Photocatalytic hydrogen peroxide splitting on metal-free powders assisted by phosphoric acid as a stabilizer

Abstract

Hydrogen peroxide (H₂O₂) has received increasing attention as an energy carrier. To achieve a sustainable energy society, photocatalytic H₂O₂ splitting (H₂O₂ (l) → H₂ (g)⁺O₂ (g); ΔG° = + 131 kJ mol⁻¹) is a desirable reaction for on-site H₂ generation. However, this reaction has not been reported because conventional photocatalysis decomposes H₂O₂ by disproportionation (H₂O₂ (l) → H₂O (l) + 1/2O₂ (g); ΔG° = −117 kJ mol⁻¹) and by promoting H₂O₂ reduction instead of H⁺ reduction. Here we report the successful example of H₂O₂ splitting. Visible light irradiation of a graphitic carbon nitride loaded with graphene quantum dots as co-catalysts (GQDs/g-C₃N₄) in a H₂O₂ solution containing phosphoric acid (H₃PO₄) produces H₂. H₃PO₄ associates with H₂O₂ via hydrogen bonding, and this stabilization of H₂O₂ suppresses its reduction, thus promoting H⁺ reduction. The all-organic photosystem with H₃PO₄ as a stabilizer may provide a basis of photocatalytic H₂O₂ splitting.

While H₂ can serve as a renewable fuel, its large scale production, storage, and transport are challenging. Here, authors show H₂O₂ to serve as a potential energy carrier via the photocatalytic production of H₂ from stabilized H₂O₂ solutions and metal-free catalysts.

Introduction

Artificial photosynthesis, which transforms earth-abundant resources into fuels by sunlight, is an urgent and challenging issue for realizing a sustainable energy society^1,2. In the last 50 years, photocatalytic water splitting for hydrogen (H₂) generation under sunlight irradiation (Eq. (1)) has been extensively studied for this purpose^3–5. However, H₂ gas has a low volumetric energy density and needs to be converted into a storable and transportable liquid energy carrier such as an organic hydride⁶ and ammonia⁷. Therefore, identifying a new artificial photosynthesis system that directly generates a liquid energy carrier is a challenge. Recently, hydrogen peroxide (H₂O₂) has received increasing attention as a new liquid energy carrier because it is storable and transportable and generates electricity in a direct peroxide–peroxide fuel cell (DPPFC), although careful handling is necessary owing to its property of being decomposed in the presence of some metal impurities or under heating conditions⁸. The most attractive feature is that H₂O₂ can be generated from earth-abundant water and oxygen (O₂) via photocatalysis^9–11. The photogenerated valence band holes (VB h⁺) oxidize water (O₂ generation) and the conduction band electrons (CB e⁻) reduce O₂ (H₂O₂ generation). These reactions generate H₂O₂ by sunlight irradiation under ambient conditions with a positive Gibbs free energy change (Eq. (2)). Therefore, photocatalytic H₂O₂ generation is a new potential candidate for artificial photosynthesis. Recently, we reported that the resorcinol–formaldehyde resins prepared by the high-temperature hydrothermal synthesis could successfully catalyze the above-mentioned reactions on absorbing visible light up to 620 nm¹². Further, these metal-free resins stably produced H₂O₂ with 0.5% of solar-to-chemical conversion efficiency, which is comparable to the highest efficiency for photocatalytic water splitting (Eq. (1)) by metal-based powder photocatalysts¹³. Therefore, H₂O₂ is a new promising liquid energy carrier candidate.

$${\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)\stackrel{hv}{\to }{\mathrm{H}}_2\left(\mathrm{g}\right)+1/2{\mathrm{O}}_2\left(\mathrm{g}\right)\left(\mathrm{\Delta }{G}^{\circ }=+237\mathrm{kJ}{\mathrm{mol}}^{-1}\right)$$ 1

$${\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)+1/2{\mathrm{O}}_2\left(\mathrm{g}\right)\stackrel{hv}{\to }{\mathrm{H}}_2{\mathrm{O}}_2\left(\mathrm{l}\right)\left(\mathrm{\Delta }{G}^{\circ }=+117\mathrm{kJ}{\mathrm{mol}}^{-1}\right)$$ 2

To realize a sustainable energy society with H₂O₂, on-site generation of H₂ from H₂O₂ solution is necessary also for its use as a hydrogen carrier. Theoretically, semiconductor photocatalysis can promote the overall H₂O₂ splitting. The photogenerated VB h⁺ oxidize H₂O₂ to produce O₂ (Eq. (3)), and the CB e⁻ reduce H⁺ to form H₂ (Eq. (4)). These redox reactions lead to the generation of H₂ from H₂O₂ under sunlight at ambient temperature (Eq. (5)). Owing to the relatively large Gibbs free energy gain (ΔG° = +131 kJ mol⁻¹), H₂O₂ splitting is potentially a new type of artificial photosynthesis reaction. Despite these advantages, photocatalytic H₂O₂ splitting is not reported. This is because H₂O₂ readily decomposes into water and O₂ over conventional H₂ generation photocatalysts even under dark conditions by disproportionation (Eq. (6)) on the surfaces of metal-oxide semiconductors (such as TiO₂) or of metal particle co-catalysts (such as Pt)^14–16. Other reasons are that H₂O₂ is decomposed into hydroxyl radicals on ultraviolet (UV) light absorption at λ < 400 nm (Eq. (7))^17,18, and CB e⁻ reduce H₂O₂ (Eqs. (8) and (9)) more efficiently than H⁺ (Eq. (4)), owing to their low-lying reduction potentials¹⁹. The design of visible-light-driven metal-free photocatalytic systems that can selectively reduce H⁺ while suppressing H₂O₂ reduction is necessary.

