Direct detection of OH populations
We resolved OH populations of synthetic Fh nanoparticles as Fe-bound surface OH and H2O groups (3620–3690 cm−1; bending region in Supplementary Fig. 3) locked in very specific chemical environments, and as core OH groups (3395 cm−1) involved in a broader network of hydrogen bonds (Fig. 2a, b). X-ray photoelectron spectroscopy of strongly dehydrated Fh nanoparticles in ultra-high vacuum (Fig. 2c) revealed nearly equivalent concentrations of OH and O groups, and an overall composition of FeO0.71(OH)0.76 · 0.05 H2O (Fe5O7.6H4.3; Fig. S2, Table S1). We will later show that this composition is consistent with the concept of a hydroxyl-rich surface region covering a hydroxyl-poor Fh core35,36,37.
We identified surface OH groups by comparing their spectral response with our previous assignments for crystalline FeOOH surfaces (Fig. 2d, Supplementary Fig. 1). This comparison shows that Fh generates flagship bands of poorly hydrogen bonded −OH groups (3667 cm−1), chemisorbed water molecules (−OH2; ~3690 cm−1) and doubly-coordinated hydroxyl groups (μ−OH; 3620 cm−1). Triply-coordinated μ3−OH sites were not detected in the expected spectral range where they occur in other crystalline iron (oxyhydr)oxide43,45,48 minerals (cf. Fig. 2a in the 3491–3578 cm−1 region, and Supplementary Fig. 1 for reference spectra). While strong overlap with core OH bands or hydrogen bonding may have hindered spectral resolution, another is possibility, also raised by Hiemstra35,36,37, is that μ3−OH sites are of lower density on Fh surfaces. We will return to this in our theoretical analyses of OH populations.
Our previous work showed that −OH groups responsible for the 3667 cm−1 band are disposed along rows on dominant faces of FeOOH minerals (Supplementary Fig. 1)43,48. Such rows are also exposed at the edges of sheets of Fe1 octahedra in our idealised representation of Fh (Fig. 1), and co-exist with chemisorbed water (−OH2) or bare Fe Lewis acid sites in acid-neutral and charge-neutral surfaces. We note that rows with greater populations of inter-site hydrogen bonds, such as in the mineral goethite (α-FeOOH)43, generate a band downshifted by 6 cm−1, and are not present in Fh. This difference is highly significant as these bands are phenomenally responsive to the slightest changes in hydrogen bond strength, and to the tune of ~150 cm−1 per pm change in O–H bond length44. This 3667 cm−1 band is therefore key evidence for a dominance of Fh −OH groups of relatively lower hydrogen bonded environments than on goethite.
To illustrate the sensitivity of −OH groups to their bonding environments, we altered their speciation by exposing Fh to HCl (Fig. 2e). The protonation reactions that followed triggered the same spectroscopic response that we have resolved in our reference crystalline FeOOH minerals43,48, and can be appreciated by the loss of our 3667 cm−1 band and its shift to lower frequencies. These changes signal that new hydrogen bonds were formed between unreacted −OH groups and −OH2 sites along the rows we have previously mentioned. It is these chemisorbed water molecules that contribute to our second flagship band at ~3690 cm−1. The high O–H stretching frequency of these water molecules is from an entirely isolated O–H bond from the hydrogen bond network of the Fh surface, and we note that we have previously detected such species on iron (oxyhydr)oxide43,45,48 nanominerals. In this previous work, we showed that while a fraction is always resilient to outgassing under dry gases, water can be removed by prolonged exposure to ultra high vacuum or heat. We will come back to this point when looking at the thermal stability of Fh.
