Report

High-affinity adsorption leads to molecularly ordered interfaces on TiO2 in air and solution

See allHide authors and affiliations

Science  24 Aug 2018:
Vol. 361, Issue 6404, pp. 786-789
DOI: 10.1126/science.aat6752

A preference for acids

When titanium dioxide surfaces are exposed to water under ambient conditions, an ordered overlayer forms. Balajka et al. studied this process with scanning tunneling microscopy and x-ray photoelectron spectroscopy for water adsorption under vacuum conditions and in air (see the Perspective by Park). The ordered overlayer was only formed in air, the result of the adsorption of organic acids (formic and acetic acids). Although other species such as alcohols were present in much higher concentrations in air, the bidentate adsorption and entropic effects favored acid adsorption.

Science, this issue p. 786; see also p. 753

Abstract

Researchers around the world have observed the formation of molecularly ordered structures of unknown origin on the surface of titanium dioxide (TiO2) photocatalysts exposed to air and solution. Using a combination of atomic-scale microscopy and spectroscopy, we show that TiO2 selectively adsorbs atmospheric carboxylic acids that are typically present in parts-per-billion concentrations while effectively repelling other adsorbates, such as alcohols, that are present in much higher concentrations. The high affinity of the surface for carboxylic acids is attributed to their bidentate binding. These self-assembled monolayers have the unusual property of being both hydrophobic and highly water-soluble, which may contribute to the self-cleaning properties of TiO2. This finding is relevant to TiO2 photocatalysis, because the self-assembled carboxylate monolayers block the undercoordinated surface cation sites typically implicated in photocatalysis.

Metal oxide photocatalysts formed from earth-abundant metals have attracted a great deal of attention, particularly for environmental remediation and photocatalysis (1). These materials are attractive because they are inexpensive, they are typically very stable under reactive conditions (e.g., in air or solution), and their semiconducting nature enables efficient generation of long-lived photocarriers that diffuse to the surface and initiate chemical reactions. Although some of these materials have been commercialized, an atomic-scale understanding of their reactivity in ambient and solution environments has proven elusive, in part because of the diversity of possible active sites on metal oxide surfaces and the complex milieu of reactants in ambient environments. Although scientists have made considerable strides in understanding the atomic-scale structure of model catalysts in vacuum (24), the study of solution/metal oxide interfaces is just emerging (59).

An early indication of complexity at the solution/metal oxide interface was the observation of photoswitching (10). When TiO2 surfaces in ambient environments are irradiated with ultraviolet (UV) light, they become hydrophilic, but they slowly revert to hydrophobicity in the dark. In contrast, clean TiO2 is unaffected by UV irradiation in vacuum. The UV-induced hydrophilicity in ambient environments is now understood to be due to facile oxidation of adsorbates, such as hydrocarbons, to produce, for example, hydroxylated species (11). The reverse dark reaction has never been explained. Although there have been suspicions that the hydrophobic dark state results from adventitious contamination, recent reports of the formation of a highly ordered surface structure with (2 × 1) symmetry on TiO2 rutile (110) in H2O (12), or after exposure to H2O (13) or air (14, 15), have called this into question. On the basis of a combination of scanning tunneling microscopy (STM) and x-ray diffraction (XRD), this structure was recently assigned to a new ordered state of adsorbed H2O—persistent in vacuum at room temperature—in which one H2O molecule adsorbs to every other surface Ti atom (12). Others attributed the (2 × 1) structure to an ordering of H2O at the solid/liquid interface (11), the reaction of CO2 with H2O to produce bicarbonate (13), or the formation of hydroperoxyl from O2 and H2O (14). We show that none of these explanations are correct.

To clarify the configuration and origin of this persistent ordered structure, we constructed a small side chamber (Fig. 1 and fig. S1), which allows the transfer of an ultrapure H2O drop to a precleaned TiO2 surface in vacuum or a controlled gas, followed by characterization in vacuum with no air exposure. This side chamber is connected by an in-vacuum sample transfer system to the surface analysis chamber. The experiments proceeded in four steps (movie S1). First, a small icicle was grown on the cooled finger by vapor transfer from liquid H2O. The side chamber was then evacuated to ~10−7 mbar, and the precleaned sample was transferred from the analysis system to directly beneath the icicle. The side chamber was isolated, and the cold finger was warmed until a drop of liquid H2O fell onto the sample. A gas was optionally added. After exposure, the H2O and any additional gases were evacuated with a LN2-cooled cryopump, and the sample was transferred back for analysis.

Fig. 1 Dosing liquid H2O inside a vacuum chamber.

(A) The finger is cooled with LN2. (B) Ice is grown from H2O vapor, and after evacuation, the sample is introduced. (C) The finger is heated, and liquid H2O falls onto the sample and (D) remains until evacuation.

