Improving the Efficiency of Hydrogen Spillover by an Alkali Treatment Strategy for Boosting Formic Acid Dehydrogenation Performance

Abstract

Defect engineering has been demonstrated to be an attractive strategy to improve the catalytic performance of g−C₃N₄−based catalysts. Herein, three graphite carbon nitrides (labeled “CN”) containing a certain number of cyano groups and nitrogen vacancies are prepared successfully by calcination of the dicyandiamide−based CN in the presence of KOH, and the performances of the corresponding Pd−based catalysts are evaluated by using the formic acid (FA) dehydrogenation as a probe reaction. The characterizations of X−ray diffraction (XRD), scanning transmission electron microscopy (STEM), X−ray photoelectron spectra (XPS), hydrogen temperature−programmed desorption (H₂−TPD), and hydrogen spillover experiments indicate that the high catalytic activity of Pd/CNK−0.5 is mainly attributed to its high efficient hydrogen spillover, relatively high dispersity of Pd species, and basicity due to the introduction of a proper amount of cyano groups and nitrogen vacancies. The low initial activity of Pd/CNK−0.75 may mainly be ascribed to its low hydrogen spillover ability and the strongly chemisorbed hydrogen on Pd single atoms or small clusters.

1. Introduction

With population surge and economic development, energy resources have been excessively consumed, which is causing increasing consumption of fossil fuels constantly, such as coal, petroleum, and natural gas [1,2]. Consequently, the development of green and sustainable renewable energy is of paramount necessity to reduce the utilization of and dependence on traditional fossil fuels [3,4]. Hydrogen is widely considered a promising clean energy carrier due to its abundance, renewability, and high energy efficiency [5]. Compared with physical hydrogen storage, which faces challenges such as high costs, low efficiency, and leakage risks, chemical hydrogen storage offers notable advantages, including lower storage costs, higher hydrogen density, and improved transportation safety [6,7]. Among various chemical hydrogen storage methods, liquid organic hydrogen carriers (LOHCs) stand out due to their compatibility with existing fuel infrastructure and their physical similarities to conventional fuels like gasoline and diesel [8,9]. As a candidate compound of LOHCs, formic acid (FA) has attracted considerable attention for its high stability and low toxicity, especially compared with other alternatives, such as hydrazine, ammonia, and methylcyclohexane [10,11,12].

Generally, the decomposition of FA proceeds in two pathways: dehydrogenation and dehydration [13,14]:

(1)

HCOOH → H₂+ CO₂, ∆G = −48.4 kJ·mol⁻¹

(2)

HCOOH → H₂O+ CO, ∆G = −28.5 kJ·mol⁻¹

It is evident that the dehydrogenation pathway is the preferred route, as the dehydration pathway generates CO, which strongly adsorbs onto the active metal sites and leads to catalyst poisoning [15,16,17]. Consequently, the development of catalysts that exhibit both high activity and excellent selectivity toward the dehydrogenation of FA has become a central focus of current research. Over recent decades, both homogeneous and heterogeneous catalysts for efficient FA dehydrogenation have been extensively investigated [12,18,19]. As early as 1967, Coffey used transition metal complex catalysts with phosphine ligands for the decomposition of FA [20]. Although homogeneous catalysts exhibit extremely high catalytic activity, and with turnover frequency (TOF) being up to tens of thousands, their development and application are restricted by high costs, easy leaching of active components, and difficulties in separation and recovery. By contrast, heterogeneous catalysts can address these challenges, so conducting research on efficient heterogeneous catalysts has great significance for hydrogen production from FA [18,21,22]. Currently, heterogeneous catalysts based on transition metals and noble metals as active centers, such as Pt [23], Au [24,25], Ag [14], Co [26], and Pd [27,28], as well as their bimetallic and multi−metallic counterparts [29,30,31], have been widely studied and reported. Among these, supported Pd−based catalysts exhibit the highest catalytic activity. In particular, the activity of FA dehydrogenation is strongly correlated with the nature and type of catalyst supports. The main physicochemical properties of the carrier—including surface acidity/basicity, heteroatom doping, pore architecture, functional groups, and defect sites—exhibit profound impacts on catalytic performance [16,27,32,33]. To clarify the role of the metal/organic interface on FA dehydrogenation activity, Wang et al. grew NPs with an active component of CrPd on various supports, including 3−aminopropyltriethoxysilane (APTES)−functionalized monochlortriazinyl β−cyclodextrin (M−β−CD−A), carbon (C), TiO₂, ZrO₂, and other materials. Experimental results showed that M−β−CD−A−supported catalysts exhibited superior activity, in the following order: M−β−CD−A > C > TiO₂ > ZrO₂ > CeO₂ [34]. Similarly, Li et al. synthesized nitrogen−doped carbon (N−MSC−30) using urea and a two−step calcination strategy. Pd nanoparticles with an average size of merely 1.4 nm were uniformly dispersed on the N−MSC support, yielding a Pd/N−MSC catalyst with excellent performance (TOF = 8414 h⁻¹ at 60 °C) [35]. Moreover, alkaline conditions are favorable for hydrogen production from FA [36,37,38]. In many reported studies, sodium formate (HCOONa) has been introduced as an additive in FA solution [39]. On one hand, the presence of HCOO⁻ facilitates the dehydrogenation pathway for hydrogen generation. On the other hand, the alkalinity of HCOONa helps maintain the pH of the reaction system, stabilizes the reaction rate, and mitigates catalyst deactivation [40,41]. Luo et al. enhanced the alkalinity of the catalyst support by grafting −NH₂ groups onto SBA−15, which promoted the cleavage of the O−H bond in FA and improved the reaction rate. Additionally, the strengthened metal−support interactions contributed to the formation of highly dispersed nanoparticles [42].

