Introduction

Hydrogen, a kind of clean and sustainable energy, possesses tremendous advantages, such as high gravimetric energy density and a wide range of use as industrial raw materials1. Global hydrogen consumption was up to 95 million tons in 2022 and has been increasing annually, more than 80% of which was gray hydrogen, produced from fossil fuels with more than 900 million tons of CO2 emissions, bringing enormous pressure on climate change2. The photocatalytic water splitting through the Honda-Fujishima effect can convert solar energy directly into chemical energy in sustainable or green hydrogen without carbon emissions3. The particle suspension reactor is one potentially economical option in photocatalytic hydrogen evolution to meet the target of a hydrogen price of $2–4 kg−1 set by the United States Department of Energy4. However, the narrow absorption spectrum of traditional particle photocatalysts, such as TiO2 and SrTiO35, limited the utilization of visible to near-infrared solar light6. In pursuit of efficient photocatalysts, organic semiconductors, such as conjugated polymer/molecules7,8,9,10,11,12,13,14,15, have shown potential to be a viable option, due to their structural diversity, broadband absorption, easily tunable optoelectronic property, and low-cost processing.

Especially more and more high-performance organic photovoltaic materials, including electron donors (D) and acceptors (A), have been developed for bulk heterojunction (BHJ) organic solar cells, conversing solar energy to electricity with efficiencies of over 20%16,17,18,19,20, which positively drove the research enthusiasm in the heterojunction nanoparticles (HJ-NPs) based on organic photovoltaic materials for solar hydrogen evolution via direct photocatalysis. Like most chemical reactions and chemical engineering processes employing particle reactants/catalysts, it is crucial to manage the size and morphology of organic HJ-NPs to achieve their high activity in photocatalysis, and the minimized size and heterogeneous surface of those NPs are preferred for maximizing the number of active sites and photoinduced-charge carrier extraction. McCulloch and coworkers have developed a surfactant modulation approach to manage the morphology of organic HJ-NPs from a common but unfavorable core-shell structure to a BHJ structure with a heterogeneous surface, and greatly improved photocatalytic hydrogen evolution rate (HER)9. However, there are few reports on minimizing the sizes of organic NP catalysts, which typically range from dozens to hundreds of nanometers beyond their charge-carrier/exciton diffusion length, imposing inherent limitations on photocatalytic activity.

Here, we reveal the size-minimized organic NPs with BHJ structure by employing all small molecule D/A pairs, and significantly enhance their photocatalytic activities for solar-driven hydrogen evolution. Compared to widely used polymer-containing D/A pairs, polymer-free systems intrinsically tend to form smaller NPs due to weaker intermolecular interactions and the absence of molecular entanglement. The polymer-free NPs exhibit (sub)nanometer-scale diameters, which are 1–2 orders of magnitude smaller than those of traditional polymer-containing NPs and situated within the exciton/carrier diffusion length of organic semiconductors. Among the candidates screened, the NPs based on BTR:L8-BO D/A pair exhibit the most promising photocatalytic activity under AM1.5 G illumination (100 mW cm−2). Then, size-minimized BTR:L8-BO NPs supported by covalent organic framework (NP@COF) show much better stability than the supportless NPs, and yield the optimized average photocatalytic mass-united HER (HER-m) of 527.2 (±34.0) ~ 3180.7 (±234.5) mmol h−1 g−1 and the area-united HER (HER-a) of 137.4 (±10.1) ~ 364.4 (±23.5) mmol h−1 m−2 at varied NP concentrations. The external quantum efficiencies (EQEs) of BTR:L8-BO NP@COF reach 18.4–32.8% under red to near-infrared light, representing a competitive performance among organic photocatalysts (OPCs)9,10,11,21,22, at which most inorganic catalysts reported in literatures fail to respond23.

Results and discussion

Design principle for size-minimized NPs

The microemulsion method stands out as one of the most extensively employed techniques for preparing NPs, in which the dispersed phase is mixed with the continuous phase, forming nanodroplets (NDs) stabilized by the surfactants, and the ND sizes dictate the NP sizes24. The size of steady-state NDs is influenced by its internal/external pressure (Pin/Pex) difference (ΔP = Pin−Pex) on NDs and interfacial tension (γ) between dispersed and continuous phase, which can be described by the Young–Laplace equation25:

$$\Delta P=\frac{2\gamma }{R}$$
(1)

where R denotes the ND radius. In a given continuous phase, the ND sizes are primarily determined by interfacial tension and intermolecular forces within the dispersed phase. The intermolecular forces are expected to correlate positively with the inward force and negatively with the magnitude of the outward ΔP at the ND interface26. Intermolecular forces include dipole–dipole interaction, induction force, and dispersion force (the details are seen in Supplementary Note 1). Compared to small molecules, polymers possess more repeating units, which confer more internal rotational degrees of freedom and a higher number of conformations. This increases the likelihood of molecular distortion, thereby enhancing the non-uniformity of electron distribution and increasing dipole–dipole interactions within polymers27. More importantly, small molecule semiconductors with finite conjugated length tend to exhibit weaker dispersion forces and induction forces when interacting with surrounding molecules, including the material itself and/or solvents. In contrast, polymers with quasi-infinite conjugated length have more delocalized and polarized electrons, resulting in stronger intermolecular interactions28. These factors restrict the size-minimization of NDs and then NPs (Fig. 1).

Fig. 1: Diagram of the design principle.
Fig. 1: Diagram of the design principle.
Full size image

The weaker intermolecular force (F) of small molecule semiconductors with finite conjugation relative to those widely-used polymers, likely induces higher ∆P within polymer-free NDs than polymer-containing NDs; the higher ∆P benefits to overcome the size-limited in those traditional polymer-containing NPs, and leads to reduced diameters (2R) of polymer-free NDs and then NPs, accompanied by BHJ-preferred morphology rather than core-shell structure. After being supported by COF, NP@COF exhibits enhanced stability, even in high-salinity/ionic water.

