Introduction

Spinodal decomposition is an atomic level mechanism in which a thermodynamically unstable solid solution decomposes into two coexisting phases with distinctly different chemical compositions, a process that is of importance both in fundamental research and practical applications1,2,3. Fundamental to spinodal decomposition phenomenon is the absence of an energy barrier on its transformation pathway such that the decomposition occurs through spontaneous uphill diffusion process to create nanoscale dispersed phases. As a notable consequence of this process, the two newly formed phases are coherent with each other, showing lattice continuity at their interface4,5,6. Spinodal decomposition phenomenon was discovered in metal alloys early in 1940’s (ref. 7). In the intervening nearly a century, this mechanism has been widely explored in a lot of examples of materials where such defined phase segregation is needed so as to improve the properties and performance of materials4,5,8,9,10,11. Here in this work, we present the first demonstration of the utility of spinodal decomposition as a unique means to produce the lattice coherent metal/semiconductor heterostructure for capture and conversion of non-equilibrium plasmonic hot charge carriers. Such a coherent metal/semiconductor is formed between the plasmonic hafnium nitride (HfN) and its native oxynitride compound—Hf2ON2, which has proven to be highly effective for directing the separation and transport of the plasmonically excited hot electrons in HfN to drive photocatalysis.

The generation of nonequilibrium hot electrons through plasmon excitation of metal nanostructures has been proposed for a wide number of applications in optical energy conversion12,13,14,15. Plasmonic excitation in metal nanostructures involves the collective oscillation of surface free electrons induced by the incident radiation, offering a viable way to enhance light-matter interactions. So far, noble metals such as Ag and Au have been virtually the only materials used in plasmonic energy conversion in that the localized surface plasmon resonance (LSPR) excitation is well located in the visible and near infrared (NIR) regions16,17,18. Known as a novel class of conductive ceramics, group IVB transition-metal nitrides (TMNs), such as titanium nitride (TiN), zirconium nitride (ZrN), and HfN show great potential as cost-effective and viable alternatives to Au and Ag. In addition to their appealing optical properties at the border of the visible and NIR regions, these nitride plasmonic materials also have more lowered work functions (ranging from 3.6 to 4.5 eV) compared to standardly used Au plasmonics, which make them more easily amenable for aligning with the electron affinity potentials of most compound semiconductors19,20. Importantly, as far as the chemical nature is concerned, unlike elemental metals, TMNs are compound plasmonic materials and there exist a family of native oxide and oxynitride counterparts, which provides a basis for the rational design and construction of plasmonic metal/semiconductor heterostructures. Taken HfN for instance, its native stoichiometric oxides and oxynitrides, namely HfO2, Hf7O8N4 and Hf2ON2 (Fig. 1a), can form a whole material family with continuously variable chemical composition and vastly different electronic band structures. Specifically, while HfN is metallic and HfO2 is an electrical insulator with a band gap of ~5.1 eV (ref. 21), the in between Hf7O8N4 and Hf2ON2 are both n-type semiconductors with theoretically predicted bandgaps of 3.66 eV (ref. 22) and 2.83 eV (refs. 22,23), respectively. Through a judicious combination of plasmonic HfN with its different oxides or oxynitrides, it should be possible to construct heterostructures for desired purpose.

Fig. 1: Schematic diagram of spinodal decomposition in HfN-Hf2ON2 system.
figure 1

a Ternary phase diagram of the Hf-O-N system. The circles filled with solid yellow denote the stoichiometric oxides and oxynitrides of Hf. They are HfO2, Hf7O8N4, Hf2ON2 and HfN, respectively. The non-stoichiometric oxynitrides, denoted as Hf2O1-xN2 (0 < x < 1), vary in composition between HfN and Hf2ON2. The spinodal decomposition process initiates from such Hf2O1-xN2 intermediate phases and eventually leads to the formation of the HfN/Hf2ON2 heterostructure. b Schematic Gibbs free energy graph and the corresponding phase diagram for the HfN-Hf2ON2 system. The Hf2O1-xN2 intermediate phase falls into the region of negative curvature \(\it \it \it \it \it \it \it \left(\frac{{\partial }^{2}{\mbox{G}}}{{\partial {\mbox{c}}}^{2}} < 0\right)\), lying within the inflection points \(\it \it \left(\frac{{\partial }^{2}{\mbox{G}}}{{\partial {\mbox{c}}}^{2}}=0\right)\) of the curve.

Full size image

A key challenge for the practical utilization of plasmonic hot carriers generated in metal nanostructures is their ultrashort lifetime on the order of 1 ps24, which therefore requires strategies for efficiently separating hot electron-hole pairs. To this end, heterostructures consisting of plasmonic metals and appropriate semiconductors have proven to be effective, which provide a channel for quickly extracting hot carriers into the semiconductor13,25,26,27. In general, there are two mechanisms accounting for this charge transfer process: the plasmon-induced hot electron transfer (PHET) and the plasmon-induced interfacial charge-transfer transition (PICTT)24,28. The PHET mechanism involves a sequential excitation-charge transfer process of carrier generation in the metal by Landau damping, followed by transfer through tunneling or traversing the Schottky barrier. In contrast, the PICTT mechanism typically occurs in plasmonic systems with strong chemical bonds between metal nanostructures and adjacent semiconductors, where the plasmon directly damps to form separated electron-hole pairs through chemical interface damping29. The PICTT alleviates the competition with hot-electron relaxation, typically showing higher efficiency and faster kinetics than PHET. In recent studies, the coexistence of both PICTT and PHET processes has been reported in several plasmonic systems, such as Au/CdSe30, Ag/TiO231, and Ag/CsPbBr332 where it significantly enhances charge transfer efficiency.

