Introduction

Transesterification is the key step for a variety of important organic reaction processes, e.g., the alcoholysis in biodiesel production and the synthesis of polyesters and cyclic carbonates1,2,3,4,5. Herein, the transesterification with alkyl carbonates and diols was recently found to be a typical green approach to synthesize cyclic carbonates with different ring sizes and functional groups6,7,8,9. Compared to the CO2-based route that heavily relies on homogeneous catalysts (e.g., ionic liquids), the transesterification pathway offers distinct advantages10,11. The diols can be obtained from biomass or coal, and the reaction is feasible within a solvent-freely heterogeneous catalysis system12,13,14,15,16,17, in which efficient heterogeneous catalysts would be of great significance.

It was reported that the transesterification of dimethyl carbonate (DMC) and glycerol achieved a yield of glycerol carbonate for 90% (space-time yield of 33.3 gGC h-1 gcat.−1), at a high temperature of 170 °C on roasted hydrotalcite/hexagonal silica (CHT-HMS), with a catalyst dosage of 15 wt.% and a DMC/glycerol molar ratio of 3/118. Mayra G. reported a glycerol carbonate yield of 93% in the transesterification using excessive diethyl carbonate over rehydrated hydrotalcite, but the space-time yield was only 0.77 gGC h-1 gcat.−1 19. Current heterogeneous catalytic systems with alkyl carbonates and polyols mostly rely on high temperature and excessive dialkyl carbonates, eminent heterogeneous catalysts for the transesterification at low temperatures are still needed20.

Previous study indicates that the transesterifications generally take place with acid-base catalysis. ZnO catalysts with sufficient acid-base properties were able to promote the activation of urea carbonyls and diols21,22, where Zn served as the acidic sites for the activation of carbonyl groups, and the basic sites activated the hydroxyl groups in diols. Strong acid sites are deemed inactive and would lead to by-products formation via side reactions, e.g., oligomerization23,24. Oxygen-based sites usually serve as the basic site in the reaction, which has profound effects on catalytic transesterification and curbs the side reactions. Liu et al. prepared a series of transition metal doped hydrotalcites and demonstrated that the catalytic activity is proportional to the density of basic centres in the synthesis of glycerol carbonate, and revealed that Ni could dramatically boost the strength of all three basic sites (O-H, M-O, and O2-) in the calcined hydrotalcite25. Despite the above endeavours, precisely modulating the distance and local environment of the acid-base sites to realize an appropriate acid-base balance remains a great challenge. Besides, a comprehensive understanding of the catalytic mechanisms, including the interfacial interactions and intermediate identification remains a key question to answer14,15.

Cerium dioxide (CeO2) serves as a versatile heterogeneous platform for heterogeneous catalysis, given its rich surface defects (oxygen vacancies Ov or Ce3+) and abundant Lewis acid-base sites (Ce3+, O2-)26,27,28. Element doping may offer extensive opportunities to regulate the surface acidity and basicity. In this work, we fabricated a single-atom bismuth catalyst supported on rod ceria (Bi1-CeO2-R), which demonstrated ultra-high activity in the transesterification synthesis of 1,2-butylene carbonates (1,2-BC) reaction. The catalyst achieved a high 1,2-BC selectivity of 99% and a space-time yield of 2499 g1,2-BC h-1 gcat.−1, which is the highest heterogeneous catalytic activity known to date for such a reaction, and it could maintain stability for 800 h in a one-batch test. The presence of Bi1 modified the chemical environment of the active sites, thereby benefiting the reaction. We also noticed that Bi1 experienced gradual deactivation after a long-term operation beyond 800 h, nevertheless, the catalyst could be readily regenerated with an in situ thermal treatment and maintained stability for an overall 1400 h.

Results

Catalyst structure and activity

Bi-based catalysts were prepared by a one-step hydrothermal method, and the details are described in the Supplementary Information. The Bi content in the samples was confirmed by inductively coupled plasma-optical emission spectrometry (ICP-OES) (Supplementary Table 2). X-ray diffraction (XRD) patterns of both x%Bi-CeO2-R and CeO2 rods (CeO2-R) showed a pure cubic fluorite phase structure of CeO2 (JCPDS-34-0394), and no characteristic peaks of Bi species were found, indicating that Bi species are highly dispersed in CeO2 (Fig. 1b)27,29,30,31,32. Further, the diffraction peak of CeO2 (111) shifts significantly from 28.81° to 28.55° for CeO2 (111) with increasing Bi content, suggesting that Bi species may exist in the interstitials of the CeO2 lattice (Fig. 1c)27,30,31,32,33.

Fig. 1: Performance and characterization of x%Bi-CeO2-R.
Fig. 1: Performance and characterization of x%Bi-CeO2-R.
Full size image

a The transesterification reaction synthesizes 1,2-butylene carbonates from dimethyl carbonate and 1,2-butanediol. b XRD patterns of catalysts. c Partial XRD of CeO2 (111) facet. d AC-HAADF-STEM images of 0.6%Bi-CeO2-R. e Catalytic performance of 1,2-BC synthesis for different catalysts18,19,25,43,46,47,48,49,50. f Catalytic performance of catalysts in the synthesis of 1,2-BC, Reaction condition: 50 mmol of DMC, 50 mmol of 1,2-BDO, 25 mg of catalyst (0.5 wt.% catalyst dosage based on 1,2-BDO mass, 120 °C, 20 min). g Images for STEM and the corresponding EDX mapping of 0.6%Bi-CeO2-R.

