Electrostatic [FeFe]-hydrogenase–carbon nitride assemblies for efficient solar hydrogen production

The assembly of semiconductors as light absorbers and enzymes as redox catalysts offers a promising approach for sustainable chemical synthesis driven by light. However, achieving the rational design of such semi-artificial systems requires a comprehensive understanding of the abiotic–biotic interface, which poses significant challenges. In this study, we demonstrate an electrostatic interaction strategy to interface negatively charged cyanamide modified graphitic carbon nitride (NCNCNX) with an [FeFe]-hydrogenase possessing a positive surface charge around the distal FeS cluster responsible for electron uptake into the enzyme. The strong electrostatic attraction enables efficient solar hydrogen (H2) production via direct interfacial electron transfer (DET), achieving a turnover frequency (TOF) of 18 669 h−1 (4 h) and a turnover number (TON) of 198 125 (24 h). Interfacial characterizations, including quartz crystal microbalance (QCM), photoelectrochemical impedance spectroscopy (PEIS), intensity-modulated photovoltage spectroscopy (IMVS), and transient photocurrent spectroscopy (TPC) have been conducted on the semi-artificial carbon nitride-enzyme system to provide a comprehensive understanding for the future development of photocatalytic hybrid assemblies.


Introduction
Converting solar energy into clean chemical fuels, such as molecular hydrogen (H 2 ), holds promise for advancing the concept of a circular economy. 1 Among various photocatalysts, carbon nitride (CN X ) has emerged as a particularly attractive candidate due to its unique advantages, including visible light absorption, cost-effective fabrication, scalability, and low toxicity. 2To further enhance the photocatalytic performance of CN X , signicant efforts have been devoted to chemical modications and the incorporation of co-catalysts. 3The introduction of ionic cyanamide functional groups into CN X ( NCN CN X ) has demonstrated substantial improvements in charge separation and photocatalytic activity, attributed to the prolonged lifetimes of photogenerated electrons. 4,5Moreover, the negatively charged cyanamide group provides a versatile platform for potential electrostatic interactions with co-catalysts.
Nature has evolved enzymes as highly specic biological catalysts to facilitate essential processes in living organisms.Among these enzymes, hydrogenases (H 2 ases) stand out for their remarkable ability to catalyze the interconversion of protons and H 2 with high efficiency at near-zero overpotential under mild conditions, 6 surpassing the capabilities of synthetic catalysts. 7H 2 ases can be classied into three main types based on their metal cofactors: [NiFe]-H 2 ase, [FeFe]-H 2 ase, and [Fe]-H 2 ase, with [FeFe]-H 2 ase being generally the most active for the hydrogen evolution reaction (HER). 6The extensive investigation of H 2 ases as model biocatalysts 8 has not only inspired the design of articial systems such as synthetic Fe 2 S 2 (CO) 6 catalysts that mimic the active site of the Fe 2 S 2 subunit of the [FeFe]-H 2 ase (Fig. S1 †), 9,10 but also paves the way for developing biohybrid assemblies in semi-articial photosynthesis systems. 11,12y interfacing CN X with H 2 ase, we combine the strengths of both articial and biological approaches, resulting in unique properties that neither system can achieve individually. 11,12This integration opens up new avenues for exploring synergistic effects and unlocking unprecedented possibilities in solar energy conversion and catalysis.The activation of [FeFe]-H 2 ase by light can be considered as a model for the development of efficient bio-hybrid systems.Such a photocatalytic system has thus far been demonstrated using either toxic and expensive CdTe nanocrystals 13 or carbon dots with a low turnover number (TON) of 20 000 over 24 h. 14Direct electron transfer (DET) between graphitic carbon nitride (g-C 3 N 4 ) and [NiFeSe]-H 2 ase has been established with non-specic interactions, resulting in a turnover frequency (TOF) of 4117 h −1 over 4 h. 15Subsequent improvements involved the incorporation of a non-diffusional electron mediator, TiO 2 , between g-C 3 N 4 and [NiFeSe]-H 2 ase, leading to an enhanced TON (4 h) of 80 000. 16In addition to its application in H 2 ase systems, CN X has predominantly been utilized for the regeneration of NADH in mediated electron transfer (MET) processes involving formate dehydrogenase 17 or alcohol dehydrogenase. 18n this work, we present an approach for biological integration with CN X by demonstrating the electrostatic interaction with enzymes to form a functional biohybrid assembly.This method establishes a benchmark for solar H 2 production, complemented by comprehensive interfacial characterizations utilizing a quartz crystal microbalance (QCM), photoelectrochemical impedance spectroscopy (PEIS), intensitymodulated photovoltage spectroscopy (IMVS), and transient photocurrent spectroscopy (TPC).Specically, we coupled negatively charged NCN CN X with H 2 ases containing different surface charges for in vitro photocatalytic H 2 production without an external electron relay.The adsorption process of H 2 ases on NCN CN X is quantied by QCM, whereas PEIS provides insights into the charge carrier dynamics at the biomaterial interface.