Figure 2A shows that exposure of a clean TiO2(110) surface to ultrapure H2O in vacuum did not lead to the (2 × 1) structure reported in (12). Instead, the surface was essentially identical to clean, room-temperature TiO2(110) after exposure to H2O gas in vacuum (16), albeit with a low density of contaminants (white blobs in Fig. 2A). In STM, the morphology was dominated by rows of undersaturated Ti atoms, which imaged as protrusions with a 3-Å spacing. Both gas and liquid water exposure hydroxylated any pretreatment-induced O vacancies, which are apparent as protrusions between the Ti rows, but resulted in no stable H2O adsorption.

Fig. 2 Effect of air on the TiO2(110) surface.

STM images of TiO2(110) after contact with an H2O drop (A) in vacuo (+1.87 V, 112 pA) and (B) in air (+1.42 V, 41 pA). Scale bars, 3 nm. (C) DFT simulation of a formate/acetate monolayer. Ti, O, C, and H atoms are shown in blue, red, black, and white, respectively. (D) STM height distribution of TiO2(110) after exposure to a H2O drop in air and fit to equal-width Gaussians (brown and gold).

This assignment was confirmed with x-ray photoemission spectroscopy (XPS) (Fig. 3, A to C). Importantly, a 534-eV O 1s transition was not observed, so no adsorbed H2O remained in vacuum (5). The only difference between the clean and H2O-exposed surfaces was minor contamination, as shown by the aliphatic or “adventitious” C transition at 285 eV. Hydroxyls from H2O dissociation at O vacancies led to the tiny shoulder at 532.5 eV.

Fig. 3 XPS spectra of TiO2(110) after exposure to a H2O drop.

(A to C) In vacuo. (D to F) In air in Vienna, Austria. The lower panels include reference spectra from a 2-min air exposure in Ithaca, NY, and a saturation HCOOH coverage [≡ 1 monolayer (ML)]. Energies of the unobserved transitions from H2O/TiO2 (5) and bicarbonate/TiO2 systems are indicated. The “1 ML” calibration bars represent a HCOOH saturation coverage.

In contrast, exposure of a clean TiO2(110) surface to a H2O drop in air resulted in a (2 × 1) structure, as shown by STM images (Fig. 2B and fig. S2). Chemical analysis of the air-exposed surface (Fig. 3, D to F, and figs. S4 and S5) revealed carboxylic acid adsorption, as evidenced by the C 1s transition at 289.2 eV and the O 1s transition at 532.5 eV. Similar results were obtained when a nominally dry rutile surface was exposed to air for 2 min (Fig. 3). The similarity is not surprising, given that clean TiO2(110) surfaces develop a several-monolayer-thick H2O layer when exposed to air (5). In these experiments, the intensity of the 285-eV transition, which we assigned to aliphatic C in the carboxylic acid and adventitious carbon, varied from run to run.

To confirm that these features are due to carboxylic acids, reference spectra were obtained from a formic acid monolayer deposited in vacuum, as shown in blue in Fig. 3, D to F. This monolayer displayed similar, albeit more intense, transitions in the O 1s and C 1s regions, but no adventitious C. Formic acid deposition self-limits at one monolayer in vacuum (2, 4). To test whether the air exposure is similarly self-limiting, we exposed a clean TiO2(110) surface to air for 30 min. This exposure led to the development of a complete monolayer of carboxylic acid (figs. S3 and S4), as determined by XPS and the production of a well-ordered monolayer with (2 × 1) symmetry.

Carboxylic acid adsorption explains the (2 × 1) surface structure. Carboxylic acids bind to TiO2(110) dissociatively in a bridged, bidentate geometry, forming an adsorbed proton (H+) on a bridging O and an adsorbed carboxylate (RCOO) bound to two adjacent Ti atoms (24), which leads to a (2 × 1) structure.

The previously suggested (13) adsorption of bicarbonate, which has similar binding, can be ruled out from the C 1s spectrum, which responds sensitively to the chemical environment (fig. S6). Carboxylate C is in a +2 oxidation state, whereas bicarbonate C is in a +4 oxidation state. Density functional theory (DFT) calculations predict a 1.32-eV shift in the C 1s core-level energy from this change (Fig. 2B and supplementary materials). Carboxylic acid adsorption is consistent with the 1.95-Å Ti–O distance observed in in situ and ex situ XRD analysis of H2O-exposed TiO2(110) (12), because these species have a 2.08-Å Ti–O distance in DFT simulations.

STM images of the (2 × 1) surface revealed two species with an apparent height difference of 0.72 Å (Fig. 2C and fig. S2), which is in good agreement with the 0.63-Å height difference predicted by simulated STM (fig. S7) of a formate/acetate monolayer. In contrast, the height difference between two longer carboxylates of the form CH3(CH2)n+1COO and CH3(CH2)nCOO (where n = 0, 1, …) would be 0.95 Å. Therefore, Fig. 2 suggests an ~80% formate/20% acetate monolayer.

Polarized infrared spectroscopy of air-exposed surfaces (fig. S8) provided further information. Formate/TiO2(110) is the only carboxylate terminated by a single C–H bond. This structure leads to a characteristic pair of p-polarized C–H stretch resonances (supplementary materials), which were observed on both air-exposed surfaces and formate monolayers. This suggested that formate, the shortest carboxylate, is the shorter species in the STM images, and acetate is the taller one.

