1 Introduction

The efficient, selective and sustainable activation of chemical bonds within (organic) molecules is one of the central challenges to optimize (photo-)chemical conversions in the chemical industry. Due to their high abundancy and involvement in many (bio)chemical processes, alcohols are important compounds, and thus it is essential to understand their catalytic reactions. Here, we mainly focus on methanol, which is not only the smallest, but also the technologically most important alcohol. It is frequently discussed as a potential hydrogen carrier that can be sourced from biomass to replace several compounds in the current fossil fuel-based energy technology [1,2,3,4]. Furthermore, it may be potentially used for the formation of higher hydrocarbons as basic chemicals in industrial production. Despite its relatively small molecular size, methanol already contains three different types of bonds (C–O, C–H and O–H) that need to be selectively activated in order to steer product distributions. Understanding the direct connection between selective bond activation by control of the molecular environment on a catalyst surface and the product formation is the scope of this perspective article.

With respect to sustainability, it is fortunate to utilize well-abundant, non-toxic catalyst materials. Furthermore, the ability to use energy from solar irradiation is an important aspect in current research. Transition metal oxides such as ceria [5,6,7,8,9,10,11], zinc oxide [12,13,14, 15], vanadia [10, 16,17,18,19,20,21,22,23,24], iron oxides [25,26,27,28,29], tungsten oxides [30,31,32,33,34], and many others are promising candidates for such purely oxidic, metal-free catalysts, though all types of considered materials have not only advantages, but also drawbacks. The by far most studied representative of this class is titanium dioxide or titania (TiO2) [35,36,37,38,39]. Anatase and rutile are the relevant modifications, which are often used as mixtures of both such as within the well-known P25 powder. However, micro- and nanostructured materials consist of manifold surface terminations, which unfortunately impedes to obtain a systematic understanding of elementary reaction steps under technical conditions. Herein, the (110) surface of the rutile modification is the most relevant surface termination due to its outstanding thermodynamic stability. Therefore, a detailed understanding of ongoing elementary steps on a molecular level can be derived from the so-called surface-science approach, in which catalytic model reactions are investigated under ultra-high vacuum (UHV) conditions on well-defined single-crystal surfaces. While this approach yields unprecedented details of the molecular activation and reaction steps, one has to be aware of possible limitations when knowledge is transferred to technological applications, which shall be one part of the following discussion.

Being the most stable titania surface, the stoichiometric rutile TiO2 (110) surface consists of alternating rows of in-plane oxygen atoms (Opl) and titanium sites [35]. Every second titanium row along the [001] direction is covered by bridging oxygen atoms (Obr). Therefore, this surface is intrinsically bifunctional in terms of its microscopic chemical behavior. The relevant Lewis acidic sites are undercoordinated (five-fold coordinated) Ti5c, to which donor molecules can bind via their lone pair in an acid base-like motif. Oxygen sites (Obr and Opl) acting as the relevant base sites are located in next neighborhood to these acidic sites. For rutile TiO2, an overwhelming amount of research concerning various aspects such as structural and electronic properties, the deposition of metals, organic or inorganic adsorbates and their conversion in thermal, photochemical, or electron-induced reactions is available. This includes experimental and theoretical work, often in combined efforts. This perspective article mainly focuses on experimental aspects of steering molecular bond activation on rutile TiO2. For a more general and comprehensive overview, the reader is encouraged to consider the excellent review articles published by Diebold [35], Linsebiegler et al. [38], Thornton et al. [36, 37], Schneider et al. [39] and Henderson [40].

Over the last couple of years, more and more studies on reduceable oxides gave evidence, that point defects are the dominant game-changers in numerous chemical reactions. Such defects arise from partial reduction by oxygen-loss. For rutile TiO2 single crystals, bridging oxygen vacancies (Ovac) are the relevant surface defects, while Ti3+ interstitials are the dominant bulk defects [42,43,44,44]. It is important to know that both types of defects are carriers of electronic charge. Thus, particularly the Ti3+ interstitials, which become mobile at elevated temperatures, appear as shuttles of electronic charge involved in adsorption and reactions at the surface [45, 46]. They are able to dilute into the bulk as a reservoir, or, depending on the detailed conditions, diffuse towards the surface to participate in chemical reactions at elevated temperatures [44, 47]. Another important implication is, that the additional charge is at least partially located in 3d-orbitals of the titanium atoms which induces a populated state in the band gap around 1 eV below the Fermi level, and thus significantly affects the electronic structure [43, 48]. Point defects and especially Ti3+ sites were initially often ignored for reactions on titania surfaces. In contrast, recent studies gave evidence, that such sites govern for example the chemisorption and dissociation of molecular oxygen on rutile TiO2 (110) [46, 49, 50] or McMurry-type aldehyde coupling [41, 45, 51]. In this work, the versatile role of point defects in the (photo-)chemical conversion of small alcohols is added to this list [53,54,55,56,56].

Starting from selected aspects of the intrinsic surface termination, a detailed analysis of the impact of point defects in activation of different chemical bonds shall be discussed. Also, the TiO2 surface may be modified by the deposition of other oxidic structures such as, e.g., oxidic clusters to tune the atomistic properties in hybrid catalysts. Together these findings reveal, that a careful implementation of reduced titania sites and other point defects in combination with tailored reaction conditions and coadsorbates allows to steer the overall reactivity as well as the population of thermal and photochemical reaction pathways which eventually is manifested in the product distribution. The deduced concepts are further extended to hybrid oxidic catalytic systems, which serve as a tool to realize e.g. stronger acid and base strength.