Graphitic carbon nitride (CN) has emerged as a highly researched catalyst support due to its unique advantages: abundant source materials, facile synthesis, physicochemical stability, abundant active sites, nitrogen−rich functional groups, etc. Despite these advantages, its practical applications are limited by drawbacks such as low specific surface area and chemically inert stacking, which hinder catalytic efficiency [43,44,45,46,47]. Currently, researchers have developed various optimization strategies to enhance its catalytic performance, including morphology regulation, oxygen/phosphorus elemental doping, defect engineering design, and functional group modification [48,49,50,51]. Among them, introducing defect structures or functional groups into the g−C₃N₄ skeleton represents a key pathway to enhancing catalytic performance. This strategy can significantly boost catalytic activity by regulating the electronic structure between the support surface and metal active components, strengthening the anchoring effect on metal active sites to improve their dispersion while optimizing the adsorption capacity for reaction substrates [52,53,54]. Modified carbon nitride materials mainly include structural defects and functional groups such as carbon vacancies [55], nitrogen vacancies [56], cyano groups (−C≡N) [57], hydroxyl groups (−OH) [58], and carboxyl groups (−COOH) [59], with each conferring unique catalytic advantages. Cyano−functionalized carbon nitride has received significant attention due to its strong electron−absorbing ability and electron storage function. Current literature reports several −C≡N incorporation methods, including molten salt synthesis, assisted thermal polymerization, self−assembly, and chemical etching [60].

Supported ultrafine metal nanoparticles have been demonstrated to be a promising heterogeneous catalyst for FA dehydrogenation due to their more available surface active atoms [35]. However, ultrafine metal nanoparticles are easy to aggregate during the reaction due to their high surface energies and/or weak interactions with the supports. Developing highly efficient and stable metal heterogeneous catalysts for catalytic hydrogen generation from FA is still highly desired. Herein, we report an effective approach to successfully introduce cyano groups and nitrogen vacancies into the structure of carbon nitride by a post−treatment with a strong base. By precisely tuning the content of cyano groups and nitrogen vacancies, we successfully prepared Pd/CNK−X catalysts, which exhibit high activity and stability as well as excellent selectivity for hydrogen.

2. Results and Discussion

2.1. Characterization of CN and CNK−X

The surface microstructures of the CN and CNK−X supports were characterized by scanning electron microscopy (SEM) (Figure S1). All samples exhibit similar irregularly stacked sheet−like morphologies, with only slight etching traces observed on the surface of the CNK−X samples, indicating that KOH treatment does not significantly alter the overall morphology of CN. Nitrogen adsorption−desorption isotherms reveal typical type II curves accompanied by H3 hysteresis loops, indicating that there are stacked mesopores within the samples (Figure S2). As summarized in Table S1, with increasing amounts of KOH, the specific surface area of the supports decreases from 13 m²/g for CN to 8 m²/g for CNK−0.75; meanwhile, the average pore size increases from 16 nm to 28 nm, revealing that the etching effect of KOH results in obvious changes in the pore structure of CN. The XRD patterns display the crystal structures of pristine CN and CNK−X samples (Figure 1a). The peaks at around 13.0° and 27.7° are attributed to the (100) diffraction plane of in−plane structural packing motifs of heptazine units in the carbon nitride structure, and the characteristic interlayer spacing diffraction peak (002) of CN aromatic units, respectively [61]. With the increasing addition of KOH, the peak intensity decrease gradually, indicating the degradation of the long−range ordering of the in−plane structure. Meanwhile, the (002) diffraction signal shifts gradually from 27.7° to 27.9°, revealing the narrowing of the interlayer stacking distance. The chemical structures of CN and CNK−X are also supported by the Fourier transform infrared spectra (FT−IR) (Figure 1b). The sharp peak at 803 cm⁻¹ is attributed to the typical vibration mode of tris−triazine units, and the broad band in the region of 1200 cm⁻¹ to 1700 cm⁻¹ and the band at around 3050–3300 cm⁻¹ are assigned to the typical stretching vibration of heptazine rings and −NH_x or absorbed H₂O [62], respectively. The intensity of the peak at 803 cm⁻¹ for CNK−X decreases gradually, further showing that the chemical structure of CN changes with the KOH treatment. It is worth noting that a new peak at 2170 cm⁻¹ appears in the samples of CNK−0.25, CNK−0.5, and CNK−0.75 and strengthens in the same order, indicating the formation of cyano groups (−C≡N) within the CN skeleton.

