4.1. Downhole Variations in N and C Concentrations and Isotopic Compositions
 Sediments are highly variable in N concentration and isotopic composition and these variations have been used to trace a wide range of processes in organic-rich sediments, including diagenesis [e.g., Sweeney et al., 1978; Peters et al., 1978; Rau et al., 1987; Libes and Deuser, 1988; Altabet and Francois, 1994; Minoura et al., 1997; Muller and Voss, 1999; Ettwein et al., 2001]. Many of these studies were carried out in relatively organic-rich systems (compared with the section at Site 1149), very near the sediment-water interface [e.g., Peters et al., 1978; Minoura et al., 1997] or nearer continents, and relatively little attention has been paid to N behavior in deep-sea sediment sections such as that at Site 1149 (notable exceptions in the literature, but lacking N and C isotope data, are the studies by Muller , Waples and Sloan , and Waples ). Here we will attempt to outline some of the possible reasons for the shifts in concentration and isotopic composition of N and C with depth at Site 1149.
4.1.1. Oceanographic Considerations
 The decrease in δ15N with depth at Site 1149 could in part reflect change in δ15N of the primary sediments being deposited at this site through time. An increase of δ15N with time requires an increase in productivity or a decrease in nutrient supply that makes the preferential uptake of 14N by organic matter less efficient [see Ettwein et al., 2001]. Samples from 90 mbsf have δ15N values similar to those at greater depths at Site 1149, and samples at around 50 mbsf are clearly higher in δ15N than those below 90 mbsf. Therefore the shift in δ15N (and less clear trends in N and C content and δ13C) would have begun within that period of deposition (Table 1, Figure 4), corresponding to ∼2.5 to 3.5 Ma [Lozar and Mussa, 2003]. The most likely change in oceanographic factors around this time is the increase in the Kuroshio current (due to the closure of the seaway between North and South America) which is documented in the biostratigraphic record at Site 1149 by three events between 75.11 and 68.46 mbsf [Lozar and Mussa, 2003]. The general motion of the site toward the higher productivity waters of this current could help account for the gradual increases in δ15N, δ13C-org, and N and C concentrations. However, the region known to have been strongly affected by the Kuroshio current is extremely high in productivity, with most samples containing 0.5 to 1.5 wt% C as documented during ODP Leg 186 [see Mora, 2002]. Furthermore, although the waters affected by the Kuroshio current are higher in productivity, which could cause an increase in δ15N, they are also higher in nutrient supply, possibly mitigating the effects of productivity on marine δ15N (see discussions by Farrell et al. , Milder et al. , Pride et al. , and Ettwein et al. ). Finally, it is unclear whether a change in productivity of this magnitude would produce a trend in Creduced/N as seen in the data for Site 1149 (see discussions by Minoura et al. , Milder et al. , and Pride et al. ).
 Variations in δ15N similar to those at Site 1149 have been attributed in other studies to varying mixtures of terrestrial and marine organic matter [e.g., Peters et al., 1978; Minoura et al., 1997], with the marine component having δ15N near +8‰, with δ13C near −20.5‰, and the terrestrial component having δ15N near +1.8‰, with δ13C near −26.5‰. The sediments analyzed in these two studies were deposited in settings more proximal to continental sources (for Peters et al. , on the NE Pacific continental shelf; for Minoura et al. , near Japan in the Japan Sea), and we consider the significant additions of terrestrial organic matter to the Site 1149 sediments (and certainly the sediments obtained on Leg 129) less likely. It is uncertain whether the volcanic ash-rich sediments deposited at Site 1149 contained appreciable terrestrial organic matter during their deposition (see discussion of sediment sources at Site 1149 by Urbat and Pletsch ). However, this possible mixing warrants further study, and more detailed biogeochemical work could help elucidate the sources of the organic matter [e.g., Madureira et al., 1997; Meyers and Doose, 1999; Shipboard Scientific Party, 2000]. In Figure 6, we compare the δ15N-δ13C data for the upper part of the Site 1149 section with the “Marine” and “Terrestrial” organic end-members proposed by Minoura et al.  (very similar to the end-members proposed by Peters et al. ). From this comparison, in particular based on the relative scatter of the Site 1149 data about the mixing line for these two end-members, it is apparent that the Site 1149 sediments do not show a simple marine-terrestrial mixing behavior. The Site 1149 data do show some correlated variation, however (see Figure 6; noted above), and when three outliers are removed from consideration (indicated in boxes on Figure 6), the data show a linear relationship (r2 = 0.60) with a slope quite different from that of the mixing line of Minoura et al. . The data in Figure 6 for Franciscan metasedimentary rocks are discussed in section 4.2. Finally, vascular land plants typically have C/N higher than that of algae [see Meyers and Doose, 1999, and references therein], and a down-section increase in the terrestrial organic component would thus have likely resulted in an increase in C/N rather than the observed decrease in C/N at Site 1149 (see Figure 5c). Furthermore, the Creduced/N values for the Site 1149 sediments (2–9 on atomic basis) fall in the range for algae (4–10 on atomic basis) and not in the range of significantly higher C/N for vascular land plants (≥20 on atomic basis [Meyers, 1994; Meyers and Doose, 1999]).
