Self-Limiting Adsorption of WO3 Oligomers on Oxide Substrates in Solution.

Electrochemical surface science of oxides is an emerging field with expected high impact in developing, for instance, rationally designed catalysts. The aim in such catalysts is to replace noble metals by earth-abundant elements, yet without sacrificing activity. Gaining an atomic-level understanding of such systems hinges on the use of experimental surface characterization techniques such as scanning tunneling microscopy (STM), in which tungsten tips have been the most widely used probes, both in vacuum and under electrochemical conditions. Here, we present an in situ STM study with atomic resolution that shows how tungsten(VI) oxide, spontaneously generated at a W STM tip, forms 1D adsorbates on oxide substrates. By comparing the behavior of rutile TiO2(110) and magnetite Fe3O4(001) in aqueous solution, we hypothesize that, below the point of zero charge of the oxide substrate, electrostatics causes water-soluble WO3 to efficiently adsorb and form linear chains in a self-limiting manner up to submonolayer coverage. The 1D oligomers can be manipulated and nanopatterned in situ with a scanning probe tip. As WO3 spontaneously forms under all conditions of potential and pH at the tungsten-aqueous solution interface, this phenomenon also identifies an important caveat regarding the usability of tungsten tips in electrochemical surface science of oxides and other highly adsorptive materials.

Introduction

Metal oxides—abundant and robust—are the prime material candidates for energy-related applications in electro-, photo-, and heterogeneous catalysis.[1] Establishing structure–reactivity relationships, to allow rational design of improved catalysts, requires a fundamental understanding of the structural basis of the processes involved, and ideally atomic-level control over defects and dopants. Surface science methods[2] offer substantial opportunities[3] but typically operate in ultrahigh vacuum (UHV). Experiments under well-defined but realistic atmospheres and conditions relevant for applications are therefore urgently needed. Electrochemical surface science pursues an atomic-level understanding of structure and changes thereof under electrochemical conditions,[4,5] with electrochemical scanning tunneling microscopy (EC-STM) as a main experimental tool.

From the early days of STM, which since has established itself as one of the key techniques for the study of surfaces in real space, tungsten tips have been the most widely used probes, both in vacuum and under electrochemical conditions.[6−8] In addition to their low cost and high hardness (Mohs 7.5), the ease of electrochemical etching,[9] compared to Pt–Ir[10] and Au,[11] to shape tips from a wire in a concentrated hydroxide solution has certainly contributed to their popularity. As different tip metals have different electrochemical stability windows, tip material and coating are decided on the basis of the system under study.[12−14] For tungsten, during the etching process, anodic oxidation yields a tungstate (WO42–) that dissolves efficiently in the etching solution at high pH.[9] The close proximity of the STM tip to the surface under study as a prerequisite for the tunneling process also implies very short diffusion paths for material originating at the tip to reach the substrate. On the basis of this concept, fast to-and-fro diffusion of a redox species between a tip and a surface has enabled the electrochemical detection of single redox molecules.[15] On a related note, metal electrodeposition on the tip, followed by jump-to-contact transfer of metal clusters on well-defined substrates, has been demonstrated.[16] Also, other modes of near-direct contact between tip and substrate have been explored for ultralocal surface modification, including alloy formation,[17] substrate micromachining,[18] and controlled scission of bonds in covalently grafted species.[19]

Here, we demonstrate, using an electrochemical surface science approach, how tungsten(VI) oxide (WO3), spontaneously generated at tungsten EC-STM tips, forms one-dimensional adsorbates on two atomically flat oxide surfaces (rutile TiO2(110) and magnetite Fe3O4(001)). The concept of an STM tip as the source of metal ions is akin to the “electrochemical evaporator” electrode proposed by Wandelt.[20] Tungsten(VI) oxide, often in combination with other oxides such as TiO2, is an important visible-light photocatalyst[21] and electrochromic material.[22] Many synthetic approaches of variously structured—from amorphous to nanocrystalline—WO3 films and composites have been proposed,[23] including electrodeposition,[24] but often with only mesoscopic materials characterization. The present study is the first to address the WO3–oxide interface under electrochemical conditions at the atomic scale.

