The net balance of soil carbon (C) accrual vs. loss is central to future climate predictions. Accumulating research has demonstrated that geochemical factors, such as secondary clay minerals and short-range-ordered (SRO) iron (Fe), and aluminum (Al) phases, in particular, are vital determinants of C accrual1,2,3. Mineral-associated organic matter (MAOM) is thought to persist because organic matter (OM) can form strong chemical bonds to minerals and can be physically protected in microaggregates or co-precipitates4,5. Once the initial association of OM with minerals has occurred, soil structural conditions (aggregate formation, macro-scale shifts in fluid flowpaths, etc.) can further isolate and compartmentalize OM from decomposer organisms and restrict the diffusion of oxygen (O2), thus further protecting soil organic matter (SOM) against decomposition6,7. These features can lead to longer turnover times for MAOM than for particulate organic matter8,9, and may explain MAOM residence times of centuries–millennia4,5,10. A large portion of MAOM in soils and sediments is adsorbed or co-precipitated with Fe minerals11,12,13. However, soil Fe plays multiple roles in ecosystem biogeochemistry aside from C protection, some of which also drive C loss.

Soil Fe serves three categorical roles in ecosystem function (Fig. 1 and Supplementary Fig. 1): the first is a structural role, where Fe (as FeIII) forms connective cements that bind minerals and SOM together in nano-, micro-, and macro-aggregates7,14; the second is a sorbent role, whereby nutrients and OM adsorb or co-precipitate with FeIII minerals or FeIII surface coatings5; and the third is an electron-transfer role, whereby FeIII accepts electrons from microbes or electron shuttles, or FeII donates electrons to various oxidants, such as O2, NO3, or H2O215. The relative impacts of these Fe functional roles on soil C cycling remain unclear.

Fig. 1: Schematic of Fe-mediated C transformation.
figure 1

Under oxic conditions, FeIII phases sorb C (a). Under anoxic conditions, FeIII can be reduced to FeII coupled with C mineralization, releasing dissolved organic matter (DOM) (b). Fe2+ oxidation yields reactive O2 species driving CO2 and DOM production (c). d Conceptual diagram of contrasting positive (blue arrows) and negative (red arrows) changes in CO2 production linked to a given mechanism.

The sorbent and structural roles of Fe may increase soil C stocks by decreasing the availability of OM to extracellular enzymes and heterotrophic microbes5,7. A commonly accepted mechanism for MAOM formation is for dissolved organic matter (DOM) of plant or microbial origin16 to sorb or co-precipitate with existing and de novo minerals5,17,18,19. One particularly important route of MAOM formation involves the oxidation of FeII to FeIII at redox interfaces and its rapid hydrolysis to SRO FeIII (oxyhydr)oxides, which co-precipitate with DOM20. This can occur wherever FeII-bearing anoxic solutions come in contact with O2, such as in periodically flooded soil horizons or across redox gradients within aggregates in upland soils20,21,22. High rates of Fe reduction have been observed in surface soils during periods of elevated moisture and high biological activity, leading to a heterogeneous distribution of iron within soil profiles23,24,25,26. Iron reduction appears to be a ubiquitous soil biogeochemical process across a broad range of terrestrial ecosystems23,24,25,26,27,28,29,30. Across these ecosystems, C:Fe molar ratios of Fe–C associations point to the dominance of co-precipitation vs. adsorption11,12. These lines of evidence place the epicenter of Fe-associated OM formation at these dynamic anoxic-oxic interfaces in surface soils.

However, the biogeochemical factors linked to Fe-associated C formation could also contribute to its decomposition. Fe electron transfer reactions can drive C solubilization, depolymerization, and loss as CO2. During anoxic periods, microbial use of FeIII as an electron acceptor directly produces CO2 from the metabolic coupling of OM oxidation to Fe reduction27,28,29, but also releases OM from FeIII–OM coprecipitates and OM occluded in FeIII-cemented micro-aggregates30,31,32. In soils that experience frequent redox fluctuations, microbial Fe reduction can account for up to 44% of anaerobic OC mineralization33. Therefore, significant portions of C protected by complexation under oxic conditions (up to 40% of total soil C11,13) can potentially be released and decomposed following Fe reduction. Conversely, the abiotic oxidation of FeII by O2 can also produce CO2. This is a consequence of reactive oxygen species production (Fenton chemistry), which can directly produce CO2 or cleave organic polymers to increase OM availability34,35,36.

Despite evidence for Fe-stimulated decomposition, the common perception of iron’s role in SOM has largely focused on Fe-mediated OM protection via adsorption, co-precipitation, or aggregation5,7,12,19,20,37,38,39,40. While it is also recognized that Fe–OM associations are formed during Fe redox cycling, and that Fe oxidation and reduction can promote C release and mineralization31,32,33,36, these processes are rarely explored concurrently. In fact, few studies have directly measured the microbial availability of Fe-associated OM in soils40,41, and studies highlighting Fe-associated C in anoxic zones do not examine why these FeIII minerals persist despite being thermodynamically poised for reductive dissolution12,20—this is a topic of separate studies explaining FeIII stability based on the thermodynamic constraints that OM composition places on FeIII respiration22,42,43. Examining these competing functional roles together remains a critical knowledge gap.

