Activation of Small Organic Molecules on Ti²⁺-Rich TiO₂ Surfaces: Deoxygenation vs. C–C Coupling

3.1 Thermal Stability of Sputtered Titania Surfaces: Reoxidation and Restructuring

Argon ion bombardment is a common technique that may lead to a reduction of transition metal oxides for model studies [9], [14], [63]. As the sputter yield is highly sensitive to the atomic weight, the removal of oxygen from the surface is much more efficient than for the metal ions. Therefore, the increased amount of oxygen vacancies leads to a reduction of the remaining metal ions as oxygen desorbs neutrally and two electrons are transferred to the remaining metal ions for each vacancy [64]. In Figure 1 the Ti2p (a), O1s (b) and valence band (c) photoelectron spectra of a clean rutile TiO₂ (110) (1×1) surface (top lines) and of the same surface after 20 min of argon ion bombardment (bottom lines) is shown. The deconvolution of the Ti2p_3/2 and the O1s peak is exemplarily labeled as dotted, colored lines. The Ti2p signal of the pristine surface exhibits two spin-orbit split signals (Ti2p_1/2 and Ti2p_3/2) for each chemical species. Therefore, we observe two doublets, each corresponding to the regular Ti⁴⁺ (Ti2p_3/2 at 459.1 eV, red line) as well as the Ti³⁺ (Ti2p_3/2 at 457.8 eV, green line) from the defects.

Figure 1:

The X-ray photoelectron spectra (XPS) of a rutile TiO₂ (110) single crystal after 20 min of argon ion bombardment (bottom lines) (excitation with monochromatic Al Kα radiation), sputtered at 300 K with a sputtering current of 2 μA/cm² (top lines) and subsequent annealing at 873 K for 15 min. (a) Ti2p region, the deconvolution of the Ti2p_3/2 signals for the three chemical species is shown. (b) O1s region, again with exemplary deconvolution for the three species. (c) Valence band spectra showing the increased in-band-gap density of states for the sample after sputtering.

To make the argumentation clearer, we will focus on the Ti2p_3/2 signal in the following, because no new information is gained from the Ti2p_1/2 signal. The defect density can easily be determined by signal fitting and integration. However, one should keep in mind that XPS has a certain information depth (in this case, around 2 nm) and is not exclusively sensitive to the very first atomic layer, but to the topmost ones. Therefore, deviations typically occur between the defect density obtained from XPS and usual titration-desorption experiments (for instance using water [42], carbon monoxide [65] or oxygen [22], [23], [66]) or quantitative scanning tunneling microscopy (STM) results [20]. Argon ion sputtering induces a dramatic change to the Ti2p spectra (Fig. 1a). A new chemical species arises at 455.1 eV (Ti2p_3/2, blue line [≈36 % of the total Ti content]). In accordance with an earlier publication [63], we attribute this signal to Ti²⁺ species. The two other signals are significantly broadened and shifted to lower BE values (Ti2p_3/2 for Ti⁴⁺ at 458.7 eV (33 % of the total Ti content) and for Ti³⁺ at 457.0 eV (31 % of total Ti). The signal broadening indicates that the surface is not only reduced but also less ordered.

Considering the three different Ti oxidation states, the stoichiometry of the surface can also be estimated to be approximately TiO_1.5. We shall show at a later point, that the roughness and the absence of long-range order needs to be considered for the detailed quantification of sputtered surfaces. Annealing at 873 K for several minutes changes the Ti2p spectrum significantly. Here, the signal attributed to Ti²⁺ species completely vanished and the amount of Ti³⁺ species is dramatically decreased. Therefore, mostly Ti⁴⁺ is left on the surface, leading to the conclusion that the surface is reoxidized.

Figure 1b exhibits the XP spectra of the O1s region after argon ion bombardment (300 K) and annealing (873 K). After argon ion sputtering, a new, broad signal at 532.8 eV arises in the O1s spectrum (see Fig. 1b). In comparison to the Ti2p spectra, this should be related to the occurrence of Ti²⁺ species. Annealing to elevated temperatures results in a slightly asymmetric peak. The deconvolution of this signal proves the superposition of two signals. The main signal at 530.4 eV can be easily attributed to regular oxygen anions of the rutile lattice. The second, much smaller signal (531.4 eV) is dependent on the intrinsic defect density. Higher defect density leads to more pronounced signals here. Therefore, we attribute this signal to oxygen that interacts with point defects, such as oxygen vacancies or Ti³⁺ interstitials. In principle, both signals are shifted to higher binding energies compared to the main peak. This could also be due to the presence of hydroxyls at the surface. However, as we do not find any change of these signals by annealing to any temperature up to 900 K for any time for the clean rutile (110) surface but on the other hand a clear correlation to the Ti³⁺ content, we assign these signals to oxygen next to the reduced Tiⁿ⁺ species.

