Enhanced hydrogen generation by reverse spillover effects over bicomponent catalysts

The contribution of the reverse spillover effect to hydrogen generation reactions is still controversial. Herein, the promotion functions for reverse spillover in the ammonia borane hydrolysis reaction are proven by constructing a spatially separated NiO/Al2O3/Pt bicomponent catalyst via atomic layer deposition and performing in situ quick X-ray absorption near-edge structure (XANES) characterization. For the NiO/Al2O3/Pt catalyst, NiO and Pt nanoparticles are attached to the outer and inner surfaces of Al2O3 nanotubes, respectively. In situ XANES results reveal that for ammonia borane hydrolysis, the H species generated at NiO sites spill across the support to the Pt sites reversely. The reverse spillover effects account for enhanced H2 generation rates for NiO/Al2O3/Pt. For the CoOx/Al2O3/Pt and NiO/TiO2/Pt catalysts, reverse spillover effects are also confirmed. We believe that an in-depth understanding of the reverse effects will be helpful to clarify the catalytic mechanisms and provide a guide for designing highly efficient catalysts for hydrogen generation reactions. H2 is regarded as an attractive green fuel and a promising energy carrier for the future. Here, the promotion functions of reverse spillover effects in the H2 generation reactions over bicomponent catalysts are revealed by performing in situ X ray absorption near edge structure characterization.

T he ever-increasing global energy demand and the detrimental effect of the CO 2 product of fossil fuels have triggered a widespread search for alternative energy sources, which are effective and renewable and do not cause further environmental issues 1 . Because of its high energy density and renewability, H 2 has been regarded as an attractive green fuel and a promising energy carrier for the future to meet increasing energy and environmental challenges 2 . Catalytic H 2 generation from hydrogen storage materials is considered a potential method of H 2 production if they can be effectively catalyzed 3,4 . The search for efficient catalytic systems would be greatly facilitated by a clearer understanding of the underlying chemical process.
Noble metal catalysts, such as Pt, Pd, and Ru, have been recognized as important classes of catalysts for hydrogen generation, due to their high catalytic activity and durability [5][6][7][8][9][10][11] . It is noted that coupling metal catalysts with secondary metals [12][13][14][15] and/or transition metal oxides [16][17][18][19][20][21][22][23][24][25] is an encouraging strategy to further enhance catalytic performance. In the past, various theories (e.g., the metaloxide interfacial sites, electron interactions, or hydrogen reverse spillover effect) have been offered to explain the enhancement of H 2 generation when different components are combined in a catalyst. For example, Francisco Zaera and coworkers argued that in the photocatalytic production of H 2 from water with semiconductor catalysts, the role of metal additives is a reverse spillover effect, not to trap excited electrons 26 . Hydrogen reverse spillover, as a form of spillover, involves the migration of adsorbed hydrogen atoms from an oxide (or other nonmetal surface) to a metal, where they recombine to molecular hydrogen [27][28][29] . However, due to the lack of well-defined catalysts with clearly separated functional components and the difficulties in performing in situ characterization technologies, researchers have not formed an agreement on the enhancement mechanism. It is still a challenging issue to reveal the promotion effects of reverse spillover in H 2 generation reactions.
In this work, taking the ammonia borane (NH 3 ·BH 3 , AB) hydrolysis reaction as an example, the promotion functions of reverse spillover in this reaction are proven using a spatially separated NiO/Al 2 O 3 /Pt catalyst as a model catalyst, in combination with in situ quick XANES characterization. The NiO/Al 2 O 3 /Pt catalyst was prepared by a facile and general template-assisted atomic layer deposition (ALD) method [30][31][32][33][34][35][36] . In situ XANES results clearly reveal that for H 2 generation from AB, the H species generated at NiO sites spill across the support to the Pt sites, i.e., reverse spillover phenomenon. This accounts for the enhanced H 2 generation rates of bicomponent oxide-metal catalysts, compared with single component Pt-based catalysts. The reverse spillover effects are also confirmed for the CoO x /Al 2 O 3 /Pt and NiO/TiO 2 /Pt catalysts. Our study provides a guide for designing highly efficient catalysts for hydrogen generation reactions.

