Volatile and N isotope chemistry of the Molucca Sea collision zone: Tracing source components along the Sangihe Arc, Indonesia



[1] Volcanic gases are sensitive indicators of subduction processes and are used to evaluate the contributions from various source components. Nitrogen isotope systematics in particular are a valuable tool for determining the fate of organic matter in subduction zones. We present the first arc-wide survey of trace gas chemistry and nitrogen isotope variations from the Sangihe Arc of northeastern Indonesia, where the narrow Molucca Sea Plate subducts beneath the Sangihe Arc to the west and the Halmahera Arc to the east. Relative volatile abundances and N isotopic compositions of volcanic gases show systematic along-arc variations. Northern volcanoes exhibit low N2/He ratios and δ15N values (northern minima 542 and −7.3‰, respectively), indicating minimal addition of sediment to source magmas. In contrast, the southern part of the arc is characterized by high N2/He and δ15N values (southern maxima 2000 and +2.1‰, respectively), consistent with greater sediment contributions in the formation of the magmas. These observations can be correlated with the complex tectonic setting of the region whereby oblique collision between the two arcs has caused sediment obduction, decoupling the accretionary wedges from the underlying oceanic plate. In the north, where the collision is more developed, the lack of trace gas and N isotope evidence of sedimentary inputs to the source of arc magmas is consistent with enhanced sediment decoupling. In the south, where collision and accretionary wedge decoupling are not yet taking place, sediments would presumably subduct normally, in agreement with higher N2/He and δ15N values. Awu volcano, at the northernmost extension of the arc, is anomalous and exhibits high N2/He (2852) coupled with low δ15N (−3.3‰). These values are suggestive of increased slab contribution in the northernmost arc, possibly by slab melting as collision stalls the progress of the subducting plate and allows it to become superheated.

1. Introduction

[2] The study of volatile emissions from arc volcanoes and hydrothermal systems provides a wealth of information on subduction zone processes, including the geochemical cycling of material from the subducting slab and mantle wedge to the atmosphere. By combining relative abundance and isotopic data, the sources of volatiles (e.g., CO2 and N2) in volcanic arcs can be resolved into contributions from the mantle wedge, subducted oceanic sediments and underlying oceanic crust and the atmosphere [Marty et al., 1989; Sano and Marty, 1995; Sano and Williams, 1996; Fischer et al., 2002]. In turn, by combining information on volatile provenance with flux estimates from arc volcanoes, it is possible to investigate the mass balance of particular species at various subduction zones, and thereby to evaluate what proportion of volatiles supplied to the trench may get transferred back to the surface via arc-related magmatism versus what proportion is lost to the deeper mantle [Hilton et al., 2002].

[3] In this communication, we further consider the utility of volcanic gas emissions as a means to characterize and quantify various contributions to subarc magma sources. The focus is the Molucca Sea region of northeastern Indonesia (Figure 1), which lies at the junction of three major lithospheric plates (the Eurasian, the Australian, and Pacific plates), making this one of the world's most tectonically complex areas [Silver and Moore, 1978; Hamilton, 1979]. We adopt a dual approach, involving both volcanic gas chemistry and N isotope systematics, to evaluate the type and relative quantities of volatile inputs via the subducting slab to the active Sangihe Arc, one of a conjugate set of arcs in the process of colliding.

Figure 1.

Detailed tectonic map of Molucca Sea area, displaying geographic features described in text. Location is indicated by red box in inset generalized geologic map of the Indonesian region. Triangles are active and Quaternary volcanoes. Numbers next to volcano names correspond to sample IDs. Purple lines represent Benioff contours of descending slabs. Adapted from Silver and Moore [1978], Hamilton [1979], and Lallemand et al. [1998].

[4] We consider the chemistry of volcanic emissions, which comprise mostly H2O, CO2, sulfur-containing compounds (SO2 and H2S), and a variety of trace gases (e.g., N2, He, O2, H2, Ar, CH4, CO, HCl, HF, and NH3 [Giggenbach, 1996]). Nitrogen and helium in volcanic systems derive mostly from subducted sediments and the mantle wedge, respectively. Argon is not found in high concentrations in either of these reservoirs, and is therefore attributed to contamination by air during the sampling procedure. Together, these three volatile species form a powerful tool for differentiating the various origins of volatiles in arc-derived magmas.

[5] We combine this approach with the nitrogen isotope systematics of gas samples which also provide a valuable means of differentiating between the contributions from different volatile sources. This results from the observation that the upper mantle δ15N value (δ15N = {[(15N/14N)sample/(15N/14N)air] − 1} × 1000), as sampled at mid-ocean ridges, is commonly believed to be −5 ± 2‰ (although a recent study by Mohapatra and Murty [2004] indicates that the mantle may be as light as −15‰) whereas oceanic sediments have a distinctly different δ15N value, falling in the range of +7 ± 2‰ [Peters et al., 1978; Marty and Humbert, 1997; Sano et al., 1998, 2001; Fischer et al., 2002; Snyder et al., 2003; Sadofsky and Bebout, 2004].

[6] Our study, and the complementary He-C isotope study of Jaffe et al. [2004], reports the first arc-wide survey of chemical and isotopic compositions of volcanic and hydrothermal gas emissions from the Sangihe Arc. We assess whether along-arc variations in N-He systematics provide information on the composition of the slab-derived volatile input in this remote and tectonically complex region. A related aim is to investigate whether the N-He systematics can provide insight to the existing tectonic model of the Sangihe Arc whereby a buoyant forearc sedimentary mélange is proposed to “float” on the subducting oceanic plate resulting in obduction of the sedimentary wedge onto the arc, leaving the dense and sediment poor oceanic crust to be subducted into the mantle [Hamilton, 1979; Hall, 1996; Lallemand et al., 1998].

2. Geologic Setting

[7] The Molucca Sea Plate is trapped between the Australian Plate and the Philippine Plate, the former of which moves north (with respect to Eurasia) at 7–8 cm/yr, while the latter moves west (with respect to Eurasia) at 8–10 cm/yr and rotates clockwise about a pole located at 48°N, 157°E [Rangin et al., 1996]. Earthquake locations image the Molucca Sea Plate dipping to the west beneath the Sangihe Arc and to the east beneath the Halmahera Arc (Figure 2). The consumption of oceanic crust from both directions and eventual collision of the two arcs is known as the Talaud Orogeny [Hatherton and Dickinson, 1969; Moore and Silver, 1983]. Tectonic models suggest the arcs were 1000–1500 km apart in the mid-Pliocene, but they are currently separated by only approximately 250 km [Morris et al., 1983].

Figure 2.

Schematic cross section through the Molucca Sea, showing subduction of the Molucca Sea plate beneath the Halmahera arc to the east and the Sangihe arc to the west. Approximate location of the subducted ocean crust is inferred from earthquake data [Silver and Moore, 1978]. Red triangles denote the positions of the active volcanic arcs. Adapted from Silver and Moore [1978].

[8] Subduction along the Sangihe Trench was initiated shortly after 25 Ma when the collision of the Australian continental block with the Philippine Sea Plate caused broad rearrangement of Indonesian tectonics [Rangin et al., 1990; Lee and Lawver, 1995; Hall, 1996]. Timing of the onset of Halmahera subduction is more speculative; estimates based on plate reconstructions range from 7 Ma [Elburg and Foden, 1998] to 10 Ma [Moore and Silver, 1983; Lallemand et al., 1998] and 15–17 Ma [Hall, 1996; Baker and Malaihollo, 1996].

