Figures and Tables

top

Introduction

The accretion of reduced building blocks during Earth’s formation released enough energy to at least partially melt the Earth, forming one or more magma oceans (Elkins-Tanton, 2012

Elkins-Tanton, L.T. (2012) Magma oceans in the inner solar system. Annual Review of Earth and Planetary Sciences 40, 113–139.

). The evolving nitrogen abundance in the degassing magma ocean, and thus in Earth’s early atmosphere, was fundamental to the development of habitable conditions and the maintenance of the terrestrial atmosphere and biosphere (Goldblatt et al., 2009

Goldblatt, C., Claire, M.W., Lenton, T.M., Matthews, A.J., Watson, A.J., Zahnle, K.J. (2009) Nitrogen-enhanced greenhouse warming on early Earth. Nature Geosciences 2, 891–896.

). Yet, the composition of Earth’s early atmosphere remains debated: it is thought to have been either i) neutral to oxidising and composed of H₂O, CO₂, and N₂ if it resembled modern volcanic gases (Kasting, 1993

Kasting, J.F. (1993) Earth’s early atmosphere. Science 259, 920–926.

), or ii) reducing and CH₄- and NH₃-rich if it degassed from the reduced materials that formed the Earth (Zahnle et al., 2020

Zahnle, K.J., Lupu, R., Catling, D.C., Wogan, N. (2020) Creation and evolution of impact-generated reduced atmospheres of early Earth. Planetary Science Journal 1, 11.

), based on geochemical evidence that magma ocean degassing contributed to the formation of the Hadean atmosphere (see review by Gaillard et al., 2021

Gaillard, F., Bouhifd, M.A., Füri, E., Malavergne, V., Marrocchi, Y., Noack, L., Ortenzi, G., Roskosz, M., Vulpius, S. (2021) The Diverse Planetary Ingassing/Outgassing Paths Produced over Billions of Years of Magmatic Activity. Space Science Reviews 217, 1–54.

). Therefore, reconstructing nitrogen degassing during Earth’s magma ocean stage(s) is critical to constraining the composition of the early atmosphere.

Nitrogen is often considered to be an inert molecule (N₂) and expected to behave somewhat like noble gases. However, recent works have shown that nitrogen is not always chemically inert and occurs as various species (N³⁻, NH₃, NH₂⁻, NH²⁻ and N₂) in silicate melts (Mikhail and Sverjensky, 2014

Mikhail, S., Sverjensky, D.A. (2014) Nitrogen speciation in upper mantle fluids and the origin of Earth’s nitrogen-rich atmosphere. Nature Geosciences 7, 816–819.

; Dalou et al., 2019a

Dalou, C., Hirschmann, M.M., Jacobsen, S.D., Le Losq, C. (2019a) Raman spectroscopy study of C-O-H-N speciation in reduced basaltic glasses: Implications for reduced planetary mantles. Geochimica et Cosmochimica Acta 265, 32–47.

; Mosenfelder et al., 2019

Mosenfelder, J.L., Von Der Handt, A., Füri, E., Dalou, C., Hervig, R.L., Rossman, G.R., Hirschmann, M.M. (2019) Nitrogen incorporation in silicates and metals: Results from SIMS, EPMA, FTIR, and laser-extraction mass spectrometry. American Mineralogist 104, 31–46.

; Boulliung et al., 2020

Boulliung, J., Füri, E., Dalou, C., Tissandier, L., Zimmermann, L., Marrocchi, Y. (2020) Oxygen fugacity and melt composition controls on nitrogen solubility in silicate melts. Geochimica et Cosmochimica Acta 284, 120–133.

; Grewal et al., 2020

Grewal, D.S., Dasgupta, R., Farnell, A. (2020) The speciation of carbon, nitrogen, and water in magma oceans and its effect on volatile partitioning between major reservoirs of the Solar System rocky bodies. Geochimica et Cosmochimica Acta 280, 281–301.

) and fluids (Li and Keppler, 2014

Li, Y., Keppler, H. (2014) Nitrogen speciation in mantle and crustal fluids. Geochimica et Cosmochimica Acta 129, 13–32.

). The speciation and solubility of nitrogen primarily depend on oxygen fugacity (Libourel et al., 2003

Libourel, G., Marty, B., Humbert, F. (2003) Nitrogen solubility in basaltic melt. Part I. Effect of oxygen fugacity. Geochimica et Cosmochimica Acta 67, 4123–4135.

