Enhanced molecular adsorption of ethylene on reduced anatase TiO₂ (001): role of surface O-vacancies

Abstract

A density functional theory (DFT)-based study of ethylene (C₂H₄) adsorption on a reduced anatase titanium dioxide (TiO₂) (001) surface, i.e., with a surface oxygen vacancy (O_vac), is presented. It was found that C₂H₄ preferably adsorbs on the O_vac-site. The excess electrons originating from the removed oxygen weaken the C₂H₄ C=C double bond by filling the lowest unoccupied molecular orbital (LUMO) or of C₂H₄, and simultaneously enhance the binding of C₂H₄ to the reduced anatase TiO₂ (001). The bonding between the two C₂H₄ C atoms and the two Ti atoms nearest to the O_vac-site produces two σ-type bonds, leading to the emergence of a new localized mid-gap state. This hybrid 3d defect state can account for the ∼0.7 eV decrease in the band-gap observed in previous optical measurements of carbon-coated TiO₂, formed after TiO₂ exposure to C₂H₄ gas (Mater. Lett. 108, 2013, 134). Subsequent calculations on two initial decomposition pathways of C₂H₄ adsorbed on the O_vac-site show that the C–H bond and C–C bond cleaving require high activation barriers where the C–H bond cleaving is slightly easier compared to the C–C bond cleaving, 2.94 eV and 3.01 eV, respectively, (as calculated using DFT-D2+U_d = 3 eV) and the final states of the two initial decomposition pathways show similar endothermic characteristics. This finding indicates that the surface O_vac-sites tend to favor the formation of molecularly adsorbed Ti-bound sp³-C₂H₄ (–Ti–CH₂CH₂–Ti–) as compared to the dissociative adsorption case.

---

1 Introduction

There has been growing interest in the study of transition metal oxide–hydrocarbon molecule interactions in the past few years. This results from the wide applicability of the interaction in various fields. Some related applications include the decomposition of organic pollutants derived from volatile organic compounds,^1–4 petroleum chemical processes,^5,6 preservation of fresh foods,⁷ and sensitizing oxide surfaces with hydrocarbon-derived substances for solar cell devices.^8–10

In this paper particular attention is given to the ethylene–titanium dioxide (C₂H₄–TiO₂) interaction. Previous studies of the photocatalytic decomposition of C₂H₄ using TiO₂ found that in the absence of light irradiation, C₂H₄ did not adsorb nor dissociate with TiO₂. Experimental measurement shows no gas pressure change occurred even after long exposure of TiO₂ to C₂H₄. More recently Fourier-transform-infrared measurements on low temperature C₂H₄ adsorption using a Degussa P25-type of TiO₂ showed that C₂H₄ interacts weakly with TiO₂. The C₂H₄ vibrational frequencies did not undergo significant modification and C₂H₄ retains its planar configuration upon adsorption.^11–14 These results agree with theoretical calculations on C₂H₄ adsorption on a pristine anatase TiO₂ (001) surface.¹⁵ It was found that C₂H₄ adsorbs on anatase TiO₂ (001) via a π-complex interaction with the unsaturated five-coordinated Ti (Ti_5c). Furthermore, it was found that the dispersion (van der Waals) interaction further enhances the binding of C₂H₄.

A number of further studies on C₂H₄–TiO₂ have since been conducted, most of them focusing on C₂H₄ decomposition using TiO₂ as a photocatalyst.^16–20 Some recent experimental work have tried to modify TiO₂ surfaces with carbon by using a hydrocarbon such as C₂H₄ and acetylene as carbon sources.^21,22 Some other researchers have also have attempted to grow carbonaceous structures on TiO₂ for solar cell device application.^23,24 Indeed, these studies have given many valuable insights to understand the nature of C₂H₄–TiO₂ interaction. However, not much is known regarding the role of intrinsic defects to C₂H₄ interaction with TiO₂. Intrinsic defects, in particular oxygen vacancy (O_vac), are expected to have important roles in determining the physics and chemistry of transition metal oxide surfaces. The excess electrons originating from the O_vac often become a critical factor that governs the interaction of oxide surfaces with various kind of electron-acceptor molecules.^25–29 Therefore, it is necessary to study further how C₂H₄ adsorption – which can be regarded as an olefin prototype – behaves when an O_vac exists in TiO₂.

In the present work, TiO₂ anatase was chosen because of its relatively higher catalytic activity compared to its rutile counterpart.^30–32 For the surface facet, research was concentrated on the bulk-terminated (1 × 1) anatase TiO₂ (001) because of its superior catalytic activity, in particular when coupled with the presence of an O_vac.^33,34 Indeed, bulk-terminated anatase TiO₂ (001) is well known for its highly unstable characteristics, and thus, it usually undergoes surface reconstruction (the so-called (1 × 4)-reconstructed) to compensate for this instability. However, recent experimental work has succeeded in synthesizing anatase nanoparticles with the dominant bulk-terminated (001) facet. This is achieved by using a particular non-metallic atom (typically fluorine), which acts as a shape-controlling agent. Furthermore, this shape-controlling agent can be cleaned using heat treatment to produce a fluorine free surface without changing the basic crystal structure and morphology.^35–37 Based on these useful results, an attempt has been made to further explore the interaction between C₂H₄ and reduced anatase TiO₂ (001) surface. In this study, it will be shown that the presence of the O_vac promotes the formation of molecularly adsorbed C₂H₄. Furthermore, it will also be shown how the coupling between the C₂H₄-reduced anatase TiO₂ (001) can possibly extend the photoabsorption of TiO₂ to the visible light range. A similar study has been performed on the more stable (101) facet, as well. There are some differences in the energetics and configurations of the adsorbed C₂H₄, in particular, on the reduced anatase TiO₂ (101) surface with that on the reduced (001) surface. This originates mainly from the difference of the preferable site of the O_vac. The (101) surface favors the formation of the O_vac in the subsurface, which is different to the (001) surface which favors the formation of the O_vac on its outer most layer (this will be explained in Section 3.1). However, it is planned to publish the results in a future work to ensure a more rigorous and concise report.

