Atomic‐Level Synergy of Dual Single‐Atom Catalysts for Photocatalytic Hydrogen Evolution Reaction

ABSTRACT

Dual‐atom catalysts (DACs) have emerged as promising candidates for various chemical transformations with excellent atom utilization and synergistic effects between adjacent metal sites. However, their controlled synthesis and detailed understanding of cooperative effects remain challenging. Here, we design Ag‐Cu dual sites embedded in a g‐C₃N₄ matrix (AgCu‐CN) through a supramolecular self‐assembly approach followed by thermal polymerization by pyrolysis. The atomically engineered catalyst exhibits a hydrogen evolution rate of 2126 µmol g⁻¹ h⁻¹, and an apparent quantum yield (AQY) of 20% at 400 nm, surpassing the other reported metal‐N coordinated photocatalysts. X‐ray absorption spectroscopy (XAS) confirms the atomic‐level dispersion and coordination with the g‐C₃N₄ framework of the Ag and Cu single atomic sites. Comprehensive characterizations including transient absorption (TA) spectroscopy and theoretical calculations based on density functional theory demonstrate that the presence of the two metal centers broadens the photoabsorption range, enhances density of states close to the Fermi level. Thus we posit that it promotes excited state electron transfer and charge separation, and facilitate H₂O activation by directing electron migration toward the protonation site, thereby stabilizing the H^* intermediate, a crucial step in hydrogen evolution reaction. The catalysts developed in this study exhibit excellent activity, stability, and cost‐effectiveness, highlighting their strong potential for practical clean hydrogen production.

Atomically dispersed Ag‐Cu dual single‐atom sites embedded in g‐C₃N₄ (AgCu‐CN) enable efficient photocatalytic hydrogen evolution, achieving 2126 µmol g⁻¹ h⁻¹ and 20% AQY at 400 nm. The synergistic interaction between Ag and Cu promotes charge separation and excited‐state reactivity, affording a stable and cost‐effective platform for scalable solar hydrogen generation.

1. Introduction

Environmental pollution and the high carbon emissions generated by fossil fuel use remain major barriers to building a sustainable future. Therefore, the transition to clean, renewable, and eco‐friendly energy sources that produce no carbon emissions is now more critical than ever [1, 2, 3]. Amazingly, the solar energy reaching the Earth in only 1 h could fulfil the world's entire energy requirements for a year [4]. Harnessing abundant solar energy offers a promising route to tackle global energy and environmental challenges. Hydrogen, with its high energy density and eco‐friendly nature, is a leading clean fuel. Among its production methods, solar‐driven photocatalytic water splitting stands out as a sustainable and efficient approach enabled by advanced photocatalysts [5, 6, 7]. Polymeric graphitic carbon nitride (g‐C₃N₄), for its nitrogen‐rich framework, environmental friendliness, high stability, strong visible‐light absorption, and low cost, has emerged as a highly promising and sustainable photocatalyst. Importantly, the porous architecture of g‐C₃N₄ nanotubes creates a distinctive confinement effect, localizing active species at the nanoscale and enhancing catalytic efficiency [8, 9, 10]. However, g‐C₃N₄ lacks well‐defined and efficient active sites on its delocalized π‐conjugated C‐N framework, which results in sluggish surface reaction kinetics. Introducing co‐catalysts onto its surface has been shown to effectively enhance photocatalytic performance by creating specific reduction sites that probably act as the active centers [11, 12, 13].

Recent developments have demonstrated the outstanding potential of single‐atom catalysts (SACs) across diverse catalytic applications, primarily due to their ability to optimize metal atom utilization by dispersing metal clusters and nanoparticles into individual atoms [14, 15, 16]. In photocatalytic water splitting, employing single‐atom catalysts (SACs) demonstrate outstanding performance due to their high density of accessible active sites, ability to lower the energy barriers, and their effectiveness in suppressing the recombination of photogenerated charge carriers, thereby enabling more efficient hydrogen production from water [17, 18, 19]. g‐C₃N₄ has high nitrogen content and abundant lone pairs of electrons, making it an excellent support for SACs, facilitating strong coordination with metal atoms [20, 21]. Among various metals, Ag is notable for its low work function, which promotes rapid electron transfer and efficient separation of photogenerated charge carriers, thereby enhancing photocatalytic performance [22]. For instance, Jiang et al. synthesized highly active and stable silver SACs anchored on carbon nitride with a well‐defined Ag‐N₂C₂ coordination structure (Ag‐N₂C₂/CN which enhances hydrogen evolution performance by broadening light absorption and promoting efficient charge transfer [23]. Accordingly, Li et al. reported that single‐atom Ag‐doped g‐C₃N₄ (SA Ag‐CN) exhibits excellent stability and efficiency for photocatalytic H₂ evolution, owing to favorable hydrogen adsorption energy and Ag─N bond formation [24]. Moreover, Cu, a 3d transition metal, possesses variable oxidation states and offers a substantially lower cost compared to noble metals, rendering it a highly efficient and economically viable cocatalyst for enhancing photocatalytic performance. Hence recently, Luo et al. reported a single‐atom Cu‐bridged graphitic carbon nitride (SA‐Cu‐N₃C₁) photocatalyst with nitrogen vacancies, which enhanced interlayer charge transport and promoted efficient charge separation. Under visible light, it achieved a hydrogen evolution rate of 605.15 µmol g⁻¹ h⁻¹ without a Pt cocatalyst [9].

