Continuous dimethyl carbonate synthesis from CO2 and methanol over BixCe1−xOδ monoliths: Effect of bismuth doping on population of oxygen vacancies, activity, and reaction pathway

We evaluated bismuth doped cerium oxide catalysts for the continuous synthesis of dimethyl carbonate (DMC) from methanol and carbon dioxide in the absence of a dehydrating agent. BixCe1−xOδ nanocomposites of various compositions (x = 0.06–0.24) were coated on a ceramic honeycomb and their structural and catalytic properties were examined. The incorporation of Bi species into the CeO2 lattice facilitated controlling of the surface population of oxygen vacancies, which is shown to play a crucial role in the mechanism of this reaction and is an important parameter for the design of ceria-based catalysts. The DMC production rate of the BixCe1−xOδ catalysts was found to be strongly enhanced with increasing Ov concentration. The concentration of oxygen vacancies exhibited a maximum for Bi0.12Ce0.88Oδ, which afforded the highest DMC production rate. Long-term tests showed stable activity and selectivity of this catalyst over 45 h on-stream at 140 °C and a gas-hourly space velocity of 2,880 mL·gcat−1·h−1. In-situ modulation excitation diffuse reflection Fourier transform infrared spectroscopy and first-principle calculations indicate that the DMC synthesis occurs through reaction of a bidentate carbonate intermediate with the activated methoxy (−OCH3) species. The activation of CO2 to form the bidentate carbonate intermediate on the oxygen vacancy sites is identified as highest energy barrier in the reaction pathway and thus is likely the rate-determining step.


Introduction
Carbon dioxide (CO2) is a major greenhouse gas, contributing to climate change and global warming [1,2]. The growing awareness of the detrimental influence of CO2 in the atmosphere has spurred research on its capture and utilization as a sustainable C1 feedstock for organic syntheses of high valueadded chemicals [3]. Prominent examples include the syntheses of methanol [4][5][6][7], dimethyl ether (DME) [4,8], and dimethyl carbonate (DMC) [4,[9][10][11]. DMC has low toxicity and is non-corrosive, facilitating its safe handling. It is a versatile chemical, which is used for various purposes, such as aprotic polar solvent, methylation and carbonylation agent, intermediate in polycarbonate synthesis, building block in the syntheses of pharmaceuticals, and alternative fuel or oxygenated additive for diesel/gasoline [12][13][14]. Various methods have been applied for DMC synthesis [9], including transesterification, oxidative carbonylation of methanol, methanolysis of phosgene, and the direct synthesis of DMC from CO2 and methanol (CO2 + 2CH3OH → CH3OC(=O)OCH3 + H2O). Among these synthetic methods the latter one is considered as an environmentally benign and sustainable green process, as water is the major byproduct [15]. However, inherent problems of this synthetic pathway are the limitation imposed by thermodynamics and the activation of methanol and CO2. The thermodynamic limitation can be alleviated by applying high CO2 pressure and most efficiently by rapid removal of the produced water. Consequently, various methods have been explored for the in-situ water removal [16], including the use of traps [17], dehydrating agents [18,19], and membrane reactors [20].
Various types of catalyst have been evaluated for the direct synthesis of DMC from methanol and CO2 [16], including transition metal oxides and ionic liquids. Among the transition metal oxides ceria-based nanomaterials and mixed oxides have gained particular interest and have shown some potential for the direct DMC synthesis [16,18,[21][22][23]. However, the methanol conversion achieved was mostly below about 13% [24] if no provisions for rapid water removal were taken. The strong beneficial effect of in-situ removal of water from the reaction mixture has recently been demonstrated for a ceria catalyst, a high methanol conversion of 95% with ~ 99% DMC selectivity was achieved using a continuous fixed-bed reactor and 2-cyanopyridine as a dehydrating agent [18]. However, the use of a dehydrating agent, such as 2-cyanopyridine, adds considerable complexity to DMC production, because the dehydrating agent has to be recycled and its interaction with reaction components can lead to undesired side reactions. Thus, there remain still two big challenges: the search for highly active and stable catalysts, and efficient practical concepts for the removal of water.
In the present study, we have evaluated various bismuth doped ceria monolithic catalysts, which were coated on a honeycomb ceramic support. To the best of our knowledge Bi-doped ceria has so far not been evaluated as potential catalyst for the direct DMC synthesis [16]. The use of honeycomb-type catalysts could offer several advantages compared to fixed-beds of related particulate catalysts [24][25][26]. Monolithic catalysts exhibit good interphase mass and heat transfer, low pressure drop at relatively large surface area, and could facilitate more efficient removal of the formed by-product (water), which unfavorably affects the thermodynamic equilibrium and limits the catalytic efficiency of the DMC synthesis [27]. Bismuth trioxide has a cubic fluorite structure like cerium dioxide, and it has been proven that the solid solution of cerium-bismuth could be well formed when the bismuth dopant is present in an appropriate proportion [28]. We synthesized a series of BixCe1-xOδ (x = 0.06 to 0.24) nanocomposites with an average particle size in the range of 5.5-7.8 nm and coated them on a ceramic honeycomb. A strong correlation of oxygen vacancies and catalytic performance was observed. Among these monolithic catalysts those based on Bi0.12Ce0.88O δ nanocomposite exhibited the best performance in the DMC synthesis, i.e., a promising 20.8% methanol conversion and a high DMC selectivity of 83.5% without using any dehydrating agents. Furthermore, the Bi0.12Ce0.88O δ also showed stable catalytic performance over ~ 45 h on stream. In-situ modulation excitation spectroscopy in tandem with diffuse reflectance Fourier transform spectroscopy and first-principle calculations were employed to shed some light on the molecular surface processes of the DMC synthesis on this catalyst.

