Main

Oxygen (O2) is prevalent in deep-sea surface sediments where its rate of consumption reflects the sum of aerobic respiration and oxidation of reduced inorganic compounds produced by anaerobic decay. These processes define sediment community O2 consumption (SCOC), and quantifying SCOC is needed to estimate fluxes of major elemental cycles through marine systems1,2,3. We undertook multiple in situ benthic chamber lander experiments to measure abyssal SCOC in the Nauru Ocean Resources Inc. (NORI)-D licence area of the Clarion–Clipperton Zone (CCZ; Extended Data Fig. 1 and Extended Data Table 1) where polymetallic nodules cover extensive areas of seafloor. Sediments and nodules were exposed to different experimental treatments, which included the addition of dead-algal biomass, dissolved inorganic carbon and ammonium (NH4+) or cold filtered surface seawater. No-injection controls were also performed. In contrast to previous deep-sea O2 flux studies that only showed SCOC, we consistently found that more O2 was accumulating in the chambers than was being consumed, resulting in net O2 production.

Constant linear decreases in O2 optode readings were observed in two experiments (Fig. 1), and SCOC determined by in situ O2 microprofiling was 0.7 mmol O2 m−2 d−1 indicating that SCOC occurs in NORI-D as in many abyssal habitats2,3,4. However, O2 concentrations in 25 benthic chamber incubations started at 185.2 ± 2.9 µmol l−1 (1 standard error (SE)) and reached O2 maxima between 201 and 819 µmol l−1 over 47 h (Fig. 1), indicating net dark O2 production (DOP) corresponding to rates of 1.7–18 mmol O2 m−2 d−1. Independent measurements of O2 concentration using the Winkler method also showed DOP (Extended Data Fig. 2), providing evidence that the optodes were not malfunctioning. No statistically significant difference in the total net O2 produced (maximum [O2] – initial [O2]; Extended Data Table 2) was found between chambers (ANOVA, F2,9 = 0.107, p = 0.900) or experimental treatments (ANOVA, F3,9 = 0.876, p = 0.489), ruling out any experimental bias. We found no difference in the total net O2 produced between cruises (ANOVA, F2,12 = 0.391, p = 0.684), though DOP was correlated to the average surface area of the nodules (Spearman’s correlation,  = 0.664, p = 0.031). A re-evaluation of in situ O2 optode data collected from 36-h benthic chamber experiments in the abyssal eastern and western CCZ (Extended Data Figs. 1 and 3) also showed DOP, indicating its occurrence in multiple locations across the CCZ. Our findings contrast with all published deep-sea benthic O2 flux studies and suggest that DOP may provide O2 for benthic respiration. Whereas the DOP measured was greater than SCOC, we would urge caution when temporally upscaling our results, as the nonlinear production of O2 suggests that DOP may not be continuous in nature. Moreover, the variance in DOP activity seen between experiments and its relationship to nodule surface area suggests DOP activity may change with nodule spatial density and type (for example, diagenetic versus hydrogenetic), so upscaling our results by area is also imprudent without additional studies.

Fig. 1: Oxygen concentrations in μmol l−1 measured by calibrated O2 optodes through time in h in the different benthic chamber incubations.
figure 1

ac, The in situ benthic chamber lander deployments were made during the 5D (a), 5E (b) and 7A (c) cruises to the NORI-D license area (Extended Data Fig. 1). Nodules were present in all incubation experiments. The green hue, blue hue and red lines in the 5D figure (a) denote dead-algal biomass, dissolved inorganic carbon + NH4+ and filtered seawater treatments, respectively. The gap in the optode data in AKS279-Ch.3 was caused by the optode periodically not logging data. The black line indicates ambient O2 concentration measured on the outside of the benthic chambers during AKS273 on the 5D cruise. The green and yellow hue lines in the 5E (b) and 7A (c) figures denote the dead-algal biomass and control (no injection) treatments, respectively. The minor drops seen in some of the O2 concentration profiles at 28, 38 and 47 h are caused by the dilution of the chamber water with 50 ml of seawater that was entrained from the outside into the chamber through a 1.5 m (0.25 cm diameter) open tube when the syringe sampler collected seawater samples from within the chamber. The constant O2 concentration measured during the first 2 h of the 5D and 7A experiments was due to the stirrers being turned off for 1 h to allow the substrates (for example, dead-algal biomass) to sink to the sediment surface. Stirrers were turned on during the 5E expedition from the moment the lander was deployed until the lander returned and power to the stirrers was disconnected.