Recent studies have shown that introducing additional single‐atom metal sites to form dual‐metal sites can effectively tune the electronic structure and generate strong synergistic effects, thereby enhancing photocatalytic performance. In comparison to SACs, dual‐atom catalysts (DACs) maintain the benefits of atomically dispersed metals while offering superior catalytic activity, owing to their distinctive bimetallic configurations, cooperative interactions, and well‐defined coordination environments [25, 26, 27, 28]. Several studies reported that DACs such as InCu/PCN [29], MnCo/CN [11], CoAg‐PCN [22], PtCo/CN [30], and CuNi/CN [31] have superior photocatalytic performance compared to SACs. Despite reports on DACs, developing efficient dual‐atom catalysts on g‐C₃N₄ for enhanced photocatalytic hydrogen evolution remains largely unexplored. Furthermore, theoretical analysis of proton reduction was mostly based on ground‐state calculations, whereas the experimental conditions require the consideration of appropriate charge‐transfer excited states.

Herein, we report a facile and cost‐efficient template‐free approach to build Ag‐Cu dual metal sites on hexagonal g‐C₃N₄ nanostructures through hydrothermal treatment and subsequent calcination. The significant improvement in photocatalytic activity arises from the synergistic combination of the g‐C₃N₄ framework and the cooperative interactions between the Ag‐Cu dual sites. This architecture not only provides abundant active sites but also facilitates efficient charge separation, markedly enhancing proton reduction kinetics. Excited‐state calculations based on Time‐Dependent Density Functional Theory (TDDFT) and constrained DFT (cDFT) further demonstrate directed charge transfer from g‐C₃N₄ to the electron‐deficient Ag center upon proton transfer from water. The presence of Cu centers provides additional occupied states near the Fermi level, thus validating the synergistic effect. These findings underscore the potential of Ag‐Cu dual‐site engineering on g‐C₃N₄ as an effective platform for designing high‐performance, visible‐light‐responsive photocatalysts.

2. Results and Discussion

2.1. Synthesis and Structural Characterization

Atomically dispersed bimetallic Ag and Cu dual sites anchored on tubular g‐C₃N₄ (AgCu‐CN) were synthesized via a hydrothermal treatment followed by thermal polymerization of supramolecular precursors, as illustrated in Figure 1a. The rod‐like supramolecular precursor was obtained through the hydrolysis of melamine and cyanuric acid, which subsequently underwent a self‐assembly process. (Figures S1 and S2). The morphology and microstructure of AgCu‐CN were further investigated using scanning electron microscopy (SEM), field emission scanning electron microscopy (FESEM), transmission electron microscopy (TEM), and aberration‐corrected high‐angle annular dark field scanning transmission electron microscopy (AC‐HAADF‐STEM). The AgCu‐CN sample displays a characteristic tubular morphology (Figure 1b–d), similar to that of pristine CN (Figures S3 and S4), indicating that the incorporation of Ag and Cu does not alter the overall structure. The corresponding selected area electron diffraction (SAED) pattern (inset in Figure 1d) confirms the absence of a significant number of crystalline phases in AgCu‐CN. Furthermore, the bright spots in the AC‐HAADF‐STEM image (Figure 1e,f) confirm that Ag and Cu are predominantly present as SAs dispersed within the carbon nitride matrix. The HAADF‐STEM images of Cu‐CN and Ag‐CN are given in Figure S5. Furthermore, energy‐dispersive X‐ray (EDX) elemental mapping demonstrates the uniform distribution of Ag and Cu within the CN matrix. This homogeneous dispersion verifies the successful incorporation of Ag‐Cu dual sites in AgCu‐CN, as illustrated by the aberration‐corrected HAADF‐STEM‐EDX mapping in Figure 1g–i.

FIGURE 1.

(a) Schematic illustration of the synthesis process of AgCu‐CN; (b) SEM image; (c) FESEM image; (d) TEM image; (e, f) AC‐HAADF‐STEM images of AgCu‐CN; (g–i) EDX elemental mapping images of Ag, Cu, and N, respectively.

