Surface chemical heterogeneity modulates silica surface hydration

Significance

Silica, conventionally known as “glass,” is a universally used material in catalysis, nanofabrication, and many other applications, but details of its surface chemistry and interactions with water are notoriously complicated and unclear––partially due to its tunable surface chemistry. We utilize this tunable surface chemistry of silica to reveal properties of surface-bound water that impact surface reactions, adhesion, and colloidal interactions. Using a combination of surface forces, hydration dynamics, and simulation techniques, we show that surface silanol groups stabilize the surface water layer, and also that variations/fluctuations are more pronounced at intermediate silanol densities on the surface. This work provides insight into fundamental interactions of water with chemically heterogeneous surfaces.

Abstract

An in-depth knowledge of the interaction of water with amorphous silica is critical to fundamental studies of interfacial hydration water, as well as to industrial processes such as catalysis, nanofabrication, and chromatography. Silica has a tunable surface comprising hydrophilic silanol groups and moderately hydrophobic siloxane groups that can be interchanged through thermal and chemical treatments. Despite extensive studies of silica surfaces, the influence of surface hydrophilicity and chemical topology on the molecular properties of interfacial water is not well understood. In this work, we controllably altered the surface silanol density, and measured surface water diffusivity using Overhauser dynamic nuclear polarization (ODNP) and complementary silica–silica interaction forces across water using a surface forces apparatus (SFA). The results show that increased silanol density generally leads to slower water diffusivity and stronger silica–silica repulsion at short aqueous separations (less than ∼4 nm). Both techniques show sharp changes in hydration properties at intermediate silanol densities (2.0–2.9 nm⁻²). Molecular dynamics simulations of model silica–water interfaces corroborate the increase in water diffusivity with silanol density, and furthermore show that even on a smooth and crystalline surface at a fixed silanol density, adjusting the spatial distribution of silanols results in a range of surface water diffusivities spanning ∼10%. We speculate that a critical silanol cluster size or connectivity parameter could explain the sharp transition in our results, and can modulate wettability, colloidal interactions, and surface reactions, and thus is a phenomenon worth further investigation on silica and chemically heterogeneous surfaces.

Silicon dioxide, SiO₂, or glass, is a ubiquitous material particularly relevant to catalysis, geology, and semiconductor processing (1). Despite the immense literature on the properties of SiO₂ surfaces, the molecular details of the SiO₂–water interface––specifically, the exact chemical makeup of the surface as a function of pretreatment conditions, and the resulting impact on the hydrophilicity and hydration barrier of the surface––are not well understood. It is accepted that the SiO₂ surface comprises silanol (SiOH) groups that are hydrophilic, and siloxane (SiOSi) groups that are moderately hydrophobic (1, 2), but any general and unifying model of the SiO₂–water interface is lacking, likely due to the existence of SiO₂ of different morphologies (amorphous, crystalline), polymorphism (e.g., quartz, stishovite, cristobalite, coesite), porosity, surface chemistry, and degrees of contamination, which depend on synthesis conditions, preparation methods (e.g., fused, fumed, Stöber/sol–gel, nanofabrication methods), and surface cleaning procedures. Here, we present experimental measurements of the surface forces and surface water diffusivity of nonporous, amorphous silica as a function of surface hydrophilicity (or silanol coverage), accompanied by molecular dynamics simulations that offer molecular insights into the factors that tune surface water diffusivity of mixed hydrophobic/hydrophilic silica surfaces.

