In-Situ X-ray Spectroscopy of the Electric Double Layer around TiO2 Nanoparticles Dispersed in Aqueous Solution: Implications for H2 Generation

We report an experimental observation of a significant amount of hydroxide (OH−) created upon water dissociation and subsequently trapped around TiO2 nanoparticles dispersed in NH4OH aqueous solution. The hydroxide species is identified and quantified by a combination of photoemission and photon emission X-ray spectroscopies conducted on liquid samples using a liquid microjet. Unlike previous X-ray studies that observed only a few monolayers of water coverage on TiO2 surfaces and found maximally submonolayer of OH−, the true aqueous environment adopted in this study enables ion mobility and the separation of the water dissociation products H+/OH−. This facilitates the formation of OH− diffused multilayer in which the trapped OH− ions are discovered to coordinate with three water molecules to form a tetrahedral hydration configuration. The negatively charged diffuse layers, together with the positive NH4 + Stern layers, constitute >0.8 nm thick electric double layers around the TiO2 nanoparticles. The large observed amount of hydroxide indicates a high efficiency of water dissociation for the TiO2 catalyst, a promising result for H2 generation in true aqueous environments.


■ INTRODUCTION
Titanium dioxide (TiO 2 ) is a promising material to catalyze water dissociation for the production of H 2 fuel as a clean and renewable energy source. 1−5 Among the three TiO 2 phases, brooktile, rutile, and anatase, the latter phase exhibits the highest photocatalytic activity 6,7 and chemical stability in various aqueous environments. 8 Many efforts have been made to improve the TiO 2 -catalyzed energy conversion efficiency in water-splitting reactions. 3,4,9−14 However, the fast backreaction of proton (H + ) and hydroxide (OH − ) recombination into water molecules reduces the efficiency significantly. 5 The spatial separation of a proton from a hydroxide is therefore very critical for the enhancement of the water dissociation efficiency and for the device performance. This separation of the dissociation products, H + and OH − ions, in the vicinity of the TiO 2 −water interface requires a true aqueous environment (rather than a few monolayers of water coverage) to enable ions to diffuse away from the interface. In fact, the electric double layer (EDL) that forms at the TiO 2 −water interface will strongly influence ion distributions and ion mobilities.
Several models have been proposed to describe the EDL, including the Helmholtz model, 15 the Gouy−Chapman model, 16,17 and the Gouy−Chapman−Stern model. 18 In the Helmholtz model, the EDL is treated analogously to a static electric capacitor 19,20 where all counterions are postulated to be confined at the TiO 2 surface, and all co-ions are repelled. The ion mobility in the solution is completely ignored in this model. 19,20 It is, however, taken into account in the Gouy− Chapman model that postulates an interfacial diffuse layer where the ion concentration follows a Boltzmann distribution away from the interface. 20 This model assumes that ions are point charges and occupy no space, which can lead to an overestimation of surface charge. 20 The Gouy−Chapman− Stern model, 18 based on the two previous models, introduces an additional layer: the Stern layer where some or all counterions are confined at the TiO 2 surface. The remaining counterions or co-ions, depending on the electrolyte concentration, are mobile and form the diffuse layer. 21 As in the Gouy−Chapman model, ions follow a Boltzmann distribution across the layer. The thickness of the EDL in the Gouy−Chapman−Stern model is defined as the Debye length, which is strongly dependent on the electrolyte concentration. 19 In this study, the Gouy−Chapman−Stern model is most appropriate for an analysis of the acquired spectra.
Although the charge parity of the EDL at the TiO 2 −water interface in a dilute aqueous solution has been extensively investigated by various methods, 22−25 we still lack a quantitative molecular-level experimental insight into the detailed EDL structure, including the Debye length (electrostatic screening length). Soft X-ray spectroscopy studies that can directly reveal element-specific electronic structure of interfacial species are still scarce for solid surfaces in contact with real (bulk) aqueous solution. In recent years, the development of ambient pressure X-ray photoelectron spectroscopy (AP-PES) has allowed for instance probing of the EDL of a thin electrolyte film on top of a polycrystalline gold electrode created by the so-called "dip-and-pull method" by using tender X-ray radiation. 26 In the aforementioned study, the Debye length, measured as a function of the KOH electrolyte concentration, has been reported to decrease monotonically from 30.4 nm at 0.1 mM to ∼1.9 nm at near 80 mM concentration. 26 Despite extensive experimental efforts, PES studies seeking to reveal the EDL dimension have remained limited to nanometers thick water coverage, in which case ion mobility and ion adsorption/desorption at the TiO 2 surface will be different than in bulk solution. 27 In-situ probing of the TiO 2 surface fully immersed in liquid water is therefore of great importance despite the experimental challenges it poses.