[9] Convergence of the two arcs has formed one large collisional complex composed of two compressed forearc sedimentary mélanges, approximately 150 km wide and perhaps 20 km thick [Silver and Moore, 1978; Hamilton, 1979; Hall and Nichols, 1990]. Seismic reflection profiles indicate that this collisional complex, too buoyant and voluminous to be subducted, is being obducted onto the facing island arcs such that the forearc sediments have become uncoupled from both arc plates and from the underlying oceanic plate [Silver and Moore, 1978; Simandjuntak and Barber, 1996]. This leaves the sedimentary wedge to “float” above the remnant old, dense oceanic crust, which is sinking into the mantle [Hamilton, 1979; Hall, 1996; Lallemand et al., 1998]. The arc systems first converged in the north, completing their collision in what is now Mindanao by early Miocene to 10 Ma [Moore and Silver, 1983; Hall and Nichols, 1990]. According to Hamilton [1979], the collision zippers southward such that the central portions of Sangihe and Halmahera are already fused, while the southern part has not yet collided. Geophysical data (gravity, magnetics, seismicity) show a remnant Molucca Sea slab deepening northward within the mantle, reflecting increasing decoupling and related slab sinking in the northern sections where collision is more developed [Lallemand et al., 1998].

3. Previous Petrological Studies

[10] This remote area has received little attention in the petrology literature; the majority of work to date has been in the form of geophysical surveys, which supply the bulk of what is known about the Molucca Sea collision zone. However, there are a small number of studies aimed at the petrological and geochemical characteristics of this unique setting [e.g., Jezek et al., 1981; Morrice et al., 1983; Morris et al., 1983; Tatsumi et al., 1991; Elburg and Foden, 1998].

[11] Petrological studies have characterized recent lavas from the Sangihe and Halmahera arcs as dominantly calcalkaline two-pyroxene andesites [Hamilton, 1979; Jezek et al., 1981; Morrice et al., 1983]. Tholeiitic suites are restricted to the southern volcanic front [Morrice et al., 1983]. Variations in along-strike composition are noted by several workers; the southern part of the arc seems to be more represented by olivine basalts to pyroxene andesites, whereas hornblende andesites are more common in the north [Jezek et al., 1981].

[12] The volcanic geochemistry of both arcs displays characteristics typical of a subduction zone, including high Al2O3 and low TiO2 concentrations, low MgO/FeO* ratios, high large ion lithophile element (LILE) and light rare earth element (LREE) contents with respect to high field strength elements (HFSE), and trends toward high 207Pb/204Pb and 208Pb/204Pb, usually explained by incorporation of sediment into the magma source [Jezek et al., 1981; Morris et al., 1983; Macpherson et al., 2003]. Wide ranges in Zr/Nb denote a variably depleted mantle wedge, most likely by prior melting [Macpherson et al., 2003]. LILE contents (K60) and LILE/HFSE ratios increase to the north, interpreted as decreasing degrees of partial melting northward along the arc caused by a gradual cessation of subduction due to collision [Jezek et al., 1981; Morrice et al., 1983].

[13] Though the composition and quantity of sediments in the Molucca Sea have not been studied (e.g., through ocean drilling), several models have been proposed based on plate reconstructions, field relationships, and geophysical and geochemical studies. Where collision has apparently uplifted the basin floor to form the Talaud, Mayu and Tifore islands, it can be seen that the oceanic crust is overlain by deformed Tertiary limestones, sandstones, and siltstones, ∼15 km thick as indicated by seismic refraction and gravity profiles [Silver and Moore, 1978; McCaffrey et al., 1980]. The terrigenous sediments, with variable volcaniclastic contents, could have eroded from a combination of the volcanic arcs, the Australia-New Guinea continental block, Irian Jaya, and/or the Philippines. Sediments have been accumulating in the Molucca Basin throughout the ∼750–1250 km of closure in the sea since the mid-Pliocene [Hamilton, 1979; Moore and Silver, 1983; Morris et al., 1983]. Along the central axis of the basin, sediments are associated with peridotites, serpentinites, and gabbros, interpreted as slivers of oceanic crust thrust onto the surface during collision [Silver and Moore, 1978]. Elburg and Foden [1998] assumed the chemical compositions of the sediments to be similar to those from other parts of the southwest Pacific [e.g., Gamble et al., 1996; Peate et al., 1997].

[14] Despite the abundant sedimentary material in the Molucca Sea available for subduction, there is little evidence for extensive involvement of sediments in the generation of the arc lavas. 87Sr/86Sr ratios, high in sediments, are low (0.7035–0.7042) in Sangihe Arc lavas compared to those measured in other arcs [Jezek et al., 1981; Morrice et al., 1983]. Lead isotope plots indicate an Indian Ocean mid-ocean ridge basalt (I-MORB) affinity for the mantle wedges beneath both arcs [Macpherson et al., 2003].

4. Sampling and Analytical Methods

[15] Volcanic gases were sampled at six active centers along the Sangihe Arc during July and August of 2001 (refer to Figure 1 for location map and to Table 1 for a summary of sample information). The Sangihe Arc comprises several offshore volcanic islands to the north of Sulawesi, of which the following were sampled: Awu Volcano (Sangihe Is.), and Ruang Volcano. Also sampled were the following locations on Sulawesi: Lokon Volcano, Leilam and Lahendong geothermal springs in the vicinity of Mahawu Volcano, Soputan Volcano, and Ambang Volcano.

Table 1. Summary of Sample Types and Locations (N to S) for Sangihe Gas Samplesa
LocationSample IDTypeLocationTemp., °CDate Collected, m.d.yy
Lat., °NLong., °EElev., m
  • a

    Locations are listed from north to south.

     Crater Loc. 1IND-15fumarole3.675125.454127896.68.3.01
     Crater Loc. 1IND-16fumarole""""8.3.01
     Crater Loc. 2IND-17spring"""
     Crater Loc. 1IND-19fumarole2.303125.36969197.98.8.01
     Crater Loc. 2IND-20fumarole2.304125.368688140.08.8.01
     Crater Loc. 2IND-20afumarole""""8.8.01
     Crater Loc. 1IND-1fumarole1.364124.800110996.17.27.01
     Crater Loc. 2IND-2fumarole""""7.27.01
     Crater Loc. 1IND-8fumarole0.748124.421131796.17.29.01
     Crater Loc. 1IND-9fumarole""""7.29.01

[16] Gases were sampled using evacuated, 200–250 mL Pyrex flasks containing approximately 60 mL of a 5N NaOH solution and sealed at a flow-through valve with a Teflon stopcock [after Giggenbach and Goguel, 1989]. The main constituents of volcanic gases (CO2, SO2, H2S, NH3, HCl and HF) are acidic and dissolve in the caustic solution, thereby enabling the trace gases in the sample (He, Ar, N2, H2, O2, CH4 and CO) to become concentrated in the headspace of the sampling flask.

