Visualizing the iron atom exchange front in the Fe(II)-catalyzed recrystallization of goethite by atom probe tomography

Significance

The bioavailability of iron in the environment, and coupled metals, nutrients, and contaminants, depends on the stability of common Fe(III) minerals such as goethite (FeOOH) and hematite (Fe₂O₃). At redox boundaries, iron isotopic tracer studies suggest that interaction with aqueous Fe(II) creates dynamic conditions of atom exchange (AE). However, mechanistic models have not advanced beyond speculation, because of the challenges of mapping AE fronts recorded in isotopic distributions in individual nanoscale crystallites. Here we demonstrate successful use of 3D atom probe tomography for this purpose. The penetration depth, spatial heterogeneity, and ties to mineral defects are visualized, helping constrain mechanistic models and setting a precedent for detailed interrogation of iron redox cycling in the environment.

Abstract

The autocatalytic redox interaction between aqueous Fe(II) and Fe(III)-(oxyhydr)oxide minerals such as goethite and hematite leads to rapid recrystallization marked, in principle, by an atom exchange (AE) front, according to bulk iron isotopic tracer studies. However, direct evidence for this AE front has been elusive given the analytical challenges of mass-resolved imaging at the nanoscale on individual crystallites. We report successful isolation and characterization of the AE front in goethite microrods by 3D atom probe tomography (APT). The microrods were reacted with Fe(II) enriched in tracer ⁵⁷Fe at conditions consistent with prior bulk studies. APT analyses and 3D reconstructions on cross-sections of the microrods reveal an AE front that is spatially heterogeneous, at times penetrating several nanometers into the lattice, in a manner consistent with defect-accelerated exchange. Evidence for exchange along microstructural domain boundaries was also found, suggesting another important link between exchange extent and initial defect content. The findings provide an unprecedented view into the spatial and temporal characteristics of Fe(II)-catalyzed recrystallization at the atomic scale, and substantiate speculation regarding the role of defects controlling the dynamics of electron transfer and AE interaction at this important redox interface.

Redox-driven transformation and recrystallization of metal oxide minerals is a fundamentally important process in geochemistry, influencing metal solubility, bioavailability, and the fate and transport of contaminants (1). For example, at redox boundaries in soils and sediments, aqueous Fe(II) juxtaposed with ubiquitous Fe(III)-(oxyhydr)oxides accelerates phase transformations from higher-solubility, nominally more bioavailable solid forms of Fe(III) such as ferrihydrite, to lower-solubility forms such as goethite, lepidocrocite, and magnetite (2, 3). These transformations entail complete turnover of structural iron, and therefore can critically impact the speciation of coassociated heavy metals, pollutants, and nutrients in the environment through their incorporation or release.

More recently, and contrary to long-held views, it was discovered that structural iron in these more stable Fe(III)-(oxyhydr)oxide products continues to be turned over while in contact with aqueous Fe(II). Hence, the prospect of impacts on coassociated elements remains. Key supporting evidence comes from batch reactor studies using stable isotopic tracers; for example, Fe(III) minerals with iron isotopic compositions at natural abundance (NA) (i.e., 91.8% ⁵⁶Fe) were contacted with Fe(II)_aq selectively enriched in ⁵⁷Fe, thereby enabling mass-sensitive techniques to follow the mixing of solution and solid-phase iron reservoirs over time. In particular, without detectable changes in phase or crystallinity, this macroscopic approach suggests that aqueous Fe(II) can recrystallize goethite (4–6) and hematite (7, 8), in some cases completely, on the time scale of days. Similar studies for other important metal oxides show that this unexpected lability due to autocatalytic redox processes could be a common geochemical phenomenon (9, 10).

Given the semiconducting nature of Fe(III)-(oxyhydr)oxide minerals, a model based on electron conduction (4, 11) has been widely adopted as the explanation. Major steps include (i) Fe(II) sorption, (ii) electron transfer (ET) between sorbed Fe(II) and lattice Fe(III) leading to their atom exchange (AE) at the interface, and (iii) conduction of injected electrons to different Fe(III) lattice sites, which then undergo (iv) reductive release as Fe(II). The sorption, ET, and conduction steps are strongly supported by ⁵⁷Fe−Mössbauer spectroscopy (12, 13) and molecular simulations (14–19). Thus, within individual mineral particles, recrystallized regions that capture the Fe isotopic composition of solution are predicted, comprising a record of the AE front.

However, this record has yet to be demonstrated. Its existence, so far, is based on the isotopic signatures of sequential bulk aqueous extractions (5, 20–22), which provide indirect support for the conduction-based mechanism of AE. Confidence in the mechanism awaits microscopic insights that can resolve the AE front in detail. For example, various models have been developed predicting the average tracer exchange rate and/or concentration profile in goethite particles, each based upon different assumptions about whether or not the tracer isotope undergoes continuous exchange or burial (10, 21, 23). Close links between the extent of AE and the initial concentration of defects at goethite surfaces have been established (18, 19, 24, 25), suggesting the “healing” of defects as a possible thermodynamic driving force. However, because of the intrinsic limitations in linking bulk isotopic signatures to atomic-level AE mechanisms, much of the mechanistic interpretation accumulated so far is mired in circumspect ambiguity.