The XRD patterns (Figure 2a) show characteristic peaks at 27.5° and 13.3°, corresponding to the (002) and (001) planes of CN. In Ag‐CN, Cu‐CN, and AgCu‐CN, only the 27.5° peak is prominent, indicating that Ag and Cu doping disrupts the regular stacking of tri‐s‐triazine units. No metallic Ag or Cu diffraction peaks are observed, suggesting their high dispersion in the CN matrix. Additionally, the (002) peak slightly shifts from 27.5° to 27.6° (Figure S6), likely due to the incorporation of Ag and Cu single atoms into the sixfold cavities of the CN framework, affecting the stacking arrangement [8]. The fourier transform infrared (FT–IR) spectra of AgCu‐CN exhibit characteristic vibrational features associated with CN structures (Figure S7). Inductively coupled plasma optical emission spectroscopy (ICP‐OES) analysis revealed that the AgCu‐CN sample contains 0.66 wt.% Ag and 0.57 wt.% Cu, as summarized in Table S1, and the percentage of C, N, and H was determined by CHNS analysis, as given in Table S2. The solid‐state ¹³C NMR spectra of AgCu‐CN exhibit two distinct peaks, which correspond to the characteristic of N═C─N(NH₂) and N═C─N₂; these results suggest that AgCu‐CN possesses a melon‐based carbon nitride structure (Figure S8) rather than a fully condensed, hydrogen‐free tris‐s‐triazine based carbon nitride [32].

FIGURE 2.

(a) PXRD pattern of the as‐synthesized catalyst, (b) high‐resolution XPS spectra of Ag in Ag‐CN and AgCu‐CN, (c) high‐resolution XPS spectra of Cu in Cu‐CN and AgCu‐CN; (d) Cu K‐edge XANES spectra of Cu foil and AgCu‐CN, (e) k²‐weighted Fourier‐transform Cu K‐edge EXAFS spectra of AgCu‐CN, (f) Cu K‐edge EXAFS fitting curves of the AgCu‐CN, (g) Ag K‐edge XANES spectra of Ag foil and AgCu‐CN, (h) k²‐weighted Fourier‐transform Ag K‐edge EXAFS spectra of AgCu‐CN, and (i) Ag K‐edge EXAFS fitting curves of the AgCu‐CN.

X‐ray photoelectron spectroscopy (XPS) was employed to investigate the surface composition and chemical states of the as‐synthesized samples. The survey spectrum distinctly confirms the presence of C, N, Ag, and Cu elements (Figure S9), consistent with the EDX analysis. The high‐resolution C 1s spectrum (Figure S10) can be deconvoluted into three prominent peaks at binding energies of 284.49, 287.84, and 293.53 eV, corresponding to C─C bonds, sp²‐hybridized N═C─N groups within the heptazine framework, and surface charging effects, respectively [33]. Similarly, the high‐resolution N 1s spectrum (Figure S11) exhibits two well‐resolved peaks at 398.38 and 400.70, attributed to pyridinic‐N and pyrrolic‐N species, respectively. The dominant intensity of the pyridinic‐N component underscores its crucial role as a coordination site, facilitating the formation of well‐defined Ag and Cu coordination environments within the catalyst framework [34]. The high‐resolution Ag 3d spectrum (Figure 2b) of Ag‐CN reveals peaks at 367.53 and 373.84 eV, corresponding to Ag 3d_5/2 and Ag 3d_3/2, confirming that silver is positively charged. In comparison, the Ag 3d peak in AgCu‐CN shifts toward lower binding energy, indicating a reduction in the Ag valence state resulting from the synergistic interaction between Ag and Cu atoms [35]. The high‐resolution Cu 2p XPS spectrum of Cu‐CN shows two prominent peaks located at 932.07 and 951.06 eV, attributed to Cu 2p_3/2 and Cu 2p_1/2, respectively (Figure 2c). This suggests that copper exists as Cu⁺¹/Cu⁰. In comparison to the Cu SAC, the Cu 2p peaks in the AgCu DAC shift toward higher binding energies, implying a transfer of electrons from Cu to Ag within the AgCu DAC framework [36, 37].