Source data

Several lines of evidence indicate that the DOP was not caused by experimental artefacts. First, the total O2 change between the experimental and control (non-injection) treatments was statistically indistinguishable, and a steady increase in O2 concentration was recorded over many hours in multiple experiments; these observations demonstrate that DOP was not attributable to the injection of exogenous fluids. Second, diffusion of O2 from trapped air bubbles within the chamber was unlikely because each chamber uses two one-way valves in the lid to purge air from the chambers as the lander sinks. Even if an air bubble could be trapped long enough to reach the seafloor, gaseous diffusion of O2 into the water phase would take < 1 s at 4,000 m depth (Extended Data Table 3), which is inconsistent with the steady increase in O2 over many hours seen in multiple experiments (Fig. 1). Third, intrusion of O2 from the plastic chambers into the water phase is unlikely (Methods) as they are built from polyoxymethylene, which is both highly inert and chemically stable in well-oxygenated settings and would not explain the variation in DOP because all experiments used identical materials. Last, DOP was also observed during 48-h ex situ sediment incubations (Extended Data Fig. 4).

Several lines of enquiry were pursued to explain the DOP. Subsurface advection of oxic bottom water from seamount flanks into seafloor sediments5,6 and then into the chambers was discounted based on in situ O2 microprofiling that showed pore water was a net sink for O2 and undersaturated compared with the O2 seen in the chambers. Furthermore, DOP was measured in sealed ex situ experiments (Extended Data Fig. 4) that prevented O2 intrusion from below. It is unlikely that biological mechanisms were responsible for the bulk of the DOP as ex situ core incubations revealed DOP in the presence of poison (HgCl2; Extended Data Fig. 4). Whereas many microbes in the CCZ are able to detoxify Hg (II) to Hg (0)7, and some microhabitat pore spaces in the core may have remained HgCl2 free, the taxa known to be capable of DOP (for example, Nitrosopumilus maritimus) are killed by its addition8. We also observed weak statistical support between the relative abundance of certain nitrogen-cycling microbial taxa and DOP (for example, Candidatus Nitrosopumilus  = 0.474, p = 0.420). The fact that DOP was detected in ex situ controls containing only polymetallic nodules (Extended Data Fig. 4) suggested that the DOP was linked to their presence. Hence, we estimated the potential contribution of radiolytic O2 production using a kinetic model9 and found 0.18 μmol l−1 O2 would be generated by this process within 48 h. We also modelled the chemical reduction of manganese (IV) oxide at in situ temperature (1.6 °C) across a range of pH and O2 conditions encountered at the seafloor to assess if this reaction (2MnO2 → 2MnO + O2;Extended Data Fig. 5) could liberate the O2 but found that <0.1 nmol of manganese (IV) oxide would be chemically reduced to manganese (II) at seafloor conditions. As such, localized radiolytic O2 production from the sediments and nodules and chemical dissolution explain only a negligible proportion (< 0.5%) of the DOP observed.

