Catalytic ammonia formation from dinitrogen, water, and visible light energy

Abstract

The development of the production method for green ammonia, which is produced only from ubiquitous and clean small molecules (i.e., dinitrogen and water) using renewable energy, has been desired for a next-generation carbon-free energy carrier to build a carbon-neutral society and solve global warming. We have herein achieved visible-light-driven catalytic ammonia formation from dinitrogen and water under ambient conditions using tertiary phosphines, which are widely-used organic compounds, as an electron donor in the presence of molybdenum complexes as molecular catalysts for ammonia formation from dinitrogen and iridium complexes as photosensitizers. In this reaction system, visible light energy enables iridium photosensitizers to trigger electron relay from tertiary phosphines (R₃P) as weak reductants to molybdenum catalysts, and the produced radical cation (R₃P^•+) activates water molecules to donate protons for ammonia formation to molybdenum catalysts via the production of a phosphine-water adducted radical cation (R₃P^•+-OH₂).

The production of green ammonia is important for the development of next-generation carbon-free energy carriers. Here, the authors demonstrate visible-light-driven catalytic ammonia formation from dinitrogen and water under ambient conditions using tertiary phosphines in the presence of molybdenum and iridium complexes.

Introduction

The development of new scientific techniques for producing energy carriers from ubiquitous small molecules utilizing renewable energy (especially solar energy) is a pressing challenge for the international community to build a sustainable and carbon-neutral society. Green ammonia produced from atmospheric dinitrogen using solar energy is one of the most desirable compounds as an energy carrier because ammonia is easily liquified to facilitate storage and transportation with high energy density^1–4. Ammonia is also essential for industry as raw materials for various chemicals such as fertilizers and as fuel without emitting carbon dioxide during combustion^5,6. Currently, ammonia is industrially produced from dinitrogen and dihydrogen in the Haber–Bosch process (Fig. 1a). Although the Haber–Bosch process is the most atom-economical process, it requires high pressure and temperature during the reaction. To overcome this issue, the requirement for new production methods for green ammonia would be as follows: (1) ammonia production should proceed under mild reaction conditions to decrease the energy consumption, (2) the activation energy for the reaction should be supplied from renewable energy, typically solar energy, (3) the electrons and protons for ammonia production should be supplied from ubiquitous and clean electron and proton sources, desirably water.

Fig. 1. Nitrogen fixation and water activation using chemical and visible light energy.

a The industrial ammonia production from dinitrogen and dihydrogen by the Haber–Bosch process. b Catalytic ammonia formation from dinitrogen and water in the presence of the molybdenum complex using SmI₂ as a reductant. c Photocatalytic ammonia formation in the presence of the molybdenum complex and a photosensitizer using 9,10-dihydroacridine as a reductant. d Photochemical activation of water to use it as a proton donor using tertiary phosphines as a reductant. e Photochemical hydrogenation of olefins using water as a proton source and tertiary phosphines as a reductant. f The current reaction system for catalytic ammonia formation from dinitrogen and water using tertiary phosphines as a reductant and visible light energy.

For the last 20 years, transition metal complexes inspired by nitrogenase have been intensively studied as one of the effective catalysts for catalytic ammonia formation under mild reaction conditions since the report by Schrock et al.^7–14. In 2019, we developed catalytic ammonia production from dinitrogen and water using samarium diiodide as an electron source under ambient pressure and temperature in the presence of molybdenum trihalide complex bearing PCP-type pincer ligands (Fig. 1b)¹⁵. In this reaction, the dinitrogen-bridged dimolybdenum complexes, which are produced by two-electron reduction of the trihalide complexes and subsequent bridging coordination of dinitrogen, are smoothly converted to the nitride complexes as key intermediates via the efficient N≡N bond cleavage^16–18. The samarium aqua complexes, which are produced in situ to activate water molecules, can donate an electron and a proton concertedly (i.e., proton-coupled electron transfer (PCET)) to the nitride complexes to produce ammonia^19–22.

By combining the aforementioned catalytic ammonia production by the molybdenum complexes and the redox photosensitizing reaction by iridium complexes, visible-light-driven catalytic ammonia formation has been developed. In 2022, we reported photocatalytic ammonia formation using 9,10-dihydroacridine as an electron and a proton source in the presence of a molybdenum triiodide complex bearing a PCP-type pincer ligand and an iridium photosensitizer (Fig. 1c)²³. In the same year, Peters and coworkers also reported photocatalytic ammonia formation using a reduced Hantzsch ester as an electron and a proton source in the presence of a molybdenum tribromide complex bearing a PNP-type pincer ligand, an iridium photosensitizer, and a mixture of collidine and its conjugate acid as an effective additive²⁴. These reaction systems were driven by visible light, i.e., the main component of sunlight, under ambient reaction conditions, even though some organic compounds were required as an electron and a proton source for the reaction.