Preparation of monolithic BixCe1-xOδ catalysts
The BixCe1-xOδ (0.06 ≤ x ≤ 0.24) bimetallic oxides were prepared by an aqueous-phase co-precipitation method [19]. Typically, 15 g (NH4)2Ce(NO3)6 and 1.475 g Bi(NO3)3·5H2O were dissolved in water (500 mL), and 70 g urea in 100 mL aqueous solution was added. This mixture was gradually heated to 90 °C under mechanical stirring (600 rpm) and the formed solution was kept at this temperature for 5 h. Afterward the precipitates were separated by filtration, washed thoroughly with hot water, dried at 80 °C for 12 h, and calcined at 400 °C for another 4 h. The as-prepared Bi-doped cerium oxide powders were then ball-milled with distilled water to form a slurry, which was coated on a cordierite honeycomb (64 cell·cm -2 , Φ = 10 mm, L = 25 mm), and the excess slurry was blown off. Finally, the coated cordierite honeycomb was dried at 100 °C and calcined at 400 °C to yield the monolithic catalysts (Fig. 1).

Characterization
Powder X-ray diffraction (XRD) patterns were recorded on a X'Pert PRO diffractometer (PANalytical) using Cu-Kα radiation of 0.15406 nm at 40 kV and 40 mA to analyze the phase structure of the samples. The XRD data was recorded in the 2θ range of 20 o to 80 o with a step size of 0.0334 o and a counting time of 47 s per step. The results were processed using the Jade-6 software. Raman spectra were recorded at room temperature using a Renishaw inVia spectrometer with a resolution of 2 cm -1 . A solid laser beam at 532 nm was used as the exciting source.
N2 adsorption-desorption isotherms were measured at liquid nitrogen temperature (-196 °C) using an ASAP 2000 instrument (Micromeritics). Before the measurement, the samples were degassed at 300 °C for 3 h. The surface area was determined by multipoint Brunauer-Emmett-Teller analysis of the N2-adsorption isotherms.
Transmission electron microscopy (TEM) images were recorded on a Philips FEI Tecnai G 2 Spirit microscope operated at an accelerating voltage of 120 kV equipped for energydispersive X-ray spectroscopy (EDX) analysis. The specimens were prepared by ultrasonically dispersing the sample in ethanol, depositing droplets of the suspensions on a carbon-coated Cu grid, and drying in air.
The Bi-content of the samples was analyzed by inductively coupled plasma atomic emission spectroscopy (ICP-AES) using an Agilent 720ES instrument. The samples of 0.05-0.06 g were dissolved in a 5 mL aqua regia solution with the addition of 30 wt.% H2O2. The as determined bismuth loadings of the BixCe1-xOδ nanocomposites were: 6 wt.%, 12 wt.%, 18 wt.%, and 24 wt.%, respectively. X-ray photoelectron spectroscopy (XPS) was performed using a Kratos XSAM-800 (UK) spectrometer and Mg Kα radiation at 13 kV and 20 mA. The binding energies were calibrated using the C 1s level (284.8 eV) as an internal reference.
Electron paramagnetic resonance (EPR) spectra were recorded on a Bruker ESP 300e spectrometer. All the samples of 50 mg were pretreated with N2 (40 mL·min −1 ) at 150 °C for 30 min, and then exposed to a flow of argon for 2 h. After pretreatment, the tubular reactor was sealed and transferred to a glove box, where catalysts were loaded into the EPR tubes (Wilmad-LabGlass). EPR spectra were recorded in the X-band (9.7 GHz) at room temperature using a microwave power of 20 mW, amplitude modulation of 5 G, and frequency modulation of 100 kHz. Temperature programmed desorption (TPD) of CH3OH, CO2, and NH3 (CH3OH-, CO2-, and NH3-TPD) was performed on a Micromeritics Autochem II Chemisorption Analyzer, connected to a mass spectrometer or a TCD analyzer. 