; Boulliung et al., 2020

Boulliung, J., Füri, E., Dalou, C., Tissandier, L., Zimmermann, L., Marrocchi, Y. (2020) Oxygen fugacity and melt composition controls on nitrogen solubility in silicate melts. Geochimica et Cosmochimica Acta 284, 120–133.

; Bernadou et al., 2021

Bernadou, F., Gaillard, F., Füri, E., Marrocchi, Y., Slodczyk, A. (2021) Nitrogen solubility in basaltic silicate melt-Implications for degassing processes. Chemical Geology 573, 120192.

), which increased during the evolution of Earth’s magma ocean (Frost et al., 2008

Frost, D.J., Mann, U., Asahara, Y., Rubie, D.C. (2008) The redox state of the mantle during and just after core formation. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 366, 4315–4337.

). Thus, the rate of nitrogen degassing and the reactions involved must have evolved similarly over geological time.

Nitrogen isotopic compositions, conventionally normalised to the present atmospheric value and reported as , are useful for reconstructing Earth’s N degassing history. For instance, the range of δ¹⁵N values observed in Archean diamonds corresponds to that in present day diamonds and mid-ocean ridge basalts (MORBs), implying that the mantle δ¹⁵N value has not evolved since the Archean (Cartigny and Marty, 2013

Cartigny, P., Marty, B. (2013) Nitrogen isotopes and mantle geodynamics: The emergence of life and the atmosphere–crust–mantle connection. Elements 9, 359–366.

). This is consistent with the observation that diffusion controlled N₂ degassing from the present day mantle, which is expected to preferentially segregate ¹⁴N into the atmosphere following Graham’s law (i.e. the rate of diffusion/effusion of a gas is inversely proportional to the square root of its molecular weight; Javoy et al., 1986

Javoy, M., Pineau, F., Delorme, H. (1986) Carbon and nitrogen isotopes in the mantle. Chemical Geology 57, 41–62.

), is limited to isotopic fractionations of 1–1.5 ‰ (Marty and Dauphas, 2003

Marty, B., Dauphas, N. (2003) The nitrogen record of crust–mantle interaction and mantle convection from Archean to present. Earth and Planetary Sciences Letters 206, 397–410.

). However, Rayleigh models of equilibrium MORB degassing (Javoy et al., 1986

Javoy, M., Pineau, F., Delorme, H. (1986) Carbon and nitrogen isotopes in the mantle. Chemical Geology 57, 41–62.

; Cartigny et al., 2001

Cartigny, P., Jendrzejewski, N., Pineau, F., Petit, E., 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 and Planetary Science Letters 194, 241–257.

) only consider the reaction . This reaction has been experimentally determined at oxygen fugacity conditions (fO₂) > IW (Libourel et al., 2003

Libourel, G., Marty, B., Humbert, F. (2003) Nitrogen solubility in basaltic melt. Part I. Effect of oxygen fugacity. Geochimica et Cosmochimica Acta 67, 4123–4135.

; Boulliung et al., 2020

Boulliung, J., Füri, E., Dalou, C., Tissandier, L., Zimmermann, L., Marrocchi, Y. (2020) Oxygen fugacity and melt composition controls on nitrogen solubility in silicate melts. Geochimica et Cosmochimica Acta 284, 120–133.

), which applies to present day conditions of upper mantle degassing. However, because the integrated fO₂ during core formation was closer to IW–2 (Frost et al., 2008

Frost, D.J., Mann, U., Asahara, Y., Rubie, D.C. (2008) The redox state of the mantle during and just after core formation. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 366, 4315–4337.

), this reaction does not hold for the more reduced N species present in the magma ocean. Larger N isotopic fractionations are expected under reducing conditions at which N does not behave inertly. Therefore, quantifying N isotopic fractionations under conditions at which N dissolves in magmas as nitride or N-H complexes is prerequisite to quantitatively modelling the evolution of N concentrations and isotopic compositions during the formation of the Hadean atmosphere.

