Thin Water Films at Multifaceted Hematite Particle Surfaces - Langmuir

Nov 11, 2015 - These can contribute to processes as varied as biogeochemical cycling of elements, cloud formation, and ice nucleation, as well as (pho...
0 downloads 10 Views 4MB Size
Download PDF
Article pubs.acs.org/Langmuir

Thin Water Films at Multifaceted Hematite Particle Surfaces Jean-François Boily,*,† Merve Yeşilbaş,† Munshi Md. Musleh Uddin,† Lu Baiqing,† Yulia Trushkina,‡ and Germàn Salazar-Alvarez‡ †

Department of Chemistry, Umeå University, SE 901 87 Umeå, Sweden Materials and Environmental Chemistry, Arrhenius Laboratory, Stockholm University, SE 106 91 Stockholm, Sweden

‡

S Supporting Information *

ABSTRACT: Mineral surfaces exposed to moist air stabilize nanometer- to micrometer-thick water films. This study resolves the nature of thin water film formation at multifaceted hematite (α-Fe2O3) nanoparticle surfaces with crystallographic faces resolved by selected area electron diffraction. Dynamic vapor adsorption (DVA) in the 0−19 Torr range at 298 K showed that these particles stabilize water films consisting of up to 4−5 monolayers. Modeling of these data predicts water loadings in terms of an “adsorption regime” (up to 16 H2O/nm2) involving direct water binding to hematite surface sites, and of a “condensation regime” (up to 34 H2O/nm2) involving water binding to hematite-bound water nanoclusters. Vibration spectroscopy identified the predominant hematite surface hydroxo groups (−OH, μ−OH, μ3−OH) through which first layer water molecules formed hydrogen bonds, as well as surface iron sites directly coordinating water molecules (i.e., as geminal η− (OH2)2 sites). Chemometric analyses of the vibration spectra also revealed a strong correspondence in the response of hematite surface hydroxo groups to DVA-derived water loadings. These findings point to a near-saturation of the hydrogen-bonding environment of surface hydroxo groups at a partial water vapor pressure of ∼8 Torr (∼40% relative humidity). Classical molecular dynamics (MD) resolved the interfacial water structures and hydrogen bonding populations at five representative crystallographic faces expressed in these nanoparticles. Simulations of single oriented slabs underscored the individual roles of all (hydro)oxo groups in donating and accepting hydrogen bonds with first layer water in the “adsorption regime”. These analyses pointed to the preponderance of hydrogen bond-donating −OH groups in the stabilization of thin water films. Contributions of μ−OH and μ3−OH groups are secondary, yet remain essential in the stabilization of thin water films. MD simulations also helped resolve crystallographic controls on water−water interactions occurring in the “condensation regime”. Water−water hydrogen bond populations are greatest on the (001) face, and decrease in importance in the order (001) > (012) ≈ (110) > (014) ≫ (100). Simulations of a single (∼5 nm × ∼ 6 nm × ∼ 6 nm) nanometric hematite particle terminated by the (001), (110), (012), and (100) faces also highlighted the key roles that sites at particle edges play in interconnecting thin water films grown along contiguous crystallographic faces. Hydroxo−water hydrogen bond populations showed that edges were the preferential loci of binding. These simulations also suggested that equilibration times for water binding at edges were slower than on crystallographic faces. In this regard, edges, and by extension roughened surfaces, are expected to play commanding roles in the stabilization of thin water films. Thus, in focusing on the properties of nanometric-thick water layers at hematite surfaces, this study revealed the nature of interactions between water and multifaced particle surfaces. Our results pave the way for furthering our understanding of mineral-thin water film interfacial structure and reactivity on a broader range of materials.

1. INTRODUCTION

Knowledge of water binding mechanisms at hematite surfaces is, moreover, central for understanding activation mechanisms toward interfacial (photocatalyzed) electron transfer reactions that can occur between adsorbed species and the (semiconductive n-doped) bulk.9 Water splitting, for instance, is one notable reaction type where direct Fe−water interactions at hematite surfaces facilitate electron transfer,5 and thus one whose efficiency is strongly related to active water-binding sites.10 The importance of hematite has thus generated a vast range of studies devoted to its interfacial structure and

Atmospheric water vapor interacts with all solid surfaces present in nature, including minerals, bacteria and plants.1−4 Interactions between gaseous water molecules and hydrophilic functional groups outcropping solid surfaces can produce nanometer- to micrometer-thick layers with various degrees of organization. Films formed at mineral surfaces are strongly relevant to atmospheric, technological, as well as biogeochemical processes. These can contribute to processes as varied as biogeochemical cycling of elements, cloud formation, and ice nucleation, as well as (photo)catalytic reactions.5−8 Films at hematite (α-Fe2O3) surfaces are of particular interest given the widespread natural occurrence and the technological interest for this mineral. © XXXX American Chemical Society

Received: August 24, 2015 Revised: November 10, 2015

A

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

(Figure S2) were determined from the anisotropic peak width PXRD by the applying the so-called Popa rules implemented in Maud, and using up to the second degree expansion harmonic.29 2.1.3. Transmission Electron Microscopy. Low-resolution transmission electron microscopy (TEM) was used to determine particle size (Figures S3). These were obtained by measuring the maximum length between two vertices of hexagonal particles and between two edges for oval-shaped particles (Figure S3). Measurements were done on ∼200 nanoparticles using the image analysis software ImageJ.30 Selected area electron diffraction (SAED) under high resolution TEM (HRTEM) was used to identify crystallographic orientations (Figure S4). The images were collected with a JEOL JEM-2100F microscope with a Schottky-type field emission gun working at 200 kV ( f = 1.9 mm, Cs = 0.5 mm, Cc = 1.1 mm, point resolution =1.9 Å, lattice resolution =1.0 Å), and equipped with a Gatan Ultrascan 1000 camera (2048 × 2048 px2 and pixel size = 14 × 14 μm2) using a hardware binning of 2 × 2. 2.1.4. Vibration Spectroscopy. Fourier Transform Infrared (FTIR) spectra of N2(g)-dry hematite powders were collected using an Attenuated Total Reflectance (ATR; Golden Gate, single bounce diamond) accessory with a Bruker Vertex 70/V spectrometer (Section 2.2.2). Results (Figure S5a, lower spectrum) confirmed the absence of iron oxyhydroxide impurities in the sample, as can be especially appreciated by the absence of any of the characteristic Fe−O−H bending modes of goethite, akaganéite, lepidocrocite, or ferrihydrite. We note, however, that the presence of nonstoichiometric bulk hydroxyl and water, seen through O−H stretching and water bending modes in the samples exposed to 0 Torr H2O, is typical for this type of hydrothermally grown hematite.31 2.1.5. X-ray Photoelectron Spectroscopy. X-ray photoelectron spectroscopy measurements of the dry solids were carried out using a Kratos Axis Ultra electron spectrometer equipped with a monochromatic Al Kα X-ray source and a delay line detector. These measurements detected no impurities at the hematite surface other than the conventional carbon impurities associated with these measurements. The Fe 2p3/2 region was also characteristic of the multiplet structure of hematite. 2.1.6. Specific Surface Area. A N2(g) adsorption/desorption isotherm was collected using a TriStar (Micromeritics) instrument. Hematite power samples were dried in the reaction chamber at 383 K for 16 h under a stream of dry N2(g) prior measurements under liquid nitrogen temperature. Results of a 90-point isotherm were used to derive a B.E.T. specific surface area of 50.1 m2/g. 2.2. Water Vapor Binding Experiments. Water (H2O or D2O) vapor was made by passing dry N2(g) into a flow-through 316 stainless steel reservoir containing spectroscopic grade water. This saturated water vapor was blended with dry N2(g) in 1/4 in. 316 stainless steel tubes to produce water vapor pressures (pw) in the 0−19 Torr range at 298 ± 1 K. The gases were blended at different ratios using mass flow controllers (MKS, 179 A) producing a total and constant gas flow of 200 square cubic centimeter per min. A calibrated nondispersive infrared gas analyzer (Li-7000, Li-Cor) was used to track the resulting pw values throughout the course of the experiments. These gas mixtures were exposed to (1) a 54 μg dry hematite film deposited on a 10 MHz Au-coated quartz crystal for quartz crystal microbalance (QCM) measurements (Section 2.2.1) and (2) an ∼10 mg dry hematite film deposited on a diamond ATR cell for FTIR measurements (Section 2.2.2). 2.2.1. Water Loading Determination. A eQCM 10 M Quartz Crystal Microbalance (Gamry Instruments) was used to determine the mass of adsorbed water on hematite particle surfaces using the DVA approach.26 A thin hematite film was deposited by evaporation of an aqueous suspension on a 10 MHz Au-coated quartz crystal. This 54 μg film was dried under N2(g) in a flow-through cell over a 12 h period. It was thereafter exposed to H2O vapor pressures ranging from 0 to 19 Torr (2.53 kPa). The masses of the dry hematite sample and of the hematite−water mixtures were then determined from the series resonant frequencies of the quartz crystal using the Sauerbrey equation.32 Additionally, the parallel resonator frequency was used to ensure that the film was not too rigid during the course of the

