Generally, the gravity cores reflected pore water zonations expected from C_org diagenesis (Froelich et al., 1979), though due to the low C_org content, the redox zones have expanded over several meters (Figure 2). Based on the appearance of dissolved Mn⁺² in the pore waters (Froelich et al., 1979), we concluded that the maximum depth of oxygen penetration was between 10 and 30 cm, similar to previous observations (Schulz et al., 1994). As Chong et al. (2014) discussed, this is consistent with the depth at which denitrification begins. All NO3- profiles reached a maximum below the SWI and then NO3- was consumed and depleted at depth. A secondary NO3- peak (Figure 4C) was a consistent feature in 6 of the 7 cores analyzed and is discussed below in the context of the NH4+ gradient.

As in many other sediment pore water studies, for a given diagenetic reaction, the diffusive flux of a reactant may be used to predict the formation rate of a product, or fluxes may be assessed in terms of possible reaction stoichiometries. In the interest of investigating flux stoichiometries and also to assess rates of product formation, we calculated fluxes of Mn⁺², and Fe⁺² using Fick's first law following Hammond et al. (1996): (1)J=-ϕnDswdCdz where ϕ is the average porosity from the gravity core horizon of interest, D_sw is the molecular diffusivity in seawater, corrected for temperature (cm² s⁻¹), and dC/dz is the concentration gradient. dC/dz was determined by fitting a line through 3–6 points within the depth range of the selected concentration gradient. We used a value n = 2.6 following previous work to correct for tortuosity (McManus et al., 1995; Hammond et al., 1996). Molecular diffusivities at 2°C were calculated following Li and Gregory (1974) and Schulz et al. (1994) and the values we used in these calculations were: Mn⁺² = 3.20 × 10⁻⁶ cm² s⁻¹, Fe⁺² = 3.20 × 10⁻⁶ cm² s⁻¹.

The flux of Mn⁺² at the position of the arrow (Figures 2B,F,J,N,R,V,Z and Figure 4E) ranged from 1.2 to 7.5 μmol m⁻² d⁻¹ with the highest flux at station 29 and the lowest at station 25 (Table 3). There does not appear to be a geographic pattern in these fluxes, which suggests that Mn diagenesis is not simply a function of the location of the river plume nor proximity to the Amazon Fan. If the dissolved Mn⁺² sink is reaction with oxygen, the strength of the flux toward this boundary is likely controlled by the intensity of Mn oxide (MnOx) reduction in the sediments just below this horizon. The amount and lability of C_org would factor into both oxygen disappearance and MnOx remineralization and we'd expect steeper dissolved Mn⁺² gradients and larger fluxes where the oxygen disappearance depth was shallow. Figure 5 demonstrates that this hypothesis is supported, stations with a higher flux generally have a shallow Mn⁺² appearance depth. Of course this outcome is also a function of MnOx availability between sites and with depth. Chong et al. (2014) defined geographic zones of river plume influence as “central” (south of 10°N and west of 50°W), “north” (north of 10°N) and “east” (east of 50°W). Generally, stations with a shallow Mn²⁺ appearance occur in the central zone, where the river plume is present year-round and the sediments receive a higher influx of organic matter (since Chong et al., 2014 determined that river plume location correlated positively with C_org export).

Most of the pore water Mn⁺² profiles reached a near constant concentration at depth, and hence a near zero flux. The value of this constant concentration ranged between 0 and 110 μM, with station 29 having the highest value. Schulz et al. (1994) studied a core from the base of the Amazon Fan (5.14°N 46.57°W, 3511 m depth), where dissolved Mn⁺² gradually decreased with depth to near zero values at 700 cm depth. The Mn⁺² profile at station 25 shows some similarity to this previous work, with concentrations dropping to ~30 μM at 300 cm depth, but the other stations do not. Since our cores are shorter than Schulz et al. it is possible that the Mn⁺² profiles do reach zero at greater depths, but it is also possible that there is spatial variability in pore water diagenesis throughout the region and/or that the Mn sink in sediments proximal to the Amazon Fan is not present in Abyssal plain sediments. This variability could stem from changes in sedimentation on the Amazon Fan through time, as sedimentation rates have been shown to fluctuate significantly between glacial and interglacial periods (Damuth, 1977; Schulz et al., 1994). It is possible that the sediments in the Demerara Abyssal Plain are more carbon limited at depth. For the sites in this study where Mn⁺² reaches a semi-constant concentration, another possibility is that Mn⁺² concentration is controlled by the mineral solubility and oxidation state of Mn oxides as discussed below.

Most of the NO3- profiles (except station 9) had a secondary NO3- peak (4–13 μM) (Figure 4C), some defined by one point, others defined by multiple points. We presume that all NO3- produced from aerobic nitrification is consumed at shallower depths (Figure 4A), thus, the production of this additional NO3- is curious. It is possible that this feature is an artifact stemming from the oxidation of NH4+ during sampling under aerobic conditions, but the fact that this secondary NO3- appears over a limited depth range and at a specific depth, associated with other solutes, across all gravity cores makes the artifact interpretation less likely. It is also possible that this is an artifact of rhizon sampling water derived from the overlying water, trapped and perhaps channeled down the inside of the gravity core tube. We rule out this artifact because dissolved Si does not show an interruption in its profile. Bioirrigation is also ruled out for the same reason.