top

Degassing Experiments and Results

To establish the effect of fO₂ on the N degassing rate and identify N degassing reactions that may have occurred in the Earth’s magma ocean, we determined the N concentration, speciation, and isotopic fractionation of a basaltic komatiite (used as magma ocean analogue melt) during degassing at 1.5 GPa and 1550 °C in pure forsterite capsules. Degassing experiments were performed using a series of Fe-free primitive basalts in which we explored fO₂ conditions ranging from IW to IW−4.2 (Table 1). These fO₂ conditions were not determined in the Fe-free experiments, but only estimated afterwards from the IW reaction in equivalent Fe-bearing runs (see Supplementary Information). The initial ¹⁵N/¹⁴N ratio of the starting material was 0.946 ± 0.003, as determined in sample V133-PB4N2Si (Table 1), in which the initial N content was equal, within errors, to that measured after the experiment. Nitrogen was saturated in five experiments, as evidenced by bubbles in the quenched glass and/or surrounding forsterite (Fig. 1, Table 1).

Exp ID	Estimated fO2 (ΔIW)	Initial N content (ppm)	Measured N content (ppm)	F (remaining N fraction)	Measured 15N/14N ratios	Calculated δ15N normalised (‰)	Bubbles present
V152-PB05N	0.0	910	798 ± 4	0.88 ± 0.01	0.871 ± 0.005	−79 ± 8
V149-PB1N	−0.3	1945	1316 ± 16	0.68 ± 0.01	0.845 ± 0.028	−107 ± 34
V129-PB2N	−1.0	5075	3084 ± 377	0.61 ± 0.07	0.916 ± 0.003	−32 ± 6	✓
V148-PB4N	−2.3	10435	4914 ± 102	0.47 ± 0.01	0.907 ± 0.011	−41 ± 13	✓
V133-PB4N2Si	−2.6	8371	8335 ± 194	1.00 ± 0.02	0.946 ± 0.003	0 ± 6	✓
V125-PB8N	−3.3	16076	13342 ± 841	0.83 ± 0.05	0.925 ± 0.004	−5 ± 7	✓
V151-PB6N3Si	−3.4	11353	7534 ± 526	0.66 ± 0.05	0.926 ± 0.007	−21 ± 10	✓
V150-PB8N6Si	−4.2	14609	13481 ± 735	0.93 ± 0.05	0.941 ± 0.006	−5 ± 9	✓

Full size image

The major element compositions of the silicate glasses were measured by electron microprobe (Table S-1). Nitrogen concentrations and ¹⁵N/¹⁴N ratios were determined by secondary ion mass spectrometry (SIMS) using the CAMECA IMS 1280-HR2 at the CRPG (Table 1). On each sample, four to six spot analyses of ¹⁴N¹⁶O⁻ and ¹⁵N¹⁶O⁻ were performed (see Supplementary Information). Measured N concentrations were homogeneous (i.e. with standard deviations below 10 %). N concentrations at 1.5 GPa and 1550 °C decreased from 13,481 ± 735 ppm (at IW−4.2) to 798 ± 4 ppm (at IW; Fig. 2a). In N saturated samples (as attested by the presence of bubbles), N concentrations decreased from 13,481 ± 735 ppm (at IW−4.2) to 3,084 ± 377 ppm (at IW–1). This decrease of N solubility with increasing fO₂ is consistent with observations in silicate melts at 1 atm (Libourel et al., 2003

Libourel, G., Marty, B., Humbert, F. (2003) Nitrogen solubility in basaltic melt. Part I. Effect of oxygen fugacity. Geochimica et Cosmochimica Acta 67, 4123–4135.

; Boulliung et al., 2020

Boulliung, J., Füri, E., Dalou, C., Tissandier, L., Zimmermann, L., Marrocchi, Y. (2020) Oxygen fugacity and melt composition controls on nitrogen solubility in silicate melts. Geochimica et Cosmochimica Acta 284, 120–133.

) and modelled at pressures up to 1 GPa (Bernadou et al., 2021

Bernadou, F., Gaillard, F., Füri, E., Marrocchi, Y., Slodczyk, A. (2021) Nitrogen solubility in basaltic silicate melt-Implications for degassing processes. Chemical Geology 573, 120192.

). Yet, the N solubilities observed herein are up to three orders of magnitude higher than those in a similar melt composition and at similar temperatures at 1 atm (Boulliung et al., 2020

Boulliung, J., Füri, E., Dalou, C., Tissandier, L., Zimmermann, L., Marrocchi, Y. (2020) Oxygen fugacity and melt composition controls on nitrogen solubility in silicate melts. Geochimica et Cosmochimica Acta 284, 120–133.

