The geobiological nitrogen cycle: From microbes to the mantle

Abstract Nitrogen forms an integral part of the main building blocks of life, including DNA, RNA, and proteins. N2 is the dominant gas in Earth's atmosphere, and nitrogen is stored in all of Earth's geological reservoirs, including the crust, the mantle, and the core. As such, nitrogen geochemistry is fundamental to the evolution of planet Earth and the life it supports. Despite the importance of nitrogen in the Earth system, large gaps remain in our knowledge of how the surface and deep nitrogen cycles have evolved over geologic time. Here, we discuss the current understanding (or lack thereof) for how the unique interaction of biological innovation, geodynamics, and mantle petrology has acted to regulate Earth's nitrogen cycle over geologic timescales. In particular, we explore how temporal variations in the external (biosphere and atmosphere) and internal (crust and mantle) nitrogen cycles could have regulated atmospheric pN2. We consider three potential scenarios for the evolution of the geobiological nitrogen cycle over Earth's history: two in which atmospheric pN2 has changed unidirectionally (increased or decreased) over geologic time and one in which pN2 could have taken a dramatic deflection following the Great Oxidation Event. It is impossible to discriminate between these scenarios with the currently available models and datasets. However, we are optimistic that this problem can be solved, following a sustained, open‐minded, and multidisciplinary effort between surface and deep Earth communities.

surface. The result is that Earth's atmosphere became highly reactive, unlike the atmospheres of Mars and Venus which are still dominated by unreactive gases (CO 2 + N 2 ). The combination of a uniquely reactivegas-rich atmosphere and hydrosphere, coupled with subduction zone plate tectonics, means that Earth injects oxidizing material into a relatively reduced mantle (Frost & Mccammon, 2008;Kelley & Cottrell, 2009). This phenomenon is seemingly unique because such biological and tectonic processes are only known to occur on Earth. Here, we compare and contrast current ideas for how this unique interaction of biological innovation, geodynamics, and mantle petrology could have acted to regulate the geobiological nitrogen cycle over Earth history.

| Surficial nitrogen cycle
Molecular nitrogen in the gas-phase (N 2 ) is the largest surficial reservoir of nitrogen and comprises around 78% of Earth's atmosphere.
Atmospheric nitrogen is incorporated into the biosphere via the process of N 2 fixation, whereby specialized prokaryotes convert this inert N 2 gas into biomolecules (as C-NH 2 ) ( Figure 1). Nitrogen fixation is an energetically expensive process which requires 16 ATP to break the triply bonded N 2 molecule. N 2 -fixing organisms (termed "diazotrophs") accomplish this feat by utilizing the nitrogenase metallo-enzyme complex, which contains Fe and Mo in its most efficient form. Ammonium oxidation ("nitrification") generally requires molecular oxygen (Carlucci & Mcnally, 1969). These "oxidized" nitrogen compounds provide an important nutrient source for eukaryotic primary producers. Alternatively, nitrate (and some nitrite) can also be reduced back to ammonium via dissimilatory nitrate reduction to ammonium (DNRA), another process that was described fairly recently in the modern system (Lam et al., 2009).
Fixed nitrogen can be returned to the atmosphere as N 2 via several chemotrophic metabolisms, including denitrification (the reduction of NO 3 − to N 2 ) and anaerobic ammonium oxidation (anammox, the oxidation of NH 4 + to N 2 using NO 2 − ; Dalsgaard, Thamdrup, & Canfield, 2005). These processes, which result in the loss of fixed N from the biosphere, occur in soils, marine sediments, and oxygen minimum zones, completing the biological nitrogen cycle ( Figure 1).
All life requires N in a reduced form, which can either be taken up directly as ammonium or enzymatically reduced during nitrate assimilation. Once assimilated into biomass, organic nitrogen is recycled via release and re-assimilation of organic N within the surface ocean, or subsequently regenerated during organic matter remineralization in sediments. Over geologic timescales, some small amount of fixed nitrogen can leak out of the biosphere-ocean system and be buried in the sediments, as organic nitrogen or as remineralized ammonium incorporated into clays during diagenesis (Schroeder & McLain, 1998). During burial, nitrogen-bearing rocks can undergo metamorphism, which can return a significant fraction of the nitrogen back to the atmosphere (e.g., Haendel, Mühle, Nitzsche, Stiehl, & Wand, 1986). The remainder of N entrained within sediments can be subducted, along with nitrogen sequestered into altered oceanic lithosphere as ammonium and organic species (Busigny, Cartigny, & Philippot, 2011;Halama, Bebout, John, & Scambelluri, 2014). This subduction of sediment and altered oceanic lithosphere constitutes the primary flux of surficial N into the deep Earth.

