Ferrihydrite transformations in flooded paddy soils: rates, pathways, and product spatial distributions

Complex interactions between redox-driven element cycles in soils influence iron mineral transformation processes. The rates and pathways of iron mineral transformation processes have been studied intensely in model systems such as mixed suspensions, but transformation in complex heterogeneous porous media is not well understood. Here, mesh bags containing 0.5 g of ferrihydrite were incubated in five water-saturated paddy soils with contrasting microbial iron-reduction potential for up to twelve weeks. Using X-ray diffraction analysis, we show near-complete transformation of the ferrihydrite to lepidocrocite and goethite within six weeks in the soil with the highest iron(ii) release, and slower transformation with higher ratios of goethite to lepidocrocite in soils with lower iron(ii) release. In the least reduced soil, no mineral transformations were observed. In soils where ferrihydrite transformation occurred, the transformation rate was one to three orders of magnitude slower than transformation in comparable mixed-suspension studies. To interpret the spatial distribution of ferrihydrite and its transformation products, we developed a novel application of confocal micro-Raman spectroscopy in which we identified and mapped minerals on selected cross sections of mesh bag contents. After two weeks of flooded incubation, ferrihydrite was still abundant in the core of some mesh bags, and as a rim at the mineral–soil interface. The reacted outer core contained unevenly mixed ferrihydrite, goethite and lepidocrocite on the micrometre scale. The slower rate of transformation and uneven distribution of product minerals highlight the influence of biogeochemically complex matrices and diffusion processes on the transformation of minerals, and the importance of studying iron mineral transformation in environmental media.


Additional pore water analysis
: Pore water concentration data presented in Figure 1. Measurements were made by in situ probe measurement ( a ), DIMATOC carbon analyser ( b ), ICP-OES ( c ) or by ion chromatography ( d ). Columns headed with 1, 2 or 3 under each element denote measurements of samples from replicate microcosms, and dashes denotes measurements that were not taken. Data points below the limit of determination (LoD) are presented in the table as ≤LoD. The , where SD y is the standard deviation of the y intercepts of the regression line through the = 3 calibration standards and m is the gradient of the regression.
Eh (mV) a pH a DOC (mg L -1 ) b Fe (mM) c SO 4 Table S4: Additional pore water concentration data (not presented in Figure 1 and    Table S5: Element concentrations in mineral material that was extracted from the interior (inner and outer core) and untransformed exterior (mostly rim) of two-week-reacted mesh bags. All concentrations are expressed in mg of element per g of dissolved mineral. Relative difference is the percentage increase of concentration in the sample from the exterior compared to the sample from the interior. Data points below the limit of determination (LoD) are presented in the table as ≤LoD. The , where SD y is the standard deviation of the y intercepts of the = 3 regression line through the calibration standards and m is the gradient of the regression. Limits of detection vary for the same element because all data are normalised to the initial mass of dissolved mineral.  Figure S3: Photographs of a mineral aggregate immediately after removal from a mesh bag, following incubation in CS soil for twelve weeks. On the left, the whole aggregate is presented, and on the right, it is possible to see the cross section, indicating that yellow mineral transformation products were present throughout the cross-section. The orange-brown rim is visible as a covering on most of the mineral aggregate.

Photographs of mineral sampling at week twelve
Electronic Supplementary Information for Grigg et al.

