Introduction

Ammonia decomposition (2NH3 → N2 + 3H2) is promising for leveraging ammonia as a viable hydrogen carrier to effectively tackle the pressing challenges associated with hydrogen storage, transportation, and distribution because of its high hydrogen storage density (106 kg·m−3), ease of liquefaction (8–10 bars, 20 °C), and globally mature storage and transportation networks1,2,3,4. What is more, compared to other emerging liquid hydrogen carriers (e.g., CH3OH, HCOOH, and cyclohexane), the zero-carbon characteristic of ammonia decomposition delivers a perspective of constructing a carbon-neutral hydrogen system2,3. In general, thermo-catalysis is favorable for practically producing H2 from ammonia because of its high productivity and reliability5. It however encounters extensive thermal input and harsh reaction conditions6. In contrast, photocatalysis offers a relatively green and eco-friendly method for on-site hydrogen generation from ammonia decomposition by utilizing solar energy as the driving force despite the limit of the intermittent nature of sunlight. Nevertheless, for pure photocatalysis, since a broad range of the solar spectrum such as infrared light can’t be absorbed to produce charge carriers because of their low energy, most of the solar energy is not effectively transformed into chemical energy7. Most recently, photo-thermal coupling has appeared as an innovative strategy for converting solar energy into chemical fuels by simultaneous utilization of charge carriers and heat8,9,10,11,12. In this case, ultraviolet photons with high energy enable the generation of charge carriers while visible- and infrared- light are highly favorable for heating the localized surface of the catalyst owing to their significant photo-thermal effect. Of note, the synergy of heat and charge carriers is highly promising for enabling a substantial reduction in the activation energy barrier of chemical process according to previous reports8,13. Such important discoveries shield light on breaking the bottleneck of sunlight-driven ammonia decomposition toward H2. The construction of a rational photo-thermal-coupled catalytic architecture is at the core yet still remains an extraordinary challenge.

Gallium nitride nanowires vertically aligned on silicon wafer (GaN NWs/Si) have emerged as a next-generation semiconductor platform for solar fuels generation from photocatalytic water splitting14,15, CO2 reduction16,17,18, as well as methanol reforming19. It can be attributed to the following reasons: (i) At first, the optical properties of the GaN NWs/Si hybrid allow for a broad range absorption of the solar spectrum with alleviated photon scattering; (ii) Well-defined one-dimensional (1D) nanostructures facilitate the achievement of high localized surface temperature in a nano-confined environment under concentrated light illumination with shortened charge diffusion pathway, thus favoring surface redox reactions; (iii) The tunable surfaces (nitrogen- or metal-terminated) offer a flexible scaffold for manipulating the behavior of various chemical species; (iv) More importantly, earth-abundant silicon (Si), owing to the merits of highlight-absorption, chemical stability, thermal conductivity, has recently become a promising photo-thermal catalytic building block, especially if illuminated by concentrated light20. Together, the hybrid of GaN nanowire and silicon may be a suited candidate for assembling a rational photo-thermal architecture for sunlight-driven ammonia decomposition toward H2, especially when coupling with an appropriate co-catalyst.

Ruthenium (Ru) is an extensively studied catalyst for ammonia decomposition with high activity because of its suited binding with the intermediate species (neither too strongly nor too weakly)21,22,23. Hence, in this study, Ru nanoparticles (Ru NPs) were coupled with GaN NWs/Si to assemble a architecture for photo-thermal-coupling NH3 decomposition toward H2. The synergy between charge carriers and photo-induced thermal energy of Ru NPs/GaN NWs/Si enabled the achievement of a light-induced substantial reduction in apparent activation energy (Ea) for NH3 decomposition, significantly decreasing from 1.08 eV to 0.22 eV. What is more, the temperature-dependent photoluminescence spectroscopy characterization showed that the recombination of charge carriers was also significantly lowered by the photo-thermal-coupled effect. As a result, a high hydrogen evolution rate of 3.98 mmol·cm−2·h−1 is attained under 5 W·cm−2, which is nearly three orders of magnitude higher than that (0.004 mmol·cm−2·h−1) of the pure thermo-catalytic activity measured under the same temperature without light irradiation. Furthermore, because of the high efficiency of active sites and the unique surface properties, this architecture exhibits a high turnover number of > 3,400,750 mol H2 per mole ruthenium over a long-term stability test of 400 h without measurable activity degradation. The operando spectroscopic characterizations and computational investigations show that Ru nanoparticles work in synergy with GaN to effectively stabilize the *NH2 intermediate from NH3 decomposition by photo-excited holes over the catalytic interface, thus switching the potential-determining step from *NH2 + *H → *NH2 to *NH + *H → *NH with a reduced Gibbs free energy. In situ irradiated X-ray photoelectron spectroscopy (ISI-XPS) and the local density of states (LDOS) analysis suggest that the photo-induced electron migrates from GaN to Ru for reducing the proton toward *H (H+ + e− → *H), further facilitating H2 generation. Outdoor tests using a low-cost resin lens validated the viability of such a multifunctional architecture for maximally utilizing natural solar energy to realize on-site H2 generation from ammonia decomposition.

