Oxygen-Cluster-Modified Anatase with Graphene Leads to Efficient and Recyclable Photo-Catalytic Conversion of CO2 to CH4 Supported by the Positron Annihilation Study

Anatase TiO2 hollow nanoboxes were synthesized and combined with the graphene oxide to get nanocomposite of TiO2/rGO (TG). Graphene oxide was used to modify the Oxygen-Clusters and bulk to surface defects. Anatase and TG composite were characterized with the positron annihilation, XPS, EPR, EIS and photocurrent response analysis. The relative affects of defects on the photocatalytic reduction (CO2 to CH4) were studied. The TG composites showed highest photo-catalytic activity after GO coupling (49 µmol g−1 h−1), 28.6 times higher photocurrent yields much higher quantum efficiency (3.17%@400 nm) when compared to the TiO2 nanoboxes. The mechanism of enhanced photo-catalytic CO2 conversion to CH4 elucidated through electrochemical and photo-catalytic experiments with traceable isotope containing carbon dioxide (13CO2). For the first time we discovered that diminishing the comparative concentration ratio of anatase from the bulk to surface defects could significantly increase the conversion of CO2 to CH4.

The carbon dioxide (CO 2 ) emission is one of the serious environmental problems, therefore its reduction and sustainable conversion to fuels by using solar energy is highly desirable 1 . Titania (TiO 2 ) is a chemically stable and biologically benign photo-catalyst, capable to convert harmful pollutants, particularly CO 2 reduction to fuel CH 4 2 . Beside the well-known bulk defects, the surface defects are more sensitive to the photocatalytic activities. The surface defects in a pure crystal can act as e-h trap-sites along with the adsorption sites on the other hand the bulk defects act only as charge carrier traps where e-h are more likely to recombine. Therefore, there is great scientific importance to understand both the bulk and surface defect states. This study highlights the defect density effects on the photocatalytic activities, which may lead to the innovative design of photocatalysts. The defects in TiO 2 have been experimentally studied by the scanning tunneling microscopy (STM) 3 , electron paramagnetic resonance (EPR) 4 , time-resolved photoluminescence spectroscopy (PL) 5 etc. These techniques allow understanding the effects of surface or subsurface defects, thereby helped to correlate the adsorption and surface reactivity.
The surface of graphene oxide consists of oxygen and hydroxyl functional groups capable to form new bonds with the TiO 2 . When combined with graphene, the composite provides an effective approach to improve the photo-catalytic efficiency and light response. This occurs because graphene layers with 2D sp 2 -hybridized carbon atoms (pi-bond conjugated structure) have superior electronic properties and large specific surface area that significantly enhances the photo-catalytic properties 6,7 . Compared with the other defect-probes techniques, positron

Results and Discussion
Structural characterization and surface composition. The XRD spectra of TiO 2 , and TG composite  Fig. 1a(ii)) 8 . The Raman spectra of composites ( Supplementary  Fig. 1b) exhibit typical features of rGO with D-band at 1354 cm −1 and G-band at 1611 cm −1 . The remaining four peaks from 100-700 cm −1 belong to the TiO 2 . When the geometric size of TiO 2 crystal approaches to nanometer range, the E g Raman active mode blue-shifts to 150 cm −1 (instead of 143 cm −1 ) for the bulk phase 9 . The other specific vibration modes at 400 cm −1 (B 1g ), 523 cm −1 (B 1g + A 1g ) and 643 cm −1 (E g ) are characteristic peaks of the anatase phase.
The C1s XPS spectra indicated the existence of chemical binding between TiO 2 and rGO (details in Supplementary information D1). Compared with the blank GO, the new peak at 285.5 eV is ascribed to the Ti-O-C bonding. Therefore there is covalent bonding between the TiO 2 and rGO. This is further confirmed by the significnt decrease in the FTIR peks at 454, 668 and 803 cm −1 . The bright-field TEM (Fig. 1a) of TiO 2 indicated hollow box like structures with narrow size distribution and good morphological uniformity. The TiO 2 nanoboxes on graphene sheets (Fig. 1b) showed well dispersed NPs on both the graphene sides. The reduced graphene oxide is transparent with wrinkles and hollow TiO 2 structures (HRTEM, lattice spacing of 0.35 nm, Fig. 1c), confirming the (101) planes of the anatase TiO 2 . The SAED pattern (Fig. 1d) revealed a [010]-axis orientation, and the lateral

Defect characterization by PALS, XPS and EPR.
