Enhanced Photocatalytic CO2 Reduction with Defective TiO2 Nanotubes Modied by Single-Atom Binary Metal Components

A binary component catalyst consists of single atoms (SAs- Pt and Au) anchored on self-doped TiO 2 nanotubes (TNTs), was developed for photocatalytic CO 2 reduction. The defects introduced TNTs substrate was stabilized with atomic Pt and Au via strong metal support interactions (MSI), due to which, the covalent interactions of Pt-O and Au-Ti facilitated an effective transfer of photo-generated electrons from the defective sites to the SAs, and in turn an enhanced separation of electron–hole pairs and charge-carrier transmission. The Pt-Au/R-TNTs with 0.33 wt% of SA metals, exhibited a maximum of 149 times higher photocatalytic performance than unmodied R-TNT and a total apparent quantum yield (AQY) of 17.9%, in which the yield of CH 4 and C 2 H 6 reached to 360.0 and 28.8 µmol g − 1 h − 1 , respectively. The metals loading shifted the oxidation path of H 2 O from •OH generation into O 2 evolution, that inhibited the self-oxidization of the photocatalyst. of the photocatalyst. The excellent reproducibility of photocatalytic CO 2 reduction was maintained due to an effective controlling of •OH generation, that has also been achieved by the carbonate termination. The present work takes a step forward in the understanding of light-induced charge transfer processes between SAs metals and semiconductor support. New opportunities of binary SACs for constructing high-performance photocatalytic system can be explored based on the results established in the present study.


Introduction
Trapping and recycling of CO 2 by means of photocatalytic route is found to be an effective strategy in coping up with the present challenge of increasing severe greenhouse effect. CO 2 reduction using heterogeneous photocatalysis is a well proven technique because of its ability to directly convert the solar energy into high-value products such as methane, methanol, formaldehyde, formic acid, urea, carbonated vinegar and polymers, etc. 1,2 . However, for a complete realization, e cient adsorption of CO 2 and su cient energy input are required in activating the linear symmetric CO 2 molecule which is thermodynamically challenging one due to high bond energy (750 kJ/mol) of C=O 3,4 . TiO 2 is widely used for photocatalytic reduction of CO 2 owing to its suitable band structure and e cient photoelectric properties 5 . Besides the reduction of CO 2 , water oxidation is allowed to take place that is actually crucial for the conversion of CO 2 into fuel molecules. In spite of all, surface of the TiO 2 is featureless to establish an effective chemisorption with CO 2 , that results in CO 2 of high activation energy and extremely poor reaction dynamics 6 . Further, the photocatalytic performance of TiO 2 is being severely limited by ine cient utilization of light source and rapid recombination of photogenerated electron-hole pairs.
Single-atom metal catalysts (SACs) have been of great attraction in the past years due to a maximum utilization of atom e ciency and a well proven catalytic performance 7,8 . Since the adsorption of reactant molecules takes place through the coordinating unsaturated sites, metal single-atom can effectively reduce the activation energy of the reaction 9 . Bringing the advantageous features of SACs and the light utilization ability of the semiconducting metal oxide together are an intriguing approach for realizing high performance and possibly a better activity in the photocatalytic CO 2 reduction process. However, due to the high surface free energy of the single atoms (SAs) metal, they tend to get aggregated during the photocatalytic process 10 , thereby a decelerated reactions. The defective substrate supported SAs prevents the migration of metal atoms and stabilizes the active centers, in which the defective sites anchor the SAs metal via the metal-support interactions (MSI), that result in much reduced surface free energy of the metals and pretty stable defects of the support as well 11,12 . The SAs metal would accept the photo-generated electrons of defective sites and directly trigger the reaction, by which an inhibited charge carrier recombination and in turn an enhanced photocatalytic activity are possible 13 . Furthermore, the MSI of metal-defect is expected to induce substantial electronic rearrangement between the dispersed metal and the support, which would derive new optical and electronic properties 14,15 . Also, the electronic rearrangement could modulate d-band centers and alter the coordination of the metals, which would affect the molecules chemisorption on the catalysts, thus a constant enhancement in the activity could be possible 16,17 .
