Report

Electron Small Polarons and Their Mobility in Iron (Oxyhydr)oxide Nanoparticles

See allHide authors and affiliations

Science  07 Sep 2012:
Vol. 337, Issue 6099, pp. 1200-1203
DOI: 10.1126/science.1223598

Abstract

Electron mobility within iron (oxyhydr)oxides enables charge transfer between widely separated surface sites. There is increasing evidence that this internal conduction influences the rates of interfacial reactions and the outcomes of redox-driven phase transformations of environmental interest. To determine the links between crystal structure and charge-transport efficiency, we used pump-probe spectroscopy to study the dynamics of electrons introduced into iron(III) (oxyhydr)oxide nanoparticles via ultrafast interfacial electron transfer. Using time-resolved x-ray spectroscopy and ab initio calculations, we observed the formation of reduced and structurally distorted metal sites consistent with small polarons. Comparisons between different phases (hematite, maghemite, and ferrihydrite) revealed that short-range structural topology, not long-range order, dominates the electron-hopping rate.

Many important geochemical and biogeochemical redox reactions are linked to the formation or transformation of iron oxide and oxyhydroxide minerals by charge-transfer processes that cycle iron between its two common oxidation states (1). Because iron(II) is substantially more soluble than iron(III), reductive transformations of iron(III) (oxyhydr)oxides can dramatically affect the chemistry and mineralogy of soils and surface waters (2). However, it remains difficult to predict the outcome of even some common reactions of these minerals. For example, after the exposure of iron(III) (oxyhydr)oxides to reducing agents, interfacial electron transfer (ET) can lead to release of iron(II) (dissolution), formation of alternative iron(III) or mixed-valence phases (transformation), or particle growth (35). Qualitatively, it is known that such diverse behavior is due to the coupling of interfacial redox reactions and interior charge conduction processes (6), but it is unclear how this interplay directs these biogeochemical outcomes. Mineral redox reactions are complex, and it is typically unknown whether interfacial charge transfer, interior conduction, or steps such as bond breaking are rate limiting. We sought to measure the mobility of electrons transferred into iron(III) oxides by using time-resolved spectroscopy.

There is an established model describing electrons in materials such as iron oxides that are characterized by short-range metal-ligand bonding (7). In this model, conduction band states are highly localized because of strong electron-phonon interactions that stabilize charge carriers in a lattice distortion (8, 9). Specifically, electrons introduced into an iron oxide become localized in unfilled metal three-dimensional states, polarize neighboring atoms, and distort the local structure, forming polarons. Because the coupling with phonons is very strong, the polaron radius is small, on the order of a lattice constant, as depicted in Fig. 1A. These electron small polarons are effectively a localized lower-valence metal site, and conduction occurs through thermally activated electron hopping from one metal site to the next. The small-polaron model is widely used for describing conduction in many semiconductors and insulators, but support for the model is typically indirect, derived from agreement between theory and measurements such as optical spectroscopy (10) or temperature-dependent conductivity (11, 12). If Arrhenius analysis reveals the existence of an activation-energy for charge transport, then phenomenological macroscopic transport models furnish an effective elementary hopping rate. For certain material classes, electron localization throughout the material accompanied by Jahn-Teller distortion marks a transition from metallic to a polaronic insulator phase (13). However, for most materials in which small-polaron formation is posited, the density of these charge carriers is too low for bulk structural analyses; alternative methods are required to characterize them at a structural level.

Fig. 1

(A) Illustration of a small polaron in iron oxide—the local structural distortion created by the reduction of an iron(III) site. (B) Scheme of the ET pathways after light-initiated interfacial ET from a surface-bound dye molecule to a iron(III) oxide or oxyhydroxide nanoparticle in aqueous suspension. The electron-transfer rate constants ki, kr, and kc are associated with interfacial ET, recombination, and internal conduction (by hopping to an adjacent iron neighbor), respectively.