The viscosity is positively correlated with intermolecular forces29, and thus serves as a macroscopic indicator of them. Here, the viscosity of several typical organic photovoltaic materials in solution was investigated by employing the Ubbelohde viscometer30. The kinematic viscosities of typical polymer photovoltaic materials, such as PM6 and PCE10 (Supplementary Fig. 1), in chloroform solution at a concentration of 0.5 mg mL−1, are 0.42 mm2 s−1 and 0.49 mm2 s−1, respectively. These viscosities are lower than those of many common organic solvents, such as chlorobenzene (0.72 mm2 s−1), methylbenzene (0.67 mm2 s−1), and tetrahydrofuran (0.59 mm2 s−1), but remain noticeably higher than that of pure chloroform (0.38 mm2 s−1)31. As a contradistinction, chloroform solutions of small molecule photovoltaic materials (B1, DRCN5T, and BTR, chemical structures shown in Fig. 2a)32,33,34 at the same concentration as polymer solutions, present viscosities of about 0.38 mm2 s−1, nearly identical to that of pure chloroform (Fig. 2b)31. These results experimentally confirm that typical small molecule photovoltaic semiconductors possess intermolecular interactions comparable to or even weaker than those in common organic solvents, in stark contrast to conjugated polymers with significantly stronger interactions. Such a weak intermolecular interaction is unlikely to hinder the reduction in ND and NP sizes. Furthermore, self-aggregation occurs in polymer solutions due to their larger conjugation length and stronger intermolecular forces, resulting in particle sizes of 10–100 nm at concentrations of 0.25–1.00 mg mL−1, as measured by dynamic light scattering (DLS, Fig. 2c), and the much larger particles are formed in the solution with a higher concentration of 2.00 mg mL−1. In contrast, no scattering signal was detected in small molecule solutions such as BTR and B1, indicating their dispersion diameter is below the DLS detection limit (ca. 0.3 nm). This absence of self-aggregation and weaker intermolecular interactions render small molecules ideal candidates for fabricating size-minimized NPs.

Fig. 2: Screening polymer-free NPs.
Fig. 2: Screening polymer-free NPs.
Full size image

a The structures of small molecule materials. b The viscosity of polymer and small molecule solution with a concentration of 0.5 mg mL−1 (the average values ± standard deviations for 3 samples). c The diameter distribution of PM6 in chloroform solutions with different concentrations obtained by DLS. d Diameters of polymer-free HJ-NPs (D:A, w:w, 4:6, 0.4 wt.% TEBS/H2O) from TEM, while the diameters of polymer-containing HJ-NPs come from previous literature. e DLS for BTR:L8-BO NPs (D:A, w:w, 4:6, 0.4 wt. % TEBS). f TEM for BTR:L8-BO NPs (D:A, w:w, 4:6, 0.4 wt.% TEBS), and the inset is their diameter distribution. g Photocatalytic HER-m of 50 μg polymer-free NPs (D:A, w:w, 4:6, 0.4 wt.% TEBS, 30 wt.% Pt loading) in 7.5 mL 0.2 M AA aqueous solution at 5 °C, for 4 h under AM 1.5 G, 100 mW cm−2.

Screening polymer-free NPs

Typical small molecule photovoltaic donors B1, DRCN5T, and BTR were employed to prepare polymer-free HJ-NPs through the microemulsion method, by blending four typical small molecule acceptors named PC71BM, EH-IDTBR, BTP-eC9, and L8-BO (Fig. 2a) with type II D/A heterojunction energy level offset (Supplementary Fig. 2, Supplementary Table 1), and here the D/A weight ratio is 4:6, and sodium 2-(3-thienyl)-ethyloxybutylsulfonate (TEBS, 0.4 wt.%) was used as the stabilizer9. The DLS volume distribution (Supplementary Fig. 3 and Supplementary Table 2) and transmission electron microscopy (TEM) analysis (Supplementary Figs. 4, 5 and Supplementary Table 3) revealed that the diameters of these polymer-free HJ-NPs ranged from 1 to 10 nm. These values are already 1–2 orders of magnitude smaller than those of traditional polymer-containing NPs (ca. 50–100 nm), which are composed of polymer donor and small molecule acceptors (Fig. 2d, Supplementary Table 4)9,10,11,14,15,35. For instance, BTR:L8-BO NPs exhibited a mean diameter of approximately 5.5 nm according to the DLS volume distribution (Fig. 2e), consistent with the statistical mean value obtained from the TEM image (Fig. 2f).

The photocatalytic activities of these polymer-free HJ-NPs, consisting of 12 D/A pairs from three donors and four acceptors, were subsequently evaluated using 0.2 M ascorbic acid (AA) as the sacrificial agent and 30 wt.% Pt loading as cocatalyst, under simulated sunlight (AM 1.5 G, 100 mW cm−2). Despite the unoptimized fabrication of these polymer-free NPs, most of them already exhibited notable HER-m (Supplementary Fig. 6 and Supplementary Table 5), surpassing those reported for typical polymer-containing HJ-NPs in the literature9,10,11,14. Specifically, when coupled with acceptor molecule L8-BO, the B1, DRCN5T, or BTR-based HJ-NPs exhibited photocatalytic HER-m of 598 mmol h−1 g−1, 393 mmol h−1 g−1, and 754 mmol h−1 g−1 for 4 h, respectively. Meanwhile, B1:EH-IDTBR also exhibited a favorable HER-m, with the value reaching 237 mmol h−1 g−1 under the same conditions (Fig. 2g). These results suggest that the size-minimized NPs do benefit the enhanced photocatalytic activity.

Examination of dimension minimization

The diameters of NDs in ultrasonically formed microemulsions of polymer-free systems, here taking the BTR:L8-BO and B1:EH-IDTBR pair as examples, are primarily concentrated at 1–10 nm for BTR:L8-BO NDs and at 10–30 nm for B1:EH-IDTBR NDs (Supplementary Fig. 7), and the resulting NPs remain small after chloroform removal (Supplementary Fig. 8). In contrast, the ND diameters of polymer-containing systems, such as PM6:Y6 and PCE10:EH-IDTBR pairs, are larger, ranging from 50 to 100 nm. These findings indicate that the diminished size of those polymer-free NPs is contingent upon their precursor NDs, ruling out the possibility of ND reaggregation during the chloroform removal.

As indicated by Eq. (1), the lower interfacial tension γ of the dispersed/continuous phase for NP fabrication promotes the formation of smaller NDs and subsequently generated NPs, alongside the weak intermolecular interaction within the dispersed phase. The surfactant concentration is a critical parameter that governs the interfacial tension, thereby regulating the ND size and consequently the size of the resulting NPs. We measured γ using the Du Noüy ring method36, in which varying aqueous solutions of TEBS surfactant (continuous phase) and chloroform solution (0.5 mg mL−1) of various polymer-containing or polymer-free D/A pairs (dispersed phase) are employed. The chloroform solution of BTR:L8-BO, B1:EH-IDTBR, PM6:Y6, and PCE10:EH-IDTBR (D:A, w:w, 4:6) exhibited similar γ values to TEBS aqueous solution (aq.) at low concentrations, such as 21 mN m−1, 24.5 mN m−1, 23.1 mN m−1, or 23.7 mN m−1 in 0.5 wt.% TEBS aq., respectively, indicating γ should not be the primary determinant of the size reduction in polymer-free NPs relative to polymer-containing NPs. As the TEBS concentration increased from 0.5 to 10 wt.%, the γ between dispersed phase and TEBS aq. phase decreased rapidly and reached almost stable values with further increasing TEBS concentration for both polymer-containing and polymer-free systems. At a concentration of 10 wt.%, γ of PCE10:EH-IDTBR, PM6:Y6, B1:EH-IDTBR or BTR:L8-BO (CHCl3 solution)/TEBS aq. was measured at 3.1 mN m−1, 11.6 mN m−1, 7.1 mN m−1, and 6.2 mN m−1, respectively (Fig. 3a). Although the size of NPs decreased as the TEBS concentration increased for all types of NPs, a significant difference remained in the minimum achievable sizes between polymer-containing and polymer-free NPs. Notably, even at 30 wt.% TEBS, PCE10:EH-IDTBR NPs and PM6:Y6 NPs still showed a diameter of 19.2 nm and 17.8 nm, respectively, and in contrast, the diameters of polymer-free NPs are all below 5 nm (Fig. 3b).