The band for μ−OH was only of small intensity, yet its appearance is of high value for understanding the surface chemistry of Fh, as it is a well-known signal for groups on the basal faces of iron (oxyhydr)oxides (Supplementary Fig. 1)43,45. This suggests that, although generally of spheroidal morphology, Fh nanoparticles could have developed a minor basal face exposing neutrally charged and proton inactive μ−OH groups43,45. Our previous work on this face shows that these groups are embedded in a highly regular hydrogen bonding environment responsible for the narrow 3620 cm−1 band. The other μ−OH groups of the dominant spheroidal surface are, in contrast, not discernable as they are involved in a broad network of hydrogen bonds. We have arrived to the same conclusion through our work on crystalline iron (oxyhydr)oxides43,45, and will support this further in molecular simulations in the last section of this study.
We can validate these findings further by inspecting the thermal response of these spectral signatures at temperatures well below the conversion temperature (e.g. 380 °C as in Xu et al.12) of Fh to haematite (Figs. 3 and S3)2,12. Exposing Fh nanoparticles to a heating gradient in vacuo progressively removed all water, as well as surface and core OH groups (Fig. 3a, b). This loss correlates first with the continual evacuation of water below ~100 °C leading to dehydration events at ~134 °C and at ~204 °C (Fig. 3d). These two events removed the great majority bound water (see Supplementary Fig. 3 for the bending mode of water), chemisorbed water (3690 cm−1) and −OH groups (3667 cm−1). These events produced a high temperature Fh core exposing the most heat-resistant core OH and surface μ-OH groups. Using a chemometric analysis49 we show that the concentration profile (components III of Fig. 3c) of the Fh core directly aligns with the dehydration events at ~134 °C and ~204 °C, both marking transition temperatures between dominant Fh hydration states. The spectral signature of this core is also dominated by surface μ−OH (3643 cm−1) and core OH (3395 cm−1) groups.
Fh nanoparticles thus chiefly exposed isolated surface −OH and hydrogen-bonded μ−OH groups of comparable bond strength in our reference crystalline iron (oxyhydr)oxide minerals (Supplementary Fig. 1)43,48. The μ−OH groups become the sole surface species at high temperatures after removal of chemisorbed water and −OH sites. The spectra offer, however, no clear evidence for the presence of μ3−OH groups at the surface. Finally, we note that no evidence could be found for the exposure of OH groups bound to a Fe3 (tetrahedral Fe) site. Drawing from the catalytic alumina literature39,40, where OH groups bound to tetrahedrally-coordinated Al3+ are well documented, we would have expected a band overlapping with that of the chemisorbed water. However, as this band shifted congruently with gradients in proton loadings (Fig. 2e) and temperature (Fig. 3b), we have no evidence for such a site on Fh surfaces.
By resolving the complex vibrational spectroscopic signals of Fh, we here offer new possibilities for tracking the catalytic reactivity of Fh from the unique viewpoint of its OH groups. We can, of instance, begin resolving the binding of critically important gases (e.g. CO2, CH4, NOx, SOx) on Fh. As these spectroscopic signals also persist under atmospheric humidity47,50, we can explore interfacial reactions and phase transformation in nanometrically thick water films formed at Fh surfaces. In the following sections, we bridge these findings with new theoretical analyses and molecular simulations underscoring the importance of −OH and μ−OH groups on the surface chemistry of Fh nanoparticles.
Populations and spatial dispositions
We support our experimental findings by identifying OH groups on simulated Fh nanoparticles of spheroidal morphology. Spheroidal nanoparticles, with no or little defined crystallographic faces, are not only more representative of real Fh but are also expected to expose greater proportions oxygens of lower Fe coordination number than on crystallographically flat surfaces. To explore this idea, nanoparticles were generated from a spherical cut of the crystalline structure of Michel et al.6, and were modified for various surface depletion depths of Fe2 and Fe3 sites (0–6 Å), atomic displacements (0–4 Å) and Fe site vacancies (vFe). All broken Fe–O bonds were healed and protonation levels were set to −OH and μ-OH, while those of μ3−O(H) sites were set to levels predicted from the core structure. Finally, all nanoparticles were made charge-neutral by protonation of a required number of randomly selected −OH sites to −OH2. Inspection of 280 computer-generated nanoparticles reveals a dominance of −OH/−OH2 and μ−OH groups but only low densities of μ3−O(H) sites at Fh surfaces (Fig. 4; Supplementary Figs. 5–7 for full set of results). These low μ3−O(H) densities align with our experimental results, and again suggest that the spherical Fh surface should be dominated by −OH/−OH2 and μ−OH groups.