X‐ray absorption spectroscopy (XAS) was utilized to confirm the atomic nature of the Ag‐Cu dual‐atom catalyst. The X‐ray absorption near edge spectra (XANES) collected at the Cu K‐edge indicated that Cu is in the +1 oxidation state from the intensity of the white line (Figure 2d), with a sharp pre‐edge peak indicating that the Cu sites are in a non‐centrosymmetric position inside the g‐C₃N₄ framework [38]. The Fourier transformed extended X‐ray absorption fine structure (FT‐EXAFS) analysis showed that there is a predominant Cu‐N shell present at a distance of 1.5 Å (Figure 2e). A minor feature at 2.2 Å was at first believed to be from the neighbouring Cu atom; however, it does not fit well with a Cu–Cu shell. Therefore, this feature is due to the scattering of the highly ordered C₃N₄ framework. The first shell fitting results (Figure 2f; Table S3) showed that Cu is surrounded by two nitrogen atoms. The XANES spectra carried out at the Ag K‐edge showed that Ag bears a positive charge, as the white light intensity is slightly higher than metallic Ag (Figure 2g). The EXAFS analysis showed Ag‐N as predominate first shell with a minor feature at 2.5 Å (Figure 2h). The first shell fitting showed that Ag is connected to two nitrogen atoms of the g‐C₃N₄ framework as shown in Figure 2i and Table S4. These results confirm that the majority of the Cu and Ag atoms are present as single atoms in the C₃N₄ framework.

To further elucidate the paramagnetic properties of the samples, the electron paramagnetic resonance (EPR) spectra of CN, Cu‐CN, Ag‐CN, and AgCu‐CN are shown in Figure 3a. All samples exhibit a prominent resonance signal with a g‐value of approximately 2.003, which is typically associated with unpaired electrons trapped on carbon atoms within the CN framework. Compared to pristine CN, the metal‐modified samples (Ag‐CN, Cu‐CN, and AgCu‐CN) display significantly stronger EPR signals, indicating an increased density of unpaired electrons introduced by the incorporation of Ag and Cu species. Among the modified catalysts, AgCu‐CN demonstrates the highest signal intensity, suggesting a synergistic electronic coupling between the dual metal sites. This result implies more efficient electron delocalization and enhanced charge transfer between the metal centers and the CN matrix [26]. Raman spectra of CN, Cu‐CN, Ag‐CN, and AgCu‐CN recorded under 785 nm excitation (Figure S13) show similar features with a prominent band at ∼1426 cm⁻¹, characteristic of the amorphous CN framework. No significant spectral changes are observed after Ag and Cu incorporation, although a noticeable decrease in Raman intensity, particularly for AgCu‐CN, indicates strong electronic and structural interactions between the metal sites and the CN framework, slightly affecting the vibrational behavior [35].

FIGURE 3.

(a) EPR spectra, (b) UV–vis diffuse reflectance spectra, (c) XPS valence band spectra, (d) schematic diagram of energy band structure, e) photoluminescence (PL) spectra, and (f) time‐resolved PL spectra of CN, Cu‐CN, Ag‐CN, and AgCu‐CN.

2.2. Optical Properties

The band structures and optical properties of CN, Cu‐CN, Ag‐CN, and AgCu‐CN were systematically examined using UV–vis diffuse reflectance spectroscopy (UV‐DRS) and valene band XPS (VB‐XPS) analysis. As illustrated in Figure 3b, the AgCu‐CN sample exhibits markedly enhanced absorption in the visible‐light region, indicative of improved photon‐harvesting efficiency. This enhancement is likely associated with band‐structure tuning arising from the simultaneous incorporation of Ag and Cu species. Using the Kubelka‐Munk function, the optical bandgap energies of CN, Cu‐CN, Ag‐CN, and AgCu‐CN were determined to be 2.76, 2.60, 2.55, and 2.55 eV, respectively (Figure S14). The valence band potentials (E_VB) of CN, Cu‐CN, Ag‐CN, and AgCu‐CN were obtained from the VB‐XPS spectra (Figure 3c) and determined to be 2.32, 2.08, 1.92, and 1.73 eV, respectively. Furthermore, Mott‐Schottky analysis was conducted to evaluate the flat‐band potentials (E_fb) of the synthesized samples.