In this study, we quantified the relative contributions of Fe in retarding and accelerating C loss in the initial stages of MAOM formation, where physical constraints (macroaggregation, etc.) on decomposition were minimized using soil slurries (Fig. 1). We hypothesized that the electron transfer roles of Fe, which accelerate C mineralization, counteract C protection by Fe’s sorbent roles during and shortly following MAOM formation. To test this, we amended soil slurries with 57FeII and/or 13C-DOM under anoxic conditions and formed Fe–MAOM by introducing O2, simulating a primary mechanism of Fe–MAOM formation in humid soils. The soil slurries were incubated under either static oxic or alternating oxic/anoxic treatments under pH-buffered conditions, and the added and extant Fe and C were tracked using isotope measurements. Comparing treatments with and without added 57Fe, we find that adding 57FeII only decreases CO2 production when added together with 13DOC, and only under static oxic conditions. When 57FeII is added alone, Fenton chemistry promotes more SOM decomposition following oxidation of FeII to FeIII than the additional de novo 57FeIII minerals protect existing SOM. In fluctuating redox treatments, the added 57Fe provides an additional electron acceptor to fuel CO2 production during anoxic periods and 13DOC trapped by de novo 57FeIII phases is released, increasing anaerobic 13CO2 production relative to static oxic treatments. This study highlights that Fe’s electron transfer roles can largely counteract the protective effect of SRO–FeIII-C associations and sustain C decomposition in redox-dynamic systems.

Results and discussion


Consistent with a protective role, under static oxic conditions we found that FeII oxidation in the presence of added 13C-DOM resulted in SRO Fe–C associations that not only inhibited the mineralization of 13C-DOM by 35% relative to controls, but also suppressed the priming of native SOM mineralization by 47%, which consequently decreased overall CO2 production by 22% (Fig. 1d). However, when 13C-DOM was not added, FeII oxidation and the production of reactive oxygen species stimulated mineralization of native SOM by 8% relative to the controls (Fig. 1d). Thus, the formation of additional SRO–Fe phases did not provide net protection to SOM unless there was additional DOM present. As might be expected, the protective role of Fe was reversible under anoxic conditions. Although CO2 production from non-Fe amended treatments during the anoxic period was 68–70% lower than in the static oxic treatment (Fig. 1d), the de novo SRO Fe–MAOM formed via FeII oxidation was disproportionately vulnerable to subsequent reduction. This consequently stimulated the mineralization of both added 13C-DOM and the native SOM by 74% and 32–41%, respectively, and thus increased overall CO2 production by 41–49% relative to both non-Fe amended treatments (with or without added DOM, Fig. 1d). As a result of Fe-stimulated C mineralization, the anaerobic 13C-DOM mineralization was 81% greater than the oxic control. Below we provide details on the production of the Fe–MAOM, discuss the data supporting Fe protection of C along with the data supporting Fe stimulation of C loss, and then provide a synthesis of the work.

Generation of FeIII-(oxyhydr)oxides

The oxidation of 57FeII after a 1-d equilibration with the soil under anoxic conditions generated SRO FeIII (oxyhydr)oxides that impacted C cycling. Exposure to O2 (day 1–6) led to the oxidation of FeII, with aqueous FeII completely oxidized within 6 h. The sorbed FeII substantially decreased by 91% over the first day and slowly declined thereafter (Fig. 2). The treatment with both 57Fe and 13C-DOM added had 10% more adsorbed 57FeII than the 57FeII-only treatment before oxidation (Fig. 2a and b), likely due to co-sorption of the Fe2+–DOM complex, as observed previously44. The variable-temperature Mössbauer spectroscopy technique that we use to track the mineral composition of the 57Fe additions, gives excellent information on the crystallinity of the Fe phases, with high crystallinity phases ordering at higher temperatures. Both 57Fe addition treatments led to the formation of de novo SRO 57FeIII phases of lower crystallinity (lower Mössbauer ordering temperature) than the bulk soil Fe (Table 1; Supplementary Fig. 2 and Fig. 3), resulting in a 26–31 mmol kg−1 increase in lepidocrocite and 3–14 mmol kg−1 increase in nanogoethite and very-disordered FeIII (oxyhydr)oxides that preclude assignment (Fig. 3; Supplementary Table 1). The addition of 57Fe and 13C-DOM together resulted in the formation of even lower crystallinity SRO FeIII (oxyhydr)oxides than the 57Fe addition-only treatment as illustrated by the lower 35K/5K and 12K/5K crystallinity ratios (Table 1; Supplementary Fig. 3). Suppression of FeIII crystallinity by co-precipitation with dissolved fulvic acids has been shown previously in synthetic pure systems44, and here we extended this finding to a complex soil system containing a mixture of aluminosilicates, FeIII (oxyhydr)oxides, and a variety of organic compounds. In general, lower crystallinity Fe (oxyhydr)oxides (often measured by oxalate-extraction) have higher surface area, sorb more OM, and are thought to be associated with persistent OM in soils45.

Fig. 2: FeII concentration and 57Fe to total Fe ratio.
figure 2

Time-dependent a aqueous FeII, b sorbed (HCl-extractable) FeII, and c dissolved 57Fe to total Fe ratio in the aqueous phase in the redox-fluctuating treatment. FeII and 57Fe were undetectable during day 1.5–6 and 17.5–22. The gray shaded region represents the anoxic phase of the redox-fluctuating treatment. Error bars indicate s.e.m. (n = 3).

Table 1 Relative abundance of magnetically ordered FeIII (oxyhydr)oxides in 57Fe Mössbauer spectra of the initial unreacted soil and the amended 57Fe (corrected to exclude the signal from the native soil Fe), as a function of temperature.
Fig. 3: Solid-phase speciation of added 57Fe.
figure 3

57Fe partition was calculated from respective Mössbauer spectra (corrected to exclude the signal from the native soil Fe) for the (a)  57Fe-only and (b13C-DOM-57Fe addition treatments, prior to the oxic phase (day 1) and at the end of the 1st oxic (day 6), anoxic (day 17) and 2nd oxic (day 22) phases. Error bars represent standard errors associated with Mössbauer data modeling (see Supplementary Information).