Since the efficiency of hot electrons transfer is intimately related to the nature of interface hybridization between the metal and semiconductor, it is reasonable to expect that an ideal lattice-matched, coherent interface with chemical continuity will be of intrinsic benefits for facilitating the hot electrons transfer within a plasmonic metal/semiconductor heterostructure. The plasmonic HfN/Hf2ON2 heterostructure formed via spinodal decomposition in our work is such a case in point. Crucially, both HfN and Hf2ON2 adopt a cubic crystal structure with relatively small lattice mismatch, which is the basis for the formation of coherent interface via spinodal decomposition. In general, spinodal decomposition is governed by the competition between local chemical free energy and composition gradient energy, such that the lattice coherency between the two newly formed phases is a natural consequence of minimizing the total energy. For a specific material system, the most direct and critical experimental criterion for assessing the occurrence of spinodal decomposition is the X-ray diffraction (XRD) characterization, as the appearance of diffraction satellite peaks and their temporal evolution directly reflect the periodic modulation of composition due to the uphill diffusion process during spinodal decomposition. For our HfN/Hf2ON2 in particular, the spinodal decomposition starts from an intermediate phase with an in-between composition, and as expected, we observed the emergence of satellite sidebands, asymmetric broadening, and the progressive shift of diffraction maxima—hallmark features that are widely recognized as signatures of spinodal decomposition. Furthermore, we applied atomic-resolution scanning transmission electron microscopy (STEM) to systematically study the chemical native relationship of HfN and Hf2ON2 within HfN/Hf2ON2 heterostructure and observed coherent interface between them. As a consequence of this structurally well-defined coherent metal/semiconductor interface, there occurs highly efficient, ultrafast hot electrons transfer from plasmonic HfN to neighboring Hf2ON2, as experimentally verified by transient absorption (TA) spectroscopy measurements that show optical signatures suggesting that both PICTT and PHET pathways contribute to the interfacial electron transfer processes. For our HfN/Hf2ON2 heterostructure in particular, it has a broad plasmonic absorption range, extending from visible further into NIR region, and due to the well-matched band alignment between HfN and Hf2ON2 the less energetic hot electrons excited by the lower energy NIR photons can be efficiently collected across the interface. These advantages of the coherent HfN/Hf2ON2 heterostructure are key to enabling its practical utility for solar-to-fuel conversion applications such as photocatalysis. As a leading manifestation, we assessed the photocatalytic performance of HfN/Hf2ON2 heterostructure in aqueous methanol solution for H2 production at room temperature. When combined with a small amount of Pt co-catalyst, the coherent HfN/Hf2ON2 heterostructure achieves a remarkably high rate of photocatalytic H2 production (14.4 mmol g−1 h−1) from methanol under visible and NIR illumination with a maximum apparent quantum yield (AQY) of 27% at 600 nm. Importantly, even with the less energetic NIR (λ > 800 nm) photons as the only energy input, it still maintains an unexpectedly high photocatalytic activity with AQY values of 13.9% at 850 nm and 6.2% at 1064 nm. Given the prominent broadband photocatalytic activity, especially that under illumination of the less-energetic NIR photons, it offers unique possibilities for the full-spectrum solar energy utilization in photocatalysis.

Results

Figure 1b displays a simplified illustration, according to the thermodynamic description typically assumed for spinodal decomposition mechanism, of the Gibbs free energy graph and corresponding phase diagram for phase separation in the HfN-Hf2ON2 system. The initial and key step for producing coherent plasmonic HfN/Hf2ON2 heterostructure through spinodal decomposition is to obtain an intermediate phase with a stoichiometric ratio lying between Hf2ON2 and HfN (Fig. 1a). In essence, such an intermediate phase, denoted as Hf2O1-xN2 (0<x < 1), can be regarded as a single-phase solid solution of HfN and Hf2ON2, both adopting a cubic crystal structure with closely matched chemical bond lengths, which will be discussed further. Experimentally, the Hf2O1-xN2 intermediate phase was synthesized by controlled nitridation process of HfO2 nanoparticles using ammonia gas at elevated temperature, typically 1100 °C (see Methods for experimental details). Through the essentially similar nitridation processes, the pure-phase HfN and Hf2ON2 were also synthesized for structural characterization and properties assessment. As compared to pure-phase HfN and Hf2ON2, the controlled synthesis of the Hf2O1-xN2 intermediate phase is intrinsically more challenging, and it is only with precise control of synthetic conditions that we can obtain the intermediate Hf2O1-xN2 phase with varying x. With the viability of such intermediate phases, the annealing treatment at 1000 °C in an Ar atmosphere will initiate spinodal decomposition process, leading to the formation of HfN/Hf2ON2 heterostructures with different ratios of HfN to Hf2ON2; typically, the intermediate Hf2O1-xN2 phase with larger value of x gives rise to higher volume fraction of HfN in the final heterostructure.

Synthesis and structural characterization of HfN and Hf2ON2

Figure 2a presents the crystal structures and local chemical environments of both HfN and Hf2ON2, with HfN adopting a rocksalt-type structure and Hf2ON2 a cubic bixbyite structure. Specifically, in crystallized HfN, each Hf atom bonds to six equivalent N atoms, forming a mixture of edge- and corner-sharing HfN6 octahedra, while Hf2ON2 has a body-centered cubic structure where the Hf4+ cation is coordinated by four N and two O anions, forming a HfO2N4 octahedra. Powder XRD characterizations in Fig. 2b confirm that both of the as-formed HfN and Hf2ON2 exhibit a highly crystalline cubic structure; the HfN could be indexed to the \({{{\rm{Fm}}}}\bar{3}{{{\rm{m}}}}\) space group and the Hf2ON2 crystallized in the \({{{\rm{Ia}}}}\bar{3}\) space group. A thorough analysis of the elemental composition of HfN and Hf2ON2 was conducted using inductively coupled plasma atomic emission spectroscopy (ICP-AES) and a nitrogen/oxygen elemental analyzer. The results, summarized in Supplementary Table 1, show that the elemental ratios of Hf, N, and O in all three samples closely match their ideal stoichiometric ratios. In addition, EDS mapping in Supplementary Figs. 1 and 2 demonstrates the uniformity of the element distribution.

Fig. 2: Structural characterization of HfN and Hf2ON2.
figure 2

a, b Crystal structures (a) and XRD patterns (b) of the single-phase HfN and Hf2ON2. c, Atomic-resolution HAADF-STEM image of an individual HfN nanoparticle viewed along the [110] zone axis. d The FFT pattern (top) obtained from the HAADF-STEM image in (c) and the corresponding simulated electron diffraction (bottom). e Enlarged STEM image of (c) overlaid with corresponding atomic model. f Simulated atomic-resolution HAADF-STEM image for comparison with (e). g Atomic-resolution HAADF-STEM image of an individual Hf2ON2 nanoparticle viewed along the [110] zone axis. h The FFT pattern (Top) and the corresponding simulated electron diffraction (bottom). i, Enlarged STEM image of (g) overlaid with atomic model. j The simulated atomic-resolution HAADF-STEM image corresponding to (i). Source data for the Fig. 2b is provided as a Source Data file.

Full size image

Based on the Rietveld refined powder XRD of HfN and Hf2ON2 (Supplementary Fig. 3) and their structural information (Supplementary Table 3), it can be seen that the Hf-N bond length in HfN is similar to the average bond length of Hf-N and Hf-O in Hf2ON2. This similarity in crystal structure as well as shared chemical elements between HfN and Hf2ON2 suggests the possibility of forming a single-phase solid solution of Hf2O1-xN2 intermediate phase, which serves as the starting point for spinodal decomposition. As for spinodal decomposition we are about to discuss in the next section, although this process does not involve a nucleation barrier, it is hindered by compositional gradient energy and lattice strain energy. Therefore, the more similar the structures of the final separated phases are, the lower the lattice strain energy to which spinodal decomposition is subjected when it occurs and thus the more favorable it is for decomposition to proceed. Taken together, the structural similarities of HfN and Hf2ON2, as well as the availability of the Hf2O1-xN2 intermediate phase, provide a basis for the occurrence of spinodal decomposition to produce a coherent heterostructure.