(electro)chemistry, and involved synthetic, natural and doped specimens ranging from the nano- to the macro-scale.11−15 Xray reflectivity,13 tomography,16,17 atomic and scanning tunneling microscopy,15 crystal-rod truncation,18 X-ray photoelectron spectroscopy,12,19 electrochemical methods,20,21 as well as theory22,23 have provided particularly insightful details on hematite surface structure, topography, composition and reactivity. Water vapor binding has also been examined, for example by ambient-pressure X-ray photoelectron spectroscopy,12 vibration spectroscopy24 and theory.22,25 Still, an underlying and ongoing need in this area lies in predicting the individual and cooperative roles of coexisting water-binding sites that impart water film properties formed at hematite (nano)particle surfaces. This is particularly important considering that hematite surface structure controls the populations, densities, and spatial distributions of reactive hydroxo (−OH, μ−OH, μ3−OH) and Fe sites responsible for stabilizing thin water films. As such, multifaceted hematite particles could possibly host and interconnect thin water films with contrasting properties. In this study, we address the impact of hematite surface structure on thin water film formation using synthetic hematite nanoparticles with well-developed crystallographic faces. We first identify representative crystallographic faces terminating these particles using selected area electron diffraction (SAED), predict plausible populations of Fe and hydroxo surface groups responsible for binding water, and identify these groups using vibration spectroscopy. We then use the dynamic vapor adsorption (DVA)26 method to monitor water loadings and the vibration spectroscopic responses of the hematite surface to water adsorption. We also describe the molecular-level nature of these interactions in terms of hydrogen bonding populations derived by classical molecular dynamics (MD) simulations of representative crystallographic faces. Finally, we present MD simulations of a single nanosized hematite particle to study competitive water adsorption at coexisting crystallographic faces, as well as to identify the roles of that particle edges play in interconnecting thin water films grown from contiguous crystallographic faces.

2. EXPERIMENTAL AND COMPUTATIONAL METHODS 2.1. Hematite Synthesis and Characterization. 2.1.1. Synthesis. Nanosized hematite particles were synthesized by a forced hydrolysis method.27 A 200 mL aliquot of a 0.2 M FeCl3 solution was added at a rate of 2.9 mL/min to 2 L of a 2 mM HCl solution heated to 371 K. The solution was continuously stirred with a propeller throughout the ∼70 min procedure while kept under an atmosphere of N2(g). The resulting suspension was then aged for 2 weeks at 371 K, then dialyzed until the resistivity of the supernatant was comparable to that of MilliQ water (18.2 MΩ·cm). All synthesis procedures were carried out in polyethylene bottles. A portion of the resulting suspension was dried at 313 K, then ground for characterization. 2.1.2. Powder X-ray Diffraction (PXRD). Hematite powder was glued with silicon grease onto a thin glass fiber, and PXRD patterns were collected in transmission mode with a dark current mode with Xcalibur III Single Crystal diffractometer (Mo Kα1 radiation with λ = 0.709317 nm), equipped with a two-dimensional Sapphire 3 CCD detector (2048 × 2048 px2 and pixel size = 30 × 30 μm2). The data (Figure S1 of the Supporting Information) were the average of four scans, each with a 2θ range of 0−49° (d = 6.97−0.85 Å), exposure time of 2400 s, and sample-to-detector distance of 120 mm. The instrumental width was determined through the calibration of the peak width of lanthanum hexaboride (LaB6) powder, which was measured using the same procedure. Indexing and Rietveld refinement were performed using the software package Maud.28 Particle size and shape B

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

absorbance at 3750 cm−1, where absorbances from the samples were negligible, then baseline-corrected using a polynominal function to remove contributions from nonstoichiometric bulk OH and H2O. This procedure flat-lined the spectra and improved resolution of the surface OH stretching bands. A singular value decomposition and a factor indicator function35 analysis of the matrix Am×n were used to determined the number of spectral components (s) required to reproduce a given variance of the data. MCR analyses were thereafter performed to obtain εm×s and Cs×n from Am×n, a procedure that rotates the s-first dominant eigenvectors of Am×n into a real chemical space, such that such Am×n = εm×s.·Cs×n + E, where εm×s ≥ 0 and Cs×n ≥ 0, and E is the matrix of unaccounted residuals. These calculations were made using the program MCR-ALS34 in the computational environment of MATLAB 8.0 (The Mathworks, Inc.). 2.3. Molecular Dynamics. Classical molecular dynamics (MD) were used to simulate water binding at crystallographically oriented slabs as well as a single nanometric hematite particle. Interfacial reactions were simulated at neutrally charged surfaces, and therefore under conditions of their respective points of zero charge. The pH of the thin water film was, however, not taken into account, as inclusion of hydronium or hydroxide species to the adjacent bulk water solution would require substantially larger simulation cells to achieve the correct concentrations. We also note that because no electrolyte counterions are present at the interface, little or no charge ought to develop. Thus, our assumption that all surfaces are neutrally charged is likely the most reasonable one that could be taken, especially given the lack of experimentally determined values of pH in thin water films. We also note that because protonation states remained fixed throughout of the course of the simulations, the approach used in this work cannot take into account possible effects of proton transfer reactions, as OH and H2O are nondissociable. Still, by starting from the most plausible hematite terminations reflecting our experimental conditions, these simulations can provide a highly insightful view of hematite/water interfacial structure and phenomena. Furthermore, the computational efficiency of this method enables us to track competitive processes in a large nanometric single particle. Bulk and surface structures were generated starting from the crystallographic structure proposed by Finger and Hazen.36 Neutrally charged slabs oriented along the (001), (100), (110), (012), and (014) faces were ∼3 nm thick slabs, contained 1152−1792 Fe atoms and had surface area of 11−17 nm2 on the upper and under sides. They were cut at their edges so that their structures could be perpetuated by periodic boundary conditions. All surfaces were oxygen-terminated to emulate ambient atmospheric conditions, and their protonation states evaluated according to their bond valence states, as described in Venema et al.37 The upper and undersides of the slabs were thus terminated by surface (hydr)oxo groups to complete the coordination environments of surface Fe and O atoms. Simulations were also carried out on a single ∼5 nm × ∼ 6 nm × ∼ 6 nm large hematite particle consisting of 7020 Fe atoms that was terminated by the (001), (100), as well as the isostructural (110) and (120) faces. Surface oxygen populations were of 4 × 130 O on the (001) faces, 2 × 259 O on the (110) faces, 2 × 273 O on the (120) faces, and of 307 O at the edges. Their protonations states were also identical to those of the slabs. We note that, while O populations on the isostructural (110) and (120) faces were nearly identical, they are separated over four planes in the former and two planes in the latter. The slabs were connected to a ∼ 5 nm thick void, while the multifaceted particle submerged in the middle of a 20 nm × 20 nm × 20 nm simulation box. The voids were subsequently filled with various concentrations of randomly distributed single point charge (SPC)38 water molecules using the GENBOX program of GROMACS (v.4.6.5).39 Simulations were carried out using classical MD with the CLAYFF27 force field for hematite, however, using modified parameters for Fe suggested by Kerisit23 for iron, and the flexible SPC model for water.38 Surface O−H bond frequencies were constrained to 450630 kJ/mol (3660 cm−1) for −OH, 440830 kJ/mol (3620 cm−1) for μ−OH and 448170 kJ/mol (3650 cm−1) for μ3−OH sites based on previous work from our group.40 A NPT (constant number of particles, constant pressure, and constant temperature) ensemble and a time step of 0.5 fs

experiment. Continuous monitoring of these frequencies helped establish that a 20 min equilibration period was sufficient to obtain time-independent values at any given pressure, suggesting that steadystate or equilibrium with respect to water vapor binding had been achieved. Reproducibility and reversibility of the water vapor adsorption experiments were confirmed in a suite of preliminary tests. The pw dependence on hematite surface water loadings was modeled using the Do−Do33 water sorption isotherm from the DVA data: n=α+1

