## Introduction

Hydrogen as a future energy carrier is the most efficient, nontoxic and environmentally friendly fuel, because water is the only by-product in the conversion into energy through combustion or fuel cells, thus eliminating emission of greenhouse gases1,2,3,4. Yet, the realization of a hydrogen economy highly relies on how to overcome various storage and transportation issues caused by H2’s low volumetric density3,5. Of all different technologies, chemical hydrogen storage by liquid organic hydrogen carrier (LOHC) has been recently recognized as the most promising solution6,7,8,9. The LOHC concept is based on a pair of high boiling-point organic compounds that can be catalytically converted from an energy-lean to energy-rich form by hydrogenation and vice versa by dehydrogenation6,10,11,12. Among LOHC systems available in the literature1,4,9,12,13,14,15,16, homocyclic compounds offer an advantage due to low cost and high compatibility with the existing infrastructures, but suffer significantly from a high heat of dehydrogenation17,18. Thus, theoretical and experimental studies validated that the incorporation of heteroatoms in the chemical structure of homocyclic LOHC compounds reduced the enthalpy of dehydrogenation and therefore favored H2 release at low temperatures19,20,21. The representative example is a pair of N-ethylcarbazole (H0-NEC) and its corresponding perhydro product (H12-NEC), which has attracted a lot of attention from an academia and industry3,5,9,10,11,13,20. However, the downside of this pair for practical approach could be the thermal stability and physical state of H0-NEC (i.e. solid at room temperature). Recent works hence demonstrated a potential of quinaldine22, indole derivatives23,24, and pyridine derivative6,25 as alternative LOHC material.

Even though the dehydrogenation of these N-heterocyclic LOHC compounds is thermodynamically favorable compared to homocyclic compounds, an active and stable catalyst is still challenging because a relatively high reaction temperature is required for fast H2 release. In most of the previous works Pd catalysts are known to work efficiently though Pt catalysts are also reported. Particularly, Smith’s3,5 and Wasserscheid’s26,27 groups investigated various supported Pd catalysts in the dehydrogenation of H12-NEC, where the common support was alumina. Also, a 5 wt% Pd/Al2O3 catalyst was applied for the dehydrogenation of perhydro alkylindoles23,24. In case of perhydro 2-(n-methylbenzyl)pyridine (H12-MBP) we documented last year, supported Pd catalysts proved to be more active than Pt analogues6. Aside from the dehydrogenation efficiency, catalyst stability is an important issue due to process operation at high temperatures above 200 °C, in which the metal-support interaction generally determines the lifetime of a tested dehydrogenation catalyst. For instance, our recent report revealed that a carbon-coated alumina supported Pd catalyst showed the enhanced stability than Pd/C and Pd/Al2O3, because the heterogeneity of support surface prevented Pd nanoparticles from sintering into larger sizes during the reaction25. Nevertheless, this strategic approach via carbon coating using an aqueous glycerol solution is rather complex and the optimal carbon content depends on the textural and acidic properties of the fresh alumina used for carbon coating.

Among numerous synthesis methods for supported Pd catalyst system, we paid much attention to a one-pot solvent deficient precipitation (SDP) method. Woodfield and colleagues28,29 proposed the original synthesis method and focused on a variety of single and mixed metal oxides, particularly, alumina29. The synthesis typically involves only two steps: the metal salt is first mixed with a simple base in a solid state to yield an intermediate hydroxide product (e.g. metal hydroxide or metal oxide hydroxide) that further undergoes dehydration for the formation of intermediate nanoparticles, followed by the high-temperature calcination for crystallization and meso-structuring. This method is not only simple and cost effective, but also produces metal oxide particles with the high surface area and mesoporosity even though surfactant or structure directing agent is absent. This uniqueness would promise to yield an active and stable supported Pd catalyst, as demonstrated in iron Fischer–Tropsch catalysts30,31,32 and nickel methane reforming catalysts33,34.