Results and discussion
The NCN CN X photocatalysts were synthesized using melamine and potassium thiocyanate following previously published methods. 4,19Detailed synthesis procedures and characterizations, including scanning electron microscopy (SEM), attenuated total reectance Fourier-transform infrared (ATR-FTIR) spectroscopy, uorescence spectroscopy, and ultraviolet-visible spectroscopy are provided in the ESI (Fig. S2-S5).† [FeFe]-H 2 ase from Clostridium pasteurianum (CpI, heterologously produced in Escherichia coli) and [NiFeSe]-H 2 ase from Desulfovibrio vulgaris Hildenborough (DvH) were expressed and puried under anaerobic conditions. 20,21g. 1 illustrates the hypothesis that the negatively charged NCN CN X possesses the ability to activate enzymes with positively charged electron entry points, such as CpI [FeFe]-H 2 ase. 22dditionally, the unique property of NCN CN X in converting alcohols selectively into aldehydes offers an opportunity to monitor the clean oxidation reaction of 4-methylbenzyl alcohol (4-MBA) to 4-methylbenzaldehyde (p-tolualdehyde), allowing the quantication of stoichiometry from the products resulting from oxidation and reduction. 5,23o construct a photocatalytic system for solar H 2 production coupled with selective alcohol oxidation, we interfaced [FeFe]-H 2 ase with NCN CN X in the presence of 50 mM 4-MBA in 1 mL aqueous MOPS buffer solution (0.1 M, pH 7).To determine the surface charge of NCN CN X , zeta potential measurements were performed.Notably, the presence of light-induced blue radicals (absorption band from 500-750 nm, Fig. S5 †) 5,24 did not impact the surface charge of NCN CN X (Fig. S6 †), indicating that these radicals are long-lived and deeply trapped photoelectrons. 23,25he negatively charged surface is primarily attributed to the cyanamide group, which maintains a negative zeta potential even at a pH below 2 (Fig. S7 †).CpI [FeFe]-H 2 ase was selected due to its distal FeS cluster ([4Fe-4S]) being surrounded by a positively charged region containing surface arginine and lysine residues.In vivo, this distal FeS cluster region is thought to interact with the negatively charged region of ferredoxin for electron transfer. 22By constructing electrostatic NCN CN X j[FeFe]-H 2 ase assemblies, we achieve efficient solar H 2 production, with NCN CN X mimicking the role of ferredoxin to deliver electrons directly into [FeFe]-H 2 ase (Fig. 1).Time-dependent photocatalytic H 2 evolution using NCN CN X j[FeFe]-H 2 ase complexes is illustrated in Fig. 2a.The TON is determined by the ratio between the number of moles of product (H 2 ) and the number of moles of catalyst (H 2 ase) and the TOF is calculated by the TON per hour.Notably, a nearly linear increase in H 2 yield is observed during the initial 4 h, reaching 3.0 ± 0.3 mmol with a TOF of 18 669 h −1 .This TOF value is approximately 4.5 times higher than the previous benchmark (4117 h −1 ) 15 and is even comparable to systems utilizing MET such as g-C 3 N 4 jTiO 2 (TOF = 20 000 h −1 ) 16 (note that previous systems used Desulfomicrobium baculatum (Dmb) [NiFeSe]-H 2 ases).Continuous irradiation of the NCN CN X j[FeFe]-H 2 ase assemblies for 24 h yielded 7.9 ± 0.6 mmol of H 2 (TON = 198 125).The efficient DET between NCN CN X and [FeFe]-H 2 ase can be attributed to the specic electrostatic interaction at the interface, which will be further evaluated by QCM and PEIS.Furthermore, 4-MBA was selectively oxidized to p-tolualdehyde (Fig. S8-S10), † with a H 2 : p-tolualdehyde ratio of 0.77.The observed ratio indicates the deep trapping of some photogenerated electrons within the CN X polymeric structure in addition to some buffer (MOPS) oxidation (see below).