Cμ = Cμs

Kμ ∑1 n=α+1

Kμ ∑1

pw n n=α+1

pw n + Kμ ∑1

pw n − 1

n=β+1

+ So

npw n n=β+1 n K f ∑1 pw

K f ∑1 1+

(1)

The left-hand term of this equation accounts for condensation of water vapor at the hematite surface, while the right-hand term accounts for adsorption reactions with the hematite surface. Parameters for condensation include saturation concentration (Cμs), association constant (Kμ), and size of water cluster (α = 6 according to Do and Do33). Those for adsorption include crystallographic densities of water-binding sites (So), association constant (Kf) and number of hydration water per site (β). A value of β = 1 was chosen to reflect the capability of first-layer water molecules to pack to a comparable density to the crystallographically available surface oxygens. Parameters were co-optimized to fit the experimental DVA data using trust region reflective algorithm with a code written in the computational environment of Matlab 8.0 (The Mathworks, Inc.). 2.2.2. FTIR Spectroscopy and Chemometric Analyses. FTIR spectroscopy was used to monitor O−H stretching modes of hematite surface hydroxo groups interacted with water vapor. All spectra were collected with a Bruker Vertex 70/V FTIR spectrometer, equipped with a DLaTGS detector in a room kept at 298 ± 1 K. Measurements were carried out in the 600−4500 cm−1 range at a resolution of 2.5 cm−1 and at a forward/reverse scanning rate of 10 Hz resulting in 1000 coadded spectra for each sample. We used the Blackman-Harris 3-term apodization function with 16 cm−1 phase resolution and the Mertz phase correction algorithm. A 2 mL aqueous dialyzed suspension of hematite was first equilibrated in a polyethylene test tube under N2(g) for 24 h at 298 ± 1 K, then centrifuged and decanted. An aliquot of the wet paste was then applied on the diamond window of the ATR accessory using a pipet, then dried under a stream of N2(g) at 298 ± 1 K. FTIR spectra were collected during the drying period to track the removal of water and the point at which the sample was completely dried, namely, when the intensities of the hematite bulk water bending modes could not be diminished any further by N2(g). This procedure was carried out over a 12 h period to ensure full dryness. The resulting dry ∼10 mg sample was then covered with a closed-loop flow-though reaction cell for the water vapor adsorption isotherm experiment. In this experiment, the sample was exposed to water (H2O or D2O) vapor pressures from 0 to 19 Torr (2.53 kPa) at 298 ± 1 K, and during which time the FTIR spectra of the samples were continuously collected. Equilibrium at each preselected water vapor pressure was typically established over a 10−15 min period, and identified by time-independent spectra, one of which was selected as a representative spectrum. The water vapor pressure was then changed to another value after a total exposure time of 20 min. Reproducibility and reversibility of the water vapor adsorption experiments were confirmed in a suite of preliminary tests. Multivariate curve resolution (MCR)34 analyses were carried out to facilitate interpretation of the changes in the FTIR spectra collected in the 0−19 H2O torr range (Am×n, containing m wavenumbers and n water vapor pressures). MCR extracts spectral components (ε), and their respective concentration profiles (C) through the relationship A = ε·C. These spectral components are thus akin to molar absorption coefficients of relatively pure chemical species, scaled for an undetermined optical path length. In order to execute this procedure, all spectra in the 3400−3750 cm−1 region were first offset to zero C

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir were used with the Verlet algorithm41 to integrate the equations of motions for all the atoms in the system, which were projected using a periodic boundary condition. The temperature of the system (300 K) was coupled to the Nosé−Hoover42 velocity-rescale thermostat with a 0.1 ps relaxation time. Pressure (1 bar) was coupled by anisotropically scaling the simulation box with the Parrinello−Rahman43 method. The O−H bonds of all the hydroxyls and water were treated by the LINCS44 algorithm, whereas the SETTLE45 algorithm was used to treat the geometry of flexible water. A 0.8 nm cutoff was used for nonbonded van der Waals interactions, and the particle-mesh Ewald46 method was used to treat long-range electrostatic interactions. Simulations cells were first energy-minimized (double precision) using a steepest descent algorithm, typically in less than 104 steps. The minimized structure was then equilibrated (single precision) using classical MD for at least 107 steps (5 ns), followed by production runs of at least 5 ns. Total energy convergence and its components, as well as temperature and atomic densities in both slabs and particle were monitored for these entire equilibration periods. This modeling strategy adequately reproduced the ideal hematite bulk within the uncertainties discussed in the work of Kerisit.23 Although a certain extent of surface relaxation is to be noted, the relative positions of all top Fe and O atoms in all slabs and in the particle are retained throughout the course of all simulations and represent the crystallographic terminations of hematite. Production runs were used to obtain salient details of the hematite/ water vapor interface. Hydrogen bond analyses were carried out by recording donor−hydrogen−acceptor distances, angles and populations for all bonds with donor−acceptor distances below 0.3 nm and donor−hydrogen-acceptor angles below 30°. Hydration numbers were obtained from the cumulative coordination number of the oxygen of a hydroxo group for neighboring water oxygens to a cutoff of 0.3 nm. This was computed using a radial distribution function for the same atom pair, from which this cutoff was determined. In a subsequent set of calculations, the optimized structures were used to generate power spectra of surface hydroxo groups. These power spectra were obtained from 60 ps simulations in which LINCS constraints on all hydroxyl (O−H) bonds were turned off. These simulations were performed with a 0.5 as (attosecond) time step with a total of 5 × 107 steps for each production run collected after an initial run of 1 × 107 steps. Power spectra were then generated by Fourier transformation of the hydrogen velocity autocorrelation function, Rx(t), with47

Figure 1. Water vapor adsorption isotherm on hematite. HRTEM image of a representative particle is shown in the inset. The data are modeled with the Do−Do33 water sorption isotherm (eq 1). A water density of 16.0 site/nm2 was extracted from this modeling procedure, a value that is in agreement with predicted crystallographic OH and Fe densities on the particles under study. The condensation regime requires a saturation concentration of 34 sites/nm2.

surface-bound water clusters at the hematite surface. Working under the assumption that one monolayer contains approximately 10−13 H2O/nm2 (taken from various packing motifs within a single layer of 0.28 nm-wide water molecules) modeling points to the formation of a thin film consisting up to 4−5 monolayers under the conditions of this study. We also note that because both adsorption and condensation regimes overlap over a broad range of pw values, the model suggests that water uptake may not necessarily proceed as a monolayer-bymonolayer growth process. It may possibly proceed via the concomitant hydrogen bonding of water vapor to hydroxo groups and growth of surface-bound water (nano)clusters via water−water interactions. Full coverage could be then achieved by coalescence of separate water clusters into one thin water film. Vibration spectroscopic measurements of the same particles exposed to water vapor (Figure 2a) revealed the appearance of liquid-like bound water molecules through O−H stretching (∼3300 cm−1) and bending (1635 cm−1) region, and through concomitant changes in the O−H stretching bands of surface hydroxo groups (3400−3750 cm−1), to be discussed in Section 3.1. Still, no bulk liquid water was detected in these experiments, as pw was kept below pw,sat. We confirmed this further in separate sets of experiments were liquid water was purposefully formed at pw,sat (not shown). The difference spectra of Figure 2b show more clearly the impact of water binding in the pw 0−19 Torr region by removing the contributions of nonstoichiometric hydroxo and aquo groups of the hematite bulk from these raw spectra (Figures 2a for the stretching region; Figure S5c for the bending region). These difference spectra revealed thin water films with O−H stretching frequencies at 3134 and 3341 cm−1, and therefore with subsets of hydrogen bonding networks of smaller (3341 cm−1) and larger (3134 cm−1) strength than in liquid bulk water (Figure 2b, upper spectrum). Exposure of dry hematite surfaces to D2O(g) confirms this finding further through isotopically shifted bands of 3434 cm−1 to 2507 cm−1 and of 3341 cm−1 to 2351 cm−1 (Figure 2c). We also note that the isotopic shifts of the nonstoichiometric hydroxo/aquo groups

∞

Sx(f ) =

∫−∞ R x(t )e−2πift

(2)

where f is frequency. This equation was solved numerically by employing a discrete Fourier transformation to Rx(t) in Matlab 8.0 (The Mathwork, Inc.).

