26. petrology and geochemistry of pliocene … · 26. petrology and geochemistry of...

Taylor, B., Fujioka, K., et al., 1992Proceedings of the Ocean Drilling Program, Scientific Results, Vol. 126

26. PETROLOGY AND GEOCHEMISTRY OF PLIOCENE-PLEISTOCENE VOLCANIC ROCKSFROM THE IZU ARC, LEG 1261

James B. Gill,2 Conrad Seales,2 Patricia Thompson,2 Alfred G. Hochstaedter,2 and Charles Dunlap2

ABSTRACT

The igneous geochemistry of lavas and breccias from the basement of Sites 790 and 791, and pumice clasts from thePliocene-Pleistocene sedimentary section of Sites 788, 790, 791, and 793 were studied. Arc volcanism became silicic about1.5 m.y. before the inception of rifting in the Sumisu Rift at 2 Ma, but eruption of these silicic magmas reflects changes in stressregime, especially during the last 130,000 yr, rather than crustal anatexis. Arc magmas have had a larger proportion of slab-derivedcomponents since the inception of rifting than before, but are otherwise similar. Rift basalts and rhyolites are derived from adifferent source than are arc andesites to rhyolites. The rift source has less slab-derived material and is an E-MORB-like source,in contrast to an N-MORB-type source overprinted with more slab-derived material beneath the arc. Rift magma types, in theform of rare pumice and lithic clasts, preceded the rift, and the earliest magmas that erupted in the rift already differed from thoseof the arc. The earliest large rift eruption produced an exotic explosion breccia ("mousse") despite eruption at >1800 mbsl.Although this rock type is attributed primarily to high magmatic water content, the clasts are more MORB-like in trace elementand isotopic composition than are modern Mariana Trough basalts. After rifting began, arc volcanism continued to bepredominantly silicic, with individual pumice deposits containing clasts that vary in composition by about 5 wt% SiO2, or aboutas much as in historical eruptions of submarine Izu Arc volcanoes. The overall variations in magma composition with time duringthe inception of arc rifting are broadly similar in the Sumisu Rift and Lau Basin, though newly tapped OIB-type mantle seems tobe present earlier during basin formation in the Sumisu than Lau case.

INTRODUCTION

At their extremes, magmas associated with the two end-membertectonic categories—divergent and convergent plate boundaries—areclearly different: mid-ocean-ridge (MORB) vs. island-arc basalts(IAB). Because these also are volumetrically the most importantmagma types on Earth, and together are responsible for the differen-tiation of the planet, they have attracted abundant attention. However,distinctions between the two tectonic categories as well as betweenthe two magma types blur in backarc basins. Because backarc basinbasalt (BABB) may become less IAB-like and more MORB-like asthe distance between convergent and divergent boundaries growsduring widening of backarc basins (Saunders and Tarney, 1984), thegreatest blurriness is expected during the inception of backarc rifting.Consequently, this study focuses on the composition of volcanic rockserupted (1) in the central Izu Arc before and after the most recentinception of backarc rifting there, and (2) within one of the backarcrifts, the Sumisu Rift (Fig. 1). Throughout this chapter, "rifting" refersto the extensional phase before formation of new oceanic crust withina backarc basin, whereas "spreading" refers to the formation of thisnew crust. Rifting is inferred to have begun in the Sumisu Rift at about2 Ma, but spreading has not yet occurred (B. Taylor, this volume).

In addition to bearing on the geodynamic history of arc rifting andthe reasons for differences between MORB and IAB, the samples alsocontribute to one of the most thorough data sets available for thePleistocene volcanic history of submarine island-arc calderas (alsosee Nishimura et al., 1991), an example of deep submarine explosiveeruption of tholeiitic basalt (Gill et al., 1990; Koyama et al., thisvolume), and background for the dredge- and A/vm-based studies ofthe volcanoes now exposed on the seafloor of the rift (Ikeda andYuasa, 1989; Fryer et al, 1990; Hochstaedter et al., 1990a, 1990b).

The samples are of four kinds. The oldest consists of pre-riftPliocene basalt scoria to rhyolite pumice from below the uncon-formity at Site 788, which dates rift inception. The second includes

1 Taylor, B., Fujioka, K., et al., 1992. Proc. ODP, Sci. Results, 126: College Station,TX (Ocean Drilling Program).

2 Earth Sciences Board and Institute of Marine Sciences, University of California,Santa Cruz, CA 95064, U.S.A.

the late Pleistocene arc-derived scoria and pumice, which postdaterift inception. These are from forearc Site 793, above the unconfor-mity at volcanic arc Site 788, and in the upper half of the section atbackarc Sites 790 and 791 (Fig. 1). The third kind includes the basaltlava, dikes, and breccia that form the basement of the rift at Sites 790and 791. They are even earlier examples of Sumisu Rift BABB thanthe seafloor samples referenced above. The fourth is the rhyolite lavafrom the summit of the en echelon ridges on the Sumisu Rift seafloor.Although not sampled during Leg 126, they are discussed here incomparison with the coeval arc-derived pumices.

The silicic volcanic rocks range from dacite to rhyolite. To helpthe discussion, they have been divided into four increasingly differ-entiated categories (D, Rl, R2, R3) on the basis of silica content andthe concentration of compatible major and trace elements (Table 1).

SAMPLE PREPARATIONAND ANALYTICAL METHODS

Lava and breccia samples from the basement of Sites 790/791were typically 10-20 cm3 splits or mini-cores that were broken into2-10 mm diameter chips, repeatedly ultrasonicated in distilled water,and crushed for 15 min in a tungsten carbide or agate ball mill. Abouthalf were crushed at sea and half at the University of California atSanta Cruz (UCSC). The tungsten carbide crushing at UCSC intro-duced no significant Nb contamination. This was demonstrated firstby grinding separate aliquots of one basalt for 10, 20, and 30 min, andfinding that Nb contents agreed within the 0.5-1.0 ppm precisionbased on counting statistics. Second, analyses at UCSC of samplesfrom adjacent pieces of core crushed in tungsten carbide at UCSC vs.crushed in agate aboard ship also agreed to within this analytical error.However, crushing in tungsten carbide apparently added 0.3-1.0 ppmTa to the samples.

Pumice and lithic clasts were picked from unconsolidated coresections, either at one discrete horizon, or from throughout an entire1.5-m core section when the stratigraphy appeared to have beendisturbed by drilling. Samples that are single large clasts are identifiedin the tables. Some, however, were multiple but similar-appearingclasts each 1-2 cm in diameter. All clasts were scrubbed with atoothbrush and repeatedly ultrasonicated in distilled water until the

383

J. B. GILL ET AL.

130°E

Figure 1. Map of the central Izu Arc illustrating the location of Leg 126 sites, after Hiscott and Gill (this volume). Arc volcanoes (open triangles) are, from northto south, Aogashima (A), Myojinsho (J), Sumisu Jima (S), Minami Sumisu Jima (M), and Tori Shima (T). Submarine calderas occur at or near each of thesevolcanoes except Tori Shima (Murakami and Ishihara, 1985).

water was no longer turbid. After drying, discolored or otherwisevisibly altered clasts were discarded. In two cases from Hole 793A(Sections 126-793 A-3H-6 and -4H-5), two populations of clasts fromthroughout an entire core section were separated on the basis of colorand density, and were analyzed separately. Pumice samples were alsocrushed in the tungsten carbide ball mill. Most of the resultingpowders were 4-25 g, but some were 1.5-4 g. Trace element analysisby X-ray fluorescence (XRF) was not attempted for the small sam-ples, and few loss on ignition (LOI) measurements were made.

Major and trace element analysis by XRF followed the methodol-ogy of Stork et al. (1987), except that a dual Cr-Au tube was used fortrace element analysis of the pumices, resulting in an inability tomeasure their Mn, Cr, V, and Rb contents. The precision of the XRFanalyses is <1% 2σfor the major elements except Na, and <\-A ppm

for trace elements (see Appendix). The data are reported on ananhydrous basis with the LOI and original totals given as backgroundinformation. The original totals for basalts averaged 100.14 ±0.97wt%. Accuracy can be judged by the agreement for internationalstandards reported by Stork et al. (1987), many of which wererepeated during this analytical period and are within twice the preci-sion. Representative data are in the Appendix. Agreement with ship-board results for trace elements in basalts is within 2 ppm except forSr (average difference is +11 ±16 ppm, with UCSC results higher),Ni (+6 ±3 ppm), and Ba (-9 ± 12 ppm).

Our biggest analytical problem was sample preparation for Na,which was measured at UCSC by XRF, using both fusion and pressedpowder pellets, and by atomic absorption for a subset. Unless other-wise noted in Tables 2, 5, and 7, Na results are by XRF on powder

384

Table 1. Characteristics of dacites and rhyolites of the IzuArc.

SiChTiθ2Fe2O3*MgOV (ppm)Se

Dacite

64-700.7-0.86.1-7.51.9-2.2

12021-24

Rl

70-710.55-0.654.5-5.81.0-1.2

3415-18

R2

71-730.45-0.55

3.0-4.00.7-0.8

—11-15

R3

73-750.37-0.50

2.6-3.00.4-0.6

29-10

pellets. Based on comparison between methods and results for rockstandards, our XRF analyses of basalts can be up to 0.5 wt% too high,and our XRF analyses of rhyolites can be ± 0.5 wt%, even though ourinternal precision is generally ± 0.05 - 0.10 wt% by all three methods.

In addition, 10 samples were analyzed for 29 trace elements,including all 14 rare earth elements (REE), by inductively coupledplasma mass spectrometry (ICP-MS) at Memorial University, New-foundland, Canada. Samples were prepared using an HF-HNO3

bomb digestion and analyzed using a standard addition technique(Jenner et al., 1990). Precision and detection limits are <50 ppb (<3relative %) for most elements. Exceptions at the concentrationsencountered in this study are Se, Rb, and Li (± 1 ppm); Cs, Ba, Sr,Ba, Zr, and Pb (±0.1-0.3 ppm); and Nb, Ta, Eu, Lu, Th, and U (±3-7 relative %). Jenner et al. (1990) show that accuracy usually iswithin these precision levels. Cross-checks with UCSC XRF datafor ODP and other samples show agreement of ±0.5 ppm for Nb; ±1-2 ppm for Rb, Se, and most La; and ±5-10 ppm or 5% for Sr andBa. REE and Y usually agree to within 5%, but ICP-MS data for Zrand Y systematically were at least 10% low for the ODP samples. Inaddition, even bomb digestions may have been inadequate for a fewsamples in which Zr, La, and Y contents all were 15%—50% lowerby ICP-MS than XRF and in which, therefore, REE contents may beunderestimated. This is especially true of Samples 126-793 A-3H-6a(dacite) and 126-791B-78R-1 (basalt). All trace element data re-ported in this chapter were obtained by XRF at UCSC, except forthe 23 elements in Table 3 obtained only by ICP-MS at Memorial.

The Sr and Nd isotope ratios were measured at UCSC. All sampleswere leached in cold 2.5 N HC1 twice for 12-24 hr before digestion toremove seawater Sr. Total processing blanks were ~l nanogram Sr and

PETROLOGY AND GEOCHEMISTRY OF VOLCANIC ROCKS, IZU ARC

-30 picogram Nd. Unleached aliquots had ^Sr/^Sr ratios 0.0003-0.0006 higher than leached samples, whereas 143Nd/144Nd ratios differedby only 27 ppm, which is just slightly outside analytical imprecision.Replication of Sr data was assessed by completely processing fivesamples twice, including leaching. The duplicates differ in ^Sr/^Sr byonly 22 ppm on average (the range is from 5 to 52 ppm). Because we arecomparing in this chapter the isotopic results from UCSC with thoseobtained earlier at DTM for A/vm-collected samples (Hochstaedter et al.,1990b), we reanalyzed one Alvin sample (1891-7). The Sr-Nd resultsagree within analytical precision.

Mineral analyses were collected at UCSC with a Princeton Gamma-Tech energy dispersive system (EDS) using a lithium-drifted detectorconnected to an ISIWB-6 scanning electron microscope. Spectra collec-tion time was 200 s, using an accelerating voltage of 15 kV, a samplecurrent in the 300-500 nanoamp range in the Faraday cup, and a beamsize of 5-10 mm. Windows used for background fits included Nb-Ho forolivine, clinopyroxene, Cr-spinel, and magnetite. Separate sets of back-ground windows were used for Plagioclase analyses. Ge-Zr, Mg-Cs, andAg-Co background windows were used for the Na-Al, Si, and Ca-K-Fepeaks, respectively.

Concentrations were calculated relative to natural mineral stand-ards. Oxide analyses relied on chromite and synthetic TiO2 standards.Based on replicate analyses of diopside, olivine, and orthoclasestandards, the lσprecision is 0.71 mol% En (N= 14), 0.16 mol% Fo(N = 21), and 0.86 mol% Ab (N = 10), respectively. Absorption oflow-energy Na X-rays results in a 10% standard error in wt% Na2Ofor feldspar.

The accuracy of the EDS results was checked by analyzing someof the same spots with a 5-spectrometer JEOL 733 wavelengthdispersive electron microprobe at Stanford University. The composi-tion of glass also was determined using this technique. The micro-probe was operated at 15 kV using a nominal 15 nanoamp samplecurrent and a spot size of 20 mm for Plagioclase and glass, and 10 mmfor other minerals. EDS and microprobe results agree to within 2mol% En for clinopyroxene, 0.5 mol% for olivine, and 4 mol% forPlagioclase. Although single oxide grains were not analyzed by bothmethods, analyses of two Cr-spinels included in the same olivine grainbut analyzed by the different methods are compared in Table 12(Sample 126-791B-63R-1, 53 cm).