We have developed a pump-probe approach to study iron(III) phases with excess electrons and the dynamics of electron transport across multiple time scales (14). In the pump step, laser-initiated interfacial ET reduces a fraction of iron atoms at the surface of iron(III) (oxyhydr)oxide nanoparticles sensitized by the dye molecule 2′,7′-dichlorofluorescein (27DCF). In the probe step, time-resolved x-ray absorption spectroscopy (XAS) at the Fe K-edge is used to monitor the oxidation state and coordination geometry (15, 16) of the iron sites affected by interfacial ET. This approach was inspired by research on dye-sensitized semiconductor nanomaterials (17) and informed by numerous accomplishments in ultrafast science, including optical-pump–x-ray probe studies of transient electronic and structural configurations in organometallic cluster compounds (1821). Here, we report measurements of the mobility of excess electrons introduced into three phases of iron(III) (oxyhydr)oxide. We studied two crystalline phases, maghemite (γ-Fe2O3) nanoparticles with average diameter 3.4 nm and 7-nm hematite (α-Fe2O3). Additionally, we studied 3-nm 2-line ferrihydrite, an environmentally (22) and physiologically (23) important nanophase. Synthesis methods were optimized in order to achieve the smallest-size nanoparticles without impurity phases and with particle size distributions below 15%, as confirmed by means of x-ray diffraction and electron microscopy (figs. S1 to S4). Nanoparticle samples were studied in aqueous suspension at pH 4 with or without surface-bound 27DCF.

Continuous illumination of suspensions of all phases of 27DCF-coated nanoparticles caused the generation of dissolved iron(II) (fig. S5). We studied the electron injection process by means of ultrafast optical transient-absorption (TA) spectroscopy using 520-nm wavelength, 130-fs duration excitation pulses and a broadband visible probe (supplementary materials). The excitation wavelength was chosen for maximum absorption by the bound dye. Unavoidably, a fraction of the excitation intensity was also absorbed by the nanoparticles, generating electron-hole pairs that were characterized separately by studies of bare nanoparticles. All uncoated nanoparticles gave an optical TA response with a lifetime less than 200 ps (fig. S6) interpreted as nonradiative recombination of electron-hole pairs (24).

The optical TA data for all 27DCF-coated nanoparticles exhibited significantly longer time-scale dynamics than uncoated nanoparticles, as illustrated for hematite in Fig. 2. The TA data are interpreted as a sum of contributions from different states of the sample after photoexcitation (supplementary materials). The absorption band appearing around 460 nm is attributed to surface-bound 27DCF that has lost one electron through interfacial ET (27DCF+). The suppression of absorption strength from 490 to 550 nm represents bleaching of the dye ground state. The signatures of 27DCF+ and the ground-state bleach show a similar decay rate, indicating that some photoinjected electrons recombine with 27DCF+ on the nanosecond time scale. (Band gap excitations of the nanoparticles also contributed to the transient response in this wavelength range but could be distinguished by their much faster recombination dynamics.) The rate of interfacial ET cannot be identified at 460 nm through the appearance of the 27DCF+ signal because we cannot distinguish it from the neutral photoexcited dye, 27DCF* (25). However, the spectral feature in the 490- to 550-nm range additionally exhibits a sub-picosecond decay immediately after photoexcitation. Global fitting analysis shows that the transient spectrum associated with this early time scale is consistent with the fluorescence of 27DCF*, whereas all later transient spectra are dominated by the loss of ground-state absorption (Fig. 2B, inset). The sub-picosecond transient thus represents the decay of stimulated emission (SE) and is strong evidence of electron loss from 27DCF* via interfacial ET. For all 27DCF-coated nanoparticles, SE decay and hence interfacial ET occurred with time constants from 200 to 250 fs, which is close to the rate measured for 27DCF on TiO2 (26). Thus, laser-initiated interfacial ET is much faster than the subsequent spontaneous ET processes described below.

Fig. 2

(A) Optical TA spectrum of uncoated hematite (Hm) at 0.3-ps delay after photoexcitation at 520 nm compared with TA spectra of 27DCF-coated hematite nanoparticles (Hm-DCF) at the indicated delays. Arrows indicate the wavelengths at which kinetics curves in (B) were extracted. (B) Transient kinetics observed for uncoated and 27DCF-coated hematite nanoparticles at the wavelengths indicated in (A). Also plotted are best-fit transients obtained through global analysis of the Hm-DCF curves with three exponential time constants and one offset. The time axis was shifted 1 ps for plotting on a logarithmic scale. (Inset) Wavelength-dependent amplitudes of the two fastest decay time constants: The transient spectrum that decays in the first 200 fs is consistent with stimulated emission from the excited state of the dye and indicates that interfacial ET occurs on this time scale. The 1.3-ps spectrum is associated with ground-state bleach of the dye and thus represents early interfacial recombination on this time scale. Ground-state absorption and emission spectra are shown in fig. S3.