Results and discussion
Synthesis and characterization of the catalysts. The NiO/Al 2 O 3 / Pt catalyst was synthesized by ALD using carbon nanocoils (CNCs) as templates ( Supplementary Fig. 1). First, Pt nanoparticles (20 ALD cycles) and an Al 2 O 3 film (50 ALD cycles) were deposited onto CNCs. The CNC templates were then removed by calcination. Finally, NiO nanoparticles (100 ALD cycles) were deposited, obtaining NiO/Al 2 O 3 /Pt. Al 2 O 3 /Pt and NiO/Al 2 O 3 were also produced as reference catalysts. Figure 1a shows transmission electron microscopy (TEM) image of NiO/Al 2 O 3 /Pt. Hollow Al 2 O 3 nanotubes with a uniform wall thickness (ca. 7 nm) can be clearly observed. The lattice distance of Pt nanoparticles was measured to be~0.226 nm ( Supplementary Fig. 2), which corresponds to the Pt(111) plane. High-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) image and energy-dispersive X-ray spectroscopy (EDX) mapping (Fig. 1d, Fig. 4). Due to the small size and low contrast of NiO nanoparticles, it is not straightforward to distinguish NiO nanoparticles in Fig. 1a, c. From the HRTEM image of NiO/ Al 2 O 3 (inset in Fig. 1c), NiO nanoparticles can be clearly observed. The Pt content in the catalysts was measured using inductively coupled plasma-atomic emission spectrometry (ICP-AES) to be 3 The X-ray photoelectron spectroscopy (XPS) results reveal the existence of Ni 2+ species in NiO/Al 2 O 3 and NiO/Al 2 O 3 /Pt (Fig. 1h). The XPS peaks for the two catalysts are similar. The XPS peaks located at binding energies of 856.1 and 874.0 eV are attributed to Ni 2p 3/2 and Ni 2p 1/2 , respectively, and the peaks located at binding energies of 861.8 and 879.7 eV are attributed to satellite peaks. From the X-ray diffraction (XRD) patterns for the Al 2 O 3 /Pt and NiO/Al 2 O 3 /Pt catalysts ( Supplementary Fig. 5), the presence of Pt nanoparticles can be confirmed. No diffraction peak assigned to NiO is detected from the XRD patterns for NiO/ Al 2 O 3 and NiO/Al 2 O 3 /Pt, which can be ascribed to the high dispersion of ALD-prepared nanoparticles. Hydrogen temperature programmed reduction (H 2 -TPR) was used to study the redox properties of the catalysts (Fig. 1i)

ARTICLE
Enhanced hydrogen generation after NiO addition. Here, the hydrolytic dehydrogenation of AB for H 2 production is selected as a model reaction to investigate the reverse spillover effect. The catalytic performances of different catalysts for the dehydrogenation reaction of AB are shown in Fig. 2. A nearly linear H 2 evolution curve is obtained for Al 2 O 3 /Pt, suggesting a zero-order reaction with respect to AB (Fig. 2a). For the NiO/Al 2 O 3 catalyst, its H 2 evolution curve exhibits a long induction period of approximately 20 min, after which the curve starts to rise gradually. The NiO particles for NiO/ Al 2 O 3 are located on the outer surfaces of the Al 2 O 3 nanotubes. The reactant molecules (H 2 O and AB) easily access the exposed NiO sites. Thus, the mass transfer in the porous structures is unlikely to lead to the induction period of the NiO/Al 2 O 3 catalyst. It is generally believed that new active species are generated during the induction period ( Fig. 2b) [42][43][44] . However, even after reaction for 60 min, the H 2 evolution volume is still only 3.8 mL, which shows an extremely poor activity for NiO/Al 2 O 3 . For NiO/Al 2 O 3 /Pt, a rapid and linear H 2 evolution curve without an induction period is obtained. The time required to complete the hydrolysis reaction for the NiO/ Al 2 O 3 /Pt catalyst is less than that for Al 2 O 3 /Pt, indicating that NiO addition can greatly enhance the activity of the Al 2 O 3 /Pt catalyst, even though NiO alone has little activity.  Fig. 8). The catalysts still retain a linear increase in the H 2 generation volume with reaction time. According to Arrhenius plots of ln k versus 1/T (Fig. 2c), the activation energies (E a ) for the hydrolysis of AB using Al 2 O 3 /Pt and NiO/Al 2 O 3 /Pt are calculated to be 49.1 and 32.8 kJ mol -1 , respectively. The effect of the AB amount on the hydrolysis of AB was investigated ( Supplementary Fig. 9). A nearly horizontal relationship in logarithmic plots between the H 2 generation rate and AB concentration is further normalized, indicating that hydrolysis over NiO/Al 2 O 3 /Pt is also a zero-order reaction with respect to the AB concentration.

