Novel structure of active platinum-bismuth site for oxidation of carbon monoxide

As the technology development, the future advanced combustion engines must be designed to perform at a low temperature. Thus, it is a great challenge to synthesize high active and stable catalysts to resolve exhaust below 100 °C. Here, we report that bismuth as a dopant added to form platinum-bismuth cluster on silica for CO oxidation. The highly reducible oxygen species provided by surface metal-oxide (M-O) interface could be activated by CO at low temperature (~ 50 °C) with a high CO 2 production rate of 487 µmol CO2 ·g Pt−1 ·s − 1 at 110 °C. Experiment data combined with density functional calculation (DFT) results demonstrate that Pt cluster with surface Pt−O−Bi structure is the active site for CO oxidation via providing moderate CO adsorption and activating CO molecules with electron transformation between platinum atom and carbon monoxide. These ndings provide a novel and general approach towards design of potential outstanding performance catalysts for redox reaction.


Introduction
The CO oxidation reaction (CO + 1/2 O 2 = CO 2 ) is a well-known model reaction in heterogeneous catalysis, as well as a key step to resolve automobile exhaust containing CO, NO and hydrocarbons [1][2][3][4][5] . Therefore, higher activity at lower temperatures with good stability, such as platinum-based catalysts, are in great demand. In order to pursue high catalytic performance for CO oxidation, a series of reducible oxides including CeO 2 2 , FeO x 3 , MnO 2 4 and Co 3 O 4 4 were used as active supports for Pt catalysts, due to their rich surface oxygen vacancy and the so-called "strong metal-support interaction" [6][7][8] . In another hand, irreducible oxide of commercial production, low cost and extensive application such as SiO 2 and Al 2 O 3 , usually showed poor ability to disperse active platinum species of ultra-small size (< 2 nm).
Many research groups have found that the addition of a secondary element, such as Sn 5 , K 9 , and Co 10 , distinctly improved the activity of silica-or alumina-supported platinum catalysts, no matter in the form of oxide clusters or metallic alloy for Pt. Zhai et al. prepared partially oxidized Pt-alkali-O x (OH) y species, which served as the active site for low temperature Pt-catalyzed water-gas shift reaction 9 . As for bismuth element, a high oxide ion conductivity dopant, has been reported to promote the transition metal (Co) 11 and noble metal (Pd) 12 catalysts for the CO oxidation reaction. However, the doping of bismuth to platinum usually focus on PtBi alloy or intermetallic compound 13,14 . So, it could be the rst case to determine the formation of platinum-bismuth oxide cluster on silica with uniform Pt-O-Bi structure showing excellent thermal stability.
On the other hand, the precise determination of active site and reaction mechanism is still in arguments 12,15 . This promotion effect has been explained previously that bismuth oxide species could produce more oxygen vacancy to enhance the CO oxidation e ciency 11 . However, the local structure of active site for Bi-promoted catalyst and structure-function relationship in practical catalysts are still complicated, because of limited characterization techniques and complex active site structure. Therefore, we employed the comprehensive characterization methods to investigate the local structure of platinumbismuth cluster catalysts and con rm the active site of Pt−[O] x −Bi structure.
reported iron nickel hydroxide-platinum nanoparticles (Pt-OH-Fe/Ni) were highly e cient for CO oxidation owing to abundant sites of Pt-OH-M interfaces 19 . However, these different interfaces are often unstable and di cult to establish precisely. Therefore, it is a big challenge to build up stable and speci c interface to provide unique catalytic properties.
In this work, we prepared silica-supported platinum-bismuth catalysts via an incipient wetness impregnation, possessing excellent sinter resistance due to the formation of oxidized Pt x Bi y O z cluster.