Otherwise, as per theory, due to the indirect relativistic orbital stretching effect (IROSE), the d-electrons of both Pt and Au susceptibly spark chemical reactions compared to that of the other metals 18 . Attempting for a better couple of metal elements, the bimetallic systems such as CuPt, AgPt, AuCu, AuAg, and AuPt would harmonize local electronic structure of active sites, which could obviously exhibit a much more excellent activity towards CO 2 reduction than monometallic systems 19,20 . In spite of these advantages of the bimetallic clusters and nanoparticles, there have been only a few attempts on binary component involving single atom metal as cocatalysts for photocatalytic CO 2 reduction till date. In particular, mechanism of the enhancement observed in the photocatalytic activity is still ambiguous, further, the migration and utilization of photo-generated carriers due to the bonding of each SAs with the support should be investigated.
In this context, to establish a state of art photocatalyst for CO 2 reduction, the defects were introduced into TiO 2 nanotubes thin lm initially by an electrochemical self-doping approach, and the atomic Pt and Au were anchored on the defects of the TiO 2 nanotubes support by an electrochemical deposition step. A binary component of Pt and Au deposited SAC having high degree of dispersion and outstanding performance on CO 2 reduction, was fabricated. A series of photo-electrochemical characterization studies indicated that the developed covalent interactions of Pt-O and Au-Ti facilitated an effective transfer of photo-generated electrons from the defective sites to the SAs of Pt, and Au, that led to an enhanced separation of electron-hole pairs and in turn effective utilization of photo-generated electrons for CO 2 reduction. The ndings of the present work would afford new possibilities for fabricating a binary SACs and its versatile applications.

Characterization of catalysts.
The morphology and the lattice orientation of the catalysts were characterized by eld emission scanning electron microscopy (FE-SEM), High-resolution transmission electron-microscopy (HR-TEM) and X-ray diffraction (XRD) techniques ( Supplementary Figures 1-2). Upon self-doping, the anatase TiO 2 nanotube arrays (TNTs) sample was found to be rich with surface OVs and Ti 3+ defects as seen in our earlier work 21 . For the metal deposited composite catalysts, no characteristic peaks for both Pt and Au were seen due to high degree of dispersion or trace quantity of the metals 12,22 . It was further characterized by aberration-corrected high-angle annular dark-eld transmission electron microscopy (HAADF-STEM) to spot out the supported metal sites of the TNTs. As appeared in Fig. 1a, the bright spots seen on the TNTs are due to a severe aggregation of the Pt and Au that were nanoparticles larger than 2 nm in size, while the aglare bright spots (Fig. 1b) on the lattice surface of R-TNTs, were the atomically anchored Pt and Au, originated from the strong Z-contrast character 23 . As appeared in Fig. 1c of the energy dispersive X-ray spectroscopy (EDS) mapping, the atomic scale Pt and Au were evenly distributed throughout the support. While for the Pt-Au/TNTs, they obviously re ected the aggregated states (Supplementary Figure 3). As listed in Supplementary Table 1, the inductively coupled plasma optical emission spectrometer (ICP-OES) analysis of the composite catalysts, indicated that the mass contents of the Au and Pt in the range of 0.15% to 0.31%.
On account of the high surface free energy, the individual metal atoms are highly mobile and form aggregates during synthetic process 10 . While on the self-doped support of R-TNTs, they would normally get accommodated on the sites of defects, resulted in a decreased quantity of Ti 3+ and OVs (Fig. 1d).

Interaction between metals and support
As seen in Fig. 2a, a prominent broad protuberance of light absorption, as a result of localized surface plasmon resonance (LSPR) effect of Au 24 , was observed for the both Au/TNTs and Pt-Au/TNTs composites in the range of 400~600 nm. While for the Au/R-TNTs and Pt-Au/R-TNTs, the absorption range was shifted upward featurelessly due to the defects but not the LSPR absorption. The LSPR effect of nano-scale Au that exhibits as the absorptive spectrum of near-infrared through the near-ultraviolet was well documented, which is mainly because of intraband transitions between the outermost electrons within the Au 6s1p hybridized atomic orbitals. As the size gets further decreased, the energy level of the hybridized orbitals would gradually get discrete and downward, and eventually merge with Ti atom of the support and lose the resonance feature 25 . Hence, corresponding the LSPR absorption gets gradually weakened and moved toward high energy region, eventually goes disappearing when the size is smaller than 2 nm 26 .