Aberration-correction high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) shows that Bi species were uniformly dispersed on the CeO2, and the bright spots were identified as atomic dispersion Bi (Fig. 1d) (0.6%Bi-CeO2-R, denoted as Bi1-CeO2). Such homogeneous dispersion of Bi species has been confirmed by the EDX mapping results (Fig. 1g). Besides, our study also reveals the mesoporous structure of the catalysts (Supplementary Figs. 3–4 and Supplementary Table 3).

The catalytic activity of the x%Bi-CeO2-R catalysts for synthetic cyclic carbonates was investigated (Fig. 1f). The conversion over 0.6%Bi-CeO2-R exhibits the most extraordinary catalytic activity of 87.4% for the conversion and 99% for the selectivity of 1,2-BC. The space-time yield over 0.6%Bi-CeO2-R was 2499 g1,2-BC h-1 gcat.−1, which achieves the record high among the heterogeneous catalytic activities to our knowledge (an improvement of more than 100 times) (Fig. 1e, f). The 0.6%Bi-CeO2-R shows excellent catalytic activity in a wide range of cyclic carbonate syntheses, e.g., glycerol carbonate (Supplementary Table 18). Furthermore, we extended the evaluation of 0.6%Bi-CeO2-R to the synthesis of N-substituted carbamates through amine carbonylation. The results demonstrate a marked enhancement compared to sole CeO2, underscoring the broad applicability of 0.6%Bi-CeO2-R in the analogues transesterification reactions. (Supplementary Fig. 59). The activity of sole Bi2O3 and CeO2-R with various Bi content were also compared in Fig. 1f. When using pure CeO2 as catalyst, the conversion of 1,2-BDO was only 26.5%. After doping 0.2% Bi on CeO2, the conversion increased to 57.4%. By increasing the Bi content to 0.6%, the optimal conversion of 87.4% was achieved. However, when the Bi content continued to increase, the conversion of 1,2-BDO eventually dropped to 24.7% with 5% Bi. The activity exhibits a volcano tendency with the loading ratios of Bi, implying an intrinsic change relevant to the doping ratios. Besides, the Bi2O3 was found to be almost inactive with trace conversion of 1,2-BDO for only 2.1%.

Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) results (Supplementary Fig. 2) show that the length of the CeO2-R was about 100 nm and the width was about 10 nm, and the addition of Bi species does not change the rod-like morphology of CeO2. Lattice spacings visible in the images of 0.19, 0.27, and 0.31 nm are ascribed to the (110), (100), and (111) facets of CeO2, respectively. Compared to 0.6%Bi-CeO2-R, cubic and octahedral CeO2 (0.6%Bi-CeO2-C and 0.6%Bi-CeO2-O) mainly processed (100) and (111) facets, respectively, which showed no activity in the transesterification (Fig. 1f). It is concluded that CeO2 (110) facet is essential for its ultra-high activity, and Bi species plays a key role in the elevated catalytic activity, while Bi1 does not show the special synergy with CeO2 (100) and CeO2 (111) (Supplementary Fig. 2).

The electronic structure and coordination environment of the Bi1-CeO2 catalyst was further investigated by X-ray absorption fine structure (XAFS). The K-edge X-ray absorption near edge spectroscopy (XANES) curve of Bi for Bi1-CeO2 locates between Bi foil and Bi2O3, indicating that the general valence of Bi ranges from 0 to +3 (Supplementary Fig. 5). The k3-weighted extended X-ray absorption fine structure (EXAFS) and the corresponding wavelet transform spectra (Fig. 2a, b and Supplementary Figs. 6–10) confirms that Bi exists as isolated atoms in the catalyst34,35,36. Fitting curves of EXAFS for Bi1-CeO2 show that the isolated Bi site is coordinated by two O atoms (coordination number: 1.8) with an average Bi-O bond length of 2.15 Å (Supplementary Table 4). Therefore, we conclude the actual micro-structure for Bi1-CeO2 should be Bi1δ+-O2. Six different coordination models for Bi1δ+-O2 active site were proposed and simulated on CeO2 (110) to investigate their stability at 774 K, using machine learning assisted Ab-initio molecular dynamics (ML-AIMD) (Supplementary Video 1). The simulation shows that the Bi1δ+-O2 with one adjacent oxygen vacancy maintains stability throughout the 10 ps tests at 774 K, implying it is supposed to be the real structure for Bi1-CeO2 (Fig. 2c & Supplementary Fig. 11), which will be adopted in our following calculations.

Fig. 2: DFT calculations and structure for x%Bi-CeO2-R catalysis.
Fig. 2: DFT calculations and structure for x%Bi-CeO2-R catalysis.
Full size image

a FT-EXAFS fitting curve of Bi1-CeO2. b WT-EXAFS plots of Bi foil, Bi2O3, and Bi1-CeO2. c Bi1-O2 bi-coordination stabilization model on the (110) facet. (Bi). (d) XPS spectra of Bi 4 f. e EPR spectra of x%Bi-CeO2. f Bi1-CeO2-Ov differential charge calculations. g, h CO2-TPD and NH3-TPD profiles of catalysts. i Bi1-CeO2-nonOv differential charge calculations. (yellow denotes electron enrichment, and cyan denotes electron loss).