Previous reports showed oxygen defects on the surface of TiO 2 raise the local-states under the edge of the conduction band. Oxygen defects densities with one or two electrons are generated due to the missing oxygen atom in the bulk or on the surface, while the Carbon doped TiO 2 forms deep impurity states in the band gap 1-3,10,11 . We observed (UV-vis spectrum, Supplementry Fig. 2) that bandgap of TiO 2 (3.2 eV) significantly decreased (2.6 eV) with increased absorption after hybridization with GO.
In this section, we used PALS (positron annihilation lifetime spectroscopy) to quantify the defect concentration obtained by the comparison of TiO 2 and TG composites. The Table 1 shows three positron lifetime components, τ 1 , τ 2 , and τ 3 , with relative intensities I 1 , I 2 , and I 3 for the TiO 2 nanoboxes and TG composites. The shortest positron lifetime τ 1 (0.1985 ns) can be assigned to the bulk defects with free annihilation of positrons inside defect-free crystal 12 . Where as the longer lived positrons τ 2 (0.3937 ns) can be attributed to the surface defects originating from the positrons trapped by the larger defects such as oxygen vacancies or clusters, homodimers, trimers or larger 10 . The second positron lifetime components (τ 2 ) for the TiO 2 nanoboxes and TG composites are much higher than their respective first positron lifetime components (τ 1 ). The variation of τ 2 value from 0.3937 ns to 0.3864 ns was evident that the TiO 2 nanoboxes contain an abundance of surface oxygen defects and compressive bonding but reduced to atomic dimensions in the TG composites. The largest component (τ 3 ) is assigned to the micropores and to the annihilation of orthopositronium atoms 11 . Besides the lifetime factor of the positron, its relative intensity (I) gives information regarding the relative concentration of the defects 13 . The coupling of graphene oxide with TiO 2 leads to the decreased I 1 /I 2 (1.23 to 1.03), thereby clearly indicates the relative bulk to surface defects concentration decrease.
The samples were further investigated by the XPS to get information about the chemical composition, oxygen vacancies and Ti oxidation-state inside TiO 2 nano-boxes after microwave treatment with the rGO. The O1s spectrum of TiO 2 nanobox (Fig. 2a) shows a peak at the binding energy of 529.3 eV, corresponding to the Ti(IV) Sample τ 1 (ns) τ 2 (ns) τ 3 (ns) I 1 (%) I 2 (%) I 3 (%) I 1 /I 2  www.nature.com/scientificreports www.nature.com/scientificreports/ bound oxygen (O 2− ). The peak at 531.6 eV can be attributed to the high binding energy component originated from the loss of oxygen (oxygen vacancies) 14 . The peak at 532.9 eV (low binding energy component), is due to the adsorption of HO − on the surface. The O1s spectrum of TG composite (Fig. 2c) showed a strong O 2− peak at 529.5 eV. The second peak related to the oxygen vacancies O* (531.6 eV), showed significant suppression after rGO hybridization (microwave treatment). A slight shift (0.2 eV) in the peak position (529.5 eV) for the oxygen anions (Ti-O-Ti) is consistent with the Ti 2p core-level due to hybridization with graphene. Furthermore, inside TG sample the third low energy peak (near 532.9 eV) related to the surface hydroxyl group almost vanished. These observations confirmed that the concentration of surface defects is related with both surface hydroxyl groups and oxygen vacancies were significantly lowered in the hybrid sample 15 . The Fig. 2b,d schematically illustrates oxygen vacancies before and after GO hybridization.
The defects were further analyzed by the EPR (Fig. 3) because of the sensitivity of EPR towards the different defects, caused by the lower or extra electrons present on different atoms (O − , O 2− , Ti 3+ , Ti 4+ ) under different surrounding atoms (spin Hamiltonian, perturbation theory). The EPR spectra for the TiO 2 shows several peaks beside the main peak (g = 2.05). This observation clearly indicates presence of several types of defects. The previously reported EPR spectra indicated that the initial hole-trapper at the lattice of oxide ions at hydrated TiO 2 surfaces are the hydroxyl radicals. These trapped holes are the defect centers termed as bridging O − species 16 . The presence of signal at g = 2.05 with broad underline peak in TiO 2 nanobox (Fig. 3a) represents the existence of significant oxygen vacancies 17 . This occurs because electrons trapped at Ti 4+ sites can form Ti 3+ (XPS, Fig. 2) and holes at subsurface oxide ions can form O −18 . The obtained EPR spectrum for the TG composites (Fig. 3b) shows a strong paramagnetic signal at g = 2.00 with loss of extra sharp-peaks. This represents the healing of oxygen vacancies (smoother spectrum) and other defects. Therefore we concluded that rGO hybridization has remarkably lowered the oxygen vacancies and other defects. These observations were further tested by the hydrofluoric acid (HF) addition experiments. Defects were introduced into TiO 2 by changing the amount of HF (details in Supplementary Information D2). The comparison showed defect introduction by the HF (0, 50 µL) changed the shape of EPR, introduced extra peaks (less smooth EPR) and modified the g value of TiO 2 .