The present study provides direct isolation of the Fe(II)-catalyzed AE front in individual goethite crystallites, marking a precedent in characterization of this important redox interface. The ⁵⁷Fe(II) tracer experiments for AE with goethite microrods were recreated, and atom probe tomography (APT) was exploited to directly visualize ⁵⁶Fe/⁵⁷Fe distributions on individual goethite crystallites in three dimensions. APT is a mass-sensitive imaging technique capable of mapping the elemental and isotopic distributions in solids and at interfaces at the atomic scale. Success was enabled, in part, by our previous study on hematite (26), a system that lent itself more readily to initial APT methodological development. For hematite, the availability of micrometer-scale euhedral crystallites allowed for straightforward preparation of APT tips, and, more so, quantification of ⁵⁷Fe enrichment, because of the absence of structural hydrogen that introduces isobaric interferences with signals of interest (e.g., ⁵⁶FeH vs. ⁵⁷Fe). For goethite, here we surmounted the additional analytical challenges associated with APT specimen preparation of nanoscale particles, as well as developing a protocol for proper characterization of the iron isotopic composition in this hydrous material.

The resulting APT tomograms show that the AE front in goethite is spatially heterogeneous, at times penetrating several nanometers into the lattice, consistent with acceleration at defective regions of crystallite surfaces. Moreover, evidence that AE can penetrate along interdomain boundaries was obtained, suggesting a further possible tie to structural imperfections. The findings help eliminate some of the ambiguities in models based on macroscopic AE experiments. Also, this approach of visualizing AE fronts using APT presents important opportunities for future research on the much broader set of redox-catalyzed mineral transformations important to geochemical systems, including more-complex transformations such as ferrihydrite to goethite.

Results and Discussion

Goethite microrods were synthesized according to the methods of Schwertmann and Cornell (27) (see Methods). Transmission electron microscopy (TEM) images illustrate the typical morphological characteristics of these microrods (Fig. 1), showing the acicular morphology with lengths of ∼1 μm to 2 µm and the main growth direction corresponding to [010] (Pnma space group). The terminating facets at rod ends were identified as (210) and (010). A section perpendicular to [010] shows dimensions of ∼30 nm by ∼100 nm to 200 nm, defined by stepped prismatic (101) and (100) facets. Due to the limited number of observed projections, other minor terminations [e.g., (012)] cannot be excluded. Importantly, scanning TEM (STEM) observations show that the microrods often comprise intergrowths of multiple coaligned crystallites, which leads to formation of antiphase boundary crystallographic domains. STEM imaging also reveals that the particles contain a high concentration of defects in the form of nanopores or voids.

Fig. 1.

(S)TEM images of synthetic goethite microrods showing (A) typical morphology, (B) cross-sectional image highlighting common nanofaceted prismatic faces, (C) rod terminations along [010], and observations of nanopores, and (D) cross-sectional image highlighting prismatic faces and occurrence of intergrowths and nanopores.

The goethite microrods were reacted with aqueous Fe(II) initially enriched in ⁵⁷Fe for 18 d at pH 7.5 (25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffer, 25 mM KBr) under anoxic conditions (see Methods). On the basis of UV−visible (UV-vis) spectroscopic measurements of the supernatant (see Methods and Eq. 1), 0.2 mM aqueous Fe(II) was sorbed onto the goethite after 18 d. This compares well with previously reported ∼0.2 mM Fe(II) sorption after 15 d (8). Based on prior studies, the sorbed Fe(II) is assumed to have oxidized to Fe(III) and deposited a goethite-like surface phase (12, 13).

These reaction conditions were chosen to specifically mimic AE experiments in previous work (5, 8), which indicate that substantial iron exchange should occur over 18 d, to varying extents. In one study, exchange equilibrium was reached in microgoethite after 6 d (5). The subsequent change in ⁵⁶Fe/⁵⁷Fe in the goethite phase was, however, small and measured to be ∼40‰; that is, relative to ⁵⁶Fe/⁵⁷Fe in the initial goethite phase, which, at NA, is 0.023 (2.119 ⁵⁷Fe/91.754 ⁵⁶Fe), the average ⁵⁶Fe/⁵⁷Fe after reaction was ∼0.024. In a separate study, ∼30% iron exchange in goethite microrods occurred after 14 d of reaction, although exchange did not reach equilibrium after 51 d (8). These studies also indicated that the extent of recrystallization and AE are influenced by major but poorly quantifiable physical properties of particle size, crystallinity, and aggregation state; for example, more-extensive AE was shown to occur on nanorods (∼110 m²⋅g⁻¹) compared with microrods (40 m²⋅g⁻¹) at lower pH (5) and higher temperatures (8), where aggregation decreases and crystallinity increases. Thus, while we chose to use microrods for this experimental study to facilitate APT specimen preparation, the observations are assumed to be applicable to understanding recrystallization mechanisms across at least the range of goethite particle types achievable via similar synthesis strategies, if not more generally for Fe(II)/Fe(III)-(oxyhydr)oxide interfaces.