In accordance with the Ti2p spectra, all signals are significantly broadened after sputtering. After ion sputtering, the O/Ti ratio is depleted by approximately 20 %, which shows the oxygen deficiency of the sputtered surface. In detail, an O/Ti ratio of 1.78 was found for the surface after ion sputtering, while this value was 2.12 for the pristine surface. The O/Ti ratio for the sputtered surface differed from the stoichiometry that was estimated by the content of the three Tiⁿ⁺ species. One possible explanation is that the surface is significantly roughened after ion sputtering and appears without any long-range ordering. Therefore, it would be impossible to perform a simple estimation of the expected oxygen content by comparison to the bulk phase stoichiometries of different titanium oxides. In principle the quantification of the defect density should also be possible by determining the ratio of the Tiⁿ⁺ and oxygen species. However, as one oxygen vacancy equals approximately two electron charges, while Ti³⁺ interstitials are attributed to only one electron, the comparison from the defect density estimation of the different photoelectron signal ratios is difficult.

For chemical reactions as well as photocatalytic reactions, the valence band region is the most important. Particularly, the O2p and the lowest Ti3d orbitals constitute the occupied valence band while the conduction band is consisting mostly of Ti3d orbitals [24]. The valence band spectra obtained from the XPS experiments are shown in Figure 1c. For the pristine surface, the VB spectra are dominated by the O2p signals around 5.5 eV and 7.6 eV. In the band gap region (0–2 eV), there is a signal at low intensity. Therefore, only a small fraction of the Ti3d states are populated by the appearance of point defects, which are responsible for delivering additional electrons and filling the 3d orbitals. This in-band-gap density of filled states should be directly proportional to the defect density, as both point defects (Ti³⁺ and Ti²⁺) lead to the population of these states.

After ion sputtering, the O2p region is broadened, similar to the effects presented for the O1s region. More important is the band gap region. Here, more Ti3d states are populated, leading to a broad shoulder in the band gap region. The gap between the Fermi energy and the O2p levels is almost fully bridged, indicating a high conductivity and a broad UV/Vis absorption feature. Therefore, a more metal-like behavior and effective absorption of the solar light should be expected. However, as this is beyond the scope of this study, these properties were not evaluated at this point. For all three spectral regions, the reported species are in good accordance with earlier publications [59], [63].

3.2 Activation of the Organic Molecules on the Sputtered Rutile Surfaces: Deoxygenation and Reductive Coupling

To probe the activity of such Ti²⁺-rich rutile surfaces in chemical reactions, two different groups of oxygen-containing organic molecules are considered, namely, alcohols and ketones. We start with the adsorption of the simplest alcohol, which is methanol, on the sputtered rutile surface. By using TPRS, the desorption of methanol and the possible reaction products were traced. For the detailed assignment to molecular methanol desorption or the possible product formation, all spectra were analyzed with respect to the particular fragmentation patterns (see Fig. S1 in the Supplementary information). The fragmentation pattern is highly dependent on the used mass spectrometer and its specific adjustments. For a clearer understanding, we have chosen fragments that allow us to distinguish between different products.

The fragment m/z = 15 (methyl fragment) is, therefore, assigned to the desorption of methane, ethane or methanol. Here, m/z = 31 is exclusively present for the desorption of methanol, while m/z = 30 is observed for the desorption of either methanol or ethane. Whereas, the fragment m/z = 27 is dominant for ethane and ethene. By combining these fragments, a distinct assignment of the desorption traces is possible. However, it is often challenging to obtain an unambiguous assignment as many of these compounds have common fragments and exhibit relatively broad desorption features.

Figure 2 shows the recorded TPRS data after the adsorption of a monolayer of methanol on a sputtered rutile TiO₂ surface at 110 K. The dosing time on the sputtered surface was chosen as equivalent to a monolayer dosing on the clean and ordered (110) surface as described before [16]. First, we discuss the desorption of intact methanol. In a second step the product formation is highlighted.