Photochemical activation of water to use it as a proton donor has been successfully achieved using tertiary phosphines (R₃P) as electron donors²⁵. In 2018, we reported stoichiometric protonation of 2,6-lutidine (Lut) using protons from water triggered by photo-induced electron transfer from R₃P to the ferrocenium cation in the excited state (Fc^+*), resulting in quantitative production of 2,6-lutidinium cation ([LutH]⁺), phosphine oxide (R₃P=O), and ferrocene (Fc) (Fig. 1d)²⁶. In this system, radical cations of phosphines (R₃P^•+), which were produced by the electron donation to Fc^+*, reacted with water to produce R₃P^•+-OH₂ radical cation intermediate, where O-H bonds were activated enough to donate a proton to Lut. The R₃P^•-OH radical species produced by the protonation of Lut donated another electron and proton to Fc⁺ and Lut, respectively, in the process of conversion to R₃P=O. In 2023, Studer and coworkers reported photocatalytic hydrogenation of olefins using similar photochemical activation of water molecules triggered by photo-induced electron transfer from tertiary phosphines using an iridium photosensitizer (Fig. 1e)²⁷. In this system, hydrogen atom transfer (HAT) from R₃P^•-OH is suggested to be the main pathway for the conversion to R₃P=O, and HAT catalyst (i.e., 2,4,6-triisopropylbenzenethiol) significantly promoted the hydrogenation of olefins.

Based on these research backdrops, we have envisaged photocatalytic ammonia formation from dinitrogen and water using tertiary phosphines as an electron source in the presence of a molybdenum complex as an ammonia production catalyst and an iridium complex as a photosensitizer (Fig. 1f). By the visible-light-induced electron transfer from R₃P to the iridium photosensitizer and the activation of water molecules by the formation of R₃P^•+-OH₂, the resulting one-electron-reduced species (OERS) of the iridium photosensitizer and R₃P^•+-OH₂ respectively functioned as an electron and a proton source for the PCET process to the molybdenum nitride complex as a key intermediate in the catalytic cycle. As opposed to the above-mentioned photochemical systems, a proton mediator was essential for the catalytic reaction to realize the proton transfer from R₃P^•+-OH₂ to the molybdenum catalysts without the competing charge recombination between OERS of the photosensitizer and R₃P^•+-OH₂. We optimized the reaction conditions and revealed that the introduction of electron-withdrawing groups on the ligands in both the molybdenum catalysts and the iridium photosensitizers was effective in achieving the electron relay from the tertiary phosphines as weak reductants to the molybdenum catalysts. Our mechanistic studies clarified that the nucleophilic attack by water to produce R₃P^•+-OH₂ is the most plausible rate-determining step in the presence of the proton mediator; thus, the tertiary phosphines with lower electron density on the phosphorous atoms showed high efficiency in the production of R₃P^•+-OH₂, even though the preceding reductive quenching process proceeds more efficiently with increasing the electron density on the phosphorous atom. From these conflicting requirements, the highest catalytic activities were observed when using Ph₃P or tri(p-tolyl)phosphines, where the phosphorus atoms have moderate electron density. To the best of our knowledge, this is a successful example of a visible-light-driven catalytic ammonia formation from dinitrogen and water using molecular catalysts under ambient reaction conditions.

Results

Photocatalytic ammonia formation from dinitrogen and water

As a typical run, we performed the photocatalytic reduction of dinitrogen at atmospheric pressure with Ph₃P (90 equiv. based on the Mo atom of the catalyst), H₂O (540 equiv. based on the Mo atom of the catalyst) and 2,4,6-Collidine (Col, 180 equiv. based on the Mo atom of the catalyst) in the presence of catalytic amounts of a molybdenum triiodide complex bearing a PCP-type pincer ligand with a trifluoromethyl group [MoI₃(PCP-CF₃)] (1a; 2 μmol; PCP-CF₃ = 1,3-bis((di-tert-butylphosphino)methyl)−5-trifluoromethyl-benzimidazol-2-ylidene) and of an iridium photosensitizer [Ir(dFCF₃ppy)₂(dtbbpy)]ONf ([2b]ONf; 4 μmol; dFCF₃ppy = 2-(2,4-difluorophenyl)−5-(trifluoromethyl)pyridine; dtbbpy = 4,4’-di-tert-butyl-2,2’-bipyridine; ONf = nonafluorobutansulfonate) in benzene (5 mL) at room temperature for 20 h under visible light irradiation (Fig. 2a, Table 1 entry 1). The amount of produced ammonia reached 31 ± 0.4 equiv. based on the Mo atom of the catalyst in 51% yield based on Ph₃P, alongside 28 ± 0.7 equiv. of dihydrogen based on the Mo atom of the catalyst in 31% yield based on Ph₃P. During the photolysis, the reaction rate of ammonia production gradually decreased, and the amount of produced ammonia reached the maximum in 20 h of irradiation (Fig. 2b, Supplementary Table 1). As described later, the gradual decrease in the photocatalytic activity was partially derived from the accumulation of the oxidation product of Ph₃P, i.e., Ph₃P=O (see Supplementary Section 7). A light-on/off experiment under the same reaction conditions revealed that ammonia formation ceased completely in the dark and a chain propagation is not the main reaction pathway; thus, continuous visible light irradiation is necessary for the reaction procedure (Fig. 2c, Supplementary Table 9). The apparent quantum yield of the ammonia production based on the number of electrons used for ammonia formation (Φ(NH₃) = 3×(produced ammonia [µmol])/(incident photon number [µmol])) in the initial stage of the photocatalysis was estimated to be 3.6% using a 405-nm monochromic light and a ferrioxalate chemical actinometer (Fig. 2e, Supplementary Table 11)²⁸.

