Effects of Fe(III) Oxide Mineralogy and Phosphate on Fe(II) Secondary Mineral Formation during Microbial Iron Reduction

: The bioreduction of Fe(III) oxides by dissimilatory iron-reducing bacteria may result in the formation of a suite of Fe(II)-bearing secondary minerals, including magnetite (a mixed Fe(II)/Fe(III) oxide), siderite (Fe(II) carbonate), vivianite (Fe(II) phosphate), chukanovite (ferrous hydroxy car-bonate), and green rusts (mixed Fe(II)/Fe(III) hydroxides). In an effort to better understand the factors controlling the formation of speciﬁc Fe(II)-bearing secondary minerals, we examined the effects of Fe(III) oxide mineralogy, phosphate concentration, and the availability of an electron shuttle (9,10-anthraquinone-2,6-disulfonate, AQDS) on the bioreduction of a series of Fe(III) oxides (akaganeite, feroxyhyte, ferric green rust, ferrihydrite, goethite, hematite, and lepidocrocite) by Shewanella putrefaciens CN32, and the resulting formation of secondary minerals, as determined by X-ray diffraction, Mössbauer spectroscopy, and scanning electron microscopy. The overall extent of Fe(II) production was highly dependent on the type of Fe(III) oxide provided. With the exception of hematite, AQDS enhanced the rate of Fe(II) production; however, the presence of AQDS did not always lead to an increase in the overall extent of Fe(II) production and did not affect the types of Fe(II)-bearing secondary minerals that formed. The effects of the presence of phosphate on the rate and extent of Fe(II) production were variable among the Fe(III) oxides, but in general, the highest loadings of phosphate resulted in decreased rates of Fe(II) production, but ultimately higher levels of Fe(II) than in the absence of phosphate. In addition, phosphate concentration had a pronounced effect on the types of secondary minerals that formed; magnetite and chukanovite formed at phosphate concentrations of ≤ 1 mM (ferrihydrite), <~100 µ M (lepidocrocite), 500 µ M (feroxyhyte and ferric green rust), while green rust, or green rust and vivianite, formed at phosphate concentrations of 10 mM (ferrihydrite), ≥ 100 µ M (lepidocrocite), and 5 mM (feroxyhyte and ferric green rust). These results further demonstrate that the bioreduction of Fe(III) oxides, and accompanying Fe(II)-bearing secondary mineral formation, is controlled by a complex interplay of mineralogical, geochemical, and microbiological factors.


Bioreduction Experiments
The bioreduction experiments were conducted in sterile 160 mL serum bottles containing 100 mL of sterile defined mineral medium (DMM) [55] with 80 mM Fe(III) in the form of the corresponding Fe(III) oxide, 75 mM formate as the electron donor, phosphate (0-10 mM depending on the experimental system), and 100 µM AQDS as an electron shuttle in AQDS-amended systems. The DMM was prepared by combining all components (except the electron donor, phosphate, and AQDS), and the pH was adjusted to 7.5 by titration with 1 M NaOH, portioned into serum bottles, and autoclaved. After the medium cooled to ambient temperature, the electron donor, phosphate, and AQDS were added from filter-sterilized stock solutions; all systems amended with phosphate also contained AQDS. The bottles were sealed with rubber septa and aluminum crimp caps and made anoxic by sparging with sterile argon. All experimental systems were prepared in duplicate. After 24 h, a 2 mL sample of suspension was removed from each of the bottles that had been amended with phosphate to determine the extent of phosphate uptake by the Fe(III) oxides. The sample was filtered through a 0.22 µm nylon filter and the filtrate was saved for the measurement of phosphate concentration. The inoculum was prepared from latelog-phase cultures of S. putrefaciens CN32 (American Type Culture Collection BAA-543), as described by O'Loughlin et al. [12]. Experiments were initiated by spiking each bottle with the volume of inoculum needed to achieve a cell density of~5 × 10 9 cells mL −1 . The bottles were placed on a roller drum and incubated at 30 • C in the dark. Samples of the suspensions-for monitoring Fe(II) production, as well as for the identification of secondary minerals by pXRD, SEM, and 57 Fe Mössbauer spectroscopy-were collected with sterile syringes. Unless otherwise indicated, sample collection and processing were conducted in a glove box containing an anoxic atmosphere.

Analytical Methods
The reduction of Fe(III) was monitored by measuring the total Fe(II) content of 0.75 M HCl extracts of the suspensions (Fe(II) tot , referred to hereafter as Fe(II)). Samples for Fe(II) analysis were prepared by adding 0.75 mL of anoxic 1 M HCl to 0.25 mL of suspension (the addition of 100 µL of concentrated HCl to samples containing magnetite was needed to ensure the dissolution of all Fe(II) phases). After 1 week, the samples were centrifuged at 25,000× g for 10 min. The Fe(II) concentrations in the supernatants were determined by the ferrozine assay [87]. Briefly, 1 mL of HEPES-buffered ferrozine reagent [88] was added to 50 µL of supernatant, and the absorbance at 562 nm was measured. Phosphate concentration was determined by inductively coupled plasma-optical emission spectroscopy using a Perkin Elmer 4300DV instrument.
Changes in the mineralogy of the suspensions were monitored by pXRD with a Rigaku MiniFlex X-ray diffractometer with Ni-filtered Cu Kα radiation. Samples for pXRD analysis were collected by filtration on 25 mm diameter, 0.22 µm nylon filters and covered with 8.4 µm thick Kapton ® film under anoxic conditions. Although the pXRD analysis was conducted under ambient atmosphere, samples prepared in this manner showed no evidence of oxidation when scanned between 5 • and 80 • 2θ at a speed of 1.25 • 2θ min −1 . The pXRD patterns were analyzed with the JADE 7 software package (MDI, Livermore, CA, USA) to remove the background through polynomial fitting and also to remove the K α2 components.
Samples for SEM imaging were prepared by placing 500 µL of suspension on aluminum specimen mounts, allowing the solids to settle, removing the overlying liquid with a pipette, and drying the film of solids in a glove box. Specimens were briefly (<30 s) exposed to air during transfer to the Hitachi S-4700-II FEG-SEM.
Transmission Mössbauer spectroscopy was performed with a variable temperature He-cooled system with a 1024 channel detector. The 57 Co source used (~50 mCi) was in a Rh matrix at room temperature. All center shifts reported are relative to an α-Fe foil at room temperature. Samples were prepared by filtering the cell suspension (approx. 4 mL) in an anoxic glove box with recoverable filter paper. The filter paper was then sealed between Minerals 2021, 11, 149 5 of 44 two pieces of 5 mm Kapton tape to avoid oxidation while the sample was mounted. No indication of inadvertent oxidation was observed. Spectral fitting was done using Recoil Software (version 1.01998, University of Ottawa, Ottawa, ON, Canada). Voigt-based fitting was used to model the spectra to determine the hyperfine parameters and the relative areas between phases. The Lorentzian linewidth was held at 0.12 mm·s −1 , as it was the linewidth measured on the spectrometer for an ideally thick α-Fe foil. The relative peak areas (1:1 for doublets, 3:2:1:1:2:3 for sextets) were held constant throughout fitting. Each phase was fitted with only a single component (i.e., multiple QS and H distributions were not allowed for a single phase in fitting).