$${\mathrm{H}}_2{\mathrm{O}}_2\left(\mathrm{l}\right)+2{\mathrm{h}}^{+}\to {\mathrm{O}}_2\left(\mathrm{g}\right)+2{\mathrm{H}}^{+}\left(\mathrm{aq}\right)\left(+0.68\mathrm{V}\mathrm{vs}\mathrm{NHE}\right)$$ 3

$$2{\mathrm{H}}^{+}\left(\mathrm{aq}\right)+2{\mathrm{e}}^{-}\to {\mathrm{H}}_2\left(\mathrm{g}\right)\left(0\mathrm{V}\mathrm{vs}\mathrm{NHE}\right)$$ 4

$${\mathrm{H}}_2{\mathrm{O}}_2\left(\mathrm{l}\right)\stackrel{hv}{\to }{\mathrm{H}}_2\left(\mathrm{g}\right)+{\mathrm{O}}_2\left(\mathrm{g}\right)\left(\mathrm{\Delta }{G}^{\circ }=+131\mathrm{kJ}{\mathrm{mol}}^{-1}\right)$$ 5

$${\mathrm{H}}_2{\mathrm{O}}_2\left(\mathrm{l}\right)\to {\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)+1/2{\mathrm{O}}_2\left(\mathrm{g}\right)\left(\mathrm{\Delta }{G}^{\circ }=-117\mathrm{kJ}{\mathrm{mol}}^{-1}\right)$$ 6

$${\mathrm{H}}_2{\mathrm{O}}_2\left(\mathrm{l}\right)\stackrel{hv\left(\lambda <400\mathrm{nm}\right)}{\to }\mathrm{2}\cdot \mathrm{OH}\left(\mathrm{aq}\right)$$ 7

$${\mathrm{H}}_2{\mathrm{O}}_2\left(\mathrm{l}\right)+{\mathrm{H}}^{+}\left(\mathrm{aq}\right)+{\mathrm{e}}^{-}\to {\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)+\cdot \mathrm{OH}\left(\mathrm{aq}\right)\left(+1.14\mathrm{V}\mathrm{vs}\mathrm{NHE}\right)$$ 8

$${\mathrm{H}}_2{\mathrm{O}}_2\left(\mathrm{l}\right)+2{\mathrm{H}}^{+}\left(\mathrm{aq}\right)+2{\mathrm{e}}^{-}\to {\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)\left(+1.76\mathrm{V}\mathrm{vs}\mathrm{NHE}\right)$$ 9

Herein, we report the successful example of photocatalytic H₂O₂ splitting. To avoid the undesirable H₂O₂ disproportionation (Eq. (6)), we used a graphitic carbon nitride (g-C₃N₄) organic semiconductor²⁰. The g-C₃N₄ is less active for H₂O₂ disproportionation¹⁴, and its VB and CB levels (+2.00 and −0.63 V (vs NHE), respectively) are sufficient for H₂O₂ oxidation (Eq. (3)) and H⁺ reduction (Eq. (4)). We also used graphene quantum dots (GQDs) as co-catalysts; these are the carbon materials belonging to the graphene family and have attracted increasing attention as non-metal H₂ evolution co-catalysts owing to their high electron conductivity and electron reservation capacity^21–24. Visible-light irradiation of the GQD/g-C₃N₄ catalyst in a H₂O₂ solution containing phosphoric acid (H₃PO₄) at a relatively low temperature (~293 K) successfully produced H₂. The addition of H₃PO₄, which is used as a stabilizer for commercially available H₂O₂ solution^25,26, plays a pivotal role in this reaction. Raman spectroscopy and ab initio calculations revealed that H₃PO₄ associates with H₂O₂ via hydrogen bonding. The stabilization of H₂O₂ by H₃PO₄ inhibited H₂O₂ reduction (Eqs. (8) and (9)) and, hence, promoted H⁺ reduction (Eq. (4)), resulting in successful H₂ generation.

Results

Preparation and characterization of catalyst

g-C₃N₄ powder was prepared by calcination of melamine²⁷. The GQDs solution was produced by the reduction of nitrated pyrene with hydrazine followed by hydrothermal treatment²³. The GQD_x/g-C₃N₄ catalyst was prepared by the hydrothermal treatment of a GQDs solution containing g-C₃N₄, where x denotes the amount of GQDs-loaded relative to that of g-C₃N₄ [x (wt%) = GQDs/g-C₃N₄ × 100]. In the UV–visible absorption spectra of the solution (Supplementary Fig. 1), the absorption band at λ > 500 nm for GQDs²⁸ almost disappears completely after hydrothermal treatment with g-C₃N₄, indicating that almost all of the GQDs in the solutions were successfully loaded on the g-C₃N₄ surface. In Fig. 1a, the diffuse-reflectance (DR) UV–visible spectrum of GQD_x/g-C₃N₄ presents a band at λ < 470 nm, which is assigned to the bandgap transition of g-C₃N₄. Increasing the amount of the GQDs loaded increases the absorbance at λ > 500 nm, indicating that the GQDs are indeed loaded on the g-C₃N₄ surface. The bandgap energies of both g-C₃N₄ and GQD₁/g-C₃N₄ were determined to be ~2.6 eV by the Tauc plot analysis (Supplementary Fig. 2), suggesting that the GQDs loading scarcely affects the band structure of g-C₃N₄. The powder X-ray diffraction (XRD) patterns of g-C₃N₄ and GQD_x/g-C₃N₄ present peaks at 2θ = 27.4° (d = 0.325 nm), which are assigned to the (002) packing of the melem sheet (Supplementary Fig. 3)²⁰, indicating that these catalysts maintain their layered stacking structure even after the GQDs loading. The scanning electron microscopy (SEM) images of both g-C₃N₄ and GQD₁/g-C₃N₄ (Supplementary Fig. 4) indicate an amorphous solid morphology with similar sizes (~30 μm diameter). The transmission electron microscopy (TEM) results of the GQDs solution (Fig. 2, top) present dispersed GQD particles with ~3-nm diameters^23,29,30. In contrast, the GQD₁/g-C₃N₄ catalyst (Fig. 2, bottom) has larger GQD particles with ~10-nm diameters, indicating that the GQD particles are loaded onto the g-C₃N₄ surface via some aggregation during the hydrothermal treatment.