Fig. 2. Photocatalytic ammonia formation using dinitrogen and water using tertiary phosphines and visible light.

a Material balance of the photocatalytic ammonia formation from dinitrogen and water using triphenylphosphine as a reductant in the presence of the molybdenum complex 1a and the photosensitizer [2b]ONf. b The time courses of the amount of produced ammonia (blue line) and dihydrogen (red line) in the photocatalysis. Data are the average of two experiments with error bars based on s.d. c Light on/off experiment. The grey area shows the reaction time without light irradiation. d Kinetic isotope effect (KIE) on the photocatalytic ammonia formation. e Estimation of quantum yields of the photocatalytic ammonia formation.

Table 1.

The scope of tertiary phosphines for photocatalytic ammonia formation using dinitrogen and water

Entry	R3P	Eox [V]a	ηq [%]b	Φ([2b]) [%]c	Φ([2b])/ηq [%]d	NH3 production [equiv./Mo]	NH3 yield e [%]	H2 production [equiv./Mo]	H2 yield e [%]
1	Ph3P	0.80	24	7.8	32	31 ± 0.4	51 ± 0.7	28 ± 0.7	31 ± 0.6
2	3a	0.51	97	8.3	8.6	20	33	33	35
3	3b	0.68	81	21	26	33 ± 4	54 ± 8	24 ± 3	26 ± 2
4	3c	0.95	7	3.1	44	21 ± 0.4	35 ± 0.2	29 ± 0.4	31 ± 0.4
5	3d	1.23	< 0.1	–	–	< 0.1	< 0.1	< 1	< 1
6	3e	0.79 f	17	6.4	38	23 ± 0.2	38 ± 1	30 ± 0.1	33 ± 0.9
7	3f	0.82	47	8.2	17	28 ± 0.1	44 ± 0.1	28 ± 2	30 ± 2
8	3g	0.85	48	5.8	12	27 ± 1	44 ± 0.7	28 ± 2	31 ± 3
9	3h	0.69	51	6.2	12	11	19	21	24
10	3i	0.87	79	4.2	5.3	9	16	31	35
11	3j	0.72	93	1.9	2.0	2	3	43	49
12	3k	0.76	98	0.3	< 1.0	< 0.1	< 0.1	< 1	< 1
13 g	Ph3P	0.80	24	7.8	32	35	28	36	19
14 h	Ph3P	0.80	24	7.8	32	12	19	36	38

^aOxidation potential vs. Fc^0/+. The cyclic voltammograms of R₃P measured in THF are shown in Supplementary Fig. 19.

^bQuenching fraction under the reaction conditions (see Supplementary Figs. 14, 15 and Supplementary Tables 13, 14).

^cQuantum yields for the production of [2b] (see Supplementary Table 15).

^dQuantum yields for the production of [2b] divided by quenching fractions (see Supplementary Table 14).

^eYield based on Ph₃P.

^fThe value measured in MeCN (Supplementary Fig. 19).

^gPh₃P (360 μmol) was used.

^hPh₃P=O (180 μmol) was added to the reaction solution.

In the ¹H NMR spectrum of the residue obtained after the photocatalysis and evaporation of the reaction solution, triphenylphosphine oxide and Ph₃P were obtained in 86% yield (Ph₃P=O,154 μmol) and in 11% yield (19 μmol), respectively (Fig. 2a, Supplementary Fig. 3). The total yield (82%) of NH₃ and H₂ was almost matched with that of Ph₃P=O (86%), demonstrating that Ph₃P functioned as a two-electron donor via the conversion to Ph₃P=O. Several control experiments revealed that the combination of dinitrogen, Ph₃P, water, Col, a molybdenum complex, an iridium complex, and visible light is essential to promote ammonia production (Supplementary Table 2). It should be noted that collidinium cation ([ColH]⁺) was not detected in the aforementioned NMR spectrum (Supplementary Fig. 3), even though the control experiment without Col revealed that Col was required to trigger the ammonia production. This result suggests that [ColH]⁺ produced by protonation from R₃P^•+-OH₂ in situ was consumed efficiently for ammonia formation and Col/[ColH]⁺ functions as a proton mediator, as discussed in detail below. The control experiment conducted under an argon atmosphere yielded 59% dihydrogen, whereas the experiment without 1a yielded 2% dihydrogen (Supplementary Table 2). Thus, it is suggested that the dihydrogen, as a side product, was produced by unidentified molybdenum species. The nitrogen source of the produced ammonia was confirmed to be the dinitrogen gas by the ¹⁵N-labeling experiments, i.e., the photocatalytic reaction under 1 atm of ¹⁵N₂ (Supplementary Fig. 2). The catalytic reaction was also conducted under typical conditions using D₂O instead of H₂O so as to estimate the kinetic isotope effect (KIE). The ratio of the reaction rates of ammonia formation (k_H2O/k_D2O) measured in two parallel reactions was approximately 1 (Fig. 2d, Supplementary Table 10), which suggests that the proton transfer processes should not be involved in the rate-determining steps of this photocatalytic reaction. When conducting the photocatalysis on a ten times larger scale, the catalytic activity was almost maintained (NH₃: 25 equiv./Mo), demonstrating the possibility of applying this photocatalytic system to a practical reaction scale (Supplementary Fig. 34).