Previous observations of deep subsurface NO3- in deep ocean sediments (Aller et al., 1998; Anschutz et al., 2000; Mortimer et al., 2004) suggest that this feature (Figure 4C) could be created by anoxic NH4+ nitrification coupled to MnOx reduction (Table 4 Equations IIA, IIB) (Aller et al., 1998; Hulth et al., 1999). (Thamdrup and Dalsgaard, 2000) found that anoxic nitrification was insignificant at a site in Skagerrak, however, Anschutz et al. (2005) demonstrated that the Mn oxides in Skagerrak basin consist of Mn(IV) oxides (Canfield et al., 1993) that are less reactive to ammonia than Mn(III) phases. Their findings show that MnOx reduction by NH4+ is possible, though anoxic nitrification may not be ubiquitous in all anoxic sediments. More recently, a study on cores from the Clarion-Clipperton Fracture Zone in the Eastern Tropical North Pacific suggests that ammonium generated during organic matter degradation may act as a reducing agent for manganese oxides in C_org starved sediments (Mogollón et al., 2016).

If anoxic NH4+ nitrification is responsible for producing the secondary NO3- peak, then the total flux (production and consumption) of NO3- at this feature should be equal to the change in the flux of NH4+. To explore the plausibility of this reaction, we constructed an N mass balance for stations 13 and 31, chosen because they most distinctly show a correlation between the appearance and disappearance of NO3- and the change in the slope of the concentration gradient of NH4+. Flux uncertainties ranging from 2 to 36% were determined (assuming most of the flux uncertainty could be attributed to uncertainty in gradient fitting) using least squares (LINEST in Microsoft Excel).

At station 13, the ΔNH4+ was 3.3 ± 0.7 μmol m⁻² d⁻¹ and the ΣNO3- was 6.8 ± 1.3 μmol m⁻² d⁻¹. At station 31, ΔNH4+ was 1.6 ± 0.6 μmol m⁻² d⁻¹ and the ΣNO3- was 3.5 ± 1.2 μmol m⁻² d⁻¹. Based on the predicted reaction stoichiometry in Table 4, anoxic nitrification would yield a ratio of 1:1 between NH4+ and NO3-. Although our calculation of nitrate production flux was about 2X the ammonium consumption flux, within our uncertainties, the estimated fluxes show that this reaction is plausible as the reason for the secondary NO3- peak. These findings support the notion that MnOx reduction by NH4+ may be occurring in these sediments resulting in the production of NO3-. In some cores (13, 25, 27), it appears possible that a reaction involving ammonium + nitrate (anammox) could be responsible for NO3- uptake. Organic carbon-based denitrification should also consume NO3-.

The secondary NO3- peak is consumed both above and below the depth of its appearance (Figures 4B,D, respectively). In feature B, the secondary NO3- diffuses upwards and is likely consumed by denitrification (Table 4 Equation III), though it is also possible that NO3- is involved in the oxidation of NH4+ or in the anoxic reoxidation of Mn⁺² at some sites (Table 4 Equations IV and VI, respectively). The lower boundary of the secondary NO3- max (feature D) could be generated by diffusion and consumption by reaction with Fe⁺² (Table 4 Equation V). The depth of feature C is nearly coincident with the depth where feature I (Fe⁺²) appears (Table 5).

The fluxes of dissolved Fe⁺² at the position of the arrow in Figure 2 (and feature I in Figure 4) ranged from 0.06 to 3.7 μmol m⁻² d⁻¹. According to Table 4, Equation 5, we'd anticipate a flux of Fe⁺² that was 5 times greater than the consumption flux of NO3-. It does not appear to be large enough to account for NO3- consumption rates presented above. Although Fe oxidation with NO3- may be occurring at this horizon, insofar as there is sufficient NO3-sink to account for the Fe⁺² gradient, there is likely a second sink necessary to account for the all the downward NO3- flux.

With the exception of station 9, the appearance of dissolved Fe⁺² was fairly consistent throughout the region, 45–65 cm. Further downcore, Fe⁺² reached a near constant concentration of ~100 μM at stations 3, 13, 29, and 31. The profiles from stations 25 and 27 showed more variation, indicating that there may be more iron recycling downcore at these sites. At station 9, very little dissolved Fe⁺² was found at depth, perhaps removed by the production of Fe-sulfides as has been observed on the Amazon Fan (Schulz et al., 1994). It was noted that sediments in this core were very dark colored, however, we do not have sulfide data from these sites to test this hypothesis. We discuss the possible relationship between Fe⁺² uptake and silica in the following section.