[17] Concentrations of gases were determined in the Volcanic and Hydrothermal Fluids Analysis Laboratory at the University of New Mexico, according to methods of Giggenbach and Goguel [1989]. The headspace gases were analyzed using a Gow Mac gas chromatograph (GC) with two 5Å, 0.53 mm diameter molecular sieve columns: one is 30 m long, uses H2 as a carrier gas, is associated with an O2 trap at 250°C, and separates Ar, N2, CH4, and CO while the other is 50 m long and uses an Ar carrier gas for the separation of H2, He, and O2. A thermal conductivity detector (TCD) measures all of the above species. In addition, a flame ionization detector (FID) with a ruthenium methanizer is used to detect CH4 and CO. The O2 trap consists of a 30 cm 5 Å molecular sieve column to remove O2 by conversion to H2O at 250°C. Analyses of samples containing high O2 (>0.16 mmol/mol [Zimmer et al., 2004]) or repeated analyses of samples with moderate O2 contents (0.05–0.1 mmol/mol) can lead to the saturation of the O2 trap and loss of efficient O2 removal. In that case, some O2 is detected with Ar, resulting in anomalously low N2/Ar ratios (i.e., <40, the value for air saturated water).

[18] A variety of wet chemistry techniques was used to analyze the major gases dissolved in the caustic solution: NH3, total sulfur (H2S and SO2), CO2, HCl, and HF [Giggenbach and Goguel, 1989].

[19] The unabsorbed gases were also analyzed for nitrogen isotopes (after Fischer et al. [2002] and described by Zimmer et al. [2004]). Aliquots were taken from each sample using a high vacuum line with a cryogenic trap. These were analyzed for nitrogen isotope ratios on a DeltaPLUS XL mass spectrometer in the Stable Isotope Laboratory at the University of New Mexico. Nitrogen isotope compositions are expressed using the delta notation relative to the AIR standard (δ15NAIR = 0.0‰), determined by using aliquots of ambient air.

5. Results

[20] A total of fifteen gas samples were collected from ten sites, covering six volcanic centers of the Sangihe Arc. Total gas chemistry for all Sangihe samples is reported in Table 2 as mmol/mol of total vapor. Approximately 95–99.9% of the total gas is water vapor, as is typical of subduction zone volatiles. The remaining fraction is mostly made up of CO2, sulfur-containing compounds (SO2 and H2S), N2, and H2, with trace amounts of other species (noble gases, CH4, NH3). The following section details the gas chemistry of samples from each volcano. Information about volcanoes in the study area, is provided in Appendix A.

Table 2. Total Gas Chemistry and N and He Isotope Data for Sangihe Arca
Sample IDH2OCO2StbSO2H2SHClHFNH3HeArH2O2N2CH4COδ15N, ‰3He/4He (RC/RA)c
  • a

    Locations are listed from north to south. Reported in mmol/mol total vapor, unless otherwise noted; nd, not detected; nm, not measured.

  • b

    Total sulfur (SO2 + H2S).

  • c

    Data from Jaffe et al. [2004]; corrected for the effects of air-derived helium.

  • d

    Duplicate of previous sample.

  • e

    Due to O2 trap saturation after repeated analyses of high O2 samples, Ar value calculated from Ar = N2/83 (see text and Zimmer et al. [2004]).

   IND-15997.581.840.0280.0160.0130.1770.000660.0010<0.000010.002390.000050.077380.2890.000137<0.0000021.2 ± 0.41.00 ± 0.01
   IND-16d997.701.810.015nd0.0150.0910.000460.0012<0.000010.003030.000030.073730.3000.000110<0.0000021.2 ± 0.61.39 ± 0.02
   IND-17955.3444.370.037nmnm0.0230.002150.00010.0000610.002950.018490.032390.1740.000409<0.000002−3.3 ± 0.46.22 ± 0.08
   IND-19996.293.380.2470.1030.1440.0320.000530.00230.0000420.000430.01553<0.000040.033<0.000001<0.000001−7.3 ± 0.54.02 ± 0.04
   IND-20997.521.460.661nd0.6610.2720.001140.00110.0000240.000310.07010<0.000020.013<0.000001<0.000001−3.1 ± 0.46.99 ± 0.07
   IND-20ad996.033.010.6790.3110.3690.1660.000580.00080.0000270.000200.095010.000540.020<0.000001<0.000001−2.7 ± 0.4nm
   IND-1994.045.170.674n.m.0.6740.0360.000400.00310.0000780.000320.044200.000200.028<0.000001<0.000001−2.1 ± 0.57.12 ± 0.07
   IND-2995.743.430.689n.m.0.6890.0350.000400.00200.0000590.000700.03810<0.000030.056<0.000001<0.000001−4.0 ± 0.4nm
   IND-4979.2220.360.262n.m.0.2620.0490.000540.00380.0000570.001220.03720<0.000040.0640.000164<0.0000011.7 ± 0.47.33 ± 0.10
   IND-6989.449.420.822n.m.0.8220.0920.000680.00150.000200.001340.00170<0.000080.2200.000248<0.000002−0.73 ± 0.47.23 ± 0.14
   IND-7d987.5411.980.182n.m.0.1820.1150.000530.00170.000260.001330.00165<0.000080.1680.000211<0.0000020.69 ± 0.46.91 ± 0.10
   IND-13997.070.650.0080.008nd0.1140.00083nm<0.000010.037160.000050.365081.7180.000337<0.0000021.7 ± 0.31.11 ± 0.02
   IND-14999.350.220.009nd0.0090.1400.000360.0030<0.000010.00253e0.000050.055840.2100.000067<0.0000022.0 ± 0.50.93 ± 0.02
   IND-8990.119.080.687nd0.6870.0890.000660.00050.0000180.000180.00018<0.000020.0360.000035<0.0000012.1 ± 0.43.77 ± 0.1
   IND-9d990.588.580.684nd0.6840.0940.000600.00040.0000590.000060.00057<0.000030.0610.000061<0.0000012.1 ± 0.34.64 ± 0.07

5.1. Awu Volcano

[21] The gas sample (IND-17) collected in the crater lake has a high CO2 content (44.37 mmol/mol), and consequently a high CO2/St ratio of 1199 and a low HCl/CO2 ratio of 0.0005. The two fumarole samples (samples IND-15 and -16, duplicates) have almost no sulfur (0.028 and 0.015 mmol/mol, respectively), N2/Ar ratios of 121 and 99, and He contents that are below detection limits. The gas sample has a N2/Ar of 59, near that of groundwater, but significant He (N2/He = 2852), and He/Ar = 0.021 (N2/Heair = 148,900 and He/Arair ∼ 5 × 10−4). N isotope values are −3.3 ± 0.4‰ for sample IND-17 and 1.2 ± 0.4‰ and 1.2 ± 0.6‰ for samples IND-15 and -16, respectively.

5.2. Ruang Volcano

[22] In the crater of Ruang volcano, one fumarole was degassing diffusely through an area of hydrothermally altered material and had a lower temperature (97.7°C, sample IND-19) compared to high flow rate and higher temperature (140°C) fumaroles sampled nearby (samples IND-20 and -20a, duplicates). Notably, the higher temperature fumarole had higher St (0.661 and 0.679 mmol/mol, for IND-20 and -20a, respectively) and lower CO2/St ratios (2.2 and 4.4), in comparison with the lower temperature sample (St = 0.247 mmol/mol and CO2/St = 13.7). The N2/He ratios range from 542 to 786. N isotope values for IND-20 and IND-20a are −3.1 ± 0.4‰ and −2.7 ± 0.4‰, respectively. Sample IND-19 has the lowest δ15N value of all samples collected (−7.3 ± 0.5‰).

5.3. Lokon-Empung Volcanoes

[23] Two gas samples (IND-1 and -2) were collected from two fumaroles in the central crater. Both samples have high St (0.674 and 0.689 mmol/mol) and low CO2/St ratios (7.7 and 5.0), similar to the fumaroles at Ruang. The N2/He ratios of IND-1 and -2 are 359 and 949, respectively, and δ15N values are −4.0 ± 0.4‰ and −2.1 ± 0.5‰, respectively.