| Deep earth nitrogen cycle
Once subducted into the deep Earth, nitrogen has historically been considered an unreactive gas in the field of mantle geochemistry, owing to the strength of the covalent triple N-bond, and abundance of N 2 in the atmosphere (e.g., Boyd & Pillinger, 1994;Javoy, Pineau, & Delorme, 1986). Consequently, the behavior of nitrogen has traditionally been assumed to be akin to the unreactive noble gases (e.g., Bergin, Blake, Ciesla, Hirschmann, & Li, 2015;Halliday, 2013;Marty, 1995Marty, , 2012Mikhail, Dobosi, Verchovsky, Kurat, & Jones, 2013 Bebout, Lazzeri, & Geiger, 2016;and Johnson & Goldblatt, 2015). All of the aforementioned species of nitrogen behave differently, where their behavior can be described as compatible or incompatible to a host mineral phase, depending on charge and radius of the ion and the occupation site in the mineral (Goldschmidt, 1924). For nitrogen in the mantle, recent experimentally and theoretically determined equilibrium constants have shown that ammonium should dominate over molecular nitrogen in aqueous fluids in the uppermost mantle where the redox state (fO 2 ) is buffered below the quartz-fayalite-magnetite buffer (QFM) (Li & Keppler, 2014;Mikhail & Sverjensky, 2014) ( Figure 2). This is a significant insight, because for nitrogen, compatibility is directly related to speciation: Molecular nitrogen (N 2 0 ) is highly incompatible, whereas ammonium (NH 4 + ) is likely compatible in potassium-bearing mineral phases (Watenphul, Wunder, & Heinrich, 2009) and can dissolve as a trace component in potassium-absent phases (Li, Wiedenbeck, Shcheka, & Keppler, 2013;Watenphul, Wunder, Wirth, & Heinrich, 2010). This is a crucial distinction from the perspective of the global nitrogen cycle, because nitrogen in the molecular form is likely to be outgassed, while ammonium could be stored in mineral phases in the deep Earth (Mikhail & Sverjensky, 2014). We explore this concept, and its potential response to changes in global redox, in the discussion below.
Because Earth is a dynamic planet with active plate tectonics, what goes into the mantle sometimes comes back out, and vice versa.
Therefore, the exchange of nitrogen between the surface and interior is governed by subduction (in-gassing) and volcanism (outgassing), and this interplay ultimately controls atmospheric N 2 levels (Cartigny & Ader, 2003;Dauphas & Marty, 2004;Kerrich & Jia, 2004;Marty & Dauphas, 2003a,b). However, the further back one looks in time the less data are available, and there is a predictable dearth of data to constrain either the fluxes of nitrogen over geological time, or the partial pressure of atmospheric nitrogen in the deep past. This means we must rely on sparse data and predictive thermodynamic models to estimate past N 2 dynamics.