In this study, X-ray photoelectron spectroscopy (XPS) and partial electron/fluorescence yield X-ray absorption spectroscopy (PEY/PFY-XAS) measurements were conducted on anatase-phase TiO 2 nanoparticles (NPs) dispersed in slightly basic aqueous NH 4 OH solution (pH ∼7.8). We have recently demonstrated dissociative water adsorption at the NP− solution interface in such a TiO 2 system. 28 Here, we report complementary electron-yield and photon-yield detections to quantify the amount of water dissociation at the TiO 2 −water interface as well as to estimate the Debye length. The NP solution has a concentration of 20 wt % corresponding to an average separation distance between the dispersed NPs of ∼80 μm. In this case of high NP concentration, the Stern layer is composed of stabilizer NH 4 + ions (discussed later with Figure  2). The stable NP solution (with no aggregation) can be well described by the Derjaguin, Landau, Verwey, and Overbeek (DLVO) theory. 19,20 At high electrolyte concentration, the NP interaction potential barrier is lowered, and a minimum in the potential curve (a secondary energy barrier with oppositely charged ions) can lead to slow NP aggregation. 20 For electrolyte concentrations above the so-called critical coagulation value, the energy barrier reduces to below zero and triggers NPs to aggregate rapidly. 20 The TiO 2 NP colloidal solution in this study is of relatively high concentration but is still below the critical value. We briefly review the expected nature of the EDL and how it depends on the electrolyte concentration. As illustrated in curve A of the left panel of Figure 1, at low electrolyte concentration, positive charge (NH 4 + cations) accumulates at the NP surface to establish charge neutrality of the NP + EDL entity. Curve B of the left panel of Figure 1 shows how at high electrolyte concentrations, negative charge from the electrolyte (OH − in our case) will be drawn closer to the NPs and stay associated with them throughout the EDL (Figure 1, right panel). The primary goal of the present study is to use PES together with the more bulksensitive PFY-XA spectroscopy to identify the molecular nature of the EDL, including the quantity and composition of oppositely charged ions, and to estimate the thickness of the EDL and infer hydration structure details. Our finding of a large quantity of hydroxides around the TiO 2 NPs not only indicates efficient water dissociation reactions catalyzed by TiO 2 NPs but also points at an efficient pathway for H 2 generation catalyzed by TiO 2 surfaces in aqueous solutions.

■ EXPERIMENTAL SECTION
Both the PES and PEY-XAS measurements were conducted with the SOL 3 PES setup, 29 and the PFY-XAS studies were performed with the LiXedrom setup, 30 at the synchrotron-radiation facility BESSY II, Berlin. For all measurements we used the soft X-ray beamline U49/2-PGM1. A 20 wt % colloidal solution of 20 nm diameter anatase-phase TiO 2 NPs dispersed in 0.3 M NH 4 OH aqueous solution was acquired from NYACOL, USA, and introduced into the experimental vacuum chambers via a liquid microjet. 31 We also investigated 0.5 and 2 M NH 4 OH and 5 M NaOH aqueous solutions (prepared by using ammonium hydroxide solution28% NH 3 in H 2 O, >99.99% purity, and sodium hydroxide, >97.0% purity, provided from Sigma-Aldrich (Merck)) as well as a 20 wt % colloidal solution of 10 nm diameter anatase-phase TiO 2 NPs (>99.9% purity) dispersed in 0.5 M HCl aqueous solution purchased from MK Nano, Canadaall of which served as spectral references. Note that the NP surface in 0.5 M HCl aqueous solution is fully covered by adsorbed Cl − ions, so water cannot interact with the TiO 2 NP. This provides a useful spectroscopic reference and has been applied in several previous liquid microjet PES studies. 