5.4. Lahendong

[24] Lahendong geothermal complex, is located in a maar near Lokon and Mahawu volcanoes and contains fumaroles, springs, and bubbling pools. One gas sample was taken from a strong, but wet, fumarole (IND-4). This sample has low St (0.262 mmol/mol) and also high CO2/St and low HCl/CO2 ratios of 77.7 and 0.002, respectively. A high CO2 content of 20.36 mmol/mol is consistent with interaction of the gases with the abundant hydrothermal water observed at the sampling site. The ratios of trace gases are N2/He = 1123 and He/Ar = 0.047, with a N isotope value of 1.7 ± 0.4‰. Samples (IND-6 and -7, duplicates) were collected from a small bubbling pool. These samples have higher CO2 contents compared to Lokon and Ruang samples: 9.42 mmol/mol and 11.98 mmol/mol for IND-6 and for IND-7, respectively. N2/He ratios are 1100 and 646, and N2/Ar are fairly consistent at 164 and 126 for IND-6 and IND-7, respectively. δ15N values are −0.73 ± 0.4‰ and 0.69 ± 0.4‰, respectively.

5.5. Soputan Volcano

[25] One gas sample was collected at the porous lava dome (IND-13) and another at a weak fumarole on the cinder cone (IND-14). Both have very low St and CO2; IND-13 has St = 0.008 mmol/mol and CO2 = 0.65 mmol/mol, while IND-14 has St = 0.009 mmol/mol and CO2 = 0.22 mmol/mol. This results in St/HCl and CO2/HCl that are orders of magnitude lower than other samples. The N2/Ar ratio is 46 for IND-13. For IND-14, the measured N2/Ar ratios is 12 but this value has been compromised due to analyses of high O2 samples prior to IND-14 (IND-13, 15 and 16) causing the decrease in the efficiency of the O2 trap and detection of O2 as Ar (see analytical methods). In Table 1 we report the calculated Ar content based on the N2/Ar value of Air (83) and the N2 content measured in the sample. Helium is below detection limit in IND-13 and IND-14. Nitrogen isotope values are 1.7 ± 0.3‰ and 2.0 ± 0.5‰, respectively.

5.6. Ambang Volcano

[26] Two gas samples (IND-8 and -9, duplicates) were collected. Ratios of the major gas species are CO2/St = 13.2 and 12.5, St/HCl = 7.72 and 7.28, and HCl/CO2 = 0.010 and 0.011 for IND-8 and -9, respectively. The N2/He ratios are 2000 and 1034, and δ15N values are 2.1 ± 0.4‰ and 2.1 ± 0.3‰, respectively.

6. Integrity of Data

[27] In order to investigate the volatile characteristics of the Sangihe subarc mantle and discuss the ultimate sources of the discharging gases, it is critical to evaluate which samples have been compromised by atmospheric interactions and/or contamination due to the nature of the hydrothermal features. At each sampling location, we targeted geothermal features with the highest temperatures and/or the highest gas flow rates in order to minimize contamination of the magmatic volatiles by air and air-saturated water. With respect to the influence of outlet temperature and type of hydrothermal feature (fumarole versus bubbling hot spring) on the N-He systematics, previous work has shown that (1) N isotopic values vary by <1‰ for samples collected at 950°C and 180°C at the same volcano [Fischer et al., 1998] and (2) variations in δ15N along strike of an arc are consistent with variations in the slab N contribution to the volatile emissions and seem independent of the type of hydrothermal feature sampled [Zimmer et al., 2004]. Variations in δ15N values through time are more difficult to assess due to the lack of time series data extending over periods longer than several months. However, repeated sampling at Poas Volcano, Costa Rica over a period of three months, showed that the δ15N values varied by ∼1‰ or less [Zimmer et al., 2004].

[28] Several samples (IND-13, -14, -15, and -16) contain no detectable He and consequently on a N2-He-Ar plot (Figure 3a) lie near the air and air-saturated water (asw; N2/Arasw = 40) compositions. Even using GC He detection limits (indicated as “<” in Table 2) to calculate maximum He*/Ar for these samples (He* henceforth denotes a maximum He value based on detection limits) their (He*/Ar) ratios are 1–2 orders of magnitude lower than those of other samples. This is consistent with severe air contamination of these samples and indicates that He partial pressure in the gas phase is too low for He to be detected by GC. Additionally, these same samples exhibit high N2/He* and O2 contents relative to other samples (Tables 2 and 3). Given this collection of characteristics, IND-13, -14 (Soputan), and duplicates IND-15, -16 (Awu) are considered to exhibit too much air-contamination to be useful for determining volatile sources.

Figure 3.

Triangular plots characterizing Sangihe gas chemistry in terms of (a) minor gas species N2, He, and Ar, with typical fields for mantle-derived (low N2/He) and arc-type (higher N2/He) samples shown, as well as the compositions of air, groundwater, and air-saturated water (asw) for reference, and (b) major gas species CO2, St, and HCl, showing typical hydrothermal (CO2/St > 10) and volcanic (CO2/St < 10) gas fields. After Giggenbach [1992].

Table 3. Data Used in Volatile Source Calculations Assuming a −5‰ Mantle End-Member: Measured δ15N Values and N2/He, as Well as Mantle, Sediment, and Air End-Member Valuesa
Sample IDδ15N, ‰N2/HeNitrogen Sourceb%Sc
  • a

    See text for references. Negative components were zeroed; numbers in parentheses are original values.

  • b

    Fractions of nitrogen derived from mantle (M), sediment (S), and air (A) sources were calculated according to δ15Nobs = fmδ15Nm + fsδ15Ns + faδ15Na, 1/(N2/He)obs = fm/(N2/He)m + fs/(N2/He)s + fa/(N2/He)a, and fm + fs + fa = 1 [Sano et al., 1998].

  • c

    Percent of nitrogen derived from sediments, 100 × S/(M + S).

  • d

    Calculated using GC He detection limits.

IND-1−4.0 ± 0.43590.4200.580
IND-2−2.1 ± 0.59490.1600.850
IND-41.7 ± 0.411230.130.330.5472
IND-6−0.73 ± 0.411000.1400.870
IND-70.69 ± 0.46460.230.510.2654
IND-82.1 ± 0.420000.070.350.5884
IND-92.1 ± 0.310340.140.400.4674.
IND-131.7 ± 0.3171,800d00.240.76100
IND-142.0 ± 0.521,875d0.000.290.7199
IND-151.2 ± 0.429,794d0.000.170.8399
IND-161.2 ± 0.632,609d0.000.170.8399
IND-17−3.3 ± 0.428520.0400.960
IND-19−7.3 ± 0.57860.1100.890
IND-20−3.1 ± 0.45420.2300.780
IND-20a−2.7 ± 0.47410.1700.840
Mantle−5 ± 2150100 
Sediment7 ± 210,000010 

[29] Although these samples effectively sample air, based on low He*/Ar, high O2 and N2/He*, the N isotopes are not that of air (0.0‰), but are positive (1–2‰). It is unclear why these samples do not have δ15N = 0.0‰. Possibilities to explain this observation include (1) a small isotopic fractionation between water and air in these systems and/or (2) small additions of surface-derived organic nitrogen. Deviance of the N isotopic composition from the air value for samples with severe air contamination by ±2‰ has been observed elsewhere (T. P. Fischer, unpublished results), and Mariner et al. [2003] suggest that air-saturated water has a δ15N value of +1‰. Carbon isotopic values of CO2 are also commonly observed to deviate from the air value (∼−8‰) in samples that are overwhelmed by air based on He/Ne ratios [Jaffe et al., 2004].