Table 2. Major and trace element analyses of Pliocene arc pumices.

HoleCore, sectionInterval (cm)Age (Ma)Depth (mbsf)

Major elements (wt%):SiO2

TiChAI2O3Fe 2 O 3

MnOMgOCaONa 2OK2OP 2 O 5

OrigLOIOrig sum

Trace elements (ppm):BaSrZrNbLaYNiSe

Sequence 3

788Ca

8H-167-69

2.869

74.750.32

13.513.37

0.421.944.161.480.04

—92.65

119146204

4.012.654

3.49.1

788Ca

12H-3104-106

3.1110

74.860.44

13.182.75

0.772.343.921.650.09—

92.91

788Ca

12H-510-13

3.1112

72.400.52

13.043.620.121.143.15

c4.501.020.103.89

99.60

115146115

1.85.2

463.1

12.5

788Ca

14H-316-18

3.1128

70.550.72

13.594.93

1.093.69

C4.780.710.16

98.61

108178102

2.57.5

514.7

17.6

Sequence 2

788C14H-343^6

3.1128

75.400.26

13.072.83

0.551.904.411.520.06—

92.34

788Ca b

14H-659-61

3.2133

73.010.55

13.363.77

1.073.134.011.000.10—

91.92

788Ca

14H-CC5-153.2134

60.320.92

14.2510.450.183.576.45

C3.380.610.111.84

100.90

5517262

0.95.7328.8

32.0

788Ca

15H-368-71

3.2138

50.781.00

19.619.02

4.5611.93C2.670.260.12—

97.51

24243

501.42.9

2020.745.3

788Ca

19H-CC10-123.45182

65.710.84

14.417.17—1.505.244.220.760.15—

98.88

8617391

1.84.9

413.3

23.7

Note: LOI = loss on ignition.a Single clast analysis; remainder are multiple clast populations.b Shipboard sample preparation and major element analyses.c Atomic absorption analysis.

385

J. B. GILL ET AL.

PRE-RIFT ARC VOLCANISM

Rifting is inferred to have begun before deposition of the oldestsediments encountered at Sites 790 and 791 in the Sumisu Rift (before1.1 Ma), and after deposition of the youngest sediments beneath theunconformity at 35.1 mbsf at Site 788 on the arc (after 2.35 Ma) (B.Taylor, this volume). Our samples of pre-rift arc volcanism are frombeneath the unconformity at Site 788.

Significant volcanism in the Izu Arc resumed at about 17 Ma,shortly after cessation of spreading in the Shikoku Basin, and hascontinued to the present (B. Taylor, this volume). Because dense basalthad been observed from Alvin to crop out on the lower graben wall only5 km west of Site 788 (Taylor et al., 1990), one of the original drillingobjectives there was to obtain lavas erupted in the arc just before the mostrecent episode of arc rifting. However, only variably indurated pumicesands and gravels, not lavas, were recovered at Site 788. These volcani-clastic sediments are the only products of pre-rift Pliocene volcanism atthis location along the arc. In retrospect, the basaltic rocks seen from Alvinare syn-rift sills in the graben wall (Hochstaedter et al, 1990a). Conse-quently, a more complete record of the magmatism preceding arc riftingwas obtained by studying the geochemistry of the sandstones (Hiscottand Gill, this volume).

The absence of lava makes it ambiguous whether or not thevolcanic front during the Neogene was along strike of Site 788 as now,or west of it at the modern remnant arc (Nishi-Shichito Ridge) fromwhich 2.0 ± 1.1 Ma basalt has been dredged (Yuasa, 1985). B. Taylor(pers. comm., 1990) argues from the seismic stratigraphy in the moatsurrounding Sumisu Jima volcano that the moat formed before thelate Miocene, so that the resumption of arc volcanism after thecessation of spreading in the Shikoku Basin occurred at the presentvolcanic front. If sso, the Pliocene pumices at Site 788 were eruptedfrom centers along strike to the south or north, were deposited in alow spot between centers, and were then uplifted during arc rifting toform the unconformity at 35.1 mbsf at Site 788.

Nine samples of pumice were studied from below that uncon-formity (Tables 2-4). Stratigraphically, most are from the middle ofthree fines-depleted, upward-coarsening pumiceous debris flows thatmark the onset of silicic volcanism near Sumisu Jima during the

Table 3. ICP-MS trace element analyses of arc and rift volcanic rocks.

period from about 3.8 to 2.35 Ma (Nishimura et al., 1991). Sequence2 at 100-140 mbsf is 2.9-3.15 Ma. Our samples include both thejuvenile silicic pumices and one each of basalt and andesite lithicclasts from near the base of this debris flow. The juvenile pumicesrange from dacite to R3 rhyolite; generally, they are more silicic thanthe bulk samples of interbedded sandstone, but they have similarratios of incompatible trace elements (Hiscott and Gill, this volume).In this one debris flow (indeed, in one section of core), the silicicpumices vary by a factor of 2 in MgO and K2O contents, and a factorof 3 in TiO2 and P2O5. In general, the silicic pumices have enrichmentsof alkaline earths relative to REE. One (Core 126-788C-12H-5) hasa flat REE pattern marked by La-Ce depletion and a negative Euanomaly (Figs. 2-3). Nonetheless, their relative alkali and alkalineearth enrichments are less than in their Quaternary analogues, and CsN

~ RbN, so that their "arc-like-ness" is muted.Two R3 rhyolite pumices were found, one within each of the

second and third debris flows. Their incompatible trace element ratiosare more similar to those of the late Pleistocene rhyolite lavas eruptedatop the en echelon ridges on the floor of the Sumisu Rift (see "SumisuRift Rhyolites" section, this chapter) than the other rhyolite pumicesdescribed above. This is most vividly illustrated by comparing the R3rhyolite in the youngest debris flow below the unconformity (Core126-788C-8H-1) with the R3 rhyolite pumice from Core 126-788C-3H-3 above the unconformity at Site 788 (i.e., postdating the incep-tion of rifting) (Fig. 3B). The older sample has much higher Zr/Baratios, Th + LREE (light rare earth element) enrichment rather thandepletion, CsN < RbN, and no Ba or Pb spikes, all of which make itsimilar to the rift rhyolites. The second apparently rift-type rhyoliteis from the second debris flow, in Core 126-788C-14H-3, but it wastoo small for trace element analysis by XRF. Both are unlike all otherpumices or sandstones analyzed, whether pre- or post-rift in age, andwhether deposited in the backarc, arc, or forearc. However, rhyoliteash with similar composition occurs at Site 792 in 0.61-0.73 Masediment (Egeberg et al., this volume). The closest chemical ana-logues erupted in the rift, where they occur as lava, not pumice (see"Sumisu Rift Rhyolites" section, this chapter). The differences be-tween the R3 and the R1-R2 rhyolites probably are not the result ofdifferentiation because fractionation of observed phenocrysts cannot

HoleCore, section

VLiRbCsHfTaPbThULaCePrNdSmEuGdTbDyHoErTmYbLu

788C8H-1

2.699.06

15.700.405.332.433.501.200.50

12.3031.60

4.6921.80

5.831.547.031.147.911.675.070.785.290.83

Pliocene arc

788C12H-5

24.307.639.140.523.371.203.020.470.295.10

14.602.44

12.704.181.265.450.936.391.424.230.644.470.68

788C14H-CC

243.0012.307.720.381.792.192.020.180.182.698.521.508.262.931.134.060.724.961.053.080.473.170.48

788C15H-3

387.003.673.310.101.291.841.200.220.452.798.281.427.352.380.933.060.523.440.732.160.302.020.30

793A3H-6b

121.0011.109.890.962.508.525.620.360.563.69

10.201.799.613.201.014.300.765.141.123.400.513.450.53

Pleistocene arc

790B10H-lc

33.809.428.040.612.822.026.600.470.374.46

12.302.11

11.403.841.215.120.916.151.354.200.624.170.65

788C3H-3

2.209.996.830.462.971.303.270.340.214.12

12.302.14

11.503.981.345.390.986.811.494.570.704.770.75

790C24R-1

277.0010.906.520.161.620.296.510.370.166.17

14.202.38

11.603.151.163.920.613.840.782.200.342.080.31

791B55R-1

290.006.543.570.081.780.121.390.290.114.28

11.401.879.422.791.053.360.553.580.762.190.332.000.32

791B59R-1

270.007.212.620.091.580.091.030.230.163.95

10.301.768.812.421.023.330.543.340.702.060.291.850.27

Rift samples

791B64R-1

321.005.351.320.021.141.260.670.150.463.257.901.457.632.310.912.930.483.230.672.010.291.830.29

791B75R-1

345.004.771.320.031.280.082.150.180.293.509.341.578.032.370.923.160.483.270.691.980.281.850.28

791B78R-ld

287.003.531.390.031.161.060.910.380.144.38

10.801.597.302.060.812.340.372.470.501.390.201.210.18

1894-9"

301.003.033.740.121.240.460.710.210.162.366.841.196.372.200.882.660.453.000.631.810.261.680.26

1895-4a

4.0312.4021.30

0.715.800.453.261.520.50

15.4040.90

6.2828.90

7.842.038.911.52

10.202.196.750.986.721.04

a See Hochstaedter, et al. (1990a, 1990b) concerning Alvin samples 1894-9 (basalt) and 1895-4 (rhyolite).b Sample 126-793A-3H-6,0-150 cm; rhyolite.c Sample 126-790B-10H-1, 0-35 cm.d Sample 126-791B-78R-1,113-115 cm.

386


Table 4. Sr and Nd isotopic analyses of arc and rift

volcanic rocks.

Sample

Pliocene pumices:126-788C-12H-5126-788C-14H-CC126-788C-15H-3

Pleistocene pumices:126-793 A-3H-6

126-790B-10H-1126-788C-3H-3

Rift basement:126-790C-24R-1

u126-791B-55R-1

126-791B-59R-1

126-791B-63R-1126-791B-64R-1

u

Alvin1891-7DTM

•"Sr/^Sr

0.7034361110.703483 ±110.703110+14

0.703596 ±80.703601 ±100.703526 ±100.703409 ±11

0.703015 ±100.703298 ±100.702919 ±130.702940 ±120.702948 ±100.702956 ±110.703036 ±110.703001 ±80.703053 ±100.703699 ±30

0.702990 ±70.703023 ±11

1 4 3Nd/1 4 4Nd

0.513091 ±60.513070 ±16

0.513104 ±9

0.513103 ±80.51310216

0.513066 ±7

0.513100±10

0.513094+80.513088 ±7

0.513115 ±7

0.513043 ±190.513012121

8.98.4

9.0

9.09.0

8.3

9.0

8.98.8

9.3

7.57.5

Note: Errors are 2 standard errors on mean of 80-150 ratios, in ppm.εNd relative to 0.512638 for bulk earth. The letter "u" indicatesunleached; all others were leached before digestion. "DTM" resultis from Hochstaedter, et al. (1990b).

account for the differences in trace element ratios noted above.Instead, these differences are the same as those between rift and arcbasalts and, therefore, reflect parentage.

Two lithic clasts also were analyzed. The basalt (Sample 126-788C-15H-3, 68-71 cm; about 3.1 Ma) is especially interestingbecause of its partial similarity in trace element concentrations andSr-Nd isotope ratios to the later basalts of the Sumisu Rift seafloor(SRB), and its difference from Quaternary basalts of the Izu Arc. Ithas very low concentrations of incompatible trace elements, includingno relative enrichment in Ba or Pb, and a differently but distinctivelyshaped REE pattern with a maximum between Pr and Sm (Figs. 2-3).Ce/Pb ratios, for example, remain MORB-like. The enrichment inLREE and Eu, and the depletion in Ce, are similar to characteristicsof the later SRB (Hochstaedter et al., 1990b). Its 87Sr/86Sr ratio(0.70311) is lower than that of the juvenile silicic pumice (0.70344)and similar to that of SRB from the graben wall (Hochstaedter et al.,1990b). The difference in e d is smaller but analytically significant(+8.4 vs. +9.0). Nonetheless, the similarity to SRB is incomplete; thehigh Al and Ca and low Fe and Ti of the lithic basalt clast suggest amore arc-like differentiation history. In contrast, the andesite lithicclast appears cogenetic with the more abundant silicic pumices. Boththe lithic andesite clast and the juvenile rhyolite clasts have Sr and,in the case of the andesite, Nd isotopic compositions identical to thoseof Quaternary andesites from the Izu Arc (Table 4).

To summarize, the geochemical characteristics of the volcanismthat immediately predated the inception of rifting can be inferred onlyfrom pumice clasts and sandstones from below the unconformity atSite 788. Most clasts are dacite to rhyolite, and most are enriched inPb and alkaline earths, yet depleted in LREE and high field strengthelements (HFSE) relative to heavy rare earth elements (HREE).However, these arc-like attributes are less extreme than in basalts torhyolites of the Quaternary arc. One basalt lithic clast, which wasdeposited 1 m.y. before a rift started to form at about 2 Ma, has manyof the geochemical characteristics of the eventual backarc basin basaltsource, although the basalt did not differentiate in the same way. Themost differentiated rhyolites from near the top of the section also arechemically more similar to rhyolites of the eventual rift rather than tomost or all of those erupted from the arc. These rhyolites weredeposited about 2.6 Ma and also predate rift inception.

POST-RIFT-INCEPTION ARC VOLCANISM

The magma eruption rate in the Izu Arc has increased duringbackarc rifting. At Sites 790 and 791 where age constraints are thebest, eruption of coarse pumice has been common only during the last130,000 yr. The geochemical characteristics of this late Quaternaryvolcanism have been studied by analyzing glass shards from ashlayers (Rodolfo et al., this volume; Egeberg et al., this volume),volcanic sands (Hiscott and Gill, this volume), and pumice clasts (thischapter; Nishimura et al., this volume). The pumice clasts are the leastabundant but most reliable records of magma composition at nearbysubmarine calderas, which are, from north to south, Sumisu Jima,Minami Sumisu Jima, and Tori Shima (Fig. 1). Our samples are fromSites 788, 790, 791, and 793.