Absorption. The UV-Vis spectra of TiO 2 nano-boxes and composites ( Supplementary Fig. 2) exhibit extended absorption therefore indicates lowered band gap of composites when compared with the standard TiO 2 . The decrease in band gap has been attributed to the formation of C-Ti bond 19 . It is well-known that the extended absorption of light for the TG composite opens the possibility of higher photo-catalytic activity of a given photo-chemical reaction. Furthermore, considering the practical applications, the recyclability of photo-catalyst is another challenge to avoid contamination and higher economic efficiency. The photo-catalyst composites containing magnetite can easily separate from the solution by an external magnet. Therefore, the designing of efficient, magnetic photo-catalytic composite with reliable and stable structure has high importance. The non-toxic SiO 2 coating layer, conjugates well with functional groups, and known to protect Fe 3 O 4 from oxidation thereby increases stability of NP-decorated photo-catalyst. Therefore, we have incorporated these materials to construct  Supplementary Fig. 4a,b).
The comparison of magnetic hysteresis loops related to the Fe 3 O 4 , Fe 3 O 4 @SiO 2 and TFSG composites ( Supplementary Fig. 5) showed room-temperature ferromagnetism for the Fe 3 O 4 and Fe 3 O 4 @SiO 2 and TFSG composites. The results show that there is a decrease in the magnetic saturation of TFSG composite due to the non-magnetic TiO 2 . The saturation magnetization value (M s ) of the TFSG composites is 3.18 emu g −1 , which is lower than those of the Fe 3 O 4 @SiO 2 (57.48 emu. g −1 ) and Fe 3 O 4 (64.03 emu. g −1 ). This value of magnetization is good for the efficient recovery of photo-catalyst.
Photocurrent and charge separation. The transient photocurrent response for the TiO 2 , TG and TFSG composites were compared (Fig. 4a). The photocurrent (photo-electrochemical) response of the TiO 2 nanobox was weak (8.2 × 10 −5 mA.cm −2 ). While the TG composite showed significantly increased photocurrent response 2.344 × 10 −2 mA.cm −2 (28.6 times higher) and 9.786 × 10 −4 mA.cm −2 for the TFSG composites (11.96 times higher). This significant photo-current enhancement attributed to the lower recombination rate of electron-hole pairs and longer lifetime due to the presence of rGO sheets (2D -conjugation structure) as compared to the TiO 2 . The comparison of EIS curves (Fig. 4b) showed the single-semicircles (Nyquist plots) in the high frequency range and a linear behavior in the low frequency region. The lowest charge transfer resistance (R ct ) was observed (4.9 Ω) for the TG composite when compared to the TFSG (6.2 Ω), and TiO 2 (6.5 Ω). Therefore, the TG composite exhibits the lowest R ct , indicating lower charge-transfer resistance resulting in much faster separation of photo-generated species 20,21 .
To verify the carbon detected in CO 2 conversion originates from the decomposition of photo-catalyst or from the added CO 2 gas, we carried out isotopic experiments by using traceable 13 CO 2 under identical photo-catalytic reaction conditions. The products of photo-catalytic reaction were identified and quantified by the GC-MS ( Supplementary Fig. 7). The 13 C, 13 CH, 13 CH 2 , 13 CH 3 and 13 CH 4 molecular fragments were detected. These observations from MS spectrum confirmed that the produced CH 4 indeed originated-from the reduction of CO 2 gas and not from the photocatalyst decomposition.