Figure 2:

The TPRS data obtained after the adsorption of a monolayer of methanol on a sputtered rutile TiO₂ (110) surface at 110 K with the fragments m/z = 31 (methanol), m/z = 30 (methanol, ethane) sputtering at 300 K with a sputtering current of 2 μA/cm², m/z = 27 (ethane, ethene) and m/z = 15 (methanol, methane, ethane).

The main fragment of methanol (m/z = 31) reveals only one small desorption maximum at 335 K beside the tailing from the multilayer desorption. The desorption feature around 335 K is in proper agreement with the methanol desorption from regular Ti_5c (corresponding to Ti⁴⁺ centers) at the ordered rutile (110) surface. The slight shifts of the desorption temperature (approximately + 50 K in this case) can be easily explained due to the presence of many reduced Ti³⁺ and Ti²⁺ centers in the neighborhood of the Ti⁴⁺ centers. By comparing the m/z = 31 to the m/z = 15 signal we conclude that only a small fraction of 20 % of the adsorbed methanol desorbs intact without conversion to another product.

More important for the conversion of methanol is the temperature regime around 550 K. Here, the main desorption features for the fragments m/z = 15 and m/z = 27 are observed, whereas there is no peak for m/z = 31 (methanol). The signal for m/z = 15 is the most intensive one. Hence, we can infer that the deoxygenation reaction forming alkanes is the preferred pathway under these conditions. The desorption of ethane can be excluded because of the missing signal in the fragment m/z = 30 for this temperature. Moreover, this is a likely assumption due to the fact that the formation of methane requires addition of a hydrogen atom to the methyl fragment that is likely to be abstracted from methanol forming surface hydroxyls. Therefore, methane is apparently the main product. In addition, the reductive coupling product ethene (m/z = 27) is observed in a significant amount.

Methanol is already a relatively reactive molecule on defective titania surfaces [16]. The activation of more inert compounds such as acetone, that does not form any products on the clean (110) surface is another important point. Therefore, the activation of acetone was also examined, and in the case of acetone, more fragments must be considered. The typical fragmentation pattern for these compounds are shown in the Supplementary information (Fig. S3). For the intact molecular desorption of acetone, m/z = 58 and m/z = 43 are considered. However, due to the fragmentation pattern and the discrimination of higher masses of our quadrupole mass spectrometer, the m/z ratios higher than m/z = 50 are not intensive. Hence, these fragments cannot be easily distinguished from the noise. For the deoxygenation reaction, two possible products are to be considered: propene and propane. Here, the assignment is more complex. Propane desorption can be identified by a desorption signal in m/z = 29, but it also has a little contribution to m/z = 39. The detection of propene is likely based on the m/z = 39, but m/z = 42 is also possible. Fortunately, propane can be distinguished from propene as the latter does not exhibit the m/z = 29 fragment that is the most intense one for propane. As reductive coupling could also occur, 2,3 dimethyl-butane (m/z = 55) and 2,3 dimethyl-butene (m/z = 84) were monitored.

The comparison of the different fragments shown in Figure 4 demonstrates that no molecular acetone is desorbing from the unordered surface. This was determined by monitoring the main fragment (m/z = 43) and the fragment with the highest m/z ratio (m/z = 58) of acetone. The absence of the fragment at m/z = 58 even when magnified 10 times indicates that no molecular acetone is desorbed. Considering also the m/z = 55 fragment we conclude that these signals in m/z = 43 are not due to acetone desorption but are caused by 2,3 dimethyl-butane formation.

Three different desorption features can be identified for the fragments m/z = 29, m/z = 39, m/z = 42, m/z = 55 and m/z = 84. This indicates the formation of several products at three different desorption temperatures at 375 K, 483 K and 640 K, respectively. To distinguish between the possible products for both expected reactions (deoxygenation and reductive coupling), every fragment is related to a desorbing product as follows.

