Photocatalytic hydrogen production using polymeric carbon nitride with a hydrogenase and a bioinspired synthetic Ni catalyst.

Solar-light-driven H2 production in water with a [NiFeSe]-hydrogenase (H2ase) and a bioinspired synthetic nickel catalyst (NiP) in combination with a heptazine carbon nitride polymer, melon (CN(x)), is reported. The semibiological and purely synthetic systems show catalytic activity during solar light irradiation with turnover numbers (TONs) of more than 50,000 mol H2(mol H2ase)(-1) and approximately 155 mol H2 (mol NiP)(-1) in redox-mediator-free aqueous solution at pH 6 and 4.5, respectively. Both systems maintained a reduced photoactivity under UV-free solar light irradiation (λ>420 nm).

Efficient and noble metal-free water photolysis using sunlight is a primary focus of research to advance sustainable solar energy generation.[1] Photocatalytic H2 production can be achieved by employing hybrid systems with a solid-state light absorber such as an inorganic semiconductor assisted by a metallic, synthetic, or enzymatic electrocatalyst.[2] These systems typically contain expensive, inefficient, and/or unstable components, but high performance solar fuel devices need to be constructed from parts without these limitations.

Hydrogenases (H2ases) are H2-cycling enzymes and are by far the most efficient noble-metal-free electrocatalysts for H2 generation with an unrivalled turnover frequency (TOF) benchmark of more than 103 s−1 even at a modest overpotential.[3] This excellent electrocatalytic activity of H2ases was exploited in photocatalytic schemes with a light absorber in the absence of a soluble redox mediator: a homogeneous photocatalytic system with a molecular organic dye,[4] and semiheterogeneous systems, in which the H2ase is immobilized on Ru dye-sensitized TiO2 nanoparticles,[5] and on Cd-containing quantum dots,[6] displaying excellent photocatalytic activity in sacrificial schemes.

An efficient class of H2ase-inspired synthetic catalysts containing non-noble metal centers have been developed by DuBois and co-workers.[7] They possess a Ni bis(diphosphine) ligand core bearing pendant amino groups, which, much like those found in the active site of [FeFe]-H2ases,[8] can act as catalytically active proton relays in the second coordination sphere of the 3d metal center. Photocatalytic H2 generation with such Ni bis(diphosphine) catalysts has only been achieved in combination with a costly Ru dye in purely aqueous solution.[9]

Amorphous polymeric carbon nitride (CN) with a poly(tri-s-triazine) (polyheptazine) building block (often referred to as melon or g-C3N4) has recently emerged as an attractive visible-light absorber and can generate H2 photocatalytically.[10] It can be easily synthesized by condensation of cyanamide, dicyandiamide, or melamine at elevated temperatures and displays high activities and photostability of more than 72 h.[10b] The material has well-suited band positions for water splitting and a band gap of approximately 2.7 eV with a conduction band potential at −0.8 V vs. RHE.[10a,b] Co-catalyst integration of non-noble metals,[11] Pt,[10b], [12] Ni(TEOA)32+ (TEOA=triethanolamine),[13] and cobaloximes[14] with CN has previously been used as a strategy to enhance H2 evolution rates.

In this study, we report a photocatalytic CN–enzyme hybrid system for visible-light-driven H2 generation (Figure 1). This CN–H2ase hybrid assembly operates in an aqueous sacrificial electron donor solution and does not require an expensive or fragile light absorber for visible light promoted photocatalysis as reported for the previous H2ase-based systems. We selected Desulfomicrobium baculatum (Dmb) [NiFeSe]-H2ase because of its well-known[15] and excellent H2 evolution rate as well as tolerance toward H2 and O2, allowing for the accumulation of H2 during irradiation and the handling of the enzyme in the absence of strictly anaerobic conditions.[4], [5], [15b]

Figure 1

Representation of the photo-H2 production with CN and Dmb [NiFeSe]-H2ase (PDB ID: 1CC1)[15a] or CN and NiP (counterions omitted) in aqueous EDTA solution. Irradiation of CN results in the photoinduced direct electron transfer to the catalysts with H2 formation and hole quenching in CN by EDTA.

