Rational Design of Synergistic Structure Between Single-Atoms and Nanoparticles for CO2 Hydrogenation to Formate Under Ambient Conditions

Single-atom catalysts (SACs) as the new frontier in heterogeneous catalysis have attracted increasing attention. However, the rational design of SACs with high catalytic activities for specified reactions still remains challenging. Herein, we report the rational design of a Pd1-PdNPs synergistic structure on 2,6-pyridinedicarbonitrile-derived covalent triazine framework (CTF) as an efficient active site for CO2 hydrogenation to formate under ambient conditions. Compared with the catalysts mainly comprising Pd1 and PdNPs, this hybrid catalyst presented significantly improved catalytic activity. By regulating the ratio of Pd1 to PdNPs, we obtained the optimal catalytic activity with a formate formation rate of 3.66 molHCOOM·molPd −1·h−1 under ambient conditions (30°C, 0.1 MPa). Moreover, as a heterogeneous catalyst, this hybrid catalyst is easily recovered and exhibits about a 20% decrease in the catalytic activity after five cycles. These findings are significant in elucidating new rational design principles for CO2 hydrogenation catalysts with superior activity and may open up the possibilities of converting CO2 under ambient conditions.


INTRODUCTION
In recent years, single-atom catalysts (SACs) have attracted significant interest in heterogeneous catalysis for their advantages of 100% metal atom use, single active site structure, and unexpected high activity and selectivity for various reactions (Yang et al., 2013;Wang et al., 2018). Since the concept of SACs was proposed in 2011 (Qiao et al., 2011), it has solved the problems of low metal loadings (Liu et al., 2016;Li et al., 2018), poor thermal stability (Jones et al., 2016;Liu et al., 2020a;Liu et al., 2020b), and difficulties in large-scale production (Liu et al., 2020b;He et al., 2020) after nearly 10 years of vigorous development. However, although SACs are presented as homogeneous active centers, their catalytic activity is often difficult to compare with that of homogeneous catalysts and enzyme catalysts, and even lower than that of the corresponding nanocatalyst system in some cases (Ding et al., 2015). One of the reasons is that it is often difficult for a single metal atom center to activate multiple reactants with different properties or a single polyatomic molecule. The catalytic activity of monatomic catalysts does not depend on their single active center, but also on their surrounding chemical environment. The design and controllable construction of the synergistic structure between the single-atom center and its surrounding active sites is the key to achieving the high-efficiency catalytic performance of SACs.
Synergistic catalysis has been widely studied in the field of nanocatalyst and has shown obvious advantages in improving the activity and product selectivity of catalysts. For example, Liu et al. (2017) designed a Schiff-base-mediated gold catalyst for hydrogenation to formate reaction. The Schiff-base functional group grafted on a SiO 2 support helps CO 2 activation by the formation of a weak carbamate zwitterionic intermediate, and Au is for H 2 dissociation. We previously introduced a CeO 2 promoter with excellent dissociation ability to water molecules and oxygen molecules into an Au/MgGa 2 O 4 catalyst to establish the synergistic catalytic effect between nano Au and the CeO 2 promoter, which significantly improved the catalytic activity for the water gas shift and catalytic combustion reaction (Ren et al., 2019). As for SACs, although the concept of an atomic scaled synergistic effect has not been clearly proposed and systematically studied, there are some research examples of synergistic effects. For example, we recently reported a highly active dual single-Pd-atom catalyst, which could catalyze the hydrogenation of CO 2 to formate under ambient conditions (Ren et al., 2022). It was found that the pore enrichment effect of microporous structures and the ternary synergetic effect among two neighboring Pd atoms and a rich nitrogen environment were the main reasons for this extraordinary catalytic activity. Liu et al. (2018) prepared a Ni-N-C SAC with metal loading up to 7.5 wt% by using N-doped C as the support, which exhibited excellent catalytic activity and cyclic stability for the hydrogenolysis of cellulose to ethylene glycol. The theoretical calculation results showed that the H 2 molecule was activated by assistance of the nearest uncoordinated pyridine N atom of Ni. Therefore, the design of the synergistic catalytic structure based on the properties of reactants provides a good idea for the rational design of highly effective SACs.
In this work, we focused on the synergistic effect between single atoms and nanoparticles for CO 2 hydrogenation to formate, an attractive reaction to achieve CO 2 emission reduction and safe hydrogen storage (Alvarez et al., 2017;Eppinger and Huang, 2017;Su et al., 2019). Through theoretical calculations, we found that Pd 1 and Pd NPs were the preferred active sites for CO 2 activation and hydrogen dissociation, respectively. Experimentally, we synthesized a Pd 1 -Pd NPs synergistic structure on 2,6pyridinedicarbonitrile-derived covalent triazine framework (CTF) by modulating the Pd loadings and reduction time, which exhibited high efficiency for the ambient hydrogenation of CO 2 to formate. The optimal catalytic performance of the catalyst was obtained by regulating the ratio of Pd 1 to Pd NPs . Furthermore, the recycling stability was also investigated. This work provides a new strategy for the rational design of highly active SACs and is very advantageous with regard to putting forward the conversion CO 2 into practical applications.

