Xiao-Long Wangab,
Xiao-Yuan Tanab,
Jian Linc,
Hua Wangd,
Jing Sunb,
Zhong-Ning Xu
*b and
Guo-Cong Guo
b
aCollege of Chemistry and Materials Science, Fujian Normal University, Fuzhou 350007, P. R. China
bState Key Laboratory of Structural Chemistry, Fujian Institute of Research on the Structure of Matter, Chinese Academy of Sciences, Fuzhou 350002, P. R. China. E-mail: znxu@fjirsm.ac.cn
cCAS Key Laboratory of Science and Technology on Applied Catalysis, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, Dalian 116023, P. R. China
dSchool of Chemistry and Chemical Engineering, Yulin University, Yulin, Shaanxi 719000, P. R. China
First published on 18th August 2025
CO esterification involves processes using CO as the starting material and ester chemicals as products. Methyl formate, dimethyl oxalate and dimethyl carbonate are the main products. Traditional Pd nanoparticle (PdNP) catalysts have excellent catalytic performance for the CO esterification reaction. However, PdNP catalysts have problems such as high metal loading, low metal utilization and high cost. In recent years, single atom catalysts have developed rapidly due to their high atomic utilization and low loading. In this work, Pd1/(ZnMn2O4@MnO2) (Pd1/ZMO) was synthesized via a sol–gel method followed by a redispersion strategy-assisted calcination process. This Pd1/ZMO catalyst exhibits exceptional catalytic performance for CO esterification to methyl formate. Compared to the PdNP/α-Al2O3 catalyst, Pd single atom catalysts significantly reduce noble metal loading from 1 wt% to 0.25 wt%, while increasing the catalyst's mass activity by 4 times. More importantly, the Pd1/ZMO catalyst is quite stable at least for 150 hours under reaction conditions. And we have demonstrated the high dispersibility of the Pd single atom through AC-TEM, in situ DRIFTS and XPS. This work adopts the strategy of single atom catalyst, which controls the cost of the catalyst while achieving a 4-fold increase in metal utilization, providing important reference significance for the efficient utilization of noble metal and catalyst cost control.
MF is a significant chemical reagent commonly employed in the production of fungicides, gasoline additives, and pharmaceutical synthesis intermediates.2 A wide array of industrial chemicals can be derived from MF, including formic acid, formamide and methyl acrylate, among other fine chemicals. Additionally, MF has been identified as an excellent hydrogen storage material due to its non-toxic nature, ease of transportation, and high hydrogen content (8.4 wt%).3 There are several methods for producing MF, such as liquid-phase methanol carbonylation, one-step syngas synthesis and other methods.4–8 Among them, liquid-phase methanol carbonylation is the most commonly used method for industrial MF synthesis.
Our research group has developed a new method for CO esterification to MF.9 Through exploration, we found that Pd-based catalysts have good catalytic activity in CO esterification. We found that the Pd/α-Al2O3 catalyst has excellent performance for CO esterification to MF. However, Pd exists in the form of nanoparticles in the catalyst, which leads to a low utilization rate and high cost of noble metals in the catalyst. Single-atom catalysts have emerged as a focal point in recent years, finding widespread applications in various fields such as electrochemistry, photocatalysis, and traditional thermal catalysis. They exhibit outstanding performance and activity in reactions including water–gas conversion,10 C–C coupling,11 photocatalytic hydrogen production,12–14 and methanol reforming.15 Compared with nanoparticle catalysts, single-atom catalysts have the characteristics of high atomic utilization efficiency and high activity.16–18
This study synthesized Pd1/(ZnMn2O4@MnO2) (denoted as Pd1/ZMO) via a sol–gel method followed by a redispersion strategy-assisted calcination process. The combination of high temperature and strong interactions with the support material leads to the formation of ZnMn2O4@MnO2 heterojunctions on the support surface during calcination, thereby enhancing the metal–support interaction.19–21 Compared to the PdNP/α-Al2O3 catalyst, Pd1/ZMO reduced Pd loading from 1 wt% to 0.25 wt%, while increasing the catalyst's mass activity by 4 times. At the same time, there was no decrease in the stable performance during the 150-hour reaction. Furthermore, the selectivity for MF reached 96%, with a weight-time yield of 1318 g kgcat−1 h−1. Furthermore, by employing in situ electron paramagnetic resonance (in situ EPR) and diffuse reflectance infrared Fourier transform spectroscopy (in situ DRIFTS) analyses, complemented by mechanistic insights from prior studies, we propose a plausible reaction pathway for the CO esterification to MF. Experimental feasibility of scaling up the catalyst for CO esterification to MF from the nanoparticle to single-atom scale further validates the advantages of single-atom catalysts, thereby broadening their application fields.