Representative scanning electron microscopy (SEM) images of as-grown HfN and Hf2ON2 nanoparticles are shown in Supplementary Fig. 4a–d. Although the high-temperature nitridation process (above 1000 °C) inevitably induces some degree of thermal sintering, the nanoparticles largely retain their individual morphology, with an average diameter of approximately 50 nm. This morphological integrity is particularly important, as we later demonstrate that particle aggregation can significantly influence the optical response through plasmonic coupling effects. To further elucidate the atomic-scale structural features, low-magnification HAADF-STEM images and corresponding FFTs of individual HfN and Hf2ON2 nanoparticles along the [110] and [001] zone axes are provided in Supplementary Figs. 5a, b and 6a, b, respectively. In these HAADF-STEM images, only the Hf atomic columns are visible due to the Z-contrast, while lighter elements such as N and O are not discernible. High-resolution HAADF-STEM images taken along the [110] direction are displayed in Fig. 2c–f for HfN and Fig. 2g–j for Hf2ON2. Despite the apparent similarity in HAADF-STEM images of HfN and Hf2ON2, their corresponding FFTs (Fig. 2d, h) display significantly different diffraction patterns, reflecting underlying differences in atomic structure. Specifically, an alternating shift of Hf atom positions along [\(1\bar{1}0\)] direction is visible, as demonstrated in Fig. 2i and Supplementary Fig. 7, suggesting the presence of octahedral tilting in Hf2ON2. The HAADF-STEM simulations (Fig. 2f, j), obtained with imaging conditions similar to those of the experiments, reproduce the key structural features observed experimentally. These distinct FFT signatures and atomic-scale arrangements serve as reliable structural fingerprints, allowing us to clearly distinguish between HfN and Hf2ON2 phases.

Spinodal decomposition process and characterization of coherent interfaces

Starting from the intermediate phase Hf2O1-xN2, spinodal decomposition takes place upon annealing at 1000 °C in an Ar atmosphere. Figure 3a shows the XRD pattern of the intermediate phase, as well as the XRD patterns of the products obtained after annealing for different durations. The magnified patterns in Fig. 3b clearly exhibit two major changes occurring during the decomposition: (1) a gradual shift of two distinct sets of diffraction peaks toward characteristic positions of the pure HfN and Hf2ON2 phases, respectively, and (2) the appearance of side bands adjacent to the Hf2ON2 (222) and HfN (200) reflections. Specifically, after 10 min of annealing, weak side bands begin to appear indicating the early stages of spinodal decomposition. These features become more pronounced at 20 min, signifying ongoing compositional modulation. By 30 min, the side bands nearly vanish, and the primary peaks clearly resolve into two separate sets corresponding to HfN and Hf2ON2, indicating the completion of phase separation. The transient appearance and subsequent disappearance of these side bands reflect the onset and evolution of the periodic compositional fluctuations that are characteristic of spinodal decomposition33,34. Meanwhile, the continuous shift of the main diffraction peaks suggests gradual adjustments in the average chemical composition of the evolving phases. These observations align well with well-established experimental signatures of spinodal decomposition. The essentially similar structural evolution features have been reported not only in alloy systems, but also incompound materials such as TiO2/SnO235,36, TiO2/VO237,38, TiN/AlN39,40 and PbS/Pb1-xSnxTe41.

Fig. 3: Spinodal decomposition and characterization of coherent interfaces.
figure 3

a XRD evolution during phase separation from Hf2O1-xN2 intermediate phase. b Detailed XRD pattern showing changes in the Hf2ON2 (222), HfN (111), Hf2ON2 (004), and HfN (200) peaks. c HAADF-STEM image of coherent interface along [110], with composite FFT inset. d Inverse FFT image revealing the spatial distribution of the Hf2ON2 phase and interface contours, which derived from the red solid circles representing the diffraction spots of Hf2ON2. e GPA strain map (εyy) with Hf2ON2 as reference. f High-resolution STEM of HfN/Hf2ON2 interface with simulated image. g HAADF-STEM image along the [111] zone axis. h, i FFT patterns of HfN and Hf2ON2 regions. j, k Enlarged STEM images of regions in (g) with simulated atomic-resolution images of HfN and Hf2ON2, respectively. Source data for the Fig. 3a, b are provided as a Source Data file.

Full size image

Here, it is worth noting that, for spinodal decomposition occurring at nanoscale particulate systems like our case, the nanoscale size effect also has a complex influence on this phase separation process. Specifically, the large free surfaces of nanoparticles provide a fast diffusion pathway and thereby accelerate the decomposition kinetics2. Additionally, the enhanced surface energy and curvature may alter the morphological evolution during phase separation, which, in turn, affects the final biphasic textural features42. For ultra-small nanoparticles (e.g., <5 nm), however, the significantly elevated surface energy can thermodynamically suppress the onset of spinodal decomposition43,44. In our study, the intermediate phase Hf2O1-xN2 was obtained as a nanoparticulate powder with particle sizes typically in the range of several tens of nanometers, predominantly falling within 30-70 nm (Supplementary Fig. 8a–c). At this scale, surface effects are not strong enough to suppress the spinodal decomposition process but remain appreciable to influence the decomposition kinetics. In our experiments, the entire spinodal decomposition process generally completes within 20 to 30 min, indicating rapid kinetics likely facilitated by surface effects. Moreover, such surface effects may also profoundly impact the morphological textural features of the resulting HfN/Hf2ON2 heterostructure, and this remains a complex issue for future investigation. The refined XRD pattern (Supplementary Fig. 9) of the final HfN/Hf2ON2 heterostructure shows that the ratio of HfN to Hf2ON2 is approximately 0.5. As mentioned earlier, the Hf2O1-xN2 intermediate phases allow for continuous and tunable adjustment of the O content by precisely controlling nitridation conditions. The XRD patterns of different intermediate phases obtained under different nitridation conditions are shown in Supplementary Fig. 10a. After the completion of spinodal decomposition, HfN/Hf2ON2 heterostructures with varying ratios of HfN to Hf2ON2 were formed. Shown in Supplementary Fig. 10b is the XRD changes during annealing process starting from another intermediate phase which was obtained at 1100 °C for 50 min. The SEM image and low-magnification STEM image in Supplementary Fig. 11 reveal the morphology of HfN/Hf2ON2 obtained by spinodal decomposition. It can be observed that despite undergoing high-temperature annealing, the particle size and morphology of the HfN/Hf2ON2 remain highly consistent with the intermediate phase prior to annealing, without further sintering or morphological changes.

The XRD characterization results presented above provide direct evidence for the occurrence of spinodal decomposition. Generally, the progression of spinodal decomposition leads to the formation of coherent or semi-coherent interfaces11,45,46. In our system, both HfN and Hf2ON2 adopt a cubic crystal structure with a relatively small lattice mismatch, which facilitates the formation of a coherent interface during spinodal decomposition. To further substantiate this, we complemented our study with advanced HAADF-STEM characterization, which provides crucial evidence confirming the formation of lattice-coherent interfaces. Shown in Fig. 3c is a typical HAADF-STEM image of the HfN/Hf2ON2 heterostructure, taken along the [110] zone axis. On each side of the interface, the upper region is composed of HfN, while the lower region is Hf2ON2, as discerned by their respective FFT patterns (Supplementary Fig. 12a). Notably, the FFT patterns obtained from the entire field of view (Supplementary Fig. 13), including both HfN, Hf2ON2, and the interface, reveal nearly overlapping two sets of diffraction spots—indicative of a coherent lattice relationship between the two phases. We further employed nanobeam electron diffraction (NBED) on the heterostructure, which allows for spatially resolved diffraction from regions as small as 1-2 nm. NBED patterns collected from both sides of the interface (Supplementary Fig. 14) clearly exhibit diffraction patterns corresponding to HfN and Hf2ON2, respectively, confirming phase identification from the reciprocal space. Additionally, the NBED results also reveal a well-matched coherence relationship between the two phases.