Herein, we report the dehydrogenation activity and stability of Pd-Al2O3 catalyst prepared by the SDP method. The prepared catalyst system show mesoporosity, thus labelled as MPdA where M stands for mesoporosity, and retain Pd nanoparticles that are mainly composed of Pd(111) plane. In the dehydrogenation of perhydro 2-(n-methylbenzyl)pyridine MPdA offers improved H2 release capability compared to Pd counterparts prepared by conventional impregnation, which is further demonstrated in four different N-heterocyclic compounds. Moreover, the dehydrogenation activity of MPdA is maintained up to five consecutive recycle runs because of lessened sintering of Pd particles. These features in catalyst property and performance stem from the generation of solvent deficient environment in the SDP synthesis, which is examined by characterization of uncalcined precursor materials.

## Results

### Synthesis and properties of mesoporous PdO–Al2O3 material

The original synthesis recipe was amended for a MPdA catalyst of 1 wt% Pd loading, which would expect to form a homogeneous Pd–Al solid solution (Supplementary Movie 1). First, palladium nitrate (Pd(NO3)2·H2O), aluminum nitrate (Al(NO3)3·9H2O) and ammonium bicarbonate (NH4HCO3) of desired amounts were mixed together in a mortar and grounded thoroughly by a pestle at room temperature, where no additional solvents including water were added for dissolving the solid salts. As depicted in Supplementary Fig. 1, the hand-grinding initially formed a brownish liquid as hydrated water molecules were released from the metal salts and CO2 bubbles were from NH4HCO3. It continued until complete Al precipitation that was witnessed by negligible formation of bubbles. This process took about 20 min. Then, the resulting wet solid was directly calcined at 600 °C for 5 h with a ramping rate of 2 °C min−1 to ensure uniform pore development and attain better homogeneity by lowering a dehydration rate and avoiding rapid grain growth. The calcined samples were subjected to H2 reduction at 350 °C for 3 h prior to the activity test. Note that the optimal Pd loading and calcination temperature were determined through preliminary tests for the dehydrogenation of H12-MBP (Supplementary Table 1): (a) when the Pd loading varied from 1 to 5 wt% with an interval of 2 wt%, the H2 yield (calculated from GC results) and H2 volume (measured by a gas burette) were the highest at the Pd loading of 1 wt% and (b) when the calcination temperature varied from 400 to 700 °C with an interval of 100 °C, the dehydrogenation efficiency was the maximum for the MPdA sample calcined at 600 °C. The measured catalytic activities agreed well with the results of Pd particle size and Pd dispersion measured by CO pulse chemisorption experiments.

First of all, the calcined samples were characterized by N2 physisorption, as listed in Table 1. The non-reduced MPdOA follows the type IV isotherm with a composite of H2 and H3 hysteresis loops representing the presence of mesopores (the most probable pore diameter of 4.1 nm calculated from the desorption branch) with ink-bottle or channel-like pore connectivity (Fig. 1a, b), which is consistent with the characteristics of the previous materials prepared by the solvent deficient precipitation method29. The very similar isotherms were taken for MA (mesoporous Al2O3 prepared without Pd source using the same procedure as MPdOA) and 1 wt% PdO/MA (prepared by incipient wetness impregnation of Pd source onto MA) samples. The BET surface area and pore volume were lower for MPdOA and PdO/MA compared to MA due to the existence of PdO particles that was evidenced by the standard reflections of PdO (101) and (112) planes (JCPDS #43-1024) in Fig. 1c. Importantly, MPdOA was more porous than PdO/MA, possibly resulting from a homogeneous solid solution made by the one-pot method, although the XRD patterns of the two samples appeared very similar. In contrast, 1 wt% PdO supported onto commercially available γ-Al2O3 (hereafter abbreviated as gA) showed the low specific surface area of 264 m2 g−1 even though the pore volume (0.59 cm3 g−1) and diameter (6.5 nm) were larger compared to MPdOA and PdO/MA because of high crystallinity of γ phase in the parent alumina.

