Warm and oxidizing slabs limit ingassing efficiency of nitrogen to the mantle

Highlights

• Biotite-rhyolitic melt-hydrous fluid partitioning of N is reported for slab dehydration and melting conditions.

• High pressure and reducing conditions limit N extraction from slabs during dehydration.

• Oxidized slabs may be crucial in maintaining an N-rich atmosphere over geologic time.

• Slab thermal and redox state may control N/⁴⁰Ar ratio of MORB mantle.

Abstract

Nitrogen is a major and essential component of Earth's atmosphere, yet relative to other volatile elements, there are relatively few experimental constraints on the pathways by which nitrogen cycles between Earth's interior and exterior. We report mineral-melt and mineral-fluid partitioning experiments to constrain the behavior of nitrogen during slab dehydration and sediment melting processes. Experiments reacted rhyolitic melts with silicate and oxide minerals, in the presence of excess aqueous fluid, over temperatures between 725-925 °C and pressures between 0.2 and 2.3 GPa. Oxygen fugacity ranged between iron metal saturation (∼NNO-5) to that in excess of primitive arc basalts (∼NNO+2). Our experiments demonstrate that hydrous fluid is the preferred phase for nitrogen over minerals (biotite, K-feldspar, and amphibole) and rhyolitic melts across all conditions explored. Relatively large effects of pressure (Δlog($D_{melt-fluid}^N$)/Δ(GPa/K) = 761 ± 68 (1σ), Δlog($D_{biotite-fluid}^N$)/Δ(GPa/K) = 462 ± 169) and moderate effects of oxygen fugacity ($\mathrm{\Delta }log(D_{melt-fluid}^N)/\mathrm{\Delta }$NNO = -0.20 ± 0.04, Δlog$(D_{biotite-fluid}^N)/\mathrm{\Delta }$NNO = -0.10 ± 0.04) modulate partitioning of nitrogen. We further document negligible partitioning effects related to mineral composition or Cl content of hydrous fluid. Of the minerals investigated, biotite has the largest affinity for N and should control the retention of N in slabs where present. Application of partitioning data to slab dehydration PT paths highlights the potential for highly incompatible behavior ($D_{biotite-fluid}^N$ < 0.1) from the slab along warmer and oxidized (NNO+1) subduction geotherms, whereas dehydration along reduced and cooler geotherms will extract moderate amounts of nitrogen ($D_{biotite-fluid}^N$ > 0.1). We find that slab melting is less effective at extracting N from slabs than fluid loss, at least under oxidized conditions (NNO+1). Ultimately, the conditions under which slabs lose fluid strongly affect the distribution of nitrogen between Earth's interior and exterior.

Introduction

Reactive, highly volatile elements (e.g. H, C, N) are essential for life and impact climate. Among these elements, N is unique in its strong partitioning into near-surface reservoirs (Fig. 1a, Marty, 2012), defined here as the sum of N contained in the crust, oceans, and atmosphere relative to the mantle that sources mid-ocean ridge basalts (MORB). Indeed, accumulation of N into near-surface reservoirs relative to MORB mantle is ∼10× and ∼5× greater than C and H (Marty, 2012; Halliday, 2013), respectively. Sufficient N is input into subduction zones, primarily residing in slab sediments (e.g., Halama et al., 2014), such that, without a return flux, the entire atmosphere would be drawn into the mantle at current plate rates over 4.5 Gyrs (Johnson and Goldblatt, 2015). Current observations suggest, however, that the partial pressure of nitrogen in the atmosphere (p_N2) may have actually increased from ∼3 Ga to the modern (Marty et al., 2013; Som et al., 2016). A thickening, or even slowly thinning, atmosphere is consistent with inefficient return of N to the deep Earth. Efficient subduction should lead to geologically rapid drawdown of atmospheric N.

Inefficient return of N to the deep Earth likely reflects the stability of N₂, and potentially other neutral species, at or near Earth's surface. Nitrogen, when speciated as N_2, behaves as a noble gas similar to Ar (e.g., Libourel et al., 2003), hindering the regassing of atmospheric N into the mantle via subduction. An important observation, however, is that the partitioning of Ar (non-radiogenic) into near-surface reservoirs is nearly 50× greater than that observed for N (Fig. 1a). Thus, N occupies a chemical middle-ground between the more reactive volatile elements, C and H, and the more inert noble gases. The exact position of N within this range of geochemical behavior likely reflects the competing influences of neutral and other, more chemically reactive, species of N (Fig. 1a).

A clue regarding the identity of the reactive N species comes from the strong correlation between N and ⁴⁰Ar in rocks derived from MORB mantle (Fig. 1b) (Marty, 1995; Johnson and Goldblatt, 2015). ⁴⁰Ar is exclusively produced by the decay of K, and the correlation between N and ⁴⁰Ar is consequently explained by the long-term coupling of N and K in mantle environments. We note that there is a mantle component, most commonly and strongly expressed in ocean island basalts (OIB), with very high N/⁴⁰Ar ratios (>10,000) compared to the MORB mantle (124±40) (Marty and Zimmermann, 1999; Johnson and Goldblatt, 2015). Here we focus on the relatively well-established distribution of N between Earth's near surface reservoirs and MORB mantle.

The distribution of K in MORB mantle is affected by recycling of surficial materials back into the mantle, and thus, the coupling of N and K in MORB mantle suggests these two elements are recycled together. Nitrogen, when speciated as the ammonium ion (NH${}_4^{+}$), is expected to couple with K (Marty and Dauphas, 2003). NH${}_4^{+}$ is monovalent and large radius, like K. Indeed, laboratory experiments have confirmed complete solid solution between K- and NH₄-endmember feldspar and muscovite (Pöter et al., 2004).