Results

Assembly and characterization of Ru-decorated GaN NWs vertically aligned onto silicon

According to our previous work, 1D GaN NWs were controllably grown on a 4-inch silicon wafer by employing plasma-assisted molecule beam epitaxy (MBE) technology under N-rich conditions24,25. Ru NPs were subsequently immobilized onto GaN NWs/Si via a simple photo-deposition method (Supplementary Fig. S1 and Experimental Section). As characterized by scanning electron microscope (SEM), the epitaxial GaN NWs decorated with Ru NPs were vertically aligned on the silicon wafer, featuring a length of about 950 nm with an averaged diameter of about 50 nm (Fig. 1A). The well-defined 1D nanostructure renders GaN NWs with high surface area and short charge diffusion length, which is highly favorable for spatially decoupling photons absorption, charges separation, as well as surface chemical reactions. Compared to the pristine GaN NWs (Supplementary Fig. S2), the overall morphology of nanowires is not obviously varied with the decoration of Ru NPs. The high-angle annular dark-field scanning transmission microscope (HAADF-STEM) image illustrates that the Ru NPs with the size of about 19 nm was randomly distributed on GaN NWs surface (Fig. 1B). The lattice spacing of 0.26 nm is assigned to the (002) plane of GaN, in line with XRD patterns (Supplementary Fig. S3), suggesting the c-axis growth direction of the nanowires26,27. The lattice spacing of 0.23 nm is attributed to the (100) plane of metallic Ru28. However, no typical peaks of Ru NPs were observed in XRD patterns, because of its low content (0.11 μmol·cm−2), as characterized by inductively coupled plasma-atomic emission spectroscopy (ICP-OES). The energy dispersive X-ray spectroscopy (EDS) mapping further validated that Ru NPs were successfully assembled on the GaN NWs/Si platform (Fig. 1B).

Fig. 1: Structural and physical properties of GaN NWs decorated with/without Ru nanoparticles.
figure 1

A 45°-tilted SEM image of Ru NPs-decorated GaN NWs/Si. HR: H2 activity; TOF: turnover frequency; TON: turnover numbers. B HAADF-STEM and EDS mapping images of Ru NPs-decorated GaN NWs. High-resolution XPS spectra of (C) Ga 2p, (D) N 1 s, (E) Ru 3d. F Bader charge analysis between Ru and GaN. The yellow and cyan regions indicate the gain and loss of electronic charge respectively, with an isosurface of 0.008 e/Å3. Ga, blue; N, yellow; and Ru, salmon. Source data are provided as a Source Data file.

Full size image

X-ray photoelectron spectroscopy (XPS) was carried out to investigate the chemical components of Ru NPs/GaN NWs/Si and the interaction between Ru NPs and GaN NWs (Supplementary Fig. S4). The characteristic peaks of Ga 3d and N 1 s are located at ~20 and ~397 eV, respectively (Figs. 1C, D)29. The high-resolution X-ray photoelectron spectroscopy (HR-XPS) spectra of Ru 3d in Fig. 1E confirmed the presence of metallic Ru species (281.3 eV). The peak located at 282.0 eV was assigned to the Ru-N bond. Further, the observable binding energy shifts of ~0.3 eV and ~0.1 eV in the HR-XPS spectra of Ga 2p and N 1s compared to that of pristine GaN are indicative of the electron redistribution between Ru NPs and GaN NWs. Bader charge analysis was conducted to confirm the above results. As shown in Fig. 1F, a discernible charge transfer is observed from Ru to GaN on the optimized geometry of Ru/GaN with a calculated value of 0.535e. As studied below, such an electronic interaction between Ru and GaN not only stabilizes the nanoparticles against agglomeration but also provides electronic transmission channels, thus catalytically facilitating the ammonia decomposition.

As one key component of the photo-thermal-coupling architecture, GaN can generate energetic carriers with sufficient redox potential for ammonia decomposition if excited by appropriate photons as characterized by photoluminescence (PL) spectroscopy (Fig. 2A). The incorporation of Ru species can inhibit the recombination of photoinduced electron-hole pairs30. Time-resolution PL (TR-PL) spectroscopy further revealed that upon decoration with Ru species (Fig. 2B), the average charge lifetime (τavr) of GaN decreased from 2.02 ns to 1.81 ns, suggesting accelerated charge transfer30,31. The above results demonstrate that the decoration of Ru NPs onto GaN NWs is favorable for mediating the charge behavior to catalyze ammonia decomposition. The photo-thermal properties of the architectures were further studied using an infrared thermograph. As shown in Fig. 2C, the surface temperature of Ru NPs/GaN NWs/Si can reach 409.7 °C under 5 W·cm−2, which is relatively higher than that of Ru NPs/Si (Supplementary Fig. S5A). Herein, GaN NWs contribute to the photothermal effect by alleviating the photon scattering due to the well-defined 1D nanostructure32,33. In contrast, when the Si substrate was replaced by sapphire (Supplementary Fig. S5B), the measured temperature dropped to 284.5 °C without varying any other conditions (Fig. 2C). The results show that Si is indeed an ideal platform for assembling a photo-thermal architecture. The well-defined 1D nanostructure of GaN is beneficial to reducing the Rayleigh scattering, thus further increasing the surface temperature of the architecture without changing the illumination. Such a hierarchical architecture does not only provide energetic charge carriers but also enables high localized temperature. It thus holds a grand promise for photo-thermal-coupling ammonia decomposition toward H2, which will be elaborately studied next.