) and one order of magnitude higher than those at 0.08 GPa (Bernardou et al., 2021), consistent with the effect of pressure on N solubility described by Bernardou et al. (2021). The variation of N solubilities with fO₂ is directly related to changes of N speciation in the silicate melts, which were determined by Raman spectroscopy (see Supplementary Information). Around IW−4 and below, N is very soluble because it is dissolved as nitride, forming Si-N bonds with the silicate network; indeed, in very N-rich samples (>1 wt. % N), Si-N vibrations around 800 cm⁻¹ were observed (Fig. 2b). In less reduced samples (fO₂ > IW−4), N is mostly dissolved as N-H complexes (Fig. 2c), and above IW−2, N is also dissolved as molecular N₂ (Fig. 2d). Consistent with previous studies (Mosenfelder et al., 2019

Mosenfelder, J.L., Von Der Handt, A., Füri, E., Dalou, C., Hervig, R.L., Rossman, G.R., Hirschmann, M.M. (2019) Nitrogen incorporation in silicates and metals: Results from SIMS, EPMA, FTIR, and laser-extraction mass spectrometry. American Mineralogist 104, 31–46.

; Grewal et al., 2020

Grewal, D.S., Dasgupta, R., Farnell, A. (2020) The speciation of carbon, nitrogen, and water in magma oceans and its effect on volatile partitioning between major reservoirs of the Solar System rocky bodies. Geochimica et Cosmochimica Acta 280, 281–301.

), N speciation in the silicate glasses controls N solubility variations with fO₂ conditions. Depending on initial N content and fO₂, the fraction of N degassed varied from 7 to 53 %. Samples V152-PB0.5N and V149-PB1N did not show any evidence of degassing, i.e. no bubbles were observed in the glass or the forsterite rim, but lost 12 and 32 % of their initial N, respectively, likely due to diffusion through cracks formed in the forsterite capsule during compression.

Full size image

The use of ¹⁵N spiked samples permits us to obtain precise ¹⁵N/¹⁴N ratios in the silicate (with low uncertainties of 0.1 ‰ on average). With increasing fraction of degassed N, ¹⁵N/¹⁴N ratios decreased from 0.946 ± 0.003 in the undegassed sample to 0.907 ± 0.011 in the most degassed sample (53 % of initial N lost). The ¹⁵N/¹⁴N ratios were normalised to the initial ¹⁵N/¹⁴N ratio of the undegassed glass V133-PB4N2Si (0.946 ± 0.003) and expressed as δ¹⁵N_norm as:

These normalised values are not calculated relative to the atmospheric N isotopic composition (¹⁵N/¹⁴N_air = 0.003676) and cannot be directly compared to the δ¹⁵N values of natural samples.

top

Degassing Models of N Isotopic Fractionation under Reduced Conditions

The normalised N isotopic composition of the glasses decreases from 0 ‰ in the undegassed glass to −41 ± 13 ‰ in the most degassed glass (Fig. 3). Although this δ¹⁵N variation is opposite to that observed during MORB degassing (Cartigny et al., 2001

Cartigny, P., Jendrzejewski, N., Pineau, F., Petit, E., 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 and Planetary Science Letters 194, 241–257.

), it is consistent with glasses becoming isotopically lighter with increased degassing as observed for H, C, and S in natural settings (Supplementary Information). An open system degassing model, in which the exsolved vapour is immediately removed from contact with the melt during degassing, is suitable with the presence of gas bubbles within the dissolved forsterite rims (Fig. 1):

where δ¹⁵N_initial is the “normalised initial” isotopic composition of the melt (here 0 ‰; sample V133-PB4N2Si), F is the fraction of N remaining in the melt, and α is the isotopic fractionation factor. In this case, open system degassing occurs in a finite N budget, conforming to a Rayleigh process (Equation 2).

Full size image

In bubbles for which the gas pressure was sufficient to obtain a Raman signal above the noise, N was observed only as N₂, even in very reduced samples in which N was dissolved as N-H complexes (V148-PB4N) and/or nitrides (V125-PB8N). Therefore, we propose that at IW to IW−1, the N degassing reaction is:

Had the least reduced samples (V152-PB0.5N and V149-PB1N) not been affected by diffusion through cracks, they would be expected to degas following Equation 3, since N₂ is the only N species dissolved in those melts. In contrast, under more reduced conditions, N degassed as (e.g., Dalou et al., 2019a

Dalou, C., Hirschmann, M.M., Jacobsen, S.D., Le Losq, C. (2019a) Raman spectroscopy study of C-O-H-N speciation in reduced basaltic glasses: Implications for reduced planetary mantles. Geochimica et Cosmochimica Acta 265, 32–47.

