Linked Reactivity at Mineral-Water Interfaces Through Bulk Crystal Conduction

The semiconducting properties of a wide range of minerals are often ignored in the study of their interfacial geochemical behavior. We show that surface-specific charge density accumulation reactions combined with bulk charge carrier diffusivity create conditions under which interfacial electron transfer reactions at one surface couple with those at another via current flow through the crystal bulk. Specifically, we observed that a chemically induced surface potential gradient across hematite (α-Fe2O3) crystals is sufficiently high and the bulk electrical resistivity sufficiently low that dissolution of edge surfaces is linked to simultaneous growth of the crystallographically distinct (001) basal plane. The apparent importance of bulk crystal conduction is likely to be generalizable to a host of naturally abundant semiconducting minerals playing varied key roles in soils, sediments, and the atmosphere.

concentration was similar to the modern value, because Eocene estimates of Os burial flux (35)  The semiconducting properties of a wide range of minerals are often ignored in the study of their interfacial geochemical behavior. We show that surface-specific charge density accumulation reactions combined with bulk charge carrier diffusivity create conditions under which interfacial electron transfer reactions at one surface couple with those at another via current flow through the crystal bulk. Specifically, we observed that a chemically induced surface potential gradient across hematite (a-Fe 2 O 3 ) crystals is sufficiently high and the bulk electrical resistivity sufficiently low that dissolution of edge surfaces is linked to simultaneous growth of the crystallographically distinct (001) basal plane. The apparent importance of bulk crystal conduction is likely to be generalizable to a host of naturally abundant semiconducting minerals playing varied key roles in soils, sediments, and the atmosphere.
T he chemical behavior of mineral-water interfaces is central to aqueous reactivity in natural waters, soil evolution, and atmospheric chemistry and is of direct relevance for maintaining the integrity of waste repositories and remediating environmental pollutants. Traditionally, explorations of fundamental reactions at these interfaces have probed the interaction of water and relevant dissolved ions with crystallographically well-defined mineral surfaces. The pursuit so far has been dominated by the assumption that distinct surfaces of any given crystal behave independently of each other. Except by diffusion through the solution phase or across sur-face planes, exchange of mass or electron equivalents between sites of differing potential energy at different locations on any given crystal is typically assumed to be negligible. This assumption is nonetheless questionable for the widespread group of minerals that are electrical semiconductors. For example, iron oxides often have moderate to low electrical resistivity (1) and have been studied as electrode materials for decades (2)(3)(4). Iron oxide crystal surfaces are chemically reactive with water and ions, leading to solution-dependent charging behavior that differs from one surface type to the next; differing points of zero charge for proton adsorption is but one example (5,6). This difference should give rise to a surface electric potential gradient (Dy 0 ) across any crystal that has two or more structurally distinct faces exposed to solution. In principle, this gradient can bias the diffusion of charge carriers (7,8). Hence, conditions could exist when the gradient across a single crystal is sufficiently large and the electrical resistivity of the material sufficiently *To whom correspondence should be addressed. E-mail: kevin.rosso@pnl.gov low that interfacial electron transfer reactions at one surface couple with those at another by a current flowing spontaneously through the crystal bulk. The situation is analogous to galvanic metal corrosion, but instead of spatially disordered anodic and cathodic electron transfer sites, the anode and cathode are spatially confined to crystallographically distinct surface planes and are therefore physically separable for measurement. We demonstrate operability of these conditions for iron oxide, uncover their effects on the surface chemical behavior, and make the case that, in nature, different surfaces of certain abundant crystals are inextricably linked.
We examined hematite (a-Fe 2 O 3 ) because it is a wide band gap semiconductor (band gap 1.9 to 2.3 eV) (1, 3) and the most stable form of iron oxide under dry oxidizing conditions; it is extremely common in nature (9). It has the corundum structure type based on hexagonal close packed oxygen planes in which 2/3 of the available octahedral cavities are occupied by Fe 3+ . This structure gives rise to anisotropic electrical resistivity that is higher in the basal plane than along the trigonal axis by up to four orders of magnitude (10,11); reported bulk resistivities range from 10 2 to 10 6 ohm·m (1). When hematite is subjected to oxygen-limited aquatic environments, particularly in acidic conditions, it can be reductively dissolved according to Eq. 1: which has a standard reduction potential E°~0.7 V (8,12). The fundamental reaction central to the overall process is Sorbed Fe 2+ from the aqueous phase is capable of reducing hematite Fe 3+ in this system (13)(14)(15), yielding an iron redox cycle in which no net reduction occurs. Introduction of dicarboxylic acids such as oxalate causes net dissolution by chelating surface Fe 3+ (ligand assisted dissolution); it also enhances reduction, possibly through the formation of ternary surface complexes with Fe 2+ (aq) , for example (16). This collective chemistry is a good test case for our main hypothesis because it involves a source of electron equivalents from Fe 2+ in solution, a range of potential-determining ions, electron transfer across hematite-solution interfaces, and the possibility of moving electron equivalents through the crystal bulk.