100 mg of catalysts were fixed in a quartz reactor and heated under a He flow of 40 mL·min −1 at 300 °C for 4 hours to remove the adsorbates. Then the sample was cooled to room temperature, saturated with methanol or CO2, and swept by flowing He overnight. In the TPD measurements the temperature of the samples was ramped to 400 °C at a heating rate of 10 °C·min -1 , and the desorbed species were monitored in-situ by mass spectrometry (MS) at m/z = 31 for the detection of CH3Ospecies and at m/z = 44 for CO2 species.
The acidity of the Bi-Ce oxide samples was analyzed by NH3-TPD. Before measurements the oxides were pretreated at 600 °C in an Ar flow for 1 h and then cooled to room temperature. Subsequently, the sample was exposed to a flow (30 mL·min −1 ) of 20 vol.% NH3 balanced with Ar for 0.5 h, followed by exposure to an argon flow until no NH3 species could be detected anymore. The reactor temperature was programmed to increase at a ramp rate of 10 °C·min −1 , and the amount of NH3 in the effluent was recorded as function of temperature.

Catalytic test
A pressurized flow-type reaction apparatus with a stainless steel reactor of ~ 11 mm inner diameter was used for the catalytic tests (Scheme S1 in the ESM). The BixCe1-xOδ and bare CeO2 monolithic catalysts were evaluated in the DMC synthesis from CH3OH and CO2 at 140 °C and 2.4 MPa. Typically, 500 mg of catalysts was coated onto the honeycomb ceramics (2.5 mL) to produce the monolithic catalysts ( Fig. 1). Prior to catalytic tests, the reactor was sealed and purged by CO2 flow for 5 min to remove the air. Then, a mixed gas flow, consisting of a 2:1 molar ratio of CH3OH (0.145 mL·min -1 ) and CO2 (40 mL·min -1 ) was introduced with a gas hourly space velocity (GHSV) of 2,880 mL·gcat -1 ·h -1 . The product gas mixture was analyzed employing a GC-7890B equipped with a flame ionization detector (FID) through a ten-way valve. The CH3OH conversion (XCH 3 OH) and DMC selectivity (SDMC) were analyzed when the reactions were at a steady-state. XCH 3 OH and SDMC was calculated as: 3 3 DMC HCHO DME CH OH CH OH DME HCHO DME Where, Ci represents the corresponding component concentration.

In-situ infrared spectroscopy investigation
Diffuse reflection infrared Fourier transform (DRIFT) spectra were recorded on a Vertex 70v spectrometer (Bruker) equipped with a liquid nitrogen cooled mercury-cadmium-telluride (MCT) detector (ID316, ZnSe Window) and an optical filter (F321). Spectra were recorded at 4 cm -1 spectral resolution and 60 kHz scanning velocity.
Modulation-excitation spectroscopy [29,30] was carried out by periodically switching between two different gases: (CO2 → He) and (CH3OH + CO2 → He) in two different experiments at 140 and 25 °C. The last five cycles were averaged into one cycle to enhance the S/N ratio and time resolution. Phasesensitive detection [30] was used to further remove the noise and to obtain kinetic information of responding surface species. The phase-domain spectra were obtained via a mathematical treatment of the time-domain spectra according to the following equation: where, T is the length of a cycle, ω is the demodulation frequency, φk is the demodulation phase angle, k is the demodulation index (k = 1 in this study), and A(t, v  ) and Ak( v  ) are the active species responses in the time and phase domains, respectively.