2 Chemistry of Alcohols on Rutile TiO2 (110): Adsorption and Conversion

2.1 Methanol as a Versatile Probe Molecule in (Photo-)Catalysis

Being the simplest alcohol, methanol is an essential compound in the chemical industry with an annual production of more than 100 million tons [57]. It is classically derived from syngas, but can also be formed from more sustainable sources such as CO2 hydrogenation or from biomass [3]. Indeed, Nobel laureate George Olah and others suggested a renewable methanol-based chemistry as a potential replacement for the current, fossil fuel based energy technology [1]. In this view, chemical tools to convert methanol and alcohols in general to different products are essential for future technology. This becomes more important as possible products of methanol conversion such as formaldehyde, ethene, or dimethyl ether enable a rich secondary chemistry based on methanol from green sources [2]. Especially reductive C–C coupling routes (forming ethene) are of great interest because they enable the implementation of green methanol as a sustainable feedstock in future production schemes to synthesize larger carbon structures by exploiting already established petrochemical processes [56].

For methanol, reductive and oxidative reactions have to be considered (Fig. 1), which may be triggered or enhanced in the presence of defects, UV-irradiation or coadsorbed oxygen at the surface. It is important to note, that all conversion routes imply the dissociation of the O–H bond. As a consequence, we shall discuss the thus formed methoxy-species as the central surface intermediate at a later point. Once this species is formed, it can undergo C–O dissociation leading to reductive pathways (top part in Fig. 1); or, in case of oxidative pathways, a C–H activation takes places (bottom part in Fig. 1). In some cases, the preformed intermediates may diffuse and eventually undergo coupling reactions. Since the desorption of unconverted methanol competes with all of these reactions, the irreversible activation of distinct bonds by the molecular environment is crucial for high overall conversion and selectivity. Moreover, it is important to distinguish the thermal chemistry from photochemical reactions. This can be complex, since thermal effects most often overlay and compete with photochemical reaction steps. In many cases, only the right interplay of thermal and photochemical steps eventually leads to the desired products.

Fig. 1
figure 1

Overview on selected reaction pathways of methanol. Note, that some reaction pathways require the presence of defects, UV-irradiation or oxygen coadsorption or are enhanced thereby. With respect to the carbon center, reductive reactions are listed at the top, while oxidative reactions are given towards the bottom. For all products, the O–H dissociation (blue, [O–H]) is a central step. Reductive reactions can be classified by activation of the C–O bond (green, [C–O]]), while oxidative reactions imply activation of C–H bonds (red, [C–H]).

Full size image

2.2 Chemistry of Methanol on Rutile TiO2: Desorption Features and Product Formation

A large toolbox of experimental approaches and methods is nowadays applied to elucidate chemical reactions at surfaces of all kinds. One important tool to study the formation and desorption of products from heterogeneous catalysts is temperature-programmed reaction spectroscopy (TPRS). In this technique, the reactants are adsorbed in controlled amounts on well-defined single-crystal surfaces prepared under UHV conditions, which are then subsequently heated up in front of a mass spectrometer. In the last few decades, this technique was successfully applied also to some powdered materials under on-stream conditions [58]. In both cases, the desorption of products is continuously monitored by selected m/z traces, which allows the assignment to different compounds based on the fragmentation pattern. In case of simple unconverted desorption, the technique is synonymously called temperature-programmed desorption (TPD) or thermal desorption spectroscopy (TDS). TPRS is not capable of monitoring continuous conversions, but rather gives combined insights about the desorption of adsorbed educts and products, which implies relevant information on the bonding situation. Moreover, TPRS can gain information on the onset temperatures of product formation from chemisorbed precursors. This implies, that assuming a reaction order of one, no mass transport limitations are expected. That changes for coupling processes, because the coverage or intermediate density usually decreases with increasing temperature, meaning that the observation of coupling reactions (at higher temperatures) may be less relevant than it would be under continuous on-stream reaction conditions. Moreover, unconverted desorption always competes with the formation of products. Therefore, apparently low conversion yields may be observed if the adsorbed educts do not strongly bound to the surface or exhibit desorption energies below the onset temperature of the conversion, respectively. Similarly, TPRS can supply information about a possible catalyst deactivation. One example would be poisoning of the catalysts by strongly bonded adsorbates at (lower) reaction temperatures, that could be detected by their desorption or decomposition at a later point of the TPRS run. However, it is important to note that such single-run approaches are very limited to follow the long-term stability of catalysts at work. Therefore, TPRS investigations can supply only one puzzle piece of the full picture of a chemical conversion in an industrially relevant process, but, especially if such studies are backboned by elemental or vibrational spectroscopy and microscopic insights, a highly systematic level of atomistic understanding can be reached from the surface science approach.

For methanol on TiO2 (110), Henderson pioneered such studies showing that after adsorption on a comparably stoichiometric rutile sample, three desorption features of unconverted methanol can be detected in TPRS [59]. These are a physisorbed multilayer; a bilayer, that is weakly adsorbed via hydrogen bonds towards bridging oxygen sites; and finally desorption from the main adsorption site, the fivefold coordinated titanium sites Ti5c. Consequently, these features can be found in every TPRS study with slightly differing values due to experimental issues (including the temperature slope, coverage, coadsorbates, experimental setups and measurement accuracy). However, further reaction products may be observed which turned out to depend on the amount of Ti3+ defects. Figure 2 exemplarily shows selected TPRS traces for methanol conversion on preoxidized rutile surfaces under different conditions. For the detailed experimental conditions and additional information, the authors refer to the original publications from which the data were extracted [55, 56]. Two reaction channels are important, namely the deoxygenation towards methane and the partial oxidation towards formaldehyde. Relevant products are labeled in bold letters, smaller product formations are denoted in grey.

Fig. 2
figure 2

Temperature-programmed reaction spectra (2 K/s) recorded after adsorption of 75 L O2 and a saturation layer of methanol at 110 K. In the bottom case, the adsorbates on the sample were irradiated with UV-Light (365 nm, > 8 mW/cm2) at 110 K for 30 min before TPRS. The data were extracted from own, earlier publications and are reproduced with permission from the Phys. Chem. Chem. Phys. owner societies [55, 56]

Full size image

While there is only a small thermal conversion to both products on TiO2 with a small Ti3+ content (top in Fig. 2), a significantly increased formation of methane in the high-temperature reaction (m/z=15) and formaldehyde (m/z =29, but not 31) in low- and high-temperature channels is observed for higher Ti3+ concentrations. Finally, the photochemical reaction with UV-light promotes the low-temperature partial oxidation and almost fully suppresses the deoxygenation.

