Dual Functionality of a Plasmonic Gold Photocatalyst in the Reversible Dehydrogenation / Hydrogenation of N‐Heterocycles

ABSTRACT

LOHC (Liquid Organic Hydrogen Carrier) systems have emerged as a promising strategy for hydrogen storage and distribution. Among these, N‐Heterocycles stand out due to their high hydrogen capacity. Their photocatalytic dehydrogenation has attracted significant attention due to the environmentally friendly conditions under which the reaction can proceed. We report here a new hybrid plasmonic material in which a covalent organic framework (COF) serves as a stabilizing matrix for gold nanoparticles, grafted in situ via an organometallic approach, with sizes ranging from 2 to 4 nm. We have employed the synthesized material as a recyclable photocatalyst in a light‐mediated dehydrogenation of N‐Heterocycles. This new carbon‐based semiconductor photocatalyst also introduces a novel photocatalyst capable of generating both dehydrogenation and hydrogenation products by simply changing the reaction conditions, offering a significant advance for hydrogen storage.

A new hybrid plasmonic material, Au@COF, serves as a recyclable photocatalyst for the light‐mediated dehydrogenation and hydrogenation of N‐Heterocycles, where simply adjusting the reaction conditions allows control over the reaction mechanism, marking a notable breakthrough for hydrogen storage.

1. Introduction

In recent years, significant efforts have been devoted to advance in hydrogen production and promote its use as a new renewable energy source to replace the traditional and widely used non‐renewable energy sources. However, hydrogen storage still remains an unresolved problem. In this sense, LOHC (Liquid Organic Hydrogen Carriers) systems have emerged as a promising strategy for hydrogen storage and release, since they can be subjected to multiple cyclic dehydrogenation/hydrogenation processes [1]. Several systems are included in this group of molecules, such as cycloalkanes [2, 3], N‐heterocycles [4], formic acid [5], or methanol [6, 7]. Notably, N‐heterocycles offer high hydrogen capacity reaching up to 6.7 wt.% in compounds like dodecahydrocarbazol [4].

If we focus on homogeneous catalysis for the dehydrogenation/hydrogenation of N‐heterocycles, transition metal‐based catalysts show high activity and selectivity under mild conditions, but their recovery and stability are disadvantageous. In contrast, heterogeneous catalysts mainly based on supported metal nanoparticles stand out for their recyclability and high stability, although they need, traditionally, harsher reaction conditions (Scheme 1) [8]. In any case, an appropriate, efficient hydrogen storage/release strategy would involve a unique bifunctional system with catalytic activity for the dehydrogenation/hydrogenation process [9]. Some bifunctional heterogeneous catalysts based on supported Pt, Pd, Ni, or Co nanoparticles have been described, but they are used to operate at high temperatures [10, 11, 12, 13].

SCHEME 1.

Advantages and disadvantages of each catalytic system in the dehydrogenation/hydrogenation of N‐Heterocycles.

During the last few years, the photocatalytic dehydrogenation of N‐Heterocycles has gained great interest due to its more environmentally friendly chemical reaction conditions. Thus, this approach usually needs less harsh reaction conditions and avoids the use of hazardous chemicals or oxidants, being assisted by atmospheric oxygen. However, the use of this methodology in homogeneous conditions usually needs a photosensitizer and a transition metal catalyst. Recently, Mata et al. proposed a new iridium (III) organometallic photocatalyst that does not need a cocatalyst and plays a dual role, being able to produce the hydrogenation reaction, changing the catalytic conditions [14]. On the heterogeneous side, the focus is on titanium dioxide‐based catalytic systems, which play a semiconductor photocatalytic role, with Ni and Rh, as the transition metals, and on Nb₂O₅ photocatalysts that proceed through an oxidative dehydrogenation mechanism [15, 16, 17]. Other photocatalytic systems based on boron carbon nitride (BCN) [18] or perylene diimide (PDI) [19] have been developed. However, they do not show catalytic activity for the reverse reaction. In this regard, photocatalytic hydrogenation has also emerged as a desirable alternative to thermal methods, as it enables reactions to proceed under milder conditions by utilizing solar energy, significantly lowering the apparent activation energy required for the hydrogenation of LOHCs. Indeed, the field of photocatalytic hydrogenation has seen remarkable progress with the development of various photosensitive catalysts, including plasmonic metals and metal–organic frameworks (MOFs), which facilitate the reduction of organic chemicals and other substrates [20].

