Pd Catalysts Supported on Bamboo ‐ Like Nitrogen ‐ Doped Carbon Nanotubes for Hydrogen Production

: Bamboo ‐ like nitrogen ‐ doped carbon nanotubes (N ‐ CNTs) were used to synthesize supported palladium catalysts (0.2–2 wt.%) for hydrogen production via gas phase formic acid decomposition. The beneficial role of nitrogen centers of N ‐ CNTs in the formation of active isolated palladium ions and dispersed palladium nanoparticles was demonstrated. It was shown that although the surface layers of N ‐С NTs are enriched with graphitic nitrogen, palladium first interacts with accessible pyridinic centers of N ‐С NTs to form stable isolated palladium ions. The activity of Pd/N ‐ CNTs catalysts is determined by the ionic capacity of N ‐ CNTs and dispersion of metallic nanoparticles stabilized on the nitrogen centers. The maximum activity was observed for the 0.2% Pd/N ‐ CNTs catalyst consisting of isolated palladium ions. A ten ‐ fold increase in the concentration of supported palladium increased the contribution of metallic nanoparticles with a mean size of 1.3 nm and decreased the reaction rate by only a factor of 1.4.


Introduction
Rational use of fossil fuel resources is a necessary condition for the long-term stable development of society. In this context, the development of efficient processes for energy production from renewable sources is a topical problem that can be solved using various fundamental studies, particularly in the field of material science. Hydrogen production from formic acid (FA) is a way to involve renewable biomass sources into the energy cycle [1,2]. The FA decomposition reaction proceeds on mono-, bi-and trimetallic catalysts deposited on oxide or carbon supports [3,4]. Activity and selectivity of the catalysts towards hydrogen production are determined by the nature of their active component, its dispersion, and electronic state of metals. The application of nitrogen-doped carbon nanomaterials (N-CNMs) as the catalyst supports makes it possible to perform the tailored synthesis of metal supported catalysts with desired properties [5].
It is known that nitrogen in N-CNMs having different structures occupies standard positions: pyridinic (NPy), pyrrolic (NPyr), graphitic (NQ), and oxidized (NOx). According to data in the literature, supported metals can stabilize on different nitrogen centers. For example, Pt stabilization on NQ centers of nitrogen-doped carbon nanotubes (N-CNTs) leads to the formation of electron-rich Pt nanoparticles, which are active towards the hydrogenation of nitroarenes [6] and aerobic oxidation of glycerol [7]. Ombaka et al. [8] reported the formation of Pd 2+ species on the NPyr centers of N-CNTs, which are active towards the hydrogenation of nitrobenzophenone [8]. However, a substantial body of data indicates the involvement of NPy centers in the stabilization of electron-deficient metals on the surface of various N-CNMs, which show the increased activity and selectivity in hydrogenation reactions or electrochemical reduction of oxygen [9][10][11][12]. The positive role of the electron-deficient metals deposited on ordered or disordered N-CNMs was also demonstrated for FA decomposition in a liquid or gaseous medium [4,[13][14][15][16]. Additionally, for platinum deposited on nitrogen-doped carbon nanofibers (N-CNFs) and N-CNTs, it was shown that the catalytic properties of metal in this reaction are determined by its locally one-type interaction with NPy centers [16]. Among various metals that are active towards FA decomposition, platinum and palladium are the most active catalysts for this reaction [3,17]. It should be noted that the use of N-CNMs as the supports opens new possibilities for enhancing the efficiency of catalysts via the stabilization of highly active and selective atomic metal species [18,19]. Among the wide variety of N-CNMs, bamboo-like N-CNTs are characterized by a unique combination of the graphene layer curvature and the availability of all nitrogen centers. At the same time, data on the use of these materials as a support for the catalysts of the FA decomposition reaction are quite scarce [18]. This paper is devoted to investigating the formation of the Pd catalyst on bamboo-like N-CNTs for the gas phase formic acid decomposition reaction. A detailed investigation of the properties of N-CNTs, variation of palladium concentration, and application of various research methods allowed us to describe the interaction of palladium with the surface of N-CNTs. Different active palladium species were shown to form as isolated ions and metallic nanoparticles with a size of 1.3 nm, the ratio of which were determined by the affinity of palladium for NPy centers and the palladium concentration in catalysts.