The effect of the incorporation of Bi1 species on CeO2 was further revealed by visible laser Raman spectroscopy (Raman) and electron paramagnetic resonance (EPR). The vibrational mode of 1076 cm-1 is inferred to be ascribed to the Bi-O bond, and the vibrational mode F2g at 462 cm-1 is ascribed to the symmetric stretching vibration of Ce-O (Supplementary Fig. 12)33,37. The weak band at 595 cm-1 is ascribed to the defect-induced mode ID of CeO2. The integral area ratio (ID/F2g) that reflect the concentration of Ov increases from 0.0398 for CeO2 to 0.0541 for Bi1-CeO2, and then decline to 0.0472 for 1.0%Bi-CeO2 (Supplementary Table 5). The results indicate that abundant oxygen defects exist in x%Bi-CeO2. More importantly, the volcano relationship reveals that the addition of Bi may enrich Ov38, which would then affect the catalytic activity39. In line with the results from Raman spectroscopy, the 0.2% Bi-CeO2 catalyst exhibited stronger spin signals compared to pristine CeO2. Further increasing the Bi content to 0.6%gives rise to the strongest spin signal in Bi1-CeO2, approximately seven times that of pristine CeO2 (Supplementary Table 6), indicating the presence of more Ov in the Bi1-doped state (Fig. 2e). However, it should be noted that when the Bi content increased to 5%Bi-CeO2, the presence of excessive oxide (not detected by XRD, Fig. 1b) may cause a significant decrease in the effective Ov concentration on CeO2 surface, leading to a corresponding reduction EPR signal and catalytic activity (Fig. 1f).

The X-ray photoelectron spectroscopy (XPS) results shows that the binding energy peaks of Bi 4 f spectra were at 163.2 and 158.3 eV, which are ascribed to the Bi 4 f 7/2 and Bi 4 f 5/2 spin orbitals of Bi3+, and no Bi0+ species were observed (Fig. 2d). In addition, by comparing the sample with different loadings, a shift of the Bi 4 f spectra towards higher binding energies was observed, indicating the electron loss from the Bi atom to the O atom in the Bi1δ+-O-Ce3+-Ov structure. Compared to the intact CeO2, Ov concentration in Bi1-CeO2 is significantly enriched, leading to the rise of Ce3+ proportion from 22.09% to 35.89% (Supplementary Table 5). A sharp concentration rise of Ov from 22.4% (CeO2) to 31.9% (Bi1-CeO2) was observed in the O 1 s spectra, confirming that Bi1 induced the increase in Ov (Supplementary Fig. 13 and Supplementary Table 4)40,41. In addition, the binding energy of OL shifts towards lower binding energies, suggesting that OL gains electrons with the addition of Bi. The effect of Ov on the charge transfer between Bi and CeO2 was quantitatively analysed by the charge density difference (CDD) (Fig. 2f–i). The results reveal that Bi with an adjacent Ov carries a positive charge of +0.62 e, and Bi1 on the intact CeO2 shows more positive charge of +0.85 e. The results confirm that electrons migrate from the Bi atom to the support via the O atom while there is more electron back-donation from the support to Bi in the presence of Ov. The addition of Bi1 alters the local electronic structure of adjacent lattice oxygen (OL) and may have a remarkable effect on the acid-base properties of the CeO2 matrix.

We investigated the change in acidity and basicity of the catalyst surface with the addition of Bi species by temperature programmed desorption (TPD) of CO2 and NH3, respectively. Figure 2g shows CO2 desorbed at 50-200 °C, 200-400 °C, and 400-700 °C, from weak Lewis basic sites, moderate Lewis basic sites, and strong Lewis basic sites, respectively. Supplementary Table 7 shows only a trace amount of strong Lewis basic sites were present in Bi2O3, and only abundant weak basic sites for CeO2. With the addition of Bi, rich strong basic sites were observed on CeO2, and the strong basic sites on CeO2 shift to higher temperature regions with the Bi content, implying the enhancement of the basic strength. NH3-TPD results in Fig. 2h show strong acidic sites were dominant in Bi2O3 and CeO2, and with the addition of Bi, the concentration of acid sites for CeO2 shows analogues increasing tendency as the basic sites. This is because the increasing concentration of Ov induces more Ce3+ sites as weak-mediate Lewis acid (Supplementary Table 7). The result is agreed by the pyridine IR tests that the Lewis acid concentration for Bi1-CeO2 is much higher than CeO2-R (Supplementary Fig. 14 and Supplementary Table 8). Nevertheless, the shift of weak acidic sites to lower temperatures indicates the acidity is weakened. Bader charge analysis shows that the presence of Ov leads to less charge on Ce adjacent to the defects ( + 2.27 e), compared to intact CeO2-nonOv ( + 2.32 e), and this change is further enhanced by the presence of Bi1 (Bi1-CeO2-Ov, +2.23 e), in line with the weaker acidity observed in experiments (Supplementary Fig. 15). Moreover, O atoms adjacent to Ov carried negative charge of −1.15 e and −1.21 e for CeO2-nonOv and CeO2-Ov, respectively. In the presence of Bi1, the negative charge on the O atom is increased to −1.23 e, in line with the strong basicity observed in experiments.