Fig. 1. Properties of GQD_x/g-C₃N₄ photocatalysts.

a DR UV-vis spectra of catalysts. b Amounts of H₂ evolved on the respective catalysts in water containing TEOA as a sacrificial electron donor performed in a closed system (0.1 MPa Ar). Conditions: catalyst (100 mg), a water/TEOA (9/1 v/v) mixture (30 mL), light irradiation (λ > 420 nm, Xe lamp), room temperature, and photoirradiation time (6 h). Error bars represent standard error (s.e.) determined by three independent experiments. c Photocurrent response of the catalysts measured on an FTO glass in 0.1 M Na₂SO₄ solution at a bias of 0.8 V vs Ag/AgCl. d Time-dependent change in the amount of H₂ evolved and the H₂ selectivity during photoirradiation of the entire wavelength light (λ > 300 nm) or λ > 420 nm light in a closed gas circulation system (3 kPa Ar). Conditions: catalyst (200 mg), H₂O₂ (10 mM, 100 mL), H₃PO₄ (1 M), temperature (293 K), and light irradiation (solar simulator with AM 1.5 G filter, 1-sun). Error bars represent standard error (s.e.) determined by three independent experiments.

Fig. 2. Typical TEM images.

a GQD solution. b GQD₁/g-C₃N₄ catalyst. The red arrows are the guide for the eyes to follow the GQD particles.

Effect of GQDs as co-catalysts

The effect of the GQDs on the photocatalytic activity for H₂ generation was studied using triethanolamine (TEOA) as a sacrificial electron donor. A TEOA/water (1/9 v/v) mixture (30 mL) containing the catalyst (100 mg) was photoirradiated for 6 h by a xenon lamp (λ > 420 nm) with magnetic stirring at room temperature under Ar atmosphere in a closed system (0.1 MPa). Figure 1b presents the amount of H₂ evolved over the respective catalysts. The loading of GQDs enhances H₂ evolution, indicating that the GQDs act as co-catalysts for H⁺ reduction. GQD₁/g-C₃N₄ exhibits the highest activity; however, further GQDs loading decrease the activity. This decrease is because a larger amount of GQDs absorb more incident light (Fig. 1a) and suppresses the photoexcitation of g-C₃N₄. Figure 1c shows the photocurrent responses of the catalysts measured on a fluorine tin oxide (FTO) electrode. The photocurrent density of GQD₁/g-C₃N₄ is higher than that of g-C₃N₄, indicating that the CB e⁻ photogenerated on g-C₃N₄ are efficiently transferred to the GQDs and electrode^31,32. The enhanced charge separation of the h⁺ and e⁻ pairs by the GQDs loading, therefore, enhances the activity for H₂ evolution. As shown by the blue bar in Fig. 1b, g-C₃N₄ when loaded with Pt particles (Pt/g-C₃N₄), a typical noble metal co-catalyst for H₂ generation³³, exhibits a much higher activity than GQD₁/g-C₃N₄, where the amount of H₂ formed is 100-fold higher than that formed on GQD₁/g-C₃N₄. The activity of the GQDs as co-catalysts for H₂ generation is lower than that of Pt.

Photocatalytic H₂O₂ splitting

Photocatalytic H₂O₂ splitting was performed in a closed gas circulation system at a reduced pressure³⁴ (3 kPa Ar) under a constant temperature. Visible light (λ > 420 nm) was irradiated by a solar simulator with an AM1.5 G filter (1-sun) to a H₂O₂ solution (1 mmol, 100 mL) containing a catalyst (200 mg) at 293 K. Table 1 lists the amounts of H₂ and O₂ generated and the amount of H₂O₂ consumed during 6 h of photoirradiation. As shown by entry 1, stirring the solution containing GQD₁/g-C₃N₄ in the dark does not produce H₂ or O₂ and does not consume H₂O₂, indicating that the catalyst is indeed inactive for the disproportionation of H₂O₂. As shown by entry 2, photoirradiation of Pt/g-C₃N₄ does not produce H₂ while decomposing almost the entire H₂O₂ and producing a very large amount of O₂, suggesting that Pt significantly promote H₂O₂ disproportionation (Eq. (6))^14,15. As shown by entry 3, photoirradiation of GQD₁/g-C₃N₄ in pure water without H₂O₂ does not produce H₂ or O₂ because g-C₃N₄ is less active for water oxidation by the VB h⁺ due to its relatively negative VB level³⁵. As exhibited by entry 4, photoirradiation of GQD₁/g-C₃N₄ in a H₂O₂ solution also does not produce H₂ while promoting O₂ production (17 μmol) and H₂O₂ consumption (30 μmol), suggesting that the CB e⁻ do not reduce H⁺ (Eq. (4)) although the VB h⁺ oxidize H₂O₂ (Eq. (3)). This is because the CB e⁻ reduce H₂O₂ (Eqs. (8) and (9)) more efficiently than H⁺ reduction (Eq. (4)), owing to their low-lying reduction potentials¹⁹. The H₂O₂ reduction by CB e⁻ must therefore be suppressed to promote H⁺ reduction.