2 Computational details

Spin-polarized density functional theory (DFT)-based calculations were carried out within the projector augmented-wave (PAW) method, as implemented in the Vienna Ab-initio Simulation Package (VASP) code.^38–40 The Generalized Gradient Approximation (GGA) as formulated by Perdew–Burke–Ernzerhof (PBE) was used to describe the exchange interaction and electronic correlation.^41,42 A plane-wave basis set was used with a kinetic cut-off energy of 520 eV. The geometric optimizations were carried out with residual forces less than 0.01 eV Å⁻¹. The calculated bulk lattice parameters for anatase TiO₂ are a = b = 3.83 Å and c = 9.66 Å, and these results are in good agreement with the results from previous work.^43,44

Approximately 16 Å of vacuum space was utilized and included a dipole correction to eliminate the artificial interaction between the periodically repeated slabs. A c(2 × 2) unit cell, was utilized containing four unsaturated Ti_5c atoms, each bonding to two-raised 2-coordinated (O_2c) and two-lowered 3-coordinated (O_3c) oxygen atoms along the [100] and [010] directions, respectively. Brillouin zone integration was sampled with 6 × 6 × 1 k-points. Each slab consists of four O–Ti–O layers. This amount of layers is sufficient to produces good convergence of the (001) surface structure^32,45 and has been widely used in previous studies.^46–48 Two upper O–Ti–O layers of the slab were allowed to relax during geometry optimization, keeping the other remaining O–Ti–O layers at the bottom side fixed to its initial bulk position. The adsorption energy is defined as:

E_ads = E_{adsorbate/surface} − (E_adsorbate + E_surface) (1)

where E_adsorbate is the energy of isolated C₂H₄ in the vacuum. E_surface is the total energy of the anatase (001) surface and E_{adsorbate/surface} is the total energy of the surface together with the adsorbed molecule. A long range dispersion interaction [or van der Waals (vdW) interaction] was included,⁴⁹ i.e.

The damping factor, S₆, is 0.75 as the GGA-PBE was utilized in all calculations. The dispersion coefficient (C₆) and the vdW radius (R₀) for O, C, and H are 0.7 J nm⁶ mol⁻¹, 1.75 J nm⁶ mol⁻¹, 0.14 J nm⁶ mol⁻¹ and 1.342 Å, 1.452 Å, 1.001 Å, respectively.⁴⁹ For Ti, C₆ and R₀ are set to be 10.8 J nm⁶ mol⁻¹ and 1.562 Å, following previous theoretical studies.⁵⁰

Previous theoretical studies on TiO₂ with an O_vac have shown that a common GGA functional failed to describe the experimentally observed localized defect states, because of inherent spurious self-interaction within the DFT formalism. This results in unphysical delocalization of the electronic states.^51–53 The utilization of DFT-D2+U_d, however, could overcome this problem. This method is implemented by introducing the on-site Coulomb interaction in the localized d-orbital and exchange interactions, by adding an effective Hubbard-U parameter to better describe the repulsion between the electrons on the same orbital.^54,55 Here, U_d = 3 eV was mainly used to determine the bonding mechanism and energetics of the initial decomposition reactions of C₂H₄ on reduced anatase TiO₂ (001). Indeed, the chosen U_d is not sufficient to fully open the band-gap of pristine anatase TiO₂ to its experimentally measured value (∼2.4 eV versus ∼3.2 eV). However, there are some considerations that have been taken into account as to why the present U_d was chosen: (1) the present choice of U_d properly describes the energy level of the reduced anatase TiO₂ (001) defect state with respect to the conduction band minimum (this will be elaborated further in the next section); (2) it gives a reasonable estimation for the reaction energies of TiO₂ with other molecules [e.g., hydrogen (H₂)];⁵⁶ (3) if the U_d-parameter (typically >5 eV) is too high it may open the band-gap more, but simultaneously it may also result in the wrong description of the adsorbate-surface bond when it comes to adsorption properties and also the energetics of corresponding chemical reactions.⁵⁶ In this work, the hybrid DFT calculation (HSE06)⁵⁷ was also performed to check the consistency of the result calculated using DFT-D2+U_d, in particular for C₂H₄ adsorbing on the O_vac site. Finally, the climbing image nudged elastic band method (CI-NEB) was utilized with four to eight transition images to determine the activation energy barriers for the initial C–H and C=C bond cleaving of adsorbed C₂H₄. Vibrational frequency calculations were then used for the confirmation of the transition state (TS).

3 Results and discussion

3.1 Oxygen vacancy (O_vac) on anatase TiO₂ (001) surface

To determine the O_vac site on the unreconstructed anatase TiO₂ (001), four different sites were considered, two on the surface and two in the sub-surface (Fig. 1). Here, ∼0.25 ML of vacancy concentration was utilized. The O_vac concentration considered gives 7.65 Å of O_vac–O_vac distance which leads to almost isolated impurities. The vacancy formation energies (E_{vac-formation}) needed to remove the oxygen from each site were then calculated. E_{vac-formation} is calculated by using the formula:

E_{vac-formation} = E_TiO2−x − (E_TiO2 + ½ E_O2) (3)

where E_TiO2−x is the total energy of the reduced slab, E_TiO2 is total energy of the pristine slab, and E_O2 is the free energy of the oxygen molecule in the gas phase. Calculated vacancy formation energies (Table 1) show that two-coordinated oxygen (2-fold or O_2c) on the surface is thermodynamically the most feasible site for O_vac formation (site #1). Upon the formation of surface O_vac, the two neighboring Ti atoms moved away from each other, resulting in an elongation of the Ti_4c–Ti_4c distance (d_Ti4c–Ti4c) to ∼4.89 Å (Ti–Ti atom distance on the pristine slab of (001) surface is ∼3.83 Å). The changes because of the oxygen vacancy also appeared in the electronic structure of the reduced surface. Local density of states (LDOS) of the defective surface shows the emergence of two defect states within the band-gap, located at around ∼0.5 eV and ∼1.2 eV from the conduction band minimum (Fig. 2a). These defect states can be attributed to the 3d conduction band of TiO₂, being occupied by the excess electrons originating from the removed O. The charge density profile of the defect states shows accumulation of excess electrons on the two adjacent Ti atoms near the O_vac (Fig. 2b). This accumulation produces a polaronic type lattice distortion to the atoms located in the vicinity of the O_vac. Commonly, this particular distortion originates from the charge carriers which exist in the polar semiconductor or insulator, producing a repulsive interaction between ions with the same charge and attractive interactions between those with the opposite charge. In the present case, this interaction originated from the excess electrons of the O_vac that reside in the nearest Ti-atoms. This is similar to what has also been observed in a previous work⁵⁸ [LDOS of TiO₂ (001) with a surface O_vac for DFT-D2 and DFT-D2+U_d = 4 eV can be found in the ESI Fig. S1 and S2†]. Overall, the present result for the reduced anatase TiO₂ (001) surface with O_vac shows good agreement with the results found for previous theoretical calculations^58–60 as well as experimental observations using photoemission spectroscopy.^61,62

Fig. 1 (a) The numbers (1–4) indicate the four possible O_vac sites considered in this work, (b) O_vac site considered for C₂H₄ adsorption in this work and the surrounding area near the vacancy site [site #1 in (a)]. The d_Ti4c–Ti4c denotes the elongation of two 4-fold Ti (Ti_4c) upon the C₂H₄ formation.

Table 1 Calculated neutral O_vac formation energy (E_{vac-formation})

Site	Evac-formation (eV)	Difference
1	3.50	—
2	4.35	0.85
3	4.47	0.97
4	4.80	1.30

---

Fig. 2 Local density of states (LDOS) of (a) defective anatase TiO₂ (001) with surface O_vac calculated DFT-D2+U_d (U_d = 3 eV), (b) 3d orbital projection of the two Ti atoms nearest to the vacancy site. Negative and positive values of the y-axis are minority and majority spin, respectively. The Fermi level is set to 0 eV (isosurface = 0.02 e Å⁻³).

3.2 C₂H₄ molecular adsorption on a surface with an O_vac

Previous studies have observed several geometric configurations for C₂H₄-solid surface interaction such as π-complex, di-σ complex, ethylidene, vinylidene, ethylidyne and ethylylidene.⁶³ This study focused on the π-complex and di-σ complex interactions. These interactions have been consistently observed in the low temperature C₂H₄-solid surface experiments, in which C₂H₄ molecularly adsorbs parallel to the surface and remains intact with respect to its gas phase.⁶⁴ The previously mentioned interactions mainly originate from the highly reactive π/π* orbitals of the C₂H₄ C=C bond.

Five different adsorption configurations of C₂H₄ on the reduced anatase TiO₂ (001) with surface O_vac as shown in Fig. 3 were studied. Note that only C₂H₄ adsorption on the reduced anatase TiO₂ (001) with O_vac created at site #1 which is the most feasible site for creating O_vac was considered. Single point calculations were first performed for each of the five configurations, where all degrees of freedom were constrained and only the vertical distance of C₂H₄ relative to the surface was varied (Fig. S3, ESI†). Based on the one-dimensional potential energy curves obtained from each of the configurations, the distance at which the C₂H₄-reduced TiO₂ (001) surface system gives the lowest energy was further fully optimized. In previous work by Shukri and Kasai, it was found that the long range dispersion interaction shows a significant contribution to the adsorption energy of C₂H₄. This affects overall C₂H₄ stability on the pristine TiO₂ surface.¹⁵ The contribution of vdW interaction was also checked by comparing the calculated adsorption energy of C₂H₄ on each adsorption site. Indeed, the vdW interaction still contributes a significant part to the overall calculated adsorption energies (Table S1, ESI†). Therefore, in the present study, the long range dispersion interaction (vdW interaction) was also included in the calculations.

Fig. 3 Top view (left panels) and side view (right panels) of C₂H₄ adsorption configurations on and in the vicinity of surface O_vac, (a) and (b) are C₂H₄ on hollow-site [Hol], (c) and (d) are C₂H₄ on O_3c-site [O_3c], (e) and (f) are tilted-C₂H₄ on O_3c-site [O_3c-tilted], (g) and (h) are C₂H₄ on Ti_4c-site [Ti_4c], (i) and (j) are C₂H₄ on O_vac-site [O_vac]. The name for each configuration is given in square brackets. The dash–line circle corresponds to the location of oxygen vacancy.

The energetics and geometric properties of C₂H₄ adsorbed on a defective anatase TiO₂ (001) surface calculated at DFT-D2+U_d (U_d = 3 eV) are shown in Table 2. C₂H₄ adsorption on Ti_5c, O_3c, hollow, and O_3c-tilted sites resulted in relatively weak adsorption energies. These weak interactions were also indicated by the geometrical structure of C₂H₄, having almost no perturbation on its initial sp²-hybridization and adsorbed at a relatively far distance from the reduced TiO₂ (001) surface. For molecularly adsorbed C₂H₄ on those four previously mentioned sites, it can be inferred that vdW is the dominant interaction that stabilizes the adsorption.