The fraction of silanol and siloxane groups on the silica surface is easily adjusted. Under wet or humid conditions, and at equilibrium, the silica surface fully comprises silanol groups (2). When heated above 200 °C, adjacent silanol groups begin to condense to form siloxane bridges via a process known as dehydroxylation. The equilibrium density of silanols on the surface, α_OH, has been studied extensively under dry conditions, using techniques such as thermogravimetric analysis (3), infrared spectroscopy (4–7), deuterium exchange (5, 8), ²⁹Si and ¹H NMR (3, 6, 9), and titration (10), among others (11). Generally, α_OH decreases steadily from ∼4.6 SiOH/nm² at temperatures <200 °C, to <0.5 SiOH/nm² at ∼1,000 °C (2). Although studies have often disagreed on the true maximum surface silanol density, the disagreement is largely due to the distinction between accessible (surface) or inaccessible (subsurface) silanols (4, 6, 12). Regardless of any quantitative discrepancies, the qualitative trend of decreasing silanol density with temperature is accepted. Siloxane bridges react with water to regenerate silanol groups, a process known as rehydroxylation, but this process can be very slow. Silica surfaces pretreated at 400 °C may fully rehydroxylate within an hour when exposed to liquid water (13, 14), whereas surfaces pretreated at 1,000 °C may take years (1, 2). Additionally, plasma or UV-ozone treatment can rapidly rehydroxylate the surface (15, 16).

Beyond the average silanol density α_OH, attempts have also been made to characterize the distribution and coordination of silanol groups, but a complete picture of the surface remains elusive. A combination of infrared spectroscopy (IR), deuterium exchange, and NMR studies has quantified the fraction of vicinal [Q₃, Si(OH)-O-Si(OH], geminal [Q₂, Si(OH)₂], and isolated [Q₃, SiO-Si(OH)-OSi] silanols on the surface as a function of pretreatment temperature (2, 17), or alternatively, the fraction of hydrogen-bonded (proximal) and non–hydrogen-bonded silanols (7, 12, 18). IR measurements of partially dehydroxylated silica in water vapor have shown (somewhat expectedly) that water preferentially adsorbs onto the silanol groups, rather than onto the siloxane bridges (1, 14). Because water is required for rehydroxylation, this result suggests that siloxane bridges adjacent to silanols are more readily hydroxylated than siloxanes distant to silanols. It follows that the growth of silanols is more likely to occur in patches (i.e., autocatalytically) than homogeneously. Recent measurements of the adsorption of Ga(CH₃)₃ dimers onto two adjacent silanol groups have shown the potential of quantifying such clustering/coordination experimentally by chemical tools (19). Molecular dynamics simulations of amorphous silica surfaces, while allowing one to clearly visualize the silanol speciation at the surface, have only rarely explored the thermodynamically favorable positioning of silanol groups (20, 21). Rather, most simulation efforts have considered predetermined, fixed chemistries or those constructed via heuristics from bulk, annealed silica (22, 23).

The hydrophilicity of silanol groups arises from their ability to both accept and donate hydrogen bonds (17), whereas siloxane bridges only weakly accept hydrogen bonds (24). However, only a few measurements have experimentally quantified the surface hydrophilicity as a function of α_OH. Macroscopic observations of the hydrophilicity of the SiO₂ surface are most clearly seen with water contact angle measurements, where the contact angle (averaged between advancing and receding angles) is <5° on a fully hydroxylated surface, and 40° on an almost fully dehydroxylated surface pretreated at 1,000 °C (25). Force measurements between silica surfaces in water using the surface forces apparatus (SFA) have shown that fully hydroxylated surfaces repel each other strongly below ∼4-nm separation (26–28). Vigil et al. (26) found that dehydroxylated surfaces exhibit a smaller repulsion. However, no systematic study of the silica–silica repulsive forces across water as a function of silanol density has been conducted with the SFA. Vigil et al. proposed that the added short-range repulsion between the hydroxylated surfaces was most likely a steric effect due to protrusion of polysilicic acid groups (or “hairs”), but acknowledged that the results were dependent on the detailed surface preparation method, even for a fixed silanol density. However, SFA measurements by Grabbe showed that coating the silica with an amine monolayer greatly reduced the short-range repulsion (27, 29), which the author(s) used to support the existence of a hydration repulsion, rather than a steric repulsion. Regardless of the differences in interpretation, the contact angle and SFA results alone do not offer conclusive insight into the chemical structure/makeup of the surfaces or the molecular behavior of water at the interface.