Preparation of the goethite microrods into needle-shaped tips is challenging but necessary for APT analysis. Currently, there are few successful approaches for the preparation of APT specimens from unsupported nanostructured materials that result in meaningful APT results (28–32). The goethite microrods in our study were prepared for APT analysis using an encapsulation, cross-sectioning approach (33). The basic steps are summarized by scanning electron microscope (SEM) images in Fig. 2 (see SI Appendix, section 1 for more detail on APT specimen preparation). In brief, the goethite microrods were embedded within a chromium matrix via ion beam sputter deposition. Chromium was chosen as the matrix material because it adheres well to oxide materials and facilitates a stable evaporation transition to iron oxides given their similar evaporation fields (26, 34). Single particles were chosen for APT analysis using a dual-beam SEM (Thermo Fisher Scientific Helios Nanolab 600i) with focused ion beam (FIB) capabilities. Each particle was extracted using conventional FIB techniques (35), and the lift-out was mounted onto a Si micropost. In this geometry, the APT specimen captures a small cross-section of the goethite particle containing several prismatic surfaces [e.g., (101)], similar to that shown in Fig. 1C. These surfaces are of particular interest, as they are predicted to undergo the most growth at the given reaction conditions relative to (210) faces at the particle ends, which may undergo dissolution (20). Needle-shaped APT specimens were obtained by annular milling with the FIB; the tips were milled so that the goethite/chromium interface was as close to the apex as possible.

Fig. 2.

SEM illustration of goethite particle lift-out and cross-sectional APT tip preparation. (A) Goethite “embedded” within a chromium matrix. (B) Lift-out showing the embedded goethite particle in cross-section. (C) APT tip where a small section of the goethite cross-section is placed at the tip apex.

Targeted goethite microrods were successfully analyzed via pulsed laser APT (see Methods for analysis conditions). Here we describe the specific results from two goethite microrods (referred to as MR1 and MR2), which are illustrative of four total successful APT runs across three goethite microrods. The reader is directed to SI Appendix for detailed descriptions of the APT reconstruction parameters and analytical procedures. In brief, 3D chemical reconstructions of the specimens were obtained by careful assignment of iron, chromium, oxygen, and hydrogen ionic species within the collected mass-to-charge state ratio (m/z) spectra; identification and assignment of ionic species was aided by supplemental APT measurements on control goethite specimens whose isotopic compositions were at NA (SI Appendix, sections 2 and 3). Specific iron isotopes within the 3D volume were delineated by assigning each isotopic mass peak within a subset of iron ionic species with a unique identity, as described in Taylor et al. (26). Successful identification, quantification, and resolution of the elemental/isotopic distribution at the subnanometer level in geologic materials has also been demonstrated in previous studies (36–38). We emphasize that, for goethite, it is important to avoid potential hydride interferences within the selected iron ionic species, as this will lead to quantification errors. For this study, we found the ⁵⁶Fe and ⁵⁷Fe isotopes of the Fe²⁺ species were the most reliable for analyses of the iron isotopic composition, as it reproduced NA ratios reliably in the goethite reference samples and did not suffer from hydride interferences (SI Appendix, sections 2 and 3). Thus, ⁵⁷Fe enrichment for the goethite microrods was observed using ⁵⁷Fe²⁺ (hereafter referred to simply as ⁵⁷Fe) and was quantified as ⁵⁷Fe²⁺/⁵⁶Fe²⁺ (hereafter referred to as ⁵⁶Fe/⁵⁷Fe).

The Iron Atom Exchange Front.

Several goethite surfaces were captured in the 3D APT reconstruction of a small cross-section from MR1 (Fig. 3A). While specific surfaces are not identifiable in the APT cross-sections, based on TEM cross-sectional analyses, the surfaces that are apparent in the APT reconstructions are likely to be the prismatic (101) or (100) faces (Fig. 1). The ⁵⁷Fe-enriched regions are clearly visualized using isotopic labeling, and a survey of all APT data consistently shows that these regions are laterally nonuniform along the microrod surfaces. As mentioned earlier, oxidative adsorption of ⁵⁷Fe coupled to dissolution of solid-associated ⁵⁶Fe can occur on separate but adjacent surfaces (20). The regions enriched in the tracer isotope are nominally the reaction fronts where oxidative adsorption of aqueous Fe(II) and associated growth of goethite occurs. Interestingly, some regions of the surface show no apparent enrichment. These surfaces could therefore comprise corresponding reaction fronts where reductive dissolution occurs, although the data do not allow for definitive proof in this regard. In general, however, the APT results are consistent with the notion that prismatic surfaces are locations of preferential goethite growth while in contact with aqueous Fe(II) (19, 20), but they also reveal that this process occurs heterogeneously across these surfaces in a fashion dominated by localized “hot spots” of AE at the nanoscale.