An obvious electron paramagnetic resonance (EPR) signal centered at g = 2.002 shows the existence of unpaired electrons on paramagnetic aromatic carbon atoms (Figure 1c), which may result from the induction of defect sites [63]. The gradual increase in the C/N ratio for CN, CNK−0.25, CNK−0.5, and CNK−0.75 validates the formation of more nitrogen vacancies in the CNK−X samples (Table S2). Admittedly, the increase in the unpaired electron concentration may also be attributed to the incorporation of cyano groups, which effectively induce the delocalization of isolated valence electrons within the π−conjugated heteroaromatic framework owing to their strong electron−withdrawing nature [64]. The signal intensities of the Lorentzian line for CNK−X are much higher than that of CN, particularly for CNK−0.5 and CNK−0.75, reflecting the increase in nitrogen vacancies and cyano groups.

The surface element state of the samples is measured by X−ray photoelectron spectrum (XPS). The C 1s signals of CN comprise three peaks located at 284.8, 286.3, and 288.2 eV (Figure 2a), corresponding to C−C in adventitious hydrocarbon species, C−NH_x at heptazine unit edges, and N−C=N in typical aromatic CN heterocycles, respectively [65]. Although the −C≡N signal overlaps with C−NH_x due to the similar binding energies, the stronger C−NH_x signals of CNK−X samples comparing to pristine CN still reveals the formation of cyano group [66]. Moreover, the N−C=N peak gradually shifts to lower binding energy with the increase in KOH, showing the formation of more nitrogen defects in CNK−0.5 and CNK−0.75, which lead to a shift in the electron cloud toward the C atoms. The N 1s spectra can also be deconvoluted into three peaks at 398.7, 400.2, and 401.1 eV (Figure 2b), which are attributed to the pyridine nitrogen in the heptazine unit (C−N=C, N_2C), the ternary nitrogen (N−(C)₃, N_3C) at the center of the heptazine units, and the primary and secondary nitrogen in the −NH_x groups, respectively. The content of C−N=C in the CNK−X samples decreases progressively with the increase in KOH loading. Meanwhile, the ratio of N_2C/N_3C decreases significantly from 4.8 for CN to 3.0 for CNK−0.75 (Table S3), indicating that a few N_2C atoms of the tri−s−triazine unit are lost during the KOH treatment, resulting in the generation of nitrogen defects within the C₃N₄ structure. At the same time, the reduction in peak intensity of −NH_x can be attributed to the KOH−induced deprotonation of terminal −NH_x and the etching−induced cyano group formation, which is also corroborated by FT−IR characterization.

Solid−state ¹³C NMR spectroscopy of CN and CNK−X samples reveals two dominant resonance signals (Figure S3). The C1 signal is attributed to the carbon atoms in the −C−NH_x environment, while the C2 signal corresponds to sp²−hybridized carbon atoms within the conjugated heptazine ring. The C1/C2 intensity ratio increases from 1.83 for CN to 3.13 for CNK−0.75, indicating the progressive disruption of the heptazine framework in CN [67]. In addition, both peaks shift towards the high field gradually due to the formation of nitrogen vacancies and cyano groups, which leads to the redistribution of local electrons around carbon atoms. These observations collectively suggest that the KOH treatment induces the formation of cyano groups and nitrogen vacancies within the CN structure. Based on these results, a schematic representation of the modified CN structure is proposed (Figure 1d).

SO₂−TPD analysis was conducted to investigate the alkalinity of CN and CNK−X (Figure 3). It can be seen that all the samples show a wide SO₂ desorption peak in the range of 50–230 °C, attributed to the adsorption of SO₂ on the alkaline sites. However, the intensities of the SO₂ desorption peaks are quite different. Among them, CNK−0.75 displays the most intense desorption peak, followed by CNK−0.5, CNK−0.25, and CN in turn, indicating that the KOH treatment results in an obvious enhancement in the alkalinity of CN due to the introduction of cyano groups and nitrogen vacancies. It is well recognized that the alkalinity of catalyst supports is beneficial for enhancing the activity of the corresponding catalyst for FA dehydrogenation [42,68].