### Dehydrogenation activity of mesoporous Pd-Al2O3 catalyst

The catalyst samples obtained after H2 reduction under the identical condition were tested in the dehydrogenation of H12-MBP for 4 h with the metal-to-reactant (M/R) ratio of 0.1 mol%, where H2 volume was monitored by a gas burette. When the catalytic performance was first compared at 250 °C, the volume of H2 released was higher for MPdA (790 cm3) than for Pd/MA and Pd/gA (Fig. 2a). MPdA also showed the higher 1st-order rate constant (0.031 min−1 gcat−1) and turnover frequency (98.9 min−1) than Pd/MA and Pd/gA (Supplementary Table 2). Moreover, the volume of H2 released was larger over MPdA under the identical condition than over the benchmark catalyst Pd/C (662 cm3) reported in our previous work (Supplementary Table 3). The next comparison at 230 °C confirmed that MPdA (496 cm3) was more active than Pd/MA and Pd/gA. This was also true in the activity tests at 270 °C. In terms of the dehydrogenation efficiency, MPdA exhibited the higher values at all the reaction temperatures (Table 2). From the Arrhenius plot, the activation energy (Ea) value of MPdA was much smaller (49.7 kJ mol−1) compared to those of the other impregnated catalysts Pd/MA (80.6 kJ mol−1) and Pd/gA (84.4 kJ mol−1), indicating that MPdA is the best among the studied catalysts.

The volumetric measurement results matched well with those calculated by analyzing the composition of the product mixtures obtained after the 4 h reaction using a gas chromatograph (GC), where the measurement errors ranged within 1–1.5% (Table 2). For the product mixtures obtained at 250 °C, GC analysis revealed that the selectivity to the full dehydrogenation product H0-MBP was 2-fold higher over MPdA than over Pd/gA whereas the difference in the conversion of H12-MBP was merely 7.4% (Fig. 2b). This was also found for Pd/MA, even if the difference was smaller compared to Pd/gA.

Since it was believed that a higher selectivity to H0-MBP greatly contributed to the improved dehydrogenation efficiency of MPdA, the activity test was conducted for a different period of 0.5, 1 or 4 h at 250 °C and 0.4 mol% M/R. In gas chromatogram of the product mixture conducted over MPdA for 0.5 h, huge GC peaks corresponding to H0-MBP were detected along with those of unconverted H12-MBP (Fig. 3a), resulting in the H12-MBP conversion of 41.9% and the outstanding H0-MBP selectivity of 96.8%. As the reaction over MPdA progressed up to 4 h, the H12-MBP conversion increased to ca. 94% with the H0-MBP selectivity maintaining near 97%. In sharp contrast, GC peaks of the reaction intermediate H6-MBP were obviously seen for all product mixtures obtained over Pd/gA (Fig. 3b). The calculation showed that the conversion of H12-MBP increased from 30.9 to 83.2% and the selectivity to H0-MBP also did from 36.7 to 64.9% as the reaction progressed to 4 h. This suggests that MPdA catalyst is selective in the dehydrogenation of H12-MBP (time-dependent H2 release curve shown in Supplementary Fig. 2).

The dehydrogenation efficiency could be enhanced to over 90% by increasing the temperature to 270 °C at 0.1 mol% M/R (Table 2) or by adding more catalyst up to 0.4 mol% M/R at 250 °C (92.4%). Note that the dehydrogenation efficiency approached nearly 99% by the former approach, which is attributed to lower dehydrogenation enthalpy values of N-heterocyclic compounds (i.e., H12-MBP (67.3 kJ/molH2)6, H12-NEC (55 kJ/molH2)35, octahydroindole (57.5 kJ/molH2)36, decahydroquinoline (61.9 kJ/molH2)37, and perhydro 2-methyl-6-(n-methylbenzyl)pyridine (66.7 kJ/molH2)6) compared to homocyclic counterparts (i.e. H12-benzyl toluene (71 kJ/molH2)6,35 that is a counterpart of H12-MBP and decaline (63.9 kJ/molH2)37 that is a counterpart of decahydroquinoline).

Furthermore, the dehydrogenation activities of MPdA and Pd/gA were compared for different N-heterocyclic compounds (Fig. 4 and Supplementary Fig. 3). When octahydroindole (i) was tested at 180 °C due to the low boiling point of 222.7 °C, MPdA exhibited the larger volume of released H2 by 10.5% than Pd/gA. Also, the similar difference of about 12% was found for decahydroquinoline (ii) and for perhydro 2-methyl-6-(n-methylbenzyl)pyridine (iii of which the theoretical H2 storage capacity is 5.8 wt%6) at 230 °C. When the last comparison was done for H12-NEC (iv), the dehydrogenation activity of MPdA at 230 °C was higher by 20.6% than that of Pd/gA. These results suggest that the catalyst MPdA prepared in this work has great potential in the dehydrogenation of various N-heterocyclic compounds.