To further investigate the phase distribution, we performed inverse FFT analysis by selecting diffraction spots corresponding to Hf2ON2 (marked with red solid circles in the inset FFT of Fig. 3c). The inverse FFT reveals the spatial distribution of Hf2ON2 in detail, allowing precise identification of the interface profile. To complement this, geometric phase analysis (GPA) was employed to evaluate the strain distribution across the coherent HfN/Hf2ON2 interface (Fig. 3e and Supplementary Fig. 15). The strain map (εyy) provides distinct contrast between the HfN and Hf2ON2 regions, thereby delineating the interface location—consistent with the boundary identified by inverse FFT analysis. The line profile of εyy strain map (Supplementary Fig. 15) shows that the interfacial strain is generally below 10%, a level that remains within the tolerance for coherent interface formation between HfN and Hf2ON2. A magnified HAADF-STEM image of the heterogeneous interface is presented in Fig. 3f, where HfN and Hf2ON2 domains are seen to seamlessly connect without obvious dislocations or extended defects, aligning with the corresponding locally simulated interface region. This coherent interface forms through uphill diffusion during spinodal decomposition, driven by the minimization of interfacial energy between the emerging phases.

Further observations along the [111] zone axis (Fig. 3g) reveal distinct atomic arrangements, with corresponding FFT patterns (Fig. 3h, i) confirming the identity of HfN and Hf2ON2. The inverse FFT patterns in Supplementary Fig. 16c clearly delineate the position of the HfN/Hf2ON2 interface. Figure 3j compares the atomic arrangements in the HfN region between experimental and simulated images, while Fig. 3k presents the corresponding comparison for Hf2ON2, collectively confirming the formation of a coherent heterostructure. It is worth mentioning that the heterostructure along the [111] zone axis is more intuitive and easier to distinguish compared to the [110] zone axis heterostructure, due to its distinct and readily recognizable atomic arrangement. The STEM characterizations were also conducted along the [001] and [121] zone axes (Supplementary Fig. 12b, d), revealing the coherent relationship of HfN and Hf2ON2 along multiple crystallographic directions. Moreover, high-resolution STEM imaging combined with EELS mapping was performed along the [001] zone axis (Supplementary Figs. 16a, b and 17). The spatial distribution of N and O near the interface closely matches the heterostructure phase contrast observed in the HAADF-STEM image. Low-magnification EELS mapping (Supplementary Fig. 18) reveals an embedded heterostructure morphology consistent with the characteristic compositional modulation of spinodal decomposition. These STEM results, together with the XRD analysis of phase evolution, provide strong and complementary evidence for spinodal decomposition and the formation of the HfN/Hf2ON2 heterostructure with coherent interfaces. As we will discuss below, such coherent interfaces play a critical role in facilitating the extraction and transport of hot electrons from plasmonically excited HfN to semiconducting Hf2ON2.

Optical properties and interfacial charge transfer dynamics of HfN/Hf2ON2

The plasmonic properties of HfN have been extensively studied both theoretically and experimentally47,48,49. As expected, the XPS valence-band spectrum of HfN shows a pronounced peak near the Fermi level (Supplementary Fig. 19), which is characteristic of a typical metallic electronic structure. In contrast, Hf2ON2 displays a typical semiconducting profile without any notable peak near the Fermi level. The metallic character of HfN and the semiconducting character of Hf2ON2 are further corroborated by their optical absorption spectra presented in Fig. 4a. Specifically, HfN displays broadband absorption with a peak centered around 500 nm, while Hf2ON2 shows typical semiconductor absorption behavior with a band edge near 480 nm, consistent with previous literature22. Upon forming the HfN/Hf2ON2 heterostructure, the spectrum shows both the band-edge absorption of Hf2ON2 and the broadband LSPR absorption of HfN, covering both the visible and NIR regions. The top panel of Fig. 4a shows the absorption spectra of HfN/Hf2ON2 heterostructures with varying HfN-to-Hf2ON2 ratios. The LSPR peaks of HfN/Hf2ON2 heterostructures exhibit a clear redshift and broaden further into the NIR region. Such changes in the LSPR band cannot be solely attributed to the dielectric; rather, they indicate the possible presence of strong interfacial damping between the metal and the semiconductor28,29. Furthermore, the intimate contact interface and electronic interaction between HfN and Hf2ON2 are also evidenced by strong photoluminescence (PL) quenching, as shown in Fig. 4b. Upon selective excitation of Hf2ON2 with a 365 nm laser, the HfN/Hf2ON2 heterostructure exhibits significant photoluminescence quenching compared to pure Hf2ON2. This is accompanied by a decrease in PL lifetime from 4.36 ns for pure Hf2ON2 to 3.67 ns when coupled with HfN in the heterostructure (Fig. 4c).

Fig. 4: Optical properties and interfacial charge transfer dynamics of HfN/Hf2ON2 heterostructure.
figure 4

a UV-Vis-NIR absorption spectra of HfN/Hf2ON2 heterostructures with different HfN: Hf2ON2 ratios (1:4.5 (light red line), 1:2.1 (dark red line), 1:1.25 (pale purple line)), alongside pure-phase Hf2ON2 (orange) and HfN (black). b Steady-state PL spectra showing quenched emission in heterostructures. c Time resolved PL lifetime curves of HfN/Hf2ON2 heterostructure and the pure-phase Hf2ON2. d A two-dimensional pseudo-color plot of HfN/Hf2ON2 heterostructure with pump wavelength of 650 nm. e TA kinetics probed at different wavelengths (420 nm, 460 nm and 480 nm). f TA kinetics probed at 720 nm. g Schematic illustration of plasmon-induced hot-electron transfer in HfN/Hf2ON2 heterostructure after femtosecond laser pumping. h Transient current-time curves of HfN/Hf2ON2 photoanode at 0.27 V (vs Ag/AgCl) under monochromatic light irradiation. Source data for the Fig. 4a–h are provided as a Source Data file.

Full size image

As discussed above, the efficiency of plasmonic hot electron extraction in a metal/semiconductor heterostructure critically depends on the quality and extent of the contact interface. The greater the extent of interface contact, the higher the electron transfer efficiency. In this work, the availability of the lattice-matched, coherent interface between the plasmonic HfN and semiconductor Hf2ON2 constitute the basis of highly efficient hot electrons extraction. The interfacial band structure of HfN/Hf2ON2 heterostructure is schematically shown in Supplementary Fig. 20. It can be seen that the interfacial barrier is small, facilitating the collection of low-energy hot electrons.