[30] Nitrogen isotopes of most duplicate samples are identical within error. Discrepancies do exist between duplicate pairs IND-6 (−0.73 ± 0.4‰) and IND-7 (0.69 ± 0.4‰) from Lahendong geothermal field. During collection of sample IND-6, spring water entered the sampling flask, introducing air-saturated water and possibly altering its composition. Samples IND-1 (δ15N = −4.0 ± 0.4‰) and IND-2 (δ15N = −2.1 ± 0.5‰) were collected at two different fumarole sites in Lokon crater. Sample IND-2 shows more air contribution than IND-1 based on He/Ar (0.24 for IND-1 and 0.08 for IND-2). A larger amount of air in sample IND-2 likely resulted in a decrease of the δ15N value toward 0.0‰.

[31] Volatiles from Soputan (IND-13), Aesoput (IND-14), and Awu (IND-15 and -16), have 3He/4He of 1 to 1.4 RA (where RA = 1.39 × 10−6, the 3He/4He ratio in air [Mamyrin et al., 1970; Clarke et al., 1976]) consistent with severe air contamination (Table 2) [also see Jaffe et al., 2004, Table 1]. Samples IND-8 and -9, duplicates from Ambang, have 3He/4He ratios of 3.8 and 4.6 RA, respectively and lower than the world-wide arc average of 5.4 RA [Hilton et al., 2002]. These low values are consistent with considerable addition of 4He from crustal sources. Nitrogen systematics do not seem to be as susceptible as He to modification by crustal contributions, however. For example, continental volcanic zones such as Valles Caldera (New Mexico) where He isotopic values of ∼6RA indicate some contribution of crustal He, still exhibit mantle-like δ15N and N2/He despite the crustal material through which the volatiles move [Fischer et al., 2003]. Snyder et al. [2003] show that δ15N correlates with CH4 in geothermal well gases from Central America suggesting a crustal source of N in these systems. The Sangihe data set, however, does not show such a correlation (Table 2).

[32] The N isotopic value of −7.3 ± 0.5‰ measured for sample IND-19 (Ruang) is an unusual outlier and anomalously lighter than the majority of volcanic volatile measurements in the literature, and also lighter than the anticipated −5 ± 2‰ value for the mantle [Marty and Humbert, 1997; Sano et al., 1998; Fischer et al., 2002]. This sample was collected in a zone of extensive hydrothermal alteration where weak and diffuse degassing occurred. In addition, IND-19 also has a low 3He/4He ratio (4.02 RC/RA) and may therefore be contaminated with a surficial component, possibly due to the very diffuse nature of that locality. Samples IND-20 and -20a taken concurrently from a stronger neighboring fumarole give identical δ15N values (within error) of −3.1 ± 0.4‰ and −2.7 ± 0.4‰. Sample IND-20 has a 3He/4He ratio of 6.99 RA.

[33] In summary, we conclude that a total of 5 samples have experienced sufficient modification to their relative N2-He-Ar abundances and nitrogen isotopes that they no longer reflect their primary magmatic character. These samples are IND-13, -14 (Soputan), IND-15, -16 (Awu) and IND -19 (Ruang). Samples IND-8 and -9 (Ambang) have 3He/4He ratios that show contamination by crustal 4He.

7. Discussion

[34] Samples collected in this study display the typical range of CO2/St and CO2/HCl ratios observed in low temperature (<200°C) gas emissions (Figure 3b). Variations in these reactive major components are caused by surficial and secondary processes such as partitioning of SO2 and HCl gas into a liquid phase to form H2SO4 and HCl, resulting in low pH hydrothermal waters, or the remobilization of previously precipitated elemental S [e.g., Giggenbach, 1996]. Inert or less reactive gases such as He, Ar, N2, and CO2 are less influenced by these secondary processes and have therefore commonly been used to characterize the volatile provenance in arc magmas, although it has been shown by Snyder et al. [2001] that CO2 contents in geothermal wells can be affected by the precipitation of calcite at depth.

7.1. Nitrogen-He-Ar Systematics of Volcanic Arcs

[35] Volcanic arc systems, being associated with sediment subduction, usually incorporate slab-derived C and N into their magma sources: this, in turn, is reflected in the composition of emitted magmatic volatiles [Marty et al., 1989; Kita et al., 1993; Giggenbach, 1996; Sano and Williams, 1996; Sano et al., 1998; Fischer et al., 1998]. In particular, N mainly traces slab-derived hemipelagic sediments [Fischer et al., 2002], while carbonate-rich sediments and altered ocean crustal basement are traced by C [Marty and Jambon, 1987; Sano and Marty, 1995; Shaw et al., 2003]. Helium, in contrast, is primarily mantle-derived [Craig et al., 1978] and it is recognized that the 3He/4He ratios of arcs approach those of MORB (8 ± 1 RA) [Poreda and Craig, 1989; Hilton et al., 2002]. Argon, on the other hand, is mainly of atmospheric origin and 40Ar/36Ar ratios in arc gases rarely exceed the air value of 295 by more than 10% [Sano et al., 2001]. Gases emitted from arc volcanoes are therefore characterized by a high N2 content with respect to He and Ar (N2/He = 1000–10,000; N2/Ar > 80), whereas other “mantle-derived” gases, emitted from MORB or ocean island basalts (OIB), lack sediment input and have N2/He < 100, and N2/Ar ratios similar to the air value (83) [Giggenbach, 1992; Marty and Zimmermann, 1999]. Any shift in N2/He to values higher than the mantle (as represented by MORB) may be due to addition of air derived N which would also result in a decrease in He/Ar from the expected mantle value of approximately 2 [Burnard, 1999] toward the air value of 0.00056. In order to distinguish between N addition from a sedimentary source and air, therefore, both N2/He and He/Ar need to be considered.

[36] There is a factor of ∼10 difference in solubility between He and Ar in basaltic melts [Jambon et al., 1986]: this could potentially lead to variations in He/Ar and N2/He ratios in the exsolved gas phase (assuming that N2 and Ar solubilities are similar in melts [Carroll and Webster, 1994]). It has been shown, however, that both N2/36Ar and N2/3He ratios correlate with δ15N values in arc gases [Sano et al., 2001]. This observation suggests that fractionation due to magma degassing is not the major factor influencing N2/He ratios, and variations in N2/He and δ15N can be interpreted to reflect the amount of hemipelagic sediments contributing to the volatile inventory of arc magmas [Sano et al., 1998, 2001; Fischer et al., 2002; Snyder et al., 2003; Zimmer et al., 2004].

7.2. Sources of Volatiles in the Sangihe Arc

[37] Variations in volcanic volatile systematics are useful in tracing magma source components. With respect to N2-He-Ar abundance ratios, the majority of (non air-contaminated) samples exhibit mantle-derived characteristics (N2/He < 1000); only samples from Ambang plot in the arc-type field (Figure 3a). This suggests that the N signal from sediments along the Sangihe arc is not as pronounced compared with other arcs world-wide [Sano et al., 2001]. Although five samples (two Soputan, two Awu, and one Ruang sample) are excluded because they show extreme air contamination or modification of their magmatic characteristics (section 6), the remaining ten samples can be used to determine volatile sources.