Fig. 3.

Three-dimensional APT reconstructions of localized ⁵⁷Fe enrichment. (A) The 3D APT reconstruction of goethite particle MR1 reacted with ⁵⁷Fe(II); 5-nm cube is used for 3D scale. Arrows indicate regions 1 and 2 highlighting (B) nonenriched and (C) ⁵⁷Fe-enriched regions, respectively. (D) Comparison with ⁵⁷Fe(II)−(001) hematite surface from Taylor et al. (26). Note that the ⁵⁶Fe/⁵⁷Fe values correspond to the y axis on the left-hand side, while the Cr and Fr concentrations correspond to the y axis on the right-hand side. The onset of the goethite phase is set to 0 nm, and the vertical, gray dashed lines mark the onset of the nominally nonrecrystallized phase (i.e., where ⁵⁶Fe/⁵⁷Fe reaches NA).

Quantification of ⁵⁷Fe Enrichment.

Quantification of the isotopic composition across different goethite surfaces of varying ⁵⁷Fe enrichment was desirable to better understand the average characteristics of the AE reaction front. For example, a feature of particular interest is the average tracer concentration with depth, as this concentration profile could, in principle, be compared with proposed extents of exchange reported from batch tracer studies and associated mechanistic modeling (10, 21, 23). The location of the goethite/chromium interface was first established by plotting the atomic concentration of iron and chromium (Fig. 3 B and C). We note that the measured iron concentration in the bulk regions away from the goethite/chromium interface [∼35 atomic (at.) %] deviates from the stoichiometry for goethite (25 at. %). This is a known issue for APT of oxides in general (39, 40) and iron oxides in particular (41). These challenges preclude accurate stoichiometric measurements for the goethite microrods and limit quantitative understanding of the elemental surface chemistry and its effect on Fe(II) adsorption. Knowledge of the surface stoichiometry would be important to determine, for example, whether iron-deficient regions and/or OH-rich surfaces promote ET and AE processes, as recently postulated (24). Improvements in the stoichiometric quantification, especially across the interface, would be invaluable, but may be at the limits of current APT detector capabilities as discussed by Baptiste et al. (40). We could, however, designate the location where the goethite/chromium interface terminated, and the onset of the goethite phase to be at the depth where the Cr concentration reaches a minimum of ∼0 at. % within error (defined at 0 nm in the corresponding APT profiles).

The iron isotopic composition across goethite/chromium interfaces was quantified as ⁵⁶Fe/⁵⁷Fe (derived from the Fe²⁺ species), considering 2σ (95%) confidence interval. The average ⁵⁶Fe/⁵⁷Fe in the bulk of the goethite microrod was determined to be 0.037 ± 0.032 (2σ), within measurement error of that expected at NA (0.023). The predicted uncertainty in the measurements can be high due to the limited volume analyzed and poor counting statistics. However, within the 2σ (95%) confidence interval, strong variations in ⁵⁷Fe enrichment across the goethite surface were confidently visualized. Some surfaces showed no detectable enrichment (Fig. 3B, region 1) while others showed up to 20× higher ⁵⁶Fe/⁵⁷Fe relative to the bulk goethite (Fig. 3C, region 2). These ⁵⁷Fe-enriched regions are thus clearly assignable to heterogeneous hot spot locations of AE reaction fronts.

To gain insight into the temporal evolution of AE, a definition for the depth of the nonrecrystallized phase was needed, which was taken to be located nominally where ⁵⁶Fe/⁵⁷Fe decays to NA levels within statistical uncertainty, although this is somewhat ambiguous in the absence of a permanent reference marker (26). This depth can either underlie or coincide with the onset of the goethite phase itself, described above. In other words, if ⁵⁷Fe enrichment is present, based on prior work, it is assumed that it was deposited in the form of goethite (12, 13). Hence, at any given location, the extent of AE within the goethite phase was estimated based on comparison of two depths: the onset of goethite compared with the onset of NA levels. For instance, if these two criteria are met at the same depth, this was taken to indicate that any recrystallization was limited to the interfacial region or was negligible. If, however, these depths were statistically different, the thickness of the overlying tracer-enriched goethite was used to characterize the extent of recrystallization, and the depth distribution of the tracer within this recrystallized region was examined to seek clues about the AE process over time.