3. RESULTS AND DISCUSSIONS Water vapor uptake by hematite particle surfaces (Figure 1) can be classified as a Type II (IUPAC) adsorption isotherm.48 It is characterized by (1) a Langmuirian-like “adsorption regime” predominating below water vapor pressures of pw ∼ 6−8 Torr, where surface hydro−water interactions prevail, and (2) a water vapor “condensation regime” at higher pressures involving water−water interactions at the hematite surface. We note that the data do not extend to pw saturation pressures (pw,sat) where liquid water is formed in the system. Modeling of these data, using the Do−Do model33 of eq 1, suggests that both regimes operate over a broad range of partial water pressures. The model suggests a maximal water binding density of So = 16.0 H2O/nm2, a value that is highly comparable to the crystallography surface density of oxygens. The condensation saturation concentration of Cμs = 34.0 H2O/ nm2 corresponds to the water density accumulated by water− water interactions, namely, water binding on pre-existing D

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

Figure 2. (a) Absolute and (b,c) difference ATR-FTIR spectra of synthetic hematite particles exposed to H2O (a,b) and D2O (c) vapor in the 0−19 Torr range. Difference spectra were obtained by subtracting raw spectra (a) from the reference at 0 Torr (a, bottom). Letters correspond to surface hydroxo/aquo groups (Table 1): G (geminal H2O), S (singly coordinated OH), D (doubly coordinated OH), and T (triply coordinated OH). See Figure S5c for water bending region.

of hematite, seen through broad 2800−3800 cm−1 region, suggest that the thin water film can also exchange with nonstoichiometric hydroxo-bearing components of the near hematite surface. The bands of surface hydroxo groups in the 3400−3750 cm−1 region reveal further detail on the nature of water adsorption reactions (Figures 2a and 3). This collection of

175 cm−1 per pm), they are highly sensitive indicators of slight changes in hydrogen bond environments. In this study, we make use of this sensitivity to identify water-binding sites at hematite surfaces. We start by assigning bands of the 3400− 3750 cm−1 region to hydroxo groups of different crystallographic faces (Section 3.1), then track their vibrational spectroscopic responses to water vapor loadings. We thereafter correlate these changes to hydrogen bonding populations pertaining to hydroxo−water (Section 3.2) and water−water (Section 3.3) interactions studied by MD simulations of crystallographically oriented slabs. Finally, we use MD simulations of a single nanometric hematite particle to monitor competitive water vapor binding at coexisting crystallographic faces, as well as to address the roles of edges in stabilizing thin water films (Section 3.4). 3.1. Hematite Surface Sites. Identification of waterbinding sites requires an appreciation of the dominant crystallographic faces of the particles under study. TEM imaging (Figures 1, S3, and S4) first showed that the particles were monodispersed in size (maximum Feret diameter of 28 ± 7 nm (1σ)) and multifaceted in shape. SAED measurements of these particles, detailed in the Supporting Information (Figure S4), revealed that the particles tends to lie on their [001] zone axis parallel to the electron beam, and are bound laterally by several faces, including the {110}, {210}, {120}, and {014} faces. We note that while the preferential orientation of the particles impeded the determination of the (001) by SAED, the presence of this face was strongly supported by modeling of the PXRD data (Figure S2). Idealized crystallographic representations of five selected crystallographic terminations (Figure 4), including the commonly occurring (012) face in hexagonal particles,51 revealed contrasting populations and dispositions of −OH (S), μ−OH (D), and μ3−OH (T) groups. While the (110) and (014) faces ideally display all three groups, the (012) face display only −OH and μ3−OH groups, and the basal (001) face only μ−OH groups. The (100) face exhibits surface Fe sites capable of forming geminal sites (η-(OH2)2), namely two aquo groups bound to the same iron octahedron. The populations of these groups are, in contrast to hematite, of minor importance in iron oxyhydroxides, where they are confined to tips of acicular particles.52,53 Finally, we note that typical oxygen site densities at hematite surfaces of 13−15 sites/nm2 are highly consistent with the So adsorption maximum extracted by modeling (eq 1) of the adsorption isotherm (Figure 1).

Figure 3. Representative O−H stretching regions of goethite (GT)53 and lath lepidocrocite (LL)53 in relation to that of hematite (H) particles considered in this work. Band assignments previously made for the iron oxyhydroxides are applied to hematite. Red and green spectra are obtained from the Fourier transform of the hydrogen autocorrelation function of classical MD simulations presented in this work for −OH groups of the (012) face (red) and μ−OH groups of the (001) face (green). Their absolute calculated positions were shifted by 20−25 cm−1 to match the experimental peaks. The triplet for the (001) face is explained by three O−H bond strengths caused by three distinct H-bond environments at that surface. Note that this triplet was generated theoretically from one single O−H bond strength, and thus the splitting was induced by the distinct H-bond environments. Molecular configurations and H-bond patterns (dashed lines) of representative surfaces are shown on the right.

narrow O−H stretching bands, with full width at half maxima of about 20−50 cm−1, arise from −OH, μ−OH, and μ3−OH groups with specific sets of hydrogen bonding environments. They are most pronounced under dry conditions, and have strong similarities with counterpart groups at iron oxyhydroxide surfaces previously studied in our group49 (Figure 3). Furthermore, as previous studies49,50 showed that these bands are strongly responsive to changes in O−H bond length (150− E

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

Figure 4. Snapshot of MD simulation (top and middle row) and selected water density profiles (bottom row) of the five dominating crystallographic faces of the hematite particles under study. Classical MD simulations were carried out with the CLAYFF27 force field using flexible SPC water molecules (NPT, 300 K, 1 bar). Water density profiles generated by the simulations (Figure 4) also showed that low Miller index faces (i.e., (001), (100), (110)) developed a separate first water monolayer, while higher index faces (i.e., (012), (014)) two closely spaced layers. These contain about 9 H2O/nm2 on low and 6 H2O/nm2 on high Miller index faces. We also note that they are ideally organized as two-dimensional arrays containing rings of 5−8 water molecules (Figure 7), a result caused by the templating effects of organized surface hydroxo groups with which they form hydrogen bonds. Greater water loadings produce, in contrast, more disorganized second to fourth layer water molecules with populations ultimately surpassing crystallographic hydroxo site densities.

OH groups of the (001) face, while the singlet generated by lepidocrocite arises from a single dominant H-bonding environment. These differences are readily captured by MDgenerated power spectra (Figure 3, labeled as “theoretical” and generated from eq 2).40 Finally, μ3−OH groups of the hematite surface are present as isolated OH groups (3550 cm−1), such as in lepidocrocite (Figure 3), and hydrogen bonded (3490 cm−1) sites, such as in goethite (Figure 3).53 MD simulations suggest that isolated μ3−OH groups are from the (110) face but that hydrogen bonded groups are from the (014) and (012) faces of hematite (Table 1). 3.2. Water-Binding in the “Adsorption Regime”. 3.2.1. Vibration Spectroscopy. Building upon the band assignments made in the preceding section, the spectral changes of the O−H stretching region (Figure 5a) are interpreted on the basis of face- and site-specific water-binding mechanisms. To help achieve this goal, we performed chemometric analyses of background-subtracted absorption data of the 3400−3750 cm−1 region, shown in Figure 5a. Singular value decomposition and MCR-ALS analyses (Section 2.2.2) showed that the first two components (eigenvectors) could reproduce 95% of variance of the absorption data (Am×n; Section 2.2.2). While inclusion of a third component improved the representation of the data to 97% of the variance, rotation of their eigenvectors to a real chemical space by MCR-ALS could not converge. We consequently modeled the data using two components (εm×2 and C2×n; Section 2.2.2), with the quality of the fit to the data demonstrated in Figure 5a. In this model (Figure 5b), one spectral component (ε) corresponds to the dry hematite particle surfaces equilibrated at 0 Torr H2O (cf. Figure 5b vs Figure 3, bottom), and a second component corresponding to humid hematite equilibrated at pw below the water saturation pressure (pw,sat = 23.77 Torr at 298 K). Salient spectral features of this latter component include attenuated bands for −OH (S), μ−OH (D) and μ3−OH (T). The spectral changes arise from hydrogen bonding of water to these groups,