Table 1.

Results of photocatalytic H₂O₂ splitting.

Entry	Temperature (K)	Catalyst	Light	Additive	H2 (μmol)	O2 (μmol)	Consumed H2O2 (μmol)
1	293	GQD1/g-C3N4			N.D.	N.D.	<0.1
2	293	Pt/g-C3N4	>420 nm		N.D.	125	>999a
3b	293	GQD1/g-C3N4	>420 nm		N.D.	N.D.
4	293	GQD1/g-C3N4	>420 nm		N.D.	17.3	30
5	293	GQD1/g-C3N4	>420 nm	H3PO4 (0.5 M)c	0.4	16.9	28
6	293	GQD1/g-C3N4	>420 nm	H3PO4 (1 M)d	1.2	7.7	20
7	293	GQD1/g-C3N4	>420 nm	H3PO4 (2 M)e	0.3	4.2	13
8	293	GQD1/g-C3N4	>420 nm	H2SO4 (0.1 M)d	N.D.	15.6	25
9	293	Pt/g-C3N4	>420 nm	H3PO4 (1 M)	0.2	251	>999a
10	293	GQD1/g-C3N4	>420 nm	NaH2PO4 (1 M)f	0.4	49.6	78
11	303	GQD1/g-C3N4	>420 nm	H3PO4 (1 M)	0.5	9.4	20
12	313	GQD1/g-C3N4	>420 nm	H3PO4 (1 M)	N.D.	10.8	21
13	323	GQD1/g-C3N4	>420 nm	H3PO4 (1 M)	N.D.	8.8	25
14	293	GQD1/g-C3N4	>300 nm	H3PO4 (1 M)	0.5	22.1	127

Reactions were performed in a closed gas circulation system (3 kPa Ar). Conditions are: catalyst (200 mg), light source (solar simulator, 1-sun), H₂O₂ solution (1 mmol (10 mM), 100 mL), and photoirradiation time (6 h). All of the data are the mean values determined by the multiple experiments (at least three times) and contain ±21% deviations.

^aMost of H₂O₂ decomposes while setting up the reactor owing to rapid disproportionation on the Pt particles.

^bPure water without H₂O₂ was used for the photoreaction.

^cThe pH of the solutions was ~1.2.

^dThe pH of the solution was ~1.0.

^eThe pH of the solution was ~0.6.

^fThe pH of the solution was adjusted to ~1.0 by the addition of H₂SO₄.

Effect of H₃PO₄

It is well known that H₃PO₄ is generally added to commercially available H₂O₂ solutions as a stabilizer to inhibit the decomposition of H₂O₂^36,37. Therefore, we studied the effect of H₃PO₄ on the photocatalytic H₂O₂ splitting. As exhibited by entries 5–7, visible-light irradiation of GQD₁/g-C₃N₄ in a H₂O₂ solution containing H₃PO₄ produces H₂. This is the first successful example of photocatalytic H₂O₂ splitting. The photocatalytic activity depends on the amount of H₃PO₄ added. The addition of 1 M H₃PO₄ (entry 6) produces the largest amount of H₂ (1.2 μmol), but further addition (entry 7) decreases the activity. Several types of reagents such as H₃PO₃³⁷_, uric acid³⁸, Na₂CO₃³⁹, KHCO₃⁴⁰, barbituric acid⁴¹, hippuric acid⁴², urea, and acetanilide⁴³ have also been reported to serve as stabilizers for H₂O₂. However, as shown in Supplementary Table 1, the photoirradiation of GQD₁/g-C₃N₄ in a H₂O₂ solution containing each of the respective reagents barely produces H₂, indicating that H₃PO₄ is specifically effective for photocatalytic H₂O₂ splitting. The pH of 1 M H₃PO₄ solution is ~1.0. As presented by entry 8, the photoreaction, when performed in a H₂O₂ solution with H₂SO₄ (0.1 M, pH ~1), does not produce H₂. This indicates that a low pH is not a critical factor for enhanced H₂ generation with H₃PO₄. As exhibited by entry 7, addition of more H₃PO₄ (2 M) decreases the amount of H₂ generated. In this case, the amounts of O₂ generated and H₂O₂ consumed also decrease because the H⁺ concentration increased by a larger amount of H₃PO₄ (pH ~0.7) suppresses the H₂O₂ oxidation by the VB h⁺ (Eq. (3)). These data imply that H₃PO₄ may interact with H₂O₂ and inhibit its reduction (Eqs. (8) and (9)), resulting in the promotion of H⁺ reduction (Eq. (4)). As shown by entry 9, the Pt/g-C₃N₄ catalyst, when photoirradiated with 1 M H₃PO₄, produces H₂; however, in the absence of H₃PO₄, H₂ is not produced (entry 2). This confirms that H₃PO₄ interacts with H₂O₂ and inhibits the H₂O₂ reduction by the CB e⁻; however, the amount of H₂ formed (0.2 μmol) is much smaller than that formed over GQD₁/g-C₃N₄ (1.2 μmol, entry 6). In this case, almost the entire H₂O₂ (1 mmol) is decomposed, similar to the case without H₃PO₄ (entry 2), indicating that Pt particles inevitably promote H₂O₂ disproportionation even in the presence of H₃PO₄ (Eq. (6)). Therefore, the use of the all-organic GQD/g-C₃N₄ catalyst, which is inactive for H₂O₂ disproportionation, together with H₃PO₄, which suppresses H₂O₂ reduction by the CB e⁻, is effective for photocatalytic H₂O₂ splitting.