To demonstrate the effectiveness of MPdA catalyst, the powder sample was pressurized with a PIKE CrushIR Hydraulic Press at 0.5 US tons using an in-house die to give a pellet of 3 mm diameter and 5 mm length. As shown in Supplementary Fig. 4, the pellet-type MPdA exhibited the similar H2 evolution curve to the corresponding powder while the dehydrogenation rate in the initial run was a little low. Furthermore, the analogous H2 evolution results were obtained in the next three consecutive cycles over the pellet-type MPdA catalyst. It is worth noting here that the low H2 evolution rate at the initial stage of dehydrogenation is possibly due to diffusion limitation resulting from the pellet structure. When the product mixture was analyzed by GC, the H2 yield was estimated to be 66.6% that is lower by only 4% than the powder sample. This result guarantees the viability of one-pot solvent deficient precipitation method for the practical application albeit optimization study needs to be done in near future.

### Pd characteristics of mesoporous Pd–Al2O3 catalyst

Since we could not distinguish the reflections of metallic Pd in XRD patterns of the reduced samples due to overlapping with the reflections of alumina (Fig. 1d), CO pulse chemisorption experiments were conducted to estimate the size and dispersion of Pd particles. They unveiled that MPdA retained the smallest Pd particles (Pd size = ca. 2.6 nm, Pd dispersion = 42% and Pd surface area = 190 m2 g−1) than the other catalyst samples (Table 1). This finding was supported by high angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images of fresh samples. When the average sizes were calculated from the particle size distribution charts (Supplementary Fig. 5), the average size of Pd particles increased in the following order: MPdA (3.4 nm) <mPd/MA (4.4 nm) m)Pd/gA (5.3 nm) (Supplementary Fig. 6a).

Furthermore, the samples were characterized by H2-TPR experiments (Fig. 5a). For general Al2O3-supported Pd catalysts, a negative peak below 130 °C was reported to correspond to the decomposition of palladium hydride species formed by PdO reduction and subsequent H2 storage around room temperature, where the intensity and position of this negative peak depend on the size of supported Pd metal particles38,39,40. In the samples studied herein, a negative peak of MPdA was smaller (volume of H2 consumed = 0.61 μmol g−1) and positioned at a higher temperature of 113 °C in comparison with Pd/MA and Pd/gA (Table 1). This indicates the existence of smaller-sized Pd particles in MPdA, which is in good accordance with CO pulse chemisorption results and HAADF-STEM images. The finding is attributed to the solvent deficient environment generated in the synthesis that inhibits the migration of reagent materials and subsequently the over growth of nanoparticles30,41. Therefore, we believed that the higher dehydrogenation performance of MPdA was attributed to the higher surface area and greater dispersion of Pd particles.

Besides, CO-chemisorbed DRIFT spectrum of MPdA showed very distinguishable spectral features from those of Pd/MA and Pd/gA. According to the report of Zhu et al.42, five IR bands were assigned as shown in Fig. 5b; linearly bound CO bands (A1 and A2) above 2000 cm−1 and multiply bound CO bands (B, C, and D) below 2000 cm−1. In the former region MPdA exhibited peak maximum at 2064 cm−1, in contrast to Pd/MA and Pd/gA showing the band at 2086 cm−1 that represents linearly bound CO on defects (mainly, corners) of Pd crystallites42,43,44. Hicks et al. assigned the band at 2064 cm−1 to CO adsorption on well dispersed but not crystalline Pd45, while Lear et al. reported that linearly bound CO on Pd particles’ edge sites related with (111) facet was responsible for this band46. In the latter region MPdA exhibited stronger bands at 1918 and 1860 cm−1 than the impregnated counterparts, where these bands are attributed to isolated-bridged and tri-coordinated CO adsorption on Pd(111) facet. In contrast, the impregnated catalysts than MPdA displayed a stronger band at 1978 cm−1 that is ascribed to compressed-bridged CO adsorbed on structurally open crystal planes such as (100) or (110)47,48. It is worth noting here that the dehydrogenation reaction proceeds dominantly over thermodynamically stable Pd(111) facet of Pd nanoparticles supported on Al2O3 due to the closed-packing of Pd atoms that favor the preferred adsorption geometry of N-heterocyclic LOHC compounds by keeping H atoms on the LOHC molecule and Pd atoms in close proximity and eventually lead to the faster dehydrogenation compared to other Pd surfaces of less packing such as Pd(100) and Pd(110)3,5,26,27. When the relative intensity of CO adsorption bands was calculated in a manner that the intensity sum of Pd(111)-related bands C and D is normalized by the intensity of non Pd(111)-related band B (Supplementary Table 4), MPdA (3.8) showed a higher value than Pd/MA (2.4) and Pd/gA (1.0). This suggests that MPdA is mainly composed of well dispersed Pd phase with a higher density of (111) facet that is preferentially exposed to the reaction mixture, which explains its improved dehydrogenation ability.