Development of a potential gradient Dy 0 of significant magnitude across the crystal requires selective interaction between potential-determining ions and specific hematite surfaces. We focus here on roles of protons (low pH) and oxalate as a representative dicarboxylic acid. The hematite (001) basal surface is structurally distinct from any edge surfaces. In water, the (001) surface is terminated predominantly by doubly coordinated hydroxyls (17)(18)(19) that are relatively inert to the protonation and deprotonation reactions needed for charge accumulation. Smaller populations of more reactive singly coordinated and triply coordinated hydroxyls, capable of positive charge accumulation, are associated with terminal Fe groups (19). Terminal Fe groups with low coordination to the underlying surface can be easily chelated by oxalate anions to form negatively charged mononuclear bidentate inner-sphere surface complexes (20,21). In contrast, many remaining low-index surfaces of hematite crystals such as (012) are dominated by higher-coordinated Fe (17,22). This higher-coordinated Fe is more difficult to chelate yet bears singly and/or triply coordinated hydroxyls for charge accumulation. In general therefore, we expect that, relative to other hematite surfaces, the (001) surface should show weaker pH-dependent charge accumulation, an observation increasingly confirmed by recent data and theory (23,24), and stronger interaction with oxalate anions that increases with decreasing pH.
We performed several experiments to potentiometrically measure Dy 0 for specific hematite surfaces and to determine its effect on their behavior in Fe 2+ and oxalate solutions. We chose a large natural specular hematite crystal with well-defined surfaces that could be isolated for study. The crystal was low in impurity content (25), a natural n-type semiconductor (1), and had a room temperature electrical resistivity of 10 5 ohm·m as measured by the four-point probe method. Generating replicate samples required cutting specific crystallographic surfaces from the crystal as rectangular prism-shaped specimens, which also yielded a vicinal surface type along cut edges. For example, we prepared millimetersized oriented prisms exposing two (001) surfaces on the top and bottom of the prism and four orthogonal (hk0) vicinal sides (25). Annealing in air under conditions where hematite is the only stable iron oxide effectively cleans and organizes the surfaces without modifying the bulk electrical conductivity. This procedure yields highly organized (001) surfaces, with accompanying (hk0) vicinal surfaces that are microfaceted with stable edge terminations (Fig. 1, A and B). Similarly, prism specimens bearing (012) and (113) surfaces (Fig. 1, C and D) with accompanying vicinal surfaces were prepared.
To determine the magnitude of Dy 0 and the roles of Fe 2+ and oxalate solution components, we measured the open-circuit potential (E OCP ) in four solution types. The E OCP is the electrode rest potential relative to a standard reference electrode. Changes in the E OCP are directly related to changes in y 0 (26,27), which in turn is sensitive to surface complexation reactions with our potential-determining ions H + , Cl − , Fe 2+ , and oxalate species (28). The measurements were performed at room temperature at effectively constant ionic strength under anaerobic conditions (25). We focused these measurements on the (001) and the accompanying (hk0) vicinal surface type. The observed approximately linear pH dependence, with predominantly negative slopes, is consistent with the accumulation of positive surface charge with decreasing pH (Fig. 2). As expected, in pure electrolyte solution the (001) surface showed a less-negative slope relative to that of the (hk0) surface, consistent with a lower density of charge accumulation sites on the (001) surface ( Fig. 2A). In contrast, oxalate anions bind preferentially to the (001) surface with decreasing pH, even to the point of sign reversal in the slope (Fig. 2B). Addition of Fe 2+ to either solution shows that its primary effect is to lower the overall potential for both the (001) and (hk0) surfaces without substantially modifying the pH dependence (Fig. 2, A and B). Taking E (hk0) -E (001) as an estimate of Dy 0 , in the presence of oxalate and irrespective of the presence of Fe 2+ (aq) we found that the potential gradient is large and positive, on the order of tenths of volts below pH = 3 (Fig.  2C). Under these conditions, we expect that mobile electrons acting as majority carriers in hematite would be directed by Dy 0 from the (001) surface to the (hk0) surface. E OCP measurements directly between identical surface types [e.g., E (001) -E (001) ] (25) showed no significant voltage.