Results
Structural characterization of bismuth-doped platinum samples. A serial of silica-supported platinum and platinum-bismuth samples were prepared by a co-incipient wetness impregnation method. The bulk concentrations of platinum and bismuth (Pt: 0.8, 0.9 and 0.9 wt.%; Bi: 2.3 and 6.1wt.% for 1Pt-SiO 2 , 1Pt2Bi-SiO 2 , 1Pt5Bi-SiO 2 respectively) are close to these designed numbers, indicating the disposition of bismuth species have no effect on the loading of platinum (Supplementary Table 1). Furthermore, the Bifree and Bi-doped samples have the similar textural properties, such as S BET values and the type of adsorption-desorption isotherms (Supplementary Table 1 and Supplementary Fig. 1), in which we can exclude the physical effect on the following catalytic performance.
Small-size oxide species were extremely stable on silica surface in Bi-promoted samples in aberrationcorrected high-angle annular dark-eld imaging scanning transmission electron microscopy (HAADF-STEM) images (Fig. 1a, Fig. 2b,c and 3b). It illustrates that the addition of bismuth element could suppress the growth of metal/oxide particles, similar to the promotion by alkali ions 9 and silica support shows poor ability to stabilize platinum species. Furthermore, the corresponding aberrationcorrected energy dispersive spectroscopy (EDS) mapping results of 1Pt2Bi-SiO 2 (Fig. 1c) show that platinum and bismuth elements distribute uniformly at the cluster level within the same areas on the surface of SiO 2 without evident separation. Xie et al. also reported that bismuth species was deposited selectively on the Pt particles rather than the carbon support 22 . Meanwhile, the X-ray diffraction (XRD) also con rmed the promotion of bismuth species, in which no obvious Pt/PtO/PtO 2 /Bi/Bi 2 O 3 phase was detected in 1Pt2Bi-SiO 2 , even in a "slow-scan" mode ( Fig. 1d and Supplementary Fig. 4). When bismuth oxide species were deposited on silica separately (2Bi-SiO 2 ), it also stabilized in small-size without any diffraction peaks of Bi/Bi 2 O 3 in XRD pro les (Fig. 1d). However, the observation of sharp Pt peaks (39.7º and 46.2º) veri es the formation of huge Pt particle in 1Pt-SiO 2 . As shown in Fig. 1c, the dopant of bismuth species reaches optimization at 2 wt.% and the formation of broad diffraction peak of Bi 2 O 3 for 1Pt5Bi-SiO 2 . Therefore, bismuth oxide species as a promoter could improve the anti-sintering ability for platinum dispersed on an inert support.
The local coordination structure of platinum and bismuth species. According to aberration-corrected HAADF-STEM images, we can observe the existence of platinum-bismuth oxide clusters. X-ray absorption ne structure (XAFS) technique was applied to clarify the local structure of this cluster in Bi-promoted samples. The X-ray absorption near-edge structure (XANES) of Pt-L 3 edge pro les (   Fig. 2e. To our best knowledge, it is the rst time to observe the formation of uniform platinum-bismuth oxide clusters to suppress the aggregation of Pt species. Catalytic performance of Pt/PtBi-SiO 2 catalysts in CO oxidation. CO oxidation was applied as a model reaction to investigate the role of bismuth-dopant. When the catalysts were pretreated at 300 °C under air, Bi-free and Bi-promoted samples shows almost same CO oxidation activity with complete CO conversion at ~ 220 °C ( Supplementary Fig. 8), may due to poor ability to adsorb CO or overhigh valence of platinum species 24 . However, we found that hydrogen reduction signi cantly enhanced CO oxidation activity for Bipromoted catalysts (Fig. 3a). The temperature of 50% CO conversion dropped off from 165 to 85 °C as the reduction temperature increasing from 0 to 210 °C. Interestingly, a platform appeared as hydrogen reduction at 150 and 180 °C, indicating a structure transformation of active site occurred during the hydrogen reduction compared with fresh Bi-doped catalysts. On the basis of the CO oxidation activity (Fig. 3b), a remarkable promotion to platinum-silica catalysts was observed by the addition of Bi oxide species with similar Pt loading (0.9 wt.