In this work, we controllably alter the surface silanol density of silica surfaces α_OH through heat treatment, and subsequently use Overhauser dynamic nuclear polarization (ODNP) to measure surface water diffusivity, the SFA to measure interaction forces between silica surfaces across water, and molecular dynamics (MD) simulations of the silica–water interface to offer plausible interpretation of the experimental results. The results show that the equilibrium surface water diffusivity correlates with conventional measures of hydrophilicity (e.g., contact angle and interaction forces) and demonstrate the significance of the distribution of chemical moieties on surfaces that comprise both hydrophilic and hydrophobic chemical groups. Even though the findings are plausible after the fact, it was not clear a priori whether differences in the heterogeneity (silanol distribution) of otherwise identical silica surfaces would manifest themselves experimentally in surface water diffusivity or surface forces averaged across several water layers. While the surface chemistry of silica is still not fully understood, we propose that its tunability and the apparent effect on surface hydration thermodynamics makes it both a critical problem for understanding chemistry and processes involving silica surfaces, and a meaningful platform to study fundamental hydration phenomena.

Results and Analysis

Measurements of Silica Surface Water Diffusivity.

To determine the effects of surface silanol density α_OH on hydration dynamics, surface water diffusivity was measured with ODNP, an NMR technique that measures the local translational water diffusivity within 1 nm of the paramagnetic spin labels in the system (30). ODNP captures equilibrium water diffusivity by single-particle dynamics by means of ¹H NMR relaxometry of water in the solution state. Details of the sample preparation and spin labeling procedure are given in SI Appendix, section S1, but briefly, we attached a stable nitroxide radical spin label [(2,2,6,6-Tetramethylpiperidin-1-yl)oxyl, TEMPO] to the surface of nonporous, fumed silica nanoparticles (Aerosil 380) at a TEMPO density of <0.03 nm⁻² attached through an aminopropyl linkage. The particles were suspended in a 15 mM NaCl solution at pH 5.0 for ODNP measurements conducted at 22 °C.

If partially dehydroxylated silica surfaces were desired, then before the functionalization chemistry described above, the particles were heated under dry nitrogen for 6 h at temperatures ranging from 600 to 1,000 °C. The silanol density was quantified on unfunctionalized particles without TEMPO. Because partial rehydroxylation was expected during the functionalization steps in the ODNP sample preparation, the unfunctionalized particles were subjected to similar aqueous conditions before the silanol density measurement. The number of surface silanol groups per mass of silica was measured by taking the average between ¹H NMR measurements using deuterium exchange (5, 8) and the change in pH upon submerging the particles in a slightly basic (pH ∼9.5) solution of 1 M NaCl (10). The techniques rely on the exchangeability of the silanol protons with deuterium (from D₂O) and Na⁺ (from the NaCl solution), and the results agreed well between these two complementary methods (SI Appendix, Table S1). The surface area per mass (specific surface area), which decreased with increased pretreatment temperature, was determined by N₂ Brunauer–Emmett–Teller (BET) measurements (31) (SI Appendix, Table S1).

ODNP measurements of surface water diffusivity D_surface as a function of surface silanol density α_OH are shown in Fig. 1. D_surface is particularly small, and thus the silica surfaces give rise to an unusually large retardation of surface water dynamics. Specifically, D_surface is (1.6–3.2) × 10⁻¹⁰ m²/s, which is ∼10× smaller than the bulk value for water (2.3 × 10⁻⁹ m²/s) (30), ∼2–4× smaller than the value on phospholipid membranes (7.2 × 10⁻¹⁰ m²/s) (32), and ∼1–5× smaller than the value on chemically heterogeneous Chemotaxis Y protein surfaces [(2–12)× 10⁻¹⁰ m²/s)] (33). As expected, D_surface generally decreases with increasing α_OH (increasing hydrophilicity), but there is a sharp decrease in D_surface going from α_OH = 2.2–2.7 nm⁻², while D_surface remains constant within error for α_OH < 2.2 and α_OH > 2.7 nm⁻². Under the same conditions as the ODNP measurements, electron paramagnetic resonance (EPR) measurements of the TEMPO-labeled silica, which give some measure of the mobility of the TEMPO moiety (32, 34), show no differences across the here-tested range of silanol densities (SI Appendix, Fig. S2), suggesting that the transition is not due to silica aggregation that occurs at a specific value of α_OH. Further, the transition does not appear to be due to a sudden change in water density near the spin labels. Electron spin echo envelope modulation (ESEEM) measurements (35) that quantify the number density of (deuterated) water in the primary hydration shell of the spin label were conducted on five of the samples, showing relatively constant water density across a range of α_OH (SI Appendix, Fig. S5). This result further suggests that the position of the TEMPO linkage (i.e., by extension or burial) does not vary substantially across the range of α_OH.

