Ammonia (NH3) is arguably the most readily available nitrogen feedstock, with an annual production over 182 million tons from elemental nitrogen and hydrogen via the Haber−Bosch process1,2. The conversion of inorganic NH3 or NH3·H2O into organic amines thus continues to attract the attention of academia and industry (Fig. 1a)3,4,5,6,7. Among those, direct catalytic syntheses of primary amines (-NH2) from NH3·H2O attracts special attention due to the cost and step economy8,9,10. Moreover, primary amines are not only commonly found in pharmaceuticals, natural products and agrochemicals11,12,13,14,15, but also could be readily derived to more complex nitrogen-containing compounds16,17. However, the development of such a process by transition-metal catalysis is hampered by the high strength of the N−H bond (107 kcal mol−1) and Lewis basicity of NH3, resulting in poisoning electrophilic metal catalysts18,19. The transition-metal-catalyzed N−H insertion reactions constitute a well-established strategy for C−N bond formation20,21,22,23,24,25,26,27,28,29, while the few known examples of N−H insertion with NH3 typically report a mixture of primary, secondary, and even tertiary amines30,31. Zhou and co-workers very recently disclosed a milestone progress of asymmetrical carbene insertion into the N−H bonds of NH3 (in MTBE) by the cooperative action of copper complexes and chiral hydrogen-bond donor, albeit the scope of this chemistry was limited to alkyl diazoesters32. NH3·H2O is a cheaper and safer nitrogen source than pressurized liquid NH3, not requiring special equipment for transportation, storage, and handling, thus at a reduced cost per mole of NH3 equivalents10,33. However, such a carbene insertion into the N−H bonds of NH3·H2O has not yet been achieved to date, presumably as the easy deactivation of transition-metal complexes by the formation of a stable Werner complex or ligand exchange with Lewis basic NH3 inhibits the generation of metal-carbene complexes18,19,32. Moreover, even if the desired metal carbenes were generated, competitive O−H insertion with water34,35 and multiple N−H insertion30,31 with initially formed primary amines still exist (Fig. 1b).

Fig. 1: Direct route to primary amines from inorganic NH3 and NH3·H2O: strategies and challenges.
figure 1

a Strategies to primary amines using NH3 and NH3·H2O as nitrogen source. b Challenges of N−H insertion of NH3·H2O. c TpBr3Ag-catalyzed two-phase system enables chemoselective carbene insertion into the N−H bonds of NH3·H2O. Tp, tris(pyrazolyl)borate; Ar, aryl; Tfs, 2-(trifluoromethyl)benzenesulfonyl.

Herein, we report a promising solution to this long-term challenge by Ag-catalyzed two-phase reaction system for the chemoselective carbene insertion into the N−H bonds of NH3·H2O using a variety of diazo compounds24,25 and N-triftosylhydrazones36,37,38,39,40 as carbene precursors (Fig. 1c). Our work stems from the following initial assumptions. First, coordination by a homoscorpionate Tp ligand protects the silver center, which enables it to react with diazo compound to generate a silver carbene even in the presence of NH332 and water41,42,43. Second, water acts as a proton-transporter to facilitate the 1,2-proton shift of N-ylide, thereby ensuring the selectivity of carbene N−H insertion43,44,45,46. This process represents an efficient and practical chemoselective carbene insertion into the N−H bonds of NH3·H2O, enabling the efficient synthesis of primary amines, including diaryl methylamines and α-amino acid esters, valuable building blocks for pharmaceuticals and agrochemicals11,12,13,14,15.

Results and discussion

Investigation of reaction conditions

Initial screening studies were conducted on the N−H insertion of NH3·H2O with methyl phenyldiazoacetate 1 (Table 1). After evaluating multiple reaction parameters, the desired N–H insertion product 2 was obtained under optimized conditions in 92% yield (in 12 h from treating 1 with 8.0 equiv of NH3·H2O at 60 oC in 1,2-dichloroethane (DCE) in the presence of 10 mol % TpBr3Ag(thf))47, along with 5% of O–H insertion product 3 (Table 1, entry 1). When TpBr3Ag(thf) was replaced with Tp(CF3)2Ag(thf)48, the yield increased to 40% (entry 7). All other tested transition-metal catalysts failed to deliver even a trace amount of the insertion product 2, instead leading to products of side reactions of carbene or diazo compounds (entries 8–12).