### Recyclability of mesoporous Pd–Al2O3 catalyst

In hydrogen release from H12-MBP, the catalyst stability was verified at 250 °C and 0.4 mol% M/R in five consecutive runs. As noted in Fig. 6a (GC results in Supplementary Table 5), the dehydrogenation efficiencies of Pd/gA and Pd/MA continuously dropped to 49.0 and 69.1% in the fifth cycle, respectively, whereas the initial efficiency of MPdA (92.3%) was maintained during five runs. Furthermore, the catalyst durability in the dehydrogenation reaction was once again verified by selecting a recently demonstrated potential LOHC, i.e., perhydro 2-methylindole (H8-MID) with a H2 storage capacity of 5.76 wt%24. Figure 6b presents the much better stability of MPdA at 250 °C and 0.6 mol% M/R in which the H2 release capacity was very stable at 5.70 wt% during five consecutive runs (GC results in Supplementary Table 6). It should be noted here that the different M/R values for H12-MBP and H8-MID were chosen in order to satisfy the liberated amount of H2 required for the mobile applications, i.e., 5.5 wt% of gravimetric density recommended by the U.S. Department of Energy4. The improved stability resulted from the preservation of smaller Pd particles in MPdA (4.0 nm) compared to Pd/MA (9.6 nm) and Pd/gA (11.5 nm) that was acknowledged by HAADF-STEM images of the spent samples after the dehydrogenation of H12-MBP at 250 °C for 4 h (Supplementary Fig. 6b).

### Feature of one-pot solvent deficient precipitation method

To understand the unique stability of MPdA, uncalcined precursors were characterized by XRD. According to the literature30,32,41, the solid-state reaction between metal nitrate and NH4HCO3 forms a solvent deficient environment and then yields an intermediate hydroxide product (e.g. metal hydroxide or metal oxide hydroxide) and a byproduct ammonium nitrate (NH4NO3). The XRD pattern for a dried MA precursor (called “D precursor”) confirmed the presence of NH4NO3 but did not show any reflections of Al phase due to strong diffraction peaks of the ammonium salt (Fig. 7a). Furthermore, the dried MA precursor was rinsed with H2O and then dried again (called “DRD precursor”), of which the XRD pattern revealed a boehmite (AlOOH) phase (Fig. 7b) that is formed according to Eq. 1. Therefore, in the precursor stage boehmite aggregates are surrounded over ammonium nitrate acting as a spatial buffer or template29,41. Subsequent thermal calcination transforms boehmite to gamma-Al2O3 while producing mesopores by the decomposition of NH4NO3.

The diffraction peaks related to palladium were not detected in XRD patterns of two different precursors of MPdA due to the strong diffraction signals of NH4NO3 and AlOOH (Fig. 7a, b). Thus, Pd precursors were prepared using Pd(NO3)2·H2O and NH4HCO3 without Al source. In XRD patterns of the D and DRD precursors of Pd, we observed three Pd-related species, such as PdO, Pd(NH3)2(NO2)2, and [(Pd(NO2)(NH3)3)2(Pd(NH3)4(NO3)4], (Fig. 7c) that are produced by chemical reactions, such as Eqs. 24 in the following:

$${{\mathrm{Al}}\left( {{\mathrm{NO}}_{\mathrm{3}}} \right)_{\mathrm{3}}{\mathrm{\cdot 9H}}_{\mathrm{2}}{\mathrm{O + 3NH}}_{\mathrm{4}}{\mathrm{HCO}}_{\mathrm{3}} \to {\mathrm{AlOOH + 3NH}}_{\mathrm{4}}{\mathrm{NO}}_{\mathrm{3}}{\mathrm{ + 10H}}_{\mathrm{2}}{\mathrm{O + 3CO}}_{\mathrm{2}}}$$
(1)
$${{\mathrm{Pd}}\left( {{\mathrm{NO}}_{\mathrm{3}}} \right)_{\mathrm{2}}{\mathrm{\cdot H}}_{\mathrm{2}}{\mathrm{O + 2NH}}_{\mathrm{4}}{\mathrm{HCO}}_{\mathrm{3}} \to {\mathrm{PdO + 2NH}}_{\mathrm{4}}{\mathrm{NO}}_{\mathrm{3}}{\mathrm{ + 2H}}_{\mathrm{2}}{\mathrm{O + 2CO}}_{\mathrm{2}}}$$
(2)
$${{\mathrm{2Pd}}\left( {{\mathrm{NO}}_{\mathrm{3}}} \right)_{\mathrm{2}}{\mathrm{\cdot H}}_{\mathrm{2}}{\mathrm{O + 6NH}}_{\mathrm{4}}{\mathrm{HCO}}_{\mathrm{3}} \to {\mathrm{2Pd}}\left( {{\mathrm{NH}}_{\mathrm{3}}} \right)_{\mathrm{2}}\left( {{\mathrm{NO}}_{\mathrm{2}}} \right)_{\mathrm{2}}{\mathrm{ + NH}}_{\mathrm{4}}{\mathrm{NO}}_{\mathrm{3}}{\mathrm{ + 9H}}_{\mathrm{2}}{\mathrm{O + 6CO}}_{\mathrm{2}}}$$
(3)
$${\mathrm{6Pd}}\left( {{\mathrm{NO}}_{\mathrm{3}}} \right)_{\mathrm{2}}{\mathrm{\cdot H}}_{\mathrm{2}}{\mathrm{O}} + {\mathrm{22NH}}_{\mathrm{4}}{\mathrm{HCO}}_{\mathrm{3}} \to \\ {\mathrm{2}}\left[ {\left( {{\mathrm{Pd}}\left( {{\mathrm{NH}}_{\mathrm{3}}} \right)_{\mathrm{3}}\left( {{\mathrm{NO}}_{\mathrm{2}}} \right)} \right)_{\mathrm{2}}\left( {{\mathrm{Pd}}\left( {{\mathrm{NH}}_{\mathrm{3}}} \right)_{\mathrm{4}}\left( {{\mathrm{NO}}_{\mathrm{3}}} \right)_{\mathrm{4}}} \right)} \right] \\ + {\mathrm{NH}}_{\mathrm{4}}{\mathrm{NO}}_{\mathrm{3}} + {\mathrm{29H}}_{\mathrm{2}}{\mathrm{O}} + {\mathrm{22CO}}_{\mathrm{2}}$$
(4)

Therefore, these Pd species, boehmite and NH4NO3 are considered to be well mixed in MPdA precursor. After calcination to bring about mesopore formation and phase transition, PdO particles would become protected by adjacent Al2O3 particles probably because the former particles of the extremely small population are located in the vicinity of the latter particles consisting of the mesopore wall, resulting in no appreciable difference in the pore diameters of MA and MPdOA (Fig. 1b). The high resolution transmission electron microscopy (HRTEM) images in Fig. 8a demonstrate that Pd particles are protected by Al2O3 layers in the fresh MPdA. In contrast, such kind of protection lacks in the HRTEM images of fresh Pd/MA (Fig. 8b) and Pd/gA (Fig. 8c). This difference is also visible for the spent MPdA, Pd/MA and Pd/gA samples (Supplementary Fig. 7). Consequently, the dissimilar environment of Pd particles in MPdA, Pd/MA and Pd/gA will influence their surface mobility in the dehydrogenation reaction operated at high temperatures; compared to Pd/MA and Pd/gA, Pd particles are less mobile in MPdA by aid of neighboring alumina particles that can prevent the aggregation of Pd particles and hence keep the initial dehydrogenation activity of MPdA unchanged in the next four runs.