Fig. 1.

Values of water diffusivity measured at the silica surface D_surface in 15 mM NaCl, pH 5.0, and 22 °C, as measured by ODNP. Errors in D_surface are the SD of at least three measurements at each value of α_OH. The surface silanol density α_OH was quantified by deuterium exchange and pH measurements following heat treatments at 1,000, 950, 900, 800, 700, 600, or 22 °C. There is a steep transition in D_surface between α_OH = 2.2 and 2.7 nm⁻² (corresponding to pretreatment temperatures of 800 and 700 °C, respectively). The upper x axis shows the average silanol spacing if one were to assume isotropically distributed silanols on the surface. The dashed gray line represents a linear trend of D_surface with α_OH, which connects the extrema of the end points in the plot, but does not pass through all of the data points. The solid gray line shows the more likely trend.

Additionally, ODNP measurements on maximally hydroxylated silica (α_OH = 4.5 ± 0.2 nm⁻²) were conducted at pH 3.5 and 9.0 to test the effect that deprotonated, charged silanols (SiO⁻) may have on hydration dynamics, as the surface density of SiO⁻ groups varies by a factor of at least 50 over that pH range (36, 37). Within error, no difference was observed in the hydration dynamics [D_surface (in units of 10⁻¹⁰ m²/s) = 2.1 ± 0.4 at pH 3.5, 1.9 ± 0.3 at pH 5.0, and 1.8 ± 0.3 at pH 9.0]. SFA studies have also shown a minimal effect of varying surface charge density (38), as corroborated by water contact angle measurements, in which we found the contact angle to be 0° on maximally hydroxylated silica at all three pH values, suggesting that charged SiO⁻ groups do not contribute significantly to the observed variation in surface hydration effects. This behavior distinguishes silica from other notable surfaces on which hydration has been extensively studied, such as muscovite mica (39), acrylic polymers (40), and lipid membranes (41, 42), where surface hydration is essentially due to charged surface groups/ions with individual hydration shells and defined hydrated excluded volumes or is modulated by interactions of hydrated ions with macromolecules.

To examine the possibility of a swollen, gel-like surface layer that grows over the scale of hours–days, ODNP measurements were also performed on maximally hydroxylated silica after the particles were submerged in the 15 mM NaCl solution for 2 d. The results were unchanged after the extra immersion period [D_surface (in units of 10⁻¹⁰ m²/s) = 1.9 ± 0.3 initially vs. 1.8 ± 0.4 after 2 d], as was also observed in our SFA measurements on maximally hydroxylated silica surfaces (described below) and the SFA measurements of Grabbe and Horn (27) on dehydroxylated silica. The negligible contribution of silica hairs in the gel layer that would correspond to Q¹ silica is further corroborated by solid-state ¹H-²⁹Si cross polarization/magic-angle spinning dynamic nuclear polarization NMR (CP-MAS DNP-NMR) (SI Appendix, Fig. S3) that shows the Q¹ species to constitute only 1.4% of the total surface or subsurface ²⁹Si signal (SI Appendix, Fig. S4). The confidence in the assignment to Q¹ species can be derived from the systematic absence of this peak in freshly submerged silica, its appearance after 16 h of hydration, disappearance upon drying, and reappearance upon rehydration for an additional 16 h. Further details are in SI Appendix, section S2. Taken together, we can neglect the steric effect of gel-like silica species contributing to the measurement of surface hydration properties in the ODNP measurements.