Starting from this idealized depiction of surface (hydr)oxo and Fe populations, we assigned the surface O−H stretching bands of the 3400−3750 cm−1 region. These efforts were constrained further by comparison with our previous efforts on well-characterized synthetic FeOOH minerals53 (Figure 3), the details of which summarized in Tables 1 and S1. One of our Table 1. Band Assignment of Surface Hydroxo/Aquo Groups on Hematitea surface group

symbol

band position

hkl

η−OH2 −OH (isolated) −OH···−OH···−OH ···μ−OH (basal) μ3−OH (isolated) μ3−OH···−OH

G (geminal) S (singly) S (singly) D (doubly) T (triply) T (triply)

3695 3672 3661 3634/3620/3609b 3532 3493, 3420

(100) (014) (110), (012) (001) (110) (014) , (012)

a

Stretching frequencies pertain to OH group in bold. H-bonding (···) environment given along rows of −OH groups. bTriplet caused by Hbonding environment of μ−OH (cf. Figure 3; cf. Table S1 for band assignments in relation to those of hydroxo groups at FeOOH surfaces).

recent studies (Boily and Song53) contains further details on the approach used to assign these bands. Our interpretations suggest that −OH groups of dry hematite (Figure 3 bottom; background-subtracted spectrum of Figure 2a bottom) are present as poorly (3678 cm−1) and partially H-bonded (3667 cm−1) sites (Figure 3). MD-derived hydrogen bond populations of representative surfaces (Figure 4) also suggest that isolated OH groups could be from the (014) face, while partially hydrogen bonded groups are from the (110) and (012) faces. Next, μ−OH groups of the (001) basal face are detected as a triplet centered at 3620 cm−1, a value that exactly matches a singlet generated by μ−OH groups of the basal (010) face of lepidocrocite.53 The triplet of hematite arises from three distinct H-bonding environments adopted by μ− F

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

Figure 5. (a) The 3400−3750 cm−1 region after background subtraction for nonstoichiometric bulk OH and water as well as water vapor in the 0− 19 Torr region in (2.15 kPa) and water vapor mixed in 102 kPa N2(g) at 298 ± 1 K. The data matrix A (a) relates to MCR components ε (b) with the concentration profile C (c) through the Beer−Lambert−Bouguer law of transmission of light in materials (A = εC). Absorbance data in panel (a) are shown as points (•) and MCR model as lines. Letters in panel a correspond to surface hydroxo/aquo groups (Table 1): G (geminal H2O), S (singly coordinated OH), D (doubly coordinated OH), T (triply coordinated OH). Red and green spectra in (b) represent MD-derived power spectra showing the doublet from the −OH groups of the (012) face and the triplet from the μ−OH groups from the (001) face. See Figure S6 for full set of power spectra. Concentration profiles (c) were normalized for value at saturation (8 Torr). MCR component fraction and DVA-derived water loadings are also shown as a function of pw (torr H2O). MD-derived total H-bond populations at different crystallographic faces are shown as a function of water loading (H2O/nm2) by scaling 1.8 SPC-H2O/nm2 per torr H2O.

Figure 6. H-bond populations at four crystallographic faces of hematite exposed to various water loadings. Obtained from MD simulations (NPT, 300 K, 1 bar, 1 ns simulation with 107 steps).

nm2 via water−water binding (“condensation”), and therefore of 1.3 H2O/nm2 of unbound hydroxo groups. We note that the absence of vibrational signatures of the latter can be explained by the strong cooperative nature of hydrogen bonding networks at the hematite surfaces. We have, in fact, previously described this phenomenon for rows of −OH groups affected by protonation,53 hydrogen bonding,53 as well as water vapor54,55 and carbon dioxide56 binding. In the following sections we will present further details on the populations of hydrogen bonds at the hematite surface. 3.2.2. Molecular Modeling. MD simulations of the five representative crystallographic faces under study provided additional constraints to our understanding of water binding mechanisms. First, the water density profiles of Figure 4

thus red-shifting their vibrational frequencies to broad ranges of indistinguishable values. Still, the second component contains bands resulting from hydroxo groups with distinct hydrogen bonded environment, as a new band at 3694 cm−1 (Figure 5b) corresponding to the coordination of water molecules at bare Fe sites of the (100) face.40,52,54 The pw dependence of the MCR-derived concentration (C) profile underscores the sensitivity of the spectral response of the O−H stretching region of surface hydroxo groups to water binding (Figure 5c). This profile reveals the complete disappearance of all original bands of the dry hematite surfaces at pw ∼ 8 Torr. According to adsorption modeling (eq 1), this corresponds to an occupation of ∼12.9 H2O/nm2 via direct surface hydroxo−water binding (“adsorption”) and 1.8 H2O/ G

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

concentration (Cs×n) versus pw, as well as (4) the MD-derived hydrogen bond populations versus theoretical water loadings (H2O/nm2). As stated in Section 3.2.2, the DVA-derived water loadings strongly correlate with a loss in the surface hydroxo O−H stretching frequencies. We can also obtain a congruent response with the MD-derived hydrogen bond populations by empirically scaling pw to 1.8 SPC-H2O/nm2 per torr H2O. The relationship of Figure 5c thus underscores the notion that water binding to hematite surfaces systematically and concurrently shifts the vibrational modes of surface hydroxo groups and its hydrogen bonding populations under the “adsorption regime”. Hematite-bound waters then become additional water-binding sites under the “condensation regime”, which becomes the predominant process at higher pw. 3.3. Water−Water Interactions of “Condensation Regime”. The “condensation regime” pertains to water binding on hematite-bound water molecules, resulting in multilayered thin film. As stated earlier in this paper, these hematite-bound water molecules acquire vibrational modes that are comparable to liquid water yet exhibit a subset of larger (3134 cm−1) and smaller (3341 cm−1) hydrogen bond strengths (Figure 2). Hydrogen bonding populations analyses (Figure 7a) add further insight into this matter. Normally, MD simulations of 1

(bottom) revealed the strong impact of crystallographic orientation on interfacial water structure, namely, with facespecific first-layer water densities, distinct positions and densities of juxtaposed adsorption layers. For example, the (012) and (014) faces display molecular-level corrugations stabilizing two-closely spaced monolayers of water at the near hematite/water interface (Figure 4). The (001) and (110) faces promote, on the other hand, a single distinct first layer. In contrast to all other faces considered in this work, the (100) face binds first-layer water molecules via direct coordination to bare Fe sites, achieving a H2O:Fe ratio of 1.6:1.0 at saturation, a value lower than the expected geminal (η-(OH2)2) sites. The second water layer at this interface is in turn exclusively controlled by water−water interactions. This strong crystallographic controls on water binding can also be appreciated in the MD-derived hydrogen bond populations between hydroxo groups and first-layer water molecules of the “adsorption regime” (Figure 6). At all surfaces, hydrogen bond populations undergo steep hikes with water loadings, reaching a plateau at levels close to the crystallographic site density. Water binding promotes, on the other hand, only small declines in the populations in intersite hydrogen bond interactions (e.g., −OH···−OH) that characterize dry hematite surfaces. While most hydroxo groups are both donating (e.g., −OH···OH2) and accepting (e.g., −O···H2O) hydrogen bonds with interfacial water molecules, steric constraints promote a stronger predisposition for donation, and therefore for the preferential orientation of first layer water hydrogens toward the gas phase. We note that −OH groups are the predominant hydrogen bonding sites (1.1−1.6 H-bond per − OH at first water layer saturation). Faces exhibiting these groups (i.e., (110), (012), and (014)) are therefore strong hydrogen bond donors (donor-to-acceptor ratios of 1.9−2.9). Hydrogen bonding involving μ−OH groups is the second most important water binding sites, with 0.4−0.7 hydrogen bond per μ−OH at saturation. We note that μ−OH groups of the (001) face have a strong propensity for both donating and accepting hydrogen bonds (donor-to-acceptor ratio of 1.3), hence the preferential orientation of first-layer water hydrogens toward and parallel to the hematite surface. Finally, μ3−OH groups are the least involved in hydrogen bonding, with ∼0.3 hydrogen bond per site at saturation. The impact of water binding on the O−H bond strength of μ−OH groups of the (001) face and of −OH groups of the (012) face were resolved in MD-derived theoretical power spectra (Figure 5b, full suite is in Figure S6). These spectra show that a distinct network of hydrogen bonds involving μ− OH (D) groups generating the triplet of the (001) face under dry conditions is disrupted by water adsorption, and results in a singlet-like state at high water loadings. Those involving −OH (S) groups of the (012) face predict, on the other hand, strongly attenuated and shifted O−H stretching frequencies upon water binding. They show, at the same time, that a portion of the bands still persist at high water loadings, as notably seen in the second MCR component pertaining to humid hematite (Figure 5b). These results therefore provide a link demonstrating how hydrogen bonding patterns and populations impact spectral shifts measured in the laboratory. Another link can also be made by comparing DVA, FTIR, and hydrogen bond populations of the “adsorption regime”, as shown in Figure 5c. This diagram compares (1) DVA-derived water loadings (H2O/nm2) versus pw, (2) adsorption modeling (eq 1) of these data versus pw, (3) the MCR component