Fe(III) Oxide Bioreduction
The rate and extent of bioreduction of the Fe(III) oxides varied substantially based on mineralogy ( Figure 1 and Table 1). Within~80 days, only 3% (on a mol Fe(III) basis) of hematite and 10% of goethite were reduced, compared to 58% of maghemite, 57% of ferrihydrite, and 49% of ferric green rust. The intermediate reduction of 36%, 32%, and 30% of added Fe(III) occurred for akaganeite, lepidocrocite, and feroxyhyte, respectively. The rate of Fe(III) oxide bioreduction was poorly correlated with oxide surface area ( Figure 2). These results are consistent with previous studies showing that highly crystalline phases, such as hematite and goethite, are less bioavailable for microbial iron reduction than less crystalline phases, such as ferrihydrite and lepidocrocite [26,84,[89][90][91].
pipette, and drying the film of solids in a glove box. Specimens were briefly (<30 s) exposed to air during transfer to the Hitachi S-4700-II FEG-SEM.
Transmission Mössbauer spectroscopy was performed with a variable temperature He-cooled system with a 1024 channel detector. The 57 Co source used (~50 mCi) was in a Rh matrix at room temperature. All center shifts reported are relative to an α-Fe foil at room temperature. Samples were prepared by filtering the cell suspension (approx. 4 mL) in an anoxic glove box with recoverable filter paper. The filter paper was then sealed between two pieces of 5 mm Kapton tape to avoid oxidation while the sample was mounted. No indication of inadvertent oxidation was observed. Spectral fitting was done using Recoil Software (version 1.01998, University of Ottawa, Ottawa, ON, Canada). Voigt-based fitting was used to model the spectra to determine the hyperfine parameters and the relative areas between phases. The Lorentzian linewidth was held at 0.12 mm·s −1 , as it was the linewidth measured on the spectrometer for an ideally thick α-Fe foil. The relative peak areas (1:1 for doublets, 3:2:1:1:2:3 for sextets) were held constant throughout fitting. Each phase was fitted with only a single component (i.e., multiple QS and H distributions were not allowed for a single phase in fitting).

Fe(III) Oxide Bioreduction
The rate and extent of bioreduction of the Fe(III) oxides varied substantially based on mineralogy ( Figure 1 and Table 1). Within ~80 days, only 3% (on a mol Fe(III) basis) of hematite and 10% of goethite were reduced, compared to 58% of maghemite, 57% of ferrihydrite, and 49% of ferric green rust. The intermediate reduction of 36%, 32%, and 30% of added Fe(III) occurred for akaganeite, lepidocrocite, and feroxyhyte, respectively. The rate of Fe(III) oxide bioreduction was poorly correlated with oxide surface area ( Figure 2). These results are consistent with previous studies showing that highly crystalline phases, such as hematite and goethite, are less bioavailable for microbial iron reduction than less crystalline phases, such as ferrihydrite and lepidocrocite [26,84,[89][90][91].   a At ~80 days after inoculation. b Fe(II) production rates were calculated by linear regression using least-squares regression of the data during the period of maximum sustained Fe(II) production. c Data for akaganeite are from O'Loughlin et al. [67]. d The value for the surface area of ferrihydrite is from Roden [90].

Hematite
Within the first 24 h of incubation, ~1 mM Fe(II) was produced from the bioreduction of hematite (Figure 3), followed by a more gradual increase in Fe(II) to 2.44 mM by day 78. A final measurement was made 928 days after inoculation, at which point 6.9 mM Fe(II) was produced, corresponding to a reduction of 8.6% of the hematite. The presence of the electron shuttle AQDS had essentially no effect on the rate of hematite bioreduction (Table 2); however, over time the extent of Fe(II) production in the AQDS-amended system was significantly greater. The presence of both AQDS and 500 µ M phosphate resulted in a significant increase in the initial rate of hematite bioreduction and higher levels of Fe(II) through the initial 74 days of the incubation, although by day 744 there was essentially no difference between either AQDS-amended system.

Hematite
Within the first 24 h of incubation,~1 mM Fe(II) was produced from the bioreduction of hematite (Figure 3), followed by a more gradual increase in Fe(II) to 2.44 mM by day 78. A final measurement was made 928 days after inoculation, at which point 6.9 mM Fe(II) was produced, corresponding to a reduction of 8.6% of the hematite. The presence of the electron shuttle AQDS had essentially no effect on the rate of hematite bioreduction (Table 2); however, over time the extent of Fe(II) production in the AQDS-amended system was significantly greater. The presence of both AQDS and 500 µM phosphate resulted in a significant increase in the initial rate of hematite bioreduction and higher levels of Fe(II) through the initial 74 days of the incubation, although by day 744 there was essentially no difference between either AQDS-amended system.    Analysis of the solids in the hematite bioreactors by pXRD initially showed no evidence of crystalline secondary minerals ( Figure 4). Following long-term incubation, indications of siderite are evident in all the treatments. Small, imperfect rhombohedral (pseudocubic) crystals were observed in the solids ( Figure 5). The crystals were 200-500 nm on the edge and were dispersed as single crystals in a hematite groundmass. The overall morphology of these crystals is similar to that of siderite formed during the bioreduction of Fe(III) oxides [17,[92][93][94]. Previous studies of the bioreduction of hematite have reported the formation of magnetite and vivianite as secondary minerals [22,95,96]; however, there was no indication that either of these phases formed in any of our hematite bioreduction systems.

Goethite
Within the first seven days after inoculation, the rate of Fe(II) production during goethite bioreduction was 0.60 ± 0.09 µM·day −1 , after which the rate decreased considerably ( Figure 6). The presence of AQDS increased the initial rate of Fe(II) production to 1.05 ± 0.28 µM·day −1 , and throughout the experiment, the amount of Fe(II) was greater compared to the unamended system ( Figure 6 and Table 2). The addition of 500 µM phosphate had no effect on the initial rate of Fe(II) production; however, after day 8, the Fe(II) levels remained essentially unchanged.
Although siderite and vivianite have been observed as secondary minerals, resulting from the bioreduction of goethite [84,93,97,98], analysis of the solids in the goethite bioreactors by pXRD provided no conclusive evidence of crystalline secondary minerals ( Figure 7). However, there is a small peak at~34 • 2θ in the diffraction pattern of the solids from both the unamended and AQDS-amended bioreactors at 1060 days. This peak corresponds to a prominent chukanovite peak in the diffraction patterns of secondary minerals, formed during the bioreduction of akaganeite, lepidocrocite, and maghemite by S. putrefaciens CN32 under conditions similar to those in the goethite bioreactors in this study [31,66,67]. 57 Fe Mössbauer analysis of the solids in both systems indicated~20% chukanovite (Table 3). In addition, SEM imaging of the solids in bioreactors with and without AQDS amendment show platy crystallites with a morphology consistent with that of biogenic chukanovite [21,41,66], among residual goethite crystals ( Figure 8B,C). There was no evidence of secondary mineral formation in the goethite bioreactors amended with both AQDS and phosphate; however, the goethite did become more crystalline (Figure 7), and goethite crystallite morphology changed from acicular to lath-like ( Figure 8A,D).