Next, the substituent effect on the phosphines was examined to optimize the reaction conditions (Table 1, Supplementary Table 5). When methoxy (3a), methyl (3b), fluoro (3c), and trifluoromethyl (3d) groups were introduced at the 4-position of the phenyl rings in Ph₃P, ammonia was produced except for when using 3d (Table 1, entries 1-5). The largest amount of ammonia was obtained when Ph₃P (31 equiv./Mo) or 3b (33 equiv./Mo) was used. The Φ(NH₃) values using 3a, 3b, and 3c were estimated to be 4.1, 8.6%, and 1.8%, respectively (Fig. 1e). Based on cyclic voltammograms, the oxidation peak potentials (E_ox vs. Fc^0/+) of the phosphines were estimated as follows: Ph₃P: 0.80 V, 3a: 0.51 V, 3b: 0.68 V, 3c: 0.95 V, and 3d: 1.23 V (Supplementary Fig. 19). These results suggest that it is necessary to use phosphines possessing enough reducing power to trigger photo-induced electron transfer to the photosensitizer [2b]ONf, of which the reduction potential in the excited state (E_red*) was 0.86 V (Supplementary Table 16). When a larger amount of Ph₃P (360 μmol) was used, the amount of produced ammonia was almost identical (35 equiv./Mo, Table 1, entry 13). The different tri-tolyl-phosphines (3e and 3f), where the methyl groups are respectively located at 2- and 3-positions, functioned as slightly less efficient electron donors compared to 3b (Table 1; entries 3, 6, 7). Other tertiary phosphines (3g, 3h, 3i, 3j), which have cyclohexyl or methyl groups on phosphorous atoms instead of phenyl groups, also functioned as electron donors for photocatalytic reactions, even though the catalytic activity became lower than those using Ph₃P (Table 1; entries 1, 8-11). A bidentate phosphine such as (R)-BINAP did not function as an electron donor in this photocatalytic system (Table 1, entry 12).

Not only tertiary phosphines but also proton mediators, reaction solvents, and the amount of water were optimized (Table 2; entries 6-8, 13-17; Supplementary Tables 3, 4, and 7). For the photocatalytic system in this study, Col was the most suitable proton mediator, probably due to the high steric hindrance around the nitrogen atom, which prevents it from coordinating to the molybdenum catalysts to suppress the interaction with dinitrogen molecules (see Supplementary Section 6). The water-immiscible solvents, i.e., benzene and toluene, were suitable for affording large amounts of ammonia after 20 h, though the maximum reaction efficiencies in the initial stage of the photocatalytic reactions were achieved when using a water-miscible solution (i.e., THF) due to the high water concentration (Supplementary Figs. 24–25). The effect of the amount of added water was different when the reaction solvent was changed. When using benzene as a reaction solvent, the larger amount of water suppressed the ammonia formation, probably because of the extraction of the proton mediator ([ColH]⁺) to the water droplet to decrease the concentration of [ColH]⁺ in the benzene phase (Supplementary Figs. 25 and 26). In contrast, when using THF as a reaction solvent, the efficiencies of the ammonia formation in the initial stage became higher with increasing the concentration of water, even though the amounts of ammonia obtained after 20 h of irradiation were lower than those in benzene solutions (Supplementary Figs. 25 and 27). It should be noted that Φ(NH₃), when using THF as a reaction solvent and 7200 µmol of water, reached 22.4%, which was much higher than those in the reported photocatalytic systems for ammonia formation (Supplementary Table 11).

Table 2.

Photocatalytic ammonia formation using dinitrogen and water using Ph₃P in the presence of molybdenum catalysts, photosensitizers and proton mediators^a

Entry	Mo catalyst	Photosensitizer	Proton mediator	NH3 production [equiv./Mo]	NH3 yield [%]b	H2 production [equiv./Mo]	H2 yield [%]b
1	1a	[2b]ONf	Col	31 ± 0.4	51 ± 0.7	28 ± 1	31 ± 1
2	1b	[2b]ONf	Col	1	2	35	38
3	1c	[2b]ONf	Col	10	16	56	61
4	1d	[2b]ONf	Col	4	7	50	57
5	1e	[2b]ONf	Col	26 ± 2	44 ± 4	32 ± 0.1	36 ± 0.1
6	1a	[2b]ONf	Lut	28	48	37	30
7	1a	[2b]ONf	Pic	24	41	33	38
8	1a	[2b]ONf	Py	17	28	26	28
9	1a	[2c]ONf	Col	36 ± 1	59 ± 4	22 ± 0.9	24 ± 0.4
10	1a	[2d]ONf	Col	29 ± 4	47 ± 6	35 ± 3	38 ± 4
11	1a	[2a]ONf	Col	< 1	< 1	1	2
12	1a	[2b]PF6	Col	18	30	22	25
13c	1a	[2b]ONf	Col	27 ± 2	45 ± 2	28 ± 0.7	31 ± 0.6
14 d	1a	[2b]ONf	Col	18	31	36	41
15e	1a	[2b]ONf	Col	8	14	21	23
16 f	1a	[2b]ONf	Col	26 ± 2	43 ± 5	26 ± 3	28 ± 2
17 g	1a	[2b]ONf	Col	19	33	28	32
18 h	1a	[2b]ONf	Col	52 ± 1	44 ± 0.9	65 ± 1	37 ± 1
19i	1a	[2b]ONf	Col	45 ± 2	19 ± 0.9	188 ± 7	53 ± 2

^aThe redox potentials of photosensitizers are summarized in Supplementary Figs. 17, 18 and Supplementary Table 16. The photophysical properties of photosensitizers are summarized in Supplementary Figs. 8, 11–13 and Supplementary Table 12.