Fig. 3: Examination of NP size-minimization and BHJ morphology.
Fig. 3: Examination of NP size-minimization and BHJ morphology.
Full size image

a Interfacial tension of dispersed phase (0.5 mg mL−1 chloroform solution)/continuous phase (varying concentrations of surfactant aqueous solution). The chemical structures of the used surfactants are depicted. b The surfactant concentration-NP diameter plot of polymer-free and polymer-containing D/A pairs. 2D SAXS patterns of BTR:L8-BO NPs stabilized by 0.05 wt.% SDS (c) and 0.4 wt.% TEBS (d). SAXS intensity profiles of BTR:L8-BO NPs fitted with DAB-fractal model (e) and core-shell model (f).

Likewise, the surfactant structure is also a critical factor that governs the interfacial tension. Subsequently, various types of surfactants with gradient concentration, including sodium dodecyl sulfate (SDS), sodium dodecylbenzenesulphonate (SDBS), and sodium cholate (SC), were further used to prepare NPs. Similar to TEBS surfactant, the concentrations of SDS, SDBS, and SC aq. exhibit a negative correlation with their interfacial tensions with polymer-free/polymer-containing chloroform solution (B1:EH-IDTBR, BTR:L8-BO, PM6:Y6, and PCE10:EH-IDTBR, 4:6, w:w, 0.5 mg mL−1), reaching minimum values when the concentrations are up to approximately 1 wt.% (Fig. 3a). At concentrations exceeding 0.01–0.05 wt.% for SDS, SDBS, or SC, the diameters of the produced polymer-free NPs (B1:EH-IDTBR and BTR:L8-BO) are as small as 5 nm, substantially smaller than those (>20 nm) of polymer-containing NPs (PM6:Y6 and PCE10:EH-IDTBR, Fig. 3b). This indicates that, regardless of the surfactant type or dosage, polymer-free NPs consistently exhibited significantly smaller sizes than polymer-containing NPs.

Then, the external factors such as ultrasound power (100–300 W for 1–5 min) did not significantly affect the size of polymer-free NPs in the range of 1–10 nm and polymer-containing NPs (>20 nm) (Supplementary 9). These findings confirm that the shear energy applied in this study (250 W for 5 min) was sufficient to generate well-dispersed NDs under thermodynamic equilibrium and was not a limiting factor for the resulting NP size. By tuning the NP preparation parameters, including interfacial tension and shear energy, we found that small molecules, due to their weaker intermolecular interactions and the absence of molecular entanglement, are inherently more likely to forming smaller NPs than polymers.

Internal morphology

In addition to NP size, the internal morphology of HJ-NPs is another major intrinsic factor influencing their photocatalytic activity, and the BHJ-like morphology is favored for the extraction of photogenerated hole/electron, according to the previous study by McCulloch and coworkers9. Small-angle X-ray scattering (SAXS) is used to analyze the NP morphology statistically, and X-ray scattering profiles provide insights into nanostructural inhomogeneities of electron density within bulk materials. Then, the analysis of SAXS data from coherent scattering events provides information about the distributions of material components. Figure 3c, d display the two-dimensional SAXS patterns of BTR:L8-BO NPs stabilized by 0.05 wt.% SDS and 0.4 wt.% TEBS, respectively, and their corresponding one-dimensional profiles are shown in Fig. 3e, f. The X-ray scattering profiles of single-component NPs are shown in Supplementary Fig. 10. Here, we use Debye–Anderson–Brumberger and the fractal-like network model37,38,39 and core-shell model40 (the details of the model seen in Supplementary Note 2) for the fitting analysis of the measured scattering profile, respectively. The SAXS profiles of BTR:L8-BO NPs stabilized by SDS or TEBS, fit well with the Debye–Anderson–Brumberger (DAB) and the fractal-like network model (fitting coefficients close to 1) (Fig. 3e, the parameters of fitting are listed in Supplementary Tables 6, 7), which indicates that the BTR:L8-BO NPs based on either SDS or TEBS have the BHJ-like continuous interpenetrating network structure. On the contrary, the model describing core-shell particles fails to fit well the SAXS profile of BTR:L8-BO NPs, which have coefficients of determination of 0.98 at most for BTR:L8-BO NPs stabilized by SDS and only 0.06 for BTR:L8-BO NPs stabilized by TEBS (Fig. 3f, the parameters of fitting seen in Supplementary Tables 8, 9), when setting the radius as the corresponding value and applicable to a Log-normal distribution (Supplementary Table 10) with polydisperse shell thickness (Gaussian distribution). Similarly, SAXS profiles of B1:EH-IDTBR NPs stabilized by SDS or TEBS can also be fitted well with the model describing the continuous interpenetrating network structure (fitting coefficient of almost 1 while failing to be fitted with the core-shell model (Supplementary Fig. 11 and Supplementary Table 11–14). The above results suggest that size-minimized polymer-free HJ-NPs tend to form a continuous interpenetrating network structure rather than a core-shell structure, independent of molecular structure or surfactant types. This is different from the typical polymer-containing HJ NPs, of which SDS-stabilized NPs and TEBS-stabilized NPs tend to exhibit core-shell or BHJ structures, respectively9,41.

To further investigate internal morphology, we examined TEBS-stabilized BTR:L8-BO NPs as a representative example using cryo-TEM. In contrast to the core-shell structure typically observed in polymer-containing HJ-NPs, the BTR:L8-BO polymer-free NPs exhibit continuous amorphous morphology with only a few crystalline domains dispersed inside. These crystalline domains exhibit two primary lattice spacings of 1.23–1.24 nm and 1.50 nm (Supplementary Fig. 12) that is corresponding to L8-BO lamellar stacking and BTR unit crystal cell, respectively33,42. This predominance of amorphous regions further supports the formation of a BHJ-like structure in the polymer-free HJ-NPs, consistent with the SAXS results. This BHJ-preferred morphology is likely strongly associated with the size minimization of polymer-free HJ-NPs. From a geometric standpoint, such small (<5 nm) polymer-free NPs can be treated as molecular clusters only composed of several molecules, and thereby it is difficult to form a core-shell structure with a continuous shell to fully cover up another paired material. Such cover is usually found in SDS-stabilized polymer-containing HJ NPs41, and the smaller particles prefer tighter polymer shell encapsulation, thereby further hindering the electron transfer to Pt, which may contribute to the decreased HER recently reported in SDS-stabilized polymer-containing NPs below 60 nm43. Then the SDS-stabilized size-minimized BTR:L8-BO NPs here with similar BHJ morphology also show lower photocatalytic activities than their TEBS-stabilized counterparts, despite the SDS concentration used is much lower than the value in the literature (Supplementary Fig. 13), which can be attributed to the decreased Pt deposition caused by the long alkyl chain of SDS44.