First-principle calculations
First-principles calculations were carried out using spinpolarized density functional theory (DFT) with generalized gradient approximation (GGA) of Perdew-Burke-Ernzerhof (PBE) implemented in VASP code [31,32]. The DFT+U methodology with a value of U = 5.0 eV was used in this work, which has been extensively utilized for ceria in the literature [33,34]. The valence electronic states were expanded in the basis of plane waves with the core-valence interaction represented using the projector augmented wave (PAW) approach and a cutoff of 400 eV [32]. To model the CeO2(111) surface, a periodic slab with a (5×5) surface unit cell was considered, containing nine atomic layers and a vacuum layer of 15 Å. The bottom three atomic layers of ceria substrate were fixed to mimic the bulk structure, whereas the other layers were allowed to relax during geometry optimizations. Due to the large supercell, k-point sampling was restricted to Γ point. (2×2×1) k-points were also used to validate our results. The climbing image nudged elastic band (CI-NEB) method was used for the transition state calculations [35]. The Gibbs adsorption free energy (ΔG) was calculated according to ΔG = ΔE + ΔZPE − TΔS, and the zero point energy (ZPE) for the adsorbed species by using the finite-displacement method with the fixed substrate. The entropy corrections for the adsorbates on the surface were considered zero, since the main contribution to the entropy is due to the translational entropy.

Characterization of BixCe1-xOδ nanocomposites
XRD patterns of the series of BixCe1-xOδ (x = 0.06 to 0.24) nanocomposites prepared by the described coprecipitation method are shown in Fig. 2 Figure 2(b) shows the N2-adsorptiondesorption isotherms of the BixCe1-xOδ samples. All samples showed Type IV isotherms. The N2-uptake increased with higher Bi content of the nanocomposites, as also reflected by the increasing BET surface areas ( Table 1). The surface properties of the BixCe1-xOδ nanocomposites were investigated by Raman spectroscopy, EPR, TEM, EDX, XPS, and TPD of CO2, CH3OH, and NH3.
The concentration of oxygen vacancies (Ov) was analysed using Raman spectroscopy, EPR, and XPS. Figure 3(a) shows the Raman spectra of the different nanocomposites. Two distinct Raman bands appeared at 465 and 600 cm -1 . The band at 465 cm -1 represents the symmetrical stretching F2g mode of Ce-O [36]. The F2g band was broaden and slightly shifted to higher wave numbers when the Bi 3+ species were incorporated into the ceria lattice forming BixCe1-xOδ bimetallic oxides. The Ov concentrations derived by evaluating the Raman spectra are presented in the ESM. Figure 3(b) depicts the EPR spectra of CeO2 and the Bi0.12Ce0.88Oδ nanocomposite, typical for the BixCe1-xOδ oxides. An intensive single line EPR signal with g 2.003 was observed with both CeO2 and the Bi0.12Ce0.88Oδ nanocomposite, indicating the presence of surface Ce 3+ with free electrons and oxygen vacancies [37]. Note that Bi0.12Ce0.88Oδ showed a stronger EPR signal compared to CeO2, indicating a much higher Ov concentration, in line with XPS results (Table 1) and the Raman results presented in the ESM.
TEM images of the BixCe1-xOδ oxides are shown in Fig. 4. The average crystal sizes of the samples with different composition were: Bi0.06Ce0.94Oδ (~ 8.9 nm), Bi0.12Ce0.88Oδ (~ 6.5 nm), Bi0.18Ce0.82Oδ (~ 6.3 nm), and Bi0.24Ce0.76Oδ (~ 6.1 nm). The decreasing particle size with higher concentration of Bi species indicates that the crystal growth of the oxide particles was inhibited due to the lattice expansion generated by the incorporation of Bi (Bi 3+ cation radius (0.103 nm) is larger than that of the Ce 4+ cation (0.087 nm)). EDX elemental mapping confirmed that Bi species were uniformly dispersed in the ceria forming bimetal oxides, as representatively shown for the Bi0.12Ce0.88Oδ nanocomposite in Fig. 5.
Next, the chemical states and oxygen vacancies of BixCe1-xOδ were investigated using XPS. The Ce 3d spectra ( Fig. 6(a)) shows characteristic peaks of v, v'' , and v''' , corresponding to Ce 3d5/2, which a r e assigned to Ce 4+ ions in BixCe1-xOδ. The peaks of v0 and v' belong to the Ce 3d5/2 of Ce 3+ species [38]. While some Ce 3+ species were also present in the bare CeO2 samples, the concentration of Ce 3+ in BixCe1-xOδ was significantly higher, ranging from 16.1% to 20.8% (Table 1). This indicates that some Ce 4+ was reduced to Ce 3+ when Bi 3+ ions were introduced into the ceria lattice. Bismuth was present as Bi 3+ species, as indicated by the XPS analyses presented in Fig. S3 in the ESM [28]. The O 1s spectra are shown in Fig. 6(b). The asymmetric spectra can be deconvoluted into three peaks with BEs of 529.3, 530.5, and 532.2 eV, respectively. These BEs have been associated to O 2species in the lattice (OL), oxygen vacancies or defects (Ov), and chemisorbed or dissociated oxygen species, such as hydroxyl groups (OH), respectively [39][40][41][42]. The oxygen vacancy (Ov) concentrations estimated from these spectra are listed in Table 1. The concentration of Ov in ceria (23.2%) increased strongly with incorporation of Bi 3+ into the ceria lattice and reached a maximum of 37.3% in Bi0.12Ce0.88Oδ, in accord with the strong EPR signal observed with this sample (Fig. 3(b)).