Table 2 Adsorption energies and selected geometric parameters of adsorbed C₂H₄ on different configurations calculated at DFT-D2+U_d (U_d = 3 eV). See Fig. 3 for the names of each adsorption configuration

Configuration	ΔEads (eV)	dC–C (Å)	dC–H (Å)	dC–Ti (Å)
Ti4c	−0.10	1.33	1.10	3.18
O3c	−0.18	1.33	1.09	3.32
Ovac	−0.94	1.47	1.09	2.10
Hol	−0.32	1.33	1.09	3.87
O3c-tilted	−0.15	1.33	1.10	3.40

---

C₂H₄ adsorption on an O-vacancy site (O_2c-site), however, has a quite distinct property when compared to the other four configurations. Upon adsorption, the initial (gas phase) sp²-hybridization of C₂H₄ changed into sp³-type structure, following the elongation of the C₂H₄ C=C bond. The C=C bond was elongated to ∼0.16 Å compared to its gas phase bond length, indicating a weakened bonding interaction between the two carbon atoms. This weakening also corresponds to the decrement of the C₂H₄ bond order from the initial double bond (C=C) to become a single bond (C–C). From now on, the weakened carbon–carbon bond of adsorbed C₂H₄ with the C–C notation will be referred to. In order to further elucidate the mechanism of C₂H₄ adsorption on the O_vac-site, the orbital-resolved LDOS of the isolated C₂H₄ gas phase and C₂H₄ adsorbed on the O_vac-site [Fig. 4 and 5] was plotted. Here, the p_z-orbitals of the two C atoms formed the highest occupied molecular orbital (HOMO; π_cc)–LUMO of C₂H₄. These two frontier molecular orbitals (MOs) often play a significant part in C₂H₄ interaction with various solid surfaces. By observing the plot of the orbital-resolved LDOS of C₂H₄ adsorbed on the O_vac site (Fig. 5), it can be seen that there is a strong hybridization between the p-states of C₂H₄ C–C with the Ti 3d-states, i.e., two adjacent Ti atoms in the vicinity of the O_vac.⁶⁵ Another noticeable feature is the emergence of a localized occupied state below the Fermi level, which formed by the mixing of the of C₂H₄ with the occupied Ti 3d defect states.⁶⁶ This hybridization implies that charges from occupied defect states of Ti 3d are transferred to the of C₂H₄, indicating a back donation from the defect states to C₂H₄ that subsequently weakened the adsorbed C₂H₄ C–C bond. The weakening because of the depletion of charges in the C–C region upon adsorption can also be observed, whereas accumulation of charges took place in the two C–Ti bonding and C–H bonding regions (Fig. 5d). The accumulation of charges in the C–H bonding regions subsequently tilted the four hydrogen atoms, and as a consequence the C₂H₄ minimized the repulsive interactions between its bonding electrons. Fig. 5 also shows the plot of the spin density profile within the energy range of Ti 3d defect states – derived-bonding localized mid-gap state that shows two σ-type bonding profiles between both C atoms of C₂H₄ with the two Ti atoms near the vacancy site. These two σ-type interactions (or the so-called di-σ complex) characterize the rehybridization of C₂H₄ into the sp³-type upon its adsorption on the O_vac site. It is interesting to note that previous experimental work has shown that coating a anatase TiO₂ surface using C₂H₄ and acetylene as carbon sources resulted in the decrease of the TiO₂ band-gap by about ∼0.7 eV, subsequently increasing the photoactivity of the C-modified TiO₂ under visible light.²² These present results clarified at least one possible origin of this band-gap decrement. The hybrid – 3d defect state resulting from C₂H₄ – reduced anatase TiO₂ (001) coupling (located at ∼1.0 eV below the conduction band minimum) can account for the increasing photoactivity of C-modified TiO₂, thus extending the system photoactivity to the visible light range. Thus, exposing the system to ultraviolet-visible light can trigger electronic excitation from the hybrid – 3d defect state to the conduction band that accounts for the experimentally observed decreasing band-gap. This is similar to other electron-acceptor molecules adsorbed on a reduced TiO₂ surface such as oxygen (O₂), where a localized mid-gap state was also observed upon adsorption.⁶⁷ In addition, a hybrid-DFT calculation was also performed to check the consistency of the present result (C₂H₄ adsorbed on the O_vac-site) obtained by DFT+U_d (U_d = 3 eV). The optimized geometry and adsorption energy results obtained using hybrid-DFT are in good agreement with the results obtained from DFT+U_d (U_d = 3 eV) (Table 3). This consistency ensures the reliability of the chosen U_d value. To this end, it can be concluded that excess electrons because of the oxygen vacancy play a vital role to stabilize C₂H₄ adsorption on the reduced anatase TiO₂ (001) surface. The present finding is in contrast to the C₂H₄ adsorption on clean/pristine anatase TiO₂ (001), in which C₂H₄ adsorption is mainly driven by the dispersion (or vdW) interaction.¹⁵

Fig. 4 Isolated molecular orbital (MO) of C₂H₄ projected onto the sum of the two C atoms' 2p-state with their corresponding orbitals. The x-axis represents energy in eV and the y-axis corresponds to the states per energy level (isosurface = 0.02 e Å⁻³).

Fig. 5 LDOS of C₂H₄ adsorbed on a O_vac-site calculated at DFT-D2+U_d (U_d = 3 eV) and the corresponding partial charge density of the newly formed mid-gap state. (a) Orbital-resolved LDOS of the two C-atoms of C₂H₄ projected onto their p-state, plotted together with total DOS, (b) orbital-resolved LDOS of the two Ti-atom near the O_vac projected onto their 3d-orbital, (c) simplified schematic interaction between 3d defect states (3d_def) with C₂H₄ LUMO , (d) charge density difference (CDD) plot of C₂H₄ adsorbed on O_vac-site. The green color corresponds to the charge depleted region, whereas the yellow corresponds to the charge accumulated region. The Fermi level is set to 0 eV (isosurface = 0.02 e Å⁻³).

