The Sangihe Arc is presently colliding with the Halmahera Arc in northeastern Indonesia, forming the world's only extant example of an arc-arc collision zone. We report the first helium and carbon isotopic and relative abundance data from the Sangihe Arc volcanoes as a means to trace magma origins in this complicated tectonic region. Results of this study define a north-south trend in 3He/4He, CO2/3He, and δ13C, suggesting that there are variations in primary magma source characteristics along the strike of the arc. The northernmost volcanoes (Awu and Karangetang) have higher CO2/3He and δ13C (up to 179 × 109 and −0.4‰, respectively) and lower 3He/4He (∼5.4 RA) than the southernmost volcanoes (Ruang, Lokon, and Mahawu). Resolving the arc CO2 into component structures (mantle-derived, plus slab-derived organic and carbonate CO2), the northern volcanoes contain an unusually high (>90%) contribution of CO2 derived from isotopically heavy carbonate associated with the subducting slab (sediment and altered oceanic basement). Furthermore, the overall slab contribution (CO2 of carbonate and organic origin) relative to carbon of mantle wedge origin is significantly enhanced in the northern segment of the arc. These observations may be caused by greater volumes of sediment subduction in the northern arc, along-strike variability in subducted sediment composition, or enhanced slab-derived fluid/melt production resulting from the superheating of the slab as collision progresses southward.
 The study of helium and carbon in arc-related volcanic emissions has provided a wealth of information on the subduction process and its involvement in geochemical cycling between the terrestrial mantle and the crust, hydrosphere and atmosphere. For example, the flux of various volatile species (including CO2) from subduction zones can be estimated through knowledge of the arc-related primordial 3He flux [Torgersen, 1989; Allard, 1992] and measurement of the relevant elemental ratio (xi/3He) where xi = element of interest (see review by Hilton et al. ). In the case of CO2, it is possible to resolve the output flux into constituent components (slab-related versus mantle wedge contributions) through the use of modeled end-member compositions with specific He and CO2 isotopic and relative abundance characteristics [Marty et al., 1989; Varekamp et al., 1992; Sano and Marty, 1995]. Helium and carbon (isotopes and/or relative abundances) can also be used to identify regional tectonic controls on magma genesis, as in the case of the Sunda-Banda arcs of Indonesia [Hilton and Craig, 1989; Hilton et al., 1992; Varekamp et al., 1992], including recognizing crustal influences and their contribution to arc-related magmatism [Gasparon et al., 1994].
 In this contribution, we apply the He-C approach to geothermal fluids collected at active volcanic centers along the Sangihe Arc of northeastern Indonesia. The Sangihe Arc is unusual because it is part of a conjugate pair of arcs, located on either side of the Molucca Sea Plate, which are in the process of colliding (Figure 1). The Sangihe and Halmahera arcs are the only extant example of an arc-arc collision zone. Although data are available on major element chemistry of arc rocks in this region [e.g., Tatsumi et al., 1991; Elburg and Foden, 1998; Macpherson et al., 2003], there is little information on the composition of subducted components. For example, there are no major or trace element studies on Molucca Sea sediment, nor is there any record of the composition of the Molucca Sea basement (as it has been subducted). Therefore, to identify various contributions to magmagenesis in the region, prior studies have had to approximate the composition of (a) Molucca Sea sediments (major and trace elements and isotopes), using sedimentary analogs from the Philippines, SW Pacific and Banda Arc, and (b) crustal basement, using material from the Celebes Sea [see Elburg and Foden, 1998]. In this work, we target He and CO2 characteristics in the volcanic arc output of this remote arc setting, with the aim of characterizing slab-related sources by considering the gross (volatile) systematics of the subducted components (both the oceanic crustal basement and its sedimentary veneer). In this respect, the present study represents an attempt to utilize geothermal fluids of the Sangihe Arc to reveal the characteristics of volatiles from the underlying source region. A related aim is to assess if along-strike variations exist in the He-C characteristics. Such variations may reflect heterogeneity in the composition of slab-related inputs, and thus provide information related to the tectonic development of the region.
2. Geological and Tectonic Background
 There are four major lithospheric plates which interact in the Molucca Sea region of northern Indonesia: the Eurasian Plate to the west, the Philippine Plate to the east, the Australian Plate to the south, and the Molucca Sea Plate trapped in the center (Figure 1a). The Molucca Sea Plate is highly unusual in that it is being subducted beneath both the Eurasian and Philippine Sea plates giving rise to two converging, subparallel volcanic arcs: the Sangihe Arc in the west and the Halmahera Arc in the east (Figures 1b and 1c). Seismic data indicate that the Benioff-zone extends as much as 600 km below the Sangihe Arc, at an approximate dip angle of 45°, and up to 250 km below the younger Halmahera Arc subducting in the opposite direction at approximately 40° [Hatherton and Dickinson, 1969; Silver and Moore, 1978; Lallemand et al., 1998]. The mean convergence rates of the Sangihe and Halmahera arcs are ∼4 cm/yr and ∼3 cm/yr, respectively, on the basis of seismic and chronological constraints [Lallemand et al., 1998].
 The origin of the Molucca Sea Plate can be traced to the collision of the Philippine Sea Plate with the Australian Plate approximately 25 Ma ago, trapping a piece of Indian Ocean lithosphere that attached to, and moved with, the Philippine Sea Plate: this fragment became the Molucca Sea Plate [Hall, 1996]. Subduction on the western side, forming the Sangihe Arc, probably began soon after collision. Eastward subduction of the Molucca Sea Plate began about 15 Ma ago when collision of the Snellius Plateau with the Sangihe-Philippine Arc system forced a subduction reversal, breaking the Molucca Sea Plate away from the Philippine Sea Plate and forming the Halmahera Arc [Hall and Nichols, 1990; Hall, 1996].
