Introduction

Hydrogen has long been regarded as one of the most promising sustainable energy carriers because it has a high mass energy density, can be efficiently transformed into electricity by fuel cells, and generates only water during combustion1,2,3,4. In recent years, significant progress has been made in the generation of hydrogen from renewable energy sources and the development of efficient hydrogen-powered fuel cells5,6,7,8. However, hydrogen has not been widely used as an energy source because its storage is challenging. There are economic and safety concerns associated with the compressed and cryogenic liquid hydrogen9,10, therefore reversible hydrogen storage in chemical bonds via catalytic hydrogenation/dehydrogenation reactions is a promising technique11,12. Formic acid13,14,15,16, formaldehyde17,18, and methanol19,20,21,22,23 have been extensively studied as molecular hydrogen carriers. However, H2 release from these compounds involves generation of CO2, and H2 cannot be readily reloaded because the liquid carriers are consumed. The need to use stoichiometric amounts of bases and the low hydrogen capacity of formic acid (4.3 wt%) further limit such approaches. To develop more efficient hydrogen storage systems, liquid organic hydrogen carriers (LOHCs) have emerged as a powerful strategy, in which a pair of H2-rich and H2-lean liquid organic compounds can repeatedly release and unload H224,25,26,27,28. Early LOHCs studies focused on the dehydrogenation of cycloalkanes and the reverse hydrogenation of aromatics, but harsh reaction conditions (usually >250 °C) were required29,30. To lower the endothermicity, LOHCs systems based on nitrogen-containing organic hydrides24, e.g., N-heterocycles, were developed. These have high H2 capacities, in the range 5.3–7.3 wt%31,32,33,34,35,36,37. These systems can be promoted by various homogeneous and heterogeneous catalysts38,39,40 under relatively mild conditions (Fig. 1a). Besides, the Milstein and Prakash groups reported LOHCs systems via Ru-catalyzed amide bond formation and hydrogenation (Fig. 1b)41,42,43,44,45. Notably, widely available and inexpensive amines and alcohols can be used as hydrogen carriers in these systems. Despite the favorable thermodynamics, usually <80% selectivity for dehydrogenative amide bond formation was reached43,44,45.

Fig. 1: Hydrogen storage systems based on N-containing organic hydrides.
figure 1

a Heterocycles hydrogenation and dehydrogenation reactions. b Amide bond formation and Hydrogenation reactions. c Mn-catalyzed hydrogenation and dehydrogenation reactions.

In addition to identifying renewable and inexpensive liquid molecular hydrogen carriers, the development of cost-effective, sustainable, and efficient catalysts for LOHCs systems is a significant goal. To the best of our knowledge, all the reported LOHCs systems relied on the use of precious metal catalysts. The development of efficient catalytic systems based on earth-abundant non-noble metals is therefore important. On the basis of recent achievements in Mn-catalyzed hydrogenation and dehydrogenation reactions46,47,48,49,50,51,52,53,54,55,56,57,58,59,60, we have developed the LOHCs system based on Mn-catalyzed dehydrogenative amide bond formation and the reverse hydrogenation reaction. Remarkably, the maximum selectivity for the dehydrogenative condensation of methanol and N,N’-dimethylethylenediamine (DMEDA) to form N,N’-(ethane-1,2-diyl)bis(N-methylformamide) was 97% (Fig. 1c).

Results and discussion

Optimization of the reaction conditions

We commenced the investigation by exploring the catalytic activities of a series of NNP- and PNP-pincer Mn complexes I to VII in the dehydrogenative condensation of DMEDA (1) and methanol (Table 1). In the presence of PhPNP-complex VI (2 mol%), and tBuOK (4 mol%) in 1,4-dioxane, full conversion of 1 was achieved along with formation of the desired product N,N’-(ethane-1,2-diyl)bis(N-methylformamide) (2a) in 86% yield (Table 1, entry 6). N-methyl-N-[2-(methylamino)ethyl]formamide (2b) and 3-methyl-1-methyleneimidazolidin-1-ium (3) were generated in 9% and 5% yield, respectively. Under the same conditions, the more electron-rich iPrPNP-complex V gave much lower selectivity for 2a (Table 1, entry 5), and the NNP-complexes IIV showed no activity (Table 1, entries 1–4).

Table 1 Mn-catalyzed dehydrogenative condensation of 1 and CH3OH.

