Adatom Fe(III) on the hematite surface: Observation of a key reactive surface species
© American Institute of Physics 2004
Received: 02 February 2004
Accepted: 26 May 2004
Published: 30 June 2004
The reactivity of a mineral surface is determined by the variety and population of different types of surface sites (e.g., step, kink, adatom, and defect sites). The concept of "adsorbed nutrient" has been built into crystal growth theories, and many other studies of mineral surface reactivity appeal to ill-defined "active sites." Despite their theoretical importance, there has been little direct experimental or analytical investigation of the structure and properties of such species. Here, we use ex-situ and in-situ scanning tunneling microcopy (STM) combined with calculated images based on a resonant tunneling model to show that observed nonperiodic protrusions and depressions on the hematite (001) surface can be explained as Fe in an adsorbed or adatom state occupying sites different from those that result from simple termination of the bulk mineral. The number of such sites varies with sample preparation history, consistent with their removal from the surface in low pH solutions.
Mineral surfaces are the medium upon which the Earth's solids and fluids interact. Their reactivity in the fundamental processes of adsorption, dissolution/growth, and electron transfer is directly tied to their atomic structure. In addition to two-dimensionally periodic surface structures, there are one-dimensional periodic and zero-dimensional structures such as steps and kink sites that play important roles in mineral reactivity. These include adsorbed or adatom sites. There are long-standing models of surface chemistry that appeal to the existence of populations of adatom species having properties distinct from both the solid and aqueous solutes. In crystal growth theories, for example, dissolved nutrient is postulated to pass through the "adsorbed nutrient" state before incorporation into the crystal structure. Despite the theoretical importance of a pool of adsorbed nutrient in mineral dissolution and growth, there is little work confirming its existence or properties on mineral oxides. Surface adatom species (or sites) are difficult to study because of their small size, restriction to surfaces and interfaces, and nonperiodic nature. Indirectly, it has been shown that transient spikes in dissolution rate occur in response to pH changes in a way consistent with the formation and dissolution of adsorbed Fe at the hematite surface, [3–6] and an isotopic exchange and Mössbauer study by Rea et al concluded that a population of kinetically labile sites characterizes the ferrihydrite surface.
Nonperiodic adatom sites at mineral surfaces can behave quite differently from other surface sites, and thus are crucial to understanding the overall chemical reactivity of mineral surfaces. Adatom sites are more sterically accessible than other sites and thus may be more easily complexed by organic molecules that are prone to formation of multidentate surface complexes. There is evidence that organic molecule adsorption can be enhanced by formation of organic-Fe ternary surface complexes, which suggests that adsorbed or adatom iron creates a surface more prone to organic molecule adsorption.[8, 9] Here, we make a first step toward direct study of such adsorbed, adatom, or nonperiodic surface material for the case of hematite (α-Fe2O3). We present atomically resolved scanning tunneling microscopy (STM) images of hematite (001) surfaces, supported by a resonant tunneling model (RTM) parametrized with ab initio calculations, that are consistent with the existence of nonperiodic adsorbed Fe on the periodic hematite (001) surface structure. We also present initial evidence that exposure of the surface to acidic conditions removes much of the nonperiodic material from the surface. This suggests a direct correspondence between the observed nonperiodic Fe and the dissolution-labile Fe observed in other studies of hematite. [3–6]
II. Hematite (001) Surface Structure
We used natural hematite from Tarascon sur Ariège, France. Sample preparation and characterization are described in more detail in Ref. 19. We used samples that had been exposed to distilled and doubly deionized (DDI) water for 96 h and either imaged subsequently or exposed to concentrated nitric acid for 1 h before imaging.
Natural hematite is typically an n-type semiconductor at 25°C due to donor impurities. Based on resistivity, our hematite had an estimated impurity concentration (donors minus acceptors) of between 1 × 10-3 and 4 × 10-2 at. %. A laser-ablation ICP-MS analysis gave a total donor minus acceptor concentration of 2.0 × 10-3 at. %, within the range of values from resistivity. These semiquantitative measurements confirm that our hematite is n-type, and show that the main donor impurities are Sn and Ti. Adding up donor, acceptor, and impurities that do not affect conduction, we have about 6 × 10-3 at. % impurities on the basis of ICP-MS. Oxygen vacancies, which act as electron donors, are included in the resistivity measurement but not in the ICP-MS measurement.
B. Scanning tunneling microscopy (STM)
A Digital instruments Nanoscope IIIa controller was used with a Molecular Imaging electrochemical STM with a 4-μm scanner to image hematite both in- and ex-situ (in pH 1 HNO3 or DDI H2O when in-situ). Imaging conditions (bias voltage, setpoint current) are given in the figure captions. For ex-situ imaging we used electrochemically etched tungsten tips, and for in-situ imaging we used commercial Pt/Ir tips insulated with Apiezon wax and tested for < 10 pA faradaic current (Molecular Imaging).