Stabilization of H₂O₂ by H₃PO₄

H₃PO₄ is used as a stabilizer to suppress the disproportionation of H₂O₂ caused by metal cation impurities^44,45, although the mechanism for their association has not been clarified yet^46–48. The interaction between H₂O₂ and H₃PO₄ was confirmed by cyclic voltammetry (CV) measurements. Figure 3 presents the CV curves obtained with a GQDs-loaded glassy carbon electrode in a solution containing 0.1 M Na₂SO₄ as an electrolyte. In 0.1 M H₂SO₄ (pH ~1.0), as denoted by a green line, the CV curve presents a weak cathodic current at ~ –1.5 V (vs RHE), which is assigned to H⁺ reduction, where no anodic current is observed. The H₃PO₄ solution (1 M, pH ~1.0) presents a similar CV curve (yellow line). Addition of H₂O₂ (0.1 M) to the H₂SO₄ solution (red line) exhibits a strong cathodic current at <–0.8 V, which is assigned to H₂O₂ reduction (Eqs. (8) and (9))⁴⁹, whereas a strong anodic current is observed at >1.2 V, which is assigned to H₂O₂ oxidation (Eq. (3)). In contrast, as shown by the blue and purple lines, addition of H₂O₂ to the H₃PO₄ solution presents cathodic current, which is weaker than that obtained in the H₂O₂/H₂SO₄ mixture (red line). These data indicate that H₂O₂ interacts with H₃PO₄ and suppresses its reduction. As reported⁵⁰, at this pH (~1), phosphoric acid exists as a non-dissociated neutral H₃PO₄ form, where its mole ratio is ~94% and a mole ratio of mono-deprotonated H₂PO₄⁻ form is ~6%, suggesting that H₂O₂ interacts with the H₃PO₄ form.

Fig. 3. CV charts measured on a GQDs-loaded grassy carbon electrode in the different solutions.

The pH of the solutions is ~1, and all the measurements were carried out under N₂ atmosphere with 0.1 M Na₂SO₄ and at a scan rate of 0.1 V s⁻¹.

Raman spectroscopy

The interaction between H₂O₂ and H₃PO₄ was further studied via Raman spectroscopy at 293 K (Fig. 4a). Pure water (purple) exhibits a strong band at 2800–3800 cm⁻¹ for the O–H stretching vibration of water, indicative of strong hydrogen (H-) bonding between the water molecules, as is schematically shown in Fig. 5a (left). The H₂O₂ solution (black) exhibits a weaker O–H band because the interaction between H₂O₂ and water weakens the H-bonding between the water molecules. The H₃PO₄ solution (orange) exhibits a much weaker O–H band because, as shown in Fig. 5a (middle), H₃PO₄ strongly interacts with water⁵⁰ and significantly weakens the H-bonding between the water molecules. The addition of H₂O₂ to the H₃PO₄ solution (blue), however, increases the O–H band. This indicates that, as shown in Fig. 5a (right), H₃PO₄ interacts with H₂O₂ more strongly than with water owing to the H-bonding interaction to form the H₂O₂–H₂PO₄⁻ bidentate complex with a structure similar to the urea–H₂PO₄⁻ complex⁵¹. This may suppress the water–H₃PO₄ interaction and regenerate strong H-bonding interaction between the water molecules. These data suggest that H₃PO₄ associates with H₂O₂ via H-bonding to form a stabilized complex (Fig. 5a (right)), which may inhibit the H₂O₂ reduction by the CB e⁻.

Fig. 4. Raman spectra (600–4000 cm⁻¹ region) of the respective solutions.

a 293 K. b 323 K. The H₂O₂ and H₃PO₄ concentrations are 2 M and 4 M, respectively.

Fig. 5. Schematic representation of H-bonding interaction and optimized structures of compounds.

a Sequential change in H-bonding interaction between water, H₂PO₄⁻, and H₂O₂. b optimized structures of H₂O₂, H₂PO₄⁻, and H₂O₂–H₂PO₄⁻ complex (DFT/B3LYP/6–31 G*(d), PCM: water).

Figure 6a presents the Raman spectroscopy results of the respective solutions in the 850–950 cm⁻¹ region measured at 293 K. The H₂O₂ solution (top) exhibits a band at 875.6 cm⁻¹, which is assigned to the O–O stretching vibration of H₂O₂⁵². The H₃PO₄ solution (middle) exhibits a band at 899.0 cm⁻¹ assigned to the P–O symmetric stretching vibration of H₃PO₄⁵³. The spectrum of the H₂O₂/H₃PO₄ mixture (bottom) can be deconvoluted into O–O and P–O stretching components. The O–O band of the mixture appears at 874.6 cm⁻¹, which lies at a lower wavenumber (Δ$\tilde{v}$ = −1.0 cm⁻¹) than that of the band of pure H₂O₂ (875.6 cm⁻¹). In addition, the P–O band of the mixture appears at 899.0 cm⁻¹, which also lies at a lower wavenumber (Δ$\tilde{v}$ = −0.7 cm⁻¹) than that of pure H₃PO₄ (899.7 cm⁻¹). These data suggest that the lengths of the O–O and P–O bonds are extended by the H-bonding between H₂O₂ and H₃PO₄ (Fig. 5a (right)).