The above results revealed the unique Bi1δ+-O-Ce3+-Ov coordination structure has changed the intrinsic electronic structure of CeO2 and activated the OL for abundant Ov creation, and the Ov would give rise to rich moderate acidic sites and strong basic sites on CeO2 in the presence of Bi1, which are important for efficient transesterification reaction.

Reaction mechanisms

Our study has proved that the exposed (110) facet of rod-CeO2 is rich in Ov, which potentially contains the most catalytically active sites. However, the mechanism for transesterification including the selective adsorption and the activation of -C = O and -O-H bonds during the synthesis of cyclic carbonates has rarely been clarified. In situ diffuse reflectance infrared Fourier transform (in situ DRIFT) was implemented to investigate the adsorption behaviour of DMC and 1,2-BDO on Bi2O3, CeO2, and Bi1-CeO2. Figures 3a-3b and Supplementary Fig. 16 show that 1,2-BDO adsorbed onto Bi1-CeO2 mainly in a bridged butoxy conformation on the Ov (1035 cm-1) at room temperature, in addition to the on-top butoxy (1091 cm-1) and bridged butoxy (1062 cm-1) conformations42. It was found that the top butoxy and bridged butoxy conformations would gradually shift to the bridged butoxy conformation with the temperature on the Ov of Bi1-CeO2 (Figs. 3d, e). In contrast, only on-top butoxy adsorption conformation (1086 cm-1) was observed on the Bi2O3 surface that is proven to be inactive for transesterification (Supplementary Fig. 17). The adsorption behaviour of 1,2-BDO on Bi1-CeO2-C and Bi1-CeO2-O was also compared, and on-top butoxy ( ~ 1100/1098 cm-1) and bridged butoxy ( ~ 1069/1071 cm-1) have been identified on both samples, which are inactive neither for the transesterification reaction (Supplementary Figs. 18, 19). We eventually find that Ce in adjacent to Ov is the major catalytic site for the activation of 1,2-BDO with the bridged butoxy-Ov conformation at 1035 cm-1 (Fig. 3e and Supplementary Figs. 20–22). The preferable adsorption behaviour and the catalytic activity of x%Bi-CeO2 catalysts with varying Bi contents further corroborate this conclusion (Supplementary Fig. 23 and Fig. 1f). DFT study further (Supplementary Fig. 32 & Supplementary Table 9) revealed that 1,2-BDO only had weak interaction with Bi1 (Ead = −0.19 eV), and the molecule was likely to adsorb through the primary hydroxyl groups at the Ov on CeO2 with the spontaneous dissociation of Oα-H (Ea = −1.39 eV), corresponding to the conformation of bridged butoxy-Ov. Although the adsorption with dual hydroxyl groups had higher adsorption energy of above 2.0 eV, it has been observed that such adsorption conformations as the major conformation on Bi2O3 and Bi1-CeO2-C (Supplementary Figs. 17, 18), thus we infer they are less reactive for the reaction.

Fig. 3: DFT calculations and reaction mechanism.
Fig. 3: DFT calculations and reaction mechanism.
Full size image

a In situ DRIFT spectra of optimal conformation of 1,2-BDO attached to Bi1-CeO2. b The optimal conformation of DMC attached to Bi1-CeO2. c The Bonding evolution of carbon and oxygen bonds of DMC on Bi1-CeO2. d In situ DRIFT spectra of 1,2-BDO on CeO2 with increasing temperature. e In situ DRIFT spectra of 1,2-BDO on Bi1-CeO2 with increasing temperature. f In situ DRIFT spectra of DMC on CeO2 with increasing temperature. g In situ DRIFT spectra of DMC on Bi1-CeO2 with increasing temperature. h Key reaction steps of the transesterification process that are shared by the CeO2 and Bi1-CeO2. i, j Projected partial density of states of Ce20 in CeO2 and Bi1-CeO2. k Regulation of acid-base sites by Bi1 enables remote catalysis for the synthesis of cyclic carbonates.

Regarding the adsorption of DMC, the catalysts of Bi1-CeO2-O and Bi2O3 with low activity gave rise to the bi-dentate conformation based on the prominent vibration of C = O-Ce-Ov (Supplementary Figs. 24, 25). On the contrary, DMC on Bi1-CeO2-C, CeO2, and ultra-highly active Bi1-CeO2-R all showed the predominant adsorption conformation of cis-cis-monodentate at 1766 cm-1 (Supplementary Figs. 26–28). With the increasing temperature, the cis-trans monodentate adsorption at 1751 cm-1 was completely shifted to cis-cis-monodentate adsorption conformation (Fig. 3f, g and Supplementary Figs. 29, 30). Further comparison of DMC adsorption results on x%Bi-CeO2 catalysts reveals that Bi content influences the adsorption conformation (Supplementary Fig. 31). The cis-cis monodentate adsorption conformation exhibits a direct correlation with the catalytic activity performance (Fig. 1f). This is consistent with the DFT results, where the cis-cis-monodentate C = O-Ce-Ov binding has the highest adsorption energy of −1.04 eV (Supplementary Table 9 and Supplementary Fig. 34). The calculations show that Ov facilitates 1,2-BDO adsorption (1,2-BDO-4 and 1,2-BDO-5), while Bi1 impacts on the DMC adsorption (DMC-2 and DMC-6, Supplementary Figs. 34, 35 and Supplementary Tables 9-10).