29 (1) making the solution basic. This reaction is dependent on temperature and pH, and an equilibrium between NH 3 and NH 4 + species is established at approximately pH 9 (pK a ∼ 9) at 24°C. 39 Figure 1, right panel), and as a result, there is no free NH 4 + in the solution. NH 4 + adsorption thus perturbs the ammonium−ammonia chemical equilibrium (eq 1) and leads to the production of more OH − . However, the experimentally measured pH of the [2:1] NH 4 + NP colloidal solution is 7.8, just slightly above neutral pH. This surprising observation implies the presence of additional reactions other than the ammonium−ammonia equilibrium that occur in the solution and affect the number of free OH − ions. To determine these additional reactions, we first investigate NH 4 + ions in the presence of TiO 2 NPs and subsequently compare with NH 4 OH aqueous solution (without TiO 2 NPs added) by measuring the PEY-XA at the N K-edge and XP spectra of the N 1s orbital. We then explore the OH − species by recording the O 1s XP spectra from the 5 M NaOH, 0.5 M NH 4 OH,  Figure 2A shows the N K-edge PEY-XA spectra of the [2:1] NH 4 + NP solution and 2 M NH 4 OH aqueous solution (pH ∼12). These two spectra are generated by integrating the signal intensity of the valence-band photoelectron spectra (not shown here) as a function of photon energy; see, e.g., ref 41 for a more detailed explanation. Observed spectral differences reveal the existence of different nitrogen-containing species in the two solutions. The spectrum of the 2 M NH 4 OH solution (Figure 2A, bottom) primarily exhibits spectral features characteristic of neutral gas-phase ammonia (NH 3 ). 42 This can be explained by the ammonium−ammonia equilibrium reaction (eq 1). At high pH ∼12, the equilibrium reaction

ACS Applied Nano Materials
Article moves toward the NH 3 (aq) side; because of ammonia's high volatility 39,40 and the rather short probing depth of soft X-ray PES, the spectrum is then dominated by the NH 3 (g) signal.
Note that Henry's constant for the hydrated NH 3 is relatively high, 59 mol/atm at 25°C. 43 Under the experimental conditions, we expect NH 3 (aq) to be below the detection limitthis is confirmed by the absence of the main feature that would be expected to appear between the first two peaks of gas phase followed by an intense absorption step edge, 42 clearly not present in the bottom spectrum of Figure 2A. Our observation is in agreement with a previous N 1s Auger electron/autoionization PE study from a 2.6 M NH 3 aqueous solution. 44 Turning now to the N K-edge PEY-XA spectrum of the [2:1] NH 4 + NP solution (Figure 2A, top), we find reasonable resemblance to the reported spectrum of a NH 4 + aqueous solution, 43 but a better agreement is obtained when comparing to the dry NH 4 + salts. 45,46 Considering the nearly neutral pH of this solution, it suggests that NH 4 + ions are adsorbed at the TiO 2 NP surfaces, acting as a stabilizer.
Our conclusion on the nature of the nitrogen-containing species is corroborated by the N 1s XP spectra from the same two solutions, also presented in Figure 2. Measurements were performed at two photon energies: 800 eV ( Figure 2B) and 1200 eV ( Figure 2C). The latter energy probes deeper into the solutions due to the larger kinetic energies (KE) of the photoelectrons. 47 Beginning with the 800 eV spectrum taken from the 2 M NH 4 OH solution ( Figure 2B, bottom), we observe a large peak at 405.9 eV binding energy (BE) arising from N 1s ionization of gas-phase ammonia, NH 3 (g). The signal from NH 3 (aq) is weak, giving rise to a small shoulder at 404.8 eV BE. In contrast, the 800 eV PE spectrum of the [2:1] NH 4 + NP solution ( Figure 2B, top) exhibits a single peak at larger BE of 407.0 eV originating from N 1s ionization of adsorbed ammonium stabilizer, NH 4 + . This is exactly what we would expect from Figure 2A, top. Note also that in both cases the energies of the detected electrons are rather similar. That is, the KEs of the N 1s photoelectrons (excited by 800 eV photon energy) and of the N 1s Auger electrons (from which the PEY-XA spectrum in Figure 2A, bottom, has been obtained) are ∼400 eV. At this energy the electron probing depth is thus too small to detect a noticeable signal from NH 3 (aq). Increasing the photon energy from 800 to 1200 eV has a minor effect on the XP spectrum of [2:1] NH 4 + solution ( Figure 2C, top), but one observes a drastic change of the spectrum from the 2 M NH 4 OH solution ( Figure 2C, bottom). Because of the increased probing depth, the spectrum of the 2 M NH 4 OH solution now exhibits, in addition to the NH 3 (g) signal at 405.9 