Catalytic performance in DMC synthesis
The catalytic performance of the BixCe1-xOδ catalysts, coated on the honeycomb ceramics, were evaluated for the DMC synthesis. The reactions were carried out at a pressure of 2.4 MPa [24][25][26]. The results of the catalytic tests are presented in Fig. 7. Note that beside the main reaction (DMC synthesis), the formation of DME 2CH3OH → CH3OCH3 + H2O, formaldehyde CH3OH + CO2 → HCHO + CO + H2O, and methanol decomposition CH3OH + 2CO2 → 3CO + 2H2O were observed [15]. The catalytic performance over the best catalyst, Bi0.12Ce0.88Oδ, was evaluated at different reaction temperatures from 100 to 180 °C. Figure 7(a) shows that the methanol conversion reached a maximum at around 160 °C, while the selectivity to SDMC decreased monotonically with increasing temperature in the investigated temperature range. The optimal reaction temperature for the production of DMC was around 140 °C, taking into account both methanol conversion and DMC selectivity. In Fig. 7(b) the catalytic performances of the BixCe1-xOδ catalysts with different Bi-content (x) are compared.
The curve of CH3OH conversion shows a distinct maximum for the Bi0.12Ce0.88Oδ catalyst (dashed black line). This nanocomposite exhibited the highest activity (~ 21% conversion of CH3OH), while, the DMC selectivity gradually decreased from 89% to 72% with increasing concentration of incorporated bismuth. Taken all together, the Bi0.12Ce0.88Oδ afforded the    best catalytic performance (highest yield of DMC), including CH3OH conversion and DMC selectivity, which is mainly due to its surface properties (vide infra). The cordierite support without Bi-Ce oxide coating was inactive for the DMC synthesis under identical reaction conditions, corroborating that the catalytic active sites are associated with the Bi-Ce oxides. Further, we examined the relationship between the catalytic activity and concentration of oxygen vacancies. Figure 7(c) shows that the DMC productivity of the BixCe1-xOδ catalysts is strongly enhanced with increasing OV concentration. Although the exact determination of the Ov concentration by XPS may be subject to some uncertainty, there is no doubt that higher Ov concentrations result in higher DMC productivity, as Raman and EPR investigations further substantiated. These results clearly indicate that oxygen vacancies play a crucial role in the catalytic process.
In DMC synthesis often deactivation of catalysts caused by the accumulation of carbonate species (e.g., CO3 2-) on the catalyst surface has been reported [8]. Therefore, we also investigated the long-term behavior of the best performing catalyst, Bi0.12Ce0.88Oδ, under identical reaction conditions. As shown in Fig. 8, the CH3OH conversion and selectivity to DMC remained stable at 20.6±0.3% and 85.1±1%, respectively, over the test duration of 45 h on stream. The DMC formation rate on Bi0.12Ce0.88Oδ reached up to ~ 40 mmol DMC·gcat -1 ·h -1 . Thus, the Bi-doped cerium oxide monolithic catalyst exhibited excellent stability in the continuous DMC synthesis, which is mainly due to the robust nature of the Bi-doped cerium oxides, as evidenced by the XRD and TEM analyses of the spent Bi0.12Ce0.88Oδ catalyst after the long-term tests, Figs. S4 and S5 in the ESM.

