Direct identification of reaction sites on ferrihydrite

Hydroxyl groups are the cornerstone species driving catalytic reactions on mineral nanoparticles of Earth’s crust, water, and atmosphere. Here we directly identify populations of these groups on ferrihydrite, a key yet misunderstood iron oxyhydroxide nanomineral in natural sciences. This is achieved by resolving an enigmatic set of vibrational spectroscopic signatures of reactive hydroxo groups and chemisorbed water molecules embedded in specific chemical environments. We assist these findings by exploring a vast array of configurations of computer-generated nanoparticles. We find that these groups are mainly disposed along rows at edges of sheets of iron octahedra. Molecular dynamics of nanoparticles as large as 10 nm show that the most reactive surface hydroxo groups are predominantly free, yet are hydrogen bond acceptors in an intricate network formed with less reactive groups. The resolved vibrational spectroscopic signatures open new possibilities for tracking catalytic reactions on ferrihydrite, directly from the unique viewpoint of its reactive hydroxyl groups.


Material Characterization
Suspensions of 14.2 g/L Fh were stored in N2(g)-filled polyethylene container at 4 ºC. A portion of the resulting washed solids was dried under a stream N2(g) at room temperature and used for phase and chemical characterisation (Fig. S1). X-ray diffraction (Bruker d8 Advance working in q-q mode with Cu Ka radiation) on the dry solids confirmed that 6-line Fh was the sole crystallographic phase in the precipitates. Vibration spectra showed no evidence for the conversion of Fh to (low crystallinity) FeOOH, as can be confirmed by the lack Fe-O-H bending modes of goethite and lepidocrocite. 1 Samples imaged by transmission electron microscopy imaging (JE-1230, JEOL) revealed particle aggregates of the order of ~20-30 nm. A 90-point adsorption/desorption isotherm (TriStar, Micromeritics) on samples previously dried in situ at 100°C for 16 h under a stream of N2(g) revealed a Brunauer-Emmet-Teller 2 specific surface area of 209 m 2 /g and 5.5% in microporosity. These measurements suggest individual particles as small as ~7-8 nm in diameter.
We used X-ray photoelectron spectroscopy to analyse the near-surface atomic composition of Fh exposed to a vacuum of less than 10 -7 Pa. The Kratos Axis Ultra electron spectrometer used for these measurements was equipped with a delay line detector, a monochromated Al Ka source operated at 150 W, a charge neutralizer, and a hybrid lens system with a magnetic lens providing an analysis area of 0.3 mm ´ 0.7 mm. We collected survey spectra from 1100 to 0 eV at a pass energy of 160 eV, while the high-resolution spectra (pass energy 20 eV) for Fe 2p, O 1s, C 1s and Cl 2p were collected at a rate of 0.1 eV/step. Using the spectra processing software of Kratos, we applied a Shirley background to all high-resolution spectra, and adjusted the binding energy scale to the 285.0 eV C 1s line of aliphatic carbon. We modeled these resulting background-subtracted spectra with a 70% Gauss/30% Lorentz function (Table S1).

Vibration Spectroscopy
Aqueous suspensions of Fh (14.2 g/L; 2968 m 2 /L) were prepared at total concentrations of 31 NaOH/nm 2 to 81 HCl/nm 2 , keeping the pH in the 2-10 range. These loadings were achieved by additional of standardised NaOH or HCl to the aqueous suspensions. All suspensions were equilibrated in sealed polyethylene test tubes for 24 h under N2(g), then exposed to a stream of N2 (g) for another 0.5 h prior centrifugation. The pastes were then applied onto an Attenuated Total Reflectance accessory (Golden Gate, single-bounce diamond cell), using a Bruker Vertex 70/V Fourier Transform Infrared spectrometer equipped with a DTGS detector. 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 co-added spectra for each sample. Blackman-Harris 3-term apodisation function was used to correct phase resolution.
Temperature programmed desorption experiments were performed in an optical chamber (AABSPEC #2000-A) equipped with CaF2 windows, and holding a vacuum of less than 0.3 Pa. Spectra were collected in the 1100-4500 cm -1 range at a resolution of 4 cm -1 . Each spectrum was the average of 50 scans.

Chemometrics
We used the multivariate curve resolution method 3 to facilitate interpretation of the changes in the vibration spectra in the temperature programmed desorption experiments. The method extracts spectral components (ε), and their respective concentration profiles (C) from a 2D matrix of spectral absorbances (A m×n of m wavenumbers and n temperatures) through the Beer-Lambert 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, we first offset all spectra in the O-H stretching region to zero absorbance at 4000 cm -1 , where absorbances from the sample are negligible. We then determined the number of chemically significant species responsible for the variance of these spectra using the Malinowski's factor indicator function 4 and starting from abstract eigenvectors and eigenvalues obtained from a singular value decomposition. 5 We then used the MCR-ALS 3 program to extract ε and C using our chosen number of species. The program rotates those number of eigenvectors of the A matrix into a real chemical space, such that such A m×n = ε m×s .·C s×n + E, where ε m×s ≥ 0 and C s×n ≥ 0, and E is the matrix of unaccounted residuals. These calculations were made in the computational environment of MATLAB 9.1. 6

Molecular Dynamics
Molecular dynamics simulations of the Fh nanoparticles were performed with program GROMACS/2019.1. 7 Simulations were carried out using the Clayff force field 8 and the revised parameter for iron. 9 Following the practice of modelling tetrahedrally coordinated metal ions in clays, we treated the tetrahedral Fe3 site with the same interaction potentials values as the octahedral iron but constrained its O-Fe3-O angle of 109.5°. A NVT (constant number of particles, constant volume and constant temperature) ensemble and a time step of 1.0 fs were used with the Verlet algorithm 10 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-Hoover 11 velocity-rescale thermostat with a 0.1 ps relaxation time. The O-H bond strength of all the hydroxyls were treated by the LINCS 12 algorithm. A 0.8 nm cutoff was used for non-bonded van der Waals interactions and the particle-mesh Ewald 13 method was used to treat long-range electrostatic interactions.
Simulations cells were first energy-minimised (double precision) using a steepest descent algorithm. The resulting structure was then equilibrated (single precision) using classical MD for at least 10 7 steps (10 ns), followed by production runs of at least another 10 ns. Total energy convergence and its components as well as temperature, and atomic densities were monitored for these entire equilibration periods. This modelling strategy retained the Fh bulk structure, yet relaxed the surface. Radial distribution functions for atomic pairs and hydrogen bond analyses were carried out using the utilities of GROMACS/2019.1. 7 Radial distribution functions with respect to the particle center of mass were however generated using a MATALB 9.1 6 code written for this work. GT=goethite (a-FeOOH); L=lepidocrocite (g-FeOOH); HEM=hematite (a-Fe2O3)). All data were extracted from previous studies from our group. 14-16 A B electron microscopy spheroidal Fh particles and/or aggregates of the order of ~20-30 nm (Fig. 1A). Individual particles could however be as small as ~7-8 nm in diameter given their B.E.T. specific surface area of 209 m 2 /g. (B) XRD pattern of the 6-line ferrihydrite sample, and compared against PDF #00-058-0900 (Fe9.5O14(OH)2). (C-D) X-ray photoelectron spectroscopy of the (C) Fe 2p and (D) the O1s regions, including fits. The O1S region reveals an OH/O ratio of 1.1. XPS spectra were shifted to the C1s peak at 285.0 eV.  a. Sensitivity factor; b. n.b. inorganic and organic carbon contaminants originate from the atmosphere, and acquired during adventitious exposure to ambient air. c. Trace chloride contaminants are remnants from the synthesis procedure.