Photoinduced charge-carrier dynamics

Preliminary screening indicates that polymer-free NPs based on BTR:L8-BO exhibit the most promising photocatalytic activity. The BTR:L8-BO HJ-NPs display broad absorption from the visible to near-infrared region (Supplementary Fig. 14a, b) and exhibit efficient charge separation, as evidenced by photoluminescence (PL) quenching compared to pristine donor or acceptor NPs (Supplementary Fig. 14c, d), which meets the prerequisite for efficient solar-driven photocatalysis. To further investigate the exciton dissociation dynamics in BTR:L8-BO HJ NPs, transient absorption (TA) measurements were first carried out with pump fluence dependence under selective excitation of the acceptor at 780 nm (Supplementary Fig. 15)45,46,47. As shown in Supplementary Fig. 16, the ground state bleaching signal of L8-BO, peaking at 784 nm and consistent with the absorption profile of L8-BO, shows no significant change in decay lifetime across different excitation intensities (0.82–3.27 μJ cm−2), indicating the absence of exciton-exciton annihilation at these weak excitation fluences11. Subsequently, TA spectra were collected for BTR:L8-BO (w:w, 4:6) NPs along with polymer-containing PM6:L8-BO (w:w, 4:6) NPs as a reference under 780 nm pump excitation at a fluence of 1.63 μJ cm−2. As illustrated in Fig. 4a–d, the three emerged excited-state absorption bands with peaks at ~550, ~900 and ~1350 nm can be assigned to the reduced form of L8-BO, the localized excitation (LE) state of L8-BO, and the intra-moiety excited (i-EX) state of L8-BO10,48,49, respectively, which are also present in the TA spectra of L8-BO NPs (Supplementary Fig. 17). Kinetic analysis of the reduced L8-BO signal reveals that the small-molecule system BTR:L8-BO exhibits a longer decay lifetime (t1/t2 = 1.2/350.8 ps) compared to the polymer-containing PM6:L8-BO system (t1/t2 = 0.8/22.4 ps), indicating slower recombination of the reduced acceptor in the small-molecule system (Fig. 4e). Meanwhile, the lifetimes of LE state (t1/t2 = 0.6/12.5 ps) and the i-Ex state (t1/t2 = 5.2/206.1 ps) in BTR:L8-BO are shorter than those in PM6:L8-BO (LE:t1/t2 = 0.6/13.4 ps; i-Ex:t1/t2 = 8.5/286.3 ps), suggesting that the continuous interpenetrating network structure in the size-minimized polymer-free HJ-NPs facilitates shorter diffusion paths and more efficient exciton dissociation compared to the polymer-containing system, which is expected to contribute to enhanced photocatalytic hydrogen evolution (Fig. 4f, g)9.

Fig. 4: TA spectra.
Fig. 4: TA spectra.
Full size image

Broadband TA data (a) and TA spectra at different time scale (b) of BTR:L8-BO (4:6) NPs stabilized by 0.4 wt.% TEBS under excitation of 780 nm. Broadband TA data (c) and TA spectra at different time scale (d) of PM6:L8-BO (4:6) NPs stabilized by 0.4 wt.% TEBS under excitation of 780 nm. TA traces of PM6:L8-BO NPs and BTR:L8-BO NPs probed at 550 nm (e), 900 nm (f), and 1350 nm (g).

Solar-driven photocatalytic hydrogen evolution

Then, BTR:L8-BO HJ NPs were optimized for photocatalytic sacrificial hydrogen evolution under simulated sunlight, in which AA (0.2 M) acted as the sacrificial agent and Pt was photo-deposited in situ on the HJ-NP surface as a cocatalyst (Fig. 5a). The highest HER-m achieved at a D/A ratio of 4:6 (w:w) (Supplementary Fig. 18), could be attributed to more efficient charge separation relative to other D/A ratios which was supported by the shortest decay lifetime of the L8-BO i-Ex signal observed in TA experiments (Supplementary Figs. 19 and 20). The optimized TEBS/H2O weight ratio and Pt-loading ratio were determined to be 0.4 wt.%, and 70 wt.% respectively, with detailed optimization results provided in Supplementary Figs. 21–24. Determined by ICP-MS according to the reported method44,50, the actual Pt loading (mass of Pt on OPC/mass of OPC) was 25.5 wt.% (Supplementary Table 15). Furthermore, TA spectra of NPs in the presence of AA and AA+Pt were measured. Upon the addition of AA, the hole in the donor BTR domain can be effectively extracted, while charge recombination of reduced L8-BO in the acceptor domain is not significantly affected, leading to a similar reduced acceptor lifetime (t1/t2 = 1.2/365.3 ps). In contrast, the lifetime of reduced L8-BO decreases markedly to t1/t2 = 0.2/1.1 ps after in-situ Pt deposition, indicating rapid electron transfer from the NPs to Pt (Supplementary Figs. 25 and 26).

Fig. 5: Photocatalytic hydrogen evolution.
Fig. 5: Photocatalytic hydrogen evolution.
Full size image

a Schematic diagram of solar-driven photocatalytic hydrogen evolution with sacrificial agent and Pt. b The structures of TEBS and PEBS. c HER-a/HER-m versus concentration for BTR:L8-BO NPs (D:A, w:w, 5:5, 0.4 wt.% PEBS, 20 wt.% Pt loading) without or with COF in 0.2 M AA under AM1.5 G, 100 mW cm−2 (the average values ± standard deviations for 3 samples). d SEM morphology of the COF and the NP@COF composite (BTR:L8-BO 5:5 NPs, 0.4 wt.% PEBS/H2O, 20 wt.% Pt loading). The dashed square in the middle panel indicates the region enlarged in the right panel, where NPs are highlighted with white circles. e Absorption spectrum and EQEs at 400, 500, 600, 700, and 808 nm of BTR:L8-BO NPs (D:A, w:w, 5:5, 0.4 wt.% PEBS/H2O, 20 wt.% Pt loading)@COF in 0.2 M AA at a NP concentration of 53.33 μg mL−1. f The photo of BTR:L8-BO NPs (D:A, w:w, 4:6, 0.4 wt.% TEBS/H2O) in (I) water; (II) 0.6 M NaCl aqueous solution and BTR:L8-BO NPs (D:A, w:w, 4:6, 0.4 wt.% TEBS/H2O, 70 wt.% Pt loading)@COF in (III) water; (IV) 0.6 M NaCl aqueous solution under 650 nm laser illumination. The NP concentration is 50 μg mL−1. g HER-m of BTR:L8-BO NPs (4:6, w:w, 0.4 wt.% TEBS/H2O)@COF in 0.2 M antibiotics aqueous solution under AM1.5 G, 100 mW cm−2 for 4 h and in a mixed solution of 0.6 M NaCl and 0.2 M AA under AM1.5 G, 100 mW cm−2 for 10 h. Notably, NPs without COF are not stable and do not generate hydrogen in such high-salinity conditions. For all the above photocatalytic reactions, the solvent volume was 7.5 mL, and the reactions were carried out at 5 °C. All the weight ratio of NP:COF in Fig. 5 is 10:8 (w:w).