EXPERIMENT Chemicals
All the chemicals used in this study were of analytical grade and were used without further purification unless otherwise noted.

Sample Preparation
The CTF-400: 2,6-DCP-derived CTF was synthesized as described elsewhere (Kuhn et al., 2008). In detail, 2,6-pyridinedicarbonitrile and anhydrous ZnCl 2 were mixed at the ratio of 1:5 (w/w) and ground under the glove box. The mixed powder was transferred into a quartz ampoule tube, and then evacuated, sealed, and heated to 400°C for 40 h. After the reaction, the mixture was subsequently ground and washed with large amounts of water and diluted HCl (2 M) to remove the residual ZnCl 2 . After that, the resulting black powder was dried in a vacuum at 150°C for 12 h, and the resulting product was denoted to be "CTF-400." Pd/[CTF-400]: Pd/[CTF-400] catalysts were prepared by soaking the CTF-400 support powder in the aqueous solution of palladium trifluoroacetate with a Pd nominal weight loading of 10 wt% and 0.5 wt% for 12 h with magnetic stirring in a N 2 atmosphere. The suspensions were then filtrated and washed with deionized water. The resulting filter cake was dried at 80°C for 12 h under a vacuum. The samples were denoted as 10Pd/[CTF-400] and 0.5Pd/[CTF-400], respectively, and the Pd actual weight loadings were detected to be 6.48 wt% and 0.28 wt% from ICP-AES analysis, respectively. 10Pd/[CTF-400]-R-t samples were achieved by hydrogen reduction at 300°C for a different time. Specifically, the sample was heated to 300°C with the programming of 5°C/min under a constant flow of nitrogen. Then, the sample was kept at 300°C for 5, 10, 15, 30, 60, and 180 min under a constant flow of 10% H 2 /90% N 2 at 10 ml/min, respectively. Subsequently, the furnace was cooled down to room temperature under the protection of nitrogen.

Characterization
Fourier-transform infrared spectra (FT-IR) were performed in transmission mode on a Bruker VERTEX 70v spectrometer equipped with a DLATGS detector. The sample was diluted with KBr powder.
Surface areas and pore size distribution analyses were measured on Quantachrome Autosorb-iQ. N 2 was used as the adsorbate and surface areas were calculated using the BET analysis method. All of the samples were degassed under a vacuum at 260°C for 8 h before measurement.
Elemental analyses were performed on a Vario El elemental analyzer.
X-ray photoelectron spectroscopy (XPS) data were analyzed on a Thermofisher ESCALAB 250Xi spectrometer using a monochromatized Al Kα X-ray source (1,486.6 eV). All the XPS data were calibrated by using C 1s binding energy at 284.8 eV.
Transmission electron microscopy (TEM) analysis was performed on FEI Tecnai G2 F20 at 200 keV. Aberration-corrected scanning transmission electron microscopy (AC-STEM) and EDX mapping analysis were performed on a JEOL JEM-ARM200F.