Figure 7. (a) Hydrogen bond populations for water−water interactions, pertaining to water molecules forming one (1 HB), two (2 HB), and three (3 HB) bonds per H2O. Arrows on the right-hand ordinate axis show the bulk SPC water limit for 1 HB, 2 HB, and 3 HB populations. (b) Snapshots of MD simulations showing representative water clusters at hematite surfaces. Obtained from MD simulations (NPT, 300 K, 1 bar, 1 ns simulation with 107 steps).

g/cm3 of flexible SPC waters at 300 K retrieve an overall hydrogen bond population of ∼3.4 hydrogen bonds (HB) per H2O, a number that can be broken down to 1.44 as 1 HB, 0.52 as 2 HB, and 0.30 as 3 HB per H2O. While populations of hematite-bound waters reach those values at water loadings exceeding crystallographic site densities, they are substantially lower below these densities (Figure 7a). These altered populations are, as emphasized in the previous section, a consequence of the predominant surface hydroxo−water interactions. Still, water−water hydrogen bond populations increase sharply with water loading are greatest on the (001) face, and decrease in importance in the order (001) > (012) ≈ (110) ≫ (100). This order is a direct consequence of the structuring of first layer water resulting in, for example, ring-like water clusters at the (001) face (Figure 7b). At the other end of this series, Fe-bound water molecules of the (100) face are the least prone to water−water interactions given the orientational constraints imposed on first layer water molecules (Figures 4 and 7b). Finally, populations of water molecules involved in 3 H

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

Figure 8. Simulation cell (a) and MD (b−e) snapshots (looking down the (001) face) of the OH-terminated ∼5 × ∼ 6 × ∼ 6 nm α-Fe2O3 multifaceted particle. Classical MD simulations were carried out with the CLAYFF27 force field using flexible SPC water molecules (NPT, 300 K, 1 bar). The sequence of snapshots at 1.5, 5.0, and 15 H2O/nm2 (b−-d) illustrate the preferential accumulation of water at the (100) face, and the formation of a two-dimensional distribution of water at the (001) face. Water molecules located at edges in the simulation of 5 H2O/nm2 are highlighted in panel e.

HB (Figure 7a) do not reach the pure flexible SPC water limit within the water loadings considered in this work. This is imposed by the limited number of water layers, impeding the extensive three-dimensional distribution of hydrogen bonds of liquid water. The trends of these data sets, however, strongly suggest that loadings that are only slightly above ∼40 SPC H2O/nm2 should reach this point. 3.4. Simulation of Water Binding at a Single Multifaceted Hematite Particle Surface. While MD simulations of idealized single crystallographic faces depicted plausible interfacial water structures and hydrogen bonded populations in thin water films, they cannot on their own fully account for water binding on multifaceted surfaces. Particle edges are of particular interest in this regard as they are regions where films grown from contiguous faces interconnect, forming a continuous thin water film over an entire particle surface. We consequently sought to expand our MD efforts on oriented slabs to a single ∼5 nm × ∼ 6 nm × ∼ 6 nm particle terminated by the (001), (100) as well as the isostructural (110) and (120) faces exposed to water loadings ranging from 1.5 to 15 H2O/ nm2 (Figure 7b−d). Simulations first underpinned the major and systematic impact of water binding during the first phases of equilibration as water molecules approached and bound to hematite surfaces (Figure 9a). This was strongly correlated with drops in the total energy of the system (Figure 9a). Water binding took place in a stepwise fashion as various clusters of gas-phase water, formed in the early stages of equilibration, and sequentially came in contact with hematite. These interactions, just as in simulations of the oriented slabs, took place in the form of hydrogen bonding with surface hydroxo groups on the (001), (110), and (120) faces, through direct coordination on surface Fe sites of the (100) face, as well as through water−water interactions. The latter can be appreciated by two-dimensional arrays of firstlayer waters in the MD simulation snapshots in Figure 8. Surface hydroxo−water binding reactions equilibrated most rapidly at the (100) face, where direct Fe−H2O binding occurs. Equilibration was, however, slower on the other faces, perhaps due to greater rates of bond formation/rupture. Edges were, in contrast, the slowest loci of equilibration with respect to Hbonding (Figure 9a). Hydrogen bond populations of the system equilibrated at ≫5 ns, where energies were simulation timeindependent, showed that edges were the preferential loci of binding, with populations decreasing in the order edge > (001) > > (110) > (120). Hydration numbers of edge sites were also up to ∼2 H2O per site (Figures 9b and S7 for details) at the highest water loadings, a value that is on par with sites of the (001) plane but greater than those of the (120) (1.5 H2O/site) and (100) (1.2 H2O/site) faces. The greater populations of

Figure 9. (a) Example of total system energy and H-bond evolution during a typical classical MD simulation of the ∼5 × ∼ 6 × ∼ 6 nm αFe2O3 multifaceted particle (5 H2O/nm2). Production runs were obtained for simulations in the 5−10 ns range where the energy profile is time independent (not shown). (b) Hydration number of sites obtained by radial distribution function analysis (Figure S7)).

sites at edges arises from the larger possibilities for binding than for equivalent sites at flat surfaces. A previous study,57 employing a dissociable water MD model, also suggested that this also leads to a larger proton-active site density at edges. Finally, we note that populations of the isostructural (120) and (110) faces are different, even if oxygen populations on the particle are nearly identical (Section 2.3). This difference stems from the stronger impact that edges exert on water binding on the (110) face, as its sites are distributed over four smaller terminations of the particle, while those of the (120) face are expressed over only two terminations. This observation consequently further illustrates the influence that edges exert on neighboring sites of a face. As such, total site density within a single face, bounded by adjacent faces, could also be a key parameter in determining mineral surface reactivity. This possibility should thus warrant further investigations.