Fig. 2: Optical and physical properties of GaN NWs decorated with/without Ru nanoparticles.
figure 2

A PL spectroscopy of GaN NWs/Si decorated with/without Ru NPs recorded with pulse excitation of 80 MHz at a wavelength of 325 nm. B TRPL spectroscopy of GaN NWs decorated with/without Ru NPs by a time-correlated single photon counting technique. τavr: average charge lifetime. C Infrared thermal images of Ru NPs/GaN NWs/Si, GaN NWs/Si, Ru NPs/Si, and Ru NPs/GaN TF/Sapphire surface under 5 W·cm−2 concentrated light illumination. Ru NPs loaded onto commercial GaN thin films on sapphire substrate is defined as Ru NPs/GaN TF/Sapphire. Source data are provided as a Source Data file.

Full size image

Photo-thermal-coupling performance of ammonia decomposition toward H2

The performance of Ru NPs/GaN NWs/Si for ammonia decomposition was tested in a sealed quartz chamber under atmospheric argon. A 300 W Xenon lamp equipped with a quartz lens was used as the light source. A commercial ammonia aqueous solution was used as the feedstock as water is a good medium for NH3 storage and transportation under ambient conditions. Under concentrated light illumination, the aqueous ammonia solution was facilely evaporated due to the strong photo-thermal effect. The critical role of each component of the ternary Ru NPs/GaN NWs/Si architecture was first investigated (Fig. 3A). In the absence of Ru species, the bare GaN NWs/Si platform is almost inactive for light-driven ammonia decomposition although charge carriers and photo-induced heat can be provided under concentrated light illumination at 5 W‧cm−2 (Fig. 3A), suggesting that Ru species are essential for the reaction by serving as active sites. The hydrogen evolution rate was significantly enhanced by the immobilization of Ru NPs (Supplementary Figs. S7 and S8). A maximum value of 1.77 mmol·cm−2·h−1 was achieved at a Ru loading of 0.11 μmol·cm−2, corresponding to an optimal turnover frequency of 16091 h−1, which is an intrinsic metric for evaluating the activity of the catalytic sites (Supplementary Figs. S6 and S7). Herein, the average diameter of Ru NPs was measured to be ~19 nm (Supplementary Fig. S8). A reduced H2 activity of 0.98 mmol·cm−2·h−1 was observed when Ru loading increased up to 0.22 μmol·cm−2 (Supplementary Fig. S6). It is primarily due to the undesired agglomeration of Ru NPs with an average diameter of > 86 nm (Supplementary Fig. S8)34. Despite the high surface temperature arising from the significant photo-thermal effect of silicon, there is almost no hydrogen production over Ru/Si, suggesting the critical role of GaN NWs in offering energetic charge carriers (Supplementary Fig. S9).

Fig. 3: On-site H2 production from photo-thermal-coupling ammonia decomposition.
figure 3

A H2 evolution rate over Ru NPs/GaN NWs/Si, GaN NWs/Si, and Ru NPs/Si illuminated by a 300 W-xenon lamp under 5 W·cm−2. B H2 activity over Ru NPs/GaN NWs/Si under dark equipped with an external heating system and concentrated light-illuminating conditions without external heating. The architecture was maintained at the same temperature for better comparison of the performance between photo-thermal-coupling catalysis and pure thermocatalysis. C Arrhenius plots for H2 evolution rate under dark and light conditions over Ru NPs/GaN NWs/Si. D H2 evolution rate over Ru NPs/GaN NWs/Si under 5 W·cm−2 with/without cooling. E H2 evolution rate over Ru NPs/GaN NWs/Si under light irradiation in different spectral ranges (full spectra, ultraviolet, visible, and infrared) with/without external heating source. The temperature of the external heating source is set to 280 °C. F Durability test over Ru NPs/GaN NWs under 4 W·cm−2. Sample, ~0.5 cm2, ~0.36 mg·cm−2; atmospheric argon; 300 W Xenon lamp. Source data are provided as a Source Data file.