Figure 6. Plot of δ15N vs. δ13C of the upper part of the sediment section at Site 1149, compared with the mixing line (thicker, red line) between “Terrestrial” and “Marine” organic components proposed by Minoura et al. [1997; cf. Peters et al., 1978]. When the three outliers indicated in the small boxes (the sediment sample from 1 mbsf, and two bioturbated clays from near 135–142 mbsf) are removed from consideration, an r2 of ∼0.60 is obtained for the Site 1149 data (see thinner, orange line), but the slope of this line differs significantly from that of the mixing line of Minoura et al. . The data for low-grade Franciscan Complex metasedimentary rocks plot near the “Terrestrial” organic component of Minoura et al. , consistent with the known deposition of the Franciscan sediments near the continental margin in W. North America.
Download figure to PowerPoint
4.1.2. Diagenetic Processes
 A third possibility is that the shift in δ15N values in the sediments recovered from the upper part of the Site 1149 section (Site 1149A), with δ15N decreasing steadily within the uppermost 120 m (Unit I) from +8.2‰ at 1.4 mbsf to +4.7 at 113 mbsf (Figure 4b), is due to complex diagenetic processes, conceivably with superimposed more minor effects of the two oceanographic alternatives presented above (changes in productivity related to ocean currents and/or deposition of differing proportions of marine and terrestrial organic fractions). The down-section change in δ15N at Site 1149 is accompanied by a decrease in total N concentration (Figure 4a), which could reflect the loss of N with δ15N significantly higher than +8‰, perhaps a nitrate component. Reduced-C concentrations and δ13C also decrease slightly with increasing depth in the upper 150 m of site 1149, contributing to a subtle down-section trend in Creduced/N, and indicating some parallel behavior throughout the organic reservoir. Together, the shifts in reduced C, total N, and Creduced/N with increasing depth at Site 1149 are compatible with the shifts in the same parameters over similar depth horizons reported for other Pacific Ocean deep-sea sediment cores and attributed to diagenesis [Muller, 1977; Waples and Sloan, 1980; Waples, 1985] (unfortunately, N and C isotope data were not obtained in these previous studies). A full explanation of these organic geochemical phenomena, and a more comprehensive reconstruction of diagenesis in this section, would require analyses of separated organic and inorganic N fractions (see studies by Muller , Waples and Sloan , and Williams et al. ) and a more comprehensive consideration of the chemical compositions of the interstitial waters from the sediment cores. However, some reasonable speculation would be that the shift in δ15N, and the associated other variations in the C-N signature, are the result of diagenesis, perhaps a microbially mediated reprocessing of organic matter such as that suggested to occur in algal mats [Lehmann et al., 2002].
 The results of Lehmann et al.  deserve particular notice in this context for the similarity of results despite experimental conditions that differ significantly from the natural conditions in Site 1149. This study consisted of experimental simulations of early diagenesis of algae under varying redox conditions. These authors found minor shifts in concentration of C and N, during early diagenesis, and fairly significant changes in isotopic composition, including a shift in δ15N of ∼3‰ under anoxic conditions after 60 days. The similarity of the results of our study, showing shifts in δ15N of ∼3‰ in a 120 m sediment column and the Lehmann et al.  study conducted at different pressures and temperatures, and at very different timescales, suggests that a shift in δ15N could be a common feature of early diagenesis in off-trench, deep-sea sediment sections, and that the magnitude of the negative shift (depending on a variety of conditions, notably the redox state) could conceivably be on the order of ∼1–3‰. At Site 1149, the release of organic N, presumably related to the N and δ15N shifts reported here, is recorded by the slight enrichments (up to ∼200 μM) in dissolved NH4+ in the interstitial waters, the overall chemistry of which has led to the conclusion that the Site 1149 diagenetic environment is slightly suboxic [Plank et al., 2000]. Dissolved NH4+ in the waters (see Figure 4c) reaches a maximum through the upper 50 mbsf, reflecting N release during organic matter degradation, and the decreased dissolved NH4+ below this interval reflects enhanced uptake during clay diagenesis [Plank et al., 2000].