Iron protection of organic matter

In the static oxic treatment, addition of 57Fe suppressed the mineralization of 13C-DOM by 35% (p < 0.01, Figs. 1d4 and Table 2): cumulative CO2 production was 25.6 ± 0.8 and 39.5 ± 1.1 mmol C kg−1 with and without 57Fe, respectively, equivalent to 17.1% and 26.3% of the added 13C-DOM (Fig. 4c and Table 2). Although 57Fe addition also inhibited net 13C-DOM mineralization in the fluctuating redox treatments (p < 0.01, Fig. 4c and Table 2), this inhibition was confined to the oxic portions (days 1–6 and days 17–22) of the incubation and was partly offset by a 74% enhanced 13C-DOM mineralization during the anoxic phase (days 6–17) relative to the treatment without added Fe (Figs. 1d, 4b, and Table 2, see below).

Fig. 4: Mineralization of the amended 13C-labeled DOM.
figure 4

a The percent contribution of 13C-labeled DOM to total CO2 production, b CO2 production rates from 13C-labeled DOM, and c cumulative CO2 production from 13C-labeled DOM. The gray shaded region represents the anoxic phase of the redox-fluctuating treatment. Error bars indicate s.e.m. (n = 3).

Table 2 Cumulative CO2 production in the fluctuating redox and static oxic treatments for all soils.

The generation of low crystallinity SRO–FeIII (oxyhydr)oxides from the oxidation of 57FeII in the presence of 13C-DOM resulted in a lower DOC concentration than in the 13C-DOM-only treatment and a concurrent increase in solid-phase 13C content (Fig. 5). This likely reflects the formation of SRO Fe–C complexes with 13C-DOM adsorbing or co-precipitating with the newly-formed SRO lepidocrocite and nanogoethite phases (Fig. 3). It is generally assumed that SRO Fe phases contribute to soil C persistence by protecting it against microbial mineralization5, but few studies have directly measured the bioavailability of Fe-associated OM39. Our study provides evidence that de novo formation of SRO Fe–C complexes inhibit the mineralization of fresh DOM inputs to soil. Others have also observed a large decrease in OM decomposition when glucose or fulvic acid sorbed to synthetic Fe minerals (ferrihydrite/goethite) was added to soils, as compared to additions of the free organic compounds40,41. The bioavailability of mineral-associated OM is generally thought to be linked to C loadings (e.g., C/Fe ratios), with a maximum adsorption capacity occuring at a C/Fe molar ratio of about one46. Co-precipitation could result in Fe–OM associations with much higher C/Fe ratios11,12,19. In our study, the initial C/Fe molar ratio of the added 13C-DOM and 57Fe was 2.1. If we assume that all DOM that was removed from the solution during the FeII oxidation event sorbed to the newly-formed FeIII (oxyhydr)oxides, the C/Fe ratio of those OM–FeIII (oxyhydr)oxide complexes would be ~1.7. Thus, there was likely 13C-DOM with a low affinity for FeIII (oxyhydr)oxides that remained as unprotected 13C-DOM in the aqueous phase and this likely led to our observation of significant 13C-DOM mineralization even in the presence of de novo FeIII (oxyhydr)oxides (Fig. 4).

Fig. 5: Dissolved organic carbon concentration and solid-phase 13C enrichment.
figure 5

a Dissolved organic carbon concentration (MES buffer concentration was subtracted) and b solid-phase 13C atom fraction from 13C-DOM addition treatments. The gray shaded region represents the anoxic phase of the redox-fluctuating treatment. Error bars represent s.e.m. (n = 3).

Labile C inputs are often observed to alter the decomposition of extant SOM, defined as priming47,48. During the oxic periods of the experiment, 57FeII oxidation in the presence of added 13C-DOM not only suppressed the mineralization of the amended 13C-DOM, but also partially inhibited the priming of native SOM decomposition compared to the DOM-only treatment (Fig. 6; Table 2; Supplementary Table 2). In the static oxic treatment, addition of 13C-DOM alone or together with 57Fe increased native SOM-derived CO2 production compared to the soil-only control (priming effect) (p < 0.01, Fig. 6a, c and e; Table 2; Supplementary Table 2). However, adding 13C-DOM and 57Fe together resulted in a significantly smaller priming effect on native SOM mineralization than adding 13C-DOM alone under the static oxic treatment (p < 0.01, Fig. 6e, Table 2 and Supplementary Table 2). With the addition of 13C-DOM, cumulative primed CO2 from native SOM under the static oxic treatment measured 10.2 ± 1.2 and 19.1 ± 0.9 mmol C kg−1 with and without 57Fe addition, respectively (Fig. 6e and Supplementary Table 2). Cumulatively, adding 13C-DOM together with 57Fe suppressed aerobic priming of native SOM by 47% relative to adding 13C-DOM alone (Fig. 1d, Table 2 and Supplementary Table 2). Collectively, 57Fe oxidation in the presence of 13C-DOM resulted in 22% less overall C mineralization, compared to addition of 13C-DOM alone (Fig. 1d and Table 2).

Fig. 6: Native soil C mineralization.
figure 6

CO2 production rates from soil C under a static-oxic and b redox-fluctuating conditions, cumulative CO2 production from soil C under c static-oxic and d redox-fluctuating conditions, and priming of soil C under e static-oxic and f redox-fluctuating conditions. The gray shaded region represents the anoxic phase of the redox-fluctuating treatment. Error bars indicate s.e.m. (n = 3).