Therefore, given the excellent recyclability of heterogeneous photocatalysts and their typically mild operating reaction conditions, it is essential to develop a single recyclable photocatalyst that could efficiently enable both dehydrogenation and hydrogenation transformations under mild conditions. Taking all the previous comments into account, we report herein a wavelength‐controlled reversible photocatalyst that uses blue LED light to dehydrogenate N‐Heterocycles for hydrogen production and then uses green LED light to hydrogenate the resulting molecule in a plasmonic‐driven process. This light‐selective dual‐mode functionality offers a precise and energy‐efficient method for managing chemical energy in a single, reusable system. For this purpose, we have targeted covalent organic frameworks (COFs) as a carbon‐based matrix due to their semiconductor characteristics, high surface areas, and well‐defined crystalline structures formed through robust covalent bonding. In addition to these outstanding properties, COF materials are very good candidates for nanoparticle deposition. Their porosity and tunable pore size can be adjusted to accommodate nanoparticles of various dimensions or even prevent them from agglomeration. In addition, the mesoporous structure, characterized by well‐defined channel cavities, ensures an efficient access to the active sites, essential for photocatalytic applications. Finally, their semiconductor nature makes them suitable candidates for facilitating the electron transport and charge carrier separation in photocatalytic processes, where it is necessary. Thus, COFs have attracted much attention during the last years in the field of heterogeneous catalysis and specifically in the gas storage [21, 22], photocatalysis [23, 24, 25], electrocatalysis [26] or photo‐electrocatalysis [26, 27]. Although several MOF materials have been shown to catalyze the reversible dehydrogenation/hydrogenation of N‐Heterocycles [28, 29], COF materials have not yet been reported in literature as catalysts capable of promoting these reactions.

We propose in this work an organometallic approach for the in situ synthesis of very small gold nanoparticles (∼3 nm) within a COF matrix, from the reduction of [Au(C₆F₅)(tht)] with triisopropylsilane (TIPS). This method offers great advantages over other nanoparticle synthesis methods, such as the employment of mild reaction conditions and high control over the size and shape of nanoparticles. The exceptional porosity of the COF matrix provides an ideal scaffold to regulate the growth of nanoparticles and enhance their photocatalytic performance. In summary, we report herein the first plasmonic COF material exhibiting photocatalytic activity in the reversible dehydrogenation/hydrogenation of N‐Heterocycles.

2. Results and Discussion

2.1. Synthesis and Characterization

Specific synthetic procedures and experimental conditions of the COF matrix are provided in the Supporting Information material. Briefly, the condensation between 1,3,5‐Tris(4‐aminophenyl)benzene (TAPB) and 1,3,5‐triacetylbenzene (BTAC) takes place in the presence of an excess of acetic acid, under stirring for 1 h, resulting in the formation of 1‐COF as a yellow powder. The condensation was confirmed by the disappearance of the signals in the infrared spectrum at 3430 and 3352 cm⁻¹, corresponding to the vibration of the −NH₂ group in TAPB monomer, and 1694 cm⁻¹, corresponding to the vibration of the −C═O group in BTAC (Figure S13). Transmission electron microscopy (TEM) images revealed the formation of COF nanospheres of 202 ± 15 nm in diameter (Figure S17, 1).

The nanomaterial 1‐COF was thermally treated with a mixture of different solvents to obtain a new material 2‐COF with higher porosity, as evidenced by the TEM images (Figure S17, 2), but still maintaining the spherical shape. These porous micro‐channels will serve as a template for the in situ controlled growth of gold nanoparticles (Scheme 1). Moreover, the nanoporous material 2‐COF was acidified following a previously reported procedure to finally obtain a new material 3‐COF as a red powder [24]. The main aim of the synthesis of 3‐COF was to improve the absorption in the visible region of the material for its use as a photocatalyst (Figure S8). The resultant materials were further characterized through a series of spectroscopic techniques such as infrared spectroscopy (FT‐IR), UV–vis spectroscopy, x‐ray photoelectron spectroscopy (XPS), and solid‐state ¹³C NMR. There was no obvious change in IR signals between the solvent‐treated 2‐COF and the bulk 1‐COF. However, the band at 1627 cm⁻¹ corresponding to the vibration of the ─C═N disappears in the modified COF matrix (3‐COF), and a new band at 1668 cm⁻¹ corresponding to the vibration of the group ─C═N─H⁺ appears (Figure S14). The UV–vis spectral comparison between the three COF matrices showed a strong absorption in the 200–450 nm range (Figure S8), with 2‐COF exhibiting a tail until 600 nm. Moreover, a clear redshift of the absorption band toward the NIR region, extending up to 700 nm, was achieved for 3‐COF.