| Evolution of the biogeochemical N cycle
The emergence and evolution of biogeochemical nitrogen cycling over Earth history remains a major question mark in the field of geobiology (e.g., see recent review by Stüeken, Kipp, Koehler, & Buick, 2016).
As the speciation and microbial cycling of nitrogen are highly redoxdependent, the biogeochemical nitrogen cycle would have changed significantly with the progressive oxygenation of Earth surface environments (e.g., Stüeken et al., 2016). Notably, in Earth's Archean oceans before the buildup of significant oxygen, processes involved in the modern N cycle that required molecular oxygen (or the presence of significant quantities of oxidized N compounds) would have been absent.
In particular, nitrification, the stepwise oxidation of ammonium to nitrite and nitrate, requires at least some free oxygen (Carlucci & Mcnally, 1969). Thus, throughout the Archean and earliest Paleoproterozoic the nitrogen cycle would likely have consisted of N 2 fixation followed by burial and regeneration of NH 4 + in sediments ( Figure 1).
Following widespread oxygenation of the marine biosphere during the Great Oxidation Event (GOE) between ~2.4 and 2.3 Ga (e.g., reviewed in Farquhar, Zerkle, & Bekker, 2014), the primary N loss pathways of denitrification and anammox proliferated to take advantage of the newly available oxidized N compounds (Zerkle et al., 2017 Thomazo, Ader, & Philippot, 2011). These studies suggest that the aerobic N pathways could have evolved immediately following oxygenic photosynthesis, to take advantage of the newly available free oxygen in highly productive coastal environments.
However, the isotope effect associated with these N loss pathways would only have been expressed once nitrite/nitrate was available at sufficient levels that their quantitative removal from the ocean did not occur. Therefore, the shift toward more positive δ 15 N values by contribution to primary productivity in the redox-stratified continental margins that that seemingly dominated for much of the Proterozoic (e.g., Poulton et al., 2010;Scott et al., 2011). Primary producers in marginal sulfidic zones could have had direct access to upwelling nutrients, including ammonium, before it reached the redox interface and was lost to nitrification-denitrification and anammox reactions.
Consumption of upwelling ammonium by anoxygenic phototrophs could have further contributed to the fixed nitrogen deficit in surface waters, such that N 2 -fixing organisms in the photic zone had a competitive advantage. Many researchers have suggested that N 2 fixation would also have been limited in the Proterozoic oceans, due to drawdown of bioactive trace metals required for nitrogenase (particularly Mo) under widespread sulfidic conditions (Anbar & Knoll, 2002 (Reinhard, Raiswell, Scott, Anbar, & Lyons, 2009;Scott et al., 2008). However, records of molybdenum in shales suggest that marine molybdenum levels rose to near modern levels at the end of the Neoproterozoic with enhanced oxidative weathering during a second rise in atmospheric O 2 (Sahoo et al., 2012), potentially further enhancing global N 2 fixation rates.
As the delivery of nitrogen to the sediments is intimately We used denitrification levels estimated from modern marine N balances (e.g., Canfield et al., 2010;Gruber, 2008). The biggest unknown in this calculation is when coupled nitrification/denitrification proliferated and how rates of denitrification responded to changes in ocean chemistry throughout the Proterozoic (as dis-

| The missing nitrogen conundrum
Our understanding of the evolution of the deep Earth nitrogen cycle and the relevant nitrogen reservoirs is even less well constrained.
One puzzle often cited is the "missing nitrogen conundrum," which follows because the abundance of nitrogen in the bulk silicate Earth (BSE) appears to be significantly lower than that of other volatile elements ( Figure 4). The most striking feature of these data is the depletions shown for Xe and N, which could mean that these two elements with contrasting behaviors and masses were both lost early in Earth's history (Bergin et al., 2015) without also depleting Ne, Ar, and Kr-elements that are far more akin to Xe in terms of geochemical behavior, and in the case of Kr, mass. Another explanation is that there exists a hidden nitrogen reservoir in the deep Earth, inferred to be the core or unspecified deep mantle domain (Barry & Hilton, 2016;Halliday, 2013;Marty, 2012). This reservoir is unlikely to be the core because carbon is more siderophile than nitrogen (Dalou, Hirschmann, von  than the present-day atmosphere (Li et al., 2013), and preliminary data suggest the transition zone could store even more (Yoshioka, Wiedenbeck, Shcheka, & Keppler, 2016). Therefore, if saturated (which is unlikely), then the upper mantle alone can enable the BSE nitrogen abundance to match the C/N ratio of carbonaceous chondrites (shown on Figure 4).
These datasets and thermodynamic models strongly imply the existence of a deep mantle-based nitrogen reservoir. Thus, contrary to previous assumptions, it is now clear that the mantle, and not the atmosphere, could form the largest nitrogen reservoir on Earth (Goldblatt et al., 2009;Mikhail & Howell, 2016;Palya, Buick, & Bebout, 2011). Various authors have attempted to provide calculated/assumed reservoir mass fractions for nitrogen in the core, silicate Earth, and the atmosphere (Bebout et al., 2013;Dalou et al. 2016;Goldblatt et al., 2009;Johnson & Goldblatt, 2015;Marty, Zimmermann, Pujol, Burgess, & Philippot, 2013;Palya et al., 2011); however, these estimates all vary significantly, and there is currently no consensus on the relative distributions.

