Geochemistry, Geophysics, Geosystems

Geochemistry of lavas from Negros Arc, west central Philippines: Insights into the contribution from the subducting slab



[1] The Negros Arc in west central Philippines is comprised of six Pliocene to Quaternary stratovolcanoes that have erupted calc-alkaline to slightly shoshonitic basalts to dacites. Petrographic, major and trace element, and Sr, Nd, and Pb isotope data suggest that crustal level differentiation processes have likely generated the range in composition of lavas. Several silicic lavas from the southern volcanoes have high Sr and moderate Sr/Y similar to melts presumed to be derived from subducted basaltic crust (adakites). However, these silicic lavas have low La/Yb unlike typical adakites, and have other incompatible trace element ratios (e.g., Nb/Ta, Ba/La, Pb/Ce) which are opposite to those expected of adakites. Modeling results also indicate that partial melting of the subducting Southeast Sulu Sea crust cannot satisfactorily reproduce the adakite-like characteristics or account for the other incompatible trace element ratios. Fractionation of amphibole (± apatite) together with variable fractionation and accumulation of plagioclase have likely imparted the adakite-like features. Trace element and isotopic variations indicate a three-component mixing, whereby variable proportions of melted sediment and aqueous fluids derived from basaltic crust were added to the subarc MORB-source type mantle to form the source of Negros Arc lavas. Southern Negros lavas reflect the strongest fluid contribution whereas northern Negros lavas best exhibit the sediment melt contribution.

1. Introduction

[2] There is consensus that the source for most arc magmas is the mantle wedge that was previously enriched in volatiles and large ion lithophile elements derived from the subducting slab [e.g., Kay, 1980; McCulloch and Gamble, 1991; Arculus, 1994; Tatsumi and Eggins, 1995]. Partial melting of the metasomatized wedge generates basaltic magmas that may then differentiate to more silicic types at higher crustal levels through processes including fractional crystallization, assimilation, and magma mixing [e.g., Gill, 1981; Ewart and Hawkesworth, 1987; Devine, 1995]. The basaltic oceanic crust and sediment cover of the subducting slab are both viable sources for the geochemical enrichment of arc magmas [e.g., White and Dupre, 1986; Miller et al., 1994; Pearce and Peate, 1995]. However, the relative importance of these subduction sources to arc magmas is still in debate. Several studies have emphasized contribution only from the basaltic crust [Yogodzinski et al., 1994; Kersting and Arculus, 1995; Sajona et al., 2000a] or sediment cover [Vroon et al., 1995; Hoogewerff et al., 1997], but most have invoked involvement of both sources [e.g., Ellam and Hawkesworth, 1988; McDermott et al., 1993; Elliot et al., 1997; Woodhead et al., 1998].

[3] The nature by which the subduction contributions are transported from the slab to the mantle wedge is also controversial. Many studies have stressed that aqueous fluids expelled during dehydration of the subducting slab carry contributions from the basaltic crust and sediments [e.g., White and Dupre, 1986; Tatsumi et al., 1986; Castillo et al., 1999; Hochstaedter et al., 2001]. Several studies also considered an aqueous fluid flux from the basaltic crust but suggested a melt phase for the transfer of sedimentary materials [e.g., Elliot et al., 1997; Class et al., 2001]. In contrast, other studies have emphasized the contribution of the basaltic crust as a melt phase [e.g., Defant et al., 1992; Yogodzinski et al., 1994; Shinjo, 1999; Sajona et al., 2000a].

[4] This study presents an evaluation of major element, trace element, and Sr, Nd, and Pb isotopic data for volcanic lavas in Negros Arc, west central Philippines, which are related to subduction of the 10–30 Ma Southeast Sulu Sea Basin. Geochemical data for the arc lavas are integrated with existing and new data for the subducting basaltic crust and sediment cover to address the following objectives: (1) to constrain the origin of andesitic to dacitic lavas in the arc - whether these are genetically related to mafic magmas derived from the mantle wedge, or are partial melts of subducted Sulu Sea crust [Sajona et al., 1993, 1994, 2000a, 2000b], and (2) to characterize the subduction components, i.e., the slab materials and their transfer processes that potentially enriched the mantle wedge source of Negros Arc magmas.

2. Geotectonic Setting of Negros Arc

[5] The Negros Arc (Figure 1) is the northern segment of a curvilinear chain of Pliocene to Quaternary volcanoes associated with the southeastward subduction of the Southeast Sulu Sea Basin along the Negros and Sulu Trenches in west central and southwestern Philippines [Hamilton, 1979; Sajona et al., 1993; Schluter et al., 1996]. The ∼300 km arc comprises six stratovolcanoes on Negros Island and the northeastern tip of Zamboanga Peninsula on Mindanao Island. These stratovolcanoes are shown in the inset map of Figure 1. The Negros Arc is built upon an oceanic terrane. The Cotabato Fault separates it from the Zamboanga/Sulu Arc segment that has continental basement [Pubellier et al., 1991; Schluter et al., 1996; Rangin et al., 1999a; Tamayo et al., 2000].

Figure 1.

Generalized tectonic map of the Philippines. Inset shows enlarged map of southern Philippines [after Sajona et al., 1994]. Symbols are as follows: open triangles, volcanoes related to active subduction zones on margins; solid triangles, volcanoes grouped as part of Central Mindanao arc possibly associated with detached remnant of Molucca Sea Plate; CF, Cotabato Fault; PF, Philippine Fault. Negros Arc volcanoes are numbered: 1, Silay; 2, Mandalagan; 3, Canlaon; 4, Cuernos; 5, Ampiro; 6, Malindang.

[6] The Southeast Sulu Sea Basin is a small backarc basin that formed by seafloor spreading in Oligocene to early late Miocene (30–10 Ma) based on magnetic anomalies [Roeser, 1991] or during early to middle Miocene (19–15 Ma) based on paleontologic dating [Rangin, 1991]. It formed behind an active arc represented by the now submerged Cagayan Ridge. The basaltic crust of the basin has trace element chemistry transitional from island arc to normal mid-ocean ridge basalt (N-MORB) and Indian Ocean MORB-like isotope compositions [Spadea et al., 1991, 1996]. A ∼1050 m thick sequence of pelagic, volcaniclastic and continent-derived sediments covers the basaltic crust at the deep part of the basin [Rangin et al., 1990]. However, only ∼200 m thick section, dominated by continent-sourced sediments, is likely being subducted along the Negros Trench [Solidum, 2002].

[7] Subduction of the eastern margin of Southeast Sulu Sea Basin along the Negros Trench may have started during the late Miocene or early Pliocene [Hall, 1996; Rangin et al., 1999a; Sajona et al., 2000b]. The age of the subducting crust, as constrained by magnetic lineation [Roeser, 1991; Schluter et al., 1996], is ∼20 Ma in front of the southern end of the trench and gets older toward the north. Average convergence rate along the Negros Trench is between 3.5 to 4.5 cm/yr based on Global Positioning System measurements [Rangin et al., 1999b]. Seismicity extends down to a depth of 150 to 160 km under Negros Island [Acharya and Aggarwal, 1980; Bautista, 1996] and indicate the top of the subducting slab is ∼100 km beneath Negros Island volcanoes [Bautista, 1996].