Table 3 Calculated adsorption energies and selected geometrical parameters of C₂H₄ adsorbed on the O_vac-site calculated at DFT-D2+U_d (U_d = 4 eV, 5 eV) and hybrid-DFT

Calculation method	Eads (eV)	C–C (Å)	C–H (Å)	C–Ti (Å)
DFT-D2+Ud (Ud = 4 eV)	−0.8	1.49	1.1	2.14
DFT-D2+Ud (Ud = 5 eV)	−0.32	1.35	1.1	2.64
Hybrid-DFT (HSE06)	−1.12	1.50	1.10	2.10

---

Previous research has shown that by controlling the work function (W), it may be possible to control the redox chemistry of oxide materials as well.^25,28 In conjunction with the present work on the C₂H₄-reduced TiO₂ system, the influence of increasing reduced anatase TiO₂ (001) work function (W_TiO2) on the adsorption of C₂H₄ was also checked. This idea is tested in an ad hoc manner, whereby the position of the defect state (which simultaneously determines the W_TiO2) is varied upon altering the U_d values. Here, the C₂H₄ adsorption properties on O_vac for two other U_d values (U_d = 4 eV and 5 eV) have been calculated. The different behaviors of C₂H₄ adsorption calculated with U_d = 5 eV were observed, whereas the result for U_d = 4 eV showed overall similarity to the result calculated with U_d = 3 eV (Fig. S4, ESI†). C₂H₄ retained its planar geometry and its adsorption energy was also significantly lowered (Table 3). In order to further elucidate the decrease of the adsorption strength of C₂H₄ on the vacancy site calculated with U_d = 5 eV, the LDOS for the present case were plotted (Fig. 6a). Compared to the DFT-D2+U_d (U_d = 3 eV and 4 eV) cases, some less significant orbital hybridization can be observed, in particular between the Ti 3d defect states with C₂H₄ (LUMO). Indeed, some fractions of C₂H₄ can still be observed within the gap states and also near the valence band edge, showing some electronic filling of the orbital. However, those hybridizations are not strong enough to produce significant stabilization for the C₂H₄-reduced TiO₂ (001) coupling, because most of the broadening of still occupies the region above the Fermi level, resulting in less significant interplay with the π_cc (HOMO). This indicates that charge transfer from the surface to the C₂H₄ is significantly hindered because of the deeper emergence of the defect states in the energy level (or more tightly bound excess electrons) (Fig. 6b). Nevertheless, the tendency of the di-σ bonding formation from the charge density difference (CDD) plot can still be seen, as the charges accumulate in both the C–Ti regions (Fig. 6c). The results obtained from DFT-D2+U_d (U_d = 5 eV) imply another important point that the deeper the energy level of excess electrons (or the more the excess electrons are tightly bound), the more difficult the charge transfer to the C₂H₄ LUMO will be. This is similar to if the work function of a substrate is controlled, as has also been demonstrated by previous work of O₂ adsorption on reduced rutile TiO₂ (110) surface.²⁹

Fig. 6 (a) Orbital-resolved LDOS of two C-atoms of C₂H₄ projected onto their 2p-state plotted together with total DOS calculated at DFT-D2+U_d (U_d = 5 eV). Projection of partial charge density profile of within two energy windows are also shown, (b) LDOS of reduced anatase TiO₂ (001) with O_vac calculated at DFT-D2+(U_d = 5 eV), and (c) CDD plot of C₂H₄ adsorbed on O_vac-site. Green color corresponds to charge depleted region, whereas yellow corresponds to charge accumulated region. The Fermi level is set to 0 eV (isosurface = 0.02 e Å⁻³).

3.3 C₂H₄ dissociative adsorption on a surface with an O-vacancy

In this sub-section, the relevant results for two initial decomposition pathways of C₂H₄ will be presented, namely: (1) C–H bond cleaving (or dehydrogenation), and (2) direct C–C bond cleaving.

The ‘*’-symbol indicates the adsorbed species on the anatase TiO₂ (001) surface; O_2c-surf corresponds to the nearest two-fold coordinated oxygen (from the O_vac-site) on the topmost layer. Note that adsorbed C₂H₄ on O_vac-site was adopted as the initial configuration for both decomposition pathways. The products for the two initial decomposition pathways are determined based on the two criteria: (1) the adsorbed C₂H₄ dissociated fragments that give the most energetically stable configuration (relative to isolated gas-phase C₂H₄ and reduced TiO₂ (001) slab); (2) the shortest diffusion pathway for C₂H₄ H-atom upon dehydrogenation. The first pathway (dehydrogenation) results in the breaking of one H atom of C₂H₄, followed by the formation of a surface hydroxyl with the nearest O_2c (two-fold coordination oxygen), whereas the remaining fraction (CHCH₂ or vinyl) makes a bond with the two Ti atoms nearest to the O_vac-site. For the second pathway (C–C bond cleaving), C₂H₄ breaks into two CH₂ fragments, where one fragment occupies the O_vac-site and another fragment forms adsorbed formaldehyde with the nearest neighbor O_2c (two-fold coordination oxygen) as well. In addition, the consideration of formaldehyde as a C–C cleaving product is based on the previous experimental work in which formaldehyde is frequently observed during hydrocarbon–oxide interaction.^68,69

As shown in Fig. 7, it can be see that both decomposition pathways are highly activated reactions with calculated activation barriers of 2.94 eV and 3.01 eV for C–H and C–C bond cleaving, respectively. The calculated reaction energies for both decomposition pathways show similar endothermic characteristics. It is interesting to see that both decomposition pathways show almost similar activation energies. We can attribute this similarity in high activation barrier, partly, to the fundamental characteristics of the C–H and C–C bonds. For dehydrogenation, note that the C–H bond is formed by a strong σ-type interaction, so that breaking this particular bond requires a large amount of energy. It is also important to note that diffusion energy needed by the H to move to the nearest O_2c (to form the surface hydroxyl) is also accounted for in the present calculated barrier. This is also in agreement with results from previous work, where atomic H needs to overcome a certain amount of activation barriers to diffuse on the surface of TiO₂, and this can range from 0.7 eV–2.06 eV, depending on the diffusion pathways and simulation environments.^70,71

Fig. 7 First two initial decomposition pathways of C₂H₄ adsorbed on O_vac-site. Reaction (4) corresponds to C₂H₄ dehydrogenation, whereas (5) corresponds to direct C=C bond cleaving. IS, TS, and FS correspond to initial, transition, and final state, respectively. All energies are calculated with respect to isolated (gas phase) C₂H₄ and reduced slab anatase TiO₂ (001) with surface O_vac.