Experimental Section

Rutile TiO2(110) samples (SurfaceNet GmbH, hat-shaped, miscut <0.1°) were prepared using a wet-chemical procedure yielding a well-defined, atomically flat bulk-truncated (1 × 1) surface. Briefly, the samples were ultrasonicated in a neutral detergent solution (Merck Extran M02; 2% v/v in water; pH ca. 8) to remove polishing debris, followed by rinsing in ultrapure water (Milli-Q, Millipore, 18.2 MOhm cm, ≤3 ppb total organic carbon). The samples were then annealed in a 20:80 oxygen:argon atmosphere, and their conductivity was increased to enable STM observation, by reduction in UHV at 750 °C. Finally, adventitious carbon was removed by heating (65 °C, 8 min) the samples in a 3:1 v/v mixture NH3 25%:H2O2 30%, followed by copious rinsing with ultrapure water and immediate transfer to the EC-STM cell.

Cyclic voltammetry was performed using a Metrohm–Autolab PGSTAT32 potentiostat and a standard two-compartment glass cell carrying a reversible hydrogen reference and Pt wire counter electrode. All electrochemical potentials are reported versus the normal hydrogen electrode (NHE). EC-STM was performed with an Agilent 5500 AFM/STM with built-in bipotentiostat, using electrochemically etched W tips (from 0.25 mm wire, 99.95%, annealed, Advent UK)[9] coated with a thermoplastic polymer to minimize capacitive and Faradaic current, and a palladium hydride reference electrode. The Pt-wire counter electrode was flame-annealed before use. The EC-STM cell was placed in an environmental chamber that was purged with high-purity Ar (99.999%, Air Liquide, additionally purified with a MicroTorr point-of-use purifier). The electrolyte was prepared from ultrapure water and ultrapure 70% HClO4 (Merck suprapur) or reagent-grade NaClO4·H2O (VWR), which were both used as received. All glassware and the Kel-F EC-STM cell were cleaned by boiling in 20% nitric acid and rinsing with ultrapure water. X-ray photoelectron spectroscopy (XPS) and UHV-STM were conducted in an Omicron UHV system with a base pressure of 1 × 10–10 mbar using Mg Kα X-rays and a SPECS PHOIBOS 100 analyzer at normal emission with a pass energy of 20 eV. The size of STM features (e.g., width of tungsten oxide adsorbates) was obtained by averaging 20 manual measurements on the same STM image, after calibration of the scanner based on known lattice parameters (here, row–row distance on rutile TiO2(110)). The indicated error bar is twice the estimated standard deviation (95% confidence interval assuming normal distribution). For electrochemistry-to-UHV transfer, the sample was removed from the electrochemical cell, rinsed copiously with ultrapure water and inserted into the loadlock of the UHV chamber, after venting the former with high-purity Ar. The loadlock was evacuated with a liquid-nitrogen-cooled sorption pump (ca. 5 min) and then opened to a turbomolecular pump running at full speed in order to minimize contamination by the oil diffusing from the rotary pump. After ca. 20 min of pumping, a pressure of 1 × 10–6 mbar was reached in the loadlock, allowing sample transfer into the main chamber.

Results

Figure a shows an EC-STM image of rutile TiO2(110) in 0.1 M perchloric acid, imaged with a tungsten tip. Large terraces and monatomic steps are easily discerned. The atomically resolved image in Figure b matches the appearance of this surface in UHV,[25] i.e., a bulk-truncated (1 × 1) structure with alternating bright and dark rows along the [001] direction. In UHV, the bright rows are assigned to 5-fold coordinated Ti4+ ions, and the contrast ensues from high local density of empty states,[26] making them appear higher than neighboring bridging oxygen rows. In aqueous solution, these Ti rows likely become fully hydrated and thus also physically higher.[27] We did not observe the (1 × 2) structure recently reported for the same substrate in pure water,[28] which was ascribed to water structuring at the solid–liquid interface. The large-scale variations in contrast are typically also seen with UHV-STM at the clean surface but have been discussed controversially in the literature.[29,30]

Figure 1

EC-STM images of rutile TiO2(110) in 0.1 M HClO4, (a, b) immediately after approaching the tip indicating a (1 × 1) bulk-truncated surface, (d) after scanning for 30–60 min, and (e, f) after scanning for several hours. (c) Fast Fourier transform (FFT) of the full high-resolution image in part b. STM sample bias (Vbias = Es – Etip) and set point current as indicated; substrate potential Es vs NHE: (a, b, f) +1.19 V; (d) +1.70 V; (e) +1.95 V.