2.2. Catalytic Activity and Stability of Pd/CN and Pd/CNK−X

The reaction curves of Pd/CN and Pd/CNK−X in FA dehydrogenation are given in Figure 4a. Among them, Pd/CNK−0.5 exhibits the highest catalytic activity, which achieves complete FA conversion within 30 min at 60 °C in the presence of a ratio of 1 mol % Pd and gives a total gas evolution of 291 mL, which is much higher than that of Pd/CN (244 mL) after 90 min under identical conditions. The activity of Pd/CNK−0.25 is also slightly higher than that of Pd/CN. These data indicate that the KOH treatment is beneficial for improving the FA dehydrogenation activity of Pd/CNK−X. However, further increasing the KOH ratio to 0.75 leads to a decline in the catalytic performance of Pd/CNK−0.75. The activities of the four catalysts at all measured time points follow the order Pd/CNK−0.5 > Pd/CNK−0.25 > Pd/CN > Pd/CNK−0.75, which is consistent with the calculated TOF value (Table S4), indicating the importance of the introduction of proper numbers of cyano groups and nitrogen vacancies in the CN framework. The activity of Pd/CNK−0.5 is better than most of the reported Pd−based catalysts (Table S5). The composition of the gas obtained from FA dehydrogenation is determined by gas chromatography (Figure S4). There is only H₂ and CO₂, and no CO is detected, indicating the high selectivity of the catalysts for FA dehydrogenation.

The stability of Pd/CN, Pd/CNK−0.25, Pd/CNK−0.5, and Pd/CNK−0.75 in FA dehydrogenation was also investigated, and the results are shown in Figure 4b. Pd/CN and Pd/CNK−0.25 exhibit an obvious decrease in reaction rate and conversion efficiency during the first three cycles, and their gas production decreases from 236 mL and 249 mL for the first cycle to 184 mL and 202 mL for the third cycle, respectively. However, Pd/CNK−0.5 shows only a slight conversion rate reduction in the first three cycles, which still leads to complete FA conversion within 40 min in the third cycle. In the fourth and fifth cycles, the deactivation of Pd/CNK−0.5 intensifies to a certain extent, but its gas production within 90 min is still up to 270 mL, implying its relatively higher activity and stability comparing to Pd/CN and Pd/CNK−0.25. Interestingly, the catalytic activity of Pd/CNK−0.75 increases gradually in the first three cycles, although its initial activity (177 mL) is low. The gas production rises to 229 mL in the third cycle. Moreover, its activity is almost maintained at this level in the further cyclic experiments, showing the highest stability compared to the three others. These results suggest that the incorporation of cyano groups and nitrogen vacancies contributes positively to the stability of the catalyst.

The effects of reaction temperature on the FA dehydrogenation over Pd/CN, Pd/CNK−0.5, and Pd/CNK−0.75 were also investigated. The gas volumes produced by Pd/CN, Pd/CNK−0.5, and Pd/CNK−0.75 under high temperature at every detecting time point are higher than that at low temperatures, indicating that the reaction rates of the three catalysts are all highly sensitive to temperature. The plots of gas volume versus reaction time reveal an initial linear relationship between reaction time and FA concentration, showing that the FA dehydrogenation follows first−order kinetics (Figure S5). Based on the Arrhenius equation, the apparent activation energies for this reaction catalyzed by Pd/CN, Pd/CNK−0.5, and Pd/CNK−0.75 are calculated to be 43.72 kJ/mol, 37.77 kJ/mol, and 54.97 kJ/mol, respectively (Figure S6). Pd/CNK−0.5 exhibits the lowest activation energy, which is consistent with it having the highest catalytic activity.

The effects of FA concentration and catalyst dosage on FA dehydrogenation were investigated further over Pd/CNK−0.5. As shown in Figure 5a, the gas evolution volumes increase with the increasing FA concentration from 1.0 mol/L to 4.0 mol/L at the same time, which can be attributed to the high number of substrate molecules available. However, the FA conversions start to decrease when the FA concentration is above 2 M within the same reaction time (Figure 5b), and the time required for complete conversion is prolonged. This may be attributed to the decrease in H₂O concentration. H₂O molecules as a weak Brønsted base can accept the H⁺ donated by Brønsted acid site and promote FA decomposition [69]. From Figure 5c,d, one can see that the FA dehydrogenation rate increases almost linearly with the catalyst dosage in the early stage of reaction due to the increase in catalytic active sites. When the catalyst dosage is 0.6%, the reaction only generates 188 mL of gas within 30 min, corresponding to a FA conversion of approximately 64%. Increasing the catalyst dosage to 1.2% results in the production of 292 mL gas within 15 min, corresponding to the almost complete conversion of FA. However, the calculated turnover frequencies (TOFs) under different catalyst dosages are quite different, increasing from 1340 h⁻¹ to 2081 h⁻¹ when the catalyst dosage increases from 0.6% to 1.2%, implying that there is a significant diffusion effect. Under high catalyst dosage, more surface Pd atoms are available, so fewer substrate molecules need to enter the catalyst pore.

