Geochimica et Cosmochimica Acta, Vol. 67, No. 21, pp. 4123– 4135, 2003 Copyright © 2003 Elsevier Ltd Printed in the USA. All rights reserved 0016-7037/03 $30.00 ⫹ .00

Pergamon

doi:10.1016/S0016-7037(03)00259-X

Nitrogen solubility in basaltic melt. Part I. Effect of oxygen fugacity G. LIBOUREL,1,2,* B. MARTY,1,2 and F. HUMBERT3 1

Centre de Recherches Pe´trographiques et Ge´ochimiques, CNRS-EPR2300, 15, rue Notre Dame des Pauvres, BP20, 54501 Vandoeuvre-les-Nancy, France 2 Ecole Nationale Supe´rieure de Ge´ologie, INPL, Rue du Doyen Marcel Roubault, BP40, 54501 Vandoeuvre-les-Nancy, France 3 CFC-DAUM, Compagnie Franc¸aise du Cristal, Usine rue Cristallerie, 54112 Vannes le Chatel, France (Received July 23, 2002; accepted in revised form March 10, 2003)

Abstract—The role of the oxygen fugacity on the incorporation of nitrogen in basaltic magmas has been investigated using one atmosphere high temperature equilibration of tholeiitic-like compositions under controlled nitrogen and oxygen partial pressures in the [C-N-O] system. Nitrogen was extracted with a CO2 laser under high vacuum and analyzed by static mass spectrometry. Over a redox range of 18 oxygen fugacity log units, this study shows that the incorporation of nitrogen in silicate melts follows two different behaviors. For log fO2 values between ⫺0.7 and ⫺10.7 (the latter corresponding to IW ⫺ 1.3), nitrogen dissolves as a N2 molecule into cavities of the silicate network (physical solubility). Nitrogen presents a constant solubility (Henry’s) coefficient of 2.21 ⫾ 0.53 ⫻ 10⫺9 mol g⫺1 atm⫺1 at 1425°C, identical within uncertainties to the solubility of argon. Further decrease in the oxygen fugacity (log fO2 between ⫺10.7 and ⫺18 corresponding to the range from IW ⫺ 1.3 to IW ⫺ 8.3) results in a drastic increase of the solubility of nitrogen by up to 5 orders of magnitude as nitrogen becomes chemically bounded with atoms of the silicate melt network (chemical solubility). The present results strongly suggest that under reducing conditions nitrogen dissolves in silicate melts as N3⫺ species rather than as CN⫺ cyanide radicals. The nitrogen content of a tholeiitic magma equilibrated with N2 is computed from thermochemical processing of our data set as [N2] (mol N2 · g⫺1) ⫽ (2.21 ⫾ 0.53) ⫻ 10–9 · PN2 ⫹ fO2–3/4 · (2.13 ⫾ 0.11) ⫻ 10–17 · PN21/ 2 High nitrogen contents in primitive meteorites, especially in glass inclusions encapsulated in magnesian olivine of chondrites, are unlikely to result from nitrogen dissolution from the solar nebula gas, unless the pressure of the latter is underestimated by several orders of magnitude. Significant amounts of nitrogen, comparable to those estimated for the present-day mantle, could have been incorporated in the early Earth by dissolution in a magma ocean, under fO2 conditions relevant to those prevailing during metal segregation. The present results also imply that the N2/Ar ratio of tholeiitic basalts (e.g., MORBs) is not fractionated during magma generation and degassing, allowing to use argon as a geochemical proxy for nitrogen. It is probable that mantle nitrogen was degassed at rates and fluxes comparable to that of argon, as the oxygen fugacity of the mantle was unlikely to have been below IW from Archean to Present. Therefore, fractionation between nitrogen and argon in the Earth-atmosphere system is more probably the result of recycling rather than of mantle-derived magma degassing. Copyright © 2003 Elsevier Ltd network, according to their atomic radius. Extension of such a rare gas model to major volatiles (e.g., H, C, N, S) requires, before their use as potential tracers, an investigation of the chemical reactivity of these elements within silicate networks (Carroll and Holloway, 1994) and documentation of the physical and chemical factors controlling their solubility. Among volatile elements, nitrogen is a major forming element of the Earth’s atmosphere and one of the key elements of biologic evolution. Knowledge of its (elemental or isotopic) behavior is therefore of prime importance to understand and quantify processes which might have controlled the exchange of nitrogen between the mantle and the atmosphere through time. Among these processes, extraction from the solid Earth through partial melting and degassing of mantle-derived magmas has certainly played a central role. However, mechanisms governing the incorporation of nitrogen in basaltic melts are not well documented,1 which hitherto precludes quantitative treatments of the evolution of atmospheric nitrogen. In oxidizing

1. INTRODUCTION

Rare gases are powerful geochemical and cosmochemical tracers, whose physical properties in silicates, including solubility, have been extensively studied in the past decades (Doremus, 1966; Kirsten, 1968; Hayatsu and Waboso, 1985; Hiyagon and Ozima, 1986; Jambon et al., 1986; Lux, 1986; Broadhurst et al., 1992; Carroll and Draper, 1994; Miyazaki et al., 1995). These studies have shown that (1) the solubility of rare gases in silicate melt is very low, of the order of 10⫺8 to 10⫺9 mol g⫺1 atm⫺1, allowing efficient rare gas degassing of magmas during eruptions, (2) their partitioning between melt and vapor follows the Henry’s law which states that the concentration of rare gases in the melt is proportional to their partial pressures in the vapor phase, and (3) their solubility in silicate melt decreases by 2 orders of magnitude from helium to xenon. This peculiar behavior is consistent with a steric effect in which rare gases fill physical holes within the silicate melt

1

* Author to whom correspondence should be addressed (libou@ crpg.cnrs-nancy.fr).

The solubility of nitrogen in industrial slags has been intensively studied since this element constitutes an impurity in melting thaws and 4123

4124

G. Libourel, B. Marty, and F. Humbert

conditions, the solubility of nitrogen in basaltic melt is very low and may be comparable to that of rare gases (Marty, 1995; Miyasaki et al., 1995; Miyasaki, 1996). In reducing conditions, Mulfinger (1966) showed that the solubility of nitrogen in synthetic silicate melts is higher by several orders of magnitude than that in oxydizing conditions and that this contrasting behavior is due to fundamental changes in the mechanism of nitrogen dissolution in silicate melt. This author suggested that, in oxidizing conditions, nitrogen dissolves physically as a N2 molecule whereas in reducing conditions nitrogen is chemically bounded to the silicate melt network. Similarly, high nitrogen solubility in reducing conditions was also documented for aubrite and E-chondrite melts (Fogel, 1994). Few other attempts of nitrogen solubility measurements, notably for albitic melts (Kesson and Holloway, 1974; Shilobreeva et al., 1994), are in general agreement with Mulfinger’s (1966) finding. Because the redox conditions prevailing during magmatic processes in planetary bodies including Earth can vary by several orders of magnitude, it is important to document quantitatively the role of the oxygen fugacity on the solubility of nitrogen in silicate melts of geological interest. To fill this gap, a systematic study of the solubility of nitrogen in basaltic melts under a wide range of oxygen fugacity (fO2) was undertaken. We have designed new experimental and analytical methods based on (1) high temperature equilibration of silicate melts under controlled nitrogen and oxygen partial pressures, and (2) laser extraction in high vacuum, purification of extracted volatiles and static mass spectrometry measurements of nitrogen. Owing to such specific procedure, we measured the solubility of nitrogen in basaltic melts over a range of 18 log units of oxygen fugacity in the [C-N-O] vapor system. In a companion paper we shall report the results of our investigation on the role of the melt and gas compositions in the [H-C-N-O] system together with planetary implications. 2. EXPERIMENTAL 2.1. Experimental Strategy To get a precise and stable control on the oxygen fugacity, we decided to use gas mixtures (CO, CO2, N2 or H2) flown in a vertical furnace containing the silicate melt at pressures close to the atmospheric one. As a result, our experiments were carried out under low nitrogen partial pressures (PN2 ⬍ 1 atm), contrary to previous equilibration experiments performed under high partial pressures of rare gases or nitrogen using external or internal heated vessels (Kesson and Holloway, 1974; Carroll and Draper, 1994; Shilobreeva et al., 1994; Miyazaki, 1996; Shibata et al., 1998). The resulting nitrogen contents of our experimental charges were extremely small and we developed a new analytical method, based on CO2-laser extraction-static mass spectrometry analysis, to get enough nitrogen sensitivity together with low blank levels. In the following, we describe the different steps of this experimental and analytical protocol. 2.2. Starting Material Nitrogen equilibration experiments on basaltic melts were undertaken on a single synthetic starting material (CM-1), which composition is that of a typical MORB (Hofmann, 1988). Due to potential iron loss of run samples in reduced conditions, this MORB-like starting composition was computed on an iron-free basis. The starting material was obtained by mixing up reagent grade oxides (SiO2, TiO2 and

Table 1. Experimental conditions and gas mixtures used for nitrogen equilibration. Depending on the imposed oxygen fugacity, different crucibles were used; C and Pt stand for graphite and for platinum, respectively. The log(fO2) was computed from the composition of the fluxed gas and the temperature of the run using the Thermodata code. Oxidation conditions are referenced to IW buffer. Gas fraction