^bYield based on Ph₃P.

^cSolvent was changed to toluene.

^dSolvent was changed to THF.

^eSolvent was changed to Et₂O.

^fWater (360 μmol) was used.

^gWater (3600 μmol) was used.

^h1a (1 μmol) was used.

ⁱ1a (0.5 μmol) was used.

The nature of the molybdenum catalysts and photosensitizers has a significant influence on the catalytic activity (Table 2, Supplementary Tables 6 and 8). The use of other molybdenum complexes, i.e., a molybdenum trichloride complex bearing PCP-CF₃ [MoCl₃(PCP-CF₃)] (1b), a molybdenum triiodide complex bearing a non-substituted PCP-type pincer ligand [MoI₃(PCP)] (1c: PCP = 1,3-bis((di-tert-butylphosphino)methyl)-benzimidazol-2-ylidene), a molybdenum triiodide complex bearing a pyridine-based PNP-type pincer ligand [MoI₃(PNP)] (1d: PNP = 2,6-bis(di-tert-butylphosphinomethyl)pyridine), led to much smaller amounts of produced ammonia (Table 2, entries 1-4), indicating that the trichloride structure is not suitable for this photocatalytic system and it is essential to use a PCP-type pincer ligand bearing the electron-withdrawing group to strengthen the coordination of pincer ligand to metal center and to promote electron transfer from the iridium complex to the molybdenum complex. As with the reported system using chemical reductants²¹, the turnover frequencies for the ammonia formation (TOF(NH₃)) in the initial stage of the photocatalytic reactions increased with increasing the bond dissociation free energy of the N-H bonds in the corresponding molybdenum imide complexes (BDFE(N-H)) (Supplementary Table 18, Supplementary Fig. 20). The BDFE(N-H) values also reflect the electron-accepting ability of the complexes; thus, the molybdenum catalysts with high BDFE(N-H) promote the electron transfer from the photosensitizers, resulting in high photocatalytic activity. In fact, from the absorption spectral changes during photoirradiation, we revealed that the molybdenum complexes bearing a PCP-type pincer ligand accepted electrons from the photosensitizer more efficiently than those bearing a PNP-type pincer ligand, probably due to the higher BDFE(N-H) value (Supplementary Table 18, Supplementary Fig. 21). Ammonia was also catalytically obtained with the use of the nitride complex [Mo(N)I(PCP-CF₃)] (1e, Table 2, entry 5), meaning that the nitride complex should also be a key intermediate in this photocatalytic system. It is noted that the KIE value was also negligible even when using 1e (Fig. 2d, Supplementary Table 10), indicating that the N≡N bond cleavage was not involved in the rate-determining steps of this photocatalytic reaction. In the reaction in the presence of 1a, the amount of produced ammonia was increased to 52 and 45 equiv./Mo by changing the amount of 1a to 1 and 0.5 μmol, respectively (Table 2, entries 18 and 19).

The effect of substituents on the iridium complexes was next examined using a series of iridium complexes, which have different numbers of electron-withdrawing groups (i.e., fluoro or trifluoromethyl groups) on the phenylpyridine ligands (Table 2; entries 1, 9-11). The photocatalytic activity using [2c]ONf and [2d]ONf (see structures in Table 2) became similar to that using [2b]ONf. In contrast, [2a]ONf bearing non-substituted phenylpyridine ligands, which functioned as an efficient photosensitizer in the reported photocatalytic systems using dihydroacridine²³ or Hantzsch ester²⁴, led to much lower photocatalytic activity. Based on cyclic voltammetry (Supplementary Fig. 18) and Franck-Condon analysis (Supplementary Table 12)²⁹, reduction potentials in the excited state (E_red* vs. Fc^0/+) of the iridium complexes were estimated as follows: [2b]ONf: +0.86 V, [2c]ONf: +0.88 V, [2d]ONf: +0.46 V, and [2a]ONf: +0.37 V (Supplementary Table 16). These results revealed that the introduction of electron-withdrawing substituents (especially, fluorine groups) on the phenyl rings was effective for increasing the oxidizing power in the excited state to trigger the photo-oxidation of phosphines (E_ox of Ph₃P: 0.80 V).