The Pd1/ZMO catalyst prepared by the co-precipitation method was as follows: the PdNP/ZMO sample was treated with a prepared 1 M dilute nitric acid solution, stirred for 30 minutes, filtered, and washed several times until the pH of the filtrate approached 7. The precipitate was then dried overnight at 80 °C. Finally, the sample was heated in a muffle furnace at a rate of 1 °C min−1 to 350 °C and calcined for 2 h to produce the Pd1/ZMO catalyst.
The PdNP/α-Al2O3 catalyst was synthesized using the conventional impregnation stirring method.
Catalyst evaluation conditions: catalyst loading 200 mg each time, reaction gas space velocity 3000 h−1, reaction gas pressure 0.1 MPa, reaction gas ratio CO:
MN
:
H2 = 1.4
:
1
:
0.6, and evaluation temperature 120 °C.
Conversion of CO (%) = ([CO]in/[Ar]in − [CO]out/[Ar]out)/([CO]in/[Ar]in) × 100% |
Selectivity of MF (%) = (AMF × R − FMF/MMF)/(3 × ADMM × R − FDMM/MDMM + AMF × R − FMF/MMF + 2 × ADMC × R − FDMC/MDMC + 2 × ADMO × R − FDMO/MDMO) × 100% |
WTY of MF (g kgcat−1 h−1) = Mass of MF yield (g)/(Catalyst quality (kg) × Time (h)) |
Mass activity (mol gPd−1 h−1) = Mole of CO conversion (mol)/(Pd quality (g) × Time (h)) |
ADMM, AMF, ADMC and ADMO represent the peak areas of dimethoxymethane (DMM), methyl formate (MF), dimethyl carbonate (DMC) and dimethyl oxalate (DMO).
R − FDMM, R − FMF, R − FDMC and R − FDMO represent the relative correction factors of DMM, MF, DMC and DMO. The relative correction factor is calculated by standard sample injection.
MDMM, MMF, MDMC and MDMO represent the relative molecular mass of DMM, MF, DMC and DMO.
The selectivity of MF is calculated based on the MN reactant.
The mass activity is calculated based on the loading of noble metal Pd in the catalysts.
As shown in Fig. 3a–f, SEM images reveal distinct differences in the surface morphology of the catalyst before and after nitric acid etching. In Fig. 3a and b, it is evident that the surface of the PdNPs/ZMO catalyst is relatively smooth prior to etching.24 Conversely, in Fig. 3c and d, the Pd1/ZMO catalyst obtained after etching and subsequent calcination exhibits numerous sheet-like structures on the spinel surface, contrasting sharply with the original smooth surface.25 In Fig. 3e and f, the Pd1/ZMO-150 catalyst still maintains the same sheet structure of the Pd1/ZMO catalyst According to the BET analysis presented in Table S2, the specific surface area of the catalyst increased from 30.79 m2 g−1 to 69.65 m2 g−1 following the etching process, indicating a more pronounced defect structure on the spinel surface. The etched catalyst demonstrates a larger specific surface area and enhanced pore structure, which facilitates the adsorption of reactant gases at the active sites for catalytic reactions. Raman spectroscopy analysis of the catalyst, illustrated in Fig. S2a–S2c, shows characteristic peaks at 305 cm−1 and 316 cm−1 corresponding to the Mn–O6 vibrations of the ZnMn2O4 spinel, while the peak at 664 cm−1 corresponds to Zn–O4 vibration.26 A decrease in the intensity of the Mn–O6 peak, along with a redshift in its position post-etching, suggests an increased density of defects on the catalyst surface.27
During the catalyst synthesis, we employed diluted nitric acid treatment to etch a portion of the Zn2+ ions from the catalyst. Subsequent high-temperature treatment resulted in the formation of a heterostructure comprising ZnMn2O4@MnO2. By controlling the etching time with diluted nitric acid, we could regulate the content of Zn2+ in the framework, thereby controlling the concentration of cation vacancies.28 The subsequent calcination introduced oxygen vacancies at the zinc-deficient cation sites, facilitating the formation of MnO2 with Mnδ+ and establishing electron transfer at the surface of the ZnMn2O4 spinel, resulting in the formation of the heterostructure.29 As depicted in Fig. S3a and S3b, TEM images of PdNP/ZMO and Pd1/ZMO reveal that the catalysts exhibit an elliptical, pebble-like morphology. Lattice spacing measurements indicate that the ZnMn2O4 (101) planes in Pd1/ZMO are aligned with MnO2 (111) planes, confirming the formation of a heterostructure on the surface following etching and calcination.