| Variations in atmospheric pN 2 over geologic time
Direct data for the partial pressure of atmospheric N 2 over time are limited, but the majority of available data support lower pN 2 in the Archean compared to the 0.79 bar of N 2 in the modern atmosphere.
For example, Marty et al. (2013)  Modeling studies of greenhouse warming on the early Earth have proposed elevated pN 2 in the Archean as a solution to the Faint Young Sun paradox (Airapetian, Glocer, Gronoff, Hebrard, & Danchi, 2016;Goldblatt et al., 2009); however, there is limited evidence to support the suggestion that the Archean had a higher atmospheric pN 2 than the present day. Fossilized raindrop imprints from ~2.7 Ga allow for air density of less than 2 bar, but suggest it was likely less than 1.1 bar (Som, Catling, Harnmeijer, Polivka, & Buick, 2012; but see Kavanagh & Goldblatt, 2015, for a different interpretation).

| The current state of confusion (aka, Schrodinger's nitrogen cycle)
Given the uncertainties described above, a complete description of the geobiological nitrogen cycle will require solutions for the following outstanding questions: (i) a consensus for the fluxes of nitrogen into and out of the mantle, and how these relate to the surficial nitrogen cycle, and (ii) the validity of the proposed missing nitrogen conundrum. We suggest that there are three plausible scenarios to reconcile these issues within the geobiological nitrogen cycle over Earth's 4.6-billion-year history, notably that atmospheric pN 2 has changed unidirectionally (either increased or decreased) over geological time Given the dearth of relevant data, it is impossible to rule out multidirectional changes in pN 2 over geologic time ( Figure 5, conceptual model 3), but there is currently no theoretical support or potential mechanism for such variations. Conceptual model 2 is exciting and is mechanistically feasible because of the disparate compatibilities of oxidized and reduced nitrogen, as described above. However, for conceptual model 2 to be true, then the nitrogen cycle has experienced unidirectional change in both directions, with a major turning point at the GOE. In the following, we explore the geobiological implications of these three scenarios.

| Scenario #1: Atmospheric pN 2 has decreased over geologic time
If atmospheric pN 2 was relatively higher in the Archean, this could have resulted in a greater supply of abiotically fixed nitrogen from the atmosphere than previously estimated (Navarro-Gonzalez et al., 2001). A higher abiotic flux of nitrogen into the biosphere on the early Earth could have allowed for a much later evolution of N 2 fixation, supporting some phylogenetic reconstructions and molecular clock data (Boyd, Anbar, et al., 2011;Sanchez-Baracaldo, Ridgwell, & Raven, 2014). Notably, the scarcity of data available for how these processes, and the dynamic nitrogen cycle, have interacted over geologic time means it is impossible to determine whether the global nitrogen cycle is currently in equilibrium. Therefore, if a progressive drawdown of N 2 remained unbalanced by mantle degassing (e.g., Busigny et al., 2011), over geologic timescales this geobiological nitrogen pump could be progressively depleting Earth's atmosphere, with currently unexplored consequences for the climate and biogeochemical evolution of a future Earth system.

| Scenario #2: Atmospheric pN 2 has increased over geologic time
On the other hand, a relatively lower pN 2 for the Archean atmosphere is in line with an interplanetary model proposed to explain an enrichment of nitrogen over the primordial noble gases in Earth's atmosphere relative to the atmospheres on the other sampled planets that are devoid of subduction zone plate tectonics (e.g., Mars and Venus; Mikhail & Sverjensky, 2014). At present, subduction zones inject oxidizing material into the mantle wedge above the downgoing slab; at these redox conditions >QFM, thermodynamic calculations predict oxidation of NH 4 to N 2 (Figure 2). Because this N 2 is neutrally charged and highly volatile, it is therefore degassed to the atmosphere. This model predicts Earth's arc systems should degas more nitrogen than mid-ocean ridge and hot spot volcanism, consistent with observations (e.g., Hilton, Fischer, & Marty, 2002). In addition, the dependence of this N 2 production on subduction can explain Earth's higher This scenario would suggest that an increase in atmospheric pN 2 over geological time was mirrored by an increase in the burial of nitrogen in sediments ( Figure 3). A positive coupling implies that the biosphere could have directly responded to changes in pN 2 caused by progressive degassing of the mantle. As the atmospheric N 2 reservoir is orders of magnitude larger than the biosphere, the role of pN 2 in regulating primary productivity over geologic timescales is generally not considered, assuming an infinite supply for biological N 2 fixation.
However, it is unclear how abiotic atmospheric sources of fixed nitrogen (particularly important for early life) could have scaled with pN 2 (Kasting & Siefert, 2001;Navarro-Gonzalez et al., 2001); if pN 2 was very low in the Archean, this could mean even higher levels of nitrogen stress and an instantaneous requirement for the evolution of N 2 fixation alongside the early biosphere (Stüeken et al., 2015;Weiss et al., 2016). Furthermore, if N 2 outgassing increased after the GOE (as discussed below), it could mean that abiotic sources became enhanced in the early Proterozoic, potentially alleviating some of the biological nitrogen stress proposed for the boring billion.