The temporal evolution of ⁵⁶Fe/⁵⁷Fe across the ⁵⁷Fe-enriched region of MR1 can thus be described (Fig. 3C): The ⁵⁷Fe enrichment was greatest closest to the goethite/chromium interface (i.e., ⁵⁶Fe/⁵⁷Fe is ∼20× higher than NA). At the onset of the goethite phase, quantitatively significant amounts of ⁵⁷Fe enrichment are still observed at a depth of 0 nm (i.e., ⁵⁶Fe/⁵⁷Fe is up to ∼10× higher than NA). The isotopic gradient decreases to NA at a depth of −3.5 nm, indicating the onset of the nonrecrystallized goethite phase. Thus, ⁵⁷Fe deposition is greatest at the interface but can penetrate ∼3.5 nm into the goethite subsurface over a period of 18 d. This can be contrasted with the tracer distribution found in prior work for the ⁵⁷Fe(II)/hematite (001) system (26) (Fig. 3D). In that study, ⁵⁷Fe enrichment was always strictly confined within the hematite/chromium interface, with no penetration into the hematite subsurface. There, the much shorter Fe(II) exposure time (1 d) and lower reactivity of these large particles suggested that the main process was simply rapid oxidative adsorption of ⁵⁷Fe(II) onto the (001) surface.

In the goethite system, both because of the higher AE reactivity of goethite relative to hematite and because of the much longer reaction time used, more substantial extents of recrystallization were expected and observed. Given the heterogeneity and gradient in the ⁵⁷Fe concentration across the interface, recrystallization via burial-like mechanisms is shown to occur; that is, the recrystallized phase exhibits distinct isotopic compositions because it is not continuously exchanging iron atoms with the aqueous reservoir (23). However, while the apparent depths of recrystallization at hot spots are consistent with this expectation, alternative explanations exist. For example, the significant penetration depth found at hot spots could also indicate regions of enhanced subsurface access, such as by tracer diffusion into channels or nanopores. In the present system, the goethite microrods clearly contain a high concentration of internal nanoscale voids/defects (Fig. 1). Although likely not directly related to these specific features, it has been postulated that goethite surface defects, such as iron vacancies, provide sites into which Fe(II) can strongly bind and there transfer electrons to lattice Fe(III), propagating a goethite-like surface (24). It was furthermore shown that this kind of reactive goethite defect could be eliminated by “annealing” under hydrothermal treatment. However, the X-ray absorption spectroscopies used to detect and characterize these atomic-scale defects were sensitive primarily to the uppermost few angstroms of the goethite surface, and therefore likely did not sample interior channels and nanopores of the kind observed in Fig. 1. In any event, given the clear internal void space in the present system, the question remains regarding unknown extents of subsurface access for the AE front mediated by channels and pores beyond the resolution of our TEM imaging. These observations collectively highlight the ongoing need to better constrain the role of defects with respect to apparent recrystallization. Given its potential importance, in our APT studies, we paid specific attention to this topic by examining the tracer distribution along the common occurrence of grain boundaries, as described in Evidence for Active Intergranular Atom Exchange.

Evidence for Active Intergranular Atom Exchange.

In addition to capturing tracer enrichment at what appeared to be primarily external surfaces of individual goethite microrods, separate APT reconstructions across a multidomain particle reveal penetration of ⁵⁷Fe into confined spaces between crystallites, likely by intergranular boundary diffusion (Fig. 4). Given that the synthetic microrods often displayed grain intergrowths with apparent subdomains of crystallinity (Fig. 1), APT characterization was also directed at microscopically identified regions of crystallites where grain boundary density was presumed high. The wedge containing the second single goethite microrod MR2 was sectioned in half, enabling analysis at two different locations (referred to as S1 and S2).

Fig. 4.

Three-dimensional APT reconstructions of microrod MR2 sections (A) S1 and (B) S2, where S2 highlights intergranular diffusion of ⁵⁷Fe. Solid arrows through full tip reconstructions indicate the ∼2-nm-thick region where cross-section measurements were taken; 5-nm cubes are used for scale.

In the 3D APT reconstruction of S1 in Fig. 4A, nanofaceted prismatic faces are captured with varying amounts of ⁵⁷Fe enrichment at the exterior surfaces, similar to that observed for MR1 in Fig. 3A. The reconstructed S2 specimen (Fig. 4B) resembles that of S1 but, more interestingly, shows features consistent with the intergrowth of two particles (referred to as P1 and P2). That is, a microstructural boundary in S2 separating P1 from P2 is clearly observed via the presence of C, CO_x, Cr, and CrO_x ionic species. The chromium diffused into the grain boundary during the coating process, and carbon may have entered either during the coating or at an earlier time during specimen handling. Furthermore, within this boundary, ⁵⁷Fe enrichment is observed, thus demonstrating intergranular diffusion of Fe(II). Given that ⁵⁷Fe is associated initially with the aqueous reservoir, ⁵⁷Fe diffusion into the grain boundary may indicate diffusion of hydrated Fe(II), as opposed to purely atomic diffusion, such as via small channels large enough to accommodate solvated Fe(II), although this is purely speculative. In any event, Fe(II) diffusion into intergranular boundaries is consistent with the notion that recrystallization is driven by the annealing of defective microstructural regions, as indicated on MR1 and previously speculated (24).