C. Resonant tunneling model (RTM)
The concept of resonant tunneling is well known in solid-state physics, and is the basis of devices known as resonant tunneling diodes.[31, 32] The models have been extended to include thin-film semiconductor devices that are physically akin to the situation described here (metal tip, resonator in the tip–sample gap, semiconductor). As applied to hematite, the RTM is a relatively quick way to evaluate both the distance dependence of current from an adatom to the substrate as well as changes in the degree of adatom solvation associated with changes in bonding to the surface. For example, an Fe atom with one bond to the surface (e.g., a IV site in Fig. 1) will likely have a different distance to nearest-neighbor Fe atoms than a 3V (or A) site and be coordinated by more inner-sphere water molecules than would an Fe atom in A, B, or C sites (Fig. 1). These factors, in addition to changes in electronic structure, will affect the kinetics of electron transfer from the surface atom to or from an STM tip.
The RTM, and the ab initio calculations used to support and parametrize it, are described in Ref. 19. Here, we present only the main concepts. The solvent structure around an aqueous Fe3+ ion changes when the Fe3+ is reduced to Fe2+. Fe–O bond lengths increase, and the hydrogen-bonded structure of outer-sphere water of solvation changes. In this reduction (a process of zero net ΔG known as self-exchange), the solvent reorganizations required to bring the donor and acceptors states to the same energy is the reorganization energy, λ. For electron transfer between aqueous Fe3+ and Fe2+ λ is relatively large (about 100 kT or 2.6 eV33 at room temperature), making it an important consideration for electron transfer to and from hematite in air or water. Methods for calculating current from an electrode (such as an STM tip) through a redox-active monolayer to a metal substrate in which the redox-active layer is subject to solvent reorganization have been presented and tested. [34–36] We have used this idea because the λ for different surface sites (A, B, and C sites as well as "adsorbed" Fe in sites other than A, B, and C; Fig. 1) can be substantially different from one another because of differing degrees of solvation, leading to different electron-transfer characteristics for different surface sites.
We only consider tunneling to and from iron atoms, in agreement with previous work.[16, 19, 37] The A, B, and C sites, as well as nonperiodic or adatom sites, are the resonators mediating electron transfer between the tip and the hematite substrate. Electron transfer (ET) occurs in two steps; for current from substrate to tip, ET occurs first from the substrate to the redox center (sr), and then from the redox center to the tip (rt). We assume that the density of states of substrate and tip are independent of bias voltage. The tunneling current, j, is then
where β is the tunneling decay constant for the couple indicated by the subscript, d is the distance between the designated couple (the nearest-neighbor Fe atoms are considered the substrate, which determines the d sr distances), ε is the electron energy, and V b is the STM sample bias voltage relative to the tip. e0 and ħ are the electron charge and Planck's constant, respectively. For given distances and electronic structures, the current is thus proportional to the density of unoccupied states on the resonator, also called the density of oxidized states D ox (ε), which can be approximated by a Gaussian
where λ is the Marcus reorganization energy, and ε r is the reduction potential of the redox center. The λ term allows us to incorporate the effect of a solvent on current through surface sites. λ can be separated into inner-sphere (λ IS ) and outer-sphere (λ OS ) components.
Unlike studies in which bias voltage was kept at a constant, small value and only the electrode potential was varied relative to a reference electrode,[36, 38] here the bias voltage is varied. The resulting image characteristics are therefore somewhat more complex than the simple increases and decreases in current when a particular site comes into and then out of resonance. However, our approach is in keeping with the simple STM imaging we performed, rather than electrochemical resonant tunneling microscopy. We discuss some of these complexities as the calculated images are presented and compared to STM images.
To calculate images, we need values for the parameters in Eqs. (1) and (2) for each type of site we wish to model. Plane-wave pseudopotential calculations and density functional theory calculations on clusters were used to predict bond lengths and thus d sr , λ (= λ IS + λ OS ), ε r , and contributed to the determination of β. Details on the calculations and the methods used can be found in Ref. 19. For the case of "adsorbed" Fe, we have not attempted to model each of the many different structures that could occur. Instead, we simply recognize that the various possible sites will have a range of λ, d rt , d sr , and ε r values. We initially use ε r from the A sites modeled in Ref. 19 (see Fig. 1) and vary the height of the Fe site above the surface (which varies d sr and d rt ) as well as the λ attributed to the site. We also vary ε r at fixed λ. This approach allows us to test a range of conditions in a simple way, but cannot be expected to model all of the possible structural variations for such adatoms. In effect, the RTM as applied here is a qualitative guide to trends, but cannot be expected to quantitatively reproduce the experimental results.
Parameters used in the RTM. Other symbols are defined in the text, except "NN" which indicates the number of equivalent "nearest neigbors."