Independent measurements of water contact angle, θ_o, were performed on silicon wafers pretreated in the same way as the ODNP samples shown in Fig. 1, but without using the functionalization reagents. A plot of cos(θ_o) vs. pretreatment temperature is shown in SI Appendix, Fig. S6, displaying a moderately steep change between 600 and 800 °C, albeit much less dramatic than that seen in Fig. 1 between samples pretreated at 700 and 800 °C.

Force–Distance Measurements Between Silica Surfaces.

To better understand how the observed trends in hydration dynamics relate to interactions between silica surfaces, the equilibrium interaction forces between silica surfaces of varying silanol density were measured with the SFA under quasi-static conditions. In the SFA experiments, the absolute distance between back-silvered silica surfaces in a cross-cylinder geometry was measured by interferometry, and the interaction force F measured by the deflection of a cantilever spring (43). The distance was controlled by moving one surface with a piezoelectric crystal in steps of 2–20 nm, and 30 s of equilibrium time was allowed before the distance was measured. Force–distance profiles are typically shown as F normalized by the radius of the surfaces R plotted against the separation distance D (i.e., F/R vs. D). Values for D = 0 correspond to silica–silica contact under dry conditions. Silica surfaces in our experiments were prepared using the “bubble method,” resulting in thin (3–6 μm) sheets of glass-blown silica, described in detail elsewhere (27, 38). The silica, being mostly dehydroxylated initially (water contact angle of 38°), was subjected to UV-ozone treatments for 0.5–10 min to give water contact angles of 33°, 26°, and 0° on the surfaces. The water contact angles were used to estimate the silanol density via the Israelachvili–Gee equation (44), giving α_OH values of 4.6, 2.9, 2.0, and 1.2 nm⁻², which roughly span the range obtained in the ODNP measurements.

Representative force–distance profiles for two sets of silica surfaces are shown in Fig. 2. Decreasing the water contact angle from 33° to 26° results most notably in an added repulsion at D < 4 nm, where one would expect surface hydration to impact forces. Force profiles corresponding to θ_o = 26° match very closely with those for θ_o = 0°. The short-range repulsion decreases significantly in going from θ_o = 26° to 33°, and then remains constant up to θ_o = 38° (SI Appendix, Fig. S7). Force profiles were fit to the standard Derjaguin–Landau–Verwey–Overbeek (DLVO) model (45) comprising an attractive van der Waals force and a repulsive electric double-layer force (details given in SI Appendix, section S3). The measured forces begin to deviate from the DLVO fit below D = 4–5 nm for the 26° surface, and below D = 2–3 nm for the 33° surface. An additional short-range force that varies exponentially with distance can be added to better fit the data. The added force component increases ∼2.5× in magnitude upon decreasing θ_o from 33° to 26° (see the exponential prefactor P in SI Appendix, Table S2).

Fig. 2.

Representative SFA measurements of interaction forces as a function of distance (0.1-nm distance resolution) between silica surfaces across a 15 mM NaCl solution at 22 °C. The two surfaces used in each force run displayed the same water contact angles––either θ_o = 26° or 33° (i.e., a symmetric experiment). Values of α_OH were calculated from the Israelachvili–Gee equation (44). The solid lines correspond to DLVO fits for each data series. A Hamaker constant A of 6.1 × 10⁻²¹ J was used (45), and fitted values for the surface potential ψ_o were −60 ± 4 mV for the 26° surface, and −55 ± 3 mV for the 33° surface. Force profiles for surfaces displaying θ_o = 0° (α_OH = 4.6 nm⁻²) essentially lie on top of those for the 26° surface, and force profiles for surfaces displaying θ_o = 38° (α_OH = 1.2 nm⁻²) similarly match with those for the 33° surface, so these additional profiles are omitted here for clarity, but are shown in SI Appendix, Fig. S7. The force measurements were conducted ∼1 h after injecting the aqueous solution between the surfaces, but measurements after 48 h showed no change for the θ_o = 0° surface.