Prolonged (>30 min) EC-STM observation, Figure d–f, reveals the gradual emergence of elongated, bright features on the TiO2 surface. The apparent height of these additional features is on the order of 0.3 nm, which suggests a monolayer species. The image further indicates that the adsorbates arrange in a pattern whose long axis is perpendicular to the [001] direction of the substrate, whereas a closer examination suggests that some substructure with a certain degree of registry with the substrate lattice may exist. Figure e shows that, eventually, uniform submonolayer coverage is obtained.

As EC-STM has no chemical sensitivity, the sample was removed from the electrolyte, rinsed with ultrapure water to remove perchloric acid, and transferred to UHV. Without any further treatment of the sample, atomic resolution of the substrate is again obtained, Figure , and the bright adsorbates are seen to persist on the surface.

Figure 2

(a, b) UHV-STM images of the WO-covered rutile TiO2(110) sample after extraction from the electrolyte. (c) XPS spectrum of the W 4d region. STM sample bias and set point current as indicated.

X-ray photoelectron spectroscopy (XPS) of the W 4d-region, Figure c, shows a clear signature of tungsten (4d3/2 at 260.0 eV; 4d5/2 at 247.5 eV; for more details, see the Supporting Information). The peak positions with respect to metallic W are shifted by 4.5 eV toward higher binding energy, indicating an oxidation state of +VI, as is the case in WO3.[31] The intensity of the W peaks is consistent with submonolayer, but uniform, coverage, as the XPS setup used averages the signal from several mm2 of the sample; highly local WO3 deposition would not yield a similar XPS intensity.

Manipulation of the WO3 oligomers was possible in situ, i.e., in the aqueous solution, with the EC-STM tip, Figure . To this end, a 200 × 200 nm2 section of the image was imaged with 5 times higher tunneling current (0.5 nA instead of 0.1 nA), keeping all other parameters constant. Immediately after scanning this area, the original 300 × 300 nm2 area was imaged using the original tunneling conditions to observe the effect. In the smaller square, the number density of WO3 oligomers decreases to a few percent of the initial coverage following passage of the EC-STM tip, demonstrating STM-assisted nanopatterning of the decorated surface.[19] The tip-assisted nanopatterning, however, was no longer possible after >3 h of contact between the substrate and the bright features (even using higher tunneling currents; see, for instance, Figure e), unless an excursion of the substrate potential into the hydrogen evolution region was performed.

Figure 3

In situ tip-assisted nanopatterning of WO/TiO2(110). The central square area of 200 × 200 nm2 was scanned with Iset = 0.5 nA, followed by zoom-out to 300 × 300 nm2 and imaging with Iset = 0.1 nA. STM sample bias +1.61 V. Left, initial surface; right, nanopatterned central area. Substrate potential vs NHE +1.22 V.

Very similar observations as described so far for rutile TiO2(110) were made for magnetite Fe3O4(001), for which to our knowledge no EC-STM studies exist in the literature. In this case, imaging took place in 0.1 M NaClO4, because the substrate is unstable at lower pH values and becomes etched.[32]Figure a shows that, after prolonged (several hours) EC-STM imaging with a tungsten tip, a high coverage of bright features is obtained. Using high tunneling currents, an image with close to atomic resolution is revealed, Figure b, in which the perpendicular orientation of bright features on neighboring terraces matches UHV observations.[33] The FFT of this image, Figure c, indicates a (1 × 1) periodicity, i.e., that the surface is unreconstructed.[34]

Figure 4

EC-STM images of magnetite Fe3O4(001) in 0.1 M NaClO4, (a) showing high coverage with WO features and (b) after scanning with high (>2 nA) tunneling current, which reveals the cleaned Fe3O4(001) surface. (c) FFT of full image b. STM sample bias and set point current as indicated; substrate potential vs NHE: (a) +0.38 V; (b) +0.57 V.