β sr (Å-1)
β rt (Å-1)
d sr (Å)
λ is (eV)
λ os (eV)
ε r (eV)
IV. Results and Discussion
A. RTM calculations
B. Bias voltage and preparation dependence of STM images
C. Dual-bias imaging
Although the RTM cannot be regarded as quantitatively accurate, the RTM results presented in Figs. 3, 8, and 9 suggest that the STM observations in Fig. 7 might be explained if the local λ and resonance energies of the sites involved differ slightly from those of the periodic surface structure. It should also be kept in mind that the RTM calculations only consider an A-type site with altered RTM parameters on an iron-terminated surface; the behavior of Fe adatoms on an oxygen-terminated surface is likely to be quite different.
D. Structural hints
E. Impurities and site densities
The measured impurity concentration allows us to ask whether the nonperiodic bumps and depressions in the STM images could be caused by impurity atoms (e.g., through local potential effects on the local electronic structure). For example, in a 20 × 20 nm STM image, there are just under 5500 Fe sites (assuming an Fe termination and including A, B, and C sites). Based on our highest impurity concentration (4 × 10-2 at. %), only about 2.2 atoms are impurities. As pointed out in Eggleston et al. (2003), we have observed up to about 60 nonperiodic features in some 20 × 20 nm images.
Therefore, only a small minority of the nonperiodic features can be explained by appeal to impurities exclusively, and it is therefore highly unlikely that most of the observed nonperiodic sites are caused by impurities.
Another useful comparison between Ref. 4 and the present work is that of site densities. On the basis of dissolution transients in response to pH jumps, Samson et al. estimated a site density for kinetically labile surface Fe sites of about 1.0 μmol m-2. This translates to about 24 sites in each 20 × 20 nm image. Figure 5 shows that our STM observations are up to a factor or 2 to 3 higher than this, but are within the same order of magnitude.
We have presented calculated and experimental STM images showing that nonperiodic sites (both protrusions and depressions) observed on hematite (001) surfaces by STM, both ex-situ and in-situ, are consistent with Fe in an adsorbed or adatom state on the hematite surface. These sites are observed as protrusions at higher negative bias voltages. Acid treatment of the surfaces both prior to imaging and during in-situ imaging removes most of these sites, consistent with their interpretation as adsorbed species. The time scale of removal during in-situ imaging at pH 1 is consistent with the pH-jump experiments of Samson et al. for dissolution of kinetically labile Fe from the hematite surface during approach to steady-state dissolution, and is roughly consistent with the time scale of isotopic exchange experiments of Rea et al. for kinetically labile surface Fe on ferrihydrite.
Adatom Fe in nonperiodic and nonbulk sites at the hematite surface represents structures whose chemical behavior differs from, and cannot be directly predicted from, an understanding of the bulk termination structure. The conditions that influence or control the production and consumption of this reservoir of reactive surface sites should thus be a subject of future investigation.
- Brown GE, et al: Chem Rev (Washington, DC). 1998, 99: 77-10.1021/cr980011z.View ArticleGoogle Scholar
- Zhang J-W, Nancollas GH: Mineral–Water Interface Geochemistry. Edited by: Hochella MF, Jr, White AF. 1990, Reviews in Mineralogy, Mineralogical Society of America, Washington, DC, 23: 365-396.Google Scholar
- Samson SD, Eggleston CM: Environ Sci Technol. 1998, 32: 2871-10.1021/es9803097.View ArticleGoogle Scholar
- Samson SD, Stillings LL, Eggleston CM: Geochim Cosmochim Acta. 2000, 64: 3471-10.1016/S0016-7037(00)00462-2.View ArticleGoogle Scholar
- Samson SD, Eggleston CM: Geochim Cosmochim Acta. 2000, 64: 3675-10.1016/S0016-7037(00)00461-0.View ArticleGoogle Scholar