Al
The effects of the distance between NiO and Pt components (i.e., the thicknesses of the Al 2 O 3 support) and NiO loadings in the catalyst on the catalytic performance were investigated ( Supplementary Fig. 10 Catalytic mechanism. Raman measurements were employed to characterize the used catalysts after reaction for 10 min, as shown in Fig. 2d. The Raman spectrum for the reference sample AB shows the B−N stretching mode at 727 and 783 cm -1 , the B-H stretching mode at 2280 and 2375 cm -1 , the N−H stretching mode at 3175, 3251, and 3316 cm -1 , the BH 3 deformation mode at 1159 and 1188 cm -1 , and the NH 3 deformation mode at 1600 cm -1 , in agreement with the literature results 45 . For the Al 2 O 3 support after reaction, these peaks can still be observed. For the NiO/Al 2 O 3 and Al 2 O 3 /Pt samples after reaction, the NH 3 deformation peak can be found at approximately 1600 cm -1 , while the B−N and B-H peaks cannot be observed. These results demonstrate that NiO and Pt can readily dissociate the B−N and B−H bonds of AB in the presence of H 2 O. The dynamic behaviour of Ni species in the catalysts under the H 2 generation reaction was probed with a quick XANES. The incident X-rays usually produce no damage to the material, as opposed to the action of electron or ion probes 46 . This capability of XANES makes it suitable for (in situ) catalyst structure studies 47 . From Fig. 3a, it can be found that the intensity of the white line peak for NiO/Al 2 O 3 decreases with the reaction time, indicating that the Ni 2+ species are gradually reduced. The Ni species are far from being fully reduced after reaction for 60 min. Furthermore, the in situ XANES spectrum was simulated by a linear combination of the ex situ spectrum of the as-prepared catalyst and the spectrum for the reference sample (Ni foil) to quantitatively reveal the dynamic behaviour of Ni species in the catalysts during the reaction. The experimental XANES spectra can be reproduced perfectly by simple linear fitting, with an extremely low R factor (Supplementary Fig. 12 and Supplementary Table 3). The reduction degrees for NiO/Al 2 O 3 after reaction for 10, 20, 30, 40, 50, and 60 min are 3.6 ± 0.3, 7.1 ± 0.4, 10.0 ± 0.2, 11.8 ± 0.3, 13.6 ± 0.3, and 14.2 ± 0.2%, respectively (Fig. 3b). The reduction degree does not increase linearly with reaction time. These results demonstrate that metallic Ni 0 species are generated gradually during the reaction, and the Ni 0 generation rate slows down with time. For NiO/Al 2 O 3 /Pt, one may expect that more NiO will be reduced into metallic Ni 0 after Pt addition due to spillover effects. Surprisingly, the in situ XANES spectra remain unchanged throughout the reaction, indicating that the reduction of Ni 2+ species is totally inhibited after Pt addition (Fig. 3c, d).

Discussion
Our  49,50 . Thus, reverse spillover is the lowest energy pathway. This is also confirmed experimentally. The H 2 evolution curve for NiO/Al 2 O 3 confirms that the H species are not released from NiO. The reduction of NiO is totally inhibited after Pt addition, revealing that the H species generated at NiO sites are not consumed for the reduction of NiO. The H species spill across the Al 2 O 3 support from NiO to Pt sites, where they can combine into H 2 and release (Fig. 3f). This is called the reverse spillover process, which accounts for the enhanced H 2 generation rate for NiO/Al 2 O 3 /Pt after NiO addition.