In the anatase phase of TiO 2 , the vibration modes of Ti-O bonding showed characteristic Raman signals (Supplementary Figure 4) at ~141 (E g (1)), ~390 (B 1g ), ~515 (A 1g ) and ~634 (E g (2)) cm -1 , while introducing the OVs, the length of the bonding was altered, consequently a slight symmetric shift of 4 cm -1 was observed for the R-TNTs 27 . Furthermore, the Raman intensity of the Ti-O in Pt-Au/TNTs was signi cantly enhanced compared to that in TNTs. Under light irradiation, photo-induced electrons generated from nano-size Au, then migrated to TiO 2 and inelastically collided with the electrons presented there, which boosted the vibration of Ti-O and in turn an enhancement of surface Raman scattering 28 .
While for SAs modi ed semiconductor catalyst, the LSPR effect was no longer presented which could be a reason for the similar Raman features observed for both the Pt-Au/R-TNTs and R-TNTs as well.
As seen in the XPS 4f spectrum of the Pt (Fig. 4b) and Au (Fig. 4c), the absorption edge position was located between 0 to +4 valence for Pt n+ (n, 0~4) species and 0 to +1 valence for Au δ+ (δ, 0~1) species, respectively. Further, compared to Pt-Au/TNTs, an up-shift 0.34~0.36 eV of bonding energy (BE) was observed for both Pt and Au in the Pt-Au/R-TNTs composite, which was identical with that for monometal supported SACs in Pt/R-TNTs and Au/R-TNTs composites (Supplementary Figure 5). As the size decreased, the core-shell screening between the metal atoms would get weaken, which re ected the increase of core-level BE, namely a nal-state effect of the species 29  and O 2p orbitals 30 , therefore, the Pt species should exist as compounds. As far as the normalized Ti Kedge XANES (Fig. 2e), the three prepeaks (4960~4980 eV, inset of the Fig. 2e) observed against both R-TNTs and SAs Pt-Au/R-TNTs, re ected the higher intensity and non-centrosymmetric characteristics unlike for the perfect TiO 2 by which the presence of defects near Ti atom was con rmed. Furthermore, the existence of three Ti-O shells (R = 1.12, 1.54, and 2.12 Å) 31 was con rmed by the fourier transform extended X-ray absorption ne structure (FT-EXAFS) of Ti (Fig. 2f), wherein the length of those bonds for R-TNTs sample was found to be shortened due to the introduction of OVs 32 . Upon anchoring the Pt and Au, two of the shortened Ti-O bondings in the R-TNTs got resumed, the bond length even exceeded for perfect TiO 2 , however, since relative instability of the third Ti-O shell (R = 2.12 Å), of which bonding interaction might be substituted by the MSI, a total disappearance of signal was observed. The introduced metals could trap electrons from defects and make bonding with the uncoordinated defects Photocatalytic reduction of CO 2 and photoelectric properties of the catalysts To evaluate the performance of the fabricated catalyst composite on CO 2 reduction, several control experiments were initially conducted, that included (i) treating the gaseous mixture of CO 2 and H 2 O with the catalyst under dark condition; (ii) irradiating the gaseous mixture without catalyst; (iii) irradiation of the (i) but without H 2 O, and (iv) irradiation of the (i) but without CO 2 . No hydrocarbons were found to be formed in all the mentioned cases even at an extended reaction period of 3 hours, proving that the sources of carbon and proton were only from the input of CO 2 and H 2 O, respectively.