[8] Negros Island volcanoes overlie basement of Paleogene volcanics and sediments overlain by Miocene limestones and clastic sediments [Bureau of Mines and Geosciences, 1982; Maturgo and Pamatian, 1994; Pamatian et al., 1995]. Negros Island volcanoes are dominantly andesites and dacites; minor basaltic andesite flows also occur at Canlaon and Cuernos [Umbal and Arboleda, 1987; Maturgo and Pamatian, 1994; Pamatian et al., 1995; von Biedersee and Pichler, 1995]. The adjoining Ampiro and Malindang volcanoes in northeastern Zamboanga overlie a basement of Cretaceous peridotite, with minor amphibolites and schists, and Tertiary metavolcanics, limestone and clastic sediments [Pubellier et al., 1991; Apuada et al., 1992; Santos and Pamatian, 1995]. Basalt to basaltic andesite volcanism dominated the history of these volcanoes but minor andesitic to dacitic eruptions also occurred [Barnett et al., 1985; Apuada et al., 1992; Sajona et al., 1994; Santos and Pamatian, 1995]. Some studies have grouped Ampiro and Malindang with the rest of the Central Mindanao Arc volcanoes farther to the east, which are possibly associated with a detached remnant of the Molucca Sea Plate [Sajona et al., 1994, 2000a]. Ampiro and Malindang volcanoes are considered part of the Negros Arc in this study because they are aligned with volcanoes lying parallel the Negros Trench. More importantly, tomographic surveys suggest the presence of an east-dipping slab associated with the Sulu Sea Basin crust beneath both Negros Island and northeastern Zamboanga Peninsula [Besana et al., 1997; Rangin et al., 1999a].

3. Samples and Methods

[9] Seventy-nine representative samples were selected from the rock collections of the Philippine National Oil Company, Philippine Institute of Volcanology and Seismology, and Scripps Institution of Oceanography (SIO) and analyzed for this study. Thin sections of most of these were examined under a petrographic microscope. All samples were analyzed for major oxides and selected trace elements (e.g., Rb, Ba, Sr, Nb, Zr, Y, V, Co and Ni) by X-ray fluorescence (XRF). Smaller sets of representative samples were selected for Pb, Nb, Ta, and rare earth element (REE) determinations by inductively coupled plasma mass spectrometry (ICPMS) and for Sr, Nd and Pb isotope ratio measurements by thermal ionization mass spectrometry (TIMS). Three representative samples of sediments likely subducted in the Negros Trench were also analyzed for their Nb and Ta concentrations by ICPMS. Details on methods for sample preparation and analytical measurements are presented in Solidum [2002]; estimated accuracy and precision are described as notes under appropriate tables.

4. Results

4.1. Petrography

[10] Negros Arc lavas range from basalts (49 wt% SiO2) to dacites (64 wt% SiO2). All samples are porphyritic. A few basalt and basaltic andesite samples are moderately phyric (with 3–15 modal% phenocrysts) but most lavas, especially the more evolved andesite and dacites, are highly phyric (usually with 15–40% to as much as 60% phenocrysts).

[11] Basalts and basaltic andesites, which are most common in Ampiro and Malindang volcanoes, are typically vesicular. Olivine and clinopyroxene are ubiquitous as phenocrysts; several samples also have minor magnetite and orthopyroxene phenocrysts. The groundmass typically has plagioclase microlites, pyroxene and titanomagnetite.

[12] In andesites and dacites, plagioclase is the main phenocryst with minor magnetite. Plagioclase typically occurs in low-silica andesites together with phenocrysts of clinopyroxene, orthopyroxene and sometimes hornblende set in a groundmass of plagioclase, pyroxene, magnetite, and glass, or in more evolved lavas together with phenocrysts of hornblende with or without minor biotite in a matrix of plagioclase, magnetite, and glass. Glomerocrysts or clots of complexly zoned plagioclases are common in andesites and dacites. Sparse olivine phenocryst is present in a few low-silica andesites. Apatite is a common accessory in the hornblende-bearing andesites and dacites from Ampiro.

[13] Several Negros Arc lavas show resorption and reaction textures indicating crystal-melt disequilibrium [Davidson et al., 1988; Defant et al., 1991a; Miklius et al., 1991], possibly caused by crystal accumulation. Olivine phenocrysts in some basalts, basaltic andesites, and andesites have corroded rims and rounded form. Plagioclase phenocrysts in a few mafic magmas, and in most andesites and dacites, have numerous glass inclusions forming a sieve-texture as a result of partial dissolution and subsequent infilling by the melt of plagioclase which is out of equilibrium with the bulk melt [Blundy and Shimizu, 1991]. Resorption in the plagioclases is also shown by the corrosion and rounding of grain boundaries.

[14] Interaction with basement rocks by some of the andesites and dacites in Ampiro and Malindang is suggested by the presence in some samples of small (≤ a few cm3) xenoliths of peridotite and chlorite schist and olivine xenocrysts. The olivine xenocrysts are different from olivine phenocrysts derived from arc basaltic magmas because they are highly fractured and/or commonly occur as polycrystalline aggregates, and some of them have hornblende reaction rims. As mentioned earlier, the peridotitic basement and associated metamorphic rocks beneath Ampiro-Malindang is part of an oceanic terrane.

4.2. Major and Trace Elements

[15] Major and trace element data are listed in Table 1. The majority of the samples from Northern Negros, Cuernos, and Ampiro-Malindang (northern, central, and southern volcano group, respectively) are medium-K to high-K calc-alkaline basalts, basaltic andesites, andesites, and dacites (Figure 2). Ampiro-Malindang, from which most samples were collected, also have high-K calc-alkaline to shoshonitic basalts and andesites [cf. Sajona et al., 1994, 2000a]. Many Ampiro-Malindang basalts are highly magnesian (>6% MgO), but their Ni and Cr contents are not significantly high to represent primary magmas (i.e., these should have Ni ∼ 250–300 ppm, Cr ∼ 500–600 ppm [Perfit et al., 1980; Wilson, 1989]). A few of the magnesian basalts and basaltic andesites have unusually high Ni and Cr abundances (up to 486 ppm and 657 ppm, respectively), but petrographic data show that these values can be attributed to olivine and spinel accumulation.

Figure 2.

K2O versus SiO2 plot of Negros Arc lavas. Fields and nomenclature after Peccerillo and Taylor [1976].

Table 1. Major and Trace Element Compositions of Representative Negros Arc Volcanic Rocksa
SampleSilayMandalaganCanlaonCuernos de NegrosCuernos de NegrosAmpiroMalindang
  • a

    Replicate analyses show major element precisions in terms of relative standard deviation (RSD, 1σ) are <1% for Si, Ti, Al, Fe, Mg, and Ca, <2% for K and P, <3% for Na, and <4% for Mn. Trace element precisions are <5% for Ba, Cr, Pb, Rb, Sr, V, Y, Zr, and REEs except for Nd and Sm (<7%), and <8% for Co, Nb, and Ni. PF, pyroclastic flow; DA, debris avalanche deposit.

SiO2 (wt.%)53.7558.4762.9154.0859.4256.758.8660.1360.0858.4255.5458.0561.459.9758.1552.0959.4156.6756.9464.0456.0451.4151.6262.4955.9259.8550.3448.5148.7850.9649.948.7559.7448.18
Rb (ppm)27466072665557607683323845423620433351377435343234343716503237733629
Cr5121833166912 572762513176356545696527633244359456188725259578500
Pb                   7.983.
La13.710.6 17.716.114.614.714.914.114.612.114.213.612.912.6914.91116.69.793.
Ce27.121.9 39.430.331.430.529.629.330.721.128.626.125.426.119.931.523.333.521.
Pr3.93 5.444.
Nd17.113 23.416.217.917.618.817.717.615.215.414.213.614.612.517.21319.311.911.
Gd                   2.632.
Tb0.620.51 0.630.50.530.550.60.60.53 0.470.420.40.48 0.43 0.630.370.470.410.40.350.570.380.440.420.440.40.450.380.290.43
Er2.42.1 1.4
Yb2.42.1 21.81.82221.
Lu0.380.35 0.320.270.270.330. 0.22
La/Yb65 998777871291076138 962284832223362

[16] Major and trace elements show some scatter in the Harker diagrams, but as a whole, they define linear to curvilinear trends (Figures 3 and 4). With increase in SiO2, TiO2, Fe2O3*, MnO, MgO, and CaO all decrease, whereas K2O and Na2O increase (Figures 2 and 3). Al2O3 and P2O5 vary in the basalts and basaltic andesites, but Al2O3 remains essentially constant whereas P2O5 exhibits a decreasing trend with increasing SiO2 starting at ∼56%. Several high silica lavas (SiO2 > 58%) from Ampiro-Malindang have slightly higher MgO content than andesites and dacites from Cuernos and Northern Negros.