Recent transient spectroscopic and electron paramagnetic resonance (EPR) studies extensively characterized the deep traps stored in NCN CN X on the time scale from ps to s. [23][24][25][26] EPR analysis showed that these long-lived and deeply trapped photoelectrons emerged as blue radicals, processing a symmetric Gaussian line near the free electron g value at Xband frequency of ∼9.6 GHz. 24These blue radicals can also be visualized by UV-vis spectroscopy (Fig. S5 †) and qualitatively by the eye (Fig. S11 †).The oxidation ability of NCN CN X has been further evaluated using glycerol (a waste product from the biodiesel industry) as the reductant on a model NCN CN X jPt (2 wt%) system.The oxidation products are quantied as glyceraldehyde at 134.2 ± 5.7 mmol h −1 g −1 and dihydroxyacetone at 54.9 ± 6.8 mmol h −1 g −1 (Fig. S12-S15) † and the H 2 yield is 94 ± 9 mmol h −1 g −1 .Replacing 4-MBA with ethylenediaminetetraacetic acid (EDTA) led to a ∼16% enhancement of H 2 yield (4 h) with NCN CN X j[FeFe]-H 2 ase assemblies (Table S1 †).This indicates that, under DET conditions, the rate of 4-MBA oxidation is limiting compared to HER.Despite this observed higher activity, it is worth noting that EDTA is considered as a sacricial electron donor, and its oxidation results in a range of products that cannot be easily characterized.As a result, the primary focus of this study is the conversion of 4-MBA to ptolualdehyde, serving as a model reaction for the selective oxidation of alcohols to aldehydes.
Exclusion controls were conducted by removing individual components from the photocatalytic system (Table S1 †).As depicted in Fig. 2a and S16, † the system exhibited signicantly reduced efficiency in the absence of any component, with H 2 yields below 0.3 mmol over 4 h.Minor H 2 evolution activity was observed with NCN CN X j[FeFe]-H 2 ase assemblies even in the absence of 4-MBA, yielding 285 ± 24 nmol H 2 in 4 h.This observation suggests that MOPS serves as a much less efficient electron donor in the photocatalytic reaction.NMR analysis (Fig. S17-S20 †) provides evidence of MOPS oxidation during photocatalysis.However, it is noteworthy that alcohol oxidation on NCN CN X is so efficient and selective that the photocatalytic activity of MOPS oxidation is only ∼5% compared to 4-MBA oxidation in 24 h (Fig. 2a).Upon replacing MOPS buffer with pH 7 phosphate buffer, the reaction is drastically reduced in the absence of 4-MBA, accompanied by a ∼27% decrease in photocatalytic activity in the presence of 4-MBA (Fig. S21, Table S2 †).This observation is in line with literature that MOPS as a standard Good's buffer can maintain high in vitro biochemical and biological activities. 27Notably, control experiments were also performed to validate the proposed electrostatic interactions using a DvH [NiFeSe]-H 2 ase, differing from the previously reported Dmb [NiFeSe]-H 2 ase. 15,16The distal FeS cluster of both [NiFeSe]-H 2 ases near its surface is surrounded by amino acids that lead to a local negative charge, serving as the electron entry point for interaction with the positively charged heme of cytochrome c 3 during electron transfer in vivo. 28Interfacing DvH [NiFeSe]-H 2 ase with NCN CN X resulted in the production of 8.6 ± 0.33 nmol of H 2 in 4 h, with a signicantly lower TON of 215 (Fig. 2a).These results indicate that electrostatic repulsion prevents DET in this system.A detailed comparison among state-of-the-art photocatalytic systems combining carbonaceous photocatalysts and H 2 ase are listed in Table 1.