To examine the effects of Dy 0 of this sign and magnitude on the surface chemical behavior, we examined surfaces of the oriented prisms from the same sample by using atomic force microscopy (AFM) before and after anaerobic reaction with Fe 2+ -oxalate solutions. Thermostated batch vessels were used with temperatures ranging from room temperature to 75°C and pH ranging from 2 to 3 (25). Fe 2+ -oxalate concentrations consistent with previously published experiments that establish net dissolution in terms of Fe 3+ (aq) release on fine-grained powders were used (16,29,30). Collectively, these conditions were selected in keeping with Eq. 1 while also accelerating surface transformations into a more easily observable time frame. Light was excluded in all cases to avoid oxalate acting as a reductant. Equilibrium thermodynamic calculations along with Eh measurements at run conditions confirm that all our reaction conditions lie within the Fe 2+ (aq) stability field (25). Hematite was the limiting reactant; total dissolution would retain undersaturation with respect to any possible iron oxide phases.
AFM examination of (001) surfaces after reaction runs showed remarkable features. In every case, for both natural and synthetic samples, (001) surfaces were overgrown with a hexagonal pseudo-pyramidal morphology of uniform orientation. Images at early stages show the island growth of these features on the initially flat (001) surface (Fig. 1E). After 12 hours, the reaction yielded merged pyramid-covered (001) surfaces with peak-to-valley heights averaging 200 nm and pyramid bases approaching a micrometer in width, imparting a distinct matte appearance to the reacted (001) surface visible to the naked eye. Transmission electron microscopy (TEM) and se-lected area diffraction measurements along [001] transects of this sample type ( fig. S1), along with x-ray photoelectron spectroscopy, x-ray diffraction, and energy dispersive x-ray spectroscopy, confirmed that the grown material is structurally and compositionally a-Fe 2 O 3 of identical orientation as the underlying material without detectable impurities. The line of intersection of apparent pyramid "facets" with the (001)    tion, but the interplanar angle varies with run duration and does not correspond to low-index planes in hematite. The large size, morphologic symmetry, and mutual orientation of these pyramids require homoepitaxy, that is, growth of additional hematite on hematite.
In contrast, all other surfaces examined show features characteristic of dissolution. For example, the four (hk0) vicinal sides of prism samples bearing (001) surfaces on top and bottom exhibited fine-scale pitting and roughening (Fig.  1F). The (012) and (113) surfaces of prism samples show development of etch pits at various length scales and with symmetry corresponding to crystallographic orientation (Fig. 1, G and H). We observed identical behavior under the same conditions with use of synthetic tabular hematite crystals bearing primarily (001) and (012) surfaces, in which case no surface preparation by annealing was required (fig. S2). The conclusion is that the hematite (001) surface grows under our conditions whereas all other surfaces sampled dissolve, as represented schematically in Fig. 3A.
We designed experiments using the prism samples to test whether or not pyramid islands are deposited on the (001) surface by precipitation of trace Fe 3+ from solution (16). In these experiments, two prism samples were used in the reaction vessel instead of one. Four (hk0) vicinal sides of one crystal were sealed with an inert epoxy (25), leaving two (001) surfaces exposed, whereas on the other crystal the two (001) surfaces were sealed, leaving four vicinal surfaces exposed. Collectively, the two crystals expose the same six kinds of surfaces to solution as in the runs above with one crystal, in the same relative proportion and surface area, but they involve (001) surfaces that are physically separated from the (hk0) surfaces (Fig. 3B). In this case, the results of reaction runs show only dissolution features on all exposed surfaces, including (001) (fig. S3A). The (001) pyramidal morphology does not develop in this separated two-crystal configuration. The same experiment performed on samples in which the pyramidal morphology had already been grown on (001) before sealing its (hk0) sides showed dissolution of the pyramidal over-growths on the (001) surface ( fig. S3B). Therefore, the (001) pyramidal overgrowths do not form by precipitation of ferric iron. Furthermore, we deduce that chemical processes at the (001)-solution interface causing pyramidal growth during reaction are facilitated by solid contact between the (001) and (hk0) surfaces; that is, these surfaces must be on the same crystal.