Photo-catalytic activities. The photo-catalytic activities of TiO 2 nanoboxes, TG composites and TFSG composites samples were evaluated for CO 2 conversion with triethanol amine as the electrons donor (Fig. 5a). The TiO 2 nanoboxes showed weak response for the photo-catalytic CO 2 conversion to CH 4 (~15.0 µmol g −1 h −1 ) after 1 h irradiation. The TG composites demonstrate the highest (~49.0 µmol g −1 h −1 ) activity for CO 2 conversion. Since, the TFSG composites were hybridized with magnetic nanoparticles (Fe 3 O 4 @SiO 2 ) for easier recovery thereby it showed lower CO 2 conversion to CH 4 activity (27.0 µmol g −1 h −1 ) than TG, but still 3.2 times higher than that of TiO 2 nanoboxes (15.0 µmol g −1 h −1 ). The main detected gas products were CH 4 , H 2 and CO (GC analysis). The Fig. 5b shows the quantitative comparison of produced H 2 during this experiment. We further performed blank tests to confirm that CH 4 comes from the reduction of CO 2 . After blank photo-catalytic experiment without CO 2 under the same conditions, we have not detected any CH 4 .
The Fig. 6 shows comparison of time dependent H 2 evolution for the TiO 2 , TG and TFSG composites. This water splitting reaction showed a sustained H 2 release rate of ~18.46 µmol/h for the TG composites. The standard solar light illumination for 8 hours, the hydrogen evolution kept increasing, demonstrating the robust photo-catalytic performance. This observation indicates stable and active photo-catalyst for long term water splitting. After conversion of CO 2 to CH 4 (5 h), samples were collected, washed several times and dried. The TEM results (Details in Supplementary information D3) showed stable composite after 5 h light irradiation.
The rate of H 2 evolution followed the following order: TG composites > TFSG composites > TiO 2 nano-boxes respectively. Quantum efficacies of these samples were calculated (Supplementary information 8) as 3.02%, www.nature.com/scientificreports www.nature.com/scientificreports/ 2.01%, and 0.3%, respectively for the TG, TFSG, and TiO 2 nanobox. In the CO 2 reduction, the photo-generated electrons-holes would migrate to the rGO surface. The conduction band of TiO 2 nano-boxes and the work function of rGO are reported as −4.2 eV and −4.42 eV, respectively 22,23 . The proximity of energy levels is helpful for the photo-generated species transfer from the TiO 2 nano-boxes conduction band to the rGO thereby produces radicals. The reason behind is the adsorption kinetics of the CO 2 (10 −8 to 10 −3 s) on TiO 2 nanoboxes is slower than the e-h recombination time (10 −9 s) 24,25 . The composites increased the lifetime of the charged species that further improved the selective formation of CH 4 gas along with other minor product gases. The photo-catalytic performance supports our proposed interpretations of the CO 2 conversion to CH 4 through reduction mechanism.
As reported the hydroxyl radicals (·OH) are recognized to be important intermediates for CH 4 conversion. The mechanism of charge separation, coumarin fluorescence method (coumarin + • OH → fluorescent 7-hydroxycoumarin) is used to detect the amount of hydroxyl radicals ( • OH) produced via photochemistry. The procedure relies on the basic principle that: the stronger the fluorescence signal being observed, the larger the • OH produced. We further compared the produced • OH for TiO 2 nanoboxes, TFSG and TG composites. The amount of produced • OH in TG and TFSG is larger than the TiO 2 nanoboxes (Fig. 7), with TG composites showing the highest fluorescence intensity. Hence, the TG composites produced highest amount of • OH thereby confirming the excellent photo-electrochemical and photo-catalytic properties.
In summary, we have explored the photo-catalytic reduction of CO 2 on TiO 2 nano-boxes, TiO 2 hybridzed with graphene (TG) and TiO 2 -(Fe 3 O 4 @SiO 2 )-rGO nanocomposite (TFSG). Oxygen clusters along with surface/ bulk defects in TiO 2 nanoboxes play vital role in efficient photocatalysis. Positron annihilation results revealed decreased Oxygen clusters (mono, dimers, trimers) and the relative concentration ratio of bulk/surface defects in TiO 2 nanoboxes efficiently enhanced the photocatalytic mechanism. The photo-catalytic experiments and traceable isotopes containing carbon dioxide ( 13 CO 2 ) analyzed through mass spectroscopy confirmed the source of CH 4 production. The TG composite exhibits 3.2 times (~49.0 µmol g −1 h −1 ) higher photo-catalytic reduction of CO 2 to CH 4 , when compared to the TiO 2 nanoboxes. The TFSG composites retained magnetization, therby provides an easy way for the catalyst recyclability. The XPS study revealed that number of oxygen vacancies significantly decreased after GO hybridization, further confirmed by the EPR result. The excellent electrochemical and photo-catalytic properties were assigned to the lowered defects, proximity of energy levels, much higher