The apparent propensities for AE between the intergrown particles P1 and P2 within S2 differ considerably, as the exterior surfaces of P2 are ⁵⁷Fe-enriched, while the exterior surfaces of P1 are effectively not enriched in ⁵⁷Fe. Similarly, looking at S2 in cross-section across the boundary, the ⁵⁷Fe is associated with the interior surface of P2, while the interior surface of P1 is not enriched. The different reactivities between P1 and P2 could possibly be attributed to exposures of different facets with different affinities for Fe(II), although, in the absence of a complementary technique to determine crystallographic orientations of domains in APT tips before analysis, this remains purely speculative. Nonetheless, these observations highlight the heterogeneity in reactivity from grain to grain and how AE and recrystallization mechanisms can vary across single crystallites.

While boundary diffusion has long been thought to play a considerable role in ion adsorption on goethite (42–45), APT provides detailed insights into this phenomenon. Akin to this topic is the importance of aggregation-based effects, as aggregation controls the accessible reactive surface area (46–48), although reconciling its impact has been difficult due to limited quantification. Previous studies have postulated that the hindered interaction of Fe(II) with occluded regions influences long-term ion adsorption and recrystallization kinetics (21); the initial rapid stages of recrystallization are thought to largely be derived from interactions of Fe(II) with well-exposed surfaces such as on the periphery of particle aggregates, while the following slower stages reflect diffusion-limited interactions in interior domains. Furthermore, distorted Fe−O bonds at strained regions at grain boundaries and associated crystal defects are thought to enhance the mobility of atoms through the goethite crystal structure (8).

The APT analyses here are consistent with these previous assertions. A number of reactive processes were directly observed in the Fe(II)-catalyzed recrystallization of goethite over 18 d of reaction, including oxidative adsorption of ⁵⁷Fe(II) onto prismatic goethite surfaces, atomic diffusion of ⁵⁷Fe several nanometers within the goethite subsurface at potential defect sites, and slower aqueous diffusion of ⁵⁷Fe into interior surfaces of intergranular boundaries. For instance, reconstructions for S2 of MR2 capture AE fronts reflective of the initial and later stages of exchange. Aqueous diffusion of the isotopic tracer into and through the grain boundary is shown to provide a pathway to penetrate deeper into the crystallite, thereby potentially enabling atoms deeper within the bulk to undergo AE. These results provide direct proof that intergranular domains in some instances can be active in the context of Fe(II)-catalyzed AE and recrystallization, although quantification of the impact of these effects requires further exploration. Given that APT is established in characterizing the chemistry at grain boundaries, interfaces, and nanoscale features, this technique is uniquely poised to help address these kinds of effects in future work.

Conclusions

Direct visualization of the AE front long hypothesized to exist as a record of Fe(II)-catalyzed recrystallization of microscale-to-nanoscale goethite crystallites has now been achieved. Fe(II) oxidative adsorption and growth on goethite appears to be a highly spatially heterogeneous process on the nanoscale, but largely consistent with area-averaged models that ascribe the highest propensity to prismatic (101) surfaces of acicular goethite. The ⁵⁷Fe enrichments varied greatly between nanoscale hot spots displaying ∼20× more ⁵⁶Fe/⁵⁷Fe relative to the bulk goethite, while other nanoscale domains exhibited no enrichment above NA. The latter domains may simply have a lower affinity for Fe(II) or represent locations where corresponding reductive dissolution events were more facile on average. AE fronts were found to penetrate past the nominal surface of goethite, in some cases, for the given run conditions up to more than 3 nm. Additionally, APT evidence reveals Fe(II) diffusion into microstructural grain boundaries and related defect-enabled penetration and the expansion of the active AE interface.

Breakthroughs in the characterization of AE fronts in individual goethite particles gained through isotope-resolved imaging from nanometer to atomic scales provides supporting evidence for the conduction-based mechanism of Fe(II)-catalyzed recrystallization as well as important insights. The data are consistent with oxidative adsorption of Fe(II) onto prismatic surfaces and autocatalytic growth of goethite incorporating the tracer isotope, and the likely importance of defective regions, broadly defined, for creating preferential hot spots of AE. APT furthermore clearly shows diffusion along intergranular boundaries, consistent with the likely involvement of diffusion-limited access to internal spaces. These results complement and aid in the explanation of previous findings. For example, insignificant morphologic changes during recrystallization may correspond to cases where most of the apparent AE is occurring at and evolving the defect content at the surface and/or within nanoporous intergranular domains. Oxidative sorption of Fe(II) and goethite growth could potentially occur much more frequently at defective regions of the goethite surface accessible to aqueous Fe(II) (19, 24, 25), thermodynamically driven by formation of more stable crystalline phases.