The optimized BTR:L8-BO NPs exhibited the highest average HER-m of up to 1786 mmol h−1 g−1 at a low concentration of 3.33 μg mL−1 for 4 h under AM 1.5 G, 100 mW cm−2. Then the concentration of the photocatalyst was increased for almost saturated absorption (26.67 μg mL−1), BTR:L8-BO NPs showed average photocatalytic HER-a of 161 mmol h−1 m−2 and HER-m of 468 mmol h−1 g−1 for 4 h. Further increasing the concentration to 53.33 μg mL−1, a HER-a of 194 mmol h−1 m−2 and a HER-m of 281 mmol h−1 g−1 are yielded (Supplementary Table 16), and under this concentration, the external quantum efficiencies (EQEs) of BTR:L8-BO NPs were 9.2%, 7.2%, 12.7% and 19.2% at 500, 600, 700, and 800 nm, respectively (Supplementary Fig. 27 and Supplementary Table 17).

Furthermore, a surfactant named 2-phenylethoxybutyl sulfonate sodium (PEBS, the structure shown in Fig. 5b) with a phenyl instead of thiophenyl in TEBS was synthesized (the details of synthesis are shown in Methods) and used to stabilize polymer-free HJ-NPs for further improving photocatalytic hydrogen evolution. The optimized D/A ratio, PEBS/H2O weight ratio, and Pt-loading ratio are 5:5 (w:w), 0.4 wt.%, and 20 wt.%, respectively, and the detailed results are shown in Supplementary Figs. 28–31. Then the actual Pt load in this condition is 11.6 wt.% from ICP-MS measurement (Supplementary Table 15). The optimized BTR:L8-BO NPs stabilized by PEBS exhibited the average HER-m of up to 2314.1 ± 306.6 mmol h−1 g−1 at a low photocatalyst concentration of 3.33 μg mL−1, and as the concentration increased to 26.67 μg mL−1 ~ 53.33 μg mL−1, the average HER-m fell to 656.9 ( ± 64.8) ~ 368.8 (±29.0) mmol h−1 g−1 and the average HER-a raised to 227.1 (±22.4) ~ 255.0 ( ± 20.0) mmol h−1 m−2 for 4 h under AM 1.5 G, 100 mW cm−2 (Fig. 5c, Supplementary Table 18). The improvement in hydrogen generation activity can be attributed to the further reduced size of NPs stabilized by PEBS (3.13 nm) compared to TEBS (5.47 nm, Supplementary Fig. 32). These notable HER-m results support that minimizing HJ-NPs by employing polymer-free pairs to the nanometer size is efficient for photocatalytic activity enhancement. Recently, Li, Zhu, Cooper, and coworkers have developed a high-throughput automated platform for screening molecular heterojunction photocatalysts, which also suggests the advantage of small molecules in flow synthesis13.

The nanoscale particle size of polymer-free HJ NPs presents a challenge in terms of stability during the photocatalytic hydrogen evolution, as they are susceptible to aggregation and precipitation, thereby limiting their practical applications. The aggregation of these NPs was prohibited by introducing COF (named as Py-TAPD-COF, Supplementary Fig. 33) which, under the mildly acidic conditions generated by AA, becomes partially protonated at the imine or amine sites51, thereby acquiring a positive surface charge that electrostatically attracts and anchors the negatively charged NPs, preventing their aggregation and sedimentation and thus enhancing stability and sustaining photocatalytic activity over time. Consequently, numerous NPs adhered to COF were observed by using scanning electron microscopy (SEM, Fig. 5d and Supplementary Fig. 34). The NPs supported by COF (NP@COF) show much better photocatalytic stability than that of the NPs without COF (Supplementary Fig. 35). Subsequently, we have carried out ultraviolet-visible (UV-vis) absorption spectra and mass spectrometry to characterize the structural stability. UV-vis absorption spectra and mass spectrometry verify that the chemical structure of NP@COF remains essentially unchanged (Supplementary Fig. 36). These results demonstrate the structural robustness of the NP@COF composite during repeated photocatalytic operation. The optimized BTR:L8-BO NP@COF, stabilized with TEBS, exhibited the highest average HER-m of up to 1735 mmol h−1 g−1 at a low concentration of 3.33 μg mL−1 and the highest average HER-a of 183 mmol h−1 m−2 at a concentration of 53.33 μg mL−1 for 4 h under AM 1.5 G, 100 mW cm−2 illumination (Supplementary Figs. 37, 38 and Supplementary Table 19). The corresponding maximum EQEs are 8.8%, 7.5%, 12.1% and 19.8% at 500, 600, 700, and 808 nm, respectively (Supplementary Fig. 39, Supplementary Table 17). The close EQEs between these NPs and NP@COF indicate that their similar photocatalytic activity in the initial stage is mainly attributed to the enhanced stability rather than the catalytic activity of Py-TAPD-COF itself, while pure Py-TAPD-COF does not produce hydrogen under the same conditions.