The oxygen evolution reaction requires an input voltage of 1.23 V plus an overpotential of approximately 0.37 V to split seawater into H2 and O2 (ref. 10) at NORI-D’s seafloor mean pH (7.41). This value can be lowered by several hundred millivolts if the reaction proceeds via the lattice-oxygen-mediated mechanism11. Use of metal catalysts such as Mn oxides enriched with transition metals (for example, Ni) found in nodules12 and characterized by large tunnel areas and abundant defect sites can optimize the adsorption of reactants and enhance conductivity and catalytic performance11,13,14. We tested the electrical potential between two platinum electrodes at 153 sites on the surfaces of 12 nodules (Fig. 2) from the UK1, NORI-D and Bundesanstalt für Geowissenschaften und Rohstoffe (BGR) license areas. Although the potentials between different positions on the nodules were highly variable, - potentials up to 0.95 V were found and high mean background-corrected potentials were detected under cold-water conditions (Fig. 2 and Extended Data Table 4). On the basis of these studies and DOP being observed in nodule-only ex situ incubations (Extended Data Fig. 4), we hypothesize that the DOP may have partly resulted from seawater electrolysis, with the necessary energy coming from the potential difference between metal ions within the nodule layers, leading to an internal redistribution of electrons. Whereas questions remain concerning this potential mechanism (such as the identity of the energy source(s), longevity of DOP, catalytic stabilities, electrochemical conditions on exposed versus buried nodules surfaces and the influence of different chemistries within the nodule layers), the ‘geo-battery’ hypothesis was supported by the link between DOP and nodule average surface area. This connection could be due to an increased abundance of anode and cathode sites or a greater abundance of high Ni and Cu dendritic porous layers in larger nodules15. Assuming the ‘geo-battery’ is partly responsible for the DOP observed, the initial high DOP rate may have been related to the ‘bow-wave’ of the lander removing sediments from the surface of the nodules and exposing electrochemically active sites on the nodules. The slowdown in DOP seen later in the incubations could have then been caused by a reduction in voltage potential and/or degradation of metal-oxide catalysts that has been observed in Mn oxide catalysts previously10. Whereas this process requires further investigation, if true, DOP activity may fluctuate with sediment coverage on the nodules inviting the urgent question of how sediment remobilization and distribution over large areas during deep-sea mining may influence DOP.

Fig. 2: Box and whisker plots of background-corrected voltage potentials on nodule surfaces.
figure 2

The nodules were collected from the NORI-D (1-5), UK1 (6-8) and the BGR (9-12) license areas. Potentials were measured at 21 °C (nodules 1–12) and 5 °C (nodules 6 and 7 cold) and between two different UK1 nodules (Tests 1 and 2) and across the surface of a metamorphosed carbonate rock (control). Means are designated by the ‘x’ symbol, medians by the line, boxes show the lower and upper quartile values (excluding the median), whereas the whisker bars refer to the minimum and maximum data values. The number of technical replicate measurements made at different points on the surface of each nodule/rock to make each box-whisker is shown by the number above each whisker bar.

Source data

Understanding the mechanism(s) behind DOP, its temporal nature and its spatial distribution will allow its role in abyssal ocean ecosystems to be better understood. Future studies of DOP in the deep sea may also shed light on broader relationships between metal-oxide deposition, biological evolution and the oxygenation of Earth16,17.

Methods

A benthic chamber lander was deployed in the NORI-D license area six times in May–June 2021 (5D cruise), five times in November–December 2021 (5E cruise) and five times in August–September 2022 (7A cruise) (Extended Data Fig. 1 and Extended Data Table 1). The lander comprised three independent, autonomous, square benthic chambers (484 cm2) separated by approximately <0.5 m. After arriving at the seafloor, the lander waited for 0.07–1.34 d before the chambers were pushed into the sediment to create an enclosed microcosm of the seafloor. Ten minutes into the incubation period, the enclosed chambers were injected with 50 ml of one of three solutions: (1) 0.45-µm-filtered, cold surface seawater containing 79.2 mg of freeze-dried Phaeodactylum tricornutum algae, (2) 32 µM Na2HCO3 and 40 µM NH4Cl dissolved in cold artificial seawater (salinity 35) and (3) 0.45-µm-filtered, cold surface seawater. On some occasions, the injection mechanism failed allowing the response to control (no injection) conditions to be measured. The seafloor in the study area had a temperature of 1.6 °C ± 0.006 °C (SE, n = 28) and a pH of 7.41 ± 0.05 (SE, n = 17). Immediately after the injection, the overlying water was mixed with a submersible stirrer at 60 rpm for 1 min before the stirrer was turned off that allowed any particulate substrates to settle for 1 h. After 1 h, the stirrer was then turned on again for the remainder of the experiments. During the 5E expedition, the stirrers were programmed to continually stir the overlying water even immediately after injection.

