 41:277-302.

Bougault, H., 1980. Contribution des elements de transition à la ge-nèse des basaltes océaniques [Ph.D. dissert.]. Université de ParisVII, Paris.

Bougault, H., Cambon, P., Corré, O., Joron, J. L., and Treuil, M.,1979. Evidence for variability of magmatic processes and uppermantle heterogeneity in the axial region of the Mid-Atlantic Ridgenear 22° and 36°N. Tectonophysics, 55:11-34.

Clague, D. A., Frey, F. A., Thompson, G., and Rindge, S., 1981. Mi-nor and trace element geochemistry of volcanic rocks dredged fromthe Galapagos Spreading Centre: role of crystal fractionation andmantle heterogeneity. J. Geophys. Res., 86:9469-9482.

Davies, G. F., 1981. The earth's neodymium budget and structure andevolution of the mantle. Nature, 290:208-213.

Flower, M. F. J., 1981. Binary mixing in ocean-ridge spreading seg-ments. Nature, 292:45-47.

Frey, F. A., Dickey, J. S., Thompson, G., Bryan, W. B., and Davies,H. L., 1980. Evidence for heterogeneous primary MORB and man-tle sources, NW Indian Ocean. Contrib. Mineral. Petrol., 74:387-402.

Fujii, T., and Bougault, H., 1983. Melting relations of a magnesianabyssal tholeiite and the origin of MORBs. Earth Planet. Sci. Lett.,62:283-295.

Hertogen, J., and Gijbels, R., 1971. Instrumental neutron activationanalysis of rocks with a Low-Energy Photon Detector. Anal. Chim.Ada, 56:61-82.

Langmuir, C. H., Bender, J. F., Bence, A. E., Hanson, G. N., andTaylor, S. R., 1977. Petrogenesis of basalts from the FAMOUSarea: Mid-Atlantic Ridge. Earth Planet. Sci. Lett., 36:133-156.

Langmuir, C. H., Vocke, R. D., Hanson, G. N., and Hart, S. R.,1978. A general mixing equation with applications to Icelandic ba-salts. Earth Planet. Sci. Lett., 37:380-392.

O'Hara, M. J., 1977. Geochemical evolution during fractional crystal-lisation of a periodically refilled magma chamber. Nature, 266:503-507.

O'Nions, R. K., Evensen, N. M., Hamilton, P. J., and Carter, S. R.,1978. Melting of the mantle past and present: isotope and trace el-ement evidence. Phil. Trans. R. Soc. London, Ser. A, 258:547-559.

Rhodes, J. M., and Dungan, M. A., 1979. The evolution of oceanfloor basalts. In Talwani, M., Harrison, C. G., and Hayes, D. E.(Eds.), Deep drilling results in the Atlantic Ocean: ocean crust:Washington (Am. Geophys. Union), Maurice Ewing Series, 2:262-272.

Schilling, J.-G., 1975. Azores mantle blob: rare earth evidence. EarthPlanet. Sci. Lett., 25:103-115.

Stolper, E., and Walker, D., 1980. Melt density and the average com-position of basalt. Contrib. Mineral. Petrol., 74:7-12.

White, W. M., and Schilling, J.-G., 1978. The nature and origin ofgeochemical variation in Mid-Atlantic Ridge basalts from the Cen-tral North Atlantic. Geochim. Cosmochim. Acta, 42:1501-1516.

White, W. M., Tapia, M. D. M., and Schilling, J.-G., 1979. The pe-trology and geochemistry of the Azores Islands. Contrib. Mineral.Petrol., 69:201-213.

Wood, D. A., 1979. A variably veined suboceanic upper mantle—ge-netic significance for mid-ocean ridge basalts from geochemicalevidence. Geology, 7:499-503.

Wood, D. A., Tarney, J., Varet, J., Saunders, A. D., Bougault, H., Jo-ron, J. L., Treuil, M., and Cann, J. R., 1979. Geochemistry of ba-salts drilled in the North Atlantic by IPOD Leg 49: implicationsfor mantle heterogeneity. Earth Planet. Sci. Lett., 42:77-97.

Date of Initial Receipt: 22 June 1983Date of Acceptance: 31 January 1984

457