Goethite
Within the first seven days after inoculation, the rate of Fe(II) production during goethite bioreduction was 0.60 ± 0.09 µ M·d -1 , after which the rate decreased considerably (Figure 6). The presence of AQDS increased the initial rate of Fe(II) production to 1.05 ± 0.28 µM·d −1 , and throughout the experiment, the amount of Fe(II) was greater compared to the unamended system ( Figure 6 and Table 2). The addition of 500 µ M phosphate had no effect on the initial rate of Fe(II) production; however, after day 8, the Fe(II) levels remained essentially unchanged.
Although siderite and vivianite have been observed as secondary minerals, resulting from the bioreduction of goethite [84,93,97,98], analysis of the solids in the goethite bioreactors by pXRD provided no conclusive evidence of crystalline secondary minerals (Figure 7). However, there is a small peak at ~34° 2θ in the diffraction pattern of the solids from both the unamended and AQDS-amended bioreactors at 1060 days. This peak corresponds to a prominent chukanovite peak in the diffraction patterns of secondary minerals, formed during the bioreduction of akaganeite, lepidocrocite, and maghemite by S. putrefaciens CN32 under conditions similar to those in the goethite bioreactors in this study [31,66,67]. 57 Fe Mössbauer analysis of the solids in both systems indicated ~20% chukanovite (Table 3). In addition, SEM imaging of the solids in bioreactors with and without AQDS amendment show platy crystallites with a morphology consistent with that of biogenic chukanovite [21,41,66], among residual goethite crystals ( Figure 8B,C). There was no evidence of secondary mineral formation in the goethite bioreactors amended with both AQDS and phosphate; however, the goethite did become more crystalline (Figure 7), and goethite crystallite morphology changed from acicular to lath-like ( Figure 8A,D).

Maghemite
Maghemite was readily reduced by CN32 ( Figure 9) with an initial rate of Fe(II) production of ~30 mM·day -1 . The presence of AQDS resulted in a 4-fold increase in the initial rate of Fe(II) production, but over time the overall extent of Fe(III) reduction was the same as the non-AQDS-amended system ( Table 2). The pXRD patterns of the solids in the AQDS-and AQDS+ systems at the end of the incubation were similar, with both showing remnant maghemite peaks, as well as peaks corresponding to chukanovite ( Figure 10). Likewise, 57 Fe Mössbauer analysis of the solids in both systems showed similar levels of chukanovite and magnetite (Table 3). Maghemite is isostructural with magnetite and is the fully oxidized end member of a solid solution with stoichiometric magnetite as the fully reduced end member. As such, maghemite can be reduced to magnetite via topotactic transformation. The transformation from maghemite to magnetite is apparent in the Mössbauer spectra, due to the appearance of an octahedral Fe 2.5+ signal, which has a center shift (~0.72 mm/s) that differs from the center shift for octahedral Fe 3+ and tetrahedral Fe 3+ (~0.38 mm/s). The octahedral Fe 2.5+ mixed valence signal is a consequence of electron hopping, occurring between octahedral sites at a rate substantially faster than the characteristic time of Mössbauer spectroscopy [99]. Although the magnetite largely retains the morphology of the parent maghemite crystallites, the edges are more irregular and there is significant pitting ( Figure 11A

Maghemite
Maghemite was readily reduced by CN32 ( Figure 9) with an initial rate of Fe(II) production of~30 mM·day −1 . The presence of AQDS resulted in a 4-fold increase in the initial rate of Fe(II) production, but over time the overall extent of Fe(III) reduction was the same as the non-AQDS-amended system ( Table 2). The pXRD patterns of the solids in the AQDS-and AQDS+ systems at the end of the incubation were similar, with both showing remnant maghemite peaks, as well as peaks corresponding to chukanovite ( Figure 10). Likewise, 57 Fe Mössbauer analysis of the solids in both systems showed similar levels of chukanovite and magnetite (Table 3). Maghemite is isostructural with magnetite and is the fully oxidized end member of a solid solution with stoichiometric magnetite as the fully reduced end member. As such, maghemite can be reduced to magnetite via topotactic transformation. The transformation from maghemite to magnetite is apparent in the Mössbauer spectra, due to the appearance of an octahedral Fe 2.5+ signal, which has a center shift (~0.72 mm/s) that differs from the center shift for octahedral Fe 3+ and tetrahedral Fe 3+ (~0.38 mm/s). The octahedral Fe 2.5+ mixed valence signal is a consequence of electron hopping, occurring between octahedral sites at a rate substantially faster than the characteristic time of Mössbauer spectroscopy [99]. Although the magnetite largely retains the morphology of the parent maghemite crystallites, the edges are more irregular and there is significant pitting ( Figure 11A,B). Bladed chukanovite crystallites are evident in both AQDS-and AQDS+ systems ( Figure 11C,D).
The effect of phosphate concentration (0-10 mM) on maghemite bioreduction was observed in AQDS-amended systems. Phosphate concentration had essentially no effect within the first 23 h of incubation ( Figure 9A). Fe(II) levels in the 10 mM phosphate system were higher from day 1.5 through day 3 than those in the systems with 1 mM phosphate or less; however, from day 31 through to the end of the incubation, the Fe(II) concentration in the 10 mM phosphate system was significantly lower (~40 mM versus 55-60 at day 184). Chukanovite and magnetite were identified as secondary minerals in the 0, 0.01, 0.1, and 1 mM phosphate systems; however, the 10 mM phosphate system contained vivianite and magnetite (Table 3, and Figures 10 and 11); the formation of magnetite in the 0.01 and 0.1 mM phosphate systems was not directly observed but inferred from the Mössbauer analysis of the 0, 1, and 10 mM phosphate systems (Table 3)

Ferrihydrite
The bioreduction of ferrihydrite produced~27 mM Fe(II) within 24 h, after which the Fe(II) concentrations remained essentially stable until the end of the incubation ( Figure 12). Analysis of the solids by pXRD 46 days after inoculation indicated the presence of magnetite and FHC ( Figure 13). The magnetite was in the form of aggregated, nominally cubic crystallites~10 nm ( Figure 14A), consistent with the nano-scale magnetite commonly observed as a secondary mineral of ferrihydrite bioreduction [17,29,100]. Irregular, bladed chukanovite crystals were dispersed among the magnetite groundmass ( Figure 14B). The initial rate of Fe(II) production was nearly double in the presence of AQDS (Table 2) and the overall extent of Fe(II) production at the end of the incubation was substantially greater than in the absence of AQDS (44 mM versus 29 mM). Ferrihydrite was transformed to magnetite and chukanovite ( Figures 13 and 14C), with chukanovite as the dominant secondary mineral (Table 2). Although the initial rate of Fe(II) production was lower in the presence of 1 mM phosphate, overall, phosphate concentrations from 0.01-1 mM had little effect on the extent of ferrihydrite bioreduction ( Figure 12). Magnetite and chukanovite were observed as secondary minerals over this range of phosphate concentrations; however, the phases became less crystalline with increasing phosphate concentration ( Figure 13) and the chukanovite crystallites became increasingly irregular ( Figure 14D-F). The addition of 10 mM phosphate resulted in a biphasic Fe(II) production profile. After the initial 6 h period of rapid Fe(II) production (71.8 ± 3.4 mM·day −1 ), a rate of 1.58 ± 0.05 mM·day −1 was sustained from day 1 through 23, resulting in the accumulation of~60 mM Fe(II). At this phosphate concentration, green rust and vivianite formed as secondary minerals ( Figure 13 and Table 2); magnetite and chukanovite were not evident. The green rust crystallites were platy and nominally hexagonal ( Figure 14G,H) but did not have the well-defined edges often reported for biogenic green rust [17,18,66,101]. The vivianite was present as tabular crystallites up to 25 µm in the longest dimension and occasionally as twins ( Figure 14G,H), consistent with previous reports of biogenic vivianite resulting from the bioreduction of ferrihydrite [19,66,92,94,102]. ) Initial (<7 days) Fe(II) production during the bioreduction of maghemite by S. putrefaciens CN32 and 84 days. Error bars indicate one standard deviation.