Fig. 6. Raman spectra (850–950 cm⁻¹) of (top) H₂O₂, (middle) H₃PO₄, and (bottom) a mixture of H₂O₂ and H₃PO₄ measured at different temperatures.

a 293 K. b 323 K. The H₂O₂ and H₃PO₄ concentrations are 2 M and 4 M, respectively. The dots are the experimental data, and the solid lines are the deconvoluted components, where the green lines are the sum of the components.

Ab initio calculations

To obtain further information on the interaction between H₂O₂ and H₃PO₄, the structure of the H₂O₂–H₂PO₄⁻ complex (Fig. 5a (right)) was calculated by density functional theory (DFT) using the polarizable continuum model (PCM) with water as the solvent (Supplementary Tables 3–6)⁵⁴. Figure 5b presents the optimized structures of H₂O₂, H₃PO₄, and the H₂O₂–H₂PO₄⁻ complex, and the lengths of the respective bonds. The O–O bond length in the complex (1.462 Å) is longer than that in H₂O₂ (1.455 Å). In addition, the lengths of the two P–O bonds in the complex (1.525 Å (P–O3) and 1.519 Å (P–O4)) are longer than those in H₃PO₄ (1.480 Å (P–O3)), where the average P–O lengths of the complex (1.580 Å) is also longer than those of H₃PO₄ (1.578 Å). The shortened lengths of the O–O bond (Δ = 0.007 Å) and P–O bonds (Δ = 0.002 Å) by the H-bonding interaction agree reasonably with the lower wavenumber shift of the O–O and P–O stretching vibrations, as confirmed via Raman analysis (Fig. 6a). The DFT frequency calculations (Supplementary Table 2) reveal that the wavenumbers of the O–O stretching in H₂O₂ and the complex are 947.7 cm⁻¹ and 942.8 cm⁻¹⁵⁵, respectively (Δ$\tilde{v}$ = −4.9 cm⁻¹), and the wavenumbers of the P–O symmetric stretching in H₃PO₄ and the complex are 1075 cm⁻¹ and 1039 cm⁻¹⁵⁶, respectively (Δ$\tilde{v}$ = −36 cm⁻¹). Although the frequencies calculated by DFT are overestimated owing to the large anharmonicity of the high-frequency modes^57,58, the lower wavenumber shifts of the O–O and P–O bonds caused by the H-bonding interaction between H₂O₂ and H₃PO₄ are consistent with the Raman data (Fig. 6a). In addition, the two O–H‧‧‧O distances of the complex are 2.749 Å (O1–H1‧‧‧O3) and 2.733 Å (O2–H2‧‧‧O4), indicative of strong and mostly electrostatic H-bonding interaction between H₂O₂ and H₃PO₄⁵⁹. Supplementary Fig. 5 summarizes the energy diagram for the frontier molecular orbitals⁶⁰ of the calculated models. The interaction of the highest-occupied molecular orbital (HOMO) of H₂PO₄⁻ with the lowest-unoccupied molecular orbital (LUMO) of H₂O₂ creates a frontier orbital of the complex (LUMO + 1) at an energy level higher than that of free H₂O₂, meaning that the H₂O₂ moiety of the complex is stabilized against its reduction upon external electron injection⁶¹. This is because the electron donation from H₂PO₄⁻ to H₂O₂ leads to a decrease in the electron affinity of the H₂O₂ moiety and stabilizes its orbital. The result is consistent with the suppressed H₂O₂ reduction (Fig. 3) and the enhanced H₂ evolution (Table 1, entry 6) by the addition of H₃PO₄. Therefore, H₃PO₄ associates with H₂O₂ via H-bonding and produce a stabilized complex (Fig. 5a (right) and Fig. 5b (right)). This interaction decreases the electron affinity of H₂O₂ and suppresses the H₂O₂ reduction by the CB e⁻ (Eqs. (8) and (9)), thus resulting in the promotion of H⁺ reduction (Eq. (4)). It must be noted that addition of H₃PO₄ is necessary for efficient stabilization of H₂O₂; addition of phosphate salts is ineffective. As shown by entry 10 (Table 1), a NaH₂PO₄ (1 M) solution (pH ~1), when used for photoreaction, produces a lower amount of H₂ than a H₃PO₄ (1 M) solution (entry 6) although fully protonated H₃PO₄ species exist in both solutions. Alkaline metal cations interact with H₂O₂ in solution⁶². The H₂O₂–Na⁺ interaction may weaken the H₂O₂–H₂PO₄⁻ complex, probably resulting in the decreased H₂ evolution activity.