Time-resolved XAS characterized the solid-phase species generated by interfacial ET (supplementary materials). For each sample, ground-state Fe K-edge XA spectra are shown in Fig. 3A, and transient XA spectra acquired 150 ps after ET (Fig. 3B), which is reported as a difference relative to the ground-state spectrum (ΔXA). Although the magnitudes of the ΔXA spectra are small (≤2% of ground-state absorbance), these spectra were observed reproducibly during our studies of dye-sensitized nanoparticles (21 samples analyzed). No ∆XA spectrum above the noise was detectable for any control (fig. S8A). XAS analysis and chemical assays of the samples after reaction showed the gradual accumulation of dissolved aqueous iron(II) (figs. S5 and S13).

Fig. 3

Iron K-edge XA spectroscopic observation of reduced iron sites in three phases of iron (oxyhydr)oxide nanoparticles: maghemite (M), ferrihydrite (F), and hematite (H). (A) Ground state XA spectra of 27DCF-sensitized nanoparticles in suspension. (B) Difference (ΔXA) spectra acquired 150 ps after light-initiated ET. The raw (markers) and three-point smoothed (thick lines) data are compared with model spectra obtained by combining the ground-state reference with either a magnetite reference (dashed black lines) or a copy of the ground-state spectrum shifted –1.4 eV to simulate valence change without structural relaxation (solid black lines). (C) Comparison of the experimental Fe K-edge XA spectra for hematite with calculated spectra for ground-state hematite (FEFF 1) and for a locally distorted small-polaron site in hematite (FEFF 2) obtained from ab initio structure prediction. (Inset) The calculated on-threshold intensity exhibits high sensitivity to first-shell Fe-O bond length. (D) Comparison of the transient hematite ΔXA spectrum with calculated spectra for reduced sites without structural relaxation (FEFF 1) and for reduced and structurally distorted sites (FEFF 2). (E and F) Comparison of corresponding experimental and simulated spectra for the maghemite phase.

All ΔXA are dominated by a positive feature that is at lower energy than the ground-state threshold, which is consistent with a model in which light-initiated ET reduced Fe3+ to Fe2+ (16). The Fe3+/Fe2+ K-edge chemical shift for iron that is octahedrally coordinated by oxygen is typically observed to be –1.4 ± 0.1 eV (27). Thus, to model iron reduction without associated structural change we simulated ΔXA spectra by mixing the ground-state spectrum in a 99:1 proportion with a duplicate to which this energy shift was applied. The resulting spectra do not match the experimental transient spectra well for any phase, indicating that structural change accompanies reduction. Because the chemical reduction of both maghemite (28) and ferrihydrite (29) can induce transformation to the mixed-valence inverse spinel magnetite, Fe3O4, we tested whether light-initiated ET led to the formation of magnetite domains but again obtained poor matches to the experimental data. Magnetite and maghemite differ in oxygen stoichiometry as well as iron valence, and the poor agreement indicates that atom diffusion to achieve local charge balance and the nucleation of a new mineral phase does not occur within 150 ps (30). Thus, interfacial ET causes iron reduction and a structural change that is less extensive than a complete phase change.

We next developed a theoretical approach to test whether the transient spectra observed for hematite and maghemite were consistent with the formation of small polarons in these phases. We used ab initio calculations for periodic systems to predict the distortion in the local structure of an iron(II) site using methods similar to previous work (9). Briefly, an extra electron was localized on an iron cation in the ground state, and all atomic coordinates were energy-minimized self-consistently with the wave function until converged. The small-polaron structure was defined relative to the atomic coordinates calculated for a completely iron(III) lattice. The consequences of structural distortion for XAS were investigated by using a full multiple scattering code (31) that has been used successfully to interpret metal K-edge spectra (32). Simulation parameters were chosen that optimized the agreement between calculated and experimental data for ground-state hematite and maghemite. The XAS calculations reproduce most features in the ground-state spectra (Fig. 3, C and E, and fig. S15). However, there are evident discrepancies in lineshape broadening and energy position that are greater in magnitude than the transient differences observed in the time-resolved spectra. Consequently, we refined the goal of the simulation study not to reproduce precisely the ΔXA data but to test whether small-polaron formation is a plausible interpretation of the observed lineshapes.

We calculated theoretical ΔXA spectra for reduced Fe sites in hematite and maghemite with or without the structural distortion predicted by simulation (Fig. 3, D and F). For both phases, the calculations predicted that small-polaron formation enhanced the intensity of the key feature of the ∆XA spectra. We sought to understand the structural origin of this feature (fig. S16), finding it to be highly sensitive to the Fe-O bond length (Fig. 3C, inset). As illustrated in Fig. 1A, bond length expansion is the major predicted feature of small-polaron formation, with a ~4% increase in hematite.