To more directly investigate whether the Py-TAPD-COF participates in charge transfer, we first measure the absorption of the Py-TAPD-COF with a relatively wide bandgap (Supplementary Fig. 40a), which exhibits limited intrinsic absorption with a peak at 500 nm (longer wavelength signal attributed to scattering). Subsequently, PL spectra show that the Py-TAPD-COF itself exhibits almost no fluorescence emission (Supplementary Fig. 40b). When single-component L8-BO NPs are loaded onto the Py-TAPD-COF and excited at 780 nm, the PL intensity decreases by 32% compared to those NPs without the Py-TAPD-COF (Supplementary Fig. 40c). Similarly, when BTR-only NPs are loaded onto the Py-TAPD-COF, the PL intensity also decreases by 18% (Supplementary Fig. 40b). The electron transfer should be limited by a type Ⅰ contact between Py-TAPD-COF and L8-BO within HJ-NPs (Supplementary Fig. 40d). Meanwhile, although the LUMO energy of BTR is a little shallower than Py-TAPD-COF, which may allow the electron transfer from BTR to Py-TAPD-COF, the spatial separation between the Py-TAPD-COF and BTR by the surfactant TEBS likely restricts the possibility of charge transfer. Therefore, the observed fluorescence reduction may be attributed to decreased excitation light absorption by L8-BO/BTR due to the light scattering by Py-TAPD-COF. Both mechanism insights and initial hydrogen evolution performance indicate that negligible charge transfer occurs between the Py-TAPD-COF and NPs, with the Py-TAPD-COF primarily functioning as a support. At the same time, when stabilized with PEBS instead of TEBS, the optimized BTR:L8-BO NP@COF exhibited further enhanced average HER-m of up to 3180.7 ± 234.5 mmol h−1 g−1 at a low concentration of 3.33 μg mL−1, and average HER-a of 364.4 ± 23.5 mmol h−1 m−2 at a concentration of 53.33 μg mL−1 for 4 h under simulated sunlight (Fig. 5c, Supplementary Fig. 41 and Supplementary Table 20), demonstrating competitive performance among reported OPCs (Supplementary Table 21). The corresponding maximum EQEs are 19.1, 23.2, 22.3, 18.4, and 32.8% at 400, 500, 600, 700, and 808 nm (Fig. 5e, Supplementary Table 17), and the notable EQE values verify the high photocatalytic activity, especially in the near-infrared region (Supplementary Table 21)21, at which most existing inorganic photocatalysts do not respond. Although the EQE test is conducted under conditions of light absorption saturation, the EQE at different wavelengths still varies due to differences in charge transfer pathways. When the acceptor is excited under near-infrared illumination, electrons transfer directly to the Pt cocatalyst, whereas donor excitation requires an additional step from donor to acceptor before reaching Pt11. The shorter transfer pathway and lower recombination probability in the acceptor-excited process account for the higher quantum efficiency observed under near-infrared illumination.

Hydrogen evolution in varied water sources

The enhanced stability of NP@COF catalysts made them promising for the direct utilization of some high-salinity water sources other than scarce fresh water resources, like some wastewater and even seawater (96.5% of the Earth’s water), in which most supportless NPs suffer severe condensation and catalytic deactivation, because the heightened concentrations of free ions disrupt the double layer surrounding the NPs, prompting their aggregation and subsequent sedimentation52. Benefiting from the support of COF, NP@COF can be stable in water as a colloid showing the obvious Tyndall phenomenon under 650 nm laser illumination, even if the electric double layer structure of NPs is disrupted in high-salinity water (Fig. 5f). Here we utilize a 0.6 M sodium chloride aqueous solution to emulate the ionic and high-salinity environment of seawater (global average salinity of 3.5%) for photocatalytic hydrogen evolution, and in such highly ionic water NP@COF still exhibit obviously photocatalytic activity with an average AA-sacrificial HER-m of 10.9 mmol h−1 g−1 under AM1.5 G, 100 mW cm−2 for 10 h (Fig. 5g). In contrast, NPs without COF are susceptible to rapid aggregation and sedimentation when stirred under dark conditions in the sodium chloride aqueous solution. Although the HER of NP@COF in simulated seawater is markedly lower than that observed in pure water due to much more sodium ions than hydrogen ions accumulated on the particle surfaces, this development represents a significant improvement over the severe instability of NPs in saline environments53.

The photocatalytic sacrificial hydrogen evolution usually suffers additional operating costs due to the consumption of chemical sacrificial agents like AA used here. Benefiting from the enhanced stability of NP@COF in high-salinity water, reducing agents widely present in high-salinity wastewater show considerable potential to replace the commercial sacrificial agents for hydrogen evolution. Here, we take several antibiotic wastewaters as examples, in which antibiotics are widely used in both human medicine and animal husbandry, posing substantial environmental risks due to their persistence and potential for adverse effects on non-target organisms. The total annual consumption of antibiotics gradually increases around the world, reaching over 106 tons, and a considerable portion is discharged into wastewater, where antibiotic concentrations detected in surface water spans from ng L−1 to mg L−1, and these biologically active reductive compounds can have detrimental impacts on the environmental matrix, necessitating thorough evaluation of their ecological ramifications54,55. The NP@COF maintains stability (Supplementary Fig. 42) and exhibits photocatalytic activity in 0.2 M antibiotic solutions without AA sacrificial agents, where lincomycin hydrochloride (LM), gentamicin sulfate (GM, a mixture of three compounds), and levofloxacin hydrochloride (LX) serve as sacrificial agents with HER-m of 6.6 mmol h−1 g−1, 7.9 mmol h−1 g−1, and 3.8 mmol h−1 g−1, respectively, under AM1.5 G illumination at 100 mW cm−2 for 4 h at a NP concentration of 6.67 μg mL−1 (Fig. 5g). In contrast, NPs without COFs aggregate after 5 min (Supplementary Fig. 43), resulting in no hydrogen evolution. At a concentration of 0.2 M, AA has a pH of 2.4 and an ion concentration of 10−3 M, and in comparison, antibiotics at 0.2 M have ion concentrations of 10−1 M, two orders of magnitude higher than AA. At a NP concentration of 53.33 μg mL−1, NP@COF remains stable in antibiotic solutions, with HER-m of 1.9 mmol h−1 g−1 in GM and 1.6 mmol h−1 g−1 in LM (Supplementary Fig. 44). This result supports that the COF-stabilized size-minimized NPs have the capability for generating hydrogen in wastewater with antibiotics and other reducing agents, and then reducing the operational cost of energy supply by using the wastewater partially instead of the commercial sacrificial agents.

In summary, we reveal and demonstrate that the NP size can be minimized in polymer-free HJ NPs employed with all small molecule D/A pairs, overcoming the limit of widely-used polymer NPs and bringing enhanced performance in photocatalytic activity; and then we stabilize size-minimized NPs by supporting them on COFs, even in high-salinity water. These polymer-free NPs exhibit diameters ranging from 1 to 10 nm at 0.4 wt.% TEBS/H2O, which have already been 1–2 orders smaller than those (ca. 50–100 nm) reported for traditional polymer-containing NPs, and the weak intermolecular forces and the absence of entanglement between small molecules mainly contribute to the smaller size of polymer-free NDs and NPs than polymer-containing counterparts in the same fabrication conditions. Profiting from size-minimization, the preferred BHJ morphology is formed within polymer-free HJ-NPs, independent of the surfactant type. As a result, both polymer-free HJ-NPs without and with COFs exhibit enhanced performance for photocatalytic hydrogen evolution under AM 1.5 G, 100 mW cm−2; the optimized polymer-free NP@COF achieves a HER-a of up to 364.4 ± 23.5 mmol h−1 m−2 with the near-infrared EQE of up to 32.8%, and HER-m values of 527.2 (±34.0) ~ 3180.7 (±234.5) mmol h−1 g−1 at varied OPC concentration. Furthermore, the NP@COF with enhanced stability exhibits promising hydrogen evolution in seawater and wastewater containing antibiotics. Our work confirms the advantage of small molecules for heterogeneous reactions as nano-catalysts, which boosts the development of organic NP photocatalysts for solar hydrogen evolution. This strategy to minimize the size of NPs and then stabilize them on frameworks should be extended to other chemical and chemical engineering processes, such as pollutant photodegradation and industrial catalysis, as well as other applications where stable nanosized particles are required or preferred.