To determine the charge transfer efficiency of DET, methyl viologen (MV) as a soluble electron mediator is used to activate MET (Fig. 2a).Note that the presence of MV may suppress DET due to the kinetic and thermodynamic favorable one electron reduction of MV molecules to MV radicals. 25,30Upon the addition of MV (2 mM), the H 2 yield reaches 24.0 ± 0.  ).These ndings align with recent observations on carbon dotj[FeFe]-H 2 ase photocatalytic systems. 14o gain deep insights into the interaction between NCN CN X and H 2 ase, QCM analysis was conducted.As shown in the schematic in Fig. 3a, the Au-coated quartz chip was functionalized with a thin layer of NCN CN X by drop casting 0.5 mL of an ultrasonicated suspension (0.1 mg mL −1 ) of NCN CN X , to mimic the operando conditions during photocatalysis.By owing a buffer solution containing enzymes on the chip, the adsorption process of H 2 ase at the surface of NCN CN X can be monitored as a function of time and quantied based on the Sauerbrey equation. 31Fig. 3b shows the QCM analysis of CpI [FeFe]-H 2 ase and DvH [NiFeSe]-H 2 ase on the NCN CN X -modied chip.Aer establishing a stable baseline by circulating 0.1 M MOPS pH 7 buffer with 50 mM 4-MBA, the enzymes were introduced separately at the same concentration as in the photocatalysis experiments.The adsorption of both H 2 ases on NCN CN X exhibits in two distinct stages, a fast adsorption process before 1.5 h and slow adsorption aer 1.5 h.Interestingly, a higher amount of [NiFeSe]-H 2 ase (39.5 pmol cm −2 ) is adsorbed onto NCN CN X compared to [FeFe]-H 2 ase (16.6 pmol cm −2 ) over 10 h.The observed adsorption proles can be explained by the proposed electrostatic interactions in Fig. 3c.Based on the electrostatic potential maps (Fig. 3c), both H 2 ases exhibit distinct surface charge distribution.By indexing the specic protein structures, CpI [FeFe]-H 2 ase (PDB: 4XDC), and DvH [NiFeSe]-H 2 ase (PDB: 5JSH), within the protein dipole moment database, 32 it is found that [NiFeSe]-H 2 ase possesses a larger dipole moment of 1972 D compared to [FeFe]-H 2 ase (1707 D), resulting in a stronger association (Fig. 3b).However, for [NiFeSe]-H 2 ase, DET can only be established via the negatively charged patch near the distal FeS cluster, which is unfavorable for the negatively charged NCN CN X and thus dramatically reduces DET to the enzyme active site for catalysis.Consequently, the strong association observed at the NCN CN X j[NiFeSe]-H 2 ase interface is non-specic and results mainly in inactive biohybrid assemblies.In contrast, absorbed [FeFe]-H 2 ase has positively charged distal FeS cluster that can specically interact with negatively charged NCN CN X for DET. 22,28The presence of other positively charged regions in the [FeFe]-H 2 ase (Fig. 3c) might also attract NCN CN X .However, due to the rigidity of the heptazine-based NCN CN X (A-B 0 stacking), proper orientation for DET near the distal FeS cluster could be hindered.This might explain the observed relatively low 25% DET/MET ratio.Therefore, QCM analysis provides valuable insights into the signicance of specic interactions in facilitating DET. 33o investigate the charge carrier dynamics between H 2 ases and NCN CN X , PEIS was performed using a three-electrode conguration.By applying a sinusoidal potential modulation to the NCN CN X -modied working electrode, which was made by depositing a NCN CN X suspension (5 mL, 24 mg mL −1 ) on FTOcoated glass, 23 the impedance was recorded as the ratio of the complex-valued potential and current. 