Lepidocrocite
The presence of AQDS significantly enhanced the rate of Fe(II) production during lepidocrocite bioreduction, but resulted in only a minor increase in the overall extent of Fe(II) production ( Figure 15 and Table 4):~80% of Fe(II) production occurred within 0.9 days in the presence of AQDS but it took 4.8 days to achieve the same extent in the absence of AQDS. pXRD analysis of the solids remaining in both systems at the end of the incubation showed the formation of magnetite and no indication of residual lepidocrocite ( Figure 16); however, SEM imaging of the solids showed both fine-grained magnetite (~50 nm) and bladed micron-sized chukanovite crystallites ( Figure 17B-D). Magnetite and chukanovite are commonly reported products of the microbial reduction of lepidocrocite [18,26,27,31,41,66,103], but there were no indications of less commonly observed secondary minerals, such as siderite or goethite [97]. Initial (<2 days) Fe(II) production during the bioreduction of ferrihydrite by S. putrefaciens CN32 of 46 days. Error bars indicate one standard deviation.

Lepidocrocite
The presence of AQDS significantly enhanced the rate of Fe(II) production during lepidocrocite bioreduction, but resulted in only a minor increase in the overall extent of Fe(II) production ( Figure 15 and Table 4): ~80% of Fe(II) production occurred within 0.9 of AQDS. pXRD analysis of the solids remaining in both systems at the end of the incubation showed the formation of magnetite and no indication of residual lepidocrocite (Figure 16); however, SEM imaging of the solids showed both fine-grained magnetite (~50 nm) and bladed micron-sized chukanovite crystallites ( Figure 17B-D). Magnetite and chukanovite are commonly reported products of the microbial reduction of lepidocrocite [18,26,27,31,41,66,103], but there were no indications of less commonly observed secondary minerals, such as siderite or goethite [97].   In a previous study, we observed that the presence of 500 µM phosphate had a substantial effect on the rate and extent of Fe(II) production and the formation of secondary minerals during the bioreduction of lepidocrocite by CN32 [31]. In this study, we examined the effects of a broad range of phosphate concentrations on lepidocrocite bioreduction in systems amended with AQDS. The addition of either 1 µM or 10 µM phosphate had little effect on the Fe(II) production profiles compared to the unamended system; the rates of Fe(II) production and total Fe(II) concentrations were essentially the same ( Figure 15 and Table 4). Likewise, magnetite and chukanovite were the only observed secondary minerals (Figures 16 and 17D). The addition of 100 µM phosphate resulted in an order of magnitude decrease in the rate of Fe(II) production (2.0 ± 0.1 mM·day −1 , compared to 29.9 ± 1.8 mM·day −1 in the 10 µM phosphate system), but the final Fe(II) concentration was nearly double (55.2 ± 1.2 mM relative to 31.1 ± 0.9 mM in the 10 µM phosphate system). Green rust was the only secondary mineral evident in the pXRD pattern of the solids remaining at the end of the incubation ( Figure 16) and there was no indication of the formation of any transient or intermediate secondary minerals during bioreduction ( Figure 18). The hexagonal, platy green rust crystallites had a morphology ( Figure 17F) consistent with that of biogenic green rust [17,18,104]. These results are essentially the same as what we previously observed in an identical experimental system that was amended with 500 µM phosphate [31]. Increasing the phosphate concentration from 100 µM to 1 mM resulted in a significant increase in the rate of Fe(II) production and a small increase in the final Fe(II) concentration (Table 4), and green rust was the only observed secondary mineral (Figures 16 and 17G). The addition of 10 mM phosphate resulted in a biphasic Fe(II) production profile ( Figure 15). After an initial pulse of 11.0 ± 0.8 mM Fe(II) within the first 0.9 days, Fe(II) production continued at a sustained rate of 1.9 ± 0.1 mM·day −1 through until day 30, ultimately leading to the accumulation of 71.4 ± 4.3 mM Fe(II) at the end of incubation. At this phosphate concentration, both green rust and vivianite formed as secondary minerals (Figures 16 and 17H).   Given that there was a clear transition in the rate and extent of Fe(II) production, as well as the formation of secondary minerals between 10 and 100 µM phosphate, we examined lepidocrocite bioreduction over this range of concentrations at a finer scale. The Fe(II) production profiles (rate and extent) and secondary mineral formation (magnetite and chukanovite) were similar for 0, 10, and 25 µM phosphate-amended systems ( Figure 19 and Table 4). The results for the systems with 50 µM phosphate amendment were similar, except for a short initial lag and slightly higher Fe(II) concentration at the end of the incubation. The addition of 75 µM phosphate resulted in distinctly different outcomes between the duplicate bottles. Replicate 1 exhibited an initial lag in Fe(II) production that was followed by a relatively lower rate of Fe(II) production (compared to the 0, 10, 25, and 50 µM treatments); at the end of the incubation, replicate 1 had a slightly higher Fe(II) concentration, and magnetite and chukanovite were the only secondary minerals identified by pXRD and SEM (Figures 20 and 21B). However, the Fe(II) production profile for replicate 2 of the 75 µM phosphate treatment was distinctly different from replicate 1. The rate of Fe(II) production was much slower and was sustained over a longer period of time, similar to the 100 µM treatment. As with the 100 µM treatment, replicate 2 had a higher final Fe(II) concentration, and green rust was the only secondary mineral observed by pXRD and SEM (Figures 20 and 21C). These results suggest a narrow tipping point with respect to phosphate concentration and lepidocrocite bioreduction kinetics, extent, and secondary mineral formation. Below the critical phosphate concentration, Fe(II) production rates are "fast" and magnetite and chukanovite form as secondary minerals, whereas above the critical phosphate concentration, Fe(II) production is slower, the overall extent is greater, and green rust is observed as the sole secondary mineral.

Feroxyhyte
The bioreduction of feroxyhyte was essentially complete within 11 days after inoculation ( Figure 22). The rate of Fe(II) production was 5 times faster with the addition of AQDS; however, both AQDS-and AQDS+ systems contained~24 mM Fe(II) at the end of the incubation (Table 4). A highly crystalline magnetite, consisting of 50-100 nm nominally cubic crystallites and well-formed bladed chukanovite crystallites, formed in both treatments (Figures 23 and 24B-D), with no indications of other secondary minerals or residual feroxyhyte ( Figure 23 and Table 5). There is little information available on secondary minerals resulting from feroxyhyte bioreduction; however, the formation of magnetite in our study is similar to the formation of magnetite during feroxyhyte bioreduction by Geobacter sulfurreducens [26].
The addition of 500 µM phosphate greatly inhibited the kinetics of feroxyhyte bioreduction-0.6 ± 0.01 mM Fe(II) day −1 compared to 23.3 ± 3.2 mM Fe(II) day −1 in the system without phosphate-however, the amount of Fe(II) produced by the end of the incubation was two-times greater (Table 4). Magnetite and chukanovite both formed as secondary minerals (Figure 23), but chukanovite was the dominant product (Table 5) and present as poorly formed, irregular crystals ( Figure 24E). Increasing the phosphate concentration to 5 mM resulted in a similar bioreduction profile with respect to Fe(II) production kinetics and final Fe(II) concentration ( Figure 22 and Table 4), but in place of magnetite and chukanovite, vivianite and green rust formed as secondary minerals (Figure 23), both of which were poorly developed (i.e., irregular crystallites not exhibiting typical morphologies for green rust or vivianite) ( Figure 24F).