The photocatalytic H2 generation systems were assembled in a photoreactor (total volume 7.74 mL) by dispersing CN (5 mg, approx. 1 μm-sized particles with a Brunauer–Emmett–Teller surface area of 9 m2 g−1; see Figures S1–S4 for FTIR, XRD, SEM, and zeta-potential measurements) in an aqueous solution of the electron donor (0.1 m, 3 mL). The catalyst (H2ase or NiP, see below) was added to the suspension and the light-protected reactor was sealed and purged with 2 % CH4 (as an internal gas chromatography standard) in N2 before irradiating the stirred mixture at 25 °C. Irradiation was provided by a solar light simulator (air mass 1.5 G, 100 mW cm−2) and headspace H2 was quantified at regular time intervals by gas chromatography. The reaction conditions were optimized for a high rate of H2 production per catalyst (as expressed by the TOF) by varying the pH of the solution, the amount of catalyst and by screening different electron donors (Table S1; Figures S5 and S6).

The optimized standard system for CN–H2ase comprises 50 pmol H2ase with 5 mg melon in 3 mL ethylenediamine tetraacetic acid (EDTA, 0.1 m) at pH 6 under simulated solar irradiation at λ>300 nm (Figure 2). Under these conditions, a TOF of (5532±553) mol H2 (mol H2ase)−1 h−1 and (55.3±5.5) μmol H2 (g CN)−1 h−1 are photogenerated with almost linear H2 evolution, producing (0.82±0.08) μmol H2 during the first four hours. Photoinduced direct electron transfer from CN to the H2ase was therefore observed, making a soluble redox mediator unnecessary. The CN–H2ase suspension was photoactive for 48 h, whereupon (2.5±0.2) μmol of H2 was produced with a TON of >50 000. Control experiments in the dark and in the absence of CN or H2ase showed no H2 formation. Only minimal amounts of H2 were produced when the electron donor buffer EDTA was replaced by potassium phosphate buffer (KPi; 41 mm, pH 7, Table S1).

Figure 2

H2 production under optimized conditions using Dmb [NiFeSe]-H2ase (50 pmol) in EDTA (pH 6, 0.1 m, 3 mL) and CN (5 mg) under 1 sun irradiation in the absence (λ>300 nm) and presence of a 420 nm UV filter. Control experiments without EDTA, CN, or H2ase are also shown.

The amount of H2ase per mg of CN and the light intensity were varied to gain insight into the performance-limiting factors of the CN–H2ase hybrid. Increasing the H2ase loading from 50 to 200 pmol per 5 mg CN resulted in a linear increase in overall H2 generation with an unchanged TOF (Table S1, Figure S6). Decreasing the solar light intensity with neutral density filters from 100 to 50 mW cm−2 did not result in a significant reduction of the photoactivity, although a further reduction to 20 mW cm−2 resulted in approximately 40 % decreased activity (Table S2; Figure S7). These experiments suggest that the optimized CN–H2ase system is not limited by light absorption at CN, and support that enzyme adsorption and interaction with the CN is performance limiting (see below).

The CN–H2ase system was also studied under visible light irradiation (λ>420 nm). A decrease in photoactivity was observed giving rise to a TOF of (768±77) h−1, which corresponds to 14 % of the activity under UV/Vis irradiation (Figure 2). This can be attributed to the significantly reduced light absorption of CN above 420 nm (Figure S8). The external quantum efficiency (EQE) of the system was determined by irradiation of samples under standard conditions using a monochromatic LED light source at two wavelengths (λ=365 nm, I=3.5 mW cm−2 and λ=460 nm, I=47 mW cm−2). UV-irradiation gave an unoptimized EQE of approximately 7×10−2 %, whereas an EQE of 5×10−3 % was obtained at λ=465 nm (Figure S8).

A centrifugation test was performed to gain insight into the strength of interaction between the enzyme and CN particles. First, H2 production was monitored for 2 h with CN–H2ase under standard conditions. The suspension was then centrifuged (5000 rpm, 5 min) followed by washing the pellet with water and redispersion of the particles in aqueous EDTA (0.1 m, pH 6). This suspension was then irradiated again after purging the headspace with 2 % CH4 in N2. The remaining activity of this mixture was 12 % relative to the activity prior to centrifugation, indicating that a relatively weak interaction suffices for electron transfer to occur from CN to the H2ase. Physical adsorption of the H2ase on the CN surface can be expected and we speculate that the H2ase[16] may form hydrogen bonds with the =NH=, terminal =NH2 or Lewis basic heptazine edge nitrogens in CN.[10a], [17] The isoelectric point of CN was determined by zeta potential measurements as 3.3[18] and, at pH 6, the surface of CN is therefore negatively charged (≈−15 mV) (Figure S4).