IW–8.3 IW–8.3 IW–8 IW–7 IW–6.8 IW–6.7 IW–4.1 IW–2 IW IW⫹2 IW⫹3.6 IW⫹4.1 IW⫹4.3 IW⫹8.7

Crucible

CO

1400 1425 1425 1400 1425 1425 1425 1425 1425 1425 1425 1425 1425 1425

C C C C C C Pt Pt Pt Pt Pt Pt Pt Pt

0.156 0.150 0.200 0.748 0.800 0.950 0.499 0.486 0.386 0.128 0.027 0.012

CO2

0.001 0.014 0.115 0.372 0.473 0.488 0.000

N2

log(fO2)

0.844 0.850 0.800 0.253 0.200 0.050 0.500 0.500 0.500 0.500 0.500 1.000 0.500 0.781

–17.7 –17.7 –17.4 –16.4 –16.2 –16.1 –13.5 –11.4 –9.4 –7.4 –5.9 –5.3 –5.1 –0.7

Al2O3) and carbonates (MgCO3, CaCO3, Na2CO3, K2CO3) in adequate proportions and finely ground in an agate mortar. After decarbonation, the mixture was fused in argon atmosphere at 1400°C for 8 h and quenched. The resulting glass was finely ground and aliquots of this powder were used as the starting material for our experimental runs. Detailed analysis of this starting material is given below. The CM-1 composition is orthopyroxene- and quartz-normative and represents that of silica saturated tholeiitic basalts. 2.3. Experimental Setup and Run Conditions The experiments were performed at atmospheric pressure in a GERO HTVR 70-250 vertical drop-quench furnace. Temperatures were measured using Pt-PtRh10 thermocouples located in the hotspot of the furnace. All experiments were performed, above the liquidus of the CM-1 starting composition, at the same constant temperature of 1425°C (only two were performed at 1400°C). To document the nitrogen solubility over a fO2range covering most all geochemical and cosmochemical environments, the selected oxygen fugacities ranged from that of air down to log fO2 ⫽ ⫺18.3, corresponding to ⬃IW ⫺ 8. Note that, for simplification, oxidation conditions are referenced here to the IW (Iron/Wustite) buffer curve. For instance, IW ⫹ 3.6 ⫽ FMQ ⫽ ⫺5.9 ⫽ log fO2 at 1 atm and 1425°C, and similarly, IW ⫹ 8.7 ⫽ Air ; IW ⫹ 4.3 ⫽ NNO; and IW ⫺ 6.6 ⫽ CCO. To establish this redox range, different mixtures of pure gases (CO, CO2, N2) in the [C-N-O] vapor system were used, buffered or not by the presence of graphite (Table 1). The oxygen fugacity and the nitrogen partial pressure were controlled using appropriate mixtures of these gases, fed to the furnace muffle tube at a flow rate of 300 cm3 min⫺1 with mixing proportions regulated by TYLAN mass flow controllers. For runs above IW ⫺ 4 (Table 1), 50% of the gas flowing through the furnace tube was made of CO-CO2 in variable proportion, the remaining 50% being N2. The IW ⫺ 4 redox state corresponds to the most reduced condition that could be achieved with mass flow controllers since they cannot regulate precisely enough gas mixtures with a CO2/CO ratio below 0.002. To perform experiments at more reduced conditions, glass samples were held in a graphite crucible under a CO-N2 atmosphere. Oxidation conditions are thus buffered in the range of the C/CO buffer curve, i.e., IW ⫺ 6.6 at 1 atm and 1425°C, according to the following reaction: C(graphite) ⫹ 1/2 O2(gas) ⫽ CO(gas)

(1)

with fO2 ⫽ fCO/aC · KEqn. 1, where fCO corresponds to the carbon monoxide fugacity, aC, to the activity of carbon in graphite, and KEqn. 1 to the equilibrium constant of Eqn. 1. Since this reaction states that 1/2

steels. However, these slags compositions are so impoverished in silica that they are not relevant to model basaltic melts. (see below).

T (°C)

Nitrogen solubility in basaltic melt fO2 is proportional to the square of PCO (assuming aC ⫽ 1 and fCO ⫽ ⌫CO · PCO, with the fugacity coefficient ⌫CO ⫽ 1), it is possible to decrease the oxygen fugacity well below the C/CO buffer curve simply by diluting the CO gas by N2. Consequently, all experiments using graphite crucibles were performed under variable nitrogen partial pressure between 0.05 ⬍ PN2 ⬍ 0.95 atm (Table 1). The oxygen fugacity was calculated for each run at 1 atm and 1425°C, using a thermodynamic code (Thermodata and JANAF data base). A fixed yttriumstabilized zirconium oxygen sensor was used to monitor some of these gas mixtures, and showed that precision of oxygen fugacity control is better than 0.1 log fO2 unit. Above IW ⫺ 4, all experiments were performed in a platinum crucible (20 mm OD) containing 300 to 500 mg of starting material. Below this value, samples were run simultaneously in three small graphite crucibles 4 mm ID and 6.5 mm high (the crucible dimensions were chosen to enhance the vapor-liquid exchange) containing 100 mg of starting material each and held in a larger graphite crucible. All experiments were terminated by drop-quenching the sample in the ambient furnace atmosphere. After each experiment, fragments of the quenched charges were mounted in epoxy, polished, and examined optically. Each glass fragment was found to be clear and free of crystals, with the notable exception of those obtained in the most reduced conditions, where osbornite, a titanium nitride (TiN), was detected (see below). In very reducing conditions which can promote evaporation of alkalis (see below), quenched glasses were systematically analyzed using a CAMECA SX50 electron microprobe to check their homogeneity and any variation in their composition. 2.4. Analytical Procedure A full description of the analytical setup is given elsewhere (Humbert, 1998; Humbert et al., 2000). Here we report a brief description of this analytical facility and a summary of its performance. The main requirement was an analytical system able to analyze extremely low amounts of nitrogen. For instance, for experimental charges of the order of 1 mg, and considering a nitrogen solubility of 2 ⫻ 10⫺9 mol N2 g⫺1 atm⫺1 in oxidizing conditions (see section 4.1), the amount of nitrogen to be analyzed is typically 2 ⫻ 10⫺12 mol N2 (0.03 ng N). It was also necessary to obtain N2 blanks comparable to, or lower than, the analyzed quantities. The only analytical method able to provide such detection limit of N is static mass spectrometry of the type developed for rare gas analysis. Nitrogen blanks reported in the literature range from 0.1 to 0.4 ng N for “cold” blanks, i.e., procedural blanks without heating extraction, and from 1 to 4.5 ng N for “hot” blanks (Frick and Pepin, 1981; Boyd et al., 1988; Hashizume and Sugiura, 1990; Marty et al., 1995; Murty et al., 1996; Cartigny, 1997; Mathew et al., 1998). In nitrogen extraction/purification lines, the major sources of blanks are hot devices such as the extraction furnace, CuO furnaces and catalysers. Among these, the extraction furnace is the most important source of blank nitrogen. To circumvent this problem, we used a defocused CO2 laser as a heat source. The samples were loaded in a stainless steel extraction cell comprising a ZnSe window transparent to the laser beam, and a Pyrex window for CCD camera observation. The sample holder has seven 3.5 mm OD hemispheric cavities in which milligram-sized samples are loaded, and internal water circulation to cool it during laser heating (Humbert et al., 2000). Quenched glass samples were ultrasonically cleaned in high grade acetone for 5 min. Samples were then gently crushed and aliquots of glass were selected under a binocular microscope for the absence of visible defects, bubbles and/or crystals. The selected glass fragments were again cleaned with acetone before loading them in the laser cell. After sample loading, the sample cell was evacuated in high vacuum and baked overnight at 150°C. The laser has a power of 30 W in the CW mode and a wavelength of 10.6 ␮m, and is monitored using a visible He-Ne laser. Both lasers are mounted on a X-Y deck. The power of the laser can be modulated to apply stepwise heating to the samples. During the course of this study, milligram-sized samples were heated to fusion during typically 2 min, resulting in a melted silicate sphere vibrating under the energetic beam. The nominal temperature was kept below ⬃1700°C to limit sample evaporation. Duplicate fusion of samples showed that nitrogen extraction was ⬎ 99% efficient. The gas was then expanded into the purification section. The purification of

4125

nitrogen was achieved classically using a CuO furnace cycled between 450 and 720°C, a cold trap in which temperature was adjusted to just above that of liquid N2, and a hot platinum foil used for catalysing oxydo-reduction reactions. The nitrogen purification section is made of Pyrex and quartz glass connected using all-metal valves. The line can also be used for simultaneous rare gas analysis (Marty and Humbert, 1997). Purified nitrogen was admitted after proper volume dilution in a static rare gas, sector-type, mass spectrometer (VG5400, Micromass) comprising a Nier-type ion source, a bent tube of 0.54 m radius, and two collectors (Faraday collector and axial electron multiplier fitted with an ion counting system). Ion currents corresponding to masses 28, 29, and 29, 30, were measured on the Faraday collector and on the electron multiplier, respectively, and signals were extrapolated to the time of gas introduction into the ion source. The sensitivity of the analyzer was determined using calibrated aliquots of atmospheric nitrogen (initially 4.14 ⫻ 10⫺10 mol N2, which was then adjusted by volume dilution to match within 20% the amount of sample nitrogen). During this study, the N2 sensitivity was 2.6 to 3.6 ⫻ 10⫺4 A/Torr, depending on the source conditions. For the purpose of illustration, a N2 amount of 1.4 ⫻ 10⫺11 mol (0.39 ng N) resulted in a tension of 4 V on the Faraday collector (R ⫽ 1011 ⍀). The isotopic composition of the sample was computed after correction for isobaric interferences and for blank contributions using the N isotopic composition of the standard gas (Marty and Humbert, 1997). Several successful tests of the efficiency of nitrogen purification were carried out using mixtures of CO and N2 and are reported in Humbert et al. (2000). After optimisation of the purification procedure, notably on the trap temperatures and on reaction durations, the final blanks were 1 to 2 ⫻ ⫺12 10 mol N2 (30 – 60 pg N) and were found to vary by ⬃5% within 1 d, and by 50% within 1 month. For the latter, the main cause of variation was residual degassing of the sample cell as the blank level decreased continuously with time after sample loading. Since this study, further work has allowed us to reduce blanks to 5 to 10 pg N to analyze single grains of lunar ilmenite (Hashizume et al., 2002). The precision on the amount of N, computed by propagating standard deviations of peak height measurements and uncertainties on the calibrated volumes of the line, was found to be ⬍ 5% and ⬃3% in most cases. This figure is comparable to the reproducibility of atmospheric N standard measurements over 1 month (Humbert et al., 2000). 3. RESULTS