We analyzed 27 Pleistocene pumices, mostly from sediments youngerthan 200,000 yr (Tables 3-5). Three are basalt scoria from sediments thatare about 2-600 k.y. old from Sites 790 and 791. They are even moresimilar in composition to Sumisu Rift basalt lavas of similar age that cropout on the modern floor of the rift than is the basaltic lithic clast describedabove. Consequently, we think that they are examples of explosivelyerupted magmas from the rift rather than from the arc, somewhat similarin mode of eruption to the explosion breccia encountered at the base ofSite 791 (Gill et al., 1990). Sediments above and below thick Pumice UnitIII at both Sites 790 and 791 (discussed below) have mafic compositions,the trace element patterns of which are similar to those of SRB (Hiscottand Gill, this volume).

The rest of the pumices consist of arc-derived dacite and rhyolite.Many of those at Sites 790 and 791 can be correlated and define fivelarge pumice eruptions at about 1, 31, 61, 67, and 131 ka (PumiceUnits I to V, respectively; Nishimura et al., 1991). Quaternary pumicesfrom the other sites are known only to be younger than 275 ka andcannot be correlated with these five eruptions in detail. However, thelate Quaternary pumices from arc Site 788 are most like Pumice UnitI, with an abundance of R3 rhyolite.

Almost as much variation exists within each pumice deposit as dodifferences between them. Both analyzed samples from Pumice UnitIV are R2 rhyolites (Table 5 and Nishimura et al., this volume). UnitsII and III contain both Rl and R2 rhyolites, and Unit I contains bothR2 and R3 rhyolites. Dacites are absent from the backarc sites butoccur in both the arc and forearc sites. R3 rhyolites are the leastcommon, occurring only in the youngest horizons at Sites 788 and790. Dacite and R2 rhyolite pumices coexist in the same core sectionsat least twice at Site 793 (Cores 126-793A-3H-6 and -4H-5; see Table5). Although this range may reflect mixtures of juvenile and lithiccomponents, individual silicic eruptions in the Izu Arc can yieldpumices that range from mafic dacite to Rl rhyolite (e.g., the 1952eruption of Miyojinsho; see Table 6), so the distribution may beprimary. In no site is there a stratigraphic pattern in the chemicalcomposition of pumice.

All the silicic pumices are characterized by very low concentra-tions of K2O and other incompatible elements, including HFSE andLREE as well as alkalis and alkaline earths (Figs. 4 and 5). Althoughthey are not as potash-poor as are some of the glass shards inPleistocene ash layers (Rodolfo et al., this volume), they are uniformly"trondhjemitic" in composition. The concentrations of all compatibleelements decrease from dacite to R2 rhyolite, but incompatible elementconcentrations (especially alkalis, LREE, HFSE, and Th) are notablyerratic. The Cs/Rb, Ba/K, and Pb/Ce ratios are consistently high, as iscommon in oceanic arc basalts and andesites, and higher than in thePliocene pumices. The REE patterns are uniformly LREE-depleted,approximately the same as in the Pliocene pumices (Fig. 5). 87Sr/86Srratios are 0.7034-0.7036, similar to or slightly less than those of Quater-nary Izu Arc volcanics (Notsu et al., 1983) and most of the Pliocenepumices. %d is +9.0 and is similar to that found in Quaternary Izu Arcbasalts (Nohda and Wasserburg, 1981).

387

J. B. GILL ET AL.

T3

Oxzü»

LULU

50

40

30

20

10

La CΘ Pr Nd PmSm Eii Gd Tb Dy Ho Er Tm Yb Lu

Figure 2. Chondrite-normalized REE patterns for pre-rift arc volcanics fromSite 788. Normalization values are from Evensen et al. (1978).

These characteristics are shared by the other presumably Quater-nary rhyolites that have been dredged from the Izu Arc (Ikeda andYuasa, 1989). Consequently, they seem representative of the volumi-nous Pleistocene volcanic materials of the central Izu Arc. However,they differ from the subaerial dacites to rhyolites erupted in thenorthern Izu Arc that have higher concentrations of incompatibleelements despite lower Ti, Na, and HREE at similar SiO2 and MgOcontents (Table 6).

Rhyolites that are enriched in LREE ±Nb-Ta also may erupt at thevolcanic front of the Izu Arc, but the evidence is inconclusive. Two

Pliocene LREE-enriched R3 rhyolite pumices from Site 788 werediscussed in the previous section. Two 0.1-Ma turbidites from Sites790 and 791 have intermediate compositions and flat REE patterns(Hiscott and Gill, this volume), as do the hemipelagic muds at thesesites (Nishimura et al., this volume). Because both the basaltic andrhyolitic magmas of the rift are LREE-enriched (Hochstaedter et al.,1990b), these pumices, turbidites, and muds may include materialerupted in the rift rather than arc. The arc, however, is expected to bea major source of detritus at all three sites. Finally, forearc ash at Site792 includes four 8-19 cm-thick, 0.6-0.7 Ma ash layers which areslightly enriched in LREE and Ta (part of the TI category of Egeberget al., this volume). Although the ashes imply shallow-water eruption,none of these four sediment types requires that LREE-enriched mag-mas erupted at the volcanic front rather than behind it.

The Leg 126 Pleistocene samples have higher concentrations ofelements commonly enriched in arc magmas, such as Ba, Cs, and Pb,especially relative to LREE, than do the Pliocene arc samples dis-cussed above, even though Zr/Y ratios are lower in both the Pleisto-cene pumice clasts and sandstones (Hiscott and Gill, this volume).Thus, arc-like characteristics increased with time in the highly differ-entiated magmas, being lower in the arc before the inception ofbackarc rifting and greater after inception began.

EARLIEST VOLCANISM IN THE SUMISU RIFT

The basement of Sites 790 and 791 consists largely of basalticrocks that formed earlier in the history of backarc basin developmentthan any rocks hitherto available from any other backarc. At both sitesthis basement lies beneath sediments about 1.1 Ma (Unit II). Althoughthe age of inception of backarc basin development is somewhatambiguous, both conceptually and temporally, the results of Leg 126suggest that the Sumisu Rift began to deepen rapidly between 1 and2 Ma (B. Taylor, this volume), so that our samples formed within thefirst 0.5-1.0 m.y. of rift development.

About 40 samples from almost all igneous stratigraphic units atboth sites have been analyzed; the results are given in Tables 3,4, and

B

Cs Rb K Ba Sr Pb Th U Nb Cθ Zr Sm Yb

4 0

30

20

10

5

-

1 l

1 l

l

1 1 1 1

1 1 1 1

\ A iIMi 'V

1895-4 _

i / /vl-


Figure 3. A. Normalized extended trace element diagram for pre-rift arc volcanics from Site 788. Normalization values are as for Figure 2 and from Hofmann(1988). B. Normalized extended trace element diagram for R3 rhyolites. Section 126-788C-8H is a pre-rift Pliocene arc pumice; Section 126-788C-3H is apost-rift-inception Pleistocene arc pumice; and 1895-4 is a Pleistocene Sumisu Rift rhyolite collected using Alvin.

388


Table 5. Major and trace element analyses of anhydrous Pleistocene pumices.

HoleCore, sectionInterval (cm)Depth (mbsf)

Major elements (wt):SiO2

TiO2

AI2O3Fe2θ3

a?Na2OK2OP2θ5

Orig. LOIOrig. sum


793Ab

1H-214-17

1.5

72.880.45

13.903.070.723.504.500.900.08

95.37

185149111

1.91.9

372.8

11.2

793Ab

2H-3115-118

8

70.010.64

13.864.611.144.034.590.%0.16

96.59

181157106

1.42.2

413.3

13.9

793Ab

2H-42-5

8

71.520.64

13.774.870.973.773.241.070.15

94.11

793A3H-61-150

21

64.070.71

15.017.491.845.963.930.850.14

98.29

168172117

1.44.2

451.7

11.3

793A3H-61-150

21

71.140.51

13.474.140.813.57

C3.371.120.13

94.68

357245167

2.49.6

64

21.3

793A4H-135-40

23

72.360.52

13.613.880.713.474.241.080.13

95.50

16116296

1.02.9

442.6

12.9

793A4H-2

137-13826

64.810.69

14.297.302.125.874.040.730.14

98.10

13916785

1.45.3

365.2

22.4

793A4H-365-75

27

68.660.61

14.075.651.244.604.031.000.14

98.01

15516198

0.02.5

413.5

17.6

793A4H4

42-4428

63.070.72

15.367.642.196.733.430.740.13

98.85

14516778

1.32.9

344.0

20.4

793A4H-50-150

30

66.140.68

14.086.102.165.874.050.780.15

98.83

14416888

0.93.3

376.9

21.9

793A4H-50-150

30

71.340.53

13.573.900.823.445.310.970.12

94.76

14915690

0.54.0

412.2

11.1

788Cb

1H-190-92

5

74.430.49

12.622.980.622.595.300.860.09

94.72

11913189

1.84.0

4333.2

10.1

Sequence 4

788C2H-130-90

14

67.150.80

13.936.991.924.913.400.750.14

90.66

23217272

1.02.5

348.2

22.0

788Cb

3H-310-12

23

75.530.37

12.602.620.432.26

C4.580.820.05

95.10

12214496

0.72.9

472.4

10.1

HoleCore, sectionInterval (cm)Depth (mbsf)Unit3

790Ab

3H-13 ^18

790A a b

4H-257-59

30

791 Ab

3H-30-150

21i

790Ba

7H-10-40

52ii

790B10H-10-35

81iii

790Ba

10H-180-150

82iii

791 A*22H-226-28

198iii

79OCa

5H-10-150

124iv

791Ba

94R-CC2-3474

791B13R-12-4502

791B34R-1

1-2705

Major elements (wt):S1O2TiO2

A12O3Fe2O3

MnOMgOCaONa2OK2OP2O5Orig. LOIOrig. sum


49.191.12

16.8711.23

—8.01

10.222.780.350.23

98.82

49.091.07

16.4311.230.168.12

10.892.440.370.19

99.81

42222

613.4

21102

72.130.43

13.553.830.110.812.295.471.300.082.11

99.06

165

70.570.56

13.624.210.141.023.875.160.750.113.80

100.26

14517484

2.42.4

412.2

14.8

70.440.60

13.594.54

1.023.845.110.730.13

95.98

16016994

2.06.7

431.3

14.8

72.660.60

13.373.850.150.793.55

a4.700.630.102.96

99.85

12617384

3.34.4

423.6

15.5

69.860.62

14.004.820.141.104.094.460.770.144.25

99.59

127

812.2

10.542

2.614.8

73.380.48

13.163.740.120.863.124.180.870.093.72

99.98

73.080.40

13.363.550.130.842.984.970.910.070.38

100.29

16816493

0.63.2

453.6

12.9

70.780.60

13.454.81

0.963.834.720.680.16

94.74

50.050.97

18.539.64

5.7211.63

2.900.410.16

97.24

5024756

4.45.0

2085.6

Note: LOI = loss on ignition.a Shipboard sample preparation and major element analyses.

Single clast analyses; remainder are multiple clast populations.c Atomic absorption analysis.d Roman numbered "units" are the coarse pumices at Sites 790/791 discussed in the text.

7. Broadly speaking, all are BABB and similar to the younger SumisuRift basalts (SRB), which are exposed on the Sumisu Rift seafloorand previously had been sampled by means of dredging and submers-ible recovery (Ikeda and Yuasa, 1989; Fryer et al., 1990; Hochstaedteret al, 1990a, 1990b).

The basement samples can be grouped into four broad categories.The first consists of the basement at Site 790 and the lavas at the topof the Site 791 basement (791-t: igneous stratigraphic units 2,3, 5,6,7, and 10). The second (791-m) consists of the main breccia unit ofSite 791 ("mousse": igneous stratigraphic units 11, 12, and 14). Thethird (791-d) consists of the dikes that crosscut the basement of Site791 (igneous stratigraphic units 4, 8, 15, 17, and 29). The fourth(791-b) is the generally altered material beneath the mousse at thebottom of Site 791. The range of differentiation in the first twocategories extends from about Mg# 63 to 47, MgO = 8.5^.7 wt%,Ni = 150-10 ppm, Cr = 270-20 ppm, and Se = 40-30 ppm. This isapproximately the same range as on the modern Sumisu Rift floor(Hochstaedter et al., 1990a) and at Site 809 (ODP Leg 132) on one ofthe exposed lava ridges (Storms, Natland, et al., 1991).

The Lavas (Site 790 and Group 791-t)

All lava samples are basalts with the phenocryst assemblage spinel(<l vol%) < olivine (l%-3%) < Plagioclase (5%-20%) (Table 8). Augitesaturation is reached in Group 791 -t (Unit 2) and in one sample from Site790, at between 6 and 7 wt% MgO. Magnetite-saturation (microphe-

nocrysts >O. 1 mm diameter) is reached within the same range of MgOcontents as augite saturation. Representative mineral analyses aregiven in Tables 9-13 for variably differentiated samples. (Only theentries for Core 126-791B-53R-1 are from a Leg 126 lava. Those forCores 126-791B-60R-1 and -63R-1 are from clasts in the brecciadescribed in the next section, and those for 1896-1 are from anA/v/rc-collected lava described by Hochstaedter et al. [1990a]. The fourhost rocks have MgO contents of 8.3,7.8,5.8, and 4.6 wt%, respectively.)