The first desorption feature at 375 K displays a signal mainly in three different fragments (m/z = 29, m/z = 39 and m/z = 42) and a small shoulder in the fragment m/z = 43. Based on the analysis of the fragmentation pattern of all possible products (see Supplementary information, Fig. S3) the desorption of propene and propane seems likely. Therefore, at the given temperature the deoxygenation reaction exclusively takes place. However, as in the fragment m/z = 29, the desorption at 375 K is more prominent than for the other shown fragments, we likely suggest that the propane yield is higher here compared to propene. The second desorption peak at 483 K is the most intensive signal. In addition to the four fragments described before, which are also present for this feature, some additional peaks for the fragments m/z = 55 and the highest measured fragment m/z = 84 are observed. Therefore, besides the deoxygenation process the reductive coupling of two acetone molecules to 2,3 dimethyl-butane (see Fig. 3: m/z = 55) and 2,3 dimethyl-butene (see Fig. 3: m/z = 84) also takes place. The last desorption species at around 640 K exhibits the formation of the same products as the feature at 483 K. Therefore, desorption signals can be detected in all given fragments without the fragment with the highest m/z ratio of acetone (m/z = 58). This means that, again, the total conversion of acetone is observed, including the formation of the deoxygenation products (propene and propane) as well as the reductive coupling products (2,3 dimethyl-butane and 2,3 dimethyl-butene). Notably, the major product formation for acetone is taking place at much lower temperatures than in the case of methanol.

Figure 3:

The temperature-programmed reaction spectra after the adsorption of a monolayer of acetone at 110 K onto a sputtered rutile TiO₂ (110) surface exhibiting the fragments m/z = 29 (propane), m/z = 39 (propane, propene, 2,3 dimethyl-butene), m/z = 42 (propene), m/z = 55 (2,3 dimethyl-butane) and m/z = 84 (2,3 dimethyl-butene) and m/z = 43 (acetone, propane, 2,3 dimethyl-butene) and m/z = 58 (acetone) sputtering at 300 K with a sputtering current of 2 μ/cm².

However, from the shown TPRS data, it is not possible to propose a reaction mechanism for the deoxygenation as well as for the reductive coupling reaction. Regarding the deoxygenation reaction of both molecules, it remains unclear how the proton addition (methane/propane) or migration (propene) occurs. In case of hydrocarbon formation from methanol we suggest that the necessary hydrogen comes from hydroxyls formed by dissociative methanol adsorption. In case of acetone, we consider the formation and dissociation of hydroxyl groups due to the keto-enol tautomerization. Our data do not allow us to distinguish between a genuine surface reaction or a recombinative desorption. FT-IRRAS experiments and additional TPRS studies, including isotopic labeling, may help to identify the reaction path in the future.

3.3 Adsorbate-Induced Reoxidation and Long-Range Ordering During the Conversion of Oxygen-Containing Organic Molecules

In both cases presented previously, the thermal conversion of oxygen-containing molecules on the sputtered rutile TiO₂ surface via a deoxygenation pathway and the reductive coupling was demonstrated. The intact molecular desorption of the reactants was observed only as minor traces. Therefore, we have to consider whether the missing oxygen atoms of these organic molecules can reoxidize and reorganize the surface. To clarify this, LEED pattern of the three different states during the temperature ramp were obtained. For that, a monolayer of acetone was adsorbed on the sputtered surface and then heated subsequently to two different temperatures (500 K and 650 K). Current studies have shown that the surface neither reoxidizes nor shows any signs of reordering in the LEED pattern after annealing at these temperatures. After cooling the sample to 110 K, another monolayer of acetone was adsorbed and again heated to the given temperatures. This procedure was repeated three times. In order to probe the surface changes during the reactions at two different temperatures, the given temperatures were chosen with respect to the TPRS data shown in Figure 3.

Figure 4 shows the recorded LEED patterns. For comparison, the LEED pattern of a clean rutile TiO₂ (110) (1×1) surface (a) and the observed LEED pattern of the sputtered surface obtained after 20 min of argon ion bombardment (b) are also given. After a monolayer adsorption of acetone at 110 K and heating to 500 K, no significant change of the pattern was observed (c). Whereas, after further heating to 650 K, clear spots appeared on the LEED pattern, indicating a starting reordering (d). The reflexes became more visible after every repetition of acetone adsorption and subsequent heating (Fig. 4e–h), leading to a further convergence towards the LEED pattern of a clean rutile TiO₂ (110) (1×1) surface.

Figure 4:

The LEED patterns and XP spectra recorded at different steps during the conversion of acetone on sputtered rutile surfaces (beam energy = 100.5 eV) sputtering at 300 K with 1 eV for 20 min. (a) Clean rutile TiO₂ (110) (1×1) surface. (b) Sputtered surface after the first adsorption of a monolayer of acetone and heating to (c) 500 K and (d) 650 K. (e) Second adsorption of a monolayer acetone and again heating to (e) 500 K and (f) 650 K as well as the third adsorption of a monolayer acetone and heating to (g) 500 K and (h) 650 K (adsorption temperature 110 K). The corresponding Ti2p XP spectra are shown on the right as indicated (excitation with monochromatic Al Kα radiation). A saturation dose at room temperature was used.