From Fig. 4a, the TEM image of the PdNP catalyst shows Pd nanoparticles, with a size of approximately 0.5 nm as measured. As shown in Fig. 4b, the AC-TEM of the Pd1/ZMO catalyst can clearly observe the highly dispersed single-atom state of Pd on its surface. Fig. 4c shows the elemental mapping images of the Pd1/ZMO catalyst, which shows that after etching and high-temperature treatment, Pd still exhibits high dispersion at a scale of 50 nm. As shown in Fig. S4, in the mapping diagram of the Pd1/ZMO catalyst after 150 h reaction, Pd still maintains a high level of decentralization. It indicates that Pd1 has good stability in the single-atom catalyst.30
ICP elemental analysis, as presented in Table S1, determined the Pd loading in Pd1/ZMO to be around 0.25 wt%. Following the etching with diluted nitric acid, a noticeable decrease in Zn content was observed. XPS analysis of the full elemental spectrum also indicated a significant reduction in the surface intensity of Zn.31 Elemental mapping of Pd1/ZMO further illustrated that some Zn2+ was etched away by the diluted nitric acid, while Mn was uniformly distributed both on the surface and within the catalyst, providing additional confirmation of the heterostructure's existence.32
Additionally, TOF-SIMS analysis in the negative ion mode, as illustrated in Fig. S6, S7 and Table S3, identifies a Pd–O–M signal, supporting the conclusion that Pd primarily exists as Pd2+ single atoms. The MnO2 signal intensity on the catalyst surface is substantially higher than that of ZnMn2O4, indicating the formation of a ZnMn2O4@MnO2 heterostructure post-etching. The presence of Pd–Zn and Pd–Mn bonds on the catalyst surface suggests that single-atom Pd is integrated within or bonded to the ZnMn2O4 lattice, contributing effectively as catalytic sites.33
Further XPS examination of Pd oxidation states, as shown in Fig. 5a and b, shows that Pd in PdNP/ZMO exists predominantly in the Pd0 state, with both Pd nanoclusters.27 In Pd1/ZMO, Pd is solely in the Pd2+ form as Pd–O–M. As shown in the wide scan in Fig. S5a, the relative peak intensity of Zn 2p significantly decreases in Pd1/ZMO after etching, indicating extensive Zn dissolution in dilute nitric acid, while the Pd 3d peak intensity remains largely unchanged, aligning with expectations. Fitting analysis for the Mn elements, as presented in Fig. S5b and S5c, reveals Mn in the +2, +3, and +4 oxidation states within Pd1/ZMO post-etching, suggesting the formation of a ZnMn2O4@MnO2 heterostructure upon re-calcination.34 SEM characterization confirms increased surface roughness and the presence of flaky structures after etching, indicative of MnO2 nanosheets on the catalyst surface. After 150 hours, the Pd 3d XPS spectra reveal the coexistence of Pd0 and Pd2+ species, indicating a dynamic valence state during the reaction (Fig. S11). Under the CO and H2 atmosphere, Pd2+ is partially reduced to the Pd0 state, whereas the MN containing reactive gas oxidizes Pd0 back to Pd2+. The absence of Pd clusters after 150 hours demonstrates that this redox cycling maintains an active electronic equilibrium around each Pd atom, thereby preserving its atomically dispersed state and ensuring stable catalytic performance.