Figure 3.

Major element oxide variations of Negros Arc lavas plotted against SiO2 content.

Figure 4.

Trace element variations of Negros Arc lavas plotted versus SiO2.

[17] Compatible trace elements such as V and Ni (and also Cr which is not shown) correlate negatively with silica (Figure 4). Among the incompatible trace elements, Rb shows the widest ranges in abundance and generally shows no correlation with SiO2. Barium positively correlates with SiO2, but the rest are correlated only up to 56–57% SiO2, and then exhibit varying trends at higher levels. Strontium and light rare earth element (LREE) La have relatively flat trends at SiO2 contents > 56%. High field strength elements (HFSE) Nb and Zr, Y and the heavy rare earth element (HREE) Yb initially increase up to about 56–57% SiO2, after which Zr essentially defines a flat trend, whereas Nb, Y, and Yb decrease with increasing silica. Several of the basaltic andesite lavas have high Sr and Al2O3 (up to ∼19%), possibly enhanced by plagioclase accumulation.

[18] Negros Arc lavas have trace element characteristics typical of calc-alkaline arc lavas. Relative to N-MORB, these show enrichment in large-ion lithophile elements (LILE, e.g., Rb, K, Ba, and Sr) over HFSE (e.g., Ba/Zr = 2.2–4.7) and LREE (e.g., Ba/La = 17–61) and depletions in Nb relative to adjoining elements (Figures 5 and 6). There is also enrichment in LREE relative to middle REE (MREE; e.g., La/Sm = 1.4–5.1) and HREE (e.g., La/Yb = 2.1–12.8). The basalts and basaltic andesites from Ampiro-Malindang have the lowest LREE whereas the most evolved samples have the highest LREE and lowest MREE and HREE, forming a steep pattern from the LREE to the MREE and slightly concave to flat pattern from the MREE to HREE. Among the samples, lavas from Northern Negros, on average, have the highest contents of HFSE, Y, and REE at similar SiO2 levels; mafic lavas (<56% SiO2) from Ampiro-Malindang display the highest enrichment in alkali elements K and Rb (Figures 2 and 4). Compared to Southeast Sulu Sea Basin basalts, Negros Arc lavas are more enriched in highly incompatible elements but their HFSE, Y, and HREE contents overlap with the basin lava concentrations (Figures 5 and 6).

Figure 5.

Normal-MORB normalized incompatible element and chondrite-normalized REE plots for Northern Negros and Cuernos lavas. Normal-MORB and chondrite normalizing values from Sun and McDonough [1989] and Evensen et al. [1978], respectively.

Figure 6.

Normal-MORB-normalized incompatible element and chondrite-normalized REE plots for Ampiro and Malindang lavas. Normal-MORB and chondrite normalizing values from Sun and McDonough [1989] and Evensen et al. [1978], respectively.

4.3. Sr, Nd, and Pb Isotopes

[19] New Sr, Nd, and Pb isotopic ratios of lavas from Negros Arc and those reported by Sajona et al. [2000a] for three Ampiro-Malindang lavas are in Table 2 and shown in Figures 7a and 7c, together with data for the Southeast Sulu Sea Basin crust [Spadea et al., 1996] and bulk sediment being subducted along the Negros Trench [Solidum, 2002].

Figure 7.

Plot of 87Sr/86Sr versus 143Nd/144Nd (A), 206Pb/204Pb versus 207Pb/204Pb (B) and 208Pb/204Pb (C) of Negros Arc lavas. Also plotted are data for Southeast Sulu Sea basalts [Spadea et al., 1996] and bulk sediment likely subducted at Negros Trench. Cross in top panel represents analytical error. NHRL is Northern Hemisphere Reference Line [Hart, 1984]. Crosses represent analytical error.

Table 2. Sr, Nd, and Pb Isotopic Ratios of Negros Arc Lavasa
SampleLocationRock Type87Sr/86Sr143Nd/144Nd206Pb/204Pb207Pb/204Pb208Pb/204Pb
  • a

    Strontium isotopic ratios were measured by dynamic multicollection, fractionation corrected to 86Sr/88Sr = 0.1194 and normalized to 87Sr/86Sr = 0.71025 for NBS 987. Analytical uncertainty for 87Sr/86Sr measurements is ±0.000018 but in-run precisions were better than ±0.000012. Neodymium isotopic ratios were measured in oxide form by dynamic multicollection, fractionation corrected to 146NdO/144NdO = 0.72225 (146Nd/144Nd = 0.7219) and are reported relative to 143Nd/144Nd = 0.511850 for the La Jolla Standard. Analytical uncertainty for 143Nd/144Nd measurements is 0.000014 (0.3 E units) but in-run precisions were better than 0.000010. Lead isotopic ratios were measured by static multicollection and are reported relative to the values of Todt et al. [1996] for NBS SRM 981; the long-term errors measured for this standard are ±0.008 for 206Pb/204Pb and 207Pb/204Pb, and ±0.024 for 208Pb/204Pb. Total procedural blanks are negligible: <10 picograms (pg) for Nd, <35 pg for Sr, and <60 pg for Pb. Some of the data have been previously presented by Castillo [1996]. Asterisk denotes samples with adakite-like features.

  • b

    Data from Sajona et al. [2000a]. Their reported Sr isotopic values were normalized to 87Sr/86Sr = 0.71025 for NBS 987, similar to the rest of Negros Arc samples.

Northern Negros
CAN-09SilayBasaltic andesite0.7035440.51297518.28515.53738.295
CNV-2CanlaonBasaltic andesite0.7035700.512977   
CRN-1CuernosBasaltic andesite0.7036590.513061   
CRN-8BCuernosBasaltic andesite0.7035260.513037   
AM-50AmpiroBasaltic andesite0.7039440.513026   
AMP-29BbAmpiroBasaltic andesite0.7039590.51302718.21915.52138.150
PH93-59bMalindangBasaltic andesite0.7037550.51302318.20815.54138.264

[20] Northern Negros, Cuernos, and Ampiro-Malindang volcano groups display a range of Sr and Nd isotopic ratios although, in general, ratios within each group greatly overlap (Table 2; Figure 7a). There are also variations between groups, which are more evident in the Nd isotopes. The 87Sr/86Sr ratios of Northern Negros lavas are similar to those of Cuernos and Ampiro-Malindang but the former have distinctively lower 143Nd/144Nd. On average, Cuernos lavas have lower 87Sr/86Sr and higher 143Nd/144Nd than Ampiro-Malindang lavas. Compared to Southeast Sulu Sea Basin crust, the 87Sr/86Sr ratios of Negros Arc lavas fall between relatively fresh basin lavas and seawater-altered basin samples. Nd isotope ratios for Cuernos and Ampiro-Malindang lavas overlap with the middle range of values of the basin lavas whereas those from Northern Negros partly overlap the lower range with one sample having distinctly lower 143Nd/144Nd.

[21] Covariations of 207Pb/204Pb and 208Pb/204Pb with 206Pb/204Pb similarly show clear differences between groups of volcanoes in the arc (Figures 7b and 7c). Northern Negros lavas have the highest Pb isotopic ratios whereas the single lava from Cuernos analyzed for Pb isotopes has the lowest values measured. Ampiro-Malindang lavas have intermediate Pb isotope ratios. Compared to Southeast Sulu Sea crust, the Pb isotopic ratios for Cuernos lava are similar to those of the basin basalts but those of Northern Negros lavas are distinctly higher. The 207Pb/204Pb and 208Pb/204Pb ratios of Ampiro-Malindang lavas overlap with those of the basin lavas but their 206Pb/204Pb ratios are slightly higher. Collectively, Negros Arc lavas define an array from low Pb isotope values similar to the Southeast Sulu Sea Basin lavas to higher ratios trending toward sediment values.