At this point we focus on more general aspects, and shall discuss the details of how to control the reactivity and selectivity by the atomistic environment at a later point of this work. Figure 3 gives an overview about selected products including an estimation of the product formation or desorption energy (whatever is higher) based on Redheads equation [52, 54,55,56, 59,60,61,62].

Fig. 3
figure 3

Temperature-programmed reaction spectroscopy (TPRS) features for unconverted (blue) desorption of methanol and desorption of the two dominating products methane (green, deoxygenation) and formaldehyde (red, partial oxidation) from rutile TiO2 (110). Note, that the product distribution is dependent on the reaction conditions including the defect density and coadsorbates. The top scale shows the desorption energy according to Redheads equation, using a pre-exponential factor derived from transition state theory (see supporting information for details). All values are listed in table S1 in the supporting information.

Full size image

Especially the dominating high-temperature reaction channels above 500 K obviously proceed via an intermediate with a high thermal stability. This species was identified to be the methoxy intermediate using FT-IRRAS, TPRS, STM and other techniques [52, 59, 63,64,65]. In view of the drastically higher dissociation energies of all bonds in the free methanol molecule (approx.: \({\Delta H}_{C-O}\) = 371 kJ mol−1; \({\Delta H}_{C-H}\) = 400 kJ mol−1; \({\Delta H}_{O-H}\) = 412 kJ mol−1, see Table S3 in the SI for further details), the vital role of the titania surface in lowering activation barriers of these reactions becomes clear. Thus, it is not surprising that the overall interaction of methanol with the TiO2 surface is comparably strong and can even trigger lifting of the surface relaxation in some molecular arrangements [66]. More important, the O–H dissociation occurs much easier on titania than in the gas phase, and methoxy- and methanol adsorbates are similar in energy [66,67,68,69,70]. Vice versa, the recombination of methoxy and surface hydrogen (Obr–H or O(ads)–H) exhibits extremely low barriers. Indeed, based on own FT-IRRAS results being consistent with a STM-based study by Dohnálek et al., molecular methanol and methoxy-species were shown to coexist for all coverages below 320 K [52, 71]. This behavior was also theoretically predicted in DFT and MD based studies [67,68,69,70]. Surprisingly, combined MD-DFT simulations revealed that activation of C–O and O–H bonds on the stoichiometric rutile (110) surface are similar in reaction energy, but C–O cleavage is inhibited due to an extremely large activation barrier [72]. At a later point of this work, Ti3+ defects shall be shown to be useful to overcome this problem.

We want to point out that these findings are not only limited to methanol, but are widely transferable to predict the chemistry of other small oxygenates on titania surfaces. This especially becomes evident, if one compares typical product desorption features from other alcohols, polyols or other C1-compounds such as formaldehyde. Also, virtually all sorts of alcohols including primary, secondary, tertiary alcohols and polyols have been studied for their adsorption and thermal conversion on stoichiometric TiO2. Given their similar chemical behavior, systematic trends e.g. by the chain length and inductive (+I) effects affecting the acidity or the stability of possible hydrocarbon fragments have been intensively studied predominantly by the groups of Dohnálek, [73,74,75,76,77,78,79,80,81] Henderson[59, 60, 82] and Madix [83, 84]. Figure 4 presents a selection of literature data, showing that a very similar pattern of oxidative and reductive reactions can be found for such compounds [53, 77, 80, 83, 85,86,87,88,89,90]. Therefore, we want to emphasize the universality of this work for the chemistry of oxygenates on titania, which appears in well agreement with other publications [74, 81, 83, 91].

Fig. 4
figure 4

Selected overview on TPRS features for other alcohols, polyols and formaldehyde. Multi- and bilayer desorption features are not listed for clarity. Note the similar pattern as for methanol (Fig. 3). A list of all used desorption features is included in the supporting information (table S2).

Full size image

3 Synergism of Acidic and Basic Sites at Oxidic Surfaces

The previous section conveys an important message: The formation of all products and therefore the overall methanol conversion is directly dependent on the activation of the O–H bond, forming the methoxy species. Assuming a full catalytic cycle, this also implies the need for hydrogen removal from the surface. The intrinsic bifunctionality of the rutile TiO2 (110) surface as a regular ordered array of Lewis acidic (Ti5c) and basic (Obr and Opl) sites is one key for this process. While methanol and other alcohols adsorb on the Ti5c sites via their oxygen lone pair in an acid–base-like fashion, the hydrogen can be easily abstracted (3,4 in scheme 1) by oxygen sites, and equally important, transferred along the oxygen rows. Indeed, the high mobility of hydrogen on rows of Obr was shown in a number of STM studies for O–H cleavage of adsorbed alcohols or water [71, 82, 92]. Therefore, the rows of Obr can be seen as a hydrogen shuttle. This enables an efficient hydrogen removal by formation of water (5 and 6 in scheme 1) from two hydroxyl groups and subsequent desorption at the typical water desorption temperatures (approx. 180–190 K and 260 − 290 K) depending on the overall surface coverage. Scheme 1 summarizes the main steps for formation of the methoxy intermediate and hydrogen removal when coadsorbing oxygen and methanol on a defect rich surface [50, 52, 83, 93].

Scheme 1
scheme 1

Reactions steps involved in the removal of hydrogen by formation of water followed by subsequent desorption. For simplicity, the abstraction of hydrogen by molecular oxygen or peroxo-species instead of atomic O(ads) is omitted. __Ti5c denotes a free adsorption site at a fivefold-coordinated titanium atom.