3.1. Attainment of Equilibrium and Data Reproducibility Preliminary experiments were run to evaluate the time necessary to reach gas-silicate melt equilibrium. Because of higher nitrogen solubility in reducing conditions, these tests were performed at IW ⫺ 8.3 during intervals of time varying from 3 to 72 h (Table 2; Humbert, 1998). The results indicate that nitrogen equilibration was complete after 48 h (Fig. 1a) and all nitrogen equilibration experiments lasted 48 h or more (Table 2). Between three and five replicate nitrogen analyses on different chips of quenched glasses were carried out for each run (Table 2). The reproducibility of run data is excellent in the case of samples equilibrated in reducing conditions which contain high amounts of nitrogen (Table 2; Fig. 1b). The overall data dispersion for replicate analyses is 10% relative in this case, e.g., samples CM-1#4A and CM-1#8. For samples containing less nitrogen, the dispersion of data is larger (Fig. 1b) and can reach 1 order of magnitude for samples having very low amounts of nitrogen (⬍0.1 ppm), e.g., CM-1#14 and CM-1#16 (Table 2). For the latter, several factors can affect the measured amount of nitrogen, such as contamination by airderived N (e.g., adsorption, organic contamination), occurrence of micro bubbles, etc. These factors were minimized by adopting the specific preparation procedures described in section 2.

4126

G. Libourel, B. Marty, and F. Humbert Table 2. Run conditions and nitrogen content of experimental samples for the CM-1 basaltic like composition.

Run (see Table 3)

Conditions (see Table 1)

Duration (hs)

CM-1#4A

IW – 8.3

48

CM-1#4B

CM-1#5

CM-1#6

IW – 8.3

IW – 8.3

IW – 8

72

48

72

T (°C)

log(fO2) (atm)

PN2 (atm)

1400

–17.7

0.84

1400

1425

1425

–17.7

–17.7

–17.4

0.84

0.85

0.80

Sample Mauve, with gold coating

Mauve, with gold coating

Mauve, with gold coating Mauve, with gold coating

CM-1#7

IW – 6.9

72

1400

–16.4

0.25

Mauve

CM-1#8

IW – 6.8

96

1425

–16.2

0.20

Mauve

CM-1#9

IW – 6.7

48

1425

–16.1

0.05

Mauve

CM-1#10

IW – 4.1

36

1425

–13.5

0.50

Mauve

CM-1#11

IW – 2

48

1425

–11.4

0.50

Mauve

CM-1#12

IW

48

1425

–9.4

0.50

Light mauve

CM-1#13

IW ⫹ 2

48

1425

–7.4

0.50

Yellow-green

CM-1#14

IW ⫹ 3.6

48

1425

–5.9

0.50

Yellow-green

CM-1#15

IW ⫹ 4.1

48

1425

–5.3

1.00

Yellow-green

CM-1#16

IW ⫹ 4.3

49

1425

–5.1

0.50

Yellow-green

CM-1#17

IW ⫹ 8.7

48

1425

–0.7

0.78

Brownish-green

Mass sample (mg)

Blank sample (%)

0.40

0.06

1389

49,589

1.10 0.43 0.88 0.50

0.02 0.04 0.02 0.04

1260 1473 1216 1368

45,003 52,594 43,433 48,844

0.52 0.68 1.43 0.86

0.04 0.03 0.02 0.01

1339 1192 1098 1579

47,803 42,578 39,196 56,396

0.57 1.25 0.90

0.00 0.00 0.03

1545 1427 1725

55,173 50,937 61,581

0.95 0.13 0.26 0.41 0.82 2.32 0.37 1.16 0.98 0.84 0.64 0.47 0.96 1.60 2.59 1.69 2.21 2.33 32.9 21.9 9.79 12.5 11.9 45.1 5.70 17.9 9.61 6.21 5.11 9.18 23.4 9.22 37.8 22.1 23.8 4.51 7.99 18.5 4.25 24.3 7.11 17.9 12.2 8.42 3.65 7.22

0.02 0.13 0.06 0.12 0.07 0.02 0.13 0.11 0.06 0.08 0.13 0.10 0.04 0.05 2.63 3.13 1.68 1.39 69.7 2.53 15.9 8.51 87.1 9.99 29.1 55.6 69.4 26.7 67.8 8.61 44.8 53.1 80.7 9.28 4.79 11.5 65.7 6.78 51.0 3.62 57.4 11.5 8.91 2.95 50.3 38.8

1747 1700 1767 651.5 505.1 586.8 537.9 481.2 423.1 427.4 455.0 133.3 143.7 129.8 1.062 1.518 1.948 0.939 0.023 0.170 0.082 0.197 0.003 0.046 0.021 0.005 0.004 0.073 0.005 0.122 0.009 0.011 0.001 0.030 0.054 0.130 0.025 0.048 0.021 0.081 0.016 0.029 0.054 0.099 0.025 0.036

62,373 60,696 63,089 23,262 18,036 20,951 19,207 17,180 15,107 15,260 16,247 4761 5130 4634 37.9 54.2 69.5 33.5 0.84 6.07 2.92 7.02 0.10 1.65 0.76 0.17 0.16 2.62 0.17 4.36 0.32 0.38 0.05 1.07 1.92 4.64 0.88 1.72 0.75 2.88 0.57 1.03 1.92 3.54 0.90 1.29

N content (ppm)

N2 content (10–9 mol g–1)

Nitrogen solubility in basaltic melt

4127

oxidizing conditions, the Henry’s coefficients for nitrogen computed from the N contents of these glasses represent an upper limit. We note however that these coefficients are comparable to those obtained by Miyazaki et al. (1995) and Miyazaki (1996) under higher nitrogen partial pressures (see “Discussion”). Duplicate experiments were also performed by keeping the oxygen fugacity almost constant but by varying the nitrogen partial pressures (i.e., CM-1#8 and CM-1#9, CM-1#15 and CM-1#16, Tables 1 and 2). The good match between N solubility data obtained in each of these duplicates are further evidence of attainment of equilibrium. 3.2. Evolution of the Melt Composition

Fig. 1. (a) Nitrogen content versus time, measured in CM-1 MORBlike compositions from a set of solubility experiments performed at 1400°C, IW ⫺ 8.3, and PN2 ⫽ 0.84 atm (see Table 1). Notice that nitrogen equilibration is completed after 48 h. (b) Nitrogen content for replicate analyses obtained on different chips of glasses belonging to the same solubility experiment. Four sets of experiments are presented: CM-1#4A, CM-1#8, CM-1#10, and CM-1#16 (see Table 1), from reduced to oxidized conditions, respectively.

Nevertheless, we cannot exclude the occurrence of tiny bubbles containing the buffer gas, nor abnormal nitrogen adsorption on glassy fragments. Since both processes would tend to increase artificially the amount of trapped nitrogen, it is possible that, in

Analysis of glass compositions after gas equilibration reveals that substantial loss of alkali (Na2O and K2O) occurred. We found that the degree of loss increased with the equilibration time for a given oxygen fugacity (Table 3), and with decreasing oxygen fugacity for a given equilibration duration (t ⬃ 48 h, Fig. 2). These results are in agreement with previous studies (Tsuchiyama et al., 1981; Tissandier et al., 1998; among others) that demonstrated that alkalies behave as volatile elements at high temperature and in reducing conditions. Even if the compositional drifts from pristine CM-1 starting material are more important in reducing conditions, these changes, i.e., slight increase in melt polymerization, may not distort significantly the present results, because of the low initial alkali contents of these glasses (⬍3 wt.%). Nevertheless, nitrogen solubility experiments were performed on a alkali-free CM-1 composition for very reducing conditions below IW ⫺ 4. More importantly, run products show noticeable changes in their coloration, from a dull yellow-green in oxidizing conditions to a light mauve around IW buffer and to dark mauve in the most reducing conditions (Table 2). In addition, all runs performed below IW ⫺ 8 present a golden (to bronze) film coating the surface of each glassy sphere. Electron microprobe analyses have revealed that such golden coatings are systematically associated with significant titanium depletions in the glass, up to ⬃40% (Table 3). To better understand these features, Electron Paramagnetic Resonance (EPR) analyses on selected samples were performed (CM-1#16, CM-1#12, CM1#4B; Humbert, 1998). The resulting spectra show that the

Table 3. Glass compositions of some representative run products. See Tables 1 and 2 for run conditions. CM-1 is the starting MORB-like composition used for this study. Sample (Table 1) CM-1 CM-1#4D CM-1#4C CM-1#4A CM-1#4B CM-1#6 CM-1#7 CM-1#12 a b