Although the direct electron transfer was observed from the photoexcited CN to the H2ase, the CN–H2ase system displayed a significantly increased photoactivity under standard conditions upon addition of an excess of the redox mediator, methyl viologen (MV),[19] producing up to 18.7 μmol H2 after 4 h (Figure S9). A long-term experiment with H2ase (50 pmol), CN (5 mg), and added MV (5 μmol) in aqueous EDTA (0.1 m) at pH 6 was also performed. The photoreactor was purged with 2 % CH4/N2 after 24 and 48 h and additional MV (5 μmol) was added at the same time intervals. After 69 h, the CN–MV–H2ase system produced 77 μmol H2 with a TON of 1.5×106 and an initial TOF of 12.3 s−1 (Figure S10). Replenishment of MV was required due to decomposition of the organic mediator during irradiation. The substantially increased H2 production activity in the presence of MV suggests that the electron transfer from CN to H2ase is not yet fully optimized, presumably due to weak and nonspecific interactions at the CN–H2ase interface.

Steady-state photoluminescence (PL) measurements were also performed with the CN in suspension upon photoexcitation at λ=365 nm and following the PL emission at 450 nm (Figure S11). The PL emission of sonicated CN (0.22 g mL−1 in 0.1 m EDTA pH 6) is more strongly quenched upon addition of 50 pmol MV compared to 50 pmol H2ase. These results further support that the photoinduced electron transfer from CN to MV is more efficient than that to the H2ase.

The reported semibiological hybrid system provides a novel “per active site” activity benchmark for a cocatalyst on a CN material.[7g], [11a,b], [20] Photocatalytic H2 generation schemes previously reported with H2ases and other light absorbers show a high TOF (approximately 106 h−1), but these systems rely on an expensive (Ru dye), toxic (Cd-based quantum dot), and/or fragile (organic dye) visible light absorber.[4], [5b], [6c] This study demonstrates that the biocompatibility of CN can be exploited to overcome these limitations and that by improving the coupling of CN to the H2ase, the photoactivity will be further enhanced.

Successful H2 production with CN–H2ase prompted us to investigate a water-soluble and functional synthetic H2ase-mimic, [NiII(PPh2{NPhCH2P(O)(OH)2}2)2]Br2 (NiP; Figure 1),[9] for comparison. Ni bis(diphosphine)[7a–e] complexes are among the most active H2 generation electrocatalysts and, importantly, NiP has recently been shown to act as an excellent electrocatalyst in aqueous solution.[9] The purely synthetic CN–NiP assembly is photoactive and conditions were optimized for the highest TOFNiP. Aqueous EDTA solutions (0.1 m) at pH 4.5 containing NiP (20 nmol) and suspended CN (5 mg) were studied under simulated solar irradiation at λ>300 nm (Table S3, Figures S12–S14). Under these conditions, solar H2 generation by CN–NiP gave an initial activity of (437.1±43.7) μmol H2 (g CN)−1 h−1 producing (2.2±0.2) μmol H2 in the first hour and giving a TOFNiP of (109.3±10.9) mol H2 (mol NiP)−1 h−1. CN-NiP was photoactive for three hours, whereupon (3.3±0.4) μmol of H2 with a TON of (166.1±20.6) mol H2 (mol NiP)−1 was produced (Figure 3). A 64 % decrease in photocatalytic H2 generation yield was observed for CN–NiP when irradiating with λ>420 nm instead of >300 nm solar light. Decomposition of NiP is the likely reason for the ceased activity after three hours, because the photoactivity is fully regenerated if additional NiP is added (Figure S15).

Figure 3

H2 production under optimized conditions using NiP (20 nmol) in aqueous EDTA (0.1 m, pH 4.5, 3 mL) and CN (5 mg) under 1 sun irradiation. Data collected under standard conditions (λ>300 nm), with UV-light-filtered irradiation (λ>420 nm) and control experiments without NiP catalyst or CN are also shown.

Photo-H2 generation with CN–NiP is thus significantly higher than for previously reported CN systems with immobilized noble-metal-free cocatalysts in aqueous solution. A TOF of <0.5 h−1 and a TON of 4 was reported for a cobaloxime, [CoCl(dimethylglyoximato)2(pyridine)], after 8 h irradiation with CN in aqueous TEOA at pH 10.4.[14a] Other systems comprising a cobaloxime with a pyrene-functionalized pyridine[14c] and NiCl2 with TEOA and CN[13] showed TONs of 160 and 281 and TOFs of approximately 40 and 6.7 h−1, respectively, but required excess organic solvent. Previously, photo-H2 generation with NiP was only reported with a molecular Ru dye, in which a TOFNiP of up to 460 h−1 and a TONNiP of up to 723 in pH 4.5 ascorbic acid solution were reported.[9]