Full size image

The light intensity-dependent H2 activity was further measured to study the photo-thermal synergy. As illustrated in Fig. 3B, the H2 evolution rate showed an increasing trend as the light intensity increased, and reached 3.98 mmol·cm−2·h−1 at 5 W·cm−2 with an appreciable turnover frequency (TOF) of 36,182 h−1 without any other energy inputs (Supplementary Fig. S10). Such a great activity is nearly 1-2 orders of magnitude higher than state-of-the-art thermal or photocatalytic ammonia decomposition systems (Table S1). To decouple the photo-excited carriers and photo-induced heat contribution, the catalytic properties of Ru NPs/GaN NWs/Si were both photo-catalytically and thermal-catalytically measured. First of all, light-induced heating was in operando recorded with an infrared thermograph as a function of light intensity (Supplementary Fig. S11). The surface temperature of the as-designed architecture showed an increasing trend as the illumination intensity increased, varying from 274.1 °C at 3.0 W·cm−2 at up to 409.7 °C at 5.0 W·cm−2. Shockingly, the photo-thermal-coupling activity is nearly 1000 times higher than that of pure thermo-catalysis under the same temperatures over the temperature range tested (Fig. 3B). Herein, the photo-thermal-coupling activity was obtained by the only input of concentrated light while the thermo-catalysis was powered by external heating. Of note, as calculated by the Arrhenius equation, the apparent activation energy (Ea) of NH3 decomposition over Ru NPs/GaN NWs/Si is significantly reduced from 1.08 eV to 0.22 eV upon light illumination (Fig. 3C). The above results suggested that the reaction proceeded via photocatalysis, which can be further promoted by the photoinduced thermal effect resulting from the concentrated visible- and infrared light. The influence of light wavelength on the reaction was also studied. As shown in Supplementary Fig. S12, under a measured reaction temperature of 270 °C set by external heating, the introduction of 275 nm under monochromatic light illumination of 29.8 mW·cm−2 in the system results in an enhanced H2 evolution rate of 1.46 μmol·cm−2·h−1 with a high apparent quantum efficiency (AQE) of 1.19% over Ru NPs/GaN NWs/Si. It is much higher than that of the 535 nm under monochromatic light illumination of 285.1 mW·cm−2. Therefore, it is reasonable to speculate that the high-energy photons of <365 nm that can produce energetic charge carriers are critical for superior activity. However, by employing an external cooling system to alleviate the photo-thermal effect of the reaction system (Fig. S13), the measured catalytic activity sharply dropped from 3.98 to 0.06 mmol·cm−2·h−1 (Fig. 3D). Thus, the photo-induced heat also plays an unignored role for promoting ammonia decomposition. Based on the results above, it is rationally hypothesized that for Ru NPs/GaN NWs/Si, upon concentrated sunlight illumination, high-energy ultraviolet is able to produce energetic charge carriers by exciting GaN NWs. Meanwhile, the remaining visible- and infrared light absorbed by the silicon substrate, accounting for a large proposition of the solar energy (~93%), contributed to heating the architecture and offered high-localized surface reaction temperature for ammonia decomposition. Charge carriers and heat work in synergy to promote the reaction by inducing a substantial reduction of activation barriers of ammonia decomposition and increasing the reaction temperature. Such a hypothesis was further validated by wavelength control experiments. The thermal contribution from various regions of the solar spectrum was first studied. As shown in Supplementary Fig. S14, the surface temperature of the architecture was measured by an infrared thermograph. It was found that the full-arc spectrum at 4 W·cm−2 can increase the architecture temperature up to 342.7 °C. By contrast, the surface temperature of the architecture dropped to 100.5 °C if an ultraviolet light filter was employed. Meanwhile, visible and infrared light can heat the architecture to 161.7 and 170.4 °C, respectively. In the absence of an external heat source, the hydrogen activity of Ru NPs/GaN NPs illuminated by ultraviolet light was only about 20% of the full spectrum (Fig. 3E). The incorporation of the infrared and visible light did not show activity for ammonia decomposition toward H2. Notably, once an external heat source was applied for heating the reaction system (280 °C), the introduction of ultraviolet light can greatly improve the activity of H2 up to 1.56 mmol·cm−2·h−1, which is nearly 90% of the activity obtained under the full spectrum. These findings reveal the synergy between charge carriers induced by ultraviolet light and photo-induced heat as a result of visible light and infrared light played a vital role in the performance of Ru NPs/GaN NWs/Si.

As a key metric for practical application, the stability of the architecture was also examined (Fig. 3F). A high turnover number (TON) of 3,400,075 moles of hydrogen per mole of Ru NPs was achieved after 400 h of light illumination. Of note, an H2:N2 stoichiometric ratio of 3:1 was observed, suggesting the absence of the undesired side reaction. As characterized by XPS and TEM, there were no significant morphology and chemical component variations for this architecture after the stability test (Supplementary Fig. S15). The aforementioned observations are indicative of the good stability.

Temperature-dependent photoluminescence (TD-PL) spectroscopy characterizations were conducted to study the photo-thermal effect on the reaction. It is observed that by increasing the measured temperature from 50 °C to 450 °C, the recombination of photoexcited e−/h+ pairs was evidently inhibited (Fig. 4A)35,36. Hence, the photo-thermal effect does not only promote the reaction by increasing the reaction temperature but also facilitates the separation of photoexcited e−/h+ pairs, which is very beneficial for the reaction. Operando diffuse reflectance infrared Fourier-transform spectroscopy (DRIFTS) was utilized to monitor the key intermediates of ammonia decomposition (Fig. 4B). It is discovered that the peaks at around ~948 cm−1 and ~3255 cm−1 are assigned to the adsorbed *NH3 species1,37,38. The peak intensity at around 1369 cm−1 arising from the *NH2 intermediate increased with the irradiation time39,40, which is well matched with the EPR results (Supplementary Fig. S16). It can be attributed to the continued deprotonation of *NH3 toward *NH2 by the photo-excited holes (NH3 + h+ → NH2 + H+), which had not yet reached steady-state condition even 25 min after since the start because of the saturation of ammonia in the cell. Of note, compared with pure thermal catalysis (Supplementary Fig. S17), the accumulation rate of *NH2 on the surface of the catalyst during the photo-thermal-coupling process is significantly faster (Supplementary Fig. S18), further suggesting that the photo-thermal effect greatly promotes the activation and deprotonation of the reactant molecules. The above results demonstrate the viability of maximally utilizing solar energy to drive NH3 decomposition toward H2 by effectively utilizing photons at various regions (Fig. 4C).

Fig. 4: TD-PL spectroscopy and operando DRIFT spectra.
figure 4

A TD-PL spectroscopy of Ru NPs/GaN NWs/Si. B Operando DRIFT spectra of ammonia decomposition over Ru NPs/GaN NWs/Si under light illumination of 4 W·cm−2. C Schematic diagram of the synergy between charge carriers and photo-induced heat for promoting ammonia decomposition over Ru NPs/GaN NWs/Si. Source data are provided as a Source Data file.