Tauc plot representation of the optical absorbance (αhʋ)² with respect to energy (hʋ) is a typical method to determine the band gap of the semiconductor materials. The experimental spectra of (1‐3)‐COF showed band gap values of 2.86, 2.78, and 2.25 eV, respectively, indicating a clear decrease in the energy band gap for the 3‐COF matrix (Figures S5–S7).

TEM analysis of 1‐COF (Figure S17) revealed a clear spherical shape of the COF matrix with a 202 ± 15 nm diameter. TEM analysis of the 2‐COF material clearly showed a nanoporous structure with internal cavities within the COF spheres. However, acid treatment to obtain 3‐COF resulted in the partial loss of the spherical morphology.

In addition, the successful protonation of 2‐COF was also confirmed through XPS) measurements (Figure S26). Deconvolution of the high‐resolution N 1s XPS spectrum of 2‐COF revealed two peaks at 399.0 and 400.2 corresponding to C═N and C─N⁺, respectively. Analysis of the 3‐COF spectrum revealed a noticeable decrease in intensity of the 399.0 peak and an increase at 400.2 eV, corresponding to the C═N⁺ group (Figure S27). Finally, the connectivity of the framework was confirmed by solid‐state ¹³C NMR spectroscopy (Figure S16).

X‐ray diffraction (XRD) analysis of the synthesized COF materials indicates that the 1‐COF sample is predominantly amorphous (Figure S35). In contrast, materials 2 and 3 displayed higher degrees of crystallinity. This enhancement in structural order may be attributed to the increased porosity that develops after solvent treatment and acidification of the 2‐COF and 3‐COF matrices.

After the preparation and characterization of the COF matrices, small‐sized gold nanoparticles were synthesized within the cavities of the covalent organic frameworks, leading to a controlled size and shape growth. The homogeneous growth of Au NPs was achieved through the mild reduction of the organometallic compound [AuR(tht)] (R = C₆F₅; tht = tetrahydrothiophene), using TIPS as reducing agent. The synthesis of the new materials on the 1‐COF, 2‐COF, and 3‐COF matrices led to the obtention of the plasmonic materials 4‐COF, 5‐COF, and 6‐COF, respectively.

The analysis of the IR signals did not reveal any significant changes in the covalent matrix with respect to the isolated COF matrices after the deposition of gold nanoparticles (Figure S15). We have further characterized the materials through UV–vis spectroscopy, XPS, TEM and scanning TEM (STEM) microscopy. First, we verified the presence of gold nanoparticles through UV–vis measurements (Figure 1A). For 4‐COF and 5‐COF materials, a new absorption band at 550 nm, characteristic of the LSPR (Localized Surface Plasmon Resonance) of gold nanoparticles, appeared. In the case of 6‐COF, the intense absorption of the semiconductor between the 200–700 nm range overlapped the region of the LSPR, making it impossible to observe any change in the absorption due to the presence of gold nanoparticles.

FIGURE 1.

Characterization of (4‐6)‐COF plasmonic materials. (A) UV–vis absorption spectra in the solid state. (B) High‐resolution Au 4f XPS spectra. (C) TEM images. (D) Histogram size distribution of gold nanoparticles. (E,F) HAADF‐STEM of gold nanoparticles at low and high magnification, respectively.

The survey XPS spectrum of (4‐6)‐COF (Figure S28) ruled out the presence of fluorine atoms and, hence, traces of the organometallic [AuR(tht)] (R = C₆F₅) precursor in the sample. We also recorded the high‐resolution XPS spectra for the Au 4f region on the plasmonic (4‐6)‐COF materials (Figures 1B). The deconvolution of the signals revealed a doublet of peaks at 84.1 and 87.7 eV, which was assigned to metallic gold.

The analysis of the TEM images was decisive in determining the shape and size of the gold nanoparticles. If we focused on the 4‐COF material, we observed the presence of gold nanoparticles mainly on the surface of the COF sphere with a mean size of 3.5 ± 0.2 nm (Figure 1C,E). Conversely, the elevated porosity of the 2‐COF matrix promoted the formation of gold nanoparticles both on the surface and within the internal cavities in the COF sphere (5‐COF), resulting in smaller particles (2.8 ± 0.1 nm) compared to those formed on the 1‐COF‐based material (4‐COF). On the other hand, the deposition of gold nanoparticles on the 3‐COF resulted in the formation of even smaller nanoparticles of 2.4 ± 0.1 nm, despite the loss of the spherical shape of the COF matrix, as can be evidenced in Figure 1C (6‐COF).