To confirm charge transfer at the metal/semiconductor interface, we employed transient absorption spectroscopy to probe the dynamics of plasmon-induced hot carriers. Figure 4d–f show the optical density change (ΔOD) spectra and time evolution of the heterostructure samples under 650 nm optical pumping. Since the 650 nm laser has a photon energy below the band gap of Hf2ON2 (~480 nm), it selectively excites the LSPR of HfN within the HfN/Hf2ON2 heterostructure. The TA spectrum of the heterostructure is presented as a two-dimensional pseudo-color plot in Fig. 4d. The kinetic decay profile at 720 nm shown in Fig. 4f exhibits characteristics consistent with the LSPR of HfN. Apart from this dominating optical feature, crucially, at 485 nm there also rises a weak yet distinguishable bleaching signal, which corresponds to the band-edge bleaching of Hf2ON2. This bleaching is more clearly visible in the dynamic curve shown in Fig. 4e and Supplementary Fig. 21a, where it can be seen that the signal reaches its peak at around 300 ps and then decays. For the pure-phase Hf2ON2, no transient signal is observed (Supplementary Fig. 22). Therefore, the bleaching signal at 485 nm observed in the HfN/Hf2ON2 heterostructure is attributed to charge transfer from HfN to Hf2ON2. According to the dynamics fitting at 720 nm with a multiple exponential, decay constants were obtained in Supplementary Table 4. For the HfN/Hf2ON2 heterostructure, an ultrafast decay component with a short lifetime (<100 fs) appears, which is absent in pure HfN (Supplementary Fig. 21b, c and Supplementary Table 4). Such an ultrafast process indicates the occurrence of PICTT with an amplitude ratio of ~30%. Furthermore, TA spectroscopy under 700 nm excitation—which efficiently excites the plasmonic resonance of HfN while minimizing spectral overlap with Hf2ON2—confirms the ultrafast charge transfer behavior (Supplementary Fig. 23). As we discussed earlier, PICTT requires a higher-quality interfacial contact compared to PHET, and typically both PICTT and PHET pathways contribute to the overall hot electron transfer process in plasmonic metal/semiconductor systems with strong coupling interfaces. This coexistence has been reported in several recent studies29,30,31,32. The transient spectrum results present evidence of the occurrence of PICTT, evoking the picture of the coexistence of both PHET and PICTT as depicted in Fig. 4g.

Another important piece of evidence for plasmonic hot electron transfer is provided by a one-electron photoelectrochemical (PEC) model reaction. The PEC measurements were carried out in an aqueous solution containing a reversible one-electron redox couple, [Fe(CN)6]3−/4− (50 mM/50 mM). In these experiments, the HfN/Hf2ON2 heterostructure served as the photoanode, where the oxidation reaction occurred; meanwhile, photogenerated electrons generated from the photoanode were transferred to the platinum electrode via an external circuit. Electrochemical impedance spectroscopy (EIS) of the HfN/Hf2ON2 film shows a small semicircle in the Nyquist plot. This feature indicates an efficient charge transfer process characterized by low charge transfer resistance at the electrode–electrolyte interface (Supplementary Fig. 24a).

To corroborate the results of the transient absorption measurement, amperometric photocurrent density-time profiles (light chopping) were obtained at an applied bias voltage of 0.27 V (vs Ag/AgCl) under monochromatic light ranging from the visible to the NIR region (Fig. 4h). Under these conditions, hot electrons generated by the photoexcitation of HfN are transferred to Hf2ON2 and then collected in the external circuit. The photocatalyst utilized to generate the photocurrent density-time profile depicted in Fig. 4h is a heterostructure with a HfN/Hf2ON2 ratio of 0.5. The wavelength-dependent photocurrent response measured on HfN/Hf2ON2 samples is shown in Supplementary Fig. 24b, together with the corresponding absorbance spectrum. The photoinduced current response qualitatively tracks the plasmonic absorbance spectra of the corresponding HfN/Hf2ON2 photoelectrodes, suggesting that plasmon excitation, possibly involving hot-electron generation, contributes significantly to the observed photocurrent.

Under 600 nm illumination, the photocurrent density reached 16.5 μA/cm2. More importantly, even under NIR illumination (λ > 800 nm), the plasmonic photoanode also exhibited appreciable photocurrent responses, with photocurrent densities of 13 μA/cm2 and 7.5 μA/cm2 under 850 nm and 900 nm illumination, respectively. Control experiments (Supplementary Fig. 24c) with HfN alone shows decreased photocurrents under both visible and NIR illuminations compared to HfN/Hf2ON2 heterostructures. Hence, the enhanced photoelectrocatalytic activity of the HfN/Hf2ON2 heterostructure photoanode can be attributed to the lattice-coherent heterointerface, which facilitates efficient charge carrier separation and transport. Even hot electrons generated under low-energy photon excitation can transfer across the interface from HfN to Hf2ON2, indicating that the heterostructure possesses significant potential for full-spectrum solar energy utilization.

Photocatalytic hydrogen generation from methanol solution

To demonstrate practical utility, we employed the HfN/Hf2ON2 heterostructure as a photocatalyst for hydrogen production from photocatalytic methanol decomposition reaction. Following a procedure similar to those previously established50,51, we investigated the H2 evolution from aqueous methanol with the plasmonic HfN/Hf2ON2 heterostructure. The reaction temperature was maintained at 25 °C using a connected reflux water condenser to prevent temperature increases from light radiation, thus minimizing interference from photothermal effects. Pt was selected as the hydrogen‑evolution cocatalyst and photodeposited onto the heterostructure from an aqueous H2PtCl6 solution. The Pt particles obtained via photodeposition exhibit a face-centered cubic crystal structure, uniform spatial distribution, and an average size of approximately 2 nm, as evidenced by Supplementary Fig. 25. Their nanometric size (~2 nm) affords a high surface area, increasing interfacial contact with the heterostructure and providing more active sites for hydrogen evolution. Given the electron flow direction in the HfN/Hf2ON2 system, Pt predominantly deposits on the Hf2ON2 side, as also verified by HAADF-STEM images (Supplementary Fig. 25d).

With different amounts of Pt loading, the HfN/Hf2ON2 with a HfN to Hf2ON2 ratio of 1:2.1 achieves varying degrees of improvement in hydrogen evolution (Fig. 5a). Under optical excitation, the plasmonically excited electrons from HfN can be transferred to Hf2ON2, and subsequently collected by Pt for initiating reaction. Specifically, in the absence of Pt, the pure HfN/Hf2ON2 heterostructure exhibited no observable photocatalytic activity; nevertheless, after loading a small amount of Pt, even 0.2 wt% loading, there was a noticeable boost in H2 evolution. An optimal Pt loading of ~1.0 wt% yields an H2 evolution rate of up to 11.2 mmol g−1 h−1 under visible and NIR illumination (λ > 400 nm). With further increase of the loading amount, the rate of H2 evolution shows no significant change. Transient absorption spectroscopy (Supplementary Fig. 26) reveals the influence of Pt loading on the carrier dynamics of both HfN and the HfN/Hf2ON2 heterostructure. Compared to Pt-HfN, the Pt-HfN/Hf2ON2 heterostructure demonstrates a longer carrier lifetime, reflecting a reduction in electron–hole recombination due to efficient electron transfer.