Time (h)

log fO2 (atm)

SiO2 (wt.%)

Al2O3

MgO

CaO

Na2O

K2O

TiO2

8 3 12 48 72 72 72 48

Air –17.7 –17.7 –17.6 –17.7 –17.4 –16.4 –9.4

58.2 (4)a 57.9 (6) 59.0 (3) 58.3 (2) 57.9 (4) 54.7 (1) 57.7 (3) 56.5 (3)

15.7 (3) 16.0 (2) 16.4 (1) 16.5 (3) 17.1 (3) 18.8 (3) 16.8 (2) 17.2 (1)

8.6 (2) 8.4 (3) 8.8 (1) 8.9 (1) 8.9 (1) 9.5 (1) 9.0 (2) 8.4 (1)

13.3 (1) 13.2 (2) 13.5 (2) 13.8 (2) 14.1 (1) 15.7 (2) 13.9 (1) 12.8 (1)

2.8 (1) 2.12 (1) 0.49 (1) — — — — 2.38 (1)

0.2 (07) 0.17 (05) —b — — — — 0.18 (03)

1.8 (2) 1.83 (1) 1.87 (1) 1.53 (06) 1.53 (06) 1.27 (05) 1.96 (05) 2.00 (06)

Numbers in parentheses indicate 1 standard deviation of replicate analyses in terms of least units cited. Dash indicates analyses below the detection limit of the electron microprobe.

4128

G. Libourel, B. Marty, and F. Humbert

Fig. 2. Sodium evolution in glass after nitrogen equilibration as function of oxygen fugacity at 1425°C. (Na2O)/(Na2O)0 stands for the ratio of the sodium content measured in the glass after equilibration over the sodium content of the CM-1 starting glass. For all the charges, the equilibration time is above 48 h. Redox conditions are indicated in the upper part of the graph. This figure illustrates that CM-1 experiences severe sodium loss in most of the runs and that sodium behaves as volatile elements in reduced conditions. A similar feature can be also seen for K2O content.

yellowish-brownish color found in oxidizing conditions is due to minor iron contamination from the Pt crucible (⬍1 wt.% FeO). In reducing conditions, EPR data indicate that the mauve color is due to the occurrence of Ti3⫹ in the glass newtwork, and that the corresponding peak intensity of EPR spectra increases with decreasing oxygen fugacity. Finally, X-ray diffractogram of a gold coating reveals peaks located at 2.4499 and 1.2253 d (Å) that can be assigned to the (111) and (222) faces of isometric osbornite crystals, a titanium nitride (TiN). All these results allow us to conclude that the mauve color after equilibration is due to partial reduction of titanium in the melt (Ti4⫹ 3 Ti3⫹), according to the following reaction: 2 TiO2 (melt) ⫽ Ti2O3 (melt) ⫹ 1/2 O2 (g)

Fig. 3. Nitrogen contents (in ppm or mol g⫺1 units) in CM-1 MORB-like melt as function of oxygen fugacity at 1425°C. Filled circles correspond to average equilibrium values calculated from Table 1 together with their standard deviation. Redox conditions are indicated in the upper part of the graph.

4. DISCUSSION

The nitrogen behavior observed in this study (Fig. 3) is consistent with the qualitative observation of Mulfinger (1966) that there exists at least two fundamentally different mechanisms of nitrogen trapping in silicate melts. In oxidizing conditions, the incorporation of nitrogen corresponds to dissolution of N2 molecule, with no covalent or ionic bounding between N atoms and the silicate network, while in reducing conditions, dissolved nitrogen is chemically bounded with atoms constituting the silicate melt network. However, Mulfinger (1966) did not investigate the speciation of nitrogen under reducing conditions, and this question will be addressed in the light of the present data. According to our results, the change between chemical solubility and physical dissolution is observed to take place around the IW buffer (Fig. 3). 4.1. Nitrogen Solubility in the Range Air to IW

(2)

Since Ti shows important affinity with N, the crystallization of titanium nitride raises the interesting possibility of a specific N speciation with Ti and will be discussed in detail in section 4.3.

3.3. Nitrogen Content The main result of this study is that the nitrogen content in silicate melts strongly depends on oxygen fugacity (Fig. 3). For fO2 from air to IW, the nitrogen content is very low and near-constant at ⬃10⫺9 mol g⫺1 (Table 2). Below the IW buffer (IW to IW ⫺ 7), the nitrogen content increases drastically by 4 to 5 orders of magnitude with decreasing fO2. A further decrease in oxygen fugacity below IW ⫺ 7 does not seem to affect the nitrogen solubility that levels off at ⬃5 ⫻ 10⫺5 mol g⫺1. This peculiar behavior is addressed in detail in the next section.

In this range of redox conditions, there is no discernable variation in the nitrogen content, suggesting a common process of N dissolution in silicate melts. Miyazaki (1996) showed that, in similar oxidizing conditions, the nitrogen solubility in the pressure range 1 to 800 bars could be approximated by a Henry’s law behavior, see also Kesson and Holloway (1974). The nearly constant N solubility between IW and air deduced from this study is consistent with the dissolution of nitrogen as N2 molecule, i.e., the most stable N species in this range of P-T-fO2 conditions, in agreement with the following reaction: N2 (gas) ⫽ N2 (melt)

(3)

In the investigated range of N2 pressures, the nitrogen fugacity is close to the nitrogen partial pressure (fN2 ⫽ ⌫N2 · PN2, with the fugacity coefficient ⌫N2 ⫽ 1) and the nitrogen content of the melt normalized to the partial pressure of N2, [N]n, phys, is equivalent to the physical solubility of nitrogen in the melt:

Nitrogen solubility in basaltic melt

Fig. 4. Nitrogen solubility in CM-1 MORB-like melt as function of oxygen fugacity at 1425°C. Filled circles correspond to average equilibrium values calculated from Table 1 and their standard deviation. Shaded box corresponds to data obtained by Fogel (1994) in aubrite and E-chondrite melts (see comments in the text). Redox conditions are indicated in the upper part of the graph. This study shows that, depending on oxygen fugacity, nitrogen dissolution in a silicate melt can be accounted for by either a physical or a chemical solubility. Nitrogen incorporation in silicate melt is hence expressed in ppm N atm⫺1 or ppm N atm⫺1/2 and in mol N2 g⫺1 atm⫺1 or mole N2 g⫺1 atm⫺1/2. See text for explanations.

4129

Fig. 5. Rare gases and nitrogen solubility in silicate melt as function of their atomic or molecular diameter. Data from Carroll and Draper (1994) and Miyazaki et al. (1995). Nitrogen physical solubility in silicate melt obeys to steric effect model of rare gas solubility and is almost equivalent to argon solubility.

1/2 N2 (gas) ⫽ N(melt)

(5)

Therefore we normalize the N content to the square root of the PN2: [N]n, phys ⫽ [N2]/PN2 ⫺1

(4) ⫺1

⫺1

and expressed in ppm N atm , or in mol N2 g atm units. The average nitrogen solubility computed for the CM-1 composition (tholeiite-type) at 1425°C is 2.21 ⫾ 0.53 ⫻ 10⫺9 mol g⫺1 atm⫺1 (Fig. 4), consistent with a value of ⬃3.6 ⫻ 10⫺9 mol g⫺1 atm⫺1 proposed by Javoy and Pineau (1991) from analysis of natural basalt, a determination of 3.7 ⫾ 1.1 ⫻ 10⫺9 mol g⫺1 atm⫺1 for natural basaltic melt equilibrated with air at 1600°C (Marty, 1995), and with a determination of 2.8 ⫾ 1.5 ⫻ 10⫺9 mol g⫺1 atm⫺1 in basaltic melt at 1300°C between IW and air (Miyazaki et al., 1995; Miyazaki, 1996). Such range for the solubility of nitrogen is comparable to that of argon as determined by several authors (Hiyagon and Ozima, 1986; Jambon et al., 1986; Lux, 1986; Carroll and Draper, 1994; Miyazaki, 1996) in basaltic melts at temperatures between 1300 and 1650°C (Fig. 5). Because N2 and Ar have also comparable atomic/molecular dimensions, the similarity in solubility coefficients supports the view that nitrogen dissolves in basaltic melt as a N2 molecule and physically fills holes within the silicate melt network as argon does. In this case, it is generally assumed (Doremus, 1966; Schackelford et al., 1972; Shelby, 1974) that the decrease of solubility with increasing gas atomic radii is due to the existence of a continuous distribution of size of holes within the silicate melt network, showing a larger number of small holes in which atoms like helium could be preferentially incorporated (Fig. 5). 4.2. Chemical Solubility In reducing conditions below IW, the nitrogen content is unlikely to be proportional to PN2 because nitrogen dissolves as a N radical (see below):

[N]n, chem ⫽ [N]/共PN2兲 1/ 2

(6)