[38] Nitrogen in volcanic and geothermal volatiles derives from a combination of three sources: the mantle wedge (M), subducted hemipelagic sediments (S), and air (A) [Giggenbach, 1995; Sano et al., 1998; Fischer et al., 2002]. Three component mixing between these sources can be approximated following the approach of Sano et al. [1998]:

equation image
equation image
equation image

where obs is the observed ratio, f is the fraction contributed by M, S and A to the total nitrogen output, δ15Nm = −5‰ [Marty and Humbert, 1997], δ15Ns = +7‰ [Peters et al., 1978], δ15Na = 0‰, (N2/He)m = 150 [Marty and Zimmermann, 1999; Fischer et al., 2002], (N2/He)s = 10,000 [Sano et al., 1998] and (N2/He)a = 148,900 [Ozima and Podosek, 2002].

[39] The relative contributions of N to Sangihe gas samples from the three end-members are presented in Table 3. It is notable that many Sangihe samples give negative values for one of the components (IND-1, -2, -6, -13, -17, -19, -20, and -20a), suggesting that end-member values typically used for M, and S are not universally representative. In order to allow comparisons among all samples, the negative component was assumed to be equal to 0 and the sum of the remaining two components was defined to equal 1. Where this adjustment was performed, the original values (including any negative numbers) are provided in parentheses below the corrected values in Table 3.

[40] Three-component mixing between these end-members is shown in Figure 4 and provides a visual representation of the calculations presented in Table 3. Following this approach, the A component can be subtracted from the mixture so that samples comprise M + S only. This “removal” of air from measured values leaves an uncontaminated binary mixture between mantle and sediment. The percentage of sediment-derived nitrogen in a binary sediment-mantle mix, = (100 × S)/(M + S), is presented as %S in Table 3 [Fischer et al., 2002].

Figure 4.

N2/He and δ15N compositions of Sangihe gas samples compared to theoretical end-member compositions for mantle (M) with δ15Nm = −5‰ and N2/Hem = 150, hemipelagic sediment (S) with δ15Ns = +7‰ and N2/Hes = 10,000, and air (A) with δ15Na = 0‰ and N2/Hea = 148,900. See text for end-member composition references. Dotted lines represent air addition to mixtures of mantle and sediment, with percentages representing fractions of sediment-derived nitrogen in a mantle-sediment mix [after Fischer et al., 2002]. Symbols and labels as in Figures 3a–3b. Error bars represent calculated laboratory error.

[41] We note in Figure 4 that addition of sedimentary N to a mantle end-member composition shifts signatures toward higher N2/He and δ15N values. Air addition greatly elevates N2/He ratios; samples that project toward the A apex have experienced more contamination by air than those that lie close to the M-S mixing curve. It is evident that many samples do not conform to the theoretical mixing relationships of the generally accepted end-members; samples with negative components in Table 3 fall outside the mixing trajectories in Figure 4. Mohapatra and Murty [2004] have suggested that the currently accepted δ15N mantle end-member value of −5‰ may instead be the composition of a mixture, due to the approach of statistical averaging of a range of MORB data. Since MORBs contain a number of geochemical components, a method that averages them may be inappropriate for determining a true end-member. Alternatively, the most extreme values measured in MORB (∼−15‰) may be interpreted as the true mantle composition, while the heavier values represent dilutions of that end-member. Following this logic, and because many samples seem to trend toward lighter N isotopic values, mixing between M, S, and A components is re-calculated using δ15Nm = −15‰. These results are presented in Table 4. With this end-member, only two of the fifteen samples (IND-17, and -19) have negative components, suggesting that a lighter mantle end-member may be more appropriate for Sangihe samples than the generally accepted MORB value (Figure 5). To date, little is known about the heterogeneity of δ15Nm in the subarc mantle wedge. Note that the GC detection limit for He was used to generate minimum N2/He* ratios for the air-contaminated samples (IND-13, -14, -15, and -16), which approach the A apex in Figure 5. Because these samples plot close to air, they should not be used to evaluate volatile sources; although they show a very high S component (>99%S), this is simply a spurious effect of projecting severe air-contaminated samples onto the M-S mixing curve. Some samples approach the mantle end-member composition, indicating that they received relatively little sedimentary N addition to their sources.

Figure 5.

N2/He and δ15N compositions of Sangihe gas samples, identical to Figure 4 with the only difference being that the N isotopic value of the mantle end-member has been shifted to −15‰ in order to best model the data points.

Table 4. Data Used in Volatile Source Calculations Assuming a −15‰ Mantle End-Member: Measured δ15N Values and N2/He, as Well as Mantle, Sediment, and Air End-Member Valuesa
Sample IDδ15N, ‰N2/HeNitrogen Sourceb%Sc
  • a

    See text for references. Negative components were zeroed; numbers in parentheses are original values.

  • b

    Fractions of nitrogen derived from mantle (M), sediment (S), and air (A) sources were calculated according to δ15Nobs = fmδ15Nm + fsδ15Ns + faδ15Na, 1/(N2/He)obs = fm/(N2/He)m + fs/(N2/He)s + fa/(N2/He)a, and fm + fs + fa = 1 [Sano et al., 1998].

  • c

    Percent of nitrogen derived from sediments, 100 × S/(M + S).

  • d

    Calculated using GC He detection limits.

IND-1−4.0 ± 0.43590.410.310.2743
IND-2−2.1 ± 0.59490.160.040.8119
IND-41.7 ± 0.411230.130.510.3680
IND-6−0.73 ± 0.411000.130.190.6858
IND-70.69 ± 0.46460.220.580.2072
IND-82.1 ± 0.420000.070.450.4987
IND-92.1 ± 0.310340.140.590.2781
IND-131.7 ± 0.3171,800d00.240.77100
IND-142.0 ± 0.521,875d0.0020.290.7199
IND-151.2 ± 0.429,794d0.0020.180.8299
IND-161.2 ± 0.632,609d0.0010.170.8399
IND-17−3.3 ± 0.428520.0400.960
IND-19−7.3 ± 0.57860.1200.880
IND-20−3.1 ± 0.45420.270.150.5834
IND-20a−2.7 ± 0.47410.200.050.7518
mantle−15 ± 2150100 
sediment7 ± 210,000010 

7.3. Along-Arc Trends

[42] In Figure 6, we show the along-strike variations in δ15N, N2/He and percent sediment contribution (%S), calculated assuming the −15‰ mantle end-member. This figure illustrates a strong latitudinal control in these parameters and on the sediment contribution, with the southern Sangihe volcanoes characterized by a significantly greater contribution than the offshore northern arc volcanoes. For example, the average %S contributions for each locality decrease systematically from S to N: Ambang (84%), Lahendong (70%), Lokon (31%) and Ruang (26%). Note, however, that sample IND-17 from Awu, the northernmost volcano, has an anomalously high N2/He ratio for its δ15N value and falls outside the mixing trajectories in Figures 4 and 5. A mantle end-member with a N2/He ratio of 650 (and δ15N = −15‰) would be required to place Awu inside the mixing trajectories (on the M-S mixing line) and a N2/He ratio of 1050 would be required for Awu volatiles to show a sedimentary N contribution of 25%, similar to Ruang to the south. Given these observations and Awu's negative δ15N (−3.3‰) value, we believe that Awu receives less hemipelagic sediment-derived N than Ruang and possibly none at all (see section 7.4).

Figure 6.

Along-strike variation in δ15N, N2/He, and %S (the percentage of N derived from a M-S mixture assuming δ15Nm = −15‰) for Sangihe samples. Gray filled circles show the average %S for each sample suite. Global arc average from Sano et al. [1998]. Costa Rica average from Zimmer et al. [2004] for comparison.