![]() | ||
Fig. 5 XPS spectra of Pd 3d of (a) PdNPs/ZMO and (b) Pd1/ZMO; the in situ CO-DRIFTS of (c) PdNPs/ZMO and (d) Pd1/ZMO. |
XPS analysis of the oxygen element in the catalyst, as depicted in Fig. S5d and S5e, reveals that Pd1/ZMO, post-etching, exhibits a greater abundance of lattice oxygen and adsorbed oxygen peaks compared to PdNP/ZMO. The more pronounced Oads peak in Pd1/ZMO can be attributed to the formation of cationic vacancies from the etching of Zn2+, which, after calcination, introduces additional oxygen coordination. Additionally, Pd1/ZMO has a larger specific surface area and porosity than PdNP/ZMO, which facilitates the exposure of Pd2+ single atoms on reactive surfaces. This exposure promotes CO adsorption and activation, enhancing further reactions with MN. Additionally, EPR measurements, as depicted in Fig. S8a, present a sinusoidal pattern in the G-value graph, with the g value determined to be 2.009, confirming the presence of oxygen vacancies.35 Oxygen vacancies act as robust anchoring sites for individual Pd atoms, thereby inhibiting their surface diffusion and agglomeration under the reaction conditions. This is evidenced by the atomically dispersed Pd signals observed in HAADF-STEM images. Furthermore, these vacancies modulate the electronic structure of Pd and its adsorption properties.
The in situ CO-DRIFTS studies, as shown in Fig. 5c and d, reveal that peaks in the 1800–2000 cm−1 range correspond to bridge-type adsorption of CO on Pd, while those between 2000 cm−1 and 2100 cm−1 are attributed to linear adsorption on Pd. Peaks above 2100 cm−1 is attributed to the linear adsorption or multiple adsorptions of Pd in the high energy state. The peak observed near 2200 cm−1 corresponds to gaseous CO and was minimized by continuously purging with N2 during the measurement to reduce interference from gas-phase species. For the PdNP/ZMO catalyst, prominent bridged CO adsorption bands appear at approximately 1916 cm−1 and 1986 cm−1, indicating that bridge-type adsorption dominates on Pd nanoparticles. Additionally, a weaker band at 2090 cm−1 is assigned to linearly adsorbed CO, suggesting that a small portion of Pd exists as isolated single sites in PdNP/ZMO. In contrast, the Pd1/ZMO catalyst exhibits only a single linear CO adsorption band at 2086 cm−1, confirming that Pd is exclusively present in a highly dispersed single-atom configuration.36
![]() | ||
Fig. 6 (a) Catalyst performance comparison diagram; (b) 150-hour reaction performance diagram of Pd1/ZMO. |
The catalyst active metal is Pd. During the CO esterification reaction to generate MF, Pd interacts with various reactant gases, leading to the formation of different intermediates, which subsequently react to form the final product MF. The overall three-component reaction is as follows:
2CH3ONO + 2CO + H2 → 2HCOOCH3 + 2NO |
In this reaction system, the primary processes involve the coupling cleavage of Pd and MN, generating the Pd–OCH3 intermediate. Pd also adsorbs CO, forming Pd–CO, and these intermediates can further generate Pd–COOCH3, a crucial intermediate in CO esterification. Additionally, H2 interacts with Pd, forming Pd–H through homolytic cleavage.
*OCH3 + *CO → *COOCH3 |
*COOCH3 + ·H → HCOOCH3 |
To validate this hypothesis, a series of characterization experiments were performed. First, in situ CO-DRIFTS tests were conducted. As seen in Fig. 5c and d, both PdNP/ZMO and Pd1/ZMO adsorb CO. However, due to better dispersion, Pd1/ZMO prefers linear adsorption, while PdNP/ZMO shows bridge adsorption. While both adsorb CO, this single characterization does not provide a conclusive explanation. Thus, in situ CO-MN DRIFTS characterization was also conducted. As shown in Fig. 7a, b and S9, the peaks at 1739 and 1192 cm−1 are ascribed to the CO and C–O stretching vibrations of adsorbed *COOCH3 intermediates. Moreover, the peak at 1444 cm−1 is assigned to the C–H deformation of adsorbed *COOCH3 intermediates on the catalyst surface. Strong new peaks appear at 1765 and 1269 and 1297 cm−1 with the passing of time, which are ascribed to the C
O and C–O stretching vibrations of gaseous MF product, respectively. Meanwhile, the peaks appeared at 1901 and 1843 cm−1 are assigned to the bimodal peaks of gaseous NO, indicating that NO was also produced during the reaction. Additionally, the peak at 2960 cm−1 is assigned to the C–H stretching vibration of the methyl group.