We propose that de novo SRO FeIII minerals protected the added 13C-DOM under static oxic conditions, decreasing DOM availability for microbial growth, and suppressing the priming of native SOM. An alternative explanation, that the sorption of native SOM onto the de novo SRO FeIII (oxyhydr)oxides inhibited priming, is unlikely because when we added 57FeII alone it actually increased the mineralization of native SOM (due to reactive oxygen species production, as discussed below) (Fig. 6a, e and Table 2). Similarly, prior studies have shown that the addition of new SRO FeIII phases to soils has little to no impact on the mineralization of native SOM41,49. Rather, it is likely that DOM–FeIII interactions create physico-chemical barriers that limit priming by decreasing microbial access to the new DOM. Thus, when reduced soils receive oxygenated water due to rainfall, snowmelt, or irrigation, the oxidation of FeII in the presence of DOM and formation of OM–FeIII complexes may contribute to C protection both directly, as previously known, and indirectly, by suppressing priming.

Iron stimulation of DOM and SOM mineralization

Only in the treatment where 13C-DOM and 57Fe were added together were we able to confirm that Fe had an overall protective effect on OM, and that protection was limited to the oxic portions of the experiment. Below, we quantified the impact of Fe on OM mineralization via Fe-stimulated Fenton chemistry during the first few days of oxic exposure and via Fe reduction-mediated reactions during the anoxic periods.

Adding 57Fe alone strongly stimulated CO2 production from native SOM during the first 3 days of the static oxic treatment (and the oxic portions of the fluctuating redox treatment) relative to the soil-only control (p < 0.01), with no stimulatory impact afterwards (Fig. 6a and e). Cumulatively, FeII oxidation stimulated CO2 production by 8% (Fig. 1d and Table 2). To confirm the role of Fenton chemistry, we performed a parallel experiment with added terephthalate—an effective hydroxyl radical scavenger—and found similar CO2 production between FeII-added treatments and soil-only controls (Supplementary Fig. 4). Recent studies have similarly shown that FeII oxidation is linked to increases in soil CO2 production via the generation of radical oxygen species50,51, which facilitate the breakdown of complex biopolymers to produce labile substrates for microbial respiration34,35. Others have also attributed increased CO2 production following FeII oxidation to an increase in acidity that can promote DOC release36. Given that we conducted these experiments in a strong buffer at a constant pH, the increased CO2 production following FeII oxidation was most likely derived from the production of reactive oxygen species such as the hydroxyl radical.

Soil C mineralization rates typically decrease as O2 becomes limiting22,52,53. In our soil-only control, CO2 production from native SOM during the anoxic period was 70% lower than in the static oxic treatment (Figs. 1d, 6 and Table 2). However, during the anoxic portions of the experiment, Fe addition stimulated native SOM mineralization relative to the no-Fe treatment (Fig. 6b and f; Table 2). In the 57Fe-addition treatment, the degree of anoxic suppression of CO2 production decreased from 70 to 58% of that under oxic conditions (p < 0.05; Table 2), as a result of a 41% higher anoxic native SOM-derived CO2 production in the Fe addition treatment than in the no-addition control (p < 0.05, Table 2; Figs. 1d, 6b, d and f). This likely resulted from enhanced microbial use of FeIII as an electron acceptor in the 57Fe addition treatments. Following the transition from oxic to anoxic conditions in the fluctuating redox treatments, substantial FeIII reduction occurred (day 6–17, Fig. 2a and b) and adding 57Fe increased the total FeII production rates (2.9 mmol kg−1 d−1) compared to the soil-only control (1.6 mmol kg−1 d−1). This was most likely due to the facile reduction of de novo 57Fe SRO lepidocrocite and nanogoethite, which had a much lower crystallinity (and thus a higher reactivity) than native soil FeIII (oxyhydr)oxides (Table 1 and Fig. 3; Supplementary Table 1 and Fig. 3). The availability of native SRO FeIII phases likely limits Fe reduction in this subtropical agricultural soil, and the de novo SRO 57FeIII phases were preferentially utilized as electron acceptors for microbial respiration as evidenced by the preferential release of 57FeII in the aqueous phase (Fig. 2c) and the measured decrease of these 57Fe mineral phases following reduction (Fig. 3a and Table 1; Supplementary Table 1).