Therefore, the deoxygenation reaction and the reductive coupling are also driven by the (partial) reoxidation and reordering of the reduced surface. The missing oxygen atoms that are not monitored in any desorbing product are integrated into the rebuilt rutile TiO₂ (110) surface during both reactions. This also implies that we are unable to observe pure catalytic reactions here but at least partially stochiometric reactions with the titania itself. This conclusion is also supported by the TPRS investigations of the repeated adsorption-heating presented here.

These TPRS data are to be found in Figure S5 in the Supplementary information. The temperature programmed desorption from the surface was followed up to 680 K with repeated acetone adsorption at 110 K. A dramatic change was observed between the first and the second desorption spectra. The deoxygenation and reductive coupling yield dropped significantly (almost no desorption for m/z = 29 and 39). Instead of the three desorption features described previously, at the second and any further repetition, only one broad desorption feature at around 350 K appeared for these products. As this feature is a bit more pronounced for m/z = 29, we likely assign this feature to be dominated by propane desorption. Beyond the first repetition of the adsorption-desorption procedure a broad desorption feature for molecular acetone appears (m/z = 43, 58) around 274 K. This gives a hint for a dramatic chemical change of the surface during the first reaction cycle.

For a more systematic understanding, we have also followed the change of the surface during the conversion of acetone by XPS. Figure 4 (right set of spectra) shows the corresponding Ti2p spectra. The spectra for the ordered rutile TiO₂ (110) surface as well as the spectra after sputtering are given as references. As already shown in the first part, three different Ti species were present (Ti⁴⁺, Ti³⁺ and Ti²⁺, respectively). During the repeated adsorption of acetone (saturation doses at room temperature and subsequent heating to the given temperatures, see Experimental for details), the Ti²⁺ signal strongly decreased, whereas the Ti⁴⁺ signal increased and narrowed. The signal corresponding to Ti³⁺ was approximately constant. Similar conclusions can be drawn from the O1s and valence band spectra taken during the conversion of acetone. These spectra are to be found in the Supplementary information, Figure S6. Here, the O1s signal narrowed and the shoulder corresponding to the presence of oxygen in the neighborhood to reduced Tiⁿ⁺ centers as well as the filled in-band-gap density of states vanished during the procedure. Figure S7 (see Supplementary information) shows the corresponding C1s spectra taken for one adsorption cycle of acetone. As for the sputtered surface, no C1s signal was observed, two chemical carbon species (284.6 eV and 286.0 eV, approximately in a 2:1 ratio) were detected after the adsorption of acetone at 300 K. From the signal position as well as its stochiometric ratio, the signals can be attributed to the carbon in the methyl and the carbonyl fragments of acetone. Heating to 405 K and 500 K led to a continuous decrease of these signals, indicating the conversion of acetone as shown before. However, no assignment to different reaction pathways can be concluded from these data, especially because the slow desorption of the adsorbates by the X-ray radiation cannot be excluded. Heating to 650 K and to higher temperatures removes all carbon.

By combining the recorded LEED, TPRS and XPS data, we can identify the formation of propene, propane and coupling products (2,3 dimethyl-butane and 2,3 dimethyl-butene) by the reaction of acetone with Ti²⁺ centers at different temperatures between 350 K and 650 K, including a long range reordering of the surface towards the typical rutile (110) (1×1) structure. The missing oxygen from the deoxygenation reaction is most likely incorporated into the titania surface during this process. The signal narrowing is consistent with the restructuring of the surface as observed in LEED. However, this does not fully explain the drastic change of the TPR spectra after the first adsorption cycle. Anyway, one possible explanation for this discrepancy could be the different adsorption temperatures for TPRS/LEED (T = 110 K) and the XPS (T = 298 K) experiments, resulting in a higher coverage for the cooled samples.

These results indicate that the sputtered surfaces cannot only be reorganized by annealing but also by the adsorption and deoxygenation of oxygen containing molecules. This is in good agreement with the STM-based study by the Besenbacher group, who followed the thermal reaction of reduced Ti ions with oxygen to form TiO₂ islands [23], [24]. Another example is a study by the group of Sears, who used STM and low energy ion scattering to study the formation of ordered structures during annealing a reduced titania surface in an oxygen containing atmosphere [67].