[43] Models of global nitrogen flux indicate that the N in arc volatiles is approximately 75% sediment-derived on average [Sano et al., 1998]. Most localities in the Sangihe Arc fall well below that value, consistent with the idea of a generally low contribution of sedimentary derived N to the source region of Sangihe Arc magmas. In fact, the northern Sangihe volcanoes receive <35% of the emitted N2 from N supplied by subducted hemipelagic sediments. This value is similar to that of Costa Rica, where sediment off-scraping has been proposed to explain the low %S contribution to geothermal N2 discharges [Fischer et al., 2002; Zimmer et al., 2004].

[44] We propose two possible explanations for these along-arc variations in N isotope and relative abundance systematics.

[45] 1. A decrease in the amount of hemipelagic sediment subducted (along-strike from south to north) could produce the observed trends in N-He systematics. A thicker pile of subducted organic sediments in the south would yield higher N2/He and δ15N values by increasing the proportion of heavier sedimentary N with respect to mantle-derived N (and He). Deep sea drilling at several trenches worldwide (e.g., Central America, E. Sunda, Java, Antilles, Japan, Aleutian) has revealed that the hemipelagic diatom-rich muds are usually found at the top of the sediment column, and tend to overlie the pelagic carbonates [Plank and Langmuir, 1998]. Off-scraping of the uppermost hemipelagic sediment layer has been proposed for Costa Rica where the N contributed from sediments is only 37% on average, significantly less than in the rest of the Central American margin [Fischer et al., 2002; Zimmer et al., 2004]. If hemipelagic muds comprise the top sedimentary layer in the Sangihe trench, then off-scraping could also lead to a diminished hemipelagic sediment signature. A greater amount of off-scraping in the north compared to the south would result in the low N2/He and δ15N exhibited by northern Sangihe volcanoes.

[46] 2. A change in sediment composition along strike from hemipelagic-dominated in the south to carbonate-dominated in the north would yield decreasing N2/He and δ15N values. This scenario has been proposed for the Sangihe Arc by Jaffe et al. [2004] on the basis of C-He systematics. Carbon sources can be resolved into contributions from carbonates (L; both carbonate-rich sediments and altered oceanic basement), organic-rich (hemipelagic) sediments (S), and the mantle wedge (M). Higher L/S ratios would indicate an increased contribution of C by the carbonate component with respect to organic sediments. The observed increase in L/S ratios from south to north (9 at Lokon to 87 at Awu) suggests that carbonate contributions are more important in the north, whereas southern Sangihe volcanoes receive relatively more CO2 from organic (hemipelagic) sediments [Jaffe et al., 2004].

7.4. Tectonic Implications

[47] In this section, we propose a model for the origin of volatiles in the Sangihe Arc based upon the existing tectonic framework of the Molucca Sea region. As discussed above, the collision complex of forearc sediments appears to be decoupled from the down going slab due to the advanced state of collision between the Sangihe and Halmahera arcs. As arc-arc convergence has been shown to be oblique and more developed in the north, it can be expected that the degree of decoupling between slab and sediments may also be more pronounced to the north. This is consistent with geophysical data that image a northward deepening slab in the north [Lallemand et al., 1998]. Sediment decoupling implies that hemipelagic sediments in the northern part of the arc are trapped in the collision complex and are not subducted. Collision in the southern Sangihe arc has not yet begun [Hamilton, 1979]; therefore it can be presumed that the sediments there remain coupled to the oceanic plate on which they rest and are subducted along with the ocean crustal basement. Samples from the north show less hemipelagic sediment involvement (lower N2/He, δ15N, %S and higher L/S ratios), while those in the south have a greater hemipelagic sedimentary signature (higher N2/He, δ15N, %S and lower L/S ratios).

[48] Awu Volcano is unusual because its volatile systematics show negative (mantle-like) δ15N yet high (sediment-like) N2/He. A negative N isotopic signature at this northernmost volcano is consistent with a lack of sediment subduction in the northern arc, as is evidenced by the other samples, yet this is seemingly contradicted by the high N2/He ratio. An elevated N2/He at Awu could be due to potential N-He fractionation during magma degassing but this is unlikely given prior observations [Sano et al., 2001]. Alternatively, high N2/He could result from the addition of excess nitrogen from a source that has negative δ15N. One possibility is a contribution of N derived from the subducted altered oceanic crust. Jaffe et al. [2004] have argued that 3He/4He values slightly less than the MORB ratio in northern Sangihe Arc magmas reflect radiogenic He released from the crustal basement of the subducting slab as sediments are unable to transport He to subarc depths due to the high diffusivity of He [Hilton et al., 1992]. If He is released by sediments in the forearc, the N2/He of a subducted slab as a whole would be elevated relative to MORB. Variations in the thermal structure of the subduction zone in the north could promote melting of the subducted crust, which would incorporate this elevated N2/He signature into the magma source to produce the observed N-He characteristics.

[49] It is noteworthy, in this respect, that a lack of seismicity in northern Sangihe has been interpreted to indicate slow-down or cessation of subduction due to advanced arc-arc collision in the region [McCaffrey, 1982]. Thermal modeling by Peacock et al. [1994] showed that slower rates of subduction allow the slab to reach higher temperatures. If collision has caused the slab to stall in the mantle, it could become heated and be more prone to melting. Significantly, trace element and radiogenic isotope data suggest that the mechanism by which mass transfer from the subduction component to the mantle beneath Sangihe takes place, trends from fluid-dominated to partial melt-dominated over time as collision progresses and subduction wanes [Elburg and Foden, 1998].

[50] The idea of a stalled slab undergoing melting may help explain the anomalous N systematics of the Awu gas sample. Melting of the slab could elevate N2/He ratios, even if off-scraping of sediments has removed sedimentary N, as in scenario (1) above. N isotopes could retain their negative mantle signatures owing to the MORB origin of the subducting oceanic basement, leading to high N2/He in association with negative δ15N values, such as are exhibited at Awu. In the southern part of the arc, on the other hand, N2/He ratios are high and δ15N values are positive, indicative of significant sediment involvement and transfer of subducted components into the mantle wedge by devolatilization of the sediment-laden slab.

[51] This idea is consistent with the interpretation of the C and He isotope systematics of Sangihe Arc gases [Jaffe et al., 2004]. The average MORB CO2/3He ratio of 1.6 × 109 [Des Marais, 1985; Marty and Jambon, 1987] is significantly lower than that of organic and carbonate sediments (average value of 1 × 1013; Sano and Marty, 1995], and an elevation in CO2/3He is considered indicative of slab involvement. The CO2/3He ratios are 1–2 orders of magnitude higher in the north (117 and 121 at Awu and Karangetang (a volcano 102 km south of Awu on Siau Island), respectively) than in the south (6.4, 5.2, 5.8 and 2.8 at Lokon, Lahendong, Leilam and Ruang, respectively), indicating greater slab influence in the northern arc (all values × 109). Contributions to the C budget from both the carbonate sediments/altered oceanic crust (L) and organic sediment (S) components can be evaluated by comparing the L/S ratios of volatiles emitted along-strike of the arc. Awu volcano shows the highest L/S ratio of all localities (87 versus 13–17 for the rest of the arc), consistent with a predominantly carbonate derived CO2 contribution to Awu volatiles. Taken together, the CO2/3He and L/S ratios indicate that volatile contributions from carbonate-rich sediments or altered ocean crust are significantly greater in the northern Sangihe arc compared to the south. Lower L/S in the south suggests greater subduction of organic-rich sediment from the uppermost portion of the sedimentary column.