2.3. Characterization of the Catalysts

To elucidate the relationship between the catalyst structure and its catalytic performance, Pd/CN, Pd/CNK−0.25, Pd/CNK−0.5, and Pd/CNK−0.75 were thoroughly characterized. Inductively coupled plasma optical emission spectroscopy (ICP−OES) results reveal that the actual palladium loadings are around 2.85–2.91%, slightly lower than the nominal loading of 3% (Table S6). Additionally, no distinct diffraction peak corresponding to palladium species is observed in the X−ray diffraction (XRD) patterns, suggesting a high degree of Pd dispersion on all catalysts (Figure S7). To further investigate the distribution and particle size of Pd species on the catalysts, high−angle annular dark−field scanning transmission electron microscopy (HAADF−STEM) was employed to characterize all the samples (Figure 6). The STEM data show that the Pd species are uniformly distributed on all the catalysts. The average sizes of Pd nanoparticles on Pd/CN, Pd/CNK−0.25, and Pd/CNK−0.5 are 2.48 nm, 1.98 nm, and 1.34 nm, respectively. Especially for Pd/CNK−0.75, only a few Pd nanoparticles are observed, indicating an excellent dispersion of Pd species on CNK−0.75 (Figure 6d). This is further confirmed by the characterization of aberration−corrected HAADF−STEM (AC−HAADF−STEM), which shows that the majority of Pd species exist in single atoms (Figure 6h). The EDX elemental mapping data further confirm the uniform distribution of Pd species on the catalysts (Figure S8). These findings suggest that the introduction of cyano groups and nitrogen vacancies can enhance the dispersion of active Pd species and increase the number of accessible catalytic active sites. The activities of Pd/CN, Pd/CNK−0.25, and Pd/CNK−0.5 are consistent with the reverse order of their Pd particle size, indicating that the high dispersion of Pd species may be one factor influencing catalytic performance. However, Pd/CNK−0.75 does not follow this trend, suggesting that there may exist other factors that determine the catalytic activity for FA decomposition.

The recovered catalysts, Pd/CN−used and Pd/CNK−0.25−used after three cycles, as well as Pd/CNK−0.5−used after five cycles and Pd/CNK−0.75−used after eight cycles, were also characterized by STEM (Figure S9). Their average Pd sizes are 2.94, 2.11, 1.87, and 1.42 nm, respectively, which are larger than those of the fresh ones, indicating the aggregation of Pd species during FA decomposition. This may be the main reason why Pd/CN, Pd/CNK−0.25, and Pd/CNK−0.5 deactivate in the cycling experiment. The leaching ratio of Pd species in the first cycle is much lower, revealing that leaching of active components is not the key factor of catalyst deactivation (Table S6). The Pd leaching ratios of Pd/CNK−X (0.016‰–0.025‰) are lower than that of Pd/CN (0.039‰), also reflecting the facilitation of cyano groups and nitrogen vacancies for catalyst stability.

The electronic states of Pd species were characterized by X−ray photoelectron spectroscopy (XPS) for the fresh and recovered catalysts (Figure 7). The Pd 3d core−level XPS spectra display four main peaks: The peaks at around 335.7 eV and 341.0 eV correspond to metallic Pd⁰, while those at 337.6 eV and 343.1 eV attribute to Pd²⁺ species. The Pd⁰/Pd²⁺ ratios of Pd/CNK−0.5 (0.91) and Pd/CNK−0.75 (0.38) are lower than those of Pd/CN (1.04) (Table S6), owing to the introduction of more cyano groups and nitrogen vacancies, which leads to a strong interaction between Pd and support. However, the data of Pd/CNK−0.25 (1.25) are higher than those of Pd/CN, and the reason is still unclear. The Pd binding energies of Pd/CNK−X moving to higher values for 0.1−0.5 eV further demonstrates the stronger interaction between Pd and the CNK−X supports. The Pd⁰/Pd²⁺ ratios of the recovered catalysts after different reaction cycles increase significantly, revealing a reduction in Pd species by formic acid or hydrogen generated during the reaction process (Figure 7b, Table S7). The order of Pd⁰/Pd²⁺ ratios for the fresh catalysts or the used catalysts is not consistent with their activities, indicating that the Pd electronic states may not be the main factor influencing catalytic performance.