Additional evidence for the link between NH${}_4^{+}$ and K-bearing minerals comes from rocks exhumed from subduction zones. Bulk rock K₂O and N are correlated within a wide range of metamorphic terranes (Bebout and Fogel, 1992; Busigny et al., 2003; Halama et al., 2016). Metasedimentary rocks can contain N concentrations upwards of 1000 ppm, with the large majority identified as NH${}_4^{+}$ within mica (Bebout and Fogel, 1992; Sadofsky and Bebout, 2000; Busigny et al., 2003; Plessen et al., 2010; Johnson and Goldblatt, 2015). MORB mantle N contents are estimated to be ∼0.25 ppm (Marty and Dauphas, 2003), for comparison. In addition, N contents of exhumed rocks decrease with increasing metamorphic grade, and fractional loss of N is typically larger than loss of K (Bebout and Fogel, 1992; Palya et al., 2011; Plessen et al., 2010). If N were perfectly coupled to K during prograde metamorphism, then the loss of N and K should be matched. Thus, the preferential loss of N compared to K suggests NH${}_4^{+}$ is not the only species controlling the behavior of N during subduction. Mixed N species stability is also implied by N isotopic variations within prograde metamophric rocks (Bebout and Fogel, 1992; Mingram and Bräuer, 2001). The likelihood that N exists as mixed species in subduction environments aligns with the degree of N partitioning into near-surface reservoirs, where it occupies the middle-ground between volatile elements that are stoichiometric components of common minerals (C and H) in the slab and noble gases (Fig. 1a).

If N is a mixed species element within subduction environments, then its geochemical behavior will be strongly dependent on the thermodynamic parameters present within slabs. Reaction (1) describes the stability of NH${}_4^{+}$ relative to N₂:

$$2N_2+6H_{2}O+4H^{+}=4N{H}_4^{+}+3O_2$$

And the relative stabilities of NH${}_4^{+}$, NH_3, and NH${}_2^{-}$ (amide anion) follow Reactions (2) and (3):

$$N{H}_3+H^{+}=N{H}_4^{+}$$

$$N{H}_3=N{H}_2^{-}+H^{+}$$

The stability of NH${}_4^{+}$ relative to other species (e.g., N_2, NH₃, NH${}_2^{-}$) depends on pressure (P) and temperature (T), as well as bulk composition, oxygen fugacity (ƒO₂) and pH (we refer to these combined compositional parameters as X), following Reactions (1)-(3) (Mikhail and Sverjensky, 2014; Li and Keppler, 2014; Chen et al., 2019; Mikhail et al., 2017; Mallik et al., 2018). Each species will also have its own, unique partitioning behavior between minerals, melts, and fluids that also depends on P-T-X conditions. Partitioning is defined here as the wt.% ratio of nitrogen in either minerals or melts over fluid (e.g., $D_{mineral-fluid}^N$ = mineral wt.% N / fluid wt.% N), when the concentrations reflect an equilibrium distribution. Nitrogen that is partitioned into minerals is retained in the slab, while nitrogen partitioned into fluid and melt can be lost to the mantle wedge and ultimately returned to the atmosphere. A portion of the N fluxed into the mantle wedge may be retained in secondary serpentinites (e.g., Cannaò et al., 2020).

Neutral species, akin to Ar, are expected to partition into fluids over melts and minerals despite some capacity of slab minerals to incorporate these species (∼1-100 ppb/bar of neutral species fugacity; Jackson et al., 2015; Krantz et al., 2019), while experiments show that NH${}_4^{+}$ partitions nearly evenly between muscovite, K-feldpsar and fluid (Pöter et al., 2004). Thus, the capacity of a slab to retain N during mass loss is likely a complex function of both the stability of NH${}_4^{+}$ (Reactions (1)-(3)) and the partitioning of NH${}_4^{+}$ between K-bearing minerals, melt, and fluid.

Using this framework, we seek to experimentally quantify how efficiently N is lost from slabs undergoing dehydration and melting by measuring the reactivity of N with silicate, both minerals and melts. The majority of N in slabs is initially reduced by biologic activity within the oceans and is therefore concentrated within sediments and, to a lesser degree, the underlying uppermost oceanic crust (Li et al., 2007; Halama et al., 2014; Busigny et al., 2019). Available evidence suggests the reactivity of N with rock is controlled by the exchange of NH${}_4^{+}$ into minerals with stoichiometric K. Micas are major K-bearing phases for metasediments and metabasalt, with muscovite being present at lower P/T conditions, biotite having a wide stability field up to 2.5 GPa and temperatures associated with slab melting, and phengite being stable beyond 2.5 GPa (e.g., Johnson and Plank, 2000; Schmidt et al., 2004). Feldspars and amphibole can also have stoichiometric K and are stable under certain slab P-T-X conditions (e.g., Johnson and Plank, 2000). Our experiments were consequently designed to produce relatively large grains of mica (biotite), feldspar (Na- and K-rich), and amphibole (hornblende) in equilibrium with hydrous fluid and rhyolitic melt to study the behavior of N during dehydration and melting processes up to 2.3 GPa and 925 °C. Our work quantifying N reactivity with minerals builds from previous efforts where nitrogen behavior in slab systems was inferred from N speciation in fluids and melt-fluid partitioning (e.g., Li and Keppler, 2014; Li et al., 2015; Mallik et al., 2018).