eV, a second peak at 404.8 eV of similar intensity arising from NH 3 (aq). This implies that the majority of NH 4 + in the [2:1] NH 4 + NP solution cannot be assigned to species solvated in bulk solution. If this were the case, a significant amount of NH 3 (g) (and some NH 3 (aq) at 1200 eV excitation energy) signal would be detected due to the ammonia−ammonium equilibrium (eq 1). Therefore, Figure 2 reveals that almost all NH 4 + ions in the [2:1] NH 4 + NP solution are adsorbed on the TiO 2 NP surface, covering about half of the available surface area, as mentioned above. The adsorbed NH 4 + ions constitute a positively charged Stern layer of the EDL. The thickness of this Stern layer is on the order of the diameter of the NH 4 + ion which is ∼0.35 nm, assuming a 0.175 nm NH 4 + ionic radius. 48 We next focus on the nature of the oxygen-containing species, with a particular interest in identifying dissociated water and estimating the amount of OH − anions, which allows for an approximation of the length of the diffuse layer. For this we have measured the oxygen 1s PE (Figure 3) and oxygen Kedge PFY-XA ( Figure 4) spectra from the same [2:1] NH 4 + NP solution. In addition, we present the respective spectra from two reference aqueous solutions: 5 M NaOH and 0.5 M NH 4 OH. The 5 M NaOH solution is used to establish the OH − PES peak position and to identify the effect of OH − on the water structure through changes of the water O 1s PE peak. Note that 0.5 M concentration is reasonably close to the 0.3 M NH 4 OH concentration used to stabilize the aqueous-phase TiO 2 NPs. All PE spectra were measured at 1200 eV photon energy, at the same energy which we have applied to increase the probing depth into the solution in the case of the N 1s PE measurements ( Figure 2C). Figure 3A shows the O 1s PE spectrum from the 5 M NaOH aqueous solution. Three major oxygen features can be observed: the dominating liquid-water peak at 538.1 eV BE, 31 the water-vapor contribution at 539.9 eV, 49,50 and the hydrated OH − giving rise to a shoulder at 535.8 eV. 51   1] NH 4 + NP + 0.5 M NaOH (D) aqueous solutions measured at 1200 eV photon energy. Note that solution in (D) has pH of only 9.8, whereas the pure 0.5 M NaOH solution has pH 13.7. All spectra are fitted with multiple Gaussian functions that each represent the different oxygen species: liquid water (H 2 O(l)), gas-phase water (H 2 O(g)), OH − , and TiO 2 . Spectra are presented on the binding energy scale, and intensities of the different spectra are displayed to yield the same peak height of the liquid-water peak. Dots are the measured data, and the solid red lines are the total fits. The detailed fitting parameters are presented in Table S1.

ACS Applied Nano Materials
Article exhibits no OH − signal at this low OH − concentration; this is also true for a NaOH aqueous solution of the same concentration. 51 Another important difference between the spectra of lower (0.5 M) and higher (5 M) OH − concentrations is the change of the water O 1s peak width. The full width at half-maximum (fwhm) of the lowconcentration spectrum is ∼1.6 eV (which is almost the same as for neat liquid water 50 ) and increases by 12.5% (0.2 eV wider) for the higher concentration. Qualitatively, such a broadening is in accord with a theoretical finding that hydration of hydroxide affects the electronic structure of water, 52 and it also agrees with a previous experimental and computational PFY-XA study of hydroxide aqueous solution 53 which will be detailed below along with the discussion of Figure 4.
The O 1s PE spectrum of the [2:1] NH 4 + NP solution, shown in Figure 3C, is rather similar to the spectrum of the 5 M NaOH aqueous solution. However, there is an additional small peak from the TiO 2 lattice oxygen at 534.8 eV BE, and three additional subtle but significant spectral differences can be observed. (1) The OH − signal intensity from the NP solution amounts to almost 70% of that from the 5 M NaOH solution. According to the spectrum of 0.5 M NH 4 OH ( Figure 3B), one would expect to observe no OH − signal from the [2:1] NH 4 + NP solution, let alone to exhibit the intensity comparable to that of the 5 M NaOH aqueous solution.