Materials and Methods
Nitrogen-doped and nitrogen-free carbon nanotubes (N-CNTs and CNTs) were synthesized by the standard CVD route [20]. An ethylene-ammonia mixture or ethylene were decomposed on the Fe-containing catalyst at 650-700 °C. The growth catalyst was removed by the boiling of N-CNTs and CNTs in 2 M hydrochloric acid. Nevertheless, X-ray photoelectron spectroscopy (XPS) revealed the presence of iron in the washed samples with the content up to approximately 0.5 at.%. Previously, we have demonstrated the low intrinsic activity of washed N-CNTs and CNTs in FA decomposition reactions compared to metal supported catalysts [16]. Palladium was deposited on carbons, which were pre-dried in Ar at 170 °С, by incipient wetness impregnation from a Pd acetate-acetone solution. The samples were dried in air at 105 °С for 8 h and then reduced in an H2/Ar flow at 200 °С for 1 h.
Nitrogen adsorption measurements were carried out at 77 K by the use of ASAP-2400. Prior to measurements, the samples were evacuated at 200 °С for 24 h. Acetone capacity of the N-CNTs and CNTs was measured by the standard incipient wetness impregnation method. Acetone was added dropwise under stirring to the sample dried in Ar at 170 °C until the pores of the material were completely filled. The volume of the absorbed acetone (cm 3 /g) was determined by weighing. CO chemisorption on the catalysts was measured at room temperature by a pulse technique. Before the experiments, the catalysts were reduced in hydrogen flow at 200 °С. X-ray diffraction patterns were recorded using an HZG-4 (Zeiss, Germany) diffractometer with monochromatic CuKα radiation. Electron microscopy investigation was performed using JEM-2200 FS (JEOL Ltd., Japan) and Titan 60-300 (FEI, Netherlands) electron microscopes. Raman spectroscopy with an excitation wavelength of 632.8 nm was performed using a HORIBA LabRAM HR800 (HORIBA Jobin Yvon Raman Division, France) Raman spectrometer. XPS study of N-CNTs was performed on an ES-300 (KRATOS Analytical, UK) photoelectron spectrometer with an AlKα source (hν = 1486.6 eV). The distribution of nitrogen in N-CNTs was studied by synchrotron-based photoelectron spectroscopy at 500 and 800 eV photon energy [21]. Pd catalysts were examined using a VG ESCALAB HP (ThermoScientific, UK) instrument with an AlKα source (hν = 1486.6 eV). In situ reduction of the catalysts under 1000 Pa H2 in a preparation chamber of the XPS unit was performed at 300 °C for 1 h [22]. After the pretreatment, the sample was transferred to an analytical chamber without contact with air.
Activity of the catalysts in the decomposition of FA (5 vol. % НСООН/Не) was measured using a flow setup. A 20 mg catalyst sample was uniformly mixed with 0.5 cm 3 of quartz sand (grain size of 0.25-0.5 mm). The reaction mixture preheated to 100 °С in a special box was fed into the reactor at a rate of 20 cm 3 /min. The experiments were carried out in the temperatureprogrammed mode at a ramp rate of 2°/min with chromatographic analysis of the gas mixture.
Before measuring the activity, the catalyst was reduced in a 10% Н2/He mixture at 200 °С for 1 h. The apparent turnover frequency (TOF) was calculated as a ratio of the reaction rate measured at low (<20%) conversions to the total number of palladium atoms in the catalysts.

Properties of N-CNTs
Nitrogen-doped carbon nanotubes were synthesized by the CVD method, which is widely applied to obtain N-СNTs [23][24][25][26]. As seen in Table 1, by increasing the ammonia concentration in the С2Н4-NH3 reaction mixture (25,40, 60 and 75% NH3), one can monotonically increase the degree of nitrogen doping of nanotubes from 1.8 to 6.6 at.% with a simultaneous decrease in the yield of N-CNTs from 38 to 5 g C/g of catalyst. Such an effect of N/C ratio on the nitrogen content in N-CNTs was described by other authors for different reaction mixtures [24,26,27]. In turn, a decrease in the yield of N-СNTs is related to deactivation of the catalyst due to its encapsulation (Figure 1), which has been reported by many authors for the CVD synthesis of N-СNTs and other carbon materials [25,26,[28][29][30].  TEM micrographs of the 4.4 at.% N-CNTs as an example and СNTs are displayed in Figure 1. Ethylene decomposition is accompanied by the predominant formation of multiwalled carbon nanotubes. In their turn, N-СNTs have a distinct bamboo-like structure, which is known to form due to the curvature of nitrogen-containing graphene layers [31,32]. According to STEM-HAADF images with elemental mapping, nitrogen is uniformly distributed in N-CNTs.