Ferric Green Rust
Ferric green rust was readily reduced by CN32 ( Figure 25), with an initial Fe(II) production rate of~19 mM·day −1 . The presence of AQDS resulted in a nearly 5-fold increase in the initial rate of Fe(II) production (~91 mM·day −1 ), but over time the overall extent of Fe(III) reduction was the same as the non-AQDS-amended system ( Figure 25 and Table 4). The addition of 500 µM phosphate had no effect on either the rate or extent of Fe(II) production; however, the maximum rate of sustained Fe(II) production decreased to~1 mM Fe(II) day −1 and the final Fe(II) concentration increased by >70% in the 5 mM phosphate-amended system ( Figure 25 and Table 4).
The bioreduction of ferric green rust by CN32 resulted in the formation of magnetite and chukanovite in all but the 5 mM phosphate system (Table 4). In the absence of added phosphate, highly crystalline magnetite formed in both the AQDS-and AQDS+ systems (Figure 26), and the crystallites were larger (50-250 nm) than the magnetite that formed in any of the other Fe(III) oxide systems we examined ( Figure 27B-D). A somewhat less crystalline, finer-grained magnetite was observed in the system amended with 500 µM phosphate ( Figure 27E). There was little indication of chukanovite in the diffraction patterns of the solids; however, SEM imaging of the solids showed well-formed bladed chukanovite crystals ( Figure 27C-E) and Mössbauer analysis indicated that~15% and~27% of the iron in the solids was present as chukanovite in the AQDS+ systems containing 0 and 500 µM phosphate, respectively (Table 5). These results are in contrast to previous studies of the bioreduction of ferric green rust that have reported green rust as the only secondary mineral [66,101]. In the 5 mM phosphate-amended system, pXRD and Mössbauer analysis indicated the transformation of ferric green rust to green rust ( Figure 26 and Table 5). The green rust crystals were substantially larger than the parent ferric green rust (in some cases, >5 µm across) and in some instances exhibiting the terracing of the layers ( Figure 27F

Feroxyhyte
The bioreduction of feroxyhyte was essentially complete within 11 days after inoculation ( Figure 22). The rate of Fe(II) production was 5 times faster with the addition of AQDS; however, both AQDS-and AQDS+ systems contained ~24 mM Fe(II) at the end of the incubation (Table 4). A highly crystalline magnetite, consisting of 50-100 nm nominally cubic crystallites and well-formed bladed chukanovite crystallites, formed in both treatments (Figures 23 and 24B-D), with no indications of other secondary minerals or residual feroxyhyte ( Figure 23 and Table 5). There is little information available on secondary minerals resulting from feroxyhyte bioreduction; however, the formation of magnetite in our study is similar to the formation of magnetite during feroxyhyte bioreduction by Geobacter sulfurreducens [26].
The addition of 500 µ M phosphate greatly inhibited the kinetics of feroxyhyte bioreduction-0.6 ± 0.01 mM Fe(II) day −1 compared to 23.3 ± 3.2 mM Fe(II) day −1 in the system without phosphate-however, the amount of Fe(II) produced by the end of the incubation was two-times greater (Table 4). Magnetite and chukanovite both formed as secondary minerals (Figure 23), but chukanovite was the dominant product (Table 5) and present as poorly formed, irregular crystals ( Figure 24E). Increasing the phosphate concentration to 5 mM resulted in a similar bioreduction profile with respect to Fe(II) production kinetics and final Fe(II) concentration ( Figure 22 and Table 4), but in place of magnetite and chukanovite, vivianite and green rust formed as secondary minerals (Figure 23), both of which were poorly developed (i.e., irregular crystallites not exhibiting typical morphologies for green rust or vivianite) ( Figure 24F).

Ferric Green Rust
Ferric green rust was readily reduced by CN32 (Figure 25), with an initial Fe(II) production rate of ~19 mM·day -1 . The presence of AQDS resulted in a nearly 5-fold increase in the initial rate of Fe(II) production (~91 mM·day −1 ), but over time the overall extent of Fe(III) reduction was the same as the non-AQDS-amended system ( Figure 25 and Table  4). The addition of 500 µ M phosphate had no effect on either the rate or extent of Fe(II) production; however, the maximum rate of sustained Fe(II) production decreased to ~1 mM Fe(II) day -1 and the final Fe(II) concentration increased by >70% in the 5 mM phosphate-amended system ( Figure 25 and Table 4).
The bioreduction of ferric green rust by CN32 resulted in the formation of magnetite and chukanovite in all but the 5 mM phosphate system (Table 4). In the absence of added phosphate, highly crystalline magnetite formed in both the AQDS-and AQDS+ systems (Figure 26), and the crystallites were larger (50-250 nm) than the magnetite that formed in any of the other Fe(III) oxide systems we examined ( Figure 27B-D). A somewhat less crystalline, finer-grained magnetite was observed in the system amended with 500 µ M phosphate ( Figure 27E). There was little indication of chukanovite in the diffraction patterns of the solids; however, SEM imaging of the solids showed well-formed bladed chukanovite crystals ( Figure 27C-E) and Mössbauer analysis indicated that ~15% and ~27% of the iron in the solids was present as chukanovite in the AQDS+ systems containing 0 and 500 µ M phosphate, respectively (Table 5). These results are in contrast to previous studies of the bioreduction of ferric green rust that have reported green rust as the only secondary mineral [66,101]. In the 5 mM phosphate-amended system, pXRD and Mössbauer analysis indicated the transformation of ferric green rust to green rust ( Figure  26 and Table 5). The green rust crystals were substantially larger than the parent ferric green rust (in some cases, >5 µ m across) and in some instances exhibiting the terracing of the layers (Figure 27F-H). Table 5. Fit parameters from Mössbauer analysis of the secondary minerals in the indicated systems.