 There is presently no surface remnant of the Molucca Sea Plate basement other than the thick collision complex consisting of the mélange wedges of the two arcs [Silver and Moore, 1978]. This wedge of unconsolidated and deformed Tertiary sediments, up to 15 km in thickness, is composed of volcaniclastic and continental debris, including peridotite, serpentinite, gabbro, basalt, chert, limestone, and greywacke [Silver and Moore, 1978; Hamilton, 1979; Sukamto, 1979], generated primarily by the continuous and ongoing collision of the Halmahera and Sangihe arcs [Silver and Moore, 1978; McCaffrey et al., 1980].
 The Sangihe Arc is the oldest extant subduction zone in the Philippine-Indonesia region: subduction-related rocks about 25 Ma old [Dow, 1976; Effendi, 1976; Apandi, 1977; Priadi, 1993] have been reported in the northeastern arm of Sulawesi. Whereas the southern portion of the arc ends at the left-lateral Sorong Fault (Figure 1a), the northern termination is marked by the collision of eastern and western Mindanao about 4–5 Ma ago [Pubellier et al., 1991]. The present-day active arc ceases at ∼4°N; however, the extinct margin extends to 5.5°N [Lallemand et al., 1998; Widiwijayanti et al., 2003], with lavas in central Mindanao thought to be of Sangihe origin [Pubellier et al., 1991]. This northern collision region is further complicated by back-arc thrusting along the Cotobato Trench in the west and the Philippine Trench in the east, leaving dissected remnants of the Sangihe (and Halmahera) arcs on Mindanao Island and the Talaud Ridge [Morrice et al., 1983]. Post-collisional volcanism remains active in central Mindanao (i.e., the relic Sangihe Arc) as a result of the continuing subduction of portions of the Halmahera Arc [Pubellier et al., 1991, 1999]. While the active portions of the Sangihe and Halmahera arcs are currently several hundred kilometers apart, the northern collision is propagating southward and will continue until the Halmahera Arc accretes onto the Eurasian margin [Hall and Nichols, 1990].
 The active Sangihe Arc can be divided into two segments: four volcanic islands north of the large island province, Sulawesi, and eight volcanoes in northeastern Sulawesi (see Figure 1b). The active volcanoes in the northern segment of the arc (Awu, Banua Wahu, Karangetang, and Ruang) occur about 50 km apart along a north-south transect. The onshore active Sangihe Arc volcanoes (Tongkoko, Mahawu, Lokon-Empung, and Soputan) are spaced less regularly. Four other volcanoes, morphologically young and hydrothermally active (Duasudara, Klabat, Manado Tua, and Ambang), also lie along the southern portion of the arc. Lavas erupted along the chain are mostly typical island arc volcanics (basaltic andesite to andesite) with the exception of Soputan that produces olivine-rich basalt [Morrice et al., 1983]. The active Sangihe Arc is bounded by Awu volcano in the north and Ambang in the south.
3. Sampling and Analytical Protocols
 In order to characterize the helium and carbon systematics of the Sangihe Arc, we collected geothermal fluids (fumaroles, gas emanations from bubbling springs, and thermal spring waters), and olivine- and/or pyroxene-rich lavas from seventeen different localities covering seven active volcanic centers along the north-south strike of the arc. Collection of geothermal samples was facilitated by use of a Ti-tube or inverted plastic funnel [after Giggenbach and Goguel, 1989], and samples were stored in either AR-glass bottles or cold-welded annealed copper tubes for transfer back to the laboratory. Normal sampling precautions were taken to minimize the possibility of air contamination [see Hilton et al., 2002].
 In the laboratory, samples were extracted using instrumentation and procedures described previously [Kulongoski and Hilton, 2002; Shaw et al., 2003]. Briefly, all fluid samples were released into an ultra-high vacuum system consisting of a series of traps held at different temperatures, thus separating water vapor from the condensable (mainly CO2) and noncondensable gases. The noncondensable fraction was aliquoted into an AR-glass breakseal for transfer to a noble gas mass spectrometer (MAP215) for He isotopic and He and Ne relative abundance analyses. The condensable fraction was transferred to a separate vacuum line for further purification, and the total amount of CO2 was measured manometrically. A fraction of the purified CO2 was then collected in another breakseal for isotopic analysis (using a VG Prism mass spectrometer).
 Rock samples were prepared by crushing and sieving into various grain sizes. Olivine and pyroxene mineral grains, ranging in diameter from 850–1700 μm, were hand-picked from the appropriate size fraction, and ultrasonically cleaned using a 50/50 methanol/acetone solution. Olivine and pyroxene grains were loaded separately into on-line, electro-magnetic crushers (see description by Scarsi ) attached to the MAP215 to release volatiles for measurement.
 Prior to inlet into the noble gas mass spectrometer, all samples underwent a purification procedure designed to isolate the He and Ne fractions by a combination of active gas gettering, sorption onto charcoal at liquid nitrogen temperatures and cryogenic separation using a liquid He cooled trap. Isotopic measurements were made in static mode, and either an air or a 3He-rich standard (Murdering Mudpots (Yellowstone) = 16.45 RA where RA = air 3He/4He) was used for normalization.