5. Discussion and Interpretation

5.1. Geochemical Variations in Negros Arc Lavas

[22] In the Negros Arc, crystal fractionation appears to have been the dominant mechanism in generating the wide range of lava types based on the following lines of evidence. (1) All samples are porphyritic; lava sequences from all volcano groups exhibit systematic change in phenocrysts with varying SiO2. (2) Lava sequences from each group show clear major and trace element trends relative to SiO2 (Figures 3 and 4), several trends have inflections that suggest crystal-melt fractionation [e.g., Wilson, 1989; Tatsumi and Eggins, 1995]. Variations in the trends can be attributed to changes in phenocrysts. Rapid decrease in MgO and Ni in basalts of Ampiro-Malindang reflects the role of olivine fractionation in the most mafic compositions. The decline in CaO and V in the basalts and basaltic andesites from all groups can be explained by clinopyroxene fractionation. Essentially flat trends for Al2O3 and Sr for andesites and dacites, which differ from positive trends observed in more mafic lavas of Ampiro-Malindang, suggest variable plagioclase fractionation with or without accumulation. Decline in P2O5 starting at 56–57% SiO2 reflects onset of apatite fractionation. The relatively flat trend of Zr and negative trends of Nb, Y, and Yb at SiO2 contents >56–57%, which contrast with positive trends in more mafic lavas, coincide with the ubiquity of hornblende in evolved lavas. Because these elements are compatible with hornblende in silicic magmas [Pearce and Norry, 1979; Bacon and Druitt, 1988; Sisson, 1994], the changes in their trends are likely related to the dominant role of hornblende fractionation in the evolution of the high-silica lavas. Decrease in TiO2 contents from basaltic andesites to dacites can be related to minor involvement of magnetite in the fractionating assemblage. (3) On process identification diagrams such as Ba versus Ba/La (Figure 8), lavas from each group define horizontal to subhorizontal arrays consistent with fractional crystallization [Allegre and Minster, 1978; Davidson et al., 1988]. (4) The fairly limited range of isotopic ratios of lavas with varying SiO2 contents from Cuernos and Ampiro-Malindang (Table 2) is compatible with their derivation from common sources and differentiation through crystal fractionation.

Figure 8.

Ba versus Ba/La plot of Negros Arc lavas. Filled symbols are basalts and basaltic andesites (SiO2 < 56%) and open symbols are andesites and dacites (SiO2 > 56%). Also shown are schematic trends during fractional crystallization, partial melting and mantle source enrichment. Composition of MORB mantle [Sun and McDonough, 1989] is plotted for comparison.

[23] Petrographic and geochemical evidence also suggests that other shallow level processes such as crystal accumulation and repeated influx of basaltic melt into a mass of fractionated lava have operated together with fractional crystallization during the evolution of Negros Arc magmas. Olivine accumulation in some mafic lavas of Ampiro-Malindang is indicated by overgrowth textures on phenocrysts and elevated MgO and Ni contents that depart from fractionation trends defined by the majority of lavas. Plagioclase accumulation is common in the high silica lavas and is petrographically manifested by glomerocrysts of complexly zoned plagioclases. Moreover, simple plagioclase fractionation typically causes negative Eu anomaly in chondrite-normalized REE patterns as Eu is strongly partitioned into plagioclase relative to other REE [Vukadinovic, 1993]. Combined with the elevated but flat Al2O3 and Sr trends in high SiO2 lavas (Figures 3 and 4), the absence of either negative or positive Eu anomaly of several andesites and dacites in Negros Arc (Figures 5 and 6) is consistent with plagioclase accumulation to parental melts of these lavas. Influx of new melt is evidenced by resorbed and sieve-textured phenocrysts.

[24] The presence of ultramafic xenoliths in the high-silica (SiO2 > 58%) andesites and dacites of Ampiro-Malindang may be responsible for their slightly elevated MgO, Ni, and Cr contents and the presence of xenocrystic olivine grains in a few samples. The xenoliths are presumed extracted from the peridotitic basement beneath Ampiro-Malindang, which is part of an oceanic terrane, and is likely depleted in incompatible elements but enriched in compatible elements. All Ampiro-Malindang lavas may have interacted with this basement rock, but xenoliths in the silicic lavas are discernable because of the inability of silicic melt, relative to mafic melt, to assimilate ultramafic rock. Note that the presence of a small amount of the ultramafic xenolith certainly will increase compatible elements such as Mg, Ni, and Cr although it may not significantly decrease highly incompatible abundances in the silicic lavas. Isotopic ratios of some high-silica lavas from Ampiro-Malindang are not significantly different from more mafic counterparts (Table 2). This suggests that either the host lava has an isotopic signature fairly similar to the enclosed peridotite or that incompatible element concentrations (including Sr, Nd, and Pb) in the host lava were high enough to render the isotopic ratios insensitive to the presence of xenolith.

[25] The above mentioned crustal level, open system magmatic differentiation processes have likely generated the wide range of lavas in Negros Arc volcanoes, and could broadly explain major and trace element variations between lava types. However, these cannot fully account for large differences in incompatible element concentrations between lavas with similar SiO2 levels from the different groups or even from within the same group (Figure 4). For example, Ba and Rb vary in basaltic andesites of Northern Negros by factors of 2 and 5, respectively, and in basaltic lavas of Ampiro-Malindang by a factor of 3. As previously noted, Northern Negros lavas, on average, are more enriched in HFSE, Y, and REE at a given SiO2 level than lavas from other groups. These large variations in trace element in lavas that have undergone similar extent of differentiation likely reflect variable parental magmas. These in turn could be due to differences in the degree of partial melting or to variations in composition of the magma source.

[26] The diversity of incompatible trace element ratios and radiogenic isotopic compositions in Negros Arc lavas, however, reflects control by magma source heterogeneity on compositional variations of parental magmas. Ratios of highly incompatible trace elements (e.g., LILE/LREE) are relatively invariant during partial melting of the mantle and are also generally retained in subsequent shallow level crystal fractionation [Sun and McDonough, 1989; Elliot et al., 1997]. For example, Ampiro-Malindang lavas have higher Ba/La ratios than those from Cuernos and Northern Negros (Figure 8). The Ampiro-Malindang lavas with high Ba/La have low La/Yb (<5) ratios whereas the Cuernos and Northern Negros lavas with low Ba/La have high La/Yb (>5) ratios. Thus the difference in Ba/La ratios between Negros Arc volcano groups is not due to differences in partial melting of a uniform source, but to differences in magma source compositions along the arc. As Negros Arc lavas collectively have higher Ba/La ratios than a typical unmetasomatized mantle (e.g., MORB-source type mantle Ba/La = 2.5 [Sun and McDonough, 1989; Tatsumi, 2000]), the variable Ba/La ratios exhibited by the lavas strongly imply variable enrichment of the mantle wedge in either Ba or both Ba and La. The variable Sr, Nd, and Pb isotopic ratios provide independent and clear evidence of magma source variability within each group and between groups of volcanoes in Negros Arc. Lavas from within each group exhibit a small but significant range in their isotopic ratios and each group is distinct in their Nd and Pb isotopic values (Figures 7a to 7c).