%). The catalytic performance reaches the maximum at the dopant of 2 wt.% ( Supplementary Fig. 9), may due to overmuch bismuth oxide species hindering CO adsorption or covering platinum active site. For comparison, pure Bi catalysts (2Bi-SiO 2 ) shows no CO oxidation activity below 160 °C ( Fig. 3b), demonstrating bismuth species are not active site just as secondary dopant to modi cation platinum active site. Furthermore, we collected the kinetic data to compare the inherent catalytic activity. The speci c activity normalized by the platinum amount for 1Pt2Bi-SiO 2 was 487 µmol CO g Pt −1 s − 1 at 110 °C, as active as the reported Pt/CeO 2 catalysts (103−518 µmol CO g Pt −1 s − 1 at 80−130 °C, see Table 1), as well as ten times higher than that of pure Pt catalyst (Supplementary Table 3). For eliminating size-effect on active site, 1Pt-SiO 2 -400 was similarly inactive for CO oxidation at Furthermore, we employed XAFS technique to detect the valance state and local coordination structure of active sites. XANES data in Fig. 4f indicated platinum species are at low valence state after CO oxidation: which may result in low activity due to no surface-active oxygen to participate in CO oxidation 30 . As a reference, 1Pt-SiO 2 -400 exhibited low activity in CO oxidation, even though possessing similar local coordination structure for Pt − O and Pt − Pt shell to the used 1Pt2Bi-SiO 2 ( Supplementary Fig. 14b). Thus, we can draw a conclude that the surface Pt−[O] x −Bi structure plays a key role in low temperature CO oxidation reaction rather than oxidized Pt x O z cluster.
The reducibility and active oxygen for Pt/PtBi-SiO 2 . As we known, the reducibility of catalysts is crucial in various redox reactions [31][32][33] . For fresh samples, a main reduction peak located at ~ 100 °C appeared on pro les of H 2 −temperature programmed reduction (H 2 −TPR) in Fig. 5a for 1Pt-SiO 2 contributed by the reduction of Pt x O z clusters 34 . However, for Bi-doped samples, the rst broad reduction peak was shifted to 162 °C (1Pt2Bi-SiO 2 ) and 197 °C (1Pt5Bi-SiO 2 ) in Supplementary Fig. 15a Supplementary Fig. 16 also demonstrated the surface-active oxygen for 1Pt2Bi-SiO 2 (~ 75 µmol/g) is almost two times than that (~ 40 µmol/g) of 1Pt-SiO 2 . However, these oxygen species only reacted with CO molecule above 100 °C, well consistent with low activity in CO oxidation with oxidative pretreatment (Supplementary Fig. 8).
We also carried out the H 2  superior active to react with CO molecule generating CO 2 from ~ 50 °C (Fig. 5b), well consistent with the CO oxidation "light off" temperature (Fig. 3b). However, 1Pt-SiO 2 did not activate the surface hydroxyls to produce CO 2 (water-gas shift reaction) until 195 °C 9 . This highly reducible oxygen species may motivate initial CO oxidation (~ 50 °C) through Mars-van Krevelen (MvK) mechanism 36 and show a strong correlation between reaction rate and active oxygen amount in Fig. 5c 9 . The CO adsorption reached saturation rapidly for 1Pt-SiO 2 at 2062 cm − 1 attributed to linear CO adsorbed on Pt 0 sites (Pt 0 -CO) 37 ( Fig. 6a and Supplementary Fig. 17), resulting in no other Pt sites dissociating gases oxygen into active oxygen species 38 . However, the peak for CO adsorption over 1Pt2Bi-SiO 2 was detected at 2043 cm − 1 , which was absolutely different linear CO adsorbed on Pt 0 sites. This red-shift phenomenon can be eliminated the possibility of either size-effect of Pt species or CO adsorption on pure Bi species. In one hand, 1Pt-SiO 2 -400 with total platinum cluster also showed the peak around 2060 cm − 1 (Supplementary Fig. 18) similar with the peak (2062 cm − 1 ) in 1Pt-SiO 2 . In another hand, there is no CO adsorption peak for 2Bi-SiO 2 except for gases CO peaks ( Supplementary Fig. 19), no matter with oxidative or reduced pretreatment. consistent with the band after hydrogen pretreatment at 210 °C (Fig. 6c). For 1Pt-SiO 2 , after CO oxidation, the CO adsorption band occurred a blue shift (2072 cm − 1 ) compared to the band (2062 cm − 1 ) after hydrogen reduction at 210 °C ( Supplementary Fig. 23), due to the surface oxidation of metallic Pt cluster.