 Helium and carbon isotopic and relative abundance characteristics of 28 samples (26 geothermal fluids and 2 phenocrysts) from the Sangihe volcanic arc are given in Table 1. Samples cover 7 distinct volcanic centers (3 offshore and 4 onshore). Results from each volcanic center, listed from north to south, are discussed in turn.
Table 1. Helium and Carbon Results Along the N-S Strike of the Sangihe Arc (Sampled in 2001)
IND-## = AR-glass bottle, I-### = Copper tube, others are rock samples.
Abbreviations: sf, thermal spring fluid phase; sg, thermal spring gas phase; fm, fumarole; sl, solfatara fumarole; ol, olivine; px, pyroxene.
RM/RA = measured 3He/4He (RM) in sample relative to air 3He/4He (RA). Error is 1σ.
X = [(He/Ne)sample/(He/Ne)air] × 1.209 (i.e., the air-normalized He/Ne ratio multiplied by βNe/βHe = 1.209 - the ratio of the Bunsen coefficients at 17°C).
RC/RA = air-corrected 3He/4He (RC) in sample relative to air (RA). RC/RA = [(RM/RA)X − 1]/(X − 1).
Values in italics have been removed from the discussion; they are considered unrepresentative of primary magmatic source (see section 5.1).
Errors on δ13C are less than ± 0.5‰, based on replicate analyses.
[4He] per gram of water in fluid samples, and per gram of mineral in rock samples.
Crater Loc. 1
1.00 ± 0.01
1.00 ± 0.01
31.3 ± 0.4
1.22 ± 0.01
1.39 ± 0.02
108 ± 1
1.08 ± 0.02
1.14 ± 0.02
96 ± 2
Crater Loc. 2
6.12 ± 0.07
6.22 ± 0.08
117 ± 2
5.49 ± 0.06
6.43 ± 0.08
63.8 ± 0.9
4.73 ± 0.06
5.08 ± 0.07
27.5 ± 0.4
4.99 ± 0.07
5.37 ± 0.08
179 ± 3
4.42 ± 0.39
4.86 ± 0.50
Crater Loc. 1
3.55 ± 0.03
4.02 ± 0.04
1.49 ± 0.08
Crater Loc. 2
6.97 ± 0.07
6.99 ± 0.07
2.83 ± 0.04
7.11 ± 0.06
7.12 ± 0.07
4.41 ± 0.05
7.27 ± 0.06
7.27 ± 0.06
8.4 ± 0.1
7.33 ± 0.11
7.50 ± 0.12
53.8 ± 1
7.11 ± 0.08
7.20 ± 0.08
6.44 ± 0.08
7.32 ± 0.07
7.33 ± 0.10
5.07 ± 0.09
7.04 ± 0.08
7.04 ± 0.10
5.37 ± 0.09
7.23 ± 0.09
7.23 ± 0.14
5.8 ± 0.1
6.90 ± 0.09
6.91 ± 0.10
5.8 ± 0.1
1.07 ± 0.02
1.11 ± 0.02
0.69 ± 0.01
1.04 ± 0.01
1.07 ± 0.01
0.95 ± 0.02
0.93 ± 0.02
0.113 ± 0.002
1.02 ± 0.01
1.03 ± 0.01
0.143 ± 0.002
5.33 ± 0.15
5.39 ± 0.34
Crater Loc. 1
4.63 ± 0.07
4.64 ± 0.07
5.7 ± 0.1
4.59 ± 0.04
4.60 ± 0.04
6.45 ± 0.08
Crater Loc. 2
3.68 ± 0.09
3.77 ± 0.10
590 ± 20
3.27 ± 0.10
3.27 ± 0.22
140 ± 10
3.97 ± 0.06
4.08 ± 0.06
292 ± 5
4.1. Awu Volcano (Sangihe Island)
 Two different crater localities were sampled. The fumarole samples (from location 1) have air-like 3He/4He (∼1 RA) and low He/Ne values (X ∼ 2, where X is the air-normalized He/Ne value multiplied by the ratio of the Bunsen coefficients; see Table 1 footnote). The second locality (location 2) was a bubbling spring located in the crater lake close to the shore: this gas sample has a He isotope value of 6.2 RA with an X value of 55. The CO2/3He value of the spring is 117 × 109. Its δ13C value is −0.4‰.
4.2. Karangetang Volcano (Siau Island)
 Two coastal volcano flank thermal spring localities were sampled, with both Temboko (6.4 RA) and Lehi (5.4 RA) showing a dominantly magmatic He input, albeit with a significant air correction (X ≤ 12). One sample (IND-17) has an anomalously high δ13C value of +1.3‰; however, the other two samples have CO2/3He values between 6.4 and 17.9 (× 109) and have similar δ13C values of ∼−2‰. The pyroxene mineral separate from Sang'01-40 has a 3He/4He ratio ∼4.9 RA; slightly lower than the thermal spring samples.
4.3. Ruang Volcano (Tagulandang Island Group)
 Two summit fumarole localities were sampled, with one characterized by a 3He/4He ratio of 7.0 RA (X = 470). The other locality has a greater degree of air contamination (X = 6.41) and its 3He/4He ratio is significantly lower (3.6 RA). Although the air-corrected 3He/4He ratios are dissimilar (4.0 and 7.0 RA), there is good agreement between δ13C values (∼−3‰) for the two localities. In addition, the two CO2/3He ratios 1.5–2.8 (×109) fall within a narrow range.