It should also be noted that for the α_OH = 4.6 nm⁻² surfaces, the forces remained unchanged after exposing the surfaces to the aqueous solution for 48 h. Furthermore, the change in silica film thickness was measured as a function of relative humidity for the most hydrophilic (α_OH = 4.6 nm⁻²) and the most hydrophobic (α_OH = 1.2 nm⁻²) surfaces. Vigil et al. observed a swelling of 1.5–2.0 nm per surface (in going from 0 to 100% relative humidity) for the hydrophilic silica surfaces used in their experiments, which was partially used to justify the existence of a gelatinous layer of silica hairs that caused the short-range repulsion. Here, we observed an increase in thickness (per surface) of 0.8 nm for the α_OH = 4.6 nm⁻² surface and 0.6 nm for the α_OH = 1.2 nm⁻² surface in going from 0 to 100% relative humidity (SI Appendix, Fig. S8), which may be due to two to three layers of bound water on each surface. Based on these results, if silica hairs contributed to the short-range repulsion in the experiments of Vigil et al., they certainly contribute less in our experiments.

Simulations of the SiO₂–Water Interface.

MD simulations of the SiO₂–water interface were performed to develop a molecular picture and rationale for the trends observed in Fig. 1. The simulated surface water diffusivity D_sim was computed for model amorphous surfaces with α_OH varied between 1.5 and 4.0 nm⁻². Methodological details related to the MD simulations and the calculation of D_sim are in Materials and Methods and SI Appendix, section S4. In general, values of D_sim were an order of magnitude larger than the ODNP D_surface values. Both techniques probe diffusive dynamics but have conceptual differences such that one should not expect the same absolute diffusivities but rather similar qualitative trends. These differences include inaccuracy in the water model selected [the four-site transferable intermolecular potential model for Ewald techniques (TIP4P-Ew) has a slightly higher diffusivity than real water], the models used to calculate diffusivity [mean-squared displacement in simulations and the force-free hard-sphere model in ODNP (30)], and the treatment of boundary conditions in developing and applying the diffusivity models. Initial simulation results showed a near-linear trend for surface water diffusivity D_sim with α_OH (SI Appendix, Fig. S9). However, we found that the spatial distribution of silanols on the surface (for a given α_OH) could result in noticeably different D_sim values (i.e., the two amorphous interfaces at 4.0 nm⁻² exhibit different D_sim values).

To understand the potential range of diffusivities accessible to a surface with a fixed composition (number) of silanol groups, due to their distinct possible arrangements, we used a genetic algorithm to adjust the patterning of silanol groups on crystalline cristobalite-$10\overline{1}$ surfaces to achieve maximum or minimum values of D_sim for four distinct α_OH values: 1.0, 1.5, 2.0, and 3.0 nm⁻² (SI Appendix, Fig. S10). Crystalline rather than amorphous surfaces were used in the genetic algorithm optimization due to greater ease in adjusting silanol patterns without the need of adjusting the underlying surface structure (Materials and Methods). This choice reduces the range of D_sim values observed at a fixed density, as the crystalline surface intrinsically presents fewer heterogeneities in its surface structure. The results of the optimization after 70 iterations (generations), are shown in Fig. 3. The difference between maximum and minimum D_sim values is ∼10% for each of the four intermediate α_OH values (the gray area of Fig. 3A), but more iterations could further increase the range. The data points for α_OH = 0 and 4.0 nm⁻² represent close-packed lattices of all siloxane or all silanol groups, for which no rearrangement of silanol groups is possible. Fig. 3B shows images of the simulated surfaces associated with the data points in Fig. 3A. The maximum diffusivity surfaces show a higher degree of clustering/clumping of the silanol groups than the minimum diffusivity counterparts. We quantified the clustering with an average silanol coordination number N_c, which is the average number of silanols within 5.5 Å of any given silanol. Further metrics of clustering are given in SI Appendix, Table S3.