The more relevant nanoparticles to be considered in our analysis are those whose mass densities (Fig. 4b; corresponding molar mass in Supplementary Fig. 6) are closer to experimental values (e.g. ~3.4 g/cm−3 for ~2−3 nm particles11). In accordance with the work of Hiemstra35,36,37 on crystalline Fh nanoparticles, we find that depleting surface Fe2 and Fe3 sites is a highly effective means for reproducing experimental mass densities. A depletion depth of no more than 4 Å from the average positions of the top O atoms is needed to obtain such values. We can also justify introducing up to ~20% additional vacancies Fe2 and Fe3 vacancies (vFe2,Fe3) throughout the Fh core (Figs. 4b and S7). We can, however, introduce only small levels of vacancies at Fe1 sites, given X-ray total scattering work51 pointing to nearly fully occupied octahedral sheets.
Another experimental constraint to consider prior discussing OH populations, is the OH/O ratio of ~1.1 obtained by X-ray photoelectron spectroscopy of our strongly dehydrated Fh nanoparticles in ultra-high vacuum. The OH/O ratios retrieved from our theoretical considerations (Fig. 4c) were strongly size-dependent as they reflect contrasting proportions of the OH-rich surface and OH-poor core. In this context, only the ratio at ~3 nm aligns with the ~1.1 ratio (Fig. 2c). We can also obtain this ratio for larger particles sizes by consideration of the uppermost region of the particles and by contributions of non-stoichiometric core OH groups generated by vacancies.
Our search within particles with a range of reasonable mass densities shows that surface depletion of Fe2 and Fe3 sites is the dominant factor altering −OH densities on Fh (Fig. 4d). We find that a surface depletion down to 4 Å from the surface more than doubled −OH sites densities, yet depletions at greater depths had no additional impact. Our simulations also show that atomic displacements, although important in the elucidation of core structure14, have no impact on −OH populations as long as Fe–O bonds are not displaced any longer than 2 Å (Figs. 4d and S7). Creating −OH in O sites of the Fh core should therefore require larger levels of disorder seen in poorly crystalline materials but not likely those of Fh nanoparticles displaying relatively more atomic order. Along the same vein, introducing vacancies at core Fe2 and Fe3 sites had no impact on −OH populations (Supplementary Fig. 7), although we note that concentrations can be effectively increased throughout the core with vacancies at Fe1 sites (Fig. 4e). Our simulations thus show that −OH on spheroidal Fh surfaces are predominantly disposed as rows at edges of Fe1 sheets as corner (~2.7 –OH/nm2) and edge (~2.0 –OH/nm2) sites (Figs. 1 and 4f). We note that these populations fall in the lower range of densities previously resolved on crystalline faces by Hiemstra35,36,37,38, still they readily account for metal ion and ligand adsorption densities reported in the literature32,33,52.
Populations of μ−OH groups were strongly size-dependent in particles <~4 nm. They reached densities that were about ~3 times less than those of −OH in larger particles (Fig. 4g). While these sites are predominantly disposed as rows at edges of Fe1 sheets (Fig. 1), they are also exposed on basal faces (Fig. 4a), which are most clearly expressed in nanoparticles of up to ~6 nm in diameter. This result thus ties further the faint 3620 cm−1 band (Fig. 2d) to μ-OH sites on the basal face of Fh. We also find that, unlike the case for −OH, vacancies in Fe2 and Fe3 sites increase μ−OH populations throughout the Fh core. As these populations become even larger with vacancies at Fe1 sites, we expect a greater variability of μ−OH populations throughout the core of poorly crystalline Fh. However, given their low ligand exchange capability53,54 and proton affinity43, we cannot expect enhanced reactivities from these sites alone.