Fe(II)-Secondary Mineral Formation in Relation to Parent Fe(III) Oxide in Unamended Systems
In our experimental systems, the bioreduction of Fe(III) oxides by S. putrefaciens CN32 resulted in the formation of siderite, magnetite, chukanovite, green rust, or vivianite, depending on the experimental conditions. Except for hematite and goethite, magnetite was the dominant secondary mineral observed in the absence of added phosphate. Among the Fe(III) oxides we examined, hematite and goethite had the lowest levels of bioreduction, consistent with previous studies also showing the diminished bioreduction of hematite and goethite, compared to phases such as ferrihydrite, lepidocrocite, akaganeite, and feroxyhyte [19,26,84,91,105,106].
Relative to other Fe(III) oxides, secondary mineral formation during hematite bioreduction has not been widely studied. Behrends and Van Cappellen [22] reported the formation of magnetite and siderite following hematite bioreduction, although magnetite was only observed at bicarbonate concentrations > 5 mM. They proposed that the formation of magnetite is dependent on the relative rates of Fe(II) and Fe(III) supply in the aqueous phase, with the rate of soluble Fe(II) production, determined by the bioreduction rate and the rate of soluble Fe(III) formation, determined by the non-reductive dissolution of the hematite-the latter increasing with increasing bicarbonate concentrations. However, Luo et al, [96] examined hematite bioreduction at <0.1 mM bicarbonate and observed only magnetite. In our experimental system, siderite was the only secondary mineral observed. Although we did not measure bicarbonate concentrations, we can estimate that our system contained at least 3.5 mM carbonate, based on the production of~7 mM Fe(II) and the stoichiometry of the reduction of Fe(III) coupled with formate oxidation (2Fe(III) + HCO 2 − + H 2 O) → 2Fe(II) + HCO 3 − + 2H + ). Therefore, bicarbonate is not the only control determining magnetite formation during the bioreduction of hematite.
As with hematite, studies of secondary mineral formation during goethite bioreduction are rather limited. Although siderite formation is commonly observed under high bicarbonate concentrations (e.g., 30 mM) [84,93,94], it has also been observed in systems containing as low as~2 mM bicarbonate [97]. We did not observe siderite in our goethite bioreduction experiment, even though at least 11 mM of bicarbonate was generated, based on the amount of Fe(II) produced; rather, chukanovite formed, despite the fact that chukanovite should be metastable, relative to siderite, under our conditions [107]. The formation of chukanovite has been reported in studies of the bioreduction of akaganeite, ferrihydrite, lepidocrocite, and maghemite [21,27,31,41,66,67,108] (as well as for feroxyhyte and ferric green rust, as observed in this study), but not goethite. In these previous studies, magnetite or green rust initially formed as secondary minerals, and chukanovite formation was at their expense, making it essentially a tertiary mineral. In our study, chukanovite truly appears to be a secondary mineral of goethite bioreduction, as no intermediary phases were observed.
Magnetite and chukanovite were secondary minerals of the bioreduction of the remaining Fe(III) oxides (Tables 2 and 4). Magnetite is commonly reported as a product of microbial reduction of akaganeite [20,21,26,67], ferrihydrite [16,17,28,29,100,[109][110][111][112], lepidocrocite [26,27,31,41,97,104], feroxyhyte [26], and maghemite [66]. The reductive transformation of Fe(III) oxides to magnetite can occur by either solid-state transformation (topotactic) or dissolution-reprecipitation (reconstruction) mechanisms, depending on the parent Fe(III) oxide and the experimental conditions. In our experimental system, the transformation of poorly crystalline maghemite to poorly crystalline magnetite was clearly topotactic, given that magnetite is isostructural with maghemite and there was no change in crystallite morphology ( Figure 11A,B; note that the pitting evident on the magnetite crystallites was likely due to continued bioreduction). Ferrihydrite is characterized by small crystal size (<10 nm) and the formation of magnetite of similar size and morphology during ferrihydrite bioreduction has been attributed to topotactic transformation [17,29]. However, magnetite can also form by reconstructive processes [113] that can involve the formation of intermediate Fe phases [114], which result in magnetite specimens having sizes/morphologies different from ferrihydrite. In our experimental system, ferrihydrite transformation resulted in a poorly crystalline and very fine-grained (~10 nm) magnetite, consistent with a topotactic transformation. However, the magnetite specimens formed during the bioreduction of lepidocrocite, feroxyhyte, and ferric green rust were highly crystalline and likely produced by a dissolution-reprecipitation process, given how dissimilar the morphologies of these magnetites are from their parent Fe(III) oxides. The lepidocrocite used in our study consisted of elongated lath-like crystallites~10 nm wide and up to 40 nm long that were transformed to nominally cubic magnetite~50 nm in size. The thin, nominally hexagonal 25-75 nm wide feroxyhyte crystallites were bioreduced to 50-100 nm cubic and highly crystalline magnetite. The bioreduction of thin, nominally hexagonal ferric green rust crystallites up to 200 nm across were reduced to cubic/octahedral magnetite ranging in size from 50 to 250 nm.
The formation of magnetite as a secondary mineral of Fe(III) oxide bioreduction has often been linked to "high" rates of Fe(III) reduction for topotactic transformation, or "high" rates of soluble Fe(II) and Fe(III) supply for transformation by dissolutionprecipitation [22,28,29] and our results are consistent with these observations. In our study, magnetite did not form during hematite or goethite bioreduction, which had maximum Fe(II) production rates of 0.17 and 0.60 mM Fe(II) day −1 , respectively. Conversely, magnetite formed during the bioreduction of all of the other Fe(III) oxides examined, which had maximum Fe(II) production rates, ranging from 4.4 to 62.2 mM Fe(II) day −1 . Magnetite also formed during the bioreduction of akaganeite in a separate study (13.7 mM Fe(II) day −1 ), under conditions identical to those in this study [67].

Impact of AQDS on Fe(II) Secondary Mineral Formation
The presence of AQDS has been shown to enhance the rate and often the extent of the bioreduction of a wide range of Fe(III) oxides [17,26,30,32,45,46,55,84,[115][116][117][118]. In our experimental systems AQDS-enhancement of the rate of Fe(II) production ranged from 11.8% for hematite to 430% for feroxyhyte. However, the presence of AQDS did not always lead to an increase in the overall extent of Fe(II) production, as there was no statistically signifi-cant difference between AQDS-and AQDS+ treatments in the bioreduction of maghemite, feroxyhyte, and ferric green rust.
Although the presence of electron shuttles almost always results in the enhanced microbial reduction of Fe(III) oxides, there does not appear to be a consistent effect of electron shuttles on secondary mineral formation. Fredrickson et al. [17] and Zachara et al. [64,84] reported that AQDS affects both the types of secondary minerals and their relative crystallinity. Cutting et al. [26] found differences between AQDS-and AQDS+ systems in the relative amounts of goethite and magnetite formed during the bioreduction of akaganeite, lepidocrocite, and schwertmannite. In a previous study, we found that the bioreduction of akaganeite in the presence of AQDS resulted in the formation of a more crystalline magnetite than in its absence. Moreover, most of the magnetite formed in the AQDS+ system was ultimately transformed to chukanovite, but only a minor amount of the magnetite in the AQDS-system was converted to siderite. Conversely, Zegeye et al. [32], O'Loughlin [55], and Coker et al. [117] reported that the presence or absence of AQDS had no effect on the formation of specific secondary minerals during the bioreduction of lepidocrocite or ferrihydrite. The apparent inconsistencies in the effect of AQDS on secondary mineral formation may be due to differences in experimental conditions other than the presence/absence of AQDS.
In this study, AQDS had little effect on the formation of specific secondary minerals; that is, the secondary minerals that formed during the bioreduction of a specific Fe(III) oxide were the same in the presence and in the absence of AQDS (Tables 2 and 4) and there was no apparent difference in their crystallinity. Previous studies identifying chukanovite as a product of Fe(III) oxide bioreduction have all been in the presence of AQDS [21,27,31,41,66,108]. However, we observe chukanovite formation even in the absence of AQDS, indicating that AQDS is not essential for chukanovite formation. The presence of AQDS appears to affect the abundance of chukanovite, as our Mössbauer analysis shows a greater proportion of chukanovite in the AQDS-amended goethite system and there are indications in the pXRD data that AQDS may have enhanced the reduction of magnetite to chukanovite in the ferrihydrite system. It is therefore likely that AQDS significantly affects SMP distribution only in cases when the rate of Fe(II) supply is the major factor in determining biomineral formation.