Fig. 5: Photocatalytic H2 generation from aqueous methanol.
figure 5

a H2 evolution rate of Pt-HfN/Hf2ON2 catalysts with different amounts of Pt loading from aqueous methanol under visible and NIR light irradiation (λ > 400 nm). The ratio of HfN/Hf2ON2 is 1:2.1. b The performance for H2 evolution of various samples with optimal Pt loading, including HfN, Hf2ON2, and HfN/Hf2ON2 heterostructures with varying ratios of HfN to Hf2ON2, under visible and NIR light irradiation (λ > 400 nm). c The effect of different methanol concentrations on H2 evolution rates. d Typical time courses of H2 evolution amount under different illumination conditions: visible and NIR light irradiation (λ > 400 nm), AM1.5 G, and NIR light (λ > 800 nm) illumination. e Wavelength-dependent AQY of H2 evolution using the Pt-HfN/Hf2ON2 photocatalyst, along with the UV–Vis–NIR absorption spectrum of the HfN/Hf2ON2 heterostructure. Error bars represent the standard deviation from three independent measurements. Source data for the Fig. 5a–e are provided as a Source Data file.

Full size image

Under the optimal Pt loading condition, the photocatalytic activity of a series of HfN/Hf2ON2 heterostructures with different ratios of HfN and Hf2ON2 were evaluated, as shown in Fig. 5b. The results reveal that the phase volume ratio of HfN to Hf2ON2 also influences the performance of the coherent HfN/Hf2ON2 photocatalyst, and the best-performing ratio was found to be about 1:1.25 of HfN to Hf2ON2. In contrast, both bare Hf2ON2 and HfN exhibited negligible activity for methanol decomposition, primarily due to the poor visible light absorption of Hf2ON2 and the limited charge separation efficiency of HfN alone. This enhanced performance is attributed to the efficient transfer of plasmonically excited electrons from HfN to Hf2ON2 under optical excitation, followed by the subsequent collection by Pt, which facilitates the hydrogen evolution reaction. Figure 5c shows the dependence of H2 evolution on methanol concentration for the HfN/Hf2ON2 (1:1.25) heterostructure loaded with 1.0 wt% Pt, under visible and NIR illumination (λ > 400 nm). The H2 evolution rate increases markedly with rising methanol concentration, with a pronounced enhancement up to 80 vol%. Even in pure methanol (100 vol%), an appreciable amount of H2 evolution is observed, possibly indicating that the H2 evolution is from methanol decomposition rather than water splitting, which is consistent with previously reported results in the literature52,53,54.

The well-matched band alignment between HfN and Hf2ON2 gives rise to a low interfacial barrier and thereby permits the collection of less-energetic hot electrons. To determine whether the less energetic photons can drive the catalytic reaction, we measured the photocatalytic activity at varied wavelengths controlled by different long-pass cut off filters, as shown in Fig. 5d. When illuminated under visible and NIR light (λ > 400 nm) for 3 h, the HfN/Hf2ON2 heterostructure can achieve a H2 evolution rate of 14.4 mmol g−1 h−1. Remarkably, solely by NIR light illumination (λ > 800 nm), the amount of H2 evolution still remains at a high level of 6.7 mmol g−1 h−1. Moreover, the HfN/Hf2ON2 sample presents a stable high activity over at least 12 h of reaction/evacuation processes. Figure 5e shows the wavelength-dependent AQY along with the absorbance spectrum of the HfN/Hf2ON2 heterostructure. The AQY closely follows the plasmonic absorbance profile, indicating that the plasmonic excitation of HfN plays a direct role in enhancing hydrogen evolution. The AQY of the HfN/Hf2ON2 photocatalyst exhibits a broad spectral response, with a peak value of 27% at approximately 600 nm. Importantly, even with the less energetic NIR (λ > 800 nm) photons as the only energy input, it still maintains an unexpectedly high photocatalytic activity with AQY values of 13.9% at 850 nm and 6.2% at 1064 nm. Such performance highlights the potential of the HfN/Hf2ON2 heterostructure for full-spectrum solar energy utilization and enhanced solar energy conversion efficiency.

For plasmonic photocatalysis, photothermal effects are also a non-negligible factor that may synergize with the hot carrier effects to drive reactions such as the methanol decomposition as investigated in our present study. Photothermal heating arises from electron-phonon scattering of unextracted hot carriers, causing localized temperature increases in plasmonic nanoparticles. Generally, the extent of this contribution varies depending on the specific reaction chemistry and operating conditions. For a nanoparticulate dispersion-based catalytic setup operated at room temperature like in our study, the efficient convective heat dissipation will limit the rise in local surface temperature55,56. Such mild photothermal heating is insufficient to drive the catalytic methanol decomposition, a reaction which, according to previous reports, requires temperatures typically above 200 °C and elevated pressures when driven purely by thermal energy. This limitation highlights the rationale behind our design of a coherent metal/semiconductor heterojunction: to enable efficient hot-electron transfer and extraction for driving the photocatalytic methanol decomposition. Indeed, as mentioned above, our transient absorption spectroscopy results provide direct evidence of efficient interfacial charge transfer from HfN to Hf2ON2. Importantly, further photoelectrochemical measurements confirm that the extracted hot electrons actively participate in the photochemical reactions. Therefore, it is the hot electron-driven process that plays a dominant role in the activation the photocatalytic methanol decomposition for high-efficiency H2 generation, although the photothermal effects also synergistically contribute to the enhancements of photocatalytic rates.

Investigation of the reaction mechanism

According to previous literature57,58,59, the reaction pathway of methanol dehydrogenation generally follows a similar mechanism under both photocatalytic and thermocatalytic conditions. The key steps involve the dissociative adsorption of methanol, the formation of intermediate species and their subsequent oxidation to produce formaldehyde (HCHO) or deeper oxidized products. To gain further insight into this process, we conducted in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) to monitor methanol decomposition reaction over the Pt-HfN/Hf2ON2 heterostructure in liquid-phase methanol. As shown in Fig. 6a, spectra were collected over 34 min of visible light illumination, with data recorded at 1-min intervals. Before illumination, the photocatalyst was immersed in methanol for 10 min to allow sufficient adsorption of methanol onto the surface.

Fig. 6: The mechanism of photocatalytic methanol decomposition.
figure 6

a In situ DRIFTS spectra of the Pt-HfN/Hf2ON2 heterostructure with and without visible light. The light gray curve represents the data obtained under dark conditions. The spectra were collected during 34 min of visible light illumination, with data recorded at 1-min intervals. b DMPO adducts recorded in a pure methanol system containing Pt-HfN/Hf2ON2 heterostructure under ambient conditions, with and without visible-light illumination. Source data for the Fig. 6a, b are provided as a Source Data file.