[52] Off-scraping of the uppermost hemipelagic sediments, due to the decoupling of the collision complex from the subducting lithosphere, can account for lower N2/He and δ15N and higher L/S ratios in the north relative to the south. Cessation of subduction resulting in heating of the slab and a more efficient addition of slab-derived melts and/or fluids can account for elevated CO2/3He ratios, and also provide an explanation for the anomalous Awu gas composition (high N2/He with negative δ15N).

8. Conclusions

[53] This work presents the first complete gas chemistry and nitrogen isotope data on geothermal fluids from the Sangihe Arc. The data give insight on how volatiles are transferred from the subducting slab to the mantle wedge and atmosphere. Along arc changes in the N2/He ratios and δ15N values of the gases can be reconciled when considering variations in the amount of N transferred to arc magmas from subducted hemipelagic sediments. The following conclusions are emphasized:

[54] 1. The 15 geothermal samples collected in this study show the general characteristics of gases discharging from arc magmas world-wide, with H2O, CO2 and St as their main constituents. Samples that have been compromised by air contamination have been identified by their extreme N2/He ratios (>20,000), low He/Ar ratios (<0.01) detectable O2 and δ15N values close to 0.0‰. Four samples fulfilled all of these criteria and are eliminated from the discussion of mantle volatile composition.

[55] 2. The N2/He ratios and N isotopic composition of the samples range from values characteristic of mantle-derived volatiles, as sampled by MORB, to values typical for other arcs world-wide. Quantification of nitrogen source components, in terms of mantle (M), subducted hemipelagic sediments (S), and air (A), reveals a range from M-dominated to S-dominated values.

[56] 3. The percentage of N derived from a hemipelagic sedimentary source (%S) has a distinct along-arc variation from <30%S in the north to 84%S in the south. Most localities show a significantly lower contribution from sediments than the global arc average of 75% of Sano et al. [1998].

[57] 4. Two possible explanations have been put forward to account for observed along-arc trends in N-He systematics: (1) variations in the degree of off-scraping of the uppermost hemipelagic sediments, with less hemipelagic material being subducted in the north, and more organic sediment subduction in the south, and (2) changes in sediment composition such that the sediments in the northern arc contain more carbonate, whereas the south has more organic material.

[58] 5. The observed trends are explained within the framework of an existing tectonic model for the Sangihe Arc [Hamilton, 1979; Hall, 1996; Lallemand et al., 1998]. Oblique arc-arc collision between Sangihe and Halmahera causes sediments to become decoupled from the underlying oceanic crust in the north where the collision is most advanced, while sediments in the south are unaffected and are subducted as at other arcs. Trends toward lower N2/He, δ15N, and %S reflect the increasing extent of sediment decoupling from the subducting oceanic crust in the north resulting in a lower amount of sediment-derived N to be transferred to arc magmas. Stalling and heating of the downgoing plate in the north, however, would be expected to enhance the contribution of slab components with respect to that of the mantle wedge as suggested by C-He relationships of the same samples [Jaffe et al., 2004]. We propose that in the north, N2 and CO2 are transferred from the altered oceanic crust, giving Awu volatiles the following characteristics: (1) negative and MORB-like δ15N (−3.3‰) at high N2/He (∼2800) and (2) carbonate δ13C (−0.4‰) at high CO2/3He (117 × 109). Further work at localities where the hemipelagic sediment input to arc magmas is minimal is required to better constrain the composition of volatiles contributed from the altered oceanic crust to arc magmas and the atmosphere.

Appendix A

[59] The information on the sampling localities provided below is based on Volcanoes of the World [Simkin and Siebert, 1994], the Smithsonian's “Global Volcanism Program” (http://www.hrw.com/science/si-science/earth/tectonics/volcano/), and the University of North Dakota's “Volcano World” (http://volcano.und.nodak.edu).

A1. Awu Volcano

[60] Awu is a massive stratovolcano rising out of a 4.5 km wide caldera at the northern end of Sangihe Island. It is one of Indonesia's most active volcanoes, erupting explosively in 1711, 1812, 1856, 1892, and 1966, and producing pyroclastic flows and lahars. The most recent eruptions took place in 1992 and at the time of writing (June 2004). In 2001, Awu contained a crater lake (pH 2.5) at its summit with strong gas bubbling near the shore, where a spring sample was collected (IND-17). In addition, we sampled weak fumaroles diffusely emitting gases at a second location (samples IND-15 and -16, duplicates).

A2. Ruang Volcano

[61] Ruang is an island stratovolcano located between the northeastern tip of Sulawesi and Sangihe Island: it is the southernmost volcanically active island in the chain. An eruption in 1949 produced lava flows and contributed to the growth of the 1904 lava dome that partially fills the crater and which is now covered in vegetation. Ruang last erupted in 2002. There are several fumaroles located between the dome and the interior crater wall, and a lava flow emerges from the crater.

A3. Lokon-Empung Volcanoes

[62] The twin volcanoes Lokon and Empung on the northeastern arm of Sulawesi have summits 2.2 km apart. In the saddle between the peaks is situated a third and younger crater. It is from this crater, Tompaluan that all eruptions have originated since the 18th century. A May 2001 eruption, accompanied by seismic tremor, sent up a 900 m ash column, and the latest eruption occurred in 2002.

A4. Mahawu Volcano

[63] Mahawu, situated immediately east of Lokon-Empung, has a crater that occasionally contains a crater lake, where fumaroles, mudpots, and small geysers have been observed since 1994. Occasional small explosive eruptions have been recorded since 1789, the last of which was in 1977. The 160 m deep crater was inaccessible for gas sampling (without climbing equipment), and therefore samples were collected instead from geothermal fields (Lahendong) near Lokon and Mahawu volcanoes.

A5. Soputan Volcano

[64] Soputan is a small stratovolcano constructed on the southern rim of Quaternary Tondano caldera, one of the most active of Sulawesi. Historically, eruptions have originated both from the summit crater and from a cinder cone that formed in 1906 on the northeast flank, called Aeseput. Beginning July 16, 2003, there have been explosions, pyroclastic flows, avalanches, and an eventual lava flow issuing from the northwestern part of the summit lava dome, built since 1991. Fumarole degassing is very diffuse and weak.

A6. Ambang Volcano

[65] Ambang is a compound volcano on the western arm of Sulawesi, having multiple closely spaced vents. Its only known historical eruption took place in the 1840s, and the nature of this eruption is unrecorded. Ambang currently has strong fumarolic activity in the summit crater.


[66] This research was funded by a Research Allocations grant from the University of New Mexico (T.P.F.), the Caswell Silver Foundation in the form of a graduate fellowship (L.C.), and an NSF grant (EAR-0100881) (D.R.H.). Additional funds for field expenses came from UCSD (Earth Sciences Program). We thank Lillie Jaffe, Justin Kulongoski, and Pat Castillo for valuable support in the field, Purnama Hilton for coordination with LIPI, and Mark Erdmann for arranging accommodations in Manado. We also thank GRDC director Bambang Dwiyanto (Bandung) for his support of the project. The help of Viorel Atudorei was critical to N isotope analyses, as was the help of John Husler with wet chemical analysis. Comments of M. J. van Bergen, G. Snyder, Gray Bebout, (associate editor) and two anonymous reviewers greatly improved the manuscript.