34A Randles circuit consisting of a series resistance (R S ) in parallel with a combination of bulk capacitance (C bulk ) and charge transfer resistance (R ct ), was proposed to t the impedance response (Fig. 4a). 35The Nyquist plot (Fig. 4a) of the impedance response measured at −0.1 V vs. the reversible hydrogen electrode (RHE), is dominated by a single semicircle with no indication of a Warburg diffusion element.The proposed equivalent circuit provided a good t (r 2 > 0.95) to the impedance response, enabling quantitative analysis of the charge transfer process.
Upon introduction of [FeFe]-H 2 ase on the working electrode, a decrease in the semicircle diameter is observed, corresponding to a decrease in R ct from 15 975 U to 12 317 U.This indicates that [FeFe]-H 2 ase, as a biocatalyst, facilitates the charge transfer from NCN CN X to the electrolyte for HER.Likewise, NCN CN X j[NiFeSe]-H 2 ase shows a R ct of 13 950 U, similar to the bare NCN CN X .Such behaviors have been widely observed when incorporating synthetic co-catalysts onto semiconductors. 36The tting results allowed determination of the pseudo rst-order rate constant for charge transfer (k ct ), based on the phenomenological model developed for an illuminated photoelectrode. 37,38Specically, the angular frequency at the maximum imaginary component of the semicircle in Nyquist plot (Fig. 4a) is equal to k ct .The addition of [FeFe]-H 2 ase signicantly enhances k ct from 6.85 s −1 to 11.99 s −1 , whereas [NiFeSe]-H 2 ase shows a negative impact on k ct with a value of 5.02 s −1 , further conrming the importance of specic interactions in facilitating the charge transfer process.The Bode phase plots (Fig. 4b) revealed that the charge transfer process occurred within the frequency range of 0.1 Hz to 1 kHz, consistent with the reported timeframe for photocatalytic HER using CN X . 36Within this range, the characteristic frequency at the maximum phase shi of NCN CN X j[FeFe]-H 2 ase are higher than NCN CN X j[NiFeSe]-H 2 ase and pristine NCN CN X , indicating [FeFe]-H 2 ase can initiate a faster charge transfer process for HER.Likewise, the characteristic frequencies of NCN CN X j[NiFeSe]-H 2 ase and pristine NCN CN X remain the same, meaning that DET cannot be established between NCN CN X and [NiFeSe]-H 2 ase.The voltage-dependent impedance response is illustrated in Fig. S23.† A more negative applied potential yields a diminished semicircular feature in the Nyquist plots, indicating reduced charge transfer resistance.This observation arises from the introduction of a larger band bending, resulting in improved separation of photogenerated charges. 39,40Consequently, a greater population of free charge carriers is localized within the semiconductor, increasing the conductivity of NCN CN X .Notably, the RC response of the conductive substrate forms a semicircle with a diameter of approximately 200 U in the Nyquist plots (Fig. S24a †) at high frequency region (10 kHz to 1 MHz, Fig. S24b †) in the Bode phase plots. 23This impedance study on CN X with H 2 ase demonstrates that a specic interaction enables efficient DET by decreasing in R ct and increasing in k ct .