):

The fugacities of the relevant N species cannot be calculated at present due to a lack of thermodynamic data (see Supplementary Information).

To understand how N speciation can affect gas-melt N isotopic fractionation, we derived the basic equation for equilibrium isotopic fractionation for stable isotopes (Young et al., 2015

Young, E.D., Manning, C.E., Schauble, E.A., Shahar, A., Macris, C.A., Lazar, C., Jordan, M. (2015) High-temperature equilibrium isotope fractionation of non-traditional stable isotopes: Experiments, theory, and applications. Chemical Geology 395, 176–195.

):

where h is Planck’s constant (J.K⁻¹), k_b is the Boltzmann constant (J.Hz⁻¹), T is temperature (K), v^gas and v^melt denote the Raman vibration modes (Hz) of N species in the gas and silicate phases, respectively, and μ_N is the reduced mass of the N-bearing molecule. Using this expression, we calculated that the N isotopic fractionation factor at 1550 °C is 0.999 for degassing by Equation 3, and, under more reduced conditions, 1.005 for Equation 4, both in agreement with Hanschmann (1981)

Hanschmann, G. (1981) Quantum chemical calculations of isotope effects in nitrogen-containing molecules. ZFI-Mitteilungen 41, 19–39.

, and 0.998 for Equation 5. These equilibrium fractionation factors cannot explain the variation of δ¹⁵N_norm values observed in the degassed samples; the high isotopic fractionation undergone by these degassed melts rather suggests a kinetic process. Indeed, such large kinetic nitrogen isotopic effects have also been observed in mantle-derived samples (Yokochi et al., 2009

Yokochi, R., Marty, B., Chazot, G., Burnard, P. (2009) Nitrogen in peridotite xenoliths: Lithophile behavior and magmatic isotope fractionation. Geochimica et Cosmochimica Acta 73, 4843–4861.

). However, kinetic fractionation via evaporation results in increasingly heavy isotopic values (Saal and Hauri, 2021

Saal, A.E., Hauri, E.H. (2021) Large sulfur isotope fractionation in lunar volcanic glasses reveals the magmatic differentiation and degassing of the Moon. Science Advances 7, eabe4641.

), opposite to our observations (Fig. 3c). The observed large isotopic fractionation may thus be explained by a variable diffusion process depending on N speciation, and therefore on fO₂ conditions. The magnitude of diffusive N isotopic fractionation during degassing is similar to that observed in natural high pressure-high temperature peridotitic systems (>25 ‰; Yokochi et al., 2009

Yokochi, R., Marty, B., Chazot, G., Burnard, P. (2009) Nitrogen in peridotite xenoliths: Lithophile behavior and magmatic isotope fractionation. Geochimica et Cosmochimica Acta 73, 4843–4861.

; see Supplementary Information). Such a fractionation effect can be estimated from the ratio of the diffusion coefficients (D) of ¹⁴N and ¹⁵N, expressed as a function of their mass (m): , where β is an empirical parameter with values between 0 and 0.5 that positively correlates with the diffusivity of the ion in silicate melts (Van Orman and Krawczynski, 2015

Van Orman, J.A., Krawczynski, M.J. (2015) Theoretical constraints on the isotope effect for diffusion in minerals. Geochimica et Cosmochimica Acta 164, 365–381.

). Hence, N₂ and NH₃, which diffuse rapidly through silicate melts relative to Si, have a large isotopic effect (high β), whereas N³⁻, which is bonded to Si, diffuses slowly (Boulliung et al., 2021

Boulliung, J., Dalou, C., Tissandier, L., Füri, E., Marrocchi, Y. (2021) Nitrogen diffusion in silicate melts under reducing conditions. American Mineralogist 106, 662–666.