Moreover, we found an unexpected Pd character in the selected area electron diffraction (SAED) patterns obtained from the image of MPdA. In Pd/MA and Pd/gA, two diffraction patterns corresponding to Pd(111) and Pd(200) planes appeared whereas the SAED pattern of Pd(111) showing a d-spacing of 0.224 nm was captured all around the positions of MPdA. This indicates more presence of Pd(111) plane in MPdA. Additionally, the overall SAED patterns of MPdA, Pd/MA and Pd/gA (right-hand sided images in Fig. 8) revealed that bright small spots to make up the rings representing Pd(111), Pd(200) and Pd(220) planes were more evident in Pd/MA and Pd/gA than in MPdA, meaning that Pd particles are polycrystalline in the former samples whereas they are well dispersed in the latter sample. These SAED results further approve the finding of CO-chemisorbed DRIFT analysis that the population of Pd(111) plane decreases in the following order: MPdA PdPd/MA MAPd/gA.

## Discussion

We demonstrate here a one-pot solvent deficient precipitation method for the preparation of mesoporous Pd-Al2O3 (MPdA) catalyst providing the improved H2 release and outstanding durability in the dehydrogenation of N-heterocyclic compounds. Catalyst characterization works manifest the high dispersion of small Pd nanoparticles mainly composed of Pd(111) plane in the fresh MPdA and the inhibited growth of Pd particles by Al2O3 neighbors in the consecutive reuse of MPdA. These unique properties are owing to the generation of solvent deficient environment in the synthesis, which not only constrains the nucleation and growth of nanoparticles but involves the participation of Pd and Al species in the construction of mesoporous wall. Consequently, the catalyst MPdA is believed to offer great potential in a variety of hydrogenation and dehydrogenation reactions, where Pd loading and calcination condition need to be specified for each reaction. In addition, the synthesis method reported herein can be applied potentially for other supported metal catalysts: for instance, we have very recently succeeded in preparing a mesoporous Pt–Al2O3 (MPtA) catalyst that shows extraordinary durability and highly greater hydrogenation and dehydrogenation performance for commercially available LOHC materials.

## Methods

### Synthesis of mesoporous Al2O3

Aluminum nitrate Al(NO3)3·9H2O and ammonium bicarbonate NH4HCO3 with a molar ratio of 1:3 were mixed together in a mortar and grounded thoroughly by a pestle at room temperature, which proceeded during ca. 20 min. The resulting precursor was directly calcined at 600 °C for 5 h with a ramping rate of 2 °C min−1 to obtain a mesoporous Al2O3 (denoted as MA).

### Synthesis of 1 wt% Pd supported on γ-Al2O3 or mesoporous Al2O3

A desired amount of Pd(NO3)2.H2O dissolved in an aqueous solution was impregnated onto γ-Al2O3 (STREM) or mesoporous alumina (MA). After drying at 105 °C in an oven, the final catalyst samples were obtained by calcination at 600 °C for 5 h with a ramping rate of 2 °C min−1 followed by H2 reduction at 350 °C for 3 h with a ramping rate of 5 °C min−1, yielding Pd/ γ-Al2O3 (Pd/gA) and Pd/MA.

### Synthesis of mesoporous 1 wt% Pd–Al2O3

In a mortar, palladium nitrate Pd(NO3)2·H2O (78.459 mg), aluminum nitrate Al(NO3)3·9H2O (22.291 g), and ammonium bicarbonate NH4HCO3 (14.235 g) were mixed together and grounded thoroughly by a pestle at room temperature for ca. 20 min. The resulting material was subjected to the identical calcination as described above for MA synthesis, while the calcined sample was reduced under the condition identical for Pd/gA and Pd/MA. The prepared catalyst is denoted as MPdA.