Full size image

Possible mechanism of ammonia decomposition

The electronic state of the catalytic interface can affect the reaction significantly. Thereby, the redistribution of photo-induced electrons at the interface between GaN and Ru was characterized by in situ irradiated XPS (ISI-XPS) (Fig. 5A). It is clear that under light illumination, the binding energy of Ru 3d illustrated a slightly negative shift. Conversely, a marked positive shift in the HR-XPS spectra of Ga 2p and N 1s was observed (Supplementary Fig. S19). Herein, these observations validated that Ru NPs behave as effective electron sinks as reported41,42. The electron redistribution from GaN to Ru under light illumination is critical for superior activity41. NH3 temperature-programmed desorption (NH3-TPD) characterizations were conducted to investigate the adsorption behavior of ammonia molecules. When GaN was decorated with Ru NPs, the desorption signal of NH3 was evidently enhanced (Fig. 5B). It is indicative that the incorporation of Ru species can promote the adsorption of NH3 molecules onto the GaN NWs surface, thus favoring hydrogen evolution from ammonia decomposition. Operando DRIFT reveals the evolution track of ammonia at the molecular level (Fig. 4B and Supplementary Fig. S20). Interestingly, the accumulation rate of *NH2 intermediates on the pristine GaN is much lower than that of Ru NPs/GaN NWs/Si (Fig. 5C), suggesting that the absence of Ru stabilizes the *NH2 for further dehydrogenation, which will be discussed by DFT calculations next. Isotope experiments were conducted in NH3/D2O medium to elucidate the origin of hydrogen. It is found that H2 almost originated from NH3 (Supplementary Fig. S21). Moreover, there was not an evident variation in reaction kinetics between NH3/H2O and NH3/D2O, which is indicative of no kinetic isotope effect (KIE) (Supplementary Fig. S22). The ratio of H2 and N2 was detected online with the use of a reaction chamber linking the gas chromatography (GC) to prevent the interference of N2 in the air. GC analysis of the product suggested a stoichiometric H2/N2 ratio of 1/3 (Supplementary Fig. S23), with no other byproducts aside from H2 and N2. Based on the above results, it is reasonably speculated that water does not actively participate in the reaction; rather, it serves primarily as an ideal storage and transportation medium of NH3 under ambient conditions, which is important for practical applications.

Fig. 5: Spectral characterization and DFT calculation.
figure 5

A ISI-XPS spectra of Ru 3d with or without light illumination. B NH3-TPD spectra of bare GaN NWs and Ru NPs/GaN NWs. C Calculated slope change of the peak intensity of *NH2 based on operando DRIFT spectra in Fig. 4B and Fig. S20. D LDOS for pristine GaN and Ru/GaN, respectively. The black dashed line indicates the position of the Fermi level. E The calculated free energy ∆G diagrams for NH3 decomposition on GaN and Ru/GaN. The values in the figures indicate the energy difference for the potential-limiting step of the reaction. F Schematic diagram of ammonia decomposition process over the Ru/GaN interface. Source data are provided as a Source Data file.

Full size image

DFT calculations were performed to gain insights into the reaction mechanism at the molecular level. Firstly, three optimal surface models of Ru (0001), GaN (10\(\bar{1}\)0), and Ru/GaN were constructed (Supplementary Fig. S24). The electronic properties were analyzed by plotting the local density of states (LDOS) for GaN and Ru (Fig. 5D). Pristine GaN with a large bandgap (3.4 eV) is generally not conducive to electron transfer in the reaction43. However, after decorating with Ru NPs, the metal states appeared near the Fermi level of GaN, enabling a high conductivity of this architecture. More importantly, the strong interaction between GaN and Ru arising from the electron redistribution formed a new state around the Fermi level for Ga and N atoms at the interface, thus facilitating the photoexcited electron transfer from GaN to Ru and its subsequent participation in the ammonia decomposition. This computational result was in well consistency with the ISI-XPS characterizations (Fig. 5A). Furthermore, the ammonia decomposition pathway on the three surfaces was calculated and the results are shown in the Gibbs free energy diagram (Fig. 5E). Upon the initial adsorption stage, the NH3 molecule was strongly attached to the GaN surface, with the N atom binding to the Ga atom, and then a H atom was captured by the near N atom in GaN (Supplementary Fig. S25). Of note, the NH3 adsorption energy over the GaN is slightly reduced from −0.98 eV to −1.15 eV after the immobilization of Ru NPs (Supplementary Fig. S26 and Table S2), indicating an enhanced adsorption capacity of Ru/GaN for NH3 molecule. It is well matched with the NH3-TPD measurements (Fig. 5B). Following the step of *NH3 → *NH2, *NH2 intermediate was stabilized onto the catalytic interface for the subsequent decomposition process. Particularly, in the case of GaN, we observed that the potential-determining step (PDS) exhibited high endergonicity with a significant energy difference of 1.78 eV. In contrast, the presence of a Ru cluster on the GaN surface stabilized the *NH2 intermediates and thus significantly reduced the barriers and shifted the PDS to the desorption of the second H atom, resulting in a smaller free energy change of only 0.58 eV (Fig. 5E and Supplementary Fig. S27). Based on the ISI-XPS and LDOS results (Fig. 5A and Fig. 5D), the transfer of photoinduced electrons enables the electron-rich Ru sites to a lower ∆GH value, which is a favor for the accumulation of *H for the final formation of H2 (Fig. 5F). What’s more, the adsorption energy of water molecule is much higher than that of NH3 molecule over the Ru NPs-decorated GaN surface (Supplementary Figs. S26, S28, and S29), and significantly increase the energy barrier for H2O dissociation on both GaN and Ru/GaN surfaces compared to NH3 decomposition (Supplementary Fig. S30). Therefore, in this study, water primarily functions as a medium of ammonia rather than a hydrogen source, in line with the isotopic experimental results.