The syringe samplers removed approximately 50 ml of seawater from the water phase of each chamber at 0.1 or 0.03, 1, 3, 9, 28, 38 and 47 h into the incubation experiment. Oxygen optodes (CONTROS HydroFlash O2 manufactured by Kongsberg Maritime Contros GmbH) mounted in the lid of each chamber logged O2 concentrations in the chamber every 10 seconds throughout each experiment. Two days before the first lander deployment of each cruise, the optodes underwent a two-point, multi-temperature calibration using 0 and 100% O2 calibration solutions at 1.2, 7, 18 and 30 °C following the recommendations of Bittig et al. (ref. 18). On the 5D cruise, we also calibrated the sensors 2 d after the last lander experiment so we could estimate optode drift, which was negligible (0.27 µmol l−1 d1) over the course of the six-week cruise. The 0% and 100% O2 saturation solutions were created by bubbling 0.45-µm-filtered surface seawater in a bottle sitting in a water-chilling/heating unit with N2 gas (0%) or an aquarium air bubbling unit (100%) for 30 min. The O2 concentration of the calibration solutions was confirmed in triplicate by Winkler titration. After incubating seafloor sediments for 47 h, the lander chambers were closed by a shutter door at the base of the chambers, and the chambers were then pulled slowly out of the sediment, which took 1 h. The lander was then recalled from the seafloor. In eight instances, the lander programme did not finish and the doors did not shut, preventing the sampling of sediment and determination of the volume of the water phase in the chambers (Extended Data Table 2). Once the lander was back and secured on deck, the chambers were opened and the water above the sediment removed via syphoning into a bucket. The distance from the top of the sediment to the base of the chamber lid was then measured in four places to get an accurate water depth for water volume estimates. Whenever possible, a photograph was then taken of the chamber sediment and nodules from directly above the opening of the chamber. All syringes containing water samples were removed and taken to the shipboard lab for immediate processing or stored in a cold lab (4 °C) before processing. The optodes were removed and their onboard data downloaded to a computer. Finally, the nodules were removed from the chambers and washed of attached organic debris with cold (4 °C), 0.45-µm-filtered surface seawater and placed in sterile Whirlpak bags to be weighed in the laboratory later. The number of polymetallic nodules at the seafloor determined from chamber counts was 1170 ± 97 m−2.

Unfiltered syringe sample seawater was carefully transferred from each 50 ml syringe to a 12 ml exetainer via a 10 cm tube attached to the syringe nozzle, ensuring no air bubbles were introduced and immediately fixed for microWinkler titration. The sample was then mixed thoroughly using a glass bead placed in the exetainer and placed in the dark in a 4 °C refrigerator for 30–45 min to allow the precipitate to settle. Once the precipitate had sedimented, the exetainers were shaken again and left for 2–3 h before Winkler titrations were performed. All titrations were completed within 12 h after sampling to determine dissolved O2 concentrations. Each Winkler sample (approximately 5 ml) was titrated twice, and duplicate measurements showed minor differences in O2 concentration (5D cruise error: 3.5 ± 0.3 μmol l−1, n = 71; 5E cruise error: 1.3 ± 0.2 μmol l1, n = 69; 7A cruise error: 2.8 ± 0.4 μmol l1, n = 84). Winkler O2 concentration data were averaged for each syringe sample. The O2 concentrations estimated by Winkler analysis were 22 ± 1% (n = 42, SE, 5D cruise), 8 ± 4% (n = 39, SE, 5E cruise) and 24 ± 2% (n = 40, SE, 7A cruise) lower than the concentrations measured by the optodes at the same time point in the same incubations most likely due to out gassing of supersaturated O2 caused by depressurization and warming of the externally mounted syringes (whose samples were used for Winkler analyses) during the lander recovery to the surface.

Back on shore, the final O2 concentration values were calculated following Bittig et al. (ref. 18) from the optode, calibration and in situ pressure data that was derived from the depth where each lander deployment was made. Time stamps in the optode data were compared to the lander computer programme times so the optode readings could be aligned to the schedule of the chamber experiment. The total change in O2 concentration in each chamber was then calculated from the volume of the water phase above the sediment and the difference in O2 concentration from when the chambers started to seal off the sediment to the point when the maximum O2 concentration was reached.