4.4. Lokon Volcano (North Sulawesi)
3He/4He values of duplicate samples from a single fumarole locality at Lokon volcano showed excellent agreement (∼7.2 ± 0.1 RA). Air contamination in each case was negligible (X > 800). CO2/3He ratios are 4–8 (×109) and the single δ13C value is −3.6‰.
4.5. Mahawu Volcano (North Sulawesi)
 Two different localities in the vicinity of Mahawu volcano were sampled: the Lahendong geothermal complex (including thermal resort) and a remote thermal spring locality at Kakaskasen. Consistent results were obtained for all samples with the exception of one duplicate from Kakaskasen (see below). 3He/4He ratios are 6.9–7.5 RA and CO2/3He values lie between 5 and 6 (×109). One sample from Kakaskasen has an anomalously high CO2/3He ratio of 54 × 109. δ13C values for Lahendong are tightly constrained at ∼−3.3‰, whereas Kakaskasen values are significantly lower (−5.8 to −6.8‰).
4.6. Soputan Volcano (North Sulawesi)
 All fumarole samples from the summit of Soputan have air-like values (3He/4He ∼ 1 RA; X ∼ 2.5). The flank cinder cone Aeseput samples also had air-like He isotopes (3He/4He ∼ 1 RA; X ∼ 2.5). The olivine crystals, separated from the ash sample, gave a 3He/4He ratio of ∼5.4 RA.
4.7. Ambang Volcano (North Sulawesi)
 Two localities were sampled on the summit of Ambang Volcano, and one on the volcano flank. The summit fumarole locality gave consistent 3He/4He ratios of 4.6 RA for 2 samples, with good agreement in both CO2/3He (5.7 × 109 and 6.4 × 109) and δ13C values (−4.5‰ and −5.2‰). The summit thermal fluid locality gave lower 3He/4He ratios (3.3 RA and 3.8 RA), and higher δ13C (−4.3‰ and −2.3‰) and CO2/3He (140 × 109 and 590 × 109) values. The flank thermal spring locality at Bonkurai has intermediate 3He/4He (4.1 RA), CO2/3He (292 × 109) and δ13C (−4.7‰) values.
 There are three principal contributors to the inventory of magmatic volatiles at arc-related settings: the subducted crustal basement, its overlying sedimentary veneer and the mantle wedge above the slab. The atmosphere as well as the arc lithosphere through which magma erupts are also potential sources of volatiles; although these sources are generally regarded as contamination and unrelated to the magma source of volatiles. Therefore, in order to quantify the various contributors of volatiles to these magmas, and to recognize any extraneous additions to, or modifications from, source characteristics, it is essential to adopt various criteria that differentiate samples that possess intrinsic magma characteristics from those that have been modified by air contamination, crustal assimilation, degassing, and/or sampling error. To this end, we adopt a number of criteria in the following section aimed at assessing the integrity of each individual sample to preserve its primary source signature. Only after we have applied this filter can we relate the He-C systematics of the Sangihe Arc to the tectonic framework of the region.
5.1. Identifying Samples With Modified He-C Characteristics
5.1.1. Air-Like 3He/4He and He/Ne Values
 There are a total of 7 geothermal samples with air-like 3He/4He and He/Ne values (Table 1). These are: all three samples from Awu crater fumaroles (location 1; IND-15, IND-16, I-048) and all four samples from Soputan fumaroles (from the main crater and the Aeseput flank cone; IND-12, I-016, IND-13, IND-14). In both cases, the diffusely flowing fumaroles could have entrained air through the highly altered (Awu) or porous and blocky (Soputan) material covering the fumarole discharge sites.
5.1.2. Low 3He/4He: Indications of Radiogenic Crustal Influence
 Helium isotopic ratios in arc-related settings around the circum-Pacific region generally fall between 6–8 RA [Poreda and Craig, 1989]. A more recent compilation of arc 3He/4He data [Hilton et al., 2002] has produced an average value of ∼5.4 RA for all arcs worldwide. As He isotopes are a sensitive tracer of volatiles of crustal provenance [Hilton et al., 1993; Gasparon et al., 1994], we suspect that 3He/4He less than the average value cited above may have incorporated a significant radiogenic He contribution, thus masking the magmatic source ratio. For this reason, we reject all five Ambang samples (IND-9, IND-10, IND-11, I-198, and I-193), as their 3He/4He are less than 4.6 RA. We reject the pyroxene separate from Karangetang (Sang'01-40), and the olivine sample from Soputan (SOP-1) as they also have low 3He/4He values.
5.1.3. High CO2/3He Values: Indications of Fractionation in the Hydrothermal System
 The average CO2/3He ratio of arc-related volcanism is 15 ± 11 (×109) [Sano and Marty, 1995; Sano and Williams, 1996]. Significant deviations from this value are usually indicative of fractionation of He from C [van Soest et al., 1998; Hilton et al., 2002]. In the case of geothermal (aqueous) fluids, the greater solubility of C relative to He [see Stephen and Stephen, 1963] can lead to higher CO2/3He values in the residual phase following vapor formation or other gas loss event. Both van Soest et al.  and Shaw et al.  report significantly higher CO2/3He ratios in gases dissolved in thermal waters compared to gas phase samples collected at the same locality. This observation is consistent with the notion that free gas samples provide a more robust means of sampling the intrinsic CO2/3He value of a degassing magma body whereas gases dissolved in waters may represent (fractionated) volatiles residual from a degassing event. The current data set allows us to make the same comparison between liquid and gas phase CO2/3He values.