Various reaction parameters were screened to further improve the efficiency of this transformation (Table 2). Use of <6 equiv of MeOH decreased the selectivity for 2a and increased production of monoamide 2b (Table 2, entry 3). The added base also crucially affected the selectivity. The best choice was tBuOK; tBuONa gave much worse results (Table 2, entries 1 and 4). When a weaker base, i.e., KOH, was used, the reaction yield of 2a decreased to 71% (Table 2, entry 6). Concentration screening showed that 0.42–0.63 M 1 gave the highest selectivity for the desired product 2a (Table 2, entries 1 and 8). Decreasing the reaction temperature from 165 to 150 °C diminished the reaction efficiency (Table 2, entry 10). We assumed that H2 generated during the reaction would accumulate in the closed system, and this would inhibit dehydrogenation. To verify this assumption, the evolved gas was released from the reaction system after reaction for 2 h, and the reaction was then performed for a further 6 h at 165 °C. The yield of 2a increased to 97% (Table 2, entry 11). The evolved gas (24 mL) was collected in a gas buret. GC analysis showed that the H2 purity in the gas mixture was 98.7% and the yield was 98%. In addition to H2, 1.3% CO was detected as a result of decarbonylation of the formaldehyde intermediate. The generation of CO was completely suppressed by adding 2 mol% of catalyst VI in two equal portions, before and after gas release, after reaction for 2 h. This procedure provided released H2 of purity over 99.9% (Table 2, entry 12).

Table 2 Optimization of dehydrogenative condensation of 1a and CH3OH.

We then turned our attention to the reverse reaction, hydrogenation of 2a to diamine 1 (Table 3). The NNP-pincer Mn catalyst IV was found to be the most active catalyst for this transformation (Supplementary Table 5). A decrease in the H2 pressure from 60 to 40 bar lowered conversion of the monoamide 2b to the fully hydrogenated product 1 at 150 °C (Table 3, entry 2). Notably, this hydrogenation reaction proceeded smoothly at 110 °C, with an excellent yield of 1, when 2 mol% of catalyst IV and 2.5 mol% of tBuOK were used (Table 3, entry 8–10). For 0.25 mmol-scale reactions, a larger amount of base was needed to achieve the same level of selectivity (Table 3, entries 1 and 6). The use of a single catalyst for both the dehydrogenation and hydrogenation reactions is desirable toward practical applications. We therefore optimized the hydrogenation of 2a with catalyst VI, the same catalyst as was used in the dehydrogenation reaction (Supplementary Table 8). Under harsh reaction conditions (180 °C, 80 bar of H2), a 94% yield of 1 was obtained with complex VI as the catalyst (Table 3, entry 11).

Table 3 Optimization of Hydrogenation of 2a.

After establishing the optimum reaction conditions for both the dehydrogenative coupling of methanol and 1, and the hydrogenation of diamide 2a, we performed Mn-catalyzed reversible interconversion between 1/methanol and 2a (Fig. 2). The cycle began with dehydrogenation using 2 mol% VI as the catalyst; this resulted in full conversion of 1. The crude reaction mixture, 2 mol% IV, and 2.5 mol% tBuOK were transferred to an autoclave for the reverse hydrogenation. The catalytic activity of IV was not decreased by the presence of the dehydrogenation catalyst VI. After one cycle, diamine 1 was recovered in 95% yield. Moreover, the reversible hydrogenation and dehydrogenation cycle could also be realized using the same catalyst VI, albeit with harsh reaction conditions for the H2-loading process. Compared with the corresponding LOHC system based on Ru-catalysis43, the Mn-catalyzed dehydrogenation process in this system required a higher reaction temperature (165 °C vs 120 °C) and catalyst loading (2 mol% vs 1 mol%), which has yet to be improved in the future works.

Fig. 2
figure 2

Reversible interconversion between 1/methanol and 2a by Mn-catalyzed hydrogenation and dehydrogenation.

Control experiments

Under the optimized reaction conditions for catalyst VI bearing the N–H moiety, the corresponding N-methyl-substituted complex VII showed no reactivity for the dehydrogenative condensation of 1 and methanol, and gave only 7% conversion for the hydrogenation of 2a to 2b. (Fig. 3) These results demonstrate the crucial role of the N–H group in metal–ligand-cooperation for the (de)hydrogenation process. Furthermore, mercury and ligand poisoning experiments indicated that both hydrogenation and dehydrogenation reaction processes proceed with involvement of homogeneous molecular catalysts (Supplementary Tables 1113).

Fig. 3: Mn-catalyzed hydrogenation and dehydrogenation using N-methyl substituted catalyst VII.
figure 3

a VII-catalyzed Dehydrogenation reaction. b VII-catalyzed hydrogenation reaction. c The structure of VII.