High‐angle annular dark field (HAADF‐STEM) analysis on the (4‐6)‐COF materials revealed further information about the COF shape and size, the position of gold nanoparticles within the matrix, and the shape and size of the metallic nanoparticles. The relative height of the NPs with respect to the matrix was inferred by changing the focus on the matrix. Figure S16 shows how we started from a focusing condition in which only the NPs located at the outermost edge of the COF were in focus and moved toward lower and higher focusing conditions. The position of the focus can give a rough value for the diameter of the particle, concluding that the COF matrix is not a perfect sphere but flat‐shaped. The shape of the matrix was modified by the treatment with solvents (5‐COF), leading to a more porous structure with apparently big cavities inside (Figure S22). This modified structure led to the formation of gold nanoparticles not only along the edges of the matrix but also throughout the porous network of the covalent organic framework. Moreover, the analysis of the 6‐COF sample evidenced that the matrix still remained unaffected after the acidic treatment, and the porous structure can be easily observable on the surface of the sphere (Figure S24).

From a morphological perspective, the Au NPs appear approximately spherical in shape and monodispersed in size. From a structural point of view, the interplanar distances and angular relationships are consistent with the structure of metallic Au. In this sense, most nanoparticles are single crystals, which means that the extent of the crystal domain corresponds to the size of the nanoparticle. On the other hand, a minority is constituted by twinned nanoparticles (indicated by broken lines in Figures S21, S23, and S25).

The elemental quantification done through XPS (Table S1) on the COF materials reveals the presence of ∼0.90% of gold in the (4‐6)‐COF, being slightly greater in 6‐COF (1.02%). Due to the exclusive presence of metallic gold in the 4–6 samples, the scarce fluorine content in the samples may be attributed to the presence of some C₆F₅ ⁻ anions on the surface of the gold nanoparticles.

The formation of AuNPS within the structure of the COF matrices was confirmed by XRD analysis, as shown in Figure S36. The main diffraction peaks at 2θ values of 38.2° and 44.4° for the (4‐6)‐COF materials correspond to the (100) and (200) planes characteristic of the face‐centered structure of gold nanoparticles, indicating the presence of highly crystalline gold NPs. Comparison of three plasmonic materials reveals that material 6 exhibits the highest crystallinity, whereas material 4 exhibits the poorest. This trend reveals that as the size of gold nanoparticles decreases, their crystallinity is higher. Furthermore, the greater intensity of the diffraction peaks of material 6 may also be attributed to its higher gold content, as evidenced by the elemental quantification from the XPS analysis (Table S1).

2.2. Photocatalytic Dehydrogenation of N‐Heterocycles

Inspired by other hybrid semiconductor materials that catalyze the light‐mediated photocatalytic dehydrogenation on N‐Heterocycles for hydrogen production, we decided to employ the newly synthesized plasmonic COF‐based materials as photocatalysts for this reaction.

We selected the 5‐COF as the reference catalyst for the dehydrogenation reaction of N‐Heterocycles, owing to the homogeneous distribution and well‐defined morphology of gold nanoparticles within the covalent matrix.

Considering that previously reported N‐Heterocyclic dehydrogenation catalytic conditions typically required high temperatures (80°C–200°C), we initially evaluated the catalytic activity of 5‐COF at 100°C using toluene as solvent for the dehydrogenation of 1,2,3,4‐tetrahydroquinoline (THQ) (Figure 2 and Table 1) [4, 30]. Under these conditions, we reached good catalytic conversions of >99% in 24 h by irradiating the reaction with a blue LED light of 460 nm (Entry 2). The employment of room temperature conditions led to diminished conversions of just 42% in 24 h (Entry 3).

FIGURE 2.

Reverse dehydrogenation/hydrogenation photocatalytic reaction for N‐Heterocycles and scope of the photocatalytic dehydrogenation.

TABLE 1.

Optimization of the catalytic reaction conditions for the dehydrogenation of THQ.