Phosphate Effects on Fe(II) Secondary Mineral Formation
The effect of phosphate on secondary mineral formation in our experimental systems varied depending on the parent Fe(III) oxide and the phosphate concentration. In the case of hematite, the presence of 500 µM phosphate had no apparent effect, as siderite was the only observed secondary mineral, with or without phosphate. This is in contrast to the study of Yan et al. [95], where the bioreduction of hematite by G. sulfurreducens led to vivianite formation; however, the Fe:PO 4 ratio was 1:2 (~2 mM Fe and 4 mM phosphate) compared to 160:1 in our study. Behrends and Van Cappellen reported that the presence of both 1 mM and 50 mM phosphate completely inhibited the formation of magnetite as a secondary mineral during the bioreduction of hematite (~32 mM Fe) by S. putrefaciens strain 200R, with the presumed formation of a ferrous phosphate phase in the system amended with 50 mM phosphate [22]. Similarly, vivianite, but not magnetite, formed during the bioreduction of hematite by S. putrefaciens CN32 [84]. However, magnetite formation during the bioreduction of hematite (167 mM Fe) by Shewanella oneidensis strain MR-1 was not inhibited in the presence of 3 mM phosphate, as both magnetite and vivianite were observed [96]. In our experimental system, chukanovite was the only secondary mineral that formed during goethite bioreduction in the presence or absence of 500 µM phosphate; however, there was a notable increase over time in the crystallinity of goethite in the presence of 500 µM phosphate that was accompanied by a change in crystallite morphology from acicular to lath-like. A study of the bioreduction of goethite (either 5, 10, or 100 mM Fe) by S. putrefaciens strain CN32 in the presence of 440 µM phosphate reported siderite as the only secondary mineral indicated by pXRD analysis [93]; however, there was no phosphate-free system for comparison. In a separate study, goethite bioreduction by S. putrefaciens CN32 in the presence of 4 mM phosphate resulted in the formation of siderite and vivianite (with vivianite forming first) compared to siderite alone in the absence of phosphate [84].
The presence of phosphate seemed to have little effect on the bioreduction of maghemite to magnetite. At phosphate concentrations ≤ 1 mM, magnetite and chukanovite were the only observed secondary minerals, with roughly equivalent amounts of magnetite and chukanovite in the phosphate-free system and the system with 1 mM phosphate. These results are consistent with a previous study showing that the presence of 0.21 wt% (as P) structural phosphate in maghemite had no effect on secondary mineral formation [66]. However, in the presence of 10 mM phosphate, magnetite and vivianite formed, with no indication of chukanovite. As the transformation of maghemite to magnetite is topotactic, it is perhaps not surprising that phosphate did not seem to inhibit magnetite formation, even as the phosphate loading on maghemite reached the sorption capacity ( Table 6). The formation of vivianite in the 10 mM phosphate system is consistent with the high concentration of phosphate remaining in solution (~9 mM) after phosphate sorption by maghemite.
As discussed in Section 4.1, magnetite and chukanovite were observed as secondary minerals during the bioreduction of ferrihydrite, lepidocrocite, feroxyhyte, and ferric green rust in the absence of phosphate. Several previous studies have shown a close association between the presence of phosphate and the formation of green rust at the expense of magnetite during the reduction of ferrihydrite, lepidocrocite, and akaganeite by IRB [17,31,40,[65][66][67] However, most of these studies have been binary in their approach; i.e., phosphate was either absent or present at a single concentration/loading, and even those that have examined phosphate effects at multiple concentrations offer limited insight as to the sorption density or aqueous concentration range over which phosphate inhibits magnetite formation; either because magnetite was inhibited even at the lowest phosphate concentration/loading examined or by the presence of other similar oxyanions (e.g., silicate) [39,40,66]. In our study, we observe clear phosphate concentration effects on the formation of magnetite versus green rust; magnetite and chukanovite formed at phosphate concentrations of ≤1 mM (ferrihydrite), <~100 µM (lepidocrocite), 500 µM (feroxyhyte and ferric green rust), while green rust, or green rust and vivianite, formed at phosphate concentrations of 10 mM (ferrihydrite), ≥100 µM (lepidocrocite), and 5 mM (feroxyhyte and ferric green rust).
Phosphate binds strongly to Fe(III) oxides through the formation of inner-sphere complexes [119] that are known to affect the stability and subsequent transformations of Fe(III) oxides (including biotic and abiotic reduction) [39,120,121]. As such, the extent of phosphate sorption to an Fe(III) oxide may have some bearing on the formation of specific secondary minerals. This is clearly the case for vivianite, a ferrous phosphate mineral. Vivianite is commonly reported in systems where phosphate loadings are high enough to maintain solution phase phosphate concentrations that support vivianite precipitation. For example, in our study, vivianite was only observed in systems where the equilibrium solution phase phosphate concentrations of >4 mM are expected (Table 6). It is less clear how phosphate sorption affects the formation of green rust versus magnetite. We observed a level of phosphate coverage for ferrihydrite, lepidocrocite, and feroxyhyte, below which magnetite (and chukanovite) formed and above which green rust (and vivianite) formed (Table 6); however, the threshold level was specific to each Fe(III) oxide. Borch et al. [39] examined secondary mineral formation during ferrihydrite bioreduction by S. putrefaciens CN32 under conditions of 0%, 50%, and 100% phosphate coverage. Both magnetite and green rust were observed at 0% and 50% coverage, but only green rust formed at 100% coverage. The differences between their study and ours are noteworthy in two respects. First, they observed green rust formation in the absence of phosphate, and second, both green rust and magnetite were observed in the same system. In our experiments, we never observed magnetite and green rust in the same system; indeed, magnetite formed in one replicate bottle in the system containing lepidocrocite and 75 µM phosphate, and green rust formed in the other (Table 4). It is also worth noting that the effect of the extent of phosphate coverage varied among the Fe(III) oxides; for example, phosphate coverage of nearly 70% capacity had no effect on the secondary minerals that formed in the hematite and goethite systems, while a completely different secondary mineral formed in the lepidocrocite system, with coverage as low as 7% capacity (Table 6). Furthermore, green rust was never observed as a product of maghemite bioreduction, even at nominal coverage > 100% of the calculated capacity ( Table 6). As we discuss in the next section, other factors, in addition to phosphate concentration, have been identified as potentially contributing to the formation of green rust during Fe(III) oxide bioreduction, and both our study and that of Borch et al. [39] did not examine phosphate effects in the absence of some of these other factors. As such, our results offer only partial insight into the effects of phosphate on secondary mineral formation.