| Scenario #3: Atmospheric pN 2 underwent a major turning point during the GOE
Either in line with or independent of Scenario #2 above, the Great Oxidation Event could have driven a relative increase in pN 2 by altering the speciation and partitioning of nitrogen in subducting sediments. This follows because the speciation and partitioning of nitrogen are redox-sensitive, as discussed in section 2.2. Therefore, the relative abundance of nitrogen degassed out of the mantle over time must have changed before and after the GOE, because the oxidizing potential of the material subducted (oceanic sediments and altered oceanic crust) must have differed. In particular, post-GOE the oxidation state of the mantle overlying the slabs would become too oxidizing to stabilize ammonic nitrogen, and molecular nitrogen would be favored ( Figure 2). As inert N 2 molecules are highly volatile and highly incompatible in all mineral phases, this molecular N 2 would be released as a gas-phase to the atmosphere, a concept not quantitatively incorporated into recent flux models (Barry & Hilton, 2016;Busigny et al., 2011;Fischer et al., 2002). Therefore, we propose (on a global scale) it is plausible for the flux of nitrogen to have dramatically shifted across the GOE, whereby ingassing dominated pre-GOE and outgassing dominated post-GOE. Thus, following the GOE, subduction zones began excessively degassing nitrogen to the atmosphere, which over time resulted in Earth's higher N 2 / 36 Ar and N 2 / 20 Ne ratios relative to the Martian and Venusian atmospheres (Mikhail & Sverjensky, 2014).
A second, as yet unexplored, link between the GOE and atmospheric pN 2 could stem from oxidative weathering. Johnson and Goldblatt (2015) estimated that continental crust could constitute a significant reservoir of nitrogen. This nitrogen is hosted both in sedimentary rocks (as organic matter or NH 4 + ) and in crystalline rocks (predominantly as NH 4 + ). There are currently no quantitative estimates we are aware of for the contribution of oxidative weathering to atmospheric N 2 in the modern Earth system; however, it potentially constitutes an additional N 2 source that would have been absent or nearly so prior to global oxygenation.

| CONCLUSIONS
In summary, subduction zones inject biologically mediated material into the mantle, which is affected by Earth surface redox. This subducted material affects mantle melting and degassing, and can significantly alter Earth's atmospheric chemistry over geologic time.
Atmospheric chemistry has been both a driving force, and a response to, biological innovation and evolution. Therefore, the geological and biological nitrogen cycles are intimately linked, albeit with long lag times possibly verging on hundreds to thousands of millions of years.
In the above discussion, we have highlighted some of the looming questions in nitrogen geobiology, and the problems and challenges involved in addressing them. What is painfully obvious from this discussion is that a cross-disciplinary approach will be crucial in tackling these issues. This requirement goes beyond the typical marriage of low-temperature geochemistry, microbiology, and sedimentology, toward a deeper connection between biogeochemistry, mantle petrology, and geodynamics. For example, additional thermodynamic and experimental constraints on N speciation under relevant deep Earth conditions are warranted, but these data are only relevant if the input (sourced primarily from biology) is considered. The geobiological nitrogen puzzle will only be solved by considering the biosphere as intimately connected with the geosphere, and vice versa. We therefore put this challenge to the Geobiology community, to don your hard hats and delve into Earth's interior (accompanied by your hard-rock colleagues).

AZ acknowledges Natural Environmental Research Council Fellowship
NE/H016805 for financial support. The authors would also like to thank M. Claire and E. Stüeken for valuable discussions and P. Barry for sharing his modeling data.