Table 1 Optimization of N–H insertion of NH3·H2O with diazo compound

Substrate scope

The scope of substrate diazo compounds was then explored under the optimized conditions (Fig. 2a). The tested aryl and heteroaryl diazoacetates resulted in the desired α-amino acid esters (426) in 53–98% yield, regardless of electronic character or position of the substituents on the aromatic ring. In addition to the methyl and ethyl esters, benzyl (27), allyl (28), propargyl (29), 2-(trimethylsilyl)ethyl (30), tert-butyl (31), and cyclohexyl (32) phenyldiazoacetates also furnished the corresponding insertion products in good yield. The reaction is not limited to donor/acceptor diazo compounds, and can be successfully expanded to donor/donor diazo compounds. A broad range of symmetric and unsymmetric diaryl diazomethanes afforded the target diaryl methylamines (3339) in good to excellent yield. Unfortunately, alkyl diazo compounds are a current limitation, undergoing competing 1,2-H shift to form alkenes.

Fig. 2: Reaction scope of carbene insertion into the N−H bonds of NH3·H2O.
figure 2

a Silver-catalyzed N–H insertion of NH3·H2O with diazo compounds. b Silver-catalyzed N–H insertion of NH3·H2O with N-triftosylhydrazones. Reactions were carried out on a 0.3-mmol scale. Isolated yield reported. aIsolated as hydrochloride salt. tBu tert-butyl, Bpin boronic acid pinacol ester, Ac acetyl, Ms methanesulfonyl, Ph phenyl, TMS trimethylsilyl, Bn benzyl.

As the toxicity and potential explosivity of high-energy diazo compounds prevent scale-up of this transformation24,25, we explored the possible use of easily prepared, bench-stable N-sulfonylhydrazones as carbene precursors36,37,49,50,51. Another round of optimization studies with diphenyl N-triftosylhydrazone as model substrate resulted in the desired diarylmethylamine (33) being obtained in 85% yield, when the reaction was performed in DCE at 80 oC with Cs2CO3 as the base (see Supplementary Table 2 for details). In contrast, N-tosylhydrazone proved unsuitable substrate, as the same product was obtained in a much lower 33% yield under identical conditions (entry 11, Supplementary Table 2). As shown in Fig. 2b, under these modified conditions, various diaryl N-triftosylhydrazones provided the desired diaryl methylamines in good to excellent yield along with a trace amount of O−H insertion products (33, 34, 37, and 4060)—the electronic and steric effects did not impact the reaction efficiency and chemoselectivity. Heteroaryl methylamines, including benzofuryl (61, 64, and 65), benzothienyl (62 and 63), furyl (66) and thienyl (67) methylamines, were analogously isolated in moderate to good yield from the corresponding N-triftosylhydrazones. Donor/acceptor N-triftosylhydrazones could also undergo effective N−H insertion reactions (8, 28, 32, 68, and 69). Notably, this in situ diazo generation protocol proved to be equally effective as the corresponding diazo-initiated reactions (8, 28, 3234, and 37). The reaction exhibited excellent functional group tolerance of a range of functional groups, including halogen, aniline, ketone, ester, nitro, olefin, alkyne, tert-butyl, methoxy, trifluoromethyl and trimethylsilyl groups, predominantly providing the desired N−H insertion products along with only trace amounts of O−H insertion products.