CO2-and CH3OH-TPD
TPD was applied to study the adsorption properties of the Bi0.12Ce0.88Oδ nanocomposite for methanol and carbon dioxide [39,43]. As shown in Fig. 9(a), a broad desorption peak ranging from ~ 70 to ~ 200 °C was observed in the CO2-TPD, which is attributed to the removal of bidentate carbonate species on the oxide surface [44]. Note that the intensity of the desorption peak of the Bi0.12Ce0.88Oδ nanocomposite is much higher than that of bare CeO2, indicating that CO2 adsorption is enhanced on Bi0.12Ce0.88Oδ, which exhibits a much higher concentration of oxygen vacancies than bare ceria (cf. Table 1). The CH3OH desorption profiles in (Fig. 9(b)) show two peaks centered at ~ 100 and 300 °C, which are assigned to associative and dissociative adsorption (CH3O-) of CH3OH, respectively [39]. Interestingly, the intensity of the latter peaks is similar, implying that the stability of methanol chemisorbed on Bi0.12Ce0.88Oδ and bare CeO2 is similar. The TPD results indicate that the Bidoping strongly enhances the CO2 adsorption uptake of ceria and that the activation of carbon dioxide plays a crucial role in the dimethyl carbonate synthesis from CO2 and CH3OH (vide infra).
The NH3-TPD analyses showed that the acidic properties  of the oxides slightly decreased with increasing concentration of the bismuth dopant (Table 1 and Table S1 in the ESM). However, no obvious correlation between acidic properties and catalytic performance was observed.

In-situ infrared spectroscopy
To gain deeper insight into the species formed on the catalyst surface, we carried out in-situ DRIFTS studies coupled with modulation excitation spectroscopy and phase sensitivity detection at the reaction temperature of 140 °C [29,30]. In the first experiments ( Fig. 10(a)), periodically switching between CO and He (CO2 → He modulation) was applied. The main species adsorbed on the surface of the catalyst were carbonate and bidentate carbonate species. The IR bands at 1,018 cm -1 are assigned to the C-O bond of bidentate carbonate species [44]. The IR bands at 1,539 and 1,392 cm -1 are attributed to "O-C-O" of bidentate carbonate species. The formation of the bidentate carbonate species involved surface hydroxyl groups (-OH), as the broad band at 1,269 cm -1 belongs to CO-H of bidentate carbonate species [45]. The bidentate carbonate also shows a broad band at 1,634 cm -1 [46]. Therefore, the bidentate carbonate species should be formed when CO2 adsorbed on the oxygen vacancy sites through insertion of one O atom into the vacancy site on the catalyst surface. Of note, bands due to the physically absorbed CO2 appeared at ~ 2,359 and 2,343 cm -1 [47]. As a side-product formate species were also detected. The newly appearing positive bands at 2,900, 2,858 and 2,855 cm -1 , combined with the negative going bands at 949 and 1,018 cm -1 , indicate the cleavage of the O-C bond of carbonate, promoting the formation of formate species [48]. Next, the in-situ DRIFTS monitoring was carried out by periodically switching between the reactant gas mixture and He (CH3OH + CO2 → He), Fig. 10(b). The introduction of the reactant gas mixture gave rise to the appearance of bands at  1,194, 1,051, and 1,007 cm −1 , which are assigned to the C-O stretching modes of terminal (t-OCH3) and bridged (b-OCH3) methoxy species, respectively [49]. Furthermore, the newly appearing positive bands at 1,601, 1,456, 1,347 and 1,287 cm -1 indicate the production of the monodentate methyl carbonate species (CH3O-C(=O)-O) [49]. The methyl carbonate quickly reacted with the activated methoxy species to yield the final product DMC, as the observation of a drastic decrease of methoxy species at 1,007 cm −1 and the simultaneously increasing band at 1,601 cm −1 indicates [44].