The experimental data and simulations provide strong evidence for the small-polaron model by identifying key electronic and structural features of this state after it has formed. Although measurements with sub-picosecond resolution (15) will be required to follow small-polaron formation, our measurements of the iron(II) lifetime on the nanosecond and microsecond time scales allowed electron mobility to be quantified. We determined the loss of iron(II) by monitoring the intensity of the transient ΔXA signal at 7.125 keV (Fig. 4). The iron(II) concentration exhibits a biphasic decay, initially dropping to ~30% of the initial value within ~2 ns [ferrihydrite and maghemite (fig. S8B)] or ~10 ns (hematite). We fitted the nanosecond–time scale kinetics data with a three-site model for the location of an electron transferred to a surface iron site after dye excitation and determined two intrinsic ET rates from fits of the nanosecond kinetics for all phases: kc, the rate at which electrons at a surface site hop away from the ET site, and kr, the recombination rate (Fig. 1B). For hematite, kc = 0.19 ± 0.04 ns–1, a value within the range predicted by ab initio simulations of small-polaron–mediated electron conduction in bulk hematite (33). For both ferrihydrite and maghemite, kc = 0.34 ± 0.09 ns–1. It is surprising that the electron-hopping rates are the same for maghemite, a well-characterized crystal phase, and ferrihydrite, a disordered nanomaterial for which a new structural model (22) is debated (34, 35). The finding adds weight to the suggestion that these phases may contain similar structural elements but with alternative long-range organization (36). For the maghemite phase, analysis of the site-specific contributions to the ground and transient spectra revealed that both octahedral and tetrahedral iron sites contributed to conductivity at room temperature, in a proportion corroborated by the ab initio calculations (supplementary materials).

Fig. 4

(A) Experimental kinetics curves showing the creation and loss of iron(II) in hematite and ferrihydrite nanoparticles. Maghemite nanosecond kinetics data are given in fig. S8b. Fits to a three-site, two-exponential model are used to obtain kc and kr. Errors in the fitted constants are ±0.05 ns−1 or less. (B) Experimental kinetic curves showing the presence and continued loss of iron(II) at the microsecond time scale, compared with the iron(II) signals at 5 ns.

The ability to probe electron small polarons is a first step toward experimentally distinguishing the multiple steps that comprise redox transformations of iron (oxyhydr)oxides and measuring their rates. The small-polaron hopping rate is a fundamental mineral phase–dependent limit on the kinetics of many iron redox reactions. As shown in Fig. 4B, we additionally observed iron(II) in the nanoparticles at microsecond time scales, representing itinerant electrons that were trapped within the iron oxides before leaving either by dissolution or recombination. Complementing the present measurements of hopping rates with rates of additional steps, such as bond breaking or the nucleation of new mineral phase, will enable more detailed mechanistic descriptions of the reductive dissolution of iron(III) (oxyhydr)oxides and other important environmental reactions. We anticipate that the combination of transient-absorption spectroscopy, conventional kinetics measurements, and ab initio and kinetic modeling offers a framework for this program.

Supplementary Materials

www.sciencemag.org/cgi/content/full/337/6099/1200/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S16

Table S1

References (3750)

References and Notes

  1. Acknowledgments: Time-resolved x-ray transient spectroscopy was performed at beamline 11-ID-D at the Advanced Photon Source (APS). Laser facilities at 11-ID-D were provided by the Solar Energy Conversion group of Chemical Sciences and Engineering Division of Argonne National Laboratory, which is funded through New Facility and Mid-scale Instrumentation grants to L. X. Chen et al. We thank L. X. Chen, G. Jennings, and C. Kurtz. PDF analysis was performed at beamline 11-ID-B at the APS. Transient absorption spectroscopy was performed at the Argonne Center for Nanoscale Materials (CNM), and we thank G. Wiederrecht and D. Gosztola. This work was supported by the Chemical Imaging program of the U.S. Department of Energy, Office of Science, and Office of Basic Energy Sciences (DOE-BES), under contract DE-AC02-05CH11231. K.M.R. and P.Z. gratefully acknowledge support from DOE-BES Geosciences program to PNNL. C.F. acknowledges support from the Danish Council for Independent Research. Use of the APS and the CNM is supported by DOE-BES under contract DE-AC02-06CH11357.
View Abstract

Stay Connected to Science

Navigate This Article