Gram-scale synthesis and synthetic applications

When the reaction of NH3·H2O with diphenyl N-triftosylhydrazone was conducted on a gram-scale, hydrochloride 33·HCl was obtained pure, without chromatography, in a two-step 74% yield from diphenyl ketone (Fig. 3a). Our silver-catalyzed protocol could also be applied to late-stage modification of bioactive molecules (Fig. 3b). For instance, natural products containing a hydroxy group, such as phytol, tocopherol, vitamin D3, and (-)-β-citronellol, were first converted into the corresponding phenyldiazoacetates, then subjected to the optimized reaction conditions, affording the corresponding α-amino acid esters in moderate to good yield (7073). Similarly, doxepin-1, a precursor of doxopin hydrochloride (a psychotropic drug) was easily converted into diarylmethylamine 74 in moderate yield through the corresponding diazo compound. The diarylmethylamine moiety is present in many agrochemical and pharmaceutical compounds, for example, the hydrochloride salts of cetirizine, hydroxyzine, and meclizine possessing the structural motif of 4612, and the antimigraine drug lomerizine containing the structural motif of 4013. Ketoprofen (an oral analgesic) and fenofibrate (an oral drug used to lower cholesterol levels) were converted to the corresponding primary amines 75 and 76, respectively, via the corresponding N-triftosylhydrazones. Compounds 77 and 78, intermediates in the synthesis of pharmaceutical compounds (GK-GKRP disruptor14 and DOR115 agonist, respectively), could also be obtained by this N−H insertion of NH3·H2O with N-triftosylhydrazones, showcasing the potential of our protocol for applications in drug discovery (Fig. 3c).

Fig. 3: Synthetic applications.
figure 3

a Gram-scale reaction. b Late-stage transformation of bioactive and drug molecules. aStarted from the corresponding diazo compound. bStarted from the corresponding N-triftosylhydrazone. c Shortened synthesis of key intermediates of drug molecule. Detailed reaction conditions are provided in the supplementary materials.

Mechanistic investigations

To gain insights into the reaction mechanism and the origin of chemoselectivity, we performed control experiments and density functional theory (DFT) calculations. A competition experiment between NH3·H2O and styrene resulted in a mixture of insertion product 33 and cyclopropane 79, suggesting that the ylide intermediate may be generated from silver carbene (Fig. 4a)21. DFT calculations were carried out to explore the possible reaction pathways of silver carbene Int1 with NH3, diphenylmethylamine 33, and H2O, respectively (see Supplementary Figs. 38 for details). Computed lowest-energy pathways for single N−H, double N−H, and O−H insertion are shown in Fig. 4b. The calculated Ag−C distance in Int1 of 2.093 Å points to a weak silver carbene Ag=C bond with an electrophilic carbenic carbon, which favors nucleophilic attack of X−H bonds onto Int121,36. DFT calculations show that the free energy activation for the formation of silver-ligated N-ylide intermediate Int2-N from NH3 (2.7 kcal mol−1) is much lower than that for the formation of Int2-N’ from primary amine 33 and Int2-O from H2O (7.9 and 7.4 kcal mol−1, respectively). These results are consistent with the nucleophilicity of NH3, primary amine 33, and H2O, respectively (for the comparison of their nucleophilicity by natural population analysis, see Supplementary Fig. 9).

Fig. 4: Experimental and DFT computational insights into the mechanism.
figure 4

a Control experiments. b Computed lowest-energy pathways for silver-catalyzed carbene insertion of NH3, H2O, and primary amine 33. The relative free energies present in parentheses and the (RDS) energy barriers (kcal mol–1) were calculated at the SMD(DCE)-M06/[6-311 + G(d,p)-SDD(Ag/Br)] level. All distances are in angstroms. c X-Ray structure of [TpBr3Ag]2(DCE). d The Color-filled RDG isosurface for Int3-N’(isovalue set to 0.5): the water molecules are omitted for clarity: (blue) areas of attraction (covalent bonding); (green) vdW interactions; (red) areas of repulsion (steric and ring effects).

The weak Ag−C bond (2.212 Å) in the silver-ligated ylide favors the ligand exchange of NH3 with Int3-N to release a free ylide Int4-N and silver complex LAg(NH3). The water-assisted 1,2-proton shift from free N-ylide Int4-N opens the lowest-energy route to the N−H insertion of NH3 (Supplementary Figs. 3 and 4)43,44,45,46. Because of the weak silver coordination, LAg(NH3) can react with the solvent (DCE) to generate reactive [TpBr3Ag]2(DCE), thus suggesting LAg(NH3) differs from the generally stable metal−ammine complexes8,9. The structure of [TpBr3Ag]2(DCE) was confirmed by X-ray and NMR on a sample of isolated material (Fig. 4c), which proved as effective as TpBr3Ag(thf) in the N−H insertion of NH3·H2O (see Supplementary Fig. 2). These results are consistent with our initial hypothesis that the homoscorpionate TpBr3 ligand renders the silver catalyst compatible with NH3, thus achieving the challenging N−H insertion of NH3·H2O.