 Unusually high CO2/3He ratios are measured in Ambang and Karangetang geothermal fluid samples. Three Ambang geothermal fluid phase samples (IND-10, I-193, and IND-11; average CO2/3He ∼ 340 × 109) had significantly higher CO2/3He values than their gas-phase counterparts from the same locality (IND-8, IND-9, and I-198; average CO2/3He ∼ 4.5 × 109). Likewise, one Karangetang thermal spring fluid phase sample (I-021) has a high CO2/3He value (179 × 109) when compared with the other fluid phase sample from the same locality (IND-17 CO2/3He = 27.5 × 109). These differences between gas and fluid phase CO2/3He values suggest that the fluid phase CO2/3He represents volatiles that have been modified due to fractionation most likely by vapor formation in the geothermal fluid system. Therefore we omit four samples with anomalously high CO2/3He values from further consideration.
5.1.4. Atypical δ13C Values: Additional Indications of Alteration in the Hydrothermal System
 Although agreement in δ13C between duplicate samples is generally good (see Table 1), there is one locality (Karangetang volcano) with samples showing markedly different δ13C values. One sample (IND-17) has an unusual δ13C value of +1.5‰ that is very different from its duplicate (I-021) (δ13C = −2.0‰). We can find no analytical reason for the discrepancy. However, we note that given the observation [e.g., Sano and Williams, 1996] that almost all arc-related δ13C values are negative (reflecting end-member mixtures of CO2 which are all ≤0‰) then this positive value is clearly anomalous. Consequently, we reject it from further consideration.
 Likewise, we measure exceptionally low δ13C values in two samples from Soputan volcano: the two Aeseput samples (IND-13 and IND-14) have exceptionally low δ13C values of ∼−20‰. This observation, coupled with air-like He-isotope ratios, leads us to believe that these samples have been compromised. Interestingly, the carbon is not air-like (air CO2 has a δ13C ∼−8‰ [Keeling, 1984]). Prolonged outgassing may be responsible for these extraordinarily light C isotope ratios [Gerlach and Taylor, 1990]. Again, we remove these values from further consideration.
5.1.5. Poor Agreement Between Duplicate Samples
 There are a number of samples that show poor agreement between duplicate samples. For example, the 3He/4He ratio of IND-19 from location 1 in Ruang crater is rejected as a result of its low He/Ne and 3He/4He ratios compared to a duplicate sample (IND-20) collected at the same locality. Similarly, sample I-150 from Kakaskasen (Mahawu volcano) has an anomalously high CO2/3He value (53.8 × 109), approximately ten times that of a duplicate from the same locality (6.4 × 109), and other samples from the same volcano (∼5.5 × 109). Additionally, its δ13C value, and the δ13C value of its duplicate (IND-3), is much lower than other samples from Mahawu (average δ13C = −3.3‰). It seems reasonable therefore to assume that both samples from Kakaskasen do not reflect the primary magma sources in the region.
 Out of a total of 26 geothermal fluid samples that were collected along the strike of the Sangihe Arc, 15 samples have experienced sufficient modification that both their He and C systematics no longer reflect primary magma characteristics. These samples are: IND-15, IND-16, I-048 (Awu); IND-17 (Karangetang); I-150, IND-3 (Mahawu); IND-12, I-016, IND-13, IND-14 (Soputan); and IND-9, I-198, IND-10, I-193, IND-11 (Ambang). Of the remaining 11 samples, there is no evidence of modification for either He or C for the following nine: I-043 (Awu); IND-18 (Karangetang); IND-20 (Ruang); IND-1, I-159 (Lokon); IND-4, IND-5, IND-6, IND-7 (Mahawu). The remaining 2 samples have unmodified data for either He or C (CO2/3He and δ13C) but not both.
5.2. Along-Strike Variations in He-C Characteristics
 In this section, all interpretations are based on the filtered data from section 5.1. Figure 2 shows 3He/4He, CO2/3He and δ13C as a function of latitude along the Sangihe Arc. Here, we consider both variations along-strike as well as the magnitude of the absolute values in relation to averages of arc-related volcanism and MORB worldwide.
5.2.1. The 3He/4He Variations
 The majority of arc-related volcanism is characterized by 3He/4He values coincident with that found in MORB mantle [Poreda and Craig, 1989; Hilton et al., 2002]. Some localities, however, record the addition of radiogenic He, either through slab-derived contributions [Hilton and Craig, 1989; Hilton et al., 1992; Marty et al., 1994] and/or contamination by arc crust [Hilton et al., 1993; Gasparon et al., 1994]. If the filtered database (section 5.1) allows consideration of magmatic 3He/4He characteristics only, then the Sangihe Arc comprises relatively low values in the northern section of the arc (Awu and Karangetang; 5.4–6.4 RA) versus more typical arc-like values in the southern segment (Ruang, Lokon and Mahawu; 6.9–7.5 RA). Addition of a small radiogenic He component to magmas of the northern Sangihe Arc may reflect contributions from either the subducted sedimentary veneer, its underlying oceanic basement or the overlying arc crust. It seems unlikely that sediments have the transport capacity (due to diffusional losses) to subduct He [Hilton et al., 1992; Hiyagon, 1994] so we can discount this possibility with a fair degree of confidence. Preliminary trace element and Sr-Nd-Pb data do not show any evidence for upper-crustal contamination along the entire strike of the Sangihe Arc [van der Meer et al., 2002]. This observation is in contrast to the southern Lesser Antilles Arc where low 3He/4He ratios are accompanied by radiogenic Sr and Pb isotope ratios [van Soest et al., 1998, 2002]. In the absence of any other constraints, therefore, we conclude that the source of radiogenic He in northern Sangihe Arc magmas may be the subducted crustal basement which modifies 3He/4He values to below typical arc values (see further discussion in section 5.4).