Numerous studies have demonstrated that the decomposition of β−PdH_x and the subsequent desorption of hydrogen from the Pd surface constitute one of the rate−determining steps in the dehydrogenation of FA [37,70]. In order to investigate the H₂ desorption capability of the catalysts, hydrogen temperature−programmed desorption (H₂−TPD) experiments were conducted (Figure 8). The H₂ desorption signals of the four pure supports are very weak, revealing that the supports show much lower H₂ molecule adsorption. All four catalysts exhibit a broad main desorption peak centered at about 100 °C, attributed to the decomposition of Pd hydride and the H₂ desorbed from metallic Pd nanoclusters or nanoparticles [71,72]. The Pd/CNK−X catalysts have higher peak intensities than that of Pd/CN, particularly for Pd/CNK−0.5, indicating that hydrogen spills over on these catalysts [73]. Notably, Pd/CNK−0.75 also exhibits another desorption peak at 232 °C except the above main desorption peak, which may be attributed to the strongly chemisorbed hydrogen on single Pd atoms or small clusters. It was found that the combination of active H and the desorption of molecular H₂ are sluggish on ultra−small Pd species [72]. This may be one main reason why Pd/CNK−0.75 shows low catalytic activity in FA dehydrogenation. The increase in its activity in the first three cycles may be attributed to the aggregation of partial Pd atoms or clusters, which enhances the H₂ desorption (Figure S9).

The hydrogen spillover of the catalysts was further investigated by using yellow WO₃ as an indicator (Figure 9) [74]. The pure WO₃ sample exhibit no observable color change after H₂ treatment. The Pd/CNK−0.5 + WO₃ mixture displays the most pronounced color change, indicating the highest hydrogen spillover capability of Pd/CNK−0.5, which is followed by Pd/CNK−0.25, Pd/CN, and then Pd/CNK−0.75. The hydrogen spillover capacities of the four catalysts are in the order Pd/CNK−0.75 < Pd/CN < Pd/CNK−0.25 < Pd/CNK−0.5, which is consistent with their activities, indicating that hydrogen spillover may be one main factor that influences the catalytic activity. Hydrogen spillover facilitates the transfer of activated H from Pd−H species to support and promote the regeneration of Pd active sites.

In order to reveal the hydrogen spillover of the supports, the color changes of the ternary systems of Pd/CNK−0.75, various supports, and WO₃ were also detected (Figure 10). The color of the system containing CN after H₂ treatment is even lighter than that of the binary system of Pd/CNK−0.75 + WO₃, indicating that the presence of CN inhibits the active H atom transfer from the surface of Pd/CNK−0.75 to WO₃ to a certain extent due to the low hydrogen spillover capability of CN. The colors of the systems containing CNK−0.25 and CNK−0.5 progressively deepen compared to the binary system of Pd/CNK−0.75 + WO₃, demonstrating the enhancement of CNK−0.25 and CNK−0.5 to the hydrogen spillover between the catalyst and WO₃, particular for CNK−0.5, due to the introduction of more cyano groups and nitrogen vacancies [75]. The color of the ternary system with CNK−0.75 is only slightly darker than that of Pd/CNK−0.75 + WO₃, implying a relatively weaker hydrogen spillover of CNK−0.75 compared to CNK−0.25 and CNK−0.5, possibly owning to the excessive distortion of the CN structure. This result indicates that the hydrogen spillover of the supports is in the order CN < CNK−0.75 < CNK−0.25 < CNK−0.5, which is not consistent with the corresponding catalysts. Among the four catalysts, Pd/CNK−0.75 shows the weakest hydrogen spillover ability, which may be attributed to the much lower hydrogen activation capability of the single Pd atom [76]. The hydrogen spillover capacities of Pd/CNK−0.75 after one reaction cycle and two reaction cycles progressively improves owning to the aggregation of Pd atoms and clusters during FA dehydrogenation (Figure S10), which may be one main reason why Pd/CNK−0.75 shows increasing activity in the first three cycles. This further demonstrates the positive effect of hydrogen spillover on the activity of Pd/CNK−X.

3. Materials and Methods

3.1. Chemicals

Dicyandiamide (C₂H₄N₄, AR) was purchased from Shanghai Haohong Biomedical Technology Co., Ltd. (Shanghai, China). Palladium chloride (PdCl₂, Pd 59.5%) was obtained from Shanghai Jiuyue Chemical Co., Ltd. (Shanghai, China). Formic acid (HCOOH, FA) was provided by Tianjin Fuchen Chemistry Co., Ltd. (Tianjin, China). Sodium hydroxide (NaOH) was purchased from the Aladdin Company (Hong Kong, China). Potassium hydroxide (KOH) was obtained from Beijing Chemical Industry Group Co., Ltd. (Beijing, China). Hydrochloric acid (HCl) was obtained from Beijing Yili Fine Chemicals Co., Ltd. (Beijing, China). Sodium borohydride (NaBH₄) was purchased from Xilong Scientific Co., Ltd. (Guangzhou, China).