The yields of photocatalytic CO 2 reduction by using the SAs Pt-Au/R-TNTs, Pt-Au/TNTs, and R-TNTs composites under illumination, were shown in Fig. 3a-b and their corresponding products were listed out in Supplementary Table 1. As seen, the products viz. C 2 H 6 and CH 4 were formed by the SAs Pt-Au/R-TNTs composite catalyst and their maximum rate of generation were shown to be 28.8 and 360.0 µmol g -1 h -1 , respectively, for which the corresponding AQYs were of 2.7 and 15.2 % (total value of 17.9%). The photocatalysis of 13 CO 2 isotope labeling further con rmed the reduction products, for which the signals (Fig. 3c) of the isotopic 13 CH 4 and 13 C 2 H 6 were seen at m/z=17 and 32, respectively, and their fragment ions were also observed at m/z = 15, 16, 30, and 31 35,36 .
A much higher photocatalytic activity towards CO 2 reduction was observed with SAs Pt-Au/R-TNTs composite, for instance, it was 5.5 and 149 times higher than that of Pt-Au/TNTs and R-TNTs, respectively. Interestingly, the synergistic effect was found with a binary combinations of atomic Pt and Au, in which 2 times higher rate of CO 2 reduction was seen compared with the monometallic system  Figure 10). If this was the case, scouring the surface with deionized water should have restored the activity by recovering the active sites, but the restoring was only partial (Supplementary Figure 9b). The deactivation phenomenon might be originated from the sacri cial oxidation of the catalyst caused by photo-induced holes. To verify the self-oxidation step, a control experiment was conducted by replacing the water gas with 10 vol % mixture of CH 3 OH/H 2 O. Interestingly, the activity of the photocatalyst was retained with a higher durability for the hydrocarbons yield (Supplementary Figure 9c). The catalyst underwent deterioration by the self-oxidation, thereby a decelerated photo-reduction of CO 2 . During the process, -OH groups that terminated on the surface of TiO 2 , were oxidized into •OH, which could in turn oxidize the catalyst itself. However, by substituting CO 3 2-with -OH, the durability and activity of the photocatalysis were high and stable for at least 4 runs (Fig.   3d) of the process. Because the CO 3 2would get oxidized into O 2 but not the •OH, thus the self-oxidation of the catalyst was controlled 38 .
Excluding the slight shift of the energy band gap of the catalysts, the introduction of defects and loading of SAs had major impact of developing an integrally enhanced system for absorption of the light in the visible and near-infrared regions (Fig. 4a), as well as extended lifetime of the carriers. As seen in Fig. 4b orbital into Pt 5d orbital, or via Ti 3d orbital into Au 6s orbital 40,41 . The MSI that existed between SAs of Pt/Au and the defects had widen the spectrum of absorptive region, extended the lifetime of the carriers and channelized the electron migration from the defective sites into the SAs metals for which could be an excellent electronic collector and in turn an e cient photocatalysis for CO 2 reduction. Furthermore, electrochemical impedance spectroscopy (EIS) for the SAs Pt-Au/R-TNTs (Fig. 4e) showed a representative Nyquist plot with the smallest arc radius among the catalysts under irradiation, the effective charge separation and transfer could also be a cause for the high performance. , 1320 and 1405 cm -1 ) and bicarbonate (HCO 3 -, 1635 cm -1 ) (Fig. 9a-b) 42 .

Mechanism of CO 2 adsorption and activation
During the CO 2 adsorption process, several distinguishable IR vibrating bands were observed at about 1140 and 2078 cm -1 corresponding to the species HCOOand CO, respectively. It was understood that the catalyst was capable enough of harvesting infrared lights 43 , which has been con rmed in the experiments of our ongoing works. Since the elimination of adsorbed species such as OH, CO 3 2-, HCO 3 -, and HCOO -, was not possible by purging the catalyst surface with He gas, the adsorption should be of chemical interactions (Supplementary Figure 11a-b). and all these were found to increase with the illumination period ( Fig. 9c-d). As the peaks for CO disappeared, it was understood that the species are converted into highly stable intermediates. It is worth noting that, though most of the earlier works have found the carbonates accumulation 45,46 , the present study did not encounter any such accumulation due to their facile conversion with the SAs Pt-Au/R-TNTs composite.