On-site hydrogen evolution under natural concentrated sunlight illumination

To assess the practical viability, the as-assembled architecture was tested under natural concentrated sunlight. The homemade experimental setup mainly consisted of a Fresnel lens, support frame, and quartz reaction chamber (Fig. 6A). The natural sunlight was concentrated by a cheap and simple lens, which is beneficial for improving the performance and reducing the usage of the catalyst. As depicted in Fig. 6B and Supplementary Fig. S31, the H2 evolution rate is directly related to the natural sunlight illumination conditions, which is varied from 0.07 mmol·cm−2·h−1 to 0.17 mmol·cm−2·h−1, as a result of the varied light intensity. This observation suggests that the H2 rate over the architecture is highly dependent on the light intensity. What’s more, the solar-to-hydrogen (STH) efficiency was also calculated, reaching an optimal value of 5% from 13:00 pm to 14:00 pm (Supplementary Fig. S32). After 14 h of operation, San illustrious TON of 51,689 moles H2 per moles Ru NPs was achieved under naturally concentrated sunlight conditions (Fig. 6C). Such an outdoor test revealed the viability of utilizing natural sunlight for hydrogen production from ammonia. To step forward to practical application, as the priority, the fabrication cost of the architecture needs to be significantly reduced. Meanwhile, long-term stability is required for the architecture.

Fig. 6: Outdoor experiments.
figure 6

A Image of outdoor test setup equipped with Ru/GaN NWs/Si. B Activity and (C) TON of Ru NPs/GaN NWs for ammonia decomposition under concentrated natural sunlight without external heating, the inset in (B) is the digital picture of Ru NPs/GaN NWs/Si. Source data are provided as a Source Data file.

Full size image

Discussion

In summary, Ru NPs/GaN NWs/Si has demonstrated a virtual success in photo-thermal-coupling ammonia decomposition toward H2 under concentrated sunlight illumination. The synergy of photo-excited charge carriers and photoinduced heat enabled a substantial reduction in the activation energy barrier for NH3 decomposition. The architecture of Ru NPs/GaN NWs/Si showed a measurable H2 rate of 3.98 mmol·cm−2·h−1 under 5 W·cm−2 by an increasing factor of nearly 1000 compared with pure thermocatalysis under the same tested temperature without light illumination. Meanwhile, the architecture exhibited a considerable TON of 3,400,750 mol H2 per mol Ru without activity variation after 400 h of uninterrupted light illumination. Experimental, spectroscopic, and computational calculations disclosed the origin of the measurable performance of the architecture. Outdoor tests further confirmed the viability of this architecture for maximally utilizing the entire solar energy to produce hydrogen from ammonia.

Methods

MBE growth of GaN NWs/Si. MBE growth of GaN NWs/Si

GaN NWs were epitaxially grown on 4-inch silicon (111) wafer using a SVTA molecular beam epitaxy (MBE) system, equipped with a dual filament Knudsen cell for gallium source (Ga purity 99.99999%), a standard Knudsen cell for magnesium source (Mg purity 99.9999%), and a Veeco Unibulb radio frequency (RF) nitrogen plasma source (N2 purity 99.9999%). The nitrogen plasma was operated with a constant N2 gas flow of 1.0 sccm and RF power of 400 W, corresponding to a growth rate of 300 nm/h for GaN NWs. All GaN NWs were grown under N-rich growth conditions with a III/V ratio of 0.2. Before growth, the silicon wafer was firstly degassed in the growth chamber at 900 °C for 30 min. Then, 60 s surface nitridation was performed at 800 °C to form a few monolayers of SixNy, which facilitates the nucleation of the following GaN base. Subsequently, unintentionally doped n-type GaN NWs were grown at 700 °C, followed by the growth of Mg-doped p-GaN NWs at 600 °C. During the growth of p-GaN NWs, Mg cell temperature was gradually varied from 230 °C to 270 °C, which is an efficient approach for controlling the diameter of GaN NWs. The resulting GaN NWs exhibit a p-n double-layer structure. The Mg element was successfully detected in the epitaxial GaN NWs by EDS mapping characterization (Supplementary Fig. S33). The consumption of Ga, Mg, and N2 is less than 200 mg, 0.6 mg, and 200 mL in each of these GaN nanowire samples growth.

Synthesis of Ru NPs/GaN NWs/Si

The deposition of Ru NPs was conducted in a sealed Pyrex chamber equipped with a quartz lid through a simple photo-deposition process. The GaN NWs supported by silicon wafer (~0.5 cm2) was fixed at the bottom of the chamber, and then 50 mL methanol aqueous solution with a methanol/H2O (methanol, Shanghai Titan Technology Co., Ltd) volume ratio of 1/5, and a specified amount (5, 10, 20 μL…) of 0.2 M Ru precursor solution (RuCl3·H2O, Shanghai Bide Pharmaceutical Technology Co., Ltd.) were added into the chamber. The chamber was subsequently evacuated and filled with argon to atmospheric pressure, followed by a full-arc 300 W Xe lamp irradiated for 30 min. At last, the sample was rinsed with distilled water and dried by compressed air. Notably, the amount of Ru was controlled by adjusting the content of Ru precursor solution (5, 10, 20 μL…), and the loading amount of Ru over GaN NWs/Si was precisely determined by ICP-OES, which is reported in Supplementary Fig. S6.