Mechanistic studies

To get an insight into the reaction mechanism, we investigated the photochemical property of [2b]ONf and conducted several stoichiometric reactions. We first performed Stern-Volmer analysis using [2b]ONf and Ph₃P in benzene (Fig. 3a, Supplementary Fig. 14). The emission intensity from [2b]ONf under stationary light was quenched by increasing the concentration of Ph₃P. From the Stern-Volmer constant (K_SV, 8.9 M⁻¹), the efficiency of emission quenching (η_q) was estimated to be 24% under the reaction conditions for photocatalytic ammonia formation ([Ph₃P] = 36 mM). This quenching efficiency was the same even in the presence of water and Col (K_SV = 8.5 M⁻¹). We also monitored the absorption spectral changes during the photo-irradiation to a benzene solution containing [2b]ONf, Ph₃P, water, and Col (Fig. 3c), and characteristic absorption bands, of which peaks were located at around 510 and 530 nm, were observed. The shape of the observed absorption spectra was exactly matched with that of [2b], which was prepared by the reduction of [2b]ONf with potassium graphite (Fig. 3d, Supplementary Fig. 7). These results clearly indicate that the quenching of photoluminescence was derived from reductive quenching to produce one-electron-reduced species of [2b]ONf ([2b]). It should be noted that the control experiment without Col revealed that Col was essential to produce [2b] efficiently (Fig. 3b), probably because R₃P^•+-OH₂ was deprotonated by Col to produce R₃P^•-OH and collidinium ion ([ColH]⁺) and suppress charge recombination processes (see below). When changing the amount of Col from 180 to 720 µmol, the efficiencies for the production of [2b] were almost unchanged, meaning that the deprotonation step should not be the rate-determining step under the above-mentioned reaction conditions (Supplementary Table 19). We next investigated a reaction between 1a and 3 equiv. of [2b] in benzene at room temperature for 30 min under an atmospheric pressure of dinitrogen. From ¹H NMR spectroscopy measured after the reaction, the corresponding molybdenum–nitride complex [Mo(N)I(PCP-CF₃)] (1e) was formed in 84% yield (Fig. 3e, Supplementary Figs. 4, 5). This result suggests that the reduction of 1a by [2b] produced the nitride complex 1e via the reported pathway, i.e., the production of dinitrogen-bridged dimolybdenum complex after two-electron reduction of 1a, and subsequent cleavage of the N≡N bond in the dinitrogen-bridged complex. The further photochemical reduction of 1e with Ph₃P and water in the presence of [2b]ONf as a photosensitizer and Col under Ar and visible light irradiation afforded ammonia in 94% yield based on used 1e (Fig. 3f), indicating that photochemical hydrogenation of 1e took place using [2b]ONf, Ph₃P, and water to produce ammonia efficiently. These results suggest that the catalytic cycle for ammonia formation by molybdenum complexes should proceed via the splitting route previously proposed by our group¹⁵. The consumption of [2b], i.e., the electron transfer from [2b]ONf to 1a, was also confirmed by the absence of absorption bands derived from [2b] in the absorption spectrum after 1 h of photo-irradiation to the benzene solution containing 1a, [2b]ONf, Ph₃P, water, and Col (Fig. 3b, c). In the absorption spectral changes during the photoirradiation, the characteristic absorption bands derived from [2b] were not observed almost completely, meaning that the efficiency of the electron transfer from the iridium photosensitizer to the molybdenum catalyst should be high enough.

Fig. 3. Stoichiometric reactions of iridium and molybdenum complexes and photosensitizing reactions by iridium complexes.

a Stern-Volmer analysis using [2b]ONf and Ph₃P. The emission spectra (middle panel) in the presence of Ph₃P (0.1 M: red line, 0.01-0.06 M: grey lines) and in the absence of Ph₃P (black line), and the Stern-Volmer plots (right panel) are also illustrated. b Absorption spectrum (red line) measured after photo-irradiation to the benzene solution containing [2b]ONf (0.8 mM), Ph₃P (36 mM), Col (72 mM), and H₂O (216 mM). The spectra recorded in the control experiments without Col (green line), with 1a (blue line), or without irradiation (black line) are also illustrated. c Absorption spectral changes measured every 30 s during photo-irradiation (before irradiation: black, after 30 s: red, after 60 s: orange, after 90 s: green, after 120 s: blue) to the benzene solution containing [2b]ONf (0.8 mM), Ph₃P (36 mM), Col (72 mM), and H₂O (216 mM) in the absence of 1a (left) or the presence of 1a (right). d Reduction of [2b]PF₆ into [2b] with KC₈. The UV-vis absorption spectra of [2b]ONf (black) and [2b] are also illustrated (right panel). e Reduction of 1a into molybdenum–nitride complex (1e) with [2b] under N₂. f Transformation of the nitride ligand in 1e to ammonia using Ph₃P and H₂O in the presence of [2b]ONf under visible light irradiation.

We also investigated the mechanism for the activity loss. As described above, the photocatalytic reaction using twice the amount of Ph₃P afforded a similar amount of ammonia (Table 1, entry 13; Supplementary Fig. 32). In contrast, when adding Ph₃P=O as an additive, the amount of produced ammonia drastically decreased (Table 1, entry 14; Supplementary Fig. 28). These results indicate that the decrease in the concentration of the electron donor is not the main reason for the activity loss and not Ph₃P, but Ph₃P=O caused an inhibitory effect. The efficiency for the production of [2b] was not affected by Ph₃P=O, and the proton transfer from [ColH]⁺ to Ph₃P=O was not observed (Supplementary Figs. 30, 31). Besides, the molybdenum complexes 1a or 1e were hardly detected in the reaction solution after the photocatalytic reactions by NMR, mass, and UV-vis absorption spectroscopy, indicating that most of the molybdenum complexes were converted to an unidentified species that has less catalytic activity (see Supplementary Fig. 33 and Supplementary Section 7). These results suggest that Ph₃P=O did not affect the efficiency for the production of [2b] and the transfer of protons and/or electrons to the molybdenum catalysts and might suppress the ammonia formation reaction by the molybdenum complexes, possibly via the coordination to an unstable intermediate (see Supplementary Fig. 29 and Supplementary Section 7).