4.2. Comparison of N Signature in Deep-Sea Sediments With That in Paleoaccretionary Rocks
 Although δ15N is relatively high and highly variable in surface/near-surface oceanic sediments (with some values >+10‰; see Figure 2), the low grade, forearc metasedimentary rocks studied to date appear to have a much narrower range of δ15N values mostly lower and between +1 and +3‰ [Bebout and Fogel, 1992; Bebout, 1997; Bebout et al., 1999a; Sadofsky and Bebout, 2003] (see comparisons in Figure 2). This “disconnect” between the δ15N values of many oceanic sediments (see Figures 23–4), mostly from relatively near the sediment-water interface, and the values for low-grade metamorphosed deep-sea sediments in forearc metamorphic suites (representative data in Figure 2) could result from several factors. First, many of the deep-sea sediment sections analyzed previously and in the present study, are not representative of the near-continent, trench settings with high sedimentation rates resulting in the voluminous trench sediment and accretionary prisms sampled by the forearc metamorphic rocks (e.g., Catalina Schist, Franciscan Complex, Western Baja Terrane [see Sadofsky and Bebout, 2003]). Enhanced contributions of terrestrial organic matter (with lower δ15N and δ13C than marine organic matter) must be considered a factor when comparing the seafloor sediment data with the N and C isotope data for the circum-Pacific forearc metamorphic rocks as the metamorphosed voluminous clastic sediments in these metamorphic suites were deposited near a continental margin. Correspondingly, the δ15N values or these low-grade forearc metamorphic rocks generally fall in the range of +1 to +3‰, and with low δ13C, quite near the “Terrestrial” organic component proposed by Peters et al.  and Minoura et al.  (see Figure 6). As another alternative, diagenesis (more pronounced with increasing depth), and conceivably also, incipient low-grade metamorphism during very early subduction history (perhaps in part accompanying mechanical fluid expulsion at shallow levels of accretionary prisms), shifts the δ15N of the initially high-δ15N seafloor sediments from their near-surface values (δ15N near +8 per mil [see Libes and Deuser, 1988; this study] (Figure 4) toward values similar to those observed for the low-grade metamorphic equivalents, which represent subduction to depths of ∼5–40 km. We propose (see section 4.1.2) that the variation in δ15N and δ13C (Figure 6) and C and N concentrations and Creduced/N) in the Site 1149 sediment section largely reflects diagenesis, possibly with some lesser superimposed variation due to changes in primary productivity and marine-terrestrial organic sources. However, it appears extremely likely that the lower whole rock δ15N (and the δ13C of the reduced C) of the forearc metasedimentary rocks, relative to that for many marine sediments and most of the sediment section at Site 1149 (see comparison in Figure 2), is largely attributable to greater proportions of terrestrial organic matter in the forearc metamorphic suites deposited nearer a continental sediment source (perhaps with superimposed minor effects of diagenesis).
4.3. Mass Balance Calculations and Implications for Fluxes Into Subduction Zones
 Because there are significant variations in the concentrations and isotopic compositions of N and C throughout these sedimentary sections, the calculation of the fluxes of these elements into the IBM subduction zone is not straightforward. The approach that we favor is to use shipboard estimates of the lithologic units, including the thickness of these units (and average densities of the materials), and average all measured values for each lithological layer (see Table 2). We present only data from Site 1149 in these calculations. We will assume that a section like that at Site 1149 is subducted and not take the apparent shift in shift in δ15N as a function of depth into account at this time, because of the difficulty of deciding the cut-off for which samples to include and which to leave out. We suspect that the actual δ15N and N concentration could be slightly lower than the values used in the calculations we present below, but that the magnitudes of the differences produced by these small changes are at this point within error in any likely application of these data (<±0.5‰ for δ15N, <5% relative for the mass of N subducted).