4. CONCLUSIONS Hematite nanoparticles stabilize a thin water film equivalent to up to 4−5 monolayers when exposed to water vapor in the pw 0−19 Torr range at 298 K. Water binding proceeds through an “adsorption regime” and a “condensation regime” concurrently operating over a broad range of water vapor pressures. The resulting thin water film exhibits liquid water-like vibration properties with, however, subsets of weaker and stronger hydrogen bonds. Water binding also results in important shifts in the O−H stretching frequencies of surface hydroxo groups of the hematite surface. These shifts provided experimental proof for the involvement of −OH, μ−OH, and μ3−OH groups of the particles under study. MD simulations of idealized slabs of these faces showed that −OH groups are the predominant sites of interactions, yet μ−OH and μ3−OH groups contribute in the stabilization and the crystallographically imparted properI

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX

Langmuir

■

ties of water films. Simulations of a single nanometric particle with the (001), (100) and the isostructural (110) and (120) faces also showed that sites at edges play key roles in interconnecting water films grown from contiguous faces. These simulations suggest that edges can effectively impact the level of water binding with adjacent faces, and thus possibly control mechanisms of water film growth at particle surfaces. As such, this study points to potential limitations in the study of isolated crystallographic faces for predicting the interfacial reactivity of multifaceted surfaces. Our empirical scaling of pw to 1.8 SPC-H2O/nm2 per torr H2O in Figure 5c could, for example, be ultimately subject to revision by a more explicit account of the contributions of edges in future studies. Based on these findings, we expect that there should much to learn in following water film growth dynamics at multifaceted and even roughened surfaces in the future. The ability of advanced molecular modeling tools58−60 in accounting for water dissociation should, moreover, play an increasing role in this area. In this respect, calculation of water adsorption free energies, as well as dynamics of bond formation/rupture, at different surfaces and particle edges exposed to different water loadings, will be particularly instrumental in constraining thermodynamic models from a molecular viewpoint. These will be of particular help for addressing charge development and interfacial proton transfer processes within the confines of thin water films, such as in the presence of interfacially confined counterions. Efforts along such lines should thus help deepen our understanding of anisotropic processes occurring at multifaceted hematite particle surfaces. They should, moreover help extend these concepts to other phases central to environmental, geochemical, atmospheric, and technological processes.

■

REFERENCES

(1) Argyris, D.; Tummala, N. R.; Striolo, A.; Cole, D. R. Molecular Structure and Dynamics in Thin Water Films at the Silica and Graphite Surfaces. J. Phys. Chem. C 2008, 112 (35), 13587−13599. (2) Ewing, G. E. Ambient Thin Film Water on Insulator Surfaces. Chem. Rev. 2006, 106 (4), 1511−1526. (3) Kimmel, G. A.; Baer, M.; Petrik, N. G.; VandeVondele, J.; Rousseau, R.; Mundy, C. J. Polarization- and Azimuth-Resolved Infrared Spectroscopy of Water on TiO2(110): Anisotropy and the Hydrogen-Bonding Network. J. Phys. Chem. Lett. 2012, 3 (6), 778− 784. (4) Petrik, N. G.; Kimmel, G. A. Hydrogen bonding, H-D exchange, and molecular mobility in thin water films on TiO2(110). Phys. Rev. Lett. 2007, 99 (19), 196103. (5) Sivula, K.; Le Formal, F.; Gratzel, M. Solar Water Splitting: Progress Using Hematite (α-Fe2O3) Photoelectrodes. ChemSusChem 2011, 4 (4), 432−449. (6) Li, Z. S.; Luo, W. J.; Zhang, M. L.; Feng, J. Y.; Zou, Z. G. Photoelectrochemical cells for solar hydrogen production: current state of promising photoelectrodes, methods to improve their properties, and outlook. Energy Environ. Sci. 2013, 6 (2), 347−370. (7) Riha, S. C.; Klahr, B. M.; Tyo, E. C.; Seifert, S.; Vajda, S.; Pellin, M. J.; Hamann, T. W.; Martinson, A. B. F. Atomic Layer Deposition of a Submonolayer Catalyst for the Enhanced Photoelectrochemical Performance of Water Oxidation with Hematite. ACS Nano 2013, 7 (3), 2396−2405. (8) Klug, J. A.; Becker, N. G.; Riha, S. C.; Martinson, A. B. F.; Elam, J. W.; Pellin, M. J.; Proslier, T. Low temperature atomic layer deposition of highly photoactive hematite using iron(III) chloride and water. J. Mater. Chem. A 2013, 1 (38), 11607−11613. (9) Yanina, S.; Rosso, K. Linked reactivity at mineral-water interfaces through bulk crystal conduction. Science 2008, 320, 218−222. (10) Wang, T. H.; Huang, M. C.; Hsieh, Y. K.; Chang, W. S.; Lin, J. C.; Lee, C. H.; Wang, C. F. Influence of sodium halides (NaF, NaCl, NaBr, NaI) on the photocatalytic performance of hydrothermally synthesized hematite photoanodes. ACS Appl. Mater. Interfaces 2013, 5 (16), 7937−7949. (11) Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Bisquert, J.; Hamann, T. W. Photoelectrochemical and Impedance Spectroscopic Investigation of Water Oxidation with ″Co-Pi″-Coated Hematite Electrodes. J. Am. Chem. Soc. 2012, 134 (40), 16693−16700. (12) Yamamoto, S.; Kendelewicz, T.; Newberg, J. T.; Ketteler, G.; Starr, D. E.; Mysak, E. R.; Andersson, K. J.; Ogasawara, H.; Bluhm, H.; Salmeron, M.; Brown, G. E.; Nilsson, A. Water Adsorption on αFe2O3(0001) at near Ambient Conditions. J. Phys. Chem. C 2010, 114 (5), 2256−2266. (13) Catalano, J. G. Weak interfacial water ordering on isostructural hematite and corundum (001) surfaces. Geochim. Cosmochim. Acta 2011, 75 (8), 2062−2071. (14) Boily, J. F.; Shchukarev, A. X-ray Photoelectron Spectroscopy of Fast-Frozen Hematite Colloids in Aqueous Solutions. 2. Tracing the relationship between surface charge and electrolyte adsorption. J. Phys. Chem. C 2010, 114 (6), 2613−2616. (15) Eggleston, C. M.; Stack, A. G.; Rosso, K. M.; Bice, A. M. Adatom Fe(III) on the hematite surface: Observation of a key reactive surface species. Geochem. Trans. 2004, 5 (2), 33−40. (16) Bachhav, M.; Danoix, F.; Hannoyer, B.; Bassat, J. M.; Danoix, R. Investigation of O-18 enriched hematite (α-Fe2O3) by laser assisted atom probe tomography. Int. J. Mass Spectrom. 2013, 335, 57−60. (17) Monsegue, N.; Jin, X.; Echigo, T.; Wang, G.; Murayama, M. Three-Dimensional Characterization of Iron Oxide (α-Fe2O3) Nanoparticles: Application of a Compressed Sensing Inspired Reconstruction Algorithm to Electron Tomography. Microsc. Microanal. 2012, 18 (6), 1362−1367. (18) Tanwar, K. S.; Petitto, S. C.; Ghose, S. K.; Eng, P. J.; Trainor, T. P. Fe(II) adsorption on hematite (0001). Geochim. Cosmochim. Acta 2009, 73 (15), 4346−4365. (19) Shimizu, K.; Shchukarev, A.; Kozin, P. A.; Boily, J. F. X-ray photoelectron spectroscopy of fast-frozen hematite colloids in aqueous

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.langmuir.5b03167. Hematite particle characterization by XRD and SAED; summary of band assignment for surface hydroxo groups; FTIR spectra of hematite exposed to water vapor; MDgenerated power spectra of hematite hydroxo functional groups exposed to water; MD-derived radial distribution function for hydroxo−water interactions at crystallographic faces of single hematite particle (PDF)

■

Article

AUTHOR INFORMATION

Corresponding Author

*Tel.: +46 90 786 5270; e-mail: [email protected]. se. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the Swedish research council to J.F.B. (VR 2012-2976), and the Carl-Tryggers and the Kempe Foundations. A portion of this research was conducted using the resources of the High Performance Computing Centre North (HPC2N). G.S.A. and Y.T. thank the Knut and Alice Wallenberg Foundation (KAW) for the financial support from the project 3DEM-NATUR and for providing the electron microscopy facilities at SU. J