Detailed reaction pathways of catalytic transesterification were analysed using DFT modelling. Figure 3h shows the plausible reaction mechanism for synthesizing 1,2-BC from DMC and 1,2-BDO on Bi1-CeO2, and the reactions over CeO2 were calculated for comparison (Supplementary Table 11). Over Bi1-CeO2, 1,2-BDO experienced dissociative adsorption and contributed to surface proton enrichment, and DMC adsorbed in a cis-cis-monodentate conformation further dissociated and produced one molecule methanol with the surface proton. This step experienced an energy barrier of 1.91 eV, smaller than 2.25 eV over CeO2. The left moiety (structure III-1) would further engage in the synthetic reaction with an energy barrier of 0.51 eV (TS-2 Bi1-CeO2), and it was 0.87 eV over CeO2. Cyclization was the final step that had an energy barrier of 1.54 eV over Bi1-CeO2 and 0.19 eV over CeO2 to yield the 1,2-BC precursor (structure V). 1,2-BC escaped from the surface of CeO2 with a desorption energy of 0.64 eV, while over Bi1-CeO2, we observed a spontaneous desorption process with an exothermic energy of −0.09 eV. During the modelling, a unique Bi1-O3 structure was noticed in the cyclization reaction of the 1,2-BC formation. The modelling reveals that Bi1 is not the conventional active site with direct contact with reactants, but gives rise to remote catalysis. DMC dissociation is the potential rate-controlling step in the transesterification reaction, which is more facilitated by Bi1-CeO2, and Bi1 also benefits the spontaneous desorption of 1,2-BC.

Crystal Orbital Hamilton Population (COHP) analysis revealed the C-O and O-H bonds of DMC on the surface of Bi1-CeO2 are less activated compared to those over CeO2-R (Supplementary Fig. 36 & Supplementary Table 12, but the strongest C = O bond experiences much less unnecessary activation during the reaction on the Bi1-CeO2 surface (-COHP, 0.67), in contrast to CeO2 (-COHP, 1.07) (Supplementary Fig. 37 & Supplementary Table 13). This implies that the endothermic contribution by the C = O bond is much less on Bi1-CeO2. In addition, COHP analysis also reveals that Ce-O bonds adjacent to the Bi site underwent potential cleavage and re-bonding, resulting in an overall elevated -COHP value of 0.33. This may result in additional exothermic contribution to the reaction of DMC decomposition over Bi1-CeO2 (Supplementary Figs. 38, 39 & Supplementary Table 14). The above results indicate that the structural evolution induced by Bi1 is supposed to be a key reason for the lower energy barrier for DMC decomposition on Bi1-CeO2.

Previous studies have suggested a competing reaction pathway during the formation of cyclic carbonates from DMC and o-diols via transesterification, involving the formation of chain carbonate intermediates (e.g., hydroxy-butyl methyl carbonate, HBMC) before cyclic carbonate production18,23,43,44,45, and it was speculated that the cyclization of alkyl carbonates and diols might be the rate-controlling step for the transesterification23,44. In-situ IR was used to track the possible intermediates in the reaction (Supplementary Figs. 40, 41), but the presence of the intermediate HBMC was not detected. DFT calculations (Supplementary Figs. 42, 43) show that the reaction pathway existed over the pristine CeO2 surface starting from the dual-hydroxyl adsorption of 1,2-BDO but with a high reaction energy barrier of 1.20 eV, and would be interrupted in the vicinity of Bi1.

Bader charge analysis provides more insights that the Ce atom adjacent to the Ov carries less positive charge in the presence of Bi1 than that on intact CeO2. Projected density of states (P-DOS) analysis of Ce and O atoms in both CeO2 and Bi1-CeO2 (Fig. 3i, j and Supplementary Figs. 44-46) further elucidates the spatial extent and mechanistic role of the electron transfer: the incorporation of Bi1 induces non-localized f-electrons at the Fermi level of Ce atoms adjacent to Ov and a richness of electrons on lattice oxygen atoms, endowing them with unexpected conductivity (Fig. 4b, d, and e) and thereby establishing efficient electron propagation pathways within the catalyst. This modulation of the electronic structure significantly enhances the intrinsic reactivity of the material, leading to the remarkable transesterification activity (Fig. 3k), which we term remote catalysis.

Fig. 4: Deactivation and regeneration of Bi1-CeO2.
Fig. 4: Deactivation and regeneration of Bi1-CeO2.
Full size image

a Catalytic performance of fresh and re-dispersed Bi1-CeO2 for the synthesis of 1,2-BC in a fixed bed for 1732 h. b Original structural modelling of Bi1-CeO2. c XRD spectra of different deactivated and re-dispersed Bi1-CeO2. d P-DOS of O1-Bi1-CeO2. e P-DOS of O1-CeO2. f Quantitative spin quantum number results for different deactivated and re-dispersed Bi1-CeO2.