In the following we use the term of solubility for both chemical and physical incorporation processes. Between IW and CCO buffer curves (Fig. 4), the nitrogen solubility varies dramatically between ⬃0.06 ppm N atm⫺1/2 at IW and ⬃1500 ppm N atm⫺1/2 at IW ⫺ 6.6. This trend is in general agreement with the range determined by Fogel (1994) at IW ⫺ 10, 1500 to 1600°C, on aubrite and E-chondrite melts (Fig. 4). However, it is important to note that these data were obtained in a [C-N-O-H] vapor system, and are hence not directly comparable to our hydrogen-free vapor, since changes in N speciation of the vapor phase may have some important effects on the N solubility. Nevertheless, the present 3 to 4 orders of magnitude increase of the N solubility with decreasing fO2 confirms the qualitative view of Mulfinger (1966) that nitrogen is chemically bound to the silicate network. In such reducing environments, the only other existing data are solubility measurements done on liquids representing industrial slags motivated by the fact that nitrogen constitutes an impurity in melting thaws and steels. In oxide (silicate or not) melts, all experiments show a similar negative correlation between N solubility and fO2 (Choh et al., 1973; Martinez and Sano, 1990). In the most reducing conditions, slags with N solubility as high as ⬃104 ppm N atm⫺1 or ⬃3 ⫻ 10⫺4 mol g⫺1 atm⫺1 have even been documented. Based on such results, metallurgists generally assume that, under very reducing conditions, the mechanism of solubility is governed by a chemical process of dissolution, in which the nitrogen gas reacts with the molten silicate to create new bonds with the atoms of the silicate network. Ito and Fruehan (1988), Martinez and Sano (1990), Schwerdtfeger et al. (1978) have thus proposed that the

4130

G. Libourel, B. Marty, and F. Humbert

dissolution reaction for N has to be formulated on an ionic basis, with nitrogen expressed in the form of (N3⫺) nitride ions, according to 1/2 N2 (gas) ⫹ 3/2 O2–(melt) ⫽ N3–(melt) ⫹ 3/4 O2(gas)

(7)

where O2⫺(melt) corresponds to non-bridging oxygen atoms of the polymerized network of the melt. In the presence of graphite and CO as in our experiments, nitrogen may also dissolve as cyanide radicals (CN⫺) in the silicate melt, following the reaction C (graphite) ⫹ 1/2 N2 (gas) ⫹ 1/2 O2–(melt) ⫽ CN–(melt) ⫹ 1/4 O2(gas) (8) Both Eqn. 7 and 8 imply dissociation of molecular nitrogen and proportionality of the N solubility to the partial pressure of nitrogen (PN2), the oxygen fugacity (fO2) and to the composition of the melt, through the activity of non-bridging oxygens (aO2⫺[melt]). In the iron making industry, graphite is systematically used as a reducing agent and, in the metallurgical literature, dissolution Eqn. 7 and 8 are re-arranged to better describe the smelter processes: 3 C (graphite) ⫹ N2 (gas) ⫹ 3 O2–(melt) ⫽ 2 N3–(melt) ⫹ 3 CO(gas) (9) 3 C (graphite) ⫹ N2 (gas) ⫹ O2–(melt) ⫽ 2 C–(melt) ⫹ CO(gas)

Fig. 6. Nitrogen solubility in CM-1 MORB-like melt as function of oxygen fugacity below IW at 1425°C. Between IW and ⬃IW ⫺ 7 (CCO buffer), nitrogen solubility in silicate melt (filled circles) increases drastically with the reducing character of the ambient atmosphere, i.e., by ⬎ 4 orders of magnitude. In these redox conditions, nitrogen is chemically dissolved in the silicate melt network. Data and equation regressions are also shown in the graph, and suggests that nitrogen solubility is almost proportional to fO2⫺3/4. Below ⬃IW ⫺ 7, all the glass samples (empty circles) contain precipitated crystals of osbornite (TiN), that explains the peculiar behavior of nitrogen solubility in these very reduced conditions. Note that, in agreement with such a crystallization, nitrogen solubility in these melts is far below the expected values. See text for explanations.

(10) In this case, the amount of dissolved N is proportional to PCO and not directly to fO2. However, since Eqn. 1 states that PCO is proportional to fO21/2, Eqn. 7 is equivalent to Eqn. 9 and Eqn. 8 similar to Eqn. 10. From Eqn. 7 and 8, the solubilities of N3⫺ ([N3⫺]n, chem) and CN⫺ ([CN⫺]n, chem) can be expressed as [N3–]n, chem ⫽ [N3–]/PN21/ 2 ⫽ K Eqn. 7 · 共aO2–[melt])3/2/fO23/4 (11) [CN–]n, chem ⫽ [CN–]/PN21/ 2 ⫽ K Eqn. 8 · 共aO2–[melt])1/2/fO21/4 (12) This formulation assumes that the activities of the nitrogen species in the melt can be conveniently approximated by their molar concentrations ([N3⫺] and [CN⫺]), due to their low concentration in the melt. KEqn. 7 and KEqn. 8 stand for the equilibrium constants of Eqn. 7 and 8, respectively. Since in our experimental conditions both the composition and the temperature are kept constant, the solubilities of N3⫺ and CN⫺ are of the forms [N3–]n, chem ⫽ ␣ · fO2–3/4

(13)

[CN–]n, chem ⫽ ␤ · fO2–1/4

(14)

where ␣ and ␤ are constants, and which, in a log-log diagram, result in straight lines with slopes of –3/4 and –1/4, respectively. Using this reasoning, it is thus possible to have direct insights on the speciation of nitrogen in basaltic melt under reducing conditions (Fig. 6). Between IW and IW ⫺ 7, the N

solubility decreases linearly in this log-log diagram with increasing fO2 and the correlation admits a slope of – 0.67. Note that if the regression in Figure 6 is performed without the data obtained at IW, the slope is ⫺0.805 with a better correlation coefficient: R2 ⫽ 0.98, suggesting that the best fit is likely between ⫺0.67 and ⫺0.81. Since this experimental slope is closer to ⫺0.75 than to ⫺0.25, we conclude that, in basaltic melt, nitrogen dissolves under reducing conditions as N3⫺ rather than as CN⫺. This finding is in agreement with metallurgical studies showing that N3⫺ in slags is more abundant by at least 1 order of magnitude than CN⫺ (Schwerdtfeger et al., 1978). Accordingly, the nitrogen chemical solubility ([N]n, chem) in basaltic melt can be calculated as follows: [N]n, chem ⬃ [N3–]n, chem ⫽ [N3–]/PN21/ 2

(15)

and expressed in ppm N atm⫺1/2, or in mol N2 g⫺1 atm⫺1/2 units. 4.3. The Effect of Titanium in the [C-N-O] Vapor System Below IW ⫺ 7, the solubility of N no longer follows the trend observed between IW and IW ⫺ 7 as the N content apparently reaches a plateau (Figs. 4 and 6). As discussed in subsection 3.2, we have detected the occurrence of TiN for samples equilibrated in these very reducing conditions, suggesting that nitrogen solubility may be buffered by the saturation of TiN in the melt. Indeed, we have shown that the mauve color in the equilibrated samples is due to partial reduction of Ti4⫹ in the melt according to the following equation (see also Eqn. 2):

Nitrogen solubility in basaltic melt

Ti4⫹ (melt) ⫹ 1/2 O2–(melt) ⫽ Ti3⫹ (melt) ⫹ 1/4 O2(gas)

4131

(16)

This reaction indicates that the decrease of the oxygen fugacity results in an increase in the Ti3⫹/Ti4⫹ ratio of the melt. Since the nitrogen solubility increases in the same direction, combination of Eqn. 7 and 16 gives the corresponding nitride precipitation reaction: 1/2 N2 (gas) ⫹ Ti4⫹(melt) ⫹ 2 O2– (melt) ⫽ TiN (osbornite) ⫹ O2 (gas) (17) or 1/2 N2 (gas) ⫹ TiO2 (melt) ⫽ TiN (osbornite) ⫹ O2 (gas)

(18)

Our experiments show that at 1425°C the saturation of TiN occurs for log fO2 between –17.4 and –16.4 (Table 2), in good agreement with a calculated value of log fO2 ⫽ ⫺16.2 atm from thermodynamic data (⌬GEqn. 18 ⫽ 461.4 kJ/mol at 1425°C from Thermodata and JANAF data base) and CM-1#6 run composition and conditions. The nitrogen solubility in melt and TiN precipitation can be envisioned as follows. When fO2 decreases beyond IW, the concentrations of N3⫺ and Ti3⫹ increase together according to Eqn. 7 and 16 (Fig. 4 or 6). It is important to note that, as demonstrated previously by Tomioka and Suito (1993) for BaO-TiOx melts, the change of titanium valence due to change in redox conditions has no influence on the solubility of nitrogen in melts. At ⬃IW ⫺ 7, the melt reaches saturation for TiN and the system contains now three phases: gas ⫹ melt ⫹ TiN. For a further decrease of the fO2, only the modal proportion of TiN is allowed to increase, the concentration of other nitrogen species, including N3⫺, remaining constant in the liquid. Since the N contents measured in this study are those of the melt (the TiN coating has been removed), it is understandable that the apparent nitrogen solubility is constant and lower than that predicted for a binary, gas ⫹ melt system (Fig. 6). All these observations highlight the potentially important role of titanium for the speciation of nitrogen under very reducing conditions. 4.4. A Model for the Nitrogen Solubility in Tholeiitic Magmas

Fig. 7. (a) Nitrogen solubility in CM-1 MORB-like melt as function of oxygen fugacity. Symbols as in Figure 4. Comparisons between measured and predicted values from our model of nitrogen solubility in basaltic melt. Note that for convenience data from TiN-bearing samples have been removed. See details in the text. (b) Modeled nitrogen content in a MORB-like melt as a function of oxygen fugacity and N2 partial pressure, ranging from PN2 ⫽ 10⫺3 to 103 atm. Small filled circles corresponds to N solubility measured in this study. Note that for convenience data from TiN-bearing samples have been removed.