![]() | ||
Fig. 7 The in situ CO-MN DRIFTS of (a) PdNP/ZMO and (b) Pd1/ZMO; the in situ H2-DRIFTS of (c) PdNP/ZMO and (d) Pd1/ZMO. |
For our reaction system, which involves H2, the formation of MF is more likely to proceed via the L–H mechanism, which has a lower reaction barrier. The process is as follows: Pd adsorbs CO linearly, forming Pd–CO. MN adsorbs and dissociates on Pd, forming free radicals that bond with Pd to form Pd–OCH3. Pd–OCH3 couples with Pd–CO and undergoes electron transfer to form Pd–COOCH3. The introduction of H2 leads to its dissociation into ·H free radicals under Pd influence. Some of these ·H free radicals bond with Pd to form Pd–H. Pd–COOCH3 and Pd–H undergo coupling to form MF.
According to reports, the rate-determining steps in the formation of DMC, DMO and MF are different. For MF, the key rate-determining step is the coupling reaction between Pd–H and Pd–COOCH3 during the hydrogenation step of the experimental reaction.38
*COOCH3 + ·H → HCOOCH3 |
To further investigate this mechanism, in situ monitoring was performed in a 3% H2 atmosphere. As seen in Fig. 7c and d, clear differences between PdNP/ZMO and Pd1/ZMO catalysts can be observed. Pd1/ZMO shows a more pronounced M–H bond vibration peak due to stronger interaction with H2. In contrast, PdNP/ZMO shows a weaker M–H bond peak, indicating weaker adsorption of H2 and consequently lower catalytic activity in the reaction.39–41
It is difficult to directly observe the ·H free radicals from H2. Therefore, free radical scavengers such as 5,5-dimethyl-1-pyrroline N-oxide (DMPO) were added to the test to capture them for testing. The in situ H2-EPR characterization was carried out to observe the interaction mechanism between the active metal Pd and H2. As shown in Fig. S8b, the intensity of hydrogen free radical peaks was compared. It is clear that, compared to catalysts with active Pd, the ZnMn2O4 support generates some free radicals, but their intensity is weaker. The overall peak shape and peak area ratio prove the presence of hydrogen free radicals and multiple equivalent hydrogen atoms coupled together.42,43 Additionally, considering that the support contains metals such as Zn and Mn that can adsorb and partially activate H2 to form hydrogen free radicals, the EPR spectra for PdNP/ZMO and Pd1/ZMO show little difference in peak shape, position, and intensity. However, based on previous in situ H2-DRIFTS analysis, PdNPs can generate ·H free radicals but with a weaker M–H peak, indirectly indicating that the generated ·H free radicals do not extensively interact with the active Pd, resulting in lower reaction activity. In contrast, Pd1 single atoms not only generate ·H free radicals but also facilitate their bonding with Pd to form Pd–H bonds, thus promoting the coupling with Pd–COOCH3 to form MF.
In summary, this study, through systematic experimental and theoretical analysis, provides an in-depth understanding of the mechanism by which single-atom catalysts catalyze the CO esterification to generate MF, as shown in Fig. 8. To further elucidate the reaction mechanism of CO esterification to methyl formate, we carried out DFT calculations and constructed a detailed energy profile (Fig. S12). The analysis reveals that the hydrogenation step requires the highest energy input among all elementary steps, indicating that it is the rate-determining step. Our results show that loading single-atom Pd onto a ZnMn2O4 support significantly lowers the activation energy of the reaction and facilitates the dissociation of H2 at the active sites, leading to the generation of highly active ·H free radicals. These ·H free radicals further couple with reaction intermediates, driving the overall reaction forward. The EPR and in situ DRIFTS data consistently support this mechanistic model. Future studies will focus on exploring the impact of the reaction conditions and catalyst structural adjustments on the mechanism to achieve more efficient and stable catalytic conversion processes.
The authors confirm that the data supporting the findings of this study are available within the article and its SI.
This journal is © The Royal Society of Chemistry 2025 |