The photo-H2 generation activity of CN–NiP is dependent on the NiP concentration (Figures S13 and S14) and reduction of the light intensity (I) with neutral density filters has a substantial impact on the photoactivity. The NiP-based TOF decreases from (71.1±7.1) h−1 (I=100 %) to (32.4±3.2) (I=50 %) and (13.1±1.4) h−1 (I=20 %; Table S4; Figure S16). The purely synthetic system is therefore limited both by catalyst concentration and light absorption. The unoptimized EQE for the CN–NiP system was determined to be (0.37±0.02) % under UV light (λ=365 nm) and (0.04±0.01) % under blue light irradiation (λ=460 nm) after 2 h. The wavelength-dependent EQE is consistent with the decrease in light absorption by CN at higher wavelengths (Figure S17).

Centrifugation experiments in analogy to the enzyme system were performed to examine the strength of the interaction between CN and NiP. After centrifugation, washing, and redispersion in fresh EDTA solution, 8 % of the photoactivity remained for the synthetic system implying a weak interaction between the CN and NiP (Figure S18). Electronic absorption spectrophotometry was used to quantify the amount of NiP adsorbed to CN. By comparing UV-visible spectra of NiP (6.7 μm; λmax=320 and 450 nm) in aqueous EDTA solution (3 mL; 0.1 m, pH 4.5) before and after the addition of CN and centrifugation, an estimate of approximately 20 % NiP was adsorbed on CN (Figure S19). The physical adsorption and H-bonding between the phosphonic acid groups in NiP and the terminal =NH2 and =NH= groups in CN are possible modes of interaction.[10a], [17]

The addition of MV (20 μmol) to a standard photocatalytic experiment showed an approximately 20 % decreased H2 production activity. The reaction mixture turned dark blue upon irradiation, indicative of the presence of reduced MV, and implies that MV successfully scavenged electrons from the photoexcited CN, but was unable to transfer them to NiP (Figure S20).

The comparison of the CN–H2ase with the CN–NiP hybrid system shows the expected higher “per active site” activity of the enzymatic system, whereas the purely synthetic system shows an overall higher H2 production rate due to the larger amount of NiP (20 nmol) used compared to H2ase (50 pmol). Thus, we also studied the CN catalyst systems with the same amount of NiP and H2ase (200 pmol) on CN (5 mg) in aqueous EDTA solution (pH 4.5 and pH 7.0, respectively). At the same concentration, the enzyme (TOF=2528 h−1) greatly outperforms the NiP cocatalyst (TOF=64 h−1), demonstrating that substantial improvements are still required to develop synthetic catalysts with activities comparable to enzymes (Figure S21, Tables S1 and S3).

Finally, we photodeposited 1 wt % Pt onto CN (5 mg) for a direct comparison of this benchmark system with CN–H2ase and CN–NiP. Following a standard procedure,[12] the platinized CN system was irradiated with visible light (λ>420 nm) in an aqueous 10 vol % TEOA solution, generating 94 μmol H2 (g CN)−1 h−1, which corresponds to a TOFPt of 4.3 mol H2 (mol Pt)−1 h−1. Thus, the CN–H2ase and CN–NiP systems compare favorably when using TOF as the metrics of system performance.

In summary, solar-light-driven H2 production with hybrid systems consisting of polymeric CN with H2ase and the bioinspired synthetic catalyst, NiP, has been demonstrated. The systems operate without a soluble redox mediator and are not limited by a photo-unstable or expensive dye. The semibiological CN–H2ase assembly achieved a record TOF of 5532 h−1 and TON of >50 000 after two days as a cocatalyst with CN. The additional use of the redox mediator MV allowed for the photogeneration of H2 with a TOF of 12.3 s−1 and a TON of >1×106, which displays the further potential of the hybrid assembly after optimization of the biomaterial interface. CN–H2ase also maintains respectable activity under visible light irradiation for more than 48 h. Recent investigations into improving the absorption profile of CN in the visible range demonstrate the potential of this material and illustrate that its use as a light-harvesting material will continue to develop, as its absorption profile is further improved.[21] The entirely synthetic CN–NiP system displays an unprecedentedly high TOF (109 h−1) and TON (166) for a hybrid system made of a molecular cocatalyst with CN in purely aqueous solution. This work advances the use of hybrid photocatalytic schemes by integrating highly active electrocatalysts with the photostable and inexpensive CN, which is shown to be compatible with biological and bioinspired electrocatalysts, namely hydrogenases and their mimics in aqueous solution.