The charge carrier dynamics between H 2 ases and NCN CN X were further examined using IMVS and TPC techniques in a three-electrode setup.IMVS is a spectroelectrochemical method widely employed in assessing electron recombination processes in photovoltaics.It monitors the open circuit voltage response to the sinusoidally modulated incident light intensity.The characteristic frequency observed at the minimum point of the Nyquist plot (f min ) directly correlates to the time constant of electron recombination.This parameter can be calculated using the following equation, providing the rst-order electron lifetime s n : 41 This model has been recently expanded to photoanodes for solar water oxidation. 42,43Although IMVS is not operando due to the distinct differences between a photoelectrode and a photocatalyst, photoelectrochemical techniques have been widely employed to gain insights into charge carrier behaviors in photochemical systems.This equation can be applied to a NCN CN X -based photoelectrode due to it functions as a photoanode in the presence of 4-MBA and under open circuit conditions. 23As depicted in Fig. 5a, the IMVS response exhibits a distinct semicircle in quadrant IV of the Nyquist plot, suggesting rapid kinetics in 4-MBA oxidation, similar to the cases of sacricial Na 2 SO 3 and H 2 O 2 oxidation on a hematite photoanode. 42,43he non-specic interaction between [NiFeSe]-H 2 ase and NCN CN X yields s n of 57.0 ms, whereas 55.1 ms is observed for pristine NCN CN X .This consistency indicates that the charge recombination process remains unaffected, matching with our  previous observations of the absence of DET between [NiFeSe]-H 2 ase and NCN CN X .Upon the introduction of [FeFe]-H 2 ase onto NCN CN X , we observe a prolonged s n of 81.6 ms that can be explained as follows: Even under open circuit conditions, where no net electron exchange occurs at the NCN CN X jelectrolyte interface, photogenerated electrons theoretically have the potential to react with protons via H 2 ases.Note that a TOF of 18 669 h −1 ( NCN CN X j[FeFe]-H 2 ase) corresponds to a frequency of 5.2 Hz.Therefore, all losses of photogenerated electrons within the measured frequency range (0.5 MHz to 0.5 Hz) are a combination of bulk recombination and catalytic reaction.The inuence of a catalytic overlayer on IMVS response remains a topic of debate.Recent studies indicate that a co-catalyst overlayer can delocalize photogenerated charge carriers from the bulk photoelectrode, promoting charge separation and prolonging electron lifetime. 42Here, despite the absence of net exchange current, we hypothesize that photogenerated electrons can be stored in H 2 ase in the form of metal hydrides and reduced FeS clusters, 44 i.e., reversible intermediates for H 2 evolution reaction.This storage mechanism reduces the probability of charge recombination with holes, resulting in an extended electron lifetime.IMVS observations are in line with the recent transient spectroscopic study on the impact of electron accumulation to charge recombination in NCN CN X . 25Thus, we demonstrate the use of IMVS on carbon nitride materials and on studying bio-hybrids.
Having gained insights into charge recombination, we conducted TPC measurements to assess the inuence of H 2 ases on the electron extraction process of NCN CN X .The normalized TPC response, illustrated in Fig. 5b, reveals that the integration of [FeFe]-H 2 ase with NCN CN X leads to a signicant reduction in electron transit time (s t ) from 0.28 s to 0.16 s.This indicates that [FeFe]-H 2 ase, acting as a co-catalyst for NCN CN X , can effectively collect photoelectrons for HER, thereby facilitating electron transport within NCN CN X . 45,46Likewise, NCN CN X j[NiFeSe]-H 2 ase assemblies are non-specic, resulting in an unchanged s t of 0.30 s.The combined results from IMVS and TPC highlight the favorable effects of DET between [FeFe]-H 2 ase and NCN CN X on both charge recombination and transport.In contrast, the non-specically interacted NCN CN X j[NiFeSe]-H 2 ase assemblies demonstrate minimal alterations in both s n and s t .