For C–C bond cleaving, a slightly higher activation barrier by about 0.07 eV can be observed, when compared to the C–H bond cleaving. However, similar to the C–H bond, the same argument also prevails in the C–C bond cleaving case, because, once again, the breaking of the σ-type interaction is being dealt with. The C–C bond cleaving consequently requires a certain amount of energy as well. Another important point is that the activation barrier for the C–C bond cleaving also involves the contribution from the energy needed to break the Ti_4c–O_2c bonding nearest to the O_vac-site, in order to facilitate the formation of adsorbed formaldehyde species. To this end, it can be inferred that this high activation barrier for C–C bond cleaving stems from two contributions: the energy needed to break the C–C σ-bond, and also, subsequently, the Ti_4c–O_2c bond in order to facilitate H₂CO formation as the final product. To some extent, this similarity is analogous to C–H and C–C bond breaking of gas phase C₂H₆ in which all the CC and CH bonds are composed by σ-type bonding (the energies needed to break CC and CH in C₂H₆ are ∼3.95 eV and ∼4.36 eV, respectively). Indeed, to the best of our knowledge, there are direct experimental data available that can be referred to and compared with the present results. However, the present tendency of the formation of C–C coupling on the O_vac is similarly observed when reduced rutile TiO₂ (110) with surface bridging O_vac is exposed to formaldehyde. By using scanning tunneling microscopy, Zhu et al. observed that, mediated by the O_vac, adsorbed formaldehyde and/or methylene can couple with other adsorbed formaldehyde/methylene species on the surface through several reaction channels, forming C₂H₄ as the final product as the temperature increased.^68,69 The present results support these experimental findings, in which coupled C–C bond (or described here as molecularly adsorbed C₂H₄) is more favorable when compared to the dissociative adsorption cases when O_vac exists on the surface of TiO₂ [as shown by the high dissociation barriers and the comparison of thermodynamic stability of the initial reactant (molecularly adsorbed C₂H₄) with the final product, namely, C–H and C–C bond cleaving]. An interesting recent work on acetylene (C₂H₂) adsorption on anatase TiO₂ (101) and (001) surfaces by Chen et al. should also be mentioned briefly.⁷² The published report focuses on the study of C₂H₂ cyclo-oligomerisation mechanism on anatase TiO₂ surfaces. In particular, it was reported that C₂H₂ spontaneously dissociates upon adsorption on the pristine anatase TiO₂ (001) via H–CCH bond cleaving. The dissociated CCH fragment then binds to the surface of the Ti, forming a Ti–CCH surface complex where the H binds to the nearest O_2c producing surface OH. Therefore, there seems to be a contradiction between the present results with those obtained by Chen et al., as a relatively high activation energy of C₂H₄ CH bond cleaving was reported in the current paper. However, it is also important to note some aspects of the two systems in order to come to terms with these seemingly different results. The present results deal with reduced anatase TiO₂ (001) with surface O_vac whereas Chen et al. deal with the pristine surface of anatase TiO₂ (001). This difference in the surface of interest may already lead to a different adsorption and reaction mechanism as well. A qualitative (and to some extent semi-quantitative) similarity that may be noticed is that in terms of the thermodynamics of the reaction, this result agrees with that obtained by Chen et al., in which reduced anatase TiO₂ (001) favors the dehydrogenation or the CH bond cleaving. Indeed, it is more reasonable and interesting to further discuss the C₂H₂ adsorption on the pristine anatase TiO₂ (001) results with the previous work of C₂H₄ on pristine anatase TiO₂ (001).¹⁵ However, this is already beyond the scope of the present work. More rigorous work related to the decomposition pathways of unsaturated hydrocarbon (both C₂H₂ and C₂H₄) on the pristine (as well as hydroxylated) surface of anatase surfaces is currently being done and will soon be covered in a future report.

4 Conclusion

Based on density functional theory calculations, C₂H₄ adsorption on an anatase TiO₂ (001) surface with O_vac has been studied. Oxygen vacancies tend to form on the topmost layer, occupying the O_2c (2-coordinated oxygen ion) site with a vacancy formation energy of ∼3.5 eV. The results presented in this paper suggest that C₂H₄ adsorbed on the O_vac-site is the most stable configuration. The adsorption of C₂H₄ on the O_vac-site is accompanied by C₂H₄ rehybridization from its initial gas-phase sp² structure to sp³-type hybridization. This is because of the filling of the C₂H₄ LUMO by the excess electrons of the O_vac residing in the vicinity of the two nearest adjacent Ti atoms. This implies that the presence of an O_vac is important for C₂H₄ binding on the TiO₂ surface. Electronic structure analysis shows that the bonding between C₂H₄ and surface Ti atoms on the O_vac site produced a new localized mid-gap state. This state could provide a direct explanation for the origin of the reduced band-gap observed in a previous optical measurement of carbon-coated TiO₂. Subsequent calculations on two initial decomposition pathways of C₂H₄ adsorbed on the O_vac-site show that the C–H bond and C–C bond cleaving require high activation barriers, with the C–H bond cleaving being slightly more favorable when compared to the C=C bond cleaving (2.94 eV and 3.01 eV, respectively) and the final states of the two initial decomposition pathways show similar endothermic characteristics. Finally, this finding indicates that surface O_vac-site tends to promote the formation of molecularly adsorbed Ti-bound sp³-C₂H₄ (–Ti–CH₂CH₂–Ti–) when compared to the dissociative adsorption case.