Full size image

Since a number of recombinative processes is involved to form and remove water from the surface, not only the formation of hydroxyls, but especially their mobility and ability to meet at the surface appear vital for high alcohol conversion yields. Although the main shuttling occurs along the bridging oxygen (Obr) rows, it is important that in-plane oxygen sites (Opl) play an important role in the actual atomic transfer processes enabling this high mobility as suggested by DFT [69, 92]. In other words, without this delicately ordered surface array forming hydroxyl (or hydrogen, respectively) shuttles, the initial activation of alcohols would be reversible and thus far less efficient. Similar findings may be observed for significant blocking of the rows by, for instance, oxidic clusters or metal particles in hybrid catalysts, if they cannot transfer hydrogen as easily. Moreover, it was found that high coverages of methanol (bilayers) on stochiometric rutile prevent itself from effective O–H dissociation by complete blocking of Obr and Ti5c rows, while the actual barrier for a single hydrogen transfer event was not drastically higher [67, 94]. From another point of view, this also identifies water as a possible catalyst poison by generation of surface hydroxyls and therefore blocking effective O–H cleavage of methanol or other adsorbates, which is in accordance with available publications [53, 59, 63, 85, 89, 90, 95]. Technically, such issues may be easily solved by a delicate reaction temperature control.

More general, the chemistry of water on rutile surfaces is similar to the one of small alcohols, which is not surprising given the similarity of HO–H and HO–R (R = Me, Et, …). Of course, the Brønsted acidity of water is much higher than of alcohols due to the + I effect of alkyl moieties, and thus, the O–H bond can be dissociated more easily in comparison. This is especially true for reduced surfaces, at which water dissociates already at temperatures as low as 150 K [96]. If water and alcohols are both present at the surface, this implies, that the rate-limiting steps are expected to be located at the alcohol if the coverage is low enough to avoid site blocking and intermolecular effects. However, at least for stoichiometric, defect-free surfaces, there is no consensus if water dissociates upon adsorption or not, or if mixed phases are dominating the surface [94, 97,98,99]. Accordingly, dissociated and molecular water are similar in energy and the activation barriers for dissociation and recombination are very low [72, 93, 99]. Similar to methanol, high coverages likely suppress the dissociation of water [97, 100]. Again, this meets the observation of significant amounts of water acting as a possible catalyst poison for all reactions involving alkoxy-species. Due to the higher molecular symmetry and the double amount of O–H bonds per molecule, hydrogen bonding and intermolecular interactions, adsorbate networks and the detailed geometric orientations often play a bigger role for water than for alcohols [101,102,103]. For example, strict adsorption in Ti–O–H–Obr motifs was found for a monolayer of undissociated water on stoichiometric surfaces, leading to a 1 × 1 overlayer in that case [98, 104]. However, the adsorbate–surface interaction may be weakened for higher coverages such as bilayers or liquid water due to formation of extended hydrogen bonding networks with other water molecules [101]. Eventually, the chemistry of water on titania is a widely studied topic with a broad range of available publications from studies under classical UHV conditions over near-ambient pressures up to the effect of liquid water, but depending on accuracy and methods, different results are gained [97, 98, 101, 104, 105]. Certainly, the behavior of water is closely related to alcohol conversion, but at this point this work shall focus on chemical bond activation of organic molecules and especially alcoholes within different conversion pathways. Thus, for more information on water chemistry on titania surfaces and implications on (photo-)catalysis, the authors refer to available review articles and other publications [36, 37, 40, 99, 106, 107].

Mobility along the surface with low or without energetic barriers is also crucial for coupling reactions, that allow to build up larger organic molecules from methoxy- or methyl fragments. Especially the formation of ethene from two equivalents of methanol, which is discussed in more detail at a later point, is an important example for such processes [54]. Similarly, in a DFT-based study, Lang et al. highlighted the high mobility of η2-bonded formaldehyde (dioxomethylene-like) along Ti5c rows as essential for the reaction of a methoxy and a dioxomethylene-like adsorbate, eventually resulting in the formation of methylformate (oxidative coupling) [108].

Furthermore, not only the arrangement, but naturally the strength of basic sites is another important point. Since basic sites are involved in both, O–H cleavage as well as C–H activation, not only the overall conversion but also the reaction selectivity can be critically tuned by the strength of these basic sites. While the basicity of regular oxygen sites on the rutile (110) surface appears to be relatively moderate, and hence, only little methanol is converted, oxygen adatoms O(ads) or even oxygen radicals involved in photochemical reactions exhibit a stronger basicity, leading to the pronunciation of oxidative pathways. Also charge from point defects can have the same effect and increase the conversion. Indeed, Diebold et al. recently presented the reduction of O2 by charge transfer from defects to the oxygen molecule in a combined STM and TDS-based study at the example of anatase (101) [49]. This allows the formation of electron-rich oxygen adsorbates (Oadδ− or O2δ−) that act as strong bases. Finally, introduction of oxidic structures such as tungsten oxide clusters leads to hybrid catalysts, that exhibit a stronger basicity and therefore can be efficiently used in hydrogen abstraction reactions [55, 109].

Hence, strong basicity is manifested in the product distributions via the high hydrogen abstraction capability. For methanol, this affects not only the overall conversion via the O–H dissociation, but also the product distribution by promoting C–H cleavage, which leads to an increased partial oxidation yield. Therefore, the formaldehyde yield directly correlates with the amount of strongly basic sites. After C–H activation, formaldehyde is not directly present at the surface, but is formed in-situ during desorption from a dioxomethylene-like precursor as we have identified by FT-IRRAS in agreement with pure adsorption of formaldehyde studied by the groups of Wöll and Dohnálek [52, 86]. The critical role of the interplay of C–H and O–H activation to steer reaction pathways as well as the increased basicity by defects is also evident from recent theoretical studies [108]. We shall show at a later point that activation of C–O (reductive pathways) and C–H bonds (oxidative pathways) compete, and illuminate a general reaction roadmap for the dominant products from methanol conversion.