The behavior strongly suggests that bulk charge transport provides the link between the two types of surfaces. As a further test, we again prepared two crystals with partial exposure of (001) on one and (hk0) surfaces on the other, except this time with an electrical connection between them (Fig.  3C). A crystal exposing only (001) surfaces was connected to a crystal beneath exposing only (hk0) surfaces by electrically conductive colloidal Ag paste, which was subsequently cured, sealed off from contact with solution using additional epoxy, and tested for ohmic behavior by resistivity measurements. In this design, the crystals are effectively wired together by the (001)-Ag-(001) junction between them. Reaction in this wired two-crystal configuration proceeds as if the crystals were one; pyramidal hematite grows on the exposed (001) surface of the upper crystal ( fig. S3C), whereas the four (hk0) sides of the lower crystal dissolve (Fig. 3C). Therefore, the nature of the interaction between the (001) and vicinal surfaces that gives rise to the pyramidal growth of hematite (001) during reaction derives from bulk charge transport. Surface diffusion along the hematite-solution interface was ruled out by painting a ring of sealant on a (001) surface so that only bulk transport could access the circumscribed region, and within that region hematite island growth also occurred ( fig. S4).
The collective behavior of the system is therefore suggestive of two distinct but coupled interfacial processes: growth at (001) by and dissolution of edge surfaces, for example, (hk0) surfaces, by with coupling mediated by charge transport from (001) to (hk0) surfaces through the crystal bulk. The process involves preferential net oxidative adsorption of Fe 2+ (aq) at the (001)-solution interface and valence interchange with structural Fe 3+ at that surface (Fig. 4). At temperatures of interest (room temperature and higher), bulk charge transport is sufficiently facile to support a small current through the bulk. Net electron equivalents injected into the (001) surface follow an electrically biased random walk through the crystal to (hk0) surfaces. At (hk0) exit points, internal reduction of Fe 3+ to Fe 2+ solubilizes and releases iron into solution. This circuit is driven by the Dy 0 gradient generated across the crystal from divergent charge accumulation at structurally distinct surface types. The sign and magnitude of Dy 0 , the conductivity of the natural crystal, and the growth rates of the pyramidal islands are all mutually consistent. For example, taking Dy 0 = 0.2 V at pH = 2, a temperatureadjusted electrical resistivity = 10 4 ohm·m for 75°C (31,32), and an electron transport path length of 1 mm, the maximum amount of additional hematite expected on the (001) surface in 12 hours is a layer~100 nm thick, the same order of magnitude as that observed. Surface potentialdriven charge carrier diffusivity has been invoked qualitatively to explain microscopic oxide transformation processes before (13,33,34) but not on the length scale examined here nor with surface specificity. Given the observation that the (001) surface continues to grow beyond the coalescence of the pyramidal islands, at the atomic scale the pyramidal (001) morphology must retain the essential structural and therefore chemical characteristics that give rise to the potential of the initial (001) surface. Furthermore, the observed process does not preclude traditionally held spatially localized dissolution in the hematite system. Rather, the evidence suggests that the processes operate in parallel and that the behavior based on the electrical circuit through the crystal dominates when chemical requirements that establish a large enough surface electric potential gradient are met.
The finding provides insight into the reductive transformation of iron oxides, which is important in the biogeochemical cycling of iron in nature and the removal of iron oxide films in industry. Because this finding can be easily generalized to a host of naturally abundant semiconducting transition metal oxide and sulfide minerals capable of dominating the interfacial surface area in soils, sediments, and among atmospheric particles, its implications are fairly widespread. Of immediate impact is the concept that the reactivity of any given surface on such materials can be coupled to that of another surface, with a dependence on crystal morphology as a whole. This phenomenon should apply to natural crystals in the environment as well as those selectively cut, broken, or otherwise prepared for laboratory study.  4. Schematic diagram depicting the inferred coupled interfacial electron transfer process operative under our conditions for the hematite single crystals. The chemically self-induced surface potential gradient across the crystal directs current flow through the bulk. The current is facilitated by sufficiently low electrical resistivity in a process that is fed by net injection of electron equivalents at (001) surfaces and net release of electron equivalents at (hk0) surfaces.