Catalytic Reactions
CO 2 hydrogenation to formate reaction was carried out in a base solution under ambient conditions (30°C and 1 bar). In general, 20 mg catalysts were added into 5 ml NaHCO 3 (1 mol/L) in a three-necked bottle connected to a balloon. Then, the feed gas comprising CO 2 (50% vol%) and H 2 (50% vol%) was introduced after purging the residual air. After stirring for 12 h under ambient conditions, formate in the reaction mixture was determined by high-performance liquid chromatography (HPLC). In recycling experiments, the catalyst was recovered by filtration, washed with water, and dried under a vacuum. The reaction rate was calculated according to the following equation:

Rate
Concentration of formate mol

Computational Methods
The calculation was performed by using the M06l/6-31G* method for nonmetal elements whereas the M06l/LANL2DZ method was used for metal atoms (Hehre et al., 1972;Hay and Wadt, 1985;Zhao and Truhlar, 2006). The metals were augmented with the corresponding LANL2DZ pseudopotential, which was both acceptable in precision and timeconsuming. Vibrational frequencies of the optimized configurations were analyzed to validate that these configurations correspond to the local minima or transition state (TS). The TS with one imaginary frequency was found and verified by the intrinsic reaction coordinate (IRC) method (Fukui, 1981). The DFT calculations were performed on the Vienna ab initio simulation package (VASP) (Kresse and Furthmuller, 1996b;Kresse and Furthmuller, 1996a) to investigate the CO 2 hydrogenation process on a Pd bulk surface and Pd 1 /CTF. The optB88-vdW was used to describe the exchange-correlation functional, which described the van der Waals forces appropriately (Dion et al., 2004;Lee et al., 2010). The projector augmented wave (PAW) potentials (Blochl, 1994) were used for electron-ion interactions, with a plane-wave kinetic energy cutoff of 400 eV. The geometry structures were relaxed until the forces on all atoms were less than 0.05 eV/Å. The transition states were searched using the Climbing Image Nudged Elastic Band (CI-NEB) method (Henkelman et al., 2000). Each transition state was relaxed until the forces on all atoms were less than 0.05 eV/Å. The Pd (1 1 1) surface was modeled by a three-layer slab with a (4 × 4) surface unit cell and a vacuum thickness of 20 Å. The bottom two atomic layers of Pd (1 1 1) were fixed while the remaining layer, together with the adsorbates, were fully relaxed during relaxation. The lattice constant for Pd 1 /CTF is the same as for Pd, with a size of 13.79 Å × 13.79 Å × 24.51 Å. The Brillouin zone was sampled using a (2 × 2×1) k-point grid based on the Monkhorst-Pack (1976) scheme.

Theoretical Predictions
To achieve the most active catalyst structure, two different Pd species, Pd 1 and Pd NPs , were investigated to carry out the CO 2 activation and H 2 dissociation process. We found that 2,6pyridinedicarbonitrile-derived CTF-coordinated Pd 1 exhibited much higher adsorption energy for the CO 2 molecule than that of the Pd (111) surface (−1.15 vs. −0.28 eV), indicating that Pd 1 could be served as the active site for CO 2 activation ( Figures 1A,B). However, further calculation showed that the dissociation of the hydrogen molecule on Pd 1 was difficult in thermodynamics. In contrast, the dissociation of the hydrogen molecule on the Pd surface is thermodynamically feasible (Supplementary Table S1). Moreover, metallic Pd has also been proved to have high activity in H 2 dissociation with a nearly zero activation barrier (Ni and Zeng, 2009;Lozano et al., 2010). Therefore, we infer that Pd 1 and Pd NPs catalyze CO 2 hydrogenation in collaboration through the hydrogen molecule dissociation on Pd NPs and CO 2 hydrogenation on Pd 1 . The potential energy surface (PES) of CO 2 hydrogenation to formic acid on this hybrid catalyst is shown in Figure 1C. The first hydrogenation step is the rate-determining step with a barrier energy of 1.37 eV. The second step of hydrogenation is exothermic, with an energy of 2.90 eV. The energy barrier of this step is only 0.12 eV. The theoretical calculations confirm that the hydrogenation of CO 2 catalyzed by the Pd 1 -Pd NPs synergistic structure has a low barrier and can occur under ambient conditions.