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir solutions. 4. Coexistence of alkali metal (Na+, K+, Rb+, Cs+) and chloride ions. Surf. Sci. 2012, 606 (13−14), 1005−1009. (20) Shimizu, K.; Boily, J. F. Electrochemical Properties and Relaxation Times of the Hematite/Water Interface. Langmuir 2014, 30 (31), 9591−9598. (21) Chatman, S.; Zarzycki, P.; Preocanin, T.; Rosso, K. M. Effect of surface site interactions on potentiometric titration of hematite (αFe2O3) crystal faces. J. Colloid Interface Sci. 2013, 391, 125−134. (22) Guo, H. B.; Xu, H. F.; Barnard, A. S. Can hematite nanoparticles be an environmental indicator? Energy Environ. Sci. 2013, 6 (2), 561− 569. (23) Kerisit, S. Water structure at hematite−water interfaces. Geochim. Cosmochim. Acta 2011, 75 (8), 2043−2061. (24) Nanayakkara, C. E.; Jayaweera, P. M.; Rubasinghege, G.; Baltrusaitis, J.; Grassian, V. H. Surface Photochemistry of Adsorbed Nitrate: The Role of Adsorbed Water in the Formation of Reduced Nitrogen Species on α-Fe2O3 Particle Surfaces. J. Phys. Chem. A 2014, 118 (1), 158−166. (25) Nguyen, M. T.; Seriani, N.; Gebauer, R. Water adsorption and dissociation on α-Fe2O3(0001): PBE+U calculations. J. Chem. Phys. 2013, 138 (19), 194709. (26) Zografi, G.; Kontny, M. J. Sorption of Water by Solids; Marcel Dekker: New York, 1995. (27) Cygan, R. T.; Liang, J. J.; Kalinichev, A. G. Molecular models of hydroxide, oxyhydroxide, and clay phases and the development of a general force field. J. Phys. Chem. B 2004, 108 (4), 1255−1266. (28) Lutterotti, L.; Bortolotti, M.; Ischia, G.; Lonardelli, I.; Wenk, H. R. Rietveld texture analysis from diffraction images. Z. Kristallogr. 2007, 2007, 125−130. (29) Popa, N. C. The (hkl) dependence of diffraction-line broadening caused by strain and size for all Laue groups in Rietveld refinement. J. Appl. Crystallogr. 1998, 31, 176−180. (30) Schneider, C. A.; Rasband, W. S.; Eliceiri, K. W. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods 2012, 9 (7), 671− 675. (31) Stanjek, H.; Schwertmann, U. The influence of aluminum on iron-oxides. 16. Hydroxyl and aluminum substitutino in synthetic hematites. Clays Clay Miner. 1992, 40 (3), 347−354. (32) Sauerbrey, G. Verwendung von schwingquarzen zur wagung dunner schichten und zur mikrowagnug. Eur. Phys. J. A 1959, 155 (2), 206−222. (33) Do, D. D.; Do, H. D. A model for water adsorption in activated carbon. Carbon 2000, 38 (5), 767−773. (34) Jaumot, J.; Gargallo, R.; de Juan, A.; Tauler, R. A graphical userfriendly interface for MCR-ALS: a new tool for multivariate curve resolution in MATLAB. Chemom. Intell. Lab. Syst. 2005, 76 (1), 101− 110. (35) Malinowski, E. R. Determination of number of factors and experimental error in a data matrix. Anal. Chem. 1977, 49 (4), 612− 617. (36) Finger, L. W.; Hazen, R. M. Crystal-structure and isothermal compression of Fe2O3, Cr2O3 and V2O3 to 50 kbars. J. Appl. Phys. 1980, 51 (10), 5362−5367. (37) Venema, P.; Hiemstra, T.; Weidler, P. G.; van Riemsdijk, W. H. Intrinsic proton affinity of reactive surface groups of metal (hydr)oxides: application to iron (hydr)oxides. J. Colloid Interface Sci. 1998, 198 (2), 282−295. (38) Berendsen, H. J. C.; Postma, J. P.; van Gunsterenand, W. F.; Hermans, J. Interaction models for water in relation to protein hydration. In Intermolecular Forces; D. Reidel Publishing Company: Dordrect, The Netherlands, 1981; pp 331−342. (39) van der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.; Berendsen, H. J. C. GROMACS: fast, flexible, and free. J. Comput. Chem. 2005, 26 (16), 1701−1718. (40) Song, X.; Boily, J. F. Water vapor interactions with FeOOH particle surfaces. Chem. Phys. Lett. 2013, 560 (0), 1−9. (41) Hockney, R. W.; Goel, S. P.; Eastwood, J. W. Quiet highresolution computer models of a plasma. J. Comput. Phys. 1974, 14 (2), 148−158.

(42) Nose, S.; Klein, M. L. Constant Pressure Molecular-Dynamics for Molecular-Systems. Mol. Phys. 1983, 50 (5), 1055−1076. (43) Parrinello, M.; Rahman, A. Polymorphic Transitions in SingleCrystals - A New Molecular-Dynamics Method. J. Appl. Phys. 1981, 52 (12), 7182−7190. (44) Hess, B.; Bekker, H.; Berendsen, H. J. C.; Fraaije, J. LINCS: a linear constraint solver for molecular simulations. J. Comput. Chem. 1997, 18 (12), 1463−1472. (45) Miyamoto, S.; Kollman, P. A. SETTLE - an analyical version of the shake and rattle algorithm for rigid water models. J. Comput. Chem. 1992, 13 (8), 952−962. (46) Darden, T.; York, D.; Pedersen, L. Particle mesh Ewald - an N.Log(N) method for Ewald sums in large systems. J. Chem. Phys. 1993, 98 (12), 10089−10092. (47) Brown, R. G.; Hwang, P. Y. C. Introduction to Random Signals and Applied Kalman Filtering; John Wiley & Sons: New York, 1997. (48) Sing, K. S. W.; Everett, D. H.; Haul, R. A. W.; Moscou, L.; Pierotti, R. A.; Rouquerol, J.; Siemieniewska, T. Reporting physisorption data for gas solid systems with special reference to the determination of surface-area and porotisy (Recommendations, 1984). Pure Appl. Chem. 1985, 57 (4), 603−619. (49) Rustad, J. R.; Boily, J. F. Density functional calculation of the infrared spectrum of surface hydroxyl groups on goethite (α-FeOOH). Am. Mineral. 2010, 95 (2−3), 414−417. (50) Ohno, K.; Okimura, M.; Akai, N.; Katsumoto, Y. The effects of cooperative hydrogen bonding on the OH stretching-band shift for water clusters studied by matrix-isolation infrared spectroscopy and density functional theory. Phys. Chem. Chem. Phys. 2005, 7 (16), 3005−3014. (51) Shchukarev, A.; Boily, J. F.; Felmy, A. R. XPS of fast-frozen hematite colloids in NaCl aqueous solutions: I. evidence for the formation of multiple layers of hydrated sodium and chloride ions induced by the (001) basal plane. J. Phys. Chem. C 2007, 111 (49), 18307−18316. (52) Song, X.; Boily, J. F. Surface Hydroxyl Identity and Reactivity in Akaganeite. J. Phys. Chem. C 2011, 115 (34), 17036−17045. (53) Song, X.; Boily, J. F. Structural controls on OH site availability and reactivity at iron oxyhydroxide particle surfaces. Phys. Chem. Chem. Phys. 2012, 14 (8), 2579−2586. (54) Song, X. W.; Boily, J. F. Water vapor diffusion into a nanostructured iron oxyhydroxide. Inorg. Chem. 2013, 52 (12), 7107− 7113. (55) Song, X.; Boily, J. F. Water vapor adsorption on goethite. Environ. Sci. Technol. 2013, 47 (13), 7171−7. (56) Song, X.; Boily, J. F. Carbon Dioxide Binding at Dry FeOOH Mineral Surfaces: Evidence for Structure-Controlled Speciation. Environ. Sci. Technol. 2013, 47 (16), 9241−9248. (57) Rustad, J. R.; Felmy, A. R. The influence of edge sites on the development of surface charge on goethite nanoparticles: A molecular dynamics investigations. Geochim. Cosmochim. Acta 2005, 69 (6), 1405−1411. (58) Motta, A.; Gaigeot, M. P.; Costa, D. Ab initio molecular dynamics study of the AlOOH boehmite/water Interface: Role of steps in interfacial Grotthus proton transfers. J. Phys. Chem. C 2012, 116 (23), 12514−12524. (59) Tocci, G.; Michaelides, A. Solvent-induced proton hopping at a water-oxide interface. J. Phys. Chem. Lett. 2014, 5 (3), 474−480. (60) Chen, J.; Li, Y. F.; Sit, P.; Selloni, A. Chemical dynamics of the first proton-coupled electron transfer of water oxidation on TiO2 Anatase. J. Am. Chem. Soc. 2013, 135 (50), 18774−18777.

K

DOI: 10.1021/acs.langmuir.5b03167 Langmuir XXXX, XXX, XXX−XXX