5.2.2. The CO2/3He Variations
 In a manner similar to the He isotope distribution, CO2/3He values of geothermal samples from Mahawu and Lokon 4–8 (×109) are typical of arc-related volcanism: all samples fall close to the worldwide arc average of 15 ± 11 × 109 [Sano and Williams, 1996]. Ruang has a slightly lower than average value (CO2/3He ∼ 2 ± 1 × 109); nevertheless, its value is close to the lower range of arc lavas. Therefore the three southern volcanoes show similar, arc-like CO2/3He characteristics.
 In contrast, the northern arc samples (Awu and Karangetang) have CO2/3He values significantly higher than the southern arc (Figure 2). The CO2/3He values are 64–180 (× 109), or between 1 and 2 orders of magnitude greater than the southern segment of the arc. Higher CO2/3He values in the northern arc would suggest a diminished mantle 3He input, a greater subducted slab influence, or both. A diminished mantle 3He input and/or a greater slab flux might also be expected to result in a lower 3He/4He ratio, as observed for this section of the arc.
5.2.3. The δ13C Variations
 Although δ13C values of all samples are higher than typical mantle values (∼−6.5‰ [Sano and Marty, 1995]), there is a marked distinction in δ13C values between the northern and southern segments of the arc. The two volcanoes in the northern arc (Awu and Karangetang) have δ13C values ≥−2‰, whereas all three southern segment volcanoes (Ruang, Lokon and Mahawu) are characterized by δ13C values ≤−3‰. It is interesting to note that if some of the rejected data from Soputan and Ambang are included, then the pattern between the northern and southern segments remains unchanged as these two volcanoes also have low values of δ13C.
5.2.4. Along-Strike Variations in He-C: Summary
 Along-strike profiles in 3He/4He, CO2/3He and δ13C of Sangihe Arc volatiles show a consistent pattern: the southern Sangihe Arc volcanoes (Ruang, Lokon, and Mahawu) show typical arc values in all three He-C parameters whereas the northern volcanoes (Awu and Karangetang) have higher CO2/3He and δ13C, and lower 3He/4He ratios relative to the southern segment. The results are consistent with either addition of a strong crustal input (presumably slab-derived) in the northern segment of the arc or a reduction in the mantle contribution. The boundary between the two segments lies between the islands of Ruang and Karangetang.
5.3. Differentiating CO2 Sources in Sangihe Arc Magmas
 Following the methodology of Marty et al.  and Sano and Marty , we can resolve the total CO2 observed in any particular Sangihe Arc sample into its component structures. We adopt the same end-member compositions of previous workers: namely, MORB mantle (CO2/3He = 2 × 109; δ13C = −6.5‰); limestone (sedimentary carbonate and altered oceanic basement) (CO2/3He = 1013; δ13C = 0‰); and organic sediment (CO2/3He = 1013; δ13C = −30‰). In Figure 3 we plot He-C results from the present study along with the end-member compositions joined by two binary mixing trajectories with mantle-derived carbon common to both mixtures. It is clear that Sangihe Arc samples do not fall on either of the binary mixing trajectories. This would imply that all three end-members must contribute to the total CO2. As the majority of samples lie close to the M-L binary mixing line, an alternative explanation is that the Sangihe samples are indeed binary mixtures of mantle- and limestone-derived carbon with some samples experiencing isotopic fractionation to lower δ13C [Snyder et al., 2001]. Dewatering of metabasalts at the slab interface and/or precipitation of calcite within the hydrothermal systems are two possibilities that may lead to lower δ13C values [Snyder et al., 2001]. Given the observation that some samples (e.g., Lehi and Temboko) appear to plot subparallel to the M-L mixing line (Figure 3), then the extent of isotopic fractionation must be approximately equal in all cases irrespective of possible along-strike variations in thermal regime at the slab interface (see next section) or differences in fluid chemistry at the various volcanic geothermal systems. Both of these possibilities seem unlikely. A more compelling argument comes from the N-isotope systematics which show some volatile contributions from organic sedimentary material along the entire strike of the Sangihe Arc (L. E. Clor et al., Volatile and N-isotopic chemistry of colliding island arcs: Tracing source components along the Sangihe Arc, Indonesia, submitted to Geochemistry Geophysics Geosystems, 2004) (hereinafter referred to as Clor et al., submitted manuscript, 2004). We conclude therefore that addition of sedimentary-derived C is responsible for δ13C values lower than predicted by simple binary M-L mixing.
 In this case, the contribution of each end-member to the total CO2 can be quantified by use of the following equations [Sano and Marty, 1995]:
where o = observed and f is the fraction contributed by L, S and M to the total carbon output.
 In Table 2, we detail the fractional contributions of the L-, M- and S-components to the total CO2 inventory as well as the ratios of (a) limestone (carbonate) to organic sediment carbon input (L/S), and (b) slab-derived carbon (L + S) to carbon of mantle derivation (M). The first and most obvious point to note is that carbon is predominantly of carbonate derivation throughout the Sangihe Arc. This could reflect contributions from both sedimentary carbonate in subducted sedimentary sequences as well as carbonate contained within the oceanic basement (e.g., as calcite veins). It is not unusual for arc-type lavas to be dominated by a carbonate component [Sano and Marty, 1995; Hilton et al., 2002]. However, it is significant that the proportion of carbonate-derived carbon is consistently >90% in the northern section of the arc, and considerably lower in the southern segment. Indeed, the southern Sangihe arc has a carbonate-derived carbon contribution similar to the average value seen at arcs worldwide (∼75%).