Catalyst Synthesis and Characterization
The 2,6-pyridinedicarbonitrile-derived covalent triazine framework is formed by the trimerization of 2,6-pyridinedicarbonitrile in molten ZnCl 2 at 400°C for 40 h, and is labeled CTF-400 ( Figure 2A). FT-IR, shown in Supplementary Figure S1, confirmed the formation of the corresponding covalent triazine rings. In addition, the prepared CTF-400 possessed a surface area of 418 m 2 g −1 , a pore size of 0.53 nm, and a total pore volume of 0.21 m 3 g −1 (Supplementary Table S2). The nitrogen content was estimated to be 19.51 wt% (Supplementary Table S3), which is high enough to load and stabilize the supported Pd species. The Pd/[CTF-400] catalyst was prepared by soaking the support powders in aqueous palladium trifluoroacetate followed by filtrating and washing after stirring for 12 h under a N 2 atmosphere. By varying the initial Pd loading percent to be 0.5 wt% and 10 wt%, the single-atomic dispersed sample and the coexistence of Pd 1 and Pd NPs samples were prepared. The samples were denoted to be 0.5Pd/[CTF-400] and 10Pd/[CTF-400], respectively. After further reduction of 10Pd/[CTF-400] at 300°C for 3 h, the Pd NPs -dominated sample was also achieved as the reference, which was denoted to be 10Pd/[CTF-400]-R-3h. Figures 2B-D    Frontiers in Chemistry | www.frontiersin.org July 2022 | Volume 10 | Article 957412 4 ( Figure 2B). Scattered Pd nanoparticles appear on the 10Pd/ [CTF-400] sample, which illustrates that Pd atoms on CTF-400 have partially aggregated to nanoparticles ( Figure 2C). Further observation of this sample using AC-STEM confirms that Pd single atoms and nanoparticles coexist with good contact (Figures 2E,F). However, for the 10Pd/[CTF-400]-R-3h sample, a large amount of Pd nanoparticles with a major size of~3 nm was presented. The formation of Pd NPs was due to the aggregation of highly dispersed Pd atoms. For the as-prepared catalyst, most of the Pd species are well dispersed as single-atoms and sub-nanoparticles (as shown in Figures 2E,F), which have high surface energy and are unstable at high temperatures and reductive atmospheres. Therefore, the percentage of Pd NPs in the 10Pd/[CTF-400]-R-3h sample increased after reduction for 3 h. In addition, electronic properties for both samples were also estimated by XPS. As shown in Figure 2G, the binding energy of Pd 3d 5/2 in 0.5Pd/[CTF-400] is 337.3 eV, which can be attributed to that for Pd 2+ presented as Pd 1 in geometry. In addition to Pd 2+ , 10Pd/[CTF-400] also shows the characteristic of metallic Pd corresponding to Pd NPs in geometry with a binding energy of Pd 3d 5/2 at 335.2 eV, and the ratio of Pd 2+ to Pd 0 is estimated to be about 90%. Different from that of 10Pd/[CTF-400], Pd 0 is dominant in the 10Pd/[CTF-400]-R-3h sample with the ratio of Pd 0 to Pd 2+ being more than 70%. On the whole, the Pd 1 , Pd 1 -Pd NPs hybrid, and Pd NPs -dominated catalysts were successfully prepared, which can be used as good model catalysts to study the synergistic effect between different Pd active sites.