XPS revealed (Figure 2) that nitrogen in N-СNTs is in pyridinic (NPy, ~ 398 eV), pyrrolic (NPyr, ~ 400 eV), graphitic (NQ, ~ 401 eV), oxidized (NOx, ~ 403 eV) and molecular (NN2, ~ 405 eV) states [24,26,33]. The presence of molecular nitrogen encapsulated between graphene layers or inside channels is characteristic of bamboo-like N-CNTs [26,34]. An increase in the total nitrogen content in N-CNTs is accompanied by an increase in the content of all structural species. However, this changes the predominant species and the ratio of nitrogen species. In the N-CNTs with the minimum doping degree, the main contribution is made by graphitic nitrogen, and the NPy/NQ ratio is equal to 0.5, Table 1. As the total nitrogen content is increased, the contribution of pyridinic nitrogen grows; as a result, the NPy/NQ value achieves 1.2 in the case of 6.6 at.% N-CNTs.
It is known that nitrogen incorporation in the pyridinic positions of graphene layers is accompanied by the formation of carbon vacancies, i.e., by an increase in structural defectivity. Indeed, the formation of such fragments was observed in TEM images of N-СNTs obtained at atomic resolution [23]. The formation of such structural elements was confirmed by a growth of the ID/IG ratio with a simultaneous decrease in the I2D/ID ratio revealed by Raman spectroscopy ( Table 1). The ID/IG ratio reflects defectivity of carbon nanomaterials [35,36], while the I2D/ID ratio is used to analyze the size of the defect-free graphene blocks [37,38] (Figure 3). In our earlier work, the XRD study of N-CNTs and their structural simulation gave grounds to suppose the formation of the fragments similar to the structural elements of the g-C3N4 layer comprising the ordered carbon vacancies and pyridinic nitrogen atoms [20]; the estimated size of such a fragment of the layer is not smaller than 5 Å.  However, it should be noted that the obtained XPS data do not provide information on the surface state of N-CNTs with respect to different nitrogen species, which is an important factor when these materials are employed as the catalyst supports. In this context, N-СNTs were additionally studied by synchrotron-based XPS with variation of the analysis depth using radiation energies of 500 and 800 eV. The data obtained were compared with the results of standard XPS study performed at 1486.6 eV. The radiation energies of 500, 800 and 1486.6 eV allowed us to acquire data at the analysis depths of ca. 1.6, ca. 3 and ca. 6 nm, respectively ( Figure 4). According to TEM data displayed in Figure  1b, the mean wall thickness of N-CNTs is ca. 5 nm. Hence, when the depth of analysis is varied from 1.6 to 6 nm, the acquired data could provide information on the nitrogen in external/internal graphene layers, and partially in internal arches.   [24]. As seen in Figure 4c, when the entire tube wall (ca. 6 nm) or only its external/subsurface layers (ca. 1.5 nm) are analyzed, NPy/NQ, ratios for 4.4 at.% N-СNTs virtually do not differ. In the case of 1.8 at.% N-СNTs, as the depth of analysis is decreased, the NPy/NQ ratio increases by a factor of ca. 1.6. Nevertheless, our study of N-CNTs with the minimum depth of analysis allows a conclusion that external layers of N-CNTs, irrespective of the doping degree, are slightly enriched with graphitic nitrogen.
To choose a method of palladium deposition on N-СNTs, their textural and adsorption properties were investigated. As seen in Table 1, nitrogen doping of CNTs does not produce changes in the textural characteristics of carbon nanotubes. Indeed, their specific surface area changes in a narrow range of 148-160 m 2 /g, and pore size distribution curves for all the samples are similar and have distinct maxima in the regions of 3-4 nm and 25-40 nm ( Figure 5). Fenelonov et al. [39] proposed that pores with the size of 3-4 nm are formed due to the curvature of graphene layers, whereas larger pores emerge due to the entangled weave of N-CNTs leading to the formation of multiple voids.
However, an increase in the nitrogen content in N-CNTs is accompanied by a considerable growth of their capacity towards acetone, from 3 to~ 6-7 cm 3 /g, which was used as a solvent in the deposition of palladium catalyst (Figure 5b). This suggests that such a high acetone capacity of N-CNTs is achieved due to the bulk filling of inner channels of the tubes in spite of multiple regular internal arches (Figure 1b). Note that the values of acetone capacity of fresh and treated in 2 M HCl carbons are the same. In turn, а good correlation between nitrogen content in N-CNTs and NPy/NQ, ID/IG and I2D/D ratios was found (Table 1). Therefore, it can be supposed that the presence of large defects (not smaller than 5 Å) in graphene layers of the synthesized N-CNTs, which are represented by the ordered carbon vacancies surrounded with pyridinic nitrogen atoms [20], improves their wettability and makes the capillary effect more pronounced. Kumar et al. [40] calculated isotherms of water adsorption on disordered N-C materials and attributed the anomalously high adsorption to the implementation of a specific mechanism, according to which 1D, 2D and 3D nucleation, accompanied by the formation of solvent clusters which resembles the growth of crystals, was followed by bulk condensation. The filling of the inner channels of N-СNTs may indicate that during their deposition on N-CNTs by wet chemistry methods, metals can be anchored not only on the external surface but also in the inner channels of nanotubes [41]. On the other hand, Hao and al. [42] conclude that high hydrophilicity of N-doped carbon nanomaterials ensures a high dispersion of supported metallic catalysts, thus increasing their activity.