Discussion
In summary, we have prepared silica-supported platinum-bismuth catalysts via an incipient wetness impregnation, possessing excellent sinter resistance due to the formation of oxidized Pt x Bi y O z cluster. (2) activating CO molecules to catalyze CO oxidation in a lower apparent activation energy. Therefore, we have provided a general approach towards design of potential active and stable platinum catalysts. and Pt(NH 3 ) 4 (NO 3 ) 2 (20 mg) in 0.1 mol/L HNO 3 solutions (3 mL) was added dropwise onto SiO 2 power (1 g) under manually stirring. The powders were standing in ambient conditions for 2 h and then dried in still air at 80 °C for 12 h, followed by air-calcination at 550 °C for 4 h (ramping rate: 2 °C/min). As comparison, we also prepared 1Pt-SiO 2 calcinated at 400 ºC, 1Pt1Bi-SiO 2 and 2Bi-SiO 2 samples with same synthetic method. The Bi and Pt contents were controlled on demand during preparation process of catalysts, and the data of these catalysts is as follows: Company, China). CO conversion was de ned as CO reaction / CO input × 100%. The related stability tests were done in the same conditions at the constant reaction temperature of 100ºC for 10 h with a GHSV of ∼134,000 mL⋅g cat −1 ⋅h − 1 . Rate measurements were made in the separate catalytic tests rather than the "light-off" mode, i. e. the same gas composition, but at speci c space velocities to ensure operation in the kinetic regime (< 20% conversion of CO).
Materials characterization. The actual platinum and bismuth concentrations of the catalysts were analyzed by inductively coupled plasma atomic emission spectroscopy (ICP-AES; Optima 5300DV, PerkinElmer). The air-calcined samples (fresh catalysts) were used directly for characterization. First, 0.1 g catalyst (accurate to 0.0001 g) was added to 2 mL hydro uoric acid under continuous stirring until the powder was dissolved adequately. Second, the as-formed SiF 4 was removed via evaporation. Then, almost 3 mL of nitric acid was introduced and the solution was kept slightly boiling for 2 h. Finally, the solution was cooled to nearly 25 ºC and diluted for the ICP-AES test.
XRD patterns were recorded on a Bruker D8 Advance diffractometer (40 kV, 40 mA) with a scanning rate of 4° min − 1 , using Cu K α1 radiation (λ = 1.5406 Å). The diffraction patterns were collected from 10 to 80°w ith a step of 0.02°. The 2θ angles were calibrated with a µm-scale Alumina disc. The powder sample after grinding was placed inside a quartz sample holder for each test. XPS analysis was performed on an Axis Ultra XPS spectrometer (Kratos, U.K.) with 225 W of Al Kα radiation. The C 1 s line at 284.8 eV was used to calibrate the binding energies.
The nitrogen adsorption-desorption measurements were performed on an ASAP 2020-HD88 analyzer (Micromeritics Co. Ltd.) at 77 K. The measured powders were degassed at 250 °C under vacuum (< 100 µmHg) for over 4 h. The BET speci c surface areas (S BET ) were calculated from data in the relative pressure range between 0.05 and 0.20. The pore diameter (D p ) distribution was calculated from the adsorption branch of the isotherms, based on the BJH method.