Besides their hydrogen abstraction capability, oxygen sites at the titania surface appear as a relevant factor to steer the orientation of chemisorbates [53, 70, 94, 110, 111]. While the Ti–E (E = O, N, P, S …) bond via the lone pair of E surely makes the largest contribution to the adsorption energy, numerous weaker interactions of Opl and Obr sites with polarized H–C or H–O moieties can actually steer the final adsorption geometry at the surface. In turn, these may have the potential to steer e.g. the (stereo-)selectivity in conversion of complex molecules in future applications. This fact shall be exemplified, starting with the coverage-dependent adsorption of isopropanol and the multidentate adsorption of methylamine [53, 110]. In both cases, the adsorbates rotate into parallel alignment with the Obr rows if enough space is available to achieve the most stable adsorption geometry. This is not only due to the sterical demand of terminal moieties, but also caused by dispersive interactions with neighboring oxygen sites. Therefore, the oxygen atoms surrounding the Ti5c site can be seen as ligands surrounding an undercoordinated metal site similar to common homogeneous catalysts. In the third case study, which is the adsorption of acetone, the coordination to oxygen sites has an even more drastic effect: On defect-rich TiO2 with coadsorbed oxygen, an oxygen-induced tilt of the chemisorbed acetone molecules towards the oxygen rows was identified as the first step towards total oxidation, i.e. eventually yields carbon dioxide and water (combustion of acetone) [112]. Though, in addition to the undercoordinated metal sites, that are most often in the main focus, oxygen sites are equally important contributors on the intrinsically bifunctional rutile TiO2 (110) surface.

4 The Versatile Role of (Sub-)Surface Defects

While stoichiometric rutile TiO2 does not efficiently convert methanol, (sub-)surface defects such as Ti3+ interstitials and oxygen vacancies (Ovac) trigger a high conversion. This arises from an increased oxophilicity and effects on the charge distribution within the semiconductor material. In this regard, especially Ti3+ sites are important, which can diffuse towards the surface at elevated temperatures, implying the crystal volume as a defect reservoir. In the following, the role of (sub-)surface defects on the adsorption and conversion of small oxygenates shall be illuminated from two different viewpoints. It is important to note, that in a full reaction, typically more than one of the following aspects applies and thus a multiple, non-linear promotion of reaction paths is observed.

4.1 Abstraction of Oxygen

Reduced titania naturally exhibits an increased oxophilicity, rendering it an ideal material for the abstraction of oxygen from almost any molecule. This is due to the high energetic benefit of forming Ti–O bonds (\({\Delta H}_{B}\) = 672 kJ mol−1). In detail, the reduced titanium sites react with the abstracted oxygen forming first substoichiometric TiOx clusters, which can eventually end up in the formation of extended TiO2 islands on top of the initial surface [42,43,44, 113, 114]. Indeed, highly reduced surfaces (Ti2+, Ti3+) prepared by ion sputtering are extremely reactive to methanol. A virtually quantitative conversion towards methane and lower amounts of reductive coupling products, here ethene, was observed. This reaction appears non-sensitive to the educt, since acetone is equally converted virtually quantitatively into propene and propane as well as the corresponding C–C coupling products [54]. Unfortunately, such highly reduced surfaces are quickly deactivated due to reoxidation and reordering of the (110) surface [54]. Therefore, it would be necessary to either recover the active sites e.g. by electroreduction of thin films or batch-wise reaction with intermediate reduction periods (catalyst reactivation). However, it provides evidence to the high potential of reduced Ti sites in oxygen abstraction [46, 50].

Another example underlining the oxygen abstraction potential of reduced Ti sites is the photochemical formation of ethene from two equivalents of methanol, which is exclusively present for high bulk defect densities [56]. This reaction likely proceeds via coupling of two dioxomethylene species at the surface, as adsorption of pure formaldehyde on highly reduced TiO2 also leads to ethene formation for which the dioxomethylene species has been identified as a relevant intermediate [13, 86, 87]. Here, this road is taken from methanol as a more sustainable starting material. This involves formation of a sufficient coverage of dioxomethylene-adsorbates, which are formed from methoxy groups after (photochemical) hydrogen abstraction from a C–H bond by oxygen sites. Similarly, the defect-mediated thermal coupling of benzaldehyde towards stilbene may serve as another example for such McMurry-type reactions [45, 51, 115]. For reductive coupling reactions, the oxygen abstraction is vital to avoid oxidative coupling [65, 116]. In turn, both compounds (here ethene and methyl formate) may be desired products depending on the context, so the oxygen abstraction potential is a powerful switch to steer the coupling product distribution. Anyway, coupling occurs on a relatively low rate and is not dominating the desorbing products in TPRS experiments, most likely due to the comparably low adsorbate coverage. As discussed above, higher coupling yields may be achieved by higher coverages of intermediates under technical reaction conditions at elevated pressures and extended irradiation times or intensities.

This knowledge has a direct impact on the choice of oxidic catalysts: Since Ti3+ defects diffuse towards the surface more likely for rutile than for anatase (in which such defects are more stable in the bulk) [117], reductive reactions are expected to be more pronounced for rutile titania in comparison. Indeed, the formation and accumulation of methoxy-species on anatase (101) single crystals and nanoparticles was recently found by Diebold et al. [65] and Dohnálek et al. [118] who did not report a significant methane formation in their TPRS experiments. The eventual consequence of this effect may be a catalyst poisoning with alkoxy sites, if they are not converted in another reaction pathway, such as photo-oxidations. Nevertheless, this allows to realize surface coverages of the methoxy intermediate that are high enough to promote coupling reactions, e.g., the formation of methyl formate [65]. However, in lack of oxygen abstraction capability, this unfortunately is not an option to guide towards increased reductive coupling to ethene on TiO2 under UV irradiation, but may be valuable to promote oxidative coupling.