Properties of Pd/N-CNTs Catalysts
A series of Pd/N-CNTs catalysts (0.2, 0.5, 1 and 2 wt.% of Pd) were synthesized using the N-СNTs that were obtained by decomposition of the 40% С2H4-60% NH3 mixture. The choice of support was based on the optimal combination of the nitrogen content in carbon nanotubes and the yield of N-CNTs (Table 1). To study the interaction of palladium with the N-CNTs surface, the concentration of palladium was varied from 0.2 to 2 wt.%. N-free catalysts 0.2% Pd/CNTs and 2% Pd/CNTs were synthesized for comparison purposes. The revealed high affinity of N-СNTs for acetone, which was used in our study as a solvent of palladium salt, explains the uniform distribution of palladium in the catalysts and the absence of agglomerates ( Figure 6). In addition, as seen in Figure 6b, it cannot be ruled out that individual palladium particles are located in the inner channels of nanotubes. In the X-ray diffraction patterns, a weak intensity peak related to Pd 111 was recorded only for the catalysts with a maximum metal content (2 wt.%; see Figure 7). The X-ray diffraction patterns of all other catalysts show only the maxima related to the carbon phase.  Table 2 and Figure 8. In the case of N-free catalysts, a ten-fold increase in palladium content was accompanied by a decrease in apparent TOF from 180 to 72 h −1 , with a simultaneous decrease in СО/Pd ratio from 33.5 to 22.1%, according to the CO chemisorption data. Indeed, the mean particle size for 0.2% Pd/CNTs was 1.2 nm, whereas for 2% Pd/CNTs it was 2.3 nm. According to the XPS, palladium was in the metallic state and the binding energy of Pd 3d was 335.6 eV (Figure 9). This value is typical of the dispersed metallic palladium particles [43,44]. The TOF values of the 0.2% Pd/CNTs and 2% Pd/CNTs catalysts calculated as a ratio of the reaction rate to the number of surface palladium atoms were 193 and 147 h −1 , respectively. The observed difference in TOF values may be due to different palladium crystallinity or its different localization in the tubes (outer surface and inner channels). Selectivity of the catalysts towards hydrogen formation did not exceed 92%.  All the nitrogen-doped catalysts, irrespective of palladium content, demonstrated the increased activity and selectivity (97-98%) in comparison with 0.2-2% Pd/CNTs catalysts. TOF values varied in a narrow range of 225-324 h −1 ; the maximum activity was observed for the 0.2% Pd/N-СNT catalyst. The dependence of СО/Pd on palladium content for this series of catalysts had a complicated shape. A decrease in palladium content from 2 to 1% was not accompanied by changes in СО/Pd, and a further decrease in Pd concentration to 0.2% produced a sharp drop in СO/Pd from 32 to 16.1%. The observed dependence of СО/Pd on palladium content did not correlate with the ТЕМ data, according to which mean particle size for the catalysts with 2% Pd and 1% Pd was equal to 1.4 and 1.3 nm, respectively (Table 2), whereas for 0.5% Pd and 0.2% Pd it could not be determined owing to their small sizes. An HAADF-STEM study of 0.2% Pd/N-CNTs showed that the catalyst consisted predominantly of isolated palladium atoms ( Figure 6). The 2% Pd/N-CNT catalyst contained not only nanoparticles but also a considerable amount of single palladium atoms. As seen in Figure 9, the binding energy of palladium in 0.2% Pd/N-CNTs was 337.7 eV, which may correspond to both the PdO [45] and the ionic palladium stabilized on pyridinic centers of N-CNTs [46]. However, oxidation of the samples was completely excluded in XPS studies; therefore, Pd 3d BE 337.7 eV can reliably be assigned to Pd 2+ -NPy species. Some studies have demonstrated that Pd 2+ -NPy species, in distinction to Pd 0 , weakly chemisorb СО [12,46]. In the case of 2% Pd/N-CNTs, palladium is in the highly dispersed Pd 0 (BE Pd 3d 335.5 eV) and Pd 2+ -NPy (BE Pd 3d 337.7 eV) states.