Deactivation and regeneration of Bi1-CeO2

We carried out an ultra-long stability test for the Bi1-CeO2 and noticed the catalyst maintained stable for 800 h in one batch and started activity decline thereafter (Fig. 4a). The deactivation by carbonaceous deposit or by-products on the catalyst was excluded by surface cleaning. TEM characterization of the deactivated catalyst suggests that Bi1 might have transformed to BiOx whilst the reaction as facet −121 of Bi2O3 was identified (Fig. 4a). The spent catalysts with different extent of deactivation (denoted as x%D-Bi1-CeO2) were tested by XRD. It has been found that the diffraction peaks of Bi2O3 (−121) at about 28° begin to appear and enhances with further deactivation (Fig. 4c). The deactivation effect of BiOx was further verified in experiments (Supplementary Fig. 47a), where the 70%D-Bi1-CeO2 catalyst showed similarly low activity as 5%Bi-CeO2 in terms of the 1,2-BC yield, of which the main Bi species was found to present as Bi2O3 in both samples, evidence by the identification of a Bi-O-Bi scattering path in the EXAFS spectra (Supplementary Figs. 47, 55–59 and Supplementary Table 15). Furthermore, a series of catalysts were synthesized on CeO2 support via the conventional wet impregnation method followed by calcination. Notably, these catalysts exhibited the same poor activity as pristine CeO2, verifying the inert behaviour of Bi2O3 clusters over CeO2. In addition, we implemented calcination for the deactivated catalyst and surprisingly observed that the BiOx species completely disappeared in 3 h at 773 K (Fig. 4c). The above result indicates BiOx may has transformed back to Bi1 during the thermal treatment. The regenerated catalyst R-Bi1-CeO2 achieves the same 1,2-BC yield as the fresh catalyst and could remain stable for another 600 h (Fig. 4a).

DFT modelling reveals that during the formation of the 1,2-BC precursor, the OL in CeO2 was activated and migrated towards Bi1 to form Bi1-O3, which contributes to the high kinetic barrier of 1.54 eV in the reaction pathway (TS-3, Supplementary Fig. 60). The desorption of 1,2-BC consists of exothermic lattice recovery (−0.93 eV) and a desorption energy of product (0.84 eV), resulting in an overall exothermic process (−0.09 eV). P-DOS for Bi1-CeO2 shows that the p-electrons of the OL adjacent to Bi1 cross the Fermi level (Fig. 4d, e), indicating the related Ce-O bonds being activated and more abundant Ov are susceptible to generate whilst the cyclization reaction of the formation of BiOx. The acidity and basicity are therefore altered by the increasing amount of Ov, leading to the change of the activity of the catalyst.

Quantitative EPR analysis regarding the Ov in Bi1-CeO2, x%D-Bi1-CeO2, and R-Bi1-CeO2 further verified the activation and dynamic migration of the lattice O (Fig. 4f). The normalized spin density gradually increased from 7.280 e^11/mm3 with the degree of deactivation, reaching 3.575 e^12/mm3 in 80%D-Bi1-CeO2. The results confirm that the cleavage of Ce-O bonds takes place to generate Ov with the deactivation. After the thermal treatment, the BiOx species was completely recovered to Bi1-O2, evidenced by the drop of spin density to 7.522 e^11/mm3 for R-Bi1-CeO2, indicating that the lattice O atoms are restored.

The evolution of Bi1 and BiOx nanoparticles was further confirmed by the XAFS by a new Bi-O-Bi scattering path at 3.51 Å (Figs. 5b, d and Supplementary Figs. 48–55) and the increased coordination number of O to 2.1 (Supplementary Table 15). The Bi-O-Bi scattering path disappeared after the thermal treatment, meanwhile the Bi-O path was regained (Fig. 5c, f) with an O coordination number of 1.6, resembling that of the fresh catalyst (1.8) (Supplementary Figs. 56–59 and Supplementary Table 16). The above conclusion has also been verified by the low catalytic activity of 5% Bi-CeO2, in which the Bi species mainly existed as metal oxide (Supplementary Figs. 47, 56–59 and Supplementary Table 16). These findings collectively demonstrate that the structural evolution from BiOₓ to the isolated Bi1 sites, driven by a straightforward thermal treatment, represents the intrinsic change for the recovery of catalytic activity.

Fig. 5: The dynamic structural evolution of catalysts.
Fig. 5: The dynamic structural evolution of catalysts.
Full size image

a Fitting curves of the FT-EXAFS for the fresh Bi1-CeO2. b FT-EXAFS fitting curve of 70% D-Bi1-CeO2. c FT-EXAFS fitting curve of Bi1-CeO2-Redispersion. d WT-EXAFS plots of Bi1-CeO2 Fresh. e WT-EXAFS plots of 70% D-Bi1-CeO2. f WT-EXAFS plots of Bi1-CeO2-Redispersion. g Deactivation-redispersion process of the catalyst.