To test whether the adsorbed carboxylates were formed by the reaction of gas-phase molecules with H2O, we coexposed clean rutile surfaces to liquid H2O and a variety of atmospheric gases. No reaction was detected with CO2 (0.4 mbar in air) or CO (10−4 mbar in air) (figs. S9 and S10).

To show that the adsorbed carboxylic acids arise from a ubiquitous component of air and not specific contaminants of our air, we obtained XPS spectra and STM images of TiO2(110) exposed to air in rural Ithaca, NY, USA, ~7000 km from the measurements taken in urban Vienna, Austria. These data (figs. S2 and S4) display the same chemical signatures, the same height difference, and a similar height distribution.

The formation of two such similar, highly ordered monolayers in ambient environments on different continents—as well as previous observations of (2 × 1) structures in England (12), Italy (11), the United States (13), and Japan (14)—can only be explained by ubiquitous environmental species with high-affinity binding. We address these requirements sequentially.

Formic and acetic acids are the dominant sources of atmospheric acidity in most parts of the world, with longer-chain acids having much lower concentrations (17). Measurements in the United States and Germany, for example, show typical partial pressures of 10−6 to 10−7 mbar for each species (16). These acids have many origins; however, biogenic sources, particularly the oxidation of isoprene from trees and shrubs, are thought to dominate (16). Both acids are rapidly and efficiently adsorbed by liquid H2O, leading to calculated concentrations of each acid of ~10 μM in ambient environments (supplementary materials). This facile trapping is also expected in the several-monolayer-thick H2O layer (5) that forms on air-exposed TiO2(110).

The hydrolysis of atmospheric CO2 and CO leads to much lower concentrations of H2CO3 and HCOOH. Thermodynamic calculations (supplementary materials) show that air-equilibrated H2O contains 10−8 M H2CO3 and 10−7 M HCOOH—orders of magnitude less than from atmospheric formic and acetic acids.

Having established that formic and acetic acids are ubiquitous in air, why do these relatively minor species—as opposed to other small molecules, such as alcohols—form monolayers on TiO2(110)? The partial pressure of atmospheric methanol (18), for example, is 10−2 to 10−3 mbar. Although many small molecules adsorb to TiO2, species with monodentate binding [e.g., alcohols (19) and amines (20)] typically desorb near room temperature, whereas carboxylic acids, which form two bonds to the substrate, are stable until much higher temperatures (21). Thus, although some alcohols and other monodentate species are present in considerably higher concentrations in air, desorption of these species will be facile.

Our finding that TiO2(110) surfaces in ambient environments are terminated by a well-ordered carboxylate monolayer (and not a persistent H2O monolayer or adventitious molecules) explains a long-standing puzzle: the spontaneous transformation of initially hydrophilic TiO2 surfaces into hydrophobic surfaces in the dark. We attribute the dark state to the spontaneous formation of mixed formate/acetate monolayers by adsorption from the atmosphere, given that formic acid solutions also produce hydrophobic, self-assembled monolayers on TiO2(110) (supplementary materials).

This finding provides further insight into the well-known self-cleaning properties of TiO2. Self-assembled carboxylate monolayers have the unusual property of being both hydrophobic (from their CHx-terminated “tails”) and highly water-soluble (from their acidic “heads”). In their hydrophobic state, the carboxylate monolayers will resist contaminant adsorption. However, their high water solubility will cause the surface to become hydrophilic under rinsing, enabling the water sheeting action that is important for self-cleaning.

Perhaps most unexpectedly, this finding suggests that surfaces in ambient environments may be more structured and controlled than previously realized. In this study, a clean metal oxide surface selectively adsorbed species present in parts-per-billion concentrations to form a molecularly ordered interface, while effectively repelling other adsorbates present in much higher concentrations. This finding may have important implications for photocatalysis, given that self-assembled carboxylate monolayers effectively block the undercoordinated cation sites typically implicated in photocatalysis.

Supplementary Materials

www.sciencemag.org/content/361/6404/786/suppl/DC1

Materials and Methods

Figs. S1 to S10

References (2229)

Movie S1

References and Notes

Acknowledgments: R. Gärtner and H. Schmid are thanked for technical assistance. Funding: This work was supported by the Austrian Science Fund FWF (Wittgenstein Prize project Z-250-N27, Doctoral College Solids4Fun project W1243-N16, and project F45), the European Research Council advanced grant Oxide Surfaces (ERC-2011-ADG-20110209), and the National Science Foundation (CHE-1708025). This work used the National Energy Research Scientific Computing Center (Department of Energy, DE-AC02-05CH11231). Author contributions: J.B., J.P., M.K., M.S., and U.D. designed the apparatus and conceived of the experiment. J.B. and M.A.H. performed Vienna experiments, W.J.I.D. performed Ithaca experiments, and M.A.H. performed calculations. M.A.H. wrote the manuscript with input from all. Competing interests: The authors declare no competing interests. Data and materials availability: All data are available in the manuscript or the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article