Albeit similar to that of NH3, the reaction pathway for the N−H insertion of diphenylmethylamine 33 (Supplementary Fig. 6) encounters a higher energy barrier for the cleavage of the Ag−C bond in ylide Int3-N’ (9.3 vs. 2.1 kcal mol−1), owing to the stronger Br···π weak interaction between the phenyl and TpBr3 ligand in Int3-N’, as determined by color-filled reduced density gradient (RDG, Fig. 4d and Supplementary Fig. 9)52,53. This hypothesis was confirmed by the reaction between primary amine 33 and diphenyldiazomethane S33 without NH3·H2O, whereby in the absence of the Br···π weak interaction AgOTf led to a higher product yield (60% vs. 98%, Fig. 4a). The dual N−H insertion proceeds through an activation barrier of 12.0 kcal mol−1, which is 8.7 kcal mol−1 higher than that for the N−H insertion of NH3, thus accounting for the absence of over N−H insertion product 33’ under the optimized reaction condition. We also investigated the reaction of the silver carbene Int1 with water (Fig. S4). The Ag−C distance in Int3-O is shorter than that in Int3-N (2.156 vs. 2.212 Å), which favors O−H insertion as the lowest-energy route for the 1,2-proton shift assisted by two molecules of water, with an activation free energy of 6.8 kcal mol−1. Therefore, the formation of ylide Int2-O, with an activation free energy of 7.4 kcal mol−1, is rate-limiting step for the O−H insertion.

The overall values for activation free energy for the turnover limiting-steps (1,2-H shift or O-ylide formation) of the N−H, dual N−H, and O−H insertion paths were 3.3, 12.0, and 7.4 kcal mol−1, respectively, consistent with the experimental 10:0:1 chemoselectivity in the respective products. The weak interaction between the silver center and the carbenic carbon is beneficial to the formation and dissociation of silver-ligated ylide, which favors N−H insertion of NH3 over O−H insertion. On the other hand, the significant Br···π weak interaction between the phenyl group in the substrate and the bulky TpBr3 ligand inhibits the dual N−H insertion.

In this work, we disclose an efficient methodology for the chemoselective carbene insertion into the N−H bonds of NH3·H2O by a silver-catalyzed two-phase system, providing access to value-added primary amines from industrial inorganic nitrogen source. Considering the easy availability and environmental safety of NH3·H2O and the significance of the obtained nitrogen-containing compounds, the discovery of the compatibility between the metal-carbene catalyst with the NH3·H2O adduct may have broader applications in transition-metal-catalyzed N−H activation of NH3·H2O.


General procedure for silver-catalyzed N–H insertion of NH3·H2O with diazo compounds

A Schlenk tube was charged with TpBr3Ag(thf) (33.0 mg, 0.03 mmol, 10 mol %). The tube was evacuated and filled with N2 for three times. A mixture of NH3·H2O (308 μL, 28–30% wt%, 0.6 mmol, 8.0 equiv) and DCE (2 mL) was injected into the tube by syringe, followed by DCE (2 mL) solution of diazo compound (0.3 mmol, 1.0 equiv). The resulting mixture was stirred at 60 °C for 12 h in the dark. When the reaction was completed, the crude reaction mixture was allowed to reach room temperature and concentrated in vacuo and purified by column chromatography on silica gel (petroleum ether/EtOAc) to afford the corresponding N–H insertion product.

General procedure for silver-catalyzed N–H insertion of NH3·H2O with N-triftosylhydrazones

A Schlenk tube was charged with TpBr3Ag(thf) (33.0 mg, 0.03 mmol, 10 mol%) and Cs2CO3 (146.6 mg, 0.45 mmol, 1.5 equiv). The tube was evacuated and filled with N2 for three times. A mixture of NH3·H2O (308 μL, 28–30% wt%, 0.6 mmol, 8.0 equiv) and DCE (2 mL) was injected into the tube by syringe, followed by DCE (2 mL) solution of N-triftosylhydrazone (0.3 mmol, 1.0 equiv). The resulting mixture was stirred at 80 °C for 24 h in the dark. When the reaction was completed, the crude reaction mixture was allowed to reach room temperature and concentrated in vacuo and purified by column chromatography on silica gel (petroleum ether/EtOAc) to afford the corresponding N–H insertion product.