Discussion

In this work, we prepared Bi1-CeO2 for the catalytic transesterification synthesis of 1,2-BC. The catalyst exhibited extraordinary activity (2499 g1,2-BC h-1 gcat.−1) and stability over 800 h. XAFS results and theoretical calculations confirmed that the atomically dispersed Bi1 was coordinated with two oxygen atoms on CeO2. Mechanistic investigations revealed that the Bi1 sites did not behave as the conventional active sites, but induced a distinct modulation of the electronic structures for the adjacent Ce and O atoms, boosting the delocalisation of f electrons of Ce3+ in the presence of Ov and endowing conductivity of lattice oxygen atoms, thus altering the intrinsic acid/basic properties of the catalyst with a prominent remote electron propagation. The presence of Bi1 not only lowered the energy barrier of the carbonyl transferring from DMC during the transesterification, but also gave rise to exothermic desorption of 1,2-BC via the remote catalysis. We observed Bi1 centred structural evolution was the key to the facile desorption of product, however, it also contributed to the gradual deactivation of the catalysts with the formation of BiOx. XRD, XAFS, and EPR results indicated that the decline of the activity was attributed to the Bi1 centred structural evolution when lattice O was excessively activated. Nevertheless, the Bi1 sites was proven to be readily regenerated by a facile thermal treatment, which shed more light to the rational design for the catalyst of transesterification.

Methods

Catalyst preparation

Chemicals

All the chemicals are used as received from the provider without further purification. Bismuth(III) nitrate pentahydrate (Bi(NO3)3·5H2O, 99.9%) and cerium(III) nitrate hexahydrate (Ce(NO3)3·6H2O, 99.9%) obtained from Sigma Aldrich were used as the precursors of x%Bi-CeO2.

Synthesis of CeO2-Rods and x%Bi-CeO2-Rods

Atomic bismuth-cerium dioxide catalysts (Bi1-CeO2-rods) were prepared by a one-step hydrothermal method. Specifically, 0.868 g of Ce(NO3)3·6H2O [Sinopharm Chemical Reagent Co.] and 9.6 g of NaOH (SCRC) were dissolved in 5 mL and 35 mL of deionized (DI) water, respectively. Subsequently, the solutions were mixed well and stirred continuously for 0.5 h. The resulting mixture was transferred to a Teflon-lined stainless steel autoclave reactor and held in an oven at 373 K for 24 h. The obtained precipitate was centrifuged and washed until the pH reached 7.0. Further, the mixture was dried at 353 K for 12 h and calcined at 773 K for 3 h to obtain CeO2. To obtain x%Bi-CeO2 catalysts with different Bi molar contents, a similar process was used, differing in that the Ce(NO3)3·6H2O and Bi(NO3)3·5H2O (SCRC) solutions were homogeneously mixed before mixing with an alkaline solution, with Bi/Ce molar ratios of 0.002. 0.004, 0.006, 0.008, and 0.01. The final catalysts were named as 0.2% Bi-CeO2, 0.4% Bi-CeO2, 0.6% Bi-CeO2, 0.8% Bi-CeO2 and 1.0% Bi-CeO2, respectively.

Synthesis of Bi1-CeO2-cubes

Bi-CeO2-cubes were prepared similarly to Bi-CeO2-Rods, with the difference that the hydrothermal treatment temperature was 453 K.

Synthesis of Bi1-CeO2-octahedrons

Bi-CeO2-octahedrons were prepared by a one-step hydrothermal method. Specifically, 3.8 g of Ce(NO3)3·6H2O [SCRC], a certain molar amount of Bi(NO3)3·5H2O [SCRC], and 3.3 g of C2H6O6 (SCRC) were dissolved in 35 mL of deionized (DI) water. Subsequently, the solutions were mixed well and stirred continuously for 0.5 h. The resulting mixture was transferred to a Teflon-lined stainless steel reactor autoclave and held at 453 K for 24 h. The obtained precipitate was centrifuged and washed with deionized water and ethanol until the pH reached 7.0. Further, the mixture was dried at 363 K for 12 h and calcined at 673 K for 2 h to obtain Bi-CeO2-octahedrons.

Characterizations

The morphology of all catalysts was investigated by Zeiss Gemini 300 scanning electron microscope (SEM) and TF20 transmission electron microscope (TEM). Subsequent image analysis and processing were carried out using the Gatan Digital Micrograph software.

Aberration-correction high-angle annular dark-field scanning transmission electron microscopy (AC HAADF-STEM) and associated EDX elemental mapping were obtained using a Thermo Fisher Scientific Titan Themis operated at 300 kV, equipped with a CEOS DCOR probe corrector and a Super-X 4-quadrant detector for EDX.

Bi content of these as-prepared catalysts was obtained by an inductively coupled plasma-optical emission spectrometer (ICP-OES).

X-ray diffraction (XRD) patterns of catalyst by D/tex ultra 250 diffraction measurements (Cu-Ka). The XRD patterns were recorded from 5 to 90° (2θ) with a scanning step of 0.01°. The employed radiation was Cu K-α with a wavelength of 0.154 nm.

X-ray photoelectron spectroscopy (XPS) was obtained by Thermo Scientific K-Alpha instrument, corrected by C 1 s = 284.8 eV binding energy. Raman spectroscopy was tested on the HORIBA Scientific LabRAM HR Evolution spectrometer.

Electron paramagnetic resonance (EPR) spectra were recorded at room temperature on Bruker EMXplus-6/1 device. Spectra have been recorded with a Continuous Wave (CW) Bruker EMX spectrometer operating at X-band (9.8 GHz), equipped with a cylindrical cavity at 100 kHz field modulation.