Entry a	Catalyst	Conditions	Light Source	Conversion (%)
1	5‐COF	Toluene, 100°C	White	35/88 (6/24 h)
2	5‐COF	Toluene, 100°C	Blue	67/99 (6/24 h)
3	5‐COF	Toluene, r.t.	Blue	0/42 (6/24 h)
4	5‐COF	MeOH, r.t.	Blue	18/99 (6/24 h)
5	2‐COF	MeOH, r.t.	Blue	8.5/40 (6/24 h)
6	5‐COF	MeOH, r.t.	Green	0/15 (6/24 h)
7	5‐COF	MeOH, r.t.	White	0/12.4 (6/24 h)
8	3‐COF	MeOH, r.t.	Blue	62.5/78 (6/8 h)
9	6‐COF	MeOH, r.t.	Blue	86.2/100 (6/8 h)
10	3‐COF	MeOH, r.t.	White	23 (8 h)
11	6‐COF	MeOH, r.t.	White	30 (8 h)

^a Reaction conditions: amine (0.02 mmol), catalyst (12 mg), solvent (3 mL), and room temperature. Yield determined by GC‐MS.

When the solvent was changed to a more environmentally friendly one, such as methanol, and the reaction was carried out at room temperature, promising results of >99% of conversion under blue light irradiation were obtained (Entry 4). As shown in Figure 3A, blue light selectively excites the semiconductor band, indicating that the observed catalytic activity primarily originates from the interaction between the COF matrix and the light. In contrast, the use of green light (532 nm) that exclusively excites the LSPR region in AuNPs yielded poor catalytic results (Entry 6) (∼15% in 24 h). It is likely that the photocatalytic activity under white LED light irradiation arises from a synergistic effect between semiconductor band excitation and the LSPR excitation of gold nanoparticles. However, the optimal catalytic conversion was achieved by irradiating the reaction mixture with blue LED light. Moreover, the presence of gold nanoparticles within the photocatalyst appears to be essential, as it led to ∼60% increase in conversion compared to the matrix 2‐COF (Entry 5).

FIGURE 3.

(A) Comparison of the solid‐state UV–vis absorption spectra for the materials 2‐COF, 5‐COF, and 6‐COF. Spectra of the visible, green (532 nm) and blue (470 nm) LED lights used in the photocatalytic experiments. (B) Recyclability test of the 5‐COF photocatalyst during 5 cycles of reaction. (C) Experimental results of mechanism research. Quinoline yields (%) of the photocatalytic dehydrogenation in the presence of different scavengers. Reaction conditions: 0.04 mmol of the quenching agent was added. (D) VB, CB, and band gap eV energies for the materials 2‐COF, 5‐COF, and 6‐COF. Active species in the reaction mechanism. (E) Proposed mechanism for the acceptor‐assisted dehydrogenation of THQ. (F) Proposed two‐step mechanism for the photocatalytic hydrogenation of quinoline: photocatalytic ammonia borane methanolysis, hydrogen production, and further plasmonic hydrogenation of quinoline.

In addition, the examination of the catalyst through TEM after a cycle of photocatalytic dehydrogenation reaction revealed that the catalyst remained unaffected, and gold nanoparticles do not show any evidence of self‐aggregation or lixiviation by the use of these reaction conditions (Figure S43). The use of the acid‐treated 3‐COF matrix and the gold deposited material 6‐COF resulted in good catalytic conversions after just 8 h of reaction. Under blue light irradiation, the plasmonic 6‐COF achieved a 22% higher conversion than the 3‐COF matrix (Entries 8 and 9). However, the irradiation with visible LED light did not lead to any further improvement in the catalytic conversion, indicating that the blue light remains the most effective wavelength for an improved conversion.

With these optimal reaction conditions, this catalytic protocol was then applied to the dehydrogenation of other N‐heterocycle amines. As shown in Figure 2, various N‐heterocycles with different hydrogen content were compatible with these photocatalytic conditions, and the expected dehydrogenated products were obtained in good yields.

As we have commented above, the reversible dehydrogenation/hydrogenation process of N‐heterocycles represents a promising strategy in the development of LOHC for hydrogen storage. Nevertheless, from a practical point of view, the use of two different catalysts for each reaction is not the best choice. Therefore, the development of a single catalyst able to efficiently catalyze both processes by only changing the reaction conditions remains an important challenge. As we have seen in the previous paragraphs, we have developed a versatile photocatalyst in the dehydrogenation of N‐heterocycles. Thus, 5‐COF enabled the dehydrogenation reaction under photocatalytic conditions of blue LED light irradiation, room temperature, air atmosphere, and methanol as solvent.

Very interestingly, 5‐COF also facilitated the reverse reaction when the reaction mixture was irradiated at 65°C with a white LED light. These photocatalytic conditions led to a ∼75% of conversion in 3 h of reaction (Table 2, Entry 1). However, the reaction under dark conditions only reached a hydrogenation conversion of just 10% after 3 h of reaction (Entry 3). A series of experiments was conducted to elucidate the hydrogenation mechanism. Notably, when the reaction was carried out using only the 2‐COF matrix as the catalyst under white LED irradiation conditions, no formation of THQ was observed (Entry 2). This result highlights again the essential role of gold nanoparticles in enabling the photocatalytic reaction again.