Discussion
Pd/N-CNTs catalysts have been synthesized and examined in the gas-phase decomposition of FA for hydrogen production. To describe the interaction of palladium with N-CNTs, palladium concentration was varied from 0.2 to 2 wt.%, and a comparison with N-free catalysts was made. XPS experiments with variation of the analysis depth demonstrated that nitrogen on the external graphene surface of N-CNTs was in all structural states. Irrespective of the doping degree, the predominance of graphitic nitrogen was observed. However, according to XPS and HAADF-STEM studies of the 0.2% Pd/N-CNT catalyst, palladium starts interacting with pyridinic nitrogen centers of N-СNTs to produce isolated ions. Indeed, theoretical calculations show that palladium preferentially interacts with the NPy centers of N-CNTs instead of the NQ centers to form a positive +0.5 е charge [47]. The binding energy of Pd 3d in 0.2% Pd/N-CNTs is 337.7 eV, which coincides well with the BE Pd 3d values for Pd/N-СNTs catalysts reported in literature [47,48]. In these studies, DFT calculations were used to propose the models of palladium anchoring on both the single NPy center and 3NPy-V (where V is the vacancy). It seems interesting that close BE Pd 3d values of 336.9-338.5 eV were also obtained in the case of palladium deposition on g-C3N4 or CN material derived from a metal-organic framework [12,19]. It was supposed that stabilization of isolated palladium ions occurs with the participation of four nitrogen centers (2NPy and 2NPyr) [19] or 6NPy, which forms structural elements of the g-C3N4 graphene layer [12]. It should be noted that in the 0.2% Pd/N-CNTs catalyst, the NPy/Pd ratio is 25, which makes it possible to implement any of the proposed models, including Pd anchoring on fragments of the layer, which are similar to g-С3N4. The possibility of the formation of such fragments in graphene layers of N-CNTs was demonstrated in our earlier study [20]. As the content of supported palladium is increased, this most likely leads to the filling of accessible NPy centers; as a result, metallic nanoparticles start to form ( Figure 10). This assumption is supported by the obtained dependence of СО/Pd on palladium content in the catalysts. It reflects the occurrence of competitive processes: an increase in CO chemisorption owing to the growing dispersion of metallic nanoparticles, and a decrease in CO chemisorption due to formation of ionic palladium. It should be noted that the involvement of NQ centers in the stabilization of metallic nanoparticles, which are more dispersed than in N-free catalysts, cannot be ruled out; their involvement was derived, for example, by Shi et al. [6] from a 0.2 eV shift of the corresponding maximum in the N1s spectrum towards higher binding energies. We did not observe any changes in the N1s spectrum of Pd/N-CNTs catalysts in comparison with N-CNTs. However, a 0.1 eV shift of BE Pd 3d for a more dispersed 2% Pd/N-CNTs (1.4 nm) towards lower binding energies in comparison with 2% Pd/CNTs (2.3 nm) provides the basis to suppose such a possibility. As a result, the involvement of nitrogen centers in the stabilization of palladium decreases the TOF value only by a factor of 1.4 at a ten-fold increase in palladium concentration. For N-free catalysts, the TOF value decreases by a factor of 2.5. As shown by the catalytic experiments, isolated palladium ions are more active than metallic palladium nanoparticles, and the maximum TOF value is observed on the 0.2% Pd/N-CNT catalyst. A strong interaction of palladium with the pyridinic centers of N-СNTs is responsible for high stability of the catalyst in the course of the reaction as compared to the N-free catalyst (Figure 8c). Presumably, activity of the catalysts with high palladium content is determined by the ratio of two palladium species and by the size of metallic nanoparticles. In comparison with N-free catalysts, selectivity on all the N-doped catalysts increased by 5%, which corresponds to nearly a four-fold decrease in CO concentration. Additionally, the discovered stability of the catalyst 2% Pd/N-CNT compared to 2% Pd/CNT confirms the stabilization of both palladium species (isolated ions and metal nanoparticles) by nitrogen centers [49].

Conclusions
Nitrogen centers of N-CNTs were shown to be efficient in the stabilization of different palladium species in the form of isolated ions and highly dispersed metal nanoparticles that are active towards formic acid decomposition for hydrogen production. This allows a conclusion that various nitrogen-containing carbon nanomaterials are promising for use as the supports of metallic catalysts for this reaction. Taking into account an increased activity of isolated palladium ions, it is more preferable to use N-СNMs with predominant pyridinic nitrogen centers in the external graphene layers.