Table 2. Estimated Annual Sedimentary N-C Subduction Flux at the Izu Margin
| ||Thickness, m||N, ppm||δ15NAir||Creduced, wt.%||δ13CVPDB||Average Densitya|
|Ash and diatom/rad clay (1149 Unit I)||120||371||5.8||0.22||−22.8||0.641|
|Dark brown pelagic clay (1149 Unit II)||60||261||4.5||0.09||−25||0.695|
|Chert and marl layers (1149 Units III, IV, and V)||230||22||3.2||0.02||−24.8||2.04|
| || ||N, g||δ15NAir||C, g||δ13CVPDB|| |
|Fluxes (per Linear km)|
|Fluxes (for C only reduced C)|| ||2.5E + 06||5.0||1.4E + 07||−24.0|| |
|Calcite C flux, marl layers|| || || ||9.2E + 08||2.3|| |
|Total C flux (full section)|| || || ||9.3E + 08||1.9|| |
|Total Fluxes (Over 1050 km Linear Trench Length)|
|This study|| ||2.6E + 09||5.0||9.8E + 11||1.9|| |
|Hilton et al. || ||5.8E + 09||7.0 subarc||6.2E + 11||0, −20b|| |
 To calculate the fluxes of C and N entering the subduction zone at the Izu trench, we begin by separating the materials into three general categories chosen to reflect the resolution of the sampling in this particular study. These categories are: Ash and diatomaceous clay (∼120 m thick, Unit I at site 1149); Dark brown pelagic clay (∼60 m, Unit II, 1149); Radiolarian chert, zeolite-rich clay and marl layers (∼230 m). Multiplying the thicknesses of these units by the orthogonal convergence rate of 5 cm/yr (see compilation of convergence rates by Plank and Langmuir ), and the average dry densities for each unit (values from Plank et al. ) allows a simple calculation of the mass of each lithology entering the subduction zone. Note that the volumes are multiplied by dry density to remove porosity (quite high in the upper 180 m) and pore water from the discussion, we assume for simplicity that all pore water is lost during the early stages of subduction. Average concentrations of C and N can then be combined with our knowledge of the mass of that lithology being subducted to determine the masses of C and N entering the subduction zone in solids. Simple averages of the δ13C and δ15N values in each lithology are used to calculate an average isotopic composition for each unit,. The mass of carbonate was calculated by averaging the shipboard analyses of carbonate content for the mixed carbonate/siliceous layers, combined with the isotopic compositions presented in Table 1.
 On the basis of these mass balance calculations, carried out using the orthogonal convergence rate of 5 cm/yr [Plank et al., 2000] (Table 2), we suggest that a sedimentary section like that at Site 1149 delivers an annual flux into the subduction zone of 2.5 × 106 g of N and 1.4 × 107 g of reduced C per linear km of trench, with average δ15N of +5.0‰ and δ13C of −24.0‰, respectively. In addition, 9.2 × 108 g/yr of C is subducted per linear km in carbonate-rich layers of 1149B with average δ13C of +2.3‰. Because of the relatively low C content of the clastic units and the fairly thick carbonate section, the carbonate (oxidized C) budget overwhelms the reduced C budget at this site and 9.3 × 108 g/yr per linear km are input into the subduction zone with an average δ13C of +1.9‰ near that of the carbonate-rich horizons. The overall Ctotal/N of the subducting sediment is near 370:1 (including both reduced and oxidized C reservoirs), the Coxidized/Creduced is ∼65:1, and the Creduced/N (organic component) is ∼5.6:1. The total N flux into the Izu-Bonin convergent margin, for a trench length of 1050 km [from Plank and Langmuir, 1998], is calculated at 2.62 × 109 g/yr, considerably smaller than the flux of 5.8 × 109 g/yr calculated by Hilton et al. , who estimated concentrations for subducting sediments (using 100 ppm for the entire sediment section of 400 m) and used a somewhat larger sediment flux into this margin (4.18 × 1013 g/yr compared with our use of 3.09 × 1013 g/yr).