Another important consequence may be that small rutile titania crystals of a few nanometers are not as appropriate for application in reductive reactions because due to their low bulk volume they cannot supply a large amount of (bulk) Ti3+ sites per surface adsorption site.

4.2 Charge Transfer and Bond Polarization

While oxygen sites can be interpreted as hydrogen- and therefore proton (or positive charge) shuttles, Ti3+ defects can be seen as negative charge shuttles. Hence they can alter the electronic structure of the surface due to the population of Ti3d-orbitals just below the Fermi-level. The partially delocalized charge becomes evident for example in blurry features in STM measurements above subsurface Ti3+ sites [119]. Adsorbates and other structures at the surface interact with that charge density, which can significantly affect adsorbate stability, the molecular charge distribution or the polarization of chemical bonds or moieties [46, 47, 120, 121]. For example, Chen et al. recently predicted by DFT that polarons on the titania surface may promote or inhibit the O–H dissociation of water and methanol (Janus-like character) by electronic interactions of polaron and adsorbate, depending on the detailed adsorption site and geometry [111]. Furthermore, enhanced hydrogen bonding in bilayers may be triggered by strong induced dipoles due to interaction between adsorbates and polarons [111]. Since organic molecules typically adsorb in an acid–base-like motif, such charge can also change the desorption energy due to an altered acidity of the Ti5c sites [120]. Moreover, as illustrated in the previous chapter, the charge from defects increases the strength of basic sites. While the oxophilicity of Ti3+ drives C–O cleavage (reductive) steps, C–H abstractions (oxidative reactions) are promoted by the higher proton affinity which is caused by the electronic charge supplied by defects. Also, DFT suggests an electronic interaction of Ti3+ sites in the bulk and molecules such as water or methanol at the surface that is strong enough to eventually attract Ti3+ if the mobility is high enough [47]. It again shows how versatile the role of such defects can be, which is followed up in the last chapter.

At this point, we want to bring up three examples to showcase the role of defect charges. The first aspect is directly related to the conversion of alcohols: Besenbacher and coworkers highlighted, that defects and in particular Ti3+ interstitials govern the chemistry of molecular oxygen on rutile (110) [46, 50]. According to the previous chapter, the high oxophilicity by Ti3+ and oxygen vacancies leads to dissociation of molecular oxygen at low temperatures, and hence no molecular oxygen is present at highly reduced rutile surfaces [50]. Surprisingly, on medium reduced samples, the charge of subsurface Ti3+ sites stabilizes additional O2 molecules on the rutile surface. The O2 desorption is therefore observed around 410 K, when the diffusion of Ti3+ sets in [46].

Second example: We have recently reported embedded argon as a common residual of the TiO2 surface preparation [122]. The major argon desorption after sputtering accompanies the surface reoxidation and reordering into the well-known (110) structure. Using the Ar2p XPS signal as a sensitive probe of the environment of the embedded argon, it was found, that the Ar2p position is equal for highly reduced surfaces directly after ion bombardment and the remaining argon in the bulk of fully reordered surfaces. Thus, the remaining argon in the rutile bulk appears interacting with Ti3+ sites and therefore must be located in its near neighborhood, indicating that additional stabilization by defect sites may also play a role here [122].

The defect-mediated reductive coupling of two benzaldehyde molecules towards stilbene may serve as our third example. As shown above, Ti3+ is involved to abstract the oxygen from those molecules, but it was also shown based on FT-IRRAS and STM combined with DFT calculations, that the coupling of the two molecules proceeds at one common subsurface Ti3+ site that stabilizes both reaction intermediates and polarizes them, so that the formation of a new C–C double bond becomes possible [45, 51, 115].

4.2.1 Charge Transfer in Hybrid Oxidic Systems

From such examples it becomes clear, that charge transfer to adsorbate species can be an absolutely critical factor to tune the reactivity and to control the product distribution. However, this effect is not limited to molecular adsorbates but may also affect other structures such as coadsorbed clusters or nanoparticles. Such complex structures are often necessary to adjust catalysts towards the reaction’s requirements, and typically involve combinations of two or more materials. Within this work, we call these materials hybrid catalysts.

While metal deposits on reducible oxide surfaces are known to exhibit enlarged electron densities at metallic films or particles [123, 124], we have recently shown the close interaction of titania point defects with tungsten oxide clusters [125]. Besides their relatively high basicity at the terminal tungstyl groups, such clusters exhibit a rich chemistry due to various stable oxidation states and the ability of the fully oxidized cluster to store additional electrons in a delocalized d-aromatic state [126]. Indeed, by XPS we observed the generation of surface near Ti3+ sites upon deposition of reduced tungsten oxide clusters. Besides, stoichiometric clusters can accumulate Ti3+ from the crystal bulk near the surface, interact electronically with them and thus form apparently electron rich structures at the surface [125].

The generation and accumulation of reduced titanium sites certainly increases the oxygen affinity of the surface. Moreover, the high basicity of W = O also boosts the ability for hydrogen abstraction, which in turn promotes the overall conversion and especially oxidative pathways. Thus, an enhanced, stronger developed bifunctionality may be created by the proper choice of surface structures supplied with charge from (sub-)surface defects. This is not exclusive to rutile TiO2, but can be transferred to other substrates or clusters. For example, Artiglia et al. found evidence for a reverse charge transfer from vanadia clusters towards stochiometric rutile TiO2 [127]. As another example, Gamba et al. reported the abstraction of hydrogen from methanol molecules at various defects sites on iron-rich Fe3O4 (001) [128].

5 One at a Time: Steps on the Reaction Roadmap

Combining all the single aspects discussed up to this points paints a comprehensive image of the reaction roadmap for methanol on rutile surfaces. Here, the main steps on the way from molecular adsorption to product release for the two dominating pathways will be illuminated, to show how tracks can be chosen for methanol conversion on pure rutile TiO2 and a hybrid model catalyst, namely, tungsten oxide clusters ((WO3)n) on top of rutile TiO2 (110).