(2) The fwhm of the water O 1s PE peak is very similar to the one from neat water, and the peak is not broadened unlike in the case of the 5 M NaOH aqueous solution.
(3) The OH − peak position is at slightly smaller (by 0.5 eV) BE compared to the PE spectrum from 5 M NaOH aqueous solution ( Figure 3A). All relevant BEs, fwhm, and peak areas are provided in Table S1.
Observation 1 indicates that in the NP solution the number of OH − significantly exceeds the number that would be provided from 0.3 M concentration, and yet the pH of the [2:1] NH 4 + NP solution is only 7.8. Observations 2 and 3 imply that OH − species in the NP solution are accommodated in a different chemical environment than the hydrated OH − in water at basic pH. In our previous work, in the same [2:1] NH 4 + NP aqueous solution, 28 we have reported that water dissociates at the TiO 2 NP surface, which can qualitatively explain the increase of OH − signals (observation 1) in the spectrum. An important consideration is that observation 1 is at odds with the fact that the pH of the [2:1] NH 4 + NP solution is only 7.8. It thus implies that most of the OH − molecules in the NP solution must be trapped around the TiO 2 NPs by the positive Stern layer, forming the diffuse layer of the EDL; these confined OH − species make no contribution to the pH measurement. Analogous to our previous photoemission study of aqueous-phase TiO 2 NPs, 28 we have also measured the [2:1] NH 4 + NP solution with 0.5 M NaOH added ( Figure 3D) to show that the intensity of the OH − peak at 536.3 eV continues to increase, becoming as large as for the 5 M NaOH solution. However, the pH is ∼9.8, and the added amount of hydroxide could not produce such signal enhancement. This further corroborates our conclusion that a large amount of OH − is indeed produced by additional water dissociation at the TiO 2 surface. The implications of observations 2 and 3 together with the findings of Figure 3D will be detailed in the following paragraphs, where we acquire additional support from PFY-XA measurements. The latter is indispensable here as it is a bulksensitive method, enabling a realistic estimate of the amount of OH − present in the NP solution.
The O K-edge PFY-XA spectrum of the [2:1] NH 4 + NP solution, with the assistance of the reference PFY-XA spectra of [1:1] Cl − NP solution, neat water, and 1.5 M NaOH aqueous solution, allows for a quantitative estimation of the OH − amount in the NP solution. All PFY-XA spectra, shown in Figure 4, were generated by integrating the X-ray emission intensities resulting from valence band to core transitions (O 2p → 1s) following resonant excitation at the O K-edge. Note that the PFY-XA spectrum of the 1.5 M NaOH aqueous solution, shown in Figures 4B and 4D, has been adapted from ref 54. This spectrum helps to quantify the amount of OH − in the NP solutions, which will be explained later. Figure 4A shows the spectra from the [2:1] NH 4 + NP solution and the  NP and 1.5 M NaOH solution spectra are presented in (C) and (D), respectively. Different filling colors refer to different oxygen species: liquid-water oxygen (no filling), gas-phase water (gray), OH − (blue and purple), and TiO 2 lattice oxygen (e g and t 2g , green). The two spectral backgrounds in (C) and (D) are fitted with a cubic function, indicated by gray curves (see also Figure S1). The detailed fitting parameters are presented in Table S2.