) and has a smaller isotopic effect (low β). Good fits to our data are β = 0.2 for N diffusing as N³⁻ and β = 0.5 for N as NH₃ (Fig. 3c). A simple diffusion process cannot explain the extremely large fractionation in undegassed samples (V152-PB0.5N and V149-PB1N), unless the diffusion of ¹⁴N and ¹⁵N is decoupled, with ¹⁵N favouring the stiffest bonds (N₂) and diffusing significantly faster than ¹⁴N, favouring weaker N-H bonds in NH₃. It is possible that N₂ diffusion in these melts is faster than degassing and controls the isotopic fractionation mechanism in these undegassed samples.

top

Evolution of the N Composition of Earth’s Primitive Atmosphere

Figure 3 shows that fO₂ conditions determine N solubility and, therefore, the fraction of N degassed and the isotopic fractionation trends. Our results suggest that the upper layer of Earth’s magma ocean degassed N as N₂, regardless of fO₂ conditions, but that the intensity of N₂ degassing increased as the magma ocean became less reduced. This is in agreement with the recent model of Sossi et al. (2020)

Sossi, P.A., Burnham, A.D., Badro, J., Lanzirotti, A., Newville, M., O’Neill, H.S.C. (2020) Redox state of Earth’s magma ocean and its Venus-like early atmosphere. Science Advances 6, eabd1387.

, which predicts that Earth’s prebiotic atmosphere was dominated by CO₂ and N₂ gases.

Whether N was dissolved in the magma ocean as nitride, ammonia, or both, N loss depleted the surface or upper parts of the magma ocean in ¹⁵N relative to the atmosphere. This could have produced the currently observed isotopic differences between the ¹⁵N-depleted mantle (δ¹⁵N = –5 ± 2 ‰ in diamonds and MORBs) and the ¹⁵N-enriched surface (δ¹⁵N = +6 ‰ in sediments and δ¹⁵N = 0 ‰, by definition, in the atmosphere) (Cartigny and Marty, 2013

Cartigny, P., Marty, B. (2013) Nitrogen isotopes and mantle geodynamics: The emergence of life and the atmosphere–crust–mantle connection. Elements 9, 359–366.

). Other explanations of the N isotopic distribution among terrestrial reservoirs include i) recycling (subduction) of surficial materials, and ii) core–mantle differentiation (Li et al., 2016

Li, Y.F., Marty, B., Shcheka, S., Zimmermann, L., Keppler, H. (2016) Nitrogen isotope fractionation during terrestrial core-mantle separation. Geochemical Perspectives Letters 2, 138–147.

; Dalou et al., 2019b

Dalou, C., Füri, E., Deligny, C., Piani, L., Caumon, M.C., Laumonier, M., Boulliung, J., Edén, M. (2019b) Redox control on nitrogen isotope fractionation during planetary core formation. Proceedings of the National Academy of Sciences 116, 14485–14494.

) However, Labidi et al. (2020)

Labidi, J., Barry, P.H., Bekaert, D.V., Broadley, M.W., Marty, B., Giunta, T., Warr, O., Sherwood Lollar, B., Fisher, T.B., Avice, G., Caracausi, A., Ballentine, C.J., Halldórsson, S.A., Stefánsson, A., Kurz, M.D., Kohl, I.E., Young, E.D. (2020) Hydrothermal ¹⁵N¹⁵N abundances constrain the origins of mantle nitrogen. Nature 580, 367–371.

recently showed that N recycling is inefficient. How core–mantle differentiation would have affected the upper layer of the magma ocean and thus the present day depleted mantle, would have depended on the depth of the magma ocean and the efficiency of post-magma ocean convection to homogenise the δ¹⁵N value of the mantle; thus this process remains disputable.

top

Acknowledgements

We thank Johan Villeneuve for his assistance and support during SIMS analyses, Jean-Luc Devidal for performing microprobe analyses on silicate glasses, and Marie-Camille Caumon for her help with Raman spectroscopy analyses. We are grateful to Robert Dennen for English editing. We thank reviewers Paolo Sossi and Sami Mikhail for their comments that greatly improved the manuscript, and editor Helen Williams for handling the manuscript. This work was supported by the European Research Council under the European Union’s Horizon 2020 research and innovation program (Grant Agreement no. 715028). This is CRPG contribution 2812.

Editor: Helen Williams

top

References

top

Supplementary Information

The Supplementary Information includes:

• 1. Methods
• 2. Supplementary Discussion
• Tables S-1 and S-2
• Figures S-1 to S-3
• Supplementary Information References

Download the Supplementary Information (PDF).