The reverse spillover phenomenon has also been confirmed in other catalytic systems, for example, in AB hydrolysis catalyzed by CoO x /Al 2 O 3 /Pt. As shown in Fig. 4a, an induction period can also be found in the H 2 evolution curve for the CoO x /Al 2 O 3 catalyst. For CoO x /Al 2 O 3 /Pt, a rapid and nearly linear H 2 evolution curve is obtained, and its activity is higher than that of Al 2 O 3 /Pt. From the in situ XANES results, it can be found that after reaction for 30 min, the position of the white line peak for CoO x /Al 2 O 3 shifts to a lower energy, and the intensity of the white line peak decreases, indicating that the Co oxide species are reduced (Fig. 4b). For CoO x /Al 2 O 3 /Pt, the change in the white line peak after the reaction is very slight (Fig. 4c). The in situ XANES spectrum was simulated by a linear combination of the ex situ spectrum for the as-prepared catalyst and the spectra obtained for the reference samples (CoO and metallic Co 0 ) ( Supplementary Fig. 14). For CoO x /Al 2 O 3 after reaction for 30 min, 9.9% extra CoO and 12.9% extra metallic Co 0 are formed. However, for CoO x /Al 2 O 3 /Pt, 9.7% of extra CoO and only 1.6% of extra metallic Co 0 are formed. It can be concluded that the reduction of Co oxide species to metallic Co 0 is mostly inhibited after Pt addition.
In addition to nonreducible Al 2 O 3 , when reducible TiO 2 is used as a support, reverse spillover effects are also confirmed ( Supplementary Figs. 15 and 16). As shown in Fig. 4d, the induction period in the H 2 evolution curve for the NiO/TiO 2 catalyst is shortened to 7 min, indicating that NiO supported on TiO 2 is easier to reduce than that supported on Al 2 O 3 . The H 2 evolution curve for NiO/TiO 2 /Pt is not linear. In the beginning, its rate is similar to that of TiO 2 /Pt. After that, the rate for NiO/ TiO 2 /Pt begins to increase rapidly, exceeding the rate for TiO 2 /Pt. This implies that NiO sites are reduced to metallic Ni 0 sites during the reaction. The in situ XANES spectra for NiO/TiO 2 and NiO/TiO 2 /Pt are slightly rough, which is due to the high activities of the catalysts. The liquid reaction system is disturbed by a large amount of H 2 bubbles, and thus, the X-ray absorption is affected. The in situ XANES (Fig. 4e, f) and its linear combination fitting results ( Supplementary Fig. 17) demonstrate that for NiO/TiO 2 after reaction for 30 min, 26.1% extra metallic Ni 0 is formed, while for NiO/TiO 2 /Pt, 13.0% extra metallic Ni 0 is formed. The reduction of NiO to metallic Ni 0 is partially inhibited after Pt addition because of the reverse spillover effects. There are two competing pathways for the H species generated at the NiO sites of NiO/TiO 2 /Pt. A fraction of the H species spill over reversely to Pt sites; the rest is consumed to reduce NiO to metallic Ni 0 .
In summary, we designed spatially separated NiO/Al 2 O 3 /Pt catalysts to clarify the contribution of the reverse spillover effect to enhanced H 2 generation rates. The in situ XANES results reveal that the H species generated at NiO sites are not consumed for the reduction of NiO to Ni 0 or released as H 2 at NiO sites. Instead, they reversely spill across the support to the Pt sites. The reverse spillover effects account for the enhanced H 2 generation rates. The effects are also confirmed for CoO x /Al 2 O 3 /Pt and NiO/ TiO 2 /Pt catalysts. In general, we believe that, with the help of an in-depth understanding of reverse spillover effects, this work can provide guidance for rationally designing highly efficient catalysts for H 2 production in the future. Sample characterization. The chemical compositions of these samples were determined by ICP-AES. The TEM and HRTEM images were taken on a JEOL-2100F microscope. The N 2 sorption measurements were performed using Micromeritics Tristar 3000 at 77 K. XRD patterns were collected on a Bruker D8 Advance X-ray diffractometer using a Cu Kα source. XPS spectra were recorded on an AXIS ULTRA DLD spectrometer (Shimadzu/Kratos) to characterize the surface composition with the Al Kα line as the excitation source. H 2 -TPR experiments were performed using a tubular quartz reactor (TP-5080, Tianjin Xianquan, China), into which a 50 mg sample was loaded. The reduction was conducted in a 10% H 2 /N 2 