[N]n, chem ⫽ [N2]/PN2 ⫽

Nitrogen dissolution in silicate melts hence obeys two different mechanisms: a physical solubility, [N]n, phys, in relatively oxidizing conditions and a chemical solubility, [N]n, chem, in more reducing environments. We have shown that the nitrogen physical solubility in the CM-1 tholeiitic-like melt is Henrian: [N]n, phys ⫽ [N2]/PN2 ⫽ 0.0611 ⫾ 0.0149 ppm N atm–1 (19) [N]n, phys ⫽ [N2]/PN2 ⫽ 2.21 ⫾ 0.53 ⫻ 10–9 mol N2 g–1 atm–1 (20)

fO2–3/4 · 2.13 ⫾ 0.11 ⫻ 10–17 mol N2 g–1 atm–1/2

The limit between these two solubility mechanisms is therefore obtained by equating these two variation laws and solving for the oxygen fugacity. The resulting log fO2 is ⫺10.7 atm, suggesting that the nitrogen physical solubility is taking place above IW ⫺ 1.3, whereas a chemical dissolution of N is expected for redox conditions below this value. Nitrogen solubility in tholeiitic magmas can also inferred from this data set, by noting that the bulk solubility of nitrogen is [N]n ⫽ [N]n, phys ⫹ [N]n, chem. Consequently, the nitrogen content in tholeiitic magmas as a function of oxygen and nitrogen fugacities/partial pressures (Fig. 7) can be calculated as follows:

In reducing conditions and when the present data are regressed with an equation of the form of y ⫽ a · xb and setting b ⫽ ⫺3/4, the solubility can be formulated as follows:

[N] ⫽ 0.0611 ⫾ 0.0149 · PN2 ⫹

[N]n, chem ⫽ [N2]/PN21/ 2 ⫽

[N2] ⫽ 2.21 ⫾ 0.53 ⫻ 10–9 · PN2 ⫹

fO2–3/4 · 5.97 ⫾ 0.31 ⫻ 10–10 ppm N atm–1/2

(21)

(22)

fO2–3/4 · 5.97 ⫾ 0.31 ⫻ 10–10 · PN21/2 ppm N

fO2–3/4 · 2.13 ⫾ 0.11 ⫻ 10–17 · PN21/2 mol N2 g–

(23)

(24)

4132

G. Libourel, B. Marty, and F. Humbert

As shown in Figure 7a, this model fits reasonably well our data set over the whole range of oxygen fugacity. We also note that the data of Miyazaki (1996) which show a slight increase in nitrogen solubility between IW and IW ⫺ 2, can be nicely fitted. As expected, the influence of PN2 on the N solubility is more important in the physical solubility regime where it is directly proportional to PN2 following the Henry’s law (Fig. 7b). In these oxidized conditions, it is noteworthy that the model reproduces almost perfectly the data of Kesson and Holloway (1974) and Miyazaki (1996) obtained however under high pressures. Interestingly, this modeling emphasizes that nitrogen degassing of a silicate liquid may proceed either by decreasing the pressure or by oxidation, or both below IW. 5. COSMOCHEMICAL AND GEOCHEMICAL IMPLICATIONS

5.1. Nitrogen Contents in Meteorites Under reducing conditions below IW ⫺ 1, this study shows that nitrogen behaves less and less as a volatile species, while inert rare gases, even if experiments are not available, are still expected to have a low solubility. It is therefore very likely that the nitrogen/rare gas ratio, i.e., N2/36Ar, increases rapidly with decreasing oxygen fugacity below that threshold. Such a fractionation is indeed observed in extraterrestrial materials since typical N2/36Ar ratios measured in chondrites are in the range of 105 to 107 (e.g., Pepin, 1991, and references therein), whereas the solar canonical value is ⬃37 (Anders and Grevesse, 1989). Although this difference suggests that N in meteorites is more likely bound to the crystal lattice in reducing conditions as documented in this study, it does not imply that meteoritic nitrogen has been incorporated in chondrites by equilibration with the nebular gas. Indeed, abundances of nitrogen in primitive chondrites, i.e., tens to thousands of parts per million (Clayton, 1981; Kerridge, 1985; Hashizume and Sugiura, 1995), are orders of magnitude higher than the abundance expected in silicate melt in equilibrium with the solar nebular gas, i.e., ⬃0.05 N ppm assuming PN2 ⬃ 1 ⫻ 10⫺7 atm and fO2 ⫽10⫺16 (Grossman, 1972; Anders and Grevesse, 1989). Similarly, high nitrogen contents from 40 to 1500 ppm documented in glass inclusions in olivines of several chondritic meteorites (Varela et al., 2003) cannot be due to equilibration in canonical nebular gas since the required total pressures would have been much higher than anticipated by any nebular models. Thus, it seems very likely that N may have been incorporated in precursor materials of primitive meteorites as nitride grains (Si2N2O, Si3N4, TiN) and organics (Robert and Epstein, 1982). Nitrogen in enstatite chondrites has raised considerable attention because this class of meteorites may represent a good cosmochemical analog for the material that formed the Earth as far as O and Cr (Javoy, 1997) and Mo (Dauphas et al., 2002) isotopes are concerned. Bulk E-chondrites contain ⬃1000 ppm N with a ␦15N value of ⫺30 ⫾ 10‰ and a N2/36Ar ratio of ⬃2 ⫻ 106 (Moore et al., 1969; Pepin, 1991). Since the solar nebula ␦15N value is thought to be ⬍ ⫺240‰ (Hashizume et al., 2000), possibly ⫺360‰ (Owen et al., 2001), on the order of 50 to 100 ppm N in E-chondrites might have been directly inherited from the solar nebula if the other N end-member was

“normal” planetary N (for the purpose of illustration, ␦15N ⬃ 0‰). This order of magnitude is again much higher than the amount of ⬃0.05 N ppm resulting from equilibration with solar gas. Therefore, either the “canonical” pressure of the solar nebula was locally much higher than previously thought by 4 orders of magnitude (e.g., transient atmospheres around parent bodies), or nitrogen, which might have been partly solar in origin, was processed into some solid compound before incorporation in E-chondrites. 5.2. Incorporation of Nitrogen in the Solid Earth The terrestrial mantle contains neon having a solar-like isotope composition, and one way to incorporate solar rare gases is by dissolution of a dense primitive atmosphere into a magma ocean, (e.g., Porcelli and Pepin, 2000), so that nitrogen could also have been incorporated in the mantle following such a process. The amount of nitrogen in the solid Earth is estimated to be ⬃2 ppm (Marty, 1995), and the mantle fO2 during metal segregation could have been as low as IW ⫺ 2.5 (Jones and Palme, 2000). A lower limit on the amount of nitrogen that could have been incorporated, computed assuming an atmospheric nitrogen partial pressure equivalent to the present-day one, is 0.4 ppm (Fig. 7b), which agrees reasonably well with the predicted N content of the solid Earth, if it is considered that the atmosphere could have been much thicker during the first tens of Ma than at present. Indeed, the difference of isotopic composition of atmospheric rare gases from that of mantle rare gases has been explained by isotopic fractionation during hydrodynamic escape of a terrestrial primitive atmosphere that was thicker by several orders of magnitude than the present-day one (e.g., Pepin, 1991; Porcelli and Pepin, 2000). Similarly, the neon content of the mantle requires the presence of a solar-like atmosphere having a pressure ⬃100 atm (Porcelli and Pepin, 2000). However, dissolution of a solar-like atmosphere does not provide the required geochemical source for nitrogen since the ␦15N value of the Earth, particularly of the mantle (⫺5‰) is clearly different from that of solar N (see above). Therefore, in addition to the occurrence of a solar-like atmosphere as the source of mantle Ne, nitrogen values strongly suggest that this primitive atmosphere also contained volatile elements supplied by “planetary” bodies analogous to chondritic meteorites, i.e., from ⫺30 to ⫹40‰ for most chondritic classes, (Kerridge, 1985; Grady et al., 1986; Javoy, 1997; Tolstikhin and Marty, 1998). It is of note however that modeling the acquisition of nitrogen both in the solid Earth and the atmosphere will require a better knowledge of the partition of nitrogen between metal and silicate under pressure, temperature and oxygen fugacity conditions relevant to those of core formation. 5.3. The Degassing Rate of Mantle Nitrogen Through Time As discussed in the preceding section, the physical solubility of nitrogen in tholeiitic melt (2.2 ⫾ 0.5 ⫻ 10⫺9 mol g⫺1 atm⫺1) is identical, within uncertainty, to that of argon, 2.6 ⫾ 0.4 ⫻ 10⫺9 mol g⫺1 atm⫺1 (Jambon et al., 1986), but lower than that of helium, 2.5 ⫾ 0.5 ⫻ 10⫺8 mol g⫺1 atm⫺1 (Jambon et al., 1986; Carroll and Draper, 1994) and than that of CO2, 1 ⫻ 10⫺8 mol g⫺1 atm⫺1 (Jendrzejewski et al., 1997). A similar