Theoretical consideration of the reaction mechanism
To gain further insight into the reaction mechanism, the reaction paths of the DMC synthesis from CO2 and CH3OH on the Bi-doped reduced ceria surface were investigated by first-principle calculations. The decreased average oxygen vacancy formation energy with increasing Bi-doping of ceria was examined, as shown in Fig. 11. The results show that doping of ceria with Bi atoms can enhance the formation of oxygen vacancies, which is consistent with the experimental observations.
The reaction pathway proposed based on the experimental and theoretical results is presented in Fig. 12. Two energetically feasible reaction pathways are shown. The adsorption of CH3OH on the Bi-doped reduced ceria surface can generate a reaction Gibbs free energy (G) of -0.76 eV, suggesting this surface can facilitate the adsorption of CH3OH, as corroborated by the CH3OH-TPD measurements shown in Fig. 9(b).
The adsorbed CH3OH can dissociate to methoxy species via subtraction of a proton (path 1). This dissociation has to overcome a small energy barrier of 0.18 eV and generates -0.41 eV reaction energy, indicating that it can easily occur. The formed methoxy species can refill an oxygen vacancy on the surface. The following adsorption of CO2 will refill another oxygen vacancy. CO2 is prone to adsorb perpendicular to the surface. Then the adsorbed methoxy and CO2 react together to form the monodentate methyl carbonate (CH3O-C(=O)-O, path 1). This is the first potential rate-determining step, which needs to overcome an energy barrier of 0.84 eV.
Alternatively, the monodentate methyl carbonate can also be directly formed from the adsorbed CH3OH and CO2 by overcoming an energy barrier of 0.69 eV (path 2). The monodentate methyl carbonate can directly react with the second adsorbed CH3OH to give the final DMC product. This process is the second potential rate determining step and needs to overcome an energy barrier of 0.98 eV. Finally, the DMC  Only the top three atomic layers are shown for clear demonstration. Cerium, surface oxygen, subsurface oxygen, Bi, C, and H atoms are in white, red, pink, blue, grey, and green, respectively. Ea represents the energy barrier. and H2O products will desorb from the surface.
It is interesting to compare the reaction paths proposed here, based on in-situ DRIFTS, TPD, and theoretical calculations for the DMC synthesis over the Bi-doped ceria with that earlier suggested by Marin et al. [10] for this reaction over ceria nanorods and commercial ceria. Based on initial reaction rate data, they derived a reaction rate law where the order with respect to ceria and CO2 was approximately +1, and that of CH3OH −1. This finding indicates that over ceria the reaction obeys Langmuir-Hinshelwood kinetics, where CO2 and CH3OH are binding to the catalyst surface in separate steps and implies that CO2 adsorption and activation is the rate-determining step. This scenario is in line with reaction path 1 in Fig. 12, indicating that the reaction paths is similar on BixCe1-xOδ and ceria. However, based on the theoretical calculations direct DMC production via reaction path 2 is energetically also feasible.

Conclusions
Bismuth doped BixCe1-xOδ nanocomposites with different Bi content were coated on a ceramic honeycomb and their catalytic properties were evaluated in the continuous synthesis of DMC from CO2 and CH3OH in the absence of any dehydrating agent. The BixCe1-xOδ nanocomposites were synthesized by an aqueous-phase co-precipitation method and characterized by XRD, TEM, EDX, EPR, XPS, TPD, and DRIFTS. The population of oxygen vacancies in the BixCe1-xOδ nanocomposites was found to depend on the concentration of the Bi dopant and proved to be crucial for this reaction. The DMC production rate over the BixCe1-xOδ nanocomposites increased with higher oxygen vacancies (Ov) concentration, reaching a maximum for Bi0.12Ce0.88Oδ. In-situ DRIFTS combined with modulation excitation spectroscopy and first-principle calculations suggest that DMC is formed via coupling of bidentate carbonate species with the activated methoxy species on the surface of these catalysts. The catalytic performance of Bi0.12Ce0.88Oδ is among the best achieved so far with ceria-based catalysts in the absence of a dehydrating agent. The performance of the Bi0.12Ce0.88Oδ catalyst was stable in long-term tests over 45 h on-stream. However, as with other catalysts, economic production of DMC can only be achieved if the thermodynamic limitation is overcome by in-situ removal of the water formed as by-product. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.