5.2. Constraining the Origin of Negros Arc “Adakites”

5.2.1. Adakite and Nonadakite Characteristics in Negros Arc Silicic Lavas

[27] Several high silica (>56% SiO2) lavas from Cuernos and Ampiro-Malindang possess some of the geochemical features of rocks termed adakites, which are explained as direct melts of subducted oceanic crust [Defant and Drummond, 1990, 1993]. These Cuernos and Ampiro-Malindang lavas have high Sr concentrations (500–680 ppm), low Y and Yb contents (11–18 ppm and 0.6–1.7 ppm, respectively), and moderate Sr/Y ratios (27–52) (Table 1; Figure 9) Sajona et al. [1993, 1994, 2000a, 2000b]. However, they lack other chemical features critical for adakites. Negros Arc silicic lavas have low La/Yb ratios (< 20), high Ba/La and Pb/Ce ratios, and relatively flat to slightly concave HREE patterns (Figures 5 and 6), unlike typical adakites. Cuernos silicic lavas, alleged to have adakite-like characteristics, have low MgO (average ∼2.5%), Ni (∼7 ppm), and Cr (∼17 ppm) similar to nonadakitic lavas of Northern Negros. Only those from Ampiro-Malindang have MgO ∼3.5%, Ni ∼41 ppm, and Cr ∼100 ppm similar to other proposed adakites [e.g., Martin, 1999]. We emphasize, as discussed previously, that this enrichment can be explained by incorporation of ultramafic basement by silicic differentiates from more mafic magmas. It is important to note that Cuernos and Ampiro-Malindang silicic lavas with adakite features have Sr, Nd, and Pb isotopic ratios that fall within or close to the range of values in Southeast Sulu Sea Basin basalts (Figure 7); this would be consistent with an origin by partial melting of the subducted crust. However, their isotopic compositions can also be explained by derivation from basaltic magmas generated from the mantle wedge having isotopic characteristics similar to the Southeast Sulu Sea basalts. This interpretation is supported by their isotopic similarity to associated mafic lavas (Table 2), which are not adakitic.

Figure 9.

Plots of (a) La/Yb versus Yb and (b) Sr/Y versus Y for silicic (SiO2 > 56%) Negros Arc lavas. Also shown are batch partial melting curves for an average Southeast Sulu Sea crust with residues of varying amphibole:garnet:clinopyroxene proportions: A (75:7:18); B (55:15:30); C (15:45:40); D (0:15:85); E (0:20:80); F (0:50:50). Numbers adjacent to tick marks in top plot (Figure 9a) are melting percentages. Typical adakite field in lower plot (Figure 9b) is from Defant et al. [1991b].

5.2.2. Melting Model Calculations

[28] To test the proposed origin for some of the silicic lavas in Cuernos and Ampiro-Malindang by melting of basaltic crust, we compare concentrations and ratios of several key trace elements in Negros Arc lavas with those calculated for partial melts of the subducted Southeast Sulu Sea Basin crust using the batch equilibrium melting equation of Shaw [1970]. Source compositions used in modeling are average concentrations of the basin's basaltic lavas reported by Spadea et al. [1996]; mineral-melt partition coefficients used are in Table 3.

Table 3. Source Basaltic Crust Compositions and Mineral-Melt Partition Coefficients Utilized in Partial Melting Models
Average CompositionSrYLaYbBaPbCeNb   
SE Sulu Sea Basalta13220.43.282.07250.928.842.5   
Partition Coefficientsb
Rutile       26.544  

[29] Figures 9a and 9b show variations in Negros Arc silicic lavas of La/Yb versus Yb and Sr/Y versus Y; these variations are criteria commonly used to distinguish melts of subducted crust from other arc lavas [Defant and Drummond, 1990; Martin, 1999]. Superimposed on these plots are calculated partial melting curves for the Southeast Sulu Sea basaltic crust at the appropriate pressure (depth) conditions beneath the arc volcanoes. Low Yb values and La/Yb ratios in most silicic lavas from Cuernos and Ampiro-Malindang can be reproduced by partial melting of the crust with residues containing variable proportions of hornblende (amphibole), garnet, and clinopyroxene (Curves A, B, and C; Figure 9a), similar to models proposed by other studies in this area and elsewhere [e.g., Sajona et al., 2000a, 2000b; Defant and Drummond, 1990, 1993]. However, calculated melting curves for the majority Negros Arc silicic lavas require significant amounts (>55%) of residual amphibole in the subducted crust beneath the Negros Arc volcanoes. Experimental data indicate that amphibole stability in subducted basaltic crust is unlikely to exceed depths of ∼80 to 85 km (pressure of ∼25 to 26 kbar [Rushmer, 1991; Tatsumi and Eggins, 1995; Drummond et al., 1996]). Cuernos, which is closer to the trench than Ampiro-Malindang, is already ∼100 km above the slab according to Bautista [1996].

[30] Model calculation results for eclogite residues composed essentially of garnet and clinopyroxene (Curves D, E, and F) are also problematic because these are inconsistent with experimental results of dehydration melting of basaltic rocks at high pressures. Rapp et al. [1991, 1999; Rapp and Watson, 1995] and Sen and Dunn [1994] have produced adakitic magmas by 10–40% partial melting of basaltic compositions at pressures ≥20 kbar leaving an eclogitic residue of 40–60% garnet, 60–40% clinopyroxene and accessory rutile. However, Yb and La/Yb values in some of Cuernos and Ampiro-Malindang silicic lavas within the 10–40% experimental melting range have only ≤ 20% residual garnet (Curves D and E). A few silicic lavas from Ampiro-Malindang that can be matched with a reasonable residual eclogite mineralogy (e.g., 50% garnet, Curve F) require very high degrees (≥50%) of partial melting, which should generate mafic melts (i.e., SiO2 < 56% [Rapp and Watson, 1995]).

[31] In general agreement, the Sr/Y and Y of silicic lavas in Negros Arc likewise cannot be reproduced by model partial melts of average Southeast Sulu Sea crust (Figure 9b). Only Curve D plots closest to the Ampiro-Malindang silicic lavas, but it has only 15% residual garnet. The other combinations of garnet amphibolite to eclogite residues (Curves A, B, C, E, F) cannot duplicate the Sr/Y and Y of Negros Arc silicic lavas. The low Y values of the model melts are due to the fact that their Y contents are significantly lower than the natural data at similar Sr/Y ratios, even though a low Y partition coefficient for garnet in silicic partial melts (∼4.7 [Jenner et al., 1993]) was used.

[32] Another useful geochemical tracer of melts of subducted crust is Nb/Ta ratio because it is not changed significantly from source value during moderate to high degrees of melting and subsequent low pressure crystallization of typical arc minerals [Stolz et al., 1996; Eggins et al., 1997; Munker, 1998]. Aqueous fluids equilibrated with residual rutile do not contain appreciable amounts of Nb and Ta and their Nb/Ta ratio are slightly decreased due to more effective retainment of Nb in rutile [Brenan et al., 1994]. In contrast, silicic melts in equilibrium with residual rutile carry small amounts of Nb and Ta and their Nb/Ta ratio are significantly increased due to higher compatibility of Ta in rutile [Green and Pearson, 1987; Jenner et al., 1993]. Thus as proposed by Stolz et al. [1996], partial melts of the oceanic crust with rutile in the residue would exhibit Nb/Ta ratios significantly greater than subducted crust and chondritic primitive mantle values (17.7 for N-MORB crust; 17.5 for primitive mantle [Sun and McDonough, 1989]). Although Prouteau et al. [2000] contend that Nb/Ta ratios cannot be an indicator for slab melting in subduction zones because of uncertainty in rutile-melt partitioning coefficients, several studies of arc lavas suggest otherwise. Metasomatism of the mantle wedge by high Nb/Ta oceanic crust melts best explain high Nb/Ta ratios in Sunda Arc basalts (up to 33 [Stolz et al., 1996]) and Cambrian arc lavas from New Zealand (up to 25 [Munker, 1998]). A similar mechanism of mantle metasomatism, involving partial melts of sediments equilibrated with rutile, has been proposed to explain the higher range of Nb/Ta values in Mariana arc lavas [Elliot et al., 1997; Peate and Pearce, 1998].