Temperature effect

The stability of the H₂O₂–H₂PO₄⁻ complex depends on the temperature of the solution. As shown by entry 6 (Table 1), the photoirradiation of GQD₁/g-C₃N₄ at 293 K in a H₂O₂ solution containing 1 M H₃PO₄ produces 1.2 μmol of H₂. However, when the reaction is performed at 303 K (entry 11), the amount of H₂ produced is decreased to 0.5 μmol, and further increase in the temperature (313 K and 323 K) does not produce H₂ (entries 12 and 13), indicating that higher temperature is ineffective. To clarify the temperature effect on the H₂O₂–H₂PO₄⁻ interaction, Raman spectroscopy was measured at a high temperature (323 K). As displayed in Fig. 4b (orange), the O–H band of water decreases by the addition of H₃PO₄; this is owing to the water–H₃PO₄ H-bonding interaction (Fig. 5a (middle)), similar to the case at 293 K (Fig. 4a). At 293 K, the decreased O–H band increases on the addition of H₂O₂; this is because the formation of the H₂O₂–H₂PO₄⁻ complex regenerates the H-bonding between the water molecules. However, at 323 K, the addition of H₂O₂ to the H₃PO₄ solution barely increases the O–H band, indicating that there is no H-bonding interaction between H₂O₂ and H₂PO₄⁻. As displayed in Fig. 6a, the Raman spectrum of the H₂O₂/H₃PO₄ mixture measured at 293 K presents O–O (874.6 cm⁻¹) and P–O (899.0 cm⁻¹) bands, both of which lie at lower wavenumbers than those of only H₂O₂ (875.6 cm⁻¹) and H₃PO₄ (899.7 cm⁻¹), owing to the H₂O₂–H₂PO₄⁻ complex formation. However, at 323 K (Fig. 6b), the H₂O₂/H₃PO₄ mixture shows O–O (875.6 cm⁻¹) and P–O (899.7 cm⁻¹) bands at identical positions to those in H₂O₂ and H₃PO₄. These data suggest that the H₂O₂–H₂PO₄⁻ complex is destabilized at higher temperatures because a rise in temperature increases the kinetic energy of the molecules and weakens their H-bonding interaction⁶³, as observed in several H-bonding systems such as dimethyl sulfoxide–water⁶⁴ and urea–water⁶⁵ at the similar temperature range (283–333 K). The weakened interaction at higher temperatures inevitably promotes H₂O₂ reduction (Eqs. (8) and (9)) and, hence, decrease H₂ generation (Eq. (4)). The photoreaction at a relatively lower temperature is therefore necessary for efficient H₂ generation.

Effect of light wavelength

Visible-light irradiation is also necessary for this system. Figure 1d shows the time course for the amount of H₂ evolved during the photocatalytic H₂O₂ splitting over the GQD₁/g-C₃N₄ catalyst with 1 M H₃PO₄ at 293 K. Further, the H₂ selectivity is defined as the ratio of the amount of H₂ evolved to that of the consumed H₂O₂ (Eq. (10)). Under visible light (λ > 420 nm) exposure, as depicted by circle keys, the amount of H₂ evolved increases almost linearly with time, and the H₂ selectivity is almost constant even after prolonged photoirradiation. This indicates that the system consisting of the all-organic GQD₁/g-C₃N₄ catalyst and H₃PO₄ as a stabilizer at a low temperature stably promotes H₂O₂ splitting. In addition, as presented in Supplementary Fig. 6, the recovered GQD₁/g-C₃N₄ catalyst, when reused for further photoreactions, maintains the activity and the H₂ selectivity, indicating that the catalyst is reusable without the loss of activity. However, the H₂ selectivity is only ~6 %, indicating that the H₂O₂ reduction (Eqs. (8) and (9)) still occurs more efficiently than the H⁺ reduction (Eq. (4)) even in the presence of H₃PO₄.

$${\mathrm{H}}_2\mathrm{selectivity}\left(\mathrm{\%}\right)=\frac{\left[{\mathrm{H}}_2\mathrm{evolved}\left(\mathrm{\mu }\mathrm{mol}\right)\right]}{\left[{\mathrm{H}}_2{\mathrm{O}}_2\mathrm{consumed}\left(\mathrm{\mu }\mathrm{mol}\right)\right]}\times 100$$ 10

The square keys in Fig. 1d denote the results obtained under irradiation of the entire wavelength light (λ > 300 nm) from the solar simulator (Supplementary Fig. 7). The amount of H₂ evolved is much smaller than that under λ > 420 nm irradiation. As summarized in Table 1 (entry 14), λ > 300 nm irradiation for 6 h produces a larger amount of O₂ (22 μmol) than the λ > 420 nm irradiation (7.7 μmol, entry 5). This indicates that UV irradiation enhances the catalyst photoexcitation and efficiently promotes H₂O₂ oxidation by the VB h⁺ (Eq. (3)), clearly suggesting that UV irradiation suppresses H⁺ reduction (Eq. (4)). The λ > 300 nm irradiation decomposes a larger amount of H₂O₂ (127 μmol) than the λ > 420 nm irradiation (20 μmol, entry 6); this is owing to the photodecomposition of H₂O₂ by the absorbed UV light (Eq. (7)), as confirmed by the absorption spectra of the H₂O₂ solution in the UV region (Supplementary Fig. 8). This suggests that the H₂O₂–H₂PO₄⁻ complex absorbs UV light and undergoes destabilization. The weakened H-bonding of the complex may therefore inevitably promote H₂O₂ reduction (Eqs. (8) and (9)), thus inhibiting the H⁺ reduction (Eq. (4)). As displayed in Fig. 1d (square), the H₂ selectivity with λ > 300 nm irradiation is only ~0.5% owing to the accelerated H₂O₂ decomposition, indicating that visible-light irradiation is necessary for efficient photocatalytic H₂O₂ splitting. Even under visible-light irradiation, the H₂ selectivity is ~6%; therefore, further selectivity enhancement is necessary for practical applications. Nevertheless, the successful example presented here based on the combination of an all-organic photocatalyst and H₃PO₄ as a stabilizer under sunlight illumination presents significant potential for photocatalytic H₂O₂ splitting.