- Samson SD, Eggleston CM: Water–rock Interaction, Ore Deposits, and Environmental Geochemistry: A Tribute to David A. Crerar. Edited by: Hellmann R, Wood SA. 2001, Geochemical Society, St. Louis, MO, 61-73.Google Scholar
- Rea BA, Davis JA, Waychunas GA: Clays Clay Miner. 1994, 42: 23-View ArticleGoogle Scholar
- Kraemer SM, Hering JG: Geochim Cosmochim Acta. 1997, 61: 2855-10.1016/S0016-7037(97)00133-6.View ArticleGoogle Scholar
- Fish W: Portland State University, personal communication. 2004Google Scholar
- Hochella MF, Eggleston CM, Eilings VB, Parks GA, Brown GE, Wu CM, Kjoller K: Am Mineral. 1989, 74: 1233-Google Scholar
- Eggleston CM, Hochella MF: Am Mineral. 1992, 77: 911-Google Scholar
- Condon NG, Murray PW, Leibsle FM, Thornton G, Lennie AR, Vaughan DJ: Surf Sci Lett. 1994, 310: L609-10.1016/0039-6028(94)91360-9.View ArticleGoogle Scholar
- Condon NG, Leibsle FM, Lennie AR, Murray PW, Vaughan DJ, Thornton G: Phys Rev Lett. 1995, 75: 1961-10.1103/PhysRevLett.75.1961.View ArticleGoogle Scholar
- Condon NG, Leibsle FM, Lennie AR, Murray PW, Parker TM, Vaughan DJ, Thornton G: Surf Sci. 1998, 397: 278-10.1016/S0039-6028(97)00744-9.View ArticleGoogle Scholar
- Wang X-G, Weiss W, Shaikhutdinov SK, Ritter M, Petersen M, Wagner F, Schlogl R, Scheffler M: Phys Rev Lett. 1998, 81: 1038-10.1103/PhysRevLett.81.1038.View ArticleGoogle Scholar
- Eggleston CM: Am Mineral. 1999, 84: 1061-Google Scholar
- Shaikhutdinov SK, Weiss W: Surf Sci Lett. 1999, 432: L627-10.1016/S0039-6028(99)00643-3.View ArticleGoogle Scholar
- Adib K, Camillone N, Fitts JP, Rim KT, Flynn GW, Joyce SA, Osgood RM: Surf Sci. 2002, 497: 127-10.1016/S0039-6028(01)01635-1.View ArticleGoogle Scholar
- Eggleston CM, Stack AG, Rosso KM, Higgins SR, Bice AM, Boese SW, Pribyl RD, Nichols JJ: Geochim Cosmochim Acta. 2003, 67: 985-10.1016/S0016-7037(02)01200-0.View ArticleGoogle Scholar
- Banwart S, Davies S, Stumm W: Colloids Surf. 1989, 39: 303-10.1016/0166-6622(89)80281-1.View ArticleGoogle Scholar
- Maurice PA, Hochella MF, Parks GA, Sposito G: Clays Clay Miner. 1995, 43: 29-View ArticleGoogle Scholar
- Hug SJ: J Colloid Interface Sci. 1997, 188: 415-10.1006/jcis.1996.4755.View ArticleGoogle Scholar
- Lenhart JJ, Honeyman BD: Geochim Cosmochim Acta. 1999, 63: 2891-10.1016/S0016-7037(99)00269-0.View ArticleGoogle Scholar
- Duckworth OW, Martin ST: Geochim Cosmochim Acta. 2001, 65: 4289-10.1016/S0016-7037(01)00696-2.View ArticleGoogle Scholar
- Kristl I, Kretzschmar R: Geochim Cosmochim Acta. 2001, 65: 3435-10.1016/S0016-7037(01)00695-0.View ArticleGoogle Scholar
- Marcus RA: J Chem Phys. 1956, 24: 966-10.1063/1.1742723.View ArticleGoogle Scholar
- Marcus RA: J Chem Phys. 1965, 43: 679-10.1063/1.1696792.View ArticleGoogle Scholar
- Marcus RA, Sutin N: Biochim Biophys Acta. 1985, 811: 265-View ArticleGoogle Scholar
- Marcus RA: J Phys Chem. 1990, 94: 1050-10.1021/j100366a005.View ArticleGoogle Scholar
- Jones F, Rohl AL, Farrow JB, van Bronswijk W: Phys Chem Chem Phys. 2000, 2: 3209-10.1039/b003380o.View ArticleGoogle Scholar
- Diez E, Dominguez-Adame F, Sanchez A: J Appl Phys. 1995, 77: 4816-10.1063/1.359404.View ArticleGoogle Scholar
- Presilla C, Sjostrand J: J Math Phys. 1996, 37: 4816-10.1063/1.531671.View ArticleGoogle Scholar
- Grätzel M: Heterogeneous Photochemical Electron Transfer. 1989, CRC Press, Boca Raton, FLGoogle Scholar
- Schmickler W, Widrig C: J Electroanal Chem. 1992, 336: 213-10.1016/0022-0728(92)80272-6.View ArticleGoogle Scholar
- Schmickler W: Surf Sci. 1993, 295: 43-10.1016/0039-6028(93)90183-K.View ArticleGoogle Scholar
- Schmickler W, Tao N: Electrochim Acta. 1997, 42 (18): 2809-2815. 10.1016/S0013-4686(97)00084-4.View ArticleGoogle Scholar
- Tarrach G, Burgler D, Schaub T, Wiesendanger R, Guntherodt HJ: Surf Sci. 1993, 285: 1-10.1016/0039-6028(93)90908-3.View ArticleGoogle Scholar
- Tao NJ: Phys Rev Lett. 1996, 76: 4066-10.1103/PhysRevLett.76.4066.View ArticleGoogle Scholar