Stations 9, 13, and 29 all had peaks (between 16 and 31 μM) in dissolved Fe⁺² that appear very shallow in the sediment column (Figures 4G,H), 20–40 cm above the major gradient of dissolved Fe⁺² (feature I) and within the zone of Mn reduction (Table 5). These primary peaks in Fe⁺² imply local production and consumption. Generally, Fe reduction is not expected to take place at the same depth as Mn reduction due to the lower energy yield of the reaction (Froelich et al., 1979). However, the free energy calculations used to generate the expected energetic yield from anaerobic diagenetic reactions generally assume that the metal oxides are well-characterized, crystalline minerals (Burdige, 1993). It could be possible for localized Fe reduction to occur at concurrent depths with Mn reduction if the Fe oxides were poorly crystallized, amorphous solids. The shallow peaks of dissolved Fe⁺² are also coincident with the depth at which the primary NO3- peak goes to zero (Figure 4A, Table 5). In feature G, the newly produced Fe⁺² is consumed as it diffuses upwards, perhaps oxidized by the NO3- (Table 4 Equation V), although this cannot be tested given the poor data coverage. In feature H, just below the primary Fe peak, it is unclear what reaction is responsible for Fe consumption, as by this depth, all NO3- has been consumed. Postma (1985) proposed a reaction between Fe⁺² and Mn-oxides (Table 4 Equation VIII) as a potential Fe sink.

It has been generally found that iron reduction occurs at a depth coinciding with the disappearance of NO3- from pore waters (Lyle, 1983; Burdige, 1993). However, here, only stations 9 and 31 displayed this type of pattern. At the other stations, dissolved Fe⁺² appeared 30–65 cm below the depth of “major” NO3- disappearance (Table 5). We also examined the depths of appearance of the major gradients of Mn⁺² and Fe⁺² and found that while stations 9 and 31 had very little separation between the appearances of the two metals (4 and 14 cm, respectively), the other stations showed a separation ranging from 40 to 70 cm. One possible cause of this large spacing between the diagenetic zones is that the C_org content of the sediments is quite low. At station 13, we measured the C_org wt% of the upper 120 cm of the gravity core, and found that concentrations decreased with depth to 70 cm, after which concentrations increased (Figure 3). This minimum in C_org correlates to the appearance of Fe⁺² in the pore waters (feature I). Others have shown that this is a ubiquitous phenomenon in the sediments of the equatorial Atlantic (Funk et al., 2003). However, it is also possible that sulfides produced below the NO3- zone sequester any dissolved Fe²⁺, and so the depth where Fe²⁺ increases indicates the depth where sulfate reduction slows. In this case, Fe²⁺ would only increase if it were being produced faster than the rate of sulfide production.

While all of the profiles of dissolved silica increased with depth, they did not reach a constant value, as seen in many other deep sea profiles (Martin et al., 1991; Zabel et al., 1998; Shibamoto and Harada, 2010). The dissolution of biogenic opal in sediments typically generates a pore water asymptotic value within 5–30 cm below the SWI that can range from 100 to 850 μM (Fanning and Pilson, 1974; Archer et al., 1993; McManus et al., 1995; Sayles et al., 1996; Rabouille et al., 1997; Dixit et al., 2001), though these values are well below the solubility of fresh plankton (Lawson et al., 1978; Van Cappellen and Qiu, 1997). At sites studied here, they reach maxima of 100–250 μM. One possible explanation is that the there is a kinetic “competition” between the dissolution of bSi and the formation of alumino-silicate minerals (Mackenzie and Garrels, 1966; Ristvet, 1978; Mackenzie et al., 1981; McManus et al., 1995). Clay formation produced by reverse weathering has been documented experimentally (Dixit et al., 2001), and has also been identified as an important reaction in the sediments of the Brazilian-French Guianan continental shelf underlying the Amazon Plume (Michalopoulos and Aller, 1995, 2004; Michalopoulos et al., 2000). Thus, it is quite possible that reverse weathering also plays an important role in opal preservation in the deep-sea sediments underlying the Amazon plume. The rapid dissolution of bSi in the surface sediments could be curtailed by the production of dissolved Al(III) from the detrital minerals raining from Amazon plume waters. Al(III) has been shown to reduce silica solubility and induce precipitation of authigenic alumino-silicates (Dixit et al., 2001). The asymptotic value at depth could represent the new balance between the kinetics of fresh opal dissolution and reverse weathering processes. The co-occurrence of the dissolved Si minimum and the steep Fe⁺² gradient may imply Fe uptake is involved as a sink for dissolved Si.

It is also possible that shelf-derived silica makes it to the deep sea. (DeMaster et al., 1996) noted that Amazon shelf waters exhibit high rates of silica production (10 mmol m⁻² d⁻¹); diatoms on the Amazon shelf may sink and be carried landward by the estuarine subsurface flow or seaward via suspension events. It is also possible that turbidity currents carry sediments off the Amazon shelf through the channel and levee system of the Amazon Fan (Damuth et al., 1988) and this process may supply additional terrigenous material to the abyssal plain. Al(III) supplied by this mechanism could contribute to dissolved Si uptake.