Benthic O2 microprofiling

Benthic O2 microprofiles were made during lander deployments AKS313, AKS316, AKS318 and AKS321 during the 5E cruise using a UNISENSE deep-sea microprofiling unit mounted <0.5 m from the benthic chambers. The microprofiles were made using 20 cm O2 microsensors that penetrated the sediment in 0.05 mm steps. The microsensors were calibrated 2 h before the lander deployments at in situ temperature (1.6 °C) at 0% and 100% O2 saturation (above). At each sampling depth, the microsensor stopped for 5 s before each measurement was made. The sensor then recorded five individual O2 concentration measurements. The average of these five measurements was taken for each depth point. The sediment surface was determined manually based on the turning point in the slope of O2 concentration with depth where O2 started to become depleted. SCOC was determined from Fick’s first law of diffusion.

Microbiology sampling

Nodule and sediment samples for microbial community analyses were collected from the 5D experimental chambers. Approximately 30 g of sediment from each of the 0–2 cm and 2–5 cm horizons and 50 g of intact nodules were placed in separate sterile Whirlpak bags with a pre-sterilized spatula and then transferred to a −80 °C freezer. DNA from approximately 10 g of nodules and 250 mg of sediment were extracted using the Qiagen PowerMax soil and PowerSoil extraction kits, respectively. Extracted DNA was then shipped on dry ice to Laragen Inc. and sequenced using a proprietary in-house method. The V4 region of 16 S rRNA genes were amplified using the Earth Microbiome Project protocol19 with the 515 F (5′‐GTGYCAGCMGCCGCGGTAA20) and 806 R (5′-GGACTACNVGGGTWTCTAAT21) primers. Raw fastq files were processed using a custom pipeline (https://github.com/Boston-University-Microbiome-Initiative/BU16s) built with QIIME 2020.2 (https://www.nature.com/articles/s41587-019-0209-9). Adaptor sequences were removed using cutadapt (https://doi.org/10.14806/ej.17.1.200), read truncation positions were determined by mineer (more below), amplicon sequence variants (ASVs) were generated using dada2 (trunc-len-r20) (https://doi.org/10.1038/nmeth.3869) and ASVs were clustered to 99% identity with the SILVA 132 database (https://academic.oup.com/nar/article/42/D1/D643/1061236) using the vsearch cluster-features-closed-reference (https://doi.org/10.7717/peerj.2584). Due to drops in sequencing quality, all reverse reads were truncated by 49 bases (from a length of 301 to 252) as determined by minERR, an algorithm for determining optimal sequence length based on sequence quality scores (https://github.com/michaelsilverstein/mineer). Family- and genus-level abundance was computed by summing the relative abundance of all ASVs with the same family/genus classification within each sample. Spearman correlations were then computed between family- and genus-level abundance and observed optode-derived total O2 changes. Sequences have been archived at National Centre for Biotechnology Information GenBank under the Bioproject ID PRJNA1117483.

Polymetallic nodule surface area measurements

Photographs of the surface sediment and nodules in the chambers were imported into Image J. The outline of each nodule in each chamber photograph was then traced and the surface area of the nodule automatically calculated in Image J (assuming each surface nodule was flat in shape) and logged as an Image J file before being exported and saved as an Excel file.

Radiolysis O2 production estimates

To estimate the potential radiolytic O2 production, published concentrations of 238U, 235U, 232Th, 40K (refs. 22,23,24,25,26) in seawater were used (Supplementary Table 1). For nodules, 238U, 235U and 232Th isotopes of three nodules from chamber experiments from the 5D cruise were measured by Multicollector-Inductively Coupled Plasma Mass Spectrometer using previously described methods27,28,29 and averaged; 40K values were derived from the literature12. Nodule and seawater contributions were calculated using a kinetic model developed by ref. 9 that incorporates 32 reactions (equation (1) in ref. 30). The nodule boundary layer was assumed to be fully integrated with the seawater, surpassing the respective ~23 to ~452 μm stopping power distance of alpha and beta particles used to model geologic materials31. Sediment radiolytic O2 was calculated as half of the previously quantified H2 production rates in equatorial Pacific subsurface sediment32, given the stoichiometry of water’s radiolytic decomposition (an equivalency that probably offers an overestimate of derived O2). Contributions from these three components (nodules, sediment and seawater) were scaled by the benthic chamber’s size and contents to produce an estimate of 0.18 μmol l−1 of O2 generated over 48 h according to the following expression.