TABLE 2.

Optimization of the catalytic reaction conditions for the hydrogenation of quinoline.

Entry a	Catalyst	Light source	Conversion (%) 3 h
1	5‐COF	White	75
2	2‐COF	White	0
3	5‐COF	Dark	46
4	5‐COF	Blue	36
5	5‐COF	Green	85
6	3‐COF	Green	0
7	6‐COF	Green	40
8	6‐COF	White	43

^a Reaction conditions: amine (0.02 mmol), 5‐COF (12 mg), NH₃BH₃ (0.3 mmol), methanol (3 mL), and ambient air atmosphere. Yield determined by GC‐MS.

It is worth highlighting that the irradiation with blue LED light (460 nm) resulted in low catalytic conversion (36%) (Entry 4). In contrast, catalytic activity significantly increased (85%) when green LED light (532 nm) was used (Entry 5). This suggests that blue LED light, which exclusively excites the semiconductor, yields poor catalytic conversions and rules out the possibility of attributing this process of hydrogenation to a photocatalytic excitation of the semiconductor. By contrast, green LED light, which selectively excites the LSPR of gold nanoparticles, results in enhanced catalytic hydrogenation conversions.

Considering the results of the different catalytic experiments, we envisaged that the photocatalyst selectively proceeds via different light‐mediated reaction mechanisms for the dehydrogenation and the hydrogenation of N‐heterocycles.

In the case of the dehydrogenation reaction and according to the achieved results regarding the irradiation with different wavelength sources, we propose a reaction mechanism in which both the semiconductor matrix and gold nanoparticles are involved. In order to determine the possible species involved and then the reaction mechanism, a series of quenching experiments was performed. We first determined the key reactive oxygen species in the dehydrogenation mechanism. The formation of these active species is limited to the band gap of the semiconductor, and their potentials must be considered (E_VB‐E_CB limits).

The valence band maximum (VBM) of the materials has been determined by usingXPS to assess their electronic structure. The VBM values were estimated by identifying the intersection point between the edges of the linear extrapolation from the valence band edge and the background signal. Figure 3C shows the E_VB‐E_CB energy limits of the COF matrix 2‐COF and the plasmonic 5‐COF, revealing that the direct formation of hydroxyl radicals (2.4 eV) is not feasible in any case because its potential is not contained within the E_VB‐E_CB range. Those that are included in the interval are the superoxide radical ·O₂ ⁻ and the holes h⁺.

In Figure 3B, we can observe the influence of superoxide radicals (·O₂ ⁻), hydroxyl radicals (·OH), and oxidative photogenerated holes (h⁺) on the photocatalytic dehydrogenation of N‐heterocycles. Thus, the reaction in the presence of t‐BuOH, a well‐known hydroxyl radical (·OH) quenching agent, proceeded without any influence on the quinoline generation, which indicates that hydroxyl radicals are not involved in the mechanism. In contrast, the catalytic performance of the proposed catalytic system was inhibited under inert atmosphere conditions (N₂), which are effective for quenching superoxide radicals (·O₂ ⁻). Triethanolamine (TEA) was employed as a scavenger for photogenerated holes (h⁺), and under these conditions, the catalytic activity was almost suppressed, resulting in minimal quinoline generation.

Considering the effects of the scavengers, it is evident that atmospheric oxygen is not merely a passive spectator, but it plays an active role in the reaction mechanism. These findings suggest the involvement of both superoxide radicals (·O₂ ⁻) and photogenerated holes (h⁺) in the photocatalytic dehydrogenation mechanism of N‐Heterocycles.

Based on the above experimental results, as well as in previous reported results, we propose a mechanism of O₂ acceptor‐assisted dehydrogenation of N‐Heterocycles (Figure 3D). The oxidative photogenerated holes (h⁺) and reductive photogenerated electrons (e⁻) are produced by the excitation of the COF matrix. The N‐heterocyclic substrate is oxidized by the generated holes of the semiconductor, leading to the formation of a cation radical. The noble metal cocatalyst subsequently serves to temporarily store electrons from the conduction band (CB) of the COF matrix, improving the charge‐carrier separation. Then, atmospheric O₂ is reduced by the photogenerated electrons transferred from the gold nanoparticle, producing the superoxide radical (·O₂ ⁻). The next step undergoes the oxidation reaction by the extraction of two protons to form the C═N bond. A second oxidation leads to the formation of a second hydrogen molecule, providing the complete final aromatic quinoline.