Article
We label this signal contribution OH − (i*). Then in Figure 4B, we compare OH − (i*) with the OH − signal obtained from the neat 1.5 M NaOH aqueous solution (no NPs contained). For that we subtract the PFY-XA spectrum of neat water (shown in blue) from the one of the NaOH solution (in brown). This difference spectrum is presented at the bottom of Figure 4B. Unlike in Figure 4A, we now observe two contributions, centered at 532.8 and 534.1 eV photon energy, labeled OH − (i) and OH − (ii), respectively. These two OH − contributions have been previously reported for 4 and 6 M KOH aqueous solutions and were attributed, based on computations, to the hydroxide hydration complexes in tricoordinated (tetrahedralcoordinated) OH − (H 2 O) 3 and in planar 4-fold-coordinated (hypercoordinated) OH − (H 2 O) 4 configurations, respectively. 53 The tetrahedral configuration produces a single preedge peak at 532.5 eV photon energy (matching the OH − (i) peak position in Figure 4B), whereas the hypercoordinated configuration gives rise to a pre-edge peak at 534.5 eV (coinciding with OH − (ii) in Figure 4B) as well as to OH − (i). 53 The good agreement between the computed energies for OH − (i) with the experimental energy of OH − (i*) strongly suggests that not only is the OH − hydration configuration near the NP surface different than in bulk water, but the tricoordinated structure dominates. Note that the occurrence of different OH − configurations would be qualitatively consistent with the different O 1s BEs found in the NP solution compared to the 5 M NaOH solution (Figures 3A and  3C). Our suggested hydroxide hydration configuration within the EDL is corroborated by a Car−Parrinello molecular dynamics (CPMD) simulation, finding that the fraction of 3and 4-coordinated structures depends strongly on the OH − concentration. At low concentration, OH − ions are surrounded by abundant water molecules, and the OH − (H 2 O) 4 hypercoordinated configuration is preferred; the tetrahedral complex OH − (H 2 O) 3 quickly (within 2−3 ps) interconverts to OH − (H 2 O) 4 . 55,56 At high concentration, fewer water molecules are available to coordinate OH − , making the OH − (H 2 O) 3 configuration more favorable. 57 We argue that the OH − concentration in the diffuse layer is high, although we have yet to quantify the concentration, and thus the 3coordinated structure dominates. This interpretation is supported by a nuclear magnetic resonance study of the water molecule distribution around TiO 2 NPs, showing strong confinement of water molecules with low mobility and reactivity within the first few layers above the TiO 2 surface. 58 The potential cation effect of Na + on the hydroxide and water spectra is found to be negligible in this study, even though Na + was considered to be bound to hydroxide and water molecules in highly concentrated solutions. 59 We have recently examined the influence of NaI salt on the electronic structure of liquid water by PES and molecular dynamical and quantum chemical calculations 60 and found no significant effect from NaI, up to 8 M concentration, on the water PE spectra. The Na + effect on O K-edge XA spectra were investigated previously as well, 61 which showed only 0.09, 0.04, and 0.02 eV energy shifts of the water prepeak at ∼535 eV for NaCl, NaBr, and NaI solutions of 3 M concentrations.
Having characterized the OH − hydration configurations and their contributions to the PFY-XA spectra, we now quantify the amount of OH − within the EDL. The O K-edge PFY-XA spectra of [2:1] NH 4 + NP ( Figure 4A) and 1.5 M NaOH solutions ( Figure 4B) solutions are then fitted with Gaussians, each representing the respective oxygen component, as shown in Figures 4C and 4D. In both cases, the well-known water preedge peak at 535.0 eV 62 is presented in blue (unfilled) and is used for intensity normalization; the relative intensity of this signal will be used to estimate the OH − concentration. In Figure 4C, the two TiO 2 oxygen features at 531.1 and 533.3 eV (green Gaussians), resulting from O 2p orbital hybridization with Ti-3d-derived t 2g and e g orbitals, 63

ACS Applied Nano Materials
Article OH − signal) shown in Figure S1 and Table S2. It is clear that the fitted Gaussian at 532.6 eV (light blue; Figure 4C), which accounts for the OH − signal, is needed to yield a good total fit (red line) of the overall spectrum. The 532.6 eV peak position perfectly matches the position of the computed tetrahedral species OH − (H 2 O) 3 (feature OH − (i) at 532.5 eV photon energy), 53 in agreement with our above, more qualitative analysis. We notice that in our previous O 1s PE spectroscopy study of TiO 2 NP aqueous solution 28 the obtained PEY-XA spectra exhibit somewhat smaller OH − signal with respect to the water pre-edge intensity. Also, the TiO 2 signal is considerably smaller in the electron-yield measurements. It is this signal-intensity discrimination intrinsic to the (rather surface-sensitive) PES study that does not allow for a quantitative estimate of the relative concentration of the (oxygen) species of interest. In contrast, the signals obtained with bulk-sensitive PFY are argued to quantitatively reflect the total TiO 2 lattice and OH − concentrations in solutions.