Methods

Materials and characterization

The chemical reagents and materials were purchased from commercial companies and were used without further purification. Except for L8-BO and Py-TAPD-COF, organic semiconductor materials were purchased from Solarmer Energy, Co., Ltd. L8-BO was purchased from Nanjing Zhiyan Technology Co., Ltd. Py-TAPD-COF, LX, LM, and GM were purchased from Beijing Innochem Technology Co., Ltd. Organic solvents were purchased from Tianjin Concord Technology Co., Ltd. UV-vis absorption spectra were measured by a UH4150 spectrophotometer. PL spectra were measured by a FLS980 fluorescence spectrometer (Edinburgh Instruments). Hydrodynamic diameters of NPs and NDs were measured by DLS (Zetasizer Nano-ZS, Malvern Instruments) and batch-to-batch variations under identical conditions were within experimental uncertainty. Interfacial tension were measured employing the du Noüy ring method. The liquid under scrutiny was ensconced within a 10 mL vitreous vessel. An organic phase concentration of 0.5 mg mL−1 was meticulously selected with a corresponding concentration for the surfactant aqueous counterpart. The interfacial tension were recorded by an automatic tensiometer (FST300C, AFES). SEM measurement was performed using a Hitachi S-4800. Silicon wafers underwent ultraviolet ozone treatment for 20 min, followed by direct deposition of 3 μL NPs onto their surfaces. After vacuum drying at 40 °C for 2 h, before SEM measurement. TEM measurement was performed using the Hitachi HT7700 TEM. Ultra-thin microgrids were subjected to ultraviolet ozone treatment for 20 min. Subsequently, 3 μL of the NP suspension was directly deposited onto the pretreated microgrids, followed by vacuum drying at 40 °C for 2 h before TEM measurement.

Synthesis of PEBS

To a solution of 2-Phenethanol (5 mL; 41.7 mmol) in freshly distilled toluene (200 mL), 1.18 g of sodium hydride (48 mmol) was added slowly, which led to a white precipitate. After the mixture was stirred at 25 °C for 0.5 h, 1,4-butanesultone (48 mmol) was added dropwise. The mixture was then refluxed for 2 h under the protection of nitrogen. By filtering and washing with toluene and ethanol, and vacuum drying at 40 °C, the final pale-yellow precipitate was obtained with a yield of 46%. 1H NMR (400 MHz, D2O): δ 7.23 (m, 5H), 3.66 (t, 2H), 3.43 (t, 2H), 2.78 (m, 4H), 1.57 (m, 4H). 13C NMR (100 MHz, D2O): δ 139.12, 128.89, 128.64, 126.44, 70.99, 69.88, 50.63, 35.02, 27.37, 20.86. ESI-MS (TOF): [M-Na]− calcd. for C12H17O4S− 257.08, found 257.14.

Viscosity measurement

Viscosities of organic semiconductor material solutions were measured by an Ubbelohde viscometer. Initially, the efflux time for a series of solvents with different viscosities, such as chloroform (0.38 mm2 s−1), dichloromethane (0.32 mm2 s−1), toluene (0.67 mm2 s−1), ethyl acetate (0.50 mm2 s−1), tetrahydrofuran (0.59 mm2 s−1), n-hexane (0.50 mm2 s−1), chlorobenzene (0.72 mm2 s−1), and dichlorobenzene (1.01 mm2 s−1)31, were determined by a Ubbelohde viscometer for three times. The efflux time is 132 ± 0.63 s, 115 ± 0.4 s, 230 ± 0.8 s, 170 ± 0.49 s, 188 ± 0.4 s, 182 ± 0.63 s, 246 ± 0.4 s, and 355 ± 0.4 s for chloroform, dichloromethane, toluene, ethyl acetate, tetrahydrofuran, n-hexane, chlorobenzene, and dichlorobenzene, respectively. The working curve of efflux time to viscosity was calculated from the above data:

$$V=0.00288\times {t}^{*}+0.00536$$
(2)

where V and t* represented viscosity and efflux time. Subsequently, the efflux time for the chloroform solutions of the organic semiconductor materials were measured by a Ubbelohde viscometer. The efflux time values are 145 ± 0.63 s, 167 ± 0.49 s, 129 ± 0.49 s, 131 ± 0.4 s, and 129 ± 0.63 s for 0.5 mg mL−1 chloroform solution of PM6, PCE10, BTR, DRCN5T, and B1, respectively. Then, the corresponding viscosity can be calculated according to formula (2).

Fabrication of NPs

The semiconductor materials were dissolved in chloroform for 2 h with a concentration of 0.5 mg mL−1. Chloroform solutions with corresponding D/A weight ratios were added to pre-configured TEBS or PEBS aqueous solutions with a chloroform/H2O ratio of 1/4 (v/v). The mixed solutions were sonicated for 5 min using the ultrasonic processor (SCIENTZ-IID, Scientz Biotechnology) to obtain the mini-emulsion. The chloroform in mini-emulsion was removed using rotary evaporation for 2–3 min to obtain a surfactant-stabilized NPs dispersion in water. Finally, the NPs dispersion was filtered (0.45 μm mixed cellulose) to remove large aggregates. The NPs stabilized by SDS, SDBS, and SC were fabricated using the same method. The final concentration was determined by the following: 2 mL of methanol was added to 1 mL of NPs for de-emulsification. Subsequently, 2 mL chloroform was introduced to extract the semiconductor material from the aforementioned mixture. Following the extraction, the organic solvents were removed using rotary evaporation. Then, 1 mL of chloroform was added to the residue. Finally, the absorbance of the solution was measured by UV–vis absorption spectra to ascertain the final mass of NPs in the solution. Notably, we did not have a long time of mechanical stirring during the formation of microemulsion or heating during the removal of organic solvents. In contrast, literature reports typically use a long-term heating or stand method to remove solvents, which usually requires 1–20 h9,11,41,44,56,57.

Fabrication of NP@COF

The optimized polymer-free NPs (BTR:L8-BO NPs stabilized by 0.4 wt.% TEBS or PEBS) were dispersed in an aqueous solution containing 0.2 M AA (7.5 mL), and then the platinum precursor was added to the NPs dispersion. The specific platinum precursor is 1 mg mL−1 K2PtCl6 aqueous solution (the amount added is calculated based on the required platinum mass ratio). Unless otherwise specified, the Pt loading ratio refers to the added Pt amount/OPC amount. The system was continuously irradiated for 4 h under LED white light source (100 mW cm−2). The 2 mg Py-TAPD-COF was added to 8 mL purified water, and was subjected to sonication using an ultrasonic power of 250 W for a duration of 5 min, resulting in an aqueous dispersion of COF with a concentration of 250 μg mL−1. Various proportions of Py-TAPD-COF were introduced into the Pt-loaded NPs dispersion. Subsequently, distillation was conducted under reduced pressure at 60 °C for a period of 1 h, with water added intermittently during the distillation process to ensure that the system does not undergo complete evaporation. Finally, the concentration of the NPs was adjusted to 50 μg mL−1 to obtain NP@COF. Following this processing, the BTR NP@COF and L8-BO NP@COF samples for photoluminescence measurement were prepared at NP:COF weight ratio of 10:8, without Pt loading.