Considering the results of catalytic experiments conducted on the hydrogenation reaction of N‐heterocycles, and supported by literature references [30, 31], we propose the photocatalytic methanolysis of ammonia borane as a method for hydrogen production. In addition, the generated hydrogen is proposed to be used in a subsequent step for the hydrogenation of N‐Heterocycles. Irradiation with green LED light resulted in the highest catalytic conversions (Table 2, Entry 5, 85%) compared to other tested wavelengths of irradiation. In this sense, green LED light irradiation selectively excites the localized surface plasmon resonance of gold nanoparticles, leading to the generation of hot electrons. Figure 3E depicts the proposed mechanism of ammonia borane adsorption to the gold nanoparticle surface and the implication of four molecules of methanol in the complete dehydrogenation process, leading to the release of three hydrogen molecules and the production of ammonium tetramethoxyborate NH₄B(OCH₃)₄. During the reaction, CH₃OH molecules and NH₃BH₃ adsorb onto the active sites of gold nanoparticles. Subsequently, a methanol molecule attacks ammonia borane, resulting in the formation of H₃NH₂B‐CH₃O and the release of one molecule of hydrogen. The successive attack by three additional methanol molecules leads to the formation of the final byproduct, NH₄B(OCH₃)₄, and the release of two more hydrogen molecules.

Photocatalytic methanolysis would imply the formation of reaction intermediates. Thus, the activation of CH₃OH and NH₃BH₃ molecules onto the surface of gold nanoparticles would lead to the formation of the intermediates proposed in Equations ((1), (2), (3), (4), (5)).

$$\mathrm{Catalyst}+\mathrm{C}{\mathrm{H}}_3\mathrm{OH}\to \mathrm{Catalyst}-\mathrm{C}{\mathrm{H}}_3\mathrm{OH}$$ (1)

$$\mathrm{Catalyst}-\mathrm{C}{\mathrm{H}}_3\mathrm{OH}\to \mathrm{Catalyst}-{\mathrm{H}}^{+}+\mathrm{Catalyst}-\mathrm{OC}{{\mathrm{H}}_3}^{-}$$ (2)

$$\mathrm{Catalyst}+\mathrm{N}{\mathrm{H}}_3\mathrm{B}{\mathrm{H}}_3\to \mathrm{Catalyst}-{\mathrm{H}}_2\mathrm{BN}{{\mathrm{H}}_3}^{+}+\mathrm{Catalyst}-{\mathrm{H}}^{-}$$ (3)

$$\mathrm{Catalyst}-{\mathrm{H}}^{-}+\mathrm{Catalyst}-{\mathrm{H}}^{+}\to {\mathrm{H}}_2$$ (4)

$$\mathrm{Catalyst}-\mathrm{OC}{{\mathrm{H}}_3}^{-}+\mathrm{Catalyst}-{\mathrm{H}}_2\mathrm{BN}{{\mathrm{H}}_3}^{+}\to \mathrm{N}{\mathrm{H}}_3\mathrm{B}{\mathrm{H}}_2\mathrm{OC}{\mathrm{H}}_3$$ (5)

$$\mathrm{Catalyst}-{\mathrm{H}}^{+}+\mathrm{Catalyst}-{\mathrm{H}}^{+}+2{\mathrm{e}}^{-}\to {\mathrm{H}}_2$$ (6)

$$\mathrm{Catalyst}-{\mathrm{H}}^{-}+\mathrm{Catalyst}-{\mathrm{H}}^{-}+2{\mathrm{h}}^{-}\to {\mathrm{H}}_2$$ (7)

NMR measurements provided significant insights into the proposed mechanism. Hydrogen generation from NH₃BH₃ methanolysis was tested under green LED light irradiation at 65°C, and ¹¹B NMR measurements determined the presence of B(OCH₃)₄ ⁻ in the sample as a product of the methanolysis (Figure S46) [32]. Moreover, the reaction was checked under catalytic conditions in the hydrogenation reaction of quinoline. The complete methanolysis of ammonia borane was observed, and it is depicted in Figure S47.