Olivine phenocrysts have Fo7 6_8 8 cores that correlate positivelywith host rock Mg# (Fig. 6). The rims and host lavas are not inequilibrium unless KD = 0.35, where KD is the iron-magnesiumexchange reaction constant for olivine-liquid. A more likely explana-tion is that KD = 0.30 and some of the olivine is accumulative,especially in the more differentiated samples. However, the lack ofglass precludes further quantification of KD or oxygen fugacity (fθ2).

Plagioclase cores are An83_95 except in the most differentiated lavas,in which sieved cores reach An74. Rims are up to 5 mol% more sodic thancores, and zoning is predominantly normal. Most phenocrysts are clear,and sieved crystals containing glass inclusions are common only in themore differentiated, augite-saturated samples. Augite is the only py-roxene; no crystals or rims of pigeonite or orthopyroxene were discov-ered, even in the most differentiated samples.

The ubiquitous chromite inclusions have Cr-numbers (Cr# =Cr/(Cr + Al) ratios) of 0.35-0.47, which go through a maximum astheir host olivines change from Fo 8 8 to Fo 7 6 (Fig. 7A). The Cr-num-bers are similar to those in spinels in MORB, which have the same

389

J. B. GILL ET AL.

Table 6. Representative chemical analyses of Quaternary dacites to rhyoiites from the Izu-Bonin Arc.

Major elements (wt%):S1O2TiO 2

AI2O3Fe 2O 3

FeOMnOMgOCaONa2OK2OP2O5

Trace elements (ppm):RbSrBaZrNbCeNdYNiSeV

Reference

Higashi-Izu

68.800.34

15.601.622.240.091.674.003.721.800.10

45256

1093.6

23

175

11

(1)

73.600.36

14.110.961.350.080.752.364.052.260.09

57226

1284.4

27

1728

(1)

Niijima

74.600.13

15.050.581.120.030.721.402.823.280.25

(2)

Kozushima

75.810.20

13.570.950.550.080.361.624.312.54

(2)

Miyojinsho

63.620.79

16.183.153.820.182.796.152.690.510.13

(2)

69.070.47

14.732.172.590.121.295.133.690.630.11

(2)

Aoga Rift

71.360.35

13.200.822.370.130.982.544.480.800.08

11130

142

14.413.145

26(3)

Sumisu-Jima

68.630.74

13.801.723.310.191.214.314.340.580.19

7.4183181

11.712.6

26(3)

73.470.29

12.340.481.410.100.721.284.641.310.04

20122

195

48

14(3)

Notes: "Higashi-Izu" is on the Izu Peninsula of Honshu; Miyojinsho (1952 eruption), Niijima, and Kozushima are islands. Results from the

Aoga Rift and near Sumisu-jima were obtained from dredged samples. References: (1) Miyajima et al. (1985); (2) Kuna (1962); (3) Ikeda

and Yuasa (1989).

Si/Al ratio (cf. Dick and Bullen, 1984), and the correlation betweenMg-numbers in coexisting olivines and spinels is the same as inMORB (Fig. 7B). However, the Cr-numbers of the spinels are lowerthan expected for the Mg-numbers of the spinels (Fig. 7A). Thisdisplacement from the pattern in MORB also characterizes some arcbasalts and seems to reflect higher pressure fractionation than inMORB (Dick and Bullen, 1984; Luhr and Carmichael, 1985).FQ+3/Σ Fe ratios are higher than in Cr-spinels from MORB, andincrease from 0.35 to 0.62 as the Mg# of the spinel phase decreases.In differentiated Sample 126-791B-53R-1, 10-12 cm, the spinelphase included in both olivine (Fo76) and augite (Wθ44En43Fs13)phenocrysts has Cr2O3 = 6 wt%, whereas the magnetite in the ground-mass has Cr2O3 = 0.1 wt%. Thus, the miscibility gap between spineland magnetite is small in these slightly oxidized differentiates.

Differentiation in the lavas is tholeiitic in character, characterizedby Fe-Ti enrichment (Fig. 8). However, neither Fe nor Ti enrichmentis as great with respect to MgO as in MORB, nor is the depletion inA12O3. That the apparent lack of Al-depletion is partly just a result ofPlagioclase accumulation is shown by the 3 wt% lower A12O3 contentof the mousse glass, discussed below. Incompatible elements, includ-ing Zr and Y, increase by about 30% from the most to the least maficlavas at both sites, implying about 20% fractional crystallization.Overall, the differentiation trend within the lavas is the same as withinthe younger SRB (Hochstaedter et al., 1990a).

Trace element characteristics and isotopic compositions also aresimilar to those of SRB (Figs. 9-11; cf. Hochstaedter et al., 1990b). TheLREE are generally enriched in contrast to the LREE-depleted patternsof coeval samples from the arc (see above). The negative Ce anomalies

La CΘ Pr Nd PmSm Eu Gd Tb Dy Ho Er Tm Yb Lu

Figure 4. Chondrite-normalized REE patterns for post-rift-inception volcanicrocks. The "rift rhyolite" was collected using Alvin, and the "arc rhyoiites" arefrom Leg 126.

40

30

20coCD

S 10ooo

•D

.N 575εoz

• 1895-4• 126-793A-3H

A 126-790B-10H

A 126-788C-3H


Figure 5. Normalized extended trace element diagram for post-rift-inceptionarc volcanic rocks; also see Figure 4.

390


that are present in basalts at both Sites 790 and 791 are absent fromthe arc samples (Figs. 2 and 4), but probably are present in the youngerA/Wn-collected rift basalts for which La data were slightly suspect andnot reported by Hochstaedter et al. (1990b). The 87Sr/86Sr ratios of0.70292-0.70302 and 143Nd/144Nd ratios of 0.51307-0.51310 are withinthe range of SRB, and the Sr ratios are distinctly lower than in the coevalarc samples (Fig. 11).

Trace elements are decoupled from each other between the sites.The 790 basement lavas have higher concentrations of Zr, Nb, andBa, greater enrichment of LREE, higher Zr/Y ratios, and lower Zr/Nbratios than do most basement lavas at Site 791. Consequently, thelavas at Site 790 are more like E-MORB, whereas lavas at Site 791are more like N-MORB. The more LREE-enriched basalts at Site 790may have slightly higher Sr and lower Nd isotope ratios, but too fewanalyses were conducted for confidence.

In detail, the lavas from Site 791 are extremely Ba-poor (5-20ppm), poorer even than in SRB (30-50 ppm). This was suspected onthe basis of shipboard XRF data, and has been confirmed by oursubsequent XRF and ICP-MS results. Stunningly, therefore, even theearliest backarc basalts lack the characteristic Ba - spikes of island-arcbasalts and are almost indistinguishable from MORB. At the otherextreme, some ODP basalts, especially at Site 790, extend to higherconcentrations of Zr, LREE, Sr, and Na than in SRB. The isotopicvariation among the ODP lavas is less than between the inner ridgesand Shadow Mountain types of SRB (Hochstaedter et al, 1990b),even though the trace element differences are greater. Percent melting,therefore, was more variable during the earliest arc rifting.

The Mousse (Group 791-m)

Igneous stratigraphic units 11, 12, and 14 constitute the secondgroup cited above and form about half the basement drilled at Site791. Collectively, these units are a 135-m-thick explosion brecciadubbed "mousse" at sea because of its dark, frothy appearance. Thisbreccia is the earliest large syn-rift volcanic unit sampled from anybackarc basin. It was described in detail by Gill et al. (1990), and itspaleomagnetic characteristics and eruption style are discussed byKoyama et al. (this volume).

The mousse consists of three highly vesicular components: juve-nile glass clasts; a fine-grained, glass-rich juvenile matrix; and acci-dental lithics. Their relative proportion has not been studied in detail,but it is approximately 5:75:20. Both of the glassy juvenile compo-nents are thought to have quenched during explosive deep-watereruption, whereas the lithic materials are considered to be accidental.We analyzed 13 bulk samples by XRF (Table 7); glass and mineralanalyses by microprobe are available for one of the freshest samples(Tables 7 and 9-13). Almost all the XRF analyses are of material thatconsists completely or mostly of the accidental clasts. Only Sample126-791B-67R-1, 61—63 cm, is mostly matrix. This is because, at thetime of shipboard sampling, the clasts seemed fresher and to be theonly unambiguously igneous material. Obtaining hundreds of milli-grams of fresh juvenile glass for analysis is difficult because smalllithic clasts are ubiquitous.

The mousse is distinctive texturally because of its clastic nature andthe vesicular, agglutinated character of its matrix. Lithic clasts are angular,monomict, variably vesicular basalt up to 6 cm in diameter (Fig. 12).(Rare clasts of indeterminant fine-grained material also occur.) Millime-ter-sized lithic fragments are widespread and cannot be distinguishedfrom the matrix without transmitted light, which reveals that the matrixwithin the lithic clasts is tachylite (opaque) formed by slower cooling,whereas the matrix surrounding the clasts is sideromelane (transparent)formed by rapid quenching (Fig. 13). (See Fisher and Schminke, 1984,for a discussion of submarine glass types.)

The juvenile glass clasts are up to 10 mm in diameter. Shards canbe nonvesicular, or contain just a few vesicles, in regions up to 25 mm2

(see the cover photograph accompanying Gill et al., 1990; also see

Fig. 13). Most commonly, however, areas of dense glass occur be-tween zones of intense vesiculation within which bubbles are close-packed (75% vesiculation; e.g., Fig. 14).

The matrix of the breccia is ubiquitously vesicular with a narrowsize distribution of bubbles centered at 0.1 mm diameter. Thesevesicles are smaller and more uniform in size than in either the glassclasts or lithic clasts (Fig. 15). In reflected light, the matrix appearsto consist of small vesicular glass shards, sometimes no larger thannecessary to enclose a few 0.1-mm-sized vesicles. There is a gradualtransition between "glass clasts" and "matrix." Seen in transmittedlight, vesicles are similarly deformed in some regions of sideromelaneglass as large as 2 mm2, but they are differently deformed in otheradjacent regions (Fig. 16). Microprobe analyses show that the glassclasts are homogeneous in composition. Consequently, the deposit isinterpreted as consisting of a single juvenile component, with thematrix being thoroughly disrupted glass clasts. The volcanologicalimplications of the texture of the mousse are discussed further by Gillet al. (1990) and Koyama et al. (this volume), who argue that it reflectsa deep submarine Strombolian eruption.

The juvenile glass is differentiated basalt (Mg# = 50) broadlysimilar in composition to comparably differentiated overlying lavas(Table 7). The matrix contains euhedral olivine and Plagioclase crys-tals, and the olivines contain spinel inclusions. Augite is absent fromthe mousse matrix although present in comparably differentiatedlavas. The olivines are zoned from Fo85.80, the latter representing anequilibrium composition for the glass, assuming KD = 0.30 and molarFe+3/2Fe = 0.15 (Fig. 6). This differs from the statement in Gill et al.(1990), that the olivines were too Fe-rich for equilibrium. Probeanalyses of glass were not then available, and the discrepancy is aresult of using the higher Mg# of the matrix-rich Sample 126-791B-67R-1, 61-63 cm. The plagioclases and spinels range from An78_89

and Cr# = 35-38, respectively, as in the overlying lavas (Table 9).For the reasons cited above, more is known about the lithic clasts

than about the juvenile component. The clasts have high Mg# (60-65), but lower Ni and Cr contents and higher Se and V contents thanmost of the overlying lavas. Zr contents are similar with respect toMgO, but low with respect to Ni when compared with the lavas,rendering the mousse clasts somewhat more like arc tholeiites thanMORB. Little range in differentiation is indicated (Ni = 40-60 ppm;TiO2= 1.0-1.1 wt%).

Although petrographically the clasts appear monolithic, and thevariation in compatible elements is small, significant variation existsin some trace elements. Broadly, two types of clast are present. Themore common type has higher P2O5 (0.2-0.6 wt%) and Y (23-42 ppm) than in SRB, whereas the other has concentrations similar tothose in SRB (0.07-0.10 wt% and 16-21 ppm, respectively). Thehigh-P group also has higher Sr (200-250 vs. 170-190 ppm) and Ba(10-30 vs. 5-6 ppm) but lower alkali contents. Based on this chemicalcontrast, the high-P group seems more altered, but this is not evidentpetrographically and no secondary phosphate minerals were ob-served. Before HC1 leaching, the 87Sr/86Sr ratio in one high-P clastwas 0.70370 (vs. 0.70330 in an unleached lava), whereas after leach-ing there was no difference between high- and low-P clasts (0.70303).Stable isotope studies showed that some clasts have been altered byseawater (Gill et al., 1990), but which type of clast is unknown.

Only one clast is exceptional. Sample 126-791B-73R-2, 84-86 cm,may be arc-derived, judging from its low Ti, Fe, and Nb, and high Si. Itis also the only sample that has a secondary magnetic component(Koyama et al., this volume). Whether or not its geochemical distinctive-ness reflects alteration could not be assessed petrographically because ithad been thermally demagnetized first.

The differences between clasts and matrix are interpreted asindicating that the clasts were accidental (Gill et al., 1990). If, as inmost pyroclastic basalt eruptions, the lithics are xenoliths entrainedduring ascent, then the clasts are samples of deeper stratigraphic levelsthan were drilled. The geochemistry of the clasts indicates that most

391

J. B. GILL ET AL.

Table 7. Major and trace element analyses of rift basement samples.