Figure 5a depicts the relevant steps after molecular adsorption of methanol on undercoordinated titanium sites (Ti5c) for the deoxygenation towards methane and the partial oxidation to formaldehyde. These are the two dominant products.

Fig. 5
figure 5

a Schematic of the main steps in the two dominant methanol conversion pathways (deoxygenation and partial oxidation) on bifunctional TiO2 surfaces and tungsten oxide clusters atop. b Radar overview of the enhancement of the dominant products formed from methanol based on the reaction conditions. All values were extracted from own, separately published studies, which provide comprehensive experimental details [55, 56]. A saturation dose of methanol dosed at 110 K was used, that saturates defect and Ti5c sites, as well as formation of a bilayer bond to bridging oxygen atoms. Herein, O2 denotes experiments with oxygen preadsorption (75 L at 110 K), UV denotes 30 min of UV irradiation (365 nm, photon flux >1.5 x 1016 s-1 cm-2, 110 K) after adsorption of oxygen and methanol where applicable. WO3 labels experiments with a tungsten oxide cluster coverage of 3.5 WO3 nm-2 and initial flash anneal to 880 K before further use. Finally, Ti3+ denotes use of a highly reduced rutile TiO2 sample (dark blue, >7.5 % Ti3+/Ti4+ based on normal emission Ti2p XPS). The least reactive surface (stoichiometric TiO2 without coadsorbates and without UV irradiation) was chosen as a reference and the relative enhancement according to the m/z=15 integral (for methane, as shown in Figure 2) and the integral area of peaks in the ratio of m/z=29/m/z=31 (for low-temp. and high-temp. formaldehyde, magnified by a factor of 5 for better visibility) is displayed in b.

Full size image

For all conversion routes, the dissociation of the O–H bond is the first step (1). Recombination with surrounding surface hydroxyls is virtually barrier-free, so water formation and release into the vacuum (2) is the next crucial step to ensure a high conversion by irreversible formation of methoxy moieties. This crucial step can be facilitated by basic sites such as electron-rich oxygen adsorbates (Oadδ- or O2δ- formed by charge transfer from Ti3+ as illustrated above). Indeed, a significantly enhanced product formation can be observed in all cases where oxygen is preadsorbed in the presence of defect sites (Fig. 5b, red boxes). This also applies to the (WO3 + O2)-labeled case, in which Ti3+ sites are formed during cluster deposition. The importance of coadsorbed oxygen also indicates that oxygen vacancies, which are healed in the presence of oxygen, are likely to play a minor role for O–H dissociation [50].

Once the methoxy-species is formed, it can undergo mainly two reaction channels on partially reduced rutile samples, that are depending on the presence of basic or oxophilic sites: If basic sites that are capable of C–H activations are present, the road towards formaldehyde is taken (3). Sites with a high enough basicity may be formed by electron-rich oxygen species after charge transfer from defect sites or photochemical generation or (O=W=O) groups of tungsten oxide clusters with an even higher basicity (Fig. 5a, step (6)). Thus, thermal formaldehyde is generated in the presence of oxygen and defects, particularly in the presence of tungsten oxide (Fig. 5b) due to their increased basicity. In case of UV irradiation, photochemically generated reactants such as excitons and oxygen radical species produced thereby are strong bases, leading to a more pronounced activation of C–H bonds at low temperature. Therefore, the product distribution after or during irradiation with UV light is altered in favor of the oxidative pathway [39]. Indeed, formaldehyde formed by partial oxidation is the dominant photochemical product, desorbing around 250 K after UV irradiation (Fig. 5b, purple). Again, the presence of coadsorbed oxygen as well as bulk Ti3+ interstitials promotes this pathway (Fig. 5a, step (4)), which appears as a combination of several effects ranging from an improved light absorption due to populated in-band-gap states to a higher conductivity and other electronic effects enabling longer exciton lifetimes [56, 129]. E.g., the higher coverage of methoxy-species on defective titania can alter the surface band bending of the semiconductor in favor of the oxidative pathway (upward band bending drives holes to the surface), while high coverages of molecular methanol may induce the reverse effect [130, 131]. Other theoretical studies suggest that the hole is more localized on methoxy-species than on molecularly adsorbed methanol, and therefore is more reactive in the first case [121]. Thus, probably a complex combination of chemical and physical effects governs the photochemistry here.

It is interesting, that in contrast to the photochemical reaction, the thermal partial oxidation towards formaldehyde appears in two features, one low-temperature channel around 230–280 K (purple in Fig. 5b) and an additional high-temperature channel above 600 K parallel to the deoxygenation (red in Fig. 5b) [52, 55]. FT-IRRAS[52] proved in consistence with adsorption of pure formaldehyde, [86] that the partial oxidation product formaldehyde is not present in its molecular form on the surface but desorbs from a dioxomethylene-like precursor formed by C–H cleavage from the methoxy species. Water release features were observed in the TPRS concomitant to both formaldehyde desorption features, and thus allow further mechanistic insights. For the low-temperature channel, this hints towards a preformed precursor, that already underwent C–H activation and is eventually decomposed to formaldehyde and water at 250 K. Note, that the desorption of pure water would be observed at slightly higher temperatures. In contrast, in the high-temperature channel, it is likely that the actual C–H activation appears in-situ and water directly gets released in parallel to formaldehyde formation. While no indications for different intermediates compared to the low-temperature channel could be found via FT-IRRAS, the thermal diffusion of Ti3+ towards the surface, that becomes dominant at temperatures above 450 K, may be one reason for the occurrence of two partial oxidation channels at different temperatures [44, 52, 125]. On the other hand, water release implies also removal of oxygen from the surface, which could also be driven by the formation of oxygen vacancies at elevated temperatures. This process is energetically favorable at high temperatures (> 600 K) as evident from the common thermal formation of clean, but defective rutile TiO2 (110) surfaces, and may be more dominant here since water is a more stable leaving molecule than release of atomic oxygen would be [35, 42].