Iron’s stimulation of C mineralization during anoxic periods was greatly enhanced when 57Fe and 13C-DOM were added together, yielding increases in mineralization of native SOM (anoxic priming) and 13C-DOM by 32 ± 3% and 74 ± 7%, respectively, relative to adding 13C-DOM alone (p < 0.05; Table 2; Figs. 1d, 4b, 4c, 6b, d and f). In fact, when13C-DOM and 57Fe were added together, anaerobic 13C-DOM mineralization in the fluctuating redox treatment was even 81% greater than the aerobic 13C-DOM mineralization in the static oxic treatment at the same point in time (p < 0.01; Fig. 4b; Table 2). The stimulation of 13C-DOM mineralization under anoxic conditions was linked in part to its molecular composition, given that thermodynamic constraints on Fe reduction limit metabolism to relatively oxidized C substrates22,42,43. During the anoxic periods, the mineralization of 13C-DOM over the native SOM (in both the 13C-DOM only and DOM–Fe addition treatments) was 2–3 times higher than that in the static oxic treatment at the same point (Fig. 4a). Characterization of the molecular composition of 13C-DOM and water-extractable native SOM using Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS) revealed that the 13C-DOM had significantly less lignin-derived materials and much more aliphatic formulae than the water-extractable native SOM (Supplementary Table 3, Figs. 5 and 6), which represents the most bioavailable fraction of native SOM54. The preferential anaerobic mineralization of 13C-DOM over SOM may be due to a lower abundance of lignin-derived compounds, which are not readily depolymerized under anoxic conditions55. In addition, compared to water-extractable native SOM, 13C-DOM contains compounds with higher nominal oxidation state of C (NOSC values > 0.5, Supplementary Fig. 6), which are associated with a higher likelihood of thermodynamic favorability (−ΔGr) when coupled to FeIII reduction than the bioavailable fraction of native SOM22,43. Fe reduction was also stimulated by the addition of 13C-DOM alone (producing 2.3 mmol kg−1 d−1 of FeII compared to the soil-only control rate of 1.6 mmol kg−1 d−1) (Fig. 2a and b), consistent with prior work26. However, when Fe and DOM were added together, Fe reduction was greatly increased to 7.4 mmol kg−1 d−1, which was even greater than the additive effect of separate 13C-DOM (2.3 mmol kg−1 d−1) and 57Fe additions (2.9 mmol kg−1 d−1) (Fig. 2a and b). This was because compared to oxidation of 57FeII alone, oxidizing 57FeII in the presence of 13C-DOM led to formation of even less-crystalline SRO lepidocrocite and nanogoethite phases (all ordering at <35 K in the Mössbauer spectra, Table 1 and Supplementary Fig. 3). These SRO 57Fe-13C-OM phases exhibited high rates of Fe reduction, releasing significant 57Fe2+(aq) and 13C-DOM (Figs. 2 and 5) when exposed to anoxic conditions, leaving the solid phase depleted in its lowest crystallinity Fe phases (Fig. 3 and Table 1; Supplementary Table 1 and Fig. 7), and preferentially stimulating anaerobic mineralization of the added 13C-DOM (Fig. 4a).

Fe reduction can solubilize significant amounts of OM adsorbed or coprecipitated with FeIII (oxyhydr)oxides directly, as shown in our experiment, or indirectly because of an increase in pH30,32,56. This re-mobilized 13C-DOM often includes biochemically labile C32,57, and may potentially offset the kinetic/thermodynamic constraints often limiting anaerobic decomposition22,43,58. We find that collectively, the reduction of SRO FeIII phases offset O2 limitations on C mineralization by 24 ± 3% relative to the non-Fe amended treatment (Table 2).


A recent survey of over 5500 soil profiles spanning continental scale environmental gradients found that SRO Fe and Al (oxyhydr)oxide abundance was the best predictor of C content in humid soils, among the geochemical and climate variables that were available45. This is consistent with other work showing that SRO FeIII phases are broadly implicated in the persistence of OM in soil1,3,59. However, the nature of the relationship between Fe and C in humid soils—and redox dynamic soils in general, which would include floodplain and perennial wetland soils from all climatic regions—is far from straightforward. Humid soils are replete with microsites that undergo dynamic anoxia in response to high labile C loads during periods of high moisture and experience appreciable FeIII reduction rates23,25,60,61. Oxidation of the FeII generated from FeIII reduction is a common mechanism for MAOM formation in humid and redox-dynamic soils, yet Fe is also responsible for OM loss and our work here illustrates two principal refinements in this regard.

First, the production of SRO Fe–MAOM via FeII oxidation will likely increase CO2 production in the short-term. Only when we formed MAOM in the presence of DOM and maintained strict oxic conditions was there a net decrease in C mineralization (both in the added 13C-DOM and the native SOM, i.e. via decreased priming). When we simply generated MAOM via FeII oxidation without added DOM, Fenton chemistry caused an 8% increase in C mineralization (Fig. 1d). Upon the inevitable return to periodic anoxia in humid soils, our work suggests that C mineralization would be accelerated by 41–49% by Fe reduction (Fig. 1d), thus counteracting the stabilization effect on OM of SRO Fe phases. The magnitude of these counteracting mechanisms may also be influenced by soil structure, which we largely eliminated in our study by conducting experiments in soil slurries. Hence, direct application of our results to in situ soil environments is tentative. However, the general principles of our work are also likely to be applicable to structurally complex soil systems. For example, Fe mineral-associated C is often released in natural soils under in-situ flow conditions as a consequence of dissimilatory Fe reduction (e.g.,62) and thus becomes more vulnerable to microbial decomposition. In our study, we even found that the added DOM was preferentially degraded under anoxic conditions relative to the oxic control (Fig. 4), which highlights how the thermodynamic constraints of anaerobic metabolism and the molecular composition of C sources can influence the fate of fresh DOM inputs22,42,43. Consequently, the net effect of Fe–C interactions in dynamic redox environments likely hinges in part on the composition of DOM inputs, a worthy topic for further research.

Second, our work here suggests that the initial SRO Fe–C associations are not likely to persist without protection from periodic Fe reduction events. Several researchers have identified or produced SRO–FeIII–OM colloids that are resistant to either microbial or chemical reduction63,64,65,66,67, however, the key components conferring this protection are variable and/or elusive. Some work has identified that SRO–FeIII–OM co-precipitates with low C/Fe ratios provide resistance to microbial reduction63,64, whereas other work has emphasized structural properties (conformation and micro-aggregation) as the mechanism that retards dissolution65,66,67. SRO Fe–OM phases are often co-precipitated with Al and Si ions68—which can retard recrystallization69—and given the co-association of Al and Fe with OM in humid soils, Al is a strong candidate for protecting Fe against reduction. However, studies that have examined Al and Si co-precipitated Fe-(oxyhydr)oxides found those ions also make the co-precipitates more susceptible to reductive dissolution70. Coward et al.67 recently proposed several mechanisms by which SRO FeIII–OM phases could become resistant to reductive dissolution, including acquiring reduction-resistant surface coatings, or becoming embedded in a composite aggregate structure6. Such a protective coating could even come from higher crystallinity Fe (oxyhydr)oxides. Hall et al. recently found that 14C-derived C residence time in humid soils was positively correlated with Fe phase crystallinity71. Consistent with that, we find here that in contrast to the initial oxidation event, the 2nd oxidation event generated more crystalline 57Fe phases (Table 1; Supplementary Fig. 8) and did not stimulate additional C mineralization (Fig. 6 and Table 2). It may be that during repeated redox fluctuations a substantial portion of the co-precipitated OM would be lost, but a core Fe–MAOM structure would remain protected from reductive dissolution.