[33] The Nb/Ta ratios of adakite-like and nonadakite silicic and mafic lavas in Negros Arc are slightly less than the chondritic primitive mantle ratio of 17.5 (silicic lavas: 13.5–16; mafic lavas: 13.7–16.4; Table 4). These are lower than the 29–23 range of Nb/Ta values for 10–40% model partial melts of the Southeast Sulu Sea crust assuming an average Nb of 2.5, an N-MORB Nb/Ta of 17.7, and a rutile-bearing eclogite residue (garnet:clinopyroxene:rutile = 59.5:39.5:1, with rutile weight% following Stolz et al. [1996] and Rapp et al. [1999] and overall phase proportions similar to Rapp et al. [1991, 1999]). It is reasonable to assume the presence of rutile in the slab as this is a common accessory mineral in the residues of high pressure dehydration and partial melting experiments of basaltic rocks [e.g., Kogiso et al., 1997; Rapp et al., 1991, 1999; Sen and Dunn, 1994] and in exhumed subducted slabs [e.g., Sorensen and Grossman, 1989, 1993]. Moreover, the presence of Nb depletion relative to other incompatible trace elements in all of the analyzed Negros lavas (Figures 5 and 6) strongly suggest residual rutile in the subducted slab [Brenan et al., 1994; Stalder et al., 1998]. Reducing the proportion of rutile to about 0.25% would only lower values of model melts at 40% melting to about 21. As mentioned earlier, significant residual amphibole is unlikely at ∼100 km slab depth beneath the Negros Arc, but even assuming presence of ∼75% amphibole, which slightly prefers Nb to Ta [Klein et al., 1997], together with small amount of rutile (0.25%) would still lead to a high Nb/Ta ratio of ∼ 19 at 40% melting. Finally, incorporation of bulk sediment with a low Nb/Ta ratio (∼12.7) in partial melts of the crust to force their Nb/Ta to lower values is inconsistent with higher Nb values of both the model melts and sediment than the Negros Arc silicic lavas. Thus Nb/Ta data do not support an oceanic crust melt origin for the silicic lavas in Negros Arc. On the contrary, the similarity in Nb/Ta ratios of silicic lavas with associated mafic lavas strongly suggests a cogenetic relationship.

Table 4. Nb, Ta, and Nb/Ta Ratios for Negros Arc Lavas and Southeast Sulu Sea Sediments Based on ICPMS Analysis
SampleLocationRock TypeNbTaNb/Ta
  • a

    Samples with adakite-like features. Analytical uncertainty for Nb/Ta ratio is ±1.1.

Northern Negros
CNV-2CanlaonBasaltic andesite3.810.2714.3
CRN-1CuernosBasaltic andesite3.150.2214.2
CRN-8BCuernosBasaltic andesite2.110.1316.1
AM-50AmpiroBasaltic andesite2.590.1715.1
PH93-59MalindangBasaltic andesite2.060.1513.7
Southeast Sulu Sea
768B-38X-1, 34-40 Claystone8.440.6612.7
768C-33R-2, 22-28 Claystone6.690.5412.5
768C-72R-1, 82-88 Brown claystone3.760.2614.6

5.2.3. Likely Origin of the Adakite-Like Trace Element Characteristics of the Silicic Lavas

[34] A cogenetic relationship between mafic and silicic lavas with adakite features from Cuernos and Ampiro-Malindang is supported by their similar Sr, Nd, and Pb isotopic compositions. Geochemical and petrographic data suggest that the adakite-like trace element characteristics exhibited by several Cuernos and Ampiro-Malindang silicic lavas are simply products of crustal level magmatic differentiation processes on more mafic parents. Among the phenocrysts in Negros Arc silicic lavas, amphibole and minor apatite have higher partition coefficients for Y and HREE than for Sr and LREE. Thus fractional crystallization dominated by amphibole together with small amounts of apatite could easily account for the depletions in Y and Yb that give adakite-like signature. These effects, combined with variable plagioclase accumulation, give rise to high Sr/Y and moderate increase in La/Yb ratios of several Cuernos and Ampiro-Malindang silicic lavas (Figure 10). Although similar crystallization processes have occurred in all groups, the predominance of low Y and Yb silicic lavas in Cuernos and Ampiro-Malindang compared to Northern Negros are likely due to the lower Y and HREE compositions of their parental melts. Most of the mafic lavas in these groups have Y and Yb contents which are already within the adakite range (Figure 9).

Figure 10.

(a) La/Yb-Yb and (b) Sr/Y-Y variations in Negros Arc lavas. Filled symbols are samples with SiO2 < 56% and open symbols with SiO2 > 56%. Vector clusters show the effect of removal of plagioclase (P), olivine (Ol), orthopyroxene (Op), clinopyroxene (C), amphibole (Am), magnetite (M), apatite (Ap), and biotite (B). Fractionation curves for a crystallizing assemblage dominated by amphibole (92.25%) with minor plagioclase (5%), magnetite (2.5%), and apatite (0.25%) and using a basaltic andesite from Cuernos as parental melt. Numbers adjacent to tick marks indicate extent of crystallization. Distribution coefficients used to determine mineral vectors and fractional crystallization curves are from Solidum [2002, and references therein].

5.3. Enrichment of Negros Arc Mantle Source

5.3.1. Presubduction Mantle Wedge

[35] It is necessary to assess the nature of the mantle wedge prior to additions from the slab to properly evaluate subduction contributions to the enrichment of arc magmas. Many studies have shown that the MORB source, or the variably depleted residue of such a source following extraction of MORB or backarc basin basalts, are the typical mantle components above subducting plates [e.g., Ellam and Hawkesworth, 1988; Hawkins and Florendo, 1992; Woodhead et al., 1993; Tatsumi, 2000] although other works suggest the additional presence of an ocean island basalt (OIB)-like incompatible trace element enriched component [e.g., Morris and Hart, 1983; Stern and Ito, 1983; Reagan and Gill, 1989; Leeman et al., 1990; Castillo et al., 2002].

[36] Tectonic and paleogeographic reconstructions of the southeast Asian region [Hall, 1996; Schluter et al., 1996] indicate that the Southeast Sulu Sea backarc basin has not migrated since its formation and essentially has tapped mantle which is proximal to that beneath Negros Island and Zamboanga Peninsula. The basaltic lavas from the basin, therefore, provide constraint to the nature of the presubduction mantle below Negros Arc. Spadea et al. [1991, 1996] showed that lavas from Southeast Sulu Sea Basin, although exhibiting minor enrichment in LILE and LREE which may be attributed to fluid input from previously subducted crust, were derived from a slightly depleted normal MORB-source mantle with Indian MORB isotopic characteristics. The presence of similar mantle beneath Negros Arc is supported by comparable contents of HFSE, Y, and HREE in least evolved arc lavas of Cuernos and Ampiro-Malindang and basin basalts (Figures 5 and 6). These elements would not be mobile in subduction zone fluids [e.g., Brenan et al., 1994; Pearce and Peate, 1995; Johnson and Plank, 1999]. This inference is reinforced by similarities of Nd isotopic compositions between Cuernos and Ampiro-Malindang lavas and the basin basalts. In addition, Pb isotopic arrays of the arc lavas overlap with the Pb isotope values of the basin lavas (Figure 7). Relatively low abundances of TiO2 (<1 wt%) and HFSE (e.g., Nb >1 to <8 ppm) in basalts and basaltic andesites and strong negative Nb anomaly relative to LILE and LREE exhibited by all samples from the arc (Figures 5 and 6) are inconsistent with an OIB-like enriched component in the Negros Arc mantle wedge. Thus it is reasonable to assume a MORB-source mantle, similar to the source of Southeast Sulu Sea Basin lavas, as supra-subduction zone mantle beneath Negros Arc. It is also reasonable to attribute the LILE enrichment of the mantle wedge source of Negros Arc magmas mainly to subducted slab contributions.