Nitrogen solubility in basaltic melt

solubility of N and Ar is also supported by the volatile analysis of vesicles from midocean ridge basalts (MORB) which showed no detectable N2-Ar fractionation over a wide range of vesicularities and therefore various degrees of magma degassing (Marty, 1995). Based on variations in the respective behavior of nitrogen and argon for a suite of MORBs from the Southwest Indian Ridge, Cartigny et al. (2001) proposed that argon is 1.2 times more soluble than nitrogen, a subtle difference which is beyond the precision reached by experimental work referenced above and by the present study, and which nevertheless support a similar degassing rate. The major process of mantle degassing occurs along midocean ridges during generation and eruption of tholeiitic magmas, and rare gases have been intensively used to constrain fluxes of volatile elements from the mantle into the atmosphere and hydrosphere. Of particular interest is argon which has one isotope (40Ar) produced during the decay (T1/2 ⫽ 1.25 Ga) of one isotope of potassium (40K), and other isotopes (38Ar and 36Ar) having been trapped during Earth’s formation. Hence the kinetics of mantle-atmosphere volatile transfer have been largely modeled from K-Ar systematics. The similarity of the N2 and Ar solubilities implies that the N2/Ar ratio is little fractionated during magma degassing, and suggests that Ar could constitute a good proxy for the behavior of nitrogen during Earth-atmosphere evolution for fO2 conditions where physical solubility dominates. Accordingly, the N2/Ar ratio can be used as a powerful tracer of sources. In contrast, CO2/N2 and He/N2 ratios are subject to fractionation during degassing since their solubility ratio are ⬃5 and 11, respectively, and therefore, their use as geochemical tracers is subject to caution if specific procedures of correction are not undertaken. In a study of a suite of MORB glasses, Marty (1995) noted that the N2/40Ar ratios (where 40Ar has been produced by the decay of 40K) of the MORB source, presumably the upper mantle, and the atmosphere were comparable. In contrast, the N2/36Ar ratio of the MORB mantle was estimated to be of the order of 106, 2 orders of magnitude higher than N2/36Ar ratio of the atmosphere (1.2 ⫻ 104). Because 36Ar is thought to have been degassed very early from the MORB mantle and, in contrast, 40Ar has been produced continuously over the Earth’s history, the contrast between the atmosphere and the mantle reservoirs suggests that nitrogen could have been degassed later than 36Ar into the atmosphere as 40Ar has been (Marty, 1995). A way to tackle this problem is to consider that nitrogen degassing was less efficient than that of argon, because the solubility of nitrogen might have been controlled by the redox state of the mantle through time. This study demonstrates that the nitrogen solubility at low pressure is independent of the oxygen fugacity in oxidizing conditions down to IW ⫺ 1.3 (log fO2 ⫽ ⫺10.7 atm). This range of redox conditions covers that of all present-day terrestrial basaltic magmas including QFM and NiNiO buffers, and therefore no drastic elemental fractionation is expected between N and Ar at present during mantle-derived magma degassing. Recent advances in experimental determination of partitioning of vanadium between mantle mineral phases suggest that the Archean mantle was as oxidized as the present-day mantle from at least 3.5 Ga ago (Canil, 2002), confirming earlier work which proposed that the chemistry of oldest basalts at ⬃ 3.8 Ga ago is consistent with a fO2 near QFM (Delano,

4133

1993). It is therefore highly unlikely that the Archean mantle was reduced below IW and, by consequence that the N2/Ar ratio could have been fractionated more than at present during mantle degassing. Hence the problem raised by N-Ar isotope systematics between the mantle and the atmosphere requires other processes having fractionated rare gases from nitrogen. The similar N2/40Ar ratios of the mantle and the atmosphere may in fact translate a comparable behavior of nitrogen and potassium during mantle-surface exchanges. Provided that nitrogen is an incompatible element as potassium is, both elements are expected to be extracted from the mantle source at a similar rate which depends on the partial melting rate. Part of the potassium is recycled back into the mantle, and there is independent evidence that nitrogen is also recycled to some extent into the mantle (Andersen et al., 1995), probably as ammonium substituting for potassium. Therefore, the N/K ratio, represented by the N2/40Ar ratio, might have been kept approximately constant over the Earth’s history. In contrast, rare gases are not quantitatively recycled into the mantle (Staudacher and Alle`gre, 1988), providing an efficient fractionation mechanism between nitrogen and rare gases at subduction zones. This mass balance of nitrogen between the mantle and the Earth’s surface is required to account for variations of N isotope ratios among the concerned reservoirs, which is beyond the scope of the present paper and is developed elsewhere (Marty and Dauphas, 2003). 6. CONCLUSIONS

Owing to a set of equilibration experiments of silicate melts under controlled nitrogen partial and oxygen pressures in the [C-N-O] system, this study has documented the solubility of nitrogen in basaltic melt over 18 oxygen fugacity log units, i.e., from air to redox conditions below the C/CO buffer. The nitrogen solubility is very sensitive to the oxygen fugacity, and its incorporation in silicate melt follows two different mechanisms. In oxidizing conditions between air and IW ⫺ 1.3, the nitrogen N solubility is found to be constant, indicative of the physical dissolution of N2 molecule in the cavities of silicate melt network in a similar manner as rare gases do. In contrast, the nitrogen solubility increases drastically by several orders of magnitude below IW ⫺ 1.3 with decreasing fO2, suggesting that nitrogen from the gas reacts with the silicate melt to form new chemical bounds with the atoms of the molten silicate network. Under reducing conditions, the slope of the linear relationship between the nitrogen solubility and the logarithm of the fO2 strongly suggests that nitrogen dissolves as N3⫺ rather than as CN⫺. Because the air to IW ⫺ 1.3 range of log fO2 covers that characterizing terrestrial magmas over most of the Earth’s history, and because N solubility was found to be in the range of that of argon, we strongly advocate the use of the N2/Ar ratio as a powerful tracer of sources, since this ratio is not fractionated during magma generation and degassing. Furthermore, because this paper clearly demonstrates that reducing conditions favor the incorporation of nitrogen in silicates melts, we suggest that high N contents in primitive meteorites are unlikely to result from nitrogen dissolution from the solar nebula gas, but instead are very likely due to the seeding of nitrides in their precursor materials. Further work is now in progress to

4134

G. Libourel, B. Marty, and F. Humbert

address the effect of composition and gas speciation on nitrogen solubility in basaltic magmas. Acknowledgments—This work is part of the PhD thesis of F. Humbert. A. Rouillier is thanked for assistance in the high-temperature experimental laboratory of the CRPG-CNRS. We also thank L. Zimmerman, P. Coget, X. Framboisier, and J.C. Demange for assistance for mass spectrometry measurements and handling of the samples in the CO2laser chamber. We are also grateful to R. Podor, F. Diot and S. Barda for assistance for electron microprobe analyses at the “Service d’Analyses” of the Universite´ Henri Poincare´ , Nancy (France). We particularly thank Y. Dusausoy for Electron Paramagnetic Resonance analyses and its interpretation of the EPR spectra. Ethan F. Baxter, John R. Holloway and the associate editor Jun-Ichi Matsuda are thanked for their very thorough and helpful reviews that helped improve the clarity of the manuscript. This work was supported by INSU through IT and PNP grants (G.L and B.M). CRPG contribution number 1627. Associate editor: J.-I. Matsuda REFERENCES Anders E. and Grevesse N. (1989) Abundances of the elements: Meteoritic and solar. Geochim. Cosmochim. Acta 53, 197–214. Andersen T., Burke E. A. J., and Neumann E. R. (1995) Nitrogen-rich fluid in the upper mantle: Fluid inclusions in spinel dunite from Lanzarote, Canary Islands. Contrib. Mineral. Petrol. 120, 20 –28. Boyd S. R., Wright I. P., Franchi I. A., and Pillinger C. T. (1988) Preparation of sub-nanomole quantities of nitrogen gas for stable isotopic analysis. J. Phys. E: Sci. Instrum. 21, 876 – 885. Broadhurst C. L., Drake M. J., Hagee B. E., and Bernatowicz T. J. (1992) Solubility and partitioning of Ne, Ar, Kr, and Xe in minerals and synthetic basaltic melts. Geochim. Cosmochim. Acta 56, 709 – 723. Canil D. (2002) Vanadium in peridotites, mantle redox and tectonic environments: Archean to present. Earth Planet. Sci. Lett. 195, 75–90. Carroll M. R. and Draper D. S. (1994) Noble gases as trace elements in magmatic processes. Chem. Geol. 117, 37–56. Carroll M. R. and Holloway J. R. (1994) Volatiles in magmas. In Reviews in Mineralogy, Vol. 30 (ed. P. H. Ribbe), pp. 517. Mineralogical Society of America, Washington, DC. Cartigny P. (1997) Concentration, Composition Isotopique et Origine de l’Azote Dans le Manteau Terrestre. Ph.D. dissertation, University of Paris 7. Cartigny P., Jendrzejewski F., Pineau F., Petit E., and Javoy M. (2001) Volatile (C, N, Ar) variability in MORB and the respective roles of mantle source heterogeneity and degassing: The case of the Southwest Indian Ridge. Earth Planet. Sci. Lett. 194, 241–257. Choh T., Hanaki Y., Kato T., and Inouye M. (1973) Nitrogen absorption of liquid CaO-Al2O3 and CaO-Al2O3-SiO2 slags under reducing atmosphere. Trans. I. S. I. J. 13, 218 –225. Clayton R. N. (1981) Isotopic variations in primitive meteorites. Philos. Trans. R. Soc. London 303, 339 –349. Dauphas N., Marty B., and Reisberg L. (2002) Inference on terrestrial genesis from molybdenum isotope systematics. Geophys. Res. Lett. 29, 6, 10.1029/2001GL014237. Delano J. W. (1993) Oxidation state of the Earth’s upper mantle during the last 3800 million years: Implications for the origin of life. Lunar Planet. Sci. 24, 395–396. Doremus R. H. (1966) Physical solubility of gases in fused silicates. J. Am. Ceram. Soc. 49, 461– 462. Fogel R. A. (1994) Nitrogen solubility in aubrite and E-chondrite melts. 25th Lunar Planet. Sci. Conf. Frick U. and Pepin R. O. (1981) Microanalysis of nitrogen isotope abundances: Association of nitrogen with noble gas carriers in Allende. Earth Planet. Sci. Lett. 56, 64 – 81. Grady M. M., Wright I. P., Carr L. P., and Pillinger C. T. (1986) Compositional differences in enstatite chondrites based on carbon and nitrogen stable isotope measurements. Geochim. Cosmochim. Acta 50, 2799 –2813. Grossman L. (1972) Condensation in the primitive solar nebula. Geochim. Cosmochim. Acta 36, 597– 619.