In Figure 4D, we present the individual contributions of OH − (i) and OH − (ii) to the PFY-XA spectrum from 1.5 M NaOH aqueous solution (of Figure 4B), where we are guided by the analysis of ref 53. We first fit the water pre-edge peak (see Figure S1 and its caption) and then include two Gaussian peaks, representing OH − (i) and OH − (ii). The obtained peak positions (532.8 and 534.1 eV photon energy) are in very good agreement with the energies computed for the hypercoordinated hydroxide. 53 For estimating the OH − amount in the NP solution, we then compare the Gaussian peak area of tetrahedral (light blue peak in Figure 4C) with the area of hypercoordinated (light violet peaks in Figure 4D) OH − species, from which an ∼0.6 M OH − concentration for the [2:1] NH 4 + NP aqueous solution is derived. Here, we have used the OH − -to-H 2 O (prepeak) intensity ratio of Figure 4D as a calibration point; see the fitting parameters in Table S2 and related calculation details in the Supporting Information.
We recall that the 0.3 M NH 4 OH aqueous solution has a pH of 11.3, whereas the pH of the [2:1] NH 4 + NP solution (prepared in 0.3 M NH 4 OH aqueous solution) is only 7.8. Despite this near-neutral pH, the OH − concentration in the NP solution is estimated to be 0.6 M, and almost all these anions are confined within the EDL. Given that at 0.3 M NH 4 + concentration about half the TiO 2 NP surface sites are covered (estimated from 5.2 ions adsorbed/nm 2 ), 37,38 0.6 M OH − (H 2 O) 3 would be enough to exceed one monolayer coverage, forming the diffuse layer. This raises several questions: (1) how thick is the diffuse layer, (2) what is the origin of the 0.6 M OH − concentration, and (3) what role does the produced H + (from water dissociation) play in this interfacial system? To answer these questions, we propose two scenarios, as schematically presented in Figure 5. Both cases build on the negatively charged NP surface as presented in Figure 1B as well as on the right-hand side of this figure. In this regime, the potential parity is reversed between the Stern layer and the diffuse layer.
Scenario A ( Figure 5A) proposes that 0.3 M NH 4 OH solution contributes maximum 0.3 M OH − concentration to the diffuse layer (bulk OH − drawn toward the NPs by Coulombic attraction of the Stern layer), and the other 0.3 M OH − is provided by water dissociation at the TiO 2 NP surface. Indeed, our previous liquid microjet PES study of the same [2:1] NH 4 + TiO 2 NP colloidal solution has already shown that OH − can be produced in basic solution from water dissociation, 28 and subsequent OH − liberation (into the diffuse layer) allows for further dissociation. The amount of OH − that can be accommodated in the diffuse layer depends on the effective charge of the Stern layer, and one crucial question is whether H + (from water dissociation) is part of the Stern layer. As discussed in the Introduction, fast H + −OH − recombination is one of the major obstacles limiting the efficiency of water dissociation reactions. In the present scenario, this problem is bypassed by assuming that H + is bound at the TiO 2 surface. This confinement of H + increases the Stern layer charge and keeps the extra negative OH − species stay close, within the diffuse layer. Such a mechanism would be consistent with a previous scanning tunneling microscopy study finding that H + bonds to the oxygen sites of TiO 2 surfaces. 68 In scenario B (Figure 5B), the 0.6 M OH − concentration is assumed to be completely produced by water dissociation at the TiO 2 NP surfaces. Because of the limited number of anchoring oxygen sites for water on the TiO 2 surface, the excess amount of OH − can be rationalized when assuming that some fraction of H + ions produced at the interface migrate through the diffuse layer and recombine with the original free OH − from the 0.3 M ammonia solution to form water. This latter neutralization would explain the observed relatively low pH 9.8 when adding NaOH to the NP solution (see Figure 3D and respective figure caption). Note that some fraction of H + will inevitably recombine with OH − in the diffuse layer, and the resulting loss of OH − molecules can be replenished because the H + release from the TiO 2 surface vacates adsorption sites for further water dissociation reaction, which in turn generates additional OH − in the diffuse layer. Even if we do not understand the detailed mechanism of such efficient transfer through the diffuse layer (although very thin; discussed below), OH − in the bulk solution, as well as in the diffuse layer, must play an important role in initiating the release of H + from the TiO 2 surface. This dynamical cycle of continuous freeing surface sites for water adsorption, and the subsequent release of H + , followed by recombination, will reach an equilibrium once the bulk solution is nearly neutralized. It is worth to note that the different origins of OH − generation proposed in scenarios A and B cannot be experimentally distinguished based on the current experimental data. To accurately understand the mechanism of water dissociation, further studies are required and will help to find ways to continuously extract H + from TiO 2 catalytic sites to generate H 2 for a sustainable catalytic reaction. Such a proposed H + dynamics has been supported by photoelectrochemical cell (PEC) studies under external electric potential. 69−71 Water dissociation was found to occur at the TiO 2 working electrode, followed by H + diffusion through an electrolyte solution to the counter electrode; the nature of the EDL at the working electrode−electrolyte interface was not addressed though.