Full size image

Upon light exposure, typical infrared absorption signals of adsorbed methoxy species (⁎CH3O) (~1106 cm⁻¹, which associated with the C-O stretch vibration) appear, with their intensity increasing as the illumination time is extended. This indicates the occurrence of the dehydrogenation of CH3OH to methoxy. The ⁎CH3O species is a key intermediate during methanol decomposition. Additionally, absorption bands at ~1284, 1434, and 1703 cm−1 were observed, which correspond to the rocking and scissoring vibrations of C-H in adsorbed formaldehyde (⁎HCHO), suggesting that the reaction pathway involves the dehydrogenation of methanol to methoxy, followed by oxidation to formaldehyde and hydrogen gas. As the reaction time increases, formaldehyde undergoes further oxidation into formic acid species, evidenced by infrared absorption signals at ~3542 cm−1 (O-H stretch of ⁎COOH), 3200 cm−1 (O-H stretch of ⁎COOH dimer), and 1644 cm−1 (O-H bending of ⁎COOH), signifying a deeper oxidation pathway. Furthermore, EPR spectroscopy was employed to detect radical species generated during the initial stages of the reaction. The measurements were performed in methanol solution using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as a spin-trapping agent for free radicals. As shown in Fig. 6b, two radical intermediates—•CH3O and hydroxymethyl (•CH2OH)—were detected under visible light illumination. These radicals are commonly observed in methanol decomposition reactions. The combined DRIFTS data and EPR results indicate the intermediate products and the oxidation pathway are consistent with typical methanol decomposition reactions for hydrogen production.

A schematic overview of the reaction mechanism is provided in Supplementary Fig. 27. The production of electron-hole pairs in the HfN, as one of the decay channels of the surface plasmon excitation, is followed by electron transfer to the conduction band of Hf2ON2. These hot electrons are subsequently collected by Pt for initiating the proton reduction reaction, generating hydrogen. The remaining hot holes in HfN are directly utilized to drive the oxidation reaction of methanol, producing oxidation products such as formaldehyde and formic acid. This efficient charge separation in the HfN/Hf2ON2 heterostructure is reflected in the enhanced photocatalytic hydrogen production performance.

Discussion

In plasmonic nanostructures, strong light-matter interactions enable efficient light harvesting, yet the ultrafast decay of plasmonic hot electrons often limits their direct and effective utilization in light-driven energy conversion. Devising suitable metal/semiconductor heterostructure by which to efficiently capture and separate the plasmonic hot carriers has undergone a resurgence of interest. Crucially, the extent to which the quality of interface can be achieved between plasmonic nanostructures and surrounding semiconductors will have direct influence in determining the efficiency of plasmonic hot electrons extraction and utilization. In this work, we present the first demonstration that spinodal decomposition offers a viable route to fabricate coherent metal/semiconductor heterostructures that enable highly efficient extraction of plasmonic hot electrons. By annealing a metastable Hf2O1-xN2 intermediate phase, spontaneous phase separation occurs, leading to the formation of a HfN/Hf2ON2 heterostructure. Owing to their similar cubic crystal structures and small lattice mismatch, HfN and Hf2ON2 form atomically coherent interfaces along multiple crystallographic orientation. This plasmonic heterostructure exhibits excellent light-harvesting capability with a broadband plasmonic absorption extending from visible further to the NIR region, and in particular it also permits direct and efficient transfer of hot electrons from the optically excited HfN to semiconducting Hf2ON2. We present the key evidence from both TA spectroscopy and PEC measurements, demonstrating efficient charge transfer across the coherent interface between plasmonic HfN and semiconducting Hf2ON2 with both PICTT and PHET pathways contributing to the hot electron transfer processes.

As an important class of liquid H2 carriers, methanol has attracted a recent resurgence of interest for its particular importance in distributed hydrogen production. To fulfill the on-demand fast H2 release from methanol, outside the traditional methanol steam reforming approach that requires relatively high temperatures (200-350 °C) and high pressure as well, the direct solar-driven H2 production from photocatalytic methanol decomposition at ambient temperature has also been pursued as a more sustainable and cost-effective alternative. However, although significant advances have been made so far, the reported photocatalysts that can achieve relatively high activity are mostly limited to wide bandgap semiconductors (e.g., TiO2) that have optical absorbance typically in the ultraviolet region. Importantly, achieving efficient full-spectrum solar-driven photocatalytic methanol decomposition requires effective light harvesting in the red and NIR regions which constitute a larger proportion of the solar spectrum. The coherent HfN/Hf2ON2 heterostructure reported in our present work is such a case in point. The major distinguishing features of HfN/Hf2ON2 heterostructure are its broadband plasmonic light harvesting and the remarkably high NIR activity for photocatalytic methanol decomposition, which provides a rational guideline for the full-spectrum solar energy utilization and photocatalysis towards high-efficiency H2 production.

Methods

Chemicals and reagents

Hafnium chloride (HfCl4, ≥98%), glucose (C6H12O6, ≥99%) and urea (CH4N2O, ≥99.0%) were purchased from Alfa Aesar Co., Ltd. Methanol (CH3OH, 99.9%), potassium ferricyanide (K3[Fe(CN)6], ≥99.5%), potassium chloride (KCl, ≥99.99%), sodium sulfate (Na2SO4, 99%) and deionized water were obtained from Beijing InnoChem Science & Technology Co., Ltd. Tetrapotassium Hexacyanoferrate Hydrate (K4[Fe(CN)6]‧3H2O, 98%) was purchased from Shanghai Macklin Biochemical Technology Co., Ltd. Platinum foil electrodes (Pt, 1 cm × 1 cm, 99.99%) and Ag/AgCl reference electrode (saturated KCl solution) were provided by Beijing Perfectlight Technology Co., Ltd. All chemicals used were of analytical grade and used without further purification.

Synthesis of HfO2 nanosheet precursor

The HfO2 nanosheets were synthesized using a one-pot glucose-blowing method with slight modifications to a previously reported procedure60. In a typical experiment, 500 mg of HfCl4, 5 g of glucose, and 6 g of urea were dissolved in 50 mL of deionized water under vigorous stirring. After vigorous stirring for 30 min, the homogeneous solution was heated to 140 °C for 8 h to form a porous foam. Then the mixture was calcined at 500 °C for 12 h in air to remove the glucose skeleton.

Synthesis of HfN, Hf2ON2 and the Hf2O1-xN2 intermediate phase

The compounds Hf2ON2, HfN, and the Hf2O1-xN2 intermediate phase were obtained starting from HfO2 at elevated temperatures with a mixture of NH3 and Ar flow. For pure HfN, the most vigorous reaction conditions were required: 1150 °C for 1 h with an NH3 flow rate of 1000 mL/min and an Ar flow rate of 1000 mL/min. The reactant Hf2ON2 was prepared at a relatively lower temperature of 1100 °C for 30 min with an NH3 flow rate of 500 mL/min and an Ar flow rate of 1000 mL/min. Based on the synthesis of the pure phases, the Hf2O1-xN2 intermediate phase was obtained through a two-step synthesis method. First, HfO2 was treated at 1100 °C for 30 min with an NH3 flow rate of 500 mL/min and an Ar flow rate of 1000 mL/min to obtain Hf2ON2. Then, the NH3 flow rate was increased to 1000 mL/min at 1100 °C for varying minutes to obtain the Hf2O1-xN2 intermediate phase, as shown in Supplementary Fig. 10a. We note that by changing the reaction time in the second step, different intermediate phases with varying x can be obtained.