Deeper insight into energetic and kinetic processes controlling recrystallization at these sites can be gained by using this knowledge of AE fronts to construct refined AE models, to mimic, for example, the spatially heterogeneous dynamics of AE. Hybrid models encompassing continuous exchange and gradual burial may be required to address the fact that burial is spatially heterogeneous at the nanoscale. Iron atoms between the surface of the original nonrecrystallized phase and the aqueous reservoir can potentially exchange rapidly (e.g., within the first 5 d for microrods) (21) and then, at hot spots, gradually be removed from the exchanging fraction by burial during net goethite growth. At all other locations, the rapid mixing fraction would leave only small isotopic variations at best, likely within error of the APT measurements. Past AE modeling has been limited by inferences made from isotopic measurements averaged across solid and aqueous reservoirs (10, 21, 23). By incorporating this knowledge of the spatial distribution of the isotopic tracer in the recrystallized phase as well as the temporal evolution of the isotopic composition, these more tightly constrained models would advance closer to understanding atomic-level phenomena controlling recrystallization. Thus, combining experimental observations from the microscopic, single-crystallite level with macroscopic AE data modeling could provide a more robust framework for understanding redox-catalyzed recrystallization mechanisms.

More generally, these observations of AE fronts obtained for the goethite system may also relate to and empower new studies of more complex mineral transformation processes associated with the Fe(II)/Fe(III) interface. For instance, in Fe(II)-accelerated transformations of ferrihydrite, highly disordered surface layers are proposed to reconfigure into a “reactive ferrihydrite” phase upon interaction with Fe(II), which facilitates the nucleation and growth of crystalline iron oxide minerals such as goethite and lepidocrocite (49, 50). In principle, using isotopic tracers, the APT approach developed here could be adapted to unraveling the mass transfer pathways at the evolving interfaces in this system. Although challenging, APT analysis of the distribution of potassium on amorphous ferrihydrite particles was previously achieved (31). In particular, nanoparticles must be embedded in a matrix without nanovoids or microvoids, which can cause premature fracture or artifacts in APT reconstructions, that preserves their structure and composition to acquire meaningful data that can be related to the original particles (32, 51, 52). Tedious but essential work entails refining and testing specimen preparation techniques, such as embedding in a material compatible with field ionization of the particles that is free of voids, strongly adhered, and free of noninterfering ions. In light of accomplishments to date, the ongoing effort appears well justified.

Methods

Goethite microrods were synthesized using the method of Schwertmann and Cornell (27); that is, 50 mL of 1 M Fe(NO₃)₃ was added to 90 mL of 5 M KOH in a 2-L polyethylene flask, vigorously mixing on a stir plate. All chemicals were of reagent-grade quality. The mixture was diluted to a final volume of 1 L using DI water, and heated in an oven at 70 °C for 60 h. Following heating, the goethite was allowed to cool and then was washed via centrifugation three times. The precipitate was dried overnight at 40 °C. X-ray powder diffraction confirmed the phase was goethite (PDF#98-000-0229) (SI Appendix, Fig. S1). The surface area was measured to be 40 m²⋅g⁻¹ using the Brunauer−Emmett−Teller method with N₂ adsorption and a degas temperature of 25 °C.

TEM was performed with aberration-corrected FEI Titan 80-300 operated at 300 kV to determine the particle morphology. The microscope incorporates CEOS GmbH double-hexapole aberration corrector for the probe-forming lens, which allows imaging with ∼0.8-nm resolution in STEM mode. The presented images were acquired in STEM mode using a high-angle annular dark-field detector with beam convergence of 17.8 mrad and collection angle of 54 mrad. All images were acquired with TIA software package. The sample preparation for TEM observations included dispersion of microrods in deionized water and subsequent drop casting on lacey-coated 100 mesh Cu TEM grids (Ted Pella). To image the particles in cross-section, the particles were embedded in LR White acrylic resin (Electron Microscopy Sciences), cut into ∼70-nm thin sections with a Leica Ultracut UCT ultramicrotome equipped with a diamond knife (Diatome US), and mounted on Ultrathin C/Au TEM grids (Ted Pella).

Batch experiments were conducted similar to that in Handler et al. (5), where 2 g⋅L⁻¹ of goethite was reacted with 1 mM ⁵⁷Fe(II) at pH 7.5 (25 mM HEPES, 25 mM KBr). The 0.8 M ⁵⁷Fe(II) stock solution was prepared by first dissolving ⁵⁷Fe metal (Cambridge Isotopes) in 5 M HCl, filtering it through a 0.22-µm membrane, and diluting it to 0.1 M HCl. The ⁵⁷Fe metal had a reported isotopic abundance of 95.93% ⁵⁷Fe (compared with an NA value of 2.12%), which was confirmed by inductively coupled mass spectrometry. An end-over-end mixer was used to react 15-mL suspensions in polypropylene centrifuge tubes, wrapped in Al foil. All experiments were conducted in an anoxic glove box (N₂/H₂ atmosphere) to avoid Fe(II) oxidation by ambient oxygen. All solutions used for the Fe(II)−goethite reaction used anoxic, degassed water prepared by boiling ultrapure Milli-Q water in Pyrex Corning glass bottles on a hot plate for ∼30 min under vacuum. Afterward, the water was immediately transferred into the anoxic glove box, where purified N₂ was blown onto and sparged into it overnight.