Hashizume K. and Sugiura N. (1990) Precise measurement of nitrogen isotopic composition using a quadrupole mass spectrometer. Mass Spectrom. 38, 269 –286. Hashizume K. and Sugiura N. (1995) Nitrogen isotopes in bulk ordinary chondrites. Geochim. Cosmochim. Acta 59, 4057– 4069. Hashizume K., Chaussidon M., Marty B., and Robert F. (2000) Solar wind record on the moon: Deciphering presolar from planetary nitrogen. Science 290, 5494, 1142–1145. Hashizume K., Marty B., and Wieler R. (2002) Analyses of nitrogen and argon in single lunar grains: Towards a quantification of the asteroidal contribution to planetary surfaces. Earth Planet. Sci. Lett. 202, 201–216. Hayatsu A. and Waboso C. E. (1985) The solubility of rare gases in silicate melts and implications for K-Ar dating. Chem. Geol. 52, 97–102. Hiyagon H. and Ozima M. (1986) Partition of rare gases between olivine and basalt melt. Geochim. Cosmochim. Acta 50, 2045–2057. Hofmann A. W. (1988) Chemical differentiation of the Earth: The relationship between mantle, continental crust and oceanic crust. Earth Planet. Sci. Lett. 90, 297–314. Humbert F. (1998) Solubilite´ de l’Azote Dans les Silicates Liquides— Influence de la Fugacite´ d’Oxyge`ne et de la Composition. Ph.D. dissertation, Inst. Nal. Polytech, Lorraine, France. Humbert F., Libourel G., France-Lanord C., Zimmermann L., and Marty B. (2000) CO2-laser extraction-static mass spectrometry analysis of ultra-low concentrations of nitrogen in silicates. Geostand. Newslett. 24, 255–260. Ito K. and Fruehan R. J. (1988) Thermodynamics of nitrogen in CaO-SiO2-Al2O3 slags and its reaction with Fe-Csat melts. Metall. Trans. B 19B, 419 – 425. Jambon A., Weber H., and Braun O. (1986) Solubilities of He, Ne, Ar, Kr and Xe in a basalts melt in the range 1250-1600°C: Geochemical implications. Geochim. Cosmochim. Acta 50, 401– 408. Javoy M. (1997) The major volatile elements of the Earth: Their origin, behavior, and fate. Geophys. Res. Lett. 24, 177–180. Javoy M. and Pineau F. (1991) The volatile record of a popping rock from the Mid-Atlantic Ridge at 14°N: Chemical and isotopic composition of gases trapped in the vesicles. Earth Planet. Sci. Lett. 107, 598 – 611. Jendrzejewski N., Trull T. W., Pineau F., and Javoy M. (1997) Carbon solubility in Mid-Ocean Ridge basaltic melt at low pressures (2501950 bar). Chem. Geol. 138, 81–92. Jones J. H. and Palme H. (2000) Geochemical constraints on the origin of the Earth and Moon. In Origin of the Earth and Moon (eds. R. M. Canup and K. Righter), pp. 197–216. University of Arizona Press, Tucson. Kerridge F. (1985) Carbon, hydrogen and nitrogen in carbonaceous chondrites: Abundances and isotopic compositions in bulk samples. Geochim. Cosmochim. Acta 49, 1707–1714. Kesson S. E. and Holloway J. R. (1974) The generation of N2-CO2H2O fluids for use in hydtothermal experimentation. II Melting of albite in a multispecies fluid. Am. Mineral. 59, 598 – 603. Kirsten T. (1968) Incorporation of rare gases in solidifying enstatite melts. J. Geophys. Res. 73, 2807–2810. Lux G. (1986) The behaviour of noble gases in silicate liquids: Solution, diffusion, bubbles and surface effects, with applications to natural samples. Geochim. Cosmochim. Acta 51, 1549 –1560. Martinez E. and Sano N. (1990) Nitrogen solubilities in CaO-CaF2 melts. Steel Res. 58, 485– 490. Marty B. (1995) Nitrogen content of the mantle inferred from N2-Ar correlation in oceanic basalts. Nature 377, 326 –329. Marty B. and Humbert F. (1997) Nitrogen and argon isotopes in oceanic basalts. Earth Planet. Sci. Lett. 152, 101–112. Marty B. and Dauphas N. (2003) The nitrogen record of crust-mantle interaction and mantle convection from Archean to Present. Earth Planet. Sci. Lett. 206, 397– 410. Marty B., Lenoble M., and Vassard N. (1995) Nitrogen, helium and argon in basalt: A static mass spectrometric study. Chem. Geol. 120, 183–195. Mathew K. J., Kim J. S., and Marti K. (1998) Martian atmospheric and indigenous components of xenon and nitrogen in the Shergotty, Nakhla, and Chassigny group meteorites. Meteorit. Planet. Sci. 33, 655– 664.

Nitrogen solubility in basaltic melt Miyazaki A. (1996) Studies on Solubilities of Nitrogen and Noble Gases in Silicate Melts. Ph.D. dissertation University of Tokyo. Miyazaki A., Hiyagon H., and Sugiura N. (1995) Solubility of nitrogen and argon in basalt melt under oxidizing condition. In Volatiles in the Earth and Solar System (ed. K. A. Farley), pp. 276 –283. AIP. Moore C. B., Gibson E. K., and Keil K. (1969) Nitrogen abundances in enstatite chondrites. Earth Planet. Sci. Lett. 6, 457– 460. Mulfinger H. O. (1966) Physical and chemical solubility of nitrogen in glass melts. J. Am. Ceram. Soc. 49, 462– 467. Murty S. V. S., Mohapatra R. K., and Clement C. J. (1996) Nitrogen and noble gas components in the Martian orthopyroxenite ALH84001. Lunar Planet. Sci. Conf. Owen T., Mahaffy P. R., Niemann H. B., Atreya S., and Wong M. (2001) Protosolar nitrogen. Astrophys. J. 553, L77–L79. Pepin R. O. (1991) On the origin and early evolution of terrestrial planetary atmospheres and meteoritic volatiles. Icarus 92, 1–79. Porcelli D. and Pepin R. O. (2000) Rare gas constraints on early Earth history. In Origin of the Earth and Moon (eds. R. M. Canup and K. Righter), pp. 435– 458. University of Arizona Press, Tucson. Robert F. and Epstein S. (1982) The concentration and isotopic composition of hydrogen, carbon and nitrogen in carbonaceous chondrites. Geochim. Cosmochim. Acta 46, 81–95. Schackelford J. F., Studt P., and Fulrath R. M. (1972) Solubility of gases in glass. II. He, Ne and H2 in fused silica. J. Appl. Phys. 43, 1619 –1626. Schwerdtfeger K., Fix W., and Schubert H. G. (1978) Solubility of nitrogen and carbon in CaO-SiO2-Al2O3 slag in presence of graphite at 1450°C. Ironmaking Steelmaking 2, 67–71.

4135

Shelby J. E. (1974) Helium diffusion and solubility in K2O-SiO2 glasses. J. Am. Ser. Soc. 57, 236 –263. Shibata T., Takahashi E., and Matsuda J. I. (1998) Solubility of neon, argon, krypton and xenon in binary and ternary silicate systems: A new view on noble gas solubility. Geochim. Cosmochim. Acta 62, 1241–1253. Shilobreeva S., Kadik A., Matveev S., and Chappyzhnicov B. (1994) Solubility of N2 in basalt and albite melt at pressures 1-3 Kbar and temperature 1250°C. International Workshop on Noble Gas Geochemistry and Cosmochemistry, Yamada Conf. XXXVIII, pp. 21–22. Staudacher T. and Alle`gre C. J. (1988) Recycling of oceanic crust and sediments: The noble gas subduction barrier. Earth Planet. Sci. Lett. 89, 173–183. Tissandier L., Libourel G., Toplis M., and Chaussidon M. (1998) Alkali volatilization at high temperature in silicate melts. Meteoritics 33, A154. Tolstikhin I. N. and Marty B. (1998) The evolution of terrestrial volatiles: A view from helium, neon, argon and nitrogen isotope modelling. Chem. Geol. 147, 27–52. Tomioka K. and Suito H. (1993) Thermodynamics of nitrogen in BaO-TiOx melts. Metall. Trans. B 24B, 131–138. Tsuchiyama A., Nagahara H., and Kushiro I. (1981) Volatilization of sodium from silicate melt spheres and its application to the formation of chondrules. Geochim. Cosmochim. Acta 45, 1357–1367. Varela M. E., Bonnin-Mosbah M., Kurat G., and Gallien J. P. (2003) Nitrogen microanalysis of glass inclusions in chondritic olivines by nuclear reaction. Geochim. Cosmochim. 67, 1247–1257.