Though partial oxidations are of high interest for technological applications, the dominant thermal pathway is the deoxygenation towards methane at temperatures above 500 K, which proceeds via the methoxy intermediate based on FT-IRRAS,[52] HREELS[59] and water coadsorption/desorption experiments [59, 83, 91]. Here, reduced titanium sites appear as oxygen abstractors, that finally cleave the C–O bonds forming reactive methyl groups which eventually build methane with hydrogen available at the surface (Fig. 5a, (6)). Indeed, a significant boost of the methane yield can be found for highly reduced titania substrates.

However, the methane formation is drastically depleted, if UV irradiation is applied. It is not surprising, that anticorrelation between thermally generated methane and photochemically formed formaldehyde is observed, because both reactions rely on methoxy intermediates [52, 56, 63, 95]. All methoxy-moieties, that are not converted towards formaldehyde by C–H abstraction, undergo thermal decomposition towards methane (C–O dissociation) during the TPRS experiment. In the photochemical case, the formation of methane is drastically depleted because methoxy-species are priorly photoconverted into a surface precursor, likely dioxomethylene, which eventually desorbs at 250 K in form of formaldehyde. In view of a potential technological application, this is especially interesting because the product desorbs at lower temperatures than the educt, and possible poisoning by highly thermally-stable species is widely circumvented because those adsorbates are converted in the photochemical reaction itself. The application of UV light is therefore an important switch for a high selectivity towards partial oxidation.

It is especially important to recognize, that defect sites in combination with oxygen are not only involved in one, but in almost every step in the alcohol conversion pathways. Therefore, these reactions are not linearly promoted, but appear multiply enhanced. However, depending on the distinct reaction step, either the supplied charge towards oxygen adsorbates, the bond polarization, the oxygen abstraction affinity, or the beneficial effects on the photochemical performance are the relevant aspects for distinct elementary steps.

Last but not least, we want to emphasize, that a complex reaction network exhibiting multiple influencing parameters is opened up here. Hence, it may not be straightforward to tune more than one parameter at a time due to interconnecting aspects. For example, it logically appears promising to combine the strong basicity of tungsten oxide clusters with photochemical reaction conditions on reduced titania to further increase the ability of C–H bond cleavage and maximize the formaldehyde yield. However, the observed formaldehyde yield in that case was far less than expected, while the thermal products were more pronounced. In this special case, the accumulation of electron density from defect sites near the oxide clusters likely quenches the photochemical reactivity due to enhanced recombination [56]. This shows, how complex and sensitive the versatile character of hybrid oxide catalysts can be even under model catalytic conditions.

On the other hand, a complex system always opens up a large playground for further catalyst tailoring involving other combinations of materials with stronger basic properties to increase the C–H cleavage activity. Here, some relevant ideas and approaches shall be conveyed, while the authors of this work do not aim in a comprehensive and complete list of the virtually infinite approaches to address bond activation on various oxide-based materials. This may include oxide-based metal single-atom catalysts[132], other oxide substrates, for example the polar surfaces of zinc oxide[15], iron oxides[133], copper oxides[134, 135], tungstates[136] as well as the use of oxidic clusters with altered acid–base properties like MgO[137] or vanadia[17, 18, 127, 138] clusters. Indeed, the partial oxidation of alcohols was found to be selectively affected by most of these approaches.

Many other strategies far beyond acid–base chemistry are in the focus of research groups to fine-adjust surface properties beyond the framework discussed herein. To be brief at this point, we want to introduce only a few relevant concepts here. The most prominent example is probably the combination of two modifications of titania in the commonly used P25. Herein, especially the photoactivity of the combination of rutile and anatase profits from the inverse charge mobilities and in principle enables an efficient charge separation for enhanced photochemical reactivity compared with the pure materials [39]. However, e.g. Stierle et al. recently showed that point defects in anatase may decrease the photoactivity, whereas the reverse trend was observed for rutile titania [139]. This underlines the general finding that it is often necessary to find the optimal compromise for the eventually desired products.

Furthermore, it is important to note numerous efforts involving various dopants. Mainly two approaches are utilized here [140]: First, doping by replacement of oxygen sites with elements such as nitrogen, carbon or phosphorous. This approach is similar to the self-n-doping with Ti3+ interstitials and creation of oxygen vacancies, but the introduction of heteroelements allows stronger effects [141,142,143]. Second, mixed oxide systems are often created by (subsurface) doping with (transition) metal ions, which again allow control of the electronic structure leading to fine-adjustable acidity at the surface [144,145,146,147]. Finally, combination with metal clusters and nanoparticles enables additional reaction channels such as hydrogenations or the production of hydrogen [62, 144, 148]. Essentially, even after decades of studies, there still is a widely unexplored multiparameter playground with a virtually infinite amount of combinatorial possibilities to tailor TiO2 based (photo-)catalysts, that is not easily predictable and therefore an essential field for fundamental research on our way to affordable and sustainable photocatalytic systems.

6 Conclusion

Within this perspective article, the complex nature of alcohol conversion on oxide surfaces was surveyed at the example of methanol conversion on rutile TiO2 (110) and hybrid catalysts based thereon. We have shown, that by choosing the right atomistic environment of the adsorbates, the (selective) activation of O–H, C–O and C–H bonds can be adjusted allowing to steer the product distribution in thermal and photochemical reactions. On the way to a more general picture, selected aspects were illuminated to understand the interwoven parameters to influence elementary reaction steps. We have thoroughly discussed the role of the intrinsically bifunctional rutile titania surface consisting of an ordered array of Lewis-acid (Ti5c) and basic (Obr and Opl) sites as well as the impact of point defects by means of oxophilicity and charge transfer phenomena. Finally, all single aspects were combined in a reaction roadmap, that depicts how distinct reaction pathways can be promoted through selective bond activation conditions on titania catalysts.