Formation of Green Rust as a Secondary Mineral during Fe(III) Oxide Bioreduction
The formation of green rust as a secondary mineral during the bioreduction of Fe(III) oxides was first reported by Fredrickson et al. during the reduction of ferrihydrite by S. putrefaciens CN32 [17]. Since then, green rusts have been reported as products of the microbial reduction of Fe(III) oxides in laboratory-based studies under a wide range of experimental conditions [12,18,19,23,27,29,[31][32][33]36,[38][39][40][41]55,65,101,108,118,[122][123][124]. In addition to the presence/concentration of phosphate, other factors have been proposed to contribute to the formation of green rusts during Fe(III) oxide bioreduction, including the presence of other oxyanions (arsenate, silicate, molybdate, tungstate, etc.) [31,41]; the presence and nature of dissolved organic carbon (including humic substances and microbially produced extracellular polymeric materials) [31,33,38]; the species and population size of IRB [31][32][33]; the type and concentration of the electron donor [18,36,37]; the rate and extent of Fe(II) production [17,19,125]; the presence of electron shuttles [17]; the sorption of Fe(II) to the parent Fe(III) oxide [25]; the extent of aggregation of Fe(III) oxide particles [27]; Fe(III) oxide mineralogy (this study). Despite over 20 years of investigation, a definitive and comprehensive understanding of the key factor(s) and mechanisms of green rust formation during microbial Fe(III) oxide reduction remains elusive. However, the factors mentioned above do suggest some common themes; in particular, mineralogical characteristics of specific Fe(III) oxides, sorbate interactions, and bioreduction rate.
Green rusts form as secondary minerals during the microbial reduction of many Fe(III) oxides, including ferrihydrite ( [19,29,33,36,39,40,65,122,126] and this study), lepidocrocite ( [12,18,27,31,32,41,55,66,101,104,108,118,123] and this study), mixtures of ferrihydrite and lepidocrocite [23,38], akaganeite [67], magnetite [124], schwertmannite [127], and ferric green rust ( [66,101] and this study). Our study shows that green rust can also form during the bioreduction of feroxyhyte. However, green rusts have not been reported to form during the bioreduction of hematite, goethite, and maghemite. Usman et al. [128] have shown that green rust can be formed via the abiotic reductive transformation of goethite and hematite by Fe(II), which they attribute to careful control of the experimental conditions, which included the addition of Fe(II) at a level consistent with the 2:1 stoichiometric ratio of Fe(II):Fe(III) in green rust and the addition of NaOH to provide an OH − :Fe(III) ratio of 3 (final pH~6.3). Given the high levels of Fe(II) required for the transformation of goethite and hematite to green rust, it is perhaps not surprising that green rust has not been observed as a product of the bioreduction of either Fe(III) oxide, given the comparatively low levels (and perhaps slow kinetics) of Fe(II) that are typically reported ( [26,84,[89][90][91] and this study). Currently, there are no reports of the biotic or abiotic reduction of maghemite to green rust and it is unclear if this is because this topic has not been well studied or because the process does not happen; however, Etique et al. [124] have reported the formation of green rust as a secondary mineral during the bioreduction of magnetite, which suggest that it is the former, as maghemite is readily reduced to magnetite.
Several studies have proposed a link between the rate of Fe(II) production and the formation of green rust relative to other secondary minerals (e.g., magnetite and siderite) [17,19,28,31,32,125]. Indeed, a clear delineation was observed between "fast" Fe(II) production kinetics and magnetite formation versus green rust formation with "slow" Fe(II) production rates during the bioreduction of lepidocrocite [31,125]. In this study we did not observe a clear correlation between Fe(II) production rate and green rust formation. During the bioreduction of ferrihydrite, feroxyhyte, ferric green rust, and lepidocrocite, we observed rates of Fe(II) production from 0.6-100 mM Fe(II) d −1 in systems where magnetite formed, and from 0.7-72 Fe(II) d −1 in systems where green rust formed (Tables 2 and 4). As such, it appears that the Fe(II) production rate alone is not the determining factor for GR formation, but in some cases it may be an indicator of other processes having more direct control on secondary mineral formation.
The surfaces of Fe(III) oxides are effective sorbents for a wide range of ligands found in various natural and engineered environments, and several of the studies showing green rust formation during Fe(III) oxide bioreduction suggest that the presence of many of these ligands favors green rust formation, typically at the expense of magnetite. As already discussed, green rust formation has been linked to phosphate concentration. The presence of other oxyanions, such as silicate, arsenate, molybdate, and tungstate, also favor green rust formation [31,41]; however, borate, even at higher surface coverage, does not [31]. Likewise, there are differences in the effects of various organic ligands. Among low molecular mass organic acids, green rust formed in the presence of citrate, but not oxalate [31]. Among humic substances (a class of naturally occurring, chemically heterogeneous organic oligoelectrolytes, derived primarily from the decomposition of bacteria, algae, and higher plant material that are ubiquitous in aquatic and terrestrial environments), the formation of green rust was favored in the presence of humics with greater aromatic content and larger molecular weights, while magnetite formed in the presence of humics with higher levels of polypeptides and polysaccharides [31]. Bacteria produce EPS (a heterogeneous mixture of polysaccharides, proteins, lipids, and nucleic acids), which sorb to iron oxides, and EPS from IRB has been shown to favor the formation of green rust [31,33]. Furthermore, IRB themselves associate with Fe(III) oxide surfaces [129,130], and high cell numbers of IRB also tend to favor green rust formation [27,31,32]; however, the specific cell density, leading to green rust formation, seems to be species dependent [31,33].
Although there have been numerous studies examining the link between ligand sorption on Fe(III) oxides and green rust formation during microbial Fe(II) oxide reduction, the processes leading to the formation of green rust versus magnetite are still unclear. Green rust has been reported as a metastable intermediate in the abiotic formation of magnetite following the sorption of Fe(II) to lepidocrocite [131] or ferrihydrite [62,132,133] in aqueous suspensions. However, the presence of ligands such as phosphate or arsenate/arsenite either slowed the rate of green rust transformation to magnetite or completely inhibited it [62,134], consistent with the stabilization of green rusts by the sorption of organic and inorganic ligands [41,58,59,[135][136][137][138]. In contrast to these abiotic systems, green rust has not been observed as an intermediate phase during magnetite formation as a secondary mineral of the bioreduction of Fe(III) oxides. In our study, either green rust or magnetite formed; there was never the co-occurrence of the phases, even in the lepidocrocite system containing 75 µM phosphate, where one replicate bottle formed only green rust and the other formed only magnetite.
The formation of green rust during the bioreduction of Fe(III) oxides could occur via either a reconstructive (dissolution-reprecipitation) or a solid-state/topotactic (structural rearrangement) process. Mann et al. [62] and Hansen et al. [139] proposed that green rust formation occurs via the solid-state rearrangement of ferrihydrite following the sorption of Fe 2+ . Solid-state transformation was also invoked for the formation of green rust during the bioreduction of ferrihydrite in the presence of phosphate, for which the sorption of phosphate to ferrihydrite was proposed to inhibit magnetite formation [17,40]. However, the morphological differences between the nanoparticulate ferrihydrite and the micron-sized green rust crystals are not consistent with a purely topotactic or pseudomorphic process for ferrihydrite transformation to green rust, and a similar argument can be made for the transformations of akaganeite, feroxyhyte, and lepidocrocite. During the bioreduction of lepidocrocite in the presence of phosphate, O'Loughlin et al. [66] reported a progressive transition in particle morphology from lepidocrocite to green rust, consistent with an initial structural reorganization of lepidocrocite to proto green rust particles that then grow to achieve the characteristic platy, hexagonal morphology of typical green rust. The formation of green rust during the bioreduction of akaganeite, ferrihydrite, and feroxyhyte might proceed by a similar process involving the sorption of phosphate or other ligands.
As satisfying as it would be to propose a consistent model for the mechanism of green rust formation during Fe(III) oxide bioreduction, such an understanding remains elusive. Moreover, the conflicting and often contradictory data suggest that there are perhaps multiple pathways to green rust formation.

Environmental Relevance
Dynamic redox conditions are ubiquitous in aquatic and terrestrial systems across all geographic zones (polar, temperate, and tropical) and are evident in a diverse range of environments, including stratified lakes, lacustrine and marine sediments, floodplains and wetland environments, groundwater-surface water interaction zones, and many others. Throughout these varied environments, the activity of Fe(III)-reducing microbes is coupled with several major ecosystem processes, including the mineralization of organic matter and accompanying the release of CO 2 , the uptake and release of nutrients (e.g., N and P), and the mobility and transformations of contaminants. Moreover, it is largely due to microbially-driven Fe(III) reduction that Fe(II) is typically one of the most abundant reductants in aquatic and terrestrial environments under suboxic and anoxic conditions [140][141][142]. However, the reactivity of Fe(II) is highly dependent on its chemical speciation [143][144][145][146][147], so insights into the factors that influence the formation of specific Fe(II) phases has important implications for understanding Fe biogeochemistry and contaminant and nutrient transformations in natural and engineered environments. Among the Fe(II)-bearing secondary minerals resulting from microbial reduction of Fe(III) oxides, green rusts are particularly effective reductants for a wide range of contaminants of concern, including nitrate, chlorinated solvents, nitroaromatics, azo dyes, toxic metals, metallolids, and radionuclides [86,143,144,. Therefore, identifying the factors that lead to the formation of green rusts and other reactive Fe(II) phases can contribute to better water quality management in natural and engineered environments.