HoleCore, sectionInterval (cm)

Major elements (wt %):S1O2TiO2AI2O3aFe2O3MnOMgOCaONa2OK2OP2O5LOIOrig. sum

Trace elements (ppm):RbSrBaZrNbLaYNiSe

79OCa

21R-13-8

51.431.35

16.8910.900.164.69

10.141 1 30.540.250.24

99.58

6.22672282

2.24.2

261329

790C23R-1

4-6

49.171.03

17.4110.23

8.829.733.020.510.171.65

100.45

7.5246

9666

5.14.7

2110838

Unit

790C24R-120-22

49.211.04

17.068.91

8.5110.37b2.300.510.30

97.20

8.1255

5162

3.18.2

2310539

1

790Ca

24R-1

48.710.97

17.579.640.137.80

11.632.440.400.270.21

99.56

790Ca

24R-1

49.610.97

17.2110.620.158.49

10.442.490.560.140.49

100.68

790C33R-1

7-8

48.660.95

17.228.82

8.5310.89b2.060.410.16

95.15

5.124740542.97.42010643

791Ba

48R-146-49

49.221.20

16.8411.270.136.81

10.712.850.260.230.56

99.52

3.5251

886

2.24.8

265037

Unit 2

791B49R-112-14

50.731.29

17.5410.47

6.3610.06b3.170.280.241.89

99.04

3.7243

1481

1.93.8

265338

791B52R-110-12

50.361.30

17.579.75

5.7710.70b3.O50.340.271.30

99.54

6.1262

2077

1.46.5

254733

791Ba

53R-110-12

50.021.24

17.3510.340.155.85

10.763.000.280.240.15

99.25

3.7265

11880.26.7

265638

Unit 3

791B54R-1

1-3

48.901.09

17.1111.44

7.6710.40b2.510.250.13

97.08

3.9206

2151

2.74.4

197142

791Ba

55R-12-5

49.331.06

17.3811.130.148.05

10.93b2.310.310.170.35

100.77

4.6213

1879

1.13.5

2110639

Unit 4

791Ba

56R-169-71

48.751.74

15.1014.61

5.649.623.060.200.270.24

99.15

1.7254

669

1.27.7

269

40

791B57R-1

5-9

51.951.65

16.0112.59

4.838.91

b3.450.240.280.95

99.40

1.2273

1785

1.26.1

296

37

Note: LOI = loss of ignition.a Shipboard sample preparation and major element analyses.b Atomic absorptüon analysis.

arc geochemical characteristics were absent not only in the oldestmajor eruptive unit recovered from the rift, but also in what liesbeneath Hole 79IB.

Most of the discussion above refers to the top half of the mousse.Below Core 791B-67R, the rocks are more altered: olivines arereplaced by calcite; smectite, zeolite, and native copper fill vesicles;and juvenile glass is predominantly smectite.

The Dikes (Group 791-d)

Five dike samples were analyzed. Units 8 and 17 span the samerange in differentiation as the lavas. Dike Units 4 and 29 are the mostdifferentiated material recovered and reflect a strong tholeiitic differ-entiation path, with Fe-Ti enrichment and Ca-Al depletion and littlechange in silica. Incompatible trace elements are not strongly en-riched in these fine-grained samples, which suggests some late-stageloss of liquid from dike interstices but retention of oxides.

The Oldest Samples (Group 791-b)

Samples recovered from the base of Site 791, below the mousse,were quite altered. This is the depth interval in which the hole crossesseismically identified faults, where two different dikes intrude within20 m of one another, and where hydrothermal alteration is greatest.Except for alteration, two of the samples (Units 24 and 27) are similarin composition and texture to the lavas of Group 791-t. Sample126-791B-78R-1, 113-115 cm, is the oldest BABB encountered. Ithas similar trace element ratios but the lowest concentration of HREEof any sample analyzed (Table 4 and Fig. 13), which may be ananalytical artifact (see "Sample Preparation and Analytical Methods"section, this chapter).

Units 20 and 28 are especially interesting because they differ fromall the rest, both in texture and composition, and appear to be arc-de-rived. Compositionally, they have high K, Rb, and Ba contents, yetlow HFSE contents. Although the high concentrations may just reflectalteration, samples immediately above and below are not similarlyaffected. Texturally, they are flattened lapilli tuffs showing plasticallydeformed and apparently welded fiamme. However, not all fiammeare parallel nor are all bubble walls flattened (Fig. 17). The uncol-lapsed matrix texture may reflect diagenetic alteration accompanyingcompaction and adjacent hydrothermal activity rather than pyroclas-

tic flow (Allen, 1990). Consequently, although the samples probablywere erupted in the arc, they may have been transported by cold debrisflows rather than hot pyroclastic flows. As a result, they may havetravelled a considerable distance before deposition in the rift. Al-though they are our only examples of arc volcanism during the1-2 Ma interval, they need not represent pre-rift arc basement, aconclusion confirmed by the seismic stratigraphy, which suggestsmore rift fill below the base of Site 791 (Klaus et al., this volume).

The similarity in composition between pre-rift and rift basalt (see"Pre-rift Arc Volcanism" section, this chapter) leads to ambiguity indefining pre-rift basement in backarcs on the basis of petrology andgeochemistry. The BABB lithic clasts in the mousse probably werederived from strata underlying Site 791. BABB compositions alsocharacterize some lithic clasts (e.g., Sample 126-788C-15H-3, 68-71 cm) and glass shards (Rodolfo et al., this volume) in the pre-riftPliocene debris flows at Site 788 in the arc. Consequently, BABB wasbeing erupted somewhere in the area before rifting began.

SUMISU RIFT RHYOLITES

Rhyolite lava caps the western and central en echelon ridges ofthe Sumisu Rift, about 20 km north of Site 791 (Taylor et al., 1990;Hochstaedter et al., 1990a). Although rhyolite has been dredged fromthe northern part of the western ridge, its distribution is unknown indetail. Consequently, this section focuses on the central ridge, whichis entirely rhyolitic at depths shallower than 1900 m below sea level(mbsl). One rhyolite was dated at 276 ka (Hochstaedter et al., 1990a).

Two types of rhyolite are present. One is dense, massive obsidianthat is more differentiated, and the other is microvesicular, blocky,and slightly more mafic. The differences in composition are shownin Table 14 and are greater in compatible elements than in REE orHFSE. The more mafic rhyolites overlie the less mafic, as is commonin subaerial pumice deposits, although uncommon in rhyolite domes(Carrigan and Eichelberger, 1990). The stratigraphy suggests sequen-tial eruption of a zoned magma reservoir that was more differentiatedat its top.

In stark contrast to the coeval arc samples, the Sumisu RiftRhyolites (SRR) are entirely lava, never pumice. The SRR eruptionsite is 1500-1900 mbsl, whereas the floor of the closest arc caldera,Minami Sumisu Jima, is 800-850 mbsl and is covered by rhyolitepumice (Taylor et al., 1990). Although the additional 100 bar pressure

392


Table 7 (continued).

Unit 5

791Ba

57R-131-34

48.551.07

16.7810.590.146.45

12.352.670.410.172.14

99.18

7.5369

1174

0.85.6

2212536

791B57R-161-63

49.621.11

17.229.84

8.8410.17b2.480.180.194.89

98.99

1.3248

1965

2.14.6

2314734

Unit 6

791Ba

59R-141^44

48.911.00

17.4610.700.138.53

10.54fc2.460.200.161.39

100.31

4.1257

1166

1.2

21124

791B60R-12-5

49.261.08

18.109.91

8.6310.47*2.590.120.192.33

100.30

2.1245

1370

1.84.0

2112939

Unit 7

791B60R-162-64

48.591.12

17.0010.81

8.3010.22b2.320.260.211.46

100.42

0.1237

1490

2.35.1

3310737

Unit 8

791Ba

61R-11 ^

50.141.17

17.6910.450.135.97

11.662.630.200.180.64

100.32

5.3254

776

2.04.9

286237

Unit 10

791Ba

62R-131-33

48.841.08

17.6210.89

7.1411.524.390.090.362.10

99.72

1.6244

2348

1.76.4

334045

791B62R-188-92

48.590.96

16.9910.670.158.25

12.112.280.150.481.49

100.63

1.4255

1052

1.05.1

284641

791B63R-160-62

50.701.34

14.7011.600.186.50

11.902.680.160.21

99.97

791B63R-176-78

48.781.08

18.2111.18

7.8410.40b2.030.260.104.29

99.82

3.1197

542

1.44.3

164640

Unit

791Ba

64R-114-18

49.080.96

16.6710.69

8.5110.92

h.\o0.130.52

96.54

2.0228

5745

1.12.4

235942

11

791B65R-124-26

49.071.01

17.0410.57

8.3210.83h.490.160.573.95

101.88

2.2229

1350

1.66.4

424842

791Ba

65R-178-81

48.990.99

16.9112.170.169.18

10.032.020.300.071.56

100.82

1.2193

652

1.02.7

186140

791B66R-181-83

49.500.98

16.2810.95

9.2310.182.350.140.374.32

99.78

0.1205

1851

0.22.7

245739

791Ba

67R-161-63

49.761.04

17.5812.020.188.728.432.410.540.076.32

100.75

6.1176

553

1.53.7

165840

on the rift seafloor might be sufficient to suppress vesiculation incomparably hydrous rhyolite magma, the pressure difference is soslight that SRR probably are drier magma. This greater dryness iseven more remarkable because the SRR are also more differentiated.

Eight rhyolite samples from two Alvin dives were studied (Tables3, 4, and 10). All except one are low-K R3 rhyolites; the exception isa medium-K rhyolite from the upper, more mafic group. AlthoughSiO2 contents are similar to those of arc rhyolites on an anhydrousbasis, SRR have lower compatible element concentrations than in anyPleistocene arc pumices.

Furthermore, rift rhyolites differ from arc rhyolites in all thesame ways that rift basalts differ from arc basalts and andesites(Hochstaedter et al, 1990a, 1990b). That is, SRR have light-enrichedREE patterns, no Ba spike (the difference is in LREE rather than Baconcentration), and higher Th, U, Zr, Hf, Nb, and Be contents (as wellas LREE) relative to HREE (Figs. 4 and 5). Also, they have signifi-cantly lower 87Sr/86Sr and slightly lower 143Nd/144Nd ratios than docoeval arc rhyolites. The significance of differences in the higherabsolute concentrations of incompatible trace elements is ambiguousbecause the SRR are more differentiated overall.

MAGMA SOURCES DURING BACKARCBASIN INCEPTION

The principal change in composition of magma during the historyof the Izu arc-backarc system occurred during the early Oligocenewhen REE patterns changed from flat to LREE-depleted, perhaps inassociation with the establishment of a volcanic arc about as far fromthe plate boundary as the volcanic front is now (Hiscott and Gill, thisvolume). Within this context, the changes during the most recentepisode of arc rifting are less dramatic, but they can be placed withina better understood context of time and tectonics. There is, nonethe-less, ambiguity about "when rifting began"; the following adopts2 Ma as the time (after B. Taylor, this volume).

One surprising result of this study is that backarc basin basalt (BABB)and related rhyolite were present before rift formation. However, thisleaves open the question whether the BABB source was always present(in which case tapping the BABB source was a passive result of exten-sion), or was recently intruded (so that extension was the result ofintrusion); see Hochstaedter et al. (1990b) for discussion.

Another surprise is that the "island arc geochemical signature" wasmore evident after than before the inception of arc rifting; that is, therecycled slab component (rich in Cs, U, Pb, Ba, and 87Sr) has in-creased in the arc during the early stages of formation of the backarcbasin. If the Pliocene volcanic front was on what is now the Nishi-Shichito Ridge, then this increase accompanied trenchward migrationof the volcanic front after the inception of rifting, perhaps as the resultof tapping fertilized mantle that previously underlay the forearc (Sternet al., 1988). Alternatively, if the volcanic front has remained fixedwith respect to the trench (see the "Pre-rift Arc Volcanism" section,this chapter), then the increase may reflect either the cumulativeeffects of slab dehydration or a greater focusing of slab derivativesrelative to the percent of melting beneath the volcanic front duringthe inception of rifting.

A third surprise is that the magma eruption rate in the arc increasedduring incipient rifting, primarily in the form of explosive silicicvolcanism. Furthermore, although volcanism in both the arc andincipient rift is compositionally bimodal, neither of the silicic modesseems to be crustally derived. Were the rhyolites the product of crustalanatexis, the simplest interpretation of the increased eruption rate andbimodality would attribute the basalt to partial melting of the mantle,attribute the rhyolite to partial melting of mafic crust caused byintrusion of the basalt, and attribute the overall increase in magmaproduction rate to decompression of the mantle accompanying rifting.

However, there are three arguments against crustal derivation ofthe rhyolites. First, the arc and rift rhyolites consistently differ in traceelement and isotopic characteristics. Those from the arc are LREE-depleted, have typical arc relative enrichments in Cs, U, Pb, Ba, and87Sr, and greater relative depletions in HFSE. The greater explosivityof arc rhyolite may also indicate higher water contents. Although twoexceptional rift-like rhyolite pumice clasts were found in pre-riftdebris flows at Site 788, the overall volcaniclastic sediments aresimilar in composition to arc rhyolites and dissimilar to rift ones(Hiscott and Gill, this volume). There is no reason to expect suchstriking geographic segregation among crustally derived melts be-cause there is no reason to expect crustal heterogeneities of this typeand distribution. The second and related argument against crustalanatexis is that the differences between rift and arc rhyolites are thesame as the differences between rift and arc basalts. The rift rhyolitesare identical in Sr-Nd isotopes and similar in trace element enrichment

393

J. B. GILL ET AL.

Table 7 (continued).