Table 2. Limestone-Mantle-Sediment Contributions to CO2 Inventory of Sangihe Arc Geothermal Fluids
 Although there is an enhanced contribution of carbonate-derived carbon to the total carbon inventory in the northern Sangihe Arc, there is (with the exception of Awu volcano) little difference in the ratio of L/S in the volcanic output, i.e., the fraction of carbon of carbonate derivation versus organic (sedimentary) carbon. We note that the same observation characterizes the Nicaragua segment of the Central America Arc compared to the adjacent Costa Rica sector [Shaw et al., 2003]. This may simply reflect uniformity in this parameter in the sedimentary input. In the case of Central America, organic carbon is dispersed throughout the subducted sedimentary pile so that it is insulated against thermal loss during early stages of subduction [Shaw et al., 2003, and references therein]. The same may hold true in the Sangihe Arc input.
 Finally, the Sangihe Arc presents a consistent picture of an enhanced slab-derived input in the north relative to carbon from the mantle wedge. The ratio (L + S)/M is considerably higher in the northern volcanoes (>40) compared the southern section of the arc (<4.3). Again, the same observation was made for the Nicaragua segment of the Central America Arc compared to the adjacent Costa Rica sector [Shaw et al., 2003]. In Central America, the difference in (L + S)/M was ascribed to differences in thermal regime along the strike of the arc with steeper subduction in Nicaragua leading to a colder slab retaining its carbon inventory until subarc depths. In contrast, the warmer thermal regime in Costa Rica leads to loss of carbon from the slab at shallower (fore-arc) depths thus lowering the relative amount of available slab-derived carbon. This possibility, i.e., a thermal control on the relative contribution of slab-derived carbon, plus other alternative explanations for along-strike differences in relative carbon contributions, is discussed in more detail in the following section.
5.4. Implications for Tectonic Development of the Sangihe Arc
 In this section, we explore the possibility that differences in the volatile systematics observed between the northern and southern sections of Sangihe Arc ultimately have a tectonic origin and are related to the collision and development of the arc. In this context, we note that the Sangihe and Halmahera arcs have already collided in the north causing volcanism to cease north of 4°N [e.g., Pubellier et al., 1991], and that the collision is propagating southward. This has caused the subduction rate to decrease along the active arc [Elburg and Foden, 1998]. Therefore we might anticipate that any variability in volatile chemistry related to the collision may be more noticeable in the northernmost section of the arc, closer to the locus of collision. We speculate that the observed along-strike variations in He-C characteristics could therefore reflect (1) an increase in the volume of subducted sediment in the northern arc; (2) variability in the composition of subducted sediment chemistry; or (3) a change in thermal regime experienced by the subducting slab related to the onset of collision. These possibilities are discussed in turn.
 1. Volume of subducted sediment: The massive accretionary wedge (∼15 km thick) formed by the collision of the Sangihe and Halmahera arcs is thickest in the northern Molucca Sea where collision is oldest [e.g., Hall, 1996]. These sedimentary sequences represent the fraction of the total sediment load involved in the collision that have obducted onto the fore-arc region. Studies at other accretionary arcs worldwide [von Huene and Scholl, 1991] estimate that only ∼20% of the sediment on the incoming plate forms the prism, whereas 80% is subducted into the mantle. Although the He-C systematics of this work are consistent with recycling of sediment-derived carbon through the entire arc system, we speculate that the thicker incoming sequences in the northern arc translates into a greater absolute volume of sediments that is subducted. Under these circumstances, then it is conceivable that the contribution of mantle wedge derived carbon could be overwhelmed by the large sediment-derived carbon contribution in the north, and this would lead to the enhanced CO2/3He and (L + S)/M ratios observed at Awu and Karangetang.
 2. Variability in sediment composition: Given the relatively high C isotopic ratios measured in the northern arc, we suggest that the subducted sediment in this region contains little hemi-pelagic, organic-rich sediment. This conclusion is reinforced by nitrogen isotope data from the same Awu locality that was sampled for the present sample suite. The δ15N value of ∼−3.3‰ (Clor et al., submitted manuscript, 2004) indicates that there is some but not much organic-derived sedimentary N contributing to the volatile flux in this region (organic N is characterized by δ15N values ∼+6 to +7‰ [Peters et al., 1978; Kienast, 2000]). Thus N in the northern arc has a predominantly upper mantle origin. Carbonate is essential devoid of N [see also Fischer et al., 2002], and therefore slab carbonate addition would not mask a mantle-like N isotope signature. In combination with the high CO2/3He and (L + S)/M, it is reasonable to conclude that there is a strong slab contribution in the north, dominated by pelagic or crustal carbonate. In contrast, the southern arc samples have heavier N isotope signature (average δ15N ∼−2.5‰) and higher N2/He, consistent with a more pronounced contribution of hemipelagic sediment derived N2 (Clor et al., submitted manuscript, 2004). Thus there is geochemical evidence for some degree of heterogeneity in the slab composition between the two segments of the arc.