To understand how the distribution of surface OH groups impact inter-site interactions we turned to molecular dynamics simulations of single Fh nanoparticles with a surface depletion of Fe2 and Fe3 sites down to the first 4 Å. For these simulations, we removed all chemisorbed water molecules (−OH2) from the surface, and retained full core Fe occupancies. This configuration was chosen to focus the discussion on the hydrogen bonding environment of OH groups of a hydroxylated Fh surface over a defect-free core. It thus emulates strongly dehydrated Fh surfaces, which were likely achieved in N2(g) dried samples studied experimentally.
Our simulations retained the core crystallographic structure of Fh while mostly relaxing its hydroxylated surface (Fig. 5a–c). The surface sites in this family of simulation cells were spread over a ~0.6-nm region of the top portion of the particle, as imposed by the surface depletion scheme (Fig. 5a). This is the region whose structure is mostly affected by the relaxation of the hydroxylated shell. This relaxation can be especially appreciated by the radial distribution functions (Fig. 5c) of the 2-nm-wide nanoparticles where contributions from the surface shell are greatest (cf. Supplementary Figs. 8 and 9 for surface-specific radial distribution functions). The atomic profiles of this region (Fig. 5a) also reveal the relative depletion of μ3-OH groups in the near surface region where the coordinatively unsaturated O sites are predominantly −OH and μ−OH. This also marks the zone in which hydrogen populations transition between a dominance of core μ3−OH···O interactions (60–70% bonds per core OH) to those involving surface OH groups.
We find that Fh surfaces acquire an intricate network of hydrogen bonds that is predominantly built from donating μ−OH sites. These μ−OH sites donate hydrogen bonds up to ~1/2 of the available −OH (μ−OH···−OH) (Fig. 5d) and up to ~1/5 of μ−OH groups (μ−OH···μ−OH). The hydrogen bonding network also involves lateral μ−OH···OH−μ bonds taking place over both the dominant spherical surfaces and over the basal face. These interactions take proportionally more importance in the 2-nm particle, given the greater representation of the basal face (Fig. 4a). We also find that the higher curvature of the 2-nm nanoparticles strongly disfavours μ−OH···−OH hydrogen bonds (Fig. 5d) over the larger nanoparticles considered in this work. Because hydrogen bonding populations can strongly affect the proton affinity of −O sites50, we anticipate that these curvature-induced populations could play important roles in the acid–base chemistry of the smaller-size Fh nanoparticles. As such, consideration of these populations in future predictions of protonation constants, for example through extensions of the surface depletion model35,36,37,38 or of atomistic simulations55, can represent an important step to take in accounting for the size-dependent chemistry of Fh. Developments in this area will be highly beneficial in our pursuit for understanding the Fh surface reactivity, now that this study has established the identity and the likely dispositions of OH populations exposed at Fh nanoparticle surfaces.
By resolving the complex vibrational spectroscopic signals of Fh, this work offers new possibilities for tracking the catalytic reactivity of this nanomineral from the unique viewpoint of its OH groups. It immediately opens new possibilities for resolving the mechanisms through which Fh alters the fate of environmentally critical gases (e.g. CO2, CH4, NOx, SOx), as well as photo(electro)chemical transformation of organics in nature and technology. As these spectroscopic signals also persist under atmospheric humidity47, this work also open new possibilities for exploring interfacial reactions and phase transformation in nanometrically thick water films at Fh surfaces. It may, additionally, be pertinent to investigate the possible roles that neutrally-charged basal faces play in the oriented aggregation of Fh in water. Future consideration of such possibilities, and especially based on our findings that hydrogen-bonded −OH and μ−OH groups dominate the surface speciation of Fh, represents some of the many anticipated implications that this work can have for understanding the surface chemistry of Fh nanoparticles in nature.