For this study, we analyzed goethite reacted with ⁵⁷Fe(II) for 18 d with APT, as AE and goethite recrystallization at various extents was previously observed to occur at this time (5, 8). Following reaction, the aqueous and solid fractions were separated via centrifugation. The supernatant was removed and analyzed with the ferrozine method (using a wavelength of 562 nm on a UV-vis spectrophotometer) to determine the amount of Fe(II) remaining in solution, [Fe(II)_aq]. Weakly bound iron on the goethite surface was removed by first rinsing and centrifuging the powders with water once and then exposing them to 0.4 M HCl (trace metal basis) solution for 10 min. The supernatant from this extraction was also analyzed to determine [Fe(II)_extr], a quantity used to estimate the amount of ⁽⁵⁷⁾Fe(II) associated with the particles that has recrystallized [Fe_recryst],

$$\left[\mathrm{F}{\mathrm{e}}_{\mathrm{recryst}}\right]=\left[\mathrm{Fe}{\left(\mathrm{II}\right)}_{\mathrm{aq,initial}}\right]-\left[\mathrm{Fe}{\left(\mathrm{II}\right)}_{\mathrm{aq,final}}\right]-\left[\mathrm{Fe}{\left(\mathrm{II}\right)}_{\mathrm{extr,final}}\right].$$ [1]

Following the extraction step, the particles were rinsed and centrifuged (16,000 × g, 5 min) with anoxic water. The goethite was dried under N₂ atmosphere and then prepared for APT analyses. From this point, anoxia was no longer necessary to maintain.

Preparation of the goethite microrods into needle-shaped tips was achieved by adopting an encapsulation and cross-sectioning approach, in which conventional FIB−SEM techniques could be applied to make needle-shaped APT tips. A detailed description of the sample preparation protocol can be found in SI Appendix, section 2 and Fig. S2. APT specimens were analyzed in a CAMECA Local Electrode Atom Probe (LEAP) 4000 X-HR at a set-point temperature of 60 K, a laser pulse repetition rate of 125 kHz, and a detection rate of 0.002 ions per pulse (maintained by varying the applied specimen voltage). The laser wavelength was 355 nm, and laser energy per pulse was 80 pJ. The data were reconstructed using the Integrated Visualization and Analysis Software (IVAS 3.8.0) developed by CAMECA.

To properly quantify the chemical and isotopic composition of the ⁵⁷Fe-reacted goethite particles, control goethite specimens that were not isotopically enriched were first systematically analyzed to establish the appropriate APT analysis conditions and procedures; that is, APT analyses were conducted within the bulk of a natural goethite single crystal, serving as a standard for elemental and isotopic characterization (SI Appendix, section 2, Fig. S3, and Table S1), and a goethite microrod specimen at NA, serving as a secondary standard enabling analysis of whether the complexity of the microrod specimens (e.g., heterogeneous interfaces) influences isotopic quantification (SI Appendix, section 3 and Fig. S4). The reader is referred to SI Appendix for detailed explanations of these analyses guiding the assignment of ionic and isotopic species to the mass spectra. Ultimately, these supplementary analyses guided the assignment of the ionic and isotopic species for the m/z spectra of the isotopically enriched goethite specimens (SI Appendix, section 3 and Fig. S5). The iron isotopic composition within the 3D volume was determined by assigning each mass peak within a subset of iron ionic species with a unique isotopic identity. Specifically, the ⁵⁶Fe and ⁵⁷Fe isotopes of the Fe²⁺ species were utilized to characterize and quantify iron isotopes, as these species consistently reproduced isotopic ratios at NA in the control specimens and thus were determined to be reliable probes for characterizing and quantifying ⁵⁷Fe enrichment (SI Appendix, sections 2 and 3 and Table S1). The Fe²⁺ accounts for ∼30% of all detected iron in the mass spectrum, and thus serves as a lower-limit approximation for the total iron content.

Variations in the iron and chromium elemental concentrations and ⁵⁶Fe/⁵⁷Fe ratios across the reconstructed goethite surfaces for ⁵⁷Fe-enriched and nonenriched regions were analyzed and quantified using the proximity histogram method (or proxigrams) (53); this technique integrates the chemical and 3D positional information and calculates the ionic/atomic concentration with respect to the distance from a defined isoconcentration surface. The sorbed ⁵⁷Fe surface concentration was also quantified utilizing the ionic counts from the proxigrams. Errors in the elemental concentrations and isotopic ratios were estimated by standard counting statistics. Details on the quantification of the elemental and isotopic concentrations and their associated errors are provided in SI Appendix, section 4.