 One possibility to account for the heterogeneity in sediment composition along the strike of the Sangihe Arc is the process of off-scraping; i.e., loss of the uppermost sedimentary veneer during subduction. Fischer et al.  interpreted variations in δ15N between Costa Rica and other segments of the Central America Arc by loss of hemipelagic (N-bearing) and shallow subducted sediments by accretion to the over-riding plate. If this scenario is applicable to the Sangihe Arc, then loss of the hemipelagic sediment would have to be greater in the northern segment of the arc, consistent with the increased thickness of the accretionary wedge toward the collision. In support of this possibility is the observation that the L/S ratio is extremely high at Awu volcano (L/S = 87.4) and then becomes more constant to the south (L/S ∼ 13). The relatively large distance between Awu and Karangetang volcanoes may account for the fact that the effect of off-scraping is not yet observed at volcanoes other than Awu.
 3. Enhanced fluid/melt generation: An enhanced slab signature (i.e., high (L + S)/M) may result from more efficient heating of the slab rather than the addition of larger volumes or different compositions of sediment. In this scenario, the slow-down or cessation of collision in the northernmost arc (as seen from seismic evidence [e.g., McCaffrey, 1983; Pubellier et al., 1991]) would lead to enhanced heating of the slab [Peacock et al., 1994], thereby promoting greater production of melt and/or fluid. Additionally, on the basis of mineral chemistry, Morrice and Gill  propose a northward increase in the depth of melting.
 Along the northern extension of the Sangihe Arc (central Mindanao, The Philippines), there is extensive volcanic activity formed by post-collisional magmatism. Sajona et al.  describe these magmas as subducted basaltic crust-derived melts, or adakites. Their study attributes the slab melting to thermal rebound of previously depressed geotherms upon cessation of subduction. In the northernmost Sangihe Arc (south of Mindanao), where subduction-related volcanism is still active but where the subduction rate is slowing [e.g., Pubellier et al., 1991], an analogous situation of more efficient heating of the subducting slab would be expected to enhance the slab contribution relative to the (mantle) wedge input. Elburg and Foden  have suggested that the Sangihe Arc has evolved from a system of fluid-derived contributions to magmagenesis to a system dominated by sediment-derived melts. They indicate that this temporal change in apparent source composition is due to the reduced rate of subduction and therefore superheating of sediments on the slab. The CO2/3He and (L + S)/M values in both Awu and Karangetang geothermal fluids, 1–2 orders of magnitude higher than average volcanic arcs, are consistent with the notion of a much more efficient transfer of a slab component to the northern Sangihe volcanics. Furthermore, the anomalously high L/S value of Awu could indicate that the thermal regime is sufficiently hot so that a significant fraction of CO2 is derived from a marine carbonate component within the oceanic basement (i.e., the source of the L-component). Kerrick and Connolly  show that significant slab decarbonation is limited to localities with high-T geotherms; therefore an enhanced CO2 flux in the northern segment of the arc may indicate initiation of slab melting, as proposed by Sajona et al. . In this scenario, the helium-carbon results are tracing changes in the thermal regime of the subducted slab as the collision complex propagates south along the Molucca Sea Plate margin.
 This work presents the first He and C results of geothermal fluids from the Sangihe Arc, Indonesia, giving insight into how volatiles are mobilized in an arc-arc collision zone. The following conclusions are emphasized:
 1. The 26 geothermal and 2 mineral samples collected along the north-south strike of the Sangihe Arc contain air-, crustal-, subducted-slab and/or mantle-derived volatiles. Samples showing either air contamination (throughout the arc) or a significant crustal signature (e.g., Ambang volcano) are identified using a number of criteria (low 3He/4He, low He/Ne values, high CO2/3He, and extreme δ13C), and are distinguished from samples that characterize the parental source magmas.
 2. He-C isotope and relative abundance results of this investigation suggest that the oblique collision of the Halmahera and Sangihe arcs has defined a distinctive pattern of variability in these parameters along the north-south strike of the Sangihe Arc. The northern volcanoes (Awu and Karangetang) show higher CO2/3He and δ13C, and lower 3He/4He than the southern volcanoes (Ruang, Lokon and Mahawu): the transition in He-C systematics occurs between Karangetang and Ruang volcanoes.
 3. Resolving the CO2 output into primary source component structures (mantle- carbonate- and organic-derived carbon) indicates that there is a dominant carbonate influence (>90%) in magmas supplying the northernmost volcanoes (Awu and Karangetang). In contrast, the southern arc volcanoes (Ruang, Lokon, and Mahawu) contain carbonate contributions (∼63%) more typical of average arc magmas (∼75%).
 4. The northern arc volcanoes record an enhanced slab contribution to the carbon inventory relative to that derived from the mantle wedge. We discuss three possible reasons for this observation: (1) increase in sediment volume contributing to arc volcanism; (2) variations in sediment composition; and (3) enhanced heating of the slab resulting from cessation of collision in the northernmost portion of the arc.
 This work was supported by the National Science Foundation (grant EAR-0100881 to DRH). Additional funds for field expenses came from UCSD (Earth Sciences Program to LJ) and UNM (Research Allocations Committee to TF). We thank Justin Kulongoski and Pat Castillo for valuable support in the field, Mark Erdmann for arranging accommodations in Manado, and Purnama Hilton for help with LIPI in Jakarta. Alison Shaw, Martin Walhen and Bruce Deck all helped with laboratory analyses. Colin Macpherson, Robert Hall and the SE Asia Research Group, London, kindly supplied Figure 1. We thank J. Varekamp, J. Foden, G. Bebout (Guest Editor) and one anonymous referee for useful comments. We also thank GRDC director Bambang Dwiyanto (Bandung) for his support of the project.