Discussion

Visible-light irradiation of the all-organic GQD/g-C₃N₄ catalyst in a H₂O₂ solution containing H₃PO₄ at a relatively low temperature (~293 K) could successfully promote H₂ generation by photocatalytic H₂O₂ splitting. The Raman spectroscopy and DFT calculations revealed that H₃PO₄ associates with H₂O₂ via H-bonding interaction and forms a H₂O₂–H₂PO₄⁻ complex at a low temperature. This decreases the electron affinity of H₂O₂ and suppresses its reduction, thus promoting H⁺ reduction (H₂ generation). The present system stably produces H₂ even after prolonged photoirradiation, and the catalyst is reusable without the loss of activity and H₂ selectivity. This photocatalytic system offers several advantages: a metal-free photocatalyst and inexpensive acid (H₃PO₄) can be used for the reaction; visible light, a main constituent of sunlight irradiation, can be used as the light source; and on-site H₂ generation from a transportable and storable energy carrier (H₂O₂) can be achieved. At present, H₂ selectivity is only ~6% relative to the amount of H₂O₂ consumed, and further improvement of its selectivity is necessary for practical applications. Nevertheless, the successful example presented here based on an all-organic catalyst with H₃PO₄ as a stabilizer may contribute to the design of a highly efficient system for the on-site photocatalytic generation of H₂ from a H₂O₂ solution.

Methods

Catalyst preparation

The g-C₃N₄ powder was prepared by calcination of melamine at 823 K (heating rate: 2.3 K min⁻¹, holding time: 4 h)²⁷. The GQDs solution was synthesized as follows:³⁰ pyrene (2 g) was stirred in concentrated HNO₃ (160 mL) at 353 K for 12 h. Water (2 L) was added to the resultant, and the solids formed were recovered by filtration. The solids were dispersed in hydrazine hydrate solution (1.0 M, 40 mL), ultrasonicated for 2 h, and left in an autoclave at 423 K for 10 h. Filtration of the resultant by a microporous membrane afforded a 0.5-g L⁻¹ GQDs solution. The GQD_x/g-C₃N₄ catalysts [x (wt%) = 0.5, 1.0, 3.0, and 5.0] were prepared as follows: g-C₃N₄ (0.4 g) was added to 40 mL of the GQDs solution (0.05, 0.1, 0.3, and 0.5 g L⁻¹) and stirred for 1 h at 298 K. The mixture was left in an autoclave at 423 K for 4 h. The solids were recovered by centrifugation, washed with water, and dried at 353 K, affording GQD_x/g-C₃N₄ as brown powders. A Pt/g-C₃N₄ catalyst (Pt loading: 0.3 wt%) was prepared as follows:³³ g-C₃N₄ (1.0 g) was added to water (30 mL) containing H₂PtCl₆‧6H₂O (8.0 mg) and evaporated with vigorous stirring at 373 K. The obtained powders were reduced under H₂ flow (0.1 mL min⁻¹) at 673 K (heating rate: 10 K min⁻¹, holding time: 2 h).

Photoreaction

H₂ generation from a TEOA solution was performed in a closed system (0.1 MPa). The catalyst (100 mg) was added to a 10 vol% TEOA solution (30 mL) within a glass tube (φ 35 mm; capacity, 50 mL). The tube was sealed with a rubber cap and ultrasonicated for 5 min, and Ar gas was bubbled through the solution for 10 min. The tube was photoirradiated (λ > 420 nm) using a 2-kW Xe lamp (USHIO Inc.) with magnetic stirring at room temperature. After the reaction, the amount of H₂ formed was determined by gas chromatography-thermal conductivity detector (GC-TCD) (Shimadzu; GC-8A). Photocatalytic H₂O₂ splitting was performed in a closed gas circulation system connected to GC-TCD⁶⁶. The catalyst (200 mg) was added to a H₂O₂ solution (1 mmol, 100 mL) within a quartz cell. The cell was connected to a gas circulation system, and the system was depressurized to 3 kPa after repeated Ar purging. The cell was immersed in a temperature-controlled water bath (±0.1 K) and photoirradiated at λ > 420 nm using a solar simulator equipped with AM1.5 G filter (SAN-EI Electric Inc.; XES-40S3, 1-sun) with magnetic stirring¹². The amounts of H₂O₂ in the solutions were determined by redox titration with KMnO₄.

Analysis

The electrochemical measurements were performed in a conventional three-electrode cell using an electrochemical system (ALS700E, BAS Inc.), where Ag/AgCl and Pt wire electrodes were used as the reference and counter electrodes, respectively⁶⁶. For the photocurrent measurements, a catalyst-loaded FTO glass was used as the working electrode. Catalyst (20 mg) was added to a mixture of water (4 mL), 2-propanol (1.5 mL), and 5 wt% Nafion solution (100 μL)⁶⁷, and ultrasonicated for 10 min. The slurry was placed onto an FTO glass and dried at room temperature. The film was formed to 0.5 cm × 0.5 cm square, where the undeposited parts were coated with an epoxy resin. For the CV measurements, a glassy carbon electrode loaded with GQD was used as the working electrode, where the GQD solution was placed on the electrode and dried at room temperature. All the measurements were performed after N₂ bubbling through the solution for 1 h. Raman spectra were measured on a confocal Raman microscope (LabRAM HR-800, HORIBA) with a YAG laser (532 nm line) as the excitation source, where the laser power was 100 mW and the total data accumulation time was 100 s, respectively. DR UV-vis, XRD, SEM, and TEM observations were performed according to the procedures described in the literature¹².

Computational details

The molecular structures were optimized by the B3LYP functional (B3LYP/6-31G*(d)) using the polarizable continuum model (PCM) with water as the solvent⁵⁴ in the Gaussian 03 program. Cartesian coordinates for the models used are summarized at the end of the Supplementary Material.

Author contributions

Y.S. directed this project and conceived the experiment. Y.U., A.S., and T.H. conducted the experimental work and analyzed the data. S.H. performed the TEM observations. The manuscript was written by Y.S. and Y.U. with contributions from the other coauthors. All authors have given approval to the final version of the manuscript.

Data availability

All experimental data within the article and its Supplementary Information are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

Supplementary information is available for this paper at 10.1038/s41467-020-17216-2.