In order to rationalize the source of the tungsten-containing adsorbates we observe on both rutile and magnetite on imaging with a W tip in aqueous solution, we consider the electrochemical behavior of tungsten metal. Figure a shows cyclic voltammograms (CVs) of a polycrystalline W wire in 0.1 M HClO4, starting at the open-circuit potential (OCP) after equilibration for 600 s. During the first cycle (solid black trace), an anodic oxidation current is obtained in which three peak-like features are discerned. These may correspond to the formation of different tungsten oxides or phases thereof. The almost featureless cathodic scan is followed by a virtually zero-current second cycle (red trace), owing to the blocking of the surface by the anodically generated oxide.[35] Monitoring of OCP following these CVs yields the curve shown in Figure b, suggesting that, after ca. 500 s, an equilibrium or steady state condition is reached. Repetition of the CVs after 500 or more seconds accurately reproduces the outer trace in Figure a, indicating that the initial effect of anodization is largely removed. Progressively shorter waiting times (indicated in Figure b) between repeat CVs yield first cycles that are intermediate between that of the pristine, equilibrated surface and the blocked response, Figure a.

Figure 5

(a) Cyclic voltammogram (two cycles, solid trace) of polycrystalline W wire in 0.1 M HClO4 after equilibration for 600 s. Scan rate, 50 mV s–1. Progressively shorter equilibration times between repeat measurements (from 500 to 100 s as indicated) yield the dashed first cycles. The small cathodic peak at ca. +0.3 V is related to hydrogen intercalation into the tungsten oxide film on the wire.[36] (b) Evolution of the open circuit potential immediately following the first two CVs in panel a. Markers along this trace indicate times when the dashed repeat CVs (panel a) were recorded. (c) Pourbaix diagram of W (calculated for an analytical tungsten concentration in solution of 10–4 M).[35] Dashed parallel lines delimit the thermodynamic stability region of water. The nominal pH–potential region where the W tip is operated during EC-STM is indicated in red (pH 1, TiO2(110)) and blue (pH 7, Fe3O4(001)).

The Pourbaix, i.e., potential–pH, diagram of tungsten and its oxides is shown in Figure c.[35] The stability region of metallic W lies outside that of water—the area delimited by the dashed parallel lines—which means that tungsten is thermodynamically unstable under all experimental conditions in aqueous solution. Anodization promotes the formation of a surface oxide, as clearly follows from the voltammograms in Figure , but also at open circuit potential oxidation is thermodynamically favorable. Similarly, the presence of oxygen in the electrolyte may enhance this process but is not a precondition. Dissolved oxygen generates a mixed potential[37] that will be higher as a function of oxygen concentration and thus promotes oxidation of the metal. However, even if EC-STM is performed in an inert atmosphere (e.g., Ar), which often leads to improved image contrast,[38] the formation of soluble tungsten compounds is not prevented.

In practice, the tip potential was chosen by minimizing the Faradaic leakage current through the tip before approach to the surface. The potential region where the tip was operated is indicated in Figure c, and lies outside the stability region of water. The fact that no substantial hydrogen evolution takes place at the tip indicates a significant overpotential and sluggish electrochemical kinetics, whereas the Pourbaix diagram is limited to thermodynamics.