atmosphere at a heating rate of 10°C/min. Hydrogen consumption was calculated by an external standard method using H 2 -TPR for CuO as the standard. The Raman spectra were performed on a LabRam HR Evolution (Horiba, France) spectrometer employing a He−Ne laser with an excitation wavelength of 532 nm. After the AB catalytic hydrolysis reaction for 10 min, the catalysts were centrifuged and dried in a vacuum oven at 30°C. Finally, the samples were loaded, and the spectra were recorded at room temperature. The in situ XANES for Ni and Co K-edge were obtained on the 1W1B beamline of the Beijing Synchrotron Radiation Facility (BSRF), Institute of High Energy Physics, Chinese Academy of Sciences, and the BL14W1 and BL11B beamlines of the Shanghai Synchrotron Radiation Facility (SSRF), Shanghai Advanced Research Institute, Chinese Academy of Sciences. A Si (111) double-crystal monochromator was used to reduce the harmonic component of the monochrome beam. Ni and Co foil, NiO, CoO, and Co 3 O 4 were used as reference samples and measured in transmission mode. The sample wafer was placed in the centre of a homemade in situ XANES cell. The spectra for the catalyst were first collected in transmission mode. After that, the AB solution (5 g L -1 ) was fed into the reactor at a speed of 5 mL min -1 by a sampling pump. The quick XANES were collected during the reaction at different times. IFEFFIT software was used to calibrate the energy scale, to correct the background signal, and to normalize the intensity. The spectra at the edge jump were simulated by a linear function of the reference Ni foil and the NiO-based catalyst before the reaction to estimate the proportion of metallic Ni 0 in the catalyst during the reaction. The following formula was used:

Methods
(in situ XANES) = f 1 ·(XANES of Ni foil) + f 2 ·(ex situ XANES), where f 1 and f 2 are the fractions of the Ni foil and the as-prepared catalyst before the reaction, respectively.
Catalytic testing. The catalytic performance of the samples was tested for AB hydrolytic dehydrogenation. Typically, the catalysts were first dispersed in deionized water (10 mL) placed in a round bottom flask with a magnetic stirrer at 25 ± 0.5°C. The reaction was initialized by adding 48 mg of AB (Aldrich, 97%) into the reaction flask under stirring (700 rpm). A gas burette filled with water was connected to the flask to measure the amount of hydrogen evolved during the reaction by monitoring the displacement of the water level. In the AB concentration-dependent study, the reaction was performed at different AB concentrations (75, 112.5, 150, and 187.5 mmol L -1 ) at 25 ± 0.5°C. To calculate the activation energy (E a ), the reaction temperature was varied in the range of 20-35°C, and the AB concentration was kept constant at 150 mmol L -1 .
Computational method. All DFT calculations were carried out using periodic spin-polarized density functional theory with the Perdew−Burke−Ernzerhof generalized gradient approximation functional 51 as implemented in the Vienna ab initio simulation package (VASP) 52,53 . The calculations were performed using a plane-wave basis set, with a cut-off kinetic energy of 400 eV. Projector-augmentedwave 54 potentials were used to describe the electron−ion interactions. Dispersion interactions were included by using the DFT-D3 (BJ) correction method of Grimme et al 55,56 . The crystal structure of γ-Al 2 O 3 proposed by Gutiérrez et al. 57 was adopted in our model system. The most stable (100) surface of γ-Al 2 O 3 with three alumina layers and a Ni 4 O 4 cluster adsorbed onto it were used for the NiO/γ-Al 2 O 3 (100) slab model. The two bottom layers of the slab were kept fixed. The thickness of the vacuum region was 20 Å. A Monkhorst-Pack grid was used for Brillouin-zone integrations with 1 × 1 × 1 k-mesh (gamma point) sampling. The solvation effect was included with an implicit solvation solvent of water using the VASPsol tool 58 . The free energies at room temperature (298.15 K) were obtained by adding to the DFT electronic energy (E), the zero-point energy, enthalpy, and entropy contribution from the vibrational modes. The transition states (TS) were calculated using the climbing image nudged elastic band method 59 , and frequency analysis was confirmed to verify the TS.