As shown in Figure S15, all samples exhibit positive slopes in the Mott–Schottky plots, confirming their n‐type semiconductor nature. The E_fb of CN, Cu‐CN, Ag‐CN, and AgCu‐CN are −0.54, −0.62, −0.73, and −0.87 V vs. Ag/AgCl (pH 7), corresponding to −0.34, −0.42, −0.53, and −0.67 V vs. NHE, respectively. The E_fb of an n‐type semiconductor is typically 0.1 V more positive than the conduction band potential (E_CB) [12]. The E_CB values for CN, Cu‐CN, Ag‐CN, and AgCu‐CN were estimated as −0.44, −0.52, −0.63, and −0.77 V (vs. NHE at pH 7). These values, summarized in Table S5, were combined with the E_VB potentials and optical bandgaps to construct the full band structures of the photocatalysts, as shown in Figure 3d. Notably, AgCu‐CN possesses a more negative E_CB potential than both Cu‐CN and Ag‐CN, indicating a stronger thermodynamic driving force for proton reduction. These results indicate that AgCu‐CN possesses a favourable band alignment for efficient visible‐light‐driven hydrogen evolution. Steady‐state photoluminescence (PL) spectra recorded at an excitation wavelength of 350 nm were used to evaluate charge separation efficiency. As shown in Figure 3e, pristine CN exhibits strong PL emission, indicating rapid charge recombination. In contrast, Ag‐CN and Cu‐CN show reduced PL intensity due to the introduction of single metal atoms. Notably, AgCu‐CN displays the lowest PL intensity, demonstrating that dual metal sites significantly enhance the separation of photogenerated electron‐hole pairs. Furthermore, the charge transfer kinetics of the materials were examined using time‐resolved photoluminescence (TRPL), as shown in Figure 3f. The average lifetimes (τ_ave) for CN, Cu‐CN, Ag‐CN, and AgCu‐CN were calculated to be 5.71, 4.82, 4.52, and 3.44 ns, respectively (Table S6). Compared to pure CN and its single‐metal counterparts, the significantly reduced lifetime observed in AgCu‐CN indicates more rapid fluorescence quenching. This accelerated quenching typically results from enhanced charge separation or increased nonradiative decay pathways. The shorter lifetime of AgCu‐CN indicates more efficient transfer of photogenerated excitons. The Ag‐Cu dual sites create effective charge transfer pathways, suppressing recombination and directing more electrons to the active sites for photocatalysis. To further understand the enhanced electron trapping ability of Ag and Cu single atoms, the electron transfer rate constant (K_et) was calculated using Equations (1)–(3) [33].

$$K_{et\left(CN\to AgCu-CN\right)}=\frac{1}{{\tau }_{AgCu-CN}}-\frac{1}{{\tau }_{CN}}$$ (1)

$$K_{et\left(CN\to Ag-CN\right)}=\frac{1}{{\tau }_{Ag-CN}}-\frac{1}{{\tau }_{CN}}$$ (2)

$$K_{et\left(CN\to Cu-CN\right)}=\frac{1}{{\tau }_{Cu-CN}}-\frac{1}{{\tau }_{CN}}$$ (3)

AgCu‐CN shows a higher emission quenching rate (0.12 ns) than Ag‐CN (0.05 ns) and Cu‐CN (0.03 ns), indicating superior electron transfer capability and efficient charge separation facilitated by the Ag‐Cu dual sites.

To further explore the effect of the dual atoms on the excited charge transfer characteristics of CN, we have conducted a state‐of‐the‐art femtosecond transient absorption (TA) spectroscopy for pristine CN and AgCu‐CN. Figure 4a represents the TA spectra of CN at different pump‐probe delay times after photoexcitation at 320 nm. The TA spectrum of CN exhibits a broad photoinduced bleach signal over the entire spectral range, having two peak maxima at ∼392 and ∼480 nm. The red region peak (∼480 nm) originated due to the state filling of band edge states (BS), whereas the blue region peak (∼392 nm) was assigned to the high‐energy state (HS) transition [39]. In the CN system, the lower effective mass of electrons compared to holes leads to their dominant contribution to the transient bleach signals [40]. Therefore, both the bleach signals (BS and HS) originated from state filling of the conduction band (CB) states. Next, we monitored the TA kinetics for pristine CN at HS (∼392 nm) and BS (∼480 nm). Both kinetics were fitted with single exponential growth and triexponential decay components, as tabulated in Table S7. As shown in Figure 4c, the growth component for BS (0.22 ps) was found to be higher than HS (< 0.1 ps). It is quite obvious that the 320 nm (3.88 eV) pump excites the electrons in the higher energy states of the CB, which first cool down to the HS, then ultimately thermalized to the BS, as schematically shown in Figure S16. The decay kinetics of both bleach signals were fitted using a triexponential function with time constants of 0.65, 9.7 ps, and >1 ns for HS, and 2.27, 30.5 ps, and >1 ns for BS. The initial two decay time constants, 0.65 and 9.7 ps for HS and 2.27 and 30.5 ps for BS, were attributed to the electron trapping in two different shallow trap states (short and long), as designated in the earlier reports [41]. The longest decay constant (>1 ns) was assigned to the defect‐mediated electron‐hole recombination [39].

FIGURE 4.

Transient absorption (TA) spectra of the (a) CN and (b) AgCu‐CN at different pump‐probe time delays after the photoexcitation at 320 nm. (c) Comparative normalised TA kinetics of CN monitored at 392 nm (High energy state; HS) and 480 nm (Band edge energy state; BS). Comparative normalised TA kinetics of CN and AgCu‐CN, monitored at (d) 392 nm (High energy state; HS) and (e) 480 nm (Band edge energy state; BS). KPFM image of AgCu‐CN (f) in dark and (g) under light illumination. (h) CPD profiles under dark and illuminated conditions.