[37] That the presubduction mantle source of Negros Arc was not depleted significantly relative to N-MORB mantle source is also evident in the Nb concentrations and Nb/Ta ratios of Negros Arc lavas. Low Nb/Ta ratios in Nb-poor arc volcanics have been interpreted to reflect removal of HFSE from the mantle wedge by previous melting events in mid-ocean ridges or backarc spreading centers [Eggins et al., 1997; Elliot et al., 1997; Woodhead et al., 1998; Class et al., 2001]. For example, frontal arcs with contemporaneous backarc spreading in the western Pacific region (e.g., Izu-Bonin, Marianas, New Britain) have lavas which exhibit extreme depletions in Nb (< 1 ppm) and wide range of Nb/Ta ratios from 16 to as low as 2 [Elliot et al., 1997; Woodhead et al., 1998; Hochstaedter et al., 2001]. In contrast, basalt and basaltic andesites in Negros Arc have >1 to <8 ppm Nb and a narrow range of higher Nb/Ta ratios between 13.7–16.4 (Table 4). Nb/Ta values of Negros Arc lavas are only slightly lower than chondritic primitive mantle ratio of 17.5 [Sun and McDonough, 1989] and comparable to the suggested upper mantle value of 15.5 [Jochum et al., 1997].

[38] Knittel and co-workers [Knittel and Oles, 1995; Knittel and Yang, 1998] have also noted that lavas from Luzon Arc, north of the study area and part of the discontinuous arc system along the western margin of the Philippines like the Negros Arc, reflect a presubduction mantle source which has trace element characteristics comparable to MORB sources. The presence in Negros and Luzon arcs of mantle less depleted than most other arcs in the western Pacific is likely due to the absence of active backarc spreading in the Philippine arc systems.

5.3.2. Subduction Components

[39] Because Negros Arc lavas are variably evolved, we use a combination of radiogenic isotopes and incompatible element ratios, which do not change appreciably during magmatic processes [e.g., Elliot et al., 1997; Shinjo et al., 2000], to constrain the characteristics of their metasomatized mantle source, and to identify the sources (basaltic crust versus sediment) and transport media (aqueous fluid versus melt) responsible for enrichment of this source. Inasmuch as certain elements (e.g., HFSE, Y, MREE, HREE) become compatible during late stage crystallization of Negros Arc silicic lavas, our interpretations using these elements are mainly based on data for basalt and basaltic andesite lavas (SiO2 < 56%).

[40] Dehydration and melting experiments of typical slab minerals, basaltic rocks, and sediments, combined with observations in arc lavas, indicate selective transport of elements by aqueous fluids and melts. LILE (e.g., Rb, K, Ba, Pb, and Sr) are highly mobile, LREE (e.g., La, Ce) are slightly mobile whereas other REE, Y, and HFSE are least mobile in fluids [e.g., Tatsumi et al., 1986; Brenan et al., 1995c; Pearce and Peate, 1995; You et al., 1996; Ayers, 1998; Stalder et al., 1998; Johnson and Plank, 1999]. Elements relatively immobile in fluids (e.g., HFSE) are more effectively transferred from the slab to the wedge by melts [e.g., Pearce and Peate, 1995; Elliot et al., 1997; Johnson and Plank, 1999]. Thus significant fractionation between LILE relative to REE or HFSE can be used to track aqueous fluid input whereas variations between REE and HFSE may be utilized to track addition by melts.

[41] Negros Arc basalts and basaltic andesites have Ba/La and La/Sm compositions significantly higher than unmodified mantle (Figure 11a). Their Ba/La ratios are likewise higher than materials likely to be subducted beneath the arc but their La/Sm values fall between those for the less LREE-enriched basaltic crust and strongly LREE-enriched bulk sediment. Negros Arc lavas manifest an inverse variation, with Ampiro-Malindang lavas exhibiting higher Ba/La but lower La/Sm ratios than those from Cuernos and Northern Negros; this suggests subarc mantle metasomatism by two distinct slab contributions. Ampiro-Malindang lavas with the least LREE enrichment, not only have the highest Ba/La but also the most elevated Pb/Ce. These high LILE/LREE ratios (e.g., Ba/La up to ∼ 60) accompanied by relatively low La/Sm ratios strongly indicate aqueous fluid additions derived from subducted basaltic crust. The trend of the data array toward low Ba/La and high La/Sm ratios close to those of bulk sediment suggests a sedimentary material input. Elliot et al. [1997] observed similar inverse relationship between Ba/La and La/Sm in Mariana Arc and attributed this variation to inputs of sediment and aqueous fluids derived from basaltic crust to the mantle source of Mariana lavas.

Figure 11.

(a) Ba/La versus La/Sm of Negros Arc basalts and basaltic andesites. Also shown are the compositions of average Southeast Sulu Sea crust, bulk sediment and MORB-source type mantle [Sun and McDonough, 1989]. (b) Ba/La versus Nd/Zr of Negros Arc basalts and basaltic andesites. Also shown are compositions of average Southeast Sulu Sea crust, MORB-source type mantle, bulk sediment and its partial melt calculated by 5% batch melting and assuming a residual eclogite mineralogy (garnet/clinopyroxene ∼ 60:40). Mineral-melt partition coefficients used for estimating sediment melt composition given in Tables 3 and 5. Note that a partial melt of the bulk sediment is required in order to fractionate Nd/Zr to values higher than exhibited by Northern Negros lavas.

[42] The sediment signature from the slab is not being added to the subarc mantle source of Negros Arc magmas as an aqueous fluid, otherwise the inverse variation between Ba/La and La/Sm would be negated (Figure 12a). Bulk sediment (i.e., solid) may also be an unsuitable end-member component because many investigators consider that transport through a liquid phase is the most efficient way of transferring sediment materials into the mantle wedge [e.g., Elliot et al., 1997; Hoogewerff et al., 1997]. More important, ratios of elements thought to be less mobile in fluids (e.g., Nd and Zr) indicate how the sediment signature is being transferred in the Negros Arc. Plot of Ba/La versus Nd/Zr (Figure 11b) shows that Negros Arc mafic lavas manifest a triangular array consistent with mixing involving the mantle wedge and two subduction inputs. The high Ba/La but low Nd/Zr ratios exhibited by several Ampiro-Malindang lavas point to an aqueous fluid addition from the basaltic crust. Northern Negros lavas, which reflect the strongest contribution from sediments, define an array displaced toward high Nd/Zr values which are significantly greater than bulk sediment Nd/Zr ratio. Because Nd is more incompatible than Zr during melting, the added sediment component is more likely a partial melt.

Figure 12.

Ba/La versus 143Nd/144Nd variations in Negros Arc lavas. In Figure 12a, also shown are compositions of average Southeast Sulu Sea crust, bulk sediment and MORB-source type mantle wedge with Nd isotopic characteristics similar to most MORB-like Southeast Sulu Sea basalt [Spadea et al., 1996]. Trends A and B reflect distinct contributions from both the subducted basaltic crust and sediment. Figure 12b illustrates a three-component mixing model involving crust-derived aqueous fluids and sediment melt additions into the mantle wedge. Solid curves are binary mixtures and numbers on tick marks indicate weight percentage of one-end-member. Dashed curves are ternary mixtures between mantle and variable proportions of the two subduction components and numbers on tick marks correspond to total amount of slab components added.