### Catalyst characterization

BET surface area of the sample (100 mg) was measured in a Micromeritics 3Flex after pretreatment at 200 °C for 2 h under vacuum. X-ray diffraction (XRD) analysis was conducted with a Rigaku miniFlex300 diffractometer using a Cu Kα source (30 kV and 10 mA), where all patterns were recorded in 2θ range 10–80° at a scan rate of 4° min−1 with a step of 0.02°. High resolution transmission electron microscopy (HRTEM) and high angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images were taken in a JEOL JEM-2100F microscope with an acceleration voltage of 200 kV, where the specimen was prepared by dropping a catalyst powder in methanol onto a copper grid and vacuum drying at 60 °C. CO pulse chemisorption experiment was done in a BELCAT-B instrument (BEL Japan), where a sample of 50 mg was reduced at 350 °C for 3 h at a ramping rate of 5 °C min−1 in a 10% H2/Ar flow of 30 cm3 min−1. After cooling to 30 °C in a He flow, 5% CO in He was repeatedly injected as pulses until the peak area became saturated. The amount of CO chemisorbed was determined by quantification of the integrated peak areas from the calibrated area obtained by injecting a 5% CO/He gas through a sample loop. From CO uptake data, the average size of Pd particles was calculated assuming a spherical geometry of Pd crystallites. Also, the percentage of exposed Pd atoms (i.e., Pd dispersion) was deduced from the assumption that the ratio CO/Pd is 1:1. In  H2 temperature-programmed reduction (H2-TPR) experiments, a sample of 60 mg was pretreated under an argon flow at 500 °C for 1 h and then cooled down to room temperature. Then, the volume of H2 consumed for supported Pd catalyst samples was measured on a Micromeritics AutoChemII 2910 instrument equipped with a thermal conductivity detector as the temperature increased up to 800 °C (10 °C min−1) in a 10% H2/Ar flow of 50 cm3 min−1. CO chemisorbed diffuse reflectance infrared fourier transform (DRIFT) spectra were taken in a Nicolet 6700 FT-IR installed with a Harrick Praying Mantis high-temperature reaction chamber. Prior to the measurement, a sample of 30 mg was reduced at 350 °C for 3 h in a 5% H2/Ar flow of 100 cm3 min−1. The sample was cooled to 50 °C and the gas flow was then switched into argon for removal of physisorbed molecules. After Ar purge for 3 h, a background spectrum was taken. Then, the sample was exposed to 5% CO/Ar flow of 50 cm3 min−1 for 30 min. Finally, a CO-chemisorbed DRIFT spectrum was acquired after evacuation for 30 min.

### Catalytic activity test

A reduced catalyst sample was tested in the dehydrogenation of H12-MBP in a batch reactor (volume 90 cm3) described previously6,25. In a typical experiment, H12-MBP of 7.3 mmol was added dropwise to the catalyst pre-loaded at the bottom of the reactor vessel, where the catalyst weight varied to obtain a desired metal-to-reactant (M/R) ratio. Prior to the reaction, the reactor was purged with N2 for 10 min to remove inner moisture and oxygen. Then, the reaction mixture was heated at 18 °C min−1 to a desired temperature using a commercial heat-transfer fluid, followed by the dehydrogenation reaction for a desired period. The volume of H2 released during the reaction was monitored by the water volume displaced in a glass burette and measured with an accuracy of ±2 cm3 (examined by repeating the experiment no less than three times under the identical condition). Notably, the close examination of gas and liquid products found no formation of other gases except H2 (Supplementary Fig. 8) and no decomposition of the substrate in the liquid phase (Supplementary Fig. 9). After completion of the reaction, the reactor was cooled to room temperature and the reaction mixture was diluted with acetone (20 cm3) and filtered via a syringe filter (0.1 μm). An aliquot of this solution was mixed with a GC calibration solvent nonane and then analyzed with a Younglin YL6500 gas chromatograph (GC) equipped with a FID detector and a HP-5 column (30 m × 0.32 mm × 0.25 μm). More detailed GC analysis was done in an Agilent Technologies 7890 A GC equipped with an auto sampler (7683D), a flame ionization detector (FID) and a Restek Rxi®-17Sil column (30 m × 0.25 mm × 0.25 μm). The conversion of H12-MBP and the selectivities to H6-MBP and H0-MBP were calculated from the quantities of the reactant consumed and the products formed, using the calibration factors calculated quantitatively from the calibration curves of three authentic samples.

The aforementioned procedure was applied for H2 release from different substrates such as H12-NEC (7.24 mmol), octahydroindole (9.74 mmol), decahydroquinoline (10.83 mmol), perhydro 2-methyl-6-(n-methylbenzyl)pyridine (8.6 mmol) and perhydro 2-methylindole (10.77 mmol).

### Catalyst stability test

Based on the steps taken in our previous study25, the catalyst sample recovered after the reaction was washed three times with a mixture of acetone and heptane and dried at 105 °C in an oven. As a final step, the sample was reduced by H2 at 350 °C for 3 h. This procedure was repeated after each reaction run.