Next, we performed the transient absorption measurements for the AgCu‐CN under similar experimental conditions to investigate the ultrafast relaxation pathways in the dual single‐atom catalyst. The TA spectrum of AgCu‐CN (Figure 4b) showcased similar spectral features as pristine CN, but the TA kinetics monitored at both states (HS and BS) were significantly modulated in the presence of dual single atoms. The TA kinetics of both signals (HS and BS) were found to be accelerated in AgCu‐CN as compared to CN, as shown in Figure 4d,e, respectively. For AgCu‐CN, the decay time constants were found to be 0.14, 2.8 ps, and > 1 ns for HS and 1.1, 23.9 ps, and > 1 ns for BS, as tabulated in Table S7. These faster kinetics for the AgCu‐CN confirm the ultrafast hot electron as well as band‐edge electron transfer from the conduction band of the CN to Ag‐Cu sites [36].

The charge transfer kinetics correlate with the photocatalytic hydrogen evolution performance, indicating that Ag‐Cu dual atomic sites enhance charge transfer and suppress recombination, thereby improving solar‐driven water splitting. Kelvin Probe Force Microscopy (KPFM) was employed to analyse the surface potential of the synthesized AgCu‐CN. Figure 4f,g displays the KPFM images acquired under dark and illuminated conditions, respectively. Upon light irradiation, a distinct decrease in surface potential was observed. This negative shift signifies the accumulation of photogenerated electrons on the AgCu‐CN surface, indicating that electrons preferentially accumulate over photogenerated holes [25]. The comparative surface potential profiles are shown in Figure 4h.

2.3. Photocatalytic Hydrogen Evolution Performances

The photocatalytic HER activities of the synthesized photocatalysts were assessed in an aqueous TEOA solution under visible‐light irradiation, without employing any noble‐metal cocatalysts. Among them, pristine CN exhibited the lowest hydrogen evolution activity (72 µmol g⁻¹ h⁻¹), indicating its poor photocatalytic performance, likely due to insufficient active sites and severe intrinsic charge recombination. Upon the incorporation of single atoms, Ag‐CN exhibits a higher photocatalytic H₂ evolution rate than Cu‐CN. This indicates that Ag single‐atom sites are more effective than their Cu counterparts. Furthermore, by optimizing the Ag and Cu molar ratio in AgCu‐CN, the Ag₁Cu₂‐CN sample achieved the highest photocatalytic performance, delivering a hydrogen production rate of 2126 µmol g⁻¹ h⁻¹, which is 29.5 times higher than that of pristine CN. This remarkable enhancement can be attributed to the synergistic effect of Ag and Cu single atoms in regulating electron transfer. The hydrogen production rate of the photocatalysts is presented in Figure 5a, while Figure 5b illustrates the time‐dependent hydrogen evolution over different catalysts for a 4 h period. The HER performance of the photocatalysts is summarized in the following order: Ag₁Cu₂‐CN> Ag₂Cu₁‐CN> Ag₁Cu₁‐CN> Ag₃Cu₂‐CN> Ag‐CN>Cu‐CN>CN. In the case of higher loading of SACs, the catalytic activity decreases due to the accumulation of SACs within the pores of CN, which leads to blockage of its surface and hinders active site accessibility [42]. Relative to reported g‐C₃N₄‐based photocatalysts, AgCu‐CN shows enhanced efficiency (Table S8). The hydrogen evolution of AgCu‐CN and its physically mixed counterparts is also presented in Figure S17. The AQY for photocatalytic H₂ evolution was analyzed to determine the solar energy conversion efficiency of AgCu‐CN. As shown in Figure 5c, AgCu‐CN exhibits a high AQY of 20% at 400 nm. Even at 550 nm, the AQY remains at 4%, indicating efficient light absorption and utilization across a wide wavelength range. The hydrogen evolution rate varies with light intensity, as shown in Figure S18. Light‐off stops H₂ generation immediately, and light‐on restores it (Figure 5d), confirming the reaction relies on light. Beyond its high photocatalytic H₂ evolution activity, AgCu‐CN also shows excellent durability. The H₂ production rate remains nearly unchanged over eight consecutive cycles, demonstrating outstanding stability during photocatalytic water splitting (Figure 5e). This stability is further affirmed by its steady H₂ output throughout 20 h of uninterrupted operation (Figure 5f). A slight decline in hydrogen production was observed in the third cycle due to the depletion of TEOA; however, the activity recovered to its original level upon the addition of 3 mL of fresh TEOA. Notably, AgCu‐CN maintained consistent hydrogen evolution performance over five consecutive cycles. Moreover, post‐reaction XRD, FTIR, SEM, TEM, and XPS spectra (Figures S19–S23) revealed no significant changes in the microstructure, confirming the excellent structural stability and sustained activity of AgCu‐CN.