Acknowledgements

This work is supported in part by: Japan Ministry of Education, Culture, Sports, Science, and Technology (MEXT) Grant-in-Aid for Scientific Research (15H05736, 24246013, 15KT0062, 26248006); NEDO Project “R&D Towards Realizing an Innovative Energy Saving Hydrogen Society based on Quantum Dynamics Applications”; and the Osaka University Joining and Welding Research Institute Cooperative. Some of the numerical calculations presented here were done using the computer facilities at the following institutes: CMC (Osaka University), ISSP, KEK, NIFS, and YITP. GS would like to express gratitude to MEXT for the financial study aid. HKD and MKA gratefully acknowledge research grant support from the Institut Teknologi Bandung – Japan International Cooperation Agency (ITB-JICA) under the “Proyek Pengembangan ITB (III)” scheme. Fruitful discussions with Dr Adhitya Gandaryus Saputro (Institute of Technology, Bandung) and Dr Allan A. B. Padama (University of the Philippines, Los Baños) are also acknowledged.

References

1. M. Baldi, V. S. Escribano, J. M. G. Amores, F. Milella and G. Busca, Appl. Catal., B, 1998, 17, L175.
2. M. Paulis, L. M. Gandia, A. Gil, J. Sambeth, J. A. Odriozola and M. Montes, Appl. Catal., B, 2000, 26, 37.
3. L. Forni and I. Rosetti, Appl. Catal., B, 2002, 38, 29.
4. M. R. Morales, B. P. Barbero and L. E. Cadús, Appl. Catal., B, 2007, 74, 1.
5. J. C. Mol, Catal. Today, 1999, 51, 289.
6. K. Tanaka, K. Miyahara and K. I. Tanaka, J. Mol. Catal., 1982, 15, 133.
7. M. Hussain, S. Bensaid, F. Geobaldo, G. Saracco and N. Russo, Ind. Eng. Chem. Res., 2011, 50, 2536.
8. M. Grätzel, Acc. Chem. Res., 2009, 42, 1788.
9. J. Liu, Y. T. Kuo, K. J. Klabunde, C. Rochford, J. Wu and J. Li, ACS Appl. Mater. Interfaces, 2009, 1, 1645.
10. G. K. Mor, S. Kim, M. Paulose, O. K. Varghese, K. Shankar, J. Basham and C. A. Grimes, Nano Lett., 2009, 9, 4250.
11. I. S. McLintock and M. Ritchie, Trans. Faraday Soc., 1965, 61, 1007.
12. G. Busca, G. Porcile and V. Lorenzelli, J. Mol. Struct., 1986, 141, 395.
13. G. Busca, V. Lorenzelli, G. Ramis and V. Sanchez Escribano, Mater. Chem. Phys., 1991, 29, 175.
14. G. Busca, V. Lorenzelli, G. Ramis, J. Saussey and J. C. Lavalley, J. Mol. Struct., 1992, 267, 315.
15. G. Shukri and H. Kasai, Surf. Sci., 2014, 619, 59.
16. X. Fu, L. A. Clark, W. A. Zeltner and M. A. Anderson, J. Photochem. Photobiol., A, 1996, 97, 181.
17. S. Yamazaki, S. Tanaka and H. Tsukamoto, J. Photochem. Photobiol., A, 1999, 121, 55.
18. D.-R. Park, J. Zhang, K. Ikeue, H. Yamashita and M. Anpo, J. Catal., 1999, 185, 114.
19. T. A. Westrich, K. A. Dahlberg, M. Kaviany and J. W. Schwank, J. Phys. Chem. C, 2011, 115, 16537.
20. B. Hauchecorne, T. Tytgat, S. W. Verbruggen, D. Hauchecorne, D. Terrens, M. Smits, K. Vinken and S. Lenaerts, Appl. Catal., B, 2011, 105, 111.
21. N. K. Memon, D. H. Anjum and S. H. Chung, Combust. Flame, 2013, 160, 1848.
22. D. H. Anjum, N. K. Memon and S. H. Chung, Mater. Lett., 2013, 108, 134.
23. S. M. Jain, J. J. Biedrzycki, V. Maurino, A. Zecchina, L. Mino and G. Spoto, J. Mater. Chem. A, 2014, 2, 12247.
24. J. J. Biedrzycki, S. Livraghi, I. Corazzari, L. Mino, G. Spoto and E. Giamello, Langmuir, 2015, 31, 569.
25. M. T. Greiner, M. G. Helander, W. M. Tang, Z. B. Wang, J. Qiu and Z. H. Lu, Nat. Mater., 2012, 11, 76.
26. M. A. Henderson, Surf. Sci. Rep., 2011, 66, 185.
27. I. Justicia, P. Ordejón, G. Canto, J. L. Mozos, J. Fraxedas, G. A. Battiston, R. Gerbasi and A. Figueras, Adv. Mater., 2002, 14, 1399.
28. M. T. Greiner and Z. H. Lu, NPG Asia Mater., 2013, 5, e55.
29. N. A. Deskins, R. Rousseau and M. Dupuis, J. Phys. Chem. C, 2010, 114, 5891.
30. A. S. Barnard and P. Zapol, Phys. Rev. B: Condens. Matter Mater. Phys., 2004, 70, 235403.
31. A. S. Barnard, P. Zapol and L. A. Curtiss, Surf. Sci., 2005, 582, 173.
32. M. Lazzeri, A. Vittadini and A. Selloni, Phys. Rev. B: Condens. Matter Mater. Phys., 2001, 63, 155409.
33. G. Liu, H. G. Yang, X. Wang, L. Cheng, H. Lu, L. Wang, G. Q. Lu and H. M. Cheng, J. Phys. Chem. C, 2009, 113, 21784.
34. D. N. Pei, L. Gong, A. Y. Zhong, X. Zhang, J. J. Chen, Y. Mu and H. Q. Yu, Nat. Commun., 2015, 6, 8696.
35. H. G. Yang, C. H. Sun, S. Z. Qiao, J. Zou, G. Liu, S. C. Smith, H. M. Cheng and G. Q. Lu, Nature, 2008, 453, 638.
36. S. Lu, J. Yu and M. Jaroniec, Chem. Mater., 2011, 23, 4085.