The temperature-programmed (TPD) instrument equipped with a thermal conductivity detector was used to measure the acidity and basicity of catalysts on the quartz U-tube reactor. Samples were pre-treated in He atmosphere at 473 K for about 1 h and cooled to 323 K. The CO2/NH3 gas was injected at a flow rate of 100 mL/min. After saturation of adsorption, He purge purged for 40 min. The temperature rose to 1073 K at a ramp rate of 10 K/min.

In-situ IR testing of all catalysts was performed on a Mettler-Toledo React IR 15 m device consisting of a liquid nitrogen reactor, and a diamond probe. The wavenumber range was 4000−800 cm-1, and the acquisition time was 15 seconds.

In situ diffuse reflectance infrared Fourier transforms spectroscopy (in situ DRIFTS) was performed at wavenumbers from 4000 to 750 cm-1 by accumulating 64 scans with 4 cm-1 resolution on a Nicolet iS50 FT-IR spectrometer fitted with a Harrick DRIFTS cell and a high-sensitivity MCT detector cooled with liquid nitrogen. The sample compartment was filled with 50 mg of the catalyst. The samples were pre-treated at 673 K for 2 h and then cooled to 293, 323, 353, 373, 393, and 423 K to collect the background spectra, respectively. Subsequently, the argon gas was passed through the 1,2-BDO or DMC into the DRIFTS cell at a flow rate of 30 mL/min, respectively. After the catalyst was saturated with adsorption at room temperature, N2 was purged for 30 min to remove the physical adsorption of DMC\1,2-BDO on the surface of the catalyst, and the spectra were recorded.

The edge energy of the XANES spectrum was determined from the inflection point in the leading edge, that is, the maximum in the first derivative of the leading edge of the XANES spectra. The structural parameters were obtained by a least-squares fit in R space for the k3-weighted Fourier-transformed data using Artemis. The amplitude reduction factor in these measurements was determined by EXAFS fitting of the Bi foil, which was measured simultaneously with the sample in transmission mode during the experiments.

DFT calculations

The first-principle density functional theory plus dispersion (DFT-D) calculations were implemented in the Vienna Ab-initio Simulation Package (VASP 6.4.1) with dispersion corrections by the D3 method of Grimme. The generalized gradient corrected approximation (GGA) treated by the Perdew−Burke−Ernzerhof (PBE) exchange-correlation potential was used to calculate the exchange-correlation energy. The PAW pseudopotential was employed as the scheme in the representation of reciprocal space for all the elements. The plane-wave cut-off energy was set to 520 eV for all the calculations. The Brillouin zone was sampled using a 2 × 2 × 1 Monkhorst-Pack k-point with a smearing of 0.3 eV. Ueff = 4.5 eV was applied to Ce element in the static electron structure calculations. Independence tests of both cut-off energy and k-points were carried out, as shown in Supplementary Fig. 1. Spin polarization has been considered, and the self-consistent field (SCF) tolerance was set to 10−5 eV/atom. All the modelling was performed with a convergence threshold of 0.03 eV/Å on maximum force. No symmetry constraint was used for any modelling. The computational method is believed to give high precision results, and has been validated by experimental data. In this work, the model predicted the lattice constant of the CeO2 lattice to be a = b = c = 5.43662 Å, in agreement with the experimental data of a = b = c = 5.411 Å, as shown in Supplementary Table 1.

Catalytic evaluation

The transesterification reaction was carried out in a 20 mL autoclave equipped with a magnetic stirrer (Anhui Kemi Machinery Technology Co., Ltd). In a typical experimental procedure, nitrogen purges and replaces the gas in the reactor before the experiment begins. At the end of the reaction, the reaction product was separated, and the appropriate amount of supernatant was taken and added to the internal standard to prepare the standard solution. The content of each substance was quantitatively analysed by gas chromatography.

The conversion of 1,2-BDO, and the selectivity of 1,2-BC are calculated as follows:

$${{\rm{Conv}}}.{{\rm{of}}}\,1,2-{{\rm{BDO}}}=\frac{{{{\rm{n}}}}_{{{\rm{initial}}}}\left(1,2-{{\rm{BDO}}}\right)-{{{\rm{n}}}}_{{{\rm{residual}}}}\left(1,2-{{\rm{BDO}}}\right)}{{{{\rm{n}}}}_{{{\rm{initial}}}}\left(1,2-{{\rm{BDO}}}\right)}\times 100\%$$
(1)
$${{\rm{Sel}}}.{{\rm{of}}}\,1,2-{{\rm{BC}}}=\frac{{{\rm{n}}}\left(1,2-{{\rm{BC}}}\right)}{{{{\rm{n}}}}_{{{\rm{initial}}}}\left(1,2-{{\rm{BDO}}}\right)\times {{\rm{Conv}}}.{{\rm{of}}}\left(1,2-{{\rm{BDO}}}\right)}\times 100\%$$
(2)

Catalyst stability test

Firstly, Bi1-CeO2 was sequentially pressed and sieved and then the 40-60 mesh Bi1-CeO2 particles were filtered out using a sieve. Then, it was loaded into a stainless-steel tube with a length of 300 mm and an inner diameter of about 8 mm, and the top and bottom were filled with quartz sand. A mixed DMC/1,2-BDO solution was configured and transported to the inside of the reactor via an advection pump. The reaction solution was collected at the bottom of the reactor and sampled every 16 h interval.