FIGURE 5.

(a) Photocatalytic H₂ generation rates of various catalysts. (a–g): CN, Cu‐CN, Ag‐CN, Ag₃Cu₂‐CN, Ag₁Cu₁‐CN, Ag₂Cu₁‐CN and Ag₁Cu₂‐CN. (b) Time‐dependent hydrogen evolution over different catalysts. (c) Quantum efficiency of Ag₁Cu₂‐CN under different wavelengths. (d) Amount of H₂ produced over Ag₁Cu₂‐CN via photocatalytic water splitting under Xe lamp on‐off condition. (e) Recycling tests (1 h per cycle) of photocatalytic H₂ generation over Ag₁Cu₂‐CN. (f) The time course of H₂ generation up to 20 h for 5 runs by Ag₁Cu₂‐CN (Reaction conditions: 420 W Xe lamp, 420 nm wavelength, 10% TEOA aqueous solution, 20 mg catalyst). (g) Photocurrent responses, (h) EIS plots, and (i) LSV plot of CN, Cu‐CN, Ag‐CN, and AgCu‐CN.

A series of controlled experiments were carried out under identical reaction conditions to evaluate hydrogen evolution in the absence of light, catalyst, and TEOA. The results clearly indicate that no hydrogen is produced unless all three components‐light, catalyst, and TEOA‐are simultaneously present, as shown in Figure S24. Furthermore, a TEOA concentration of 10% v/v was found to deliver the best performance (Figure S25). Furthermore, the hydrogen evolution rates of AgCu‐CN with various sacrificial electron donors (SEDs) are shown in Figure S26, revealing that TEOA delivers the highest hydrogen evolution rate among all the tested SEDs.

Moreover, photoelectrochemical (PEC) measurements were conducted to investigate the photogenerated charge carrier dynamics of the photocatalysts. As shown in the chronoamperometric current–time (I–t) curves (Figure 5g), AgCu‐CN exhibits a much higher photocurrent density than Ag‐CN, Cu‐CN, and pristine CN, indicating superior charge separation efficiency. The interfacial charge transfer was further examined using electrochemical impedance spectroscopy (EIS). As depicted in Figure 5h, AgCu‐CN shows the smallest semicircle arc in the Nyquist plot, signifying the lowest charge‐transfer resistance and fastest charge migration among the samples. Linear sweep voltammetry (LSV) measurements (Figure 5i) further reveal that AgCu‐CN achieves a higher current density than the other samples, attributed to efficient charge separation enabled by the well‐integrated Ag‐Cu dual active sites within the CN framework. Overall, these PEC results demonstrate that the synergistic effect of Ag and Cu dual‐atom sites in AgCu‐CN promotes charge separation and transfer, enhancing charge utilization efficiency.

2.4. Proposed Mechanism

Based on the above experimental observations, the following photocatalytic mechanism is proposed for the AgCu‐CN system (Figure 6a). Under visible‐light irradiation, electron‐hole pairs are generated within the CN matrix, primarily at the nitrogen sites [22]. The photogenerated holes (h⁺) are scavenged by the sacrificial agent TEOA, while the e⁻ migrate toward the carbon atoms and participate in the HER. In the dual atom catalyst system, the introduction of Ag and Cu atoms significantly alters the charge transfer dynamics. Spectroscopic Characterization and DFT results confirm that Ag centers serve as the primary electron acceptor sites. The presence of Cu centers, however, plays a crucial role in modulating the electronic environment around the Ag sites. The electronic interaction between Ag and Cu induces local charge redistribution, which accelerates electron transfer and facilitates HER. This cooperative behavior between Ag and Cu atom doping effectively demonstrates the potential of dual‐atom catalysts in improving the photocatalytic performance of g‐C₃N₄‐based systems.

FIGURE 6.

(a) Schematic illustration of the photocatalytic H₂ evolution mechanism in AgCu‐CN. (b) Projected density of states (PDOS) in the absence and presence of the Cu center. (c) Band structure of Ag‐CN. (d) Band structure of AgCu‐CN.

3. Theoretical Analysis of Excited State Proton Transfer

The band structure calculations in the absence and presence of the Cu center (Figure 6c,d) as well as the corresponding projected density of states (Figure 6b) clearly demonstrate that the presence of the Cu atom adds additional occupied states near the Fermi level, thus providing easily accessible electron density, which can be utilized for catalysis. Furthermore, we developed a methodology based on time‐dependent DFT (TDDFT) and constrained DFT (cDFT) calculations to (i) model the charge distribution in the excited state, and (ii) estimate whether the proton transfer from water to the substrate is facilitated in the excited state.