XRD patterns
Powder X-ray diffraction measurements were performed on dried, crushed and homogenised mineral samples from each soil microcosm. Samples were resuspended in ethanol, transferred onto polished Si wafers without XRD background (711 cut, Sil'tronix Silicon Technologies, France), allowed to dry in place and closed under anoxic seal. Anoxic conditions are maintained for up to 3.5 hours, during which time the sample is measured between 5° and 70° 2θ with a step size of 0.02° 2θ for 4 s at each step (X5 Advance, Bruker, USA). Initial ferrihydrite wafers were prepared for measurement in the same way, but not closed under anoxic seal, and measured between 10° and 70° 2θ with a step size of 0.02° 2θ for 10 s at each step. Soil CS and BD were measured on mechanically milled soil which was gently packed into a holder, and measured between 5° and 70° 2θ with a step size of 0.02° 2θ for 6 s at each step. A knife was used in all measurements to reduce the detection of scattered radiation at low goniometer angles. Samples were analysed in Bragg-Brentano geometry using Cu Kα1 and Cu K α2 radiation (λ = 1.5418 Å, 40 kV, and 40 mA) and a high-resolution energy-dispersive 1D detector (LYNXEYE).
Rietveld quantitative phase analysis (QPA) was performed on TOPAS software (Version 5, Bruker, USA). Ferrihydrite was quantified using PONKCS 3 phase calibration of the synthetic ferrihydrite minerals that were used in the experiment. 4 The XRD pattern of the initial ferrihydrite samples are included in Figures S5 -S9. The PONKCS calibration was carried out on mixtures of 50% (w/w) ferrihydrite and 50% (w/w) corundum (measured in the same way as the pure ferrihydrite samples described above), to obtain an empirical description of the ferrihydrite diffraction pattern as a mass-calibrated hkl-phase. Ferrihydrite from the same synthesis batch was used in the experiments that finished at weeks one, two and six. Ferrihydrite from a second batch was used in microcosms that were sampled at the twelve-week timepoint. Other phases were only fit if they could be attributed to crystalline peaks that were observed in the diffraction patterns, and if their fitted area made up more than 1% of the total mineral abundance in the sample (except for lepidocrocite which produces a very strong characteristic (0 2 0) peak and could be confidently Gaussian curve fitting models. The models were fit to whole spectra, and therefore average the crystallite size across all crystal diffraction orientations.
Electronic Supplementary Information for Grigg et al. S11 Figure S4: Example Rietveld fitting for three XRD patterns from PT-T soil. The blue curve is the plot of the measured data, the red is the fitted model, the bold black line is the fitted background scatter, and the grey plot indicates the residual. All plots are normalised to the range of the respective raw data. Major diffraction peaks are assigned to minerals used in the Rietveld fitting (Lp for lepidocrocite, Gt for goethite and Qz for quartz), excluding ferrihydrite (Fh), and the QPA results are reported next to the plots to two significant figures. 'GoF refers to the goodness of fit of the model (see Table S6 for explanation and comparison). The labels 'rep. 1' and 'rep. 2' denote results from mesh bags that were incubated in replicate microcosms.

Electronic Supplementary Information for Grigg et al.
S12 Figure S5: XRD patterns of minerals in mesh bags from PT-T soil microcosms (including those plotted with more details of the Rietveld fitting in Figure S3). Raw data is plotted in blue and the Rietveld fit (total of all modelled phases) is plotted in red. The intensities of all spectra are normalised by the range. 'Initial Fh (no. 1)' refers to the starting material for microcosms that was removed from soil after one, two and six weeks and 'initial Fh (no. 2)' for material removed from soil after twelve weeks. 'Rep. 1' and 'rep. 2' denote results from mesh bags that were incubated in replicate microcosms. Major diffraction peaks are identified above the plots, excluding ferrihydrite.

Electronic Supplementary Information for Grigg et al.
S13 Figure S6: XRD patterns of minerals in mesh bags from PT-T soil microcosms (including those plotted with more details of the Rietveld fitting in Figure S3). Raw data is plotted in blue and the Rietveld fit (total of all modelled phases) is plotted in red. The intensities of all spectra are normalised by the range. 'Initial Fh (no. 1)' refers to the starting material for microcosms that was removed from soil after one, two and six weeks and 'initial Fh (no. 2)' for material removed from soil after twelve weeks. 'Rep. 1' and 'rep. 2' denote results from mesh bags that were incubated in replicate microcosms. Major diffraction peaks are identified above the plots, excluding ferrihydrite.