HoleCore, sectionInterval (cm)

Major elements (wt %):S1O2TiO2AI2O3aFe2O3MnOMgOCaONa2OK2OP2O5LOIOrig. sum


791B67R-1

142-144

48.541.06

17.3710.66

7.9710.054.210.110.333.60

98.98

1.1234

2946

0.65.8

254544

Unit 11

791Ba

72R-343-45

50.210.92

16.5610.940.17

10.834.784.521.150.106.40

100.18

5.6247

144

0.12.8

176431

791B73R-184-86

57.350.61

18.867.76

3.588.78b2.490.110.201.39

100.88

2.41873245

0.12.4

378

34

79 IB"73R-3

112

49.611.01

16.5910.790.127.95

10.172.900.520.232.10

99.89

2.8292

749

2.24.1

264843

Unit 12

791Ba

75R-157-59

49.680.98

17.1110.130.118.94

10.742.410.150.191.92

100.44

0.1211

547

1.03.4

196142

Unit 14

791B76R-18-11

49.271.08

17.4910.73

9.209.932.200.150.405.39

101.74

2.3212

448

1.25.8

234540

Unit 17

791Ba

77R-192-95

48.511.09

18.6011.690.207.82

10.052.600.150.151.50

100.86

0.9241

1464

0.51.4

2110938

Unit 20

791B"77R-2

128

60.660.96

15.668.920.127.221.641.763.260.244.48

100.44

71.359

14784

1.6

455

Unit 24

791Ba

78R-179-82

52.621.00

17.0910.120.099.586.603.020.470.232.94

100.82

4.6272

3564

1.57.4

216730

Unit 27

79 IB"78R-1

113-115

51.591.05

17.9810.480.148.717.592.820.220.212.45

100.79

2.6305

2684

1.66.3

175834

Unit 28

791Ba

79R-117-20

51.040.68

18.7811.160.218.981.673.963.410.064.32

99.95

25.9308186

170.12.3

142529

Unit 29

791B79R-194-97

49.091.59

16.9014.%0.225.435.905.150.730.163.43

100.13

9.8363

2155

3.03.6

222

pattern to rift basalts, as are the arc rhyolites similar to arc basalts andandesites. Although the range in differentiation and the space-timevariability preclude meaningful quantitative tests of consanguinity foreither rift or arc basalt-rhyolite suites, nothing in our data qualitativelyprecludes either silicic mode from being derived from their associatedbasalts by fractional crystallization. The final argument is that internalvariations within the Pleistocene pumices of the forearc at Site 793,which range from 63 to 73 wt% SiO2, are more explicable in termsof fractional crystallization than partial melting. That is, with theexception of one anomalous mixture of pumice clasts from Section126-793A-3H-6, compatible element concentrations (e.g., Se, Ti, Mg)drop by a factor of 2 whereas alteration-resistant incompatible ele-ment concentrations (e.g., Zr, Y) increase by <40%. Although thisdoes not preclude the dacite from being a primary anatectic melt,variation within the silicic mode is a result of fractional crystallization.Because all the rhyolites of both suites are low-K, the mafic parentmust have had very low K2O, and differentiation must have involvedrapid increase in SiO2 as the result of crystallizing calcic Plagioclaseand magnetite.

The preceding paragraph implicitly assumed that "rift rhyolites"do not erupt at the volcanic front. Possible evidence against thisassumption was noted earlier (see "Post-rift-inception Arc Volcan-ism" section, this chapter), and the distinction could be gradational,both geochemically and geographically. Although indiscriminant in-termingling of backarc and arc basalts along the volcanic front wouldaffect geodynamical models of the mantle wedge, the eruption of "riftrhyolites" at the volcanic front would not undercut the three conclu-sions reached above.

Within the backarc, another surprising observation is that theearliest Sumisu Rift volcanism is even less arc-like than the previouslystudied surface samples (SRB). Apart from high H2O and U contentsrelative to other incompatible elements, and negative Ce, Nb, and Zrspikes relative to REE, there is little to distinguish any basalts in theSumisu Rift from E-MORB. The ODP samples appear similar in mineralcompositions and differentiation style, and therefore in water pressureand oxygen fugacity, to SRB. However, the earliest basalts at Site 791formed exotic explosive pyroclastic deposits >1800 mbsl, as a result inpart of their high water contents (Gill et al., 1990; Koyama et al., thisvolume). Despite this, alkali and alkaline earth concentrations are evenlower in basaltic rocks from Site 791 than in SRB. Hence, the striking

geochemical contrast between arc and backarc magmas was presentduring the earliest phase of rift development.

In addition to eruption style and alkaline earth concentrations,which are subduction-related parameters, the ODP samples providesomewhat more extreme examples of E-MORB than do the A/vm-col-lected suite. The LREE and HFSE enrichment is greater in the basaltsof Site 790 than in those of Shadow Mountain on the seafloor,although %d in the ODP samples is not as low.

As with the surface samples, a key issue is whether the apparentdifferences between more and less depleted samples has the sameexplanation as between N-MORB and E-MORB, or whether it is aresult of recent subduction-related metasomatism (cf. Hochstaedteret al., 1990b, with Fryer et al., 1990). Because the more E-MORB-likecharacter of Site 790 basalts is accompanied by higher Zr, Nb, andLREE, we continue to think that the contrasts reflect differences inaddition to those attributable to recent subduction. The basalts fromSite 790 are from a more enriched, more OIB-like, source.

COMPARISON WITH THE LAU BASIN

Comparable information about magma compositions before, dur-ing, and after incipient rifting is known for the Lau Basin (Gill, 1976a;Cole et al., 1985, 1990; Hawkins and Melchior, 1985; Gill andWhelan, 1989a, 1989b). There, rifting apparently began at about 5 Maand spreading began at about 3 Ma.3 The spreading center is now asouthward-propagating axial rift in the northern Lau Basin, and aridge close behind the arc volcanoes in the southern Lau Basin (Parsonet al., 1990).

Although similar, the sources of data for the two arc-backarc pairsdiffer in two respects. For the Sumisu case, our results for pre- andsyn-rift arc rocks are drawn entirely from the trenchward side of thebackarc basin, and we know that our backarc basalts were eruptedwithin the rift during the first million years of its existence. In contrast,in the Lau case, the samples of pre-rift arc rocks are from the remnantarc, the samples of syn-rift arc rocks are initially from the remnant arc

3 Copies of the "Leg 135 Preliminary Report, Lau Basin" are available from ScienceOperations, Ocean DrillingProgram, 1000 Discovery Drive, College Station, TX 77845-9547, U.S.A.

394


Table 8. Modal phenocryst mineralogy of rift basement sam-ples.

Core,section

126-790C-21X-CC

126-791B-47R-147R-147R-147R-]48R-153R-155R-156R-156R-156R-157R-157R-57R-159R-60R-60R-61R-62R-62R-162R-163R-163R-164R-164R-265R-165R-166R-67R-67R-67R-68R-70R-70R-72R-72R-.73R-.75R-76R-76R-:76R-.77R-77R-77R-'77R-77R-78R-78R-78R-78R-78R-79R-

1

i

!i

!II

79R-179R-1

Piece

2

310111795112

13IB

4851

131

1AIB5B

10B13B

1315

17155

1099

101218

2A1010126B4A

17

2A6A1A

23B

810A12A

367

Unit

1

111122

3344556678

1010111111111111111111111111111111111112141415161719202022

24262728

29

Olivine

<

<<<

(<)<X<

(<)<<

« )X« )<<<

—(<)

(<)<

(<)<

« )(<)« )(<)

(<)(<)

(<)« )« )

(<)(<)(<)« )

(<)

(X)(<)(X)

(X)

Plagio-clase

X

<<<<<<XXXXXXX<<<X<XXX<<X<X<<<<<<<<<XX<XXXX<XXXX—X

(X)XX

(X)

Clino-pyroxene

—

X

<<

XX

—

—

<

—

—

———

X—X<

XNX<

(X)

Mag-netite

—

NN<N<<NNN<<N<N

N

NN—NNN—N

NNN——NNNNN<N<<<<<<

<NNNN

Notes: Phenocrysts are >0.5 mm except for magnetite where >O.l mm. Xindicates >5% by volume; < indicates l%-5%; — indicates notpresent. N indicates that the silicate phenocrysts were tabulated inSites 790/791, table 11 of Taylor, Fujioka, et al. (1990), but the thinsections were unavailable for further study. Parentheses indicate somealteration. Units are the shipboard igneous stratigraphic units (Taylor,Fujioka, et al., 1990).

and later from the volcanic arc, and the age and tectonic context of theoldest backarc basin basalts (Leg 135, Site 834) are more speculative.

In the Lau case, arc volcanism immediately before rifting constitutesthe "mature arc stage" of Fiji (Gill, 1987). Closest to the eventual LauBasin, these volcanic rocks form the Lau Volcanic Group (LVG). At thevolcanic front, they consist mostly of medium-K tholeiitic andesitescharacterized by slight LREE-enrichment, high %d (+8.0-8.4), low^Sr/^Sr (0.7027-0.7034), and low Sr/Zr ratios (Gill, 1976a, 1987; Coleet al., 1985,1990). In both cases, pre-rift arc volcanism is more enrichedin LREE, Zr, and Th, but more depleted in Ba, Sr, Pb, and U than is arcvolcanism, which postdates rift inception.

The oldest basalts in the backarc basin are MORB-like in bothcases. In the Sumisu Rift, this is the "mousse" at Site 791; in the LauBasin, it is the basement at Site 834 (see footnote 3). In the Lau case,an isotopically enriched component is present in the northern basin

(Volpe et al., 1988); this component appeared in the remnant arc whenthe spreading stage of backarc basin formation began (Gill andWhelan, 1989b; Cole et al., 1990). The enriched component is moresubtle in the Sumisu case, but concentrations of HFSE and LREE arehigher in the initial backarc than in pre- or syn-rift arc lavas.

In the Lau case, rifting began in the forearc and syn-rift arcvolcanism continued on the proto-remnant arc (the Lau islands),where it is preserved as the Korobasaga Volcanic Group (KVG; Gill,1976a; Cole et al., 1985, 1990; Gill and Whelan, 1989a). The KVGis characterized by high relative enrichments of alkalis, alkaline earths(including 87Sr), and U, implying a large slab component. It also ischaracterized by flat REE, low concentrations of HFSE and Th, andhigh %d and % of about +8.4 and +15.4, respectively (also see Saltersand Hart, 1991), implying a depleted source apart from the recent slabcomponent. Much less is known about the Sumisu case, but volcanismalso appears to continue on the proto-remnant arc; although one dacitecap is known, Ba enrichment is absent (Fryer et al., 1990).

In summary, pre-rift arc magmas in both cases had relatively smallamounts of slab-derived components. In both cases, the rifting stage wascharacterized by the immediate tapping of a MORB-type source witheven less slab-derived material (mostly water). A source characterized byan older enrichment appeared during the spreading stage in the Lau case,at the free northern end of the basin. In the Sumisu case, an olderenrichment is present even during the rifting stage. In both, syn-riftvolcanism continues on the remnant arc; this is clearly arc volcanism inthe Lau case, but not in the Sumisu. In both, subsequent frontal arcvolcanism that accompanies rifting or spreading is especially enriched incomponents apparently derived from the slab.

CONCLUSIONS

Starting about 2 m.y. before rifting began in the Pliocene, andcontinuing afterward, frontal arc volcanism in the Izu Arc becameprimarily dacitic to rhyolitic in composition. This reflects develop-ment of large submarine calderas, seven of which now exist in the arc(Murakami and Ishihara, 1985). Because the silicic magmas appearto be differentiates of subcrustal basalt rather than crustally derived,this upsurge in silicic volcanism is attributed to changes in stresswithin the arc that permitted eruption of highly viscous magma as amore tensional regime developed. Judging from the record at Sites790 and 791, the rate of this transition in the vicinity of the SumisuRift accelerated rapidly at about 130 ka, since which time large-vol-ume rhyolite eruptions have been common. This increase in eruptionrate apparently differs from the situation in the Izu Arc during riftingin the Oligocene, when arc volcanism reached a minimum duringspreading within the Shikoku Basin (B. Taylor, this volume). Thedifference may, however, merely reflect greater time resolution in themore recent case; that is, the late Quaternary spurt may be a last gasp.

This is the first andesite-to-rhyolite transition to be well-docu-mented, both geodynamically and geochemically, in an oceanic arc.In continental environments, such transitions have been interpretedas reflecting a reduction in magma supply rate associated with thecessation of subduction (e.g., Wark et al., 1990). The petrologicalevidence and interpretations are similar in both realms; the differencesin geodynamic context, in magma supply vs. eruption rate, and in theeffects of stress on magma transport and storage where crust is thickvs. thin await further study.

During the 2 m.y. before rifting, juvenile magmas continued to beLREE-depleted, but there was a diminution in the amount of uppercrustal components (as estimated by Ba/La, Sr/Nd, and Cs/Rb ratios)in these magmas. The Pleistocene rhyolites are more arc-like thanwere their Pliocene counterparts.