Fig. 3.

MD simulations of the cristobalite–water interface. (A) Surface water diffusivities calculated from experiment and simulation vs. α_OH. The error bars for simulation are the 99% confidence interval assuming a Student’s t distribution. The solid black lines trace out the minimum and maximum diffusivity surfaces for simulation after optimization using a genetic algorithm. (B) Images of the α_OH = 1.0, 1.5, and 2.0 nm⁻² surfaces associated with the max and min data points in A. The silanol coordination number N_c associated with each surface is displayed in blue below each image.

Discussion and Conclusions

The ODNP and SFA measurements together reveal a significant transition in both hydration dynamics and silica–silica short-range forces between intermediate silanol densities of α_OH = 2.0 and 2.9 nm⁻², corresponding to a water contact angle of θ_o = 33° and 26°, respectively. The agreement between the ODNP and SFA results implies that the SFA-measured short-range forces comprise, to a significant degree, a hydration force in this system. The agreement further implies that the ODNP measurements were indeed probing the average hydration behavior on the surface (i.e., the spin-label distribution was uncorrelated with the surface chemical environment). Although the exact theoretical connection between hydration dynamics and surface forces is unclear (46), the agreement suggests that the equilibrium hydration dynamics as derived from ODNP can qualitatively reflect the properties of surface water that govern hydration forces on silica–water surfaces. Such connections have been previously demonstrated in the study of liposome–water surfaces by parallel SFA-ODNP measurements (42, 47).

The MD simulations show that a range of surface water diffusivities can be obtained at intermediate values of α_OH (1.0–2.0 nm⁻²) through the precise patterning and spatial distribution of silanols. They also show that the connectivity of surface silanol groups has a particular impact on dynamics, with higher water diffusivity arrangements manifesting close clustering. The notion that the contribution to average surface hydration from individual surfaces groups may not be additive has precedent. By measuring water density fluctuations near a mixed self-assembled monolayer (SAM) surface by MD simulations, Acharya et al. (48) found that substituting a single hydrophilic group into a lattice of hydrophobic groups (case 1) had a far greater impact on surface hydration than the reverse substitution of a single hydrophobic group into a lattice of hydrophilic ones (case 2). Similar conclusions have been reached in other simulation studies that emphasize the importance of patterning on chemically heterogeneous surfaces (49, 50). These literature results might suggest that in Fig. 1, the slope (D_surface vs. α_OH) at α_OH < 2.2 nm⁻² (similar to case 1 but without truly hydrophobic groups) should be steeper than that at α_OH > 2.7 nm⁻² (similar to case 2), but any such difference in slopes could not be observed due to the experimental uncertainty. Our simulations similarly found that the surface-projected area of the volume in which any given silanol may hydrogen bond with a water molecule decreases with increasing silanol clumping/coordination (SI Appendix, Table S3).

It is possible that the experimentally observed transition originates from a 2D percolation threshold, wherein all silanol groups are connected through a fluctuating but pronounced hydrogen-bond network (either silanol–silanol or silanol–water-silanol) once the clusters become large enough (51). Similar 2D percolation networks have been studied in the context of protein hydration, particularly lysozyme. Such simulations have found a critical water concentration where water forms a connected hydrogen-bond network around the curved protein surface, which is thought to correspond to sudden changes in experimentally observed physical properties such as solution capacitance, as well as the onset of some biological functions (52, 53).