Discussion

All experimental evidence presented, together with the thermodynamics of tungsten and its oxides in aqueous solution, is consistent with the spontaneous formation of a tungsten oxide at the EC-STM tip, followed by dissolution of the oxide in the electrolyte[39] and adsorption at the oxide substrate. A detailed study combining surface-enhanced Raman scattering and electrochemical impedance measurements concluded that the surface oxide formed on W metal consists of a compact, anhydrous inner layer and an outer, hydrated layer.[40] In acidic solutions, dissolution of the hydrated layer has been shown to be the rate-determining step. Once in solution, the exact tungsten species that prevails is governed by complex solution equilibria,[35,39,41,42] but inside the stability region of water, all of these contain W in the +6 form. On the basis of the Pourbaix diagram and our XPS data, at pH values below 2, the main product formed is tungsten(VI) oxide, WO3, which forms tungstic acid on hydration:[35]A recent study, based on direct high-resolution transmission electron microscopy of crystalline regions in precipitated tungstic acid,[43] identified a corner-sharing WO5(H2O) octahedron as the fundamental building block, condensed into triangular units of formula W3O6(OH)6(H2O)3, Figure (see also the Supporting Information). Simplification of this formula shows its equivalence with H2WO4·H2O in eq . Importantly, the proposed structure is consistent with the existing infrared and Raman studies on tungstic acid solutions and gels,[43] and with in situ Raman studies of anodically oxidized W.[40]

Figure 6

(a) Possible structure of triangular W3O6(OH)6(H2O)3 units (adapted from ref (43); W violet, O red, H white). Neighboring units may oligomerize by splitting off water. (b) Proposed mechanism for the formation of 1D WO3 oligomers on oxide substrates. Triangles represent the (negatively charged) tungsten oxide units, and the TiO2(110) surface is schematically represented by its partially protonated bridging O rows.

In order to elucidate the mechanism of tungsten oxide adsorption, we consider the rutile (110) surface, featuring rows of 5-fold coordinated Ti4+ ions that alternate with rows of bridging oxygens.[25] If water is dosed on this surface, the molecules bind to the initially 5-fold coordinated Ti4+ ions in the surface, thereby resolving their undercoordination.[44] The extent to which the water dissociates on this surface in a vacuum is the subject of controversy; recent UHV results indicate a very slight preference for molecular water and a sizable activation barrier for dissociation.[45] By contrast, in electrolyte solutions, autodissociation of water and extensive hydrogen bonding within the liquid support efficient channels for the redistribution of protons, for instance, through the Grotthuss mechanism.

For the rutile (110) surface, two processes are relevant for its acid–base behavior,[46−49]Figure :We take here the view that, since the coordinated water molecule is a neutral species, its adduct with the surface Ti can be considered neutral too.

Figure 7

Schematic view of rutile (110) exposed to aqueous solution. The charge-determining species are protonated bridging oxygen (Ti2OH+, top left) and terminal hydroxyl (Ti–OH–) groups.

Protonation/deprotonation of bridging oxygens:

Dissociation of the water bound to the 5-fold coordinated Ti4+ ions:

In eqs and 3, Ka1 and Ka2 are the relevant dissociation constants. Since the ratio between the number of coordinated water molecules on the fully hydrated surface and the bridging oxygen atoms is 1:1, internal acid–base equilibration of the surface is possible by combining eqs and 3:Because the overall charge at the surface remains zero during this equilibration, this situation corresponds to the point of zero charge (PZC), at which the surface can be considered in its “zwitterionic state”, by analogy with the acid–base behavior of amino acids close to their isoelectric point. The PZC of oxides as a pH-driven property should not be confused with the potential of zero charge of free-electron metals, which is the unique electric potential value where the immersed electrode carries neither positive nor negative excess charge; the latter is of pivotal significance in explaining electrochemical phenomena, ranging from anion adsorption[50] to self-assembly[51] and nanoparticle charging.[52] The acid–base equilibria of oxide surfaces are decisive for much of their chemical properties,[53] including stability of colloids, and as such also of vast practical importance. As the number density and microscopic environment of the surface hydroxyls differ among crystallographic planes, PZC values are facet-dependent.[54,55]

For rutile (110), the PZC = 5.4[56] is related to the two acid dissociation equilibria and 3 by PZC = (pKa1 + pKa1)/2, and can be determined from electrokinetic measurements and acid–base titrations. The individual protonation constants, however, are not experimentally accessible, and substantial theoretical efforts have been invested to estimate them at pKa1 = −1 up to 5 and pKa2 = 8–9 from first-principles and electric double layer considerations.[46,47,49] On the basis of these equilibrium constants, at pH 1 (0.1 M HClO4), the rutile TiO2(110) is extensively protonated and therefore overall positively charged. With reference to Figure and eqs and 3, this leaves the coordinated water neutral and the protonated bridging oxygens the locus of the positive charge. For the three oxides we consider, WO3 has the most acidic PZC of ∼0.8,[56] which implies that the tungsten(VI) oxide species occur in anionic form at all pH values encountered here.