As noted previously, there is a striking contrast between magmaserupted within the Sumisu Rift and those from the adjacent arc (Ikedaand Yuasa, 1989; Fryer et al., 1990; Hochstaedter et al, 1990a,1990b). The differences between rift and arc magmas have beeninterpreted as indicating the presence of a source beneath the rift that

395

J. B.GILL ET AL.

90

85 -

080 -

75 -

70

cores

ALVIN •

ODP

I I

rims andg-mass

D

O

•α><A Λ

O

/ / /

&// / ^

/ \ I

I

t1

i

—

A 100

35 40 45 50 55 60 65

100 Mg/(Mg+Fθt>

in host rock

Figure 6. The composition of olivine phenocrysts in Sumisu Rift basalts (SRB).Circles are ODP samples, and squares are for A/v;n-collected basalts (Hochstaedteret al., 1990a). Closed symbols are cores; open symbols are rims and groundmass.The solid lines show equilibrium compositions, assuming KD = 0.30 (Roedder andEmslie, 1970) and molar Fe+3/(Fe+2 + Fe+3) = 0.05 and 0.15, as shown. MeasuredFe oxidation states lie within this range (Hochstaedter et al., 1990a). The dashed linesassume RD = 0.35.

differs not only in lacking as much slab-derived component, but alsoin having greater "old enrichment"; that is, reflecting more of anE-MORB source in the rift vs. an N-MORB source in the arc (Ikedaand Yuasa, 1989; Hochstaedter et al., 1990b). Our results generallyconfirm this suggestion. However, even though LREE, Zr, and Nbcontents are especially high in basalts from Site 790, their εNd ratiosare not as low as the lowest ratios of SRB. Also, we show that the riftbasalts have clear negative Ce anomalies, just as do the early Sea ofJapan basalts (Cousens et al., 1990). Thus, the earliest basalts eruptedin the rift had the same Nd isotopic composition as the arc magmasas well as some subduction-related characteristics, including highwater contents and negative Ce and Nb anomalies. However, theylacked other subduction-related traits such as 87Sr/86Sr ratios to theright of the mantle array, positive Ba spikes, and high δ^S (Gill et al.,1990). This is a much more rapid transition from arc to backarcmagma types, in time as well as space, than had been imagined.Despite the apparently high water content and high eruption rate ofthe oldest rift basalts (the mousse), they are more MORB-like in traceelement and isotopic composition than are the modern MarianaTrough basalts, for example (Hawkins and Melchior, 1985).

Mineral compositions are similar in the early (ODP) and later(Alvin) rift basalts. Plagioclase and spinel compositions confirmearlier suggestions of higher water pressure and somewhat higher f θ 2

than in MORB. Differences from MORB in the correlation betweenMg# and Cr# in the spinels may reflect higher pressure fractionationbecause of the thicker crust during incipient rifting.

Individual pumice clasts, but not bulk volcaniclastic sediments,show that rift-type basalts and rhyolites were present before riftingbegan. If, as argued above, some of the differences in compositionare the result of a different source, then either a new E-MORB sourceintruded before a morphological rift developed, or E-MORB compo-

396

80 -

<£ooo

60 -

20 -

0

100 80 60 40 20 02+v

100 Mg/(Mg+Fe")B

90

85

80

7 *

D

COo

o

1

10 30 50 70

100 Mg/(Mg+Fθ+)

90

Figure 7. A. The composition of spinel inclusions in olivines in Sumisu Riftbasalts (SRB). The symbols denote the host olivine compositions: square,>Fo8 5; hexagon, Fo g 4; triangle, Fo 8 1_ 8 2; diamond, Fo 7 6. The field for MORBis from Dick and Bullen (1984). B. Mg# of spinels and their host olivines inSumisu Rift basalts. Symbols as in Figure 7A.


4 6 8MgO (7.)

D

4 6 8MgO (%)

126-790C

126-791B

above Mousse

126-791B

126-791B

below Mousse

Figure 8. Major element composition of basement basalts from Sites 790 and 791. The fields for MORB, QuaternaryIzu Arc (arc), and Sumisu Rift basalts (SRB) are from Hochstaedter et al. (1990a).

397

J. B. GILL ET AL.

20

<D

RE

E/C

ho

nd

ri

6

5

i i I I I I 1

/SRB

-

1 1 1 1 1 1 1 1

-

-

La Ce Pr Nd PmSm Eu Gd Tb Dy Ho Er Tm Yb Lu

Figure 9. Chondrite-normalized REE patterns for basement basalts from Sites790 and 791. The field for Sumisu Rift basalts is from Hochstaedter et al.(1990b). Filled circles represent Site 790 basalts, filled squares are Site 791basalts, and filled triangles are Site 791 mousse lithic clasts. Not shown is basaltSample 126-791B-78R-1 because of possible analytical problems (see text).

30 -

20 -

o i o

1 -

Cs Rb K Ba Sr Pb Th U Nb Ce Zr SmYb

Figure 10. Normalized extended trace element diagram for basement basaltsfrom Sites 790 and 791. Symbols as in Figure 9.

nents (e.g., dikes or other plums in the pudding) had always beenpresent but began to be preferentially tapped as the rift developed.

Pumice eruptions during the last 130,000 yr have been entirelylow-K dacite-rhyolite in composition. Most of the pumice depositsare compositionally heterogeneous, either because of secondary re-working, or because the range in juvenile clasts usually is as great asthat in the 1952 eruption of nearby Myojinsho volcano.

No simple correlation exists between clast composition and tec-tonic environment. Clasts with rift-type compositions occur in pre-riftdebris flows in the arc. Clasts with arc-type compositions occur in the

0.5133

0.5132

0.5131

0.5130 -

- 12

0.5129

- β

0.7025

Figure 11. Sr and Nd isotope ratios for arc and rift samples from Leg 126. Thefields for Pacific MORB, the Quaternary Izu Arc, and Sumisu Rift basalts arefrom Hochstaedter et al. (1990b). Open circles are for Quaternary arc pumices,open squares are for Pliocene arc pumices, and solid triangles are for riftbasement basalts and lithic clasts.

rift; those near the top at Sites 790 and 791 clearly are in arc-deriveddebris, but those in the basement of Site 791 (Group 791-b) areambiguous. Their chemistry and texture may indicate proximal depo-sition in the arc and, therefore, that pre-rift basement had been reached.However, we interpret them as analogous in depositional mechanismto those near the top, and we attribute the difference in texture todiagenetic flattening.

Variations of magma composition before, during, and after arcrifting are broadly similar in the two best documented cases, Sumisuand Lau. In both, slab components are more in evidence in arcmagmas after rifting begins than before. In both, the earliest basaltsin the backarc rifts are BABB or even MORB, not IAB. Sourcescharacterized by old enrichments, which are not major contributorsto the arc magmas, are present in both backarcs.

ACKNOWLEDGMENTS

We thank B. Taylor, A. Nishimura, R. Hiscott, and M. Koyama fordiscussion and data that bear on these topics, K. Collerson and R.Williams for assistance with mass spectrometry, E. Widom for assis-tance with column chemistry, and P. Fryer and R. Hickey-Vargas forhelpful reviews. Simon Jackson oversaw the ICP-MS analyses atMemorial University.

REFERENCES

Allen, J. F., Sack, R. O., and Batiza, R., 1988. Cr-rich spinels as petrogeneticindicators: MORB-type lavas from the Lamont Seamount chain, easternPacific. Am. Mineral, 73:741-753.

Allen, R. L., 1990. Subaqueous welding or alteration, diagenetic compaction,and tectonic dissolution? LA VCEI Congress, Mainz, Abstracts.

Carrigan, C. R., and Eichelberger, J. C, 1990. Zoning of magmas by viscosityin volcanic conduits. Nature, 343:248-251.

Cole, J. W., Gill, J. B., and Woodhall, D., 1985. Petrologic history of the LauRidge, Fiji. In Scholl, D., and Vallier, T. (Eds.), Geology and OffshoreResources of Pacific Island Arcs—Tonga Region. Circum-Pac. Counc.Energy Miner. Resour., Earth Sci. Ser., 2:379^414.

Cole, J. W., Graham, I. J., and Gibson, I. L., 1990. Magmatic evolution of LateCenozoic volcanic rocks of the Lau Ridge, Fiji. Contrib. Mineral. Petrol.,104:540-554.

Cousens, B. L., Allan, J. F, and Gorton, M. P., 1990. Sediment contaminationof the mantle beneath the Sea of Japan, ODP Leg 127. Eos, 71:1704.

398


Table 9. Representative Plagioclase analyses. Table 12. Representative chromian spinel analyses.

Sample

Interval (cm)

Major elements (wt%):S1O2AI2O3FeOCaONa2OSum

Trace elements (ppm):SiAlFeCaNaSum

An

Note: C = core, R = rim,

126-79IB-60R-162-64

C

47.4633.44

0.5417.70

1.77100.91

2.1681.8000.0210.8660.1575.011

84.68

and S = siev

R

48.2732.86

0.6316.%

2.34101.06

2.1991.7640.0240.8280.2075.022

80.02

ed.

126-79IB-63R-131

C

47.4033.23

0.6017.53

1.69100.45

33

R

46.5933.31

0.6317.92

1.5499.99

126-79IB-53R110-12

C<S)

48.9932.81

0.8416.512.64

101 79

Cations based on 8 oxygens2.1731.7960.0230.8610.1505.004

85.15

1.8120.0240.8860.1385.012

86.54

fable 10. Representative clinopyroxene analyses.

2.2151.7480.0320.8000.2315.026

77.56

R

49.1232.53

0.7516.642.38

101.42

2.2261.7380.0280.8080.2095.009

79.44

18%-1

C

46.7433.46

0.8417.84

1.54100.42

2.1501.8140.0320.8790.1375.012

86.49

Sample

Major elements (wt%):

Tiθ2AI2O3Cr2O3FeOc F e 2 O 3MnOMgOSum

Trace elements (ppm):TiAl

P " +

F e 3 +

MnMgSum

Mg#

Fe‰

126-791B-60R-1,62

I

86.800.90

29.7329.9414.709.920

14.85100.04

0.0201.0380.701n 1A4.U. JO**

0.22100.6563.000

0.6430.4030.11

I

86.800.91

30.8626.6415.2610.680

14.3198.66

126-791B-63R-1, 53

I

84.001.11

29.9826.1817.8513.330

13.26101.71

I

84.000 70

28.9425.6817.8612.280.20

11.9297.58

Cations based on stoichiometry0.0201.0880.6300 3820^24100.6383.000

0.6260.3670.12

0.0251.0430.6110 4410^29600.5843.000

0.5700.3690.15

U.Uló1.0540.6270 4620^2860.0050.5493.000

0.5430.3730.15

Dl-2

I

82.001.84

21.5528.0220.3017.460

10.8299.99

0.0440.8000.6980 5350^41400.5083.000

0.4870.4660.22

Sample

Major elments (wt%):SiO2T1O2AI2O3

FeOMnOMgOCaOSum

Trace elements (ppm):SiTiAlFeMnMgCaSum

EnFsWo

126-791B-53R-1,

C

51.610.763.875.810.33

15.9122.68

100.97

R

50.411.004.906.690.32

15.1522.39

100.86

10-12

C

50.501.093.849.090.30

14.9821.14

100.94

Cations based on 6 oxygens1.8840.0210.1670.1770.0100.8660.8874.012

44.859.19

45.96

1.8510.0280.2120.2050.0100.8290.8814.016

43.2910.7345.99

1.8670.0300.1670.2810.0090.8260.8384.019

42.4714.4643.08

1896-1

C

50.960.943.978.280.29

15.1221.61

101.17

1.8730.0260.1720.2550.0090.8280.8514.015

42.8313.1644.01

R

48.911.746.01

10.490.39

14.0420.74

102.32

1.7970.0480.2600.3220.0120.7690.8164.025

40.3116.9042.80

Note: C = core and R = rim.

Table 11. Representative olivine analyses.

Sample 126-79IB-60R-I.62 791B-63R-1.53

Major elements (wt%):S1O2FeOMnOMgOCaOSum

Trace elements (ppm):

FeMnMgCaSum

Fo

40.10 39.5512.52 14.580.27 0.32

46.48 45.020.39 0.38

99.76 99.85

0.9990.2610.0061.7260.0103.001

86.87

0.9940.307

1.6S70.0103.006

84.62

41.48 40.86 38.2214.83 17.90 21.800.37 0.41 0.45

46.74 43.82 38.120.45 0.54 0.42

103.87 103.53 99.01

Cations based on 4 oxygens1.0010.2990.0081.6810.0122.999

84.89

1.0030.3680.0091.6040.0142.997

81.35

1.0040.4790.0101.4920.0122.996

75.71

40.0416.58

44.410.42

101.77

0.9960.3450.0071.6460.0113.004

82.68

25.680.60

37.750.49

103.49

0.9950.5480.0131.4360.0133.003

72.38

Note: C = core, R = rim, and G = groundmass.' 5 3 - 1 0 = 126-791B-53RI. 10-l2cm.

Dick, H.J.B., and Bullen, T, 1984. Chromian spinel as a petrogenetic indicatorin abyssal and alpine-type peridotites and spatially associated lavas. Con-trib. Mineral. Petrol, 86:54-76.

Evensen, N. M., Hamilton, P. J., and O'Nions, R. K., 1978. Rare-earth abundanceson chondritic meteorites. Geochim. Cosmochim. Acta, 42:1199-1212.

Fisher, R. V., and Schmincke, H.-U., 1984. Pyrodastic Rocks: Berlin (Sprin-ger-Verlag).

Note: I = inclusion.a Fo = Forsterite content of enclosing olivine core; see Table 11.b Probe analysis.c Fβ2θ2 calculated from charge balance.

Table 13. Representative Fe-Ti oxide analyses.

Sample 126-791B-53R-1, 10-12

Major elements (wt%):Tiθ2Cr 2O 3AJ2O3Te2θ3FeOMnOMgOCaOSum

Trace elements (ppm):TiCr

kF e 2 +

MnMgCaOSum

%Mgt%Ulvsp%Other

a8.476.69.04

31.3640.10

0.435.360

a10.806.139.46

29.9737.37

0.385.930

a14.5102.01

37.4145.81

0.490.770.35

"1.440.142.64

45.2140.89

0.381.920.22

101.35 100.04 101.35 102.84

0.2790.2070.4441.0031.1530.0150.3330

3.432

50.418.131.6

Cations based on 4 oxygens0.3570.1930.4660.9421.0550.0130.37003.396

46.617.036.4

0.5070.0000.1100.1561.5750.0190.0530.0173.420

50.744.5

4.8

0.3940.0050.1420.4131.4220.0150.1310.0113.522

60.233.5

6.3

1896-1

10.7205.58

45.0938.45

03.420

103.26

0.3460.0000.2821.3681.29800.21903.513

58.429.512.1

Note: I = inclusion and G = groundmass.a Probe analyses.b Fe2θ3 calculated after Carmichael (1967).