3.1. Proton Transfer in the Ground State

The model structures employed for the calculation of the proton transfer event are shown in Figure 7a. The details of how these structures were generated are provided in Figure S27. The energy of the hydroxide structure is estimated with cDFT by imposing −1.2|e| and +0.2|e| charges on the oxygen site and the associated proton center, respectively. Due to the absence of additional water molecules that can explicitly H‐bond to the hydroxide ion, this complex is 32.38 eV higher in energy than the corresponding water complex. Even though this energy difference is a gross overestimation, it does indicate that the proton transfer from water is highly disfavoured in the ground state. Furthermore, the difference in the charge distribution (Hirshfeld atomic charges) between the water and the hydroxide complexes (shown as a heatmap in Figure 7b) suggests that the Ag center gains substantial electron density upon the proton transfer (See Tables S9 and S10 for the individual Hirshfeld charges corresponding to the water complex and the hydroxide structure, respectively).

FIGURE 7.

(a) Models for estimating the energetics of proton transfer from water to a Ag center (b) Difference in Hirshfeld atomic charges between the ground states after and prior to the proton transfer (c) Key MOs involved in the (the oscillator strengths of the first and the third excited states are much smaller than that of the second excited state) excited state as obtained from TDDFT calculation (d) Difference in Hirshfeld atomic charges between (I) the excited state and the ground state prior to proton transfer (II) the excited states after and prior to the proton transfer. The charges on the external water molecule, either as H₂O or as a combination of a hydroxide ion and a proton, are not included in the charge density difference analysis. Note that the range in the heatmap (d) is the same as in (b).

3.2. Excited State Charge Distribution

To model the excited state charge distribution, we first computed the probable electronic transitions in the water complex with TDDFT. Our calculations indicate that among the low‐lying excited states, the second one, which primarily involves charge transfer from the HOMO to the LUMO and the LUMO+1 orbitals (Figure 6c), has the highest oscillator strength (Figure S28). Based on this observation, we developed a methodology involving cDFT to model the charge distribution in the excited state. In this methodology, we first estimated the partial charge on each atom, considering that one electron is transferred from the HOMO to a combination of the LUMO and the LUMO+1, where the relative contributions of these unoccupied orbitals are calculated from the transition amplitudes in the second excited state as obtained from the TDDFT calculations (Table S11). Then all the atoms are divided into two groups based on whether their partial charge is positive or negative. These atom groups and their collective charges are used as constraints in a subsequent cDFT calculation to represent the excited state. The difference in the charge distribution between this excited state and the ground state is depicted in Figure 7dI. It can be clearly seen that the charge‐transfer excited state results in an increase in partial negative charge on the Ag atom.

3.3. Excited State Proton Transfer

Having generated a reasonable protocol to calculate the charge‐separated excited state, we aim to estimate whether it is energetically more feasible to transfer the proton in the excited state as compared to the ground state. In order to compute the energy of the hydroxide structure in the excited state, we kept the constraint on the positively charged atoms the same as the water complex in the excited state. In addition, we also maintained the constraint on the oxygen and hydrogen atoms of the hydroxide ion, similar to the ground state. The constraint on the negatively charged atom group is lifted because we expect the negative charge density to redistribute in order to stabilize the transferred proton on the Ag site. The difference in the atomic charges between the hydroxide and water complexes in the excited state is shown in Figure 7dII. While there is still a large increase in the magnitude of the negative partial charge on Ag, the enhancement is less than the corresponding change for the ground state (compare Figure 7b,dII). Consequently, the difference in energy between the hydroxide complex and the water complex decreases by ∼2.86 eV as compared to the ground state. In short, our calculations suggest that proton transfer from water to the Ag site is facilitated in the excited state because the charge‐separated excited state already has a substantial electron density on the Ag atom.

4. Conclusion

In this study, we developed a template‐free and cost‐effective strategy to construct atomically dispersed Ag‐Cu dual sites on porous tubular g‐C₃N₄. The synergistic coordination between Ag and Cu atoms extended visible‐light absorption, narrowed the bandgap, and significantly enhanced charge separation and transfer. TA spectroscopy demonstrated the ultrafast hot electron migration from the CN to the dual‐atom sites. Comprehensive characterizations combined with DFT calculations revealed that the dual sites increase the availability of electron density close to the Fermi level, induce charge redistribution to facilitate H^* intermediate formation, thereby driving efficient hydrogen evolution. The optimized Ag₁Cu₂‐CN catalyst exhibited an H₂ evolution rate of 2126 µmol g⁻¹ h⁻¹ with an AQY of 20% at 400 nm, surpassing most reported g‐C₃N₄ single‐atom catalysts while maintaining excellent stability. These findings highlight dual single‐atom engineering as an effective strategy to tune the electronic structure of carbon nitride and enhance photocatalytic solar‐to‐hydrogen production.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: smll74100‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.