Further mechanistic insights into the dehydrogenative condensation reaction were obtained by performing a series of control experiments (Fig. 4). The Mn complex VI-catalyzed dehydrogenation of methanol in the absence of diamine 1 only produced methyl formate (4) in 4% yield. This shows that dehydrogenative ester formation is unfavorable, therefore aminolysis of 4 with 1 is not the major reaction pathway for Mn-catalyzed dehydrogenative amide formation (Fig. 4a). Significant amounts of H2 (14.6 mL) and CO (7.4 mL) in a 2:1 ratio were generated in the dehydrogenation of methanol (1.5 mmol). The use of paraformaldehyde instead of methanol led to the formation of 2a, 2b, and 3, albeit with lower selectivities (Fig. 4b). These results indicate that formaldehyde is a possible reaction intermediate in the dehydrogenative condensation of methanol with 1. Use of monoamide 2b as the starting material gave 2a in 92% yield. This shows that 2b is a likely intermediate in diamide formation (Fig. 4c). In the absence of a Mn catalyst, 1 reacted with paraformaldehyde to form 3-methyl-1-methyleneimidazolidin-1-ium (3) (Fig. 4d). In contrast, 3 was not produced from 2b under the same conditions (Fig. 4e). This shows that 3 is formed by aminalization of 1 and formaldehyde formed in situ.

Fig. 4: Control experiments for mechanistic study.
figure 4

a VI-catalyzed dehydrogenation of methanol. b Dehydrogenative condensation of 1 with paraformaldehyde. c Dehydrogenative condensation of 2b with methanol. d Condensation of 1 and paraformaldehyde without Mn catalyst. e Reactivity of 2b without Mn catalyst.

Accordingly, to prevent the formation of byproduct 3 in the catalytic dehydrogenative condensation reaction, the Mn-catalyzed dehydrogenation of the hemiaminal intermediate 5 must be much faster than its dehydration under these reaction conditions (Fig. 5).

Fig. 5
figure 5

Analysis of the competitive reaction pathways of intermediate 5.

Proposed mechanism

Based on the above results and our previous studies of Mn-catalyzed alcohol dehydrogenation61,62,63,64, plausible reaction pathways for the dehydrogenative condensation of methanol and 1 are proposed (Fig. 6). The dehydrogenation of methanol gives formaldehyde as a reaction intermediate. This reacts with one amine group of 1 to generate the hemiaminal species 5, which eliminates one molecule of H2 to give the monoamide intermediate 2b. Reaction of 2b with another molecule of formaldehyde affords the final product 2a and one molecule of H2 is released. A possible minor reaction pathway is dehydrogenative ester formation from methanol to form methyl formate (4), which undergoes aminolysis with 1 to deliver diamide 2a. The decarbonylation of formaldehyde to CO and H2 can be prevented in this reaction system by fast condensation of formaldehyde with the amino groups of 1 and monoamide 2b.

Fig. 6
figure 6

Plausible reaction pathways for dehydrogenative condensation of methanol and 1.

In summary, a LOHC system based on non-noble-metal catalysis was developed. The Mn-catalyzed dehydrogenative condensation of methanol and DMEDA to form N,N’-(ethane-1,2-diyl)bis(N-methylformamide), and the reverse hydrogenation reaction, gave unloading and loading of H2 in excellent yields with a theoretical hydrogen capacity of 5.3 wt%. Reversible hydrogen storage was shown by subsequent hydrogenation of the resulting dehydrogenation mixture to regenerate H2-rich compounds. The maximum selectivity for the dehydrogenation reaction was 97%.

Methods

General procedure for the dehydrogenation

All dehydrogenation experiments were carried out in a 25 mL pressure seal tube. In the argon atmosphere glovebox, [Mn], tBuOK, solvent, N,N’-dimethylethylenediamine 1, and MeOH were added sequentially to the seal tube equipped with a magnetic stir bar, The reaction mixture was stirred at given temperature for 16 h and cooled to room temperature. After the gas was released, the conversion of 1 was determined by NMR and the yield of products 2a, 2b, and 3 was determined by GC wiith 1,1,2,2-tetrachloroethane as the internal standard.

General procedure for the hydrogenation

All hydrogenation experiments were carried out in a Parr Instruments 4560 series autoclave (300 mL) containing an alloy plate with wells for seven 4 mL glass vials. In the argon atmosphere glovebox, N,N’-(Ethane-1,2-diyl)bis(N-methylformamide) 2a, [Mn], tBuOK, and solvent were added sequentially to the vial equipped with a magnetic stir bar, which was capped with a septum threaded with a syringe. The vial was placed in the alloy plate, which was then placed to the predried autoclave. Once sealed, the autoclave was purged three times with hydrogen, then pressurized to given pressure and heated at given temperature for 16 h. After reaction, the autoclave was cooled to 0 °C, depressurized. The conversion and yield were determined by GC and NMR with 1,1,2,2-tetrachloroethane as the internal standard.