Once H₂ is photocatalytically released in the methanolysis process, we propose that H₂ would be adsorbed and activated on the Au NPs surface, with this being a key step in the hydrogenation reaction. Thus, although in thermal‐induced hydrogenation, it is difficult to dissociate H₂ on Au NPs due to the fully occupied d‐orbitals of Au, this drawback can be overcome under photocatalytic conditions, such as green LED light irradiation. Previous studies have shown the possibility of transferring LSPR hot electrons to the H₂ LUMO antibonding orbital, favouring the H₂ splitting at the Au NP surface [33, 34]. The dissociated and highly reactive H atoms would produce the hydrogenation of quinoline, leading to the H₂‐charged LOHC THQ.

Moreover, the dehydrogenation of ammonia borane was also checked through a series of experiments (Figure S48). A two‐neck flask with a septum cap and a connection to a pressure sensor and a data transmitter was used. This setup allowed us to monitor the pressure increase during the NH₃BH₃ dehydrogenation process.

In order to consider the pressure increase arising from methanol when the temperature rises, we carried out the irradiation of 3 mL methanol solution in the absence (0.30 bar) and in the presence of the hydrogen precursor (0.76 bar) for 90 min. In the next experiments, we tested the dehydrogenation of NH₃BH₃ in the presence of the matrix (0.85 bar) and the photocatalyst (1.60 bar), and, finally, with green LED light irradiation (1.62 bar). We observed a pressure increase up to 0.84 bar when the dehydrogenation of NH₃BH₃ was performed in the presence of the photocatalyst 5‐COF. Moreover, we observed an increase of 0.75 bar when 5‐COF is used instead of the matrix 2‐COF, showing the need for the presence of gold nanoparticles for the dehydrogenation reaction. However, the irradiation with green LED light does not evidence any improvement in the H₂ release since the experiment under dark conditions (Table 2, Entry 3) also shows catalytic conversions.

2.3. Computational Calculations

In order to gain insights into the proposed mechanisms of dehydrogenation and hydrogenation of N‐Heterocycles, we have performed DFT and TDDFT calculations on a Au₁₃‐COF (1a) model system consisting of a thirteen‐gold atom cluster within the pore structure of the COF material made by the condensation of three amines (TAPB) and three aldehydes (BTAC). The optimized model reveals a pore size of 14–17 Å. Then, as shown in Figure 4, we have computed a Au₁₃‐COF‐O₂ model in which an O₂ molecule interacts with the surface of the Au₁₃ cluster (2a). We have chosen the most intense excitation, and we have computed the electron‐hole (e⁻ −h⁺) (green‐red) distributions to predict the electron density polarization in the excited state. We observe a clear shift of the electron density from the COF to the Au₁₃ cluster and, notably, an accumulation of electron density on the O₂ molecule. This analysis evidences the possible formation of a O₂ ⁻ radical on the surface of the metal nanoparticle due to a semiconductor excitation, supporting the proposed mechanism. On the other hand, we have optimized a Au₁₃‐COF‐H₂ (3a) model system with a hydrogen molecule activated on the surface of the Au₁₃ cluster to verify the mechanism of plasmonic hydrogenation of N‐Heterocycles.

FIGURE 4.

Representation of the electron‐hole density associated with the HOMO‐LUMO transitions for the 2 and 3a model systems.

TDDFT calculations on 3a reveal that the electronic polarization of the most intense excited state is dominated by the HOMO‐LUMO transition. In this excitation, electronic density shifts from the thirteen‐gold atom cluster (HOMO) toward the H₂ molecule (LUMO). This finding supports the proposed transfer of hot electrons from the gold nanoparticle to the LUMO orbital, providing the electronic density needed to facilitate H‐H cleavage.

3. Conclusion

In summary, we have developed an efficient hydrogen storage method using a new plasmonic semiconductor material AuNPs@COF, which exhibits excellent photocatalytic performance in the dehydrogenation/hydrogenation reaction of N‐heterocycles through a dual and selective blue/green light mediated process.

Previous heterogeneous catalysts dealt with the need for harsh reaction conditions or the use of hazardous oxidants for the acceptor‐assisted dehydrogenation reaction of N‐Heterocycles. In this study, we have taken a step forward, and we have shown a reusable photocatalyst which proceeds under very mild conditions: in ambient air atmosphere, without the need for external heating, oxidizing agents, or non‐green solvents. From a noteworthy perspective, the most notable achievement is that, through simple adjustment of the wavelength irradiation and thereby controlling the reaction mechanism, we have selectively achieved efficient photocatalytic activity in both dehydrogenation and hydrogenation reactions of N‐heterocycles, all using a single photocatalyst. This makes this catalytic system a promising basis for developing more efficient photocatalysts for the hydrogen storage problem.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: smll74034‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are openly available in the Supporting Information.