 A more complete inventory of the incoming N and C budgets in the IBM convergent margin will require analyses of AOC from Sites 801 and 1149. δ15N for MORB glass from the East Pacific Rise ranges from −1.0‰ to −6.5‰ (mean of ∼−3.0‰, n = 7 [Marty and Humbert, 1997]), with N2 concentrations of 3–65 × 10−10 moles/gram (<0.5 ppm) and concentrations of NH4+ in spilites (perhaps analogues to some seafloor basalts) from SW England are as high as 182 ppm (range of 1–182 ppm, with mean N concentration of 53 ppm [Hall, 1989, 1990]). These concentrations may represent the extremes for fresh seafloor basalt (with the baseline of 0.01–1.0 ppm N) and the most highly altered oceanic basalt (with up to ∼200 ppm N), and the alteration in AOC at Sites 801c and 1149 is likely to be intermediate in extent. The high K2O values reported for the crustal sections from ODP Site 801c (which experienced an increase of K2O by 17% due to alteration) and 1149 (which is more heavily altered than 801c) provide an indication that AOC may contain appreciable N that could figure significantly in the subduction flux models. This may be especially true when one considers the large volume/mass of oceanic crust being subducted (∼6 km section); any elevation in N concentration (due to hydrothermal alteration) above the maximum ∼0.5 ppm levels observed in fresh basalt glass [Marty and Humbert, 1997] could result in the subduction of N in AOC rivaling that subducted in thin, shale-poor (e.g., carbonate or chert rich) sediment sections (see Bebout [1995, Table 2] and discussions below). However, these elevated N concentrations are likely to occur only within the uppermost, more hydrothermally altered 1 km of the subducting oceanic crust (see discussion below).
 Our study demonstrates the importance of evaluating subduction inputs through analyses of the materials outboard of individual trenches and thought to be subducting in any attempts to chemically mass balance individual convergent margins. As in the case of the considerations of pre-subduction sediment δ15N, which can vary with a number of sedimentological and diagenetic factors (discussed above), the use of a single sediment N concentration as applicable in considerations of inputs and outputs at multiple convergent margins from quite different tectonic and sedimentological environments could easily yield flawed results. Our C and N input flux estimates, based on analyses of the Site 1149 section, differ significantly from recent estimates of C-N inputs based on generalized assumptions regarding C and N concentration made without the benefit of data for the sediment at this margin (latter estimates by Hilton et al. ; see comparisons in Table 2). Attempts to deduce overall efficiencies of N and C return in arc volcanic gases are highly susceptible to uncertainties in the input (and output) estimates, as can be demonstrated for the Central American arc-trench system for which Fischer et al.  recently claimed extremely efficient volcanic arc return of sedimentary N. Using N concentration data for only the upper 160 m of the section at Site 1039, Li et al.  obtained a sedimentary N subduction rate of 9.9 × 109 g/yr (for the 1100 km of trench length). Use of even a quite low concentration for the lower, carbonate-rich section would significantly increase this flux estimate (work on the lower section is in progress). In comparison, Fischer et al.  employed a smaller input rate of 2.4 × 109 g/yr, using an averaged N concentration of 100 ppm for the upper 175 m of the section (and assuming no N in the lower section) in their comparison with arc N outputs in Central America. Fischer et al.  claimed similarity of their input of 2.4 × 109 g/yr with their estimate of arc output of 4.1 × 109 g/yr, and speculated that this similarity reflects extremely efficient return of subducted sedimentary N.
 It is worth noting that the δ15N of +7‰ used by Fischer et al.  and Snyder et al.  for sediment contributing to arc magmatism is higher than those of both the aggregate sediment sections at Sites 1149 (IB) and 1039 (Central America; bulk δ15N of +5.0‰ and +5.6‰, respectively for the two margins [see Li et al., 2003]) and the shallowly subducted paleoaccretionary rocks (+1 to +3‰; see Figure 2), a difference that could be related to increase in δ15N during metamorphic N losses (see examples of this relationship by Haendel et al. , Bebout and Fogel , Bebout et al. [1999b], and Mingram and Brauer ). These metamorphic N losses would affect the efficiency with which the seafloor sediment N inventory is conveyed to depths beneath arcs, and thus any mass balance employing seafloor sediment as input and arc gases as output. Regarding the input fluxes in subduction zones, it worth noting that all of the estimates of sediment geochemical inputs assume uniformity in the incoming sediment sections along-strike in active trenches, known not to be the case.