Formation of coherent HfN/Hf2ON2 heterostructures via spinodal decomposition

The spinodal decomposition process was initiated by annealing the Hf2O1-xN2 intermediate phase at 1000 °C with Ar flow rate of 1000 mL/min. Annealing for different durations yields varying degree of phase separation. Usually, the phase separation will complete at 1000 °C annealing for 50 min. Despite the high-temperature nitridation and annealing process, the samples maintained their nanosheet morphology and avoided severe particle sintering. We used a mortar and pestle to grind the samples into individual particles for subsequent characterization.

Material characterizations

XRD measurements were performed on a Rigaku diffractometer using Cu-Kα irradiation, operated at 40 kV and 40 mA. SEM images were obtained using Hitachi SU8220 microscope. High-resolution Cs-corrected TEM images were acquired using a JEOL JEM-ARM300F (300 kV) equipped with a spherical aberration corrector at an accelerating voltage of 300 kV. To prepare samples for TEM imaging, nanomaterials were ultrasonically dispersed in ethanol, and a small volume of the suspension was drop-cast onto lacey carbon support films on copper grids, then air-dried before observation. UV-Vis-NIR absorption spectra were recorded using an Agilent Cary 5000 spectrophotometer. PL spectra were measured at room temperature on a Hitachi F-7000 fluorescence spectrometer with 350 nm excitation. The femtosecond TAS experiments were carried out on a Helios spectrometer, utilizing pump and probe pulses generated by a regenerative amplified Ti: sapphire laser system (Coherent, 25 fs, 4 mJ/pulse, 1 kHz repetition rate). In this work, transient absorption spectra spanning 350 to 800 nm were recorded following plasmon excitation with 650 nm or 700 nm pump pulses. Samples were prepared by dispersing finely milled powders into a PMMA solution, which was then spin-coated onto 1 cm×1 cm quartz substrate. The kinetic data for the HfN/Hf2ON2 heterostructure at wavelengths of 720 nm and 750 nm were analyzed using Surface Xplorer software.

Diffuse reflectance infrared Fourier transform spectroscopy measurements were performed using a custom-built accessory attached to a Bruker TENSOR II spectrometer, which is equipped with a liquid-nitrogen-cooled mercury-cadmium-telluride (MCT) detector. Electron paramagnetic resonance spectra were collected at room temperature on a Bruker EPR A-300 spectrometer. Free radicals generated during formaldehyde reforming were captured at ambient conditions using DMPO as a trapping agent, followed by recording the corresponding EPR signals of these trapped radicals.

Photoelectrochemical measurements

The photocatalyst ink was prepared by grinding 2.5 mg of the powder catalyst in a mortar, followed by the addition of 50 µL Nafion solution and 5 mL methanol. The mixture was ultrasonicated for 30 min and then drop-cast onto Indium Tin Oxide (ITO) glass with a geometric area of 1 cm2, resulting in a catalyst loading of 2.5 mg/cm2, calculated as the total catalyst mass divided by the electrode area. The coated electrode was then dried at room temperature in ambient air. Photoelectrochemical measurements were performed in a three-electrode glass cell connected to a Autolab PGSTAT 128 N Instruments. Data acquisition was performed using the Autolab PGSTAT 128 N potentiostat software. An Ag/AgCl (KCl saturated) electrode and platinum electrode were used as the reference and counter electrodes, respectively. Electrochemical impedance spectroscopy was carried out in 0.5 M Na2SO4 aqueous solution to evaluate the resistance of the electrochemical cell. The spectra were collected over a frequency range of 100 kHz to 0.1 Hz at open-circuit potential. The total resistance (Rs + Rct) was extracted by fitting the Nyquist plot with an equivalent circuit model and was determined to be XX Ω.

Photocurrent–time measurements were conducted at a fixed potential of 0.27 V (vs Ag/AgCl) under chopped monochromatic light illumination. The electrolyte used was an aqueous solution containing 50 mM K3[Fe(CN)6], 50 mM K4[Fe(CN)6], and 0.5 M KCl. The electrolyte was freshly prepared before each measurement by dissolving the reagents in deionized water and stirring until fully dissolved. It was used immediately without storage to minimize degradation, and the total volume was approximately 40 mL. The pH of the prepared solution was measured to be 7.3 ± 0.1. For fixed wavelength of the photocurrent measurement, bandpass filters with a bandwidth of less than 10 nm at the full width at half-maximum were used. All the electrochemical tests were performed at room temperature.

Photocatalytic H2 generation from methanol solution

Pt-decorated HfN/Hf2ON2 (Pt-HfN/Hf2ON2) was synthesized via a commonly used photo-deposition strategy, which has been widely adopted for noble metal loading in previous studies61,62,63,64. Typically, 50 mg of HfN/Hf2ON2 nanoparticles were ultrasonically dispersed in a 100 mL H2O/CH3OH mixture (90 mL H2O and 10 mL CH3OH). Then, a measured volume of H2PtCl6·6H2O solution was added under continuous stirring. The suspension was then irradiated with 365 nm UV light for 2 h at ambient temperature to photo-reduce the Pt precursor onto the nanostructure surface. After irradiation, the solid was separated by vacuum filtration, washed thoroughly with deionized water, and dried for subsequent use.

The photocatalytic hydrogen evolution reactions were conducted in a sealed gas circulation setup (MC-SPH20, Marry Change Co., Ltd). In a typical run, 50 mg of photocatalyst was dispersed in 150 mL of a methanol-water solution. Before illumination, the system was evacuated multiple times to ensure complete removal of residual air. The photocatalytic reaction was initiated by irradiating the suspension with a 300 W Xe lamp (MC-XUV300, Marry Change Co., Ltd). Light intensity was adjusted to approximately 100 mW cm−2 (λ > 400 nm or λ > 800 nm) using a Xenon lamp or 100 mW cm−2 under simulated sunlight (AM 1.5 G). To maintain thermal stability during the reaction, the suspension temperature was controlled at 25 °C via circulating cooling water. The generated hydrogen was quantified using gas chromatography (GC) equipped with a thermal conductivity detector (TCD) and a 5 Å molecular sieve column, with argon as the carrier gas. AQY for H2 evolution was measured under monochromatic light using band-pass filters, and calculated as follows:

$${AQY}={N}_{e}/{N}_{p}\times 100 \%=2M/{N}_{p}\times 100\%$$

where Ne is the amount of reaction electrons, Np is the incident photons, M is the amount of H2 molecules.