37. H. G. Yang, G. Liu, S. Z. Qiao, C. H. Sun, Y. G. Jin, S. C. Smith, J. Zou, H. M. Cheng and G. Q. Lu, J. Am. Chem. Soc., 2009, 131, 4078.
38. G. Kreese and J. Hafner, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 558.
39. G. Kreese and J. Furthmuller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169.
40. G. Kreese and J. Furthmuller, Comput. Mater. Sci., 1996, 6, 15.
41. J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R. Pederson, D. J. Singh and C. Fiolhais, Phys. Rev. B: Condens. Matter Mater. Phys., 1992, 46, 6671.
42. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865.
43. A. Fahmi, C. Minot, B. Silvi and M. Causa, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 11717.
44. C. J. Horward, T. M. Sabine and F. Dickson, Acta Crystallogr., Sect. B: Struct. Sci., 1991, 47, 462468.
45. A. Vittadini and M. Casarin, Theor. Chem. Acc., 2008, 120, 551.
46. X. Q. Gong and A. Selloni, J. Phys. Chem. B, 2005, 109, 19560.
47. A. Hussain, J. Gracia, B. E. Nieuwenhuys and J. W. Niemantsverdriet, ChemPhysChem, 2010, 11, 2375.
48. V. M. Bermudez, J. Phys. Chem. C, 2011, 115, 6741.
49. S. Grimme, J. Comput. Chem., 2006, 15, 1787.
50. J. R. B. Gomez, J. P. Prates Ramalho and F. Illas, Surf. Sci., 2010, 604, 428.
51. P. Mori-Sánchez, A. J. Cohen and W. Yang, Phys. Rev. Lett., 2008, 100, 146401.
52. B. J. Morgan and G. W. Watson, Phys. Rev. B: Condens. Matter Mater. Phys., 2009, 80, 233102.
53. E. Finazzi, C. D. Valentin, G. Pacchioni and A. Selloni, J. Chem. Phys., 2008, 129, 154113.
54. V. I. Anisimov, F. Aryasetiawan and A. I. Lichtenstein, J. Phys.: Condens. Matter, 1997, 9, 767.
55. B. Himmetoglu, A. Floris, S. Gironcoli and M. Cococcioni, Int. J. Quantum Chem., 2014, 114, 14.
56. M. García-Mota, A. Vojvodic, F. Abild-Pedersen and J. K. Nørskov, J. Phys. Chem. C, 2013, 117, 460.
57. J. Heyd, G. E. Scuseria and M. Ernzerhof, J. Chem. Phys., 2003, 118, 8207.
58. Y. Ortega, D. F. Hevia, J. Oviedo and M. A. San-Miguel, Appl. Surf. Sci., 2014, 294, 42.
59. T. Yamamoto and T. Ohno, Phys. Chem. Chem. Phys., 2012, 14, 589.
60. B. J. Morgan and G. W. Watson, J. Phys. Chem. C, 2010, 114, 2321.
61. A. G. Thomas, W. R. Flavell, A. R. Kumarasinghe, A. K. Mallick, D. Tsoutsou, G. C. Smith, R. Stockbauer, S. Patel, M. Grätzel and R. Hengerer, Phys. Rev. B: Condens. Matter Mater. Phys., 2003, 67, 035110.
62. A. G. Thomas, W. R. Flavell, A. K. Mallick, A. R. Kumarasinghe, D. Tsoutsou, N. Khan, C. Chatwin, S. Rayner, G. C. Smith, R. L. Stockbauer, S. Warren, T. K. Johal, S. Patel, D. Holland, A. Taleb and F. Wiame, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 75, 035105.
63. C. De La Cruz and N. Sheppard, J. Chem. Soc., Faraday Trans., 1997, 93, 3569.
64. A. Kokalj, A. Dal Corso, S. Gironcoli and S. Baroni, J. Phys. Chem. B, 2006, 110, 367.
65. By further resolving the p-states of C=C molecular orbital into its components (Fig. 5a), we can observe that it is not only the HOMO–LUMO of the ethylene (p_z-state) that hybridize with Ti 3d-states, but also the p_x and p_y. Hence resulting in an overall strong resonance between the p-states of C=C bond of ethylene with the adjacent two Ti 3d-state atoms near O_vac.
66. LDOS projected onto each orbital components of both Ti 3d and ethylene C=C molecular orbital shows this localized state is formed by the mixing of derived-bonding of C₂H₄ with the d_yz, d_x2–y2, and d_zz orbitals of the Ti 3d defect states (Fig. 5a and b). It also can be seen that d_yz orbital is the most dominant interacting state as compared to d_x2–y2, and d_zz. Two factors are the origin: (i) excess electrons from O-vacancy mostly occupy d_yz-orbital, (ii) symmetry-wise, d_yz-orbital has good compatibility with –in particular– p_z-anti bonding orbital of ethylene's CC.
67. N. H. Linh, N. T. Quang, W. A. Diño and H. Kasai, Surf. Sci., 2015, 633, 38.
68. K. Zhu, Y. Xia, M. Tang, Z. Wang, I. Lyubinetsky, Q. Ge, Z. Dohnálek, K. T. Park and Z. Zhang, J. Phys. Chem. C, 2015, 119, 18452.
69. K. Zhu, Y. Xia, M. Tang, Z. Wang, B. Jan, I. Lyubinetsky, Q. Ge, Z. Dohnálek, K. T. Park and Z. Zhang, J. Phys. Chem. C, 2015, 119, 14267.
70. M. M. Islam, M. Calatayud and G. Pacchioni, J. Phys. Chem. C, 2011, 115, 6809.
71. U. Aschauer and A. Selloni, Phys. Chem. Chem. Phys., 2012, 14, 16595.
72. H.-Y. T. Chen, S. Livraghi, E. Giamello and G. Pacchioni, ChemPlusChem, 2016, 81, 64.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra13633h

---