Characterization

X-ray diffraction (XRD) patterns were measured by a Bruker D8 Advance diffractometer (with Cu Kα, at 60 kV and 80 mA). Induced coupled plasma (ICP) characterization was measured to determine the loaded amount of Ru NPs by employing an AGILENT ICP-OES 730. X-ray photoelectron spectroscopy (XPS) was performed using ESCALAB 250xi non-monochromatic Al anodes, and the C 1 s peak at 284.8 eV was used as the internal standard for calibration. Scanning electron microscopy (SEM) images were captured with a Quattro ESEM (Thermo Fisher). PL and TR-PL spectroscopies were measured using an FLS980 (Edinburgh Instruments). TD-PL were measured using an FLS980 (Edinburgh Instruments) equipped with ADVANCED RESEARCH SYSTEMS (ARS) LT4 series standard continuous flow thermostats. Transmission electron microscopy (TEM) images were captured by employing a JEOL 2100 F microscope. HAADF-STEM images were obtained by a Thermo Fisher Scientific Talos F200X S/TEM with a Super-X EDS detector and operated at 200 kV. NH3-TPD characterization was conducted on an Autosorb-iQ-C chemisorption analyzer (Quantachrome, USA). Isotope characterizations were carried out using a TRACE 1310 gas chromatograph equipped with a 253Plus mass spectrometry module. In situ irradiated XPS was analyzed by ESCALAB 250xi non-monochromatic Al anodes using a 300 W Xe lamp as an irradiation source. In situ EPR measurement was conducted on a Bruker A300 (Germany) using 5,5-Dimethyl-1-pyrroline-N-oxide (DMPO) as a spin trap. DRIFT spectra were collected by using a Frontier FT-IR Spectrometer, PerkinElmer, equipped with an MCT detector and a 10-cm Demountable Gas Cell. In the process of the DRIFT characterization, Ru NPs/GaN NWs/Si were fixed in the Cell. The ammonia aqueous solution was injected into the Cell through a bubbling system. The background spectrum of the system was recorded when the reactant reached adsorption-desorption equilibrium on the catalyst surface. Of note, ammonia was in a saturated state in the cell at this time. The Cell was subsequently closed by cutting off the injection of the reactant. Then the photocatalyst surface was illuminated by a 300 W Xenon lamp with a light intensity of 5 W·cm−2. The DRIFT data was recorded once every three minutes.

Performance evaluation

In the lab, hydrogen production from ammonia aqueous solution was evaluated in a 0.44 L homemade sealed quartz chamber. Prior to being placed at the bottom of the chamber, the photocatalyst (~0.5 cm−2, ~0.36 mg cm−2) was thoroughly rinsed with distilled water and dried by compressed air. Subsequently, the chamber was evacuated and filled with atmospheric argon. 1 mL of ammonia aqueous solution (AR, 28%, Shanghai Titan Technology Co., Ltd.) was injected into the chamber. A 300 W Xe lamp equipped with a quartz lens was used as the light source and the illumination time was 30 min if not specifically noted. The surface temperature of the photocatalyst was recorded using infrared thermography (FOTRFIC 315, Shanghai Thermal Imaging Technology Co., Ltd.). The emissivity calibration procedure can be seen below: (i) A specific area of the catalysts was uniformly sprayed with acrylic resin. (ii) After waiting for 3 min, the catalyst was illuminated by a 300 W Xe lamp with a quartz lens. (iii) The emissivity of the infrared thermograph was set to 0.96, and the temperature of the coating position was recorded. (iv) An infrared thermograph was used to monitor the temperature of the uncoated position and adjust the emissivity ε until the recorded temperature was consistent with that of the coating position. The ε is, namely, the emissivity of the catalysts. The emissivity of all samples (Ru NPs/GaN NWs/Si, GaN NWs/Si, Ru NPs/Si, and Ru NPs/GaN TF/Sapphire) was roughly estimated to be around 0.8 according to the above procedure. For comparison, the thermo-catalytic experiments were carried out in an argon-filled tube furnace without light illumination. The wavelength-dependence experiments were measured using a 0.042 L stainless-steel cell equipped with a top sapphire window under different monochromatic light (275 nm and 535 nm) irradiation. During the reaction process, an external heating system was employed to maintain the stainless-steel cell at 270 °C monitored by a thermocouple. The light intensities of monochromatic light equipped with a quartz lens at wavelengths of 275 and 535 nm are 29.8 and 285.1 mW·cm−2, respectively. For all measurements, the light irradiation area is determined as the area of the photocatalyst (0.42 cm2). The reaction time of the monochromatic light was controlled at 3600 s. The outdoor test was performed by using a homemade setup (Fig. 6A). The catalyst wafer was placed at the bottom of the chamber in a support frame. The Fresnel lens was used for concentrating sunlight. During the testing process, the Fresnel lens was adjusted to enable the catalyst wafer to obtain the maximum light intensity. Prior to the next cycle, the chamber was evacuated and filled with argon to exclude the generated hydrogen. The products were analyzed by employing a gas chromatograph (GC-9080, Sun) equipped with a thermal conductivity detector (TCD) for the quantitative detection of hydrogen and nitrogen components in samples. The H2 evolution rate, TOF, and TON were calculated according to the following equations:

$${{{{{\rm{H}}}}}_{2} \, {{{\rm{evolution}}}}\; {{{\rm{rate}}}}}_{{surface}}=\frac{{{{{\rm{H}}}}}_{2} \, {{{\rm{yield}}}}\; {{{\rm{over}}}}\; {{{\rm{the}}}}\; {{{\rm{catalyst}}}}\; {{{\rm{per}}}}\; {{{\rm{hour}}}}}{{{{\rm{Area}}}}\; {{{\rm{of}}}}\; {{{\rm{the}}}}\; {{{\rm{catalyst}}}}}$$
(S1)
$${{{\rm{TOF}}}}_{{Ru}}=\frac{{{{\rm{H}}}}_{2} \, {{{\rm{production}}}}\;{{{\rm{amount}}}} \, \left({{{\rm{\mu }}}}{{{\rm{mol}}}}\right)}{{{{\rm{Content}}}}\, {{{\rm{of}}}}\, {{{\rm{immobilized}}}} \, {{{\rm{Ru}}}}\, {{{\rm{nanoparticles}}}}\, \left({{{\rm{\mu }}}}{{{\rm{mol}}}}\right){{\times }}{{{\rm{Reaction}}}} \, {{{\rm{time}}}} \, ({{{\rm{h}}}})}$$
(S2)
$${{{\rm{TON}}}}_{{Ru}}=\frac{{{{\rm{H}}}}_{2} \, {{{\rm{production}}}} {{{\rm{amount}}}}\, \left({{{\rm{\mu }}}}{{{\rm{mol}}}}\right)}{{{{\rm{Content }}}}\, {{{\rm{of }}}}\, {{{\rm{immobilized }}}}\, {{{\rm{Ru}}}} \, {{{\rm{nanoparticles}}}}\, \left({{{\rm{\mu }}}}{{{\rm{mol}}}}\right)}$$
(S3)

The optimal content of Ru over GaN NWs/Si was determined to be 0.11 μmol·cm−2 measured by ICP-OES.

The AQE (φ) was calculated by this equation:

$$\varphi=\frac{{nR}}{I}$$
(S4)

where n is the number of involved electrons, the R is hydrogen evolution rate (Supplementary Fig. S12), and I is the number of incident protons8,44. The light intensities (P) of monochromatic light equipped with a quartz lens at wavelengths of 275 and 535 nm are 29.8 and 285.1 mW·cm−2, respectively. For all measurements, the light irradiation area (S) is determined as the area of the photocatalyst (0.42 cm−2). The reaction time (T) of the monochromatic light was controlled at 3600 s. The detailed calculation is shown below8,44

$$I=\frac{E\lambda }{{hc}}{{\times }}S=\frac{{PT}\lambda }{{hc}}{{\times }}S$$
(S5)

Where E is the energy of photons, λ is the wavelength of using monochromatic light, h is the Plank constant, and c is the speed of light.

The kinetic isotope effects (KIE) were calculated by dividing the reaction rates of H2 production from NH3/H2O by the reaction rates of H2 production from NH3/D2O.

The solar energy conversion efficiency was calculated by the following equation:

$$\eta=\frac{R{\times \varDelta H}^{0}}{{{\mbox{Light intensity}}}\times S\times 1{\mbox{hour}}} \times 100\%$$
(S6)

Where the ΔH0 (94.6 kJ·mol−1) is the standard reaction enthalpy changes of ammonia decomposition. The area of catalyst (S) is about 30.8 cm−2. Considering the fluctuation of natural light, the light intensity is averaged within one hour, for example, the average light intensity from 13:00 to 14:00 was determined to be 7.97 W·cm−2.

Computational methods. We constructed 4 × 4 Ru (0001) surface with 4 atomic layers and 4 × 2 GaN (10\(\bar{1}\)0) surface with six atomic layers to represent the pure Ru and GaN systems, respectively. Additionally, Ru/GaN was created by depositing a small cluster of 6 Ru atoms on GaN (10\(\bar{1}\)0). Various Ru/GaN configurations were examined, and we identified the most stable configuration, as illustrated in Supplementary Fig. S24, for subsequent calculations. During the structural relaxation process, the bottom two layers of Ru (0001) and the bottom three layers of GaN slabs were kept fixed in their bulk positions. To minimize image interactions, a vacuum spacing of at least 12 Å was added along the normal direction of the surface in all slab models. The reaction intermediate for NH3 decomposition and H2O dissociation on those models are shown in Supplementary Figs. S25, S27–S29.

Spin-polarized density functional theory (DFT) calculations were performed using the Vienna Ab Initio Simulation Package (VASP)45,46. The projector-augmented wave (PAW) method was adopted to describe the interaction between valence electrons and ions47. Perdew-Burke-Ernzerhof (PBE) functional was employed to present the exchange-correlation in the Kohn-Sham equation48,49. Van der Waals (vdW) interactions were included using Grimme’s DFT-D3 method50,51. A plane-wave basic set with a kinetic energy cutoff of 500 eV was used. For all slabs, a 2 × 2 × 1 Mokhorst-Pack k-point grid was implemented to sample the Brillouin zone52. All structures converged until the atomic forces and the total energies reached 0.02 eV/Å and 10−5 eV, respectively.

In this study, the free energies were computed using the computational hydrogen electrode (CHE) model53. The Gibbs free energy of adsorption ∆G was calculated by:

$$\Delta G={E}_{{ad}}+\Delta {ZPE}-T\Delta S$$
(S7)

where Ead is the adsorption energy defined by:

$${E}_{{ad}}={E}_{{surface}+{adsorbate}}-{E}_{{surface}}-{E}_{{adsorbate}}$$
(S8)

ΔZPE and ΔS represent the changes in zero-point energy and entropy, respectively. The temperature, denoted as T, was set to the room temperature.