Plausible reaction mechanism

As mentioned above, the Stern-Volmer analysis, the electrochemical investigation, and absorption spectral changes during irradiation revealed that Ph₃P donated electrons to the excited state of [2b]ONf to produce [2b] and Ph₃P^•+ with 24% efficiency. However, [2b] could not be observed in the absence of Col; thus, the charge recombination (i.e., the reduction of Ph₃P^•+ by [2b] to reproduce [2b]⁺ and Ph₃P) competes with the subsequent chemical reaction of Ph₃P^•+ with water and Col (Fig. 4c), suggesting that Col is essential for the photocatalytic reaction to suppress the charge recombination and promote the photosensitizing reaction by [2b]ONf. Taking into account the negligible KIE effect shown in Fig. 2a, the slow and reversible addition of H₂O to Ph₃P^•+ might determine the efficiency of the production of [2b] (see Supplementary Sections 6.3–6.7). The quantum yield for the production of [2b] (Φ([2b])) in the absence of 1a was estimated to be 7.8% (Supplementary Fig. 16) using a Shimadzu QYM-01 system (Supplementary Fig. 6), indicating that more than half of [2b] produced by the reductive quenching was wasted by the charge recombination. In the photocatalytic system, the molybdenum complexes might cause an inner filter effect³⁰ (i.e., suppression of absorbing irradiation light by [2b]ONf due to the strong visible-light absorption ability of the molybdenum complexes) to decrease the Φ([2b]) value (Supplementary Figs. 9, 10). Under the reaction conditions for photocatalytic ammonia formation, due to the inner filter effect caused by 1a, [2b]ONf can absorb only 45% of absorbed photons and Φ([2b]) will decrease to 3.5% ( = 7.8% × 0.45). This estimated value is matched with the observed Φ(NH₃) value (3.6%, Fig. 2e), indicating that the efficiency for photocatalytic ammonia formation is determined mainly by the efficiency of the photosensitizing cycle by [2b]ONf and that of the catalytic cycle for ammonia formation on molybdenum complexes should be high enough for the photocatalytic systems. These hypotheses are also supported by the fact that the turnover numbers became higher when smaller amounts of 1a were used (Table 2, entries 18 and 19). Moreover, a clear and straight relationship between Φ([2b]) and Φ(NH₃) when using various derivatives of Ph₃P also confirmed that the efficiency of the photosensitizing cycle for producing [2b] determined the net efficiencies for catalytic ammonia formation (Fig. 4a, Supplementary Figs. 22). This tendency was also observed when using the other tertiary phosphines (Table 1, Supplementary Figs. 23). It is noteworthy that Φ([2b])/η_q, which reflects the efficiency for the subsequent electron transfer from [2b] produced by reductive quenching to the molybdenum catalyst without charge recombination, decreased when E_ox of tertiary phosphines decreased, indicating that the high electron density on the phosphorous atom (i.e., negative E_ox value) lowered the efficiency of the photosensitizing cycle for the production of [2b] (Fig. 4b, Supplementary Figs. 22). If the nucleophilic attack by water to R₃P^•+ is the rate-determining step as mentioned above, this straight relationship between Φ([2b])/η_q and E_ox should be reasonable because the high electron density on the phosphorous atom will suppress the nucleophilic attack by water and thus promote the competing charge recombination to waste [2b]. The above-mentioned hypothesis is confirmed by the fact that both Φ([2b]) and TOF(NH₃) increased with increasing the concentration of water in THF solutions (Supplementary Figs. 24 and 25).

Fig. 4. Possible reaction pathway of molybdenum- and photoredox-catalyzed ammonia formation from dinitrogen and water.

a Relationship between Φ([2b]) and Φ(NH₃). b Relationship between Φ([2b])/η_q and E_ox. c Proposed mechanism of photosensitizing processes with [2b]ONf containing photochemical water activation to supply electrons and protons to the ammonia formation cycle. d Proposed reaction pathway of photocatalytic ammonia formation from dinitrogen and water using Ph₃P as an electron source and visible light energy. The photocatalytic system is constructed with molybdenum-catalyzed ammonia formation cycles and photosensitizing reactions shown in c.