3.2. Synthesis of Graphite Carbon Nitride (CN) and Potassium Hydroxide−Treated CN (CNK−X)

Synthesis of CN: Typically, 10 g of dicyandiamide was placed in a lidded alumina crucible and calcined at 550 °C for 4 h in a muffle furnace with a heating rate of 5 °C/min. After natural cooling to room temperature, the sample obtained from the crucible was thoroughly ground in an agate mortar, which was denoted as CN.

Synthesis of CNK−X: 2 g of pre−prepared CN was thoroughly ground with different amounts of KOH (0.25 g, 0.5 g, and 0.75 g) and placed in an alumina porcelain boat, then it was calcinated under N₂ flow at 600 °C for 2 h with a heating rate of 10 °C/min in a quartz tube furnace. After cooling down to room temperature, the sample was washed with HCl solution (10 wt%) and water to dislodge K⁺ and adjust the samples to a neutral pH. Lastly, the samples were dried at 80 °C overnight and collected. The samples prepared with 0.25 g, 0.5 g, and 0.75 g of KOH were denoted as CNK−0.25, CNK−0.5, and CNK−0.75, respectively.

3.3. Synthesis of Pd/CN and Pd/CNK−X

The Pd/CN catalyst was prepared using a deposition−precipitation method. Specifically, 2 g of the as−prepared CN support was dispersed in 50 mL of deionized water and stirred vigorously at room temperature for 1 h. Then, PdCl₂ (100 mg) solution in dilute HCl was added. The pH value of the mixture was gradually adjusted to 10 by dropwise adding NaOH solution (10 wt%) and stirred for an additional 4 h. Subsequently, 1 mol/L NaBH₄ solution was added to the system with a 5:1 molar ratio relative to palladium. The mixture was then filtered, washed with abundant deionized water, and dried in an oven overnight at 60 °C. The Pd/CNK−X samples were prepared under a similar process.

3.4. Characterization of Catalysts and Supports

Crystal structures of the samples were obtained by means of the powder X−ray diffraction (XRD) technique on a Rigaku SmartLab SE instrument (Rigaku Corporation, Tokyo, Japan) using Cu Kα radiation (40 kV, 40 mA, λ = 0.15406 nm) with a scan range of 10 to 50°, and the scanning speed was 5°/min. The specific surface area and pore size distribution plots were determined using the Brunauer−Emmett−Teller (BET) and Barrett−Joyner−Halenda (BJH) methods of analysis, respectively, by nitrogen adsorption on a Micromeritics ASAP 2460 (Micromeritics Instrument Corporation, Norcross, GA, USA) analyzer at −196 °C. All samples were degassed at 150 °C for 6 h under vacuum before characterization. Elemental analysis (EA) of C, N, O, and H was performed on an elemental analyzer Elementar Vario MACRO (Elementar Analysensysteme GmbH, Langenselbold, Germany). Scanning transmission electron microscopic (STEM) images were taken with a JEM−2010 instrument (JEOL Ltd., Yamagata, Japan), and the sizes of the Pd nanoparticles were calculated by a statistical method from the sizes of about 150 particles in the selected area. The aberration−corrected HAADF−STEM (AC−HAADF−STEM) images were taken with a JEM−ARM300F instrument (JEOL Ltd., Yamagata, Japan). The FE-STEM SU9000 (Hitachi High-Tech Corporation, Tokyo, Japan). scanning electron microscope (SEM) was used to observe the morphology and microstructure of the samples. Electron paramagnetic resonance (EPR) was obtained on a Bruker EMXnano ESR spectrometer (Bruker Corporation, Rheinstetten, Germany). ¹³C cross−polarization−magic angle spinning nuclear magnetic resonance spectroscopy (¹³C CP−MAS NMR) was used to analyze the structure of samples on a Bruker Avance NEO 400 WB spectrometer (Bruker Corporation, Rheinstetten, Germany). Electron−binding energies of the as−obtained samples and the recovered samples were characterized with X−ray photoelectron spectroscopy (XPS, Thermo Fisher Scientific ESCALAB 250 Xi spectrometer, Waltham, MA, USA), with Mg Kα (hv = 1253.6 eV) as the excitation source.