The TEM and high-resolution TEM (HRTEM) experiments were carried out on a FEI Tecnai G 2 F20 microscope operating at 200 kV. All the tested samples were suspended in ethanol, and then a drop of this dispersed suspension was placed on an ultra-thin (3−5 nm in thickness) carbon lm-coated Cu grid.
The as-formed sample grid was dried naturally under ambient conditions before loaded into the TEM sample holder. The aberration-corrected HAADF-STEM images were carried out on a JEOL ARM200F microscope equipped with probe-forming spherical-aberration corrector.  in-situ chamber (30 mL min − 1 ) and heated in a stepped way (every 40 K); DRIFTS spectra were obtained by subtracting the background spectrum from subsequent spectra. The IR spectra for every step were recorded continuously for 40 min to reach the equilibrium. Analysis of the spectra has been carried out by using OPUS software. For further investigation of the process of adsorption-desorption of CO over Pt/PtBi-SiO 2 catalysts, a "CO − N 2 −CO − O 2 " test was measured with in situ DRIFTS. The process of activation was carried out as described above. Then a background spectrum was collected at a certain temperature (100 °C) under pure N 2 (30 mL min − 1 ). The catalyst was exposed continuously to 2% CO in N 2 for CO adsorption for 30 min. Once CO gas was switched to an N 2 stream, also the corresponding IR spectra were recorded for 30 min. Then the catalyst was exposed to 2% CO in N 2 for CO re-adsorption for 30 min; ultimately 1% O 2 in an N 2 stream was introduced, in order to follow the surface changes during the CO removal process. The CO adsorption for Pt/PtBi-SiO 2 samples was also carried out in the mode "CO adsorption → CO oxidation at 200 °C → CO adsorption" to detect structure evolution of adsorption site after CO oxidation.
DFT calculations. First-principles calculations were performed within the framework of the density functional theory (DFT) using the Vienna ab initio simulation package (VASP) 44,45 . The projector augmented wave (PAW) method was employed 46 , and the wave functions were expanded by plane-wave basis sets with an energy cutoff of 400 eV. The exchange-correlation effects were described by the optPBE-vdW functional 47,48 , which explicitly takes Van der Waals interactions into account. The simulation model of the PtO x system was constructed by nesting an icosahedral Pt 55 cluster inside a platinum oxide layer that contains 32 Pt atoms and 66 O atoms, and the deposited Bi atoms (10 atoms in the model) were placed on the outer layer of the platinum oxide. The unit cell contains enough vacuum along all three directions to eliminate spurious interactions between periodic images. Since the model of the system is an isolated cluster, the rst Brillouin zone was sampled using the Γ point only. In the geometry optimizations, all atoms were allowed to relax until the maximum force was below 0.03 Å /eV.
The vibrational frequencies of CO were calculated through a nite differential approach, in which only the CO adsorbate, the Pt atom that connects it, and the surrounding Pt, O and Bi atoms were allowed to move while all other atoms were kept frozen.
Reducible property and surface oxygen. Hydrogen temperature-programmed reduction (H 2 −TPR) was applied to determine the pretreatment temperature under hydrogen and reducibility. The measurements were carried out on a Micromeritics Autochem II 2920 instrument. Fresh (as calcinated in air) catalysts were used for characterization. Prior to the measurement, the catalyst (20-40 mesh, ~ 100 mg) was pretreated for 30 min in a ow of 5% O 2 /He (30 mL·min − 1 ) at 300 ºC (10 ºC·min − 1 ). The test was carried out from room temperature to 600 ºC (10 ºC·min − 1 ) at a ramp of 10 ºC·min − 1 under 5% H 2 /He (30 mL·min − 1 ). A thermal conductivity detector (TCD) was used to detect the changes of hydrogen concentration. The in situ H 2 -TPR for used samples was also performed on Autochem II 2920 instrument after CO oxidation without air exposure with He purging to remove reactant gas completely.

Declarations
Data availability All data generated or analysed during this study are included in this published article (and its Supplementary Information les) or can be obtained from the authors upon reasonable request.