Hydrogen evolution

Various volumes of photocatalyst NP solutions or NP@COF solutions were added to a pre-prepared 0.2 M aqueous solution of sacrificial agents (such as AA, LM, GM, or LX), either in the presence or absence of 0.6 M NaCl, to achieve different NP concentrations. The total volume of the reaction system was maintained at 7.5 mL. Depending on the specific reaction conditions, an appropriate quantity of K2PtCl6 aqueous solution (1 mg mL−1) was added to the corresponding NP solutions. For conducting hydrogen evolution under AM1.5 G light, the combined reaction mixture was transferred to a temperature-controlled recirculating batch reactor (cross-sectional area: 5.786 cm2; light penetration depth:1.3 cm) and the temperature was maintained at 5 °C. This reactor was evacuated to eliminate oxygen, and the pressure was adjusted to 1.2 kPa. The NP suspensions were exposed to a 300 W xenon lamp fitted with an AM 1.5 G filter, and the light intensity onto samples is 100 mW cm−2. The light intensity was calibrated using an OHSP-350S radiometer (HOPOO light&color technology Co., Ltd.). No additional side-products were detected during the test. To enhance the efficiency of screening for the optimal hydrogen evolution condition in terms of the NPs:COF ratio and PEBS NP screening, multiple-channel reactors (PCX-50C, PerfectLight) were employed. The NP@COF composites formed under various conditions were individually introduced into separate offline reactors, each featuring a cross-sectional area of 3.1 cm2 and a light penetration depth of 2.4 cm. These reactors were filled with 7.5 mL of an aqueous solution containing 0.2 M AA and subsequently purged repeatedly with nitrogen to eliminate atmospheric oxygen. The suspensions containing the photocatalyst NPs were agitated at room temperature and illuminated using a light source that spanned the spectrum of 400–800 nm (100 mW cm−2). Hydrogen evolution was comprehensively assessed using an all-glass automatic online trace gas analysis system (Labsolar 6 A, PLS-SXE300 + , PerfectLight), equipped with an online gas chromatograph featuring a thermal conductivity detector and N2 as the carrier gas. Notably, the multiple-channel photochemical reaction system incorporated an LED lamp (400–800 nm, 100 mW cm−2) instead of the reference spectrum (AM 1.5 G, 100 mW cm−2) typically used for standard tests of solar energy conversion systems. Additionally, the reactor exhibited a light-concentrating effect, resulting in an actual irradiation intensity exceeding 100 mW cm−2. Consequently, to ascertain the optimal conditions, all data obtained from the multiple-channel reactors were normalized to the maximum value recorded within each set of parallel experiments.

SAXS

SAXS measurements were performed on a Xeuss 2.0 SAXS/WAXS System (Xenocs) utilizing a X-ray of λCu = 0.152 nm. Sample preparation was conducted as follows: excess surfactants in the NP dispersion were removed via ultrafiltration centrifugation, and the dispersion was concentrated to a final concentration of 1 mg mL−1. The concentrated sample was dropped onto a polyimide film substrate, followed by drying at 40 °C for 2 h. The intensity data were subjected to analysis predicated upon the scattering vector (q), which is meticulously centered around the beam’s coordinates.

Cryo-TEM

Cryo-TEM measurements were performed on a Themis 300 TEM (ThermoFisher Scientific). Samples were prepared using an automated plunge freezer: 3.5 μL of the NP dispersion was deposited onto the grids, followed by a 1 s blotting duration and immediate plunging into liquid ethane. Subsequently, Cryo-TEM analysis was conducted with the microscope operating at an acceleration voltage of 300 kV.

TA

Data was collected using a femtosecond TA spectroscopy system (Vitara T-Legend Elite-TOPAS-Helios-EOS-Omni). The method of fabrication of samples was illustrated in NP Fabrication. The optical path length of the aqueous NP suspension was 1 mm. The relative polarizations of pump and probe pulses were parallel. The instrument response function was approximately 120–140 fs. The pump beam was modulated at 500 Hz with a pulse duration of 100 fs and focused to a spot diameter of 2–3 mm. With a pump power ranging from 20 to 80 μW, the pump fluence was calculated to be 0.82–3.27 μJ cm−2, based on the average spot diameter of 2.5 mm.

ICP-MS

ICP-MS (5800 VDV ICP-OES, Agilent) was employed for quantifying the concentration of residual Pt cations after the photodeposition process. NPs (stabilized by TEBS or PEBS) were subjected to platinization for 4 h under the same experimental conditions as the photocatalytic tests including AM1.5 G simulated sunlight (100 mW cm⁻²), 0.05 mg NPs, 7.5 mL reaction system volume, 0.2 M AA as the sacrificial agent, and a corresponding quantity of Pt derived from K2PtCl6 solution. Then, the NPs were isolated by ultrafiltration centrifugation, and the filtrate was used for ICP-MS analysis.

EQE

EQE measurements were conducted in a reactor with a cross-sectional area of 5.786 cm² and a light penetration depth of 1.3 cm, where the total volume of the reaction solution was maintained at 7.5 mL and assessed using an all-glass automatic online trace gas analysis system (Labsolar 6 A, PLS-SXE300 + , PerfectLight), equipped with an online gas chromatograph featuring a thermal conductivity detector and N2 as the carrier gas. Monochromatic lights at wavelengths of 400, 500, 600, 700, and 800/808 nm were sourced from a 300 W xenon lamp equipped with corresponding band-pass filters respectively. Prior to the EQE test, the catalyst sample was irradiated with AM 1.5 G simulated solar light (100 mW cm−2) for 2 h to realize in-situ photodeposition of Pt. Subsequently, the reactor was evacuated to eliminate all hydrogen gas produced during this process. The EQE was calculated using equations10:

$${{\rm{EQE}}}=2\times \frac{{{\rm{moles}}}\; {{\rm{of}}}\; {{\rm{hydrogen}}}}{{{\rm{moles}}}\; {{\rm{of}}}\; {{\rm{incident}}}\; {{\rm{photons}}}}$$
(3)
$${n}_{p}=\frac{{It}\lambda }{{N}_{A}{hc}}$$
(4)

where np represented the moles of incident photons and I was the radiant power. λ and t were the wavelength and the irradiation time (excluding the induction time). h and NA represented Planck constant and Avogadro constant, which are 6.626 × 10−34 J·s and 6.022 × 1023 mol−1, respectively. c was the speed of light, which is 3 × 108 m s−1. The uncertainty of the EQE measurements was estimated to be 5–10%.