The temperature−programmed desorption of H₂ (H₂−TPD) was carried out on a chemical adsorption analyzer (AutoChem II 2920, Micromeritics Instrument Corporation, Georgia, USA). Prior to the TPD test, a 50 mg sample (40–60 mesh) was loaded in a fixed−bed U−shaped quartz microreactor. The samples were initially pretreated with helium flow (30 mL/min) by heating to 200 °C at a rate of 5 °C/min, and this temperature was maintained for 30 min. Subsequently, the samples were exposed to a 10 vol% H₂/N₂ gas mixture at room temperature for 1.5 h to facilitate hydrogen adsorption, whereafter the sample was flushed with helium flow (30 mL/min) until baseline was smooth. The system was heated up to 400 °C with a ramp of 5 °C/min in the helium atmosphere. The products were detected by mass spectrometer. The SO₂−TPD (Temperature−Programmed Desorption) procedure was identical to that described above, with the exception that a mixed gas of 10 vol% SO₂/He was used in the adsorption step.

Hydrogen spillover test of the catalysts: A mixture of WO₃ (500 mg) and catalysts (10 mg) was ground and then placed in a tube furnace and maintained at 30 °C for 10 min under an H₂ flow (40 mL/min). After the H₂ treatment, a photo was taken of the collected mixture for comparison.

Hydrogen spillover test of the support: The Pd/CNK−0.75 catalyst (5 mg) was thoroughly ground and mixed with different support materials (10 mg) (CN, CNK−0.25, CNK−0.5, and CNK−0.75) firstly, then WO₃ (500 mg) was added and ground further. The subsequent H₂ treatment was consistent with the above procedure.

3.5. Catalytic Activity Test

The catalytic performance of the as−synthesized samples was evaluated by the decomposition of FA. Typically, 210 mg of catalyst was added to a 15 mL two−neck round−bottom flask placed in a water bath maintained at 60 °C. One neck of the flask was connected to a gas burette for gas collection, while the other was used to introduce 3.0 mL of FA solution (2.0 mol/L). The reaction commenced upon the addition of FA, and the evolved gas volume was measured by monitoring the water displacement in the gas burette.

The conversion and the selectivity of products were analyzed on an Agilent 5977A gas chromatography analyzer with a thermal conductivity detector (TCD) and a hydrogen flame ionization detector (FID), which has a 1 ppm detection limit for CO. The pure CO₂ (99.9 vol %) and H₂ (99.99 vol %) as well as a mixture gas of CO/N₂ (1.02 vol %) were used as the reference gases.

The conversion is determined by the ratio of the total gas volume (V_t) produced at desired time t to the theoretical gas volume (V₀) under complete conversion, where the gases are treated as ideal gas. One mole FA gives two moles gas (H₂ + CO₂). Thus:

Conversion = V_t/V₀ = V_t·P/2nRT

where P = 101,325 Pa, T = 298 K, and n is the number of moles of FA.

The turnover frequencies (TOFs) are calculated according to the released gas volume, which is considered the ideal gas:

$$\mathrm{T}\mathrm{O}\mathrm{F}={\displaystyle \frac{{\mathrm{P}}_0\mathrm{V}}{2\mathrm{R}\mathrm{T}{\mathrm{n}}_{\mathrm{P}\mathrm{d}}\mathrm{t}}}$$

where P₀ is the atmospheric pressure (101,325 Pa), V is the released gas volume (H₂ + CO₂), R is the universal gas constant (8.3145 m³·Pa·mol⁻¹·K⁻¹), T is the room temperature (298 K), n_Pd is the total mole number of Pd atoms in the catalyst, and t is the time required to achieve 25% conversion.

3.6. Durability Tests of the Catalysts

The catalysts were recovered by filtration after the first reaction, thoroughly washed with deionized water, and dried in air, and were reused in the FA dehydrogenation under the same condition as the first cycle.

4. Conclusions

In summary, three CNK−X supports containing a certain number of cyano groups and nitrogen vacancies are prepared successfully by CN calcination in the presence of KOH, and the performances of the corresponding Pd−based catalysts, Pd/CNK−0.25, Pd/CNK−0.5, and Pd/CNK−0.75, are evaluated for FA dehydrogenation. It is found that the introduction of cyano groups and nitrogen vacancies in CN skeleton enhances the support alkalinity and improves the Pd dispersion significantly. The high catalytic activity of Pd/CNK−0.5 is mainly attributed to its highly efficient hydrogen spillover and relatively high dispersity of Pd species, as well as its high basicity due to the introduction of proper number of cyano groups and nitrogen vacancies. The relatively high stability of Pd/CNK−0.75 is related to the strongest interaction between Pd species and the support, while its low initial activity may be owing to its low hydrogen spillover ability and the strongly chemisorbed hydrogen on single Pd atoms or small clusters.