[43] That sediment melts and aqueous fluids derived from the oceanic crust can account for the enrichment of Negros Arc mantle is again illustrated in Ba/La versus 143Nd/144Nd plot (Figure 12a). Negros Arc lavas form a triangular array diverging from the low Ba/La, high 143Nd/144Nd values of MORB-source type mantle wedge. Ampiro-Malindang lavas (array A), trend to higher Ba/La ratios with slight decrease in Nd isotope ratios that are comparable to the high 143Nd/144Nd of the average Southeast Sulu Sea crust. This array largely reflects aqueous fluid contributed by subducted basaltic crust. Northern Negros lavas (array B) trend to lower Nd isotopic ratios with smaller increase in Ba/La ratio. The negative correlation suggests that a low 143Nd/144Nd component, different from fluids dehydrated from the oceanic crust, is being added to the mantle wedge. Clearly, this is derived from subducted sediment because Northern Negros trends to 143Nd/144Nd ratio lower than those of Southeast Sulu Sea lavas (Figures 7a and 12a). Furthermore, a sediment component is consistent with the displacement of Northern Negros lavas from the fields of Southeast Sulu Sea crust toward higher values in Pb-Pb isotopic plots (Figures 7b and 7c).

5.3.3. Variable Slab Contributions

[44] We postulate that the composition of the presubduction subarc mantle wedge beneath Negros Arc's short extent is broadly similar along its length. If true, it is apparent from Figures 11 and 12 that the compositional differences between lavas from the different groups of volcanic centers must reflect variable proportions of sediment melt and oceanic crust-derived fluid fluxes from the subducting slab.

[45] The relative contributions of the subduction components to the mantle source of Negros Arc magmas were estimated from covariation of Ba/La versus 143Nd/144Nd (Figure 12b) using the following assumptions. (1) The mantle wedge has MORB-source trace element composition similar to those calculated by Tatsumi [2000] from average N-MORB [Sun and McDonough, 1989] and a Nd isotopic ratio equivalent to the most radiogenic, age-corrected value exhibited by the Southeast Sulu Sea crust [Spadea et al., 1996]. (2) 2 wt% fluid is expelled from eclogitic crust (garnet:clinopyroxene ∼ 60:40); the fluid composition was calculated by batch dehydration [e.g., Brenan et al., 1995a; Shinjo et al., 2000; Hochstaedter et al., 2001] of average Southeast Sulu Sea basalts [Spadea et al., 1996]. (3) Sediment component is calculated by 5% batch partial melting of bulk sediment (c.f. Class et al. [2001]; this is also constrained by the highest Nd/Zr ratio of the Negros Arc lavas). Mineral-fluid and sediment-melt partition coefficients and compositions of bulk subducted materials and mixing end-members used are in Table 5. Results of mixing calculations (Figure 12b) indicate that aqueous fluid from subducted crust contributes 60–95% of the slab components in Negros Arc lavas. The proportion of partial melt contribution of sediment is greater for Northern Negros lavas (20–40%) and less for Cuernos (10–15%) and Ampiro-Malindang (5–15%). Although Cuernos lavas have comparable sediment addition as Ampiro-Malindang lavas, it appears that the former were derived from mantle with lesser slab additions than for Ampiro-Malindang lavas. A greater sediment melt component in Northern Negros lavas is consistent with their slightly higher contents of the less fluid-mobile HFSE, Y, and REE than those from other groups.

Table 5. Partition Coefficients, Compositions of Materials Being Subducted, and End-Member Components Used in Mixing Model
 Partition Coefficients
  • a

    Mineral/fluid partition coefficients from Brenan et al. [1995c], Ayers [1998], Stalder et al. [1998], and Tatsumi [2000].

  • b

    Sediment/melt partition coefficients from Johnson and Plank [1999].

  • c

    Average composition of Southeast Sulu Sea basalt estimated from data reported by Spadea et al. [1996].

  • d

    Composition of bulk sediment likely subducted at Negros Trench from Solidum [2002].

  • e

    Mantle wedge with MORB-source composition from Tatsumi [2000] and Nd isotope equal to most radiogenic value for Southeast Sulu Sea basin basalt.

  • f

    Calculated basaltic crust-derived fluid composition assuming 2% fluid release from average Sulu Sea crust with garnet/clinopyroxene residue = 60:40.

  • g

    Calculated sediment melt composition assuming 5% melting of bulk sediment.

 Bulk Compositions
Southeast Sulu Sea crustc253.37.40.51305
 Compositions of Mixing Components
Mantle wedgee0.950.381.10.51310
Basaltic crust-derived fluidsf2601.530.170.51305
Sediment meltg33519.315.40.51236

6. Summary and Conclusions

[46] Negros Arc has six stratovolcanoes comprising the Northern Negros, Cuernos, and Ampiro-Malindang volcano groups. Our petrographic, major and trace element, and Sr-Nd-Pb isotopic ratio investigation suggests the following:

[47] 1. Negros Arc lavas are calc-alkaline to slightly shoshonitic basalts to dacites that exhibit typical arc compositional signatures. These have low TiO2 contents, enrichment in LILE relative to HFSE and REE, enrichment of LREE over MREE and HREE, and negative Nb anomalies.

[48] 2. The wide range of lava types in Negros Arc was likely generated by partial melting of the mantle wedge and crustal level differentiation dominated by fractional crystallization. Petrographic and geochemical data indicate additional influence of olivine accumulation in some mafic lavas from Ampiro-Malindang, plagioclase accumulation in silicic lavas, and inclusion of ultramafic xenoliths in high-silica andesites and dacites of Ampiro-Malindang.

[49] 3. Several silicic lavas in Cuernos and Ampiro-Malindang have high Sr contents, low Y and Yb concentrations, and moderate Sr/Y ratios comparable to adakites presumed to be derived from melting of subducted basaltic crust. However, their relatively flat HREE patterns and low La/Yb are unlike typical adakites; they have high LILE/LREE and slightly subchondritic Nb/Ta ratios which are comparable with associated mafic lavas, but opposite to that expected of melts of basaltic oceanic crust. Moreover, model partial melts of average Southeast Sulu Sea basaltic crust cannot satisfactorily reproduce the La/Yb-Yb and Sr/Y-Y relationships and the low Nb/Ta ratios of the silicic lavas. Fractional crystallization dominated by amphibole with small amounts of apatite, together with variable fractionation and accumulation of plagioclase, can easily reproduce the adakite-like geochemical features of the silicic lavas in the arc.

[50] 4. Lavas from within and between groups of volcanoes in the arc exhibit variations in incompatible element abundances at similar SiO2 contents, incompatible element ratios, and isotopic composition. These variations reflect inherent differences in geochemical enrichment of their MORB-source mantle wedge and are best explained by variable proportion of contributions from subducted basaltic crust and sediment. The basaltic crust input, which contributes a significant proportion of the slab components in Negros Arc lavas, is added by aqueous fluids. This is best reflected by Ampiro-Malindang lavas which exhibit high 143Nd/144Nd, largest Ba/La and Pb/Ce ratios, and lowest LREE enrichment. Subducted sediment contribution is being transferred most likely as a melt phase. This is best manifested in Northern Negros lavas which have the least radiogenic 143Nd/144Nd, most radiogenic Pb isotopic compositions, low Ba/La and Pb/Ce ratios, high La/Sm values, and slight enrichment in less fluid-mobile HFSE, Y, and REE.

[51] 5. A basaltic crust-derived melt component is not necessary for the enrichment of Negros Arc mantle. This is compatible with the lack of evidence for true subducted crust melts among the silicic lavas in the arc. These geochemical observations are consistent with the age (∼20–30 Ma) of the crust subducting beneath the arc. Melting of this crust is precluded by thermal models.


[52] We thank C. MacIsaac and K. Walda for help during the analyses, the Philippine National Oil Company and Philippine Institute of Volcanology and Seismology for providing some of the samples, and G. Bebout, E. Bourdon, and two anonymous reviewers for their helpful comments and suggestions to improve the manuscript. This work was funded by NSF grant EAR00-01212 to P.C.