Combining all data, we propose the electrostatic interaction between these oppositely charged species as the first step in the mechanism for the formation of self-limited linear WO3 adsorbates. Along similar lines, an electrostatic mechanism has been successfully considered for adsorption of small oligopeptides on negatively charged hydroxylated rutile surfaces.[57]

Step 1: electrostatics-driven nucleation of WO3 (negative) on protonated rutile (positive):The elongated bright features seen in STM, Figures d and 2a, of virtually uniform width of 1.3 ± 0.1 nm, are reminiscent of the one-dimensional oligomeric tungsten oxide chains that form on oxidized copper surfaces[58] in a vacuum. In the present case, we propose that, following nucleation of hydrated WO3 adsorbates, 1D growth takes place by adsorption of further HWO4– units followed by condensation.

Step 2: growth of polyanionic adsorbates by on-surface condensation of an integer number m triangular (HWO4–)3 subunits (written out here for one condensation reaction between every pair of triangles):As the oligomerization proceeds, the charge density of the surface decreases because of the changing composition, by one unit charge per added HWO4–, eventually terminating growth (Step 3). This self-limiting, overall electrostatic mechanism explains why no multilayers are formed, and equally applies for magnetite (with PZC 6.8[56]) in near-neutral WO3 solutions.

The fact that the adsorbates are initially easily removed with the STM tip but become more strongly bound over time may indicate the eventual formation of a covalent Ti–O–W bond by condensation:Electrochemical hydrogen evolution at the rutile surface leads to a local increase of pH, which may cause hydrolysis of this bond and, again, increases mobility of the adsorbates.

These principles, summarized in Figure , in view of the ubiquity of oxide hydration and acid–base equilibria in aqueous solution, could be of universal validity, and may find use in preparing thin-layer systems of unlike oxides with atomically defined interfaces and in electrostatic layer-by-layer strategies for the preparation of nanoparticle assemblies.

Finally, the question arises of why “adventitious tungsten” has not been, to the best of our knowledge, reported before, even though tungsten tips have been the most widely used in EC-STM. When considering the EC-STM literature to date, the most studied systems have been the adsorption and self-organization of inorganic anions on the one hand[59] and of organic molecules (tectons) on the other,[60] both on noble metals. Specific adsorption of ions at the metal–electrolyte interface determines much of the behavior of the electrochemical double layer, and has therefore been studied extensively for over a century.[59]

Taken together, systems that expose a surface with pronounced anionic character to the electrolyte represent a clear majority in EC-STM; the opposite is encountered more seldom.[51,61] The emerging field of electrochemical surface science of oxides[28,62] and other highly adsorptive materials such as hexagonal boron nitride,[63] however, may change this ratio very soon. If present, the anionic character of a substrate renders it immune toward adsorption of also negatively charged tungstates, which form under all but the most acidic pH conditions (vide supra), and explains its conspicuous absence in the EC-STM literature. This absence also suggests the cationic character of the substrate as an essential condition for adsorption, and lends further support to the mechanism we propose.

Conclusions

We have demonstrated that the use of tungsten EC-STM tips unavoidably leads to the generation of soluble tungsten oxides. In electrochemical surface science of oxides as an emerging field, and of other highly adsorptive materials, W tips therefore can be used as an “electrochemical evaporator”. Under pH conditions where the oxide substrate under study and the dissolved tungsten oxide carry opposite charges, progressive but self-limiting adsorption of low-dimensional tungsten oxide oligomers can be observed.