$${({{\rm{O}}}_{2})}_{{{t}}}=({{{Q}}}_{{\mathrm{iz}}}\times {{{E}}}_{{\rm{a}}}\times{{G}}({{\rm{O}}}_{2})\times {{{M}}}_{{\rm{O}}2}\times{{{{A}}}_{{\mathrm{iz}}}}^{-1}\times {10}^{-2})\times (1-{{\rm{e}}}^{-\lambda {{t}}})$$

Here (O2)t is the mass (kg) of O2 produced over a given time t (yr), Qiz is the mass (g) of the isotope, Ea is the average energy (eV) released from the decay of one atom; G(O2) is the radiation chemical yield of molecules per 100 eV of the radiation energy; MO2 is the O2 molecular mass (g), Aiz is the isotope atomic mass (g) and λ is the isotope-specific decay constant (y−1). The overall (O2)t value summed the contributions from 238U, 235U, 232Th and 40K across water, nodule and sediment sources.

Electrochemistry measurements

Voltage potentials were measured using a Keithley DMM6500 digital multimeter on nodules previously collected by coring in the UK1, NORI-D and BGR license areas. Nodules were initially immersed for seven days in Instant Ocean artificial seawater (salinity 35). To measure the potentials, two electrodes (platinum wire, 99.9% purity) were first washed in perchloric acid, rinsed in Milli-Q water and dried before being attached to alligator clamps attached to the multimeter. The platinum wires were then immersed in Instant Ocean artificial seawater in a glass petri dish to measure background voltages (0.003 ± 0.001 V, SE, n = 17) until stable. Once stable, a nodule was placed in the petri dish and the platinum probes placed on the nodule at random locations, ensuring contact in one of two ways. We either carefully drilled a hole into some nodules so one platinum wire could be fixed inside it while the second platinum wire was firmly pressed against the nodule surface using a clamp. Alternatively, the platinum wires were pressed firmly against two different spots on the nodule surface and held in place using a clamp. Voltages were then recorded for 1–2 min until the signal was stable. This procedure was repeated up to 20 times in different randomly selected regions of the nodules depending on their size. Measurements were undertaken on 12 nodules at 21 °C (n = 153) and a single control rock composed of metamorphosed carbonate (n = 10). Two nodules from UK1 were also retested after being cooled to 5 °C (n = 18) by placing them in Instant Ocean water in a refrigerator overnight. Voltage potentials (n = 20) between two nodules were measured using four nodules collected from UK1. Potentials measured during each measurement were averaged and corrected for the background seawater voltage measured using only Instant Ocean seawater in the absence of a nodule. Measured resistances inside some of the nodules that were broken up were in the kΩ to 100s of kΩ range, though it is unclear if these resistivities change at the nano- or microscale requiring further investigation.

Geochemistry modelling

The chemical stability and solubility of manganese (IV) oxide (birnessite) to dissolved Mn2+ as a function of pH and O2 activity was modelled using the Geochemist Workbench Professional (version 12) software, with the in-built and internally consistent THERMO database. The conditions used for generating the phase diagram (Extended Data Fig. 5) represent bottom seawater as measured in the eastern CCZ with a temperature of 1.6 °C and chlorine and manganese concentrations of 0.55 M Cl and 2e−10 M Mn, respectively.