 We suggest that the apparent similarity in sediment N input with arc volcanic outputs reported by Fischer et al.  for Central America could be partly coincidental, reflecting significant contributions from both sediments and devolatilizating AOC or significant errors in input or output estimates. Given the uncertainty of this N flux in AOC (see consideration by Bebout ), the overall Central American mass balance of N inputs and arc outputs cannot be addressed in entirety; we estimate a N subduction rate of 1.7 × 1010 g/yr in AOC (four times the sedimentary flux for the same margin), using an estimated N concentration of 10 ppm for the crustal lithology. Use of the total sediment + AOC input (2.7 × 1010 g/yr), based on this 1.7 × 1010 g/yr flux in AOC and the sediment N input flux of 9.9 × 109 g/yr (see above), a ∼15% arc return of subducted N would be indicated (incorporating the arc output rate of 4.1 × 109 g/yr from Fischer et al. ). If the AOC N concentration is reduced to 5 ppm, a ∼22% return in arcs would be indicated, and an ∼36% N return in arcs is indicated if a concentration of 1 ppm N is assumed for the AOC. This crude set of comparisons highlights the dire need for any constraints on the total flux and isotopic composition (for both N and the noble gases) of the AOC volatiles component. Direct use of our information regarding N inputs into the Izu-Bonin convergent margin to mass balance with N outputs in the corresponding arc awaits investigation of arc volcanic gases in this margin.
4.4. Fate of Subducted Sedimentary Nitrogen in the Izu-Bonin Margin (and Other Convergent Margins)
 Studies of low-grade metasedimentary suites in paleoaccretionary complexes in Western North America shed some light on the degrees of deep subduction of N in sediment and the stable isotope compositions of this N. Recent study of devolatilization in sedimentary lithologies subducted to 5–40 km depths in the Catalina Schist, California, the Franciscan Complex, California (Coast Ranges), and the Western Baja Terrane, Mexico [Sadofsky and Bebout, 2003] (see estimated peak P-T ranges in Figure 7), affords an assessment of the entrainment of N into forearc regions of a relatively “cool” subduction zone. Interestingly, samples from the Coastal Belt, the lowest-grade unit in the Coast Ranges subducted to only ∼5 km depths, show correlated δ15N and N concentration (see Figure 8a), conceivably reflecting differential loss of “heavy” N (as NO3−?) during diagenesis and extremely low-grade metamorphism accompanying subduction to these extremely shallow levels [see Sadofsky and Bebout, 2003]. Moving upgrade in the Coast Ranges, the more uniform δ15N of the somewhat higher-grade (more deeply subducted; see Figure 7) Central and Eastern Belt metasedimentary rocks (near +1.5‰) could reflect the more complete loss of this “heavy” component (see data for the three Coast Ranges units in Figure 8a).
Figure 7. Pressure-temperature diagram showing estimates of peak metamorphism for low-grade paleoaccretionary prism rocks from California, USA, and Mexico (see Grove and Bebout  for details regarding the generalized phase equilibria and stability fields compiled on this diagram). Arrows are schematic prograde P-T paths rocks might take in subduction zones, reflecting a wide range in thermal structure; for the Catalina Schist, California, these paths are inferred to reflect rapid cooling in an incipiently formed subduction zone, with warmer conditions first producing the highest-temperature epidote-amphibolite-facies unit (labeled as “EA”) and latest-stage, cool subduction producing the lawsonite-blueschist and lawsonite-albite facies units (labeled in this figure as “LBS” and “LA”, respectively). Various units of the Coast Ranges Franciscan Complex are indicated (“Coastal Belt”, “Central Belt”, and “Eastern Belt”, in order of increasing peak metamorphic pressures thus depths of underthrusting [see Sadofsky and Bebout, 2003; Blake et al., 1987]). Also indicated is the peak P-T for the Franciscan Complex metagreywackes at Pacheco Pass, California (estimates from Ernst ). Fields labeled “ST1”, “ST2”, and “ST3” indicate peak conditions for tectonometamorphic units of the Western Baja Terrane [from Sedlock, 1988], with “ST3” representing the lower-P conditions in this suite.