Perhaps most compelling is the growing evidence that various aggregation, conformation, and structural characteristics of soils confer protection for OM5,6,7,10. Even the protective surface coatings66,67 or conformational changes in OM at low C/Fe ratios64 discussed above are examples of micro-aggregate structures not unlike the encasement of SRO Fe–OM phases by aluminosilicate clays or other processes that generate micro-aggregates of minerals and OM during pedogenesis6,7,10,72. These aggregation processes can structure microaggregates with core SRO Fe phases and outer aluminosilicate or other phases that are not susceptible to reductive dissolution—as observed in Andisols by dithionite-resistant SRO–Fe phases66. Our soil slurry approach was designed to minimize the physical constraints (macro-pore flow, spatial arrangement of microbes, minerals and OM, and the development of aggregates) on C decomposition and thereby isolate the sorbent and electron-transfer roles of Fe in C dynamics (Supplementary Fig. 1). Under these conditions, we find that Fe does not confer intrinsic protection for OM in redox-dynamic soils. In an in situ soil environment—where MAOM emerges in a dynamic three-dimensional space—structural and physical protection of MAOM is thus likely a key protective mechanism for reconciling the comparatively large proportions of SRO-OM associations in soil of very old age based on 14C-dating1,4,5,59. Future studies should thus assess the extent that the formation and destruction of Fe-cemented microaggregates contribute to OM persistence in redox-dynamic soils. Our work demonstrates that the inherent persistence of SRO Fe-associated C cannot be guaranteed. Biological and geochemical context is critical for understanding the long-term fate of FeIII-associated SOM under a changing climate, given the dual roles of FeIII phases in both accelerating and inhibiting OM decomposition.



We employed a dual isotope approach in a soil slurry to test our hypothesis that the electron transfer roles of Fe that accelerate C mineralization will counteract C protection by Fe’s sorbent roles during, and shortly following, MAOM formation. We used soil slurries (i.e., homogenized mixture of soil and water) to minimize the physical-protection mechanisms of aggregation and the spatial separation of decomposers, substrates, and mineral surfaces, and thus focus on Fe’s sorbent and electron-transfer roles. Our dual isotope approach allowed us to distinguish between native SOM and fresh plant-derived DOM via 13C labeling, as well as between neo-formed reactive Fe minerals formed in situ and the different forms of Fe minerals in the native soil via 57Fe labeling coupled with 57Fe Mössbauer spectroscopy.

Preparation of 13C-labeled plant-derived DOM

13C-DOM was extracted from 13C-labeled bermudagrass. DOM is inherently heterogeneous, diverse and dynamic in composition6, and here we used bermudagrass-extracted DOM to encompass a mixture of organic molecules representative of those that derive from early stage herbaceous litter decomposition. A pulse-labeling method was used to label Tifton-85 bermudagrass (Cynodon dactylon x Cynodon nlemfuencis) with 13CO2 (99.999 atom%, Cambridge Isotope Laboratories Inc; see Supplementary Methods for additional information). After labeling, aboveground biomass was harvested, immediately frozen, freeze-dried, and then ground using a Wiley mill to <1 mm. DOM extractions were conducted in a shaker at 140 rpm for two days with a solid-to-water ratio of 1:5, followed by centrifugation. The supernatant was filtered through a 0.2 µm membrane filter. The derived DOM solution had 10.3% 13C. Characterization of the molecular composition using ultrahigh resolution mass spectrometry (FTICR-MS) revealed that this 13C-enriched DOM was comprised of predominantly aliphatic compounds (76%) and lignin-derived/carboxyl-rich alicyclic molecules (23%), with mean population O/C, H/C and DBE values of 0.44 ± 0.12, 1.60 ± 0.22, and 6.31 ± 3.04, respectively (Supplementary Fig. 6 and Table 3). Compared to water-extractable natural SOM, the bermudagrass-derived DOM had significantly more aliphatic compounds with less lignin-derived materials (Supplementary Notes). In addition, the 13C-containing population of DOM formulas displayed chemical composition distribution indistinguishable from that of 12C-only formulae, suggesting no preferential incorporation of 13C atoms across molecular compounds (Supplementary Fig. 5 and Table 3).

Study site and soil sampling

Our study site is located in the Calhoun Critical Zone Observatory (CZO) in Union County, South Carolina, USA (34.611 N;-81.727, IGSN: IEJCA0013). This site has a humid warm temperate climate, with mean annual precipitation and mean annual temperature of about 1212 mm and 17 °C, respectively (Southeast Regional Climate Center, 2016). The soil used is classified as fine kaolinitic, thermic Typic Kanhapludults of the Appling series, derived from granitic gneiss. We collected soils from cultivated land on an interfluve managed for hay and a few annual crops (e.g., Zea mais, Triticum aestivum). Current management practice includes annual plowing and disking, the addition of ~4 Mg ha−1 of lime in the last eight years and fertilization of NPK at the rate of 160, 40, and 70 kg ha−1 yr−1, respectively73. Interfluves across the Calhoun CZO are characterized by deep soils with pronounced subsurface redoximorphic features23,74 and seasonal fluctuations in Fe reduction events corresponding with antecedent moisture and labile organic C75. During the early spring, surface soils in particular experience a peak in FeII associated with Fe reduction, which subsequently subsides as the soils become drier later in the spring/summer75. Soil pits were dug by backhoe. Surface soils (0–20 cm) were collected and transported overnight to the University of Georgia under ambient conditions. Soils were homogenized and visible plant debris, rocks, and soil macro-fauna were removed manually.