Fryer, P., Taylor, B., Langmuir, C. H., and Hochstaedter, A. G., 1990. Petrologyand geochemistry of lavas from the Sumisu and Torishima backarc rifts.Earth Planet. Sci. Lett., 100:161-178.

Gill, J., Torssander, P., Lapierre, H., Taylor, R., Kaiho, K., Koyama, M., Kusakabe,ML, Aitchison, J., Cisowski, S., Dadey, K., Fujioka, K., Klaus, A., Lovely M,Marsaglia, K., Pezard, P., Taylor, B., and Tazaki, K., 1990. Explosive deepwater basalt in the Sumisu backarc rift. Science, 248:1214-1217.

Gill, J., and Whelan, P., 1989. Early rifting of an oceanic island arc (Fiji)produced shoshonitic to tholeiitic basalts. J. Geophys. Res., 94:4561^578.

Gill, J. B., 1976a. Composition and ages of Lau basin and ridge volcanic rocks:implications for evolution of an interarc basin and remnant arc. Geol. Soc.Am. Bull, 87:1384-1395.

, 1976b. From island arc to oceanic islands: Fiji, southwesternPacific. Geology, 4:123-126.

1987. Early geochemical evolution of an oceanic island arc andbackarc: Fiji and the South Fiji Basin. J. Geol., 95:589-615.

Govindaraju, K., 1990. 1989 compilation of working values and sampledescription for 272 geostandards. Geostand. Newsl., 13:1-113.

399

J. B. GILL ET AL.

15

20

25

30

35 •— I

Figure 12. Photograph of core of the explosion breccia ("mousse"); Section791B-73R-2. This is from the altered base of the section so that the lithic clastsstand out visually and are particularly abundant. The exceptional sample witha secondary magnetic overprint (see Koyama et al., this volume, fig. 2C) andpossibly arc-like composition is from slightly lower in this core section.

Hawkins, J. W., and Melchior, J. T, 1985. Petrology of Mariana Trough andLau Basin basalts. J. Geophys. Res., 90:11431-11468.

Hochstaedter, A. G., Gill, J. B., Kusakabe, M., Newman, S., Pringle, M., Taylor, B.,and Fryer, P., 1990a. Volcanism in the Sumisu Rift. I. Major element, volatile,and stable isotope geochemistry. Earth Planet. Sci. Lett, 100:179-194.

Hochstaedter, A. G., Gill, J. B., and Morris, J. D., 1990b. Volcanism in theSumisu Rift. II. Subduction and non-subduction related components. EarthPlanet. Sci. Lett., 100:195-209.

Hofmann, A. W., 1988. Chemical differentiation of the Earth: the relationshipbetween mantle, continental crust, and oceanic crust. Earth Planet. Sci.Lett., 90:291-3U.

Ikeda, Y, and Yuasa, M., 1989. Volcanism in nascent back-arc basins behind theShichito Ridge and adjacent areas in the Izu-Ogasawara arc, northwestPacific: evidence from mixing between E-type MORB and island arc magmasat the initiation of back arc rifting. Contrib. Mineral. Petrol, 101:377-393.

Jenner, G. A., Longerich, H. P., Jackson, S. E., and Fryer, B. J., 1990. ICP-MS:a powerful new tool for high precision trace element analysis in earthsciences: evidence from analysis of selected U.S.G.S. reference samples.Chem.Geol, 83:133-148.

Kuno, H. (Ed.), 1962. Catalogue of the Active Volcanoes of the World, Part II:Japan, Taiwan, and Marianas. Int. Assembly Vole. Chem. Earth's Interior.

Luhr, J. F, and Carmichael, I.S.E., 1985. Jorullo Volcano, Michoacan, Mexico(1759-1774): the earliest stages of fractionation in calc-alkaline magmas.Contrib. Mineral. Petrol, 90:142-161.

Miyajima, H. T, Yoshida, T, and Aoki, K., 1985. Geochemical study of theHigashi-Izu monogenetic volcano group. Rep. Nuclear Res. Inst., 18:158-174.

Murakami, F., and Ishihara, T, 1985. Submarine calderas discovered in thenorthern part of the Ogasawara Arc. Earth Monthly, 7:638-646.

Nishimura, A., Marsaglia, K. M., Rodolfo, K. S., Collela, A., Hiscott, R. N.,Tazaki, K., Gill, J. B., Janecek, T, Firth, J., Isminger-Kelso, M., Herman, Y,Taylor, R. N., Taylor, B., Fujioka, K., and Leg 126 Scientific Party, 1991.Pliocene-Quaternary submarine pumice deposits in the Sumisu Rift area,Izu-Bonin arc. In Fisher, R. V, and Smith, G. A. (Eds.), Sedimentation inVolcanic Settings. Spec. Publ.—Soc. Econ. Paleontol. Mineral., 45:201-208.

Nohoda, S., and Wasserburg, G. J., 1981. Nd and Sr isotopic study of volcanicrocks from Japan. Earth Planet. Sci. Lett., 52:264-276.

Notsu, K., Isshiki, N., and Hirano, M., 1983. Comprehensive strontium isotopestudy of Quaternary volcanic rocks from the Izu-Ogasawara Arc. Geo-chem. J., 17:289-302.

Parson, L. M., Pearce, J. A., Murton, B. J., Hodkinson, R. A., Bloomer, S.,Ernewein, M., Huggett, Q. J., Miller, S., Johnson, L., Rodda, P., and Helu, S.,1990. Role of ridge jumps and ridge propagation in the tectonic evolution ofthe Lau back-arc basin, southwest Pacific. Geology, 18:470-473.

Roedder, R. L., and Emslie, R. F, 1970. Olivine-liquid equilibrium. Contrib.Mineral. Petrol, 29:275-289.

Salter, V.J.M., and Hart, S. R., in press. The mantle sources of ocean ridges,islands and arcs: the Hf-isotope connection. Earth Planet. Sci. Lett.

Saunders, A. D., and Tarney, J., 1984. Geochemical characteristics of basalticvolcanism within back-arc basins. In Kokelaar, B. P., and Howells, M. F.(Eds.), Marginal Basin Geology. Spec. Publ., Geol. Soc. London, 16:59-76.

Stern, R. J., Bloomer, S. H., Lin, P.-N., Ito, E., andMorris, J., 1988. Shoshoniticmagmas in nascent arcs: new evidence from submarine volcanoes in thenorthern Marianas. Geology, 16:426-430.

Stork, A. L., Smith, D. K., and Gill, J. B., 1987. Evaluation of geochemical referencestandards by X-ray fluorescence analysis. Geostand. Newsl, 11:107-113.

Storms, M. A.,Natland, J.H.,etal., 1991. Proc. ODP, Init. Repts., 132: CollegeStation, TX (Ocean Drilling Program).

Taylor, B., Brown, G., Fryer, P., Gill, J. B., Hochstaedter, A. G., Hotta, H., Langmuir,C. H., Leinen, M., Nishimura, A., and Urabe, T, 1990. i4/vm-SeaBeam studiesof the Sumisu Rift, Izu-Bonin Arc. Earth Planet. Sci. Lett., 100:127-147.

Volpe, A. M., Macdougall, J. D., and Hawkins, J. W., 1988. Lau Basin basalts(LBB): trace element and Sr-Nd isotopic evidence for heterogeneity inbackarc basin mantle. Earth Planet. Sci. Lett., 90:174-186.

Wark, D. A., Kempter, K. A., and McDowell, F. W., 1990. Evolution of waning,subduction-related magmatism, northern Sierra Madre Occidental, Mex-ico. Geol. Soc. Am. Bull, 102:1555-1564.

Yuasa, M., 1985. Sofugan Tectonic Line, a new tectonic boundary separatingnorthern and southern parts of the Ogasawara (Bonin) Arc, northwest Pacific.In Nasu, N., Kobayashi, K., Uyeda, S., Kushiro, I., and Kagami, H. (Eds.),Formation of Active Ocean Margins: Tokyo (Terra Scientific), 483-496.

Date of initial receipt: 2 January 1991Date of acceptance: 20 August 1991Ms 126B-125

400


Clast

Matrix

Figure 13. Photomicrograph of breccia in transmitted light showing the boundary between lithic clast (dark, upperleft) and matrix (light, lower right). The euhedral Plagioclase phenocrysts of the lithic clast are clearly visible as areits 0.2-mm vesicles lined with 1-2-mm-thick smectite. The black tachylitic groundmass consists of devitrified glass,Plagioclase microlites, and submicroscopic crystals. Euhedral Plagioclase and olivine crystals are visible in the matrixcomponent, where the size and linings of vesicles are approximately the same as in the clast. The matrix componentis lighter in color because its glass is fresh sideromelane. The contact between matrix and clast exhibits complex0.05-mm-wide zones of alteration in the lower left and upper right of the figure. Sample 126-791B-62R-1, Piece 5b,for which a chemical analysis is given in Table 7; >40 magnification; the field of view is 3 mm in the long dimension.

Table 14. Chemical analyses of anhydrous Sumisu Rift rhyolites collected using Alvin.

Sample

Major elements (wt%)

SiO 2

TiO2

A12O3

Fβ2θ3MgOCaONa2OK2OP 2 O 5

H2OSum

Trace elements (ppm):BRbBaSrZrNbY

NiCrSeV

LaCeNdSmEuGdDyErYb

1895-14

73.020.42

14.362.530.492.026.071.050.06

—98.55

17157169267

4.665

10

10—

17

Type A

1895-10

73.200.42

14.432.530.411.975.971.010.06

—98.74

17163163265

4.460

6

10—

16

1888-10"

72.470.41

14.522.780.371.896.241.060.060.20

99.37

1018

151139305

5.468

61

117

1940.029.0

7.92.29.2

10.76.77.3

1888-13"

70.950.40

14.222.750.322.045.661.6S0.060.91

99.36

930

154136303

4.672

61

106

1638.928.4

7.72.18.9

10.46.67.2

1888-4

72.620.39

14.632.81

1.936.271.030.05

—100.13

23150132291

4.769

51

116

16

TypeB

1888-8

72.450.39

14.552.850.311.896.401.120.05

—98.28

23153132291

5.768

13

10—

15

1895^

73.810.35

13.972.610.241.676.081.200.06

—98.04

23155113272

4.771

10

104

1640.928.9

7.82.08.9

10.26.76.7

a Data from Hochstaedter et al. (1990a, 1990b).

401

J. B.GILL ET AL.

-

1 1 ,

•

•H

i

j

1

I

*

. s

•Jh. '

*,

1>

<••- 1 .

fyM~ •: -

SP

Figure 14. Photomicrograph of a glass clast in the matrix component. Vesicu-lation primarily occurs along fractures, leading to close-packing of vesicle andfragmentation of the glass clast. Sample 126-791B-63R-1, Piece 8; × 200magnification; the field of view is 7.5 mm in the long dimension.

402


120

4.6% vesicularity

0.8 1 1.2Diameter (mm)

1.4 1.6 1.8 0.8 1 1.2Diameter (mm)

ü

0.4 0.6 0.8 1 1.2Diameter (mm)

1.4 1.6 1.8

Figure 15. Vesicle size distribution in basaltic rocks in the basement of Site 791. Photomicrographs 8 × 10.6 mm were analyzed using Image 1.29q software. Thevertical axis of the histogram is raw, cumulative number of vesicles. The horizontal axis diameters assume that measured areas were of circles. The "percentvesicularity" numbers are integrated 2-dimensional areas of vesicles. A. Matrix of the explosion breccia (mousse) excluding lithic and glass clasts. Sample126-791B-63R-1, Piece 8. B. Lithic clast from the explosion breccia. Sample 791B-70R-1, Piece 12. C. Glass clast within the explosion breccia, similar to clastin Figure 14. Sample 126-791B-63R-1, Piece 8. D. Lava overlying the explosion breccia. Sample 126-791B-47R-1, Piece 11. Note the narrow size distributionand smaller size of vesicles in the matrix vs. the similarity in size distribution between the lava, lithic clast, and even overall glass clast.

403

J. B. GILL ET AL.

Domain 1

Figure 16. Photomicrograph of mousse matrix showing an annealed contactbetween scoriaceous domains, one more plastically deformed than the other.Sample 126-791B-67R-1, Piece 5, for which a chemical analysis is given inTable 7.

Appendix 1. XRF analytical background.

Major elements (wt%):SiO2TiO2AI2O3Fe2O3MgOCaONa2OK2OP2O5Sum


Precision8

0.330.020.110.080.030.150.100.030.010.69

0.5<1%1-41-3

11-2<l<l<l

Accuracy

-0.18-0.13-0.16-0.06+0.02-0.12+0.15-0.07+0.01

—

<2<3%2-102-5<l<2<l1-3

2

Calibrationrange

—

—

—

<150<500<200<130<15<50<502-1701-40

Unrotated clast Unflattened vesicle

Figure 17. Photomicrograph of the crystal-lithic tuff in the oldest samples atthe bottom of Site 791. Sample 126-791B-77R-2, Piece 6a, Unit 20; ×95magnification. Notice that vesicles are less flattened than pumice fragments,and that the clast in the bottom of the picture lies at a high angle to the others.Both suggest diagenetic alteration during burial or local heating rather thanpyroclastic flow.

' Replication of BCR1 daily during 2 months, as wt%. Trace elemententries are a summary of replications of several standards.

' Major elements are differences from Govindaraju (1990) for JB-2;+ means UCSC was higher on an anhydrous basis.

404

26. petrology and geochemistry of pliocene … · 26. petrology and geochemistry of...

Documents