Conclusions
We present an electrostatic strategy for linking enzymes with carbon nitrides, demonstrating a benchmark for DET between NCN CN X and [FeFe]-H 2 ase for solar H 2 production with a TON of 2 × 10 5 and a DET/MET ratio of 25% over 24 h.In contrast, the electrostatic repulsion between [NiFeSe]-H 2 ase and NCN CN X drastically reduced DET, leading to a DET/MET ratio of 0.5%.QCM analysis demonstrates that specic interactions play a pivotal role in enabling DET, irrespective of the observed differences in the adsorption proles.Complementary spectroelectrochemical analysis using PEIS, IMVS, and TPC show that interfacing [FeFe]-H 2 ase with NCN CN X facilitates charge transfer and suppresses charge recombination, as evidenced by a 23% less resistive R ct , a 75% faster k ct , a 48% longer s n , and a 43% shorter s t than bare NCN CN X .This study provides a promising and straightforward approach for achieving efficient electron transfer between carbon nitride and enzymes and serves as a reference for studying the charge carrier behavior of enzyme-photocatalyst assemblies using interfacial characterizations.

Fig. 1
Fig.1Schematic of photocatalytic H 2 evolution coupled with alcohol oxidation to aldehyde using an electrostatic NCN CN X j[FeFe]-H 2 ase (CpI, PDB: 4XDC) assembly.Scale bar refers to enzyme and CN X is not shown to scale.
1 mmol for CpI [FeFe]-H 2 ase and 25.6 ± 1.3 mmol for DvH [NiFeSe]-H 2 ase aer 24 h of irradiation.The comparable H 2 yields, despite differences in specic activity, indicate that the rate limiting step during MET is not enzyme turnover but the photoactivity of NCN CN X , as evidenced by control experiments (Table S1 †).Specically, by doubling the H 2 ase loading from 40 pmol to 80 pmol, no signicant changes in H 2 yields were observed over

Table 1
State-of-the-art photocatalytic systems combining carbonaceous light absorbers and H 2 ases and 24 hour periods under MET conditions.The efficiency of DET is qualitatively determined by the DET : MET ratio, dened by the ratio of H 2 yield in the absence of MV (DET) and in the presence of MV (MET).In the case of [FeFe]-H 2 ase and [NiFeSe]-H 2 ase, DET : MET ratios of approximately 25% and 0.5% are observed, respectively (Fig.S22 †).By comparing the results obtained from DET and MET, a schematic representation can be depicted in Fig.2b.It highlights the establishment of efficient electron transfer directly between NCN CN X and [FeFe]-H 2 ase, with a DET : MET ratio of 25%.This nding emphasizes the effectiveness and benets of electrostatic interactions in facilitating DET.Conversely, the electrostatic repulsion between [NiFeSe]-H 2 ase and NCN CN X prevents DET, resulting in a low DET : MET ratio of 0.5%.Notably, the apparent quantum efficiency (AQE) measured at 450 nm with CN CN X j[FeFe]-H 2 ase assemblies under DET and MET conditions are 0.35% and 1.4% (TableS3†), respectively.In contrast, a model NCN CN X jPt (2 wt%) system yields 0.28 ± 0.05 mmol h−1 g −1 H 2 and 0.51 ± 0.09 mmol h −1 g −1 p-tolualdehyde with an AQE of 1.92% at 450 nm.In terms of the overall stability of the designed systems, the DET system exhibited a rather linear photocatalytic activity up to 12 h.While MET systems are fully inactive only aer 20 h.Long-term experiments up to 36 h conrmed these trends, with no further H 2 production in DET aer 24 h and minimal H 2 yield increases in MET (24.0 to 25.0 mmol for [FeFe] and 25.6 to 26.7 mmol for[NiFeSe]