Download figure to PowerPoint
Figure 8. Nitrogen concentration and isotope compositions and Creduced/N of subduction-zone metasedimentary rocks, and comparison with data for the Site 1149 sediments. (a) δ15N versus whole-sediment N concentration, illustrating differences among the tectonometamorphic units of the Franciscan Complex and Western Baja Terrane, Mexico [from Sadofsky and Bebout, 2003]. (b) Creduced/N versus N concentration for metasedimentary rocks from the same units, demonstrating the extremely uniform Creduced/N for the lowest grade rocks in the Coastal Belt, for which peak metamorphic temperatures are <200°C at approximately 5 km maximum depths of underthrusting. δ15N values are similar to those expected in sediments with organic matter derived primarily from photosynthesizing organisms [see Rau et al., 1987]. Metamorphic devolatilization of N would be expected to produce a trend of increasing δ15N with decreasing N content as N bearing fluids (likely with N2 as the dominant N fluid species) preferentially fractionate the lighter isotope [see Bebout et al., 1999a, 1999b]. Also indicated in this figure is the range of Creduced/N and N concentrations in the Site 1149 sediment section (data for the upper 180 m only).
Download figure to PowerPoint
 Carbonreduced/nitrogen ratios are another useful indicator of the devolatilization of these elements, as both of these elements are provided to the sedimentary rocks largely by organic processes. Most deep-ocean sediments should have Creduced/N ratios <20:1 [Muller, 1977; Sweeney et al., 1978; Waples and Sloan, 1980], and the vast majority of the high-P/T metasedimentary suites fall into that range (see Figure 8b). The Coastal Belt samples have Creduced/N more uniform than that of the higher-grade units (see the line for Coastal Belt samples on Figure 8b), fully within the range for modern seafloor sediments, and perhaps indicating that the loss of isotopically “heavy” N producing the decrease in whole rock δ15N did not result in significant shifts in the Creduced/N of the same rocks (i.e., the δ15N of the N that was lost was significantly higher than that left behind in the sediments). The higher-grade units in the Coast Ranges show scatter in Creduced/N and shifts mostly to higher Creduced/N relative to Creduced/N of the lowest-grade Central Belt, perhaps reflecting the effects of deeper metamorphic devolatilization (i.e., at depths of 10–40 km).
 Yet unknown is the extent of N loss, and accompanying shift in δ15N, that occur during subduction of sedimentary lithologies to depths greater than those represented by the circum-Pacific paleoaccretionary suites we've examined (i.e., at depths >40 km). Busigny et al.  reported whole rock and mica separate δ15N values of +2.6 to +4.8‰ for metasedimentary samples of the Schistes Lustres (and metasedimentary rocks at Lago di Cignana, both exposed in NW Italy) subducted to depths corresponding to pressures of 1.5 to 2.5 GPa (approximately 60–90 km). Although the interpretation of the data for the Schistes Lustres and at Lago di Cignana is complicated by the variable, in some cases extreme, overprinting that occurred during exhumation of these rocks [Reinecke, 1998; Agard et al., 2002; Bebout et al., 2003], these somewhat higher values (relative to those for the lower-P suites studied by Bebout and Fogel  and Sadofsky and Bebout ) could conceivably reflect some increase due to small amounts of N loss during subduction to these greater depths. However, Busigny et al.  also analyzed presumed non-metamorphic equivalents, and because the δ15N values of these sediments are similar to those of the metamorphosed rocks, argued for no change in δ15N in sediments subducted to depths approaching 90 km. Confirmation of whether these data for the Schistes Lustres do reflect deep subduction, without appreciable geochemical effects of exhumation-related overprinting, awaits a more detailed investigation of mineral chemistry (and single-grain δ15N and trace element compositions) in these rocks. As briefly discussed above, recent studies of arc volcanic gases have employed δ15N values of near +7‰ for the deeply subducted sedimentary N component [see Fischer et al., 2002; Snyder et al., 2003], and calculations of N isotope shifts due to metamorphic devolatilization at temperatures of less than 300°C (by Rayleigh or batch loss) can easily produce shifts of 3–4‰ (i.e., shifts in δ15N from +3 to +7‰, perhaps with loss of <25% of the initially subducted N (see calculations by Bebout and Fogel  and Bebout et al. [1999a, 1999b]; at T < 300°C, 103lnαNH4+−N2 is >4‰, based on the calculations of Hanschmann ). These relatively small amounts of loss could be difficult to identify in metamorphosed sediments, given the large degree of variability thought to represent variation in the isotopic composition of the seafloor sediment protoliths.