Generally, the conversion of CO 2 into CO 2 • − is considered as an initiation and a rate-limiting step of the reduction process 47,48 . As shown in Fig. 10a On the basis of the results obtained in the in-situ DRIFTS and EPR spectroscopy studies, the mechanism of alkane (CH 4 and C 2 H 6 ) evolution with SAs Pt-Au/R-TNTs was proposed as Scheme 1. At rst, a large amount of the CO 2 molecules was adsorbed on the surface of the catalyst and underwent a reaction to produce CO 2 • − under illumination (path 1), followed by an addition reaction with H to generate COOH* (path 2) 50  In summary, a state of the art composite catalyst of SAs Pt-Au/R-TNTs, was developed by anchoring of atomic scale Pt and Au on the OVs of self-doped TiO 2 nanotubes support. The binary components were dispersed on the surface of the catalyst via strong MSI. The as-prepared SAs Pt-Au/R-TNTs with a composition of 0.33 wt % of SAs metals, performed extremely well on the photocatalytic CO 2 reduction, in which the CO 2 molecule was initially protonated to form •CH 3 , and converted further into CH 4 and a C-C coupled product of C 2 H 6 with the AQY of 15.2 and 2.7 %, respectively. The e ciency was about 5.5 and 149 times higher than that of the Pt-Au/TNTs and R-TNTs, respectively. The remarkable performance was ascribed to the signi cant enhancement in separation of photo-generated electron-hole pairs and chargecarrier transmission by the MSI of Pt-O and Au-Ti covalent bonding. Furthermore, the pathway of photocatalytic H 2 O oxidation changed from •OH generation to O 2 evolution upon loading of Pt and Au, resulting in inhibition of the self-oxidization of the photocatalyst. The excellent reproducibility of photocatalytic CO 2 reduction was maintained due to an effective controlling of •OH generation, that has also been achieved by the carbonate termination. The present work takes a step forward in the understanding of light-induced charge transfer processes between SAs metals and semiconductor support. New opportunities of binary SACs for constructing high-performance photocatalytic system can be explored based on the results established in the present study.

Methods
Chemicals.

Preparation of TNTs.
The self-organized TiO 2 nanotubes (TNTs) were prepared by electrochemical anodization technique reported in our earlier work 55 . Titanium foil used in the anodization experiment was thoroughly cleaned by sequential ultrasonication using a mixture of acetone and ethanol, followed by a chemical polishing with a solution containing HF/HNO 3 /H 2 O at a volume ratio of 1:3:6 for a period of 1 min. It was nally rinsed with deionized water and dried in a nitrogen stream. The well-cleaned Ti foil was anodized with a two-electrode system using Pt mesh as cathode at a constant potential of 60 V for 8 h. The electrolyte used was a mixture of ethylene glycol with 0.25 wt % of NH 4 F and 12.5 wt % of H 2 O. The gap between the electrodes was xed as 2.5 cm, and the electrolyte temperature was maintained constantly at 25 °C using a thermostat. The as-prepared TNTs were calcined at 450 °C in ambient air for a period of 2 h, followed by a self-doping process by an electro-reduction step.
A similar procedure of Pt/TNTs fabrication, was adopted for SAs Pt/TNTs with a replacement of TNTs by R-TNTs support.
A similar synthetic procedure of Pt/TNTs, was adopted for the fabrication of Au/TNTs with a change in the operating potential, which was repeated for six runs to step at -0.2 V vs. SCE for 5 s, then back to 0.2 V for 5 s in the 0.3 mM HAuCl 4 electrolyte.
The synthetic procedure of SAs Au/TNTs was similar to Au/TNTs, except for using R-TNTs as support instead of TNTs.
The SAs Pt-Au/R-TNTs catalyst was obtained by similar deposition of both Pt and Au for 3 runs each.
The carbonate modi cation on the surface of SAs Pt-Au/R-TNTs was carried out by immersing the catalyst in 10 mL of 0.1M Na 2 CO 3 aqueous solution for a period of 18 h at 298 K.
Reactor setup and photochemical measurement.