Catalytic Hydrogenation of CO 2 to Formic Acid Under Ambient Conditions
The catalytic performances of the as-prepared Pd catalysts for CO 2 hydrogenation were studied at 30°C in a H 2 /CO 2 mixture (0.1 MPa) with NaHCO 3 (1 mol/L) as an additive in the liquid phase. After a reaction of 12 h, the formate was detected by using HPLC. As shown in Figure 3A and Table 1, compared with the Pd 1 and Pd NPs nanoparticle-dominated catalysts, the CO 2 hydrogenation activity of the 10Pd/[CTF-400] catalyst was significantly enhanced (entries 1, 2 and 8 in Table 1 Moreover, the heterogeneous nature of this catalyst allows it to be easily recovered by centrifugation, and the recycling tests indicated that there was around a 20% decrease in the catalytic activity after five uses ( Figure 3B). This result demonstrates that the catalyst can be reused after a simple separation process, which is very advantageous with regard to practical applications. The aforementioned result confirms the theoretical predictions that the coexistence of Pd 1 and Pd NPs is necessary for the activation of H 2 and CO 2 at the same time. To obtain the optimized catalytic activity, the ratio of Pd 1 to Pd NPs was optimized. We attempted to get the sample with optimal Pd 1 /Pd NPs by lowering the 10Pd/[CTF-400]  sample into a tubular furnace at 300°C and reducing it in 10 vol% H 2 for 5, 10, 15, 30, and 60 min, respectively. The corresponding samples are denoted as 10Pd/[CTF-400]-R-t, where "t" represents the reduction time. X-ray photoelectron spectroscopy (XPS) was performed to get the ratios between Pd single atoms and nanoparticles ( Figure 4A and Supplementary Figures  S3-S7). With careful deconvolution from the overlapped peaks of Pd 2+ 3d and Pd 0 3d, the ratios of Pd 1 ion to the total palladium were estimated to be 89.9, 46.7, 43.1, 38.7, 34.3, and 33.7% with the continuous extension of hydrogen reduction time from 0 to 60 min ( Figure 4A). Figure 4B shows the catalytic performances of 10Pd/[CTF-400]-R-t under ambient conditions (30°C, 0.1 MPa). It can be seen that the catalytic activities exhibit a volcano-type curve with a decrease in the Pd 1 ratio. 10Pd/[CTF-400]-R-10 min with a Pd 1 ratio of 43.1% performs the best catalytic activity, and the formate formation rate reaches 3.66 mol HCOOM ·mol Pd −1 ·h −1 . The existence of the optimal Pd 1 ratio is due to the rate equilibrium of hydrogen dissociation on Pd 0 nanoparticles and carbon dioxide activation on Pd 1 .
Based on the aforementioned analysis, the synergistic effect mechanism of 10Pd/[CTF-400] for the hydrogenation of CO 2 to formate is shown in Figure 5. The 10Pd/[CTF-400] catalyst integrates both Pd NPs and Pd 1 into one catalyst system, where the Pd NPs boosts the dissociation of H 2 whereas Pd 1 ions undertake the activation task of CO 2 . Through the atom diffusion process, H atoms generated at Pd NPs move to the adsorbed CO 2 on Pd 1 for high-efficiency hydrogenation. Theoretical calculations shown in Figure 1 confirmed the rationality of the tasks over each active site, that is, hydrogen dissociation occurs more easily on Pd NP , and carbon dioxide hydrogenation activation prefers to occur on Pd 1 . Experiments further verified the synergistic effect between Pd 1 and Pd NPs , and the 10Pd/[CTF-400] catalyst performed nearly two orders of magnitude higher activity than 0.5Pd/[CTF-400] and twice the reactivity of 10Pd/[CTF-400]-R-3 h.

CONCLUSION
In summary, we have rationally designed a highly efficient catalytic system for the hydrogenation of carbon dioxide to formate under ambient conditions based on theoretical predictions. Through modeling the CO 2 adsorption and hydrogen dissociation process on both Pd 1 and Pd NPs , it was found that Pd 1 performed the higher adsorption energy for CO 2 and could be a potential candidate for CO 2 activation. Compared with Pd 1 , hydrogen dissociation occurred more easily on Pd nanoparticles. Based on this prediction, the Pd/ [CTF-400] catalyst integrating both Pd 1 and Pd NPs on one catalyst system was synthesized and realized the hydrogenation of CO 2 to formate with a formate formation rate of 3.66 mol HCOOM ·mol Pd −1 ·h −1 under ambient conditions (30°C, 1 bar). This hybrid catalyst presented nearly two orders of magnitude higher than the catalyst containing bare Pd 1 and twice the reactivity of that containing bare Pd 0 nanoparticles. These discoveries may pave the way for the construction of active SACs  Frontiers in Chemistry | www.frontiersin.org July 2022 | Volume 10 | Article 957412