Ex situ core incubations

Opportunistic ex situ experiments were undertaken during the 5D cruise using sediment cores retrieved by a multi-corer from the CTA area (Extended Data Fig. 1). Immediately after the multi-corer arrived back at the surface, cores were removed and transferred to a cold lab held at in situ temperature. The cores were then exposed to the following five treatments (administered using a 60 ml syringe), which included (1) Na2HCO3 (0.3 μM final concentration, n = 3), (2) NH4Cl (10 μM final concentration, n = 3) and (3) NH4Cl (50 μM final concentration, n = 3), (4) 0.3 μM Na2HCO3 + 10 μM NH4Cl (final concentration, n = 3) and (5) HgCl2 (1.1 μM final concentration, n = 3). No-injection controls (n = 3) were also performed and separate core experiments in which four nodules were incubated for 48 h by themselves with no additions. After addition, the water phase of each core was stirred and a 50-ml sample of top water was taken for microWinkler analysis (as above). Stoppers were then placed on the top of the cores, ensuring no air bubbles were present. The stoppers were secured tightly and the cores fully submerged in a large bucket containing 0.45-µm-filtered, cold, surface seawater (salinity 35). The bucket was covered with five black plastic bags and secured in the cold room with the lights turned off. After 48 h, the cores were removed from the bucket, and the cores were inspected for the presence of air bubbles. Only one core, a HgCl2 treatment, had a gas bubble beneath the bung, which was rejected from further analysis, leaving n = 2 for this treatment. The other cores were then re-sampled for dissolved O2 and analysed as before. Core-specific water volume measurements were used together with the change in O2 concentration to calculate the total net O2 change per core.

To determine if our ex situ DOP detection was affected by intrusion of O2 from the atmosphere into the core tube, two controls were performed: a shipboard test with an O2 microprofiler and a lab-based test using the Winkler method. Shipboard, a clean core tube was filled with Milli-Q water and sparged with N2 for 10 min before beginning the test. A Metrohm 8663 Multimeter was inserted through a predrilled hole in the rubber stopper, allowing for O2 concentration to be recorded every 5 s. An increase from 39 to 69 µmol l−1 was observed over ~5 h, corresponding to a rate of 0.14 mmol m−2 d−1 or 4% of the 3.5 mmol m−2 d−1 mean net DOP measured in the ex situ experiments. Back in the home laboratory, three of the original core tubes were filled with 4 °C, 0.2-µm-filtered artificial seawater (salinity 35) and sparged with N2 for 8 min through a filtered pipette tip to achieve an initial dissolved O2 concentration of ~100 µmol l−1 (for example, the approximate starting O2 concentrations for the shipboard experiments). The tubes were sealed with rubber stoppers and electrical tape, being careful to avoid bubble formation. They were then submerged in a 32-gallon plastic garbage can of unfiltered seawater (O2 concentration: 228.12 µmol l−1) in a dark cold room (8 °C) for 48 h. After 48 h, the tubes were quickly unsealed and analysed one at a time to prevent additional O2 dissolution from the air. A 50-ml sterile syringe was used to slowly collect 10 ml of seawater from the centre of the core tube, being sure to avoid bubble entrainment into the syringe. The sample was carefully expelled into a 10-ml reaction vial and fixed using the adjusted values for a 10-ml sample according to a volume-scaled Winkler titration protocol33 and the reagents from the LaMotte Dissolved Oxygen Test Kit. The fixation of each collected sample was done in less than 2 min in a fume hood. Dissolved O2 increased by 0.11 mmol m−2 d−1 during the 48 h, which corresponds to between 3.2% of the mean net DOP rate observed in the ex situ experiments (3.5 mmol m−2 d−1). Both of our control experiments provide high confidence that the diffusion of external O2 into the core tubes did not cause the O2 production measured in the ex situ core incubations.

Calculations to quantify intrusion of O2 from the polyoxymethylene chambers and lids

Oxygen intrusion was estimated from Stephens34 who calculated that 20.66 µmol l−1 of O2 could diffuse out of 428 cm2 of polyoxymethylene plastic when immersed for 48 h in hypoxic water (O2 diffusion rate: 0.02 µmol O2 cm−2 d−1). To determine the total area of plastic that would be available for diffusion (869–1,584 cm2), we added the surface area of the lid to the surface area of the four walls that would be exposed at the seafloor (based on the depth of the water phase—above). The minimum and maximum areas available for diffusion were multiplied by 0.02 µmol O2 cm−2 d−1 to estimate that 41.9–76.5 µmol O2 l−1 would diffuse out of the polyoxymethylene chamber walls and lid in 48 h under hypoxic conditions. Thus, we are highly confident that O2 leakage from the plastic chambers could not replicate the high O2 concentration seen in some of our oxygenated experiments (Fig. 1).