Finally, the thickness of the diffuse layer is estimated based on the 0.22 nm O−O bond length reported for the OH − (H 2 O) n complex. 53 The 0.6 M tetrahedral OH − (H 2 O) 3 complex derived from the quantitative analysis of the PFY-XA spectra in Figure 4 can then form an at least 0. 44

■ CONCLUSIONS
We have applied in-situ X-ray spectroscopy to characterize the electric double layer (EDL) around 20 nm TiO 2 NPs dispersed and stabilized in NH 4 OH aqueous solution. Our studies thus capture the interaction of bulk electrolyte solution with TiO 2 NP surfaces. By combining PES, PEY-XAS, and PFY-XAS, we find that at low concentration of NH 4 + stabilizer ionsbut still assuring that the NPs do not aggregatethere remains sufficient NP surface area to interact with water. The PES and PEY-XAS measurements at the N 1s and O 1s core levels, along with the measured pH, identify the Stern layer as consisting of NH 4 + stabilizer ions and the diffuse layer being composed of OH − . Moreover, the PFY measurements at the O K-edge reveal an unexpectedly large amount of OH − in the NP aqueous solution (despite near-neutral pH) that can only be explained by water dissociation. These OH − species are proposed to form an approximately >0.44 nm thick diffuse layer, charge-balanced by an ∼0.35 nm Stern layer of NH 4 + . The PFY measurements furthermore strongly suggest a prevailing 3-coordinated OH − hydration structure, OH − (H 2 O) 3 , within the diffuse layerwhich is not the case for OH − hydration in bulk solution where the 4-coordinated structure, OH − (H 2 O) 4 , is more probable. We conclude by proposing two models in which the dissociation product H + ions are either completely anchored at the TiO 2 −solution interface (scenario A) or partially escape into the bulk solution (scenario B), both consequently preventing the unwanted H + −OH − recombination. Scenario B is probably a more realistic description of the dynamics of dissociative water adsorption accompanied by H + release and recombination cycle, although elements of both models are likely relevant. The continuous H + release from TiO 2 NP surfaces, stimulated by OH − in the bulk solution or external electric potential applied from electrode, implies efficient water dissociation reactions and sustainable H 2 generation that can be realized in aqueous solutions. In addition, the present work marks an important advance in the spectroscopic characterization of the EDL, including the molecular interactions of aqueous solution with solid surfaces that are relevant for energy-material research. Our method can be readily applied to surfaces other than those studied here.

* S Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsanm.9b01939.
Detailed fitting parameters of the PE spectra in Figure 3, PFY-XA spectra in Figure 4 and the estimation details of OH − molarity based the fitted peak areas (PDF)

■ AUTHOR INFORMATION
Corresponding Authors *E-mail hebatallah@fhi-berlin.mpg.de. *E-mail jie.xiao@helmholtz-berlin.de.  Understanding of metal oxide/water systems at the molecular scale: structural evolution, interfaces, and dissolution". R.S. also gratefully acknowledges an Emmy Noether Young Investigator stipend through the DFG (project SE 2253/3-1). Open Access funding was provided by the Max Planck Society.

Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
We thank Stephen E. Bradforth and Thomas Kuḧne for their fruitful discussions. We also thank Aaron Ghrist for language editing and proofreading the manuscript. The authors thank the staff at the Helmholtz-Zentrum Berlin and BESSY II for assistance during measurements.

ACS Applied Nano Materials
Article