Total OC content and its δ13C measured via an elemental analyzer-stable isotope ratio mass spectrometer (EA-IRMS) were 2.1% and −22.3‰, respectively. Water-extractable native SOM, extracted by mixing field soils with high purity water (see details in Supplementary Methods), was 33.8 mg C kg−1. Water-extractable native SOM was comprised of largely polycyclic aromatic (21.5%), lignin-derived/carboxyl-rich alicyclic molecules (49.1%) and aliphatic compounds (22.5%) with mean population O/C, H/C and DBE values of 0.21 ± 0.09, 1.21 ± 0.35 and 13.73 ± 7.51, respectively (Supplementary Fig. 6 and Table 3). Total soil Fe content, measured by ICP-MS following Li-metaborate fusion (Acme Labs, Vancouver, BC Canada)76, was 308 mmol kg−1. The concentration of SRO FeIII oxides based on ascorbic acid/citrate extraction77 was ~25.5 mmol kg−1. Soil pH (1:2 ratio of soil: water) was 6.2. XRD analysis revealed that the clay mineralogy of this soil is dominated by kaolinite and illite78.

Laboratory incubation

The experiment had four amendment treatments: soil amended with 13C-DOM, soil amended with 57FeII, soil amended with both 13C-DOM and 57FeII, and control soils with no additions. Each received two redox treatments: the first CO2-free air (static oxic) treatment; the second treatment with 5 days of CO2-free air, then 11 days of N2 followed by 5 days of CO2-free air again (fluctuating redox treatment). Experiments were performed using soil slurries at a soil:water ratio of 1:10 in triplicate and in the dark under ambient laboratory temperatures of ~24 °C. Field-moist soil (3.5 g, equivalent to 3 g of dry soil) was added to a 125 ml brown amber flask in an anoxic glovebox (Coylabs, Grasslake, MI) with a 95%/5% N2/H2 atmosphere and stored for 2 h to remove O2, followed by mixing with 30 mL of anoxic MES buffer solution (10 mM, pH = 6). Soil slurries received either 1 ml of anoxic water (controls), or 97 atom% 57Fe-enriched Fe2+aq added as FeCl2 to reach the initial FeII concentration of 70 mmol kg–1 soil (57Fe-addition treatments). This isotopic enrichment allowed us to monitor the fate of the added 57FeII. The added 57Fe corresponds to 91% of total 57Fe (added and native) based on native soil 57Fe abundance (2.1%), although it is 18% of the total soil Fe. Anoxic 13C-DOM solution was added to achieve an initial concentration of 150 mmol C kg−1 soil of the added DOM, and thus the added C/Fe ratio is 2.1 in the 13C-DOM-57FeII addition treatment. Buffering soil slurries with MES at a constant pH excludes confounding effects of associated pH shifts. The pH of the soil slurries was adjusted to 6 using anoxic HCl or NaOH solutions. The soil slurries were then mixed on a rotary shaker (250 rpm) in the anoxic glovebox for 1 day to equilibrate the added 57FeII across the aqueous and solid phases under anoxic conditions prior to exposure to O2. Then the reactors were either exposed to static oxic or fluctuating redox treatments by placing the reactors on end-over-end shakers in custom-built, sealed atmospheric chambers (fully contained within the anoxic glovebox) with a continuous flow of either CO2-free air (static oxic treatment) or CO2-free air/N2 alternating treatment. Two sets (3 replicates per set) of parallel samples were prepared: one was used for destructively sampling the soil slurry with the other one reserved for sampling the evolved gas.

To test the effect of hydroxyl radicals on SOM mineralization, 10 mM of terephthalic acid (TPA, an effective hydroxyl radical scavenger) was added to the treatments of soil slurry-only and FeII-amended soil slurry. The reactors were equilibrated under anoxic conditions for 1 d, followed by oxidation with CO2-free air for 5 d. Gas samples were collected for CO2 analysis.

Soil slurry sampling and analysis

Sampling of anoxic reactors was performed within the anoxic glovebox. All chemical reagents were prepared in advance with degassed water to preserve Fe oxidation state and for samples collected during anoxic sampling periods. Samples were collected using wide-orifice pipette tips that allowed complete collection of soil particles in the slurry. Aqueous FeII was extracted from the soil slurries by centrifuging the samples at 14,000 rcf for 10 min, and acid-extractable (sorbed) FeII was solubilized by suspending the remaining pellet in 0.5 M HCl and shaking it for 2 h on a horizontal shaker at 150 rpm. The extracts were then centrifuged at 14,000 rcf for 10 min and the supernatants analyzed for FeII using a modified ferrozine protocol77. Fe isotope compositions in the aqueous phase and acid-extracts were measured by inductively coupled plasma mass spectrometry (ICP-MS, Perkin Elmer, Elan 9000). Soil slurries were sampled at the end of oxic or anoxic incubations for C and isotope analysis and centrifuged at 14,000 rcf for 10 mins. The supernatant was carefully removed and filtered through a 0.2 µm membrane filter for DOC analysis. DOC was measured with a Shimadzu TOC analyzer. The pellet was washed with anoxic DI water three times, freeze-dried and analyzed for total C and 13C analysis using EA-IRMS.