It is not entirely clear whether the silica surface would become patchy during hydroxylation from a dehydroxylated state, or during dehydroxylation from a hydroxylated state. Water in the vapor phase selectively, or at least preferably, adsorbs to silanol groups (1), perhaps catalyzing hydroxylation of adjacent siloxane groups, suggesting autocatalytic growth of silanol patches during hydroxylation. If that is the case, the consistency between the ODNP and SFA results is striking given that partial rehydroxylation of the nanoparticles occurred in water, whereas rehydroxylation of the SFA surfaces occurred primarily during UV-ozone treatment (albeit under humid conditions). The consistency between ODNP and SFA is also intriguing given the difference in curvature of the silica surfaces used in the two techniques (average radii of curvature of ∼3.5 nm for the nanoparticles vs. ∼2 cm for the SFA surfaces). Thus, the agreement between ODNP and SFA suggests that the characteristic size of a critical cluster would potentially be no more than a few nanometers, but the fumed particles are branched and irregular, making conclusive statements difficult. Clearly, the effect of silanol distribution at intermediate silanol densities invokes robust effects on silica surface hydration that are expected to tune silica surface properties relevant to chemistry and processes involving silica surfaces.

Together, the ODNP, SFA, and MD simulation results indicate that the average surface silanol density α_OH alone is not sufficient to characterize the hydrophilicity of a silica surface, even if it is smooth, and further uncover the possibility that the steep change in hydration dynamics (Fig. 1) may be caused by a critical silanol cluster size or coordination number. Given the limitations of our simulations, such as the use of a model crystalline surface and the limited number of surface patterns explored (which still may not fully span the range of accessible diffusivities), the results cannot quantitatively fit those of Fig. 1. Rather, they bring to light a possible explanation for the experimental trend, and invite further investigation into the relevant metrics of silanol distributions and silanol speciation as a function of rehydroxylation procedures, as is relevant to practical applications in catalysis, geology, and nanofabrication.

Materials and Methods

Upon initial purchase of the commercial silica (Aerosil 380), the particles were heated to 85 °C in water for 1 wk to ensure maximum hydroxylation. If partially dehydroxylated silica was desired for a given experiment, ∼200 mg of silica was heated in a tube furnace under dry N₂ for 6 h at the target temperature using a ramp rate of 10 °C/min. Details of the silane and spin-labeling procedures are given in SI Appendix, section S1. For ESEEM measurements, the 15 mM NaCl, pH 5 solution used in the ODNP measurements was replaced with a 70% D₂0, 30% glycerol (vol/vol) solution in all cases, and the measurements were conducted at 80 K. Further details of the EPR and NMR measurements are in SI Appendix, section S2.

Nanoparticle-specific surface area was measured with the standard BET N₂ adsorption technique using a MicroMeritics TriStar 3000 Porosimeter. Silanol density was measured through deuterium exchange by immersing dry particles in D₂O containing trace (but known) amounts of HDO and 1 mM DMSO as a concentration reference. After sonicating for 10 min, the particles were removed with centrifugation (15,000 g × 5 min), and the concentration of protons in the resultant liquid was quantified with ¹H NMR. Silanol density was also measured by immersing particles in a deaerated and slightly basic 1 M NaCl solution (pH ∼ 9.5, but measured before each experiment under Ar gas), sonicating and separating out the particles as above, and measuring the pH of the resultant liquid under Ar gas. The NaCl, 3-(ethoxydimethylsilyl)propylamine (APDMES), 4-carboxy TEMPO, PBS buffer, and MES buffer reagents used were purchased from commercial suppliers and were used without further purification.

The silica SFA surfaces were prepared using the established bubble method (27, 38). Because, when using this method, the two apposing silica surfaces have different thicknesses, calculating the distance was done as in ref. 54. Water contact angle measurements were conducted using a DataPhysics OCA 25 video-based optical contact-angle measuring system. The contact angles were measured 15 s after deposition of a 2 μL droplet, and the droplet shapes were fit to a spherical cap model.

All MD simulations were carried out using the package GROMACS (release 2016.1) (55, 56). The TIP4P-Ew water model (57) was used throughout in conjunction with force-field parameters developed for crystalline and amorphous models of silica interfaces by Heinz and coworkers (58). The patterning of silanols on the surface was adjusted according to a genetic algorithm. Further details are in SI Appendix, section S4.