Considering the above-mentioned photosensitizing processes, a plausible reaction pathway for the cooperative photoredox- and molybdenum-catalyzed ammonia formation from dinitrogen and water with Ph₃P can be summarized in Fig. 4d. The photocatalytic system in this study comprises two catalytic cycles: the photosensitizing cycle and the ammonia formation cycle. In the photosensitizing cycle, the iridium photosensitizer [2]⁺ is excited by absorbing visible light to produce the excited state [2]⁺*. Subsequently, reductive quenching of [2]⁺* by tertiary phosphines (R₃P) produces the reduced iridium complex [2] and the radical cation of R₃P (R₃P^•+). R₃P^•+ reacts with water to produce R₃P^•+-OH₂ and activate the water molecule. Owing to the high acidity, R₃P^•+-OH₂ donates a proton to produce R₃P^•-OH and [ColH]⁺. An electron transfer from [2] to the molybdenum-triiodide complex 1a produces the nitride complex 1e as a key intermediate via the bridging coordination of a dinitrogen molecule and N≡N bond cleavage. [2] and [ColH]⁺ donate an electron and a proton, respectively, to trigger PCET to 1e. By the PCET processes, hydrogenation of 1e proceeds to produce an ammonia complex via the stepwise formation of an imide and an amide complex. The BDFE value of the N-H bonds in the imide (35.8 kcal/mol), amide (53.2 kcal/mol), or ammonia complex (42.8 kcal/mol) is larger than the effective BDFE value of the combination of [2] and [ColH]⁺, suggesting that the PCET processes can be thermodynamically favorable processes (Supplementary Table 17). The bridging coordination of a dinitrogen molecule and the dissociation of ammonia accomplish the catalytic cycle. R₃P^•-OH donates another electron and proton, resulting in the formation of R₃P=O. Although the exact reaction pathway from R₃P^•-OH to R₃P=O is unclear at this stage, there are two possible pathways, according to the literature. One is the stepwise processes, i.e., electron transfer to [2]⁺* (or possibly [2]⁺) for producing R₃P⁺-OH and subsequent deprotonation to Col for producing R₃P=O and [ColH]⁺²⁶. The other is hydrogen atom transfer (HAT) to the molybdenum complexes because of the quite low bond dissociation enthalpy of the O-H bond in R₃P^•-OH (9.4 kcal/mol for Ph₃P^•-OH)²⁷. The accumulation of R₃P=O leads to lower efficiencies for the photocatalytic reactions.

Discussion

In this study, we have developed photocatalytic ammonia production using water as a proton source by utilizing the photoinduced electron transfer from tertiary phosphines. This is a successful example of visible-light-driven catalytic ammonia production from dinitrogen and water using visible light energy in the presence of transition metal complexes as catalysts under ambient reaction conditions. Besides, the photocatalytic activity (especially Φ(NH₃)) was higher than those in the reported photocatalytic systems for ammonia formation, and the scale-up experiments (Supplementary Fig. 34) and the combination reactions with organic transformations (Supplementary Figs. 35–38) clearly demonstrated the possibility of applying it to practical applications. It should be noted that the phosphines, which functioned as electron donors in our reaction system, have been widely and deeply investigated in various research areas, and direct electroreduction of phosphine oxides, which were formed as oxidized products after the catalytic reaction of our reaction system, to phosphines with a gram-scale has recently been achieved by other research groups^31–33. This means phosphine oxides can be converted to phosphines using water as an electron donor and renewable solar energy with solar cells. Therefore, we believe that the achievement described in the present manuscript represents a research breakthrough with respect to the eco-friendly methods for green ammonia production from the most ubiquitous and clean small molecules, i.e., dinitrogen and water, utilizing solar energy.

Methods

General information

Detailed experimental procedures, characterization of compounds, photophysical and photochemical properties of photosensitizers, and the detailed results of photoreactions can be found in the Supplementary Methods, Supplementary Figs., and Supplementary Tables.

Catalytic ammonia formation from dinitrogen and water under visible light irradiation

A typical experimental procedure for the photocatalytic reaction is described below. To a 20 mL Schlenk flask, a molybdenum complex was added in a glove box filled with argon gas, and the flask was then taken out of the glove box. The atmosphere in the flask was changed to dinitrogen gas, and a photosensitizer and a tertiary phosphine were added to the flask as solids under a positive flow of dinitrogen gas. A solvent (5 mL), a proton mediator, and water were sequentially added using syringes in dim light to prepare the solution in the flask. The flask was sealed with a ground glass stopper applied with vacuum grease and was sonicated for approximately one minute. The solution was irradiated by four kinds of LED light (λ = 405, 430, 530, 590 nm) in a merry-go-round irradiation apparatus (Cell System Co., Ltd., Iris-MG-S, Supplementary Fig. 1). The solution was stirred by a magnetic stirrer during the photolysis. The inside of the apparatus was continuously cooled by a motor fan to maintain the temperature within 5 °C from room temperature. After the reaction, the amount of generated dihydrogen was quantified by gas chromatography using a Shimadzu GC-8A equipped with a TCD detector and a SHINCARBON ST (6 m × 3 mm). The reaction solution was evaporated in vacuo, and the distillate was trapped in a dilute H₂SO₄ solution (0.5 M, 10 mL). To the residue in the flask, an aqueous potassium hydroxide solution (30 wt%, 5 mL) was added. Then, the mixture was evaporated in vacuo, and the distillate was trapped in another dilute H₂SO₄ solution (0.5 M, 10 mL). The amounts of ammonia present in the H₂SO₄ solutions were determined by the indophenol method³⁴. No hydrazine was detected by the p-(dimethylamino)benzaldehyde method³⁵.

Author contributions

Y.N. conceived and designed this project. Y.E. mainly conducted the experimental work. Y.Y. mainly analyzed experimental data and prepared the first draft of the manuscript. All authors discussed the results and drafted the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data on photocatalytic reactions, photophysical properties, electrochemical properties, NMR spectroscopy, and UV-vis absorption spectroscopy generated in this study are provided in the Supplementary Information. All data are available from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-59727-w.