Electronic Supplementary Information for Grigg et al.
S14 Figure S7: XRD patterns of minerals in mesh bags from PT-T soil microcosms (including those plotted with more details of the Rietveld fitting in Figure S3). Raw data is plotted in blue and the Rietveld fit (total of all modelled phases) is plotted in red. The intensities of all spectra are normalised by the range. 'Initial Fh (no. 1)' refers to the starting material for microcosms that was removed from soil after one, two and six weeks and 'initial Fh (no. 2)' for material removed from soil after twelve weeks. 'Rep. 1' and 'rep. 2' denote results from mesh bags that were incubated in replicate microcosms. Major diffraction peaks are identified above the plots, excluding ferrihydrite.

Electronic Supplementary Information for Grigg et al.
S15 Figure S8: XRD patterns of minerals in mesh bags from PT-T soil microcosms (including those plotted with more details of the Rietveld fitting in Figure S3). Raw data is plotted in blue and the Rietveld fit (total of all modelled phases) is plotted in red. The intensities of all spectra are normalised by the range. 'Initial Fh (no. 1)' refers to the starting material for microcosms that was removed from soil after one, two and six weeks and 'initial Fh (no. 2)' for material removed from soil after twelve weeks. 'Rep. 1' and 'rep. 2' denote results from mesh bags that were incubated in replicate microcosms. Major diffraction peaks are identified above the plots, excluding ferrihydrite.

Electronic Supplementary Information for Grigg et al.
S16 Figure S9: XRD patterns of minerals in mesh bags from PT-T soil microcosms (including those plotted with more details of the Rietveld fitting in Figure S3). Raw data is plotted in blue and the Rietveld fit (total of all modelled phases) is plotted in red. The intensities of all spectra are normalised by the range. 'Initial Fh (no. 1)' refers to the starting material for microcosms that was removed from soil after one, two and six weeks and 'initial Fh (no. 2)' for material removed from soil after twelve weeks. 'Rep. 1' and 'rep. 2' denote results from mesh bags that were incubated in replicate microcosms. Major diffraction peaks are identified above the plots, excluding ferrihydrite.  S18 Figure S10: Exponential decay of ferrihydrite in mesh bags, as estimated by Rietveld fitting of XRD patterns, and fit for samples measured after one, two and six weeks. The exponential decay constants are abbreviated as k.

Additional secondary electron (SE) images
Figure S13: Secondary electron (SE) images of ferrihydrite before the incubation experiments. A and B are images of the ferrihydrite synthesis batch used in microcosms that were sampled after one, two and six weeks for bulk analysis. Images C and D are of the ferrihydrite synthesis batch used in microcosms that were sampled after twelve weeks for bulk analysis, and two weeks for Raman spatial analysis.
Electronic Supplementary Information for Grigg et al.

Reference materials used in Raman study
Reference spectra were the average of at least 1500 spectra, each collected for 8 s from independent locations on the same reference mineral sample, using the same instrument and optical set-up as used for the sample collections. Sample minerals were distributed on silicon wafers, using the same preparation method as used for XRD measurements. The ferrihydrite reference is the starting mineral from this experiment. The lepidocrocite, goethite and hematite reference spectra are measurement of minerals acquired from a commercial source (Bayferrox 943, Bayferrox 910, and Bayferrox 105M, respectively; Bayer, Germany; additional characterisation in ref 13 ; Figure 4G and Figure S18). The reference for laser damaged mineral was measured on a section of severely laser-damaged ferrihydrite.

Limitations of Raman spectroscopy in this study
The principle of Raman spectrometry is fundamentally quantitative but numerous factors can affect the quantification process. Firstly, measured Raman spectra can depend on specific features of the crystals or measurement set-up. For example, spectral peak intensity and integrated area can vary according to the specific properties of the crystals being measured and the crystal orientation with regard to the laser light. 14  S23 Table S7: Summary of mineral abundance estimates from µ-Raman component analyses and lepidocrocite particle statistics from Raman maps presented in Figure 4C, Figure 4E and Figures S14-S46        S53 Figure S46: Black and white representation of a 1 µm Raman spectral map of a mineral aggregate cross section from soil BD as presented in Figure S43.