 It is appropriate to briefly discuss the extent to which our data, and recent work on subduction-zone metamorphic suites, can elucidate N cycling at the Izu-Bonin margin. For this margin (at 32°N), Peacock  calculated temperatures at the slab-mantle interface of ∼245°C at 50 km depths and ∼540°C at sub-arc depths (∼120 km). Thus, for reference, only the lowest-T paths shown in Figure 7 (resulting in temperatures of just over 200°C at depths approaching 40 km) mimic the prograde paths thought to be experienced in this relatively “cool” convergent margin. Even the P-T path experienced by the somewhat higher-grade lawsonite-blueschist-facies unit of the Catalina Schist (patterned field labeled as “LBS” on Figure 7) is somewhat higher-T than would be expected in this modern subduction zone. On the basis of the calculated P-T paths of Peacock , nearly all of the Cottian Alps Schistes Lustres (peak conditions, 300–625°C, 50–70 km [Agard et al., 2002]) and the rocks at Lago di Cignana (peak conditions, ∼625°C, 90 km [see Bebout and Nakamura, 2003]), appear to have experienced peak metamorphic temperatures somewhat higher than those indicated for their respective maximum depths in the modern Izu-Bonin margin. However, the tectonometamorphic units in the Cottian Alps Schistes Lustres traverse studied by Agard et al. , Busigny et al. , and Bebout et al. , combined with the Lago di Cignana rocks [studied by Bebout and Nakamura  and Busigny et al. ), certainly do cover a range of peak P-T broadly compatible with the P-T calculated for the slab-mantle interface in modern subduction zones (see the P-T field for only the lowest-grade unit of the Cottian Alps Schistes Lustres on Figure 7).
 In the Catalina Schist, even in the lawsonite-blueschist-facies metasedimentary unit (labeled “LBS” in Figures 2 and 7; peak conditions of 350°C at pressures corresponding to ∼40 km), δ15N values appear slightly higher than those in lawsonite-albite-facies equivalents experiencing “cooler” prograde P-T paths and lower peak metamorphic temperatures of <250°C (see schematic prograde P-T paths in Figure 7). The Schistes Lustres rocks, metamorphosed at temperatures >300°C but at higher pressures, are similarly somewhat higher in δ15N than the units of the Franciscan Complex and Western Baja Terrane, the latter for which peak temperatures are roughly consistent with peak recrystallization at shallower levels of 5–40 km in the modern Izu-Bonin subduction zone (again, 245°C at 50 km calculated by Peacock ). However, Busigny et al.  suggest that the protoliths for the Schistes Lustres had δ15N higher than that of the protoliths for the metasedimentary rocks of the Catalina Schist, Franciscan Complex, and Western Baja Terrane. Together, the work on the circum-Pacific suites [Bebout and Fogel, 1992; Sadofsky and Bebout, 2003] and the work on the Schistes Lustres [Busigny et al., 2003] appear to demonstrate the impressive compatibility of ammonium ions in micas (see discussion by Boyd ), suggesting that efficient deep entrainment of N into convergent margins is facilitated by the extremely broad stability range of particularly the white micas (see experimental study of the stability relations of phengitic muscovite by Domanik and Holloway ).
 Overall, given the likelihood that even small amounts of N loss can potentially result in significant shifts in sediment δ15N (discussion above), and taking into account the systematics in circum-Pacific paleoaccretionary suites (Catalina Schist, Franciscan Complex, Western Baja Terrane; +1 to +3‰ in sediments subducted to 5–40 km) and the results presented by Busigny et al.  (somewhat higher values of +2.6 to +4.8‰ in sediments subducted to 50–90 km), it appears that the use of a mean δ15N value of +7‰ (perhaps ±2‰) is reasonable in the studies of volcanic gases (see recent studies by Fischer et al. , Hilton et al. , and Snyder et al. ). However, as highlighted above, knowledge of the N concentrations and isotopic compositions of subducting AOC is critical in any further comparisons of the seafloor, metamorphic, and volcanic gas records of convergent margin N cycling; also critical is the efficiency with which N and noble gases released from devolatilizing AOC (and sediment, for that matter) can be mobilized into the subarc mantle wedge. Integration of the N results with other data (trace element, isotopic) for individual arcs indicating relative slab contributions from AOC and sediment could potentially allow further delineation of AOC and sediment N (and C) sources.