Gas sampling and measurements

Each reactor was flushed with CO2-free air or N2 gas for 15 min at 500 mL min−1 every 4 h to 1 d during the oxic phases and every 2–3 days during the anoxic period immediately following each headspace gas measurement. We collected gas samples for measurements of CO2 and their 13C values immediately prior to flushing, enabling us to quantify cumulative CO2 losses and their 13C values over the entire experiment. A 5 ml gas sample was collected with gastight syringes and injected to pre-evacuated 3 mL glass vials (Exetainer, Labco Inc., UK) for CO2 concentration analysis. A 30 ml gas sample was collected from each reactor and stored in helium-purged and evacuated 20-ml glass serum bottles with teflon septa sealed with aluminum crimps for 13C measurements. Concentrations of CO2 were measured with a gas chromatograph and thermal conductivity detector (Shimadzu, Kyoto, Japan). Dissolved CO2 in the slurry was calculated based on Henry’s law. The 13C/12C isotope ratio of CO2 was determined by injecting 20 ml gas using a gas-tight syringe to Piccaro G2201-i via an ultra-zero grade CO2-free air carrier gas. δ13C values of CO2 from the 57FeII-added soils and soil-only controls was corrected using three CO2 tank standards with δ13C values of −8.0‰, −23.8‰, and −39.7‰ respectively. The 13C atom fraction of CO2 from 13C-DOM-added soils was calibrated using 5 standards varying from 2 to 18% x(13C). These standards were created by mixing 99% x(13C) Na2CO3 with natural abundance Na2CO313C = 1.42‰), digesting with an excess of 12 M HCl and removing aliquots of headspace79. Concentrations of CH4 were analyzed by gas chromatography with a flame ionization detector (Shimadzu, Kyoto, Japan). However, CH4 production in this experiment was minimal, accounting for <1% of total C mineralization. Therefore the effect of CH4 production on 13C mass balance was negligible80.

The percent contribution of added 13C-DOM to CO2 respiration (PDOM) was estimated using a two-source mixing model:

$$P_{{\mathrm{DOM}}} = (x^{{\mathrm{13}}}\left[ {{\mathrm{CO}}_{\mathrm{2}}} \right]_{{\mathrm{D}} + {\mathrm{S}}} \,-\, x^{{\mathrm{13}}}\left[ {{\mathrm{CO}}_{\mathrm{2}}} \right]_{\mathrm{S}})/(x^{{\mathrm{13}}}{\mathrm{C}}_{\mathrm{D}} - x^{{\mathrm{13}}}{\mathrm{C}}_{\mathrm{S}})* {\mathrm{1}}00$$

where x13[CO2]D+S and x13[CO2]S are atom fraction 13C of CO2 respired in the 13C-DOM amended soils and the treatments with no C addition, respectively; x13CD is the initial atom fraction 13C of 13C-DOM and x13CS is the initial soil 13C. The fraction of CO2 derived from SOM was calculated by difference:

$$P_{{\mathrm{soil}}} = {\mathrm{1}}00 - P_{{\mathrm{DOM}}}$$

Fluxes of CO2 derived from the added DOM and native SOM were calculated by multiplying total CO2 fluxes by their fractional contributions. We calculated priming as the difference in soil-derived CO2 losses between treatments that received 13C-DOM and/or 57Fe additions and soil-only control treatment:

$$C_{{\mathrm{primed}}} = C_{{\mathrm{soil}}\_{\mathrm{amended}}} - C_{{\mathrm{soil}}\_{\mathrm{control}}}$$

57Fe Mössbauer analysis

Fe speciation was determined using 57Fe Mossbauer analysis. Use of 57Fe isotopes allows us to track the amended 57Fe using Mössbauer spectroscopy, which detects only 57Fe atoms and no other Fe isotopes. The Mössbauer spectra of the amended 57Fe was calculated as the difference between the spectra from the 57FeII-enriched treatment and the baseline spectra from soils with natural isotopic abundance, after taking into account the different total 57Fe concentrations in the 57FeII-enriched treatment and the control soils81. Therefore, the resulting Mössbauer spectra of the amended 57Fe excluded the spectral signal from the native soil Fe atoms (Supplementary Figs. 3, 79). To prevent FeII oxidation, solid samples for 57Fe Mössbauer analysis were collected in the anoxic glove box following centrifugation at 14,000 g for 10 min, preserved between layers of O2-impermeable Kapton tape, and immediately frozen in a −20 °C freezer81. The samples were then placed within the spectrometer cryostat (pre-cooled to <140 K), which operated in a He atmosphere to prevent FeII oxidation by O2. 57Fe Mössbauer spectra were recorded in transmission mode with a variable-temperature He-cooled cryostat (Janis Research Co.) and a 1024 channel detector. Detailed information regarding the Mössbauer spectra modeling is provided in the Supplementary Methods. The detailed fitting parameters are presented in Supplementary Tables 411.

Statistical analysis

A one-way ANOVA (Turkey’s HSD) was used to assess the effects of redox treatment on DOM- and SOM-derived CO2 production. A two-way ANOVA was performed to assess effects of DOM and Fe additions on the CO2 production and priming effect. Statistical analysis was performed using SPSS 16.0 for Windows and the differences were considered significant at p < 0.05. Iron reduction (FeII production) rates were calculated from the slope of the linear regression (R2 > 0.9) of FeII concentration over time during the anoxic period of the fluctuating treatment.