The photocatalytic reduction of CO 2 was carried out in a 15 ml quartz cell reactor with a magnetic stirrer placed at the bottom of the reactor to circulate the gas in it. The photocatalytic lm materials were prepared with a dimension of 1 cm × 2 cm. The CO 2 photoreduction was carried out under LED light source (365±10 nm) irradiation and the distance kept between the surface of the photocatalyst and the LED lamp was 1 cm. Before each experiment, the reactor was purged with high-pure CO 2 bubbles through a gas-washing bottle containing aqueous solution for 30 min. And the ow rate of gas was maintained at 50 ml/min. The average irradiation intensity of 365 nm LED (0.208 Wcm -2 ) was determined by using spectroradiometer. The AQY was calculated by the following equation 56 : Where [N(Hydrocarbon)×n] and N(photos) are the number of electrons reduced from the CO 2 molecule into hydrocarbon and incident photons, respectively.
The gaseous products were separated in HP-PLOT/Q and MolSieve 5A column, measured by thermal conductivity detector (TCD) and ame ionization detector (FID) with a gas chromatograph system. The TCD temperature was xed at 200 °C and the oven at 45 °C. The carrier gas used for quanti cation of H 2 , O 2 and CO, was Ar with a ow rate of 4.2 mL min −1 . The FID temperature was xed at 200 °C and the oven at 70 °C. The carrier gas used for the analysis of hydrocarbons was Ar with a ow rate of 10 mL min −1 .
The products of CO 2 isotope experiment, viz. CH 4 and C 2 H 6 , were detected by gas chromatograph-mass spectrometer (GC-MS) equipped with SH-Rtx-wax column. The oven temperature was maintained at 70 °C and the carrier gas was He with a ow rate of 10 mL min -1 .

Characterizations
The morphology of the catalysts was analysed by using a FE-SEM operating at an accelerating voltage of 10 kV. HR-TEM images were recorded at an acceleration voltage of 300 keV. The phase and crystal structure of the catalysts were examined by XRD equipped with Cu Kα radiation (40kV, λ=1.5406Å) and a Raman spectroscopy equipped with an argon ion laser at 532 nm. The optical properties of the catalysts were studied by UV−vis DRS, PL, and ATR-FTIR spectroscopy, techniques. Time-resolved uorescence emission decay spectra were recorded at 525 nm using excitation with a light pulse of 325 nm by a HORIBA Scienti c DeltaPro uorimeter. By using XPS, the surface elemental analysis were carried out based on the C 1s peak at 285.0 eV. The radical species generated were analyzed by EPR measurement.
The metal analysis was carried out by ICP-OES.
Hard XAFS measurement was recorded at the XAS station (BL11) of the Saga Synchrotron Light Research Center, Japan. The Pt and Au L 3 -edge XANES data were recorded in a uorescence mode. The hard X-ray was monochromatized with Si (111) double-crystals. Ti K-edge XANES was recorded in transmission mode.
Electrochemical impedance spectroscopy (EIS) was carried out between 0.1 MHz and 0.01 Hz with 10 mV AC amplitude at −0.05 V DC potential by using an electrochemical station (CHI760E, US) of threeelectrode system. The fabricated catalysts, SCE and Pt mesh were xed as working, reference, and counter electrodes, respectively. The electrolyte used was 0.1 M Na 2 SO 4 aqueous solution. A quartz cell was used as reactor for photocurrent test in which 50 mL electrolyte was placed and the reactor set-up was kept 5 cm apart from the 365 nm LED light source. The area of working electrode under illumination was 1 cm 2 .
In-situ DRIFTS measurements were conducted by the TENSOR II FT-IR spectrometer equipped with an insitu diffuse-re ectance cell. The reaction chamber was equipped with three gas ports. A mixture of highpure He, high-pure CO 2 , and H 2 O vapor could pass into the reaction chamber in which the ux of the target gas (CO 2 and H 2 O mixture) and purge gas (He) was controlled by a three-way ball valve system.
The chamber has three dome windows, each of which used for separate purpose viz. IR light source, analytical detection, and photocatalyst illumination. And a high pressure mercury lamp was used as UV light source.